799 files changed, 94320 insertions, 33159 deletions

diff --git a/Makefile b/Makefile
index 91515d6fdb..146c52093b 100644
--- a/Makefile
+++ b/Makefile
@@ -182,7 +182,7 @@ ultrix-gcc:
 
 # Rules for making release tarballs
 
-VERSION=7.5
+VERSION=7.6-devel
 DIRECTORY = Mesa-$(VERSION)
 LIB_NAME = MesaLib-$(VERSION)
 DEMO_NAME = MesaDemos-$(VERSION)
diff --git a/SConstruct b/SConstruct
index 89144f7579..4a0fb94deb 100644
--- a/SConstruct
+++ b/SConstruct
@@ -32,10 +32,10 @@ import common
 default_statetrackers = 'mesa'
 
 if common.default_platform in ('linux', 'freebsd', 'darwin'):
-	default_drivers = 'softpipe,failover,i915simple,trace'
+	default_drivers = 'softpipe,failover,i915simple,trace,identity'
 	default_winsys = 'xlib'
 elif common.default_platform in ('winddk',):
-	default_drivers = 'softpipe,i915simple,trace'
+	default_drivers = 'softpipe,i915simple,trace,identity'
 	default_winsys = 'all'
 else:
 	default_drivers = 'all'
@@ -46,7 +46,7 @@ common.AddOptions(opts)
 opts.Add(ListVariable('statetrackers', 'state trackers to build', default_statetrackers,
                      ['mesa', 'python']))
 opts.Add(ListVariable('drivers', 'pipe drivers to build', default_drivers,
-                     ['softpipe', 'failover', 'i915simple', 'i965simple', 'cell', 'trace', 'r300']))
+                     ['softpipe', 'failover', 'i915simple', 'i965simple', 'cell', 'trace', 'r300', 'identity']))
 opts.Add(ListVariable('winsys', 'winsys drivers to build', default_winsys,
                      ['xlib', 'intel', 'gdi', 'radeon']))
 
diff --git a/configs/autoconf.in b/configs/autoconf.in
index b61d7f33f0..83737e3a7d 100644
--- a/configs/autoconf.in
+++ b/configs/autoconf.in
@@ -20,6 +20,8 @@ CXXFLAGS = @CPPFLAGS@ @CXXFLAGS@ \
 	$(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) $(DEFINES)
 LDFLAGS = @LDFLAGS@
 EXTRA_LIB_PATH = @EXTRA_LIB_PATH@
+RADEON_CFLAGS = @RADEON_CFLAGS@
+RADEON_LDFLAGS = @RADEON_LDFLAGS@
 
 # Assembler
 MESA_ASM_SOURCES = @MESA_ASM_SOURCES@
@@ -31,6 +33,7 @@ SHELL = @SHELL@
 MKLIB_OPTIONS = @MKLIB_OPTIONS@
 MKDEP = @MKDEP@
 MKDEP_OPTIONS = @MKDEP_OPTIONS@
+INSTALL = @INSTALL@
 
 # Python and flags (generally only needed by the developers)
 PYTHON2 = python
diff --git a/configs/darwin b/configs/darwin
index 29fa0f13ff..213c397287 100644
--- a/configs/darwin
+++ b/configs/darwin
@@ -10,7 +10,7 @@ X11_DIR = $(INSTALL_DIR)
 
 # Compiler and flags
 CC = gcc
-CXX = gcc
+CXX = g++
 PIC_FLAGS = -fPIC
 DEFINES =  -D_DARWIN_C_SOURCE -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L \
 	   -D_BSD_SOURCE -D_SVID_SOURCE -D_GNU_SOURCE -DPTHREADS \
diff --git a/configs/default b/configs/default
index b5ef80afb3..60638d739d 100644
--- a/configs/default
+++ b/configs/default
@@ -9,7 +9,7 @@ CONFIG_NAME = default
 
 # Version info
 MESA_MAJOR=7
-MESA_MINOR=5
+MESA_MINOR=6
 MESA_TINY=0
 MESA_VERSION = $(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY)
 
@@ -36,7 +36,10 @@ MKLIB_OPTIONS =
 MKDEP = makedepend
 MKDEP_OPTIONS = -fdepend
 MAKE = make
-INSTALL = $(SHELL) $(TOP)/bin/minstall
+
+# Use MINSTALL for installing libraries, INSTALL for everything else
+MINSTALL = $(SHELL) $(TOP)/bin/minstall
+INSTALL = $(MINSTALL)
 
 # Tools for regenerating glapi (generally only needed by the developers)
 PYTHON2 = python
@@ -90,9 +93,9 @@ EGL_DRIVERS_DIRS = demo
 
 # Gallium directories and 
 GALLIUM_DIRS = auxiliary drivers state_trackers
-GALLIUM_AUXILIARY_DIRS = draw translate cso_cache pipebuffer tgsi sct rtasm util indices
+GALLIUM_AUXILIARY_DIRS = rbug draw translate cso_cache pipebuffer tgsi sct rtasm util indices
 GALLIUM_AUXILIARIES = $(foreach DIR,$(GALLIUM_AUXILIARY_DIRS),$(TOP)/src/gallium/auxiliary/$(DIR)/lib$(DIR).a)
-GALLIUM_DRIVERS_DIRS = softpipe i915simple failover trace
+GALLIUM_DRIVERS_DIRS = softpipe i915simple failover trace identity
 GALLIUM_DRIVERS = $(foreach DIR,$(GALLIUM_DRIVERS_DIRS),$(TOP)/src/gallium/drivers/$(DIR)/lib$(DIR).a)
 GALLIUM_WINSYS_DIRS = xlib egl_xlib
 GALLIUM_WINSYS_DRM_DIRS =
@@ -102,7 +105,7 @@ GALLIUM_STATE_TRACKERS_DIRS = glx
 # Library dependencies
 #EXTRA_LIB_PATH ?=
 GL_LIB_DEPS     = $(EXTRA_LIB_PATH) -lX11 -lXext -lm -lpthread
-EGL_LIB_DEPS    = $(EXTRA_LIB_PATH) -lX11 -ldl
+EGL_LIB_DEPS    = $(EXTRA_LIB_PATH) -lX11 -ldl -lpthread
 OSMESA_LIB_DEPS = $(EXTRA_LIB_PATH) -L$(TOP)/$(LIB_DIR) -l$(GL_LIB)
 GLU_LIB_DEPS    = $(EXTRA_LIB_PATH) -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) -lm
 GLUT_LIB_DEPS   = $(EXTRA_LIB_PATH) -L$(TOP)/$(LIB_DIR) -l$(GLU_LIB) -l$(GL_LIB) -lX11 -lXmu -lXi -lm
diff --git a/configs/linux-x86-64-debug b/configs/linux-x86-64-debug
index 46e72c990a..6f631c0e87 100644
--- a/configs/linux-x86-64-debug
+++ b/configs/linux-x86-64-debug
@@ -4,4 +4,5 @@ include $(TOP)/configs/linux-x86-64
 
 CONFIG_NAME = linux-x86-64-debug
 
+OPT_FLAGS = -g
 DEFINES += -DDEBUG -DDEBUG_MATH
diff --git a/configure.ac b/configure.ac
index f1ff3d4f47..9b65d96b39 100644
--- a/configure.ac
+++ b/configure.ac
@@ -29,6 +29,11 @@ AC_CHECK_PROGS([MAKE], [gmake make])
 AC_PATH_PROG([MKDEP], [makedepend])
 AC_PATH_PROG([SED], [sed])
 
+dnl Our fallback install-sh is a symlink to minstall. Use the existing
+dnl configuration in that case.
+AC_PROG_INSTALL
+test "x$INSTALL" = "x$ac_install_sh" && INSTALL='$(MINSTALL)'
+
 dnl We need a POSIX shell for parts of the build. Assume we have one
 dnl in most cases.
 case "$host_os" in
@@ -411,8 +416,8 @@ WINDOW_SYSTEM=""
 GALLIUM_DIRS="auxiliary drivers state_trackers"
 GALLIUM_WINSYS_DIRS=""
 GALLIUM_WINSYS_DRM_DIRS=""
-GALLIUM_AUXILIARY_DIRS="draw translate cso_cache pipebuffer tgsi sct rtasm util indices"
-GALLIUM_DRIVERS_DIRS="softpipe failover trace"
+GALLIUM_AUXILIARY_DIRS="rbug draw translate cso_cache pipebuffer tgsi sct rtasm util indices"
+GALLIUM_DRIVERS_DIRS="softpipe failover trace identity"
 GALLIUM_STATE_TRACKERS_DIRS=""
 
 case "$mesa_driver" in
@@ -440,6 +445,8 @@ AC_SUBST([GALLIUM_WINSYS_DRM_DIRS])
 AC_SUBST([GALLIUM_DRIVERS_DIRS])
 AC_SUBST([GALLIUM_AUXILIARY_DIRS])
 AC_SUBST([GALLIUM_STATE_TRACKERS_DIRS])
+AC_SUBST([RADEON_CFLAGS])
+AC_SUBST([RADEON_LDFLAGS])
 
 dnl
 dnl User supplied program configuration
@@ -567,6 +574,13 @@ dri)
     GL_PC_REQ_PRIV="libdrm >= $LIBDRM_REQUIRED dri2proto >= $DRI2PROTO_REQUIRED"
     DRI_PC_REQ_PRIV="libdrm >= $LIBDRM_REQUIRED"
 
+    PKG_CHECK_MODULES([LIBDRM_RADEON], [libdrm_radeon], HAVE_LIBDRM_RADEON=yes, HAVE_LIBDRM_RADEON=no)
+
+    if test "$HAVE_LIBDRM_RADEON" = yes; then
+	RADEON_CFLAGS="-DHAVE_LIBDRM_RADEON=1 $LIBDRM_RADEON_CFLAGS"
+	RADEON_LDFLAGS=$LIBDRM_RADEON_LIBS
+    fi
+
     # find the DRI deps for libGL
     if test "$x11_pkgconfig" = yes; then
         # add xcb modules if necessary
@@ -699,7 +713,7 @@ if test "$mesa_driver" = dri; then
             # because there is no x86-64 system where they could *ever*
             # be used.
             if test "x$DRI_DIRS" = "xyes"; then
-                DRI_DIRS="i915 i965 mach64 mga r128 r200 r300 radeon \
+                DRI_DIRS="i915 i965 mach64 mga r128 r200 r300 r600 radeon \
                     savage tdfx unichrome swrast"
             fi
             ;;
@@ -707,13 +721,13 @@ if test "$mesa_driver" = dri; then
             # Build only the drivers for cards that exist on PowerPC.
             # At some point MGA will be added, but not yet.
             if test "x$DRI_DIRS" = "xyes"; then
-                DRI_DIRS="mach64 r128 r200 r300 radeon tdfx swrast"
+                DRI_DIRS="mach64 r128 r200 r300 r600 radeon tdfx swrast"
             fi
             ;;
         sparc*)
             # Build only the drivers for cards that exist on sparc`
             if test "x$DRI_DIRS" = "xyes"; then
-                DRI_DIRS="mach64 r128 r200 r300 radeon ffb swrast"
+                DRI_DIRS="mach64 r128 r200 r300 r600 radeon ffb swrast"
             fi
             ;;
         esac
@@ -732,7 +746,7 @@ if test "$mesa_driver" = dri; then
         # ffb and gamma are missing because they have not been converted
         # to use the new interface.
         if test "x$DRI_DIRS" = "xyes"; then
-            DRI_DIRS="i810 i915 i965 mach64 mga r128 r200 r300 radeon tdfx \
+            DRI_DIRS="i810 i915 i965 mach64 mga r128 r200 r300 r600 radeon tdfx \
                 unichrome savage sis swrast"
         fi
         ;;
@@ -751,7 +765,7 @@ if test "$mesa_driver" = dri; then
 
     # default drivers
     if test "x$DRI_DIRS" = "xyes"; then
-        DRI_DIRS="i810 i915 i965 mach64 mga r128 r200 r300 radeon s3v \
+        DRI_DIRS="i810 i915 i965 mach64 mga r128 r200 r300 r600 radeon s3v \
             savage sis tdfx trident unichrome ffb swrast"
     fi
 
@@ -1181,7 +1195,7 @@ AC_ARG_ENABLE([gallium-nouveau],
     [enable_gallium_nouveau=no])
 if test "x$enable_gallium_nouveau" = xyes; then
     GALLIUM_WINSYS_DRM_DIRS="$GALLIUM_WINSYS_DRM_DIRS nouveau"
-    GALLIUM_DRIVERS_DIRS="$GALLIUM_DRIVERS_DIRS nv04 nv10 nv20 nv30 nv40 nv50"
+    GALLIUM_DRIVERS_DIRS="$GALLIUM_DRIVERS_DIRS nouveau nv04 nv10 nv20 nv30 nv40 nv50"
 fi
 
 
diff --git a/docs/envvars.html b/docs/envvars.html
index 7fd9fe7c0a..6d7a321da9 100644
--- a/docs/envvars.html
+++ b/docs/envvars.html
@@ -25,6 +25,9 @@ fragment programs (intended for developers only)
 <li>MESA_TNL_PROG - if set, implement conventional vertex transformation
 operations with vertex programs (intended for developers only).
 Setting this variable automatically sets the MESA_TEX_PROG variable as well.
+<li>MESA_EXTENSION_OVERRIDE - can be used to enable/disable extensions.
+A value such as "GL_EXT_foo -GL_EXT_bar" will enable the GL_EXT_foo extension
+and disable the GL_EXT_bar extension.
 </ul>
 
 <p>
diff --git a/docs/openvg.html b/docs/openvg.html
new file mode 100644
index 0000000000..442ee522f1
--- /dev/null
+++ b/docs/openvg.html
@@ -0,0 +1,68 @@
+<HTML>
+
+<TITLE>Mesa Release Notes</TITLE>
+
+<head><link rel="stylesheet" type="text/css" href="mesa.css"></head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>OpenVG State Tracker</H1>
+
+<p>
+The current version of the OpenVG state tracker implements OpenVG 1.0.
+</p>
+<p>
+More informations about OpenVG can be found at
+<a href="http://www.khronos.org/openvg/" target="_parent">
+http://www.khronos.org/openvg/</a> .
+</p>
+<p>
+The OpenVG state tracker depends on the Gallium architecture and a working EGL implementation.
+</p>
+
+
+<h2>Building the library</h2>
+<ol>
+<li>Build Mesa3D with Gallium3D. Any build that builds Gallium3D libraries and EGL will suffice</li>
+<li>cd src/gallium/state_trackers/vega; make</li>
+<li>The last step will build libOpenVG library. You can add the libdir to LD_LIBRARY_PATH or install libOpenVG</li>
+</ol>
+
+<h3>Sample build</h3>
+A sample build looks as follows:
+<pre>
+  make linux-x86-64-debug
+  cd src/gallium/state_trackers/vega
+  make
+  cd ../../../..
+  export LD_LIBRARY_PATH=$PWD/lib64
+  export EGL_DRIVER="egl_softpipe"
+</pre>
+
+<h2>OpenVG Demos</h2>
+
+<p>
+To build the OpenVG demos:
+</p>
+<pre>
+  cd progs/openvg
+  make
+</pre>
+<p>
+To run a demo:
+</p>
+<pre>
+  cd openvg/demos
+  ./lion
+</pre>
+
+
+<h2>Notes</h2>
+<ul>
+<li>EGL_DRIVER environmental variable: forces usage of a specific EGL driver. Unless you force egl_softpipe the implementation will look for a DRI hardware accelerate driver and unless you have a Gallium driver that supports it, you'll see crashes</li>
+</ul>
+
+</body>
+</html>
diff --git a/docs/relnotes-7.6.html b/docs/relnotes-7.6.html
new file mode 100644
index 0000000000..1e7ccf88bf
--- /dev/null
+++ b/docs/relnotes-7.6.html
@@ -0,0 +1,64 @@
+<HTML>
+
+<TITLE>Mesa Release Notes</TITLE>
+
+<head><link rel="stylesheet" type="text/css" href="mesa.css"></head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 7.6 Release Notes / date TBD</H1>
+
+<p>
+Mesa 7.6 is a new development release.
+People who are concerned with stability and reliability should stick
+with a previous release or wait for Mesa 7.6.1.
+</p>
+<p>
+Mesa 7.6 implements the OpenGL 2.1 API, but the version reported by
+glGetString(GL_VERSION) depends on the particular driver being used.
+Some drivers don't support all the features required in OpenGL 2.1.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+tbd
+</pre>
+
+
+<h2>New features</h2>
+<ul>
+<li><a href="openvg.html">OpenVG</a> front-end (state tracker for Gallium).
+This was written by Zack Rusin at Tungsten Graphics.
+<li>GL_APPLE_vertex_array_object for Gallium drivers and Intel DRI drivers.
+<li>GL_ARB_vertex_array_object for Gallium drivers, software drivers and
+    Intel DRI drivers.
+<li>GL_ARB_copy_buffer extension (supported in Gallium and swrast drivers)
+<LI>GL_ARB_map_buffer_range extension (supported in Gallium and software drivers)
+<li>GL_EXT_provoking_vertex extension (supported in Gallium and software drivers)
+<li>Rewritten radeon/r200/r300 driver using a buffer manager
+<li>radeon/r200/r300 GL_EXT_framebuffer_object support when used with
+    kernel memory manager
+<li>r300 driver support for GL_EXT_vertex_array_bgra, GL_EXT_texture_sRGB
+</ul>
+
+
+<h2>Bug fixes</h2>
+<ul>
+<li>i965 DRI driver fixes, including support for "unlimited" size constant
+    buffers (GLSL uniforms)
+</ul>
+
+
+<h2>Changes</h2>
+<ul>
+</ul>
+
+</body>
+</html>
diff --git a/docs/relnotes.html b/docs/relnotes.html
index 4764eb689d..35a4c7385f 100644
--- a/docs/relnotes.html
+++ b/docs/relnotes.html
@@ -13,6 +13,7 @@ The release notes summarize what's new or changed in each Mesa release.
 </p>
 
 <UL>
+<LI><A HREF="relnotes-7.6.html">7.6 release notes</A>
 <LI><A HREF="relnotes-7.5.html">7.5 release notes</A>
 <LI><A HREF="relnotes-7.4.4.html">7.4.4 release notes</A>
 <LI><A HREF="relnotes-7.4.3.html">7.4.3 release notes</A>
diff --git a/include/GL/gl.h b/include/GL/gl.h
index aad51214fb..05362b9776 100644
--- a/include/GL/gl.h
+++ b/include/GL/gl.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -1739,41 +1740,6 @@ GLAPI void GLAPIENTRY glSeparableFilter2D( GLenum target,
 GLAPI void GLAPIENTRY glGetSeparableFilter( GLenum target, GLenum format,
 	GLenum type, GLvoid *row, GLvoid *column, GLvoid *span );
 
-typedef void (APIENTRYP PFNGLBLENDCOLORPROC) (GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha);
-typedef void (APIENTRYP PFNGLBLENDEQUATIONPROC) (GLenum mode);
-typedef void (APIENTRYP PFNGLCOLORTABLEPROC) (GLenum target, GLenum internalformat, GLsizei width, GLenum format, GLenum type, const GLvoid *table);
-typedef void (APIENTRYP PFNGLCOLORTABLEPARAMETERFVPROC) (GLenum target, GLenum pname, const GLfloat *params);
-typedef void (APIENTRYP PFNGLCOLORTABLEPARAMETERIVPROC) (GLenum target, GLenum pname, const GLint *params);
-typedef void (APIENTRYP PFNGLCOPYCOLORTABLEPROC) (GLenum target, GLenum internalformat, GLint x, GLint y, GLsizei width);
-typedef void (APIENTRYP PFNGLGETCOLORTABLEPROC) (GLenum target, GLenum format, GLenum type, GLvoid *table);
-typedef void (APIENTRYP PFNGLGETCOLORTABLEPARAMETERFVPROC) (GLenum target, GLenum pname, GLfloat *params);
-typedef void (APIENTRYP PFNGLGETCOLORTABLEPARAMETERIVPROC) (GLenum target, GLenum pname, GLint *params);
-typedef void (APIENTRYP PFNGLCOLORSUBTABLEPROC) (GLenum target, GLsizei start, GLsizei count, GLenum format, GLenum type, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOPYCOLORSUBTABLEPROC) (GLenum target, GLsizei start, GLint x, GLint y, GLsizei width);
-typedef void (APIENTRYP PFNGLCONVOLUTIONFILTER1DPROC) (GLenum target, GLenum internalformat, GLsizei width, GLenum format, GLenum type, const GLvoid *image);
-typedef void (APIENTRYP PFNGLCONVOLUTIONFILTER2DPROC) (GLenum target, GLenum internalformat, GLsizei width, GLsizei height, GLenum format, GLenum type, const GLvoid *image);
-typedef void (APIENTRYP PFNGLCONVOLUTIONPARAMETERFPROC) (GLenum target, GLenum pname, GLfloat params);
-typedef void (APIENTRYP PFNGLCONVOLUTIONPARAMETERFVPROC) (GLenum target, GLenum pname, const GLfloat *params);
-typedef void (APIENTRYP PFNGLCONVOLUTIONPARAMETERIPROC) (GLenum target, GLenum pname, GLint params);
-typedef void (APIENTRYP PFNGLCONVOLUTIONPARAMETERIVPROC) (GLenum target, GLenum pname, const GLint *params);
-typedef void (APIENTRYP PFNGLCOPYCONVOLUTIONFILTER1DPROC) (GLenum target, GLenum internalformat, GLint x, GLint y, GLsizei width);
-typedef void (APIENTRYP PFNGLCOPYCONVOLUTIONFILTER2DPROC) (GLenum target, GLenum internalformat, GLint x, GLint y, GLsizei width, GLsizei height);
-typedef void (APIENTRYP PFNGLGETCONVOLUTIONFILTERPROC) (GLenum target, GLenum format, GLenum type, GLvoid *image);
-typedef void (APIENTRYP PFNGLGETCONVOLUTIONPARAMETERFVPROC) (GLenum target, GLenum pname, GLfloat *params);
-typedef void (APIENTRYP PFNGLGETCONVOLUTIONPARAMETERIVPROC) (GLenum target, GLenum pname, GLint *params);
-typedef void (APIENTRYP PFNGLGETSEPARABLEFILTERPROC) (GLenum target, GLenum format, GLenum type, GLvoid *row, GLvoid *column, GLvoid *span);
-typedef void (APIENTRYP PFNGLSEPARABLEFILTER2DPROC) (GLenum target, GLenum internalformat, GLsizei width, GLsizei height, GLenum format, GLenum type, const GLvoid *row, const GLvoid *column);
-typedef void (APIENTRYP PFNGLGETHISTOGRAMPROC) (GLenum target, GLboolean reset, GLenum format, GLenum type, GLvoid *values);
-typedef void (APIENTRYP PFNGLGETHISTOGRAMPARAMETERFVPROC) (GLenum target, GLenum pname, GLfloat *params);
-typedef void (APIENTRYP PFNGLGETHISTOGRAMPARAMETERIVPROC) (GLenum target, GLenum pname, GLint *params);
-typedef void (APIENTRYP PFNGLGETMINMAXPROC) (GLenum target, GLboolean reset, GLenum format, GLenum type, GLvoid *values);
-typedef void (APIENTRYP PFNGLGETMINMAXPARAMETERFVPROC) (GLenum target, GLenum pname, GLfloat *params);
-typedef void (APIENTRYP PFNGLGETMINMAXPARAMETERIVPROC) (GLenum target, GLenum pname, GLint *params);
-typedef void (APIENTRYP PFNGLHISTOGRAMPROC) (GLenum target, GLsizei width, GLenum internalformat, GLboolean sink);
-typedef void (APIENTRYP PFNGLMINMAXPROC) (GLenum target, GLenum internalformat, GLboolean sink);
-typedef void (APIENTRYP PFNGLRESETHISTOGRAMPROC) (GLenum target);
-typedef void (APIENTRYP PFNGLRESETMINMAXPROC) (GLenum target);
-
 
 
 /*
@@ -1978,53 +1944,6 @@ GLAPI void GLAPIENTRY glMultTransposeMatrixf( const GLfloat m[16] );
 
 GLAPI void GLAPIENTRY glSampleCoverage( GLclampf value, GLboolean invert );
 
-typedef void (APIENTRYP PFNGLACTIVETEXTUREPROC) (GLenum texture);
-typedef void (APIENTRYP PFNGLCLIENTACTIVETEXTUREPROC) (GLenum texture);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1DPROC) (GLenum target, GLdouble s);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1DVPROC) (GLenum target, const GLdouble *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1FPROC) (GLenum target, GLfloat s);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1FVPROC) (GLenum target, const GLfloat *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1IPROC) (GLenum target, GLint s);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1IVPROC) (GLenum target, const GLint *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1SPROC) (GLenum target, GLshort s);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD1SVPROC) (GLenum target, const GLshort *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2DPROC) (GLenum target, GLdouble s, GLdouble t);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2DVPROC) (GLenum target, const GLdouble *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2FPROC) (GLenum target, GLfloat s, GLfloat t);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2FVPROC) (GLenum target, const GLfloat *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2IPROC) (GLenum target, GLint s, GLint t);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2IVPROC) (GLenum target, const GLint *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2SPROC) (GLenum target, GLshort s, GLshort t);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD2SVPROC) (GLenum target, const GLshort *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3DPROC) (GLenum target, GLdouble s, GLdouble t, GLdouble r);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3DVPROC) (GLenum target, const GLdouble *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3FPROC) (GLenum target, GLfloat s, GLfloat t, GLfloat r);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3FVPROC) (GLenum target, const GLfloat *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3IPROC) (GLenum target, GLint s, GLint t, GLint r);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3IVPROC) (GLenum target, const GLint *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3SPROC) (GLenum target, GLshort s, GLshort t, GLshort r);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD3SVPROC) (GLenum target, const GLshort *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4DPROC) (GLenum target, GLdouble s, GLdouble t, GLdouble r, GLdouble q);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4DVPROC) (GLenum target, const GLdouble *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4FPROC) (GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4FVPROC) (GLenum target, const GLfloat *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4IPROC) (GLenum target, GLint s, GLint t, GLint r, GLint q);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4IVPROC) (GLenum target, const GLint *v);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4SPROC) (GLenum target, GLshort s, GLshort t, GLshort r, GLshort q);
-typedef void (APIENTRYP PFNGLMULTITEXCOORD4SVPROC) (GLenum target, const GLshort *v);
-typedef void (APIENTRYP PFNGLLOADTRANSPOSEMATRIXFPROC) (const GLfloat *m);
-typedef void (APIENTRYP PFNGLLOADTRANSPOSEMATRIXDPROC) (const GLdouble *m);
-typedef void (APIENTRYP PFNGLMULTTRANSPOSEMATRIXFPROC) (const GLfloat *m);
-typedef void (APIENTRYP PFNGLMULTTRANSPOSEMATRIXDPROC) (const GLdouble *m);
-typedef void (APIENTRYP PFNGLSAMPLECOVERAGEPROC) (GLclampf value, GLboolean invert);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE3DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE2DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE1DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC) (GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLGETCOMPRESSEDTEXIMAGEPROC) (GLenum target, GLint level, void *img);
-
 
 /*
  * GL_ARB_multitexture (ARB extension 1 and OpenGL 1.2.1)
diff --git a/include/GL/gl_mangle.h b/include/GL/gl_mangle.h
index 639d7bb7a4..54147f7338 100644
--- a/include/GL/gl_mangle.h
+++ b/include/GL/gl_mangle.h
@@ -53,6 +53,7 @@
 #define glBeginFragmentShaderATI		MANGLE(BeginFragmentShaderATI)
 #define glBegin		MANGLE(Begin)
 #define glBeginOcclusionQueryNV		MANGLE(BeginOcclusionQueryNV)
+#define glBeginPerfMonitorAMD		MANGLE(BeginPerfMonitorAMD)
 #define glBeginQueryARB		MANGLE(BeginQueryARB)
 #define glBeginQuery		MANGLE(BeginQuery)
 #define glBeginTransformFeedbackEXT		MANGLE(BeginTransformFeedbackEXT)
@@ -257,6 +258,7 @@
 #define glConvolutionParameteri		MANGLE(ConvolutionParameteri)
 #define glConvolutionParameterivEXT		MANGLE(ConvolutionParameterivEXT)
 #define glConvolutionParameteriv		MANGLE(ConvolutionParameteriv)
+#define glCopyBufferSubData		MANGLE(CopyBufferSubData)
 #define glCopyColorSubTableEXT		MANGLE(CopyColorSubTableEXT)
 #define glCopyColorSubTable		MANGLE(CopyColorSubTable)
 #define glCopyColorTable		MANGLE(CopyColorTable)
@@ -309,6 +311,7 @@
 #define glDeleteLists		MANGLE(DeleteLists)
 #define glDeleteObjectARB		MANGLE(DeleteObjectARB)
 #define glDeleteOcclusionQueriesNV		MANGLE(DeleteOcclusionQueriesNV)
+#define glDeletePerfMonitorsAMD		MANGLE(DeletePerfMonitorsAMD)
 #define glDeleteProgram		MANGLE(DeleteProgram)
 #define glDeleteProgramsARB		MANGLE(DeleteProgramsARB)
 #define glDeleteProgramsNV		MANGLE(DeleteProgramsNV)
@@ -343,6 +346,7 @@
 #define glDrawArraysEXT		MANGLE(DrawArraysEXT)
 #define glDrawArraysInstancedARB		MANGLE(DrawArraysInstancedARB)
 #define glDrawArraysInstancedEXT		MANGLE(DrawArraysInstancedEXT)
+#define glDrawArraysInstanced		MANGLE(DrawArraysInstanced)
 #define glDrawArrays		MANGLE(DrawArrays)
 #define glDrawBuffer		MANGLE(DrawBuffer)
 #define glDrawBuffersARB		MANGLE(DrawBuffersARB)
@@ -352,6 +356,7 @@
 #define glDrawElementArrayATI		MANGLE(DrawElementArrayATI)
 #define glDrawElementsInstancedARB		MANGLE(DrawElementsInstancedARB)
 #define glDrawElementsInstancedEXT		MANGLE(DrawElementsInstancedEXT)
+#define glDrawElementsInstanced		MANGLE(DrawElementsInstanced)
 #define glDrawElements		MANGLE(DrawElements)
 #define glDrawMeshArraysSUN		MANGLE(DrawMeshArraysSUN)
 #define glDrawPixels		MANGLE(DrawPixels)
@@ -381,6 +386,7 @@
 #define glEndList		MANGLE(EndList)
 #define glEnd		MANGLE(End)
 #define glEndOcclusionQueryNV		MANGLE(EndOcclusionQueryNV)
+#define glEndPerfMonitorAMD		MANGLE(EndPerfMonitorAMD)
 #define glEndQueryARB		MANGLE(EndQueryARB)
 #define glEndQuery		MANGLE(EndQuery)
 #define glEndTransformFeedbackEXT		MANGLE(EndTransformFeedbackEXT)
@@ -485,6 +491,7 @@
 #define glGenFramebuffers		MANGLE(GenFramebuffers)
 #define glGenLists		MANGLE(GenLists)
 #define glGenOcclusionQueriesNV		MANGLE(GenOcclusionQueriesNV)
+#define glGenPerfMonitorsAMD		MANGLE(GenPerfMonitorsAMD)
 #define glGenProgramsARB		MANGLE(GenProgramsARB)
 #define glGenProgramsNV		MANGLE(GenProgramsNV)
 #define glGenQueriesARB		MANGLE(GenQueriesARB)
@@ -501,7 +508,11 @@
 #define glGetActiveAttribARB		MANGLE(GetActiveAttribARB)
 #define glGetActiveAttrib		MANGLE(GetActiveAttrib)
 #define glGetActiveUniformARB		MANGLE(GetActiveUniformARB)
+#define glGetActiveUniformBlockiv		MANGLE(GetActiveUniformBlockiv)
+#define glGetActiveUniformBlockName		MANGLE(GetActiveUniformBlockName)
 #define glGetActiveUniform		MANGLE(GetActiveUniform)
+#define glGetActiveUniformName		MANGLE(GetActiveUniformName)
+#define glGetActiveUniformsiv		MANGLE(GetActiveUniformsiv)
 #define glGetActiveVaryingNV		MANGLE(GetActiveVaryingNV)
 #define glGetArrayObjectfvATI		MANGLE(GetArrayObjectfvATI)
 #define glGetArrayObjectivATI		MANGLE(GetArrayObjectivATI)
@@ -634,6 +645,12 @@
 #define glGetObjectParameterivARB		MANGLE(GetObjectParameterivARB)
 #define glGetOcclusionQueryivNV		MANGLE(GetOcclusionQueryivNV)
 #define glGetOcclusionQueryuivNV		MANGLE(GetOcclusionQueryuivNV)
+#define glGetPerfMonitorCounterDataAMD		MANGLE(GetPerfMonitorCounterDataAMD)
+#define glGetPerfMonitorCounterInfoAMD		MANGLE(GetPerfMonitorCounterInfoAMD)
+#define glGetPerfMonitorCountersAMD		MANGLE(GetPerfMonitorCountersAMD)
+#define glGetPerfMonitorCounterStringAMD		MANGLE(GetPerfMonitorCounterStringAMD)
+#define glGetPerfMonitorGroupsAMD		MANGLE(GetPerfMonitorGroupsAMD)
+#define glGetPerfMonitorGroupStringAMD		MANGLE(GetPerfMonitorGroupStringAMD)
 #define glGetPixelMapfv		MANGLE(GetPixelMapfv)
 #define glGetPixelMapuiv		MANGLE(GetPixelMapuiv)
 #define glGetPixelMapusv		MANGLE(GetPixelMapusv)
@@ -709,9 +726,11 @@
 #define glGetTransformFeedbackVaryingEXT		MANGLE(GetTransformFeedbackVaryingEXT)
 #define glGetTransformFeedbackVarying		MANGLE(GetTransformFeedbackVarying)
 #define glGetTransformFeedbackVaryingNV		MANGLE(GetTransformFeedbackVaryingNV)
+#define glGetUniformBlockIndex		MANGLE(GetUniformBlockIndex)
 #define glGetUniformBufferSizeEXT		MANGLE(GetUniformBufferSizeEXT)
 #define glGetUniformfvARB		MANGLE(GetUniformfvARB)
 #define glGetUniformfv		MANGLE(GetUniformfv)
+#define glGetUniformIndices		MANGLE(GetUniformIndices)
 #define glGetUniformivARB		MANGLE(GetUniformivARB)
 #define glGetUniformiv		MANGLE(GetUniformiv)
 #define glGetUniformLocationARB		MANGLE(GetUniformLocationARB)
@@ -744,6 +763,10 @@
 #define glGetVertexAttribPointervARB		MANGLE(GetVertexAttribPointervARB)
 #define glGetVertexAttribPointerv		MANGLE(GetVertexAttribPointerv)
 #define glGetVertexAttribPointervNV		MANGLE(GetVertexAttribPointervNV)
+#define glGetVideoi64vNV		MANGLE(GetVideoi64vNV)
+#define glGetVideoivNV		MANGLE(GetVideoivNV)
+#define glGetVideoui64vNV		MANGLE(GetVideoui64vNV)
+#define glGetVideouivNV		MANGLE(GetVideouivNV)
 #define glGlobalAlphaFactorbSUN		MANGLE(GlobalAlphaFactorbSUN)
 #define glGlobalAlphaFactordSUN		MANGLE(GlobalAlphaFactordSUN)
 #define glGlobalAlphaFactorfSUN		MANGLE(GlobalAlphaFactorfSUN)
@@ -1096,6 +1119,9 @@
 #define glPopClientAttrib		MANGLE(PopClientAttrib)
 #define glPopMatrix		MANGLE(PopMatrix)
 #define glPopName		MANGLE(PopName)
+#define glPresentFrameDualFillNV		MANGLE(PresentFrameDualFillNV)
+#define glPresentFrameKeyedNV		MANGLE(PresentFrameKeyedNV)
+#define glPrimitiveRestartIndex		MANGLE(PrimitiveRestartIndex)
 #define glPrimitiveRestartIndexNV		MANGLE(PrimitiveRestartIndexNV)
 #define glPrimitiveRestartNV		MANGLE(PrimitiveRestartNV)
 #define glPrioritizeTexturesEXT		MANGLE(PrioritizeTexturesEXT)
@@ -1173,6 +1199,7 @@
 #define glProgramUniformMatrix4x2fvEXT		MANGLE(ProgramUniformMatrix4x2fvEXT)
 #define glProgramUniformMatrix4x3fvEXT		MANGLE(ProgramUniformMatrix4x3fvEXT)
 #define glProgramVertexLimitNV		MANGLE(ProgramVertexLimitNV)
+#define glProvokingVertexEXT		MANGLE(ProvokingVertexEXT)
 #define glPushAttrib		MANGLE(PushAttrib)
 #define glPushClientAttribDefaultEXT		MANGLE(PushClientAttribDefaultEXT)
 #define glPushClientAttrib		MANGLE(PushClientAttrib)
@@ -1301,6 +1328,7 @@
 #define glSecondaryColorPointerListIBM		MANGLE(SecondaryColorPointerListIBM)
 #define glSecondaryColorPointer		MANGLE(SecondaryColorPointer)
 #define glSelectBuffer		MANGLE(SelectBuffer)
+#define glSelectPerfMonitorCountersAMD		MANGLE(SelectPerfMonitorCountersAMD)
 #define glSeparableFilter2DEXT		MANGLE(SeparableFilter2DEXT)
 #define glSeparableFilter2D		MANGLE(SeparableFilter2D)
 #define glSetFenceAPPLE		MANGLE(SetFenceAPPLE)
@@ -1345,11 +1373,14 @@
 #define glTangent3svEXT		MANGLE(Tangent3svEXT)
 #define glTangentPointerEXT		MANGLE(TangentPointerEXT)
 #define glTbufferMask3DFX		MANGLE(TbufferMask3DFX)
+#define glTessellationFactorAMD		MANGLE(TessellationFactorAMD)
+#define glTessellationModeAMD		MANGLE(TessellationModeAMD)
 #define glTestFenceAPPLE		MANGLE(TestFenceAPPLE)
 #define glTestFenceNV		MANGLE(TestFenceNV)
 #define glTestObjectAPPLE		MANGLE(TestObjectAPPLE)
 #define glTexBufferARB		MANGLE(TexBufferARB)
 #define glTexBufferEXT		MANGLE(TexBufferEXT)
+#define glTexBuffer		MANGLE(TexBuffer)
 #define glTexBumpParameterfvATI		MANGLE(TexBumpParameterfvATI)
 #define glTexBumpParameterivATI		MANGLE(TexBumpParameterivATI)
 #define glTexCoord1d		MANGLE(TexCoord1d)
@@ -1515,6 +1546,7 @@
 #define glUniform4ui		MANGLE(Uniform4ui)
 #define glUniform4uivEXT		MANGLE(Uniform4uivEXT)
 #define glUniform4uiv		MANGLE(Uniform4uiv)
+#define glUniformBlockBinding		MANGLE(UniformBlockBinding)
 #define glUniformBufferEXT		MANGLE(UniformBufferEXT)
 #define glUniformMatrix2fvARB		MANGLE(UniformMatrix2fvARB)
 #define glUniformMatrix2fv		MANGLE(UniformMatrix2fv)
@@ -1688,7 +1720,7 @@
 #define glVertexAttrib4usvARB		MANGLE(VertexAttrib4usvARB)
 #define glVertexAttrib4usv		MANGLE(VertexAttrib4usv)
 #define glVertexAttribArrayObjectATI		MANGLE(VertexAttribArrayObjectATI)
-#define glVertexAttribDivisor		MANGLE(VertexAttribDivisor)
+#define glVertexAttribDivisorARB		MANGLE(VertexAttribDivisorARB)
 #define glVertexAttribI1iEXT		MANGLE(VertexAttribI1iEXT)
 #define glVertexAttribI1i		MANGLE(VertexAttribI1i)
 #define glVertexAttribI1ivEXT		MANGLE(VertexAttribI1ivEXT)
diff --git a/include/GL/glext.h b/include/GL/glext.h
index 41149c935a..5429287d3f 100644
--- a/include/GL/glext.h
+++ b/include/GL/glext.h
@@ -6,7 +6,7 @@ extern "C" {
 #endif
 
 /*
-** Copyright (c) 2007 The Khronos Group Inc.
+** Copyright (c) 2007-2009 The Khronos Group Inc.
 ** 
 ** Permission is hereby granted, free of charge, to any person obtaining a
 ** copy of this software and/or associated documentation files (the
@@ -28,6 +28,13 @@ extern "C" {
 ** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
 */
 
+/* Header file version number, required by OpenGL ABI for Linux */
+/* glext.h last updated $Date: 2009-05-05 15:27:42 -0700 (Tue, 05 May 2009) $ */
+/* Current version at http://www.opengl.org/registry/ */
+#define GL_GLEXT_VERSION 52
+
+/* Function declaration macros - to move into glplatform.h */
+
 #if defined(_WIN32) && !defined(APIENTRY) && !defined(__CYGWIN__) && !defined(__SCITECH_SNAP__)
 #define WIN32_LEAN_AND_MEAN 1
 #include <windows.h>
@@ -45,18 +52,12 @@ extern "C" {
 
 /*************************************************************/
 
-/* Header file version number, required by OpenGL ABI for Linux */
-/* glext.h last updated 2009/03/19 */
-/* Current version at http://www.opengl.org/registry/ */
-#define GL_GLEXT_VERSION 48
-
 #ifndef GL_VERSION_1_2
 #define GL_UNSIGNED_BYTE_3_3_2            0x8032
 #define GL_UNSIGNED_SHORT_4_4_4_4         0x8033
 #define GL_UNSIGNED_SHORT_5_5_5_1         0x8034
 #define GL_UNSIGNED_INT_8_8_8_8           0x8035
 #define GL_UNSIGNED_INT_10_10_10_2        0x8036
-#define GL_RESCALE_NORMAL                 0x803A
 #define GL_TEXTURE_BINDING_3D             0x806A
 #define GL_PACK_SKIP_IMAGES               0x806B
 #define GL_PACK_IMAGE_HEIGHT              0x806C
@@ -83,17 +84,21 @@ extern "C" {
 #define GL_TEXTURE_MAX_LOD                0x813B
 #define GL_TEXTURE_BASE_LEVEL             0x813C
 #define GL_TEXTURE_MAX_LEVEL              0x813D
-#define GL_LIGHT_MODEL_COLOR_CONTROL      0x81F8
-#define GL_SINGLE_COLOR                   0x81F9
-#define GL_SEPARATE_SPECULAR_COLOR        0x81FA
 #define GL_SMOOTH_POINT_SIZE_RANGE        0x0B12
 #define GL_SMOOTH_POINT_SIZE_GRANULARITY  0x0B13
 #define GL_SMOOTH_LINE_WIDTH_RANGE        0x0B22
 #define GL_SMOOTH_LINE_WIDTH_GRANULARITY  0x0B23
-#define GL_ALIASED_POINT_SIZE_RANGE       0x846D
 #define GL_ALIASED_LINE_WIDTH_RANGE       0x846E
 #endif
 
+#ifndef GL_VERSION_1_2_DEPRECATED
+#define GL_RESCALE_NORMAL                 0x803A
+#define GL_LIGHT_MODEL_COLOR_CONTROL      0x81F8
+#define GL_SINGLE_COLOR                   0x81F9
+#define GL_SEPARATE_SPECULAR_COLOR        0x81FA
+#define GL_ALIASED_POINT_SIZE_RANGE       0x846D
+#endif
+
 #ifndef GL_ARB_imaging
 #define GL_CONSTANT_COLOR                 0x8001
 #define GL_ONE_MINUS_CONSTANT_COLOR       0x8002
@@ -106,6 +111,9 @@ extern "C" {
 #define GL_BLEND_EQUATION                 0x8009
 #define GL_FUNC_SUBTRACT                  0x800A
 #define GL_FUNC_REVERSE_SUBTRACT          0x800B
+#endif
+
+#ifndef GL_ARB_imaging_DEPRECATED
 #define GL_CONVOLUTION_1D                 0x8010
 #define GL_CONVOLUTION_2D                 0x8011
 #define GL_SEPARABLE_2D                   0x8012
@@ -206,12 +214,6 @@ extern "C" {
 #define GL_TEXTURE30                      0x84DE
 #define GL_TEXTURE31                      0x84DF
 #define GL_ACTIVE_TEXTURE                 0x84E0
-#define GL_CLIENT_ACTIVE_TEXTURE          0x84E1
-#define GL_MAX_TEXTURE_UNITS              0x84E2
-#define GL_TRANSPOSE_MODELVIEW_MATRIX     0x84E3
-#define GL_TRANSPOSE_PROJECTION_MATRIX    0x84E4
-#define GL_TRANSPOSE_TEXTURE_MATRIX       0x84E5
-#define GL_TRANSPOSE_COLOR_MATRIX         0x84E6
 #define GL_MULTISAMPLE                    0x809D
 #define GL_SAMPLE_ALPHA_TO_COVERAGE       0x809E
 #define GL_SAMPLE_ALPHA_TO_ONE            0x809F
@@ -220,9 +222,6 @@ extern "C" {
 #define GL_SAMPLES                        0x80A9
 #define GL_SAMPLE_COVERAGE_VALUE          0x80AA
 #define GL_SAMPLE_COVERAGE_INVERT         0x80AB
-#define GL_MULTISAMPLE_BIT                0x20000000
-#define GL_NORMAL_MAP                     0x8511
-#define GL_REFLECTION_MAP                 0x8512
 #define GL_TEXTURE_CUBE_MAP               0x8513
 #define GL_TEXTURE_BINDING_CUBE_MAP       0x8514
 #define GL_TEXTURE_CUBE_MAP_POSITIVE_X    0x8515
@@ -233,10 +232,6 @@ extern "C" {
 #define GL_TEXTURE_CUBE_MAP_NEGATIVE_Z    0x851A
 #define GL_PROXY_TEXTURE_CUBE_MAP         0x851B
 #define GL_MAX_CUBE_MAP_TEXTURE_SIZE      0x851C
-#define GL_COMPRESSED_ALPHA               0x84E9
-#define GL_COMPRESSED_LUMINANCE           0x84EA
-#define GL_COMPRESSED_LUMINANCE_ALPHA     0x84EB
-#define GL_COMPRESSED_INTENSITY           0x84EC
 #define GL_COMPRESSED_RGB                 0x84ED
 #define GL_COMPRESSED_RGBA                0x84EE
 #define GL_TEXTURE_COMPRESSION_HINT       0x84EF
@@ -245,6 +240,22 @@ extern "C" {
 #define GL_NUM_COMPRESSED_TEXTURE_FORMATS 0x86A2
 #define GL_COMPRESSED_TEXTURE_FORMATS     0x86A3
 #define GL_CLAMP_TO_BORDER                0x812D
+#endif
+
+#ifndef GL_VERSION_1_3_DEPRECATED
+#define GL_CLIENT_ACTIVE_TEXTURE          0x84E1
+#define GL_MAX_TEXTURE_UNITS              0x84E2
+#define GL_TRANSPOSE_MODELVIEW_MATRIX     0x84E3
+#define GL_TRANSPOSE_PROJECTION_MATRIX    0x84E4
+#define GL_TRANSPOSE_TEXTURE_MATRIX       0x84E5
+#define GL_TRANSPOSE_COLOR_MATRIX         0x84E6
+#define GL_MULTISAMPLE_BIT                0x20000000
+#define GL_NORMAL_MAP                     0x8511
+#define GL_REFLECTION_MAP                 0x8512
+#define GL_COMPRESSED_ALPHA               0x84E9
+#define GL_COMPRESSED_LUMINANCE           0x84EA
+#define GL_COMPRESSED_LUMINANCE_ALPHA     0x84EB
+#define GL_COMPRESSED_INTENSITY           0x84EC
 #define GL_COMBINE                        0x8570
 #define GL_COMBINE_RGB                    0x8571
 #define GL_COMBINE_ALPHA                  0x8572
@@ -276,16 +287,26 @@ extern "C" {
 #define GL_BLEND_SRC_RGB                  0x80C9
 #define GL_BLEND_DST_ALPHA                0x80CA
 #define GL_BLEND_SRC_ALPHA                0x80CB
-#define GL_POINT_SIZE_MIN                 0x8126
-#define GL_POINT_SIZE_MAX                 0x8127
 #define GL_POINT_FADE_THRESHOLD_SIZE      0x8128
-#define GL_POINT_DISTANCE_ATTENUATION     0x8129
-#define GL_GENERATE_MIPMAP                0x8191
-#define GL_GENERATE_MIPMAP_HINT           0x8192
 #define GL_DEPTH_COMPONENT16              0x81A5
 #define GL_DEPTH_COMPONENT24              0x81A6
 #define GL_DEPTH_COMPONENT32              0x81A7
 #define GL_MIRRORED_REPEAT                0x8370
+#define GL_MAX_TEXTURE_LOD_BIAS           0x84FD
+#define GL_TEXTURE_LOD_BIAS               0x8501
+#define GL_INCR_WRAP                      0x8507
+#define GL_DECR_WRAP                      0x8508
+#define GL_TEXTURE_DEPTH_SIZE             0x884A
+#define GL_TEXTURE_COMPARE_MODE           0x884C
+#define GL_TEXTURE_COMPARE_FUNC           0x884D
+#endif
+
+#ifndef GL_VERSION_1_4_DEPRECATED
+#define GL_POINT_SIZE_MIN                 0x8126
+#define GL_POINT_SIZE_MAX                 0x8127
+#define GL_POINT_DISTANCE_ATTENUATION     0x8129
+#define GL_GENERATE_MIPMAP                0x8191
+#define GL_GENERATE_MIPMAP_HINT           0x8192
 #define GL_FOG_COORDINATE_SOURCE          0x8450
 #define GL_FOG_COORDINATE                 0x8451
 #define GL_FRAGMENT_DEPTH                 0x8452
@@ -301,15 +322,8 @@ extern "C" {
 #define GL_SECONDARY_COLOR_ARRAY_STRIDE   0x845C
 #define GL_SECONDARY_COLOR_ARRAY_POINTER  0x845D
 #define GL_SECONDARY_COLOR_ARRAY          0x845E
-#define GL_MAX_TEXTURE_LOD_BIAS           0x84FD
 #define GL_TEXTURE_FILTER_CONTROL         0x8500
-#define GL_TEXTURE_LOD_BIAS               0x8501
-#define GL_INCR_WRAP                      0x8507
-#define GL_DECR_WRAP                      0x8508
-#define GL_TEXTURE_DEPTH_SIZE             0x884A
 #define GL_DEPTH_TEXTURE_MODE             0x884B
-#define GL_TEXTURE_COMPARE_MODE           0x884C
-#define GL_TEXTURE_COMPARE_FUNC           0x884D
 #define GL_COMPARE_R_TO_TEXTURE           0x884E
 #endif
 
@@ -324,15 +338,6 @@ extern "C" {
 #define GL_ELEMENT_ARRAY_BUFFER           0x8893
 #define GL_ARRAY_BUFFER_BINDING           0x8894
 #define GL_ELEMENT_ARRAY_BUFFER_BINDING   0x8895
-#define GL_VERTEX_ARRAY_BUFFER_BINDING    0x8896
-#define GL_NORMAL_ARRAY_BUFFER_BINDING    0x8897
-#define GL_COLOR_ARRAY_BUFFER_BINDING     0x8898
-#define GL_INDEX_ARRAY_BUFFER_BINDING     0x8899
-#define GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING 0x889A
-#define GL_EDGE_FLAG_ARRAY_BUFFER_BINDING 0x889B
-#define GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING 0x889C
-#define GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING 0x889D
-#define GL_WEIGHT_ARRAY_BUFFER_BINDING    0x889E
 #define GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING 0x889F
 #define GL_READ_ONLY                      0x88B8
 #define GL_WRITE_ONLY                     0x88B9
@@ -350,31 +355,42 @@ extern "C" {
 #define GL_DYNAMIC_READ                   0x88E9
 #define GL_DYNAMIC_COPY                   0x88EA
 #define GL_SAMPLES_PASSED                 0x8914
-#define GL_FOG_COORD_SRC                  GL_FOG_COORDINATE_SOURCE
-#define GL_FOG_COORD                      GL_FOG_COORDINATE
-#define GL_CURRENT_FOG_COORD              GL_CURRENT_FOG_COORDINATE
-#define GL_FOG_COORD_ARRAY_TYPE           GL_FOG_COORDINATE_ARRAY_TYPE
-#define GL_FOG_COORD_ARRAY_STRIDE         GL_FOG_COORDINATE_ARRAY_STRIDE
-#define GL_FOG_COORD_ARRAY_POINTER        GL_FOG_COORDINATE_ARRAY_POINTER
-#define GL_FOG_COORD_ARRAY                GL_FOG_COORDINATE_ARRAY
-#define GL_FOG_COORD_ARRAY_BUFFER_BINDING GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING
-#define GL_SRC0_RGB                       GL_SOURCE0_RGB
-#define GL_SRC1_RGB                       GL_SOURCE1_RGB
-#define GL_SRC2_RGB                       GL_SOURCE2_RGB
-#define GL_SRC0_ALPHA                     GL_SOURCE0_ALPHA
-#define GL_SRC1_ALPHA                     GL_SOURCE1_ALPHA
-#define GL_SRC2_ALPHA                     GL_SOURCE2_ALPHA
+#endif
+
+#ifndef GL_VERSION_1_5_DEPRECATED
+#define GL_VERTEX_ARRAY_BUFFER_BINDING    0x8896
+#define GL_NORMAL_ARRAY_BUFFER_BINDING    0x8897
+#define GL_COLOR_ARRAY_BUFFER_BINDING     0x8898
+#define GL_INDEX_ARRAY_BUFFER_BINDING     0x8899
+#define GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING 0x889A
+#define GL_EDGE_FLAG_ARRAY_BUFFER_BINDING 0x889B
+#define GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING 0x889C
+#define GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING 0x889D
+#define GL_WEIGHT_ARRAY_BUFFER_BINDING    0x889E
+#define GL_FOG_COORD_SRC                  0x8450
+#define GL_FOG_COORD                      0x8451
+#define GL_CURRENT_FOG_COORD              0x8453
+#define GL_FOG_COORD_ARRAY_TYPE           0x8454
+#define GL_FOG_COORD_ARRAY_STRIDE         0x8455
+#define GL_FOG_COORD_ARRAY_POINTER        0x8456
+#define GL_FOG_COORD_ARRAY                0x8457
+#define GL_FOG_COORD_ARRAY_BUFFER_BINDING 0x889D
+#define GL_SRC0_RGB                       0x8580
+#define GL_SRC1_RGB                       0x8581
+#define GL_SRC2_RGB                       0x8582
+#define GL_SRC0_ALPHA                     0x8588
+#define GL_SRC1_ALPHA                     0x8589
+#define GL_SRC2_ALPHA                     0x858A
 #endif
 
 #ifndef GL_VERSION_2_0
-#define GL_BLEND_EQUATION_RGB             GL_BLEND_EQUATION
+#define GL_BLEND_EQUATION_RGB             0x8009
 #define GL_VERTEX_ATTRIB_ARRAY_ENABLED    0x8622
 #define GL_VERTEX_ATTRIB_ARRAY_SIZE       0x8623
 #define GL_VERTEX_ATTRIB_ARRAY_STRIDE     0x8624
 #define GL_VERTEX_ATTRIB_ARRAY_TYPE       0x8625
 #define GL_CURRENT_VERTEX_ATTRIB          0x8626
 #define GL_VERTEX_PROGRAM_POINT_SIZE      0x8642
-#define GL_VERTEX_PROGRAM_TWO_SIDE        0x8643
 #define GL_VERTEX_ATTRIB_ARRAY_POINTER    0x8645
 #define GL_STENCIL_BACK_FUNC              0x8800
 #define GL_STENCIL_BACK_FAIL              0x8801
@@ -398,11 +414,8 @@ extern "C" {
 #define GL_DRAW_BUFFER14                  0x8833
 #define GL_DRAW_BUFFER15                  0x8834
 #define GL_BLEND_EQUATION_ALPHA           0x883D
-#define GL_POINT_SPRITE                   0x8861
-#define GL_COORD_REPLACE                  0x8862
 #define GL_MAX_VERTEX_ATTRIBS             0x8869
 #define GL_VERTEX_ATTRIB_ARRAY_NORMALIZED 0x886A
-#define GL_MAX_TEXTURE_COORDS             0x8871
 #define GL_MAX_TEXTURE_IMAGE_UNITS        0x8872
 #define GL_FRAGMENT_SHADER                0x8B30
 #define GL_VERTEX_SHADER                  0x8B31
@@ -453,8 +466,14 @@ extern "C" {
 #define GL_STENCIL_BACK_WRITEMASK         0x8CA5
 #endif
 
+#ifndef GL_VERSION_2_0_DEPRECATED
+#define GL_VERTEX_PROGRAM_TWO_SIDE        0x8643
+#define GL_POINT_SPRITE                   0x8861
+#define GL_COORD_REPLACE                  0x8862
+#define GL_MAX_TEXTURE_COORDS             0x8871
+#endif
+
 #ifndef GL_VERSION_2_1
-#define GL_CURRENT_RASTER_SECONDARY_COLOR 0x845F
 #define GL_PIXEL_PACK_BUFFER              0x88EB
 #define GL_PIXEL_UNPACK_BUFFER            0x88EC
 #define GL_PIXEL_PACK_BUFFER_BINDING      0x88ED
@@ -469,25 +488,29 @@ extern "C" {
 #define GL_SRGB8                          0x8C41
 #define GL_SRGB_ALPHA                     0x8C42
 #define GL_SRGB8_ALPHA8                   0x8C43
+#define GL_COMPRESSED_SRGB                0x8C48
+#define GL_COMPRESSED_SRGB_ALPHA          0x8C49
+#endif
+
+#ifndef GL_VERSION_2_1_DEPRECATED
+#define GL_CURRENT_RASTER_SECONDARY_COLOR 0x845F
 #define GL_SLUMINANCE_ALPHA               0x8C44
 #define GL_SLUMINANCE8_ALPHA8             0x8C45
 #define GL_SLUMINANCE                     0x8C46
 #define GL_SLUMINANCE8                    0x8C47
-#define GL_COMPRESSED_SRGB                0x8C48
-#define GL_COMPRESSED_SRGB_ALPHA          0x8C49
 #define GL_COMPRESSED_SLUMINANCE          0x8C4A
 #define GL_COMPRESSED_SLUMINANCE_ALPHA    0x8C4B
 #endif
 
 #ifndef GL_VERSION_3_0
-#define GL_COMPARE_REF_TO_TEXTURE         GL_COMPARE_R_TO_TEXTURE_ARB
-#define GL_CLIP_DISTANCE0                 GL_CLIP_PLANE0
-#define GL_CLIP_DISTANCE1                 GL_CLIP_PLANE1
-#define GL_CLIP_DISTANCE2                 GL_CLIP_PLANE2
-#define GL_CLIP_DISTANCE3                 GL_CLIP_PLANE3
-#define GL_CLIP_DISTANCE4                 GL_CLIP_PLANE4
-#define GL_CLIP_DISTANCE5                 GL_CLIP_PLANE5
-#define GL_MAX_CLIP_DISTANCES             GL_MAX_CLIP_PLANES
+#define GL_COMPARE_REF_TO_TEXTURE         0x884E
+#define GL_CLIP_DISTANCE0                 0x3000
+#define GL_CLIP_DISTANCE1                 0x3001
+#define GL_CLIP_DISTANCE2                 0x3002
+#define GL_CLIP_DISTANCE3                 0x3003
+#define GL_CLIP_DISTANCE4                 0x3004
+#define GL_CLIP_DISTANCE5                 0x3005
+#define GL_MAX_CLIP_DISTANCES             0x0D32
 #define GL_MAJOR_VERSION                  0x821B
 #define GL_MINOR_VERSION                  0x821C
 #define GL_NUM_EXTENSIONS                 0x821D
@@ -505,19 +528,9 @@ extern "C" {
 #define GL_MAX_ARRAY_TEXTURE_LAYERS       0x88FF
 #define GL_MIN_PROGRAM_TEXEL_OFFSET       0x8904
 #define GL_MAX_PROGRAM_TEXEL_OFFSET       0x8905
-#define GL_CLAMP_VERTEX_COLOR             0x891A
-#define GL_CLAMP_FRAGMENT_COLOR           0x891B
 #define GL_CLAMP_READ_COLOR               0x891C
 #define GL_FIXED_ONLY                     0x891D
-#define GL_MAX_VARYING_COMPONENTS         GL_MAX_VARYING_FLOATS
-#define GL_TEXTURE_RED_TYPE               0x8C10
-#define GL_TEXTURE_GREEN_TYPE             0x8C11
-#define GL_TEXTURE_BLUE_TYPE              0x8C12
-#define GL_TEXTURE_ALPHA_TYPE             0x8C13
-#define GL_TEXTURE_LUMINANCE_TYPE         0x8C14
-#define GL_TEXTURE_INTENSITY_TYPE         0x8C15
-#define GL_TEXTURE_DEPTH_TYPE             0x8C16
-#define GL_UNSIGNED_NORMALIZED            0x8C17
+#define GL_MAX_VARYING_COMPONENTS         0x8B4B
 #define GL_TEXTURE_1D_ARRAY               0x8C18
 #define GL_PROXY_TEXTURE_1D_ARRAY         0x8C19
 #define GL_TEXTURE_2D_ARRAY               0x8C1A
@@ -559,7 +572,6 @@ extern "C" {
 #define GL_RED_INTEGER                    0x8D94
 #define GL_GREEN_INTEGER                  0x8D95
 #define GL_BLUE_INTEGER                   0x8D96
-#define GL_ALPHA_INTEGER                  0x8D97
 #define GL_RGB_INTEGER                    0x8D98
 #define GL_RGBA_INTEGER                   0x8D99
 #define GL_BGR_INTEGER                    0x8D9A
@@ -615,8 +627,6 @@ extern "C" {
 /* reuse GL_TEXTURE_GREEN_TYPE */
 /* reuse GL_TEXTURE_BLUE_TYPE */
 /* reuse GL_TEXTURE_ALPHA_TYPE */
-/* reuse GL_TEXTURE_LUMINANCE_TYPE */
-/* reuse GL_TEXTURE_INTENSITY_TYPE */
 /* reuse GL_TEXTURE_DEPTH_TYPE */
 /* reuse GL_UNSIGNED_NORMALIZED */
 /* reuse GL_FRAMEBUFFER_BINDING */
@@ -716,7 +726,32 @@ extern "C" {
 /* reuse GL_VERTEX_ARRAY_BINDING */
 #endif
 
+#ifndef GL_VERSION_3_0_DEPRECATED
+#define GL_CLAMP_VERTEX_COLOR             0x891A
+#define GL_CLAMP_FRAGMENT_COLOR           0x891B
+#define GL_ALPHA_INTEGER                  0x8D97
+/* Reuse tokens from ARB_framebuffer_object */
+/* reuse GL_TEXTURE_LUMINANCE_TYPE */
+/* reuse GL_TEXTURE_INTENSITY_TYPE */
+#endif
+
 #ifndef GL_VERSION_3_1
+#define GL_SAMPLER_2D_RECT                0x8B63
+#define GL_SAMPLER_2D_RECT_SHADOW         0x8B64
+#define GL_SAMPLER_BUFFER                 0x8DC2
+#define GL_INT_SAMPLER_2D_RECT            0x8DCD
+#define GL_INT_SAMPLER_BUFFER             0x8DD0
+#define GL_UNSIGNED_INT_SAMPLER_2D_RECT   0x8DD5
+#define GL_UNSIGNED_INT_SAMPLER_BUFFER    0x8DD8
+#define GL_TEXTURE_BUFFER                 0x8C2A
+#define GL_MAX_TEXTURE_BUFFER_SIZE        0x8C2B
+#define GL_TEXTURE_BINDING_BUFFER         0x8C2C
+#define GL_TEXTURE_BUFFER_DATA_STORE_BINDING 0x8C2D
+#define GL_TEXTURE_BUFFER_FORMAT          0x8C2E
+#define GL_TEXTURE_RECTANGLE              0x84F5
+#define GL_TEXTURE_BINDING_RECTANGLE      0x84F6
+#define GL_PROXY_TEXTURE_RECTANGLE        0x84F7
+#define GL_MAX_RECTANGLE_TEXTURE_SIZE     0x84F8
 #define GL_RED_SNORM                      0x8F90
 #define GL_RG_SNORM                       0x8F91
 #define GL_RGB_SNORM                      0x8F92
@@ -732,6 +767,41 @@ extern "C" {
 #define GL_SIGNED_NORMALIZED              0x8F9C
 #define GL_PRIMITIVE_RESTART              0x8F9D
 #define GL_PRIMITIVE_RESTART_INDEX        0x8F9E
+/* Reuse tokens from ARB_copy_buffer */
+/* reuse GL_COPY_READ_BUFFER */
+/* reuse GL_COPY_WRITE_BUFFER */
+/* Would reuse tokens from ARB_draw_instanced, but it has none */
+/* Reuse tokens from ARB_uniform_buffer_object */
+/* reuse GL_UNIFORM_BUFFER */
+/* reuse GL_UNIFORM_BUFFER_BINDING */
+/* reuse GL_UNIFORM_BUFFER_START */
+/* reuse GL_UNIFORM_BUFFER_SIZE */
+/* reuse GL_MAX_VERTEX_UNIFORM_BLOCKS */
+/* reuse GL_MAX_FRAGMENT_UNIFORM_BLOCKS */
+/* reuse GL_MAX_COMBINED_UNIFORM_BLOCKS */
+/* reuse GL_MAX_UNIFORM_BUFFER_BINDINGS */
+/* reuse GL_MAX_UNIFORM_BLOCK_SIZE */
+/* reuse GL_MAX_COMBINED_VERTEX_UNIFORM_COMPONENTS */
+/* reuse GL_MAX_COMBINED_FRAGMENT_UNIFORM_COMPONENTS */
+/* reuse GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT */
+/* reuse GL_ACTIVE_UNIFORM_BLOCK_MAX_NAME_LENGTH */
+/* reuse GL_ACTIVE_UNIFORM_BLOCKS */
+/* reuse GL_UNIFORM_TYPE */
+/* reuse GL_UNIFORM_SIZE */
+/* reuse GL_UNIFORM_NAME_LENGTH */
+/* reuse GL_UNIFORM_BLOCK_INDEX */
+/* reuse GL_UNIFORM_OFFSET */
+/* reuse GL_UNIFORM_ARRAY_STRIDE */
+/* reuse GL_UNIFORM_MATRIX_STRIDE */
+/* reuse GL_UNIFORM_IS_ROW_MAJOR */
+/* reuse GL_UNIFORM_BLOCK_BINDING */
+/* reuse GL_UNIFORM_BLOCK_DATA_SIZE */
+/* reuse GL_UNIFORM_BLOCK_NAME_LENGTH */
+/* reuse GL_UNIFORM_BLOCK_ACTIVE_UNIFORMS */
+/* reuse GL_UNIFORM_BLOCK_ACTIVE_UNIFORM_INDICES */
+/* reuse GL_UNIFORM_BLOCK_REFERENCED_BY_VERTEX_SHADER */
+/* reuse GL_UNIFORM_BLOCK_REFERENCED_BY_FRAGMENT_SHADER */
+/* reuse GL_INVALID_INDEX */
 #endif
 
 #ifndef GL_ARB_multitexture
@@ -1251,12 +1321,17 @@ extern "C" {
 #define GL_FRAMEBUFFER_DEFAULT            0x8218
 #define GL_FRAMEBUFFER_UNDEFINED          0x8219
 #define GL_DEPTH_STENCIL_ATTACHMENT       0x821A
-#define GL_INDEX                          0x8222
 #define GL_MAX_RENDERBUFFER_SIZE          0x84E8
 #define GL_DEPTH_STENCIL                  0x84F9
 #define GL_UNSIGNED_INT_24_8              0x84FA
 #define GL_DEPTH24_STENCIL8               0x88F0
 #define GL_TEXTURE_STENCIL_SIZE           0x88F1
+#define GL_TEXTURE_RED_TYPE               0x8C10
+#define GL_TEXTURE_GREEN_TYPE             0x8C11
+#define GL_TEXTURE_BLUE_TYPE              0x8C12
+#define GL_TEXTURE_ALPHA_TYPE             0x8C13
+#define GL_TEXTURE_DEPTH_TYPE             0x8C16
+#define GL_UNSIGNED_NORMALIZED            0x8C17
 #define GL_FRAMEBUFFER_BINDING            0x8CA6
 #define GL_DRAW_FRAMEBUFFER_BINDING       GL_FRAMEBUFFER_BINDING
 #define GL_RENDERBUFFER_BINDING           0x8CA7
@@ -1313,6 +1388,12 @@ extern "C" {
 #define GL_MAX_SAMPLES                    0x8D57
 #endif
 
+#ifndef GL_ARB_framebuffer_object_DEPRECATED
+#define GL_INDEX                          0x8222
+#define GL_TEXTURE_LUMINANCE_TYPE         0x8C14
+#define GL_TEXTURE_INTENSITY_TYPE         0x8C15
+#endif
+
 #ifndef GL_ARB_framebuffer_sRGB
 #define GL_FRAMEBUFFER_SRGB               0x8DB9
 #endif
@@ -1434,7 +1515,7 @@ extern "C" {
 #define GL_UNIFORM_BLOCK_REFERENCED_BY_VERTEX_SHADER 0x8A44
 #define GL_UNIFORM_BLOCK_REFERENCED_BY_GEOMETRY_SHADER 0x8A45
 #define GL_UNIFORM_BLOCK_REFERENCED_BY_FRAGMENT_SHADER 0x8A46
-#define GL_INVALID_INDEX_ARB              0xFFFFFFFFu
+#define GL_INVALID_INDEX                  0xFFFFFFFFu
 #endif
 
 #ifndef GL_ARB_compatibility
@@ -1446,6 +1527,9 @@ extern "C" {
 #define GL_COPY_WRITE_BUFFER              0x8F37
 #endif
 
+#ifndef GL_ARB_shader_texture_lod
+#endif
+
 #ifndef GL_EXT_abgr
 #define GL_ABGR_EXT                       0x8000
 #endif
@@ -3431,7 +3515,7 @@ extern "C" {
 #endif
 
 #ifndef GL_EXT_blend_equation_separate
-#define GL_BLEND_EQUATION_RGB_EXT         GL_BLEND_EQUATION
+#define GL_BLEND_EQUATION_RGB_EXT         0x8009
 #define GL_BLEND_EQUATION_ALPHA_EXT       0x883D
 #endif
 
@@ -3968,7 +4052,7 @@ extern "C" {
 #include <stddef.h>
 #ifndef GL_VERSION_2_0
 /* GL type for program/shader text */
-typedef char GLchar;			/* native character */
+typedef char GLchar;
 #endif
 
 #ifndef GL_VERSION_1_5
@@ -3984,12 +4068,12 @@ typedef ptrdiff_t GLsizeiptrARB;
 #endif
 
 #ifndef GL_ARB_shader_objects
-/* GL types for handling shader object handles and program/shader text */
-typedef char GLcharARB;		/* native character */
-typedef unsigned int GLhandleARB;	/* shader object handle */
+/* GL types for program/shader text and shader object handles */
+typedef char GLcharARB;
+typedef unsigned int GLhandleARB;
 #endif
 
-/* GL types for "half" precision (s10e5) float data in host memory */
+/* GL type for "half" precision (s10e5) float data in host memory */
 #ifndef GL_ARB_half_float_pixel
 typedef unsigned short GLhalfARB;
 #endif
@@ -4031,7 +4115,8 @@ typedef __int32 int32_t;
 typedef __int64 int64_t;
 typedef unsigned __int64 uint64_t;
 #else
-#include <inttypes.h>     /* Fallback option */
+/* Fallback if nothing above works */
+#include <inttypes.h>
 #endif
 #endif
 
@@ -4046,6 +4131,21 @@ typedef uint64_t GLuint64EXT;
 GLAPI void APIENTRY glBlendColor (GLclampf, GLclampf, GLclampf, GLclampf);
 GLAPI void APIENTRY glBlendEquation (GLenum);
 GLAPI void APIENTRY glDrawRangeElements (GLenum, GLuint, GLuint, GLsizei, GLenum, const GLvoid *);
+GLAPI void APIENTRY glTexImage3D (GLenum, GLint, GLint, GLsizei, GLsizei, GLsizei, GLint, GLenum, GLenum, const GLvoid *);
+GLAPI void APIENTRY glTexSubImage3D (GLenum, GLint, GLint, GLint, GLint, GLsizei, GLsizei, GLsizei, GLenum, GLenum, const GLvoid *);
+GLAPI void APIENTRY glCopyTexSubImage3D (GLenum, GLint, GLint, GLint, GLint, GLint, GLint, GLsizei, GLsizei);
+#endif /* GL_GLEXT_PROTOTYPES */
+typedef void (APIENTRYP PFNGLBLENDCOLORPROC) (GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha);
+typedef void (APIENTRYP PFNGLBLENDEQUATIONPROC) (GLenum mode);
+typedef void (APIENTRYP PFNGLDRAWRANGEELEMENTSPROC) (GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const GLvoid *indices);
+typedef void (APIENTRYP PFNGLTEXIMAGE3DPROC) (GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLenum format, GLenum type, const GLvoid *pixels);
+typedef void (APIENTRYP PFNGLTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const GLvoid *pixels);
+typedef void (APIENTRYP PFNGLCOPYTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+#endif
+
+#ifndef GL_VERSION_1_2_DEPRECATED
+#define GL_VERSION_1_2_DEPRECATED 1
+#ifdef GL_GLEXT_PROTOTYPES
 GLAPI void APIENTRY glColorTable (GLenum, GLenum, GLsizei, GLenum, GLenum, const GLvoid *);
 GLAPI void APIENTRY glColorTableParameterfv (GLenum, GLenum, const GLfloat *);
 GLAPI void APIENTRY glColorTableParameteriv (GLenum, GLenum, const GLint *);
@@ -4078,13 +4178,7 @@ GLAPI void APIENTRY glHistogram (GLenum, GLsizei, GLenum, GLboolean);
 GLAPI void APIENTRY glMinmax (GLenum, GLenum, GLboolean);
 GLAPI void APIENTRY glResetHistogram (GLenum);
 GLAPI void APIENTRY glResetMinmax (GLenum);
-GLAPI void APIENTRY glTexImage3D (GLenum, GLint, GLint, GLsizei, GLsizei, GLsizei, GLint, GLenum, GLenum, const GLvoid *);
-GLAPI void APIENTRY glTexSubImage3D (GLenum, GLint, GLint, GLint, GLint, GLsizei, GLsizei, GLsizei, GLenum, GLenum, const GLvoid *);
-GLAPI void APIENTRY glCopyTexSubImage3D (GLenum, GLint, GLint, GLint, GLint, GLint, GLint, GLsizei, GLsizei);
 #endif /* GL_GLEXT_PROTOTYPES */
-typedef void (APIENTRYP PFNGLBLENDCOLORPROC) (GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha);
-typedef void (APIENTRYP PFNGLBLENDEQUATIONPROC) (GLenum mode);
-typedef void (APIENTRYP PFNGLDRAWRANGEELEMENTSPROC) (GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const GLvoid *indices);
 typedef void (APIENTRYP PFNGLCOLORTABLEPROC) (GLenum target, GLenum internalformat, GLsizei width, GLenum format, GLenum type, const GLvoid *table);
 typedef void (APIENTRYP PFNGLCOLORTABLEPARAMETERFVPROC) (GLenum target, GLenum pname, const GLfloat *params);
 typedef void (APIENTRYP PFNGLCOLORTABLEPARAMETERIVPROC) (GLenum target, GLenum pname, const GLint *params);
@@ -4117,15 +4211,35 @@ typedef void (APIENTRYP PFNGLHISTOGRAMPROC) (GLenum target, GLsizei width, GLenu
 typedef void (APIENTRYP PFNGLMINMAXPROC) (GLenum target, GLenum internalformat, GLboolean sink);
 typedef void (APIENTRYP PFNGLRESETHISTOGRAMPROC) (GLenum target);
 typedef void (APIENTRYP PFNGLRESETMINMAXPROC) (GLenum target);
-typedef void (APIENTRYP PFNGLTEXIMAGE3DPROC) (GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLenum format, GLenum type, const GLvoid *pixels);
-typedef void (APIENTRYP PFNGLTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const GLvoid *pixels);
-typedef void (APIENTRYP PFNGLCOPYTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height);
 #endif
 
 #ifndef GL_VERSION_1_3
 #define GL_VERSION_1_3 1
 #ifdef GL_GLEXT_PROTOTYPES
 GLAPI void APIENTRY glActiveTexture (GLenum);
+GLAPI void APIENTRY glSampleCoverage (GLclampf, GLboolean);
+GLAPI void APIENTRY glCompressedTexImage3D (GLenum, GLint, GLenum, GLsizei, GLsizei, GLsizei, GLint, GLsizei, const GLvoid *);
+GLAPI void APIENTRY glCompressedTexImage2D (GLenum, GLint, GLenum, GLsizei, GLsizei, GLint, GLsizei, const GLvoid *);
+GLAPI void APIENTRY glCompressedTexImage1D (GLenum, GLint, GLenum, GLsizei, GLint, GLsizei, const GLvoid *);
+GLAPI void APIENTRY glCompressedTexSubImage3D (GLenum, GLint, GLint, GLint, GLint, GLsizei, GLsizei, GLsizei, GLenum, GLsizei, const GLvoid *);
+GLAPI void APIENTRY glCompressedTexSubImage2D (GLenum, GLint, GLint, GLint, GLsizei, GLsizei, GLenum, GLsizei, const GLvoid *);
+GLAPI void APIENTRY glCompressedTexSubImage1D (GLenum, GLint, GLint, GLsizei, GLenum, GLsizei, const GLvoid *);
+GLAPI void APIENTRY glGetCompressedTexImage (GLenum, GLint, GLvoid *);
+#endif /* GL_GLEXT_PROTOTYPES */
+typedef void (APIENTRYP PFNGLACTIVETEXTUREPROC) (GLenum texture);
+typedef void (APIENTRYP PFNGLSAMPLECOVERAGEPROC) (GLclampf value, GLboolean invert);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE3DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE2DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE1DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC) (GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const GLvoid *data);
+typedef void (APIENTRYP PFNGLGETCOMPRESSEDTEXIMAGEPROC) (GLenum target, GLint level, GLvoid *img);
+#endif
+
+#ifndef GL_VERSION_1_3_DEPRECATED
+#define GL_VERSION_1_3_DEPRECATED 1
+#ifdef GL_GLEXT_PROTOTYPES
 GLAPI void APIENTRY glClientActiveTexture (GLenum);
 GLAPI void APIENTRY glMultiTexCoord1d (GLenum, GLdouble);
 GLAPI void APIENTRY glMultiTexCoord1dv (GLenum, const GLdouble *);
@@ -4163,16 +4277,7 @@ GLAPI void APIENTRY glLoadTransposeMatrixf (const GLfloat *);
 GLAPI void APIENTRY glLoadTransposeMatrixd (const GLdouble *);
 GLAPI void APIENTRY glMultTransposeMatrixf (const GLfloat *);
 GLAPI void APIENTRY glMultTransposeMatrixd (const GLdouble *);
-GLAPI void APIENTRY glSampleCoverage (GLclampf, GLboolean);
-GLAPI void APIENTRY glCompressedTexImage3D (GLenum, GLint, GLenum, GLsizei, GLsizei, GLsizei, GLint, GLsizei, const GLvoid *);
-GLAPI void APIENTRY glCompressedTexImage2D (GLenum, GLint, GLenum, GLsizei, GLsizei, GLint, GLsizei, const GLvoid *);
-GLAPI void APIENTRY glCompressedTexImage1D (GLenum, GLint, GLenum, GLsizei, GLint, GLsizei, const GLvoid *);
-GLAPI void APIENTRY glCompressedTexSubImage3D (GLenum, GLint, GLint, GLint, GLint, GLsizei, GLsizei, GLsizei, GLenum, GLsizei, const GLvoid *);
-GLAPI void APIENTRY glCompressedTexSubImage2D (GLenum, GLint, GLint, GLint, GLsizei, GLsizei, GLenum, GLsizei, const GLvoid *);
-GLAPI void APIENTRY glCompressedTexSubImage1D (GLenum, GLint, GLint, GLsizei, GLenum, GLsizei, const GLvoid *);
-GLAPI void APIENTRY glGetCompressedTexImage (GLenum, GLint, GLvoid *);
 #endif /* GL_GLEXT_PROTOTYPES */
-typedef void (APIENTRYP PFNGLACTIVETEXTUREPROC) (GLenum texture);
 typedef void (APIENTRYP PFNGLCLIENTACTIVETEXTUREPROC) (GLenum texture);
 typedef void (APIENTRYP PFNGLMULTITEXCOORD1DPROC) (GLenum target, GLdouble s);
 typedef void (APIENTRYP PFNGLMULTITEXCOORD1DVPROC) (GLenum target, const GLdouble *v);
@@ -4210,31 +4315,36 @@ typedef void (APIENTRYP PFNGLLOADTRANSPOSEMATRIXFPROC) (const GLfloat *m);
 typedef void (APIENTRYP PFNGLLOADTRANSPOSEMATRIXDPROC) (const GLdouble *m);
 typedef void (APIENTRYP PFNGLMULTTRANSPOSEMATRIXFPROC) (const GLfloat *m);
 typedef void (APIENTRYP PFNGLMULTTRANSPOSEMATRIXDPROC) (const GLdouble *m);
-typedef void (APIENTRYP PFNGLSAMPLECOVERAGEPROC) (GLclampf value, GLboolean invert);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE3DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE2DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXIMAGE1DPROC) (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC) (GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const GLvoid *data);
-typedef void (APIENTRYP PFNGLGETCOMPRESSEDTEXIMAGEPROC) (GLenum target, GLint level, GLvoid *img);
 #endif
 
 #ifndef GL_VERSION_1_4
 #define GL_VERSION_1_4 1
 #ifdef GL_GLEXT_PROTOTYPES
 GLAPI void APIENTRY glBlendFuncSeparate (GLenum, GLenum, GLenum, GLenum);
-GLAPI void APIENTRY glFogCoordf (GLfloat);
-GLAPI void APIENTRY glFogCoordfv (const GLfloat *);
-GLAPI void APIENTRY glFogCoordd (GLdouble);
-GLAPI void APIENTRY glFogCoorddv (const GLdouble *);
-GLAPI void APIENTRY glFogCoordPointer (GLenum, GLsizei, const GLvoid *);
 GLAPI void APIENTRY glMultiDrawArrays (GLenum, GLint *, GLsizei *, GLsizei);
 GLAPI void APIENTRY glMultiDrawElements (GLenum, const GLsizei *, GLenum, const GLvoid* *, GLsizei);
 GLAPI void APIENTRY glPointParameterf (GLenum, GLfloat);
 GLAPI void APIENTRY glPointParameterfv (GLenum, const GLfloat *);
 GLAPI void APIENTRY glPointParameteri (GLenum, GLint);
 GLAPI void APIENTRY glPointParameteriv (GLenum, const GLint *);
+#endif /* GL_GLEXT_PROTOTYPES */
+typedef void (APIENTRYP PFNGLBLENDFUNCSEPARATEPROC) (GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+typedef void (APIENTRYP PFNGLMULTIDRAWARRAYSPROC) (GLenum mode, GLint *first, GLsizei *count, GLsizei primcount);
+typedef void (APIENTRYP PFNGLMULTIDRAWELEMENTSPROC) (GLenum mode, const GLsizei *count, GLenum type, const GLvoid* *indices, GLsizei primcount);
+typedef void (APIENTRYP PFNGLPOINTPARAMETERFPROC) (GLenum pname, GLfloat param);
+typedef void (APIENTRYP PFNGLPOINTPARAMETERFVPROC) (GLenum pname, const GLfloat *params);
+typedef void (APIENTRYP PFNGLPOINTPARAMETERIPROC) (GLenum pname, GLint param);
+typedef void (APIENTRYP PFNGLPOINTPARAMETERIVPROC) (GLenum pname, const GLint *params);
+#endif
+
+#ifndef GL_VERSION_1_4_DEPRECATED
+#define GL_VERSION_1_4_DEPRECATED 1
+#ifdef GL_GLEXT_PROTOTYPES
+GLAPI void APIENTRY glFogCoordf (GLfloat);
+GLAPI void APIENTRY glFogCoordfv (const GLfloat *);
+GLAPI void APIENTRY glFogCoordd (GLdouble);
+GLAPI void APIENTRY glFogCoorddv (const GLdouble *);
+GLAPI void APIENTRY glFogCoordPointer (GLenum, GLsizei, const GLvoid *);
 GLAPI void APIENTRY glSecondaryColor3b (GLbyte, GLbyte, GLbyte);
 GLAPI void APIENTRY glSecondaryColor3bv (const GLbyte *);
 GLAPI void APIENTRY glSecondaryColor3d (GLdouble, GLdouble, GLdouble);
@@ -4269,18 +4379,11 @@ GLAPI void APIENTRY glWindowPos3iv (const GLint *);
 GLAPI void APIENTRY glWindowPos3s (GLshort, GLshort, GLshort);
 GLAPI void APIENTRY glWindowPos3sv (const GLshort *);
 #endif /* GL_GLEXT_PROTOTYPES */
-typedef void (APIENTRYP PFNGLBLENDFUNCSEPARATEPROC) (GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
 typedef void (APIENTRYP PFNGLFOGCOORDFPROC) (GLfloat coord);
 typedef void (APIENTRYP PFNGLFOGCOORDFVPROC) (const GLfloat *coord);
 typedef void (APIENTRYP PFNGLFOGCOORDDPROC) (GLdouble coord);
 typedef void (APIENTRYP PFNGLFOGCOORDDVPROC) (const GLdouble *coord);
 typedef void (APIENTRYP PFNGLFOGCOORDPOINTERPROC) (GLenum type, GLsizei stride, const GLvoid *pointer);
-typedef void (APIENTRYP PFNGLMULTIDRAWARRAYSPROC) (GLenum mode, GLint *first, GLsizei *count, GLsizei primcount);
-typedef void (APIENTRYP PFNGLMULTIDRAWELEMENTSPROC) (GLenum mode, const GLsizei *count, GLenum type, const GLvoid* *indices, GLsizei primcount);
-typedef void (APIENTRYP PFNGLPOINTPARAMETERFPROC) (GLenum pname, GLfloat param);
-typedef void (APIENTRYP PFNGLPOINTPARAMETERFVPROC) (GLenum pname, const GLfloat *params);
-typedef void (APIENTRYP PFNGLPOINTPARAMETERIPROC) (GLenum pname, GLint param);
-typedef void (APIENTRYP PFNGLPOINTPARAMETERIVPROC) (GLenum pname, const GLint *params);
 typedef void (APIENTRYP PFNGLSECONDARYCOLOR3BPROC) (GLbyte red, GLbyte green, GLbyte blue);
 typedef void (APIENTRYP PFNGLSECONDARYCOLOR3BVPROC) (const GLbyte *v);
 typedef void (APIENTRYP PFNGLSECONDARYCOLOR3DPROC) (GLdouble red, GLdouble green, GLdouble blue);
@@ -4592,26 +4695,6 @@ GLAPI void APIENTRY glGetTransformFeedbackVarying (GLuint, GLuint, GLsizei, GLsi
 GLAPI void APIENTRY glClampColor (GLenum, GLenum);
 GLAPI void APIENTRY glBeginConditionalRender (GLuint, GLenum);
 GLAPI void APIENTRY glEndConditionalRender (void);
-GLAPI void APIENTRY glVertexAttribI1i (GLuint, GLint);
-GLAPI void APIENTRY glVertexAttribI2i (GLuint, GLint, GLint);
-GLAPI void APIENTRY glVertexAttribI3i (GLuint, GLint, GLint, GLint);
-GLAPI void APIENTRY glVertexAttribI4i (GLuint, GLint, GLint, GLint, GLint);
-GLAPI void APIENTRY glVertexAttribI1ui (GLuint, GLuint);
-GLAPI void APIENTRY glVertexAttribI2ui (GLuint, GLuint, GLuint);
-GLAPI void APIENTRY glVertexAttribI3ui (GLuint, GLuint, GLuint, GLuint);
-GLAPI void APIENTRY glVertexAttribI4ui (GLuint, GLuint, GLuint, GLuint, GLuint);
-GLAPI void APIENTRY glVertexAttribI1iv (GLuint, const GLint *);
-GLAPI void APIENTRY glVertexAttribI2iv (GLuint, const GLint *);
-GLAPI void APIENTRY glVertexAttribI3iv (GLuint, const GLint *);
-GLAPI void APIENTRY glVertexAttribI4iv (GLuint, const GLint *);
-GLAPI void APIENTRY glVertexAttribI1uiv (GLuint, const GLuint *);
-GLAPI void APIENTRY glVertexAttribI2uiv (GLuint, const GLuint *);
-GLAPI void APIENTRY glVertexAttribI3uiv (GLuint, const GLuint *);
-GLAPI void APIENTRY glVertexAttribI4uiv (GLuint, const GLuint *);
-GLAPI void APIENTRY glVertexAttribI4bv (GLuint, const GLbyte *);
-GLAPI void APIENTRY glVertexAttribI4sv (GLuint, const GLshort *);
-GLAPI void APIENTRY glVertexAttribI4ubv (GLuint, const GLubyte *);
-GLAPI void APIENTRY glVertexAttribI4usv (GLuint, const GLushort *);
 GLAPI void APIENTRY glVertexAttribIPointer (GLuint, GLint, GLenum, GLsizei, const GLvoid *);
 GLAPI void APIENTRY glGetVertexAttribIiv (GLuint, GLenum, GLint *);
 GLAPI void APIENTRY glGetVertexAttribIuiv (GLuint, GLenum, GLuint *);
@@ -4651,26 +4734,6 @@ typedef void (APIENTRYP PFNGLGETTRANSFORMFEEDBACKVARYINGPROC) (GLuint program, G
 typedef void (APIENTRYP PFNGLCLAMPCOLORPROC) (GLenum target, GLenum clamp);
 typedef void (APIENTRYP PFNGLBEGINCONDITIONALRENDERPROC) (GLuint id, GLenum mode);
 typedef void (APIENTRYP PFNGLENDCONDITIONALRENDERPROC) (void);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI1IPROC) (GLuint index, GLint x);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI2IPROC) (GLuint index, GLint x, GLint y);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI3IPROC) (GLuint index, GLint x, GLint y, GLint z);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4IPROC) (GLuint index, GLint x, GLint y, GLint z, GLint w);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI1UIPROC) (GLuint index, GLuint x);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI2UIPROC) (GLuint index, GLuint x, GLuint y);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI3UIPROC) (GLuint index, GLuint x, GLuint y, GLuint z);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4UIPROC) (GLuint index, GLuint x, GLuint y, GLuint z, GLuint w);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI1IVPROC) (GLuint index, const GLint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI2IVPROC) (GLuint index, const GLint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI3IVPROC) (GLuint index, const GLint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4IVPROC) (GLuint index, const GLint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI1UIVPROC) (GLuint index, const GLuint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI2UIVPROC) (GLuint index, const GLuint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI3UIVPROC) (GLuint index, const GLuint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4UIVPROC) (GLuint index, const GLuint *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4BVPROC) (GLuint index, const GLbyte *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4SVPROC) (GLuint index, const GLshort *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4UBVPROC) (GLuint index, const GLubyte *v);
-typedef void (APIENTRYP PFNGLVERTEXATTRIBI4USVPROC) (GLuint index, const GLushort *v);
 typedef void (APIENTRYP PFNGLVERTEXATTRIBIPOINTERPROC) (GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid *pointer);
 typedef void (APIENTRYP PFNGLGETVERTEXATTRIBIIVPROC) (GLuint index, GLenum pname, GLint *params);
 typedef void (APIENTRYP PFNGLGETVERTEXATTRIBIUIVPROC) (GLuint index, GLenum pname, GLuint *params);
@@ -4696,6 +4759,69 @@ typedef void (APIENTRYP PFNGLCLEARBUFFERFIPROC) (GLenum buffer, GLint drawbuffer
 typedef const GLubyte * (APIENTRYP PFNGLGETSTRINGIPROC) (GLenum name, GLuint index);
 #endif
 
+#ifndef GL_VERSION_3_0_DEPRECATED
+#define GL_VERSION_3_0_DEPRECATED 1
+#ifdef GL_GLEXT_PROTOTYPES
+GLAPI void APIENTRY glVertexAttribI1i (GLuint, GLint);
+GLAPI void APIENTRY glVertexAttribI2i (GLuint, GLint, GLint);
+GLAPI void APIENTRY glVertexAttribI3i (GLuint, GLint, GLint, GLint);
+GLAPI void APIENTRY glVertexAttribI4i (GLuint, GLint, GLint, GLint, GLint);
+GLAPI void APIENTRY glVertexAttribI1ui (GLuint, GLuint);
+GLAPI void APIENTRY glVertexAttribI2ui (GLuint, GLuint, GLuint);
+GLAPI void APIENTRY glVertexAttribI3ui (GLuint, GLuint, GLuint, GLuint);
+GLAPI void APIENTRY glVertexAttribI4ui (GLuint, GLuint, GLuint, GLuint, GLuint);
+GLAPI void APIENTRY glVertexAttribI1iv (GLuint, const GLint *);
+GLAPI void APIENTRY glVertexAttribI2iv (GLuint, const GLint *);
+GLAPI void APIENTRY glVertexAttribI3iv (GLuint, const GLint *);
+GLAPI void APIENTRY glVertexAttribI4iv (GLuint, const GLint *);
+GLAPI void APIENTRY glVertexAttribI1uiv (GLuint, const GLuint *);
+GLAPI void APIENTRY glVertexAttribI2uiv (GLuint, const GLuint *);
+GLAPI void APIENTRY glVertexAttribI3uiv (GLuint, const GLuint *);
+GLAPI void APIENTRY glVertexAttribI4uiv (GLuint, const GLuint *);
+GLAPI void APIENTRY glVertexAttribI4bv (GLuint, const GLbyte *);
+GLAPI void APIENTRY glVertexAttribI4sv (GLuint, const GLshort *);
+GLAPI void APIENTRY glVertexAttribI4ubv (GLuint, const GLubyte *);
+GLAPI void APIENTRY glVertexAttribI4usv (GLuint, const GLushort *);
+#endif /* GL_GLEXT_PROTOTYPES */
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI1IPROC) (GLuint index, GLint x);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI2IPROC) (GLuint index, GLint x, GLint y);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI3IPROC) (GLuint index, GLint x, GLint y, GLint z);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4IPROC) (GLuint index, GLint x, GLint y, GLint z, GLint w);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI1UIPROC) (GLuint index, GLuint x);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI2UIPROC) (GLuint index, GLuint x, GLuint y);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI3UIPROC) (GLuint index, GLuint x, GLuint y, GLuint z);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4UIPROC) (GLuint index, GLuint x, GLuint y, GLuint z, GLuint w);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI1IVPROC) (GLuint index, const GLint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI2IVPROC) (GLuint index, const GLint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI3IVPROC) (GLuint index, const GLint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4IVPROC) (GLuint index, const GLint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI1UIVPROC) (GLuint index, const GLuint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI2UIVPROC) (GLuint index, const GLuint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI3UIVPROC) (GLuint index, const GLuint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4UIVPROC) (GLuint index, const GLuint *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4BVPROC) (GLuint index, const GLbyte *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4SVPROC) (GLuint index, const GLshort *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4UBVPROC) (GLuint index, const GLubyte *v);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBI4USVPROC) (GLuint index, const GLushort *v);
+#endif
+
+#ifndef GL_VERSION_3_1
+#define GL_VERSION_3_1 1
+/* OpenGL 3.1 also reuses entry points from these extensions: */
+/* ARB_copy_buffer */
+/* ARB_uniform_buffer_object */
+#ifdef GL_GLEXT_PROTOTYPES
+GLAPI void APIENTRY glDrawArraysInstanced (GLenum, GLint, GLsizei, GLsizei);
+GLAPI void APIENTRY glDrawElementsInstanced (GLenum, GLsizei, GLenum, const GLvoid *, GLsizei);
+GLAPI void APIENTRY glTexBuffer (GLenum, GLenum, GLuint);
+GLAPI void APIENTRY glPrimitiveRestartIndex (GLuint);
+#endif /* GL_GLEXT_PROTOTYPES */
+typedef void (APIENTRYP PFNGLDRAWARRAYSINSTANCEDPROC) (GLenum mode, GLint first, GLsizei count, GLsizei primcount);
+typedef void (APIENTRYP PFNGLDRAWELEMENTSINSTANCEDPROC) (GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, GLsizei primcount);
+typedef void (APIENTRYP PFNGLTEXBUFFERPROC) (GLenum target, GLenum internalformat, GLuint buffer);
+typedef void (APIENTRYP PFNGLPRIMITIVERESTARTINDEXPROC) (GLuint index);
+#endif
+
 #ifndef GL_ARB_multitexture
 #define GL_ARB_multitexture 1
 #ifdef GL_GLEXT_PROTOTYPES
@@ -5360,9 +5486,9 @@ typedef void (APIENTRYP PFNGLFRAMEBUFFERTEXTUREFACEARBPROC) (GLenum target, GLen
 #ifndef GL_ARB_instanced_arrays
 #define GL_ARB_instanced_arrays 1
 #ifdef GL_GLEXT_PROTOTYPES
-GLAPI void APIENTRY glVertexAttribDivisor (GLuint, GLuint);
+GLAPI void APIENTRY glVertexAttribDivisorARB (GLuint, GLuint);
 #endif /* GL_GLEXT_PROTOTYPES */
-typedef void (APIENTRYP PFNGLVERTEXATTRIBDIVISORPROC) (GLuint index, GLuint divisor);
+typedef void (APIENTRYP PFNGLVERTEXATTRIBDIVISORARBPROC) (GLuint index, GLuint divisor);
 #endif
 
 #ifndef GL_ARB_map_buffer_range
@@ -5437,6 +5563,10 @@ GLAPI void APIENTRY glCopyBufferSubData (GLenum, GLenum, GLintptr, GLintptr, GLs
 typedef void (APIENTRYP PFNGLCOPYBUFFERSUBDATAPROC) (GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size);
 #endif
 
+#ifndef GL_ARB_shader_texture_lod
+#define GL_ARB_shader_texture_lod 1
+#endif
+
 #ifndef GL_EXT_abgr
 #define GL_EXT_abgr 1
 #endif
@@ -8150,7 +8280,6 @@ GLAPI void APIENTRY glGetVideoivNV (GLuint, GLenum, GLint *);
 GLAPI void APIENTRY glGetVideouivNV (GLuint, GLenum, GLuint *);
 GLAPI void APIENTRY glGetVideoi64vNV (GLuint, GLenum, GLint64EXT *);
 GLAPI void APIENTRY glGetVideoui64vNV (GLuint, GLenum, GLuint64EXT *);
-GLAPI void APIENTRY glVideoParameterivNV (GLuint, GLenum, const GLint *);
 #endif /* GL_GLEXT_PROTOTYPES */
 typedef void (APIENTRYP PFNGLPRESENTFRAMEKEYEDNVPROC) (GLuint video_slot, GLuint64EXT minPresentTime, GLuint beginPresentTimeId, GLuint presentDurationId, GLenum type, GLenum target0, GLuint fill0, GLuint key0, GLenum target1, GLuint fill1, GLuint key1);
 typedef void (APIENTRYP PFNGLPRESENTFRAMEDUALFILLNVPROC) (GLuint video_slot, GLuint64EXT minPresentTime, GLuint beginPresentTimeId, GLuint presentDurationId, GLenum type, GLenum target0, GLuint fill0, GLenum target1, GLuint fill1, GLenum target2, GLuint fill2, GLenum target3, GLuint fill3);
@@ -8158,7 +8287,6 @@ typedef void (APIENTRYP PFNGLGETVIDEOIVNVPROC) (GLuint video_slot, GLenum pname,
 typedef void (APIENTRYP PFNGLGETVIDEOUIVNVPROC) (GLuint video_slot, GLenum pname, GLuint *params);
 typedef void (APIENTRYP PFNGLGETVIDEOI64VNVPROC) (GLuint video_slot, GLenum pname, GLint64EXT *params);
 typedef void (APIENTRYP PFNGLGETVIDEOUI64VNVPROC) (GLuint video_slot, GLenum pname, GLuint64EXT *params);
-typedef void (APIENTRYP PFNGLVIDEOPARAMETERIVNVPROC) (GLuint video_slot, GLenum pname, const GLint *params);
 #endif
 
 #ifndef GL_EXT_transform_feedback
diff --git a/include/GL/wglext.h b/include/GL/wglext.h
index d54e0a4d50..3c492dd634 100644
--- a/include/GL/wglext.h
+++ b/include/GL/wglext.h
@@ -6,7 +6,7 @@ extern "C" {
 #endif
 
 /*
-** Copyright (c) 2007 The Khronos Group Inc.
+** Copyright (c) 2007-2009 The Khronos Group Inc.
 ** 
 ** Permission is hereby granted, free of charge, to any person obtaining a
 ** copy of this software and/or associated documentation files (the
@@ -28,6 +28,8 @@ extern "C" {
 ** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
 */
 
+/* Function declaration macros - to move into glplatform.h */
+
 #if defined(_WIN32) && !defined(APIENTRY) && !defined(__CYGWIN__) && !defined(__SCITECH_SNAP__)
 #define WIN32_LEAN_AND_MEAN 1
 #include <windows.h>
@@ -709,6 +711,14 @@ typedef BOOL (WINAPI * PFNWGLQUERYFRAMETRACKINGI3DPROC) (DWORD *pFrameCount, DWO
 #define WGL_NV_float_buffer 1
 #endif
 
+#ifndef WGL_3DL_stereo_control
+#define WGL_3DL_stereo_control 1
+#ifdef WGL_WGLEXT_PROTOTYPES
+extern BOOL WINAPI wglSetStereoEmitterState3DL (HDC, UINT);
+#endif /* WGL_WGLEXT_PROTOTYPES */
+typedef BOOL (WINAPI * PFNWGLSETSTEREOEMITTERSTATE3DLPROC) (HDC hDC, UINT uState);
+#endif
+
 #ifndef WGL_EXT_pixel_format_packed_float
 #define WGL_EXT_pixel_format_packed_float 1
 #endif
diff --git a/include/VG/openvg.h b/include/VG/openvg.h
new file mode 100644
index 0000000000..60167e45d6
--- /dev/null
+++ b/include/VG/openvg.h
@@ -0,0 +1,686 @@
+/* $Revision: 6822 $ on $Date:: 2008-10-30 05:14:19 -0400 #$ */
+
+/*------------------------------------------------------------------------
+ *
+ * OpenVG 1.0.1 Reference Implementation
+ * -------------------------------------
+ *
+ * Copyright (c) 2008 The Khronos Group Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and /or associated documentation files
+ * (the "Materials "), to deal in the Materials without restriction,
+ * including without limitation the rights to use, copy, modify, merge,
+ * publish, distribute, sublicense, and/or sell copies of the Materials,
+ * and to permit persons to whom the Materials are furnished to do so,
+ * subject to the following conditions: 
+ *
+ * The above copyright notice and this permission notice shall be included 
+ * in all copies or substantial portions of the Materials. 
+ *
+ * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE MATERIALS OR
+ * THE USE OR OTHER DEALINGS IN THE MATERIALS.
+ *
+ *//**
+ * \file
+ * \brief	OpenVG 1.0.1 API.
+ *//*-------------------------------------------------------------------*/
+
+#ifndef _OPENVG_H
+#define _OPENVG_H
+
+#include <VG/vgplatform.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define OPENVG_VERSION_1_0		1
+#define OPENVG_VERSION_1_0_1	1
+
+#ifndef VG_MAXSHORT
+#define VG_MAXSHORT 0x7FFF
+#endif
+
+#ifndef VG_MAXINT
+#define VG_MAXINT 0x7FFFFFFF
+#endif
+
+#ifndef VG_MAX_ENUM
+#define VG_MAX_ENUM 0x7FFFFFFF
+#endif
+
+typedef long VGHandle;
+
+typedef VGHandle VGPath;
+typedef VGHandle VGImage;
+typedef VGHandle VGPaint;
+
+#define VG_INVALID_HANDLE ((VGHandle)0)
+
+typedef enum {
+  VG_FALSE               = 0,
+  VG_TRUE                = 1,
+
+  VG_BOOLEAN_FORCE_SIZE  = VG_MAX_ENUM
+} VGboolean;
+
+typedef enum {
+  VG_NO_ERROR                                 = 0,
+  VG_BAD_HANDLE_ERROR                         = 0x1000,
+  VG_ILLEGAL_ARGUMENT_ERROR                   = 0x1001,
+  VG_OUT_OF_MEMORY_ERROR                      = 0x1002,
+  VG_PATH_CAPABILITY_ERROR                    = 0x1003,
+  VG_UNSUPPORTED_IMAGE_FORMAT_ERROR           = 0x1004,
+  VG_UNSUPPORTED_PATH_FORMAT_ERROR            = 0x1005,
+  VG_IMAGE_IN_USE_ERROR                       = 0x1006,
+  VG_NO_CONTEXT_ERROR                         = 0x1007,
+
+  VG_ERROR_CODE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGErrorCode;
+
+typedef enum {
+  /* Mode settings */
+  VG_MATRIX_MODE                              = 0x1100,
+  VG_FILL_RULE                                = 0x1101,
+  VG_IMAGE_QUALITY                            = 0x1102,
+  VG_RENDERING_QUALITY                        = 0x1103,
+  VG_BLEND_MODE                               = 0x1104,
+  VG_IMAGE_MODE                               = 0x1105,
+
+  /* Scissoring rectangles */
+  VG_SCISSOR_RECTS                            = 0x1106,
+
+  /* Stroke parameters */
+  VG_STROKE_LINE_WIDTH                        = 0x1110,
+  VG_STROKE_CAP_STYLE                         = 0x1111,
+  VG_STROKE_JOIN_STYLE                        = 0x1112,
+  VG_STROKE_MITER_LIMIT                       = 0x1113,
+  VG_STROKE_DASH_PATTERN                      = 0x1114,
+  VG_STROKE_DASH_PHASE                        = 0x1115,
+  VG_STROKE_DASH_PHASE_RESET                  = 0x1116,
+
+  /* Edge fill color for VG_TILE_FILL tiling mode */
+  VG_TILE_FILL_COLOR                          = 0x1120,
+
+  /* Color for vgClear */
+  VG_CLEAR_COLOR                              = 0x1121,
+
+  /* Enable/disable alpha masking and scissoring */
+  VG_MASKING                                  = 0x1130,
+  VG_SCISSORING                               = 0x1131,
+
+  /* Pixel layout information */
+  VG_PIXEL_LAYOUT                             = 0x1140,
+  VG_SCREEN_LAYOUT                            = 0x1141,
+
+  /* Source format selection for image filters */
+  VG_FILTER_FORMAT_LINEAR                     = 0x1150,
+  VG_FILTER_FORMAT_PREMULTIPLIED              = 0x1151,
+
+  /* Destination write enable mask for image filters */
+  VG_FILTER_CHANNEL_MASK                      = 0x1152,
+
+  /* Implementation limits (read-only) */
+  VG_MAX_SCISSOR_RECTS                        = 0x1160,
+  VG_MAX_DASH_COUNT                           = 0x1161,
+  VG_MAX_KERNEL_SIZE                          = 0x1162,
+  VG_MAX_SEPARABLE_KERNEL_SIZE                = 0x1163,
+  VG_MAX_COLOR_RAMP_STOPS                     = 0x1164,
+  VG_MAX_IMAGE_WIDTH                          = 0x1165,
+  VG_MAX_IMAGE_HEIGHT                         = 0x1166,
+  VG_MAX_IMAGE_PIXELS                         = 0x1167,
+  VG_MAX_IMAGE_BYTES                          = 0x1168,
+  VG_MAX_FLOAT                                = 0x1169,
+  VG_MAX_GAUSSIAN_STD_DEVIATION               = 0x116A,
+
+  VG_PARAM_TYPE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGParamType;
+
+typedef enum {
+  VG_RENDERING_QUALITY_NONANTIALIASED         = 0x1200,
+  VG_RENDERING_QUALITY_FASTER                 = 0x1201,
+  VG_RENDERING_QUALITY_BETTER                 = 0x1202, /* Default */
+
+  VG_RENDERING_QUALITY_FORCE_SIZE             = VG_MAX_ENUM
+} VGRenderingQuality;
+
+typedef enum {
+  VG_PIXEL_LAYOUT_UNKNOWN                     = 0x1300,
+  VG_PIXEL_LAYOUT_RGB_VERTICAL                = 0x1301,
+  VG_PIXEL_LAYOUT_BGR_VERTICAL                = 0x1302,
+  VG_PIXEL_LAYOUT_RGB_HORIZONTAL              = 0x1303,
+  VG_PIXEL_LAYOUT_BGR_HORIZONTAL              = 0x1304,
+
+  VG_PIXEL_LAYOUT_FORCE_SIZE                  = VG_MAX_ENUM
+} VGPixelLayout;
+
+typedef enum {
+  VG_MATRIX_PATH_USER_TO_SURFACE              = 0x1400,
+  VG_MATRIX_IMAGE_USER_TO_SURFACE             = 0x1401,
+  VG_MATRIX_FILL_PAINT_TO_USER                = 0x1402,
+  VG_MATRIX_STROKE_PAINT_TO_USER              = 0x1403,
+
+  VG_MATRIX_MODE_FORCE_SIZE                   = VG_MAX_ENUM
+} VGMatrixMode;
+
+typedef enum {
+  VG_CLEAR_MASK                               = 0x1500,
+  VG_FILL_MASK                                = 0x1501,
+  VG_SET_MASK                                 = 0x1502,
+  VG_UNION_MASK                               = 0x1503,
+  VG_INTERSECT_MASK                           = 0x1504,
+  VG_SUBTRACT_MASK                            = 0x1505,
+
+  VG_MASK_OPERATION_FORCE_SIZE                = VG_MAX_ENUM
+} VGMaskOperation;
+
+#define VG_PATH_FORMAT_STANDARD 0
+
+typedef enum {
+  VG_PATH_DATATYPE_S_8                        =  0,
+  VG_PATH_DATATYPE_S_16                       =  1,
+  VG_PATH_DATATYPE_S_32                       =  2,
+  VG_PATH_DATATYPE_F                          =  3,
+
+  VG_PATH_DATATYPE_FORCE_SIZE                 = VG_MAX_ENUM
+} VGPathDatatype;
+
+typedef enum {
+  VG_ABSOLUTE                                 = 0,
+  VG_RELATIVE                                 = 1,
+
+  VG_PATH_ABS_REL_FORCE_SIZE                  = VG_MAX_ENUM
+} VGPathAbsRel;
+
+typedef enum {
+  VG_CLOSE_PATH                               = ( 0 << 1),
+  VG_MOVE_TO                                  = ( 1 << 1),
+  VG_LINE_TO                                  = ( 2 << 1),
+  VG_HLINE_TO                                 = ( 3 << 1),
+  VG_VLINE_TO                                 = ( 4 << 1),
+  VG_QUAD_TO                                  = ( 5 << 1),
+  VG_CUBIC_TO                                 = ( 6 << 1),
+  VG_SQUAD_TO                                 = ( 7 << 1),
+  VG_SCUBIC_TO                                = ( 8 << 1),
+  VG_SCCWARC_TO                               = ( 9 << 1),
+  VG_SCWARC_TO                                = (10 << 1),
+  VG_LCCWARC_TO                               = (11 << 1),
+  VG_LCWARC_TO                                = (12 << 1),
+
+  VG_PATH_SEGMENT_FORCE_SIZE                  = VG_MAX_ENUM
+} VGPathSegment;
+
+typedef enum {
+  VG_MOVE_TO_ABS                              = VG_MOVE_TO    | VG_ABSOLUTE,
+  VG_MOVE_TO_REL                              = VG_MOVE_TO    | VG_RELATIVE,
+  VG_LINE_TO_ABS                              = VG_LINE_TO    | VG_ABSOLUTE,
+  VG_LINE_TO_REL                              = VG_LINE_TO    | VG_RELATIVE,
+  VG_HLINE_TO_ABS                             = VG_HLINE_TO   | VG_ABSOLUTE,
+  VG_HLINE_TO_REL                             = VG_HLINE_TO   | VG_RELATIVE,
+  VG_VLINE_TO_ABS                             = VG_VLINE_TO   | VG_ABSOLUTE,
+  VG_VLINE_TO_REL                             = VG_VLINE_TO   | VG_RELATIVE,
+  VG_QUAD_TO_ABS                              = VG_QUAD_TO    | VG_ABSOLUTE,
+  VG_QUAD_TO_REL                              = VG_QUAD_TO    | VG_RELATIVE,
+  VG_CUBIC_TO_ABS                             = VG_CUBIC_TO   | VG_ABSOLUTE,
+  VG_CUBIC_TO_REL                             = VG_CUBIC_TO   | VG_RELATIVE,
+  VG_SQUAD_TO_ABS                             = VG_SQUAD_TO   | VG_ABSOLUTE,
+  VG_SQUAD_TO_REL                             = VG_SQUAD_TO   | VG_RELATIVE,
+  VG_SCUBIC_TO_ABS                            = VG_SCUBIC_TO  | VG_ABSOLUTE,
+  VG_SCUBIC_TO_REL                            = VG_SCUBIC_TO  | VG_RELATIVE,
+  VG_SCCWARC_TO_ABS                           = VG_SCCWARC_TO | VG_ABSOLUTE,
+  VG_SCCWARC_TO_REL                           = VG_SCCWARC_TO | VG_RELATIVE,
+  VG_SCWARC_TO_ABS                            = VG_SCWARC_TO  | VG_ABSOLUTE,
+  VG_SCWARC_TO_REL                            = VG_SCWARC_TO  | VG_RELATIVE,
+  VG_LCCWARC_TO_ABS                           = VG_LCCWARC_TO | VG_ABSOLUTE,
+  VG_LCCWARC_TO_REL                           = VG_LCCWARC_TO | VG_RELATIVE,
+  VG_LCWARC_TO_ABS                            = VG_LCWARC_TO  | VG_ABSOLUTE,
+  VG_LCWARC_TO_REL                            = VG_LCWARC_TO  | VG_RELATIVE,
+
+  VG_PATH_COMMAND_FORCE_SIZE                  = VG_MAX_ENUM
+} VGPathCommand;
+
+typedef enum {
+  VG_PATH_CAPABILITY_APPEND_FROM              = (1 <<  0),
+  VG_PATH_CAPABILITY_APPEND_TO                = (1 <<  1),
+  VG_PATH_CAPABILITY_MODIFY                   = (1 <<  2),
+  VG_PATH_CAPABILITY_TRANSFORM_FROM           = (1 <<  3),
+  VG_PATH_CAPABILITY_TRANSFORM_TO             = (1 <<  4),
+  VG_PATH_CAPABILITY_INTERPOLATE_FROM         = (1 <<  5),
+  VG_PATH_CAPABILITY_INTERPOLATE_TO           = (1 <<  6),
+  VG_PATH_CAPABILITY_PATH_LENGTH              = (1 <<  7),
+  VG_PATH_CAPABILITY_POINT_ALONG_PATH         = (1 <<  8),
+  VG_PATH_CAPABILITY_TANGENT_ALONG_PATH       = (1 <<  9),
+  VG_PATH_CAPABILITY_PATH_BOUNDS              = (1 << 10),
+  VG_PATH_CAPABILITY_PATH_TRANSFORMED_BOUNDS  = (1 << 11),
+  VG_PATH_CAPABILITY_ALL                      = (1 << 12) - 1,
+
+  VG_PATH_CAPABILITIES_FORCE_SIZE             = VG_MAX_ENUM
+} VGPathCapabilities;
+
+typedef enum {
+  VG_PATH_FORMAT                              = 0x1600,
+  VG_PATH_DATATYPE                            = 0x1601,
+  VG_PATH_SCALE                               = 0x1602,
+  VG_PATH_BIAS                                = 0x1603,
+  VG_PATH_NUM_SEGMENTS                        = 0x1604,
+  VG_PATH_NUM_COORDS                          = 0x1605,
+
+  VG_PATH_PARAM_TYPE_FORCE_SIZE               = VG_MAX_ENUM
+} VGPathParamType;
+
+typedef enum {
+  VG_CAP_BUTT                                 = 0x1700,
+  VG_CAP_ROUND                                = 0x1701,
+  VG_CAP_SQUARE                               = 0x1702,
+
+  VG_CAP_STYLE_FORCE_SIZE                     = VG_MAX_ENUM
+} VGCapStyle;
+
+typedef enum {
+  VG_JOIN_MITER                               = 0x1800,
+  VG_JOIN_ROUND                               = 0x1801,
+  VG_JOIN_BEVEL                               = 0x1802,
+
+  VG_JOIN_STYLE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGJoinStyle;
+
+typedef enum {
+  VG_EVEN_ODD                                 = 0x1900,
+  VG_NON_ZERO                                 = 0x1901,
+
+  VG_FILL_RULE_FORCE_SIZE                     = VG_MAX_ENUM
+} VGFillRule;
+
+typedef enum {
+  VG_STROKE_PATH                              = (1 << 0),
+  VG_FILL_PATH                                = (1 << 1),
+
+  VG_PAINT_MODE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGPaintMode;
+
+typedef enum {
+  /* Color paint parameters */
+  VG_PAINT_TYPE                               = 0x1A00,
+  VG_PAINT_COLOR                              = 0x1A01,
+  VG_PAINT_COLOR_RAMP_SPREAD_MODE             = 0x1A02,
+  VG_PAINT_COLOR_RAMP_PREMULTIPLIED           = 0x1A07,
+  VG_PAINT_COLOR_RAMP_STOPS                   = 0x1A03,
+
+  /* Linear gradient paint parameters */
+  VG_PAINT_LINEAR_GRADIENT                    = 0x1A04,
+
+  /* Radial gradient paint parameters */
+  VG_PAINT_RADIAL_GRADIENT                    = 0x1A05,
+
+  /* Pattern paint parameters */
+  VG_PAINT_PATTERN_TILING_MODE                = 0x1A06,
+
+  VG_PAINT_PARAM_TYPE_FORCE_SIZE              = VG_MAX_ENUM
+} VGPaintParamType;
+
+typedef enum {
+  VG_PAINT_TYPE_COLOR                         = 0x1B00,
+  VG_PAINT_TYPE_LINEAR_GRADIENT               = 0x1B01,
+  VG_PAINT_TYPE_RADIAL_GRADIENT               = 0x1B02,
+  VG_PAINT_TYPE_PATTERN                       = 0x1B03,
+
+  VG_PAINT_TYPE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGPaintType;
+
+typedef enum {
+  VG_COLOR_RAMP_SPREAD_PAD                    = 0x1C00,
+  VG_COLOR_RAMP_SPREAD_REPEAT                 = 0x1C01,
+  VG_COLOR_RAMP_SPREAD_REFLECT                = 0x1C02,
+
+  VG_COLOR_RAMP_SPREAD_MODE_FORCE_SIZE        = VG_MAX_ENUM
+} VGColorRampSpreadMode;
+
+typedef enum {
+  VG_TILE_FILL                                = 0x1D00,
+  VG_TILE_PAD                                 = 0x1D01,
+  VG_TILE_REPEAT                              = 0x1D02,
+  VG_TILE_REFLECT                             = 0x1D03,
+
+  VG_TILING_MODE_FORCE_SIZE                   = VG_MAX_ENUM
+} VGTilingMode;
+
+typedef enum {
+  /* RGB{A,X} channel ordering */
+  VG_sRGBX_8888                               =  0,
+  VG_sRGBA_8888                               =  1,
+  VG_sRGBA_8888_PRE                           =  2,
+  VG_sRGB_565                                 =  3,
+  VG_sRGBA_5551                               =  4,
+  VG_sRGBA_4444                               =  5,
+  VG_sL_8                                     =  6,
+  VG_lRGBX_8888                               =  7,
+  VG_lRGBA_8888                               =  8,
+  VG_lRGBA_8888_PRE                           =  9,
+  VG_lL_8                                     = 10,
+  VG_A_8                                      = 11,
+  VG_BW_1                                     = 12,
+
+  /* {A,X}RGB channel ordering */
+  VG_sXRGB_8888                               =  0 | (1 << 6),
+  VG_sARGB_8888                               =  1 | (1 << 6),
+  VG_sARGB_8888_PRE                           =  2 | (1 << 6),
+  VG_sARGB_1555                               =  4 | (1 << 6),
+  VG_sARGB_4444                               =  5 | (1 << 6),
+  VG_lXRGB_8888                               =  7 | (1 << 6),
+  VG_lARGB_8888                               =  8 | (1 << 6),
+  VG_lARGB_8888_PRE                           =  9 | (1 << 6),
+
+  /* BGR{A,X} channel ordering */
+  VG_sBGRX_8888                               =  0 | (1 << 7),
+  VG_sBGRA_8888                               =  1 | (1 << 7),
+  VG_sBGRA_8888_PRE                           =  2 | (1 << 7),
+  VG_sBGR_565                                 =  3 | (1 << 7),
+  VG_sBGRA_5551                               =  4 | (1 << 7),
+  VG_sBGRA_4444                               =  5 | (1 << 7),
+  VG_lBGRX_8888                               =  7 | (1 << 7),
+  VG_lBGRA_8888                               =  8 | (1 << 7),
+  VG_lBGRA_8888_PRE                           =  9 | (1 << 7),
+
+  /* {A,X}BGR channel ordering */
+  VG_sXBGR_8888                               =  0 | (1 << 6) | (1 << 7),
+  VG_sABGR_8888                               =  1 | (1 << 6) | (1 << 7),
+  VG_sABGR_8888_PRE                           =  2 | (1 << 6) | (1 << 7),
+  VG_sABGR_1555                               =  4 | (1 << 6) | (1 << 7),
+  VG_sABGR_4444                               =  5 | (1 << 6) | (1 << 7),
+  VG_lXBGR_8888                               =  7 | (1 << 6) | (1 << 7),
+  VG_lABGR_8888                               =  8 | (1 << 6) | (1 << 7),
+  VG_lABGR_8888_PRE                           =  9 | (1 << 6) | (1 << 7),
+
+  VG_IMAGE_FORMAT_FORCE_SIZE                  = VG_MAX_ENUM
+} VGImageFormat;
+
+typedef enum {
+  VG_IMAGE_QUALITY_NONANTIALIASED             = (1 << 0),
+  VG_IMAGE_QUALITY_FASTER                     = (1 << 1),
+  VG_IMAGE_QUALITY_BETTER                     = (1 << 2),
+
+  VG_IMAGE_QUALITY_FORCE_SIZE                 = VG_MAX_ENUM
+} VGImageQuality;
+
+typedef enum {
+  VG_IMAGE_FORMAT                             = 0x1E00,
+  VG_IMAGE_WIDTH                              = 0x1E01,
+  VG_IMAGE_HEIGHT                             = 0x1E02,
+
+  VG_IMAGE_PARAM_TYPE_FORCE_SIZE              = VG_MAX_ENUM
+} VGImageParamType;
+
+typedef enum {
+  VG_DRAW_IMAGE_NORMAL                        = 0x1F00,
+  VG_DRAW_IMAGE_MULTIPLY                      = 0x1F01,
+  VG_DRAW_IMAGE_STENCIL                       = 0x1F02,
+
+  VG_IMAGE_MODE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGImageMode;
+
+typedef enum {
+  VG_RED                                      = (1 << 3),
+  VG_GREEN                                    = (1 << 2),
+  VG_BLUE                                     = (1 << 1),
+  VG_ALPHA                                    = (1 << 0),
+
+  VG_IMAGE_CHANNEL_FORCE_SIZE                 = VG_MAX_ENUM
+} VGImageChannel;
+
+typedef enum {
+  VG_BLEND_SRC                                = 0x2000,
+  VG_BLEND_SRC_OVER                           = 0x2001,
+  VG_BLEND_DST_OVER                           = 0x2002,
+  VG_BLEND_SRC_IN                             = 0x2003,
+  VG_BLEND_DST_IN                             = 0x2004,
+  VG_BLEND_MULTIPLY                           = 0x2005,
+  VG_BLEND_SCREEN                             = 0x2006,
+  VG_BLEND_DARKEN                             = 0x2007,
+  VG_BLEND_LIGHTEN                            = 0x2008,
+  VG_BLEND_ADDITIVE                           = 0x2009,
+
+  VG_BLEND_MODE_FORCE_SIZE                    = VG_MAX_ENUM
+} VGBlendMode;
+
+typedef enum {
+  VG_IMAGE_FORMAT_QUERY                       = 0x2100,
+  VG_PATH_DATATYPE_QUERY                      = 0x2101,
+
+  VG_HARDWARE_QUERY_TYPE_FORCE_SIZE           = VG_MAX_ENUM
+} VGHardwareQueryType;
+
+typedef enum {
+  VG_HARDWARE_ACCELERATED                     = 0x2200,
+  VG_HARDWARE_UNACCELERATED                   = 0x2201,
+
+  VG_HARDWARE_QUERY_RESULT_FORCE_SIZE         = VG_MAX_ENUM
+} VGHardwareQueryResult;
+
+typedef enum {
+  VG_VENDOR                                   = 0x2300,
+  VG_RENDERER                                 = 0x2301,
+  VG_VERSION                                  = 0x2302,
+  VG_EXTENSIONS                               = 0x2303,
+
+  VG_STRING_ID_FORCE_SIZE                     = VG_MAX_ENUM
+} VGStringID;
+
+/* Function Prototypes */
+
+#ifndef VG_API_CALL
+#	error VG_API_CALL must be defined
+#endif
+
+#ifndef VG_API_ENTRY
+#   error VG_API_ENTRY must be defined 
+#endif
+
+#ifndef VG_API_EXIT
+#   error VG_API_EXIT must be defined 
+#endif
+
+VG_API_CALL VGErrorCode VG_API_ENTRY vgGetError(void) VG_API_EXIT;
+
+VG_API_CALL void VG_API_ENTRY vgFlush(void) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgFinish(void) VG_API_EXIT;
+
+/* Getters and Setters */
+VG_API_CALL void VG_API_ENTRY vgSetf (VGParamType type, VGfloat value) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSeti (VGParamType type, VGint value) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetfv(VGParamType type, VGint count,
+                         const VGfloat * values) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetiv(VGParamType type, VGint count,
+                         const VGint * values) VG_API_EXIT;
+
+VG_API_CALL VGfloat VG_API_ENTRY vgGetf(VGParamType type) VG_API_EXIT;
+VG_API_CALL VGint VG_API_ENTRY vgGeti(VGParamType type) VG_API_EXIT;
+VG_API_CALL VGint VG_API_ENTRY vgGetVectorSize(VGParamType type) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetfv(VGParamType type, VGint count, VGfloat * values) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetiv(VGParamType type, VGint count, VGint * values) VG_API_EXIT;
+
+VG_API_CALL void VG_API_ENTRY vgSetParameterf(VGHandle object,
+                                 VGint paramType,
+                                 VGfloat value) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetParameteri(VGHandle object,
+                                 VGint paramType,
+                                 VGint value) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetParameterfv(VGHandle object,
+                                  VGint paramType,
+                                  VGint count, const VGfloat * values) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetParameteriv(VGHandle object,
+                                  VGint paramType,
+                                  VGint count, const VGint * values) VG_API_EXIT;
+
+VG_API_CALL VGfloat VG_API_ENTRY vgGetParameterf(VGHandle object,
+                                    VGint paramType) VG_API_EXIT;
+VG_API_CALL VGint VG_API_ENTRY vgGetParameteri(VGHandle object,
+                                  VGint paramType);
+VG_API_CALL VGint VG_API_ENTRY vgGetParameterVectorSize(VGHandle object,
+                                           VGint paramType) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetParameterfv(VGHandle object,
+                                  VGint paramType,
+                                  VGint count, VGfloat * values) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetParameteriv(VGHandle object,
+                                  VGint paramType,
+                                  VGint count, VGint * values) VG_API_EXIT;
+
+/* Matrix Manipulation */
+VG_API_CALL void VG_API_ENTRY vgLoadIdentity(void) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgLoadMatrix(const VGfloat * m) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetMatrix(VGfloat * m) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgMultMatrix(const VGfloat * m) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgTranslate(VGfloat tx, VGfloat ty) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgScale(VGfloat sx, VGfloat sy) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgShear(VGfloat shx, VGfloat shy) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgRotate(VGfloat angle) VG_API_EXIT;
+
+/* Masking and Clearing */
+VG_API_CALL void VG_API_ENTRY vgMask(VGImage mask, VGMaskOperation operation,
+                        VGint x, VGint y, VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgClear(VGint x, VGint y, VGint width, VGint height) VG_API_EXIT;
+
+/* Paths */
+VG_API_CALL VGPath VG_API_ENTRY vgCreatePath(VGint pathFormat,
+                                VGPathDatatype datatype,
+                                VGfloat scale, VGfloat bias,
+                                VGint segmentCapacityHint,
+                                VGint coordCapacityHint,
+                                VGbitfield capabilities) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgClearPath(VGPath path, VGbitfield capabilities) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgDestroyPath(VGPath path) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgRemovePathCapabilities(VGPath path,
+                                          VGbitfield capabilities) VG_API_EXIT;
+VG_API_CALL VGbitfield VG_API_ENTRY vgGetPathCapabilities(VGPath path) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgAppendPath(VGPath dstPath, VGPath srcPath) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgAppendPathData(VGPath dstPath,
+                                  VGint numSegments,
+                                  const VGubyte * pathSegments,
+                                  const void * pathData) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgModifyPathCoords(VGPath dstPath, VGint startIndex,
+                                    VGint numSegments,
+                                    const void * pathData) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgTransformPath(VGPath dstPath, VGPath srcPath) VG_API_EXIT;
+VG_API_CALL VGboolean VG_API_ENTRY vgInterpolatePath(VGPath dstPath,
+                                        VGPath startPath,
+                                        VGPath endPath,
+                                        VGfloat amount) VG_API_EXIT;
+VG_API_CALL VGfloat VG_API_ENTRY vgPathLength(VGPath path,
+                                 VGint startSegment, VGint numSegments) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgPointAlongPath(VGPath path,
+                                  VGint startSegment, VGint numSegments,
+                                  VGfloat distance,
+                                  VGfloat * x, VGfloat * y,
+                                  VGfloat * tangentX, VGfloat * tangentY) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgPathBounds(VGPath path,
+                              VGfloat * minX, VGfloat * minY,
+                              VGfloat * width, VGfloat * height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgPathTransformedBounds(VGPath path,
+                                         VGfloat * minX, VGfloat * minY,
+                                         VGfloat * width, VGfloat * height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgDrawPath(VGPath path, VGbitfield paintModes) VG_API_EXIT;
+
+/* Paint */
+VG_API_CALL VGPaint VG_API_ENTRY vgCreatePaint(void) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgDestroyPaint(VGPaint paint) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetPaint(VGPaint paint, VGbitfield paintModes) VG_API_EXIT;
+VG_API_CALL VGPaint VG_API_ENTRY vgGetPaint(VGPaintMode paintMode) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetColor(VGPaint paint, VGuint rgba) VG_API_EXIT;
+VG_API_CALL VGuint VG_API_ENTRY vgGetColor(VGPaint paint) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgPaintPattern(VGPaint paint, VGImage pattern) VG_API_EXIT;
+
+/* Images */
+VG_API_CALL VGImage VG_API_ENTRY vgCreateImage(VGImageFormat format,
+                                  VGint width, VGint height,
+                                  VGbitfield allowedQuality) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgDestroyImage(VGImage image) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgClearImage(VGImage image,
+                              VGint x, VGint y, VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgImageSubData(VGImage image,
+                                const void * data, VGint dataStride,
+                                VGImageFormat dataFormat,
+                                VGint x, VGint y, VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetImageSubData(VGImage image,
+                                   void * data, VGint dataStride,
+                                   VGImageFormat dataFormat,
+                                   VGint x, VGint y,
+                                   VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL VGImage VG_API_ENTRY vgChildImage(VGImage parent,
+                                 VGint x, VGint y, VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL VGImage VG_API_ENTRY vgGetParent(VGImage image) VG_API_EXIT; 
+VG_API_CALL void VG_API_ENTRY vgCopyImage(VGImage dst, VGint dx, VGint dy,
+                             VGImage src, VGint sx, VGint sy,
+                             VGint width, VGint height,
+                             VGboolean dither) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgDrawImage(VGImage image) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSetPixels(VGint dx, VGint dy,
+                             VGImage src, VGint sx, VGint sy,
+                             VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgWritePixels(const void * data, VGint dataStride,
+                               VGImageFormat dataFormat,
+                               VGint dx, VGint dy,
+                               VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGetPixels(VGImage dst, VGint dx, VGint dy,
+                             VGint sx, VGint sy,
+                             VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgReadPixels(void * data, VGint dataStride,
+                              VGImageFormat dataFormat,
+                              VGint sx, VGint sy,
+                              VGint width, VGint height) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgCopyPixels(VGint dx, VGint dy,
+                              VGint sx, VGint sy,
+                              VGint width, VGint height) VG_API_EXIT;
+
+/* Image Filters */
+VG_API_CALL void VG_API_ENTRY vgColorMatrix(VGImage dst, VGImage src,
+                               const VGfloat * matrix) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgConvolve(VGImage dst, VGImage src,
+                            VGint kernelWidth, VGint kernelHeight,
+                            VGint shiftX, VGint shiftY,
+                            const VGshort * kernel,
+                            VGfloat scale,
+                            VGfloat bias,
+                            VGTilingMode tilingMode) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgSeparableConvolve(VGImage dst, VGImage src,
+                                     VGint kernelWidth,
+                                     VGint kernelHeight,
+                                     VGint shiftX, VGint shiftY,
+                                     const VGshort * kernelX,
+                                     const VGshort * kernelY,
+                                     VGfloat scale,
+                                     VGfloat bias,
+                                     VGTilingMode tilingMode) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgGaussianBlur(VGImage dst, VGImage src,
+                                VGfloat stdDeviationX,
+                                VGfloat stdDeviationY,
+                                VGTilingMode tilingMode) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgLookup(VGImage dst, VGImage src,
+                          const VGubyte * redLUT,
+                          const VGubyte * greenLUT,
+                          const VGubyte * blueLUT,
+                          const VGubyte * alphaLUT,
+                          VGboolean outputLinear,
+                          VGboolean outputPremultiplied) VG_API_EXIT;
+VG_API_CALL void VG_API_ENTRY vgLookupSingle(VGImage dst, VGImage src,
+                                const VGuint * lookupTable,
+                                VGImageChannel sourceChannel,
+                                VGboolean outputLinear,
+                                VGboolean outputPremultiplied) VG_API_EXIT;
+
+/* Hardware Queries */
+VG_API_CALL VGHardwareQueryResult VG_API_ENTRY vgHardwareQuery(VGHardwareQueryType key,
+                                                  VGint setting) VG_API_EXIT;
+
+/* Renderer and Extension Information */
+VG_API_CALL const VGubyte * VG_API_ENTRY vgGetString(VGStringID name) VG_API_EXIT;
+
+#ifdef __cplusplus 
+} /* extern "C" */
+#endif
+
+#endif /* _OPENVG_H */
diff --git a/include/VG/vgext.h b/include/VG/vgext.h
new file mode 100644
index 0000000000..97e3e779e1
--- /dev/null
+++ b/include/VG/vgext.h
@@ -0,0 +1,233 @@
+/* $Revision: 6810 $ on $Date:: 2008-10-29 10:31:37 -0400 #$ */

+

+/*------------------------------------------------------------------------

+ * 

+ * VG extensions Reference Implementation

+ * -------------------------------------

+ *

+ * Copyright (c) 2008 The Khronos Group Inc.

+ *

+ * Permission is hereby granted, free of charge, to any person obtaining a

+ * copy of this software and /or associated documentation files

+ * (the "Materials "), to deal in the Materials without restriction,

+ * including without limitation the rights to use, copy, modify, merge,

+ * publish, distribute, sublicense, and/or sell copies of the Materials,

+ * and to permit persons to whom the Materials are furnished to do so,

+ * subject to the following conditions: 

+ *

+ * The above copyright notice and this permission notice shall be included 

+ * in all copies or substantial portions of the Materials. 

+ *

+ * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,

+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.

+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,

+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR

+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE MATERIALS OR

+ * THE USE OR OTHER DEALINGS IN THE MATERIALS.

+ *

+ *//**

+ * \file

+ * \brief	VG extensions

+ *//*-------------------------------------------------------------------*/

+

+

+

+#ifndef _VGEXT_H

+#define _VGEXT_H

+

+#ifdef __cplusplus

+extern "C" {

+#endif

+

+#include <VG/openvg.h>

+#include <VG/vgu.h>

+

+#ifndef VG_API_ENTRYP

+#   define VG_API_ENTRYP VG_API_ENTRY*

+#endif

+

+#ifndef VGU_API_ENTRYP

+#   define VGU_API_ENTRYP VGU_API_ENTRY*

+#endif

+

+/*-------------------------------------------------------------------------------

+ * KHR extensions

+ *------------------------------------------------------------------------------*/

+

+typedef enum  {

+

+#ifndef VG_KHR_iterative_average_blur

+  VG_MAX_AVERAGE_BLUR_DIMENSION_KHR        = 0x116B,

+  VG_AVERAGE_BLUR_DIMENSION_RESOLUTION_KHR = 0x116C,

+  VG_MAX_AVERAGE_BLUR_ITERATIONS_KHR       = 0x116D,

+#endif

+

+  VG_PARAM_TYPE_KHR_FORCE_SIZE             = VG_MAX_ENUM

+} VGParamTypeKHR;

+

+#ifndef VG_KHR_EGL_image

+#define VG_KHR_EGL_image 1

+/* VGEGLImageKHR is an opaque handle to an EGLImage */

+typedef void* VGeglImageKHR; 

+

+#ifdef VG_VGEXT_PROTOTYPES

+VG_API_CALL VGImage VG_API_ENTRY vgCreateEGLImageTargetKHR(VGeglImageKHR image);

+#endif

+typedef VGImage (VG_API_ENTRYP PFNVGCREATEEGLIMAGETARGETKHRPROC) (VGeglImageKHR image);

+

+#endif

+

+

+#ifndef VG_KHR_iterative_average_blur

+#define VG_KHR_iterative_average_blur 1

+

+#ifdef VG_VGEXT_PROTOTYPES

+VG_API_CALL void vgIterativeAverageBlurKHR(VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGTilingMode tilingMode);

+#endif 

+typedef void (VG_API_ENTRYP PFNVGITERATIVEAVERAGEBLURKHRPROC) (VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGTilingMode tilingMode);

+

+#endif

+

+

+#ifndef VG_KHR_advanced_blending

+#define VG_KHR_advanced_blending 1

+

+typedef enum {

+  VG_BLEND_OVERLAY_KHR        = 0x2010,

+  VG_BLEND_HARDLIGHT_KHR      = 0x2011,

+  VG_BLEND_SOFTLIGHT_SVG_KHR  = 0x2012,

+  VG_BLEND_SOFTLIGHT_KHR      = 0x2013,

+  VG_BLEND_COLORDODGE_KHR     = 0x2014,

+  VG_BLEND_COLORBURN_KHR      = 0x2015,

+  VG_BLEND_DIFFERENCE_KHR     = 0x2016,

+  VG_BLEND_SUBTRACT_KHR       = 0x2017,

+  VG_BLEND_INVERT_KHR         = 0x2018,

+  VG_BLEND_EXCLUSION_KHR      = 0x2019,

+  VG_BLEND_LINEARDODGE_KHR    = 0x201a,

+  VG_BLEND_LINEARBURN_KHR     = 0x201b,

+  VG_BLEND_VIVIDLIGHT_KHR     = 0x201c,

+  VG_BLEND_LINEARLIGHT_KHR    = 0x201d,

+  VG_BLEND_PINLIGHT_KHR       = 0x201e,

+  VG_BLEND_HARDMIX_KHR        = 0x201f,

+  VG_BLEND_CLEAR_KHR          = 0x2020,

+  VG_BLEND_DST_KHR            = 0x2021,

+  VG_BLEND_SRC_OUT_KHR        = 0x2022,

+  VG_BLEND_DST_OUT_KHR        = 0x2023,

+  VG_BLEND_SRC_ATOP_KHR       = 0x2024,

+  VG_BLEND_DST_ATOP_KHR       = 0x2025,

+  VG_BLEND_XOR_KHR            = 0x2026,

+

+  VG_BLEND_MODE_KHR_FORCE_SIZE= VG_MAX_ENUM

+} VGBlendModeKHR;

+#endif

+

+#ifndef VG_KHR_parametric_filter

+#define VG_KHR_parametric_filter 1 

+

+typedef enum {

+  VG_PF_OBJECT_VISIBLE_FLAG_KHR = (1 << 0),

+  VG_PF_KNOCKOUT_FLAG_KHR       = (1 << 1),

+  VG_PF_OUTER_FLAG_KHR          = (1 << 2),

+  VG_PF_INNER_FLAG_KHR          = (1 << 3),

+

+  VG_PF_TYPE_KHR_FORCE_SIZE     = VG_MAX_ENUM

+} VGPfTypeKHR;

+

+typedef enum {

+  VGU_IMAGE_IN_USE_ERROR           = 0xF010,

+

+  VGU_ERROR_CODE_KHR_FORCE_SIZE    = VG_MAX_ENUM

+} VGUErrorCodeKHR;

+

+#ifdef VG_VGEXT_PROTOTYPES

+VG_API_CALL void VG_API_ENTRY vgParametricFilterKHR(VGImage dst,VGImage src,VGImage blur,VGfloat strength,VGfloat offsetX,VGfloat offsetY,VGbitfield filterFlags,VGPaint highlightPaint,VGPaint shadowPaint);

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguDropShadowKHR(VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint shadowColorRGBA);

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguGlowKHR(VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint glowColorRGBA) ;

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguBevelKHR(VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint highlightColorRGBA,VGuint shadowColorRGBA);

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguGradientGlowKHR(VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint stopsCount,const VGfloat* glowColorRampStops);

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguGradientBevelKHR(VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint stopsCount,const VGfloat* bevelColorRampStops);

+#endif

+typedef void (VG_API_ENTRYP PFNVGPARAMETRICFILTERKHRPROC) (VGImage dst,VGImage src,VGImage blur,VGfloat strength,VGfloat offsetX,VGfloat offsetY,VGbitfield filterFlags,VGPaint highlightPaint,VGPaint shadowPaint);

+typedef VGUErrorCode (VGU_API_ENTRYP PFNVGUDROPSHADOWKHRPROC) (VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint shadowColorRGBA);

+typedef VGUErrorCode (VGU_API_ENTRYP PFNVGUGLOWKHRPROC) (VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint glowColorRGBA);

+typedef VGUErrorCode (VGU_API_ENTRYP PFNVGUBEVELKHRPROC) (VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint highlightColorRGBA,VGuint shadowColorRGBA);

+typedef VGUErrorCode (VGU_API_ENTRYP PFNVGUGRADIENTGLOWKHRPROC) (VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint stopsCount,const VGfloat* glowColorRampStops);

+typedef VGUErrorCode (VGU_API_ENTRYP PFNVGUGRADIENTBEVELKHRPROC) (VGImage dst,VGImage src,VGfloat dimX,VGfloat dimY,VGuint iterative,VGfloat strength,VGfloat distance,VGfloat angle,VGbitfield filterFlags,VGbitfield allowedQuality,VGuint stopsCount,const VGfloat* bevelColorRampStops);

+

+#endif

+

+

+/*-------------------------------------------------------------------------------

+ * NDS extensions

+ *------------------------------------------------------------------------------*/

+

+#ifndef VG_NDS_paint_generation

+#define VG_NDS_paint_generation 1

+

+typedef enum { 

+  VG_PAINT_COLOR_RAMP_LINEAR_NDS            = 0x1A10,

+  VG_COLOR_MATRIX_NDS                       = 0x1A11,

+  VG_PAINT_COLOR_TRANSFORM_LINEAR_NDS       = 0x1A12,

+

+  VG_PAINT_PARAM_TYPE_NDS_FORCE_SIZE        = VG_MAX_ENUM

+} VGPaintParamTypeNds;

+

+typedef enum {

+  VG_DRAW_IMAGE_COLOR_MATRIX_NDS            = 0x1F10,

+

+  VG_IMAGE_MODE_NDS_FORCE_SIZE              = VG_MAX_ENUM

+} VGImageModeNds;

+#endif 

+

+

+#ifndef VG_NDS_projective_geometry

+#define VG_NDS_projective_geometry 1

+

+typedef enum {

+  VG_CLIP_MODE_NDS                          = 0x1180,

+  VG_CLIP_LINES_NDS                         = 0x1181,

+  VG_MAX_CLIP_LINES_NDS                     = 0x1182,

+

+  VG_PARAM_TYPE_NDS_FORCE_SIZE        = VG_MAX_ENUM

+} VGParamTypeNds;

+

+typedef enum {

+  VG_CLIPMODE_NONE_NDS                      = 0x3000,

+  VG_CLIPMODE_CLIP_CLOSED_NDS               = 0x3001,

+  VG_CLIPMODE_CLIP_OPEN_NDS                 = 0x3002,

+  VG_CLIPMODE_CULL_NDS                      = 0x3003,

+

+  VG_CLIPMODE_NDS_FORCE_SIZE = VG_MAX_ENUM

+} VGClipModeNds;

+

+typedef enum {

+  VG_RQUAD_TO_NDS              = ( 13 << 1 ),

+  VG_RCUBIC_TO_NDS             = ( 14 << 1 ),

+  

+  VG_PATH_SEGMENT_NDS_FORCE_SIZE = VG_MAX_ENUM

+} VGPathSegmentNds;

+

+typedef enum {

+  VG_RQUAD_TO_ABS_NDS            = (VG_RQUAD_TO_NDS  | VG_ABSOLUTE),

+  VG_RQUAD_TO_REL_NDS            = (VG_RQUAD_TO_NDS  | VG_RELATIVE),

+  VG_RCUBIC_TO_ABS_NDS           = (VG_RCUBIC_TO_NDS | VG_ABSOLUTE),

+  VG_RCUBIC_TO_REL_NDS           = (VG_RCUBIC_TO_NDS | VG_RELATIVE),

+

+  VG_PATH_COMMAND_NDS_FORCE_SIZE = VG_MAX_ENUM

+} VGPathCommandNds;

+

+#ifdef VG_VGEXT_PROTOTYPES

+VG_API_CALL void VG_API_ENTRY vgProjectiveMatrixNDS(VGboolean enable) ;

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguTransformClipLineNDS(const VGfloat Ain,const VGfloat Bin,const VGfloat Cin,const VGfloat* matrix,const VGboolean inverse,VGfloat* Aout,VGfloat* Bout,VGfloat* Cout);

+#endif 

+typedef void (VG_API_ENTRYP PFNVGPROJECTIVEMATRIXNDSPROC) (VGboolean enable) ;

+typedef VGUErrorCode (VGU_API_ENTRYP PFNVGUTRANSFORMCLIPLINENDSPROC) (const VGfloat Ain,const VGfloat Bin,const VGfloat Cin,const VGfloat* matrix,const VGboolean inverse,VGfloat* Aout,VGfloat* Bout,VGfloat* Cout);

+

+#endif

+

+#ifdef __cplusplus 

+} /* extern "C" */

+#endif

+

+#endif /* _VGEXT_H */

diff --git a/include/VG/vgplatform.h b/include/VG/vgplatform.h
new file mode 100644
index 0000000000..e4f269f658
--- /dev/null
+++ b/include/VG/vgplatform.h
@@ -0,0 +1,106 @@
+/* $Revision: 6810 $ on $Date:: 2008-10-29 10:31:37 -0400 #$ */

+

+/*------------------------------------------------------------------------

+ *

+ * VG platform specific header Reference Implementation

+ * ----------------------------------------------------

+ *

+ * Copyright (c) 2008 The Khronos Group Inc.

+ *

+ * Permission is hereby granted, free of charge, to any person obtaining a

+ * copy of this software and /or associated documentation files

+ * (the "Materials "), to deal in the Materials without restriction,

+ * including without limitation the rights to use, copy, modify, merge,

+ * publish, distribute, sublicense, and/or sell copies of the Materials,

+ * and to permit persons to whom the Materials are furnished to do so,

+ * subject to the following conditions: 

+ *

+ * The above copyright notice and this permission notice shall be included 

+ * in all copies or substantial portions of the Materials. 

+ *

+ * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,

+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.

+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,

+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR

+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE MATERIALS OR

+ * THE USE OR OTHER DEALINGS IN THE MATERIALS.

+ *

+ *//**

+ * \file

+ * \brief VG platform specific header

+ *//*-------------------------------------------------------------------*/

+

+#ifndef _VGPLATFORM_H

+#define _VGPLATFORM_H

+

+#ifdef __cplusplus

+extern "C" {

+#endif

+

+#ifndef VG_API_CALL 

+#if defined(OPENVG_STATIC_LIBRARY)

+#	define VG_API_CALL

+#else

+#	if defined(_WIN32) || defined(__VC32__)				/* Win32 */

+#		if defined (OPENVG_DLL_EXPORTS)

+#			define VG_API_CALL __declspec(dllexport)

+#		else

+#			define VG_API_CALL __declspec(dllimport)

+#		endif

+#	else 

+#		define VG_API_CALL extern

+#	endif /* defined(_WIN32) ||... */

+#endif /* defined OPENVG_STATIC_LIBRARY */

+#endif /* ifndef VG_API_CALL */

+

+#ifndef VGU_API_CALL 

+#if defined(OPENVG_STATIC_LIBRARY)

+#	define VGU_API_CALL

+#else

+#	if defined(_WIN32) || defined(__VC32__)				/* Win32 */

+#		if defined (OPENVG_DLL_EXPORTS)

+#			define VGU_API_CALL __declspec(dllexport)

+#		else

+#			define VGU_API_CALL __declspec(dllimport)

+#		endif

+#	else 

+#		define VGU_API_CALL extern

+#	endif /* defined(_WIN32) ||... */

+#endif /* defined OPENVG_STATIC_LIBRARY */

+#endif /* ifndef VGU_API_CALL */

+

+

+#ifndef VG_API_ENTRY

+#define VG_API_ENTRY

+#endif

+

+#ifndef VG_API_EXIT

+#define VG_API_EXIT

+#endif

+

+#ifndef VGU_API_ENTRY

+#define VGU_API_ENTRY

+#endif

+

+#ifndef VGU_API_EXIT

+#define VGU_API_EXIT

+#endif

+

+typedef float          VGfloat;

+typedef signed char    VGbyte;

+typedef unsigned char  VGubyte;

+typedef signed short   VGshort;

+typedef signed int     VGint;

+typedef unsigned int   VGuint;

+typedef unsigned int   VGbitfield;

+

+#ifndef VG_VGEXT_PROTOTYPES

+#define VG_VGEXT_PROTOTYPES

+#endif 

+

+#ifdef __cplusplus 

+} /* extern "C" */

+#endif

+

+#endif /* _VGPLATFORM_H */

diff --git a/include/VG/vgu.h b/include/VG/vgu.h
new file mode 100644
index 0000000000..2799684e5e
--- /dev/null
+++ b/include/VG/vgu.h
@@ -0,0 +1,130 @@
+/* $Revision: 6810 $ on $Date:: 2008-10-29 10:31:37 -0400 #$ */

+

+/*------------------------------------------------------------------------

+ * 

+ * VGU 1.0.1 Reference Implementation

+ * -------------------------------------

+ *

+ * Copyright (c) 2008 The Khronos Group Inc.

+ *

+ * Permission is hereby granted, free of charge, to any person obtaining a

+ * copy of this software and /or associated documentation files

+ * (the "Materials "), to deal in the Materials without restriction,

+ * including without limitation the rights to use, copy, modify, merge,

+ * publish, distribute, sublicense, and/or sell copies of the Materials,

+ * and to permit persons to whom the Materials are furnished to do so,

+ * subject to the following conditions: 

+ *

+ * The above copyright notice and this permission notice shall be included 

+ * in all copies or substantial portions of the Materials. 

+ *

+ * THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,

+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF

+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.

+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,

+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR

+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE MATERIALS OR

+ * THE USE OR OTHER DEALINGS IN THE MATERIALS.

+ *

+ *//**

+ * \file

+ * \brief	VGU 1.0.1 API.

+ *//*-------------------------------------------------------------------*/

+

+#ifndef _VGU_H

+#define _VGU_H

+

+#ifdef __cplusplus 

+extern "C" { 

+#endif

+

+#include <VG/openvg.h>

+

+#define VGU_VERSION_1_0 1

+

+#ifndef VGU_API_CALL

+#	error VGU_API_CALL must be defined

+#endif

+

+#ifndef VGU_API_ENTRY

+#   error VGU_API_ENTRY must be defined 

+#endif

+

+#ifndef VGU_API_EXIT

+#   error VGU_API_EXIT must be defined 

+#endif

+

+

+typedef enum {

+  VGU_NO_ERROR                                 = 0,

+  VGU_BAD_HANDLE_ERROR                         = 0xF000,

+  VGU_ILLEGAL_ARGUMENT_ERROR                   = 0xF001,

+  VGU_OUT_OF_MEMORY_ERROR                      = 0xF002,

+  VGU_PATH_CAPABILITY_ERROR                    = 0xF003,

+  VGU_BAD_WARP_ERROR                           = 0xF004,

+

+  VGU_ERROR_CODE_FORCE_SIZE                    = VG_MAX_ENUM

+} VGUErrorCode;

+

+typedef enum {

+  VGU_ARC_OPEN                                 = 0xF100,

+  VGU_ARC_CHORD                                = 0xF101,

+  VGU_ARC_PIE                                  = 0xF102,

+

+  VGU_ARC_TYPE_FORCE_SIZE                      = VG_MAX_ENUM

+} VGUArcType;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguLine(VGPath path,

+                                  VGfloat x0, VGfloat y0,

+                                  VGfloat x1, VGfloat y1) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguPolygon(VGPath path,

+                                     const VGfloat * points, VGint count,

+                                     VGboolean closed) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguRect(VGPath path,

+                                  VGfloat x, VGfloat y,

+                                  VGfloat width, VGfloat height) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguRoundRect(VGPath path,

+                                       VGfloat x, VGfloat y,

+                                       VGfloat width, VGfloat height,

+                                       VGfloat arcWidth, VGfloat arcHeight) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguEllipse(VGPath path,

+                                     VGfloat cx, VGfloat cy,

+                                     VGfloat width, VGfloat height) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguArc(VGPath path,

+                                 VGfloat x, VGfloat y,

+                                 VGfloat width, VGfloat height,

+                                 VGfloat startAngle, VGfloat angleExtent,

+                                 VGUArcType arcType) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguComputeWarpQuadToSquare(VGfloat sx0, VGfloat sy0,

+                                                     VGfloat sx1, VGfloat sy1,

+                                                     VGfloat sx2, VGfloat sy2,

+                                                     VGfloat sx3, VGfloat sy3,

+                                                     VGfloat * matrix) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguComputeWarpSquareToQuad(VGfloat dx0, VGfloat dy0,

+                                                     VGfloat dx1, VGfloat dy1,

+                                                     VGfloat dx2, VGfloat dy2,

+                                                     VGfloat dx3, VGfloat dy3,

+                                                     VGfloat * matrix) VGU_API_EXIT;

+

+VGU_API_CALL VGUErrorCode VGU_API_ENTRY vguComputeWarpQuadToQuad(VGfloat dx0, VGfloat dy0,

+                                                   VGfloat dx1, VGfloat dy1,

+                                                   VGfloat dx2, VGfloat dy2,

+                                                   VGfloat dx3, VGfloat dy3,

+                                                   VGfloat sx0, VGfloat sy0,

+                                                   VGfloat sx1, VGfloat sy1,

+                                                   VGfloat sx2, VGfloat sy2,

+                                                   VGfloat sx3, VGfloat sy3,

+                                                   VGfloat * matrix) VGU_API_EXIT;

+

+#ifdef __cplusplus 

+} /* extern "C" */

+#endif

+

+#endif /* #ifndef _VGU_H */

diff --git a/progs/demos/cubemap.c b/progs/demos/cubemap.c
index 26db42aed5..1f9f290575 100644
--- a/progs/demos/cubemap.c
+++ b/progs/demos/cubemap.c
@@ -52,6 +52,7 @@ static GLboolean NoClear = GL_FALSE;
 static GLint FrameParity = 0;
 static GLenum FilterIndex = 0;
 static GLint ClampIndex = 0;
+static GLboolean supportFBO = GL_FALSE;
 
 
 static struct {
@@ -403,6 +404,10 @@ static void init_checkers( void )
 
    glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
 
+   if (!supportFBO)
+      glTexParameteri(GL_TEXTURE_CUBE_MAP_ARB, GL_GENERATE_MIPMAP_SGIS, GL_TRUE);
+
+
    /* make colored checkerboard cube faces */
    for (f = 0; f < 6; f++) {
       for (i = 0; i < CUBE_TEX_SIZE; i++) {
@@ -426,7 +431,8 @@ static void init_checkers( void )
                    GL_BGRA, GL_UNSIGNED_BYTE, image);
    }
 
-   glGenerateMipmapEXT(GL_TEXTURE_CUBE_MAP_ARB);
+   if (supportFBO)
+      glGenerateMipmapEXT(GL_TEXTURE_CUBE_MAP_ARB);
 }
 
 
@@ -503,10 +509,13 @@ static void init( GLboolean useImageFiles )
          exit(0);
       }
 
-      /* Needed for glGenerateMipmapEXT
+      /* Needed for glGenerateMipmapEXT / auto mipmapping
        */
-      if (!strstr(exten, "GL_EXT_framebuffer_object")) {
-         printf("Sorry, this demo requires GL_EXT_framebuffer_object\n");
+      if (strstr(exten, "GL_EXT_framebuffer_object")) {
+         supportFBO = GL_TRUE;
+      }
+      else if (!strstr(exten, "GL_SGIS_generate_mipmap")) {
+         printf("Sorry, this demo requires GL_EXT_framebuffer_object or GL_SGIS_generate_mipmap\n");
          exit(0);
       }
    }
diff --git a/progs/demos/isosurf.c b/progs/demos/isosurf.c
index 6923ca2bba..2e9dff1726 100644
--- a/progs/demos/isosurf.c
+++ b/progs/demos/isosurf.c
@@ -847,8 +847,8 @@ static void Init(int argc, char *argv[])
 
    glClearColor(0.0, 0.0, 1.0, 0.0);
    glEnable( GL_DEPTH_TEST );
-   glEnable( GL_VERTEX_ARRAY_EXT );
-   glEnable( GL_NORMAL_ARRAY_EXT );
+   glEnableClientState( GL_VERTEX_ARRAY );
+   glEnableClientState( GL_NORMAL_ARRAY );
 
    glMatrixMode(GL_PROJECTION);
    glLoadIdentity();
diff --git a/progs/demos/pointblast.c b/progs/demos/pointblast.c
index 2a91b76ad3..2d70b72589 100644
--- a/progs/demos/pointblast.c
+++ b/progs/demos/pointblast.c
@@ -194,11 +194,11 @@ redraw(void)
 {
   int i;
 
+  glDepthMask(GL_TRUE);
   glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
   if (newModel)
     recalcModelView();
 
-  glDepthMask(GL_FALSE);
 
   /* Draw the floor. */
 /*  glEnable(GL_TEXTURE_2D);*/
@@ -215,7 +215,7 @@ redraw(void)
   glEnd();
 
   /* Allow particles to blend with each other. */
-  glDepthMask(GL_TRUE);
+  glDepthMask(GL_FALSE);
 
   if (blend)
      glEnable(GL_BLEND);
diff --git a/progs/demos/spriteblast.c b/progs/demos/spriteblast.c
index f6630c25d0..d73b680b79 100644
--- a/progs/demos/spriteblast.c
+++ b/progs/demos/spriteblast.c
@@ -209,13 +209,13 @@ redraw(void)
 {
   int i;
 
+  glDepthMask(GL_TRUE);
   glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
 
   glPushMatrix();
   glRotatef(15.0, 1.0, 0.0, 0.0);
   glRotatef(angle, 0.0, 1.0, 0.0);
 
-  glDepthMask(GL_FALSE);
 
   /* Draw the floor. */
 /*  glEnable(GL_TEXTURE_2D);*/
@@ -232,7 +232,7 @@ redraw(void)
   glEnd();
 
   /* Allow particles to blend with each other. */
-  glDepthMask(GL_TRUE);
+  glDepthMask(GL_FALSE);
 
   if (blend)
      glEnable(GL_BLEND);
diff --git a/progs/demos/vao_demo.c b/progs/demos/vao_demo.c
index ce416712fe..206e06fc6c 100644
--- a/progs/demos/vao_demo.c
+++ b/progs/demos/vao_demo.c
@@ -260,6 +260,8 @@ static void Key( unsigned char key, int x, int y )
    (void) y;
    switch (key) {
       case 27:
+         (*delete_vertex_arrays)( 1, & cube_array_obj );
+         (*delete_vertex_arrays)( 1, & oct_array_obj );
          glutDestroyWindow(Win);
          exit(0);
          break;
diff --git a/progs/egl/.gitignore b/progs/egl/.gitignore
index 793c6c0f61..f417aefb93 100644
--- a/progs/egl/.gitignore
+++ b/progs/egl/.gitignore
@@ -7,4 +7,5 @@ eglscreen
 egltri
 peglgears
 xeglgears
+xeglthreads
 xegl_tri
diff --git a/progs/egl/Makefile b/progs/egl/Makefile
index e1fdb1ce63..d3c32d46f7 100644
--- a/progs/egl/Makefile
+++ b/progs/egl/Makefile
@@ -18,6 +18,7 @@ PROGRAMS = \
 	eglscreen \
 	peglgears \
 	xeglgears \
+	xeglthreads \
 	xegl_tri
 
 
@@ -90,6 +91,13 @@ xeglgears.o: xeglgears.c $(HEADERS)
 	$(CC) -c $(CFLAGS) -I$(TOP)/include xeglgears.c
 
 
+xeglthreads: xeglthreads.o $(TOP)/$(LIB_DIR)/libEGL.so
+	$(CC) $(CFLAGS) xeglthreads.o -L$(TOP)/$(LIB_DIR) -lEGL -lGL $(LIBDRM_LIB) $(APP_LIB_DEPS) -o $@
+
+xeglthreads.o: xeglthreads.c $(HEADERS)
+	$(CC) -c $(CFLAGS) -I$(TOP)/include xeglthreads.c
+
+
 xegl_tri: xegl_tri.o $(TOP)/$(LIB_DIR)/libEGL.so
 	$(CC) $(CFLAGS) xegl_tri.o -L$(TOP)/$(LIB_DIR) -lEGL -lGL $(LIBDRM_LIB) $(APP_LIB_DEPS) -o $@
 
diff --git a/progs/egl/xeglthreads.c b/progs/egl/xeglthreads.c
new file mode 100644
index 0000000000..508dbc0943
--- /dev/null
+++ b/progs/egl/xeglthreads.c
@@ -0,0 +1,773 @@
+/*
+ * Copyright (C) 2000  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Ported to EGL by Chia-I Wu <olvaffe@gmail.com>
+ */
+
+
+/*
+ * This program tests EGL thread safety.
+ * Command line options:
+ *  -p                       Open a display connection for each thread
+ *  -l                       Enable application-side locking
+ *  -n <num threads>         Number of threads to create (default is 2)
+ *  -display <display name>  Specify X display (default is $DISPLAY)
+ *  -t                       Use texture mapping
+ *
+ * Brian Paul  20 July 2000
+ */
+
+
+/*
+ * Notes:
+ * - Each thread gets its own EGL context.
+ *
+ * - The EGL contexts share texture objects.
+ *
+ * - When 't' is pressed to update the texture image, the window/thread which
+ *   has input focus is signalled to change the texture.  The other threads
+ *   should see the updated texture the next time they call glBindTexture.
+ */
+
+
+#if defined(PTHREADS)   /* defined by Mesa on Linux and other platforms */
+
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include <GL/gl.h>
+#include <EGL/egl.h>
+#include <math.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <pthread.h>
+
+
+/*
+ * Each window/thread/context:
+ */
+struct winthread {
+   Display *Dpy;
+   int Index;
+   pthread_t Thread;
+   Window Win;
+   EGLDisplay Display;
+   EGLContext Context;
+   EGLSurface Surface;
+   float Angle;
+   int WinWidth, WinHeight;
+   GLboolean NewSize;
+   GLboolean Initialized;
+   GLboolean MakeNewTexture;
+};
+
+
+#define MAX_WINTHREADS 100
+static struct winthread WinThreads[MAX_WINTHREADS];
+static int NumWinThreads = 0;
+static volatile GLboolean ExitFlag = GL_FALSE;
+
+static GLboolean MultiDisplays = 0;
+static GLboolean Locking = 0;
+static GLboolean Texture = GL_FALSE;
+static GLuint TexObj = 12;
+static GLboolean Animate = GL_TRUE;
+
+static pthread_mutex_t Mutex;
+static pthread_cond_t CondVar;
+static pthread_mutex_t CondMutex;
+
+
+static void
+Error(const char *msg)
+{
+   fprintf(stderr, "Error: %s\n", msg);
+   exit(1);
+}
+
+
+static void
+signal_redraw(void)
+{
+   pthread_mutex_lock(&CondMutex);
+   pthread_cond_broadcast(&CondVar);
+   pthread_mutex_unlock(&CondMutex);
+}
+
+
+static void
+MakeNewTexture(struct winthread *wt)
+{
+#define TEX_SIZE 128
+   static float step = 0.0;
+   GLfloat image[TEX_SIZE][TEX_SIZE][4];
+   GLint width;
+   int i, j;
+
+   for (j = 0; j < TEX_SIZE; j++) {
+      for (i = 0; i < TEX_SIZE; i++) {
+         float dt = 5.0 * (j - 0.5 * TEX_SIZE) / TEX_SIZE;
+         float ds = 5.0 * (i - 0.5 * TEX_SIZE) / TEX_SIZE;
+         float r = dt * dt + ds * ds + step;
+         image[j][i][0] =
+         image[j][i][1] =
+         image[j][i][2] = 0.75 + 0.25 * cos(r);
+         image[j][i][3] = 1.0;
+      }
+   }
+
+   step += 0.5;
+
+   glBindTexture(GL_TEXTURE_2D, TexObj);
+
+   glGetTexLevelParameteriv(GL_TEXTURE_2D, 0, GL_TEXTURE_WIDTH, &width);
+   if (width) {
+      assert(width == TEX_SIZE);
+      /* sub-tex replace */
+      glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, TEX_SIZE, TEX_SIZE,
+                   GL_RGBA, GL_FLOAT, image);
+   }
+   else {
+      /* create new */
+      glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+      glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+
+      glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, TEX_SIZE, TEX_SIZE, 0,
+                   GL_RGBA, GL_FLOAT, image);
+   }
+}
+
+
+
+/* draw a colored cube */
+static void
+draw_object(void)
+{
+   glPushMatrix();
+   glScalef(0.75, 0.75, 0.75);
+
+   glColor3f(1, 0, 0);
+
+   if (Texture) {
+      glBindTexture(GL_TEXTURE_2D, TexObj);
+      glEnable(GL_TEXTURE_2D);
+   }
+   else {
+      glDisable(GL_TEXTURE_2D);
+   }
+
+   glBegin(GL_QUADS);
+
+   /* -X */
+   glColor3f(0, 1, 1);
+   glTexCoord2f(0, 0);  glVertex3f(-1, -1, -1);
+   glTexCoord2f(1, 0);  glVertex3f(-1,  1, -1);
+   glTexCoord2f(1, 1);  glVertex3f(-1,  1,  1);
+   glTexCoord2f(0, 1);  glVertex3f(-1, -1,  1);
+
+   /* +X */
+   glColor3f(1, 0, 0);
+   glTexCoord2f(0, 0);  glVertex3f(1, -1, -1);
+   glTexCoord2f(1, 0);  glVertex3f(1,  1, -1);
+   glTexCoord2f(1, 1);  glVertex3f(1,  1,  1);
+   glTexCoord2f(0, 1);  glVertex3f(1, -1,  1);
+
+   /* -Y */
+   glColor3f(1, 0, 1);
+   glTexCoord2f(0, 0);  glVertex3f(-1, -1, -1);
+   glTexCoord2f(1, 0);  glVertex3f( 1, -1, -1);
+   glTexCoord2f(1, 1);  glVertex3f( 1, -1,  1);
+   glTexCoord2f(0, 1);  glVertex3f(-1, -1,  1);
+
+   /* +Y */
+   glColor3f(0, 1, 0);
+   glTexCoord2f(0, 0);  glVertex3f(-1, 1, -1);
+   glTexCoord2f(1, 0);  glVertex3f( 1, 1, -1);
+   glTexCoord2f(1, 1);  glVertex3f( 1, 1,  1);
+   glTexCoord2f(0, 1);  glVertex3f(-1, 1,  1);
+
+   /* -Z */
+   glColor3f(1, 1, 0);
+   glTexCoord2f(0, 0);  glVertex3f(-1, -1, -1);
+   glTexCoord2f(1, 0);  glVertex3f( 1, -1, -1);
+   glTexCoord2f(1, 1);  glVertex3f( 1,  1, -1);
+   glTexCoord2f(0, 1);  glVertex3f(-1,  1, -1);
+
+   /* +Y */
+   glColor3f(0, 0, 1);
+   glTexCoord2f(0, 0);  glVertex3f(-1, -1, 1);
+   glTexCoord2f(1, 0);  glVertex3f( 1, -1, 1);
+   glTexCoord2f(1, 1);  glVertex3f( 1,  1, 1);
+   glTexCoord2f(0, 1);  glVertex3f(-1,  1, 1);
+
+   glEnd();
+
+   glPopMatrix();
+}
+
+
+/* signal resize of given window */
+static void
+resize(struct winthread *wt, int w, int h)
+{
+   wt->NewSize = GL_TRUE;
+   wt->WinWidth = w;
+   wt->WinHeight = h;
+   if (!Animate)
+      signal_redraw();
+}
+
+
+/*
+ * We have an instance of this for each thread.
+ */
+static void
+draw_loop(struct winthread *wt)
+{
+   while (!ExitFlag) {
+
+      if (Locking)
+         pthread_mutex_lock(&Mutex);
+
+      if (!wt->Initialized) {
+         eglMakeCurrent(wt->Display, wt->Surface, wt->Surface, wt->Context);
+         printf("xeglthreads: %d: GL_RENDERER = %s\n", wt->Index,
+                (char *) glGetString(GL_RENDERER));
+         if (Texture /*&& wt->Index == 0*/) {
+            MakeNewTexture(wt);
+         }
+         wt->Initialized = GL_TRUE;
+      }
+
+      if (Locking)
+         pthread_mutex_unlock(&Mutex);
+
+      eglBindAPI(EGL_OPENGL_API);
+      if (eglGetCurrentContext() != wt->Context) {
+         printf("xeglthreads: current context %p != %p\n",
+               eglGetCurrentContext(), wt->Context);
+      }
+
+      glEnable(GL_DEPTH_TEST);
+
+      if (wt->NewSize) {
+         GLfloat w = (float) wt->WinWidth / (float) wt->WinHeight;
+         glViewport(0, 0, wt->WinWidth, wt->WinHeight);
+         glMatrixMode(GL_PROJECTION);
+         glLoadIdentity();
+         glFrustum(-w, w, -1.0, 1.0, 1.5, 10);
+         glMatrixMode(GL_MODELVIEW);
+         glLoadIdentity();
+         glTranslatef(0, 0, -2.5);
+         wt->NewSize = GL_FALSE;
+      }
+
+      if (wt->MakeNewTexture) {
+         MakeNewTexture(wt);
+         wt->MakeNewTexture = GL_FALSE;
+      }
+
+      glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+      glPushMatrix();
+         glRotatef(wt->Angle, 0, 1, 0);
+         glRotatef(wt->Angle, 1, 0, 0);
+         glScalef(0.7, 0.7, 0.7);
+         draw_object();
+      glPopMatrix();
+
+      if (Locking)
+         pthread_mutex_lock(&Mutex);
+
+      eglSwapBuffers(wt->Display, wt->Surface);
+
+      if (Locking)
+         pthread_mutex_unlock(&Mutex);
+
+      if (Animate) {
+         usleep(5000);
+      }
+      else {
+         /* wait for signal to draw */
+         pthread_mutex_lock(&CondMutex);
+         pthread_cond_wait(&CondVar, &CondMutex);
+         pthread_mutex_unlock(&CondMutex);
+      }
+      wt->Angle += 1.0;
+   }
+   eglMakeCurrent(wt->Display, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
+}
+
+
+static void
+keypress(XEvent *event, struct winthread *wt)
+{
+   char buf[100];
+   KeySym keySym;
+   XComposeStatus stat;
+
+   XLookupString(&event->xkey, buf, sizeof(buf), &keySym, &stat);
+
+   switch (keySym) {
+   case XK_Escape:
+      /* tell all threads to exit */
+      if (!Animate) {
+         signal_redraw();
+      }
+      ExitFlag = GL_TRUE;
+      /*printf("exit draw_loop %d\n", wt->Index);*/
+      return;
+   case XK_t:
+   case XK_T:
+      if (Texture) {
+         wt->MakeNewTexture = GL_TRUE;
+         if (!Animate)
+            signal_redraw();
+      }
+      break;
+   case XK_a:
+   case XK_A:
+      Animate = !Animate;
+      if (Animate)  /* yes, prev Animate state! */
+         signal_redraw();
+      break;
+   case XK_s:
+   case XK_S:
+      if (!Animate)
+         signal_redraw();
+      break;
+   default:
+      ; /* nop */
+   }
+}
+
+
+/*
+ * The main process thread runs this loop.
+ * Single display connection for all threads.
+ */
+static void
+event_loop(Display *dpy)
+{
+   XEvent event;
+   int i;
+
+   assert(!MultiDisplays);
+
+   while (!ExitFlag) {
+
+      if (Locking) {
+         while (1) {
+            int k;
+            pthread_mutex_lock(&Mutex);
+            k = XPending(dpy);
+            if (k) {
+               XNextEvent(dpy, &event);
+               pthread_mutex_unlock(&Mutex);
+               break;
+            }
+            pthread_mutex_unlock(&Mutex);
+            usleep(5000);
+         }
+      }
+      else {
+         XNextEvent(dpy, &event);
+      }
+
+      switch (event.type) {
+         case ConfigureNotify:
+            /* Find winthread for this event's window */
+            for (i = 0; i < NumWinThreads; i++) {
+               struct winthread *wt = &WinThreads[i];
+               if (event.xconfigure.window == wt->Win) {
+                  resize(wt, event.xconfigure.width,
+                         event.xconfigure.height);
+                  break;
+               }
+            }
+            break;
+         case KeyPress:
+            for (i = 0; i < NumWinThreads; i++) {
+               struct winthread *wt = &WinThreads[i];
+               if (event.xkey.window == wt->Win) {
+                  keypress(&event, wt);
+                  break;
+               }
+            }
+            break;
+         default:
+            /*no-op*/ ;
+      }
+   }
+}
+
+
+/*
+ * Separate display connection for each thread.
+ */
+static void
+event_loop_multi(void)
+{
+   XEvent event;
+   int w = 0;
+
+   assert(MultiDisplays);
+
+   while (!ExitFlag) {
+      struct winthread *wt = &WinThreads[w];
+      if (XPending(wt->Dpy)) {
+         XNextEvent(wt->Dpy, &event);
+         switch (event.type) {
+         case ConfigureNotify:
+            resize(wt, event.xconfigure.width, event.xconfigure.height);
+            break;
+         case KeyPress:
+            keypress(&event, wt);
+            break;
+         default:
+            ; /* nop */
+         }
+      }
+      w = (w + 1) % NumWinThreads;
+      usleep(5000);
+   }
+}
+
+
+
+/*
+ * we'll call this once for each thread, before the threads are created.
+ */
+static void
+create_window(struct winthread *wt, EGLContext shareCtx)
+{
+   Window win;
+   EGLContext ctx;
+   EGLSurface surf;
+   EGLint attribs[] = { EGL_RED_SIZE, 1,
+                        EGL_GREEN_SIZE, 1,
+                        EGL_BLUE_SIZE, 1,
+                        EGL_DEPTH_SIZE, 1,
+                        EGL_NONE };
+   EGLConfig config;
+   EGLint num_configs;
+   EGLint vid;
+   int scrnum;
+   XSetWindowAttributes attr;
+   unsigned long mask;
+   Window root;
+   XVisualInfo *visinfo, visTemplate;
+   int num_visuals;
+   int width = 160, height = 160;
+   int xpos = (wt->Index % 8) * (width + 10);
+   int ypos = (wt->Index / 8) * (width + 20);
+
+   scrnum = DefaultScreen(wt->Dpy);
+   root = RootWindow(wt->Dpy, scrnum);
+
+   if (!eglChooseConfig(wt->Display, attribs, &config, 1, &num_configs)) {
+      Error("Unable to choose an EGL config");
+   }
+
+   assert(config);
+   assert(num_configs > 0);
+
+   if (!eglGetConfigAttrib(wt->Display, config, EGL_NATIVE_VISUAL_ID, &vid)) {
+      Error("Unable to get visual id of EGL config\n");
+   }
+
+   visTemplate.visualid = vid;
+   visinfo = XGetVisualInfo(wt->Dpy, VisualIDMask,
+                        &visTemplate, &num_visuals);
+   if (!visinfo) {
+      Error("Unable to find RGB, Z, double-buffered visual");
+   }
+
+   /* window attributes */
+   attr.background_pixel = 0;
+   attr.border_pixel = 0;
+   attr.colormap = XCreateColormap(wt->Dpy, root, visinfo->visual, AllocNone);
+   attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask;
+   mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask;
+
+   win = XCreateWindow(wt->Dpy, root, xpos, ypos, width, height,
+                        0, visinfo->depth, InputOutput,
+                        visinfo->visual, mask, &attr);
+   if (!win) {
+      Error("Couldn't create window");
+   }
+
+   XFree(visinfo);
+
+   {
+      XSizeHints sizehints;
+      sizehints.x = xpos;
+      sizehints.y = ypos;
+      sizehints.width  = width;
+      sizehints.height = height;
+      sizehints.flags = USSize | USPosition;
+      XSetNormalHints(wt->Dpy, win, &sizehints);
+      XSetStandardProperties(wt->Dpy, win, "xeglthreads", "xeglthreads",
+                              None, (char **)NULL, 0, &sizehints);
+   }
+
+   eglBindAPI(EGL_OPENGL_API);
+
+   ctx = eglCreateContext(wt->Display, config, shareCtx, NULL);
+   if (!ctx) {
+      Error("Couldn't create EGL context");
+   }
+   surf = eglCreateWindowSurface(wt->Display, config, win, NULL);
+   if (!surf) {
+      Error("Couldn't create EGL surface");
+   }
+
+   XMapWindow(wt->Dpy, win);
+   XSync(wt->Dpy, 0);
+
+   /* save the info for this window/context */
+   wt->Win = win;
+   wt->Context = ctx;
+   wt->Surface = surf;
+   wt->Angle = 0.0;
+   wt->WinWidth = width;
+   wt->WinHeight = height;
+   wt->NewSize = GL_TRUE;
+}
+
+
+/*
+ * Called by pthread_create()
+ */
+static void *
+thread_function(void *p)
+{
+   struct winthread *wt = (struct winthread *) p;
+   draw_loop(wt);
+   return NULL;
+}
+
+
+/*
+ * called before exit to wait for all threads to finish
+ */
+static void
+clean_up(void)
+{
+   int i;
+
+   /* wait for threads to finish */
+   for (i = 0; i < NumWinThreads; i++) {
+      pthread_join(WinThreads[i].Thread, NULL);
+   }
+
+   for (i = 0; i < NumWinThreads; i++) {
+      eglDestroyContext(WinThreads[i].Display, WinThreads[i].Context);
+      XDestroyWindow(WinThreads[i].Dpy, WinThreads[i].Win);
+   }
+}
+
+
+static void
+usage(void)
+{
+   printf("xeglthreads: test of EGL/GL thread safety (any key = exit)\n");
+   printf("Usage:\n");
+   printf("  xeglthreads [options]\n");
+   printf("Options:\n");
+   printf("   -display DISPLAYNAME  Specify display string\n");
+   printf("   -n NUMTHREADS  Number of threads to create\n");
+   printf("   -p  Use a separate display connection for each thread\n");
+   printf("   -l  Use application-side locking\n");
+   printf("   -t  Enable texturing\n");
+   printf("Keyboard:\n");
+   printf("   Esc  Exit\n");
+   printf("   t    Change texture image (requires -t option)\n");
+   printf("   a    Toggle animation\n");
+   printf("   s    Step rotation (when not animating)\n");
+}
+
+
+int
+main(int argc, char *argv[])
+{
+   char *displayName = NULL;
+   int numThreads = 2;
+   Display *dpy = NULL;
+   EGLDisplay *egl_dpy = NULL;
+   int i;
+   Status threadStat;
+
+   if (argc == 1) {
+      usage();
+   }
+   else {
+      int i;
+      for (i = 1; i < argc; i++) {
+         if (strcmp(argv[i], "-display") == 0 && i + 1 < argc) {
+            displayName = argv[i + 1];
+            i++;
+         }
+         else if (strcmp(argv[i], "-p") == 0) {
+            MultiDisplays = 1;
+         }
+         else if (strcmp(argv[i], "-l") == 0) {
+            Locking = 1;
+         }
+         else if (strcmp(argv[i], "-t") == 0) {
+            Texture = 1;
+         }
+         else if (strcmp(argv[i], "-n") == 0 && i + 1 < argc) {
+            numThreads = atoi(argv[i + 1]);
+            if (numThreads < 1)
+               numThreads = 1;
+            else if (numThreads > MAX_WINTHREADS)
+               numThreads = MAX_WINTHREADS;
+            i++;
+         }
+         else {
+            usage();
+            exit(1);
+         }
+      }
+   }
+
+   if (Locking)
+      printf("xeglthreads: Using explicit locks around Xlib calls.\n");
+   else
+      printf("xeglthreads: No explict locking.\n");
+
+   if (MultiDisplays)
+      printf("xeglthreads: Per-thread display connections.\n");
+   else
+      printf("xeglthreads: Single display connection.\n");
+
+   /*
+    * VERY IMPORTANT: call XInitThreads() before any other Xlib functions.
+    */
+   if (!MultiDisplays) {
+       if (!Locking) {
+           threadStat = XInitThreads();
+           if (threadStat) {
+               printf("XInitThreads() returned %d (success)\n",
+                      (int) threadStat);
+           }
+           else {
+               printf("XInitThreads() returned 0 "
+                      "(failure- this program may fail)\n");
+           }
+       }
+
+      dpy = XOpenDisplay(displayName);
+      if (!dpy) {
+         fprintf(stderr, "Unable to open display %s\n",
+                 XDisplayName(displayName));
+         return -1;
+      }
+      egl_dpy = eglGetDisplay(dpy);
+      if (!egl_dpy) {
+         fprintf(stderr, "Unable to get EGL display\n");
+         XCloseDisplay(dpy);
+         return -1;
+      }
+      if (!eglInitialize(egl_dpy, NULL, NULL)) {
+          fprintf(stderr, "Unable to initialize EGL display\n");
+          return -1;
+      }
+   }
+
+   pthread_mutex_init(&Mutex, NULL);
+   pthread_mutex_init(&CondMutex, NULL);
+   pthread_cond_init(&CondVar, NULL);
+
+   printf("xeglthreads: creating windows\n");
+
+   NumWinThreads = numThreads;
+
+   /* Create the EGL windows and contexts */
+   for (i = 0; i < numThreads; i++) {
+      EGLContext share;
+
+      if (MultiDisplays) {
+         WinThreads[i].Dpy = XOpenDisplay(displayName);
+         assert(WinThreads[i].Dpy);
+         WinThreads[i].Display = eglGetDisplay(WinThreads[i].Dpy);
+         assert(eglInitialize(WinThreads[i].Display, NULL, NULL));
+      }
+      else {
+         WinThreads[i].Dpy = dpy;
+         WinThreads[i].Display = egl_dpy;
+      }
+      WinThreads[i].Index = i;
+      WinThreads[i].Initialized = GL_FALSE;
+
+      share = (Texture && i > 0) ? WinThreads[0].Context : 0;
+
+      create_window(&WinThreads[i], share);
+   }
+
+   printf("xeglthreads: creating threads\n");
+
+   /* Create the threads */
+   for (i = 0; i < numThreads; i++) {
+      pthread_create(&WinThreads[i].Thread, NULL, thread_function,
+                     (void*) &WinThreads[i]);
+      printf("xeglthreads: Created thread %p\n",
+              (void *) WinThreads[i].Thread);
+   }
+
+   if (MultiDisplays)
+      event_loop_multi();
+   else
+      event_loop(dpy);
+
+   clean_up();
+
+   if (MultiDisplays) {
+      for (i = 0; i < numThreads; i++) {
+          eglTerminate(WinThreads[i].Display);
+          XCloseDisplay(WinThreads[i].Dpy);
+      }
+   }
+   else {
+      eglTerminate(egl_dpy);
+      XCloseDisplay(dpy);
+   }
+
+   return 0;
+}
+
+
+#else /* PTHREADS */
+
+
+#include <stdio.h>
+
+int
+main(int argc, char *argv[])
+{
+   printf("Sorry, this program wasn't compiled with PTHREADS defined.\n");
+   return 0;
+}
+
+
+#endif /* PTHREADS */
diff --git a/progs/glsl/multitex.c b/progs/glsl/multitex.c
index 913f73be17..5e971716ad 100644
--- a/progs/glsl/multitex.c
+++ b/progs/glsl/multitex.c
@@ -51,6 +51,8 @@ static GLfloat Xrot = 0.0, Yrot = .0, Zrot = 0.0;
 static GLfloat EyeDist = 10;
 static GLboolean Anim = GL_TRUE;
 static GLboolean UseArrays = GL_TRUE;
+static GLboolean UseVBO = GL_TRUE;
+static GLuint VBO = 0;
 
 static GLint VertCoord_attr = -1, TexCoord0_attr = -1, TexCoord1_attr = -1;
 
@@ -76,17 +78,68 @@ static const GLfloat VertCoords[4][2] = {
 };
 
 
+
+static void
+SetupVertexBuffer(void)
+{
+   glGenBuffersARB(1, &VBO);
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, VBO);
+
+   glBufferDataARB(GL_ARRAY_BUFFER_ARB,
+                        sizeof(VertCoords) +
+                        sizeof(Tex0Coords) +
+                        sizeof(Tex1Coords),
+                        NULL,
+                        GL_STATIC_DRAW_ARB);
+
+   /* non-interleaved vertex arrays */
+
+   glBufferSubDataARB(GL_ARRAY_BUFFER_ARB,
+                           0,                   /* offset */
+                           sizeof(VertCoords),  /* size */
+                           VertCoords);         /* data */
+
+   glBufferSubDataARB(GL_ARRAY_BUFFER_ARB,
+                           sizeof(VertCoords),  /* offset */
+                           sizeof(Tex0Coords),  /* size */
+                           Tex0Coords);         /* data */
+
+   glBufferSubDataARB(GL_ARRAY_BUFFER_ARB,
+                           sizeof(VertCoords) +
+                           sizeof(Tex0Coords),  /* offset */
+                           sizeof(Tex1Coords),  /* size */
+                           Tex1Coords);         /* data */
+
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+}
+
+
 static void
 DrawPolygonArray(void)
 {
+   void *vertPtr, *tex0Ptr, *tex1Ptr;
+
+   if (UseVBO) {
+      glBindBufferARB(GL_ARRAY_BUFFER_ARB, VBO);
+      vertPtr = (void *) 0;
+      tex0Ptr = (void *) sizeof(VertCoords);
+      tex1Ptr = (void *) (sizeof(VertCoords) + sizeof(Tex0Coords));
+   }
+   else {
+      glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+      vertPtr = VertCoords;
+      tex0Ptr = Tex0Coords;
+      tex1Ptr = Tex1Coords;
+   }
+
    if (VertCoord_attr >= 0) {
       glVertexAttribPointer(VertCoord_attr, 2, GL_FLOAT, GL_FALSE,
                                  0, VertCoords);
       glEnableVertexAttribArray(VertCoord_attr);
    }
    else {
-      glVertexPointer(2, GL_FLOAT, 0, VertCoords);
-      glEnable(GL_VERTEX_ARRAY);
+      glVertexPointer(2, GL_FLOAT, 0, vertPtr);
+      glEnableClientState(GL_VERTEX_ARRAY);
    }
 
    glVertexAttribPointer(TexCoord0_attr, 2, GL_FLOAT, GL_FALSE,
@@ -98,6 +151,8 @@ DrawPolygonArray(void)
    glEnableVertexAttribArray(TexCoord1_attr);
 
    glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
 }
 
 
@@ -163,6 +218,10 @@ key(unsigned char k, int x, int y)
       UseArrays = !UseArrays;
       printf("Arrays: %d\n", UseArrays);
       break;
+   case 'v':
+      UseVBO = !UseVBO;
+      printf("Use VBO: %d\n", UseVBO);
+      break;
    case ' ':
       Anim = !Anim;
       if (Anim)
@@ -314,10 +373,18 @@ InitGL(void)
       /*exit(1);*/
    }
    printf("GL_RENDERER = %s\n",(const char *) glGetString(GL_RENDERER));
+   printf("Usage:\n");
+   printf("  a     - toggle arrays vs. immediate mode rendering\n");
+   printf("  v     - toggle VBO usage for array rendering\n");
+   printf("  z/Z   - change viewing distance\n");
+   printf("  SPACE - toggle animation\n");
+   printf("  Esc   - exit\n");
 
    InitTextures();
    InitPrograms();
 
+   SetupVertexBuffer();
+
    glEnable(GL_DEPTH_TEST);
 
    glClearColor(.6, .6, .9, 0);
diff --git a/progs/openvg/demos/Makefile b/progs/openvg/demos/Makefile
new file mode 100644
index 0000000000..6e15342c7f
--- /dev/null
+++ b/progs/openvg/demos/Makefile
@@ -0,0 +1,39 @@
+# progs/vg/Makefile
+
+TOP = ../../..
+include $(TOP)/configs/current
+
+VG_LIBS=-lm -pthread -lEGL -lOpenVG
+INCLUDE_DIRS = -I$(TOP)/include
+
+PROGRAMS = \
+        lion \
+        sp
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+
+default: $(PROGRAMS)
+
+lion: lion.o lion-render.o
+	$(CC) $(CFLAGS) lion.o lion-render.o -L$(TOP)/$(LIB_DIR) $(VG_LIBS) -o $@
+
+lion.o: lion.c lion-render.h $(HEADERS)
+	$(CC) -c $(CFLAGS) -I$(TOP)/include lion.c
+lion-render.o: lion-render.c lion-render.h $(HEADERS)
+	$(CC) -c $(CFLAGS) -I$(TOP)/include lion-render.c
+
+
+sp: sp.c eglcommon.o
+	$(CC) $(INCLUDE_DIRS) $(CFLAGS) $^  -L$(TOP)/$(LIB_DIR) $(LIBS) $(VG_LIBS)  $(APP_LIB_DEPS) -o $@
+
+eglcommon.o: eglcommon.c $(HEADERS)
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) eglcommon.c
+
+
+clean:
+	rm -f *.o *~
+	rm -f *.so
+	rm -f $(PROGRAMS)
diff --git a/progs/openvg/demos/eglcommon.c b/progs/openvg/demos/eglcommon.c
new file mode 100644
index 0000000000..bacd5685d7
--- /dev/null
+++ b/progs/openvg/demos/eglcommon.c
@@ -0,0 +1,288 @@
+#include "eglcommon.h"
+
+
+#include <assert.h>
+#include <math.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include <X11/keysym.h>
+#include <VG/openvg.h>  /* using full OpenGL for now */
+#include <GLES/egl.h>
+
+
+static init_func    init = 0;
+static draw_func    draw = 0;
+static reshape_func reshape = 0;
+static key_func     keyPress = 0;
+static VGint width = 300, height = 300;
+
+
+void set_window_size(int w, int h)
+{
+   width = w;
+   height = h;
+}
+
+/*
+ * Create an RGB, double-buffered X window.
+ * Return the window and context handles.
+ */
+static void
+make_x_window(Display *x_dpy, EGLDisplay egl_dpy,
+              const char *name,
+              int x, int y, int width, int height,
+              Window *winRet,
+              EGLContext *ctxRet,
+              EGLSurface *surfRet)
+{
+   static const EGLint attribs[] = {
+      EGL_RED_SIZE, 1,
+      EGL_GREEN_SIZE, 1,
+      EGL_BLUE_SIZE, 1,
+      EGL_NONE
+   };
+
+   int scrnum;
+   XSetWindowAttributes attr;
+   unsigned long mask;
+   Window root;
+   Window win;
+   XVisualInfo *visInfo, visTemplate;
+   int num_visuals;
+   EGLContext ctx;
+   EGLConfig config;
+   EGLint num_configs;
+   EGLint vid;
+
+   scrnum = DefaultScreen( x_dpy );
+   root = RootWindow( x_dpy, scrnum );
+
+   if (!eglChooseConfig( egl_dpy, attribs, &config, 1, &num_configs)) {
+      printf("Error: couldn't get an EGL visual config\n");
+      exit(1);
+   }
+
+   assert(config);
+   assert(num_configs > 0);
+
+   if (!eglGetConfigAttrib(egl_dpy, config, EGL_NATIVE_VISUAL_ID, &vid)) {
+      printf("Error: eglGetConfigAttrib() failed\n");
+      exit(1);
+   }
+
+   /* The X window visual must match the EGL config */
+   visTemplate.visualid = vid;
+   visInfo = XGetVisualInfo(x_dpy, VisualIDMask, &visTemplate, &num_visuals);
+   if (!visInfo) {
+      printf("Error: couldn't get X visual\n");
+      exit(1);
+   }
+
+   /* window attributes */
+   attr.background_pixel = 0;
+   attr.border_pixel = 0;
+   attr.colormap = XCreateColormap( x_dpy, root, visInfo->visual, AllocNone);
+   attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask;
+   mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask;
+
+   win = XCreateWindow( x_dpy, root, 0, 0, width, height,
+		        0, visInfo->depth, InputOutput,
+		        visInfo->visual, mask, &attr );
+
+   /* set hints and properties */
+   {
+      XSizeHints sizehints;
+      sizehints.x = x;
+      sizehints.y = y;
+      sizehints.width  = width;
+      sizehints.height = height;
+      sizehints.flags = USSize | USPosition;
+      XSetNormalHints(x_dpy, win, &sizehints);
+      XSetStandardProperties(x_dpy, win, name, name,
+                              None, (char **)NULL, 0, &sizehints);
+   }
+
+   eglBindAPI(EGL_OPENVG_API);
+
+   ctx = eglCreateContext(egl_dpy, config, EGL_NO_CONTEXT, NULL );
+   if (!ctx) {
+      printf("Error: eglCreateContext failed\n");
+      exit(1);
+   }
+
+   *surfRet = eglCreateWindowSurface(egl_dpy, config, win, NULL);
+
+   if (!*surfRet) {
+      printf("Error: eglCreateWindowSurface failed\n");
+      exit(1);
+   }
+
+   XFree(visInfo);
+
+   *winRet = win;
+   *ctxRet = ctx;
+}
+
+static void
+event_loop(Display *dpy, Window win,
+           EGLDisplay egl_dpy, EGLSurface egl_surf)
+{
+   while (1) {
+      int redraw = 0;
+      XEvent event;
+
+      XNextEvent(dpy, &event);
+
+      switch (event.type) {
+      case Expose:
+         redraw = 1;
+         break;
+      case ConfigureNotify:
+         if (reshape) {
+            width = event.xconfigure.width;
+            height = event.xconfigure.height;
+            reshape(event.xconfigure.width, event.xconfigure.height);
+         }
+         break;
+      case KeyPress:
+      {
+         char buffer[10];
+         int r, code;
+         code = XLookupKeysym(&event.xkey, 0);
+         if (!keyPress || !keyPress(code)) {
+            r = XLookupString(&event.xkey, buffer, sizeof(buffer),
+                              NULL, NULL);
+            if (buffer[0] == 27) {
+               /* escape */
+               return;
+            }
+         }
+      }
+      redraw = 1;
+      break;
+      default:
+         ; /*no-op*/
+      }
+
+      if (redraw) {
+         draw();
+         eglSwapBuffers(egl_dpy, egl_surf);
+      }
+   }
+}
+
+int window_width(void)
+{
+   return width;
+}
+
+int window_height(void)
+{
+   return height;
+}
+
+static void
+usage(void)
+{
+   printf("Usage:\n");
+   printf("  -display <displayname>  set the display to run on\n");
+   printf("  -info                   display OpenGL renderer info\n");
+}
+
+int run(int argc, char **argv,
+        init_func init_f,
+        reshape_func resh_f,
+        draw_func draw_f,
+        key_func key_f)
+{
+   const int winWidth = width, winHeight = height;
+   Display *x_dpy;
+   Window win;
+   EGLSurface egl_surf;
+   EGLContext egl_ctx;
+   EGLDisplay egl_dpy;
+   char *dpyName = NULL;
+   GLboolean printInfo = GL_FALSE;
+   EGLint egl_major, egl_minor;
+   int i;
+   const char *s;
+
+   init = init_f;
+   draw = draw_f;
+   reshape = resh_f;
+   keyPress = key_f;
+
+   for (i = 1; i < argc; i++) {
+      if (strcmp(argv[i], "-display") == 0) {
+         dpyName = argv[i+1];
+         i++;
+      }
+      else if (strcmp(argv[i], "-info") == 0) {
+         printInfo = GL_TRUE;
+      }
+   }
+
+   x_dpy = XOpenDisplay(dpyName);
+   if (!x_dpy) {
+      printf("Error: couldn't open display %s\n",
+	     dpyName ? dpyName : getenv("DISPLAY"));
+      return -1;
+   }
+
+   egl_dpy = eglGetDisplay(x_dpy);
+   if (!egl_dpy) {
+      printf("Error: eglGetDisplay() failed\n");
+      return -1;
+   }
+
+   if (!eglInitialize(egl_dpy, &egl_major, &egl_minor)) {
+      printf("Error: eglInitialize() failed\n");
+      return -1;
+   }
+
+   s = eglQueryString(egl_dpy, EGL_VERSION);
+   printf("EGL_VERSION = %s\n", s);
+
+   make_x_window(x_dpy, egl_dpy,
+                 "OpenVG Example", 0, 0, winWidth, winHeight,
+                 &win, &egl_ctx, &egl_surf);
+
+   XMapWindow(x_dpy, win);
+   if (!eglMakeCurrent(egl_dpy, egl_surf, egl_surf, egl_ctx)) {
+      printf("Error: eglMakeCurrent() failed\n");
+      return -1;
+   }
+
+   if (printInfo) {
+      printf("VG_RENDERER   = %s\n", (char *) vgGetString(VG_RENDERER));
+      printf("VG_VERSION    = %s\n", (char *) vgGetString(VG_VERSION));
+      printf("VG_VENDOR     = %s\n", (char *) vgGetString(VG_VENDOR));
+   }
+
+   if (init)
+      init();
+
+   /* Set initial projection/viewing transformation.
+    * We can't be sure we'll get a ConfigureNotify event when the window
+    * first appears.
+    */
+   if (reshape)
+      reshape(winWidth, winHeight);
+
+   event_loop(x_dpy, win, egl_dpy, egl_surf);
+
+   eglMakeCurrent(egl_dpy, 0, 0, 0);
+   eglDestroyContext(egl_dpy, egl_ctx);
+   eglDestroySurface(egl_dpy, egl_surf);
+   eglTerminate(egl_dpy);
+
+
+   XDestroyWindow(x_dpy, win);
+   XCloseDisplay(x_dpy);
+
+   return 0;
+}
+
diff --git a/progs/openvg/demos/eglcommon.h b/progs/openvg/demos/eglcommon.h
new file mode 100644
index 0000000000..958dae9f98
--- /dev/null
+++ b/progs/openvg/demos/eglcommon.h
@@ -0,0 +1,20 @@
+#ifndef EGLCOMMON_H
+#define EGLCOMMON_H
+
+typedef void (*init_func)();
+typedef void (*reshape_func)(int, int);
+typedef void (*draw_func)();
+typedef int  (*key_func)(unsigned key);
+
+
+void set_window_size(int width, int height);
+int window_width(void);
+int window_height(void);
+
+int run(int argc, char **argv,
+        init_func init,
+        reshape_func resh,
+        draw_func draw,
+        key_func key);
+
+#endif
diff --git a/progs/openvg/demos/lion-render.c b/progs/openvg/demos/lion-render.c
new file mode 100644
index 0000000000..f3f151f552
--- /dev/null
+++ b/progs/openvg/demos/lion-render.c
@@ -0,0 +1,1573 @@
+#include "lion-render.h"
+
+#include <stdlib.h>
+#include <stdio.h>
+
+#define ELEMENTS(x) (sizeof(x)/sizeof((x)[0]))
+
+static void init(struct lion *l, int i, VGint hexColor, const VGfloat *coords, int elems)
+{
+   static VGubyte cmds[128];
+   VGfloat color[4];
+   VGint j;
+
+   color[0] = ((hexColor >> 16) & 0xff) / 255.f;
+   color[1] = ((hexColor >> 8) & 0xff) / 255.f;
+   color[2] = ((hexColor >> 0) & 0xff) / 255.f;
+   color[3] = 1.0;
+
+   l->paths[i] = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1.0f, 0.0f,
+                              0, 0, (unsigned int)VG_PATH_CAPABILITY_ALL);
+   l->fills[i] = vgCreatePaint();
+   vgSetParameterfv(l->fills[i], VG_PAINT_COLOR, 4, color);
+
+   cmds[0] = VG_MOVE_TO_ABS;
+   for (j = 1; j < elems; ++j) {
+      cmds[j] = VG_LINE_TO_ABS;
+   }
+
+   vgAppendPathData(l->paths[i], elems, cmds, coords);
+}
+
+static void poly0(struct lion *l)
+{
+   VGfloat color = 0xf2cc99;
+   static const VGfloat coords[] = {69,18, 82,8, 99,3, 118,5, 135,12, 149,21, 156,13, 165,9, 177,13, 183,28,
+                                    180,50, 164,91, 155,107, 154,114, 151,121, 141,127, 139,136, 155,206, 157,251, 126,342,
+                                    133,357, 128,376, 83,376, 75,368, 67,350, 61,350, 53,369, 4,369, 2,361, 5,354,
+                                    12,342, 16,321, 4,257, 4,244, 7,218, 9,179, 26,127, 43,93, 32,77, 30,70,
+                                    24,67, 16,49, 17,35, 18,23, 30,12, 40,7, 53,7, 62,12
+   };
+
+   init(l, 0, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly1(struct lion *l)
+{
+   VGfloat color = 0xe5b27f;
+   static const VGfloat coords[] = {142,79, 136,74, 138,82, 133,78, 133,84, 127,78, 128,85,
+                                    124,80, 125,87, 119,82, 119,90, 125,99, 125,96, 128,100, 128,94,
+                                    131,98, 132,93, 135,97, 136,93, 138,97, 139,94, 141,98, 143,94,
+                                    144,85
+   };
+
+   init(l, 1, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly2(struct lion *l)
+{
+   VGfloat color = 0xeb8080;
+   static const VGfloat coords[] = {127,101, 132,100, 137,99, 144,101, 143,105, 135,110
+   };
+
+   init(l, 2, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly3(struct lion *l)
+{
+   VGfloat color = 0xf2cc99;
+   static const VGfloat coords[] = {178,229, 157,248, 139,296, 126,349, 137,356,
+                                    158,357, 183,342, 212,332, 235,288, 235,261,
+                                    228,252, 212,250, 188,251
+   };
+
+   init(l, 3, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly4(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {56,229, 48,241, 48,250, 57,281, 63,325, 71,338,
+                                    81,315, 76,321, 79,311, 83,301, 75,308, 80,298,
+                                    73,303, 76,296, 71,298, 74,292, 69,293, 74,284,
+                                    78,278, 71,278, 74,274, 68,273, 70,268, 66,267,
+                                    68,261, 60,266, 62,259, 65,253, 57,258, 59,251,
+                                    55,254, 55,248, 60,237, 54,240, 58,234, 54,236
+   };
+
+   init(l, 4, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly5(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {74,363, 79,368, 81,368, 85,362, 89,363, 92,370, 96,373,
+                                    101,372, 108,361, 110,371, 113,373, 116,371, 120,358, 122,363,
+                                    123,371, 126,371, 129,367, 132,357, 135,361, 130,376, 127,377,
+                                    94,378, 84,376, 76,371
+   };
+
+   init(l, 5, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly6(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {212,250, 219,251, 228,258, 236,270, 235,287, 225,304,
+                                    205,332, 177,343, 171,352, 158,357, 166,352, 168,346,
+                                    168,339, 165,333, 155,327, 155,323, 161,320, 165,316,
+                                    169,316, 167,312, 171,313, 168,308, 173,309, 170,306,
+                                    177,306, 175,308, 177,311, 174,311, 176,316, 171,315,
+                                    174,319, 168,320, 168,323, 175,327, 179,332, 183,326,
+                                    184,332, 189,323, 190,328, 194,320, 194,325, 199,316,
+                                    201,320, 204,313, 206,316, 208,310, 211,305, 219,298,
+                                    226,288, 229,279, 228,266, 224,259, 217,253
+   };
+
+   init(l, 6, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly7(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {151,205, 151,238, 149,252, 141,268, 128,282, 121,301,
+                                    130,300, 126,313, 118,324, 116,337, 120,346, 133,352,
+                                    133,340, 137,333, 145,329, 156,327, 153,319, 153,291,
+                                    157,271, 170,259, 178,277, 193,250, 174,216
+   };
+
+   init(l, 7, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly8(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {78,127, 90,142, 95,155, 108,164, 125,167, 139,175,
+                                    150,206, 152,191, 141,140, 121,148, 100,136
+   };
+
+   init(l, 8, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly9(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {21,58, 35,63, 38,68, 32,69, 42,74, 40,79, 47,80, 54,83,
+                                    45,94, 34,81, 32,73, 24,66
+   };
+
+   init(l, 9, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly10(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {71,34, 67,34, 66,27, 59,24, 54,17, 48,17, 39,22,
+                                    30,26, 28,31, 31,39, 38,46, 29,45, 36,54, 41,61,
+                                    41,70, 50,69, 54,71, 55,58, 67,52, 76,43, 76,39,
+                                    68,44
+   };
+
+   init(l, 10, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly11(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {139,74, 141,83, 143,89, 144,104, 148,104, 155,106,
+                                    154,86, 157,77, 155,72, 150,77, 144,77
+   };
+
+   init(l, 11, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly12(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {105,44, 102,53, 108,58, 111,62, 112,55
+   };
+
+   init(l, 12, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly13(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {141,48, 141,54, 144,58, 139,62, 137,66, 136,59, 137,52
+   };
+
+   init(l, 13, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly14(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {98,135, 104,130, 105,134, 108,132, 108,135, 112,134,
+                                    113,137, 116,136, 116,139, 119,139, 124,141, 128,140,
+                                    133,138, 140,133, 139,140, 126,146, 104,144
+   };
+
+   init(l, 14, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly15(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {97,116, 103,119, 103,116, 111,118, 116,117, 122,114,
+                                    127,107, 135,111, 142,107, 141,114, 145,118, 149,121,
+                                    145,125, 140,124, 127,121, 113,125, 100,124
+   };
+
+   init(l, 15, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly16(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {147,33, 152,35, 157,34, 153,31, 160,31, 156,28, 161,28,
+                                    159,24, 163,25, 163,21, 165,22, 170,23, 167,17, 172,21,
+                                    174,18, 175,23, 176,22, 177,28, 177,33, 174,37, 176,39,
+                                    174,44, 171,49, 168,53, 164,57, 159,68, 156,70, 154,60,
+                                    150,51, 146,43, 144,35
+   };
+
+   init(l, 16, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly17(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {85,72, 89,74, 93,75, 100,76, 105,75, 102,79, 94,79, 88,76
+   };
+
+   init(l, 17, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly18(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {86,214, 79,221, 76,232, 82,225, 78,239, 82,234, 78,245,
+                                    81,243, 79,255, 84,250, 84,267, 87,254, 90,271, 90,257,
+                                    95,271, 93,256, 95,249, 92,252, 93,243, 89,253, 89,241,
+                                    86,250, 87,236, 83,245, 87,231, 82,231, 90,219, 84,221
+   };
+
+   init(l, 18, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly19(struct lion *l)
+{
+   VGfloat color = 0xffcc7f;
+   static const VGfloat coords[] = {93,68, 96,72, 100,73, 106,72, 108,66, 105,63, 100,62
+   };
+
+   init(l, 19, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly20(struct lion *l)
+{
+   VGfloat color = 0xffcc7f;
+   static const VGfloat coords[] = {144,64, 142,68, 142,73, 146,74, 150,73, 154,64, 149,62
+   };
+
+   init(l, 20, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly21(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {57,91, 42,111, 52,105, 41,117, 53,112, 46,120, 53,116,
+                                    50,124, 57,119, 55,127, 61,122, 60,130, 67,126, 66,134,
+                                    71,129, 72,136, 77,130, 76,137, 80,133, 82,138, 86,135,
+                                    96,135, 94,129, 86,124, 83,117, 77,123, 79,117, 73,120,
+                                    75,112, 68,116, 71,111, 65,114, 69,107, 63,110, 68,102,
+                                    61,107, 66,98, 61,103, 63,97, 57,99
+   };
+
+   init(l, 21, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly22(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {83,79, 76,79, 67,82, 75,83, 65,88, 76,87, 65,92, 76,91,
+                                    68,96, 77,95, 70,99, 80,98, 72,104, 80,102, 76,108, 85,103,
+                                    92,101, 87,98, 93,96, 86,94, 91,93, 85,91, 93,89, 99,89, 105,93,
+                                    107,85, 102,82, 92,80
+   };
+
+   init(l, 22, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly23(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {109,77, 111,83, 109,89, 113,94, 117,90, 117,81, 114,78
+   };
+
+   init(l, 23, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly24(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {122,128, 127,126, 134,127, 136,129, 134,130, 130,128, 124,129
+   };
+
+   init(l, 24, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly25(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {78,27, 82,32, 80,33, 82,36, 78,37, 82,40, 78,42, 81,46, 76,47,
+                                    78,49, 74,50, 82,52, 87,50, 83,48, 91,46, 86,45, 91,42, 88,40,
+                                    92,37, 86,34, 90,31, 86,29, 89,26
+   };
+
+   init(l, 25, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly26(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {82,17, 92,20, 79,21, 90,25, 81,25, 94,28, 93,26, 101,30,
+                                    101,26, 107,33, 108,28, 111,40, 113,34, 115,45, 117,39,
+                                    119,54, 121,46, 124,58, 126,47, 129,59, 130,49, 134,58,
+                                    133,44, 137,48, 133,37, 137,40, 133,32, 126,20, 135,26,
+                                    132,19, 138,23, 135,17, 142,18, 132,11, 116,6, 94,6, 78,11,
+                                    92,12, 80,14, 90,16
+   };
+
+   init(l, 26, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly27(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {142,234, 132,227, 124,223, 115,220, 110,225, 118,224, 127,229,
+                                    135,236, 122,234, 115,237, 113,242, 121,238, 139,243, 121,245,
+                                    111,254, 95,254, 102,244, 104,235, 110,229, 100,231, 104,224,
+                                    113,216, 122,215, 132,217, 141,224, 145,230, 149,240
+   };
+
+   init(l, 27, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly28(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {115,252, 125,248, 137,249, 143,258, 134,255, 125,254
+   };
+
+   init(l, 28, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly29(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {114,212, 130,213, 140,219, 147,225, 144,214, 137,209, 128,207
+   };
+
+   init(l, 29, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly30(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {102,263, 108,258, 117,257, 131,258, 116,260, 109,265
+   };
+
+   init(l, 30, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly31(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {51,241, 35,224, 40,238, 23,224, 31,242, 19,239, 28,247, 17,246,
+                                    25,250, 37,254, 39,263, 44,271, 47,294, 48,317, 51,328, 60,351,
+                                    60,323, 53,262, 47,246
+   };
+
+   init(l, 31, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly32(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {2,364, 9,367, 14,366, 18,355, 20,364, 26,366, 31,357, 35,364,
+                                    39,364, 42,357, 47,363, 53,360, 59,357, 54,369, 7,373
+   };
+
+   init(l, 32, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly33(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {7,349, 19,345, 25,339, 18,341, 23,333, 28,326, 23,326, 27,320,
+                                    23,316, 25,311, 20,298, 15,277, 12,264, 9,249, 10,223, 3,248,
+                                    5,261, 15,307, 17,326, 11,343
+   };
+
+   init(l, 33, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly34(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {11,226, 15,231, 25,236, 18,227
+   };
+
+   init(l, 34, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly35(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {13,214, 19,217, 32,227, 23,214, 16,208, 15,190, 24,148,
+                                    31,121, 24,137, 14,170, 8,189
+   };
+
+   init(l, 35, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly36(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {202,254, 195,258, 199,260, 193,263, 197,263, 190,268,
+                                    196,268, 191,273, 188,282, 200,272, 194,272, 201,266,
+                                    197,265, 204,262, 200,258, 204,256
+   };
+
+   init(l, 36, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly37(struct lion *l)
+{
+   VGfloat color = 0x845433;
+   static const VGfloat coords[] = {151,213, 165,212, 179,225, 189,246, 187,262, 179,275,
+                                    176,263, 177,247, 171,233, 163,230, 165,251, 157,264,
+                                    146,298, 145,321, 133,326, 143,285, 154,260, 153,240
+   };
+
+   init(l, 37, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly38(struct lion *l)
+{
+   VGfloat color = 0x845433;
+   static const VGfloat coords[] = {91,132, 95,145, 97,154, 104,148, 107,155, 109,150, 111,158,
+                                    115,152, 118,159, 120,153, 125,161, 126,155, 133,164, 132,154,
+                                    137,163, 137,152, 142,163, 147,186, 152,192, 148,167, 141,143,
+                                    124,145, 105,143
+   };
+
+   init(l, 38, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly39(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {31,57, 23,52, 26,51, 20,44, 23,42, 21,36, 22,29, 25,23,
+                                    24,32, 30,43, 26,41, 30,50, 26,48
+   };
+
+   init(l, 39, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly40(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {147,21, 149,28, 155,21, 161,16, 167,14, 175,15, 173,11, 161,9
+   };
+
+   init(l, 40, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly41(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {181,39, 175,51, 169,57, 171,65, 165,68, 165,75, 160,76,
+                                    162,91, 171,71, 180,51
+   };
+
+   init(l, 41, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly42(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {132,346, 139,348, 141,346, 142,341, 147,342, 143,355, 133,350
+   };
+
+   init(l, 42, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly43(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {146,355, 151,352, 155,348, 157,343, 160,349, 151,356, 147,357
+   };
+
+   init(l, 43, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly44(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {99,266, 100,281, 94,305, 86,322, 78,332, 72,346, 73,331, 91,291
+   };
+
+   init(l, 44, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly45(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {20,347, 32,342, 45,340, 54,345, 45,350, 42,353, 38,350,
+                                    31,353, 29,356, 23,350, 19,353, 15,349
+   };
+
+   init(l, 45, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly46(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {78,344, 86,344, 92,349, 88,358, 84,352
+   };
+
+   init(l, 46, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly47(struct lion *l)
+{
+   VGfloat color = 0x9c826b;
+   static const VGfloat coords[] = {93,347, 104,344, 117,345, 124,354, 121,357, 116,351,
+                                    112,351, 108,355, 102,351
+   };
+
+   init(l, 47, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly48(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {105,12, 111,18, 113,24, 113,29, 119,34, 116,23, 112,16
+   };
+
+   init(l, 48, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly49(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {122,27, 125,34, 127,43, 128,34, 125,29
+   };
+
+   init(l, 49, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly50(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {115,13, 122,19, 122,15, 113,10
+   };
+
+   init(l, 50, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly51(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {116,172, 107,182, 98,193, 98,183, 90,199, 89,189, 84,207,
+                                    88,206, 87,215, 95,206, 93,219, 91,230, 98,216, 97,226,
+                                    104,214, 112,209, 104,208, 113,202, 126,200, 139,207, 132,198,
+                                    142,203, 134,192, 142,195, 134,187, 140,185, 130,181, 136,177,
+                                    126,177, 125,171, 116,180
+   };
+
+   init(l, 51, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly52(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {74,220, 67,230, 67,221, 59,235, 63,233, 60,248, 70,232, 65,249,
+                                    71,243, 67,256, 73,250, 69,262, 73,259, 71,267, 76,262, 72,271,
+                                    78,270, 76,275, 82,274, 78,290, 86,279, 86,289, 92,274, 88,275,
+                                    87,264, 82,270, 82,258, 77,257, 78,247, 73,246, 77,233, 72,236
+   };
+
+   init(l, 52, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly53(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {133,230, 147,242, 148,250, 145,254, 138,247, 129,246, 142,245,
+                                    138,241, 128,237, 137,238
+   };
+
+   init(l, 53, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly54(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {133,261, 125,261, 116,263, 111,267, 125,265
+   };
+
+   init(l, 54, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly55(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {121,271, 109,273, 103,279, 99,305, 92,316, 85,327, 83,335,
+                                    89,340, 97,341, 94,336, 101,336, 96,331, 103,330, 97,327, 108,325,
+                                    99,322, 109,321, 100,318, 110,317, 105,314, 110,312, 107,310, 113,308,
+                                    105,306, 114,303, 105,301, 115,298, 107,295, 115,294, 108,293, 117,291,
+                                    109,289, 117,286, 109,286, 118,283, 112,281, 118,279, 114,278,
+                                    119,276, 115,274
+   };
+
+   init(l, 55, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly56(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {79,364, 74,359, 74,353, 76,347, 80,351, 83,356, 82,360
+   };
+
+   init(l, 56, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly57(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {91,363, 93,356, 97,353, 103,355, 105,360, 103,366, 99,371, 94,368
+   };
+
+   init(l, 57, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly58(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {110,355, 114,353, 118,357, 117,363, 113,369, 111,362
+   };
+
+   init(l, 58, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly59(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {126,354, 123,358, 124,367, 126,369, 129,361, 129,357
+   };
+
+   init(l, 59, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly60(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {30,154, 24,166, 20,182, 23,194, 29,208, 37,218, 41,210, 41,223,
+                                    46,214, 46,227, 52,216, 52,227, 61,216, 59,225, 68,213, 73,219,
+                                    70,207, 77,212, 69,200, 77,202, 70,194, 78,197, 68,187, 76,182,
+                                    64,182, 58,175, 58,185, 53,177, 50,186, 46,171, 44,182, 39,167,
+                                    36,172, 36,162, 30,166
+   };
+
+   init(l, 60, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly61(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {44,130, 41,137, 45,136, 43,150, 48,142, 48,157, 53,150,
+                                    52,164, 60,156, 61,169, 64,165, 66,175, 70,167, 74,176,
+                                    77,168, 80,183, 85,172, 90,182, 93,174, 98,181, 99,173,
+                                    104,175, 105,169, 114,168, 102,163, 95,157, 94,166, 90,154,
+                                    87,162, 82,149, 75,159, 72,148, 68,155, 67,143, 62,148, 62,138,
+                                    58,145, 56,133, 52,142, 52,128, 49,134, 47,125
+   };
+
+   init(l, 61, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly62(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {13,216, 19,219, 36,231, 22,223, 16,222, 22,227, 12,224, 13,220, 16,220
+   };
+
+   init(l, 62, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly63(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {10,231, 14,236, 25,239, 27,237, 19,234
+   };
+
+   init(l, 63, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly64(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {9,245, 14,242, 25,245, 13,245
+   };
+
+   init(l, 64, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly65(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {33,255, 26,253, 18,254, 25,256, 18,258, 27,260, 18,263,
+                                    27,265, 19,267, 29,270, 21,272, 29,276, 21,278, 30,281,
+                                    22,283, 31,287, 24,288, 32,292, 23,293, 34,298, 26,299,
+                                    37,303, 32,305, 39,309, 33,309, 39,314, 34,314, 40,318,
+                                    34,317, 40,321, 34,321, 41,326, 33,326, 40,330, 33,332,
+                                    39,333, 33,337, 42,337, 54,341, 49,337, 52,335, 47,330,
+                                    50,330, 45,325, 49,325, 45,321, 48,321, 45,316, 46,306,
+                                    45,286, 43,274, 36,261
+   };
+
+   init(l, 65, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly66(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {7,358, 9,351, 14,351, 17,359, 11,364
+   };
+
+   init(l, 66, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly67(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {44,354, 49,351, 52,355, 49,361
+   };
+
+   init(l, 67, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly68(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {32,357, 37,353, 40,358, 36,361
+   };
+
+   init(l, 68, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly69(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {139,334, 145,330, 154,330, 158,334, 154,341, 152,348,
+                                    145,350, 149,340, 147,336, 141,339, 139,345, 136,342,
+                                    136,339
+   };
+
+   init(l, 69, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly70(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {208,259, 215,259, 212,255, 220,259, 224,263, 225,274, 224,283,
+                                    220,292, 208,300, 206,308, 203,304, 199,315, 197,309, 195,318,
+                                    193,313, 190,322, 190,316, 185,325, 182,318, 180,325, 172,321,
+                                    178,320, 176,313, 186,312, 180,307, 188,307, 184,303, 191,302,
+                                    186,299, 195,294, 187,290, 197,288, 192,286, 201,283, 194,280,
+                                    203,277, 198,275, 207,271, 200,269, 209,265, 204,265, 212,262
+   };
+
+   init(l, 70, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly71(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {106,126, 106,131, 109,132, 111,134, 115,132, 115,135, 119,133, 118,137,
+                                    123,137, 128,137, 133,134, 136,130, 136,127, 132,124, 118,128, 112,128,
+                                    106,126, 106,126, 106,126
+   };
+
+   init(l, 71, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly72(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {107,114, 101,110, 98,102, 105,97, 111,98, 119,102, 121,108, 118,112, 113,115
+   };
+
+   init(l, 72, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly73(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {148,106, 145,110, 146,116, 150,118, 152,111, 151,107
+   };
+
+   init(l, 73, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly74(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {80,55, 70,52, 75,58, 63,57, 72,61, 57,61, 67,66, 57,67, 62,69, 54,71,
+                                    61,73, 54,77, 63,78, 53,85, 60,84, 56,90, 69,84, 63,82, 75,76, 70,75,
+                                    77,72, 72,71, 78,69, 72,66, 81,67, 78,64, 82,63, 80,60, 86,62
+   };
+
+   init(l, 74, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly75(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {87,56, 91,52, 96,50, 102,56, 98,56, 92,60
+   };
+
+   init(l, 75, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly76(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {85,68, 89,73, 98,76, 106,74, 96,73, 91,70
+   };
+
+   init(l, 76, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly77(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {115,57, 114,64, 111,64, 115,75, 122,81, 122,74, 126,79,
+                                    126,74, 131,78, 130,72, 133,77, 131,68, 126,61, 119,57
+   };
+
+   init(l, 77, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly78(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {145,48, 143,53, 147,59, 151,59, 150,55
+   };
+
+   init(l, 78, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly79(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {26,22, 34,15, 43,10, 52,10, 59,16, 47,15, 32,22
+   };
+
+   init(l, 79, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly80(struct lion *l)
+{
+   VGfloat color = 0xffe5b2;
+   static const VGfloat coords[] = {160,19, 152,26, 149,34, 154,33, 152,30, 157,30, 155,26, 158,27,
+                                    157,23, 161,23
+   };
+
+   init(l, 80, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly81(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {98,117, 105,122, 109,122, 105,117, 113,120, 121,120, 130,112, 128,108,
+                                    123,103, 123,99, 128,101, 132,106, 135,109, 142,105, 142,101, 145,101,
+                                    145,91, 148,101, 145,105, 136,112, 135,116, 143,124, 148,120, 150,122,
+                                    142,128, 133,122, 121,125, 112,126, 103,125, 100,129, 96,124
+   };
+
+   init(l, 81, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly82(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {146,118, 152,118, 152,115, 149,115
+   };
+
+   init(l, 82, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly83(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {148,112, 154,111, 154,109, 149,109
+   };
+
+   init(l, 83, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly84(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {106,112, 108,115, 114,116, 118,114
+   };
+
+   init(l, 84, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly85(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {108,108, 111,110, 116,110, 119,108
+   };
+
+   init(l, 85, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly86(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {106,104, 109,105, 117,106, 115,104
+   };
+
+   init(l, 86, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly87(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {50,25, 41,26, 34,33, 39,43, 49,58, 36,51, 47,68, 55,69, 54,59,
+                                    61,57, 74,46, 60,52, 67,42, 57,48, 61,40, 54,45, 60,36, 59,29,
+                                    48,38, 52,30, 47,32
+   };
+
+   init(l, 87, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly88(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {147,34, 152,41, 155,49, 161,53, 157,47, 164,47, 158,43, 168,44,
+                                    159,40, 164,37, 169,37, 164,33, 169,34, 165,28, 170,30, 170,25,
+                                    173,29, 175,27, 176,32, 173,36, 175,39, 172,42, 172,46, 168,49,
+                                    170,55, 162,57, 158,63, 155,58, 153,50, 149,46
+   };
+
+   init(l, 88, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly89(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {155,71, 159,80, 157,93, 157,102, 155,108, 150,101, 149,93,
+                                    154,101, 152,91, 151,83, 155,79
+   };
+
+   init(l, 89, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly90(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {112,78, 115,81, 114,91, 112,87, 113,82
+   };
+
+   init(l, 90, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly91(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {78,28, 64,17, 58,11, 47,9, 36,10, 28,16, 21,26, 18,41,
+                                    20,51, 23,61, 33,65, 28,68, 37,74, 36,81, 43,87, 48,90,
+                                    43,100, 40,98, 39,90, 31,80, 30,72, 22,71, 17,61, 14,46,
+                                    16,28, 23,17, 33,9, 45,6, 54,6, 65,12
+   };
+
+   init(l, 91, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly92(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {67,18, 76,9, 87,5, 101,2, 118,3, 135,8, 149,20, 149,26,
+                                    144,19, 132,12, 121,9, 105,7, 89,8, 76,14, 70,20
+   };
+
+   init(l, 92, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly93(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {56,98, 48,106, 56,103, 47,112, 56,110, 52,115, 57,113, 52,121, 62,115,
+                                    58,123, 65,119, 63,125, 69,121, 68,127, 74,125, 74,129, 79,128, 83,132,
+                                    94,135, 93,129, 85,127, 81,122, 76,126, 75,121, 71,124, 71,117, 66,121,
+                                    66,117, 62,117, 64,112, 60,113, 60,110, 57,111, 61,105, 57,107, 60,101,
+                                    55,102
+   };
+
+   init(l, 93, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly94(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {101,132, 103,138, 106,134, 106,139, 112,136, 111,142, 115,139,
+                                    114,143, 119,142, 125,145, 131,142, 135,138, 140,134, 140,129,
+                                    143,135, 145,149, 150,171, 149,184, 145,165, 141,150, 136,147,
+                                    132,151, 131,149, 126,152, 125,150, 121,152, 117,148, 111,152,
+                                    110,148, 105,149, 104,145, 98,150, 96,138, 94,132, 94,130, 98,132
+   };
+
+   init(l, 94, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly95(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {41,94, 32,110, 23,132, 12,163, 6,190, 7,217, 5,236,
+                                    3,247, 9,230, 12,211, 12,185, 18,160, 26,134, 35,110,
+                                    43,99
+   };
+
+   init(l, 95, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly96(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {32,246, 41,250, 50,257, 52,267, 53,295, 53,323, 59,350,
+                                    54,363, 51,365, 44,366, 42,360, 40,372, 54,372, 59,366,
+                                    62,353, 71,352, 75,335, 73,330, 66,318, 68,302, 64,294,
+                                    67,288, 63,286, 63,279, 59,275, 58,267, 56,262, 50,247,
+                                    42,235, 44,246, 32,236, 35,244
+   };
+
+   init(l, 96, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly97(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {134,324, 146,320, 159,322, 173,327, 179,337, 179,349,
+                                    172,355, 158,357, 170,350, 174,343, 170,333, 163,328, 152,326,
+                                    134,329
+   };
+
+   init(l, 97, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly98(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {173,339, 183,334, 184,338, 191,329, 194,332, 199,323, 202,325,
+                                    206,318, 209,320, 213,309, 221,303, 228,296, 232,289, 234,279,
+                                    233,269, 230,262, 225,256, 219,253, 208,252, 198,252, 210,249,
+                                    223,250, 232,257, 237,265, 238,277, 238,291, 232,305, 221,323,
+                                    218,335, 212,342, 200,349, 178,348
+   };
+
+   init(l, 98, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly99(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {165,296, 158,301, 156,310, 156,323, 162,324, 159,318,
+                                    162,308, 162,304
+   };
+
+   init(l, 99, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly100(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {99,252, 105,244, 107,234, 115,228, 121,228, 131,235,
+                                    122,233, 113,235, 109,246, 121,239, 133,243, 121,243,
+                                    110,251
+   };
+
+   init(l, 100, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly101(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {117,252, 124,247, 134,249, 136,253, 126,252
+   };
+
+   init(l, 101, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly102(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {117,218, 132,224, 144,233, 140,225, 132,219, 117,218,
+                                    117,218, 117,218
+   };
+
+   init(l, 102, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly103(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {122,212, 134,214, 143,221, 141,213, 132,210
+   };
+
+   init(l, 103, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly104(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {69,352, 70,363, 76,373, 86,378, 97,379, 108,379, 120,377,
+                                    128,378, 132,373, 135,361, 133,358, 132,366, 127,375, 121,374,
+                                    121,362, 119,367, 117,374, 110,376, 110,362, 107,357, 106,371,
+                                    104,375, 97,376, 90,375, 90,368, 86,362, 83,364, 86,369, 85,373,
+                                    78,370, 73,362, 71,351
+   };
+
+   init(l, 104, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly105(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {100,360, 96,363, 99,369, 102,364
+   };
+
+   init(l, 105, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly106(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {115,360, 112,363, 114,369, 117,364
+   };
+
+   init(l, 106, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly107(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {127,362, 125,364, 126,369, 128,365
+   };
+
+   init(l, 107, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly108(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {5,255, 7,276, 11,304, 15,320, 13,334, 6,348, 2,353, 0,363,
+                                    5,372, 12,374, 25,372, 38,372, 44,369, 42,367, 36,368, 31,369,
+                                    30,360, 27,368, 20,370, 16,361, 15,368, 10,369, 3,366, 3,359, 6,352,
+                                    11,348, 17,331, 19,316, 12,291, 9,274
+   };
+
+   init(l, 108, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly109(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {10,358, 7,362, 10,366, 11,362
+   };
+
+   init(l, 109, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly110(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {25,357, 22,360, 24,366, 27,360
+   };
+
+   init(l, 110, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly111(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {37,357, 34,361, 36,365, 38,361
+   };
+
+   init(l, 111, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly112(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {49,356, 46,359, 47,364, 50,360
+   };
+
+   init(l, 112, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly113(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {130,101, 132,102, 135,101, 139,102, 143,103,
+                                    142,101, 137,100, 133,100
+   };
+
+   init(l, 113, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly114(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {106,48, 105,52, 108,56, 109,52
+   };
+
+   init(l, 114, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly115(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {139,52, 139,56, 140,60, 142,58, 141,56
+   };
+
+   init(l, 115, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly116(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {25,349, 29,351, 30,355, 33,350, 37,348, 42,351, 45,347,
+                                    49,345, 44,343, 36,345
+   };
+
+   init(l, 116, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly117(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {98,347, 105,351, 107,354, 109,349, 115,349, 120,353, 118,349,
+                                    113,346, 104,346
+   };
+
+   init(l, 117, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly118(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {83,348, 87,352, 87,357, 89,351, 87,348
+   };
+
+   init(l, 118, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly119(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {155,107, 163,107, 170,107, 186,108, 175,109, 155,109
+   };
+
+   init(l, 119, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly120(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {153,114, 162,113, 175,112, 192,114, 173,114, 154,115
+   };
+
+   init(l, 120, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly121(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {152,118, 164,120, 180,123, 197,129, 169,123, 151,120
+   };
+
+   init(l, 121, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly122(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {68,109, 87,106, 107,106, 106,108, 88,108
+   };
+
+   init(l, 122, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly123(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {105,111, 95,112, 79,114, 71,116, 85,115, 102,113
+   };
+
+   init(l, 123, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly124(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {108,101, 98,99, 87,99, 78,99, 93,100, 105,102
+   };
+
+   init(l, 124, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly125(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {85,63, 91,63, 97,60, 104,60, 108,62, 111,69, 112,75,
+                                    110,74, 108,71, 103,73, 106,69, 105,65, 103,64, 103,67,
+                                    102,70, 99,70, 97,66, 94,67, 97,72, 88,67, 84,66
+   };
+
+   init(l, 125, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly126(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {140,74, 141,66, 144,61, 150,61, 156,62, 153,70, 150,73,
+                                    152,65, 150,65, 151,68, 149,71, 146,71, 144,66, 143,70,
+                                    143,74
+   };
+
+   init(l, 126, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly127(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {146,20, 156,11, 163,9, 172,9, 178,14, 182,18, 184,32, 182,42,
+                                    182,52, 177,58, 176,67, 171,76, 165,90, 157,105, 160,92, 164,85,
+                                    168,78, 167,73, 173,66, 172,62, 175,59, 174,55, 177,53, 180,46,
+                                    181,29, 179,21, 173,13, 166,11, 159,13, 153,18, 148,23
+   };
+
+   init(l, 127, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly128(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {150,187, 148,211, 150,233, 153,247, 148,267, 135,283, 125,299,
+                                    136,292, 131,313, 122,328, 122,345, 129,352, 133,359, 133,367,
+                                    137,359, 148,356, 140,350, 131,347, 129,340, 132,332, 140,328,
+                                    137,322, 140,304, 154,265, 157,244, 155,223, 161,220, 175,229,
+                                    186,247, 185,260, 176,275, 178,287, 185,277, 188,261, 196,253,
+                                    189,236, 174,213
+   };
+
+   init(l, 128, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly129(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {147,338, 142,341, 143,345, 141,354, 147,343
+   };
+
+   init(l, 129, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly130(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {157,342, 156,349, 150,356, 157,353, 163,346, 162,342
+   };
+
+   init(l, 130, color, coords, ELEMENTS(coords)/2);
+}
+
+static void poly131(struct lion *l)
+{
+   VGfloat color = 0x000000;
+   static const VGfloat coords[] = {99,265, 96,284, 92,299, 73,339, 73,333, 87,300
+   };
+
+   init(l, 131, color, coords, ELEMENTS(coords)/2);
+}
+
+
+struct lion * lion_create(void)
+{
+   struct lion *l = calloc(1, sizeof(struct lion));
+
+   poly0(l);
+   poly1(l);
+   poly2(l);
+   poly3(l);
+   poly4(l);
+   poly5(l);
+   poly6(l);
+   poly7(l);
+   poly8(l);
+   poly9(l);
+
+   poly10(l);
+   poly11(l);
+   poly12(l);
+   poly13(l);
+   poly14(l);
+   poly15(l);
+   poly16(l);
+   poly17(l);
+   poly18(l);
+   poly19(l);
+
+   poly20(l);
+   poly21(l);
+   poly22(l);
+   poly23(l);
+   poly24(l);
+   poly25(l);
+   poly26(l);
+   poly27(l);
+   poly28(l);
+   poly29(l);
+
+   poly30(l);
+   poly31(l);
+   poly32(l);
+   poly33(l);
+   poly34(l);
+   poly35(l);
+   poly36(l);
+   poly37(l);
+   poly38(l);
+   poly39(l);
+
+   poly40(l);
+   poly41(l);
+   poly42(l);
+   poly43(l);
+   poly44(l);
+   poly45(l);
+   poly46(l);
+   poly47(l);
+   poly48(l);
+   poly49(l);
+
+   poly50(l);
+   poly51(l);
+   poly52(l);
+   poly53(l);
+   poly54(l);
+   poly55(l);
+   poly56(l);
+   poly57(l);
+   poly58(l);
+   poly59(l);
+
+   poly60(l);
+   poly61(l);
+   poly62(l);
+   poly63(l);
+   poly64(l);
+   poly65(l);
+   poly66(l);
+   poly67(l);
+   poly68(l);
+   poly69(l);
+
+   poly70(l);
+   poly71(l);
+   poly72(l);
+   poly73(l);
+   poly74(l);
+   poly75(l);
+   poly76(l);
+   poly77(l);
+   poly78(l);
+   poly79(l);
+
+   poly80(l);
+   poly81(l);
+   poly82(l);
+   poly83(l);
+   poly84(l);
+   poly85(l);
+   poly86(l);
+   poly87(l);
+   poly88(l);
+   poly89(l);
+
+   poly90(l);
+   poly91(l);
+   poly92(l);
+   poly93(l);
+   poly94(l);
+   poly95(l);
+   poly96(l);
+   poly97(l);
+   poly98(l);
+   poly99(l);
+
+   poly100(l);
+   poly101(l);
+   poly102(l);
+   poly103(l);
+   poly104(l);
+   poly105(l);
+   poly106(l);
+   poly107(l);
+   poly108(l);
+   poly109(l);
+
+   poly110(l);
+   poly111(l);
+   poly112(l);
+   poly113(l);
+   poly114(l);
+   poly115(l);
+   poly116(l);
+   poly117(l);
+   poly118(l);
+   poly119(l);
+
+   poly120(l);
+   poly121(l);
+   poly122(l);
+   poly123(l);
+   poly124(l);
+   poly125(l);
+   poly126(l);
+   poly127(l);
+   poly128(l);
+   poly129(l);
+
+   poly130(l);
+   poly131(l);
+
+   return l;
+}
+
+void lion_render(struct lion *l)
+{
+   VGint i;
+
+   for (i = 0; i < LION_SIZE; ++i) {
+      vgSetPaint(l->fills[i], VG_FILL_PATH);
+      vgDrawPath(l->paths[i], VG_FILL_PATH);
+   }
+}
+
+void lion_destroy(struct lion *l)
+{
+   VGint i;
+   for (i = 0; i < LION_SIZE; ++i) {
+      vgDestroyPaint(l->fills[i]);
+      vgDestroyPath(l->paths[i]);
+   }
+   free(l);
+}
diff --git a/progs/openvg/demos/lion-render.h b/progs/openvg/demos/lion-render.h
new file mode 100644
index 0000000000..c4c020b7ed
--- /dev/null
+++ b/progs/openvg/demos/lion-render.h
@@ -0,0 +1,16 @@
+#ifndef LION_RENDER_H
+#define LION_RENDER_H
+
+#include <VG/openvg.h>
+
+#define LION_SIZE 132
+struct lion {
+   VGPath paths[LION_SIZE];
+   VGPaint fills[LION_SIZE];
+};
+
+struct lion *lion_create(void);
+void lion_render(struct lion *l);
+void lion_destroy(struct lion *l);
+
+#endif
diff --git a/progs/openvg/demos/lion.c b/progs/openvg/demos/lion.c
new file mode 100644
index 0000000000..7224fed399
--- /dev/null
+++ b/progs/openvg/demos/lion.c
@@ -0,0 +1,288 @@
+#include <assert.h>
+#include <math.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include <X11/keysym.h>
+#include <VG/openvg.h>
+#include <GLES/egl.h>
+
+#include "lion-render.h"
+
+static VGint width, height;
+struct lion *lion = 0;
+VGfloat angle = 0;
+
+static void
+draw(void)
+{
+   vgClear(0, 0, width, height);
+
+   vgSeti(VG_MATRIX_MODE, VG_MATRIX_PATH_USER_TO_SURFACE);
+   vgLoadIdentity();
+   vgTranslate(width/2, height/2);
+   vgRotate(angle);
+   vgTranslate(-width/2, -height/2);
+
+   lion_render(lion);
+
+   ++angle;
+}
+
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   width  = w;
+   height = h;
+}
+
+
+static void
+init(void)
+{
+   float clear_color[4] = {1.0, 1.0, 1.0, 1.0};
+   vgSetfv(VG_CLEAR_COLOR, 4, clear_color);
+
+   lion = lion_create();
+}
+
+
+/*
+ * Create an RGB, double-buffered X window.
+ * Return the window and context handles.
+ */
+static void
+make_x_window(Display *x_dpy, EGLDisplay egl_dpy,
+              const char *name,
+              int x, int y, int width, int height,
+              Window *winRet,
+              EGLContext *ctxRet,
+              EGLSurface *surfRet)
+{
+   static const EGLint attribs[] = {
+      EGL_RED_SIZE, 1,
+      EGL_GREEN_SIZE, 1,
+      EGL_BLUE_SIZE, 1,
+      EGL_NONE
+   };
+
+   int scrnum;
+   XSetWindowAttributes attr;
+   unsigned long mask;
+   Window root;
+   Window win;
+   XVisualInfo *visInfo, visTemplate;
+   int num_visuals;
+   EGLContext ctx;
+   EGLConfig config;
+   EGLint num_configs;
+   EGLint vid;
+
+   scrnum = DefaultScreen( x_dpy );
+   root = RootWindow( x_dpy, scrnum );
+
+   if (!eglChooseConfig( egl_dpy, attribs, &config, 1, &num_configs)) {
+      printf("Error: couldn't get an EGL visual config\n");
+      exit(1);
+   }
+
+   assert(config);
+   assert(num_configs > 0);
+
+   if (!eglGetConfigAttrib(egl_dpy, config, EGL_NATIVE_VISUAL_ID, &vid)) {
+      printf("Error: eglGetConfigAttrib() failed\n");
+      exit(1);
+   }
+
+   /* The X window visual must match the EGL config */
+   visTemplate.visualid = vid;
+   visInfo = XGetVisualInfo(x_dpy, VisualIDMask, &visTemplate, &num_visuals);
+   if (!visInfo) {
+      printf("Error: couldn't get X visual\n");
+      exit(1);
+   }
+
+   /* window attributes */
+   attr.background_pixel = 0;
+   attr.border_pixel = 0;
+   attr.colormap = XCreateColormap( x_dpy, root, visInfo->visual, AllocNone);
+   attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask;
+   mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask;
+
+   win = XCreateWindow( x_dpy, root, 0, 0, width, height,
+		        0, visInfo->depth, InputOutput,
+		        visInfo->visual, mask, &attr );
+
+   /* set hints and properties */
+   {
+      XSizeHints sizehints;
+      sizehints.x = x;
+      sizehints.y = y;
+      sizehints.width  = width;
+      sizehints.height = height;
+      sizehints.flags = USSize | USPosition;
+      XSetNormalHints(x_dpy, win, &sizehints);
+      XSetStandardProperties(x_dpy, win, name, name,
+                              None, (char **)NULL, 0, &sizehints);
+   }
+
+   eglBindAPI(EGL_OPENVG_API);
+
+   ctx = eglCreateContext(egl_dpy, config, EGL_NO_CONTEXT, NULL );
+   if (!ctx) {
+      printf("Error: eglCreateContext failed\n");
+      exit(1);
+   }
+
+   *surfRet = eglCreateWindowSurface(egl_dpy, config, win, NULL);
+
+   if (!*surfRet) {
+      printf("Error: eglCreateWindowSurface failed\n");
+      exit(1);
+   }
+
+   XFree(visInfo);
+
+   *winRet = win;
+   *ctxRet = ctx;
+}
+
+
+static void
+event_loop(Display *dpy, Window win,
+           EGLDisplay egl_dpy, EGLSurface egl_surf)
+{
+   while (1) {
+      XEvent event;
+
+      while (XPending(dpy) > 0) {
+         XNextEvent(dpy, &event);
+
+         switch (event.type) {
+         case Expose:
+            break;
+         case ConfigureNotify:
+            reshape(event.xconfigure.width, event.xconfigure.height);
+            break;
+         case KeyPress:
+         {
+            char buffer[10];
+            int r, code;
+            code = XLookupKeysym(&event.xkey, 0);
+            r = XLookupString(&event.xkey, buffer, sizeof(buffer),
+                              NULL, NULL);
+            if (buffer[0] == 27) {
+               /* escape */
+               return;
+            }
+         }
+         break;
+         default:
+            ; /*no-op*/
+         }
+      }
+
+      draw();
+      eglSwapBuffers(egl_dpy, egl_surf);
+   }
+}
+
+
+static void
+usage(void)
+{
+   printf("Usage:\n");
+   printf("  -display <displayname>  set the display to run on\n");
+   printf("  -info                   display OpenGL renderer info\n");
+}
+
+int
+main(int argc, char *argv[])
+{
+   const int winWidth = 350, winHeight = 450;
+   Display *x_dpy;
+   Window win;
+   EGLSurface egl_surf;
+   EGLContext egl_ctx;
+   EGLDisplay egl_dpy;
+   char *dpyName = NULL;
+   GLboolean printInfo = GL_FALSE;
+   EGLint egl_major, egl_minor;
+   int i;
+   const char *s;
+
+   for (i = 1; i < argc; i++) {
+      if (strcmp(argv[i], "-display") == 0) {
+         dpyName = argv[i+1];
+         i++;
+      }
+      else if (strcmp(argv[i], "-info") == 0) {
+         printInfo = GL_TRUE;
+      }
+      else {
+         usage();
+         return -1;
+      }
+   }
+
+   x_dpy = XOpenDisplay(dpyName);
+   if (!x_dpy) {
+      printf("Error: couldn't open display %s\n",
+	     dpyName ? dpyName : getenv("DISPLAY"));
+      return -1;
+   }
+
+   egl_dpy = eglGetDisplay(x_dpy);
+   if (!egl_dpy) {
+      printf("Error: eglGetDisplay() failed\n");
+      return -1;
+   }
+
+   if (!eglInitialize(egl_dpy, &egl_major, &egl_minor)) {
+      printf("Error: eglInitialize() failed\n");
+      return -1;
+   }
+
+   s = eglQueryString(egl_dpy, EGL_VERSION);
+   printf("EGL_VERSION = %s\n", s);
+
+   make_x_window(x_dpy, egl_dpy,
+                 "Lion Example", 0, 0, winWidth, winHeight,
+                 &win, &egl_ctx, &egl_surf);
+
+   XMapWindow(x_dpy, win);
+   if (!eglMakeCurrent(egl_dpy, egl_surf, egl_surf, egl_ctx)) {
+      printf("Error: eglMakeCurrent() failed\n");
+      return -1;
+   }
+
+   if (printInfo) {
+      printf("VG_RENDERER   = %s\n", (char *) vgGetString(VG_RENDERER));
+      printf("VG_VERSION    = %s\n", (char *) vgGetString(VG_VERSION));
+      printf("VG_VENDOR     = %s\n", (char *) vgGetString(VG_VENDOR));
+   }
+
+   init();
+
+   /* Set initial projection/viewing transformation.
+    * We can't be sure we'll get a ConfigureNotify event when the window
+    * first appears.
+    */
+   reshape(winWidth, winHeight);
+
+   event_loop(x_dpy, win, egl_dpy, egl_surf);
+
+   eglDestroyContext(egl_dpy, egl_ctx);
+   eglDestroySurface(egl_dpy, egl_surf);
+   eglTerminate(egl_dpy);
+
+
+   XDestroyWindow(x_dpy, win);
+   XCloseDisplay(x_dpy);
+
+   return 0;
+}
diff --git a/progs/openvg/demos/sp.c b/progs/openvg/demos/sp.c
new file mode 100644
index 0000000000..424ec47d69
--- /dev/null
+++ b/progs/openvg/demos/sp.c
@@ -0,0 +1,537 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <VG/vgu.h>
+#include <stdio.h>
+#include <math.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <X11/keysym.h>
+
+#define ELEMENTS(x) (sizeof(x)/sizeof((x)[0]))
+
+struct object {
+   VGPath path;
+   VGPaint fill;
+   VGPaint stroke;
+   VGint draw_mode;
+   VGfloat matrix[9];
+   VGfloat stroke_width;
+};
+
+struct character {
+   struct object objects[32];
+   VGint num_objects;
+};
+VGfloat identity_matrix[] = {1, 0, 0, 0, 1, 0, 0, 0, 1};
+
+struct character cartman;
+
+static void add_object_fill(const VGubyte *segments, VGint num_segments,
+                            const VGfloat *coords,
+                            VGuint color)
+{
+   struct object object;
+
+   object.path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                              1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+   vgAppendPathData(object.path, num_segments, segments, coords);
+
+   object.fill = vgCreatePaint();
+   vgSetColor(object.fill, color);
+   memcpy(object.matrix, identity_matrix, 9 * sizeof(VGfloat));
+   object.draw_mode = VG_FILL_PATH;
+
+   cartman.objects[cartman.num_objects] = object;
+   ++cartman.num_objects;
+}
+
+
+static void add_object_stroke(const VGubyte *segments, VGint num_segments,
+                              const VGfloat *coords,
+                              VGuint color, VGfloat width)
+{
+   struct object object;
+
+   object.path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                              1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+   vgAppendPathData(object.path, num_segments, segments, coords);
+
+   object.stroke = vgCreatePaint();
+   vgSetColor(object.stroke, color);
+   memcpy(object.matrix, identity_matrix, 9 * sizeof(VGfloat));
+   object.draw_mode = VG_STROKE_PATH;
+   object.stroke_width = width;
+
+   cartman.objects[cartman.num_objects] = object;
+   ++cartman.num_objects;
+}
+
+
+static void add_object_fillm(const VGubyte *segments, VGint num_segments,
+                             const VGfloat *coords,
+                             VGuint color,
+                             VGfloat *matrix)
+{
+   struct object object;
+
+   object.path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                              1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+   vgAppendPathData(object.path, num_segments, segments, coords);
+
+   object.fill = vgCreatePaint();
+   vgSetColor(object.fill, color);
+   memcpy(object.matrix, matrix, 9 * sizeof(VGfloat));
+   object.draw_mode = VG_FILL_PATH;
+
+   cartman.objects[cartman.num_objects] = object;
+   ++cartman.num_objects;
+}
+
+
+static void add_object_m(const VGubyte *segments, VGint num_segments,
+                         const VGfloat *coords,
+                         VGuint fill_color,
+                         VGuint stroke_color, VGfloat stroke_width,
+                         VGfloat *matrix)
+{
+   struct object object;
+
+   object.path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                              1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+   vgAppendPathData(object.path, num_segments, segments, coords);
+   memcpy(object.matrix, matrix, 9 * sizeof(VGfloat));
+
+   object.fill = vgCreatePaint();
+   vgSetColor(object.fill, fill_color);
+   object.draw_mode = VG_FILL_PATH | VG_STROKE_PATH;
+
+   object.stroke = vgCreatePaint();
+   vgSetColor(object.stroke, stroke_color);
+   object.stroke_width = stroke_width;
+
+   cartman.objects[cartman.num_objects] = object;
+   ++cartman.num_objects;
+}
+
+static void init_character()
+{
+   {
+      const VGubyte segments[] = {VG_MOVE_TO_ABS,
+                                  VG_CUBIC_TO_ABS,
+                                  VG_CUBIC_TO_ABS,
+                                  VG_CUBIC_TO_ABS,
+                                  VG_CUBIC_TO_ABS,
+                                  VG_CLOSE_PATH};
+      const VGfloat coords[] = {181.83267, 102.60408,
+                                181.83267,102.60408, 185.53793,114.5749, 186.5355,115.00243,
+                                187.53306,115.42996, 286.0073,115.00243, 286.0073,115.00243,
+                                286.0073,115.00243, 292.70526,103.45914, 290.85263,101.03648,
+                                289.00001,98.61381, 181.54765,102.31906, 181.83267,102.60408
+      };
+      VGuint color = 0x7c4e32ff;
+      add_object_fill(segments, ELEMENTS(segments),
+                      coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_LINE_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {188.62208,50.604156,
+                                188.62208,50.604156, 176.73127,60.479579, 170.68509,69.548844,
+                                164.63892,78.618109, 175.11895,79.827344, 175.11895,79.827344,
+                                176.52973,98.368952,
+                                176.52973,98.368952, 189.83131,110.05823, 208.97754,110.25976,
+                                228.12377,110.46131, 244.24691,111.67054, 247.06846,110.25976,
+                                249.89,108.849, 258.95927,106.8336, 260.16851,105.01975,
+                                261.37774,103.2059, 296.84865,106.43053, 297.05019,91.919698,
+                                297.25172,77.408874, 306.11945,64.308824, 282.13628,51.611853,
+                                258.15311,38.914882, 189.2267,49.999539, 188.62208,50.604156
+      };
+
+      VGuint color = 0xe30000ff;
+      add_object_fill(segments, ELEMENTS(segments),
+                      coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,
+         VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         68.25, 78.875,
+         68.25,93.296, 54.642,105, 37.875,105,
+         21.108,105, 7.5,93.296, 7.5,78.875,
+         7.5,64.454, 21.108,52.75, 37.875,52.75,
+         54.642,52.75, 68.25,64.454, 68.25,78.875
+      };
+
+      VGuint color = 0xffe1c4ff;
+      VGfloat matrix[] = {
+         1.6529, 0, 0,
+         0, 1.582037, 0,
+         172.9649,-90.0116, 1
+      };
+      add_object_fillm(segments, ELEMENTS(segments),
+                       coords, color, matrix);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         170.14687,71.536958,
+         173.53626,68.814326, 176.70232,68.971782, 180.55009,71.679467,
+         184.39785,74.387153, 199.19294,80.036105, 191.52334,86.500482,
+         189.02942,88.6025, 183.97032,85.787933, 180.26507,86.928011,
+         178.8737,87.356121, 174.71827,89.783259, 171.8028,87.494856,
+         166.95426,83.689139, 163.51779,76.861986, 170.14687,71.536958
+      };
+
+      VGuint color = 0xfff200ff;
+      add_object_fill(segments, ELEMENTS(segments),
+                      coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,  VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         299.83075,66.834136,
+         299.83075,66.834136, 287.85993,64.69649, 284.15467,72.962055,
+         280.44942,81.227621, 280.1644,78.234916, 280.1644,79.374994,
+         280.1644,80.515072, 278.16927,84.077816, 284.86722,83.792796,
+         291.56518,83.507777, 291.99271,86.785501, 294.84291,86.642991,
+         297.6931,86.500482, 303.536,85.645423, 303.67851,80.657582,
+         303.82102,75.66974, 302.68094,65.551548, 299.83075,66.834136
+      };
+
+      VGuint color = 0xfff200ff;
+      add_object_fill(segments, ELEMENTS(segments),
+                      coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS
+      };
+      const VGfloat coords[] = {
+         240.83171,75.81225,
+         240.83171,75.81225, 241.54426,88.495618, 242.25681,91.488323,
+         242.96936,94.481028, 240.6892,108.01945, 240.83171,110.01459,
+         240.97422,112.00973, 240.97422,111.01216, 240.97422,111.01216
+      };
+      VGuint color = 0x000000ff;
+      VGfloat swidth = 1.14007807;
+      add_object_stroke(segments, ELEMENTS(segments), coords, color, swidth);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,  VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_LINE_TO_ABS,  VG_LINE_TO_ABS,  VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         83.375, 95.5,
+         83.375,96.121, 83.067,96.625, 82.6875,96.625,
+         82.308,96.625, 82,96.121, 82,95.5,
+         82,94.879, 82.308,94.375, 82.6875,94.375,
+         83.066677,94.375, 83.374492,94.878024, 83.374999,95.498494,
+         82.6875,95.5,
+         83.375,95.5
+      };
+      VGuint fill_color = 0x000000ff;
+      VGuint stroke_color = 0x000000ff;
+      VGfloat swidth = 0.60000002;
+      VGfloat matrix1[] = {
+         1.140078, 0, 0,
+         0, 1.140078, 0,
+         145.4927, -15.10897, 1
+      };
+      VGfloat matrix2[] = {
+         1.140078,0, 0,
+         0,1.140078, 0,
+         144.2814,-27.93485, 1
+      };
+      VGfloat matrix3[] = {
+         1.140078,0, 0,
+         0,1.140078, 0,
+         144.1388,-3.70819, 1
+      };
+      add_object_m(segments, ELEMENTS(segments), coords,
+                   fill_color, stroke_color, swidth, matrix1);
+      add_object_m(segments, ELEMENTS(segments), coords,
+                   fill_color, stroke_color, swidth, matrix2);
+      add_object_m(segments, ELEMENTS(segments), coords,
+                   fill_color, stroke_color, swidth, matrix3);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_LINE_TO_ABS, VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         179.41001,115.28745,
+         179.41001,115.28745, 207.48443,109.30204, 236.84144,115.14494,
+         236.84144,115.14494, 274.74903,109.87208, 291.8502,115.42996,
+         179.41001,115.28745
+      };
+
+      VGuint color = 0x000000ff;
+      add_object_fill(segments, ELEMENTS(segments),
+                      coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,  VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS,  VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         83.792156,68.157364,
+         83.792156,69.669865, 82.72301,70.897403, 81.40567,70.897403,
+         80.08833,70.897403, 79.019185,69.669865, 79.019185,68.157364,
+         79.019185,66.644862, 80.08833,65.417325, 81.40567,65.417325,
+         82.721887,65.417325, 83.790391,66.642485, 83.792153,68.153696,
+         81.40567,68.157364,
+         83.792156,68.157364
+      };
+      VGuint fill_color = 0x000000ff;
+      VGuint stroke_color = 0x000000ff;
+      VGfloat swidth = 0.52891117;
+      VGfloat matrix1[] = {
+         1.140078,0, 0,
+         0,1.140078, 0,
+         145.2489,-15.58714, 1
+      };
+      add_object_m(segments, ELEMENTS(segments), coords,
+                   fill_color, stroke_color, swidth, matrix1);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS
+      };
+      const VGfloat coords[] = {
+         232.28113,66.976646,
+         232.28113,66.976646, 237.98152,70.539389, 245.39202,66.549116
+      };
+      VGuint color = 0x000000ff;
+      VGfloat swidth = 0.60299999;
+      add_object_stroke(segments, ELEMENTS(segments), coords, color, swidth);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         185.96908,30.061986,
+         185.96908,30.061986, 187.76995,14.508377, 203.23909,3.7427917,
+         209.95028,-0.92779696, 219.37764,-4.9841866, 232.1078,-6.00046,
+         246.13578,-7.1203411, 256.92106,-2.8560739, 264.81774,1.9451947,
+         280.60485,11.543934, 284.31582,25.937274, 284.08015,26.526452,
+         283.7266,27.410336, 240.83461,1.9346323, 185.96908,30.061986
+      };
+      VGuint color = 0x8ed8f8ff;
+      add_object_fill(segments, ELEMENTS(segments), coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_LINE_TO_ABS, VG_CUBIC_TO_ABS, VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         185.39542,32.061757,
+         185.82295,29.211562,
+         185.82295,29.211562, 234.70379,2.277219, 284.01217,25.078779,
+         284.86722,27.643954,
+         284.86722,27.643954, 236.69893,4.5573746, 185.39542,32.061757
+      };
+      VGuint color = 0xfff200ff;
+      add_object_fill(segments, ELEMENTS(segments), coords, color);
+   }
+
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS,  VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         219.74027,-5.917093,
+         220.49206,-8.44929, 225.15564,-10.904934, 230.21473,-11.189954,
+         235.27383,-11.474973, 243.27521,-13.287236, 249.21385,-5.724198,
+         249.89961,-4.850868, 249.28247,-4.332166, 248.62298,-3.971398,
+         247.79117,-3.516361, 247.13703,-3.392737, 246.16222,-3.408047,
+         243.63973,-3.447664, 242.54183,-3.850701, 242.54183,-3.850701,
+         242.54183,-3.850701, 238.78367,-1.737343, 236.20014,-3.565682,
+         233.88436,-5.204544, 234.27626,-4.56325, 234.27626,-4.56325,
+         234.27626,-4.56325, 232.33303,-2.975658, 230.85603,-2.995643,
+         228.59433,-3.025282, 227.73672,-4.501857, 227.21966,-4.93027,
+         226.76318,-4.932008, 226.50948,-4.491995, 226.50948,-4.491995,
+         226.50948,-4.491995, 224.53199,-2.085883, 222.51431,-2.467064,
+         221.48814,-2.66093, 218.91968,-3.15318, 219.74027,-5.917093
+      };
+      VGuint color = 0xfff200ff;
+      add_object_fill(segments, ELEMENTS(segments), coords, color);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS,  VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,  VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         178.97347,166.06432,
+         178.97347,181.2154, 168.0245,193.51193, 154.53381,193.51193,
+         141.04312,193.51193, 130.09416,181.2154, 130.09416,166.06432,
+         130.09416,150.91323, 141.04312,138.6167, 154.53381,138.6167,
+         168.0245,138.6167, 178.97347,150.91323, 178.97347,166.06432
+      };
+      VGuint color = 0xffffffff;
+      VGfloat matrix1[] = {
+         0.466614,-0.23492,  0,
+         0.108683,0.436638,  0,
+         134.5504,-0.901632, 1
+      };
+      VGfloat matrix2[] = {
+         -0.466614,-0.23492, 0,
+         -0.108683,0.436638, 0,
+         338.4496,-0.512182, 1
+      };
+      add_object_fillm(segments, ELEMENTS(segments), coords, color, matrix1);
+      add_object_fillm(segments, ELEMENTS(segments), coords, color, matrix2);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,
+         VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS,  VG_CLOSE_PATH
+      };
+      const VGfloat coords[] = {
+         123.82758,165.06168,
+         123.82758,166.79125, 122.59232,168.19497, 121.07029,168.19497,
+         119.54826,168.19497, 118.313,166.79125, 118.313,165.06168,
+         118.313,163.3321, 119.54826,161.92839, 121.07029,161.92839,
+         122.59232,161.92839, 123.82758,163.3321, 123.82758,165.06168
+      };
+      VGuint color = 0x000000ff;
+      VGfloat matrix1[] = {
+         0.525719,0, 0,
+         0,0.479931, 0,
+         178.9702,-43.3532, 1
+      };
+      VGfloat matrix2[] = {
+         0.525719,0, 0,
+         0,0.479931, 0,
+         165.258,-43.46162, 1
+      };
+      add_object_fillm(segments, ELEMENTS(segments), coords, color, matrix1);
+      add_object_fillm(segments, ELEMENTS(segments), coords, color, matrix2);
+   }
+   {
+      const VGubyte segments[] = {
+         VG_MOVE_TO_ABS, VG_CUBIC_TO_ABS, VG_CUBIC_TO_ABS
+      };
+      const VGfloat coords[] = {
+         197.25,54.5,
+         197.25,54.5, 211.75,71.5, 229.25,71.5,
+         246.75,71.5, 261.74147,71.132714, 277.75,50.75
+      };
+      VGuint color = 0x000000ff;
+      VGfloat swidth = 0.60299999;
+      add_object_stroke(segments, ELEMENTS(segments), coords, color, swidth);
+   }
+}
+
+
+static void
+init(void)
+{
+   float clear_color[4] = {1.0, 1.0, 1.0, 1.0};
+   vgSetfv(VG_CLEAR_COLOR, 4, clear_color);
+
+   init_character();
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static int
+key_press(unsigned key)
+{
+    switch(key) {
+    case XK_Right:
+
+        break;
+    case XK_Left:
+        break;
+    case XK_Up:
+        break;
+    case XK_Down:
+        break;
+    case 'a':
+        break;
+    case 's':
+        break;
+    default:
+        break;
+    }
+    return VG_FALSE;
+}
+
+static void
+draw(void)
+{
+   VGint i;
+   VGfloat save_matrix[9];
+
+   vgClear(0, 0, window_width(), window_height());
+
+   vgSeti(VG_MATRIX_MODE, VG_MATRIX_PATH_USER_TO_SURFACE);
+   vgLoadIdentity();
+   vgScale(2, 2);
+   vgTranslate(160, 60);
+   vgRotate(180);
+   vgTranslate(-160, -100);
+   vgGetMatrix(save_matrix);
+   for (i = 0; i < cartman.num_objects; ++i) {
+      struct object object = cartman.objects[i];
+      if ((object.draw_mode & VG_STROKE_PATH)) {
+         vgSetf(VG_STROKE_LINE_WIDTH, object.stroke_width);
+         vgSetPaint(object.stroke, VG_STROKE_PATH);
+      }
+      if ((object.draw_mode & VG_FILL_PATH))
+         vgSetPaint(object.fill, VG_FILL_PATH);
+      vgMultMatrix(object.matrix);
+      vgDrawPath(object.path, object.draw_mode);
+      vgLoadMatrix(save_matrix);
+   }
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+    set_window_size(400, 400);
+    return run(argc, argv, init, reshape, draw, key_press);
+}
diff --git a/progs/openvg/trivial/Makefile b/progs/openvg/trivial/Makefile
new file mode 100644
index 0000000000..362360e596
--- /dev/null
+++ b/progs/openvg/trivial/Makefile
@@ -0,0 +1,127 @@
+# These programs aren't intended to be included with the normal distro.
+# They're not too interesting but they're good for testing.
+
+TOP = ../../../
+include $(TOP)/configs/current
+
+INCLUDES = -I. -I$(TOP)/include
+LIBS=-L$(TOP)/$(LIB_DIR) -lm -lEGL -lOpenVG -lpthread
+CFLAGS += $(INCLUDES)
+
+HEADERS=eglcommon.h
+
+PROGRAMS = \
+        arc \
+        cap \
+        clear \
+        coord \
+        dash \
+        ellipse \
+        filter \
+        gradorigin \
+        lineto \
+        lingrad \
+        lookup \
+        mask4 \
+        mask \
+        path3 \
+        radialgrad \
+        readpixels \
+        roundedrect \
+        star-nonzero \
+        star-oddeven \
+        stroke2 \
+        stroke \
+        vguarc
+
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+
+default: $(PROGRAMS)
+
+
+arc: arc.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+cap: cap.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+clear: clear.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+coord: coord.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+dash: dash.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+ellipse: ellipse.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+filter: filter.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+gradorigin: gradorigin.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+image: image.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+lineto: lineto.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+lingrad: lingrad.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+lookup: lookup.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+mask: mask.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+mask4: mask4.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+path3: path3.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+pattern: pattern.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+radialgrad: radialgrad.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+readpixels: readpixels.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+roundedrect: roundedrect.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+star-nonzero: star-nonzero.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+star-oddeven: star-oddeven.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+stroke: stroke.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+stroke2: stroke2.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+vguarc: vguarc.c eglcommon.o
+	$(CC) $(CFLAGS) $^ $(LIBS)  $(APP_LIB_DEPS) -o $@
+
+
+
+eglcommon.o: eglcommon.c $(HEADERS)
+	$(CC) -c $(CFLAGS) eglcommon.c
+
+
+clean:
+	rm -f *.o *~
+	rm -f *.so
+	rm -f $(PROGRAMS)
diff --git a/progs/openvg/trivial/arc.c b/progs/openvg/trivial/arc.c
new file mode 100644
index 0000000000..db686bea6b
--- /dev/null
+++ b/progs/openvg/trivial/arc.c
@@ -0,0 +1,139 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <math.h>
+
+const VGfloat clear_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {1.0, 1.0, 1.0, 0.5};
+
+VGPath vgPath;
+
+static void ellipse(VGPath vgPath, VGfloat rx, VGfloat ry, VGfloat angle)
+{
+    static const VGubyte cmd[] =
+    { VG_MOVE_TO_ABS, VG_SCCWARC_TO_REL, VG_SCCWARC_TO_REL, VG_CLOSE_PATH };
+
+    VGfloat val[12];
+    VGfloat c = cos(angle) * rx;
+    VGfloat s = sin(angle) * rx;
+
+    val[0] = c;
+    val[1] = s;
+    val[2] = rx;
+    val[3] = ry;
+    val[4] = angle;
+    val[5] = -2.0f * c;
+    val[6] = -2.0f * s;
+    val[7] = rx;
+    val[8] = ry;
+    val[9] = angle;
+    val[10] = 2.0f * c;
+    val[11] = 2.0f * s;
+
+    vgClearPath(vgPath, VG_PATH_CAPABILITY_ALL);
+    vgAppendPathData(vgPath, sizeof(cmd), cmd, val);
+    vgDrawPath(vgPath, VG_FILL_PATH | VG_STROKE_PATH);
+}
+
+static void
+init(void)
+{
+    VGPaint vgPaint;
+
+    vgSetfv(VG_CLEAR_COLOR, 4, clear_color);
+    vgPath = vgCreatePath(VG_PATH_FORMAT_STANDARD,
+                          VG_PATH_DATATYPE_F, 1.0f, 0.0f, 0, 0,
+                          VG_PATH_CAPABILITY_ALL);
+
+    vgPaint = vgCreatePaint();
+    vgSetParameteri(vgPaint, VG_PAINT_TYPE, VG_PAINT_TYPE_COLOR);
+    vgSetColor(vgPaint, 0x00ff00ff);
+    vgSetPaint(vgPaint, VG_FILL_PATH);
+
+    vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+    vgSeti(VG_BLEND_MODE, VG_BLEND_SRC_OVER);
+    vgSetf(VG_STROKE_LINE_WIDTH, 2.0f);
+    vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_SQUARE);
+    vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_MITER);
+    vgSetf(VG_STROKE_MITER_LIMIT, 4.0f);
+    vgSeti(VG_MATRIX_MODE, VG_MATRIX_PATH_USER_TO_SURFACE);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static void
+draw(void)
+{
+    vgClear(0, 0, window_width(), window_height());
+
+#if 0
+    vgLoadIdentity();
+    vgTranslate(40.0f, 24.0f);
+    vgScale(0.61804f, 0.61804f);
+    vgShear(-1.0f, 0.0f);
+    vgDrawPath(vgPath, VG_FILL_PATH | VG_STROKE_PATH);
+#else
+
+    /* row 1, col 1: Identity transform. */
+
+    vgLoadIdentity();
+    vgTranslate(8.0f, 8.0f);
+    ellipse(vgPath, 4.0f, 4.0f, 0.0f);
+
+    /* row 1, col 2: 10^3 horizontal squeeze. */
+
+    vgLoadIdentity();
+    vgTranslate(24.0f, 8.0f);
+    vgScale(1.0e-3f, 1.0f);
+    ellipse(vgPath, 4.0e3f, 4.0f, 0.0f);
+
+    /* row 1, col 3: 10^6 horizontal squeeze. */
+
+    vgLoadIdentity();
+    vgTranslate(40.0f, 8.0f);
+    vgScale(1.0e-6f, 1.0f);
+    ellipse(vgPath, 4.0e6f, 4.0f, 0.0f);
+
+    /* row 1, col 4: 10^9 horizontal squeeze. */
+
+    vgLoadIdentity();
+    vgTranslate(56.0f, 8.0f);
+    vgScale(1.0e-9f, 1.0f);
+    ellipse(vgPath, 4.0e9f, 4.0f, 0.0f);
+
+    /* row 2, col 1: 10^3 vertical squeeze. */
+
+    vgLoadIdentity();
+    vgTranslate(8.0f, 24.0f);
+    vgScale(1.0f, 1.0e-3f);
+    ellipse(vgPath, 4.0f, 4.0e3f, 0.0f);
+
+    /* row 2, col 2: Shear 0. */
+
+    vgLoadIdentity();
+    vgTranslate(24.0f, 24.0f);
+    vgShear(0.0f, 0.0f);
+    ellipse(vgPath, 4.0f, 4.0f, 0.0f);
+
+    /* row 2, col 3: Horizontal shear -1. */
+
+    vgLoadIdentity();
+    vgTranslate(40.0f, 24.0f);
+    vgScale(0.61804f, 0.61804f);
+    vgShear(-1.0f, 0.0f);
+    ellipse(vgPath, 10.47213f, 4.0f, 31.717f);
+#endif
+    vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+    set_window_size(64, 64);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/cap.c b/progs/openvg/trivial/cap.c
new file mode 100644
index 0000000000..cd84fe3ac0
--- /dev/null
+++ b/progs/openvg/trivial/cap.c
@@ -0,0 +1,75 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <math.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+static void
+init(void)
+{
+
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+const int subtest = 0;
+static void
+draw(void)
+{
+    VGPath line;
+    VGPaint fillPaint;
+    VGubyte lineCommands[3] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS};
+    VGfloat lineCoords[] =   {-2.0f,-1.0f, 0.0f,0.0f, -1.0f, -2.0f};
+    VGfloat clearColor[] = {0.0f, 0.0f, 0.0f, 1.0f};/* black color */
+    VGfloat fillColor[] = {1.0f, 1.0f, 1.0f, 1.0f};/* white color */
+    //VGfloat testRadius = 60.0f;
+    VGfloat testRadius = 10.0f;
+    int WINDSIZEX = window_width();
+    int WINDSIZEY = window_height();
+
+    line = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                        1.0f, 0.0f, 0, 0, VG_PATH_CAPABILITY_ALL);
+    fillPaint = vgCreatePaint();
+
+    vgSetf(VG_STROKE_LINE_WIDTH, 1.0f);
+    //vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_ROUND);
+    vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_BUTT);
+    vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_ROUND);
+    //vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_BEVEL);
+
+    vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_BETTER);
+
+    vgSeti(VG_MATRIX_MODE, VG_MATRIX_PATH_USER_TO_SURFACE);
+    vgLoadIdentity();
+    vgTranslate(60, 60);
+    vgScale(testRadius * 2, testRadius * 2);
+
+    vgAppendPathData(line, 3, lineCommands, lineCoords);
+
+    vgSetfv(VG_CLEAR_COLOR, 4, clearColor);
+
+    vgSetPaint(fillPaint, VG_STROKE_PATH);
+
+    vgSetParameterfv(fillPaint, VG_PAINT_COLOR, 4, fillColor);
+    vgSetParameteri( fillPaint, VG_PAINT_TYPE, VG_PAINT_TYPE_COLOR);
+
+    vgClear(0, 0, WINDSIZEX, WINDSIZEY);
+    vgDrawPath(line, VG_STROKE_PATH);
+
+    vgDestroyPath(line);
+    vgDestroyPaint(fillPaint);
+}
+
+
+int main(int argc, char **argv)
+{
+   set_window_size(100, 100);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/clear.c b/progs/openvg/trivial/clear.c
new file mode 100644
index 0000000000..efb6bf4182
--- /dev/null
+++ b/progs/openvg/trivial/clear.c
@@ -0,0 +1,42 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <stdio.h>
+
+float red_color[4] = {1.0, 0.0, 0.0, 1.0};
+float blue_color[4] = {0.0, 0.0, 1.0, 1.0};
+
+static void
+init(void)
+{
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+    VGint scissor[4] = {100, 100, 25, 25};
+    vgSetfv(VG_CLEAR_COLOR, 4, red_color);
+    vgClear(0, 0, window_width(), window_height());
+
+    vgSetfv(VG_CLEAR_COLOR, 4, blue_color);
+    vgClear(50, 50, 50, 50);
+
+    //vgSetiv(VG_SCISSOR_RECTS, 4, scissor);
+    //vgSeti(VG_SCISSORING, VG_TRUE);
+    vgCopyPixels(100, 100, 50, 50, 50, 50);
+    vgClear(150, 150, 50, 50);
+}
+
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/coord.c b/progs/openvg/trivial/coord.c
new file mode 100644
index 0000000000..81f7cb6fc9
--- /dev/null
+++ b/progs/openvg/trivial/coord.c
@@ -0,0 +1,66 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.4, 0.1, 1.0, 1.0};
+
+VGPath path;
+VGPaint fill;
+
+
+static void
+init(void)
+{
+   /* Absent VG_CLOSE_PATH */
+   VGubyte commands[] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS,
+                         VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS};
+   VGfloat clearColor[] = {1.0f, 1.0f, 1.0f, 1.0f};/* white color */
+   VGfloat fillColor[] = {1.0f, 0.0f, 0.0f, 1.0f};/* red color */
+   VGfloat coords[] = {-16.0f, -16.0f, 0.0f, -16.0f, 0.0f, 0.0f, -16.0f, 0.0f,
+                       0.0f, 0.0f, 16.0f, 0.0f, 16.0f, 16.0f, 0.0f, 16.0f};
+
+   vgSetfv(VG_CLEAR_COLOR, 4, clearColor);
+   vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+
+   vgSeti(VG_MATRIX_MODE, VG_MATRIX_PATH_USER_TO_SURFACE);
+   vgLoadIdentity();
+   vgTranslate(32.0f, 32.0f);
+
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1.0f, 0.0f, 0, 0,
+                       VG_PATH_CAPABILITY_ALL);
+   if (path == VG_INVALID_HANDLE)
+      return;
+   fill = vgCreatePaint();
+   if (fill == VG_INVALID_HANDLE) {
+      vgDestroyPath(path);
+      return;
+   }
+   vgAppendPathData(path, 8, commands, coords);
+   vgSetPaint(fill, VG_FILL_PATH);
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, fillColor);
+   vgSetParameteri(fill, VG_PAINT_TYPE, VG_PAINT_TYPE_COLOR);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_FILL_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   set_window_size(64, 64);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/dash.c b/progs/openvg/trivial/dash.c
new file mode 100644
index 0000000000..2e84ddbd4e
--- /dev/null
+++ b/progs/openvg/trivial/dash.c
@@ -0,0 +1,95 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <X11/keysym.h>
+#include <stdio.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.4, 0.1, 1.0, 1.0};
+
+VGPath path;
+VGPaint fill;
+
+VGint cap_style = VG_CAP_BUTT;
+
+static void
+init(void)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_LINE_TO_ABS,
+                                  VG_LINE_TO_ABS
+   };
+#if 1
+   static const VGfloat coords[]   = {100, 100, 150, 100,
+                                      150, 200
+   };
+#else
+   static const VGfloat coords[]   = {100, 20, 100, 220,
+   };
+#endif
+   VGfloat dash_pattern[2] = { 20.f, 20.f };
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 3, cmds, coords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+   vgSetf(VG_STROKE_LINE_WIDTH, 20);
+   vgSeti(VG_STROKE_CAP_STYLE, cap_style);
+   vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_ROUND);
+   vgSetfv(VG_STROKE_DASH_PATTERN, 2, dash_pattern);
+   vgSetf(VG_STROKE_DASH_PHASE, 0.0f);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_STROKE_PATH);
+
+   vgFlush();
+}
+
+static int  key_press(unsigned key)
+{
+    switch(key) {
+    case XK_c:
+    case XK_C:
+        ++cap_style;
+        if (cap_style > VG_CAP_SQUARE)
+            cap_style = VG_CAP_BUTT;
+        switch(cap_style) {
+        case VG_CAP_BUTT:
+            fprintf(stderr, "Cap style 'butt'\n");
+            break;
+        case VG_CAP_ROUND:
+            fprintf(stderr, "Cap style 'round'\n");
+            break;
+        case VG_CAP_SQUARE:
+            fprintf(stderr, "Cap style 'square'\n");
+            break;
+        }
+        vgSeti(VG_STROKE_CAP_STYLE, cap_style);
+        break;
+    default:
+        break;
+    }
+
+    return VG_TRUE;
+}
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, key_press);
+}
diff --git a/progs/openvg/trivial/eglcommon.c b/progs/openvg/trivial/eglcommon.c
new file mode 100644
index 0000000000..bacd5685d7
--- /dev/null
+++ b/progs/openvg/trivial/eglcommon.c
@@ -0,0 +1,288 @@
+#include "eglcommon.h"
+
+
+#include <assert.h>
+#include <math.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include <X11/keysym.h>
+#include <VG/openvg.h>  /* using full OpenGL for now */
+#include <GLES/egl.h>
+
+
+static init_func    init = 0;
+static draw_func    draw = 0;
+static reshape_func reshape = 0;
+static key_func     keyPress = 0;
+static VGint width = 300, height = 300;
+
+
+void set_window_size(int w, int h)
+{
+   width = w;
+   height = h;
+}
+
+/*
+ * Create an RGB, double-buffered X window.
+ * Return the window and context handles.
+ */
+static void
+make_x_window(Display *x_dpy, EGLDisplay egl_dpy,
+              const char *name,
+              int x, int y, int width, int height,
+              Window *winRet,
+              EGLContext *ctxRet,
+              EGLSurface *surfRet)
+{
+   static const EGLint attribs[] = {
+      EGL_RED_SIZE, 1,
+      EGL_GREEN_SIZE, 1,
+      EGL_BLUE_SIZE, 1,
+      EGL_NONE
+   };
+
+   int scrnum;
+   XSetWindowAttributes attr;
+   unsigned long mask;
+   Window root;
+   Window win;
+   XVisualInfo *visInfo, visTemplate;
+   int num_visuals;
+   EGLContext ctx;
+   EGLConfig config;
+   EGLint num_configs;
+   EGLint vid;
+
+   scrnum = DefaultScreen( x_dpy );
+   root = RootWindow( x_dpy, scrnum );
+
+   if (!eglChooseConfig( egl_dpy, attribs, &config, 1, &num_configs)) {
+      printf("Error: couldn't get an EGL visual config\n");
+      exit(1);
+   }
+
+   assert(config);
+   assert(num_configs > 0);
+
+   if (!eglGetConfigAttrib(egl_dpy, config, EGL_NATIVE_VISUAL_ID, &vid)) {
+      printf("Error: eglGetConfigAttrib() failed\n");
+      exit(1);
+   }
+
+   /* The X window visual must match the EGL config */
+   visTemplate.visualid = vid;
+   visInfo = XGetVisualInfo(x_dpy, VisualIDMask, &visTemplate, &num_visuals);
+   if (!visInfo) {
+      printf("Error: couldn't get X visual\n");
+      exit(1);
+   }
+
+   /* window attributes */
+   attr.background_pixel = 0;
+   attr.border_pixel = 0;
+   attr.colormap = XCreateColormap( x_dpy, root, visInfo->visual, AllocNone);
+   attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask;
+   mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask;
+
+   win = XCreateWindow( x_dpy, root, 0, 0, width, height,
+		        0, visInfo->depth, InputOutput,
+		        visInfo->visual, mask, &attr );
+
+   /* set hints and properties */
+   {
+      XSizeHints sizehints;
+      sizehints.x = x;
+      sizehints.y = y;
+      sizehints.width  = width;
+      sizehints.height = height;
+      sizehints.flags = USSize | USPosition;
+      XSetNormalHints(x_dpy, win, &sizehints);
+      XSetStandardProperties(x_dpy, win, name, name,
+                              None, (char **)NULL, 0, &sizehints);
+   }
+
+   eglBindAPI(EGL_OPENVG_API);
+
+   ctx = eglCreateContext(egl_dpy, config, EGL_NO_CONTEXT, NULL );
+   if (!ctx) {
+      printf("Error: eglCreateContext failed\n");
+      exit(1);
+   }
+
+   *surfRet = eglCreateWindowSurface(egl_dpy, config, win, NULL);
+
+   if (!*surfRet) {
+      printf("Error: eglCreateWindowSurface failed\n");
+      exit(1);
+   }
+
+   XFree(visInfo);
+
+   *winRet = win;
+   *ctxRet = ctx;
+}
+
+static void
+event_loop(Display *dpy, Window win,
+           EGLDisplay egl_dpy, EGLSurface egl_surf)
+{
+   while (1) {
+      int redraw = 0;
+      XEvent event;
+
+      XNextEvent(dpy, &event);
+
+      switch (event.type) {
+      case Expose:
+         redraw = 1;
+         break;
+      case ConfigureNotify:
+         if (reshape) {
+            width = event.xconfigure.width;
+            height = event.xconfigure.height;
+            reshape(event.xconfigure.width, event.xconfigure.height);
+         }
+         break;
+      case KeyPress:
+      {
+         char buffer[10];
+         int r, code;
+         code = XLookupKeysym(&event.xkey, 0);
+         if (!keyPress || !keyPress(code)) {
+            r = XLookupString(&event.xkey, buffer, sizeof(buffer),
+                              NULL, NULL);
+            if (buffer[0] == 27) {
+               /* escape */
+               return;
+            }
+         }
+      }
+      redraw = 1;
+      break;
+      default:
+         ; /*no-op*/
+      }
+
+      if (redraw) {
+         draw();
+         eglSwapBuffers(egl_dpy, egl_surf);
+      }
+   }
+}
+
+int window_width(void)
+{
+   return width;
+}
+
+int window_height(void)
+{
+   return height;
+}
+
+static void
+usage(void)
+{
+   printf("Usage:\n");
+   printf("  -display <displayname>  set the display to run on\n");
+   printf("  -info                   display OpenGL renderer info\n");
+}
+
+int run(int argc, char **argv,
+        init_func init_f,
+        reshape_func resh_f,
+        draw_func draw_f,
+        key_func key_f)
+{
+   const int winWidth = width, winHeight = height;
+   Display *x_dpy;
+   Window win;
+   EGLSurface egl_surf;
+   EGLContext egl_ctx;
+   EGLDisplay egl_dpy;
+   char *dpyName = NULL;
+   GLboolean printInfo = GL_FALSE;
+   EGLint egl_major, egl_minor;
+   int i;
+   const char *s;
+
+   init = init_f;
+   draw = draw_f;
+   reshape = resh_f;
+   keyPress = key_f;
+
+   for (i = 1; i < argc; i++) {
+      if (strcmp(argv[i], "-display") == 0) {
+         dpyName = argv[i+1];
+         i++;
+      }
+      else if (strcmp(argv[i], "-info") == 0) {
+         printInfo = GL_TRUE;
+      }
+   }
+
+   x_dpy = XOpenDisplay(dpyName);
+   if (!x_dpy) {
+      printf("Error: couldn't open display %s\n",
+	     dpyName ? dpyName : getenv("DISPLAY"));
+      return -1;
+   }
+
+   egl_dpy = eglGetDisplay(x_dpy);
+   if (!egl_dpy) {
+      printf("Error: eglGetDisplay() failed\n");
+      return -1;
+   }
+
+   if (!eglInitialize(egl_dpy, &egl_major, &egl_minor)) {
+      printf("Error: eglInitialize() failed\n");
+      return -1;
+   }
+
+   s = eglQueryString(egl_dpy, EGL_VERSION);
+   printf("EGL_VERSION = %s\n", s);
+
+   make_x_window(x_dpy, egl_dpy,
+                 "OpenVG Example", 0, 0, winWidth, winHeight,
+                 &win, &egl_ctx, &egl_surf);
+
+   XMapWindow(x_dpy, win);
+   if (!eglMakeCurrent(egl_dpy, egl_surf, egl_surf, egl_ctx)) {
+      printf("Error: eglMakeCurrent() failed\n");
+      return -1;
+   }
+
+   if (printInfo) {
+      printf("VG_RENDERER   = %s\n", (char *) vgGetString(VG_RENDERER));
+      printf("VG_VERSION    = %s\n", (char *) vgGetString(VG_VERSION));
+      printf("VG_VENDOR     = %s\n", (char *) vgGetString(VG_VENDOR));
+   }
+
+   if (init)
+      init();
+
+   /* Set initial projection/viewing transformation.
+    * We can't be sure we'll get a ConfigureNotify event when the window
+    * first appears.
+    */
+   if (reshape)
+      reshape(winWidth, winHeight);
+
+   event_loop(x_dpy, win, egl_dpy, egl_surf);
+
+   eglMakeCurrent(egl_dpy, 0, 0, 0);
+   eglDestroyContext(egl_dpy, egl_ctx);
+   eglDestroySurface(egl_dpy, egl_surf);
+   eglTerminate(egl_dpy);
+
+
+   XDestroyWindow(x_dpy, win);
+   XCloseDisplay(x_dpy);
+
+   return 0;
+}
+
diff --git a/progs/openvg/trivial/eglcommon.h b/progs/openvg/trivial/eglcommon.h
new file mode 100644
index 0000000000..958dae9f98
--- /dev/null
+++ b/progs/openvg/trivial/eglcommon.h
@@ -0,0 +1,20 @@
+#ifndef EGLCOMMON_H
+#define EGLCOMMON_H
+
+typedef void (*init_func)();
+typedef void (*reshape_func)(int, int);
+typedef void (*draw_func)();
+typedef int  (*key_func)(unsigned key);
+
+
+void set_window_size(int width, int height);
+int window_width(void);
+int window_height(void);
+
+int run(int argc, char **argv,
+        init_func init,
+        reshape_func resh,
+        draw_func draw,
+        key_func key);
+
+#endif
diff --git a/progs/openvg/trivial/ellipse.c b/progs/openvg/trivial/ellipse.c
new file mode 100644
index 0000000000..4c7d4904f8
--- /dev/null
+++ b/progs/openvg/trivial/ellipse.c
@@ -0,0 +1,84 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <math.h>
+#include <stdlib.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.0, 0.0, 0.0, 1.0};
+
+VGPath path;
+VGPaint paint;
+
+static void
+init(void)
+{
+   VGfloat clearColor[] = {1.0f, 1.0f, 1.0f, 1.0f};/* white color */
+   VGfloat fillColor[] = {1.0f, 0.0f, 0.0f, 1.0f};/* red color */
+   static const VGubyte segments[4] = {VG_MOVE_TO_ABS,
+                                       VG_SCCWARC_TO_ABS,
+                                       VG_SCCWARC_TO_ABS,
+                                       VG_CLOSE_PATH};
+   VGfloat data[12];
+   const VGfloat cx = 0, cy=29, width=80, height=40;
+   const VGfloat hw = width * 0.5f;
+   const VGfloat hh = height * 0.5f;
+
+   data[0] = cx + hw;
+   data[1] = cy;
+   data[2] = hw;
+   data[3] = hh;
+   data[4] = 0;
+   data[5] = cx - hw;
+   data[6] = cy;
+   data[7] = hw;
+   data[8] = hh;
+   data[9] = 0;
+   data[10] = data[0];
+   data[11] = cy;
+
+   vgSetfv(VG_CLEAR_COLOR, 4, clearColor);
+   vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+
+
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                       1.0f, 0.0f, 0, 0, VG_PATH_CAPABILITY_ALL);
+   if (path == VG_INVALID_HANDLE) {
+      return;
+   }
+   paint = vgCreatePaint();
+   if (paint == VG_INVALID_HANDLE) {
+      vgDestroyPath(path);
+      return;
+   }
+
+   vgAppendPathData(path, 4, segments, data);
+   vgSetParameterfv(paint, VG_PAINT_COLOR, 4, fillColor);
+   vgSetParameteri( paint, VG_PAINT_TYPE, VG_PAINT_TYPE_COLOR);
+   vgSetPaint(paint, VG_FILL_PATH);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgLoadIdentity();
+   vgTranslate(50, 21);
+   vgDrawPath(path, VG_FILL_PATH);
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   set_window_size(100, 100);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/filter.c b/progs/openvg/trivial/filter.c
new file mode 100644
index 0000000000..d96257a933
--- /dev/null
+++ b/progs/openvg/trivial/filter.c
@@ -0,0 +1,107 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <math.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {1.0, 1.0, 1.0, 0.5};
+
+VGImage srcImg;
+VGImage dstImg;
+
+VGPaint fill;
+
+VGfloat bgCol[4] = {0.906f, 0.914f, 0.761f, 1.0f};
+
+static void
+init(void)
+{
+    VGfloat red[4];
+    VGfloat grey[4];
+    VGfloat orange[4];
+    VGfloat blue[4];
+    VGfloat black[4];
+    VGfloat white[4];
+    VGshort transKernel[49] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+    red[0] = 0.6710f;
+    red[1] = 0.1060f;
+    red[2] = 0.1330f;
+    red[3] = 1.0f;
+
+    grey[0] = 0.6347f;
+    grey[1] = 0.6561f;
+    grey[2] = 0.6057f;
+    grey[3] = 1.0f;
+
+    orange[0] = 1.0000f;
+    orange[1] = 0.8227f;
+    orange[2] = 0.5057f;
+    orange[3] = 1.0f;
+
+    blue[0] = 0.0000f;
+    blue[1] = 0.6908f;
+    blue[2] = 0.8595f;
+    blue[3] = 1.0f;
+
+    black[0] = 0;
+    black[1] = 0;
+    black[2] = 0;
+    black[3] = 1.0f;
+
+    white[0] = 1;
+    white[1] = 1;
+    white[2] = 1;
+    white[3] = 1.0f;
+
+    vgSetfv(VG_TILE_FILL_COLOR, 4, blue);
+
+    vgSeti(VG_FILTER_CHANNEL_MASK, 14);
+
+    /* Setup images */
+    srcImg = vgCreateImage(VG_sRGBA_8888, 32, 32,
+                           VG_IMAGE_QUALITY_NONANTIALIASED);
+    dstImg = vgCreateImage(VG_sRGBA_8888, 32, 32,
+                           VG_IMAGE_QUALITY_NONANTIALIASED);
+
+    vgSetfv(VG_CLEAR_COLOR, 4, black);
+    vgClearImage(srcImg, 0, 0, 32, 32);
+    vgSetfv(VG_CLEAR_COLOR, 4, red);
+    vgClearImage(srcImg, 3, 3, 27, 27);
+
+    vgSetfv(VG_CLEAR_COLOR, 4, orange);
+    vgClearImage(dstImg, 0, 0, 32, 32);
+
+    transKernel[8] = 1;
+    vgConvolve(dstImg, srcImg, 3, 3, 3, 0, transKernel,
+               1, 0, VG_TILE_FILL);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static void
+draw(void)
+{
+   vgSetfv(VG_CLEAR_COLOR, 4, bgCol);
+   vgClear(0, 0, window_width(), window_height());
+   vgSeti(VG_MATRIX_MODE, VG_MATRIX_IMAGE_USER_TO_SURFACE);
+   vgLoadIdentity();
+   vgTranslate(10, 10);
+   vgDrawImage(dstImg);
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   set_window_size(64, 64);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/gradorigin.c b/progs/openvg/trivial/gradorigin.c
new file mode 100644
index 0000000000..b376263fe5
--- /dev/null
+++ b/progs/openvg/trivial/gradorigin.c
@@ -0,0 +1,98 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <stdio.h>
+#include <string.h>
+
+static const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+
+static VGPath path;
+static VGPaint fill;
+
+VGColorRampSpreadMode spread = VG_COLOR_RAMP_SPREAD_PAD;
+
+static void
+init(void)
+{
+   VGubyte commands[5] =  {VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS, VG_CLOSE_PATH};
+   VGfloat coords[8] =    {0.0f,0.0f, 32.0f,0.0f, 32.0f,32.0f, 0.0f,32.0f };
+
+   VGfloat rampStop[20] = {-0.5f,  1.0f, 1.0f, 1.0f, 1.0f,
+                           0.25f, 1.0f, 0.0f, 0.0f, 1.0f,
+                           0.75f, 0.0f, 0.0f, 1.0f, 1.0f,
+                           1.5f,  0.0f, 0.0f, 0.0f, 0.0f};
+
+   VGfloat defaultColor[] = {1.0f, 1.0f, 1.0f, 1.0f};
+   VGfloat linearGradient[4] = {0.0f, 0.0f, 0.0f, 32.0f};
+
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                       1.0f, 0.0f, 0, 0, VG_PATH_CAPABILITY_ALL);
+   if (path == VG_INVALID_HANDLE)
+      return;
+
+   fill = vgCreatePaint();
+   if (fill == VG_INVALID_HANDLE) {
+      vgDestroyPath(path);
+      return;
+   }
+
+   vgSetfv(VG_CLEAR_COLOR, 4, defaultColor);
+   vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+
+   vgAppendPathData(path, 5, commands, coords);
+
+   vgSetPaint(fill, VG_FILL_PATH);
+   vgSetParameteri(fill, VG_PAINT_TYPE, VG_PAINT_TYPE_LINEAR_GRADIENT);
+   vgSetParameteri(fill, VG_PAINT_COLOR_RAMP_SPREAD_MODE,
+                   VG_COLOR_RAMP_SPREAD_REPEAT);
+   vgSetParameterfv(fill, VG_PAINT_LINEAR_GRADIENT, 4, linearGradient);
+   vgSetParameterfv(fill, VG_PAINT_COLOR_RAMP_STOPS, 20, rampStop);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+
+   vgDrawPath(path, VG_FILL_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   if (argc > 1) {
+      const char *arg = argv[1];
+      if (!strcmp("-pad", arg))
+         spread = VG_COLOR_RAMP_SPREAD_PAD;
+      else if (!strcmp("-repeat", arg))
+         spread = VG_COLOR_RAMP_SPREAD_REPEAT;
+      else if (!strcmp("-reflect", arg))
+         spread = VG_COLOR_RAMP_SPREAD_REFLECT;
+   }
+
+   switch(spread) {
+   case VG_COLOR_RAMP_SPREAD_PAD:
+      printf("Using spread mode: pad\n");
+      break;
+   case VG_COLOR_RAMP_SPREAD_REPEAT:
+      printf("Using spread mode: repeat\n");
+      break;
+   case VG_COLOR_RAMP_SPREAD_REFLECT:
+      printf("Using spread mode: reflect\n");
+   }
+
+   set_window_size(200, 200);
+
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/lineto.c b/progs/openvg/trivial/lineto.c
new file mode 100644
index 0000000000..94e2981811
--- /dev/null
+++ b/progs/openvg/trivial/lineto.c
@@ -0,0 +1,56 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.4, 0.1, 1.0, 1.0};
+
+VGPath path;
+VGPaint fill;
+
+
+static void
+init(void)
+{
+   static const VGubyte sqrCmds[5] = {VG_MOVE_TO_ABS, VG_HLINE_TO_ABS, VG_VLINE_TO_ABS, VG_HLINE_TO_ABS, VG_CLOSE_PATH};
+   static const VGfloat sqrCoords[5]   = {50.0f, 50.0f, 250.0f, 250.0f, 50.0f};
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 5, sqrCmds, sqrCoords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+   vgSetf(VG_STROKE_LINE_WIDTH, 10);
+   vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_BUTT);
+   vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_ROUND);
+   vgSetf(VG_STROKE_MITER_LIMIT, 4.0f);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgSeti(VG_MATRIX_MODE, VG_MATRIX_STROKE_PAINT_TO_USER);
+   vgLoadIdentity();
+   vgScale(2.25, 2.25);
+   vgDrawPath(path, VG_STROKE_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/lingrad.c b/progs/openvg/trivial/lingrad.c
new file mode 100644
index 0000000000..bcaad1f101
--- /dev/null
+++ b/progs/openvg/trivial/lingrad.c
@@ -0,0 +1,87 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <stdio.h>
+#include <string.h>
+
+static const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+
+static VGPath path;
+static VGPaint fill;
+
+VGColorRampSpreadMode spread = VG_COLOR_RAMP_SPREAD_PAD;
+
+static void
+init(void)
+{
+   static const VGubyte sqrCmds[5] = {VG_MOVE_TO_ABS, VG_HLINE_TO_ABS, VG_VLINE_TO_ABS, VG_HLINE_TO_ABS, VG_CLOSE_PATH};
+   static const VGfloat sqrCoords[5]   = {0.0f, 0.0f, 400.0f, 400.0f, 0.0f};
+
+   VGfloat rampStop[] = {0.00f, 1.0f, 1.0f, 1.0f, 1.0f,
+                         0.33f, 1.0f, 0.0f, 0.0f, 1.0f,
+                         0.66f, 0.0f, 1.0f, 0.0f, 1.0f,
+                         1.00f, 0.0f, 0.0f,  1.0f, 1.0f};
+   VGfloat linearGradient[4] = {100.0f, 100.0f, 300.0f, 300.0f};
+
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 5, sqrCmds, sqrCoords);
+
+   fill = vgCreatePaint();
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetParameteri(fill, VG_PAINT_TYPE, VG_PAINT_TYPE_LINEAR_GRADIENT);
+   vgSetParameteri(fill, VG_PAINT_COLOR_RAMP_SPREAD_MODE, spread);
+   vgSetParameterfv(fill, VG_PAINT_LINEAR_GRADIENT, 4, linearGradient);
+   vgSetParameterfv(fill, VG_PAINT_COLOR_RAMP_STOPS, 20, rampStop);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+
+   vgDrawPath(path, VG_FILL_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   if (argc > 1) {
+      const char *arg = argv[1];
+      if (!strcmp("-pad", arg))
+         spread = VG_COLOR_RAMP_SPREAD_PAD;
+      else if (!strcmp("-repeat", arg))
+         spread = VG_COLOR_RAMP_SPREAD_REPEAT;
+      else if (!strcmp("-reflect", arg))
+         spread = VG_COLOR_RAMP_SPREAD_REFLECT;
+   }
+
+   switch(spread) {
+   case VG_COLOR_RAMP_SPREAD_PAD:
+      printf("Using spread mode: pad\n");
+      break;
+   case VG_COLOR_RAMP_SPREAD_REPEAT:
+      printf("Using spread mode: repeat\n");
+      break;
+   case VG_COLOR_RAMP_SPREAD_REFLECT:
+      printf("Using spread mode: reflect\n");
+   }
+
+   set_window_size(400, 400);
+
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/lookup.c b/progs/openvg/trivial/lookup.c
new file mode 100644
index 0000000000..a103ba4488
--- /dev/null
+++ b/progs/openvg/trivial/lookup.c
@@ -0,0 +1,71 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <math.h>
+#include <stdlib.h>
+#include <stdio.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {1.0, 1.0, 1.0, 0.5};
+VGfloat clearColor[] = {1.0f, 0.0f, 0.0f, 1.0f};/* red color */
+VGImage parent;
+
+VGPaint fill;
+
+static void
+init(void)
+{
+    VGImage child1, child2;
+    VGubyte *data;
+    VGuint  LUT[256];
+    VGint i;
+
+    data = (VGubyte *)malloc(sizeof(VGubyte)*window_width()*window_height());
+
+    for (i=0;i<window_width()*window_height();i++) {
+        data[i] = 0x00;
+    }
+
+    for (i=0; i<256; i++) {
+        if ( i == 0 )
+            LUT[0] = 0xFFFFFFFF;
+        else
+            LUT[i] = 0xFF00FFFF;
+    }
+
+    parent = vgCreateImage( VG_A_8, 64, 64, VG_IMAGE_QUALITY_NONANTIALIASED );
+
+    vgImageSubData(parent, data, window_width(), VG_A_8, 0, 0,
+                   window_width(), window_height());
+    child1 = vgChildImage(parent, 0, 0, 32, 64);
+    child2 = vgChildImage(parent, 32, 0, 32, 64);
+
+    vgLookupSingle(child2, child1, LUT, VG_GREEN, VG_FALSE, VG_TRUE);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static void
+draw(void)
+{
+   vgSetfv(VG_CLEAR_COLOR, 4, clearColor);
+   vgClear(0, 0, window_width(), window_height());
+   //vgSeti(VG_MATRIX_MODE, VG_MATRIX_IMAGE_USER_TO_SURFACE);
+   //vgLoadIdentity();
+   //vgTranslate(10, 10);
+   vgDrawImage(parent);
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   set_window_size(64, 64);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/mask.c b/progs/openvg/trivial/mask.c
new file mode 100644
index 0000000000..e5c00c5699
--- /dev/null
+++ b/progs/openvg/trivial/mask.c
@@ -0,0 +1,58 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.4, 0.1, 1.0, 1.0};
+
+VGPath path;
+VGPaint fill;
+
+
+static void
+init(void)
+{
+   static const VGubyte sqrCmds[5] = {VG_MOVE_TO_ABS, VG_HLINE_TO_ABS, VG_VLINE_TO_ABS, VG_HLINE_TO_ABS, VG_CLOSE_PATH};
+   static const VGfloat sqrCoords[5]   = {50.0f, 50.0f, 250.0f, 250.0f, 50.0f};
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 5, sqrCmds, sqrCoords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+   vgSetf(VG_STROKE_LINE_WIDTH, 10);
+   vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_BUTT);
+   vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_ROUND);
+   vgSetf(VG_STROKE_MITER_LIMIT, 4.0f);
+
+   vgSeti(VG_MASKING, VG_TRUE);
+
+   vgMask(VG_INVALID_HANDLE, VG_CLEAR_MASK,
+          25, 25, 100, 100);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_FILL_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/mask4.c b/progs/openvg/trivial/mask4.c
new file mode 100644
index 0000000000..fe6db39648
--- /dev/null
+++ b/progs/openvg/trivial/mask4.c
@@ -0,0 +1,132 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <VG/vgu.h>
+#include <stdio.h>
+#include <math.h>
+#include <stdlib.h>
+
+#include <X11/keysym.h>
+
+//VGint x_pos = -10, y_pos = -10;
+VGint x_pos = 0, y_pos = 4;
+VGint img_width = 120, img_height = 120;
+
+static void RectToPath(VGPath path, VGfloat x, VGfloat y, VGfloat width, VGfloat height)
+{
+    static const VGubyte segments[5] = {VG_MOVE_TO_ABS,
+                                        VG_HLINE_TO_ABS,
+                                        VG_VLINE_TO_ABS,
+                                        VG_HLINE_TO_ABS,
+                                        VG_CLOSE_PATH};
+    VGfloat data[5];
+
+    data[0] = x;
+    data[1] = y;
+    data[2] = x + width;
+    data[3] = y + height;
+    data[4] = x;
+
+    vgAppendPathData(path, 5, segments, data);
+}
+
+static void
+init(void)
+{
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+int  key_press(unsigned key)
+{
+    switch(key) {
+    case XK_Right:
+        x_pos +=1;
+        break;
+    case XK_Left:
+        x_pos -=1;
+        break;
+    case XK_Up:
+        y_pos +=1;
+        break;
+    case XK_Down:
+        y_pos -=1;
+        break;
+    case 'a':
+        img_width  -= 5;
+        img_height -= 5;
+        break;
+    case 's':
+        img_width  += 5;
+        img_height += 5;
+        break;
+    default:
+        break;
+    }
+    fprintf(stderr, "Posi = %dx%d\n", x_pos, y_pos);
+    fprintf(stderr, "Size = %dx%d\n", img_width, img_height);
+    return VG_FALSE;
+}
+
+static void
+draw(void)
+{
+    VGint WINDSIZEX = window_width();
+    VGint WINDSIZEY = window_height();
+
+    VGPaint fill;
+    VGPath box;
+    VGfloat color[4]		= {1.f, 0.f, 0.f, 1.f};
+    VGfloat bgCol[4]		= {0.7f, 0.7f, 0.7f, 1.0f};
+    VGfloat transCol[4]         = {0.f, 0.f, 0.f, 0.f};
+    VGImage image = vgCreateImage(VG_sRGBA_8888, img_width, img_height,
+                                  VG_IMAGE_QUALITY_NONANTIALIASED);
+
+    /* Background clear */
+    fill = vgCreatePaint();
+    vgSetParameterfv(fill, VG_PAINT_COLOR, 4, color);
+    vgSetPaint(fill, VG_FILL_PATH);
+
+    box = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                       1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+    /* Rectangle to cover completely 16x16 pixel area. */
+    RectToPath(box, 0, 0, 64, 64);
+
+    vgSetfv(VG_CLEAR_COLOR, 4, transCol);
+    vgClearImage(image, 0, 0, img_width, img_height);
+    vgSetfv(VG_CLEAR_COLOR, 4, color);
+    vgClearImage(image, 10, 10, 12, 12);
+    //vgImageSubData(image, pukki_64x64_data, pukki_64x64_stride,
+    //               VG_sRGBA_8888, 0, 0, 32, 32);
+    vgSeti(VG_MASKING, VG_TRUE);
+    vgLoadIdentity();
+
+    vgSetfv(VG_CLEAR_COLOR, 4, bgCol);
+    vgClear(0, 0, WINDSIZEX, WINDSIZEY);
+
+
+    vgMask(image, VG_FILL_MASK, 0, 0, window_width(), window_height());
+    vgMask(image, VG_SET_MASK, x_pos, y_pos, 100, 100);
+
+    vgDrawPath(box, VG_FILL_PATH);
+
+    //vgSeti(VG_MATRIX_MODE, VG_MATRIX_IMAGE_USER_TO_SURFACE);
+    //vgTranslate(-10, -10);
+    //vgDrawImage(image);
+
+
+    vgDestroyPaint(fill);
+    vgDestroyPath(box);
+}
+
+
+int main(int argc, char **argv)
+{
+    set_window_size(64, 64);
+    return run(argc, argv, init, reshape,
+               draw, key_press);
+}
diff --git a/progs/openvg/trivial/path3.c b/progs/openvg/trivial/path3.c
new file mode 100644
index 0000000000..5ce600f65a
--- /dev/null
+++ b/progs/openvg/trivial/path3.c
@@ -0,0 +1,77 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <VG/vgu.h>
+#include <stdio.h>
+#include <math.h>
+#include <stdlib.h>
+
+static void
+init(void)
+{
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+
+static void
+draw(void)
+{
+    VGint WINDSIZEX = window_width();
+    VGint WINDSIZEY = window_height();
+    VGPath path;
+    VGPaint paint;
+
+    VGfloat clearColor[] = {1.0f, 1.0f, 1.0f, 0.0f};/* white color */
+    VGfloat fillColor[] = {1.0f, 0.0f, 0.0f, 1.0f};/* red color */
+
+#if 1
+    VGubyte commands[4] = {VG_MOVE_TO_ABS, VG_LCWARC_TO_ABS, VG_SCWARC_TO_ABS, VG_CLOSE_PATH};
+#else
+    VGubyte commands[4] = {VG_MOVE_TO_ABS, VG_SCCWARC_TO_ABS, VG_LCCWARC_TO_ABS,VG_CLOSE_PATH};
+#endif
+    VGfloat coords[] = {32.0f,   0.0f,
+                        -32.0f, -32.0f, 0.0f, 64.0f, 32.0f,
+                        -32.0f, -32.0f, 0.0f, 32.0f, 0.0f};
+
+
+    vgSetfv(VG_CLEAR_COLOR, 4, clearColor);
+    vgClear(0, 0, WINDSIZEX, WINDSIZEY);
+    vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+
+    vgSeti(VG_MATRIX_MODE, VG_MATRIX_PATH_USER_TO_SURFACE);
+    vgLoadIdentity();
+    //vgTranslate(32.0f, 32.0f);
+
+    path = vgCreatePath( VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                         1.0f, 0.0f, 0, 0, VG_PATH_CAPABILITY_ALL );
+    if ( path == VG_INVALID_HANDLE ) {
+        return;
+    }
+    paint = vgCreatePaint();
+    if ( paint == VG_INVALID_HANDLE ) {
+        vgDestroyPath(path);
+        return;
+    }
+
+    vgAppendPathData(path, 4, commands, coords);
+    vgSetParameterfv(paint, VG_PAINT_COLOR, 4, fillColor);
+    vgSetParameteri( paint, VG_PAINT_TYPE, VG_PAINT_TYPE_COLOR);
+    vgSetPaint(paint, VG_FILL_PATH);
+    vgDrawPath(path, VG_FILL_PATH);
+
+    vgDestroyPath(path);
+    vgDestroyPaint(paint);
+}
+
+
+int main(int argc, char **argv)
+{
+    set_window_size(64, 64);
+    return run(argc, argv, init, reshape,
+               draw, 0);
+}
diff --git a/progs/openvg/trivial/radialgrad.c b/progs/openvg/trivial/radialgrad.c
new file mode 100644
index 0000000000..cf3b1d522d
--- /dev/null
+++ b/progs/openvg/trivial/radialgrad.c
@@ -0,0 +1,99 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <stdio.h>
+#include <string.h>
+
+static const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+
+static VGPath path;
+static VGPaint fill;
+
+
+VGfloat centeredGradient[5] = {200.0f, 200.0f, 200.0f, 200.0f, 100};
+VGfloat noncenteredGradient[5] = {200.0f, 200.0f, 250.0f, 250.0f, 100};
+VGfloat *radialGradient = centeredGradient;
+
+VGColorRampSpreadMode spread = VG_COLOR_RAMP_SPREAD_PAD;
+
+static void
+init(void)
+{
+   static const VGubyte sqrCmds[5] = {VG_MOVE_TO_ABS, VG_HLINE_TO_ABS, VG_VLINE_TO_ABS, VG_HLINE_TO_ABS, VG_CLOSE_PATH};
+   static const VGfloat sqrCoords[5]   = {0.0f, 0.0f, 400.0f, 400.0f, 0.0f};
+
+   VGfloat rampStop[] = {0.00f, 1.0f, 1.0f, 1.0f, 1.0f,
+                         0.33f, 1.0f, 0.0f, 0.0f, 1.0f,
+                         0.66f, 0.0f, 1.0f, 0.0f, 1.0f,
+                         1.00f, 0.0f, 0.0f,  1.0f, 1.0f};
+
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 5, sqrCmds, sqrCoords);
+
+   fill = vgCreatePaint();
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetParameteri(fill, VG_PAINT_TYPE, VG_PAINT_TYPE_RADIAL_GRADIENT);
+   vgSetParameteri(fill, VG_PAINT_COLOR_RAMP_SPREAD_MODE, spread);
+   vgSetParameterfv(fill, VG_PAINT_RADIAL_GRADIENT, 5, radialGradient);
+   vgSetParameterfv(fill, VG_PAINT_COLOR_RAMP_STOPS, 20, rampStop);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+
+   vgDrawPath(path, VG_FILL_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   VGint i;
+   for (i = 1; i < argc; ++i) {
+      const char *arg = argv[i];
+      if (!strcmp("-pad", arg))
+         spread = VG_COLOR_RAMP_SPREAD_PAD;
+      else if (!strcmp("-repeat", arg))
+         spread = VG_COLOR_RAMP_SPREAD_REPEAT;
+      else if (!strcmp("-reflect", arg))
+         spread = VG_COLOR_RAMP_SPREAD_REFLECT;
+      else if (!strcmp("-center", arg)) {
+         printf("Centered radial gradient\n");
+         radialGradient = centeredGradient;
+      } else if (!strcmp("-noncenter", arg)) {
+         printf("Non centered radial gradient\n");
+         radialGradient = noncenteredGradient;
+      }
+   }
+
+   switch(spread) {
+   case VG_COLOR_RAMP_SPREAD_PAD:
+      printf("Using spread mode: pad\n");
+      break;
+   case VG_COLOR_RAMP_SPREAD_REPEAT:
+      printf("Using spread mode: repeat\n");
+      break;
+   case VG_COLOR_RAMP_SPREAD_REFLECT:
+      printf("Using spread mode: reflect\n");
+   }
+
+   set_window_size(400, 400);
+
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/readpixels.c b/progs/openvg/trivial/readpixels.c
new file mode 100644
index 0000000000..c8e286db9a
--- /dev/null
+++ b/progs/openvg/trivial/readpixels.c
@@ -0,0 +1,75 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+
+float red_color[4] = {1.0, 0.0, 0.0, 1.0};
+float blue_color[4] = {0.0, 0.0, 1.0, 1.0};
+VGint *data;
+
+static void
+init(void)
+{
+   data = malloc(sizeof(VGint)*2048*2048);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   static const VGint red_pixel  = 255 << 24 | 255 << 16 | 0 << 8 |   0;
+   static const VGint blue_pixel = 255 << 24 |   0 << 16 | 0 << 8 | 255;
+   VGint i;
+
+   vgSetfv(VG_CLEAR_COLOR, 4, red_color);
+   vgClear(0, 0, window_width(), window_height());
+   vgFlush();
+
+   memset(data, 0, window_width() * window_height() * sizeof(VGint));
+
+   vgReadPixels(data, window_width() * sizeof(VGint),
+                VG_lARGB_8888,
+                0, 0, window_width(), window_height());
+
+   fprintf(stderr, "Red 0 = 0x%x and at 600 = 0x%x\n",
+           data[0], data[600]);
+   for (i = 0; i < window_width() * window_height(); ++i) {
+      assert(data[i] == red_pixel);
+   }
+
+   vgSetfv(VG_CLEAR_COLOR, 4, blue_color);
+   vgClear(50, 50, 50, 50);
+   vgFlush();
+
+   memset(data, 0, window_width() * window_height() * sizeof(VGint));
+
+   vgReadPixels(data, 50 * sizeof(VGint),
+                VG_lARGB_8888,
+                50, 50, 50, 50);
+
+   fprintf(stderr, "Blue 0 = 0x%x and at 100 = 0x%x\n",
+           data[0], data[100]);
+   for (i = 0; i < 50 * 50; ++i) {
+      assert(data[i] == blue_pixel);
+   }
+}
+
+
+int main(int argc, char **argv)
+{
+   int ret = run(argc, argv, init, reshape,
+                 draw, 0);
+
+   free(data);
+   return ret;
+}
diff --git a/progs/openvg/trivial/roundedrect.c b/progs/openvg/trivial/roundedrect.c
new file mode 100644
index 0000000000..c80a4ed299
--- /dev/null
+++ b/progs/openvg/trivial/roundedrect.c
@@ -0,0 +1,67 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.9, 0.1, 0.1, 0.8};
+
+VGPath path;
+VGPaint fill;
+
+
+static void
+init(void)
+{
+   static const VGubyte sqrCmds[10] = {VG_MOVE_TO_ABS,
+                                       VG_LINE_TO_ABS,
+                                       VG_CUBIC_TO_ABS,
+                                       VG_LINE_TO_ABS,
+                                       VG_CUBIC_TO_ABS,
+                                       VG_LINE_TO_ABS,
+                                       VG_CUBIC_TO_ABS,
+                                       VG_LINE_TO_ABS,
+                                       VG_CUBIC_TO_ABS,
+                                       VG_CLOSE_PATH};
+   static const VGfloat sqrCoords[]   = {
+      45.885571, 62.857143,
+      154.11442, 62.857143,
+      162.1236, 62.857143, 168.57142, 70.260744, 168.57142, 79.457144,
+      168.57142, 123.4,
+      168.57142, 132.5964, 162.1236,  140, 154.11442, 140,
+      45.885571, 140,
+      37.876394, 140, 31.428572, 132.5964, 31.428572, 123.4,
+      31.428572, 79.457144,
+      31.428572, 70.260744, 37.876394,62.857143, 45.885571,62.857143
+   };
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 10, sqrCmds, sqrCoords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+   vgSetf(VG_STROKE_LINE_WIDTH, 6);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_STROKE_PATH);
+}
+
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/star-nonzero.c b/progs/openvg/trivial/star-nonzero.c
new file mode 100644
index 0000000000..012fbd3929
--- /dev/null
+++ b/progs/openvg/trivial/star-nonzero.c
@@ -0,0 +1,55 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat green_color[4] = {0.0, 1.0, 0.0, 0.8};
+
+VGPath path;
+VGPaint fill;
+
+
+static void
+init(void)
+{
+   static const VGubyte cmds[6] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS,
+                                   VG_LINE_TO_ABS, VG_CLOSE_PATH};
+   static const VGfloat coords[]   = {  0,  200,
+                                        300,  200,
+                                        50,   0,
+                                        150, 300,
+                                        250,   0};
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 6, cmds, coords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, green_color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+   vgSeti(VG_FILL_RULE, VG_NON_ZERO);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_FILL_PATH | VG_STROKE_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/star-oddeven.c b/progs/openvg/trivial/star-oddeven.c
new file mode 100644
index 0000000000..17311cf720
--- /dev/null
+++ b/progs/openvg/trivial/star-oddeven.c
@@ -0,0 +1,102 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat green_color[4] = {0.0, 1.0, 0.0, 0.8};
+const VGfloat black_color[4] = {0.0, 0.0, 0.0, 1.0};
+
+VGPath path;
+VGPaint fill;
+
+
+static void draw_point(VGfloat x, VGfloat y)
+{
+
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS,
+                                  VG_LINE_TO_ABS, VG_CLOSE_PATH};
+   const VGfloat coords[]   = {  x - 2,  y - 2,
+                                 x + 2,  y - 2,
+                                 x + 2,  y + 2,
+                                 x - 2,  y + 2};
+   VGPath path;
+   VGPaint fill;
+
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_ALL);
+   vgAppendPathData(path, 5, cmds, coords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, black_color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgDrawPath(path, VG_FILL_PATH);
+
+   vgDestroyPath(path);
+   vgDestroyPaint(fill);
+}
+
+static void draw_marks(VGPath path)
+{
+    VGfloat point[2], tangent[2];
+    int i = 0;
+
+    for (i = 0; i < 1300; i += 50) {
+        vgPointAlongPath(path, 0, 6, i,
+                         point + 0, point + 1,
+                         tangent + 0, tangent + 1);
+        draw_point(point[0], point[1]);
+    }
+}
+
+static void
+init(void)
+{
+   static const VGubyte cmds[6] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS, VG_LINE_TO_ABS,
+                                   VG_LINE_TO_ABS, VG_CLOSE_PATH};
+   static const VGfloat coords[]   = {  0,  200,
+                                        300,  200,
+                                        50,   0,
+                                        150, 300,
+                                        250,   0};
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_ALL);
+   vgAppendPathData(path, 6, cmds, coords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, green_color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+    VGfloat point[2], tangent[2];
+    int i = 0;
+
+    vgClear(0, 0, window_width(), window_height());
+
+    vgSetPaint(fill, VG_FILL_PATH);
+    vgDrawPath(path, VG_FILL_PATH);
+
+    draw_marks(path);
+
+
+    vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/openvg/trivial/stroke.c b/progs/openvg/trivial/stroke.c
new file mode 100644
index 0000000000..58ae5b7bc8
--- /dev/null
+++ b/progs/openvg/trivial/stroke.c
@@ -0,0 +1,116 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <X11/keysym.h>
+#include <stdio.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.4, 0.1, 1.0, 1.0};
+
+VGPath path;
+VGPaint fill;
+
+VGint cap_style = VG_CAP_BUTT;
+VGint join_style = VG_JOIN_MITER;
+
+static void
+init(void)
+{
+#if 0
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_CUBIC_TO_ABS,
+   };
+   static const VGfloat coords[]   = {30, 30, 264, 0, 0, 264, 234, 234
+   };
+#else
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_LINE_TO_ABS,
+                                  VG_LINE_TO_ABS
+   };
+   static const VGfloat coords[]   = {30, 30, 202, 30, 150, 224
+   };
+#endif
+   VGfloat dash_pattern[2] = { 20.f, 20.f };
+   path = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1, 0, 0, 0,
+                       VG_PATH_CAPABILITY_APPEND_TO);
+   vgAppendPathData(path, 3, cmds, coords);
+
+   fill = vgCreatePaint();
+   vgSetParameterfv(fill, VG_PAINT_COLOR, 4, color);
+   vgSetPaint(fill, VG_FILL_PATH);
+
+   vgSetfv(VG_CLEAR_COLOR, 4, white_color);
+   vgSetf(VG_STROKE_LINE_WIDTH, 20);
+   vgSeti(VG_STROKE_CAP_STYLE, cap_style);
+   vgSeti(VG_STROKE_JOIN_STYLE, join_style);
+   vgSetfv(VG_STROKE_DASH_PATTERN, 2, dash_pattern);
+   vgSetf(VG_STROKE_DASH_PHASE, 0.0f);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_STROKE_PATH);
+
+   vgFlush();
+}
+
+static int  key_press(unsigned key)
+{
+    switch(key) {
+    case XK_c:
+    case XK_C:
+        ++cap_style;
+        if (cap_style > VG_CAP_SQUARE)
+            cap_style = VG_CAP_BUTT;
+        switch(cap_style) {
+        case VG_CAP_BUTT:
+            fprintf(stderr, "Cap style 'butt'\n");
+            break;
+        case VG_CAP_ROUND:
+            fprintf(stderr, "Cap style 'round'\n");
+            break;
+        case VG_CAP_SQUARE:
+            fprintf(stderr, "Cap style 'square'\n");
+            break;
+        }
+        vgSeti(VG_STROKE_CAP_STYLE, cap_style);
+        break;
+    case XK_j:
+    case XK_J:
+        ++join_style;
+        if (join_style > VG_JOIN_BEVEL)
+            join_style = VG_JOIN_MITER;
+        switch(join_style) {
+        case VG_JOIN_MITER:
+            fprintf(stderr, "Join style 'miter'\n");
+            break;
+        case VG_JOIN_ROUND:
+            fprintf(stderr, "Join style 'round'\n");
+            break;
+        case VG_JOIN_BEVEL:
+            fprintf(stderr, "Join style 'bevel'\n");
+            break;
+        }
+        vgSeti(VG_STROKE_JOIN_STYLE, join_style);
+        break;
+    default:
+        break;
+    }
+
+    return VG_TRUE;
+}
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, key_press);
+}
diff --git a/progs/openvg/trivial/stroke2.c b/progs/openvg/trivial/stroke2.c
new file mode 100644
index 0000000000..ce950c1886
--- /dev/null
+++ b/progs/openvg/trivial/stroke2.c
@@ -0,0 +1,207 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <X11/keysym.h>
+#include <stdio.h>
+
+VGPaint stroke;
+VGPath target;
+VGPath lines;
+
+VGfloat xform[9];
+VGfloat color1[4];
+VGfloat color2[4];
+VGfloat bgCol[4];
+
+static void
+init(void)
+{
+    VGubyte lineCmds[6];
+    VGfloat lineCoords[8];
+    VGfloat arcCoords[5];
+    VGubyte sccCmd[1];
+    VGubyte scCmd[1];
+    VGubyte lccCmd[1];
+    VGubyte lcCmd[1];
+    VGubyte moveCmd[1];
+    VGfloat moveCoords[2];
+    VGint i;
+
+    bgCol[0] = 1.0f;
+    bgCol[1] = 1.0f;
+    bgCol[2] = 1.0f;
+    bgCol[3] = 1.0f;
+
+    vgSetfv(VG_CLEAR_COLOR, 4, bgCol);
+    vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+
+    stroke = vgCreatePaint();
+    /* Red */
+    color1[0] = 1.0f;
+    color1[1] = 0.0f;
+    color1[2] = 0.0f;
+    color1[3] = 1.0f;
+
+    /* Orange */
+    color2[0] = 1.0000f;
+    color2[1] = 1.0f;
+    color2[2] = 0.0f;
+    color2[3] = 1.0f;
+    vgSetPaint(stroke, VG_STROKE_PATH);
+
+    vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_SQUARE);
+
+    {
+        VGfloat temp[9] = {0, 0, 0, 0, 0, 0, 0, 0, 0};
+        for (i = 0; i < 9; i++)
+        {
+            xform[i] = temp[i];
+        }
+    }
+    vgGetMatrix(xform);
+
+    target = vgCreatePath(VG_PATH_FORMAT_STANDARD,
+                          VG_PATH_DATATYPE_F, 1, 0, 0, 0, VG_PATH_CAPABILITY_TRANSFORM_TO);
+
+#if 0
+    /* Line path */
+    {
+        VGubyte temp[6] = {VG_MOVE_TO_ABS, VG_VLINE_TO_REL,
+                           VG_MOVE_TO_ABS, VG_VLINE_TO_REL,
+                           VG_HLINE_TO_REL, VG_VLINE_TO_REL};
+        for (i = 0; i < 6; i++)
+        {
+            lineCmds[i] = temp[i];
+        }
+    }
+    {
+        VGfloat temp[8] = {0.5f, 0.8f, -0.6f, 0.28f, 0.6f, -0.4f, 0.44f, 0.4f};
+        for (i = 0; i < 8; i++)
+        {
+            lineCoords[i] = temp[i] * window_width();
+        }
+    }
+#else
+    {
+        VGfloat temp[5] = {0.35f, 0.15f, 29, 0.3f, 0.4f};
+        for (i = 0; i < 5; i++)
+        {
+            arcCoords[i] = temp[i] * window_width();
+        }
+        arcCoords[2] = 29;
+    }
+
+    {
+        VGubyte temp[1] = {VG_SCCWARC_TO_ABS};
+        for (i = 0; i < 1; i++)
+        {
+            sccCmd[i] = temp[i];
+        }
+    }
+    {
+        VGubyte temp[1] = {VG_SCWARC_TO_ABS};
+        for (i = 0; i < 1; i++)
+        {
+            scCmd[i] = temp[i];
+        }
+    }
+    {
+        VGubyte temp[1] = {VG_LCCWARC_TO_ABS};
+        for (i = 0; i < 1; i++)
+        {
+            lccCmd[i] = temp[i];
+        }
+    }
+    {
+        VGubyte temp[1] = {VG_LCWARC_TO_ABS};
+        for (i = 0; i < 1; i++)
+        {
+            lcCmd[i] = temp[i];
+        }
+    }
+
+    {
+        VGubyte temp[1] = {VG_MOVE_TO_ABS};
+        for (i = 0; i < 1; i++)
+        {
+            moveCmd[i] = temp[i];
+        }
+    }
+    {
+        VGfloat temp[2] = {0.7f, 0.6f};
+        for (i = 0; i < 2; i++)
+        {
+            moveCoords[i] = temp[i] * window_width();
+        }
+    }
+#endif
+
+    lines = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F, 1,
+                         0, 0, 0,
+                         VG_PATH_CAPABILITY_APPEND_TO|
+                         VG_PATH_CAPABILITY_TRANSFORM_FROM);
+#if 0
+    vgAppendPathData(lines, 6, lineCmds, lineCoords);
+#else
+    vgAppendPathData(lines, 1, moveCmd, moveCoords);
+    vgAppendPathData(lines, 1, sccCmd, arcCoords);
+    vgAppendPathData(lines, 1, moveCmd, moveCoords);
+    vgAppendPathData(lines, 1, scCmd, arcCoords);
+    vgAppendPathData(lines, 1, moveCmd, moveCoords);
+    vgAppendPathData(lines, 1, lccCmd, arcCoords);
+    vgAppendPathData(lines, 1, moveCmd, moveCoords);
+    vgAppendPathData(lines, 1, lcCmd, arcCoords);
+#endif
+
+    vgLoadIdentity();
+    vgTranslate(0.25f * window_width(), 0.25f * window_height());
+    vgRotate(30);
+    vgTranslate(-0.25f * window_width(), -0.25f * window_height());
+    vgTransformPath(target, lines);}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+}
+
+static void
+draw(void)
+{
+    vgClear(0, 0, window_width(), window_height());
+    vgLoadMatrix(xform);
+    vgLoadIdentity();
+    vgTranslate(0.25f * window_width(), 0.25f * window_height());
+    vgRotate(30);
+    vgTranslate(-0.25f * window_width(), -0.25f * window_height());
+    vgSetf(VG_STROKE_LINE_WIDTH, 7);
+    vgSetParameterfv(stroke, VG_PAINT_COLOR, 4, color1);
+    vgDrawPath(lines, VG_STROKE_PATH);
+
+    vgLoadMatrix(xform);
+    vgSetParameterfv(stroke, VG_PAINT_COLOR, 4, color2);
+    vgSetf(VG_STROKE_LINE_WIDTH, 3);
+    vgDrawPath(target, VG_STROKE_PATH);
+}
+
+static int  key_press(unsigned key)
+{
+    switch(key) {
+    case XK_c:
+    case XK_C:
+        break;
+    case XK_j:
+    case XK_J:
+        break;
+    default:
+        break;
+    }
+
+    return VG_TRUE;
+}
+
+int main(int argc, char **argv)
+{
+   return run(argc, argv, init, reshape,
+              draw, key_press);
+}
diff --git a/progs/openvg/trivial/vguarc.c b/progs/openvg/trivial/vguarc.c
new file mode 100644
index 0000000000..8d971d5c09
--- /dev/null
+++ b/progs/openvg/trivial/vguarc.c
@@ -0,0 +1,74 @@
+#include "eglcommon.h"
+
+#include <VG/openvg.h>
+#include <VG/vgu.h>
+
+const VGfloat white_color[4] = {1.0, 1.0, 1.0, 1.0};
+const VGfloat color[4] = {0.4, 0.1, 1.0, 1.0};
+
+VGPath path;
+VGPaint paint;
+
+
+static void
+init(void)
+{
+    VGfloat clearColor[] = {0.0f, 0.0f, 0.0f, 1.0f};/* black color */
+    VGfloat greenColor[] = {0.0f, 1.0f, 0.0f, 1.0f};/* green color */
+    VGint arcType = VGU_ARC_OPEN;
+    VGfloat x, y, w, h, startAngle, angleExtent;
+
+    x = 150;
+    y = 150;
+    w = 150;
+    h = 150;
+#if 0
+    startAngle  = -540.0f;
+    angleExtent = 270.0f;
+#else
+    startAngle  = 270.0f;
+    angleExtent = 90.0f;
+#endif
+
+    paint = vgCreatePaint();
+
+    vgSetPaint(paint, VG_STROKE_PATH);
+    vgSetParameterfv(paint, VG_PAINT_COLOR, 4, greenColor);
+    vgSetParameteri( paint, VG_PAINT_TYPE, VG_PAINT_TYPE_COLOR);
+    vgSetf(VG_STROKE_LINE_WIDTH, 6.0f);
+    vgSeti(VG_RENDERING_QUALITY, VG_RENDERING_QUALITY_NONANTIALIASED);
+    vgSetfv(VG_CLEAR_COLOR, 4, clearColor);
+
+    path  = vgCreatePath(VG_PATH_FORMAT_STANDARD, VG_PATH_DATATYPE_F,
+                         1.0f, 0.0f, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+    vguArc(path, x, y, w, h, startAngle, angleExtent, arcType);
+
+    vgSeti(VG_STROKE_CAP_STYLE, VG_CAP_BUTT);
+    vgSeti(VG_STROKE_JOIN_STYLE, VG_JOIN_BEVEL);
+    vgSetf(VG_STROKE_MITER_LIMIT, 4.0f);
+}
+
+/* new window size or exposure */
+static void
+reshape(int w, int h)
+{
+   vgLoadIdentity();
+}
+
+static void
+draw(void)
+{
+   vgClear(0, 0, window_width(), window_height());
+   vgDrawPath(path, VG_STROKE_PATH);
+
+   vgFlush();
+}
+
+
+int main(int argc, char **argv)
+{
+    // set_window_size(64, 63);
+   return run(argc, argv, init, reshape,
+              draw, 0);
+}
diff --git a/progs/rbug/.gitignore b/progs/rbug/.gitignore
new file mode 100644
index 0000000000..174fe42aa7
--- /dev/null
+++ b/progs/rbug/.gitignore
@@ -0,0 +1,12 @@
+bin_to_bmp
+simple_client
+simple_server
+shdr_info
+shdr_dump
+shdr_disable
+ctx_info
+ctx_rule
+tex_dump
+tex_info
+*.bmp
+*.bin
diff --git a/progs/rbug/Makefile b/progs/rbug/Makefile
new file mode 100644
index 0000000000..fb4e91d1af
--- /dev/null
+++ b/progs/rbug/Makefile
@@ -0,0 +1,48 @@
+# progs/rbug/Makefile
+
+TOP = ../..
+include $(TOP)/configs/current
+
+INCLUDES = \
+	-I. \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	$(PROG_INCLUDES)
+
+LINKS = \
+	$(GALLIUM_AUXILIARIES) \
+	$(PROG_LINKS)
+
+SOURCES = \
+	bin_to_bmp.c \
+	simple_client.c \
+	simple_server.c \
+	shdr_info.c \
+	shdr_dump.c \
+	shdr_disable.c \
+	ctx_info.c \
+	ctx_rule.c \
+	tex_info.c \
+	tex_dump.c
+
+
+OBJECTS = $(SOURCES:.c=.o)
+
+PROGS = $(OBJECTS:.o=)
+
+##### TARGETS #####
+
+default: $(OBJECTS) $(PROGS)
+
+clean:
+	-rm -f $(PROGS)
+	-rm -f *.o
+
+##### RULES #####
+
+$(OBJECTS): %.o: %.c
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DEFINES) $(PROG_DEFINES) $< -o $@
+
+$(PROGS): %: %.o
+	$(CC) $(LDFLAGS) $< $(LINKS) -o $@
diff --git a/progs/rbug/README b/progs/rbug/README
new file mode 100644
index 0000000000..0eb0a5de9a
--- /dev/null
+++ b/progs/rbug/README
@@ -0,0 +1,39 @@
+                        REMOTE DEBUGGING CLI APPLICATIONS
+
+
+= About =
+
+This directory contains a Gallium3D remote debugging cli applications.
+
+
+= Build Instructions =
+
+To build, build a normal gallium build and from this directory do the following.
+
+ make
+
+= Usage =
+
+Make sure that you driver has trace integration, see
+src/gallium/driver/trace/README for more information about that. Then from on
+the computer that you want to debug do:
+
+ export GALLIUM_RBUG=true
+
+ <launch app>
+
+From the debugging computer launch apps form this directory. Currently ip
+addresses are hardcoded and you need to edit the application, but that will
+change in the future.
+
+= Testing =
+
+The two apps simple_client and simple_server. Are unit testing of the
+connection and (de)marsheler. Just run the server first and then the client:
+
+ ./simple_server &
+ ./simple_client
+
+
+--
+Jakob Bornecrantz <jakob@vmware.com>
diff --git a/progs/rbug/bin_to_bmp.c b/progs/rbug/bin_to_bmp.c
new file mode 100644
index 0000000000..cdae3486ce
--- /dev/null
+++ b/progs/rbug/bin_to_bmp.c
@@ -0,0 +1,110 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+#include "util/u_tile.h"
+
+static uint8_t* read(const char *filename, unsigned size);
+static void dump(unsigned src_width, unsigned src_height,
+                 unsigned src_stride, enum pipe_format src_format,
+                 uint8_t *data, unsigned src_size);
+
+int main(int argc, char** argv)
+{
+   /* change these */
+   unsigned width = 64;
+   unsigned height = 64;
+   unsigned stride = width * 4;
+   unsigned size = stride * height;
+   const char *filename = "mybin.bin";
+   enum pipe_format format = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+   dump(width, height, stride, format, read(filename, size), size);
+
+   return 0;
+}
+
+static void dump(unsigned width, unsigned height,
+                 unsigned src_stride, enum pipe_format src_format,
+                 uint8_t *data, unsigned src_size)
+{
+   struct pipe_format_block src_block;
+
+   enum pipe_format dst_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+   struct pipe_format_block dst_block;
+   unsigned dst_stride;
+   unsigned dst_size;
+   float *rgba;
+   int i;
+   char filename[512];
+
+   {
+      pf_get_block(src_format, &src_block);
+      assert(src_stride >= pf_get_stride(&src_block, width));
+      assert(src_size >= pf_get_2d_size(&src_block, src_stride, width));
+   }
+   {
+      pf_get_block(dst_format, &dst_block);
+      dst_stride = pf_get_stride(&dst_block, width);
+      dst_size = pf_get_2d_size(&dst_block, dst_stride, width);
+      rgba = MALLOC(dst_size);
+   }
+
+   util_snprintf(filename, 512, "%s.bmp", pf_name(src_format));
+
+   if (pf_is_compressed(src_format)) {
+      debug_printf("skipping: %s\n", filename);
+      return;
+   }
+
+   debug_printf("saving: %s\n", filename);
+
+   for (i = 0; i < height; i++) {
+      pipe_tile_raw_to_rgba(src_format, data + src_stride * i,
+                            width, 1,
+                            &rgba[width*4*i], dst_stride);
+   }
+
+   debug_dump_float_rgba_bmp(filename, width, height, rgba, width);
+
+   FREE(rgba);
+}
+
+static uint8_t* read(const char *filename, unsigned size)
+{
+   uint8_t *data;
+   FILE *file = fopen(filename, "rb");
+
+   data = MALLOC(size);
+
+   fread(data, 1, size, file);
+   fclose(file);
+
+   return data;
+}
diff --git a/progs/rbug/ctx_info.c b/progs/rbug/ctx_info.c
new file mode 100644
index 0000000000..d72c326719
--- /dev/null
+++ b/progs/rbug/ctx_info.c
@@ -0,0 +1,80 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void talk()
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   struct rbug_proto_context_list_reply *list;
+   struct rbug_proto_context_info_reply *info;
+   int i;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Sending get contexts\n");
+   rbug_send_context_list(con, NULL);
+
+   debug_printf("Waiting for contexts\n");
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_CONTEXT_LIST_REPLY);
+   list = (struct rbug_proto_context_list_reply *)header;
+
+   debug_printf("Got contexts:\n");
+   for (i = 0; i < list->contexts_len; i++) {
+#if 0
+      rbug_send_contexts_info(con, list->contexts[i], NULL);
+
+      header = rbug_get_message(con, NULL);
+      assert(header->opcode == RBUG_OP_CONTEXT_INFO_REPLY);
+      info = (struct rbug_proto_context_info_reply *)header;
+#else
+      (void)info;
+      header = NULL;
+#endif
+
+      debug_printf("%llu\n",
+                   (unsigned long long)list->contexts[i]);
+      rbug_free_header(header);
+   }
+
+   rbug_free_header(&list->header);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   talk();
+   return 0;
+}
diff --git a/progs/rbug/ctx_rule.c b/progs/rbug/ctx_rule.c
new file mode 100644
index 0000000000..e38b7b4e9b
--- /dev/null
+++ b/progs/rbug/ctx_rule.c
@@ -0,0 +1,86 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void talk(rbug_context_t ctx, rbug_shader_t shdr)
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con;
+   struct rbug_header *header;
+
+   if (c < 0)
+      c = u_socket_connect("localhost", 13370);
+
+   con = rbug_from_socket(c);
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   rbug_send_context_draw_rule(con, ctx, 0, shdr, 0, 0, RBUG_BLOCK_AFTER, NULL);
+
+   rbug_send_ping(con, NULL);
+
+   debug_printf("Sent waiting for reply\n");
+   header = rbug_get_message(con, NULL);
+
+   if (header->opcode != RBUG_OP_PING_REPLY)
+      debug_printf("Error\n");
+   else
+      debug_printf("Ok!\n");
+
+   rbug_free_header(header);
+   rbug_disconnect(con);
+}
+
+static void print_usage()
+{
+   printf("Usage ctx_rule <context> <fragment>\n");
+   exit(-1);
+}
+
+int main(int argc, char** argv)
+{
+   long ctx;
+   long shdr;
+
+   if (argc < 3)
+      print_usage();
+
+   ctx = atol(argv[1]);
+   shdr = atol(argv[2]);
+
+   if (ctx <= 0 && ctx <= 0)
+      print_usage();
+
+   talk((uint64_t)ctx, (uint64_t)shdr);
+
+   return 0;
+}
diff --git a/progs/rbug/shdr_disable.c b/progs/rbug/shdr_disable.c
new file mode 100644
index 0000000000..e6b12073d8
--- /dev/null
+++ b/progs/rbug/shdr_disable.c
@@ -0,0 +1,82 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void talk(rbug_context_t ctx, rbug_shader_t shdr)
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   rbug_send_shader_disable(con, ctx, shdr, true, NULL);
+
+   rbug_send_ping(con, NULL);
+
+   debug_printf("Sent waiting for reply\n");
+   header = rbug_get_message(con, NULL);
+
+   if (header->opcode != RBUG_OP_PING_REPLY)
+      debug_printf("Error\n");
+   else
+      debug_printf("Ok!\n");
+
+   rbug_free_header(header);
+   rbug_disconnect(con);
+}
+
+static void print_usage()
+{
+   printf("Usage shdr_disable <context> <shader>\n");
+   exit(-1);
+}
+
+int main(int argc, char** argv)
+{
+   long ctx;
+   long shdr;
+
+   if (argc < 3)
+      print_usage();
+
+   ctx = atol(argv[1]);
+   shdr = atol(argv[2]);
+
+   if (ctx <= 0 && ctx <= 0)
+      print_usage();
+
+   talk((uint64_t)ctx, (uint64_t)shdr);
+
+   return 0;
+}
diff --git a/progs/rbug/shdr_dump.c b/progs/rbug/shdr_dump.c
new file mode 100644
index 0000000000..8f9d758d51
--- /dev/null
+++ b/progs/rbug/shdr_dump.c
@@ -0,0 +1,115 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+#include "tgsi/tgsi_dump.h"
+
+static void shader_info(struct rbug_connection *con, rbug_context_t ctx)
+{
+   struct rbug_header *header;
+   struct rbug_proto_shader_list_reply *list;
+   struct rbug_proto_shader_info_reply *info;
+   int i;
+
+   debug_printf("Sending get shaders to %llu\n", (unsigned long long)ctx);
+   rbug_send_shader_list(con, ctx, NULL);
+
+   debug_printf("Waiting for shaders from %llu\n", (unsigned long long)ctx);
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_SHADER_LIST_REPLY);
+   list = (struct rbug_proto_shader_list_reply *)header;
+
+   debug_printf("Got shaders:\n");
+   for (i = 0; i < list->shaders_len; i++) {
+      rbug_send_shader_info(con, ctx, list->shaders[i], NULL);
+
+      header = rbug_get_message(con, NULL);
+      assert(header->opcode == RBUG_OP_SHADER_INFO_REPLY);
+      info = (struct rbug_proto_shader_info_reply *)header;
+
+      debug_printf("#####################################################\n");
+      debug_printf("ctx: %llu shdr: %llu disabled %u\n",
+                   (unsigned long long)ctx,
+                   (unsigned long long)list->shaders[i],
+                   info->disabled);
+
+      /* just to be sure */
+      assert(sizeof(struct tgsi_token) == 4);
+
+      debug_printf("-----------------------------------------------------\n");
+      tgsi_dump((struct tgsi_token *)info->original, 0);
+
+      if (info->replaced_len > 0) {
+         debug_printf("-----------------------------------------------------\n");
+         tgsi_dump((struct tgsi_token *)info->replaced, 0);
+      }
+
+      rbug_free_header(header);
+   }
+
+   debug_printf("#####################################################\n");
+   rbug_free_header(&list->header);
+}
+
+static void talk()
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   struct rbug_proto_context_list_reply *list;
+   int i;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Sending get contexts\n");
+   rbug_send_context_list(con, NULL);
+
+   debug_printf("Waiting for contexts\n");
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_CONTEXT_LIST_REPLY);
+   list = (struct rbug_proto_context_list_reply *)header;
+
+   debug_printf("Got contexts:\n");
+   for (i = 0; i < list->contexts_len; i++) {
+      shader_info(con, list->contexts[i]);
+   }
+
+   rbug_free_header(&list->header);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   talk();
+   return 0;
+}
diff --git a/progs/rbug/shdr_info.c b/progs/rbug/shdr_info.c
new file mode 100644
index 0000000000..b6864e988e
--- /dev/null
+++ b/progs/rbug/shdr_info.c
@@ -0,0 +1,98 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void shader_info(struct rbug_connection *con, rbug_context_t ctx)
+{
+   struct rbug_header *header;
+   struct rbug_proto_shader_list_reply *list;
+   struct rbug_proto_shader_info_reply *info;
+   int i;
+
+   rbug_send_shader_list(con, ctx, NULL);
+
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_SHADER_LIST_REPLY);
+   list = (struct rbug_proto_shader_list_reply *)header;
+
+   debug_printf("        context |         shader |       disabled |\n");
+   for (i = 0; i < list->shaders_len; i++) {
+      rbug_send_shader_info(con, ctx, list->shaders[i], NULL);
+
+      header = rbug_get_message(con, NULL);
+      assert(header->opcode == RBUG_OP_SHADER_INFO_REPLY);
+      info = (struct rbug_proto_shader_info_reply *)header;
+
+      debug_printf("% 15llu |% 15llu |% 15u |\n",
+                   (unsigned long long)ctx,
+                   (unsigned long long)list->shaders[i],
+                   (unsigned)info->disabled);
+
+      rbug_free_header(header);
+   }
+
+   rbug_free_header(&list->header);
+}
+
+static void talk()
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   struct rbug_proto_context_list_reply *list;
+   int i;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Sending get contexts\n");
+   rbug_send_context_list(con, NULL);
+
+   debug_printf("Waiting for contexts\n");
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_CONTEXT_LIST_REPLY);
+   list = (struct rbug_proto_context_list_reply *)header;
+
+   debug_printf("Got contexts:\n");
+   for (i = 0; i < list->contexts_len; i++) {
+      shader_info(con, list->contexts[i]);
+   }
+
+   rbug_free_header(&list->header);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   talk();
+   return 0;
+}
diff --git a/progs/rbug/simple_client.c b/progs/rbug/simple_client.c
new file mode 100644
index 0000000000..38929fa796
--- /dev/null
+++ b/progs/rbug/simple_client.c
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void talk()
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   struct rbug_proto_texture_list_reply *list;
+   int i;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Sending get textures\n");
+   rbug_send_texture_list(con, NULL);
+
+   debug_printf("Waiting for textures\n");
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_TEXTURE_LIST_REPLY);
+   list = (struct rbug_proto_texture_list_reply *)header;
+
+   debug_printf("Got textures:\n");
+   for (i = 0; i < list->textures_len; i++)
+      debug_printf("\ttex %llu\n", (unsigned long long)list->textures[i]);
+
+   rbug_free_header(header);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   talk();
+   return 0;
+}
diff --git a/progs/rbug/simple_server.c b/progs/rbug/simple_server.c
new file mode 100644
index 0000000000..04380c3310
--- /dev/null
+++ b/progs/rbug/simple_server.c
@@ -0,0 +1,62 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void wait()
+{
+   int s = u_socket_listen_on_port(13370);
+   int c = u_socket_accept(s);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   rbug_texture_t texs[2];
+   uint32_t serial;
+   texs[0] = 1337;
+   texs[1] = 7331;
+
+   assert(s >= 0);
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Waiting for get textures\n");
+   header = rbug_get_message(con, &serial);
+   assert(header);
+   assert(header->opcode == RBUG_OP_TEXTURE_LIST);
+   rbug_free_header(header);
+
+   rbug_send_texture_list_reply(con, serial, texs, 2, NULL);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   wait();
+   return 0;
+}
diff --git a/progs/rbug/tex_dump.c b/progs/rbug/tex_dump.c
new file mode 100644
index 0000000000..f9e06ee994
--- /dev/null
+++ b/progs/rbug/tex_dump.c
@@ -0,0 +1,127 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+#include "util/u_tile.h"
+#include "rbug/rbug.h"
+
+static void dump(rbug_texture_t tex,
+                 struct rbug_proto_texture_info_reply *info,
+                 struct rbug_proto_texture_read_reply *read,
+                 int mip)
+{
+   enum pipe_format format = info->format;
+   uint8_t *data = read->data;
+   unsigned width = info->width[mip];
+   unsigned height = info->height[mip];
+   unsigned dst_stride = width * 4 * 4;
+   unsigned src_stride = read->stride;
+   float *rgba = MALLOC(dst_stride * height);
+   int i;
+   char filename[512];
+
+   util_snprintf(filename, 512, "%llu_%s_%u.bmp",
+                 (unsigned long long)tex, pf_name(info->format), mip);
+
+   if (pf_is_compressed(info->format)) {
+      debug_printf("skipping: %s\n", filename);
+      return;
+   }
+
+   debug_printf("saving: %s\n", filename);
+
+   for (i = 0; i < height; i++) {
+      pipe_tile_raw_to_rgba(format, data + src_stride * i,
+                            width, 1,
+                            &rgba[width*4*i], dst_stride);
+   }
+
+   debug_dump_float_rgba_bmp(filename, width, height, rgba, width);
+
+   FREE(rgba);
+}
+
+static void talk()
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   struct rbug_header *header2;
+   struct rbug_proto_texture_list_reply *list;
+   struct rbug_proto_texture_info_reply *info;
+   struct rbug_proto_texture_read_reply *read;
+   int i, j;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Sending get textures\n");
+   rbug_send_texture_list(con, NULL);
+
+   debug_printf("Waiting for textures\n");
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_TEXTURE_LIST_REPLY);
+   list = (struct rbug_proto_texture_list_reply *)header;
+
+   debug_printf("Got textures:\n");
+   for (i = 0; i < list->textures_len; i++) {
+      rbug_send_texture_info(con, list->textures[i], NULL);
+
+      header = rbug_get_message(con, NULL);
+      assert(header->opcode == RBUG_OP_TEXTURE_INFO_REPLY);
+      info = (struct rbug_proto_texture_info_reply *)header;
+
+      for (j = 0; j <= info->last_level; j++) {
+         rbug_send_texture_read(con, list->textures[i],
+                                0, j, 0,
+                                0, 0, info->width[j], info->height[j],
+                                NULL);
+
+         header2 = rbug_get_message(con, NULL);
+         assert(header2->opcode == RBUG_OP_TEXTURE_READ_REPLY);
+         read = (struct rbug_proto_texture_read_reply *)header2;
+
+         dump(list->textures[i], info, read, j);
+
+         rbug_free_header(header2);
+      }
+
+      rbug_free_header(header);
+
+   }
+   rbug_free_header(&list->header);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   talk();
+   return 0;
+}
diff --git a/progs/rbug/tex_info.c b/progs/rbug/tex_info.c
new file mode 100644
index 0000000000..4a21bae359
--- /dev/null
+++ b/progs/rbug/tex_info.c
@@ -0,0 +1,78 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include "util/u_network.h"
+
+#include "rbug/rbug.h"
+
+static void talk()
+{
+   int c = u_socket_connect("localhost", 13370);
+   struct rbug_connection *con = rbug_from_socket(c);
+   struct rbug_header *header;
+   struct rbug_proto_texture_list_reply *list;
+   struct rbug_proto_texture_info_reply *info;
+   int i;
+
+   assert(c >= 0);
+   assert(con);
+   debug_printf("Connection get!\n");
+
+   debug_printf("Sending get textures\n");
+   rbug_send_texture_list(con, NULL);
+
+   debug_printf("Waiting for textures\n");
+   header = rbug_get_message(con, NULL);
+   assert(header->opcode == RBUG_OP_TEXTURE_LIST_REPLY);
+   list = (struct rbug_proto_texture_list_reply *)header;
+
+   debug_printf("Got textures:\n");
+   for (i = 0; i < list->textures_len; i++) {
+      rbug_send_texture_info(con, list->textures[i], NULL);
+
+      header = rbug_get_message(con, NULL);
+      assert(header->opcode == RBUG_OP_TEXTURE_INFO_REPLY);
+      info = (struct rbug_proto_texture_info_reply *)header;
+
+      debug_printf("%llu %s %u x %u x %u, block(%ux%u %u), last_level: %u, nr_samples: %u, usage: %u\n",
+                   (unsigned long long)list->textures[i], pf_name(info->format),
+                   info->width[0], info->height[0], info->depth[0],
+                   info->blockw, info->blockh, info->blocksize,
+                   info->last_level, info->nr_samples, info->tex_usage);
+      rbug_free_header(header);
+   }
+
+   rbug_free_header(&list->header);
+   rbug_disconnect(con);
+}
+
+int main(int argc, char** argv)
+{
+   talk();
+   return 0;
+}
diff --git a/progs/samples/prim.c b/progs/samples/prim.c
index f47c60faef..c04750725f 100644
--- a/progs/samples/prim.c
+++ b/progs/samples/prim.c
@@ -466,25 +466,22 @@ static void Draw(void)
     } else {
 	glPolygonMode(GL_FRONT_AND_BACK, GL_LINE);
     }
-#if 01
+
     Viewport(0, 0); Point();
     Viewport(0, 1); Lines();
     Viewport(0, 2); LineStrip();
     Viewport(0, 3); LineLoop();
 
     Viewport(1, 0); Bitmap();
-
     Viewport(1, 1); TriangleFan();
     Viewport(1, 2); Triangles();
     Viewport(1, 3); TriangleStrip();
 
     Viewport(2, 0); Rect();
-#endif
     Viewport(2, 1); PolygonFunc();
-#if 01
     Viewport(2, 2); Quads();
     Viewport(2, 3); QuadStrip();
-#endif
+
     glFlush();
 
     if (doubleBuffer) {
diff --git a/progs/slang/vstest.c b/progs/slang/vstest.c
index 5108d15742..472ea9b62e 100644
--- a/progs/slang/vstest.c
+++ b/progs/slang/vstest.c
@@ -128,29 +128,29 @@ static void va_render ()
       {
       case C:
          glColorPointer (4, GL_FLOAT, 0, att->data);
-         glEnable (GL_COLOR_ARRAY);
+         glEnableClientState (GL_COLOR_ARRAY);
          break;
       case S:
          glSecondaryColorPointerEXT (4, GL_FLOAT, 0, att->data);
-         glEnable (GL_SECONDARY_COLOR_ARRAY_EXT);
+         glEnableClientState (GL_SECONDARY_COLOR_ARRAY_EXT);
          break;
       case N:
          glNormalPointer (GL_FLOAT, 0, att->data);
-         glEnable (GL_NORMAL_ARRAY);
+         glEnableClientState (GL_NORMAL_ARRAY);
          break;
       case V:
          glVertexPointer (4, GL_FLOAT, 0, att->data);
-         glEnable (GL_VERTEX_ARRAY);
+         glEnableClientState (GL_VERTEX_ARRAY);
          break;
       case T:
          assert (att->index >= 0 && att->index < 8);
          glClientActiveTextureARB (GL_TEXTURE0_ARB + att->index);
          glTexCoordPointer (4, GL_FLOAT, 0, att->data);
-         glEnable (GL_TEXTURE_COORD_ARRAY);
+         glEnableClientState (GL_TEXTURE_COORD_ARRAY);
          break;
       case F:
          glFogCoordPointerEXT (GL_FLOAT, 0, att->data);
-         glEnable (GL_FOG_COORDINATE_ARRAY_EXT);
+         glEnableClientState (GL_FOG_COORDINATE_ARRAY_EXT);
          break;
       case A:
          assert (att->index > 0 && att->index < 16);
@@ -169,23 +169,23 @@ static void va_render ()
       switch (att->dispatch)
       {
       case C:
-         glDisable (GL_COLOR_ARRAY);
+         glDisableClientState (GL_COLOR_ARRAY);
          break;
       case S:
-         glDisable (GL_SECONDARY_COLOR_ARRAY_EXT);
+         glDisableClientState (GL_SECONDARY_COLOR_ARRAY_EXT);
          break;
       case N:
-         glDisable (GL_NORMAL_ARRAY);
+         glDisableClientState (GL_NORMAL_ARRAY);
          break;
       case V:
-         glDisable (GL_VERTEX_ARRAY);
+         glDisableClientState (GL_VERTEX_ARRAY);
          break;
       case T:
          glClientActiveTextureARB (GL_TEXTURE0_ARB + att->index);
-         glDisable (GL_TEXTURE_COORD_ARRAY);
+         glDisableClientState (GL_TEXTURE_COORD_ARRAY);
          break;
       case F:
-         glDisable (GL_FOG_COORDINATE_ARRAY_EXT);
+         glDisableClientState (GL_FOG_COORDINATE_ARRAY_EXT);
          break;
       case A:
          glDisableVertexAttribArrayARB (att->index);
diff --git a/progs/tests/Makefile b/progs/tests/Makefile
index b58d7b25e4..4d9b4e8388 100644
--- a/progs/tests/Makefile
+++ b/progs/tests/Makefile
@@ -17,6 +17,7 @@ SOURCES = \
 	arbfptest1.c \
 	arbfptexture.c \
 	arbfptrig.c \
+	arbgpuprog.c \
 	arbnpot.c \
 	arbnpot-mipmap.c \
 	arbvptest1.c \
@@ -48,7 +49,8 @@ SOURCES = \
 	fptest1.c \
 	fptexture.c \
 	getprocaddress.c \
-	glutfx \
+	getteximage.c \
+	glutfx.c \
 	interleave.c \
 	invert.c \
 	jkrahntest.c \
@@ -66,6 +68,8 @@ SOURCES = \
 	no_s3tc.c \
 	packedpixels.c \
 	pbo.c \
+	persp_hint.c \
+	prim.c \
 	prog_parameter.c \
 	quads.c \
 	random.c \
diff --git a/progs/tests/SConscript b/progs/tests/SConscript
index a4b5a45573..b17fa90593 100644
--- a/progs/tests/SConscript
+++ b/progs/tests/SConscript
@@ -91,6 +91,7 @@ progs = [
     'no_s3tc',
     'packedpixels',
     'pbo',
+    'persp_hint',
     'prog_parameter',
     'quads',
     'random',
diff --git a/progs/tests/arbgpuprog.c b/progs/tests/arbgpuprog.c
new file mode 100644
index 0000000000..23aa899d96
--- /dev/null
+++ b/progs/tests/arbgpuprog.c
@@ -0,0 +1,230 @@
+/**
+ * Just compile ARB vert/frag program from named file(s).
+ */
+
+#include <assert.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <GL/glut.h>
+
+
+static GLuint FragProg;
+static GLuint VertProg;
+static GLint Win;
+
+static PFNGLPROGRAMLOCALPARAMETER4FVARBPROC glProgramLocalParameter4fvARB_func;
+static PFNGLPROGRAMLOCALPARAMETER4DARBPROC glProgramLocalParameter4dARB_func;
+static PFNGLGETPROGRAMLOCALPARAMETERDVARBPROC glGetProgramLocalParameterdvARB_func;
+static PFNGLGENPROGRAMSARBPROC glGenProgramsARB_func;
+static PFNGLPROGRAMSTRINGARBPROC glProgramStringARB_func;
+static PFNGLBINDPROGRAMARBPROC glBindProgramARB_func;
+static PFNGLISPROGRAMARBPROC glIsProgramARB_func;
+static PFNGLDELETEPROGRAMSARBPROC glDeleteProgramsARB_func;
+
+
+static void Redisplay( void )
+{
+   glClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );
+   glutSwapBuffers();
+   exit(0);
+}
+
+
+static void Reshape( int width, int height )
+{
+   glViewport( 0, 0, width, height );
+   glMatrixMode( GL_PROJECTION );
+   glLoadIdentity();
+   glFrustum( -1.0, 1.0, -1.0, 1.0, 5.0, 25.0 );
+   glMatrixMode( GL_MODELVIEW );
+   glLoadIdentity();
+   glTranslatef( 0.0, 0.0, -15.0 );
+}
+
+
+static void Key( unsigned char key, int x, int y )
+{
+   (void) x;
+   (void) y;
+   switch (key) {
+      case 27:
+         glDeleteProgramsARB_func(1, &VertProg);
+         glDeleteProgramsARB_func(1, &FragProg);
+         glutDestroyWindow(Win);
+         exit(0);
+         break;
+   }
+   glutPostRedisplay();
+}
+
+
+/* A helper for finding errors in program strings */
+static int FindLine( const char *program, int position )
+{
+   int i, line = 1;
+   for (i = 0; i < position; i++) {
+      if (program[i] == '\n')
+         line++;
+   }
+   return line;
+}
+
+
+static void Init( const char *vertProgFile,
+                  const char *fragProgFile )
+{
+   GLint errorPos;
+   char buf[10*1000];
+
+   if (!glutExtensionSupported("GL_ARB_vertex_program")) {
+      printf("Sorry, this demo requires GL_ARB_vertex_program\n");
+      exit(1);
+   }
+   if (!glutExtensionSupported("GL_ARB_fragment_program")) {
+      printf("Sorry, this demo requires GL_ARB_fragment_program\n");
+      exit(1);
+   }
+         
+   printf("GL_RENDERER = %s\n", (char *) glGetString(GL_RENDERER));
+
+   /*
+    * Get extension function pointers.
+    */
+   glProgramLocalParameter4fvARB_func = (PFNGLPROGRAMLOCALPARAMETER4FVARBPROC) glutGetProcAddress("glProgramLocalParameter4fvARB");
+   assert(glProgramLocalParameter4fvARB_func);
+
+   glProgramLocalParameter4dARB_func = (PFNGLPROGRAMLOCALPARAMETER4DARBPROC) glutGetProcAddress("glProgramLocalParameter4dARB");
+   assert(glProgramLocalParameter4dARB_func);
+
+   glGetProgramLocalParameterdvARB_func = (PFNGLGETPROGRAMLOCALPARAMETERDVARBPROC) glutGetProcAddress("glGetProgramLocalParameterdvARB");
+   assert(glGetProgramLocalParameterdvARB_func);
+
+   glGenProgramsARB_func = (PFNGLGENPROGRAMSARBPROC) glutGetProcAddress("glGenProgramsARB");
+   assert(glGenProgramsARB_func);
+
+   glProgramStringARB_func = (PFNGLPROGRAMSTRINGARBPROC) glutGetProcAddress("glProgramStringARB");
+   assert(glProgramStringARB_func);
+
+   glBindProgramARB_func = (PFNGLBINDPROGRAMARBPROC) glutGetProcAddress("glBindProgramARB");
+   assert(glBindProgramARB_func);
+
+   glIsProgramARB_func = (PFNGLISPROGRAMARBPROC) glutGetProcAddress("glIsProgramARB");
+   assert(glIsProgramARB_func);
+
+   glDeleteProgramsARB_func = (PFNGLDELETEPROGRAMSARBPROC) glutGetProcAddress("glDeleteProgramsARB");
+   assert(glDeleteProgramsARB_func);
+
+   /*
+    * Vertex program
+    */
+   if (vertProgFile) {
+      FILE *f;
+      int len;
+
+      glGenProgramsARB_func(1, &VertProg);
+      assert(VertProg > 0);
+      glBindProgramARB_func(GL_VERTEX_PROGRAM_ARB, VertProg);
+
+      f = fopen(vertProgFile, "r");
+      if (!f) {
+         printf("Unable to open %s\n", fragProgFile);
+         exit(1);
+      }
+
+      len = fread(buf, 1, 10*1000,f);
+      glProgramStringARB_func(GL_VERTEX_PROGRAM_ARB,
+                              GL_PROGRAM_FORMAT_ASCII_ARB,
+                              len,
+                              (const GLubyte *) buf);
+
+      glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errorPos);
+      if (glGetError() != GL_NO_ERROR || errorPos != -1) {
+         int l = FindLine(buf, errorPos);
+         printf("Vertex Program Error (pos=%d line=%d): %s\n", errorPos, l,
+                (char *) glGetString(GL_PROGRAM_ERROR_STRING_ARB));
+         exit(0);
+      }
+      else {
+         glEnable(GL_VERTEX_PROGRAM_ARB);
+         printf("Vertex Program OK\n");
+      }
+   }
+
+   /*
+    * Fragment program
+    */
+   if (fragProgFile) {
+      FILE *f;
+      int len;
+
+      glGenProgramsARB_func(1, &FragProg);
+      assert(FragProg > 0);
+      glBindProgramARB_func(GL_FRAGMENT_PROGRAM_ARB, FragProg);
+
+      f = fopen(fragProgFile, "r");
+      if (!f) {
+         printf("Unable to open %s\n", fragProgFile);
+         exit(1);
+      }
+
+      len = fread(buf, 1, 10*1000,f);
+      glProgramStringARB_func(GL_FRAGMENT_PROGRAM_ARB,
+                              GL_PROGRAM_FORMAT_ASCII_ARB,
+                              len,
+                              (const GLubyte *) buf);
+
+      glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errorPos);
+      if (glGetError() != GL_NO_ERROR || errorPos != -1) {
+         int l = FindLine(buf, errorPos);
+         printf("Fragment Program Error (pos=%d line=%d): %s\n", errorPos, l,
+                (char *) glGetString(GL_PROGRAM_ERROR_STRING_ARB));
+         exit(0);
+      }
+      else {
+         glEnable(GL_FRAGMENT_PROGRAM_ARB);
+         printf("Fragment Program OK\n");
+      }
+   }
+}
+
+
+int main( int argc, char *argv[] )
+{
+   const char *vertProgFile = NULL, *fragProgFile = NULL;
+   int i;
+
+   glutInit( &argc, argv );
+   glutInitWindowPosition( 0, 0 );
+   glutInitWindowSize( 200, 200 );
+   glutInitDisplayMode( GLUT_RGB | GLUT_DOUBLE | GLUT_DEPTH );
+   Win = glutCreateWindow(argv[0]);
+   glutReshapeFunc( Reshape );
+   glutKeyboardFunc( Key );
+   glutDisplayFunc( Redisplay );
+
+   if (argc == 1) {
+      printf("arbgpuprog:\n");
+      printf("  Compile GL_ARB_vertex/fragment_programs, report any errors.\n");
+      printf("Usage:\n");
+      printf("  arbgpuprog [--vp vertprogfile] [--fp fragprogfile]\n");
+      exit(1);
+   }
+
+   for (i = 1; i < argc; i++) {
+      if (strcmp(argv[i], "--vp") == 0) {
+         vertProgFile = argv[i+1];
+         i++;
+      }
+      else if (strcmp(argv[i], "--fp") == 0) {
+         fragProgFile = argv[i+1];
+         i++;
+      }
+   }
+
+   Init(vertProgFile, fragProgFile);
+
+   glutMainLoop();
+   return 0;
+}
diff --git a/progs/tests/bufferobj.c b/progs/tests/bufferobj.c
index 1d97b060ef..d4ca270016 100644
--- a/progs/tests/bufferobj.c
+++ b/progs/tests/bufferobj.c
@@ -1,5 +1,6 @@
 /*
  * Test GL_ARB_vertex_buffer_object
+ * Also test GL_ARB_vertex_array_object if supported
  *
  * Brian Paul
  * 16 Sep 2003
@@ -9,6 +10,7 @@
 #include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <string.h>
 #include <math.h>
 #include <GL/glew.h>
 #include <GL/glut.h>
@@ -17,6 +19,7 @@
 
 struct object
 {
+   GLuint ArrayObjectID; /** GL_ARB_vertex_array_object */
    GLuint VertexBufferID;
    GLuint ColorBufferID;
    GLuint ElementsBufferID;
@@ -35,6 +38,7 @@ static GLuint Win;
 
 static GLfloat Xrot = 0, Yrot = 0, Zrot = 0;
 static GLboolean Anim = GL_TRUE;
+static GLboolean Have_ARB_vertex_array_object = GL_FALSE;
 
 
 static void CheckError(int line)
@@ -48,34 +52,54 @@ static void CheckError(int line)
 
 static void DrawObject( const struct object *obj )
 {
-   glBindBufferARB(GL_ARRAY_BUFFER_ARB, obj->VertexBufferID);
-   glVertexPointer(3, GL_FLOAT, obj->VertexStride, (void *) obj->VertexOffset);
-   glEnable(GL_VERTEX_ARRAY);
+   if (Have_ARB_vertex_array_object && obj->ArrayObjectID) {
+      glBindVertexArray(obj->ArrayObjectID);
+
+      if (obj->NumElements > 0) {
+         /* indexed arrays */
+         glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, obj->ElementsBufferID);
+         glDrawElements(GL_LINE_LOOP, obj->NumElements, GL_UNSIGNED_INT, NULL);
+      }
+      else {
+         /* non-indexed arrays */
+         glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
+         glDrawArrays(GL_LINE_LOOP, 0, obj->NumVerts);
+      }
+
+      glBindVertexArray(0);
+   }
+   else {
+      /* no vertex array objects, must set vertex/color pointers per draw */
+
+      glBindBufferARB(GL_ARRAY_BUFFER_ARB, obj->VertexBufferID);
+      glVertexPointer(3, GL_FLOAT, obj->VertexStride, (void *) obj->VertexOffset);
+      glEnableClientState(GL_VERTEX_ARRAY);
 
-   /* test push/pop attrib */
-   /* XXX this leads to a segfault with NVIDIA's 53.36 driver */
+      /* test push/pop attrib */
+      /* XXX this leads to a segfault with NVIDIA's 53.36 driver */
 #if 0
-   if (1)
-   {
-      glPushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
-      /*glVertexPointer(3, GL_FLOAT, 0, (void *) (obj->VertexOffset + 10000));*/
-      glBindBufferARB(GL_ARRAY_BUFFER_ARB, 999999);
-      glPopClientAttrib();
-   }
+      if (1)
+      {
+         glPushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+         /*glVertexPointer(3, GL_FLOAT, 0, (void *) (obj->VertexOffset + 10000));*/
+         glBindBufferARB(GL_ARRAY_BUFFER_ARB, 999999);
+         glPopClientAttrib();
+      }
 #endif
-   glBindBufferARB(GL_ARRAY_BUFFER_ARB, obj->ColorBufferID);
-   glColorPointer(3, GL_FLOAT, obj->ColorStride, (void *) obj->ColorOffset);
-   glEnable(GL_COLOR_ARRAY);
-
-   if (obj->NumElements > 0) {
-      /* indexed arrays */
-      glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, obj->ElementsBufferID);
-      glDrawElements(GL_LINE_LOOP, obj->NumElements, GL_UNSIGNED_INT, NULL);
-   }
-   else {
-      /* non-indexed arrays */
-      glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
-      glDrawArrays(GL_LINE_LOOP, 0, obj->NumVerts);
+      glBindBufferARB(GL_ARRAY_BUFFER_ARB, obj->ColorBufferID);
+      glColorPointer(3, GL_FLOAT, obj->ColorStride, (void *) obj->ColorOffset);
+      glEnableClientState(GL_COLOR_ARRAY);
+
+      if (obj->NumElements > 0) {
+         /* indexed arrays */
+         glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, obj->ElementsBufferID);
+         glDrawElements(GL_LINE_LOOP, obj->NumElements, GL_UNSIGNED_INT, NULL);
+      }
+      else {
+         /* non-indexed arrays */
+         glBindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
+         glDrawArrays(GL_LINE_LOOP, 0, obj->NumVerts);
+      }
    }
 }
 
@@ -187,6 +211,28 @@ static void SpecialKey( int key, int x, int y )
 }
 
 
+/**
+ * If GL_ARB_vertex_array_object is supported, create an array object
+ * and set all the per-array state.
+ */
+static void
+CreateVertexArrayObject(struct object *obj)
+{
+   glGenVertexArrays(1, &obj->ArrayObjectID);
+   glBindVertexArray(obj->ArrayObjectID);
+
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, obj->VertexBufferID);
+   glVertexPointer(3, GL_FLOAT, obj->VertexStride, (void *) obj->VertexOffset);
+   glEnableClientState(GL_VERTEX_ARRAY);
+
+   glBindBufferARB(GL_ARRAY_BUFFER_ARB, obj->ColorBufferID);
+   glColorPointer(3, GL_FLOAT, obj->ColorStride, (void *) obj->ColorOffset);
+   glEnableClientState(GL_COLOR_ARRAY);
+
+   glBindVertexArray(0);
+}
+
+
 /*
  * Non-interleaved position/color data.
  */
@@ -262,6 +308,10 @@ static void MakeObject1(struct object *obj)
 
    glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_MAPPED_ARB, &i);
    assert(!i);
+
+   if (Have_ARB_vertex_array_object) {
+      CreateVertexArrayObject(obj);
+   }
 }
 
 
@@ -297,6 +347,10 @@ static void MakeObject2(struct object *obj)
    obj->NumElements = 0;
 
    glUnmapBufferARB(GL_ARRAY_BUFFER_ARB);
+
+   if (Have_ARB_vertex_array_object) {
+      CreateVertexArrayObject(obj);
+   }
 }
 
 
@@ -347,6 +401,10 @@ static void MakeObject3(struct object *obj)
    i[3] = 3;
    glUnmapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB);
    obj->NumElements = 4;
+
+   if (Have_ARB_vertex_array_object) {
+      CreateVertexArrayObject(obj);
+   }
 }
 
 
@@ -387,6 +445,10 @@ static void MakeObject4(struct object *obj)
    /* Setup a buffer of indices to test the ELEMENTS path */
    obj->ElementsBufferID = 0;
    obj->NumElements = 0;
+
+   if (Have_ARB_vertex_array_object) {
+      CreateVertexArrayObject(obj);
+   }
 }
 
 
@@ -399,6 +461,13 @@ static void Init( void )
    }
    printf("GL_RENDERER = %s\n", (char *) glGetString(GL_RENDERER));
 
+   Have_ARB_vertex_array_object =
+      glutExtensionSupported("GL_ARB_vertex_array_object");
+
+   printf("Using GL_ARB_vertex_array_object: %s\n",
+          (Have_ARB_vertex_array_object ? "yes" : "no"));
+
+
    /* Test buffer object deletion */
    if (1) {
       static GLubyte data[1000];
@@ -413,6 +482,7 @@ static void Init( void )
       assert(!glIsBufferARB(id));
    }
 
+   memset(Objects, 0, sizeof(Objects));
    MakeObject1(Objects + 0);
    MakeObject2(Objects + 1);
    MakeObject3(Objects + 2);
diff --git a/progs/tests/getteximage.c b/progs/tests/getteximage.c
new file mode 100644
index 0000000000..e4818a8fab
--- /dev/null
+++ b/progs/tests/getteximage.c
@@ -0,0 +1,217 @@
+/**
+ * Test glGetTexImage()
+ * Brian Paul
+ * 9 June 2009
+ */
+
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <GL/glew.h>
+#include <GL/glut.h>
+
+static int Win;
+
+
+static void
+TestGetTexImage(void)
+{
+   GLuint iter;
+   GLubyte *data = (GLubyte *) malloc(1024 * 1024 * 4);
+   GLubyte *data2 = (GLubyte *) malloc(1024 * 1024 * 4);
+
+   glEnable(GL_TEXTURE_2D);
+
+   printf("glTexImage2D + glGetTexImage:\n");
+
+   for (iter = 0; iter < 8; iter++) {
+      GLint p = (iter % 8) + 3;
+      GLint w = (1 << p);
+      GLint h = (1 << p);
+      GLuint i;
+      GLint level = 0;
+
+      printf("  Testing %d x %d tex image\n", w, h);
+
+      /* fill data */
+      for (i = 0; i < w * h * 4; i++) {
+         data[i] = i & 0xff;
+         data2[i] = 0;
+      }
+
+      glTexImage2D(GL_TEXTURE_2D, level, GL_RGBA, w, h, 0,
+                   GL_RGBA, GL_UNSIGNED_BYTE, data);
+
+      glBegin(GL_POINTS);
+      glVertex2f(0, 0);
+      glEnd();
+
+      /* get */
+      glGetTexImage(GL_TEXTURE_2D, level, GL_RGBA, GL_UNSIGNED_BYTE, data2);
+
+      /* compare */
+      for (i = 0; i < w * h * 4; i++) {
+         if (data2[i] != data[i]) {
+            printf("glTexImage + glGetTexImage failure!\n");
+            printf("Expected value %d, found %d\n", data[i], data2[i]);
+            abort();
+         }
+      }
+   }
+
+   printf("Passed\n");
+   glDisable(GL_TEXTURE_2D);
+   free(data);
+   free(data2);
+}
+
+
+static GLboolean
+ColorsEqual(const GLubyte ref[4], const GLubyte act[4])
+{
+   if (abs((int) ref[0] - (int) act[0]) > 1 ||
+       abs((int) ref[1] - (int) act[1]) > 1 ||
+       abs((int) ref[2] - (int) act[2]) > 1 ||
+       abs((int) ref[3] - (int) act[3]) > 1) {
+      printf("expected %d %d %d %d\n", ref[0], ref[1], ref[2], ref[3]);
+      printf("found    %d %d %d %d\n", act[0], act[1], act[2], act[3]);
+      return GL_FALSE;
+   }
+   return GL_TRUE;
+}
+
+
+static void
+TestGetTexImageRTT(void)
+{
+   GLuint iter;
+   GLuint fb, tex;
+   GLint w = 512;
+   GLint h = 256;
+   GLint level = 0;
+   
+   glGenTextures(1, &tex);
+   glGenFramebuffersEXT(1, &fb);
+
+   glBindTexture(GL_TEXTURE_2D, tex);
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+   glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+   glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, w, h, 0,
+                GL_RGBA, GL_UNSIGNED_BYTE, NULL);
+
+   glBindFramebuffer(GL_FRAMEBUFFER_EXT, fb);
+   glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT,
+                             GL_TEXTURE_2D, tex, level);
+
+   printf("Render to texture + glGetTexImage:\n");
+   printf("  Testing %d x %d tex image\n", w, h);
+   for (iter = 0; iter < 8; iter++) {
+      GLubyte color[4];
+      GLubyte *data2 = (GLubyte *) malloc(w * h * 4);
+      GLuint i;
+
+      /* random clear color */
+      for (i = 0; i < 4; i++) {
+         color[i] = rand() % 256;
+      }
+
+      glClearColor(color[0] / 255.0,
+                   color[1] / 255.0,
+                   color[2] / 255.0,
+                   color[3] / 255.0);
+
+      glClear(GL_COLOR_BUFFER_BIT);
+
+      /* get */
+      glGetTexImage(GL_TEXTURE_2D, level, GL_RGBA, GL_UNSIGNED_BYTE, data2);
+
+      /* compare */
+      for (i = 0; i < w * h; i += 4) {
+         if (!ColorsEqual(color, data2 + i * 4)) {
+            printf("Render to texture failure!\n");
+            abort();
+         }
+      }
+
+      free(data2);
+   }
+
+   glBindFramebuffer(GL_FRAMEBUFFER_EXT, 0);
+   glDeleteFramebuffersEXT(1, &fb);
+   glDeleteTextures(1, &tex);
+
+   printf("Passed\n");
+}
+
+
+
+
+static void
+Draw(void)
+{
+   glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
+
+   TestGetTexImage();
+
+   if (glutExtensionSupported("GL_EXT_framebuffer_object") ||
+       glutExtensionSupported("GL_ARB_framebuffer_object"))
+      TestGetTexImageRTT();
+
+   glutDestroyWindow(Win);
+   exit(0);
+
+   glutSwapBuffers();
+}
+
+
+static void
+Reshape(int width, int height)
+{
+   glViewport(0, 0, width, height);
+   glMatrixMode(GL_PROJECTION);
+   glLoadIdentity();
+   glFrustum(-1.0, 1.0, -1.0, 1.0, 5.0, 25.0);
+   glMatrixMode(GL_MODELVIEW);
+   glLoadIdentity();
+   glTranslatef(0.0, 0.0, -15.0);
+}
+
+
+static void
+Key(unsigned char key, int x, int y)
+{
+   (void) x;
+   (void) y;
+   switch (key) {
+      case 27:
+         glutDestroyWindow(Win);
+         exit(0);
+         break;
+   }
+   glutPostRedisplay();
+}
+
+
+static void
+Init(void)
+{
+}
+
+
+int
+main(int argc, char *argv[])
+{
+   glutInit(&argc, argv);
+   glutInitWindowPosition(0, 0);
+   glutInitWindowSize(400, 400);
+   glutInitDisplayMode(GLUT_RGB | GLUT_DOUBLE | GLUT_DEPTH);
+   Win = glutCreateWindow(argv[0]);
+   glewInit();
+   glutReshapeFunc(Reshape);
+   glutKeyboardFunc(Key);
+   glutDisplayFunc(Draw);
+   Init();
+   glutMainLoop();
+   return 0;
+}
diff --git a/progs/tests/mapbufrange.c b/progs/tests/mapbufrange.c
index 0021bb2607..76e02dd406 100644
--- a/progs/tests/mapbufrange.c
+++ b/progs/tests/mapbufrange.c
@@ -98,10 +98,10 @@ Draw(void)
 {
    glBindBufferARB(GL_ARRAY_BUFFER_ARB, BufferID);
    glVertexPointer(3, GL_FLOAT, 24, 0);
-   glEnable(GL_VERTEX_ARRAY);
+   glEnableClientState(GL_VERTEX_ARRAY);
 
    glColorPointer(3, GL_FLOAT, 24, (void*) 12);
-   glEnable(GL_COLOR_ARRAY);
+   glEnableClientState(GL_COLOR_ARRAY);
 
    glDrawArrays(GL_QUADS, 0, NumRects * 4);
 
diff --git a/progs/tests/mapvbo.c b/progs/tests/mapvbo.c
index c392e76835..52a22a5e79 100644
--- a/progs/tests/mapvbo.c
+++ b/progs/tests/mapvbo.c
@@ -54,10 +54,10 @@ Draw(void)
 
    glBindBufferARB(GL_ARRAY_BUFFER_ARB, BufferID);
    glVertexPointer(3, GL_FLOAT, 24, 0);
-   glEnable(GL_VERTEX_ARRAY);
+   glEnableClientState(GL_VERTEX_ARRAY);
 
    glColorPointer(3, GL_FLOAT, 24, (void*) 12);
-   glEnable(GL_COLOR_ARRAY);
+   glEnableClientState(GL_COLOR_ARRAY);
 
    glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
diff --git a/progs/tests/persp_hint.c b/progs/tests/persp_hint.c
new file mode 100644
index 0000000000..27140d1f56
--- /dev/null
+++ b/progs/tests/persp_hint.c
@@ -0,0 +1,149 @@
+/*
+ * Test the GL_PERSPECTIVE_CORRECTION_HINT setting and its effect on
+ * color interpolation.
+ *
+ * Press 'i' to toggle between GL_NICEST/GL_FASTEST/GL_DONT_CARE.
+ *
+ * Depending on the driver, the hint may make a difference, or not.
+ */
+
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <GL/glut.h>
+
+
+static GLenum PerspHint = GL_DONT_CARE;
+
+
+static void
+init(void)
+{
+   GLubyte image[256][256][4];
+   GLuint i, j;
+   for (i = 0; i < 256; i++) {
+      for (j = 0; j < 256; j++) {
+         image[i][j][0] = j;
+         image[i][j][1] = j;
+         image[i][j][2] = j;
+         image[i][j][3] = 255;
+      }
+   }
+   glTexImage2D(GL_TEXTURE_2D, 0, 4, 256, 256, 0,
+                GL_RGBA, GL_UNSIGNED_BYTE, image);
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+   glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+   glTexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_DECAL);
+
+   printf("GL_RENDERER = %s\n", (char *) glGetString(GL_RENDERER));
+}
+
+
+static void
+display(void)
+{
+   switch (PerspHint) {
+   case GL_NICEST:
+      printf("hint = GL_NICEST\n");
+      break;
+   case GL_FASTEST:
+      printf("hint = GL_FASTEST\n");
+      break;
+   case GL_DONT_CARE:
+      printf("hint = GL_DONT_CARE\n");
+      break;
+   default:
+      ;
+   }
+
+   glClear(GL_COLOR_BUFFER_BIT);
+
+#if 1
+   glBegin(GL_QUADS);
+   /* exercise perspective interpolation */
+   glColor3f(0.0, 0.0, 0.0); glVertex3f(-1.0, -1.0, -1.0);
+   glColor3f(0.0, 0.0, 0.0); glVertex3f(-1.0, 1.0, -1.0);
+   glColor3f(0.0, 1.0, 0.0); glVertex3f( 7.0, 1.0, -7.0);
+   glColor3f(0.0, 1.0, 0.0); glVertex3f( 7.0, -1.0, -7.0);
+
+   /* stripe of linear interpolation */
+   glColor3f(0.0, 0.0, 0.0); glVertex3f(-1.0, -0.1, -1.001);
+   glColor3f(0.0, 0.0, 0.0); glVertex3f(-1.0,  0.1, -1.001);
+   glColor3f(0.0, 1.0, 0.0); glVertex3f( 1.0,  0.1, -1.001);
+   glColor3f(0.0, 1.0, 0.0); glVertex3f( 1.0, -0.1, -1.001);
+   glEnd();
+#else
+   glEnable(GL_TEXTURE_2D);
+   glBegin(GL_QUADS);
+   glTexCoord2f(0.0, 0.0); glVertex3f(-1.0, 0.0, -1.0);
+   glTexCoord2f(0.0, 1.0); glVertex3f(-1.0, 1.0, -1.0);
+   glTexCoord2f(1.0, 1.0); glVertex3f( 5.0, 1.0, -7.0);
+   glTexCoord2f(1.0, 0.0); glVertex3f( 5.0, 0.0, -7.0);
+
+   glTexCoord2f(0.0, 0.0); glVertex3f(-1.0, -1.0, -1.001);
+   glTexCoord2f(0.0, 1.0); glVertex3f(-1.0, 0.0, -1.001);
+   glTexCoord2f(1.0, 1.0); glVertex3f( 1.0, 0.0, -1.001);
+   glTexCoord2f(1.0, 0.0); glVertex3f( 1.0, -1.0, -1.001);
+   glEnd();
+#endif
+
+   glFlush();
+}
+
+
+static void
+reshape(int w, int h)
+{
+   glViewport(0, 0, w, h);
+   glMatrixMode(GL_PROJECTION);
+   glLoadIdentity();
+   gluPerspective(90.0, (GLfloat) w / h, 1.0, 300.0);
+   glMatrixMode(GL_MODELVIEW);
+   glLoadIdentity();
+}
+
+
+static void
+key(unsigned char k, int x, int y)
+{
+   switch (k) {
+   case 27:  /* Escape */
+      exit(0);
+      break;
+   case 'i':
+      if (PerspHint == GL_FASTEST)
+         PerspHint = GL_NICEST;
+      else if (PerspHint == GL_NICEST)
+         PerspHint = GL_DONT_CARE;
+      else
+         PerspHint = GL_FASTEST;
+      glHint(GL_PERSPECTIVE_CORRECTION_HINT, PerspHint);
+      break;
+   default:
+      return;
+   }
+   glutPostRedisplay();
+}
+
+
+int
+main(int argc, char** argv)
+{
+    glutInit(&argc, argv);
+    glutInitDisplayMode (GLUT_SINGLE | GLUT_RGB);
+    glutInitWindowSize (500, 500);
+    glutCreateWindow (argv[0]);
+    glutReshapeFunc (reshape);
+    glutDisplayFunc(display);
+    glutKeyboardFunc(key);
+
+    printf("Main quad: perspective projection\n");
+    printf("Middle stripe: linear interpolation\n");
+    printf("Press 'i' to toggle interpolation hint\n");
+    init();
+
+    glutMainLoop();
+    return 0;
+}
diff --git a/progs/tests/prim.c b/progs/tests/prim.c
new file mode 100644
index 0000000000..3e006e823d
--- /dev/null
+++ b/progs/tests/prim.c
@@ -0,0 +1,559 @@
+#define GL_GLEXT_PROTOTYPES
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <GL/glut.h>
+
+
+#define PIXEL_CENTER(x) ((long)(x) + 0.5)
+
+#define GAP 10
+#define ROWS 3
+#define COLS 4
+
+#define OPENGL_WIDTH 48
+#define OPENGL_HEIGHT 13
+
+
+GLenum provoking = GL_LAST_VERTEX_CONVENTION_EXT;
+GLenum rgb, doubleBuffer, windType;
+GLint windW, windH;
+
+GLenum mode1, mode2;
+GLint boxW, boxH;
+GLubyte OpenGL_bits[] = {
+   0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 
+   0x7f, 0xfb, 0xff, 0xff, 0xff, 0x01,
+   0x7f, 0xfb, 0xff, 0xff, 0xff, 0x01, 
+   0x00, 0x03, 0x00, 0x00, 0x00, 0x00,
+   0x3e, 0x8f, 0xb7, 0xf9, 0xfc, 0x01, 
+   0x63, 0xdb, 0xb0, 0x8d, 0x0d, 0x00,
+   0x63, 0xdb, 0xb7, 0x8d, 0x0d, 0x00, 
+   0x63, 0xdb, 0xb6, 0x8d, 0x0d, 0x00,
+   0x63, 0x8f, 0xf3, 0xcc, 0x0d, 0x00, 
+   0x63, 0x00, 0x00, 0x0c, 0x4c, 0x0a,
+   0x63, 0x00, 0x00, 0x0c, 0x4c, 0x0e, 
+   0x63, 0x00, 0x00, 0x8c, 0xed, 0x0e,
+   0x3e, 0x00, 0x00, 0xf8, 0x0c, 0x00, 
+};
+
+
+#include "tkmap.c"
+
+static void Init(void)
+{
+
+    mode1 = GL_TRUE;
+    mode2 = GL_TRUE;
+}
+
+static void Reshape(int width, int height)
+{
+
+    windW = (GLint)width;
+    windH = (GLint)height;
+}
+
+static void RotateColorMask(void)
+{
+    static GLint rotation = 0;
+    
+    rotation = (rotation + 1) & 0x3;
+    switch (rotation) {
+      case 0:
+	glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+	glIndexMask( 0xff );
+	break;
+      case 1:
+	glColorMask(GL_FALSE, GL_TRUE, GL_TRUE, GL_TRUE);
+	glIndexMask(0xFE);
+	break;
+      case 2:
+	glColorMask(GL_TRUE, GL_FALSE, GL_TRUE, GL_TRUE);
+	glIndexMask(0xFD);
+	break;
+      case 3:
+	glColorMask(GL_TRUE, GL_TRUE, GL_FALSE, GL_TRUE);
+	glIndexMask(0xFB);
+	break;
+    }
+}
+
+static void Key(unsigned char key, int x, int y)
+{
+
+    switch (key) {
+      case 27:
+	exit(1);
+      case '1':
+	mode1 = !mode1;
+	break;
+      case '2':
+	mode2 = !mode2;
+	break;
+      case '3':
+	RotateColorMask();
+	break;
+      case 'p':
+         if (provoking == GL_FIRST_VERTEX_CONVENTION_EXT) {
+            printf("provoke last\n");
+            provoking = GL_LAST_VERTEX_CONVENTION_EXT;
+         }
+         else {
+            printf("provoke first\n");
+            provoking = GL_FIRST_VERTEX_CONVENTION_EXT;
+         }
+         glProvokingVertexEXT(provoking);
+         break;
+      default:
+	return;
+    }
+
+    glutPostRedisplay();
+}
+
+static void Viewport(GLint row, GLint column)
+{
+    GLint x, y;
+
+    boxW = (windW - (COLS + 1) * GAP) / COLS;
+    boxH = (windH - (ROWS + 1) * GAP) / ROWS;
+
+    x = GAP + column * (boxW + GAP);
+    y = GAP + row * (boxH + GAP);
+
+    glViewport(x, y, boxW, boxH);
+
+    glMatrixMode(GL_PROJECTION);
+    glLoadIdentity();
+    glOrtho(-boxW/2, boxW/2, -boxH/2, boxH/2, 0.0, 1.0);
+    glMatrixMode(GL_MODELVIEW);
+
+    glEnable(GL_SCISSOR_TEST);
+    glScissor(x, y, boxW, boxH);
+}
+
+static void Point(void)
+{
+    GLint i;
+
+    glBegin(GL_POINTS);
+	SetColor(COLOR_WHITE);
+	glVertex2i(0, 0);
+	for (i = 1; i < 8; i++) {
+	    GLint j = i * 2;
+	    SetColor(COLOR_BLACK+i);
+	    glVertex2i(-j, -j);
+	    glVertex2i(-j, 0);
+	    glVertex2i(-j, j);
+	    glVertex2i(0, j);
+	    glVertex2i(j, j);
+	    glVertex2i(j, 0);
+	    glVertex2i(j, -j);
+	    glVertex2i(0, -j);
+	}
+    glEnd();
+}
+
+static void Lines(void)
+{
+    GLint i;
+
+    glPushMatrix();
+
+    glTranslatef(-12, 0, 0);
+    for (i = 1; i < 8; i++) {
+	glBegin(GL_LINES);
+	    SetColor(COLOR_BLACK+i);
+	    glVertex2i(-boxW/4, -boxH/4);
+	    SetColor(COLOR_BLACK+i+1);
+	    glVertex2i(boxW/4, boxH/4);
+	glEnd();
+	glTranslatef(4, 0, 0);
+    }
+
+    glPopMatrix();
+
+    glBegin(GL_LINES);
+	glVertex2i(0, 0);
+    glEnd();
+}
+
+static void LineStrip(void)
+{
+
+    glBegin(GL_LINE_STRIP);
+	SetColor(COLOR_RED);
+	glVertex2f(PIXEL_CENTER(-boxW/4), PIXEL_CENTER(-boxH/4));
+	SetColor(COLOR_GREEN);
+	glVertex2f(PIXEL_CENTER(-boxW/4), PIXEL_CENTER(boxH/4));
+	SetColor(COLOR_BLUE);
+	glVertex2f(PIXEL_CENTER(boxW/4), PIXEL_CENTER(boxH/4));
+	SetColor(COLOR_WHITE);
+	glVertex2f(PIXEL_CENTER(boxW/4), PIXEL_CENTER(-boxH/4));
+    glEnd();
+
+    glBegin(GL_LINE_STRIP);
+	glVertex2i(0, 0);
+    glEnd();
+}
+
+static void LineLoop(void)
+{
+
+    glBegin(GL_LINE_LOOP);
+	SetColor(COLOR_RED);
+	glVertex2f(PIXEL_CENTER(-boxW/4), PIXEL_CENTER(-boxH/4));
+	SetColor(COLOR_GREEN);
+	glVertex2f(PIXEL_CENTER(-boxW/4), PIXEL_CENTER(boxH/4));
+	SetColor(COLOR_BLUE);
+	glVertex2f(PIXEL_CENTER(boxW/4), PIXEL_CENTER(boxH/4));
+	SetColor(COLOR_WHITE);
+	glVertex2f(PIXEL_CENTER(boxW/4), PIXEL_CENTER(-boxH/4));
+    glEnd();
+
+    glEnable(GL_LOGIC_OP);
+    glLogicOp(GL_XOR);
+
+    glEnable(GL_BLEND);
+    glBlendFunc(GL_ONE, GL_ONE);
+
+    SetColor(COLOR_MAGENTA);
+    glBegin(GL_LINE_LOOP);
+	glVertex2f(PIXEL_CENTER(-boxW/8), PIXEL_CENTER(-boxH/8));
+	glVertex2f(PIXEL_CENTER(-boxW/8), PIXEL_CENTER(boxH/8));
+    glEnd();
+    glBegin(GL_LINE_LOOP);
+	glVertex2f(PIXEL_CENTER(-boxW/8), PIXEL_CENTER(boxH/8+5));
+	glVertex2f(PIXEL_CENTER(boxW/8), PIXEL_CENTER(boxH/8+5));
+    glEnd();
+    glDisable(GL_LOGIC_OP);
+    glDisable(GL_BLEND);
+
+    SetColor(COLOR_GREEN);
+    glBegin(GL_POINTS);
+	glVertex2i(0, 0);
+    glEnd();
+
+    glBegin(GL_LINE_LOOP);
+	glVertex2i(0, 0);
+    glEnd();
+}
+
+static void Bitmap(void)
+{
+
+    glBegin(GL_LINES);
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/2, 0);
+	glVertex2i(boxW/2, 0);
+	glVertex2i(0, -boxH/2);
+	glVertex2i(0, boxH/2);
+	SetColor(COLOR_RED);
+	glVertex2i(0, -3);
+	glVertex2i(0, -3+OPENGL_HEIGHT);
+	SetColor(COLOR_BLUE);
+	glVertex2i(0, -3);
+	glVertex2i(OPENGL_WIDTH, -3);
+    glEnd();
+
+    SetColor(COLOR_GREEN);
+
+    glPixelStorei(GL_UNPACK_LSB_FIRST, GL_TRUE);
+    glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+
+    glRasterPos2i(0, 0);
+    glBitmap(OPENGL_WIDTH, OPENGL_HEIGHT, 0, 3, 0.0, 0.0, OpenGL_bits);
+}
+
+static void Triangles(void)
+{
+
+    glBegin(GL_TRIANGLES);
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/4, -boxH/4);
+	SetColor(COLOR_RED);
+	glVertex2i(-boxW/8, -boxH/16);
+	SetColor(COLOR_BLUE);
+	glVertex2i(boxW/8, -boxH/16);
+
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/4, boxH/4);
+	SetColor(COLOR_RED);
+	glVertex2i(-boxW/8, boxH/16);
+	SetColor(COLOR_BLUE);
+	glVertex2i(boxW/8, boxH/16);
+    glEnd();
+
+    glBegin(GL_TRIANGLES);
+	glVertex2i(0, 0);
+	glVertex2i(-100, 100);
+    glEnd();
+}
+
+static void TriangleStrip(void)
+{
+
+    glBegin(GL_TRIANGLE_STRIP);
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/4, -boxH/4);
+	SetColor(COLOR_RED);
+	glVertex2i(-boxW/4, boxH/4);
+	SetColor(COLOR_BLUE);
+	glVertex2i(0, -boxH/4);
+	SetColor(COLOR_WHITE);
+	glVertex2i(0, boxH/4);
+	SetColor(COLOR_CYAN);
+	glVertex2i(boxW/4, -boxH/4);
+	SetColor(COLOR_YELLOW);
+	glVertex2i(boxW/4, boxH/4);
+    glEnd();
+
+    glBegin(GL_TRIANGLE_STRIP);
+	glVertex2i(0, 0);
+	glVertex2i(-100, 100);
+    glEnd();
+}
+
+static void TriangleFan(void)
+{
+    GLint vx[8][2];
+    GLint x0, y0, x1, y1, x2, y2, x3, y3;
+    GLint i;
+
+    y0 = -boxH/4;
+    y1 = y0 + boxH/2/3;
+    y2 = y1 + boxH/2/3;
+    y3 = boxH/4;
+    x0 = -boxW/4;
+    x1 = x0 + boxW/2/3;
+    x2 = x1 + boxW/2/3;
+    x3 = boxW/4;
+
+    vx[0][0] = x0; vx[0][1] = y1;
+    vx[1][0] = x0; vx[1][1] = y2;
+    vx[2][0] = x1; vx[2][1] = y3;
+    vx[3][0] = x2; vx[3][1] = y3;
+    vx[4][0] = x3; vx[4][1] = y2;
+    vx[5][0] = x3; vx[5][1] = y1;
+    vx[6][0] = x2; vx[6][1] = y0;
+    vx[7][0] = x1; vx[7][1] = y0;
+
+    glBegin(GL_TRIANGLE_FAN);
+	SetColor(COLOR_WHITE);
+	glVertex2i(0, 0);
+	for (i = 0; i < 8; i++) {
+	    SetColor(COLOR_WHITE-i);
+	    glVertex2iv(vx[i]);
+	}
+    glEnd();
+
+    glBegin(GL_TRIANGLE_FAN);
+	glVertex2i(0, 0);
+	glVertex2i(-100, 100);
+    glEnd();
+}
+
+static void Rect(void)
+{
+
+    SetColor(COLOR_GREEN);
+    glRecti(-boxW/4, -boxH/4, boxW/4, boxH/4);
+}
+
+static void PolygonFunc(void)
+{
+    GLint vx[8][2];
+    GLint x0, y0, x1, y1, x2, y2, x3, y3;
+    GLint i;
+
+    y0 = -boxH/4;
+    y1 = y0 + boxH/2/3;
+    y2 = y1 + boxH/2/3;
+    y3 = boxH/4;
+    x0 = -boxW/4;
+    x1 = x0 + boxW/2/3;
+    x2 = x1 + boxW/2/3;
+    x3 = boxW/4;
+
+    vx[0][0] = x0; vx[0][1] = y1;
+    vx[1][0] = x0; vx[1][1] = y2;
+    vx[2][0] = x1; vx[2][1] = y3;
+    vx[3][0] = x2; vx[3][1] = y3;
+    vx[4][0] = x3; vx[4][1] = y2;
+    vx[5][0] = x3; vx[5][1] = y1;
+    vx[6][0] = x2; vx[6][1] = y0;
+    vx[7][0] = x1; vx[7][1] = y0;
+
+    glBegin(GL_POLYGON);
+	for (i = 0; i < 8; i++) {
+	    SetColor(COLOR_WHITE-i);
+	    glVertex2iv(vx[i]);
+	}
+    glEnd();
+
+    glBegin(GL_POLYGON);
+	glVertex2i(0, 0);
+	glVertex2i(100, 100);
+    glEnd();
+}
+
+static void Quads(void)
+{
+
+    glBegin(GL_QUADS);
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/4, -boxH/4);
+	SetColor(COLOR_RED);
+	glVertex2i(-boxW/8, -boxH/16);
+	SetColor(COLOR_BLUE);
+	glVertex2i(boxW/8, -boxH/16);
+	SetColor(COLOR_WHITE);
+	glVertex2i(boxW/4, -boxH/4);
+
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/4, boxH/4);
+	SetColor(COLOR_RED);
+	glVertex2i(-boxW/8, boxH/16);
+	SetColor(COLOR_BLUE);
+	glVertex2i(boxW/8, boxH/16);
+	SetColor(COLOR_WHITE);
+	glVertex2i(boxW/4, boxH/4);
+    glEnd();
+
+    glBegin(GL_QUADS);
+	glVertex2i(0, 0);
+	glVertex2i(100, 100);
+	glVertex2i(-100, 100);
+    glEnd();
+}
+
+static void QuadStrip(void)
+{
+
+    glBegin(GL_QUAD_STRIP);
+	SetColor(COLOR_GREEN);
+	glVertex2i(-boxW/4, -boxH/4);
+	SetColor(COLOR_RED);
+	glVertex2i(-boxW/4, boxH/4);
+	SetColor(COLOR_BLUE);
+	glVertex2i(0, -boxH/4);
+	SetColor(COLOR_WHITE);
+	glVertex2i(0, boxH/4);
+	SetColor(COLOR_CYAN);
+	glVertex2i(boxW/4, -boxH/4);
+	SetColor(COLOR_YELLOW);
+	glVertex2i(boxW/4, boxH/4);
+    glEnd();
+
+    glBegin(GL_QUAD_STRIP);
+	glVertex2i(0, 0);
+	glVertex2i(100, 100);
+	glVertex2i(-100, 100);
+    glEnd();
+}
+
+static void Draw(void)
+{
+
+    glViewport(0, 0, windW, windH);
+    glDisable(GL_SCISSOR_TEST);
+
+    glPushAttrib(GL_COLOR_BUFFER_BIT);
+
+    glColorMask(1, 1, 1, 1);
+    glIndexMask(~0);
+
+    glClearColor(0.0, 0.0, 0.0, 0.0);
+    glClear(GL_COLOR_BUFFER_BIT);
+
+    glPopAttrib();
+
+    if (mode1) {
+	glShadeModel(GL_SMOOTH);
+    } else {
+	glShadeModel(GL_FLAT);
+    }
+
+    if (mode2) {
+	glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
+    } else {
+	glPolygonMode(GL_FRONT_AND_BACK, GL_LINE);
+    }
+
+    Viewport(0, 0); Point();
+    Viewport(0, 1); Lines();
+    Viewport(0, 2); LineStrip();
+    Viewport(0, 3); LineLoop();
+
+    Viewport(1, 0); Bitmap();
+    Viewport(1, 1); TriangleFan();
+    Viewport(1, 2); Triangles();
+    Viewport(1, 3); TriangleStrip();
+
+    Viewport(2, 0); Rect();
+    Viewport(2, 1); PolygonFunc();
+    Viewport(2, 2); Quads();
+    Viewport(2, 3); QuadStrip();
+
+    glFlush();
+
+    if (doubleBuffer) {
+	glutSwapBuffers();
+    }
+}
+
+static GLenum Args(int argc, char **argv)
+{
+    GLint i;
+
+    rgb = GL_TRUE;
+    doubleBuffer = GL_FALSE;
+
+    for (i = 1; i < argc; i++) {
+	if (strcmp(argv[i], "-ci") == 0) {
+	    rgb = GL_FALSE;
+	} else if (strcmp(argv[i], "-rgb") == 0) {
+	    rgb = GL_TRUE;
+	} else if (strcmp(argv[i], "-sb") == 0) {
+	    doubleBuffer = GL_FALSE;
+	} else if (strcmp(argv[i], "-db") == 0) {
+	    doubleBuffer = GL_TRUE;
+	} else {
+	    printf("%s (Bad option).\n", argv[i]);
+	    return GL_FALSE;
+	}
+    }
+    return GL_TRUE;
+}
+
+int main(int argc, char **argv)
+{
+    glutInit(&argc, argv);
+
+    if (Args(argc, argv) == GL_FALSE) {
+	exit(1);
+    }
+
+    windW = 600;
+    windH = 300;
+    glutInitWindowPosition(0, 0); glutInitWindowSize( windW, windH);
+
+    windType = (rgb) ? GLUT_RGB : GLUT_INDEX;
+    windType |= (doubleBuffer) ? GLUT_DOUBLE : GLUT_SINGLE;
+    glutInitDisplayMode(windType);
+
+    if (glutCreateWindow("Primitive Test") == GL_FALSE) {
+	exit(1);
+    }
+
+    InitMap();
+
+    Init();
+
+    glutReshapeFunc(Reshape);
+    glutKeyboardFunc(Key);
+    glutDisplayFunc(Draw);
+    glutMainLoop();
+	return 0;
+}
diff --git a/progs/tests/tkmap.c b/progs/tests/tkmap.c
new file mode 100644
index 0000000000..3ded79caca
--- /dev/null
+++ b/progs/tests/tkmap.c
@@ -0,0 +1,71 @@
+
+enum {
+    COLOR_BLACK = 0,
+    COLOR_RED,
+    COLOR_GREEN,
+    COLOR_YELLOW,
+    COLOR_BLUE,
+    COLOR_MAGENTA,
+    COLOR_CYAN,
+    COLOR_WHITE
+};
+
+static float RGBMap[9][3] = {
+    {0, 0, 0},
+    {1, 0, 0},
+    {0, 1, 0},
+    {1, 1, 0},
+    {0, 0, 1},
+    {1, 0, 1},
+    {0, 1, 1},
+    {1, 1, 1},
+    {0.5, 0.5, 0.5}
+};
+
+static void SetColor(int c)
+{
+    if (glutGet(GLUT_WINDOW_RGBA))
+        glColor3fv(RGBMap[c]);
+    else
+        glIndexf(c);
+}
+
+static void InitMap(void)
+{
+    int i;
+
+    if (rgb)
+	return;
+
+    for (i = 0; i < 9; i++)
+	    glutSetColor(i, RGBMap[i][0], RGBMap[i][1], RGBMap[i][2]);
+}
+
+static void SetFogRamp(int density, int startIndex)
+{
+    int fogValues, colorValues;
+    int i, j, k;
+    float intensity;
+
+    fogValues = 1 << density;
+    colorValues = 1 << startIndex;
+    for (i = 0; i < colorValues; i++) {
+	for (j = 0; j < fogValues; j++) {
+	    k = i * fogValues + j;
+	    intensity = (i * fogValues + j * colorValues) / 255.0;
+	    glutSetColor(k, intensity, intensity, intensity);
+	}
+    }
+}
+
+static void SetGreyRamp(void)
+{
+    int i;
+    float intensity;
+
+    for (i = 0; i < 255; i++) {
+	intensity = i / 255.0;
+	glutSetColor(i, intensity, intensity, intensity);
+    }
+}
+
diff --git a/progs/trivial/.gitignore b/progs/trivial/.gitignore
index 8dcb20a68f..dce733a70a 100644
--- a/progs/trivial/.gitignore
+++ b/progs/trivial/.gitignore
@@ -19,6 +19,7 @@ fs-tri
 line
 line-clip
 line-cull
+line-flat
 line-smooth
 line-stipple-wide
 line-userclip
@@ -130,6 +131,7 @@ tristrip-flat
 vbo-drawarrays
 vbo-drawelements
 vbo-drawrange
+vbo-noninterleaved
 vp-array
 vp-array-int
 vp-clip
@@ -139,6 +141,7 @@ vp-tri-cb
 vp-tri-cb-pos
 vp-tri-cb-tex
 vp-tri-imm
+vp-tri-invariant
 vp-tri-swap
 vp-tri-tex
 vp-unfilled
diff --git a/progs/util/extfuncs.h b/progs/util/extfuncs.h
index 0469e2f2c4..2bb57030a8 100644
--- a/progs/util/extfuncs.h
+++ b/progs/util/extfuncs.h
@@ -137,7 +137,6 @@ static PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glRenderbufferStorageMultisample_
 static PFNGLFRAMEBUFFERTEXTURELAYERPROC glFramebufferTextureLayer_func = NULL;
 
 
-
 static void
 GetExtensionFuncs(void)
 {
diff --git a/progs/xdemos/glxswapcontrol.c b/progs/xdemos/glxswapcontrol.c
index 5a5d084f90..df9f7ad784 100644
--- a/progs/xdemos/glxswapcontrol.c
+++ b/progs/xdemos/glxswapcontrol.c
@@ -303,13 +303,13 @@ draw(void)
       glDisable(GL_LIGHTING);
       glShadeModel(GL_SMOOTH);
 
-      glEnable( GL_VERTEX_ARRAY );
-      glEnable( GL_COLOR_ARRAY );
+      glEnableClientState( GL_VERTEX_ARRAY );
+      glEnableClientState( GL_COLOR_ARRAY );
       glVertexPointer( 3, GL_FLOAT, 0, vert );
       glColorPointer( 3, GL_FLOAT, 0, col );
       glDrawArrays( GL_POLYGON, 0, 4 );
-      glDisable( GL_COLOR_ARRAY );
-      glDisable( GL_VERTEX_ARRAY );
+      glDisableClientState( GL_COLOR_ARRAY );
+      glDisableClientState( GL_VERTEX_ARRAY );
 
       glMatrixMode(GL_PROJECTION);
       glLoadIdentity();
diff --git a/scons/gallium.py b/scons/gallium.py
index 8082aba6f2..e9e799dc78 100644
--- a/scons/gallium.py
+++ b/scons/gallium.py
@@ -44,17 +44,18 @@ import fixes
 def quietCommandLines(env):
     # Quiet command lines
     # See also http://www.scons.org/wiki/HidingCommandLinesInOutput
-    env['ASCOMSTR'] = "Assembling $SOURCE ..."
-    env['CCCOMSTR'] = "Compiling $SOURCE ..."
-    env['SHCCCOMSTR'] = "Compiling $SOURCE ..."
-    env['CXXCOMSTR'] = "Compiling $SOURCE ..."
-    env['SHCXXCOMSTR'] = "Compiling $SOURCE ..."
-    env['ARCOMSTR'] = "Archiving $TARGET ..."
-    env['RANLIBCOMSTR'] = "Indexing $TARGET ..."
-    env['LINKCOMSTR'] = "Linking $TARGET ..."
-    env['SHLINKCOMSTR'] = "Linking $TARGET ..."
-    env['LDMODULECOMSTR'] = "Linking $TARGET ..."
-    env['SWIGCOMSTR'] = "Generating $TARGET ..."
+    env['ASCOMSTR'] = "  Assembling $SOURCE ..."
+    env['ASPPCOMSTR'] = "  Assembling $SOURCE ..."
+    env['CCCOMSTR'] = "  Compiling $SOURCE ..."
+    env['SHCCCOMSTR'] = "  Compiling $SOURCE ..."
+    env['CXXCOMSTR'] = "  Compiling $SOURCE ..."
+    env['SHCXXCOMSTR'] = "  Compiling $SOURCE ..."
+    env['ARCOMSTR'] = "  Archiving $TARGET ..."
+    env['RANLIBCOMSTR'] = "  Indexing $TARGET ..."
+    env['LINKCOMSTR'] = "  Linking $TARGET ..."
+    env['SHLINKCOMSTR'] = "  Linking $TARGET ..."
+    env['LDMODULECOMSTR'] = "  Linking $TARGET ..."
+    env['SWIGCOMSTR'] = "  Generating $TARGET ..."
 
 
 def createConvenienceLibBuilder(env):
diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c
index 1750e976b8..f316974d83 100644
--- a/src/egl/drivers/demo/demo.c
+++ b/src/egl/drivers/demo/demo.c
@@ -146,12 +146,12 @@ demoCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext
    if (!c)
       return EGL_NO_CONTEXT;
 
-   _eglInitContext(drv, dpy, &c->Base, config, attrib_list);
+   _eglInitContext(drv, &c->Base, conf, attrib_list);
    c->DemoStuff = 1;
    printf("demoCreateContext\n");
 
-   /* generate handle and insert into hash table */
-   _eglSaveContext(&c->Base);
+   /* link to display */
+   _eglLinkContext(&c->Base, _eglLookupDisplay(dpy));
    assert(_eglGetContextHandle(&c->Base));
 
    return _eglGetContextHandle(&c->Base);
@@ -213,11 +213,14 @@ demoCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
                          const EGLint *attrib_list)
 {
    DemoSurface *surf = (DemoSurface *) calloc(1, sizeof(DemoSurface));
+   _EGLConfig *conf;
+
    if (!surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   conf = _eglLookupConfig(drv, dpy, config);
+   if (!_eglInitSurface(drv, &surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
@@ -232,13 +235,9 @@ static EGLBoolean
 demoDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    DemoSurface *fs = LookupDemoSurface(surface);
-   _eglRemoveSurface(&fs->Base);
-   if (fs->Base.IsBound) {
-      fs->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   _eglUnlinkSurface(&fs->Base);
+   if (!fs->Base.IsBound)
       free(fs);
-   }
    return EGL_TRUE;
 }
 
@@ -247,13 +246,9 @@ static EGLBoolean
 demoDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
 {
    DemoContext *fc = LookupDemoContext(context);
-   _eglRemoveContext(&fc->Base);
-   if (fc->Base.IsBound) {
-      fc->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   _eglUnlinkContext(&fc->Base);
+   if (!fc->Base.IsBound)
       free(fc);
-   }
    return EGL_TRUE;
 }
 
diff --git a/src/egl/drivers/dri/Makefile b/src/egl/drivers/dri/Makefile
index 4041d5c906..567edfae97 100644
--- a/src/egl/drivers/dri/Makefile
+++ b/src/egl/drivers/dri/Makefile
@@ -50,7 +50,7 @@ $(TOP)/$(LIB_DIR)/libEGLdri.so: $(OBJECTS)
 
 install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libEGLdri.so $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libEGLdri.so $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o
diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c
index 57661cc3ab..3f9617a8f9 100644
--- a/src/egl/drivers/dri/egldri.c
+++ b/src/egl/drivers/dri/egldri.c
@@ -171,7 +171,10 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    if (!c)
       return EGL_NO_CONTEXT;
 
-   if (!_eglInitContext(drv, dpy, &c->Base, config, attrib_list)) {
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
+
+   if (!_eglInitContext(drv, &c->Base, conf, attrib_list)) {
       free(c);
       return EGL_NO_CONTEXT;
    }
@@ -189,8 +192,6 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    else
       sharePriv = NULL;
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   assert(conf);
    _eglConfigToContextModesRec(conf, &visMode);
 
    c->driContext.private = disp->driScreen.createNewContext(disp, &visMode,
@@ -200,8 +201,8 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       return EGL_FALSE;
    }
 
-   /* generate handle and insert into hash table */
-   _eglSaveContext(&c->Base);
+   /* link to display */
+   _eglLinkContext(&c->Base, &disp->Base);
 
    return _eglGetContextHandle(&c->Base);
 }
@@ -237,14 +238,18 @@ _eglDRICreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
                             const EGLint *attrib_list)
 {
    driSurface *surf;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
 
    surf = (driSurface *) calloc(1, sizeof(*surf));
    if (!surf) {
       return EGL_NO_SURFACE;
    }
 
-   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
@@ -275,7 +280,7 @@ _eglDRICreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 #endif
    }
 
-   _eglSaveSurface(&surf->Base);
+   _eglLinkSurface(&surf->Base, _eglLookupDisplay(dpy));
 
    return surf->Base.Handle;
 }
@@ -287,16 +292,12 @@ _eglDRIDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
    driDisplay *disp = Lookup_driDisplay(dpy);
    driSurface *fs = Lookup_driSurface(surface);
 
-   _eglRemoveSurface(&fs->Base);
+   _eglUnlinkSurface(&fs->Base);
 
    fs->drawable.destroyDrawable(disp, fs->drawable.private);
 
-   if (fs->Base.IsBound) {
-      fs->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   if (!fs->Base.IsBound)
       free(fs);
-   }
    return EGL_TRUE;
 }
 
@@ -307,16 +308,12 @@ _eglDRIDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
    driDisplay *disp = Lookup_driDisplay(dpy);
    driContext *fc = Lookup_driContext(context);
 
-   _eglRemoveContext(&fc->Base);
+   _eglUnlinkContext(&fc->Base);
 
    fc->driContext.destroyContext(disp, 0, fc->driContext.private);
 
-   if (fc->Base.IsBound) {
-      fc->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   if (!fc->Base.IsBound)
       free(fc);
-   }
    return EGL_TRUE;
 }
 
@@ -340,13 +337,13 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    }
 
    /* init base class, do error checking, etc. */
-   if (!_eglInitSurface(drv, dpy, &surface->Base, EGL_SCREEN_BIT_MESA,
-                        cfg, attrib_list)) {
+   if (!_eglInitSurface(drv, &surface->Base, EGL_SCREEN_BIT_MESA,
+                        config, attrib_list)) {
       free(surface);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(&surface->Base);
+   _eglLinkSurface(&surface->Base &disp->Base);
 
 
    /*
@@ -363,7 +360,7 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    if (!disp->driScreen.createNewDrawable(disp, &visMode, drawBuf,
                                           &surface->drawable, GLX_WINDOW_BIT,
                                           empty_attribute_list)) {
-      _eglRemoveSurface(&surface->Base);
+      _eglUnlinkSurface(&surface->Base);
       free(surface);
       return EGL_NO_SURFACE;
    }
diff --git a/src/egl/drivers/glx/Makefile b/src/egl/drivers/glx/Makefile
index 5f041a268f..20ef0352ad 100644
--- a/src/egl/drivers/glx/Makefile
+++ b/src/egl/drivers/glx/Makefile
@@ -58,7 +58,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS)
 
 install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	rm -f *.o
diff --git a/src/egl/drivers/glx/egl_glx.c b/src/egl/drivers/glx/egl_glx.c
index 155caa413c..207b1ea779 100644
--- a/src/egl/drivers/glx/egl_glx.c
+++ b/src/egl/drivers/glx/egl_glx.c
@@ -532,7 +532,10 @@ GLX_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    if (!GLX_ctx)
       return EGL_NO_CONTEXT;
 
-   if (!_eglInitContext(drv, dpy, &GLX_ctx->Base, config, attrib_list)) {
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
+
+   if (!_eglInitContext(drv, &GLX_ctx->Base, conf, attrib_list)) {
       free(GLX_ctx);
       return EGL_NO_CONTEXT;
    }
@@ -546,9 +549,6 @@ GLX_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       GLX_ctx_shared = GLX_egl_context(shareCtx);
    }
 
-   conf = _eglLookupConfig(drv, dpy, config);
-   assert(conf);
-
 #ifdef GLX_VERSION_1_3
    if (GLX_drv->fbconfigs)
       GLX_ctx->context = glXCreateNewContext(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], GLX_RGBA_TYPE, GLX_ctx_shared ? GLX_ctx_shared->context : NULL, GL_TRUE);
@@ -564,7 +564,7 @@ GLX_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       return EGL_FALSE;
 #endif
 
-   return _eglGetContextHandle(&GLX_ctx->Base);
+   return _eglLinkContext(&GLX_ctx->Base, disp);
 }
 
 
@@ -619,18 +619,22 @@ GLX_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    _EGLDisplay *disp = _eglLookupDisplay(dpy);
    struct GLX_egl_surface *GLX_surf;
    uint width, height;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
 
    GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
    if (!GLX_surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_WINDOW_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &GLX_surf->Base, EGL_WINDOW_BIT,
+                        conf, attrib_list)) {
       free(GLX_surf);
       return EGL_FALSE;
    }
 
-   _eglSaveSurface(&GLX_surf->Base);
+   _eglLinkSurface(&GLX_surf->Base, disp);
 
    GLX_surf->drawable = window;
    get_drawable_size(disp->Xdpy, window, &width, &height);
@@ -648,19 +652,23 @@ GLX_eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
    _EGLDisplay *disp = _eglLookupDisplay(dpy);
    struct GLX_egl_surface *GLX_surf;
+   _EGLConfig *conf;
    int i;
 
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
+
    GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
    if (!GLX_surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_PIXMAP_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &GLX_surf->Base, EGL_PIXMAP_BIT,
+                        conf, attrib_list)) {
       free(GLX_surf);
       return EGL_FALSE;
    }
 
-   _eglSaveSurface(&GLX_surf->Base);
+   _eglLinkSurface(&GLX_surf->Base, disp);
 
    for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
       switch (attrib_list[i]) {
@@ -683,20 +691,24 @@ GLX_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
    _EGLDisplay *disp = _eglLookupDisplay(dpy);
    struct GLX_egl_surface *GLX_surf;
+   _EGLConfig *conf;
    int attribs[5];
    int i = 0, j = 0;
 
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
+
    GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
    if (!GLX_surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &GLX_surf->Base, EGL_PBUFFER_BIT,
+                        conf, attrib_list)) {
       free(GLX_surf);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(&GLX_surf->Base);
+   _eglLinkSurface(&GLX_surf->Base, disp);
 
    while(attrib_list[i] != EGL_NONE) {
       switch (attrib_list[i]) {
@@ -726,13 +738,9 @@ GLX_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
    _EGLSurface *surf = _eglLookupSurface(surface);
    return EGL_TRUE;
    if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (surf->IsBound) {
-         surf->DeletePending = EGL_TRUE;
-      }
-      else {
+      _eglUnlinkSurface(surf);
+      if (!surf->IsBound)
          free(surf);
-      }
 
       return EGL_TRUE;
    }
diff --git a/src/egl/drivers/xdri/Makefile b/src/egl/drivers/xdri/Makefile
index eb83867b71..8a14027fc7 100644
--- a/src/egl/drivers/xdri/Makefile
+++ b/src/egl/drivers/xdri/Makefile
@@ -54,7 +54,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS)
 
 install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	rm -f *.o
diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c
index 3b3e312746..e040efdd43 100644
--- a/src/egl/drivers/xdri/egl_xdri.c
+++ b/src/egl/drivers/xdri/egl_xdri.c
@@ -770,7 +770,7 @@ xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    if (!xdri_ctx)
       return EGL_NO_CONTEXT;
 
-   if (!_eglInitContext(drv, dpy, &xdri_ctx->Base, config, attrib_list)) {
+   if (!_eglInitContext(drv, &xdri_ctx->Base, &xdri_config->Base, attrib_list)) {
       free(xdri_ctx);
       return EGL_NO_CONTEXT;
    }
@@ -794,7 +794,7 @@ xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    xdri_ctx->driContext.mode = xdri_config->mode;
 
-   return _eglGetContextHandle(&xdri_ctx->Base);
+   return _eglLinkContext(&xdri_ctx->Base, &disp);
 }
 
 
@@ -835,6 +835,7 @@ xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
                             NativeWindowType window, const EGLint *attrib_list)
 {
    _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config);
    struct xdri_egl_surface *xdri_surf;
    int scrn = DefaultScreen(disp->Xdpy);
    uint width, height;
@@ -843,8 +844,8 @@ xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    if (!xdri_surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_WINDOW_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &xdri_surf->Base, EGL_WINDOW_BIT,
+                        &xdri_config->Base, attrib_list)) {
       free(xdri_surf);
       return EGL_FALSE;
    }
@@ -856,7 +857,7 @@ xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    xdri_surf->driDrawable = window;
 
-   _eglSaveSurface(&xdri_surf->Base);
+   _eglLinkSurface(&xdri_surf->Base, disp);
 
    get_drawable_size(disp->Xdpy, window, &width, &height);
    xdri_surf->Base.Width = width;
@@ -888,8 +889,8 @@ xdri_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    if (!xdri_surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_PBUFFER_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &xdri_surf->Base, EGL_PBUFFER_BIT,
+                        &xdri_config->Base, attrib_list)) {
       free(xdri_surf);
       return EGL_FALSE;
    }
@@ -939,7 +940,7 @@ xdri_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 
    xdri_surf->driDrawable = window;
 
-   _eglSaveSurface(&xdri_surf->Base);
+   _eglLinkSurface(&xdri_surf->Base, disp);
 
    _eglLog(_EGL_DEBUG,
            "XDRI: CreatePbufferSurface handle %d  hDrawable %d",
@@ -956,11 +957,8 @@ xdri_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
    if (xdri_surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (xdri_surf->Base.IsBound) {
-         xdri_surf->Base.DeletePending = EGL_TRUE;
-      }
-      else {
+      _eglUnlinkSurface(&xdri_surf->Base);
+      if (!xdri_surf->Base.IsBound) {
          /*
          st_unreference_framebuffer(surf->Framebuffer);
          */
diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile
index 8cfa25ca16..e1ff8794b3 100644
--- a/src/egl/main/Makefile
+++ b/src/egl/main/Makefile
@@ -7,9 +7,11 @@ include $(TOP)/configs/current
 INCLUDE_DIRS = -I$(TOP)/include -I$(TOP)/src/mesa/glapi
 
 HEADERS = \
+	eglcompiler.h \
 	eglconfig.h \
 	eglconfigutil.h \
 	eglcontext.h \
+	eglcurrent.h \
 	egldefines.h \
 	egldisplay.h \
 	egldriver.h \
@@ -28,6 +30,7 @@ SOURCES = \
 	eglconfig.c \
 	eglconfigutil.c \
 	eglcontext.c \
+	eglcurrent.c \
 	egldisplay.c \
 	egldriver.c \
 	eglglobals.c \
@@ -66,7 +69,7 @@ $(TOP)/$(LIB_DIR)/libEGL.so: $(OBJECTS)
 
 install: default
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libEGL.so* $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libEGL.so* $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o *.so*
diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c
index 9df938e188..f0a6f7f935 100644
--- a/src/egl/main/eglapi.c
+++ b/src/egl/main/eglapi.c
@@ -51,7 +51,12 @@ eglGetDisplay(NativeDisplayType nativeDisplay)
 {
    _EGLDisplay *dpy;
    _eglInitGlobals();
-   dpy = _eglNewDisplay(nativeDisplay);
+   dpy = _eglFindDisplay(nativeDisplay);
+   if (!dpy) {
+      dpy = _eglNewDisplay(nativeDisplay);
+      if (dpy)
+         _eglLinkDisplay(dpy);
+   }
    return _eglGetDisplayHandle(dpy);
 }
 
@@ -321,7 +326,8 @@ eglGetError(void)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
    EGLint e = t->LastError;
-   t->LastError = EGL_SUCCESS;
+   if (!_eglIsCurrentThreadDummy())
+      t->LastError = EGL_SUCCESS;
    return e;
 }
 
@@ -546,11 +552,17 @@ eglBindAPI(EGLenum api)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
 
+   if (_eglIsCurrentThreadDummy())
+      return _eglError(EGL_BAD_ALLOC, "eglBindAPI");
+
+   if (!_eglIsApiValid(api))
+      return _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
+
    switch (api) {
 #ifdef EGL_VERSION_1_4
    case EGL_OPENGL_API:
       if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT) {
-         t->CurrentAPI = api;
+         t->CurrentAPIIndex = _eglConvertApiToIndex(api);
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
@@ -558,14 +570,14 @@ eglBindAPI(EGLenum api)
 #endif
    case EGL_OPENGL_ES_API:
       if (_eglGlobal.ClientAPIsMask & (EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT)) {
-         t->CurrentAPI = api;
+         t->CurrentAPIIndex = _eglConvertApiToIndex(api);
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
       return EGL_FALSE;
    case EGL_OPENVG_API:
       if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT) {
-         t->CurrentAPI = api;
+         t->CurrentAPIIndex = _eglConvertApiToIndex(api);
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
@@ -585,7 +597,7 @@ eglQueryAPI(void)
 {
    /* returns one of EGL_OPENGL_API, EGL_OPENGL_ES_API or EGL_OPENVG_API */
    _EGLThreadInfo *t = _eglGetCurrentThread();
-   return t->CurrentAPI;
+   return _eglConvertApiFromIndex(t->CurrentAPIIndex);
 }
 
 
@@ -603,15 +615,19 @@ eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype,
 EGLBoolean
 eglReleaseThread(void)
 {
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   EGLDisplay dpy = eglGetCurrentDisplay();
+   EGLDisplay dpy;
+
+   if (_eglIsCurrentThreadDummy())
+      return EGL_TRUE;
+
+   dpy = eglGetCurrentDisplay();
    if (dpy) {
       _EGLDriver *drv = _eglLookupDriver(dpy);
       /* unbind context */
       (void) drv->API.MakeCurrent(drv, dpy, EGL_NO_SURFACE,
                                   EGL_NO_SURFACE, EGL_NO_CONTEXT);
    }
-   _eglDeleteThreadData(t);
+   _eglDestroyCurrentThread();
    return EGL_TRUE;
 }
 
diff --git a/src/egl/main/eglcompiler.h b/src/egl/main/eglcompiler.h
new file mode 100644
index 0000000000..0b19afedfd
--- /dev/null
+++ b/src/egl/main/eglcompiler.h
@@ -0,0 +1,31 @@
+#ifndef EGLCOMPILER_INCLUDED
+#define EGLCOMPILER_INCLUDED
+
+
+/**
+ * Function inlining
+ */
+#if defined(__GNUC__)
+#  define INLINE __inline__
+#elif defined(__MSC__)
+#  define INLINE __inline
+#elif defined(_MSC_VER)
+#  define INLINE __inline
+#elif defined(__ICL)
+#  define INLINE __inline
+#elif defined(__INTEL_COMPILER)
+#  define INLINE inline
+#elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100)
+#  define INLINE __inline
+#elif defined(__SUNPRO_C) && defined(__C99FEATURES__)
+#  define INLINE inline
+#  define __inline inline
+#  define __inline__ inline
+#elif (__STDC_VERSION__ >= 199901L) /* C99 */
+#  define INLINE inline
+#else
+#  define INLINE
+#endif
+
+
+#endif /* EGLCOMPILER_INCLUDED */
diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c
index 461679db09..9ab4286d3a 100644
--- a/src/egl/main/eglcontext.c
+++ b/src/egl/main/eglcontext.c
@@ -14,11 +14,9 @@
  * in the attrib_list.
  */
 EGLBoolean
-_eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
-                EGLConfig config, const EGLint *attrib_list)
+_eglInitContext(_EGLDriver *drv, _EGLContext *ctx,
+                _EGLConfig *conf, const EGLint *attrib_list)
 {
-   _EGLConfig *conf;
-   _EGLDisplay *display = _eglLookupDisplay(dpy);
    EGLint i;
    const EGLenum api = eglQueryAPI();
 
@@ -27,7 +25,6 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
       return EGL_FALSE;
    }
 
-   conf = _eglLookupConfig(drv, dpy, config);
    if (!conf) {
       _eglError(EGL_BAD_CONFIG, "_eglInitContext");
       return EGL_FALSE;
@@ -49,7 +46,6 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
       }
    }
 
-   ctx->Display = display;
    ctx->Config = conf;
    ctx->DrawSurface = EGL_NO_SURFACE;
    ctx->ReadSurface = EGL_NO_SURFACE;
@@ -60,66 +56,6 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
 
 
 /**
- * Save a new _EGLContext into the hash table.
- */
-void
-_eglSaveContext(_EGLContext *ctx)
-{
-   /* no-op.
-    * Public EGLContext handle and private _EGLContext are the same.
-    */
-}
-
-
-/**
- * Remove the given _EGLContext object from the hash table.
- */
-void
-_eglRemoveContext(_EGLContext *ctx)
-{
-   /* no-op.
-    * Public EGLContext handle and private _EGLContext are the same.
-    */
-}
-
-
-/**
- * Return the public handle for the given private context ptr.
- * This is the inverse of _eglLookupContext().
- */
-EGLContext
-_eglGetContextHandle(_EGLContext *ctx)
-{
-   /* just a cast! */
-   return (EGLContext) ctx;
-}
-
-
-/**
- * Return the _EGLContext object that corresponds to the given
- * EGLContext handle.
- * This is the inverse of _eglGetContextHandle().
- */
-_EGLContext *
-_eglLookupContext(EGLContext ctx)
-{
-   /* just a cast since EGLContext is just a void ptr */
-   return (_EGLContext *) ctx;
-}
-
-
-/**
- * Return the currently bound _EGLContext object, or NULL.
- */
-_EGLContext *
-_eglGetCurrentContext(void)
-{
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   return t->CurrentContext;
-}
-
-
-/**
  * Just a placeholder/demo function.  Real driver will never use this!
  */
 EGLContext
@@ -128,18 +64,24 @@ _eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 {
 #if 0 /* example code */
    _EGLContext *context;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   if (!conf) {
+      _eglError(EGL_BAD_CONFIG, "eglCreateContext");
+      return EGL_NO_CONTEXT;
+   }
 
    context = (_EGLContext *) calloc(1, sizeof(_EGLContext));
    if (!context)
       return EGL_NO_CONTEXT;
 
-   if (!_eglInitContext(drv, dpy, context, config, attrib_list)) {
+   if (!_eglInitContext(drv, context, conf, attrib_list)) {
       free(context);
       return EGL_NO_CONTEXT;
    }
 
-   _eglSaveContext(context);
-   return (EGLContext) context;
+   return _eglLinkContext(context, _eglLookupDisplay(dpy));
 #endif
    return EGL_NO_CONTEXT;
 }
@@ -153,12 +95,9 @@ _eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
 {
    _EGLContext *context = _eglLookupContext(ctx);
    if (context) {
-      if (context->IsBound) {
-         context->DeletePending = EGL_TRUE;
-      }
-      else {
+      _eglUnlinkContext(context);
+      if (!context->IsBound)
          free(context);
-      }
       return EGL_TRUE;
    }
    else {
@@ -203,7 +142,7 @@ _eglQueryContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx,
 
 /**
  * Drivers will typically call this to do the error checking and
- * update the various IsBound and DeletePending flags.
+ * update the various flags.
  * Then, the driver will do its device-dependent Make-Current stuff.
  */
 EGLBoolean
@@ -214,10 +153,13 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
    _EGLContext *ctx = _eglLookupContext(context);
    _EGLSurface *draw = _eglLookupSurface(d);
    _EGLSurface *read = _eglLookupSurface(r);
+   _EGLContext *oldContext = NULL;
+   _EGLSurface *oldDrawSurface = NULL;
+   _EGLSurface *oldReadSurface = NULL;
+   EGLint apiIndex;
 
-   _EGLContext *oldContext = _eglGetCurrentContext();
-   _EGLSurface *oldDrawSurface = _eglGetCurrentSurface(EGL_DRAW);
-   _EGLSurface *oldReadSurface = _eglGetCurrentSurface(EGL_READ);
+   if (_eglIsCurrentThreadDummy())
+      return _eglError(EGL_BAD_ALLOC, "eglMakeCurrent");
 
    /* error checking */
    if (ctx) {
@@ -233,6 +175,32 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
          _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
          return EGL_FALSE;
       }
+
+#ifdef EGL_VERSION_1_4
+      /* OpenGL and OpenGL ES are conflicting */
+      switch (ctx->ClientAPI) {
+      case EGL_OPENGL_ES_API:
+         if (t->CurrentContexts[_eglConvertApiToIndex(EGL_OPENGL_API)])
+            return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
+         break;
+      case EGL_OPENGL_API:
+         if (t->CurrentContexts[_eglConvertApiToIndex(EGL_OPENGL_ES_API)])
+            return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
+         break;
+      default:
+         break;
+      }
+#endif
+      apiIndex = _eglConvertApiToIndex(ctx->ClientAPI);
+   }
+   else {
+      apiIndex = t->CurrentAPIIndex;
+   }
+
+   oldContext = t->CurrentContexts[apiIndex];
+   if (oldContext) {
+      oldDrawSurface = oldContext->DrawSurface;
+      oldReadSurface = oldContext->ReadSurface;
    }
 
    /*
@@ -240,7 +208,7 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
     */
    if (oldDrawSurface != NULL) {
       oldDrawSurface->IsBound = EGL_FALSE;
-      if (oldDrawSurface->DeletePending) {
+      if (!_eglIsSurfaceLinked(oldDrawSurface)) {
          /* make sure we don't try to rebind a deleted surface */
          if (draw == oldDrawSurface || draw == oldReadSurface) {
             draw = NULL;
@@ -251,7 +219,7 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
    }
    if (oldReadSurface != NULL && oldReadSurface != oldDrawSurface) {
       oldReadSurface->IsBound = EGL_FALSE;
-      if (oldReadSurface->DeletePending) {
+      if (!_eglIsSurfaceLinked(oldReadSurface)) {
          /* make sure we don't try to rebind a deleted surface */
          if (read == oldDrawSurface || read == oldReadSurface) {
             read = NULL;
@@ -262,7 +230,7 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
    }
    if (oldContext != NULL) {
       oldContext->IsBound = EGL_FALSE;
-      if (oldContext->DeletePending) {
+      if (!_eglIsContextLinked(oldContext)) {
          /* make sure we don't try to rebind a deleted context */
          if (ctx == oldContext) {
             ctx = NULL;
@@ -283,9 +251,11 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
       ctx->IsBound = EGL_TRUE;
       draw->IsBound = EGL_TRUE;
       read->IsBound = EGL_TRUE;
+      t->CurrentContexts[apiIndex] = ctx;
+   }
+   else {
+      t->CurrentContexts[apiIndex] = NULL;
    }
-
-   t->CurrentContext = ctx;
 
    return EGL_TRUE;
 }
diff --git a/src/egl/main/eglcontext.h b/src/egl/main/eglcontext.h
index 34fee9c637..2fb28d38b9 100644
--- a/src/egl/main/eglcontext.h
+++ b/src/egl/main/eglcontext.h
@@ -11,7 +11,9 @@
  */
 struct _egl_context
 {
-   _EGLDisplay *Display; /* who do I belong to? */
+   /* Managed by EGLDisplay for linking */
+   _EGLDisplay *Display;
+   _EGLContext *Next;
 
    _EGLConfig *Config;
 
@@ -19,7 +21,6 @@ struct _egl_context
    _EGLSurface *ReadSurface;
 
    EGLBoolean IsBound;
-   EGLBoolean DeletePending;
 
    EGLint ClientAPI; /**< EGL_OPENGL_ES_API, EGL_OPENGL_API, EGL_OPENVG_API */
    EGLint ClientVersion; /**< 1 = OpenGLES 1.x, 2 = OpenGLES 2.x */
@@ -27,28 +28,8 @@ struct _egl_context
 
 
 extern EGLBoolean
-_eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
-                EGLConfig config, const EGLint *attrib_list);
-
-
-extern void
-_eglSaveContext(_EGLContext *ctx);
-
-
-extern void
-_eglRemoveContext(_EGLContext *ctx);
-
-
-extern EGLContext
-_eglGetContextHandle(_EGLContext *ctx);
-
-
-extern _EGLContext *
-_eglLookupContext(EGLContext ctx);
- 
-
-extern _EGLContext *
-_eglGetCurrentContext(void);
+_eglInitContext(_EGLDriver *drv, _EGLContext *ctx,
+                _EGLConfig *config, const EGLint *attrib_list);
 
 
 extern EGLContext
diff --git a/src/egl/main/eglcurrent.c b/src/egl/main/eglcurrent.c
new file mode 100644
index 0000000000..96152db19f
--- /dev/null
+++ b/src/egl/main/eglcurrent.c
@@ -0,0 +1,283 @@
+#include <stdlib.h>
+#include <string.h>
+#include "eglcurrent.h"
+#include "eglcontext.h"
+#include "egllog.h"
+
+
+/* a fallback thread info to guarantee that every thread always has one */
+static _EGLThreadInfo dummy_thread;
+
+
+#ifdef GLX_USE_TLS
+static __thread const _EGLThreadInfo *_egl_TSD;
+   __attribute__ ((tls_model("initial-exec")));
+
+static INLINE EGLBoolean _eglInitTSD(void) { return EGL_TRUE; }
+static INLINE void _eglFiniTSD(void) { }
+static INLINE void _eglSetTSD(const _EGLThreadInfo *t) { _egl_TSD = t; }
+
+static INLINE _EGLThreadInfo *_eglGetTSD(void)
+{
+   return (_EGLThreadInfo *) _egl_TSD;
+}
+
+#elif PTHREADS
+#include <pthread.h>
+
+static pthread_key_t _egl_TSD;
+
+static INLINE EGLBoolean _eglInitTSD(void)
+{
+   return (pthread_key_create(&_egl_TSD, NULL) == 0);
+}
+
+static INLINE void _eglFiniTSD(void)
+{
+   pthread_key_delete(_egl_TSD);
+}
+
+static INLINE void _eglSetTSD(const _EGLThreadInfo *t)
+{
+   pthread_setspecific(_egl_TSD, (const void *) t);
+}
+
+static INLINE _EGLThreadInfo *_eglGetTSD(void)
+{
+   return (_EGLThreadInfo *) pthread_getspecific(_egl_TSD);
+}
+
+#else /* PTHREADS */
+static const _EGLThreadInfo *_egl_TSD;
+
+static INLINE EGLBoolean _eglInitTSD(void) { return EGL_TRUE; }
+static INLINE void _eglFiniTSD(void) { }
+static INLINE void _eglSetTSD(const _EGLThreadInfo *t) { _egl_TSD = t; }
+
+static INLINE _EGLThreadInfo *_eglGetTSD(void)
+{
+   return (_EGLThreadInfo *) _egl_TSD;
+}
+#endif /* !PTHREADS */
+
+
+static void
+_eglInitThreadInfo(_EGLThreadInfo *t)
+{
+   memset(t, 0, sizeof(*t));
+   t->LastError = EGL_SUCCESS;
+   /* default, per EGL spec */
+   t->CurrentAPIIndex = _eglConvertApiToIndex(EGL_OPENGL_ES_API);
+}
+
+
+/**
+ * Allocate and init a new _EGLThreadInfo object.
+ */
+static _EGLThreadInfo *
+_eglCreateThreadInfo(void)
+{
+   _EGLThreadInfo *t = (_EGLThreadInfo *) calloc(1, sizeof(_EGLThreadInfo));
+   if (t)
+      _eglInitThreadInfo(t);
+   else
+      t = &dummy_thread;
+   return t;
+}
+
+
+/**
+ * Delete/free a _EGLThreadInfo object.
+ */
+static void
+_eglDestroyThreadInfo(_EGLThreadInfo *t)
+{
+   if (t != &dummy_thread)
+      free(t);
+}
+
+
+/**
+ * Initialize "current thread" management.
+ */
+EGLBoolean
+_eglInitCurrent(void)
+{
+   _eglInitThreadInfo(&dummy_thread);
+   return _eglInitTSD();
+}
+
+
+/**
+ * Finish "current thread" management.
+ */
+void
+_eglFiniCurrent(void)
+{
+   /* TODO trace and release all threads... */
+   _eglFiniTSD();
+}
+
+
+/**
+ * Return the calling thread's thread info.
+ * If the calling thread nevers calls this function before, or if its thread
+ * info was destroyed, a new one is created.  This function never returns NULL.
+ * In the case allocation fails, a dummy one is returned.  See also
+ * _eglIsCurrentThreadDummy.
+ */
+_EGLThreadInfo *
+_eglGetCurrentThread(void)
+{
+   _EGLThreadInfo *t = _eglGetTSD();
+   if (!t) {
+      t = _eglCreateThreadInfo();
+      _eglSetTSD(t);
+   }
+
+   return t;
+}
+
+
+/**
+ * Destroy the calling thread's thread info.
+ */
+void
+_eglDestroyCurrentThread(void)
+{
+   _EGLThreadInfo *t = _eglGetTSD();
+   if (t) {
+      _eglDestroyThreadInfo(t);
+      _eglSetTSD(NULL);
+   }
+}
+
+
+/**
+ * Return true if the calling thread's thread info is dummy.
+ * A dummy thread info is shared by all threads and should not be modified.
+ * Functions like eglBindAPI or eglMakeCurrent should check for dummy-ness
+ * before updating the thread info.
+ */
+EGLBoolean
+_eglIsCurrentThreadDummy(void)
+{
+   _EGLThreadInfo *t = _eglGetTSD();
+   return (!t || t == &dummy_thread);
+}
+
+
+/**
+ * Return the currently bound context, or NULL.
+ */
+_EGLContext *
+_eglGetCurrentContext(void)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   return t->CurrentContexts[t->CurrentAPIIndex];
+}
+
+
+/**
+ * Return the display of the currently bound context, or NULL.
+ */
+_EGLDisplay *
+_eglGetCurrentDisplay(void)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   _EGLContext *ctx = t->CurrentContexts[t->CurrentAPIIndex];
+   if (ctx)
+      return ctx->Display;
+   else
+      return NULL;
+}
+
+
+/**
+ * Return the read or write surface of the currently bound context, or NULL.
+ */
+_EGLSurface *
+_eglGetCurrentSurface(EGLint readdraw)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   _EGLContext *ctx = t->CurrentContexts[t->CurrentAPIIndex];
+   if (ctx) {
+      switch (readdraw) {
+      case EGL_DRAW:
+         return ctx->DrawSurface;
+      case EGL_READ:
+         return ctx->ReadSurface;
+      default:
+         return NULL;
+      }
+   }
+   return NULL;
+}
+
+
+/**
+ * Record EGL error code.
+ */
+EGLBoolean
+_eglError(EGLint errCode, const char *msg)
+{
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   const char *s;
+
+   if (t == &dummy_thread)
+      return EGL_FALSE;
+
+   if (t->LastError == EGL_SUCCESS) {
+      t->LastError = errCode;
+
+      switch (errCode) {
+      case EGL_BAD_ACCESS:
+         s = "EGL_BAD_ACCESS";
+         break;
+      case EGL_BAD_ALLOC:
+         s = "EGL_BAD_ALLOC";
+         break;
+      case EGL_BAD_ATTRIBUTE:
+         s = "EGL_BAD_ATTRIBUTE";
+         break;
+      case EGL_BAD_CONFIG:
+         s = "EGL_BAD_CONFIG";
+         break;
+      case EGL_BAD_CONTEXT:
+         s = "EGL_BAD_CONTEXT";
+         break;
+      case EGL_BAD_CURRENT_SURFACE:
+         s = "EGL_BAD_CURRENT_SURFACE";
+         break;
+      case EGL_BAD_DISPLAY:
+         s = "EGL_BAD_DISPLAY";
+         break;
+      case EGL_BAD_MATCH:
+         s = "EGL_BAD_MATCH";
+         break;
+      case EGL_BAD_NATIVE_PIXMAP:
+         s = "EGL_BAD_NATIVE_PIXMAP";
+         break;
+      case EGL_BAD_NATIVE_WINDOW:
+         s = "EGL_BAD_NATIVE_WINDOW";
+         break;
+      case EGL_BAD_PARAMETER:
+         s = "EGL_BAD_PARAMETER";
+         break;
+      case EGL_BAD_SURFACE:
+         s = "EGL_BAD_SURFACE";
+         break;
+      case EGL_BAD_SCREEN_MESA:
+         s = "EGL_BAD_SCREEN_MESA";
+         break;
+      case EGL_BAD_MODE_MESA:
+         s = "EGL_BAD_MODE_MESA";
+         break;
+      default:
+         s = "other";
+      }
+      _eglLog(_EGL_DEBUG, "EGL user error 0x%x (%s) in %s\n", errCode, s, msg);
+   }
+
+   return EGL_FALSE;
+}
diff --git a/src/egl/main/eglcurrent.h b/src/egl/main/eglcurrent.h
new file mode 100644
index 0000000000..f9fdf7bd0f
--- /dev/null
+++ b/src/egl/main/eglcurrent.h
@@ -0,0 +1,92 @@
+#ifndef EGLCURRENT_INCLUDED
+#define EGLCURRENT_INCLUDED
+
+#include "egltypedefs.h"
+
+
+#define _EGL_API_NUM_INDICES \
+   (EGL_OPENGL_API - EGL_OPENGL_ES_API + 2) /* idx 0 is for EGL_NONE */
+
+
+/**
+ * Per-thread info
+ */
+struct _egl_thread_info
+{
+   EGLint LastError;
+   _EGLContext *CurrentContexts[_EGL_API_NUM_INDICES];
+   /* use index for fast access to current context */
+   EGLint CurrentAPIIndex;
+};
+
+
+extern EGLBoolean
+_eglInitCurrent(void);
+
+
+extern void
+_eglFiniCurrent(void);
+
+
+/**
+ * Return true if a client API enum can be converted to an index.
+ */
+static INLINE EGLBoolean
+_eglIsApiValid(EGLenum api)
+{
+   return ((api >= EGL_OPENGL_ES_API && api <= EGL_OPENGL_API) ||
+           api == EGL_NONE);
+}
+
+
+/**
+ * Convert a client API enum to an index, for use by thread info.
+ * The client API enum is assumed to be valid.
+ */
+static INLINE EGLint
+_eglConvertApiToIndex(EGLenum api)
+{
+   return (api != EGL_NONE) ? api - EGL_OPENGL_ES_API + 1 : 0;
+}
+
+
+/**
+ * Convert an index, used by thread info, to a client API enum.
+ * The index is assumed to be valid.
+ */
+static INLINE EGLenum
+_eglConvertApiFromIndex(EGLint idx)
+{
+   return (idx) ? EGL_OPENGL_ES_API + idx - 1 : EGL_NONE;
+}
+
+
+extern _EGLThreadInfo *
+_eglGetCurrentThread(void);
+
+
+extern void
+_eglDestroyCurrentThread(void);
+
+
+extern EGLBoolean
+_eglIsCurrentThreadDummy(void);
+
+
+extern _EGLContext *
+_eglGetCurrentContext(void);
+
+
+extern _EGLDisplay *
+_eglGetCurrentDisplay(void);
+
+
+extern _EGLSurface *
+_eglGetCurrentSurface(EGLint readdraw);
+
+
+extern EGLBoolean
+_eglError(EGLint errCode, const char *msg);
+
+
+#endif /* EGLCURRENT_INCLUDED */
diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c
index 47a2323eaf..89de609d0b 100644
--- a/src/egl/main/egldisplay.c
+++ b/src/egl/main/egldisplay.c
@@ -7,6 +7,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include "eglcontext.h"
+#include "eglsurface.h"
 #include "egldisplay.h"
 #include "egldriver.h"
 #include "eglglobals.h"
@@ -25,11 +26,6 @@ _eglNewDisplay(NativeDisplayType nativeDisplay)
 {
    _EGLDisplay *dpy = (_EGLDisplay *) calloc(1, sizeof(_EGLDisplay));
    if (dpy) {
-      EGLuint key = _eglHashGenKey(_eglGlobal.Displays);
-
-      dpy->Handle = (EGLDisplay) key;
-      _eglHashInsert(_eglGlobal.Displays, key, dpy);
-
       dpy->NativeDisplay = nativeDisplay;
 #if defined(_EGL_PLATFORM_X)
       dpy->Xdpy = (Display *) nativeDisplay;
@@ -46,8 +42,37 @@ _eglNewDisplay(NativeDisplayType nativeDisplay)
 
 
 /**
- * Return the public handle for an internal _EGLDisplay.
- * This is the inverse of _eglLookupDisplay().
+ * Link a display to itself and return the handle of the link.
+ * The handle can be passed to client directly.
+ */
+EGLDisplay
+_eglLinkDisplay(_EGLDisplay *dpy)
+{
+   EGLuint key;
+   key = _eglHashGenKey(_eglGlobal.Displays);
+   assert(key);
+   /* "link" the display to the hash table */
+   _eglHashInsert(_eglGlobal.Displays, key, dpy);
+   dpy->Handle = (EGLDisplay) key;
+
+   return dpy->Handle;
+}
+
+
+/**
+ * Unlink a linked display from itself.
+ * Accessing an unlinked display should generate EGL_BAD_DISPLAY error.
+ */
+void
+_eglUnlinkDisplay(_EGLDisplay *dpy)
+{
+   _eglHashRemove(_eglGlobal.Displays, (EGLuint) dpy->Handle);
+   dpy->Handle = EGL_NO_DISPLAY;
+}
+
+
+/**
+ * Return the handle of a linked display, or EGL_NO_DISPLAY.
  */
 EGLDisplay
 _eglGetDisplayHandle(_EGLDisplay *display)
@@ -60,40 +85,70 @@ _eglGetDisplayHandle(_EGLDisplay *display)
 
  
 /**
- * Return the _EGLDisplay object that corresponds to the given public/
- * opaque display handle.
- * This is the inverse of _eglGetDisplayHandle().
+ * Lookup a handle to find the linked display.
+ * Return NULL if the handle has no corresponding linked display.
  */
 _EGLDisplay *
 _eglLookupDisplay(EGLDisplay dpy)
 {
    EGLuint key = (EGLuint) dpy;
-   if (!_eglGlobal.Displays)
-      return NULL;
    return (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, key);
 }
 
 
-void
-_eglSaveDisplay(_EGLDisplay *dpy)
+/**
+ * Find the display corresponding to the specified native display id in all
+ * linked displays.
+ */
+_EGLDisplay *
+_eglFindDisplay(NativeDisplayType nativeDisplay)
 {
-   EGLuint key = _eglHashGenKey(_eglGlobal.Displays);
-   assert(dpy);
-   assert(!dpy->Handle);
-   dpy->Handle = (EGLDisplay) key;
-   assert(dpy->Handle);
-   _eglHashInsert(_eglGlobal.Displays, key, dpy);
+   EGLuint key = _eglHashFirstEntry(_eglGlobal.Displays);
+
+   /* Walk the hash table.  Should switch to list if it is a problem. */
+   while (key) {
+      _EGLDisplay *dpy = (_EGLDisplay *)
+            _eglHashLookup(_eglGlobal.Displays, key);
+      assert(dpy);
+
+      if (dpy->NativeDisplay == nativeDisplay)
+         return dpy;
+      key = _eglHashNextEntry(_eglGlobal.Displays, key);
+   }
+
+   return NULL;
 }
 
 
-_EGLDisplay *
-_eglGetCurrentDisplay(void)
+/**
+ * Destroy the contexts and surfaces that are linked to the display.
+ */
+void
+_eglReleaseDisplayResources(_EGLDriver *drv, EGLDisplay dpy)
 {
-   _EGLContext *ctx = _eglGetCurrentContext();
-   if (ctx)
-      return ctx->Display;
-   else
-      return NULL;
+   _EGLDisplay *display;
+   _EGLContext *contexts;
+   _EGLSurface *surfaces;
+
+   display = _eglLookupDisplay(dpy);
+   if (!display)
+      return;
+   contexts = display->ContextList;
+   surfaces = display->SurfaceList;
+
+   while (contexts) {
+      EGLContext handle = _eglGetContextHandle(contexts);
+      contexts = contexts->Next;
+      drv->API.DestroyContext(drv, dpy, handle);
+   }
+   assert(!display->ContextList);
+
+   while (surfaces) {
+      EGLSurface handle = _eglGetSurfaceHandle(surfaces);
+      surfaces = surfaces->Next;
+      drv->API.DestroySurface(drv, dpy, handle);
+   }
+   assert(!display->SurfaceList);
 }
 
 
@@ -119,3 +174,147 @@ _eglCleanupDisplay(_EGLDisplay *disp)
 
    /* driver deletes the _EGLDisplay object */
 }
+
+
+/**
+ * Link a context to a display and return the handle of the link.
+ * The handle can be passed to client directly.
+ */
+EGLContext
+_eglLinkContext(_EGLContext *ctx, _EGLDisplay *dpy)
+{
+   ctx->Display = dpy;
+   ctx->Next = dpy->ContextList;
+   dpy->ContextList = ctx;
+   return (EGLContext) ctx;
+}
+
+
+/**
+ * Unlink a linked context from its display.
+ * Accessing an unlinked context should generate EGL_BAD_CONTEXT error.
+ */
+void
+_eglUnlinkContext(_EGLContext *ctx)
+{
+   _EGLContext *prev;
+
+   prev = ctx->Display->ContextList;
+   if (prev != ctx) {
+      while (prev) {
+         if (prev->Next == ctx)
+            break;
+         prev = prev->Next;
+      }
+      assert(prev);
+      prev->Next = ctx->Next;
+   }
+   else {
+      ctx->Display->ContextList = ctx->Next;
+   }
+
+   ctx->Next = NULL;
+   ctx->Display = NULL;
+}
+
+
+/**
+ * Return the handle of a linked context, or EGL_NO_CONTEXT.
+ */
+EGLContext
+_eglGetContextHandle(_EGLContext *ctx)
+{
+   return (EGLContext) (ctx && ctx->Display) ? ctx : EGL_NO_CONTEXT;
+}
+
+
+/**
+ * Lookup a handle to find the linked context.
+ * Return NULL if the handle has no corresponding linked context.
+ */
+_EGLContext *
+_eglLookupContext(EGLContext ctx)
+{
+   _EGLContext *context = (_EGLContext *) ctx;
+   return (context && context->Display) ? context : NULL;
+}
+
+
+/**
+ * Link a surface to a display and return the handle of the link.
+ * The handle can be passed to client directly.
+ */
+EGLSurface
+_eglLinkSurface(_EGLSurface *surf, _EGLDisplay *dpy)
+{
+   EGLuint key;
+
+   surf->Display = dpy;
+   surf->Next = dpy->SurfaceList;
+   dpy->SurfaceList = surf;
+
+   key = _eglHashGenKey(_eglGlobal.Surfaces);
+   assert(key);
+   _eglHashInsert(_eglGlobal.Surfaces, key, surf);
+
+   surf->Handle = (EGLSurface) key;
+   return surf->Handle;
+}
+
+
+/**
+ * Unlink a linked surface from its display.
+ * Accessing an unlinked surface should generate EGL_BAD_SURFACE error.
+ */
+void
+_eglUnlinkSurface(_EGLSurface *surf)
+{
+   _EGLSurface *prev;
+
+   _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surf->Handle);
+   surf->Handle = EGL_NO_SURFACE;
+
+   prev = surf->Display->SurfaceList;
+   if (prev != surf) {
+      while (prev) {
+         if (prev->Next == surf)
+            break;
+         prev = prev->Next;
+      }
+      assert(prev);
+      prev->Next = surf->Next;
+   }
+   else {
+      prev = NULL;
+      surf->Display->SurfaceList = surf->Next;
+   }
+
+   surf->Next = NULL;
+   surf->Display = NULL;
+}
+
+
+/**
+ * Return the handle of a linked surface, or EGL_NO_SURFACE.
+ */
+EGLSurface
+_eglGetSurfaceHandle(_EGLSurface *surface)
+{
+   if (surface)
+      return surface->Handle;
+   else
+      return EGL_NO_SURFACE;
+}
+
+
+/**
+ * Lookup a handle to find the linked surface.
+ * Return NULL if the handle has no corresponding linked surface.
+ */
+_EGLSurface *
+_eglLookupSurface(EGLSurface surf)
+{
+   _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces,
+                                                   (EGLuint) surf);
+   return c;
+}
diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h
index ff623ee1c6..372ed3cd79 100644
--- a/src/egl/main/egldisplay.h
+++ b/src/egl/main/egldisplay.h
@@ -23,6 +23,9 @@ struct _egl_display
    EGLint NumConfigs;
    _EGLConfig **Configs;  /* array [NumConfigs] of ptr to _EGLConfig */
 
+   /* lists of linked contexts and surface */
+   _EGLContext *ContextList;
+   _EGLSurface *SurfaceList;
 #ifdef _EGL_PLATFORM_X
    Display *Xdpy;
 #endif
@@ -33,7 +36,15 @@ extern _EGLDisplay *
 _eglNewDisplay(NativeDisplayType displayName);
 
 
-EGLDisplay
+extern EGLDisplay
+_eglLinkDisplay(_EGLDisplay *dpy);
+
+
+extern void
+_eglUnlinkDisplay(_EGLDisplay *dpy);
+
+
+extern EGLDisplay
 _eglGetDisplayHandle(_EGLDisplay *display);
 
 
@@ -41,21 +52,77 @@ extern _EGLDisplay *
 _eglLookupDisplay(EGLDisplay dpy);
 
 
-extern void
-_eglSaveDisplay(_EGLDisplay *dpy);
+/**
+ * Return true if the display is linked.
+ */
+static INLINE EGLBoolean
+_eglIsDisplayLinked(_EGLDisplay *dpy)
+{
+   return (EGLBoolean) (_eglGetDisplayHandle(dpy) != EGL_NO_DISPLAY);
+}
 
 
 extern _EGLDisplay *
-_eglGetCurrentDisplay(void);
+_eglFindDisplay(NativeDisplayType nativeDisplay);
+
+
+extern void
+_eglReleaseDisplayResources(_EGLDriver *drv, EGLDisplay dpy);
 
 
 extern void
 _eglCleanupDisplay(_EGLDisplay *disp);
 
 
-extern EGLBoolean 
-_eglQueryDisplayMESA(_EGLDriver *drv, EGLDisplay dpy, EGLint attrib, EGLint *value);
+extern EGLContext
+_eglLinkContext(_EGLContext *ctx, _EGLDisplay *dpy);
+
+
+extern void
+_eglUnlinkContext(_EGLContext *ctx);
+
+
+extern EGLContext
+_eglGetContextHandle(_EGLContext *ctx);
 
 
+extern _EGLContext *
+_eglLookupContext(EGLContext ctx);
+
+
+/**
+ * Return true if the context is linked to a display.
+ */
+static INLINE EGLBoolean
+_eglIsContextLinked(_EGLContext *ctx)
+{
+   return (EGLBoolean) (_eglGetContextHandle(ctx) != EGL_NO_CONTEXT);
+}
+
+extern EGLSurface
+_eglLinkSurface(_EGLSurface *surf, _EGLDisplay *dpy);
+
+
+extern void
+_eglUnlinkSurface(_EGLSurface *surf);
+
+
+extern EGLSurface
+_eglGetSurfaceHandle(_EGLSurface *);
+
+
+extern _EGLSurface *
+_eglLookupSurface(EGLSurface surf);
+
+
+/**
+ * Return true if the surface is linked to a display.
+ */
+static INLINE EGLBoolean
+_eglIsSurfaceLinked(_EGLSurface *surf)
+{
+   return (EGLBoolean) (_eglGetSurfaceHandle(surf) != EGL_NO_SURFACE);
+}
+
 
 #endif /* EGLDISPLAY_INCLUDED */
diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c
index 43b1f51903..f2a864cd8a 100644
--- a/src/egl/main/egldriver.c
+++ b/src/egl/main/egldriver.c
@@ -284,9 +284,7 @@ _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy)
 
    _eglLog(_EGL_DEBUG, "Closing %s", drv->Name);
 
-   /*
-    * XXX check for currently bound context/surfaces and delete them?
-    */
+   _eglReleaseDisplayResources(drv, dpy);
 
    b = drv->API.Terminate(drv, dpy);
 
diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c
index b770e55dbd..55de394ef5 100644
--- a/src/egl/main/eglglobals.c
+++ b/src/egl/main/eglglobals.c
@@ -1,6 +1,6 @@
-#include <stdio.h>
 #include <stdlib.h>
 #include "eglglobals.h"
+#include "egllog.h"
 
 struct _egl_global _eglGlobal = 
 {
@@ -22,8 +22,8 @@ _eglInitGlobals(void)
 
       _eglGlobal.ClientAPIsMask = 0x0;
 
-      /* XXX temporary */
-      _eglGlobal.ThreadInfo = _eglNewThreadInfo();
+      if (!_eglInitCurrent())
+         _eglLog(_EGL_FATAL, "failed to initialize \"current\" system");
    }
 }
 
@@ -34,113 +34,8 @@ _eglInitGlobals(void)
 void
 _eglDestroyGlobals(void)
 {
+   _eglFiniCurrent();
    /* XXX TODO walk over table entries, deleting each */
    _eglDeleteHashTable(_eglGlobal.Displays);
    _eglDeleteHashTable(_eglGlobal.Surfaces);
 }
-
-
-/**
- * Allocate and init a new _EGLThreadInfo object.
- */
-_EGLThreadInfo *
-_eglNewThreadInfo(void)
-{
-   _EGLThreadInfo *t = (_EGLThreadInfo *) calloc(1, sizeof(_EGLThreadInfo));
-   if (t) {
-      t->CurrentContext = EGL_NO_CONTEXT;
-      t->LastError = EGL_SUCCESS;
-      t->CurrentAPI = EGL_OPENGL_ES_API;  /* default, per EGL spec */
-   }
-   return t;
-}
-
-
-/**
- * Delete/free a _EGLThreadInfo object.
- */
-void
-_eglDeleteThreadData(_EGLThreadInfo *t)
-{
-   free(t);
-}
-
-
-
-/**
- * Return pointer to calling thread's _EGLThreadInfo object.
- * Create a new one if needed.
- * Should never return NULL.
- */
-_EGLThreadInfo *
-_eglGetCurrentThread(void)
-{
-   _eglInitGlobals();
-
-   /* XXX temporary */
-   return _eglGlobal.ThreadInfo;
-}
-
-
-/**
- * Record EGL error code.
- */
-void
-_eglError(EGLint errCode, const char *msg)
-{
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   const char *s;
-
-   if (t->LastError == EGL_SUCCESS) {
-      t->LastError = errCode;
-
-      switch (errCode) {
-      case EGL_BAD_ACCESS:
-         s = "EGL_BAD_ACCESS";
-         break;
-      case EGL_BAD_ALLOC:
-         s = "EGL_BAD_ALLOC";
-         break;
-      case EGL_BAD_ATTRIBUTE:
-         s = "EGL_BAD_ATTRIBUTE";
-         break;
-      case EGL_BAD_CONFIG:
-         s = "EGL_BAD_CONFIG";
-         break;
-      case EGL_BAD_CONTEXT:
-         s = "EGL_BAD_CONTEXT";
-         break;
-      case EGL_BAD_CURRENT_SURFACE:
-         s = "EGL_BAD_CURRENT_SURFACE";
-         break;
-      case EGL_BAD_DISPLAY:
-         s = "EGL_BAD_DISPLAY";
-         break;
-      case EGL_BAD_MATCH:
-         s = "EGL_BAD_MATCH";
-         break;
-      case EGL_BAD_NATIVE_PIXMAP:
-         s = "EGL_BAD_NATIVE_PIXMAP";
-         break;
-      case EGL_BAD_NATIVE_WINDOW:
-         s = "EGL_BAD_NATIVE_WINDOW";
-         break;
-      case EGL_BAD_PARAMETER:
-         s = "EGL_BAD_PARAMETER";
-         break;
-      case EGL_BAD_SURFACE:
-         s = "EGL_BAD_SURFACE";
-         break;
-      case EGL_BAD_SCREEN_MESA:
-         s = "EGL_BAD_SCREEN_MESA";
-         break;
-      case EGL_BAD_MODE_MESA:
-         s = "EGL_BAD_MODE_MESA";
-         break;
-      default:
-         s = "other";
-      }
-      /* XXX temporary */
-      fprintf(stderr, "EGL user error 0x%x (%s) in %s\n", errCode, s, msg);
-   }
-}
diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h
index 14d8ea487a..a9443cfbdd 100644
--- a/src/egl/main/eglglobals.h
+++ b/src/egl/main/eglglobals.h
@@ -3,17 +3,7 @@
 
 #include "egltypedefs.h"
 #include "eglhash.h"
-
-
-/**
- * Per-thread info
- */
-struct _egl_thread_info
-{
-   EGLint LastError;
-   _EGLContext *CurrentContext;
-   EGLenum CurrentAPI;
-};
+#include "eglcurrent.h"
 
 
 /**
@@ -23,6 +13,7 @@ struct _egl_global
 {
    EGLBoolean Initialized;
 
+   /* these are private to egldisplay.c */
    _EGLHashtable *Displays;
    _EGLHashtable *Surfaces;
 
@@ -33,9 +24,6 @@ struct _egl_global
 
    char ClientAPIs[1000];   /**< updated by eglQueryString */
 
-   /* XXX temporary - should be thread-specific data (TSD) */
-   _EGLThreadInfo *ThreadInfo;
-
    EGLint NumDrivers;
    _EGLDriver *Drivers[10];
 };
@@ -52,20 +40,4 @@ extern void
 _eglDestroyGlobals(void);
 
 
-extern _EGLThreadInfo *
-_eglNewThreadInfo(void);
-
-
-extern void
-_eglDeleteThreadData(_EGLThreadInfo *t);
-
-
-extern _EGLThreadInfo *
-_eglGetCurrentThread(void);
-
-
-extern void
-_eglError(EGLint errCode, const char *msg);
-
-
 #endif /* EGLGLOBALS_INCLUDED */
diff --git a/src/egl/main/eglscreen.c b/src/egl/main/eglscreen.c
index 9c9a8377bf..b6bde65e8b 100644
--- a/src/egl/main/eglscreen.c
+++ b/src/egl/main/eglscreen.c
@@ -128,20 +128,25 @@ _eglCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
    _EGLSurface *surf;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   if (!conf) {
+      _eglError(EGL_BAD_CONFIG, "eglCreateScreenSurfaceMESA");
+      return EGL_NO_SURFACE;
+   }
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_SCREEN_BIT_MESA,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_SCREEN_BIT_MESA,
+                        conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return _eglLinkSurface(surf, _eglLookupDisplay(dpy));
 #endif
    return EGL_NO_SURFACE;
 }
diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c
index 6905acac50..9821e63628 100644
--- a/src/egl/main/eglsurface.c
+++ b/src/egl/main/eglsurface.c
@@ -6,6 +6,7 @@
 #include <assert.h>
 #include <stdlib.h>
 #include <string.h>
+#include "egldisplay.h"
 #include "eglcontext.h"
 #include "eglconfig.h"
 #include "egldriver.h"
@@ -20,12 +21,10 @@
  * \return EGL_TRUE if no errors, EGL_FALSE otherwise.
  */
 EGLBoolean
-_eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
-                _EGLSurface *surf, EGLint type, EGLConfig config,
-                const EGLint *attrib_list)
+_eglInitSurface(_EGLDriver *drv, _EGLSurface *surf, EGLint type,
+                _EGLConfig *conf, const EGLint *attrib_list)
 {
    const char *func;
-   _EGLConfig *conf;
    EGLint width = 0, height = 0, largest = 0;
    EGLint texFormat = 0, texTarget = 0, mipmapTex = 0;
    EGLint renderBuffer = EGL_BACK_BUFFER;
@@ -55,7 +54,6 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
       return EGL_FALSE;
    }
 
-   conf = _eglLookupConfig(drv, dpy, config);
    if (!conf) {
       _eglError(EGL_BAD_CONFIG, func);
       return EGL_FALSE;
@@ -211,72 +209,6 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
 }
 
 
-void
-_eglSaveSurface(_EGLSurface *surf)
-{
-   EGLuint key = _eglHashGenKey(_eglGlobal.Surfaces);
-   assert(surf);
-   assert(!surf->Handle);
-   surf->Handle = (EGLSurface) key;
-   assert(surf->Handle);
-   _eglHashInsert(_eglGlobal.Surfaces, key, surf);
-}
-
-
-void
-_eglRemoveSurface(_EGLSurface *surf)
-{
-   _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surf->Handle);
-}
-
-
-
-/**
- * Return the public handle for an internal _EGLSurface.
- * This is the inverse of _eglLookupSurface().
- */
-EGLSurface
-_eglGetSurfaceHandle(_EGLSurface *surface)
-{
-   if (surface)
-      return surface->Handle;
-   else
-      return EGL_NO_SURFACE;
-}
-
-
-/**
- * Return the private _EGLSurface which corresponds to a public EGLSurface
- * handle.
- * This is the inverse of _eglGetSurfaceHandle().
- */
-_EGLSurface *
-_eglLookupSurface(EGLSurface surf)
-{
-   _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces,
-                                                   (EGLuint) surf);
-   return c;
-}
-
-
-_EGLSurface *
-_eglGetCurrentSurface(EGLint readdraw)
-{
-   _EGLContext *ctx = _eglGetCurrentContext();
-   if (ctx) {
-      switch (readdraw) {
-      case EGL_DRAW:
-         return ctx->DrawSurface;
-      case EGL_READ:
-         return ctx->ReadSurface;
-      default:
-         return NULL;
-      }
-   }
-   return NULL;
-}
-
-
 EGLBoolean
 _eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
 {
@@ -385,19 +317,24 @@ _eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
    _EGLSurface *surf;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   if (!conf) {
+      _eglError(EGL_BAD_CONFIG, "eglCreateWindowSurface");
+      return EGL_NO_SURFACE;
+   }
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_WINDOW_BIT, config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_WINDOW_BIT, conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return _eglLinkSurface(surf, _eglLookupDisplay(dpy));
 #endif
    return EGL_NO_SURFACE;
 }
@@ -412,19 +349,24 @@ _eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
    _EGLSurface *surf;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   if (!conf) {
+      _eglError(EGL_BAD_CONFIG, "eglCreatePixmapSurface");
+      return EGL_NO_SURFACE;
+   }
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_PIXMAP_BIT, config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_PIXMAP_BIT, conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return _eglLinkSurface(surf, _eglLookupDisplay(dpy));
 #endif
    return EGL_NO_SURFACE;
 }
@@ -439,19 +381,24 @@ _eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 {
 #if 0 /* THIS IS JUST EXAMPLE CODE */
    _EGLSurface *surf;
+   _EGLConfig *conf;
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   if (!conf) {
+      _eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface");
+      return EGL_NO_SURFACE;
+   }
 
    surf = (_EGLSurface *) calloc(1, sizeof(_EGLSurface));
    if (!surf)
       return EGL_NO_SURFACE;
 
-   if (!_eglInitSurface(drv, dpy, surf, EGL_PBUFFER_BIT, config, attrib_list)) {
+   if (!_eglInitSurface(drv, surf, EGL_PBUFFER_BIT, conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(surf);
-
-   return surf->Handle;
+   return _eglLinkSurface(surf, _eglLookupDisplay(dpy));
 #endif
    return EGL_NO_SURFACE;
 }
@@ -465,13 +412,9 @@ _eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    _EGLSurface *surf = _eglLookupSurface(surface);
    if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (surf->IsBound) {
-         surf->DeletePending = EGL_TRUE;
-      }
-      else {
+      _eglUnlinkSurface(surf);
+      if (!surf->IsBound)
          free(surf);
-      }
       return EGL_TRUE;
    }
    else {
diff --git a/src/egl/main/eglsurface.h b/src/egl/main/eglsurface.h
index 50f965b5cb..f9413eb9d7 100644
--- a/src/egl/main/eglsurface.h
+++ b/src/egl/main/eglsurface.h
@@ -10,12 +10,15 @@
  */
 struct _egl_surface
 {
-   EGLSurface Handle;  /* The public/opaque handle which names this object */
+   /* Managed by EGLDisplay for linking */
+   _EGLDisplay *Display;
+   _EGLSurface *Next;
+   EGLSurface Handle;
+
    _EGLConfig *Config;
 
    /* May need reference counting here */
    EGLBoolean IsBound;
-   EGLBoolean DeletePending;
    EGLBoolean BoundToTexture;
 
    EGLint Type; /* one of EGL_WINDOW_BIT, EGL_PIXMAP_BIT or EGL_PBUFFER_BIT */
@@ -39,29 +42,8 @@ struct _egl_surface
 
 
 extern EGLBoolean
-_eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
-                _EGLSurface *surf, EGLint type, EGLConfig config,
-                const EGLint *attrib_list);
-
-
-extern void
-_eglSaveSurface(_EGLSurface *surf);
-
-
-extern void
-_eglRemoveSurface(_EGLSurface *surf);
-
-
-extern EGLSurface
-_eglGetSurfaceHandle(_EGLSurface *surface);
-
-
-extern _EGLSurface *
-_eglLookupSurface(EGLSurface surf);
- 
-
-extern _EGLSurface *
-_eglGetCurrentSurface(EGLint readdraw);
+_eglInitSurface(_EGLDriver *drv, _EGLSurface *surf, EGLint type,
+                _EGLConfig *config, const EGLint *attrib_list);
 
 
 extern EGLBoolean
diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h
index 9fbc55352c..0a770dec0c 100644
--- a/src/egl/main/egltypedefs.h
+++ b/src/egl/main/egltypedefs.h
@@ -6,6 +6,7 @@
 #include <EGL/egl.h>
 #include <EGL/eglext.h>
 
+#include "eglcompiler.h"
 
 typedef struct _egl_api _EGLAPI;
 
diff --git a/src/gallium/SConscript b/src/gallium/SConscript
index b6ceaf3edf..32b7c2980a 100644
--- a/src/gallium/SConscript
+++ b/src/gallium/SConscript
@@ -22,6 +22,7 @@ SConscript([
 	'auxiliary/draw/SConscript',
 	'auxiliary/pipebuffer/SConscript',
 	'auxiliary/indices/SConscript',
+	'auxiliary/rbug/SConscript',
 ])
 
 for driver in env['drivers']:
diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c
index 0bc77a5728..e6dce3f0e5 100644
--- a/src/gallium/auxiliary/cso_cache/cso_cache.c
+++ b/src/gallium/auxiliary/cso_cache/cso_cache.c
@@ -361,6 +361,10 @@ void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type,
 void cso_cache_delete(struct cso_cache *sc)
 {
    assert(sc);
+
+   if (!sc)
+      return;
+
    /* delete driver data */
    cso_for_each_state(sc, CSO_BLEND, delete_blend_state, 0);
    cso_for_each_state(sc, CSO_DEPTH_STENCIL_ALPHA, delete_depth_stencil_state, 0);
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index f388bf5d95..36c882acb7 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -55,14 +55,14 @@ struct cso_context {
    void *samplers[PIPE_MAX_SAMPLERS];
    unsigned nr_samplers;
 
-   void *samplers_saved[PIPE_MAX_SAMPLERS];
    unsigned nr_samplers_saved;
+   void *samplers_saved[PIPE_MAX_SAMPLERS];
 
    struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
    uint nr_textures;
 
-   struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
    uint nr_textures_saved;
+   struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
 
    /** Current and saved state.
     * The saved state is used as a 1-deep stack.
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c
index 3cde9d36d3..2e3f5b2fc0 100644
--- a/src/gallium/auxiliary/draw/draw_pipe.c
+++ b/src/gallium/auxiliary/draw/draw_pipe.c
@@ -216,22 +216,22 @@ void draw_pipeline_run( struct draw_context *draw,
                 ( DRAW_PIPE_RESET_STIPPLE |                      \
                   DRAW_PIPE_EDGE_FLAG_0 |                        \
                   DRAW_PIPE_EDGE_FLAG_2 ),                       \
-                 verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
-                 verts + stride * (i1),                          \
+                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
+                verts + stride * (i1),                           \
                 verts + stride * (i3));                          \
-      do_triangle( draw,                                         \
-                   ( DRAW_PIPE_EDGE_FLAG_0 |                     \
-                     DRAW_PIPE_EDGE_FLAG_1 ),                    \
-                 verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \
-                 verts + stride * (i2),                          \
-                 verts + stride * (i3))
+   do_triangle( draw,                                            \
+                ( DRAW_PIPE_EDGE_FLAG_0 |                        \
+                  DRAW_PIPE_EDGE_FLAG_1 ),                       \
+                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
+                verts + stride * (i2),                           \
+                verts + stride * (i3))
 
 #define TRIANGLE(flags,i0,i1,i2)                                 \
    do_triangle( draw,                                            \
                 flags,  /* flags */                              \
-                 verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
-                 verts + stride * (i1),                          \
-                 verts + stride * (i2))
+                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
+                verts + stride * (i1),                           \
+                verts + stride * (i2))
 
 #define LINE(flags,i0,i1)                                   \
    do_line( draw,                                           \
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
index a5d840b96e..1a5269c0de 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -159,8 +159,19 @@ vbuf_tri( struct draw_stage *stage,
 
    check_space( vbuf, 3 );
 
-   for (i = 0; i < 3; i++) {
-      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] );
+   if (vbuf->stage.draw->rasterizer->flatshade_first) {
+      /* Put provoking vertex in position expected by the driver.
+       * Emit last provoking vertex in first pos.
+       * Swap verts 0 & 1 to preserve polygon winding.
+       */
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[2] );
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] );
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[1] );
+   }
+   else {
+      for (i = 0; i < 3; i++) {
+         vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] );
+      }
    }
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index d84bab9eaa..014d8c7346 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -181,6 +181,16 @@ static void widepoint_point( struct draw_stage *stage,
 }
 
 
+static int
+find_fog_input_attrib(struct draw_context *draw)
+{
+   /* Scan the fragment program's input decls to find the fogcoord
+    * attribute.  The z/w components will store the point coord.
+    */
+   return 0; /* XXX fix this */
+}
+
+
 static void widepoint_first_point( struct draw_stage *stage, 
 			      struct prim_header *header )
 {
@@ -220,7 +230,7 @@ static void widepoint_first_point( struct draw_stage *stage,
       wide->num_texcoords = j;
 
       /* find fragment shader PointCoord/Fog input */
-      wide->point_coord_fs_input = 0; /* XXX fix this! */
+      wide->point_coord_fs_input = find_fog_input_attrib(draw);
 
       /* setup extra vp output (point coord implemented as a texcoord) */
       draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 81e4eae401..41fcb16a0a 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -44,7 +44,6 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 
-#include "tgsi/tgsi_exec.h"
 #include "tgsi/tgsi_scan.h"
 
 
@@ -55,6 +54,8 @@ struct draw_vertex_shader;
 struct draw_context;
 struct draw_stage;
 struct vbuf_render;
+struct tgsi_exec_machine;
+struct tgsi_sampler;
 
 
 /**
@@ -185,7 +186,7 @@ struct draw_context
       uint position_output;
 
       /** TGSI program interpreter runtime state */
-      struct tgsi_exec_machine machine;
+      struct tgsi_exec_machine *machine;
 
       uint num_samplers;
       struct tgsi_sampler **samplers;
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index 9ea0cbe599..dbb5ac7182 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -36,6 +36,7 @@
 #include "draw/draw_vs.h"
 #include "tgsi/tgsi_dump.h"
 #include "util/u_math.h"
+#include "util/u_prim.h"
 
 static unsigned trim( unsigned count, unsigned first, unsigned incr )
 {
@@ -278,7 +279,7 @@ void
 draw_arrays(struct draw_context *draw, unsigned prim,
             unsigned start, unsigned count)
 {
-   unsigned reduced_prim = draw_pt_reduced_prim(prim);
+   unsigned reduced_prim = u_reduced_prim(prim);
    if (reduced_prim != draw->reduced_prim) {
       draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
       draw->reduced_prim = reduced_prim;
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index 8ef0ea8011..7a17a9fb6b 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -228,7 +228,6 @@ void draw_pt_post_vs_destroy( struct pt_post_vs *pvs );
  * Utils: 
  */
 void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr);
-unsigned draw_pt_reduced_prim(unsigned prim);
 
 
 #endif
diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h
index 3fb0695687..4ca5b52020 100644
--- a/src/gallium/auxiliary/draw/draw_pt_decompose.h
+++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h
@@ -47,10 +47,19 @@ static void FUNC( ARGS,
 
    case PIPE_PRIM_TRIANGLES:
       for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                   (i + 0),
-                   (i + 1),
-                   (i + 2 ));
+         if (flatfirst) {
+            /* put provoking vertex in last pos for clipper */
+            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                      (i + 1),
+                      (i + 2),
+                      (i + 0 ));
+         }
+         else {
+            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                      (i + 0),
+                      (i + 1),
+                      (i + 2 ));
+         }
       }
       break;
 
@@ -58,9 +67,9 @@ static void FUNC( ARGS,
       if (flatfirst) {
          for (i = 0; i+2 < count; i++) {
             TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                      (i + 0),
                       (i + 1 + (i&1)),
-                      (i + 2 - (i&1)));
+                      (i + 2 - (i&1)),
+                      (i + 0) );
          }
       }
       else {
@@ -78,9 +87,9 @@ static void FUNC( ARGS,
          if (flatfirst) {
             for (i = 0; i+2 < count; i++) {
                TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                         (i + 1),
                          (i + 2),
-                         (0 ));
+                         0,
+                         (i + 1) );
             }
          }
          else {
@@ -96,20 +105,40 @@ static void FUNC( ARGS,
 
 
    case PIPE_PRIM_QUADS:
-      for (i = 0; i+3 < count; i += 4) {
-         QUAD( (i + 0),
-               (i + 1),
-               (i + 2),
-               (i + 3));
+      if (flatfirst) {
+         for (i = 0; i+3 < count; i += 4) {
+            QUAD( (i + 1),
+                  (i + 2),
+                  (i + 3),
+                  (i + 0) );
+         }
+      }
+      else {
+         for (i = 0; i+3 < count; i += 4) {
+            QUAD( (i + 0),
+                  (i + 1),
+                  (i + 2),
+                  (i + 3));
+         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      for (i = 0; i+3 < count; i += 2) {
-         QUAD( (i + 2),
-               (i + 0),
-               (i + 1),
-               (i + 3));
+      if (flatfirst) {
+         for (i = 0; i+3 < count; i += 2) {
+            QUAD( (i + 1),
+                  (i + 3),
+                  (i + 2),
+                  (i + 0) );
+         }
+      }
+      else {
+         for (i = 0; i+3 < count; i += 2) {
+            QUAD( (i + 2),
+                  (i + 0),
+                  (i + 1),
+                  (i + 3));
+         }
       }
       break;
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
index 6b7d02a19b..e7fe6b3b76 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
@@ -130,6 +130,10 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
       unsigned output_format;
 
       switch (vinfo->attrib[i].emit) {
+      case EMIT_4UB:
+	 output_format = PIPE_FORMAT_R8G8B8A8_UNORM;
+	 emit_sz = 4 * sizeof(unsigned char);
+         break;
       case EMIT_4F:
 	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 	 emit_sz = 4 * sizeof(float);
@@ -153,6 +157,8 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
 	 output_format = PIPE_FORMAT_R32_FLOAT;
 	 emit_sz = 1 * sizeof(float);
          break;
+      case EMIT_OMIT:
+         continue;
       default:
          assert(0);
 	 output_format = PIPE_FORMAT_NONE;
diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c
index 3bc7939c55..b61fa29143 100644
--- a/src/gallium/auxiliary/draw/draw_pt_util.c
+++ b/src/gallium/auxiliary/draw/draw_pt_util.c
@@ -75,28 +75,3 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       break;
    }
 }
-
-
-unsigned draw_pt_reduced_prim(unsigned prim)
-{
-   switch (prim) {
-   case PIPE_PRIM_POINTS:
-      return PIPE_PRIM_POINTS;
-   case PIPE_PRIM_LINES:
-   case PIPE_PRIM_LINE_STRIP:
-   case PIPE_PRIM_LINE_LOOP:
-      return PIPE_PRIM_LINES;
-   case PIPE_PRIM_TRIANGLES:
-   case PIPE_PRIM_TRIANGLE_STRIP:
-   case PIPE_PRIM_TRIANGLE_FAN:
-   case PIPE_PRIM_POLYGON:
-   case PIPE_PRIM_QUADS:
-   case PIPE_PRIM_QUAD_STRIP:
-      return PIPE_PRIM_TRIANGLES;
-   default:
-      assert(0);
-      return PIPE_PRIM_POINTS;
-   }
-}
-
-
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
index 5d268a2226..1a0527be63 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -31,6 +31,7 @@
   */
 
 #include "util/u_memory.h"
+#include "util/u_prim.h"
 #include "draw/draw_context.h"
 #include "draw/draw_private.h"
 #include "draw/draw_pt.h"
@@ -193,16 +194,30 @@ vcache_ef_quad( struct vcache_frontend *vcache,
                 unsigned i2,
                 unsigned i3 )
 {
-   vcache_triangle_flags( vcache,
-                          ( DRAW_PIPE_RESET_STIPPLE |
-                            DRAW_PIPE_EDGE_FLAG_0 |
-                            DRAW_PIPE_EDGE_FLAG_2 ),
-                          i0, i1, i3 );
-
-   vcache_triangle_flags( vcache,
-                          ( DRAW_PIPE_EDGE_FLAG_0 |
-                            DRAW_PIPE_EDGE_FLAG_1 ),
-                          i1, i2, i3 );
+   if (vcache->draw->rasterizer->flatshade_first) {
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_RESET_STIPPLE |
+                               DRAW_PIPE_EDGE_FLAG_0 |
+                               DRAW_PIPE_EDGE_FLAG_1 ),
+                             i0, i1, i2 );
+
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_EDGE_FLAG_2 |
+                               DRAW_PIPE_EDGE_FLAG_1 ),
+                             i0, i2, i3 );
+   }
+   else {
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_RESET_STIPPLE |
+                               DRAW_PIPE_EDGE_FLAG_0 |
+                               DRAW_PIPE_EDGE_FLAG_2 ),
+                             i0, i1, i3 );
+
+      vcache_triangle_flags( vcache,
+                             ( DRAW_PIPE_EDGE_FLAG_0 |
+                               DRAW_PIPE_EDGE_FLAG_1 ),
+                             i1, i2, i3 );
+   }
 }
 
 /* At least for now, we're back to using a template include file for
@@ -453,7 +468,7 @@ vcache_prepare( struct draw_pt_front_end *frontend,
    }
 
    vcache->input_prim = prim;
-   vcache->output_prim = draw_pt_reduced_prim(prim);
+   vcache->output_prim = u_reduced_prim(prim);
 
    vcache->middle = middle;
    vcache->opt = opt;
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
index ec05bbeab4..62822a3d56 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
@@ -118,21 +118,39 @@ static void FUNC( struct draw_pt_front_end *frontend,
 
    case PIPE_PRIM_QUADS:
       for (i = 0; i+3 < count; i += 4) {
-         QUAD( vcache,
-               get_elt(elts, i + 0),
-               get_elt(elts, i + 1),
-               get_elt(elts, i + 2),
-               get_elt(elts, i + 3));
+         if (flatfirst) {
+            QUAD( vcache,
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 2),
+                  get_elt(elts, i + 3) );
+         }
+         else {
+            QUAD( vcache,
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 2),
+                  get_elt(elts, i + 3) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 0; i+3 < count; i += 2) {
-         QUAD( vcache,
-               get_elt(elts, i + 2),
-               get_elt(elts, i + 0),
-               get_elt(elts, i + 1),
-               get_elt(elts, i + 3));
+         if (flatfirst) {
+            QUAD( vcache,
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 3),
+                  get_elt(elts, i + 2) );
+         }
+         else {
+            QUAD( vcache,
+                  get_elt(elts, i + 2),
+                  get_elt(elts, i + 0),
+                  get_elt(elts, i + 1),
+                  get_elt(elts, i + 3) );
+         }
       }
       break;
 
@@ -144,19 +162,38 @@ static void FUNC( struct draw_pt_front_end *frontend,
          const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
          const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
          const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
+         ushort edge_next, edge_finish;
 
-         flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+         if (flatfirst) {
+            flags = DRAW_PIPE_RESET_STIPPLE | edge_middle | edge_last;
+            edge_next = edge_last;
+            edge_finish = edge_first;
+         }
+         else {
+            flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+            edge_next = edge_middle;
+            edge_finish = edge_last;
+         }
 
-	 for (i = 0; i+2 < count; i++, flags = edge_middle) {
+	 for (i = 0; i+2 < count; i++, flags = edge_next) {
 
             if (i + 3 == count)
-               flags |= edge_last;
+               flags |= edge_finish;
 
-	    TRIANGLE( vcache,
-                      flags,
-                      get_elt(elts, i + 1),
-                      get_elt(elts, i + 2),
-                      get_elt(elts, 0));
+            if (flatfirst) {
+               TRIANGLE( vcache,
+                         flags,
+                         get_elt(elts, 0),
+                         get_elt(elts, i + 1),
+                         get_elt(elts, i + 2) );
+            }
+            else {
+               TRIANGLE( vcache,
+                         flags,
+                         get_elt(elts, i + 1),
+                         get_elt(elts, i + 2),
+                         get_elt(elts, 0));
+            }
 	 }
       }
       break;
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index c057cd67fd..790e89ed82 100644
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -43,6 +43,8 @@
 #include "translate/translate.h"
 #include "translate/translate_cache.h"
 
+#include "tgsi/tgsi_exec.h"
+
 
 
 
@@ -146,16 +148,8 @@ draw_delete_vertex_shader(struct draw_context *draw,
 boolean 
 draw_vs_init( struct draw_context *draw )
 {
-   tgsi_exec_machine_init(&draw->vs.machine);
-
-   /* FIXME: give this machine thing a proper constructor:
-    */
-   draw->vs.machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16);
-   if (!draw->vs.machine.Inputs)
-      return FALSE;
-
-   draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16);
-   if (!draw->vs.machine.Outputs)
+   draw->vs.machine = tgsi_exec_machine_create();
+   if (!draw->vs.machine)
       return FALSE;
 
    draw->vs.emit_cache = translate_cache_create();
@@ -178,12 +172,6 @@ draw_vs_init( struct draw_context *draw )
 void
 draw_vs_destroy( struct draw_context *draw )
 {
-   if (draw->vs.machine.Inputs)
-      align_free(draw->vs.machine.Inputs);
-
-   if (draw->vs.machine.Outputs)
-      align_free(draw->vs.machine.Outputs);
-
    if (draw->vs.fetch_cache)
       translate_cache_destroy(draw->vs.fetch_cache);
 
@@ -196,8 +184,7 @@ draw_vs_destroy( struct draw_context *draw )
    if (draw->vs.aligned_constant_storage)
       align_free((void*)draw->vs.aligned_constant_storage);
 
-   tgsi_exec_machine_free_data(&draw->vs.machine);
-
+   tgsi_exec_machine_destroy(draw->vs.machine);
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index dbbc33fffa..41cc802613 100644
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -41,6 +41,7 @@
 
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_scan.h"
+#include "tgsi/tgsi_exec.h"
 
 
 struct exec_vertex_shader {
@@ -114,6 +115,12 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
 #endif
 
          for (slot = 0; slot < shader->info.num_inputs; slot++) {
+#if 0
+            assert(!util_is_inf_or_nan(input[slot][0]));
+            assert(!util_is_inf_or_nan(input[slot][1]));
+            assert(!util_is_inf_or_nan(input[slot][2]));
+            assert(!util_is_inf_or_nan(input[slot][3]));
+#endif
             machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
             machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
             machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
@@ -195,7 +202,7 @@ draw_create_vs_exec(struct draw_context *draw,
    vs->base.run_linear = vs_exec_run_linear;
    vs->base.delete = vs_exec_delete;
    vs->base.create_varient = draw_vs_varient_generic;
-   vs->machine = &draw->vs.machine;
+   vs->machine = draw->vs.machine;
 
    return &vs->base;
 }
diff --git a/src/gallium/auxiliary/draw/draw_vs_ppc.c b/src/gallium/auxiliary/draw/draw_vs_ppc.c
index d35db57d57..ad184bd696 100644
--- a/src/gallium/auxiliary/draw/draw_vs_ppc.c
+++ b/src/gallium/auxiliary/draw/draw_vs_ppc.c
@@ -48,6 +48,7 @@
 #include "rtasm/rtasm_ppc.h"
 #include "tgsi/tgsi_ppc.h"
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
 
 
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index 77ba5152f9..fb58983e01 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -48,27 +48,16 @@
 #include "rtasm/rtasm_x86sse.h"
 #include "tgsi/tgsi_sse2.h"
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
 
 #define SSE_MAX_VERTICES 4
 
-typedef void (PIPE_CDECL *codegen_function) (
-   const struct tgsi_exec_vector *input, /* 1 */
-   struct tgsi_exec_vector *output, /* 2 */
-   float (*constant)[4],        /* 3 */
-   struct tgsi_exec_vector *temporary, /* 4 */
-   float (*immediates)[4],      /* 5 */
-   const float (*aos_input)[4], /* 6 */
-   uint num_inputs,             /* 7 */
-   uint input_stride,           /* 8 */
-   float (*aos_output)[4],      /* 9 */
-   uint num_outputs,            /* 10 */
-   uint output_stride );        /* 11 */
 
 struct draw_sse_vertex_shader {
    struct draw_vertex_shader base;
    struct x86_function sse2_program;
 
-   codegen_function func;
+   tgsi_sse2_vs_func func;
    
    struct tgsi_exec_machine *machine;
 };
@@ -118,11 +107,9 @@ vs_sse_run_linear( struct draw_vertex_shader *base,
 
       /* run compiled shader
        */
-      shader->func(machine->Inputs,
-		   machine->Outputs,
-		   (float (*)[4])constants,
-		   machine->Temps,
-		   (float (*)[4])shader->base.immediates,
+      shader->func(machine,
+		   constants,
+		   shader->base.immediates,
                    input,
                    base->info.num_inputs,
                    input_stride,
@@ -184,7 +171,7 @@ draw_create_vs_sse(struct draw_context *draw,
    vs->base.immediates = align_malloc(TGSI_EXEC_NUM_IMMEDIATES * 4 *
                                       sizeof(float), 16);
 
-   vs->machine = &draw->vs.machine;
+   vs->machine = draw->vs.machine;
    
    x86_init_func( &vs->sse2_program );
 
@@ -194,7 +181,7 @@ draw_create_vs_sse(struct draw_context *draw,
                         TRUE )) 
       goto fail;
       
-   vs->func = (codegen_function) x86_get_func( &vs->sse2_program );
+   vs->func = (tgsi_sse2_vs_func) x86_get_func( &vs->sse2_program );
    if (!vs->func) {
       goto fail;
    }
diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp
index 925e948763..2d2af3085e 100644
--- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp
+++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp
@@ -24,6 +24,8 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  **************************************************************************/
+
+#include <cstdio>
 #include "instructionssoa.h"
 
 #include "storagesoa.h"
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
index 0760d60716..6bdce5fcb0 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
@@ -53,6 +53,8 @@ static INLINE struct malloc_buffer *
 malloc_buffer(struct pb_buffer *buf)
 {
    assert(buf);
+   if (!buf)
+      return NULL;
    assert(buf->vtbl == &malloc_buffer_vtbl);
    return (struct malloc_buffer *)buf;
 }
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
index bc3093f620..cb32d25136 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
@@ -74,6 +74,8 @@ static INLINE struct pb_ondemand_buffer *
 pb_ondemand_buffer(struct pb_buffer *buf)
 {
    assert(buf);
+   if (!buf)
+      return NULL;
    assert(buf->vtbl == &pb_ondemand_buffer_vtbl);
    return (struct pb_ondemand_buffer *)buf;
 }
diff --git a/src/gallium/auxiliary/rbug/Makefile b/src/gallium/auxiliary/rbug/Makefile
new file mode 100644
index 0000000000..cd12e8468f
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/Makefile
@@ -0,0 +1,14 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = rbug
+
+C_SOURCES = \
+	rbug_connection.c \
+	rbug_core.c \
+	rbug_texture.c \
+	rbug_context.c \
+	rbug_shader.c \
+	rbug_demarshal.c
+
+include ../../Makefile.template
diff --git a/src/gallium/auxiliary/rbug/README b/src/gallium/auxiliary/rbug/README
new file mode 100644
index 0000000000..33d76371de
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/README
@@ -0,0 +1,21 @@
+                 GALLIUM REMOTE DEBUGGING COMMON CODE
+
+= About =
+
+This directory contains the common code for the Gallium 3D remote debugging
+driver and clients. The code is two parts the connection managment code and
+the (de)marsheller.
+
+The code currently uses tcp and ip4v for connections.
+
+Information about driver integration can be found in:
+
+src/gallium/drivers/trace/README
+
+for information about applications look in:
+
+progs/rbug/README
+
+
+--
+Jakob Bornecrantz <jakob@vmware.com>
diff --git a/src/gallium/auxiliary/rbug/SConscript b/src/gallium/auxiliary/rbug/SConscript
new file mode 100644
index 0000000000..4a9afb45d3
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/SConscript
@@ -0,0 +1,14 @@
+Import('*')
+
+rbug = env.ConvenienceLibrary(
+	target = 'rbug',
+	source = [
+		'rbug_core.c',
+		'rbug_shader.c',
+		'rbug_context.c',
+		'rbug_texture.c',
+		'rbug_demarshal.c',
+		'rbug_connection.c',
+	])
+
+auxiliaries.insert(0, rbug)
diff --git a/src/gallium/auxiliary/rbug/rbug.h b/src/gallium/auxiliary/rbug/rbug.h
new file mode 100644
index 0000000000..259bfc6c79
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Include all for users the remote debugger protocol code.
+ */
+
+#include "rbug/rbug_core.h"
+#include "rbug/rbug_shader.h"
+#include "rbug/rbug_context.h"
+#include "rbug/rbug_texture.h"
+#include "rbug/rbug_connection.h"
diff --git a/src/gallium/auxiliary/rbug/rbug_connection.c b/src/gallium/auxiliary/rbug/rbug_connection.c
new file mode 100644
index 0000000000..52acb700af
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_connection.c
@@ -0,0 +1,167 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "rbug/rbug.h"
+#include "rbug/rbug_internal.h"
+
+#include "util/u_network.h"
+
+struct rbug_connection
+{
+   int socket;
+   uint32_t send_serial;
+   uint32_t recv_serial;
+   enum rbug_opcode opcode;
+};
+
+/**
+ * Create a rbug connection from a socket created with u_socket.
+ *
+ * Result:
+ *    A new allocated connection using socket as communication path
+ */
+struct rbug_connection *
+rbug_from_socket(int socket)
+{
+   struct rbug_connection *c = CALLOC_STRUCT(rbug_connection);
+   c->socket = socket;
+   return c;
+}
+
+/**
+ * Free a connection, also closes socket.
+ */
+void
+rbug_disconnect(struct rbug_connection *c)
+{
+   u_socket_close(c->socket);
+   FREE(c);
+}
+
+/**
+ * Waits for a message to be fully received.
+ * Also returns the serial for the message, serial is not touched for replys.
+ *
+ * Result:
+ *    demarshaled message on success, NULL on connection error
+ */
+struct rbug_header *
+rbug_get_message(struct rbug_connection *c, uint32_t *serial)
+{
+   struct rbug_proto_header header;
+   struct rbug_header *out;
+   struct rbug_proto_header *data;
+   size_t length = 0;
+   size_t read = 0;
+   int ret;
+
+
+   ret = u_socket_peek(c->socket, &header, sizeof(header));
+   if (ret <= 0) {
+      return NULL;
+   }
+
+   length = (size_t)header.length * 4;
+   data = MALLOC(length);
+   if (!data) {
+      return NULL;
+   }
+
+   do {
+      uint8_t *ptr = ((uint8_t*)data) + read;
+      ret = u_socket_recv(c->socket, ptr, length - read);
+
+      if (ret <= 0) {
+         FREE(data);
+         return NULL;
+      }
+
+      read += ret;
+   } while(read < length);
+
+   out = rbug_demarshal(data);
+   if (!out)
+      FREE(data);
+   else if (serial)
+      *serial = c->recv_serial++;
+   else
+      c->recv_serial++;
+
+   return out;
+}
+
+/**
+ * Frees a message and associated data.
+ */
+void
+rbug_free_header(struct rbug_header *header)
+{
+   if (!header)
+      return;
+
+   FREE(header->__message);
+   FREE(header);
+}
+
+/**
+ * Internal function used by rbug_send_* functions.
+ *
+ * Start sending a message.
+ */
+int
+rbug_connection_send_start(struct rbug_connection *c, enum rbug_opcode opcode, uint32_t length)
+{
+   c->opcode = opcode;
+   return 0;
+}
+
+/**
+ * Internal function used by rbug_send_* functions.
+ *
+ * Write data to the socket.
+ */
+int
+rbug_connection_write(struct rbug_connection *c, void *to, uint32_t size)
+{
+   int ret = u_socket_send(c->socket, to, size);
+   return ret;
+}
+
+/**
+ * Internal function used by rbug_send_* functions.
+ *
+ * Finish writeing data to the socket.
+ * Ups the send_serial and sets the serial argument if supplied.
+ */
+int rbug_connection_send_finish(struct rbug_connection *c, uint32_t *serial)
+{
+   if (c->opcode < 0)
+      return 0;
+   else if (serial)
+      *serial = c->send_serial++;
+   else
+      c->send_serial++;
+
+   return 0;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_connection.h b/src/gallium/auxiliary/rbug/rbug_connection.h
new file mode 100644
index 0000000000..1f2c9ff347
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_connection.h
@@ -0,0 +1,45 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file contains the function defentions for connection see c file for
+ * more comments covering function use.
+ */
+
+#ifndef _RBUG_CONNECTION_H_
+#define _RBUG_CONNECTION_H_
+
+#include "rbug/rbug_proto.h"
+
+struct rbug_connection * rbug_from_socket(int socket);
+
+void rbug_disconnect(struct rbug_connection *c);
+
+struct rbug_header * rbug_get_message(struct rbug_connection *c, uint32_t *serial);
+
+void rbug_free_header(struct rbug_header *header);
+
+struct rbug_header * rbug_demarshal(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_context.c b/src/gallium/auxiliary/rbug/rbug_context.c
new file mode 100644
index 0000000000..1832425658
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_context.c
@@ -0,0 +1,759 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_context.h"
+
+int rbug_send_context_list(struct rbug_connection *__con,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_LIST));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_LIST, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_info(struct rbug_connection *__con,
+                           rbug_context_t context,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_INFO));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_INFO, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_block(struct rbug_connection *__con,
+                                 rbug_context_t context,
+                                 rbug_block_t block,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* block */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_BLOCK));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, block); /* block */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_BLOCK, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_step(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_block_t step,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* step */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_STEP));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, step); /* step */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_STEP, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_unblock(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t unblock,
+                                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* unblock */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_UNBLOCK));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, unblock); /* unblock */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_UNBLOCK, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_rule(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_shader_t vertex,
+                                rbug_shader_t fragment,
+                                rbug_texture_t texture,
+                                rbug_texture_t surface,
+                                rbug_block_t block,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* vertex */
+	LEN(8); /* fragment */
+	LEN(8); /* texture */
+	LEN(8); /* surface */
+	LEN(4); /* block */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_RULE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, vertex); /* vertex */
+	WRITE(8, rbug_shader_t, fragment); /* fragment */
+	WRITE(8, rbug_texture_t, texture); /* texture */
+	WRITE(8, rbug_texture_t, surface); /* surface */
+	WRITE(4, rbug_block_t, block); /* block */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_RULE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_flush(struct rbug_connection *__con,
+                            rbug_context_t context,
+                            int32_t flags,
+                            uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* flags */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_FLUSH));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, int32_t, flags); /* flags */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_FLUSH, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_context_t *contexts,
+                                 uint32_t contexts_len,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(8, contexts); /* contexts */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_LIST_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(8, rbug_context_t, contexts); /* contexts */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_LIST_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_shader_t vertex,
+                                 rbug_shader_t fragment,
+                                 rbug_texture_t *texs,
+                                 uint32_t texs_len,
+                                 rbug_texture_t *cbufs,
+                                 uint32_t cbufs_len,
+                                 rbug_texture_t zsbuf,
+                                 rbug_block_t blocker,
+                                 rbug_block_t blocked,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(8); /* vertex */
+	LEN(8); /* fragment */
+	LEN_ARRAY(8, texs); /* texs */
+	LEN_ARRAY(8, cbufs); /* cbufs */
+	LEN(8); /* zsbuf */
+	LEN(4); /* blocker */
+	LEN(4); /* blocked */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_INFO_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(8, rbug_shader_t, vertex); /* vertex */
+	WRITE(8, rbug_shader_t, fragment); /* fragment */
+	WRITE_ARRAY(8, rbug_texture_t, texs); /* texs */
+	WRITE_ARRAY(8, rbug_texture_t, cbufs); /* cbufs */
+	WRITE(8, rbug_texture_t, zsbuf); /* zsbuf */
+	WRITE(4, rbug_block_t, blocker); /* blocker */
+	WRITE(4, rbug_block_t, blocked); /* blocked */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_INFO_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_context_draw_blocked(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t block,
+                                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(4); /* block */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_CONTEXT_DRAW_BLOCKED));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(4, rbug_block_t, block); /* block */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_CONTEXT_DRAW_BLOCKED, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_context_list * rbug_demarshal_context_list(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_list *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_LIST)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_context_info * rbug_demarshal_context_info(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_info *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_INFO)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_block * rbug_demarshal_context_draw_block(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_block *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_BLOCK)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, block); /* block */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_step * rbug_demarshal_context_draw_step(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_step *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_STEP)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, step); /* step */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_unblock * rbug_demarshal_context_draw_unblock(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_unblock *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_UNBLOCK)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, unblock); /* unblock */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_rule * rbug_demarshal_context_draw_rule(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_rule *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_RULE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, vertex); /* vertex */
+	READ(8, rbug_shader_t, fragment); /* fragment */
+	READ(8, rbug_texture_t, texture); /* texture */
+	READ(8, rbug_texture_t, surface); /* surface */
+	READ(4, rbug_block_t, block); /* block */
+
+	return ret;
+}
+
+struct rbug_proto_context_flush * rbug_demarshal_context_flush(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_flush *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_FLUSH)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, int32_t, flags); /* flags */
+
+	return ret;
+}
+
+struct rbug_proto_context_list_reply * rbug_demarshal_context_list_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_list_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_LIST_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(8, rbug_context_t, contexts); /* contexts */
+
+	return ret;
+}
+
+struct rbug_proto_context_info_reply * rbug_demarshal_context_info_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_info_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_INFO_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(8, rbug_shader_t, vertex); /* vertex */
+	READ(8, rbug_shader_t, fragment); /* fragment */
+	READ_ARRAY(8, rbug_texture_t, texs); /* texs */
+	READ_ARRAY(8, rbug_texture_t, cbufs); /* cbufs */
+	READ(8, rbug_texture_t, zsbuf); /* zsbuf */
+	READ(4, rbug_block_t, blocker); /* blocker */
+	READ(4, rbug_block_t, blocked); /* blocked */
+
+	return ret;
+}
+
+struct rbug_proto_context_draw_blocked * rbug_demarshal_context_draw_blocked(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_context_draw_blocked *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_CONTEXT_DRAW_BLOCKED)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(4, rbug_block_t, block); /* block */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_context.h b/src/gallium/auxiliary/rbug/rbug_context.h
new file mode 100644
index 0000000000..da61c2365b
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_context.h
@@ -0,0 +1,212 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_CONTEXT_H_
+#define _RBUG_PROTO_CONTEXT_H_
+
+#include "rbug/rbug_proto.h"
+#include "rbug/rbug_core.h"
+
+typedef enum
+{
+	RBUG_BLOCK_BEFORE = 1,
+	RBUG_BLOCK_AFTER = 2,
+	RBUG_BLOCK_RULE = 4,
+	RBUG_BLOCK_MASK = 7,
+} rbug_block_t;
+
+struct rbug_proto_context_list
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_context_info
+{
+	struct rbug_header header;
+	rbug_context_t context;
+};
+
+struct rbug_proto_context_draw_block
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t block;
+};
+
+struct rbug_proto_context_draw_step
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t step;
+};
+
+struct rbug_proto_context_draw_unblock
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t unblock;
+};
+
+struct rbug_proto_context_draw_rule
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t vertex;
+	rbug_shader_t fragment;
+	rbug_texture_t texture;
+	rbug_texture_t surface;
+	rbug_block_t block;
+};
+
+struct rbug_proto_context_flush
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	int32_t flags;
+};
+
+struct rbug_proto_context_list_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_context_t *contexts;
+	uint32_t contexts_len;
+};
+
+struct rbug_proto_context_info_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_shader_t vertex;
+	rbug_shader_t fragment;
+	rbug_texture_t *texs;
+	uint32_t texs_len;
+	rbug_texture_t *cbufs;
+	uint32_t cbufs_len;
+	rbug_texture_t zsbuf;
+	rbug_block_t blocker;
+	rbug_block_t blocked;
+};
+
+struct rbug_proto_context_draw_blocked
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_block_t block;
+};
+
+int rbug_send_context_list(struct rbug_connection *__con,
+                           uint32_t *__serial);
+
+int rbug_send_context_info(struct rbug_connection *__con,
+                           rbug_context_t context,
+                           uint32_t *__serial);
+
+int rbug_send_context_draw_block(struct rbug_connection *__con,
+                                 rbug_context_t context,
+                                 rbug_block_t block,
+                                 uint32_t *__serial);
+
+int rbug_send_context_draw_step(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_block_t step,
+                                uint32_t *__serial);
+
+int rbug_send_context_draw_unblock(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t unblock,
+                                   uint32_t *__serial);
+
+int rbug_send_context_draw_rule(struct rbug_connection *__con,
+                                rbug_context_t context,
+                                rbug_shader_t vertex,
+                                rbug_shader_t fragment,
+                                rbug_texture_t texture,
+                                rbug_texture_t surface,
+                                rbug_block_t block,
+                                uint32_t *__serial);
+
+int rbug_send_context_flush(struct rbug_connection *__con,
+                            rbug_context_t context,
+                            int32_t flags,
+                            uint32_t *__serial);
+
+int rbug_send_context_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_context_t *contexts,
+                                 uint32_t contexts_len,
+                                 uint32_t *__serial);
+
+int rbug_send_context_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_shader_t vertex,
+                                 rbug_shader_t fragment,
+                                 rbug_texture_t *texs,
+                                 uint32_t texs_len,
+                                 rbug_texture_t *cbufs,
+                                 uint32_t cbufs_len,
+                                 rbug_texture_t zsbuf,
+                                 rbug_block_t blocker,
+                                 rbug_block_t blocked,
+                                 uint32_t *__serial);
+
+int rbug_send_context_draw_blocked(struct rbug_connection *__con,
+                                   rbug_context_t context,
+                                   rbug_block_t block,
+                                   uint32_t *__serial);
+
+struct rbug_proto_context_list * rbug_demarshal_context_list(struct rbug_proto_header *header);
+
+struct rbug_proto_context_info * rbug_demarshal_context_info(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_block * rbug_demarshal_context_draw_block(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_step * rbug_demarshal_context_draw_step(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_unblock * rbug_demarshal_context_draw_unblock(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_rule * rbug_demarshal_context_draw_rule(struct rbug_proto_header *header);
+
+struct rbug_proto_context_flush * rbug_demarshal_context_flush(struct rbug_proto_header *header);
+
+struct rbug_proto_context_list_reply * rbug_demarshal_context_list_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_context_info_reply * rbug_demarshal_context_info_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_context_draw_blocked * rbug_demarshal_context_draw_blocked(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_core.c b/src/gallium/auxiliary/rbug/rbug_core.c
new file mode 100644
index 0000000000..876ae5a0ce
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_core.c
@@ -0,0 +1,359 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_core.h"
+
+int rbug_send_noop(struct rbug_connection *__con,
+                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_NOOP));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_NOOP, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_ping(struct rbug_connection *__con,
+                   uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_PING));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_PING, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_error(struct rbug_connection *__con,
+                    uint32_t error,
+                    uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* error */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_ERROR));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, error); /* error */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_ERROR, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_ping_reply(struct rbug_connection *__con,
+                         uint32_t serial,
+                         uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_PING_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_PING_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_error_reply(struct rbug_connection *__con,
+                          uint32_t serial,
+                          uint32_t error,
+                          uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(4); /* error */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_ERROR_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(4, uint32_t, error); /* error */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_ERROR_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_noop * rbug_demarshal_noop(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_noop *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_NOOP)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_ping * rbug_demarshal_ping(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_ping *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_PING)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_error * rbug_demarshal_error(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_error *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_ERROR)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, error); /* error */
+
+	return ret;
+}
+
+struct rbug_proto_ping_reply * rbug_demarshal_ping_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_ping_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_PING_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+
+	return ret;
+}
+
+struct rbug_proto_error_reply * rbug_demarshal_error_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_error_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_ERROR_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(4, uint32_t, error); /* error */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_core.h b/src/gallium/auxiliary/rbug/rbug_core.h
new file mode 100644
index 0000000000..99a36a0163
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_core.h
@@ -0,0 +1,105 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_CORE_H_
+#define _RBUG_PROTO_CORE_H_
+
+#include "rbug/rbug_proto.h"
+
+typedef uint64_t rbug_shader_t;
+typedef uint64_t rbug_context_t;
+typedef uint64_t rbug_texture_t;
+
+struct rbug_proto_noop
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_ping
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_error
+{
+	struct rbug_header header;
+	uint32_t error;
+};
+
+struct rbug_proto_ping_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+};
+
+struct rbug_proto_error_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t error;
+};
+
+int rbug_send_noop(struct rbug_connection *__con,
+                   uint32_t *__serial);
+
+int rbug_send_ping(struct rbug_connection *__con,
+                   uint32_t *__serial);
+
+int rbug_send_error(struct rbug_connection *__con,
+                    uint32_t error,
+                    uint32_t *__serial);
+
+int rbug_send_ping_reply(struct rbug_connection *__con,
+                         uint32_t serial,
+                         uint32_t *__serial);
+
+int rbug_send_error_reply(struct rbug_connection *__con,
+                          uint32_t serial,
+                          uint32_t error,
+                          uint32_t *__serial);
+
+struct rbug_proto_noop * rbug_demarshal_noop(struct rbug_proto_header *header);
+
+struct rbug_proto_ping * rbug_demarshal_ping(struct rbug_proto_header *header);
+
+struct rbug_proto_error * rbug_demarshal_error(struct rbug_proto_header *header);
+
+struct rbug_proto_ping_reply * rbug_demarshal_ping_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_error_reply * rbug_demarshal_error_reply(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_demarshal.c b/src/gallium/auxiliary/rbug/rbug_demarshal.c
new file mode 100644
index 0000000000..47390fbcee
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_demarshal.c
@@ -0,0 +1,93 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "rbug.h"
+
+/**
+ * Small function that looks at the proto_header and selects the correct
+ * demarshal functions and return the result.
+ */
+struct rbug_header * rbug_demarshal(struct rbug_proto_header *header)
+{
+	switch(header->opcode) {
+	case RBUG_OP_NOOP:
+		return (struct rbug_header *)rbug_demarshal_noop(header);
+	case RBUG_OP_PING:
+		return (struct rbug_header *)rbug_demarshal_ping(header);
+	case RBUG_OP_ERROR:
+		return (struct rbug_header *)rbug_demarshal_error(header);
+	case RBUG_OP_PING_REPLY:
+		return (struct rbug_header *)rbug_demarshal_ping_reply(header);
+	case RBUG_OP_ERROR_REPLY:
+		return (struct rbug_header *)rbug_demarshal_error_reply(header);
+	case RBUG_OP_TEXTURE_LIST:
+		return (struct rbug_header *)rbug_demarshal_texture_list(header);
+	case RBUG_OP_TEXTURE_INFO:
+		return (struct rbug_header *)rbug_demarshal_texture_info(header);
+	case RBUG_OP_TEXTURE_WRITE:
+		return (struct rbug_header *)rbug_demarshal_texture_write(header);
+	case RBUG_OP_TEXTURE_READ:
+		return (struct rbug_header *)rbug_demarshal_texture_read(header);
+	case RBUG_OP_TEXTURE_LIST_REPLY:
+		return (struct rbug_header *)rbug_demarshal_texture_list_reply(header);
+	case RBUG_OP_TEXTURE_INFO_REPLY:
+		return (struct rbug_header *)rbug_demarshal_texture_info_reply(header);
+	case RBUG_OP_TEXTURE_READ_REPLY:
+		return (struct rbug_header *)rbug_demarshal_texture_read_reply(header);
+	case RBUG_OP_CONTEXT_LIST:
+		return (struct rbug_header *)rbug_demarshal_context_list(header);
+	case RBUG_OP_CONTEXT_INFO:
+		return (struct rbug_header *)rbug_demarshal_context_info(header);
+	case RBUG_OP_CONTEXT_DRAW_BLOCK:
+		return (struct rbug_header *)rbug_demarshal_context_draw_block(header);
+	case RBUG_OP_CONTEXT_DRAW_STEP:
+		return (struct rbug_header *)rbug_demarshal_context_draw_step(header);
+	case RBUG_OP_CONTEXT_DRAW_UNBLOCK:
+		return (struct rbug_header *)rbug_demarshal_context_draw_unblock(header);
+	case RBUG_OP_CONTEXT_DRAW_RULE:
+		return (struct rbug_header *)rbug_demarshal_context_draw_rule(header);
+	case RBUG_OP_CONTEXT_FLUSH:
+		return (struct rbug_header *)rbug_demarshal_context_flush(header);
+	case RBUG_OP_CONTEXT_LIST_REPLY:
+		return (struct rbug_header *)rbug_demarshal_context_list_reply(header);
+	case RBUG_OP_CONTEXT_INFO_REPLY:
+		return (struct rbug_header *)rbug_demarshal_context_info_reply(header);
+	case RBUG_OP_CONTEXT_DRAW_BLOCKED:
+		return (struct rbug_header *)rbug_demarshal_context_draw_blocked(header);
+	case RBUG_OP_SHADER_LIST:
+		return (struct rbug_header *)rbug_demarshal_shader_list(header);
+	case RBUG_OP_SHADER_INFO:
+		return (struct rbug_header *)rbug_demarshal_shader_info(header);
+	case RBUG_OP_SHADER_DISABLE:
+		return (struct rbug_header *)rbug_demarshal_shader_disable(header);
+	case RBUG_OP_SHADER_REPLACE:
+		return (struct rbug_header *)rbug_demarshal_shader_replace(header);
+	case RBUG_OP_SHADER_LIST_REPLY:
+		return (struct rbug_header *)rbug_demarshal_shader_list_reply(header);
+	case RBUG_OP_SHADER_INFO_REPLY:
+		return (struct rbug_header *)rbug_demarshal_shader_info_reply(header);
+	default:
+		return NULL;
+	}
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_internal.h b/src/gallium/auxiliary/rbug/rbug_internal.h
new file mode 100644
index 0000000000..4aba1a810f
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_internal.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file is internal to the rbug protocol code, and contains asorted
+ * features needed by the code.
+ */
+
+#ifndef _RBUG_INTERNAL_H_
+#define _RBUG_INTERNAL_H_
+
+#include "rbug/rbug_proto.h"
+
+#include "util/u_memory.h"
+#include "util/u_debug.h"
+#include <errno.h>
+
+int rbug_connection_send_start(struct rbug_connection *con, enum rbug_opcode opcode, uint32_t length);
+int rbug_connection_write(struct rbug_connection *con, void *data, uint32_t size);
+int rbug_connection_send_finish(struct rbug_connection *con, uint32_t *c);
+
+/**
+ * Only works with multiples of 2
+ */
+#define PAD(from, to)                       \
+do {                                        \
+	from = (from + to - 1) & ~(to - 1); \
+} while(0)
+
+#define LEN(size)         \
+do {                      \
+	PAD(__len, size); \
+	__len += size;    \
+} while(0)
+
+#define LEN_ARRAY(size, name)       \
+do {                                \
+	LEN(4);                     \
+	PAD(__len, size);           \
+	__len += size * name##_len; \
+} while(0)
+
+#define WRITE(size, type, name)             \
+do {                                        \
+	PAD(__pos, size);                   \
+	*((type *)(&__data[__pos])) = name; \
+	__pos += size;                      \
+} while(0)
+
+#define WRITE_ARRAY(size, type, name)                    \
+do {                                                     \
+	WRITE(4, uint32_t, name##_len);                  \
+	PAD(__pos, size);                                \
+	memcpy(&__data[__pos], name, size * name##_len); \
+	__pos += size * name##_len;                      \
+} while(0)
+
+#define READ(size, type, name)                      \
+do {                                                \
+	PAD(pos, size);                             \
+	pos += size;                                \
+	if (pos > len)                              \
+		break;                              \
+	ret->name = *((type *)(&data[pos - size])); \
+} while(0)
+
+#define READ_ARRAY(size, type, name)                             \
+do {                                                             \
+	READ(4, uint32_t, name##_len);                           \
+	if (pos > len)                                           \
+		break;                                           \
+	PAD(pos, size);                                          \
+	pos += size * ret->name##_len;                           \
+	if (pos > len)                                           \
+		break;                                           \
+	ret->name = (type *)&data[pos - size * ret->name##_len]; \
+} while(0)
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_proto.h b/src/gallium/auxiliary/rbug/rbug_proto.h
new file mode 100644
index 0000000000..d273be0166
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_proto.h
@@ -0,0 +1,94 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds common definitions of the gallium remote debugging protocol.
+ */
+
+#ifndef _RBUG_PROTO_H_
+#define _RBUG_PROTO_H_
+
+#include "pipe/p_compiler.h"
+
+/**
+ * Uniqe indentifier for each command.
+ *
+ * Replys are designated by negative.
+ */
+enum rbug_opcode
+{
+	RBUG_OP_NOOP = 0,
+	RBUG_OP_PING = 1,
+	RBUG_OP_ERROR = 2,
+	RBUG_OP_PING_REPLY = -1,
+	RBUG_OP_ERROR_REPLY = -2,
+	RBUG_OP_TEXTURE_LIST = 256,
+	RBUG_OP_TEXTURE_INFO = 257,
+	RBUG_OP_TEXTURE_WRITE = 258,
+	RBUG_OP_TEXTURE_READ = 259,
+	RBUG_OP_TEXTURE_LIST_REPLY = -256,
+	RBUG_OP_TEXTURE_INFO_REPLY = -257,
+	RBUG_OP_TEXTURE_READ_REPLY = -259,
+	RBUG_OP_CONTEXT_LIST = 512,
+	RBUG_OP_CONTEXT_INFO = 513,
+	RBUG_OP_CONTEXT_DRAW_BLOCK = 514,
+	RBUG_OP_CONTEXT_DRAW_STEP = 515,
+	RBUG_OP_CONTEXT_DRAW_UNBLOCK = 516,
+	RBUG_OP_CONTEXT_DRAW_RULE = 518,
+	RBUG_OP_CONTEXT_FLUSH = 519,
+	RBUG_OP_CONTEXT_LIST_REPLY = -512,
+	RBUG_OP_CONTEXT_INFO_REPLY = -513,
+	RBUG_OP_CONTEXT_DRAW_BLOCKED = 517,
+	RBUG_OP_SHADER_LIST = 768,
+	RBUG_OP_SHADER_INFO = 769,
+	RBUG_OP_SHADER_DISABLE = 770,
+	RBUG_OP_SHADER_REPLACE = 771,
+	RBUG_OP_SHADER_LIST_REPLY = -768,
+	RBUG_OP_SHADER_INFO_REPLY = -769,
+};
+
+/**
+ * Header for demarshaled message.
+ */
+struct rbug_header
+{
+	enum rbug_opcode opcode;
+	void *__message;
+};
+
+/**
+ * Header for a message in wire format.
+ */
+struct rbug_proto_header
+{
+	int32_t opcode;
+	uint32_t length;
+};
+
+/**
+ * Forward declare connection here, as this file is included by all users.
+ */
+struct rbug_connection;
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_shader.c b/src/gallium/auxiliary/rbug/rbug_shader.c
new file mode 100644
index 0000000000..fccd2f55ef
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_shader.c
@@ -0,0 +1,468 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_shader.h"
+
+int rbug_send_shader_list(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_LIST));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_LIST, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_info(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          rbug_shader_t shader,
+                          uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* shader */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_INFO));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, shader); /* shader */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_INFO, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_disable(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint8_t disable,
+                             uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* shader */
+	LEN(1); /* disable */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_DISABLE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, shader); /* shader */
+	WRITE(1, uint8_t, disable); /* disable */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_DISABLE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_replace(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint32_t *tokens,
+                             uint32_t tokens_len,
+                             uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* context */
+	LEN(8); /* shader */
+	LEN_ARRAY(4, tokens); /* tokens */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_REPLACE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_context_t, context); /* context */
+	WRITE(8, rbug_shader_t, shader); /* shader */
+	WRITE_ARRAY(4, uint32_t, tokens); /* tokens */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_REPLACE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_list_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                rbug_shader_t *shaders,
+                                uint32_t shaders_len,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(8, shaders); /* shaders */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_LIST_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(8, rbug_shader_t, shaders); /* shaders */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_LIST_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_shader_info_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                uint32_t *original,
+                                uint32_t original_len,
+                                uint32_t *replaced,
+                                uint32_t replaced_len,
+                                uint8_t disabled,
+                                uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(4, original); /* original */
+	LEN_ARRAY(4, replaced); /* replaced */
+	LEN(1); /* disabled */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_SHADER_INFO_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(4, uint32_t, original); /* original */
+	WRITE_ARRAY(4, uint32_t, replaced); /* replaced */
+	WRITE(1, uint8_t, disabled); /* disabled */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_SHADER_INFO_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_shader_list * rbug_demarshal_shader_list(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_list *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_LIST)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+
+	return ret;
+}
+
+struct rbug_proto_shader_info * rbug_demarshal_shader_info(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_info *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_INFO)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, shader); /* shader */
+
+	return ret;
+}
+
+struct rbug_proto_shader_disable * rbug_demarshal_shader_disable(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_disable *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_DISABLE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, shader); /* shader */
+	READ(1, uint8_t, disable); /* disable */
+
+	return ret;
+}
+
+struct rbug_proto_shader_replace * rbug_demarshal_shader_replace(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_replace *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_REPLACE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_context_t, context); /* context */
+	READ(8, rbug_shader_t, shader); /* shader */
+	READ_ARRAY(4, uint32_t, tokens); /* tokens */
+
+	return ret;
+}
+
+struct rbug_proto_shader_list_reply * rbug_demarshal_shader_list_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_list_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_LIST_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(8, rbug_shader_t, shaders); /* shaders */
+
+	return ret;
+}
+
+struct rbug_proto_shader_info_reply * rbug_demarshal_shader_info_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_shader_info_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_SHADER_INFO_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(4, uint32_t, original); /* original */
+	READ_ARRAY(4, uint32_t, replaced); /* replaced */
+	READ(1, uint8_t, disabled); /* disabled */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_shader.h b/src/gallium/auxiliary/rbug/rbug_shader.h
new file mode 100644
index 0000000000..b5d886781d
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_shader.h
@@ -0,0 +1,142 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_SHADER_H_
+#define _RBUG_PROTO_SHADER_H_
+
+#include "rbug/rbug_proto.h"
+#include "rbug/rbug_core.h"
+
+struct rbug_proto_shader_list
+{
+	struct rbug_header header;
+	rbug_context_t context;
+};
+
+struct rbug_proto_shader_info
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t shader;
+};
+
+struct rbug_proto_shader_disable
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t shader;
+	uint8_t disable;
+};
+
+struct rbug_proto_shader_replace
+{
+	struct rbug_header header;
+	rbug_context_t context;
+	rbug_shader_t shader;
+	uint32_t *tokens;
+	uint32_t tokens_len;
+};
+
+struct rbug_proto_shader_list_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_shader_t *shaders;
+	uint32_t shaders_len;
+};
+
+struct rbug_proto_shader_info_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t *original;
+	uint32_t original_len;
+	uint32_t *replaced;
+	uint32_t replaced_len;
+	uint8_t disabled;
+};
+
+int rbug_send_shader_list(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          uint32_t *__serial);
+
+int rbug_send_shader_info(struct rbug_connection *__con,
+                          rbug_context_t context,
+                          rbug_shader_t shader,
+                          uint32_t *__serial);
+
+int rbug_send_shader_disable(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint8_t disable,
+                             uint32_t *__serial);
+
+int rbug_send_shader_replace(struct rbug_connection *__con,
+                             rbug_context_t context,
+                             rbug_shader_t shader,
+                             uint32_t *tokens,
+                             uint32_t tokens_len,
+                             uint32_t *__serial);
+
+int rbug_send_shader_list_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                rbug_shader_t *shaders,
+                                uint32_t shaders_len,
+                                uint32_t *__serial);
+
+int rbug_send_shader_info_reply(struct rbug_connection *__con,
+                                uint32_t serial,
+                                uint32_t *original,
+                                uint32_t original_len,
+                                uint32_t *replaced,
+                                uint32_t replaced_len,
+                                uint8_t disabled,
+                                uint32_t *__serial);
+
+struct rbug_proto_shader_list * rbug_demarshal_shader_list(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_info * rbug_demarshal_shader_info(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_disable * rbug_demarshal_shader_disable(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_replace * rbug_demarshal_shader_replace(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_list_reply * rbug_demarshal_shader_list_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_shader_info_reply * rbug_demarshal_shader_info_reply(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rbug/rbug_texture.c b/src/gallium/auxiliary/rbug/rbug_texture.c
new file mode 100644
index 0000000000..5a918fe6bc
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_texture.c
@@ -0,0 +1,631 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds the function implementation for one of the rbug extensions.
+ * Prototypes and declerations of functions and structs is in the same folder
+ * in the header file matching this file's name.
+ *
+ * The functions starting rbug_send_* encodes a call to the write format and
+ * sends that to the supplied connection, while functions starting with
+ * rbug_demarshal_* demarshal data in the wire protocol.
+ *
+ * Functions ending with _reply are replies to requests.
+ */
+
+#include "rbug_internal.h"
+#include "rbug/rbug_texture.h"
+
+int rbug_send_texture_list(struct rbug_connection *__con,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_LIST));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_LIST, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_info(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* texture */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_INFO));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_texture_t, texture); /* texture */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_INFO, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_write(struct rbug_connection *__con,
+                            rbug_texture_t texture,
+                            uint32_t face,
+                            uint32_t level,
+                            uint32_t zslice,
+                            uint32_t x,
+                            uint32_t y,
+                            uint32_t w,
+                            uint32_t h,
+                            uint8_t *data,
+                            uint32_t data_len,
+                            uint32_t stride,
+                            uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* texture */
+	LEN(4); /* face */
+	LEN(4); /* level */
+	LEN(4); /* zslice */
+	LEN(4); /* x */
+	LEN(4); /* y */
+	LEN(4); /* w */
+	LEN(4); /* h */
+	LEN_ARRAY(1, data); /* data */
+	LEN(4); /* stride */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_WRITE));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_texture_t, texture); /* texture */
+	WRITE(4, uint32_t, face); /* face */
+	WRITE(4, uint32_t, level); /* level */
+	WRITE(4, uint32_t, zslice); /* zslice */
+	WRITE(4, uint32_t, x); /* x */
+	WRITE(4, uint32_t, y); /* y */
+	WRITE(4, uint32_t, w); /* w */
+	WRITE(4, uint32_t, h); /* h */
+	WRITE_ARRAY(1, uint8_t, data); /* data */
+	WRITE(4, uint32_t, stride); /* stride */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_WRITE, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_read(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t face,
+                           uint32_t level,
+                           uint32_t zslice,
+                           uint32_t x,
+                           uint32_t y,
+                           uint32_t w,
+                           uint32_t h,
+                           uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(8); /* texture */
+	LEN(4); /* face */
+	LEN(4); /* level */
+	LEN(4); /* zslice */
+	LEN(4); /* x */
+	LEN(4); /* y */
+	LEN(4); /* w */
+	LEN(4); /* h */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_READ));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(8, rbug_texture_t, texture); /* texture */
+	WRITE(4, uint32_t, face); /* face */
+	WRITE(4, uint32_t, level); /* level */
+	WRITE(4, uint32_t, zslice); /* zslice */
+	WRITE(4, uint32_t, x); /* x */
+	WRITE(4, uint32_t, y); /* y */
+	WRITE(4, uint32_t, w); /* w */
+	WRITE(4, uint32_t, h); /* h */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_READ, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_texture_t *textures,
+                                 uint32_t textures_len,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN_ARRAY(8, textures); /* textures */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_LIST_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE_ARRAY(8, rbug_texture_t, textures); /* textures */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_LIST_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t target,
+                                 uint32_t format,
+                                 uint32_t *width,
+                                 uint32_t width_len,
+                                 uint32_t *height,
+                                 uint32_t height_len,
+                                 uint32_t *depth,
+                                 uint32_t depth_len,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint32_t last_level,
+                                 uint32_t nr_samples,
+                                 uint32_t tex_usage,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(4); /* target */
+	LEN(4); /* format */
+	LEN_ARRAY(4, width); /* width */
+	LEN_ARRAY(4, height); /* height */
+	LEN_ARRAY(4, depth); /* depth */
+	LEN(4); /* blockw */
+	LEN(4); /* blockh */
+	LEN(4); /* blocksize */
+	LEN(4); /* last_level */
+	LEN(4); /* nr_samples */
+	LEN(4); /* tex_usage */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_INFO_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(4, uint32_t, target); /* target */
+	WRITE(4, uint32_t, format); /* format */
+	WRITE_ARRAY(4, uint32_t, width); /* width */
+	WRITE_ARRAY(4, uint32_t, height); /* height */
+	WRITE_ARRAY(4, uint32_t, depth); /* depth */
+	WRITE(4, uint32_t, blockw); /* blockw */
+	WRITE(4, uint32_t, blockh); /* blockh */
+	WRITE(4, uint32_t, blocksize); /* blocksize */
+	WRITE(4, uint32_t, last_level); /* last_level */
+	WRITE(4, uint32_t, nr_samples); /* nr_samples */
+	WRITE(4, uint32_t, tex_usage); /* tex_usage */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_INFO_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+int rbug_send_texture_read_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t format,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint8_t *data,
+                                 uint32_t data_len,
+                                 uint32_t stride,
+                                 uint32_t *__serial)
+{
+	uint32_t __len = 0;
+	uint32_t __pos = 0;
+	uint8_t *__data = NULL;
+	int __ret = 0;
+
+	LEN(8); /* header */
+	LEN(4); /* serial */
+	LEN(4); /* format */
+	LEN(4); /* blockw */
+	LEN(4); /* blockh */
+	LEN(4); /* blocksize */
+	LEN_ARRAY(1, data); /* data */
+	LEN(4); /* stride */
+
+	/* align */
+	PAD(__len, 8);
+
+	__data = (uint8_t*)MALLOC(__len);
+	if (!__data)
+		return -ENOMEM;
+
+	WRITE(4, int32_t, ((int32_t)RBUG_OP_TEXTURE_READ_REPLY));
+	WRITE(4, uint32_t, ((uint32_t)(__len / 4)));
+	WRITE(4, uint32_t, serial); /* serial */
+	WRITE(4, uint32_t, format); /* format */
+	WRITE(4, uint32_t, blockw); /* blockw */
+	WRITE(4, uint32_t, blockh); /* blockh */
+	WRITE(4, uint32_t, blocksize); /* blocksize */
+	WRITE_ARRAY(1, uint8_t, data); /* data */
+	WRITE(4, uint32_t, stride); /* stride */
+
+	/* final pad */
+	PAD(__pos, 8);
+
+	if (__pos != __len) {
+		__ret = -EINVAL;
+	} else {
+		rbug_connection_send_start(__con, RBUG_OP_TEXTURE_READ_REPLY, __len);
+		rbug_connection_write(__con, __data, __len);
+		__ret = rbug_connection_send_finish(__con, __serial);
+	}
+
+	FREE(__data);
+	return __ret;
+}
+
+struct rbug_proto_texture_list * rbug_demarshal_texture_list(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_list *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_LIST)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+
+	return ret;
+}
+
+struct rbug_proto_texture_info * rbug_demarshal_texture_info(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_info *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_INFO)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_texture_t, texture); /* texture */
+
+	return ret;
+}
+
+struct rbug_proto_texture_write * rbug_demarshal_texture_write(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_write *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_WRITE)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_texture_t, texture); /* texture */
+	READ(4, uint32_t, face); /* face */
+	READ(4, uint32_t, level); /* level */
+	READ(4, uint32_t, zslice); /* zslice */
+	READ(4, uint32_t, x); /* x */
+	READ(4, uint32_t, y); /* y */
+	READ(4, uint32_t, w); /* w */
+	READ(4, uint32_t, h); /* h */
+	READ_ARRAY(1, uint8_t, data); /* data */
+	READ(4, uint32_t, stride); /* stride */
+
+	return ret;
+}
+
+struct rbug_proto_texture_read * rbug_demarshal_texture_read(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_read *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_READ)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(8, rbug_texture_t, texture); /* texture */
+	READ(4, uint32_t, face); /* face */
+	READ(4, uint32_t, level); /* level */
+	READ(4, uint32_t, zslice); /* zslice */
+	READ(4, uint32_t, x); /* x */
+	READ(4, uint32_t, y); /* y */
+	READ(4, uint32_t, w); /* w */
+	READ(4, uint32_t, h); /* h */
+
+	return ret;
+}
+
+struct rbug_proto_texture_list_reply * rbug_demarshal_texture_list_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_list_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_LIST_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ_ARRAY(8, rbug_texture_t, textures); /* textures */
+
+	return ret;
+}
+
+struct rbug_proto_texture_info_reply * rbug_demarshal_texture_info_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_info_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_INFO_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(4, uint32_t, target); /* target */
+	READ(4, uint32_t, format); /* format */
+	READ_ARRAY(4, uint32_t, width); /* width */
+	READ_ARRAY(4, uint32_t, height); /* height */
+	READ_ARRAY(4, uint32_t, depth); /* depth */
+	READ(4, uint32_t, blockw); /* blockw */
+	READ(4, uint32_t, blockh); /* blockh */
+	READ(4, uint32_t, blocksize); /* blocksize */
+	READ(4, uint32_t, last_level); /* last_level */
+	READ(4, uint32_t, nr_samples); /* nr_samples */
+	READ(4, uint32_t, tex_usage); /* tex_usage */
+
+	return ret;
+}
+
+struct rbug_proto_texture_read_reply * rbug_demarshal_texture_read_reply(struct rbug_proto_header *header)
+{
+	uint32_t len = 0;
+	uint32_t pos = 0;
+	uint8_t *data =  NULL;
+	struct rbug_proto_texture_read_reply *ret;
+
+	if (!header)
+		return NULL;
+	if (header->opcode != (int16_t)RBUG_OP_TEXTURE_READ_REPLY)
+		return NULL;
+
+	pos = 0;
+	len = header->length * 4;
+	data = (uint8_t*)&header[1];
+	ret = MALLOC(sizeof(*ret));
+	if (!ret)
+		return NULL;
+
+	ret->header.__message = header;
+	ret->header.opcode = header->opcode;
+
+	READ(4, uint32_t, serial); /* serial */
+	READ(4, uint32_t, format); /* format */
+	READ(4, uint32_t, blockw); /* blockw */
+	READ(4, uint32_t, blockh); /* blockh */
+	READ(4, uint32_t, blocksize); /* blocksize */
+	READ_ARRAY(1, uint8_t, data); /* data */
+	READ(4, uint32_t, stride); /* stride */
+
+	return ret;
+}
diff --git a/src/gallium/auxiliary/rbug/rbug_texture.h b/src/gallium/auxiliary/rbug/rbug_texture.h
new file mode 100644
index 0000000000..fbb247e1d4
--- /dev/null
+++ b/src/gallium/auxiliary/rbug/rbug_texture.h
@@ -0,0 +1,206 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * This file holds structs decelerations and function prototypes for one of
+ * the rbug extensions. Implementation of the functions is in the same folder
+ * in the c file matching this file's name.
+ *
+ * The structs what is returned from the demarshal functions. The functions
+ * starting rbug_send_* encodes a call to the write format and sends that to
+ * the supplied connection, while functions starting with rbug_demarshal_*
+ * demarshal data from the wire protocol.
+ *
+ * Structs and functions ending with _reply are replies to requests.
+ */
+
+#ifndef _RBUG_PROTO_TEXTURE_H_
+#define _RBUG_PROTO_TEXTURE_H_
+
+#include "rbug/rbug_proto.h"
+#include "rbug/rbug_core.h"
+
+struct rbug_proto_texture_list
+{
+	struct rbug_header header;
+};
+
+struct rbug_proto_texture_info
+{
+	struct rbug_header header;
+	rbug_texture_t texture;
+};
+
+struct rbug_proto_texture_write
+{
+	struct rbug_header header;
+	rbug_texture_t texture;
+	uint32_t face;
+	uint32_t level;
+	uint32_t zslice;
+	uint32_t x;
+	uint32_t y;
+	uint32_t w;
+	uint32_t h;
+	uint8_t *data;
+	uint32_t data_len;
+	uint32_t stride;
+};
+
+struct rbug_proto_texture_read
+{
+	struct rbug_header header;
+	rbug_texture_t texture;
+	uint32_t face;
+	uint32_t level;
+	uint32_t zslice;
+	uint32_t x;
+	uint32_t y;
+	uint32_t w;
+	uint32_t h;
+};
+
+struct rbug_proto_texture_list_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	rbug_texture_t *textures;
+	uint32_t textures_len;
+};
+
+struct rbug_proto_texture_info_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t target;
+	uint32_t format;
+	uint32_t *width;
+	uint32_t width_len;
+	uint32_t *height;
+	uint32_t height_len;
+	uint32_t *depth;
+	uint32_t depth_len;
+	uint32_t blockw;
+	uint32_t blockh;
+	uint32_t blocksize;
+	uint32_t last_level;
+	uint32_t nr_samples;
+	uint32_t tex_usage;
+};
+
+struct rbug_proto_texture_read_reply
+{
+	struct rbug_header header;
+	uint32_t serial;
+	uint32_t format;
+	uint32_t blockw;
+	uint32_t blockh;
+	uint32_t blocksize;
+	uint8_t *data;
+	uint32_t data_len;
+	uint32_t stride;
+};
+
+int rbug_send_texture_list(struct rbug_connection *__con,
+                           uint32_t *__serial);
+
+int rbug_send_texture_info(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t *__serial);
+
+int rbug_send_texture_write(struct rbug_connection *__con,
+                            rbug_texture_t texture,
+                            uint32_t face,
+                            uint32_t level,
+                            uint32_t zslice,
+                            uint32_t x,
+                            uint32_t y,
+                            uint32_t w,
+                            uint32_t h,
+                            uint8_t *data,
+                            uint32_t data_len,
+                            uint32_t stride,
+                            uint32_t *__serial);
+
+int rbug_send_texture_read(struct rbug_connection *__con,
+                           rbug_texture_t texture,
+                           uint32_t face,
+                           uint32_t level,
+                           uint32_t zslice,
+                           uint32_t x,
+                           uint32_t y,
+                           uint32_t w,
+                           uint32_t h,
+                           uint32_t *__serial);
+
+int rbug_send_texture_list_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 rbug_texture_t *textures,
+                                 uint32_t textures_len,
+                                 uint32_t *__serial);
+
+int rbug_send_texture_info_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t target,
+                                 uint32_t format,
+                                 uint32_t *width,
+                                 uint32_t width_len,
+                                 uint32_t *height,
+                                 uint32_t height_len,
+                                 uint32_t *depth,
+                                 uint32_t depth_len,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint32_t last_level,
+                                 uint32_t nr_samples,
+                                 uint32_t tex_usage,
+                                 uint32_t *__serial);
+
+int rbug_send_texture_read_reply(struct rbug_connection *__con,
+                                 uint32_t serial,
+                                 uint32_t format,
+                                 uint32_t blockw,
+                                 uint32_t blockh,
+                                 uint32_t blocksize,
+                                 uint8_t *data,
+                                 uint32_t data_len,
+                                 uint32_t stride,
+                                 uint32_t *__serial);
+
+struct rbug_proto_texture_list * rbug_demarshal_texture_list(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_info * rbug_demarshal_texture_info(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_write * rbug_demarshal_texture_write(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_read * rbug_demarshal_texture_read(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_list_reply * rbug_demarshal_texture_list_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_info_reply * rbug_demarshal_texture_info_reply(struct rbug_proto_header *header);
+
+struct rbug_proto_texture_read_reply * rbug_demarshal_texture_read_reply(struct rbug_proto_header *header);
+
+#endif
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
index 57fcf6de2a..1acf3c373e 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
@@ -993,6 +993,15 @@ void sse_pmovmskb( struct x86_function *p,
    emit_modrm(p, dst, src);
 }
 
+void sse_movmskps( struct x86_function *p,
+                   struct x86_reg dst,
+                   struct x86_reg src)
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x50);
+   emit_modrm(p, dst, src);
+}
+
 /***********************************************************************
  * SSE2 instructions
  */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
index 1b5eaaca85..731a651796 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
@@ -223,6 +223,7 @@ void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg sr
 void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src );
 void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movmskps( struct x86_function *p, struct x86_reg dst, struct x86_reg src);
 
 void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 5cb322a5fa..fe571a86bc 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -365,15 +365,26 @@ tgsi_exec_machine_bind_shader(
 }
 
 
-void
-tgsi_exec_machine_init(
-   struct tgsi_exec_machine *mach )
+struct tgsi_exec_machine *
+tgsi_exec_machine_create( void )
 {
+   struct tgsi_exec_machine *mach;
    uint i;
 
-   mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps);
+   mach = align_malloc( sizeof *mach, 16 );
+   if (!mach)
+      goto fail;
+
    mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR];
 
+   mach->Samplers = NULL;
+   mach->Consts = NULL;
+   mach->Tokens = NULL;
+   mach->Primitives = NULL;
+   mach->InterpCoefs = NULL;
+   mach->Instructions = NULL;
+   mach->Declarations = NULL;
+
    /* Setup constants. */
    for( i = 0; i < 4; i++ ) {
       mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000;
@@ -393,22 +404,24 @@ tgsi_exec_machine_init(
    (void) print_chan;
    (void) print_temp;
 #endif
+
+   return mach;
+
+fail:
+   align_free(mach);
+   return NULL;
 }
 
 
 void
-tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach)
+tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach)
 {
-   if (mach->Instructions) {
+   if (mach) {
       FREE(mach->Instructions);
-      mach->Instructions = NULL;
-      mach->NumInstructions = 0;
-   }
-   if (mach->Declarations) {
       FREE(mach->Declarations);
-      mach->Declarations = NULL;
-      mach->NumDeclarations = 0;
    }
+
+   align_free(mach);
 }
 
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index da22baad3e..8a9100f4c3 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -29,6 +29,7 @@
 #define TGSI_EXEC_H
 
 #include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
 
 #if defined __cplusplus
 extern "C" {
@@ -94,7 +95,6 @@ struct tgsi_exec_labels
 
 
 #define TGSI_EXEC_NUM_TEMPS       128
-#define TGSI_EXEC_NUM_TEMP_EXTRAS   6
 #define TGSI_EXEC_NUM_IMMEDIATES  256
 
 /*
@@ -162,9 +162,14 @@ struct tgsi_exec_labels
 #define TGSI_EXEC_MASK_I            (TGSI_EXEC_NUM_TEMPS + 3)
 #define TGSI_EXEC_MASK_C            2
 
+/* 4 register buffer for various purposes */
 #define TGSI_EXEC_TEMP_R0           (TGSI_EXEC_NUM_TEMPS + 4)
+#define TGSI_EXEC_NUM_TEMP_R        4
+
+#define TGSI_EXEC_TEMP_ADDR         (TGSI_EXEC_NUM_TEMPS + 8)
+#define TGSI_EXEC_NUM_ADDRS         1
+#define TGSI_EXEC_NUM_TEMP_EXTRAS   9
 
-#define TGSI_EXEC_TEMP_ADDR         (TGSI_EXEC_NUM_TEMPS + 5)
 
 
 #define TGSI_EXEC_MAX_COND_NESTING  20
@@ -187,24 +192,21 @@ struct tgsi_exec_labels
 struct tgsi_exec_machine
 {
    /* Total = program temporaries + internal temporaries
-    *         + 1 padding to align to 16 bytes
     */
-   struct tgsi_exec_vector       _Temps[TGSI_EXEC_NUM_TEMPS +
-                                        TGSI_EXEC_NUM_TEMP_EXTRAS + 1];
+   struct tgsi_exec_vector       Temps[TGSI_EXEC_NUM_TEMPS +
+                                       TGSI_EXEC_NUM_TEMP_EXTRAS];
+
+   float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
+
+   struct tgsi_exec_vector       Inputs[PIPE_MAX_ATTRIBS];
+   struct tgsi_exec_vector       Outputs[PIPE_MAX_ATTRIBS];
 
-   /*
-    * This will point to _Temps after aligning to 16B boundary.
-    */
-   struct tgsi_exec_vector       *Temps;
    struct tgsi_exec_vector       *Addrs;
 
    struct tgsi_sampler           **Samplers;
 
-   float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
    unsigned                      ImmLimit;
    const float                   (*Consts)[4];
-   struct tgsi_exec_vector       *Inputs;
-   struct tgsi_exec_vector       *Outputs;
    const struct tgsi_token       *Tokens;   /**< Declarations, instructions */
    unsigned                      Processor; /**< TGSI_PROCESSOR_x */
 
@@ -251,9 +253,11 @@ struct tgsi_exec_machine
    struct tgsi_exec_labels Labels;
 };
 
+struct tgsi_exec_machine *
+tgsi_exec_machine_create( void );
+
 void
-tgsi_exec_machine_init(
-   struct tgsi_exec_machine *mach );
+tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach);
 
 
 void 
@@ -268,10 +272,6 @@ tgsi_exec_machine_run(
    struct tgsi_exec_machine *mach );
 
 
-void
-tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach);
-
-
 static INLINE void
 tgsi_set_kill_mask(struct tgsi_exec_machine *mach, unsigned mask)
 {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
index ba2bfdef06..df49638d51 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -32,6 +32,7 @@
 #include "util/u_debug.h"
 #include "pipe/p_shader_tokens.h"
 #include "util/u_math.h"
+#include "util/u_memory.h"
 #if defined(PIPE_ARCH_SSE)
 #include "util/u_sse.h"
 #endif
@@ -100,37 +101,55 @@ get_const_base( void )
 {
    return x86_make_reg(
       file_REG32,
-      reg_CX );
+      reg_AX );
 }
 
 static struct x86_reg
-get_input_base( void )
+get_machine_base( void )
 {
    return x86_make_reg(
       file_REG32,
-      reg_AX );
+      reg_CX );
+}
+
+static struct x86_reg
+get_input_base( void )
+{
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, Inputs) );
 }
 
 static struct x86_reg
 get_output_base( void )
 {
-   return x86_make_reg(
-      file_REG32,
-      reg_DX );
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, Outputs) );
 }
 
 static struct x86_reg
 get_temp_base( void )
 {
+   return x86_make_disp(
+      get_machine_base(),
+      Offset(struct tgsi_exec_machine, Temps) );
+}
+
+static struct x86_reg
+get_coef_base( void )
+{
    return x86_make_reg(
       file_REG32,
       reg_BX );
 }
 
 static struct x86_reg
-get_coef_base( void )
+get_sampler_base( void )
 {
-   return get_output_base();
+   return x86_make_reg(
+      file_REG32,
+      reg_DI );
 }
 
 static struct x86_reg
@@ -138,7 +157,7 @@ get_immediate_base( void )
 {
    return x86_make_reg(
       file_REG32,
-      reg_DI );
+      reg_DX );
 }
 
 
@@ -168,6 +187,15 @@ get_const(
 }
 
 static struct x86_reg
+get_sampler_ptr(
+   unsigned unit )
+{
+   return x86_make_disp(
+      get_sampler_base(),
+      unit * sizeof( struct tgsi_sampler * ) );
+}
+
+static struct x86_reg
 get_input(
    unsigned vec,
    unsigned chan )
@@ -520,24 +548,15 @@ emit_coef_dady(
  * that the stack pointer is 16 byte aligned, as expected.
  */
 static void
-emit_func_call_dst(
+emit_func_call(
    struct x86_function *func,
-   unsigned xmm_save,
-   unsigned xmm_dst,
+   unsigned xmm_save_mask,
+   const struct x86_reg *arg,
+   unsigned nr_args,
    void (PIPE_CDECL *code)() )
 {
    struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
    unsigned i, n;
-   unsigned xmm_mask;
-   
-   /* Bitmask of the xmm registers to save */
-   xmm_mask = (1 << xmm_save) - 1;
-   xmm_mask &= ~(1 << xmm_dst);
-
-   sse_movaps(
-      func,
-      get_temp( TEMP_R0, 0 ),
-      make_xmm( xmm_dst ) );
 
    x86_push(
       func,
@@ -549,8 +568,10 @@ emit_func_call_dst(
       func,
       x86_make_reg( file_REG32, reg_DX) );
    
+   /* Store XMM regs to the stack
+    */
    for(i = 0, n = 0; i < 8; ++i)
-      if(xmm_mask & (1 << i))
+      if(xmm_save_mask & (1 << i))
          ++n;
    
    x86_sub_imm(
@@ -559,26 +580,42 @@ emit_func_call_dst(
       n*16);
 
    for(i = 0, n = 0; i < 8; ++i)
-      if(xmm_mask & (1 << i)) {
+      if(xmm_save_mask & (1 << i)) {
          sse_movups(
             func,
             x86_make_disp( x86_make_reg( file_REG32, reg_SP ), n*16 ),
             make_xmm( i ) );
          ++n;
       }
+
+   for (i = 0; i < nr_args; i++) {
+      /* Load the address of the buffer we use for passing arguments and
+       * receiving results:
+       */
+      x86_lea(
+	 func,
+	 ecx,
+	 arg[i] );
    
-   x86_lea(
-      func,
-      ecx,
-      get_temp( TEMP_R0, 0 ) );
-   
-   x86_push( func, ecx );
+      /* Push actual function arguments (currently just the pointer to
+       * the buffer above), and call the function:
+       */
+      x86_push( func, ecx );
+   }
+
    x86_mov_reg_imm( func, ecx, (unsigned long) code );
    x86_call( func, ecx );
-   x86_pop(func, ecx );
-   
+
+   /* Pop the arguments (or just add an immediate to esp)
+    */
+   for (i = 0; i < nr_args; i++) {
+      x86_pop(func, ecx );
+   }
+
+   /* Pop the saved XMM regs:
+    */
    for(i = 0, n = 0; i < 8; ++i)
-      if(xmm_mask & (1 << i)) {
+      if(xmm_save_mask & (1 << i)) {
          sse_movups(
             func,
             make_xmm( i ),
@@ -602,34 +639,86 @@ emit_func_call_dst(
    x86_pop(
       func,
       x86_make_reg( file_REG32, reg_AX) );
+}
+
+static void
+emit_func_call_dst_src1(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst,
+   unsigned xmm_src0,
+   void (PIPE_CDECL *code)() )
+{
+   struct x86_reg store = get_temp( TEMP_R0, 0 );
+   unsigned xmm_mask = ((1 << xmm_save) - 1) & ~(1 << xmm_dst);
+   
+   /* Store our input parameters (in xmm regs) to the buffer we use
+    * for passing arguments.  We will pass a pointer to this buffer as
+    * the actual function argument.
+    */
+   sse_movaps(
+      func,
+      store,
+      make_xmm( xmm_src0 ) );
+
+   emit_func_call( func,
+                   xmm_mask,
+                   &store,
+                   1,
+                   code );
 
    sse_movaps(
       func,
       make_xmm( xmm_dst ),
-      get_temp( TEMP_R0, 0 ) );
+      store );
 }
 
+
 static void
-emit_func_call_dst_src(
+emit_func_call_dst_src2(
    struct x86_function *func,
    unsigned xmm_save, 
    unsigned xmm_dst,
-   unsigned xmm_src,
+   unsigned xmm_src0,
+   unsigned xmm_src1,
    void (PIPE_CDECL *code)() )
 {
+   struct x86_reg store = get_temp( TEMP_R0, 0 );
+   unsigned xmm_mask = ((1 << xmm_save) - 1) & ~(1 << xmm_dst);
+
+   /* Store two inputs to parameter buffer.
+    */
    sse_movaps(
       func,
-      get_temp( TEMP_R0, 1 ),
-      make_xmm( xmm_src ) );
+      store,
+      make_xmm( xmm_src0 ) );
 
-   emit_func_call_dst(
+   sse_movaps(
       func,
-      xmm_save,
-      xmm_dst,
-      code );
+      x86_make_disp( store, 4 * sizeof(float) ),
+      make_xmm( xmm_src1 ) );
+
+
+   /* Emit the call
+    */
+   emit_func_call( func,
+                   xmm_mask,
+                   &store,
+                   1,
+                   code );
+
+   /* Retrieve the results:
+    */
+   sse_movaps(
+      func,
+      make_xmm( xmm_dst ),
+      store );
 }
 
 
+
+
+
 #if defined(PIPE_ARCH_SSE)
 
 /*
@@ -782,10 +871,11 @@ emit_cos(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save, 
       xmm_dst,
+      xmm_dst,
       cos4f );
 }
 
@@ -812,10 +902,11 @@ emit_ex2(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       ex24f );
 }
 
@@ -857,10 +948,11 @@ emit_flr(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       flr4f );
 }
 
@@ -880,10 +972,11 @@ emit_frc(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       frc4f );
 }
 
@@ -910,10 +1003,11 @@ emit_lg2(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       lg24f );
 }
 
@@ -975,13 +1069,15 @@ emit_pow(
    struct x86_function *func,
    unsigned xmm_save, 
    unsigned xmm_dst,
-   unsigned xmm_src )
+   unsigned xmm_src0,
+   unsigned xmm_src1 )
 {
-   emit_func_call_dst_src(
+   emit_func_call_dst_src2(
       func,
       xmm_save,
       xmm_dst,
-      xmm_src,
+      xmm_src0,
+      xmm_src1,
       pow4f );
 }
 
@@ -1017,10 +1113,11 @@ emit_rnd(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       rnd4f );
 }
 
@@ -1099,10 +1196,11 @@ emit_sgn(
    unsigned xmm_save, 
    unsigned xmm_dst )
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       sgn4f );
 }
 
@@ -1121,10 +1219,11 @@ emit_sin (struct x86_function *func,
           unsigned xmm_save, 
           unsigned xmm_dst)
 {
-   emit_func_call_dst(
+   emit_func_call_dst_src1(
       func,
       xmm_save,
       xmm_dst,
+      xmm_dst,
       sin4f );
 }
 
@@ -1140,6 +1239,12 @@ emit_sub(
       make_xmm( xmm_src ) );
 }
 
+
+
+
+
+
+
 /**
  * Register fetch.
  */
@@ -1298,20 +1403,164 @@ emit_store(
 #define STORE( FUNC, INST, XMM, INDEX, CHAN )\
    emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN )
 
+
+static void PIPE_CDECL
+fetch_texel( struct tgsi_sampler **sampler,
+             float *store )
+{
+#if 0
+   uint j;
+
+   debug_printf("%s sampler: %p (%p) store: %p\n", 
+                __FUNCTION__,
+                sampler, *sampler,
+                store );
+
+   debug_printf("lodbias %f\n", store[12]);
+
+   for (j = 0; j < 4; j++)
+      debug_printf("sample %d texcoord %f %f\n", 
+                   j, 
+                   store[0+j],
+                   store[4+j]);
+#endif
+
+   {
+      float rgba[NUM_CHANNELS][QUAD_SIZE];
+      (*sampler)->get_samples(*sampler, 
+                              &store[0], 
+                              &store[4], 
+                              &store[8], 
+                              0.0f, /*store[12],  lodbias */
+                              rgba);
+
+      memcpy( store, rgba, 16 * sizeof(float));
+   }
+
+#if 0
+   for (j = 0; j < 4; j++)
+      debug_printf("sample %d result %f %f %f %f\n", 
+                   j, 
+                   store[0+j],
+                   store[4+j],
+                   store[8+j],
+                   store[12+j]);
+#endif
+}
+
 /**
  * High-level instruction translators.
  */
 
 static void
+emit_tex( struct x86_function *func,
+          const struct tgsi_full_instruction *inst,
+          boolean lodbias,
+          boolean projected)
+{
+   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   struct x86_reg args[2];
+   unsigned count;
+   unsigned i;
+
+   switch (inst->InstructionExtTexture.Texture) {
+   case TGSI_TEXTURE_1D:
+   case TGSI_TEXTURE_SHADOW1D:
+      count = 1;
+      break;
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+   case TGSI_TEXTURE_SHADOW2D:
+   case TGSI_TEXTURE_SHADOWRECT:
+      count = 2;
+      break;
+   case TGSI_TEXTURE_3D:
+   case TGSI_TEXTURE_CUBE:
+      count = 3;
+      break;
+   default:
+      assert(0);
+      return;
+   }
+
+   if (lodbias) {
+      FETCH( func, *inst, 3, 0, 3 );
+   }
+   else {
+      emit_tempf(
+         func,
+         3,
+         TGSI_EXEC_TEMP_00000000_I,
+         TGSI_EXEC_TEMP_00000000_C );
+
+   }
+
+   /* store lodbias whether enabled or not -- fetch_texel currently
+    * respects it always.
+    */
+   sse_movaps( func,
+               get_temp( TEMP_R0, 3 ),
+               make_xmm( 3 ) );
+
+   
+   if (projected) {
+      FETCH( func, *inst, 3, 0, 3 );
+
+      emit_rcp( func, 3, 3 );
+   }
+
+   for (i = 0; i < count; i++) {
+      FETCH( func, *inst, i, 0, i );
+
+      if (projected) {
+         sse_mulps(
+            func,
+            make_xmm( i ),
+            make_xmm( 3 ) );
+      }
+      
+      /* Store in the argument buffer:
+       */
+      sse_movaps(
+         func,
+         get_temp( TEMP_R0, i ),
+         make_xmm( i ) );
+   }
+
+   args[0] = get_temp( TEMP_R0, 0 );
+   args[1] = get_sampler_ptr( unit );
+
+
+   emit_func_call( func,
+                   0,
+                   args,
+                   Elements(args),
+                   fetch_texel );
+
+   /* If all four channels are enabled, could use a pointer to
+    * dst[0].x instead of TEMP_R0 for store?
+    */
+   FOR_EACH_DST0_ENABLED_CHANNEL( *inst, i ) {
+
+      sse_movaps(
+         func,
+         make_xmm( 0 ),
+         get_temp( TEMP_R0, i ) );
+
+      STORE( func, *inst, 0, 0, i );
+   }
+}
+
+
+static void
 emit_kil(
    struct x86_function *func,
    const struct tgsi_full_src_register *reg )
 {
    unsigned uniquemask;
-   unsigned registers[4];
-   unsigned nextregister = 0;
-   unsigned firstchan = ~0;
+   unsigned unique_count = 0;
    unsigned chan_index;
+   unsigned i;
 
    /* This mask stores component bits that were already tested. Note that
     * we test if the value is less than zero, so 1.0 and 0.0 need not to be
@@ -1331,18 +1580,11 @@ emit_kil(
          uniquemask |= 1 << swizzle;
 
          /* allocate register */
-         registers[chan_index] = nextregister;
          emit_fetch(
             func,
-            nextregister,
+            unique_count++,
             reg,
             chan_index );
-         nextregister++;
-
-         /* mark the first channel used */
-         if( firstchan == ~0 ) {
-            firstchan = chan_index;
-         }
       }
    }
 
@@ -1353,32 +1595,32 @@ emit_kil(
       func,
       x86_make_reg( file_REG32, reg_DX ) );
 
-   FOR_EACH_CHANNEL( chan_index ) {
-      if( uniquemask & (1 << chan_index) ) {
-         sse_cmpps(
+   for (i = 0 ; i < unique_count; i++ ) {
+      struct x86_reg dataXMM = make_xmm(i);
+
+      sse_cmpps(
+         func,
+         dataXMM,
+         get_temp(
+            TGSI_EXEC_TEMP_00000000_I,
+            TGSI_EXEC_TEMP_00000000_C ),
+         cc_LessThan );
+      
+      if( i == 0 ) {
+         sse_movmskps(
             func,
-            make_xmm( registers[chan_index] ),
-            get_temp(
-               TGSI_EXEC_TEMP_00000000_I,
-               TGSI_EXEC_TEMP_00000000_C ),
-            cc_LessThan );
-
-         if( chan_index == firstchan ) {
-            sse_pmovmskb(
-               func,
-               x86_make_reg( file_REG32, reg_AX ),
-               make_xmm( registers[chan_index] ) );
-         }
-         else {
-            sse_pmovmskb(
-               func,
-               x86_make_reg( file_REG32, reg_DX ),
-               make_xmm( registers[chan_index] ) );
-            x86_or(
-               func,
-               x86_make_reg( file_REG32, reg_AX ),
-               x86_make_reg( file_REG32, reg_DX ) );
-         }
+            x86_make_reg( file_REG32, reg_AX ),
+            dataXMM );
+      }
+      else {
+         sse_movmskps(
+            func,
+            x86_make_reg( file_REG32, reg_DX ),
+            dataXMM );
+         x86_or(
+            func,
+            x86_make_reg( file_REG32, reg_AX ),
+            x86_make_reg( file_REG32, reg_DX ) );
       }
    }
 
@@ -1573,7 +1815,7 @@ emit_instruction(
                get_temp(
                   TGSI_EXEC_TEMP_MINUS_128_I,
                   TGSI_EXEC_TEMP_MINUS_128_C ) );
-            emit_pow( func, 3, 1, 2 );
+            emit_pow( func, 3, 1, 1, 2 );
             FETCH( func, *inst, 0, 0, CHAN_X );
             sse_xorps(
                func,
@@ -1917,7 +2159,7 @@ emit_instruction(
    /* TGSI_OPCODE_POW */
       FETCH( func, *inst, 0, 0, CHAN_X );
       FETCH( func, *inst, 1, 1, CHAN_X );
-      emit_pow( func, 0, 0, 1 );
+      emit_pow( func, 0, 0, 0, 1 );
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          STORE( func, *inst, 0, 0, chan_index );
       }
@@ -2086,21 +2328,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TEX:
-      if (0) {
-	 /* Disable dummy texture code: 
-	  */
-	 emit_tempf(
-	    func,
-	    0,
-	    TEMP_ONE_I,
-	    TEMP_ONE_C );
-	 FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
-	    STORE( func, *inst, 0, 0, chan_index );
-	 }
-      }
-      else {
-	 return 0;
-      }
+      emit_tex( func, inst, FALSE, FALSE );
       break;
 
    case TGSI_OPCODE_TXD:
@@ -2198,7 +2426,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXB:
-      return 0;
+      emit_tex( func, inst, TRUE, FALSE );
       break;
 
    case TGSI_OPCODE_NRM:
@@ -2306,9 +2534,13 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXL:
-      return 0;
+      emit_tex( func, inst, TRUE, FALSE );
       break;
 
+   case TGSI_OPCODE_TXP:
+      emit_tex( func, inst, FALSE, TRUE );
+      break;
+      
    case TGSI_OPCODE_BRK:
       return 0;
       break;
@@ -2488,7 +2720,7 @@ emit_declaration(
 
 static void aos_to_soa( struct x86_function *func, 
                         uint arg_aos,
-                        uint arg_soa, 
+                        uint arg_machine, 
                         uint arg_num, 
                         uint arg_stride )
 {
@@ -2503,7 +2735,10 @@ static void aos_to_soa( struct x86_function *func,
    x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
 
    x86_mov( func, aos_input,  x86_fn_arg( func, arg_aos ) );
-   x86_mov( func, soa_input,  x86_fn_arg( func, arg_soa ) );
+   x86_mov( func, soa_input,  x86_fn_arg( func, arg_machine ) );
+   x86_lea( func, soa_input,  
+	    x86_make_disp( soa_input, 
+			   Offset(struct tgsi_exec_machine, Inputs) ) );
    x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) );
    x86_mov( func, stride,     x86_fn_arg( func, arg_stride ) );
 
@@ -2545,28 +2780,30 @@ static void aos_to_soa( struct x86_function *func,
    x86_jcc( func, cc_NE, inner_loop );
 
    /* Restore EBX */
-   x86_pop( func, aos_input );
+   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
 }
 
-static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride )
+static void soa_to_aos( struct x86_function *func, 
+			uint arg_aos, 
+			uint arg_machine, 
+			uint arg_num, 
+			uint arg_stride )
 {
-   struct x86_reg soa_output;
-   struct x86_reg aos_output;
-   struct x86_reg num_outputs;
-   struct x86_reg temp;
+   struct x86_reg soa_output = x86_make_reg( file_REG32, reg_AX );
+   struct x86_reg aos_output = x86_make_reg( file_REG32, reg_BX );
+   struct x86_reg num_outputs = x86_make_reg( file_REG32, reg_CX );
+   struct x86_reg temp = x86_make_reg( file_REG32, reg_DX );
    int inner_loop;
 
-   soa_output = x86_make_reg( file_REG32, reg_AX );
-   aos_output = x86_make_reg( file_REG32, reg_BX );
-   num_outputs = x86_make_reg( file_REG32, reg_CX );
-   temp = x86_make_reg( file_REG32, reg_DX );
-
    /* Save EBX */
-   x86_push( func, aos_output );
+   x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
 
-   x86_mov( func, soa_output, x86_fn_arg( func, soa ) );
-   x86_mov( func, aos_output, x86_fn_arg( func, aos ) );
-   x86_mov( func, num_outputs, x86_fn_arg( func, num ) );
+   x86_mov( func, aos_output, x86_fn_arg( func, arg_aos ) );
+   x86_mov( func, soa_output, x86_fn_arg( func, arg_machine ) );
+   x86_lea( func, soa_output, 
+	    x86_make_disp( soa_output, 
+			   Offset(struct tgsi_exec_machine, Outputs) ) );
+   x86_mov( func, num_outputs, x86_fn_arg( func, arg_num ) );
 
    /* do */
    inner_loop = x86_get_label( func );
@@ -2583,7 +2820,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
       sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) );
       sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) );
 
-      x86_mov( func, temp, x86_fn_arg( func, stride ) );
+      x86_mov( func, temp, x86_fn_arg( func, arg_stride ) );
       x86_push( func, aos_output );
       sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) );
       sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) );
@@ -2607,20 +2844,13 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num,
    x86_jcc( func, cc_NE, inner_loop );
 
    /* Restore EBX */
-   x86_pop( func, aos_output );
+   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
 }
 
 /**
  * Translate a TGSI vertex/fragment shader to SSE2 code.
  * Slightly different things are done for vertex vs. fragment shaders.
  *
- * Note that fragment shaders are responsible for interpolating shader
- * inputs. Because on x86 we have only 4 GP registers, and here we
- * have 5 shader arguments (input, output, const, temp and coef), the
- * code is split into two phases -- DECLARATION and INSTRUCTION phase.
- * GP register holding the output argument is aliased with the coeff
- * argument, as outputs are not needed in the DECLARATION phase.
- *
  * \param tokens  the TGSI input shader
  * \param func  the output SSE code/function
  * \param immediates  buffer to place immediates, later passed to SSE func
@@ -2634,7 +2864,6 @@ tgsi_emit_sse2(
    boolean do_swizzles )
 {
    struct tgsi_parse_context parse;
-   boolean instruction_phase = FALSE;
    unsigned ok = 1;
    uint num_immediates = 0;
 
@@ -2646,74 +2875,48 @@ tgsi_emit_sse2(
 
    /* Can't just use EDI, EBX without save/restoring them:
     */
-   x86_push(
-      func,
-      get_immediate_base() );
-
-   x86_push(
-      func,
-      get_temp_base() );
-
+   x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
+   x86_push( func, x86_make_reg( file_REG32, reg_DI ) );
 
    /*
     * Different function args for vertex/fragment shaders:
     */
-   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
-      /* DECLARATION phase, do not load output argument. */
-      x86_mov(
-         func,
-         get_input_base(),
-         x86_fn_arg( func, 1 ) );
-      /* skipping outputs argument here */
-      x86_mov(
-         func,
-         get_const_base(),
-         x86_fn_arg( func, 3 ) );
-      x86_mov(
-         func,
-         get_temp_base(),
-         x86_fn_arg( func, 4 ) );
-      x86_mov(
-         func,
-         get_coef_base(),
-         x86_fn_arg( func, 5 ) );
-      x86_mov(
-         func,
-         get_immediate_base(),
-         x86_fn_arg( func, 6 ) );
-   }
-   else {
-      assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX);
-
+   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) {
       if (do_swizzles)
          aos_to_soa( func, 
-                     6,         /* aos_input */
-                     1,         /* machine->input */
-                     7,         /* num_inputs */
-                     8 );       /* input_stride */
+                     4,         /* aos_input */
+                     1,         /* machine */
+                     5,         /* num_inputs */
+                     6 );       /* input_stride */
+   }
 
+   x86_mov(
+      func,
+      get_machine_base(),
+      x86_fn_arg( func, 1 ) );
+   x86_mov(
+      func,
+      get_const_base(),
+      x86_fn_arg( func, 2 ) );
+   x86_mov(
+      func,
+      get_immediate_base(),
+      x86_fn_arg( func, 3 ) );
+
+   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
       x86_mov(
-         func,
-         get_input_base(),
-         x86_fn_arg( func, 1 ) );
-      x86_mov(
-         func,
-         get_output_base(),
-         x86_fn_arg( func, 2 ) );
-      x86_mov(
-         func,
-         get_const_base(),
-         x86_fn_arg( func, 3 ) );
-      x86_mov(
-         func,
-         get_temp_base(),
-         x86_fn_arg( func, 4 ) );
+	 func,
+	 get_coef_base(),
+	 x86_fn_arg( func, 4 ) );
+
       x86_mov(
-         func,
-         get_immediate_base(),
-         x86_fn_arg( func, 5 ) );
+	 func,
+	 get_sampler_base(),
+	 x86_make_disp( get_machine_base(),
+                        Offset( struct tgsi_exec_machine, Samplers ) ) );
    }
 
+
    while( !tgsi_parse_end_of_tokens( &parse ) && ok ) {
       tgsi_parse_token( &parse );
 
@@ -2727,17 +2930,6 @@ tgsi_emit_sse2(
          break;
 
       case TGSI_TOKEN_TYPE_INSTRUCTION:
-         if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
-            if( !instruction_phase ) {
-               /* INSTRUCTION phase, overwrite coeff with output. */
-               instruction_phase = TRUE;
-               x86_mov(
-                  func,
-                  get_output_base(),
-                  x86_fn_arg( func, 2 ) );
-            }
-         }
-
          ok = emit_instruction(
             func,
             &parse.FullToken.FullInstruction );
@@ -2781,18 +2973,17 @@ tgsi_emit_sse2(
 
    if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) {
       if (do_swizzles)
-         soa_to_aos( func, 9, 2, 10, 11 );
+         soa_to_aos( func, 
+		     7, 	/* aos_output */
+		     1, 	/* machine */
+		     8, 	/* num_outputs */
+		     9 );	/* output_stride */
    }
 
    /* Can't just use EBX, EDI without save/restoring them:
     */
-   x86_pop(
-      func,
-      get_temp_base() );
-
-   x86_pop(
-      func,
-      get_immediate_base() );
+   x86_pop( func, x86_make_reg( file_REG32, reg_DI ) );
+   x86_pop( func, x86_make_reg( file_REG32, reg_BX ) );
 
    emit_ret( func );
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/tgsi_sse2.h
index af838b2a25..d81ee3d00e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.h
@@ -34,6 +34,7 @@ extern "C" {
 
 struct tgsi_token;
 struct x86_function;
+struct tgsi_interp_coef;
 
 unsigned
 tgsi_emit_sse2(
@@ -42,6 +43,33 @@ tgsi_emit_sse2(
    float (*immediates)[4],
    boolean do_swizzles );
 
+
+/* This is the function prototype generated when do_swizzles is false
+ * -- effectively for fragment shaders.
+ */
+typedef void (PIPE_CDECL *tgsi_sse2_fs_function) (
+   struct tgsi_exec_machine *machine, /* 1 */
+   const float (*constant)[4],		    /* 2 */
+   const float (*immediate)[4],		    /* 3 */
+   const struct tgsi_interp_coef *coef	    /* 4 */
+   );
+
+
+/* This is the function prototype generated when do_swizzles is true
+ * -- effectively for vertex shaders.
+ */
+typedef void (PIPE_CDECL *tgsi_sse2_vs_func) (
+   struct tgsi_exec_machine *machine, /* 1 */
+   const float (*constant)[4],        /* 2 */
+   const float (*immediate)[4],       /* 3 */
+   const float (*aos_input)[4], /* 4 */
+   uint num_inputs,             /* 5 */
+   uint input_stride,           /* 6 */
+   float (*aos_output)[4],      /* 7 */
+   uint num_outputs,            /* 8 */
+   uint output_stride );        /* 9 */
+
+
 #if defined __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile
index 2995aba1b9..6a8eb73e84 100644
--- a/src/gallium/auxiliary/util/Makefile
+++ b/src/gallium/auxiliary/util/Makefile
@@ -16,6 +16,7 @@ C_SOURCES = \
 	u_hash.c \
 	u_keymap.c \
 	u_linear.c \
+	u_network.c \
 	u_math.c \
 	u_mm.c \
 	u_rect.c \
diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript
index d3ac7f747f..fb142eebca 100644
--- a/src/gallium/auxiliary/util/SConscript
+++ b/src/gallium/auxiliary/util/SConscript
@@ -17,6 +17,7 @@ util = env.ConvenienceLibrary(
 		'u_hash.c',
 		'u_hash_table.c',
 		'u_keymap.c',
+		'u_network.c',
 		'u_math.c',
 		'u_mm.c',
 		'u_rect.c',
diff --git a/src/gallium/auxiliary/util/u_cache.c b/src/gallium/auxiliary/util/u_cache.c
index 41cd38171f..47c16b1c92 100644
--- a/src/gallium/auxiliary/util/u_cache.c
+++ b/src/gallium/auxiliary/util/u_cache.c
@@ -137,6 +137,8 @@ util_cache_set(struct util_cache *cache,
    struct util_cache_entry *entry;
 
    assert(cache);
+   if (!cache)
+      return;
 
    entry = util_cache_entry_get(cache, key);
    util_cache_entry_destroy(cache, entry);
@@ -158,6 +160,8 @@ util_cache_get(struct util_cache *cache,
    struct util_cache_entry *entry;
 
    assert(cache);
+   if (!cache)
+      return NULL;
 
    entry = util_cache_entry_get(cache, key);
    if(!entry->key && !entry->value)
@@ -176,7 +180,9 @@ util_cache_clear(struct util_cache *cache)
    uint32_t i;
 
    assert(cache);
-   
+   if (!cache)
+      return;
+
    for(i = 0; i < cache->size; ++i)
       util_cache_entry_destroy(cache, &cache->entries[i]);
 }
@@ -186,6 +192,8 @@ void
 util_cache_destroy(struct util_cache *cache)
 {
    assert(cache);
+   if (!cache)
+      return;
 
 #ifdef DEBUG
    if(cache->count >= 20*cache->size) {
diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c
index 6da7353e25..3703718a62 100644
--- a/src/gallium/auxiliary/util/u_handle_table.c
+++ b/src/gallium/auxiliary/util/u_handle_table.c
@@ -87,6 +87,8 @@ handle_table_set_destroy(struct handle_table *ht,
                          void (*destroy)(void *object))
 {
    assert(ht);
+   if (!ht)
+      return;
    ht->destroy = destroy;
 }
 
@@ -155,7 +157,7 @@ handle_table_add(struct handle_table *ht,
    
    assert(ht);
    assert(object);
-   if(!object)
+   if(!object || !ht)
       return 0;
 
    /* linear search for an empty handle */
@@ -193,7 +195,7 @@ handle_table_set(struct handle_table *ht,
    
    assert(ht);
    assert(handle);
-   if(!handle)
+   if(!handle || !ht)
       return 0;
 
    assert(object);
@@ -222,7 +224,7 @@ handle_table_get(struct handle_table *ht,
    
    assert(ht);
    assert(handle);
-   if(!handle || handle > ht->size)
+   if(!handle || !ht || handle > ht->size)
       return NULL;
 
    object = ht->objects[handle - 1];
@@ -240,7 +242,7 @@ handle_table_remove(struct handle_table *ht,
    
    assert(ht);
    assert(handle);
-   if(!handle || handle > ht->size)
+   if(!handle || !ht || handle > ht->size)
       return;
 
    index = handle - 1;
@@ -283,6 +285,9 @@ handle_table_destroy(struct handle_table *ht)
    unsigned index;
    assert(ht);
 
+   if (!ht)
+      return;
+
    if(ht->destroy)
       for(index = 0; index < ht->size; ++index)
          handle_table_clear(ht, index);
diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c
index 2f83e318e4..8c2a8f454c 100644
--- a/src/gallium/auxiliary/util/u_hash_table.c
+++ b/src/gallium/auxiliary/util/u_hash_table.c
@@ -148,6 +148,8 @@ hash_table_set(struct hash_table *ht,
    struct cso_hash_iter iter;
 
    assert(ht);
+   if (!ht)
+      return PIPE_ERROR_BAD_INPUT;
 
    key_hash = ht->hash(key);
 
@@ -183,6 +185,8 @@ hash_table_get(struct hash_table *ht,
    struct hash_table_item *item;
 
    assert(ht);
+   if (!ht)
+      return NULL;
 
    key_hash = ht->hash(key);
 
@@ -203,6 +207,8 @@ hash_table_remove(struct hash_table *ht,
    struct hash_table_item *item;
 
    assert(ht);
+   if (!ht)
+      return;
 
    key_hash = ht->hash(key);
 
@@ -225,7 +231,9 @@ hash_table_clear(struct hash_table *ht)
    struct hash_table_item *item;
 
    assert(ht);
-   
+   if (!ht)
+      return;
+
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter));
@@ -243,9 +251,11 @@ hash_table_foreach(struct hash_table *ht,
    struct cso_hash_iter iter;
    struct hash_table_item *item;
    enum pipe_error result;
-   
+
    assert(ht);
-   
+   if (!ht)
+      return PIPE_ERROR_BAD_INPUT;
+
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_iter_data(iter);
@@ -264,9 +274,11 @@ hash_table_destroy(struct hash_table *ht)
 {
    struct cso_hash_iter iter;
    struct hash_table_item *item;
-   
+
    assert(ht);
-   
+   if (!ht)
+      return;
+
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct hash_table_item *)cso_hash_iter_data(iter);
diff --git a/src/gallium/auxiliary/util/u_keymap.c b/src/gallium/auxiliary/util/u_keymap.c
index 3f70809efd..508a2ee063 100644
--- a/src/gallium/auxiliary/util/u_keymap.c
+++ b/src/gallium/auxiliary/util/u_keymap.c
@@ -194,6 +194,8 @@ util_keymap_insert(struct keymap *map, const void *key,
    struct cso_hash_iter iter;
 
    assert(map);
+   if (!map)
+      return FALSE;
 
    key_hash = hash(key, map->key_size);
 
@@ -234,6 +236,8 @@ util_keymap_lookup(const struct keymap *map, const void *key)
    struct keymap_item *item;
 
    assert(map);
+   if (!map)
+      return NULL;
 
    key_hash = hash(key, map->key_size);
 
@@ -258,6 +262,8 @@ util_keymap_remove(struct keymap *map, const void *key, void *user)
    struct keymap_item *item;
 
    assert(map);
+   if (!map)
+      return;
 
    key_hash = hash(key, map->key_size);
 
@@ -267,6 +273,8 @@ util_keymap_remove(struct keymap *map, const void *key, void *user)
    
    item = hash_table_item(iter);
    assert(item);
+   if (!item)
+      return;
    map->delete_func(map, item->key, item->value, user);
    FREE(item->key);
    FREE(item);
@@ -288,7 +296,9 @@ util_keymap_remove_all(struct keymap *map, void *user)
    struct keymap_item *item;
 
    assert(map);
-   
+   if (!map)
+      return;
+
    iter = cso_hash_first_node(map->cso);
    while (!cso_hash_iter_is_null(iter)) {
       item = (struct keymap_item *)
diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c
new file mode 100644
index 0000000000..07d804ecdb
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_network.c
@@ -0,0 +1,188 @@
+
+#include "pipe/p_compiler.h"
+#include "util/u_network.h"
+#include "util/u_debug.h"
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+#  include <winsock2.h>
+#  include <windows.h>
+#elif defined(PIPE_OS_LINUX)
+#  include <sys/socket.h>
+#  include <netinet/in.h>
+#  include <unistd.h>
+#  include <fcntl.h>
+#  include <netdb.h>
+#else
+#  warning "No socket implementation"
+#endif
+
+boolean
+u_socket_init()
+{
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   WORD wVersionRequested;
+   WSADATA wsaData;
+   int err;
+
+   /* Use the MAKEWORD(lowbyte, highbyte) macro declared in Windef.h */
+   wVersionRequested = MAKEWORD(1, 1);
+
+   err = WSAStartup(wVersionRequested, &wsaData);
+   if (err != 0) {
+      debug_printf("WSAStartup failed with error: %d\n", err);
+      return FALSE;
+   }
+   return TRUE;
+#elif defined(PIPE_HAVE_SOCKETS)
+   return TRUE;
+#else
+   return FALSE;
+#endif
+}
+
+void
+u_socket_stop()
+{
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   WSACleanup();
+#endif
+}
+
+void
+u_socket_close(int s)
+{
+   if (s < 0)
+      return;
+
+#if defined(PIPE_OS_LINUX)
+   shutdown(s, SHUT_RDWR);
+   close(s);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   shutdown(s, SD_BOTH);
+   closesocket(s);
+#else
+   assert(0);
+#endif
+}
+
+int u_socket_accept(int s)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return accept(s, NULL, NULL);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_send(int s, void *data, size_t size)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return send(s, data, size, 0);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_peek(int s, void *data, size_t size)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return recv(s, data, size, MSG_PEEK);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_recv(int s, void *data, size_t size)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   return recv(s, data, size, 0);
+#else
+   return -1;
+#endif
+}
+
+int
+u_socket_connect(const char *hostname, uint16_t port)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   int s;
+   struct sockaddr_in sa;
+   struct hostent *host = NULL;
+
+   memset(&sa, 0, sizeof(struct sockaddr_in));
+   host = gethostbyname(hostname);
+   if (!host)
+      return -1;
+
+   memcpy((char *)&sa.sin_addr,host->h_addr,host->h_length);
+   sa.sin_family= host->h_addrtype;
+   sa.sin_port = htons(port);
+
+   s = socket(host->h_addrtype, SOCK_STREAM, IPPROTO_TCP);
+   if (s < 0)
+      return -1;
+
+   if (connect(s, (struct sockaddr *)&sa, sizeof(sa))) {
+      u_socket_close(s);
+      return -1;
+   }
+
+   return s;
+#else
+   assert(0);
+   return -1;
+#endif
+}
+
+int
+u_socket_listen_on_port(uint16_t portnum)
+{
+#if defined(PIPE_HAVE_SOCKETS)
+   int s;
+   struct sockaddr_in sa;
+   memset(&sa, 0, sizeof(struct sockaddr_in));
+
+   sa.sin_family = AF_INET;
+   sa.sin_port = htons(portnum);
+
+   s = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+   if (s < 0)
+      return -1;
+
+   if (bind(s, (struct sockaddr *)&sa, sizeof(struct sockaddr_in)) == -1) {
+      u_socket_close(s);
+      return -1;
+   }
+
+   listen(s, 0);
+
+   return s;
+#else
+   assert(0);
+   return -1;
+#endif
+}
+
+void
+u_socket_block(int s, boolean block)
+{
+#if defined(PIPE_OS_LINUX)
+   int old = fcntl(s, F_GETFL, 0);
+   if (old == -1)
+      return;
+
+   /* TODO obey block */
+   if (block)
+      fcntl(s, F_SETFL, old & ~O_NONBLOCK);
+   else
+      fcntl(s, F_SETFL, old | O_NONBLOCK);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   u_long iMode = block ? 0 : 1;
+   ioctlsocket(s, FIONBIO, &iMode);
+#else
+   assert(0);
+#endif
+}
diff --git a/src/gallium/auxiliary/util/u_network.h b/src/gallium/auxiliary/util/u_network.h
new file mode 100644
index 0000000000..14d3884427
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_network.h
@@ -0,0 +1,24 @@
+
+#ifndef _U_NETWORK_H_
+#define _U_NETWORK_H_
+
+#include "pipe/p_compiler.h"
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+#  define PIPE_HAVE_SOCKETS
+#elif defined(PIPE_OS_LINUX)
+#  define PIPE_HAVE_SOCKETS
+#endif
+
+boolean u_socket_init(void);
+void u_socket_stop(void);
+void u_socket_close(int s);
+int u_socket_listen_on_port(uint16_t portnum);
+int u_socket_accept(int s);
+int u_socket_connect(const char *host, uint16_t port);
+int u_socket_send(int s, void *data, size_t size);
+int u_socket_peek(int s, void *data, size_t size);
+int u_socket_recv(int s, void *data, size_t size);
+void u_socket_block(int s, boolean block);
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h
index d7c3995dbf..a9b533eea7 100644
--- a/src/gallium/auxiliary/util/u_prim.h
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -119,7 +119,7 @@ static INLINE boolean u_trim_pipe_prim( unsigned pipe_prim, unsigned *nr )
 }
 
 
-static INLINE boolean u_reduced_prim( unsigned pipe_prim )
+static INLINE unsigned u_reduced_prim( unsigned pipe_prim )
 {
    switch (pipe_prim) {
    case PIPE_PRIM_POINTS:
diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c
index 7c225e2f27..5ffb7073ab 100644
--- a/src/gallium/drivers/cell/spu/spu_render.c
+++ b/src/gallium/drivers/cell/spu/spu_render.c
@@ -32,6 +32,7 @@
 
 #include "spu_main.h"
 #include "spu_render.h"
+#include "spu_shuffle.h"
 #include "spu_tri.h"
 #include "spu_tile.h"
 #include "cell/common.h"
@@ -267,15 +268,75 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr)
 
       uint drawn = 0;
 
-      /* loop over tris */
-      for (j = 0; j < render->num_indexes; j += 3) {
-         const float *v0, *v1, *v2;
-
-         v0 = (const float *) (vertices + indexes[j+0] * vertex_size);
-         v1 = (const float *) (vertices + indexes[j+1] * vertex_size);
-         v2 = (const float *) (vertices + indexes[j+2] * vertex_size);
-
-         drawn += tri_draw(v0, v1, v2, tx, ty);
+      const qword vertex_sizes = (qword)spu_splats(vertex_size);
+      const qword verticess = (qword)spu_splats((uint)vertices);
+
+      ASSERT_ALIGN16(&indexes[0]);
+
+      const uint num_indexes = render->num_indexes;
+
+      /* loop over tris
+	   * &indexes[0] will be 16 byte aligned.  This loop is heavily unrolled
+	   * avoiding variable rotates when extracting vertex indices.
+	   */
+      for (j = 0; j < num_indexes; j += 24) {
+         /* Load three vectors, containing 24 ushort indices */
+         const qword* lower_qword = (qword*)&indexes[j];
+         const qword indices0 = lower_qword[0];
+         const qword indices1 = lower_qword[1];
+         const qword indices2 = lower_qword[2];
+
+         /* stores three indices for each tri n in slots 0, 1 and 2 of vsn */
+		 /* Straightforward rotates for these */
+         qword vs0 = indices0;
+         qword vs1 = si_shlqbyi(indices0, 6);
+         qword vs3 = si_shlqbyi(indices1, 2);
+         qword vs4 = si_shlqbyi(indices1, 8);
+         qword vs6 = si_shlqbyi(indices2, 4);
+         qword vs7 = si_shlqbyi(indices2, 10);
+
+         /* For tri 2 and 5, the three indices are split across two machine
+		  * words - rotate and combine */
+         const qword tmp2a = si_shlqbyi(indices0, 12);
+         const qword tmp2b = si_rotqmbyi(indices1, 12|16);
+         qword vs2 = si_selb(tmp2a, tmp2b, si_fsmh(si_from_uint(0x20)));
+
+         const qword tmp5a = si_shlqbyi(indices1, 14);
+         const qword tmp5b = si_rotqmbyi(indices2, 14|16);
+         qword vs5 = si_selb(tmp5a, tmp5b, si_fsmh(si_from_uint(0x60)));
+
+         /* unpack indices from halfword slots to word slots */
+         vs0 = si_shufb(vs0, vs0, SHUFB8(0,A,0,B,0,C,0,0));
+         vs1 = si_shufb(vs1, vs1, SHUFB8(0,A,0,B,0,C,0,0));
+         vs2 = si_shufb(vs2, vs2, SHUFB8(0,A,0,B,0,C,0,0));
+         vs3 = si_shufb(vs3, vs3, SHUFB8(0,A,0,B,0,C,0,0));
+         vs4 = si_shufb(vs4, vs4, SHUFB8(0,A,0,B,0,C,0,0));
+         vs5 = si_shufb(vs5, vs5, SHUFB8(0,A,0,B,0,C,0,0));
+         vs6 = si_shufb(vs6, vs6, SHUFB8(0,A,0,B,0,C,0,0));
+         vs7 = si_shufb(vs7, vs7, SHUFB8(0,A,0,B,0,C,0,0));
+
+         /* Calculate address of vertex in vertices[] */
+         vs0 = si_mpya(vs0, vertex_sizes, verticess);
+         vs1 = si_mpya(vs1, vertex_sizes, verticess);
+         vs2 = si_mpya(vs2, vertex_sizes, verticess);
+         vs3 = si_mpya(vs3, vertex_sizes, verticess);
+         vs4 = si_mpya(vs4, vertex_sizes, verticess);
+         vs5 = si_mpya(vs5, vertex_sizes, verticess);
+         vs6 = si_mpya(vs6, vertex_sizes, verticess);
+         vs7 = si_mpya(vs7, vertex_sizes, verticess);
+
+         /* Select the appropriate call based on the number of vertices 
+		  * remaining */
+         switch(num_indexes - j) {
+            default: drawn += tri_draw(vs7, tx, ty);
+            case 21: drawn += tri_draw(vs6, tx, ty);
+            case 18: drawn += tri_draw(vs5, tx, ty);
+            case 15: drawn += tri_draw(vs4, tx, ty);
+            case 12: drawn += tri_draw(vs3, tx, ty);
+            case 9:  drawn += tri_draw(vs2, tx, ty);
+            case 6:  drawn += tri_draw(vs1, tx, ty);
+            case 3:  drawn += tri_draw(vs0, tx, ty);
+         }
       }
 
       //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3);
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
index d727268475..58be001be4 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.c
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
@@ -133,7 +133,15 @@ struct setup_stage {
 
    uint tx, ty;  /**< position of current tile (x, y) */
 
-   int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy;
+   union {
+      struct {
+         int cliprect_minx;
+         int cliprect_miny;
+         int cliprect_maxx;
+         int cliprect_maxy;
+      };
+      qword cliprect;
+   };
 
    struct interp_coef coef[PIPE_MAX_SHADER_INPUTS];
 
@@ -432,6 +440,41 @@ print_vertex(const struct vertex_header *v)
 }
 #endif
 
+/* Returns the minimum of each slot of two vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n]);
+ */
+static qword
+minfq(qword q0, qword q1)
+{
+   const qword q0q1m = si_fcgt(q0, q1);
+   return si_selb(q0, q1, q0q1m);
+}
+
+/* Returns the minimum of each slot of three vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n],q2[n]);
+ */
+static qword
+min3fq(qword q0, qword q1, qword q2)
+{
+   return minfq(minfq(q0, q1), q2);
+}
+
+/* Returns the maximum of each slot of two vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n],q2[n]);
+ */
+static qword
+maxfq(qword q0, qword q1) {
+   const qword q0q1m = si_fcgt(q0, q1);
+   return si_selb(q1, q0, q0q1m);
+}
+
+/* Returns the maximum of each slot of three vec_float4s as qwords.
+ * i.e. return[n] = min(q0[n],q1[n],q2[n]);
+ */
+static qword
+max3fq(qword q0, qword q1, qword q2) {
+   return maxfq(maxfq(q0, q1), q2);
+}
 
 /**
  * Sort vertices from top to bottom.
@@ -440,9 +483,7 @@ print_vertex(const struct vertex_header *v)
  * \return  FALSE if tri is totally outside tile, TRUE otherwise
  */
 static boolean
-setup_sort_vertices(const struct vertex_header *v0,
-                    const struct vertex_header *v1,
-                    const struct vertex_header *v2)
+setup_sort_vertices(const qword vs)
 {
    float area, sign;
 
@@ -455,57 +496,57 @@ setup_sort_vertices(const struct vertex_header *v0,
    }
 #endif
 
-   /* determine bottom to top order of vertices */
    {
+      /* Load the float values for various processing... */
+      const qword f0 = (qword)(((const struct vertex_header*)si_to_ptr(vs))->data[0]);
+      const qword f1 = (qword)(((const struct vertex_header*)si_to_ptr(si_rotqbyi(vs, 4)))->data[0]);
+      const qword f2 = (qword)(((const struct vertex_header*)si_to_ptr(si_rotqbyi(vs, 8)))->data[0]);
+
+      /* Check if triangle is completely outside the tile bounds
+       * Find the min and max x and y positions of the three poits */
+      const qword minf = min3fq(f0, f1, f2);
+      const qword maxf = max3fq(f0, f1, f2);
+
+      /* Compare min and max against cliprect vals */
+      const qword maxsmins = si_shufb(maxf, minf, SHUFB4(A,B,a,b));
+      const qword outside = si_fcgt(maxsmins, si_csflt(setup.cliprect, 0));
+
+      /* Use a little magic to work out of the tri is visible or not */
+      if(si_to_uint(si_xori(si_gb(outside), 0xc))) return FALSE;
+
+      /* determine bottom to top order of vertices */
       /* A table of shuffle patterns for putting vertex_header pointers into
          correct order.  Quite magical. */
-      const vec_uchar16 sort_order_patterns[] = {
-         SHUFFLE4(A,B,C,C),
-         SHUFFLE4(C,A,B,C),
-         SHUFFLE4(A,C,B,C),
-         SHUFFLE4(B,C,A,C),
-         SHUFFLE4(B,A,C,C),
-         SHUFFLE4(C,B,A,C) };
-
-      /* The vertex_header pointers, packed for easy shuffling later */
-      const vec_uint4 vs = {(unsigned)v0, (unsigned)v1, (unsigned)v2};
+      const qword sort_order_patterns[] = {
+         SHUFB4(A,B,C,C),
+         SHUFB4(C,A,B,C),
+         SHUFB4(A,C,B,C),
+         SHUFB4(B,C,A,C),
+         SHUFB4(B,A,C,C),
+         SHUFB4(C,B,A,C) };
 
       /* Collate y values into two vectors for comparison.
          Using only one shuffle constant! ;) */
-      const vec_float4 y_02_ = spu_shuffle(v0->data[0], v2->data[0], SHUFFLE4(0,B,b,C));
-      const vec_float4 y_10_ = spu_shuffle(v1->data[0], v0->data[0], SHUFFLE4(0,B,b,C));
-      const vec_float4 y_012 = spu_shuffle(y_02_, v1->data[0], SHUFFLE4(0,B,b,C));
-      const vec_float4 y_120 = spu_shuffle(y_10_, v2->data[0], SHUFFLE4(0,B,b,C));
+      const qword y_02_ = si_shufb(f0, f2, SHUFB4(0,B,b,C));
+      const qword y_10_ = si_shufb(f1, f0, SHUFB4(0,B,b,C));
+      const qword y_012 = si_shufb(y_02_, f1, SHUFB4(0,B,b,C));
+      const qword y_120 = si_shufb(y_10_, f2, SHUFB4(0,B,b,C));
 
       /* Perform comparison: {y0,y1,y2} > {y1,y2,y0} */
-      const vec_uint4 compare = spu_cmpgt(y_012, y_120);
+      const qword compare = si_fcgt(y_012, y_120);
       /* Compress the result of the comparison into 4 bits */
-      const vec_uint4 gather = spu_gather(compare);
+      const qword gather = si_gb(compare);
       /* Subtract one to attain the index into the LUT.  Magical. */
-      const unsigned int index = spu_extract(gather, 0) - 1;
+      const unsigned int index = si_to_uint(gather) - 1;
 
       /* Load the appropriate pattern and construct the desired vector. */
-      setup.vertex_headers = (qword)spu_shuffle(vs, vs, sort_order_patterns[index]);
+      setup.vertex_headers = si_shufb(vs, vs, sort_order_patterns[index]);
 
       /* Using the result of the comparison, set sign.
          Very magical. */
-      sign = ((si_to_uint(si_cntb((qword)gather)) == 2) ? 1.0f : -1.0f);
+      sign = ((si_to_uint(si_cntb(gather)) == 2) ? 1.0f : -1.0f);
    }
 
-   /* Check if triangle is completely outside the tile bounds */
-   if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy)
-      return FALSE;
-   if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny)
-      return FALSE;
-   if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx &&
-       spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx &&
-       spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx)
-      return FALSE;
-   if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx &&
-       spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx &&
-       spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx)
-      return FALSE;
-
    setup.ebot.ds = spu_sub(setup.vmid->data[0], setup.vmin->data[0]);
    setup.emaj.ds = spu_sub(setup.vmax->data[0], setup.vmin->data[0]);
    setup.etop.ds = spu_sub(setup.vmax->data[0], setup.vmid->data[0]);
@@ -761,21 +802,19 @@ subtriangle(struct edge *eleft, struct edge *eright, unsigned lines)
  * The tile data should have already been fetched.
  */
 boolean
-tri_draw(const float *v0, const float *v1, const float *v2,
+tri_draw(const qword vs,
          uint tx, uint ty)
 {
    setup.tx = tx;
    setup.ty = ty;
 
    /* set clipping bounds to tile bounds */
-   setup.cliprect_minx = tx * TILE_SIZE;
-   setup.cliprect_miny = ty * TILE_SIZE;
-   setup.cliprect_maxx = (tx + 1) * TILE_SIZE;
-   setup.cliprect_maxy = (ty + 1) * TILE_SIZE;
-
-   if (!setup_sort_vertices((struct vertex_header *) v0,
-                            (struct vertex_header *) v1,
-                            (struct vertex_header *) v2)) {
+   const qword clipbase = (qword)((vec_uint4){tx, ty});
+   const qword clipmin = si_mpyui(clipbase, TILE_SIZE);
+   const qword clipmax = si_ai(clipmin, TILE_SIZE);
+   setup.cliprect = si_shufb(clipmin, clipmax, SHUFB4(A,B,a,b));
+
+   if(!setup_sort_vertices(vs)) {
       return FALSE; /* totally clipped */
    }
 
diff --git a/src/gallium/drivers/cell/spu/spu_tri.h b/src/gallium/drivers/cell/spu/spu_tri.h
index aa694dd7c9..82e3b19ad7 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.h
+++ b/src/gallium/drivers/cell/spu/spu_tri.h
@@ -31,7 +31,7 @@
 
 
 extern boolean
-tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty);
+tri_draw(const qword vs, uint tx, uint ty);
 
 
 #endif /* SPU_TRI_H */
diff --git a/src/gallium/drivers/i965simple/brw_draw.c b/src/gallium/drivers/i965simple/brw_draw.c
index 648aaa0da5..49d80cb41c 100644
--- a/src/gallium/drivers/i965simple/brw_draw.c
+++ b/src/gallium/drivers/i965simple/brw_draw.c
@@ -35,6 +35,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/internal/p_winsys_screen.h"
+#include "util/u_prim.h"
 
 static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = {
    _3DPRIM_POINTLIST,
@@ -50,20 +51,6 @@ static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = {
 };
 
 
-static const int reduced_prim[PIPE_PRIM_POLYGON+1] = {
-   PIPE_PRIM_POINTS,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES
-};
-
-
 /* When the primitive changes, set a state bit and re-validate.  Not
  * the nicest and would rather deal with this by having all the
  * programs be immune to the active primitive (ie. cope with all
@@ -85,8 +72,8 @@ static void brw_set_prim(struct brw_context *brw, int prim)
       brw->primitive = prim;
       brw->state.dirty.brw |= BRW_NEW_PRIMITIVE;
 
-      if (reduced_prim[prim] != brw->reduced_primitive) {
-	 brw->reduced_primitive = reduced_prim[prim];
+      if (u_reduced_prim(prim) != brw->reduced_primitive) {
+	 brw->reduced_primitive = u_reduced_prim(prim);
 	 brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE;
       }
 
diff --git a/src/gallium/drivers/identity/Makefile b/src/gallium/drivers/identity/Makefile
new file mode 100644
index 0000000000..e32b9102e5
--- /dev/null
+++ b/src/gallium/drivers/identity/Makefile
@@ -0,0 +1,12 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = identity
+
+C_SOURCES = \
+	id_objects.c \
+	id_context.c \
+	id_screen.c \
+	id_drm.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/identity/SConscript b/src/gallium/drivers/identity/SConscript
new file mode 100644
index 0000000000..7f079dd0a8
--- /dev/null
+++ b/src/gallium/drivers/identity/SConscript
@@ -0,0 +1,13 @@
+Import('*')
+
+env = env.Clone()
+
+identity = env.ConvenienceLibrary(
+	target = 'identity',
+	source = [
+		'id_screen.c',
+		'id_context.c',
+		'id_objects.c',
+	])
+
+Export('identity')
diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c
new file mode 100644
index 0000000000..a500ec6045
--- /dev/null
+++ b/src/gallium/drivers/identity/id_context.c
@@ -0,0 +1,719 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_context.h"
+#include "util/u_memory.h"
+
+#include "id_public.h"
+#include "id_context.h"
+#include "id_objects.h"
+
+
+static void
+identity_destroy(struct pipe_context *_pipe)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->destroy(pipe);
+
+   free(id_pipe);
+}
+
+static void
+identity_set_edgeflags(struct pipe_context *_pipe,
+                       const unsigned *bitfield)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_edgeflags(pipe,
+                       bitfield);
+}
+
+static boolean
+identity_draw_arrays(struct pipe_context *_pipe,
+                     unsigned prim,
+                     unsigned start,
+                     unsigned count)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->draw_arrays(pipe,
+                            prim,
+                            start,
+                            count);
+}
+
+static boolean
+identity_draw_elements(struct pipe_context *_pipe,
+                       struct pipe_buffer *_indexBuffer,
+                       unsigned indexSize,
+                       unsigned prim,
+                       unsigned start,
+                       unsigned count)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_buffer *id_buffer = identity_buffer(_indexBuffer);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_buffer *indexBuffer = id_buffer->buffer;
+
+   return pipe->draw_elements(pipe,
+                              indexBuffer,
+                              indexSize,
+                              prim,
+                              start,
+                              count);
+}
+
+static boolean
+identity_draw_range_elements(struct pipe_context *_pipe,
+                             struct pipe_buffer *_indexBuffer,
+                             unsigned indexSize,
+                             unsigned minIndex,
+                             unsigned maxIndex,
+                             unsigned mode,
+                             unsigned start,
+                             unsigned count)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_buffer *id_buffer = identity_buffer(_indexBuffer);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_buffer *indexBuffer = id_buffer->buffer;
+
+   return pipe->draw_range_elements(pipe,
+                                    indexBuffer,
+                                    indexSize,
+                                    minIndex,
+                                    maxIndex,
+                                    mode,
+                                    start,
+                                    count);
+}
+
+static struct pipe_query *
+identity_create_query(struct pipe_context *_pipe,
+                      unsigned query_type)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_query(pipe,
+                             query_type);
+}
+
+static void
+identity_destroy_query(struct pipe_context *_pipe,
+                       struct pipe_query *query)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->destroy_query(pipe,
+                       query);
+}
+
+static void
+identity_begin_query(struct pipe_context *_pipe,
+                     struct pipe_query *query)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->begin_query(pipe,
+                     query);
+}
+
+static void
+identity_end_query(struct pipe_context *_pipe,
+                   struct pipe_query *query)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->end_query(pipe,
+                   query);
+}
+
+static boolean
+identity_get_query_result(struct pipe_context *_pipe,
+                          struct pipe_query *query,
+                          boolean wait,
+                          uint64_t *result)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->get_query_result(pipe,
+                                 query,
+                                 wait,
+                                 result);
+}
+
+static void *
+identity_create_blend_state(struct pipe_context *_pipe,
+                            const struct pipe_blend_state *blend)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_blend_state(pipe,
+                                   blend);
+}
+
+static void
+identity_bind_blend_state(struct pipe_context *_pipe,
+                          void *blend)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_blend_state(pipe,
+                              blend);
+}
+
+static void
+identity_delete_blend_state(struct pipe_context *_pipe,
+                            void *blend)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_blend_state(pipe,
+                            blend);
+}
+
+static void *
+identity_create_sampler_state(struct pipe_context *_pipe,
+                              const struct pipe_sampler_state *sampler)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_sampler_state(pipe,
+                                     sampler);
+}
+
+static void
+identity_bind_sampler_states(struct pipe_context *_pipe,
+                             unsigned num,
+                             void **samplers)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_sampler_states(pipe,
+                             num,
+                             samplers);
+}
+
+static void
+identity_delete_sampler_state(struct pipe_context *_pipe,
+                              void *sampler)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_sampler_state(pipe,
+                              sampler);
+}
+
+static void *
+identity_create_rasterizer_state(struct pipe_context *_pipe,
+                                 const struct pipe_rasterizer_state *rasterizer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_rasterizer_state(pipe,
+                                        rasterizer);
+}
+
+static void
+identity_bind_rasterizer_state(struct pipe_context *_pipe,
+                               void *rasterizer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_rasterizer_state(pipe,
+                               rasterizer);
+}
+
+static void
+identity_delete_rasterizer_state(struct pipe_context *_pipe,
+                                 void *rasterizer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_rasterizer_state(pipe,
+                                 rasterizer);
+}
+
+static void *
+identity_create_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                          const struct pipe_depth_stencil_alpha_state *depth_stencil_alpha)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_depth_stencil_alpha_state(pipe,
+                                                 depth_stencil_alpha);
+}
+
+static void
+identity_bind_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                        void *depth_stencil_alpha)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_depth_stencil_alpha_state(pipe,
+                                        depth_stencil_alpha);
+}
+
+static void
+identity_delete_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                          void *depth_stencil_alpha)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_depth_stencil_alpha_state(pipe,
+                                          depth_stencil_alpha);
+}
+
+static void *
+identity_create_fs_state(struct pipe_context *_pipe,
+                         const struct pipe_shader_state *fs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_fs_state(pipe,
+                                fs);
+}
+
+static void
+identity_bind_fs_state(struct pipe_context *_pipe,
+                       void *fs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_fs_state(pipe,
+                       fs);
+}
+
+static void
+identity_delete_fs_state(struct pipe_context *_pipe,
+                         void *fs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_fs_state(pipe,
+                         fs);
+}
+
+static void *
+identity_create_vs_state(struct pipe_context *_pipe,
+                         const struct pipe_shader_state *vs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   return pipe->create_vs_state(pipe,
+                                vs);
+}
+
+static void
+identity_bind_vs_state(struct pipe_context *_pipe,
+                       void *vs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->bind_vs_state(pipe,
+                       vs);
+}
+
+static void
+identity_delete_vs_state(struct pipe_context *_pipe,
+                         void *vs)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->delete_vs_state(pipe,
+                         vs);
+}
+
+static void
+identity_set_blend_color(struct pipe_context *_pipe,
+                         const struct pipe_blend_color *blend_color)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_blend_color(pipe,
+                         blend_color);
+}
+
+static void
+identity_set_clip_state(struct pipe_context *_pipe,
+                        const struct pipe_clip_state *clip)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_clip_state(pipe,
+                        clip);
+}
+
+static void
+identity_set_constant_buffer(struct pipe_context *_pipe,
+                             uint shader,
+                             uint index,
+                             const struct pipe_constant_buffer *_buffer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_constant_buffer unwrapped_buffer;
+   struct pipe_constant_buffer *buffer = NULL;
+
+   /* unwrap the input state */
+   if (_buffer) {
+      unwrapped_buffer.buffer = identity_buffer_unwrap(_buffer->buffer);
+      buffer = &unwrapped_buffer;
+   }
+
+   pipe->set_constant_buffer(pipe,
+                             shader,
+                             index,
+                             buffer);
+}
+
+static void
+identity_set_framebuffer_state(struct pipe_context *_pipe,
+                               const struct pipe_framebuffer_state *_state)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_framebuffer_state unwrapped_state;
+   struct pipe_framebuffer_state *state = NULL;
+   unsigned i;
+
+   /* unwrap the input state */
+   if (_state) {
+      memcpy(&unwrapped_state, _state, sizeof(unwrapped_state));
+      for(i = 0; i < _state->nr_cbufs; i++)
+         unwrapped_state.cbufs[i] = identity_surface_unwrap(_state->cbufs[i]);
+      for (; i < PIPE_MAX_COLOR_BUFS; i++)
+         unwrapped_state.cbufs[i] = NULL;
+      unwrapped_state.zsbuf = identity_surface_unwrap(_state->zsbuf);
+      state = &unwrapped_state;
+   }
+
+   pipe->set_framebuffer_state(pipe,
+                               state);
+}
+
+static void
+identity_set_polygon_stipple(struct pipe_context *_pipe,
+                             const struct pipe_poly_stipple *poly_stipple)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_polygon_stipple(pipe,
+                             poly_stipple);
+}
+
+static void
+identity_set_scissor_state(struct pipe_context *_pipe,
+                           const struct pipe_scissor_state *scissor)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_scissor_state(pipe,
+                           scissor);
+}
+
+static void
+identity_set_viewport_state(struct pipe_context *_pipe,
+                            const struct pipe_viewport_state *viewport)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_viewport_state(pipe,
+                            viewport);
+}
+
+static void
+identity_set_sampler_textures(struct pipe_context *_pipe,
+                              unsigned num_textures,
+                              struct pipe_texture **_textures)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
+   struct pipe_texture **textures = NULL;
+   unsigned i;
+
+   if (_textures) {
+      for (i = 0; i < num_textures; i++)
+         unwrapped_textures[i] = identity_texture_unwrap(_textures[i]);
+      for (; i < PIPE_MAX_SAMPLERS; i++)
+         unwrapped_textures[i] = NULL;
+
+      textures = unwrapped_textures;
+   }
+
+   pipe->set_sampler_textures(pipe,
+                              num_textures,
+                              _textures);
+}
+
+static void
+identity_set_vertex_buffers(struct pipe_context *_pipe,
+                            unsigned num_buffers,
+                            const struct pipe_vertex_buffer *_buffers)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_vertex_buffer unwrapped_buffers[PIPE_MAX_SHADER_INPUTS];
+   struct pipe_vertex_buffer *buffers = NULL;
+   unsigned i;
+
+   if (num_buffers) {
+      memcpy(unwrapped_buffers, _buffers, num_buffers * sizeof(*_buffers));
+      for (i = 0; i < num_buffers; i++)
+         unwrapped_buffers[i].buffer = identity_buffer_unwrap(_buffers[i].buffer);
+      buffers = unwrapped_buffers;
+   }
+
+   pipe->set_vertex_buffers(pipe,
+                            num_buffers,
+                            buffers);
+}
+
+static void
+identity_set_vertex_elements(struct pipe_context *_pipe,
+                             unsigned num_elements,
+                             const struct pipe_vertex_element *vertex_elements)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->set_vertex_elements(pipe,
+                             num_elements,
+                             vertex_elements);
+}
+
+static void
+identity_surface_copy(struct pipe_context *_pipe,
+                      struct pipe_surface *_dst,
+                      unsigned dstx,
+                      unsigned dsty,
+                      struct pipe_surface *_src,
+                      unsigned srcx,
+                      unsigned srcy,
+                      unsigned width,
+                      unsigned height)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_surface *id_surface_dst = identity_surface(_dst);
+   struct identity_surface *id_surface_src = identity_surface(_src);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_surface *dst = id_surface_dst->surface;
+   struct pipe_surface *src = id_surface_src->surface;
+
+   pipe->surface_copy(pipe,
+                      dst,
+                      dstx,
+                      dsty,
+                      src,
+                      srcx,
+                      srcy,
+                      width,
+                      height);
+}
+
+static void
+identity_surface_fill(struct pipe_context *_pipe,
+                      struct pipe_surface *_dst,
+                      unsigned dstx,
+                      unsigned dsty,
+                      unsigned width,
+                      unsigned height,
+                      unsigned value)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_surface *id_surface_dst = identity_surface(_dst);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_surface *dst = id_surface_dst->surface;
+
+   pipe->surface_fill(pipe,
+                      dst,
+                      dstx,
+                      dsty,
+                      width,
+                      height,
+                      value);
+}
+
+static void
+identity_clear(struct pipe_context *_pipe,
+               unsigned buffers,
+               const float *rgba,
+               double depth,
+               unsigned stencil)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->clear(pipe,
+               buffers,
+               rgba,
+               depth,
+               stencil);
+}
+
+static void
+identity_flush(struct pipe_context *_pipe,
+               unsigned flags,
+               struct pipe_fence_handle **fence)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+
+   pipe->flush(pipe,
+               flags,
+               fence);
+}
+
+static unsigned int
+identity_is_texture_referenced(struct pipe_context *_pipe,
+                               struct pipe_texture *_texture,
+                               unsigned face,
+                               unsigned level)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_texture *texture = id_texture->texture;
+
+   return pipe->is_texture_referenced(pipe,
+                                      texture,
+                                      face,
+                                      level);
+}
+
+static unsigned int
+identity_is_buffer_referenced(struct pipe_context *_pipe,
+                              struct pipe_buffer *_buffer)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   return pipe->is_buffer_referenced(pipe,
+                                     buffer);
+}
+
+struct pipe_context *
+identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
+{
+   struct identity_context *id_pipe;
+   (void)identity_screen(_screen);
+
+   id_pipe = CALLOC_STRUCT(identity_context);
+   if (!id_pipe) {
+      return NULL;
+   }
+
+   id_pipe->base.winsys = NULL;
+   id_pipe->base.screen = _screen;
+   id_pipe->base.priv = pipe->priv;
+   id_pipe->base.draw = NULL;
+
+   id_pipe->base.destroy = identity_destroy;
+   id_pipe->base.set_edgeflags = identity_set_edgeflags;
+   id_pipe->base.draw_arrays = identity_draw_arrays;
+   id_pipe->base.draw_elements = identity_draw_elements;
+   id_pipe->base.draw_range_elements = identity_draw_range_elements;
+   id_pipe->base.create_query = identity_create_query;
+   id_pipe->base.destroy_query = identity_destroy_query;
+   id_pipe->base.begin_query = identity_begin_query;
+   id_pipe->base.end_query = identity_end_query;
+   id_pipe->base.get_query_result = identity_get_query_result;
+   id_pipe->base.create_blend_state = identity_create_blend_state;
+   id_pipe->base.bind_blend_state = identity_bind_blend_state;
+   id_pipe->base.delete_blend_state = identity_delete_blend_state;
+   id_pipe->base.create_sampler_state = identity_create_sampler_state;
+   id_pipe->base.bind_sampler_states = identity_bind_sampler_states;
+   id_pipe->base.delete_sampler_state = identity_delete_sampler_state;
+   id_pipe->base.create_rasterizer_state = identity_create_rasterizer_state;
+   id_pipe->base.bind_rasterizer_state = identity_bind_rasterizer_state;
+   id_pipe->base.delete_rasterizer_state = identity_delete_rasterizer_state;
+   id_pipe->base.create_depth_stencil_alpha_state = identity_create_depth_stencil_alpha_state;
+   id_pipe->base.bind_depth_stencil_alpha_state = identity_bind_depth_stencil_alpha_state;
+   id_pipe->base.delete_depth_stencil_alpha_state = identity_delete_depth_stencil_alpha_state;
+   id_pipe->base.create_fs_state = identity_create_fs_state;
+   id_pipe->base.bind_fs_state = identity_bind_fs_state;
+   id_pipe->base.delete_fs_state = identity_delete_fs_state;
+   id_pipe->base.create_vs_state = identity_create_vs_state;
+   id_pipe->base.bind_vs_state = identity_bind_vs_state;
+   id_pipe->base.delete_vs_state = identity_delete_vs_state;
+   id_pipe->base.set_blend_color = identity_set_blend_color;
+   id_pipe->base.set_clip_state = identity_set_clip_state;
+   id_pipe->base.set_constant_buffer = identity_set_constant_buffer;
+   id_pipe->base.set_framebuffer_state = identity_set_framebuffer_state;
+   id_pipe->base.set_polygon_stipple = identity_set_polygon_stipple;
+   id_pipe->base.set_scissor_state = identity_set_scissor_state;
+   id_pipe->base.set_viewport_state = identity_set_viewport_state;
+   id_pipe->base.set_sampler_textures = identity_set_sampler_textures;
+   id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers;
+   id_pipe->base.set_vertex_elements = identity_set_vertex_elements;
+   id_pipe->base.surface_copy = identity_surface_copy;
+   id_pipe->base.surface_fill = identity_surface_fill;
+   id_pipe->base.clear = identity_clear;
+   id_pipe->base.flush = identity_flush;
+   id_pipe->base.is_texture_referenced = identity_is_texture_referenced;
+   id_pipe->base.is_buffer_referenced = identity_is_buffer_referenced;
+
+   id_pipe->pipe = pipe;
+
+   return &id_pipe->base;
+}
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c b/src/gallium/drivers/identity/id_context.h
index 92f5bd09c9..75b73fc7df 100644
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_lock.c
+++ b/src/gallium/drivers/identity/id_context.h
@@ -1,8 +1,8 @@
 /**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ *
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -10,64 +10,39 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * The above copyright notice and this permission notice (including the
  * next paragraph) shall be included in all copies or substantial portions
  * of the Software.
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
+ *
  **************************************************************************/
 
-#include <pipe/p_thread.h>
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_drmif.h"
+#ifndef ID_CONTEXT_H
+#define ID_CONTEXT_H
 
-pipe_static_mutex(lockMutex);
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
 
-/* Lock the hardware and validate our state.
- */
-void
-LOCK_HARDWARE(struct nouveau_context *nv)
-{
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct nouveau_device *dev = nv_screen->device;
-	struct nouveau_device_priv *nvdev = nouveau_device(dev);
-	char __ret=0;
 
-	assert(!nv->locked);
-	pipe_mutex_lock(lockMutex);
+struct identity_context {
+   struct pipe_context base;  /**< base class */
 
-	DRM_CAS(nvdev->lock, nvdev->ctx,
-		(DRM_LOCK_HELD | nvdev->ctx), __ret);
+   struct pipe_context *pipe;
+};
 
-	if (__ret) {
-		drmGetLock(nvdev->fd, nvdev->ctx, 0);
-		nouveau_contended_lock(nv);
-	}
-	nv->locked = 1;
-}
 
-/* Unlock the hardware using the global current context 
- */
-void
-UNLOCK_HARDWARE(struct nouveau_context *nv)
+static INLINE struct identity_context *
+identity_context(struct pipe_context *pipe)
 {
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct nouveau_device *dev = nv_screen->device;
-	struct nouveau_device_priv *nvdev = nouveau_device(dev);
-
-	assert(nv->locked);
-	nv->locked = 0;
-
-	DRM_UNLOCK(nvdev->fd, nvdev->lock, nvdev->ctx);
+   return (struct identity_context *)pipe;
+}
 
-	pipe_mutex_unlock(lockMutex);
-} 
+#endif /* ID_CONTEXT_H */
diff --git a/src/gallium/drivers/identity/id_drm.c b/src/gallium/drivers/identity/id_drm.c
new file mode 100644
index 0000000000..555220f853
--- /dev/null
+++ b/src/gallium/drivers/identity/id_drm.c
@@ -0,0 +1,187 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "state_tracker/drm_api.h"
+
+#include "util/u_memory.h"
+#include "identity/id_drm.h"
+#include "identity/id_public.h"
+#include "identity/id_screen.h"
+#include "identity/id_objects.h"
+
+struct identity_drm_api
+{
+   struct drm_api base;
+
+   struct drm_api *api;
+};
+
+static INLINE struct identity_drm_api *
+identity_drm_api(struct drm_api *_api)
+{
+   return (struct identity_drm_api *)_api;
+}
+
+static struct pipe_screen *
+identity_drm_create_screen(struct drm_api *_api, int fd,
+                           struct drm_create_screen_arg *arg)
+{
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct drm_api *api = id_api->api;
+   struct pipe_screen *screen;
+
+   if (arg && arg->mode != DRM_CREATE_NORMAL)
+      return NULL;
+
+   screen = api->create_screen(api, fd, arg);
+
+   return identity_screen_create(screen);
+};
+
+static struct pipe_context *
+identity_drm_create_context(struct drm_api *_api,
+                            struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct drm_api *api = id_api->api;
+   struct pipe_context *pipe;
+
+   pipe = api->create_context(api, screen);
+
+   pipe = identity_context_create(_screen, pipe);
+
+   return pipe;
+};
+
+static boolean
+identity_drm_buffer_from_texture(struct drm_api *_api,
+                                 struct pipe_texture *_texture,
+                                 struct pipe_buffer **_buffer,
+                                 unsigned *stride)
+{
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_texture *texture = id_texture->texture;
+   struct drm_api *api = id_api->api;
+   struct pipe_buffer *buffer = NULL;
+   boolean result;
+
+   result = api->buffer_from_texture(api, texture, &buffer, stride);
+
+   if (result && _buffer)
+      buffer = identity_buffer_create(identity_screen(texture->screen), buffer);
+
+   if (_buffer)
+      *_buffer = buffer;
+   else
+      pipe_buffer_reference(&buffer, NULL);
+
+   return result;
+}
+
+static struct pipe_buffer *
+identity_drm_buffer_from_handle(struct drm_api *_api,
+                                struct pipe_screen *_screen,
+                                const char *name,
+                                unsigned handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct drm_api *api = id_api->api;
+   struct pipe_buffer *result;
+
+   result = api->buffer_from_handle(api, screen, name, handle);
+
+   result = identity_buffer_create(identity_screen(_screen), result);
+
+   return result;
+}
+
+static boolean
+identity_drm_handle_from_buffer(struct drm_api *_api,
+                                struct pipe_screen *_screen,
+                                struct pipe_buffer *_buffer,
+                                unsigned *handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+   struct drm_api *api = id_api->api;
+
+   return api->handle_from_buffer(api, screen, buffer, handle);
+}
+
+static boolean
+identity_drm_global_handle_from_buffer(struct drm_api *_api,
+                                       struct pipe_screen *_screen,
+                                       struct pipe_buffer *_buffer,
+                                       unsigned *handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+   struct drm_api *api = id_api->api;
+
+   return api->global_handle_from_buffer(api, screen, buffer, handle);
+}
+
+static void
+identity_drm_destroy(struct drm_api *_api)
+{
+   struct identity_drm_api *id_api = identity_drm_api(_api);
+   struct drm_api *api = id_api->api;
+   api->destroy(api);
+
+   free(id_api);
+}
+
+struct drm_api *
+identity_drm_create(struct drm_api *api)
+{
+   struct identity_drm_api *id_api = CALLOC_STRUCT(identity_drm_api);
+
+   if (!id_api)
+      return NULL;
+
+   id_api->base.create_screen = identity_drm_create_screen;
+   id_api->base.create_context = identity_drm_create_context;
+   id_api->base.buffer_from_texture = identity_drm_buffer_from_texture;
+   id_api->base.buffer_from_handle = identity_drm_buffer_from_handle;
+   id_api->base.handle_from_buffer = identity_drm_handle_from_buffer;
+   id_api->base.global_handle_from_buffer = identity_drm_global_handle_from_buffer;
+   id_api->base.destroy = identity_drm_destroy;
+   id_api->api = api;
+
+   return &id_api->base;
+}
diff --git a/src/gallium/drivers/identity/id_drm.h b/src/gallium/drivers/identity/id_drm.h
new file mode 100644
index 0000000000..cf2ad2ce07
--- /dev/null
+++ b/src/gallium/drivers/identity/id_drm.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_DRM_H
+#define ID_DRM_H
+
+struct drm_api;
+
+struct drm_api* identity_drm_create(struct drm_api *api);
+
+#endif /* ID_DRM_H */
diff --git a/src/gallium/drivers/identity/id_objects.c b/src/gallium/drivers/identity/id_objects.c
new file mode 100644
index 0000000000..e893e59940
--- /dev/null
+++ b/src/gallium/drivers/identity/id_objects.c
@@ -0,0 +1,182 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+
+#include "id_public.h"
+#include "id_screen.h"
+#include "id_objects.h"
+
+struct pipe_buffer *
+identity_buffer_create(struct identity_screen *id_screen,
+                       struct pipe_buffer *buffer)
+{
+   struct identity_buffer *id_buffer;
+
+   if(!buffer)
+      goto error;
+
+   assert(buffer->screen == id_screen->screen);
+
+   id_buffer = CALLOC_STRUCT(identity_buffer);
+   if(!id_buffer)
+      goto error;
+
+   memcpy(&id_buffer->base, buffer, sizeof(struct pipe_buffer));
+
+   pipe_reference_init(&id_buffer->base.reference, 1);
+   id_buffer->base.screen = &id_screen->base;
+   id_buffer->buffer = buffer;
+
+   return &id_buffer->base;
+
+error:
+   pipe_buffer_reference(&buffer, NULL);
+   return NULL;
+}
+
+void
+identity_buffer_destroy(struct identity_buffer *id_buffer)
+{
+   pipe_buffer_reference(&id_buffer->buffer, NULL);
+   FREE(id_buffer);
+}
+
+
+struct pipe_texture *
+identity_texture_create(struct identity_screen *id_screen,
+                        struct pipe_texture *texture)
+{
+   struct identity_texture *id_texture;
+
+   if(!texture)
+      goto error;
+
+   assert(texture->screen == id_screen->screen);
+
+   id_texture = CALLOC_STRUCT(identity_texture);
+   if(!id_texture)
+      goto error;
+
+   memcpy(&id_texture->base, texture, sizeof(struct pipe_texture));
+
+   pipe_reference_init(&id_texture->base.reference, 1);
+   id_texture->base.screen = &id_screen->base;
+   id_texture->texture = texture;
+
+   return &id_texture->base;
+
+error:
+   pipe_texture_reference(&texture, NULL);
+   return NULL;
+}
+
+void
+identity_texture_destroy(struct identity_texture *id_texture)
+{
+   pipe_texture_reference(&id_texture->texture, NULL);
+   FREE(id_texture);
+}
+
+
+struct pipe_surface *
+identity_surface_create(struct identity_texture *id_texture,
+                        struct pipe_surface *surface)
+{
+   struct identity_surface *id_surface;
+
+   if(!surface)
+      goto error;
+
+   assert(surface->texture == id_texture->texture);
+
+   id_surface = CALLOC_STRUCT(identity_surface);
+   if(!id_surface)
+      goto error;
+
+   memcpy(&id_surface->base, surface, sizeof(struct pipe_surface));
+
+   pipe_reference_init(&id_surface->base.reference, 1);
+   id_surface->base.texture = NULL;
+   pipe_texture_reference(&id_surface->base.texture, &id_texture->base);
+   id_surface->surface = surface;
+
+   return &id_surface->base;
+
+error:
+   pipe_surface_reference(&surface, NULL);
+   return NULL;
+}
+
+void
+identity_surface_destroy(struct identity_surface *id_surface)
+{
+   pipe_texture_reference(&id_surface->base.texture, NULL);
+   pipe_surface_reference(&id_surface->surface, NULL);
+   FREE(id_surface);
+}
+
+
+struct pipe_transfer *
+identity_transfer_create(struct identity_texture *id_texture,
+                         struct pipe_transfer *transfer)
+{
+   struct identity_transfer *id_transfer;
+
+   if(!transfer)
+      goto error;
+
+   assert(transfer->texture == id_texture->texture);
+
+   id_transfer = CALLOC_STRUCT(identity_transfer);
+   if(!id_transfer)
+      goto error;
+
+   memcpy(&id_transfer->base, transfer, sizeof(struct pipe_transfer));
+
+   id_transfer->base.texture = NULL;
+   pipe_texture_reference(&id_transfer->base.texture, &id_texture->base);
+   id_transfer->transfer = transfer;
+   assert(id_transfer->base.texture == &id_texture->base);
+
+   return &id_transfer->base;
+
+error:
+   transfer->texture->screen->tex_transfer_destroy(transfer);
+   return NULL;
+}
+
+void
+identity_transfer_destroy(struct identity_transfer *id_transfer)
+{
+   struct identity_screen *id_screen = identity_screen(id_transfer->base.texture->screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   pipe_texture_reference(&id_transfer->base.texture, NULL);
+   screen->tex_transfer_destroy(id_transfer->transfer);
+   FREE(id_transfer);
+}
diff --git a/src/gallium/drivers/identity/id_objects.h b/src/gallium/drivers/identity/id_objects.h
new file mode 100644
index 0000000000..ce58faa3c7
--- /dev/null
+++ b/src/gallium/drivers/identity/id_objects.h
@@ -0,0 +1,169 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_OBJECTS_H
+#define ID_OBJECTS_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+
+#include "id_screen.h"
+
+
+struct identity_buffer
+{
+   struct pipe_buffer base;
+
+   struct pipe_buffer *buffer;
+};
+
+
+struct identity_texture
+{
+   struct pipe_texture base;
+
+   struct pipe_texture *texture;
+};
+
+
+struct identity_surface
+{
+   struct pipe_surface base;
+
+   struct pipe_surface *surface;
+};
+
+
+struct identity_transfer
+{
+   struct pipe_transfer base;
+
+   struct pipe_transfer *transfer;
+};
+
+
+static INLINE struct identity_buffer *
+identity_buffer(struct pipe_buffer *_buffer)
+{
+   if(!_buffer)
+      return NULL;
+   (void)identity_screen(_buffer->screen);
+   return (struct identity_buffer *)_buffer;
+}
+
+static INLINE struct identity_texture *
+identity_texture(struct pipe_texture *_texture)
+{
+   if(!_texture)
+      return NULL;
+   (void)identity_screen(_texture->screen);
+   return (struct identity_texture *)_texture;
+}
+
+static INLINE struct identity_surface *
+identity_surface(struct pipe_surface *_surface)
+{
+   if(!_surface)
+      return NULL;
+   (void)identity_texture(_surface->texture);
+   return (struct identity_surface *)_surface;
+}
+
+static INLINE struct identity_transfer *
+identity_transfer(struct pipe_transfer *_transfer)
+{
+   if(!_transfer)
+      return NULL;
+   (void)identity_texture(_transfer->texture);
+   return (struct identity_transfer *)_transfer;
+}
+
+
+static INLINE struct pipe_buffer *
+identity_buffer_unwrap(struct pipe_buffer *_buffer)
+{
+   if(!_buffer)
+      return NULL;
+   return identity_buffer(_buffer)->buffer;
+}
+
+static INLINE struct pipe_texture *
+identity_texture_unwrap(struct pipe_texture *_texture)
+{
+   if(!_texture)
+      return NULL;
+   return identity_texture(_texture)->texture;
+}
+
+static INLINE struct pipe_surface *
+identity_surface_unwrap(struct pipe_surface *_surface)
+{
+   if(!_surface)
+      return NULL;
+   return identity_surface(_surface)->surface;
+}
+
+static INLINE struct pipe_transfer *
+identity_transfer_unwrap(struct pipe_transfer *_transfer)
+{
+   if(!_transfer)
+      return NULL;
+   return identity_transfer(_transfer)->transfer;
+}
+
+
+struct pipe_buffer *
+identity_buffer_create(struct identity_screen *id_screen,
+                       struct pipe_buffer *buffer);
+
+void
+identity_buffer_destroy(struct identity_buffer *id_buffer);
+
+struct pipe_texture *
+identity_texture_create(struct identity_screen *id_screen,
+                        struct pipe_texture *texture);
+
+void
+identity_texture_destroy(struct identity_texture *id_texture);
+
+struct pipe_surface *
+identity_surface_create(struct identity_texture *id_texture,
+                        struct pipe_surface *surface);
+
+void
+identity_surface_destroy(struct identity_surface *id_surface);
+
+struct pipe_transfer *
+identity_transfer_create(struct identity_texture *id_texture,
+                         struct pipe_transfer *transfer);
+
+void
+identity_transfer_destroy(struct identity_transfer *id_transfer);
+
+
+#endif /* ID_OBJECTS_H */
diff --git a/src/gallium/drivers/identity/id_public.h b/src/gallium/drivers/identity/id_public.h
new file mode 100644
index 0000000000..cac14cfd60
--- /dev/null
+++ b/src/gallium/drivers/identity/id_public.h
@@ -0,0 +1,40 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_PUBLIC_H
+#define ID_PUBLIC_H
+
+struct pipe_screen;
+struct pipe_context;
+
+struct pipe_screen *
+identity_screen_create(struct pipe_screen *screen);
+
+struct pipe_context *
+identity_context_create(struct pipe_screen *screen, struct pipe_context *pipe);
+
+#endif /* PT_PUBLIC_H */
diff --git a/src/gallium/drivers/identity/id_screen.c b/src/gallium/drivers/identity/id_screen.c
new file mode 100644
index 0000000000..259f1be36e
--- /dev/null
+++ b/src/gallium/drivers/identity/id_screen.c
@@ -0,0 +1,481 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_screen.h"
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+
+#include "id_public.h"
+#include "id_screen.h"
+#include "id_objects.h"
+
+
+static void
+identity_screen_destroy(struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   screen->destroy(screen);
+
+   FREE(id_screen);
+}
+
+static const char *
+identity_screen_get_name(struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_name(screen);
+}
+
+static const char *
+identity_screen_get_vendor(struct pipe_screen *_screen)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_vendor(screen);
+}
+
+static int
+identity_screen_get_param(struct pipe_screen *_screen,
+                          int param)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_param(screen,
+                            param);
+}
+
+static float
+identity_screen_get_paramf(struct pipe_screen *_screen,
+                           int param)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->get_paramf(screen,
+                             param);
+}
+
+static boolean
+identity_screen_is_format_supported(struct pipe_screen *_screen,
+                                    enum pipe_format format,
+                                    enum pipe_texture_target target,
+                                    unsigned tex_usage,
+                                    unsigned geom_flags)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->is_format_supported(screen,
+                                      format,
+                                      target,
+                                      tex_usage,
+                                      geom_flags);
+}
+
+static struct pipe_texture *
+identity_screen_texture_create(struct pipe_screen *_screen,
+                               const struct pipe_texture *templat)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *result;
+
+   result = screen->texture_create(screen,
+                                   templat);
+
+   if (result)
+      return identity_texture_create(id_screen, result);
+   return NULL;
+}
+
+static struct pipe_texture *
+identity_screen_texture_blanket(struct pipe_screen *_screen,
+                                const struct pipe_texture *templat,
+                                const unsigned *stride,
+                                struct pipe_buffer *_buffer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+   struct pipe_texture *result;
+
+   result = screen->texture_blanket(screen,
+                                    templat,
+                                    stride,
+                                    buffer);
+
+   if (result)
+      return identity_texture_create(id_screen, result);
+   return NULL;
+}
+
+static void
+identity_screen_texture_destroy(struct pipe_texture *_texture)
+{
+   identity_texture_destroy(identity_texture(_texture));
+}
+
+static struct pipe_surface *
+identity_screen_get_tex_surface(struct pipe_screen *_screen,
+                                struct pipe_texture *_texture,
+                                unsigned face,
+                                unsigned level,
+                                unsigned zslice,
+                                unsigned usage)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+   struct pipe_surface *result;
+
+   result = screen->get_tex_surface(screen,
+                                    texture,
+                                    face,
+                                    level,
+                                    zslice,
+                                    usage);
+
+   if (result)
+      return identity_surface_create(id_texture, result);
+   return NULL;
+}
+
+static void
+identity_screen_tex_surface_destroy(struct pipe_surface *_surface)
+{
+   identity_surface_destroy(identity_surface(_surface));
+}
+
+static struct pipe_transfer *
+identity_screen_get_tex_transfer(struct pipe_screen *_screen,
+                                 struct pipe_texture *_texture,
+                                 unsigned face,
+                                 unsigned level,
+                                 unsigned zslice,
+                                 enum pipe_transfer_usage usage,
+                                 unsigned x,
+                                 unsigned y,
+                                 unsigned w,
+                                 unsigned h)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+   struct pipe_transfer *result;
+
+   result = screen->get_tex_transfer(screen,
+                                     texture,
+                                     face,
+                                     level,
+                                     zslice,
+                                     usage,
+                                     x,
+                                     y,
+                                     w,
+                                     h);
+
+   if (result)
+      return identity_transfer_create(id_texture, result);
+   return NULL;
+}
+
+static void
+identity_screen_tex_transfer_destroy(struct pipe_transfer *_transfer)
+{
+   identity_transfer_destroy(identity_transfer(_transfer));
+}
+
+static void *
+identity_screen_transfer_map(struct pipe_screen *_screen,
+                             struct pipe_transfer *_transfer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_transfer *id_transfer = identity_transfer(_transfer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_transfer *transfer = id_transfer->transfer;
+
+   return screen->transfer_map(screen,
+                               transfer);
+}
+
+static void
+identity_screen_transfer_unmap(struct pipe_screen *_screen,
+                               struct pipe_transfer *_transfer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_transfer *id_transfer = identity_transfer(_transfer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_transfer *transfer = id_transfer->transfer;
+
+   screen->transfer_unmap(screen,
+                          transfer);
+}
+
+static struct pipe_buffer *
+identity_screen_buffer_create(struct pipe_screen *_screen,
+                              unsigned alignment,
+                              unsigned usage,
+                              unsigned size)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *result;
+
+   result = screen->buffer_create(screen,
+                                  alignment,
+                                  usage,
+                                  size);
+
+   if (result)
+      return identity_buffer_create(id_screen, result);
+   return NULL;
+}
+
+static struct pipe_buffer *
+identity_screen_user_buffer_create(struct pipe_screen *_screen,
+                                   void *ptr,
+                                   unsigned bytes)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *result;
+
+   result = screen->user_buffer_create(screen,
+                                       ptr,
+                                       bytes);
+
+   if (result)
+      return identity_buffer_create(id_screen, result);
+   return NULL;
+}
+
+static struct pipe_buffer *
+identity_screen_surface_buffer_create(struct pipe_screen *_screen,
+                                      unsigned width,
+                                      unsigned height,
+                                      enum pipe_format format,
+                                      unsigned usage,
+                                      unsigned *stride)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *result;
+
+   result = screen->surface_buffer_create(screen,
+                                          width,
+                                          height,
+                                          format,
+                                          usage,
+                                          stride);
+
+   if (result)
+      return identity_buffer_create(id_screen, result);
+   return NULL;
+}
+
+static void *
+identity_screen_buffer_map(struct pipe_screen *_screen,
+                           struct pipe_buffer *_buffer,
+                           unsigned usage)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   return screen->buffer_map(screen,
+                             buffer,
+                             usage);
+}
+
+static void *
+identity_screen_buffer_map_range(struct pipe_screen *_screen,
+                                 struct pipe_buffer *_buffer,
+                                 unsigned offset,
+                                 unsigned length,
+                                 unsigned usage)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   return screen->buffer_map_range(screen,
+                                   buffer,
+                                   offset,
+                                   length,
+                                   usage);
+}
+
+static void
+identity_screen_buffer_flush_mapped_range(struct pipe_screen *_screen,
+                                          struct pipe_buffer *_buffer,
+                                          unsigned offset,
+                                          unsigned length)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   screen->buffer_flush_mapped_range(screen,
+                                     buffer,
+                                     offset,
+                                     length);
+}
+
+static void
+identity_screen_buffer_unmap(struct pipe_screen *_screen,
+                             struct pipe_buffer *_buffer)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_buffer *id_buffer = identity_buffer(_buffer);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_buffer *buffer = id_buffer->buffer;
+
+   screen->buffer_unmap(screen,
+                        buffer);
+}
+
+static void
+identity_screen_buffer_destroy(struct pipe_buffer *_buffer)
+{
+   identity_buffer_destroy(identity_buffer(_buffer));
+}
+
+static void
+identity_screen_flush_frontbuffer(struct pipe_screen *_screen,
+                                  struct pipe_surface *_surface,
+                                  void *context_private)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_surface *id_surface = identity_surface(_surface);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_surface *surface = id_surface->surface;
+
+   screen->flush_frontbuffer(screen,
+                             surface,
+                             context_private);
+}
+
+static void
+identity_screen_fence_reference(struct pipe_screen *_screen,
+                                struct pipe_fence_handle **ptr,
+                                struct pipe_fence_handle *fence)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   screen->fence_reference(screen,
+                           ptr,
+                           fence);
+}
+
+static int
+identity_screen_fence_signalled(struct pipe_screen *_screen,
+                                struct pipe_fence_handle *fence,
+                                unsigned flags)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->fence_signalled(screen,
+                                  fence,
+                                  flags);
+}
+
+static int
+identity_screen_fence_finish(struct pipe_screen *_screen,
+                             struct pipe_fence_handle *fence,
+                             unsigned flags)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct pipe_screen *screen = id_screen->screen;
+
+   return screen->fence_finish(screen,
+                               fence,
+                               flags);
+}
+
+struct pipe_screen *
+identity_screen_create(struct pipe_screen *screen)
+{
+   struct identity_screen *id_screen;
+
+   id_screen = CALLOC_STRUCT(identity_screen);
+   if (!id_screen) {
+      return NULL;
+   }
+
+   id_screen->base.winsys = NULL;
+
+   id_screen->base.destroy = identity_screen_destroy;
+   id_screen->base.get_name = identity_screen_get_name;
+   id_screen->base.get_vendor = identity_screen_get_vendor;
+   id_screen->base.get_param = identity_screen_get_param;
+   id_screen->base.get_paramf = identity_screen_get_paramf;
+   id_screen->base.is_format_supported = identity_screen_is_format_supported;
+   id_screen->base.texture_create = identity_screen_texture_create;
+   id_screen->base.texture_blanket = identity_screen_texture_blanket;
+   id_screen->base.texture_destroy = identity_screen_texture_destroy;
+   id_screen->base.get_tex_surface = identity_screen_get_tex_surface;
+   id_screen->base.tex_surface_destroy = identity_screen_tex_surface_destroy;
+   id_screen->base.get_tex_transfer = identity_screen_get_tex_transfer;
+   id_screen->base.tex_transfer_destroy = identity_screen_tex_transfer_destroy;
+   id_screen->base.transfer_map = identity_screen_transfer_map;
+   id_screen->base.transfer_unmap = identity_screen_transfer_unmap;
+   id_screen->base.buffer_create = identity_screen_buffer_create;
+   id_screen->base.user_buffer_create = identity_screen_user_buffer_create;
+   id_screen->base.surface_buffer_create = identity_screen_surface_buffer_create;
+   if (screen->buffer_map)
+      id_screen->base.buffer_map = identity_screen_buffer_map;
+   if (screen->buffer_map_range)
+      id_screen->base.buffer_map_range = identity_screen_buffer_map_range;
+   if (screen->buffer_flush_mapped_range)
+      id_screen->base.buffer_flush_mapped_range = identity_screen_buffer_flush_mapped_range;
+   if (screen->buffer_unmap)
+      id_screen->base.buffer_unmap = identity_screen_buffer_unmap;
+   id_screen->base.buffer_destroy = identity_screen_buffer_destroy;
+   id_screen->base.flush_frontbuffer = identity_screen_flush_frontbuffer;
+   id_screen->base.fence_reference = identity_screen_fence_reference;
+   id_screen->base.fence_signalled = identity_screen_fence_signalled;
+   id_screen->base.fence_finish = identity_screen_fence_finish;
+
+   id_screen->screen = screen;
+
+   return &id_screen->base;
+}
diff --git a/src/gallium/drivers/identity/id_screen.h b/src/gallium/drivers/identity/id_screen.h
new file mode 100644
index 0000000000..2c4f129089
--- /dev/null
+++ b/src/gallium/drivers/identity/id_screen.h
@@ -0,0 +1,48 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ID_SCREEN_H
+#define ID_SCREEN_H
+
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+
+
+struct identity_screen {
+   struct pipe_screen base;
+
+   struct pipe_screen *screen;
+};
+
+
+static INLINE struct identity_screen *
+identity_screen(struct pipe_screen *screen)
+{
+   return (struct identity_screen *)screen;
+}
+
+#endif /* ID_SCREEN_H */
diff --git a/src/gallium/drivers/nouveau/Makefile b/src/gallium/drivers/nouveau/Makefile
new file mode 100644
index 0000000000..dbe8a6e7bf
--- /dev/null
+++ b/src/gallium/drivers/nouveau/Makefile
@@ -0,0 +1,8 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nouveau
+
+C_SOURCES = nouveau_screen.c
+
+include ../../Makefile.template
diff --git a/src/gallium/drivers/nouveau/nouveau_push.h b/src/gallium/drivers/nouveau/nouveau_push.h
index 54ef1c1291..9c235080a5 100644
--- a/src/gallium/drivers/nouveau/nouveau_push.h
+++ b/src/gallium/drivers/nouveau/nouveau_push.h
@@ -9,13 +9,13 @@
 
 #define OUT_RING(data) do {                                                    \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	(*pc->nvws->channel->pushbuf->cur++) = (data);                         \
+	(*pc->base.channel->pushbuf->cur++) = (data);                          \
 } while(0)
 
 #define OUT_RINGp(src,size) do {                                               \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	memcpy(pc->nvws->channel->pushbuf->cur, (src), (size) * 4);            \
-	pc->nvws->channel->pushbuf->cur += (size);                             \
+	memcpy(pc->base.channel->pushbuf->cur, (src), (size) * 4);             \
+	pc->base.channel->pushbuf->cur += (size);                              \
 } while(0)
 
 #define OUT_RINGf(data) do {                                                   \
@@ -26,25 +26,35 @@
 
 #define BEGIN_RING(obj,mthd,size) do {                                         \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	if (pc->nvws->channel->pushbuf->remaining < ((size) + 1))              \
-		pc->nvws->push_flush(pc->nvws, ((size) + 1), NULL);            \
+	struct nouveau_channel *chan = pc->base.channel;                       \
+	if (chan->pushbuf->remaining < ((size) + 1))                           \
+		nouveau_pushbuf_flush(chan, ((size) + 1));                     \
 	OUT_RING((pc->obj->subc << 13) | ((size) << 18) | (mthd));             \
-	pc->nvws->channel->pushbuf->remaining -= ((size) + 1);                 \
+	chan->pushbuf->remaining -= ((size) + 1);                              \
 } while(0)
 
 #define BEGIN_RING_NI(obj,mthd,size) do {                                      \
 	BEGIN_RING(obj, (mthd) | 0x40000000, (size));                          \
 } while(0)
 
+static inline void
+DO_FIRE_RING(struct nouveau_channel *chan, struct pipe_fence_handle **fence)
+{
+	nouveau_pushbuf_flush(chan, 0);
+	if (fence)
+		*fence = NULL;
+}
+
 #define FIRE_RING(fence) do {                                                  \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	pc->nvws->push_flush(pc->nvws, 0, fence);                              \
+	DO_FIRE_RING(pc->base.channel, fence);                                 \
 } while(0)
 
 #define OUT_RELOC(bo,data,flags,vor,tor) do {                                  \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	pc->nvws->push_reloc(pc->nvws, pc->nvws->channel->pushbuf->cur++,      \
-			     (bo), (data), (flags), (vor), (tor));             \
+	struct nouveau_channel *chan = pc->base.channel;                       \
+	nouveau_pushbuf_emit_reloc(chan, chan->pushbuf->cur++, nouveau_bo(bo), \
+				   (data), 0, (flags), (vor), (tor));          \
 } while(0)
 
 /* Raw data + flags depending on FB/TT buffer */
@@ -55,8 +65,8 @@
 /* FB/TT object handle */
 #define OUT_RELOCo(bo,flags) do {                                              \
 	OUT_RELOC((bo), 0, (flags) | NOUVEAU_BO_OR,                            \
-		  pc->nvws->channel->vram->handle,                             \
-		  pc->nvws->channel->gart->handle);                            \
+		  pc->base.channel->vram->handle,                              \
+		  pc->base.channel->gart->handle);                             \
 } while(0)
 
 /* Low 32-bits of offset */
@@ -72,11 +82,12 @@
 /* A reloc which'll recombine into a NV_DMA_METHOD packet header */
 #define OUT_RELOCm(bo, flags, obj, mthd, size) do {                            \
 	NOUVEAU_PUSH_CONTEXT(pc);                                              \
-	if (pc->nvws->channel->pushbuf->remaining < ((size) + 1))              \
-		pc->nvws->push_flush(pc->nvws->channel, ((size) + 1), NULL);   \
+	struct nouveau_channel *chan = pc->base.channel;                       \
+	if (chan->pushbuf->remaining < ((size) + 1))                           \
+		nouveau_pushbuf_flush(chan, ((size) + 1));                     \
 	OUT_RELOCd((bo), (pc->obj->subc << 13) | ((size) << 18) | (mthd),      \
 		   (flags), 0, 0);                                             \
-	pc->nvws->channel->pushbuf->remaining -= ((size) + 1);                 \
+	chan->pushbuf->remaining -= ((size) + 1);                              \
 } while(0)
 
 #endif
diff --git a/src/gallium/drivers/nouveau/nouveau_screen.c b/src/gallium/drivers/nouveau/nouveau_screen.c
new file mode 100644
index 0000000000..832366e646
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_screen.c
@@ -0,0 +1,240 @@
+#include <pipe/p_defines.h>
+#include <pipe/p_screen.h>
+#include <pipe/p_state.h>
+
+#include <util/u_memory.h>
+
+#include "nouveau/nouveau_bo.h"
+#include "nouveau_winsys.h"
+#include "nouveau_screen.h"
+
+static const char *
+nouveau_screen_get_name(struct pipe_screen *pscreen)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nouveau_screen_get_vendor(struct pipe_screen *pscreen)
+{
+	return "nouveau";
+}
+
+static struct pipe_buffer *
+nouveau_screen_bo_skel(struct pipe_screen *pscreen, struct nouveau_bo *bo,
+		       unsigned alignment, unsigned usage, unsigned size)
+{
+	struct pipe_buffer *pb;
+	
+	pb = CALLOC(1, sizeof(struct pipe_buffer)+sizeof(struct nouveau_bo *));
+	if (!pb) {
+		nouveau_bo_ref(NULL, &bo);
+		return NULL;
+	}
+
+	pipe_reference_init(&pb->reference, 1);
+	pb->screen = pscreen;
+	pb->alignment = alignment;
+	pb->usage = usage;
+	pb->size = size;
+	*(struct nouveau_bo **)(pb + 1) = bo;
+	return pb;
+}
+
+static struct pipe_buffer *
+nouveau_screen_bo_new(struct pipe_screen *pscreen, unsigned alignment,
+		      unsigned usage, unsigned size)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct nouveau_bo *bo = NULL;
+	uint32_t flags = NOUVEAU_BO_MAP;
+	int ret;
+
+	if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER)
+		flags |= NOUVEAU_BO_GART;
+	else
+	if (usage & PIPE_BUFFER_USAGE_VERTEX) {
+		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF))
+			flags |= NOUVEAU_BO_GART;
+	} else
+	if (usage & PIPE_BUFFER_USAGE_INDEX) {
+		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF))
+			flags |= NOUVEAU_BO_GART;
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_PIXEL) {
+		if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE)
+			flags |= NOUVEAU_BO_GART;
+		if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE))
+			flags |= NOUVEAU_BO_VRAM;
+
+		if (dev->chipset == 0x50 || dev->chipset >= 0x80) {
+			flags |= NOUVEAU_BO_TILED;
+			if (usage & NOUVEAU_BUFFER_USAGE_ZETA)
+				flags |= NOUVEAU_BO_ZTILE;
+		}
+	}
+
+	ret = nouveau_bo_new(dev, flags, alignment, size, &bo);
+	if (ret)
+		return NULL;
+
+	return nouveau_screen_bo_skel(pscreen, bo, alignment, usage, size);
+}
+
+static struct pipe_buffer *
+nouveau_screen_bo_user(struct pipe_screen *pscreen, void *ptr, unsigned bytes)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct nouveau_bo *bo = NULL;
+	int ret;
+
+	ret = nouveau_bo_user(dev, ptr, bytes, &bo);
+	if (ret)
+		return NULL;
+
+	return nouveau_screen_bo_skel(pscreen, bo, 0, 0, bytes);
+}
+
+static inline uint32_t
+nouveau_screen_map_flags(unsigned pipe)
+{
+	uint32_t flags = 0;
+
+	if (pipe & PIPE_BUFFER_USAGE_CPU_READ)
+		flags |= NOUVEAU_BO_RD;
+	if (pipe & PIPE_BUFFER_USAGE_CPU_WRITE)
+		flags |= NOUVEAU_BO_WR;
+	if (pipe & PIPE_BUFFER_USAGE_DISCARD)
+		flags |= NOUVEAU_BO_INVAL;
+	if (pipe & PIPE_BUFFER_USAGE_DONTBLOCK)
+		flags |= NOUVEAU_BO_NOWAIT;
+	else
+	if (pipe & 0 /*PIPE_BUFFER_USAGE_UNSYNCHRONIZED*/)
+		flags |= NOUVEAU_BO_NOSYNC;
+
+	return flags;
+}
+
+static void *
+nouveau_screen_bo_map(struct pipe_screen *pscreen, struct pipe_buffer *pb,
+		      unsigned usage)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+	int ret;
+
+	ret = nouveau_bo_map(bo, nouveau_screen_map_flags(usage));
+	if (ret) {
+		debug_printf("map failed: %d\n", ret);
+		return NULL;
+	}
+
+	return bo->map;
+}
+
+static void *
+nouveau_screen_bo_map_range(struct pipe_screen *pscreen, struct pipe_buffer *pb,
+			    unsigned offset, unsigned length, unsigned usage)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+	int ret;
+
+	ret = nouveau_bo_map_range(bo, offset, length,
+				   nouveau_screen_map_flags(usage));
+	if (ret) {
+		debug_printf("map_range failed: %d\n", ret);
+		return NULL;
+	}
+
+	return (char *)bo->map - offset; /* why gallium? why? */
+}
+
+static void
+nouveau_screen_bo_map_flush(struct pipe_screen *pscreen, struct pipe_buffer *pb,
+			    unsigned offset, unsigned length)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+
+	nouveau_bo_map_flush(bo, offset, length);
+}
+
+static void
+nouveau_screen_bo_unmap(struct pipe_screen *pscreen, struct pipe_buffer *pb)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+
+	nouveau_bo_unmap(bo);
+}
+
+static void
+nouveau_screen_bo_del(struct pipe_buffer *pb)
+{
+	struct nouveau_bo *bo = nouveau_bo(pb);
+
+	nouveau_bo_ref(NULL, &bo);
+	FREE(pb);
+}
+
+static void
+nouveau_screen_fence_ref(struct pipe_screen *pscreen,
+			 struct pipe_fence_handle **ptr,
+			 struct pipe_fence_handle *pfence)
+{
+	*ptr = pfence;
+}
+
+static int
+nouveau_screen_fence_signalled(struct pipe_screen *screen,
+			       struct pipe_fence_handle *pfence,
+			       unsigned flags)
+{
+	return 0;
+}
+
+static int
+nouveau_screen_fence_finish(struct pipe_screen *screen,
+			    struct pipe_fence_handle *pfence,
+			    unsigned flags)
+{
+	return 0;
+}
+
+int
+nouveau_screen_init(struct nouveau_screen *screen, struct nouveau_device *dev)
+{
+	struct pipe_screen *pscreen = &screen->base;
+	int ret;
+
+	ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202,
+				    &screen->channel);
+	if (ret)
+		return ret;
+	screen->device = dev;
+
+	pscreen->get_name = nouveau_screen_get_name;
+	pscreen->get_vendor = nouveau_screen_get_vendor;
+
+	pscreen->buffer_create = nouveau_screen_bo_new;
+	pscreen->user_buffer_create = nouveau_screen_bo_user;
+	pscreen->buffer_map = nouveau_screen_bo_map;
+	pscreen->buffer_map_range = nouveau_screen_bo_map_range;
+	pscreen->buffer_flush_mapped_range = nouveau_screen_bo_map_flush;
+	pscreen->buffer_unmap = nouveau_screen_bo_unmap;
+	pscreen->buffer_destroy = nouveau_screen_bo_del;
+
+	pscreen->fence_reference = nouveau_screen_fence_ref;
+	pscreen->fence_signalled = nouveau_screen_fence_signalled;
+	pscreen->fence_finish = nouveau_screen_fence_finish;
+
+	return 0;
+}
+
+void
+nouveau_screen_fini(struct nouveau_screen *screen)
+{
+}
+
diff --git a/src/gallium/drivers/nouveau/nouveau_screen.h b/src/gallium/drivers/nouveau/nouveau_screen.h
new file mode 100644
index 0000000000..9968b07896
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_screen.h
@@ -0,0 +1,25 @@
+#ifndef __NOUVEAU_SCREEN_H__
+#define __NOUVEAU_SCREEN_H__
+
+struct nouveau_screen {
+	struct pipe_screen base;
+	struct nouveau_device *device;
+	struct nouveau_channel *channel;
+};
+
+static inline struct nouveau_screen *
+nouveau_screen(struct pipe_screen *pscreen)
+{
+	return (struct nouveau_screen *)pscreen;
+}
+
+static inline struct nouveau_bo *
+nouveau_bo(struct pipe_buffer *pb)
+{
+	return pb ? *(struct nouveau_bo **)(pb + 1) : NULL;
+}
+
+int nouveau_screen_init(struct nouveau_screen *, struct nouveau_device *);
+void nouveau_screen_fini(struct nouveau_screen *);
+
+#endif
diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h
index a54820e851..b595405357 100644
--- a/src/gallium/drivers/nouveau/nouveau_stateobj.h
+++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h
@@ -4,7 +4,7 @@
 #include "util/u_debug.h"
 
 struct nouveau_stateobj_reloc {
-	struct pipe_buffer *bo;
+	struct nouveau_bo *bo;
 
 	unsigned offset;
 	unsigned packet;
@@ -51,7 +51,7 @@ so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso)
         if (pipe_reference((struct pipe_reference**)pso, &ref->reference)) {
 		free(so->push);
 		for (i = 0; i < so->cur_reloc; i++)
-			pipe_buffer_reference(&so->reloc[i].bo, NULL);
+			nouveau_bo_ref(NULL, &so->reloc[i].bo);
 		free(so->reloc);
 		free(so);
 	}
@@ -81,13 +81,13 @@ so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr,
 }
 
 static INLINE void
-so_reloc(struct nouveau_stateobj *so, struct pipe_buffer *bo,
+so_reloc(struct nouveau_stateobj *so, struct nouveau_bo *bo,
 	 unsigned data, unsigned flags, unsigned vor, unsigned tor)
 {
 	struct nouveau_stateobj_reloc *r = &so->reloc[so->cur_reloc++];
 	
 	r->bo = NULL;
-	pipe_buffer_reference(&r->bo, bo);
+	nouveau_bo_ref(bo, &r->bo);
 	r->offset = so->cur - so->push;
 	r->packet = so->cur_packet;
 	r->data = data;
@@ -107,50 +107,52 @@ so_dump(struct nouveau_stateobj *so)
 }
 
 static INLINE void
-so_emit(struct nouveau_winsys *nvws, struct nouveau_stateobj *so)
+so_emit(struct nouveau_channel *chan, struct nouveau_stateobj *so)
 {
-	struct nouveau_pushbuf *pb = nvws->channel->pushbuf;
+	struct nouveau_pushbuf *pb = chan->pushbuf;
 	unsigned nr, i;
 
 	nr = so->cur - so->push;
 	if (pb->remaining < nr)
-		nvws->push_flush(nvws, nr, NULL);
+		nouveau_pushbuf_flush(chan, nr);
 	pb->remaining -= nr;
 
 	memcpy(pb->cur, so->push, nr * 4);
 	for (i = 0; i < so->cur_reloc; i++) {
 		struct nouveau_stateobj_reloc *r = &so->reloc[i];
 
-		nvws->push_reloc(nvws, pb->cur + r->offset, r->bo,
-				 r->data, r->flags, r->vor, r->tor);
+		nouveau_pushbuf_emit_reloc(chan, pb->cur + r->offset,
+					   r->bo, r->data, 0, r->flags,
+					   r->vor, r->tor);
 	}
 	pb->cur += nr;
 }
 
 static INLINE void
-so_emit_reloc_markers(struct nouveau_winsys *nvws, struct nouveau_stateobj *so)
+so_emit_reloc_markers(struct nouveau_channel *chan, struct nouveau_stateobj *so)
 {
-	struct nouveau_pushbuf *pb = nvws->channel->pushbuf;
+	struct nouveau_pushbuf *pb = chan->pushbuf;
 	unsigned i;
 
 	if (!so)
 		return;
 
 	i = so->cur_reloc << 1;
-	if (nvws->channel->pushbuf->remaining < i)
-		nvws->push_flush(nvws, i, NULL);
-	nvws->channel->pushbuf->remaining -= i;
+	if (pb->remaining < i)
+		nouveau_pushbuf_flush(chan, i);
+	pb->remaining -= i;
 
 	for (i = 0; i < so->cur_reloc; i++) {
 		struct nouveau_stateobj_reloc *r = &so->reloc[i];
 
-		nvws->push_reloc(nvws, pb->cur++, r->bo, r->packet,
-				 (r->flags & (NOUVEAU_BO_VRAM |
-					      NOUVEAU_BO_GART |
-					      NOUVEAU_BO_RDWR)) |
-				 NOUVEAU_BO_DUMMY, 0, 0);
-		nvws->push_reloc(nvws, pb->cur++, r->bo, r->data,
-				 r->flags | NOUVEAU_BO_DUMMY, r->vor, r->tor);
+		nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->packet, 0,
+					   (r->flags & (NOUVEAU_BO_VRAM |
+							NOUVEAU_BO_GART |
+							NOUVEAU_BO_RDWR)) |
+					   NOUVEAU_BO_DUMMY, 0, 0);
+		nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, r->data, 0,
+					   r->flags | NOUVEAU_BO_DUMMY,
+					   r->vor, r->tor);
 	}
 }
 
diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h
index ff7dd1c51c..42c77e5e77 100644
--- a/src/gallium/drivers/nouveau/nouveau_winsys.h
+++ b/src/gallium/drivers/nouveau/nouveau_winsys.h
@@ -23,77 +23,38 @@
 #define NOUVEAU_BUFFER_USAGE_ZETA     (1 << 17)
 #define NOUVEAU_BUFFER_USAGE_TRANSFER (1 << 18)
 
-struct nouveau_winsys {
-	struct pipe_winsys *ws;
-
-	struct nouveau_channel *channel;
-
-	int  (*res_init)(struct nouveau_resource **heap, unsigned start,
-			 unsigned size);
-	int  (*res_alloc)(struct nouveau_resource *heap, int size, void *priv,
-			  struct nouveau_resource **);
-	void (*res_free)(struct nouveau_resource **);
-
-	int  (*push_reloc)(struct nouveau_winsys *, void *ptr,
-			   struct pipe_buffer *, uint32_t data,
-			   uint32_t flags, uint32_t vor, uint32_t tor);
-	int  (*push_flush)(struct nouveau_winsys *, unsigned size,
-			   struct pipe_fence_handle **fence);
-			       
-	int       (*grobj_alloc)(struct nouveau_winsys *, int grclass,
-				 struct nouveau_grobj **);
-	void      (*grobj_free)(struct nouveau_grobj **);
-
-	int       (*notifier_alloc)(struct nouveau_winsys *, int count,
-				    struct nouveau_notifier **);
-	void      (*notifier_free)(struct nouveau_notifier **);
-	void      (*notifier_reset)(struct nouveau_notifier *, int id);
-	uint32_t  (*notifier_status)(struct nouveau_notifier *, int id);
-	uint32_t  (*notifier_retval)(struct nouveau_notifier *, int id);
-	int       (*notifier_wait)(struct nouveau_notifier *, int id,
-				   int status, double timeout);
-
-	int (*surface_copy)(struct nouveau_winsys *, struct pipe_surface *,
-			    unsigned, unsigned, struct pipe_surface *,
-			    unsigned, unsigned, unsigned, unsigned);
-	int (*surface_fill)(struct nouveau_winsys *, struct pipe_surface *,
-			    unsigned, unsigned, unsigned, unsigned, unsigned);
-
-	struct nouveau_bo *(*get_bo)(struct pipe_buffer *);
-};
-
 extern struct pipe_screen *
-nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv04_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv04_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv10_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv20_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv30_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv40_create(struct pipe_screen *, unsigned pctx_id);
 
 extern struct pipe_screen *
-nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *);
 
 extern struct pipe_context *
 nv50_create(struct pipe_screen *, unsigned pctx_id);
diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c
index 4da833c25e..93f752faec 100644
--- a/src/gallium/drivers/nv04/nv04_miptree.c
+++ b/src/gallium/drivers/nv04/nv04_miptree.c
@@ -31,7 +31,8 @@ nv04_miptree_layout(struct nv04_miptree *nv04mt)
 
 	for (l = 0; l <= pt->last_level; l++) {
 
-		nv04mt->level[l].image_offset = offset;
+		nv04mt->level[l].image_offset = 
+			CALLOC(nr_faces, sizeof(unsigned));
 		offset += nv04mt->level[l].pitch * pt->height[l];
 	}
 
diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c
index f9f6d97426..ff2febb668 100644
--- a/src/gallium/drivers/nv04/nv04_screen.c
+++ b/src/gallium/drivers/nv04/nv04_screen.c
@@ -1,27 +1,9 @@
 #include "pipe/p_screen.h"
 #include "pipe/p_inlines.h"
-#include "util/u_simple_screen.h"
 
 #include "nv04_context.h"
 #include "nv04_screen.h"
 
-static const char *
-nv04_screen_get_name(struct pipe_screen *screen)
-{
-	struct nv04_screen *nv04screen = nv04_screen(screen);
-	struct nouveau_device *dev = nv04screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv04_screen_get_vendor(struct pipe_screen *screen)
-{
-	return "nouveau";
-}
-
 static int
 nv04_screen_get_param(struct pipe_screen *screen, int param)
 {
@@ -58,6 +40,10 @@ nv04_screen_get_param(struct pipe_screen *screen, int param)
 		return 0;
 	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
 		return 1;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 0;
@@ -97,6 +83,13 @@ nv04_screen_is_format_supported(struct pipe_screen *screen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -123,10 +116,9 @@ static void
 nv04_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv04_screen *screen = nv04_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->fahrenheit);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->fahrenheit);
 	nv04_surface_2d_takedown(&screen->eng2d);
 
 	FREE(pscreen);
@@ -141,21 +133,38 @@ nv04_surface_buffer(struct pipe_surface *surf)
 }
 
 struct pipe_screen *
-nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv04_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv04_screen *screen = CALLOC_STRUCT(nv04_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	unsigned fahrenheit_class = 0, sub3d_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
+
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv04_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv04_screen_destroy;
+	pscreen->get_param = nv04_screen_get_param;
+	pscreen->get_paramf = nv04_screen_get_paramf;
+	pscreen->is_format_supported = nv04_screen_is_format_supported;
 
-	if (chipset>=0x20) {
+	nv04_screen_init_miptree_functions(pscreen);
+	nv04_screen_init_transfer_functions(pscreen);
+
+	if (dev->chipset >= 0x20) {
 		fahrenheit_class = 0;
 		sub3d_class = 0;
-	} else if (chipset>=0x10) {
+	} else if (dev->chipset >= 0x10) {
 		fahrenheit_class = NV10_DX5_TEXTURED_TRIANGLE;
 		sub3d_class = NV10_CONTEXT_SURFACES_3D;
 	} else {
@@ -164,50 +173,40 @@ nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	}
 
 	if (!fahrenheit_class) {
-		NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv04_surface_buffer;
-
 	/* 3D object */
-	ret = nvws->grobj_alloc(nvws, fahrenheit_class, &screen->fahrenheit);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0001, fahrenheit_class,
+				  &screen->fahrenheit);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return NULL;
 	}
+	BIND_RING(chan, screen->fahrenheit, 7);
 
 	/* 3D surface object */
-	ret = nvws->grobj_alloc(nvws, sub3d_class, &screen->context_surfaces_3d);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0002, sub3d_class,
+				  &screen->context_surfaces_3d);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D surface object: %d\n", ret);
 		return NULL;
 	}
+	BIND_RING(chan, screen->context_surfaces_3d, 6);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv04_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv04_screen_destroy(&screen->pipe);
+		nv04_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv04_screen_destroy;
-
-	screen->pipe.get_name = nv04_screen_get_name;
-	screen->pipe.get_vendor = nv04_screen_get_vendor;
-	screen->pipe.get_param = nv04_screen_get_param;
-	screen->pipe.get_paramf = nv04_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv04_screen_is_format_supported;
-
-	nv04_screen_init_miptree_functions(&screen->pipe);
-	nv04_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv04/nv04_screen.h b/src/gallium/drivers/nv04/nv04_screen.h
index ee6fb6db44..11466b9442 100644
--- a/src/gallium/drivers/nv04/nv04_screen.h
+++ b/src/gallium/drivers/nv04/nv04_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV04_SCREEN_H__
 #define __NV04_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04_surface_2d.h"
 
 struct nv04_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 	unsigned chipset;
diff --git a/src/gallium/drivers/nv04/nv04_state.c b/src/gallium/drivers/nv04/nv04_state.c
index 87c635f962..d356ebd8b3 100644
--- a/src/gallium/drivers/nv04/nv04_state.c
+++ b/src/gallium/drivers/nv04/nv04_state.c
@@ -2,6 +2,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -334,7 +335,7 @@ nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 			 const struct pipe_constant_buffer *buf )
 {
 	struct nv04_context *nv04 = nv04_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 
 	assert(shader < PIPE_SHADER_TYPES);
 	assert(index == 0);
@@ -342,12 +343,12 @@ nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 	if (buf) {
 		void *mapped;
 		if (buf->buffer && buf->buffer->size &&
-                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+                    (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
 		{
 			memcpy(nv04->constbuf[shader], mapped, buf->buffer->size);
 			nv04->constbuf_nr[shader] =
 				buf->buffer->size / (4 * sizeof(float));
-			ws->buffer_unmap(ws, buf->buffer);
+			pipe_buffer_unmap(pscreen, buf->buffer);
 		}
 	}
 }
diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h
index 0d51439e3f..399f750dbe 100644
--- a/src/gallium/drivers/nv04/nv04_state.h
+++ b/src/gallium/drivers/nv04/nv04_state.h
@@ -37,7 +37,7 @@ struct nv04_miptree {
 
 	struct {
 		uint pitch;
-		uint image_offset;
+		uint *image_offset;
 	} level[PIPE_MAX_TEXTURE_LEVELS];
 };
 
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c
index f3a8d7efee..f315cf54f0 100644
--- a/src/gallium/drivers/nv04/nv04_surface_2d.c
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.c
@@ -4,6 +4,7 @@
 
 #include "nouveau/nouveau_winsys.h"
 #include "nouveau/nouveau_util.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04_surface_2d.h"
 
 static INLINE int
@@ -96,11 +97,11 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
 			  struct pipe_surface *src, int sx, int sy,
 			  int w, int h)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->swzsurf->channel;
 	struct nouveau_grobj *swzsurf = ctx->swzsurf;
 	struct nouveau_grobj *sifm = ctx->sifm;
-	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	const unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	const unsigned max_w = 1024;
 	const unsigned max_h = 1024;
@@ -167,10 +168,10 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx,
 		       struct pipe_surface *dst, int dx, int dy,
 		       struct pipe_surface *src, int sx, int sy, int w, int h)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->m2mf->channel;
 	struct nouveau_grobj *m2mf = ctx->m2mf;
-	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	unsigned dst_offset = dst->offset + dy * dst_pitch +
@@ -213,11 +214,11 @@ nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 		       int dx, int dy, struct pipe_surface *src, int sx, int sy,
 		       int w, int h)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->surf2d->channel;
 	struct nouveau_grobj *surf2d = ctx->surf2d;
 	struct nouveau_grobj *blit = ctx->blit;
-	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	unsigned src_pitch = ((struct nv04_surface *)src)->pitch;
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	int format;
@@ -266,8 +267,7 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 	 * to NV_MEMORY_TO_MEMORY_FORMAT in this case.
 	 */
 	if ((src->offset & 63) || (dst->offset & 63) ||
-	    (src_pitch & 63) || (dst_pitch & 63) ||
-	    debug_get_bool_option("NOUVEAU_NO_COPYBLIT", FALSE)) {
+	    (src_pitch & 63) || (dst_pitch & 63)) {
 		nv04_surface_copy_m2mf(ctx, dst, dx, dy, src, sx, sy, w, h);
 		return;
 	}
@@ -279,10 +279,10 @@ static void
 nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
 		  int dx, int dy, int w, int h, unsigned value)
 {
-	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_channel *chan = ctx->surf2d->channel;
 	struct nouveau_grobj *surf2d = ctx->surf2d;
 	struct nouveau_grobj *rect = ctx->rect;
-	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst));
 	unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch;
 	int cs2d_format, gdirect_format;
 
@@ -334,10 +334,10 @@ nv04_surface_2d_takedown(struct nv04_surface_2d **pctx)
 }
 
 struct nv04_surface_2d *
-nv04_surface_2d_init(struct nouveau_winsys *nvws)
+nv04_surface_2d_init(struct nouveau_screen *screen)
 {
 	struct nv04_surface_2d *ctx = CALLOC_STRUCT(nv04_surface_2d);
-	struct nouveau_channel *chan = nvws->channel;
+	struct nouveau_channel *chan = screen->channel;
 	unsigned handle = 0x88000000, class;
 	int ret;
 
@@ -460,7 +460,6 @@ nv04_surface_2d_init(struct nouveau_winsys *nvws)
 		return NULL;
 	}
 
-	ctx->nvws = nvws;
 	ctx->copy = nv04_surface_copy;
 	ctx->fill = nv04_surface_fill;
 	return ctx;
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.h b/src/gallium/drivers/nv04/nv04_surface_2d.h
index 82ce7189c8..02b3f56ba8 100644
--- a/src/gallium/drivers/nv04/nv04_surface_2d.h
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.h
@@ -7,7 +7,6 @@ struct nv04_surface {
 };
 
 struct nv04_surface_2d {
-	struct nouveau_winsys *nvws;
 	struct nouveau_notifier *ntfy;
 	struct nouveau_grobj *surf2d;
 	struct nouveau_grobj *swzsurf;
@@ -26,7 +25,7 @@ struct nv04_surface_2d {
 };
 
 struct nv04_surface_2d *
-nv04_surface_2d_init(struct nouveau_winsys *nvws);
+nv04_surface_2d_init(struct nouveau_screen *screen);
 
 void
 nv04_surface_2d_takedown(struct nv04_surface_2d **);
diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c
index d21a0e34f7..e3167814f2 100644
--- a/src/gallium/drivers/nv04/nv04_vbo.c
+++ b/src/gallium/drivers/nv04/nv04_vbo.c
@@ -1,6 +1,7 @@
 #include "draw/draw_context.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv04_context.h"
 #include "nv04_state.h"
@@ -13,6 +14,7 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
                     unsigned indexSize,
                     unsigned prim, unsigned start, unsigned count)
 {
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nv04_context *nv04 = nv04_context( pipe );
 	struct draw_context *draw = nv04->draw;
 	unsigned i;
@@ -25,17 +27,17 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv04->vtxbuf[i].buffer) {
 			void *buf
-				= pipe->winsys->buffer_map(pipe->winsys,
-						nv04->vtxbuf[i].buffer,
-						PIPE_BUFFER_USAGE_CPU_READ);
+				= pipe_buffer_map(pscreen,
+						  nv04->vtxbuf[i].buffer,
+						  PIPE_BUFFER_USAGE_CPU_READ);
 			draw_set_mapped_vertex_buffer(draw, i, buf);
 		}
 	}
 	/* Map index buffer, if present */
 	if (indexBuffer) {
 		void *mapped_indexes
-			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
-					PIPE_BUFFER_USAGE_CPU_READ);
+			= pipe_buffer_map(pscreen, indexBuffer,
+					  PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
 	}
 	else {
@@ -55,12 +57,12 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv04->vtxbuf[i].buffer) {
-			pipe->winsys->buffer_unmap(pipe->winsys, nv04->vtxbuf[i].buffer);
+			pipe_buffer_unmap(pscreen, nv04->vtxbuf[i].buffer);
 			draw_set_mapped_vertex_buffer(draw, i, NULL);
 		}
 	}
 	if (indexBuffer) {
-		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		pipe_buffer_unmap(pscreen, indexBuffer);
 		draw_set_mapped_element_buffer(draw, 0, NULL);
 	}
 
diff --git a/src/gallium/drivers/nv10/nv10_context.c b/src/gallium/drivers/nv10/nv10_context.c
index 3da8d2f568..a127b134ec 100644
--- a/src/gallium/drivers/nv10/nv10_context.c
+++ b/src/gallium/drivers/nv10/nv10_context.c
@@ -30,18 +30,18 @@ nv10_destroy(struct pipe_context *pipe)
 static void nv10_init_hwctx(struct nv10_context *nv10)
 {
 	struct nv10_screen *screen = nv10->screen;
-	struct nouveau_winsys *nvws = screen->nvws;
+	struct nouveau_channel *chan = screen->base.channel;
 	int i;
 	float projectionmatrix[16];
 
 	BEGIN_RING(celsius, NV10TCL_DMA_NOTIFY, 1);
 	OUT_RING  (screen->sync->handle);
 	BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY0, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->gart->handle);
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->gart->handle);
 	BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY2, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->vram->handle);
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->vram->handle);
 
 	BEGIN_RING(celsius, NV10TCL_NOP, 1);
 	OUT_RING  (0);
diff --git a/src/gallium/drivers/nv10/nv10_prim_vbuf.c b/src/gallium/drivers/nv10/nv10_prim_vbuf.c
index 089c236302..1806d5f8cc 100644
--- a/src/gallium/drivers/nv10/nv10_prim_vbuf.c
+++ b/src/gallium/drivers/nv10/nv10_prim_vbuf.c
@@ -40,7 +40,6 @@
 
 #include "util/u_debug.h"
 #include "pipe/p_inlines.h"
-#include "pipe/internal/p_winsys_screen.h"
 
 #include "nv10_context.h"
 #include "nv10_state.h"
@@ -124,11 +123,10 @@ nv10_vbuf_render_map_vertices( struct vbuf_render *render )
 {
 	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
 	struct nv10_context *nv10 = nv10_render->nv10;
-	struct pipe_winsys *winsys = nv10->pipe.winsys;
+	struct pipe_screen *pscreen = nv10->pipe.screen;
 
-	return winsys->buffer_map(winsys, 
-			nv10_render->buffer, 
-			PIPE_BUFFER_USAGE_CPU_WRITE);
+	return pipe_buffer_map(pscreen, nv10_render->buffer,
+			       PIPE_BUFFER_USAGE_CPU_WRITE);
 }
 
 static void
@@ -138,10 +136,10 @@ nv10_vbuf_render_unmap_vertices( struct vbuf_render *render,
 {
 	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
 	struct nv10_context *nv10 = nv10_render->nv10;
-	struct pipe_winsys *winsys = nv10->pipe.winsys;
+	struct pipe_screen *pscreen = nv10->pipe.screen;
 
 	assert(!nv10_render->buffer);
-	winsys->buffer_unmap(winsys, nv10_render->buffer);
+	pipe_buffer_unmap(pscreen, nv10_render->buffer);
 }
 
 static boolean
@@ -202,8 +200,6 @@ static void
 nv10_vbuf_render_release_vertices( struct vbuf_render *render )
 {
 	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
-	struct nv10_context *nv10 = nv10_render->nv10;
-	struct pipe_screen *pscreen = &nv10->screen->pipe;
 
 	assert(nv10_render->buffer);
 	pipe_buffer_reference(&nv10_render->buffer, NULL);
diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c
index 6532a93c7b..4469b22d91 100644
--- a/src/gallium/drivers/nv10/nv10_screen.c
+++ b/src/gallium/drivers/nv10/nv10_screen.c
@@ -1,26 +1,8 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
 
 #include "nv10_context.h"
 #include "nv10_screen.h"
 
-static const char *
-nv10_screen_get_name(struct pipe_screen *screen)
-{
-	struct nv10_screen *nv10screen = nv10_screen(screen);
-	struct nouveau_device *dev = nv10screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv10_screen_get_vendor(struct pipe_screen *screen)
-{
-	return "nouveau";
-}
-
 static int
 nv10_screen_get_param(struct pipe_screen *screen, int param)
 {
@@ -53,6 +35,10 @@ nv10_screen_get_param(struct pipe_screen *screen, int param)
 		return 12;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 0;
@@ -92,7 +78,15 @@ nv10_screen_is_format_supported(struct pipe_screen *screen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			 return TRUE;
+		default:
+			 break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -120,10 +114,9 @@ static void
 nv10_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv10_screen *screen = nv10_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->celsius);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->celsius);
 
 	FREE(pscreen);
 }
@@ -137,64 +130,69 @@ nv10_surface_buffer(struct pipe_surface *surf)
 }
 
 struct pipe_screen *
-nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv10_screen *screen = CALLOC_STRUCT(nv10_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	unsigned celsius_class;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv10_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv10_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv10_screen_destroy;
+	pscreen->get_param = nv10_screen_get_param;
+	pscreen->get_paramf = nv10_screen_get_paramf;
+	pscreen->is_format_supported = nv10_screen_is_format_supported;
+
+	nv10_screen_init_miptree_functions(pscreen);
+	nv10_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	if (chipset>=0x20)
-		celsius_class=NV11TCL;
-	else if (chipset>=0x17)
-		celsius_class=NV17TCL;
-	else if (chipset>=0x11)
-		celsius_class=NV11TCL;
+	if (dev->chipset >= 0x20)
+		celsius_class = NV11TCL;
+	else if (dev->chipset >= 0x17)
+		celsius_class = NV17TCL;
+	else if (dev->chipset >= 0x11)
+		celsius_class = NV11TCL;
 	else
-		celsius_class=NV10TCL;
+		celsius_class = NV10TCL;
 
 	if (!celsius_class) {
-		NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, celsius_class, &screen->celsius);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0001, celsius_class,
+				  &screen->celsius);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->celsius, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv10_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv10_screen_destroy(&screen->pipe);
+		nv10_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv10_screen_destroy;
-
-	screen->pipe.get_name = nv10_screen_get_name;
-	screen->pipe.get_vendor = nv10_screen_get_vendor;
-	screen->pipe.get_param = nv10_screen_get_param;
-	screen->pipe.get_paramf = nv10_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv10_screen_is_format_supported;
-
-	nv10_screen_init_miptree_functions(&screen->pipe);
-	nv10_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv10/nv10_screen.h b/src/gallium/drivers/nv10/nv10_screen.h
index ad829ee3fd..86b6d8def5 100644
--- a/src/gallium/drivers/nv10/nv10_screen.h
+++ b/src/gallium/drivers/nv10/nv10_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV10_SCREEN_H__
 #define __NV10_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04/nv04_surface_2d.h"
 
 struct nv10_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv10/nv10_state.c b/src/gallium/drivers/nv10/nv10_state.c
index 119af66dfd..9b38219b99 100644
--- a/src/gallium/drivers/nv10/nv10_state.c
+++ b/src/gallium/drivers/nv10/nv10_state.c
@@ -2,6 +2,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -460,7 +461,7 @@ nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 			 const struct pipe_constant_buffer *buf )
 {
 	struct nv10_context *nv10 = nv10_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 
 	assert(shader < PIPE_SHADER_TYPES);
 	assert(index == 0);
@@ -468,12 +469,12 @@ nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 	if (buf) {
 		void *mapped;
 		if (buf->buffer && buf->buffer->size &&
-                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+                    (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
 		{
 			memcpy(nv10->constbuf[shader], mapped, buf->buffer->size);
 			nv10->constbuf_nr[shader] =
 				buf->buffer->size / (4 * sizeof(float));
-			ws->buffer_unmap(ws, buf->buffer);
+			pipe_buffer_unmap(pscreen, buf->buffer);
 		}
 	}
 }
diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c
index d0e788ac03..441a4f75f3 100644
--- a/src/gallium/drivers/nv10/nv10_vbo.c
+++ b/src/gallium/drivers/nv10/nv10_vbo.c
@@ -1,6 +1,7 @@
 #include "draw/draw_context.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv10_context.h"
 #include "nv10_state.h"
@@ -15,6 +16,7 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 {
 	struct nv10_context *nv10 = nv10_context( pipe );
 	struct draw_context *draw = nv10->draw;
+	struct pipe_screen *pscreen = pipe->screen;
 	unsigned i;
 
 	nv10_emit_hw_state(nv10);
@@ -24,9 +26,8 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv10->vtxbuf[i].buffer) {
-			void *buf
-				= pipe->winsys->buffer_map(pipe->winsys,
-						nv10->vtxbuf[i].buffer,
+			void *buf =
+				pipe_buffer_map(pscreen, nv10->vtxbuf[i].buffer,
 						PIPE_BUFFER_USAGE_CPU_READ);
 			draw_set_mapped_vertex_buffer(draw, i, buf);
 		}
@@ -34,8 +35,8 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 	/* Map index buffer, if present */
 	if (indexBuffer) {
 		void *mapped_indexes
-			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
-					PIPE_BUFFER_USAGE_CPU_READ);
+			= pipe_buffer_map(pscreen, indexBuffer,
+					  PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
 	}
 	else {
@@ -55,12 +56,12 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv10->vtxbuf[i].buffer) {
-			pipe->winsys->buffer_unmap(pipe->winsys, nv10->vtxbuf[i].buffer);
+			pipe_buffer_unmap(pscreen, nv10->vtxbuf[i].buffer);
 			draw_set_mapped_vertex_buffer(draw, i, NULL);
 		}
 	}
 	if (indexBuffer) {
-		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		pipe_buffer_unmap(pscreen, indexBuffer);
 		draw_set_mapped_element_buffer(draw, 0, NULL);
 	}
 
diff --git a/src/gallium/drivers/nv20/nv20_context.c b/src/gallium/drivers/nv20/nv20_context.c
index cbc41707d5..b32d0d83ba 100644
--- a/src/gallium/drivers/nv20/nv20_context.c
+++ b/src/gallium/drivers/nv20/nv20_context.c
@@ -30,7 +30,7 @@ nv20_destroy(struct pipe_context *pipe)
 static void nv20_init_hwctx(struct nv20_context *nv20)
 {
 	struct nv20_screen *screen = nv20->screen;
-	struct nouveau_winsys *nvws = screen->nvws;
+	struct nouveau_channel *chan = screen->base.channel;
 	int i;
 	float projectionmatrix[16];
 	const boolean is_nv25tcl = (nv20->screen->kelvin->grclass == NV25TCL);
@@ -38,11 +38,11 @@ static void nv20_init_hwctx(struct nv20_context *nv20)
 	BEGIN_RING(kelvin, NV20TCL_DMA_NOTIFY, 1);
 	OUT_RING  (screen->sync->handle);
 	BEGIN_RING(kelvin, NV20TCL_DMA_TEXTURE0, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->gart->handle); /* TEXTURE1 */
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->gart->handle); /* TEXTURE1 */
 	BEGIN_RING(kelvin, NV20TCL_DMA_COLOR, 2);
-	OUT_RING  (nvws->channel->vram->handle);
-	OUT_RING  (nvws->channel->vram->handle); /* ZETA */
+	OUT_RING  (chan->vram->handle);
+	OUT_RING  (chan->vram->handle); /* ZETA */
 
 	BEGIN_RING(kelvin, NV20TCL_DMA_QUERY, 1);
 	OUT_RING  (0); /* renouveau: beef0351, unique */
@@ -99,9 +99,9 @@ static void nv20_init_hwctx(struct nv20_context *nv20)
 		OUT_RING  (3);
 
 		BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY9, 1);
-		OUT_RING  (nvws->channel->vram->handle);
+		OUT_RING  (chan->vram->handle);
 		BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY8, 1);
-		OUT_RING  (nvws->channel->vram->handle);
+		OUT_RING  (chan->vram->handle);
 	}
 	BEGIN_RING(kelvin, NV20TCL_DMA_FENCE, 1);
 	OUT_RING  (0);	/* renouveau: beef1e10 */
diff --git a/src/gallium/drivers/nv20/nv20_prim_vbuf.c b/src/gallium/drivers/nv20/nv20_prim_vbuf.c
index 8aa342cd2d..ddfcdb8057 100644
--- a/src/gallium/drivers/nv20/nv20_prim_vbuf.c
+++ b/src/gallium/drivers/nv20/nv20_prim_vbuf.c
@@ -152,12 +152,11 @@ static void *
 nv20_vbuf_render_map_vertices( struct vbuf_render *render )
 {
 	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
-	struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys;
+	struct pipe_screen *pscreen = nv20_render->nv20->pipe.screen;
 
 	if (nv20_render->pbuffer) {
-		return winsys->buffer_map(winsys,
-				nv20_render->pbuffer,
-				PIPE_BUFFER_USAGE_CPU_WRITE);
+		return pipe_buffer_map(pscreen, nv20_render->pbuffer,
+				       PIPE_BUFFER_USAGE_CPU_WRITE);
 	} else if (nv20_render->mbuffer) {
 		return nv20_render->mbuffer;
 	} else
@@ -173,10 +172,10 @@ nv20_vbuf_render_unmap_vertices( struct vbuf_render *render,
 		ushort max_index )
 {
 	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
-	struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys;
+	struct pipe_screen *pscreen = nv20_render->nv20->pipe.screen;
 
 	if (nv20_render->pbuffer)
-		winsys->buffer_unmap(winsys, nv20_render->pbuffer);
+		pipe_buffer_unmap(pscreen, nv20_render->pbuffer);
 }
 
 static boolean
@@ -358,7 +357,6 @@ nv20_vbuf_render_release_vertices( struct vbuf_render *render )
 {
 	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
 	struct nv20_context *nv20 = nv20_render->nv20;
-	struct pipe_screen *pscreen = &nv20->screen->pipe;
 
 	if (nv20_render->pbuffer) {
 		pipe_buffer_reference(&nv20_render->pbuffer, NULL);
diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c
index 7760ae27c0..e6924ad71e 100644
--- a/src/gallium/drivers/nv20/nv20_screen.c
+++ b/src/gallium/drivers/nv20/nv20_screen.c
@@ -1,26 +1,8 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
 
 #include "nv20_context.h"
 #include "nv20_screen.h"
 
-static const char *
-nv20_screen_get_name(struct pipe_screen *screen)
-{
-	struct nv20_screen *nv20screen = nv20_screen(screen);
-	struct nouveau_device *dev = nv20screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv20_screen_get_vendor(struct pipe_screen *screen)
-{
-	return "nouveau";
-}
-
 static int
 nv20_screen_get_param(struct pipe_screen *screen, int param)
 {
@@ -53,6 +35,10 @@ nv20_screen_get_param(struct pipe_screen *screen, int param)
 		return 12;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 0;
@@ -92,7 +78,15 @@ nv20_screen_is_format_supported(struct pipe_screen *screen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			return TRUE;
+		default:
+			 break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -120,10 +114,9 @@ static void
 nv20_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv20_screen *screen = nv20_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->kelvin);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->kelvin);
 
 	FREE(pscreen);
 }
@@ -137,60 +130,65 @@ nv20_surface_buffer(struct pipe_surface *surf)
 }
 
 struct pipe_screen *
-nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv20_screen *screen = CALLOC_STRUCT(nv20_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	unsigned kelvin_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv20_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv20_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv20_screen_destroy;
+	pscreen->get_param = nv20_screen_get_param;
+	pscreen->get_paramf = nv20_screen_get_paramf;
+	pscreen->is_format_supported = nv20_screen_is_format_supported;
+
+	nv20_screen_init_miptree_functions(pscreen);
+	nv20_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	if (chipset >= 0x25)
+	if (dev->chipset >= 0x25)
 		kelvin_class = NV25TCL;
-	else if (chipset >= 0x20)
+	else if (dev->chipset >= 0x20)
 		kelvin_class = NV20TCL;
 
-	if (!kelvin_class || chipset >= 0x30) {
-		NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", chipset);
+	if (!kelvin_class || dev->chipset >= 0x30) {
+		NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, kelvin_class, &screen->kelvin);
+	ret = nouveau_grobj_alloc(chan, 0xbeef0097, kelvin_class,
+				  &screen->kelvin);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->kelvin, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv20_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv20_screen_destroy(&screen->pipe);
+		nv20_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv20_screen_destroy;
-
-	screen->pipe.get_name = nv20_screen_get_name;
-	screen->pipe.get_vendor = nv20_screen_get_vendor;
-	screen->pipe.get_param = nv20_screen_get_param;
-	screen->pipe.get_paramf = nv20_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv20_screen_is_format_supported;
-
-	nv20_screen_init_miptree_functions(&screen->pipe);
-	nv20_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv20/nv20_screen.h b/src/gallium/drivers/nv20/nv20_screen.h
index d9fce2bced..fc7bb05033 100644
--- a/src/gallium/drivers/nv20/nv20_screen.h
+++ b/src/gallium/drivers/nv20/nv20_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV20_SCREEN_H__
 #define __NV20_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04/nv04_surface_2d.h"
 
 struct nv20_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv20/nv20_state.c b/src/gallium/drivers/nv20/nv20_state.c
index ecec4f49a0..ed4084980f 100644
--- a/src/gallium/drivers/nv20/nv20_state.c
+++ b/src/gallium/drivers/nv20/nv20_state.c
@@ -2,6 +2,7 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -453,7 +454,7 @@ nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 			 const struct pipe_constant_buffer *buf )
 {
 	struct nv20_context *nv20 = nv20_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 
 	assert(shader < PIPE_SHADER_TYPES);
 	assert(index == 0);
@@ -461,12 +462,12 @@ nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 	if (buf) {
 		void *mapped;
 		if (buf->buffer && buf->buffer->size &&
-                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+                    (mapped = pipe_buffer_map(pscreen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
 		{
 			memcpy(nv20->constbuf[shader], mapped, buf->buffer->size);
 			nv20->constbuf_nr[shader] =
 				buf->buffer->size / (4 * sizeof(float));
-			ws->buffer_unmap(ws, buf->buffer);
+			pipe_buffer_unmap(pscreen, buf->buffer);
 		}
 	}
 }
diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c
index 24d8f4bef0..84d7db6c5e 100644
--- a/src/gallium/drivers/nv20/nv20_vbo.c
+++ b/src/gallium/drivers/nv20/nv20_vbo.c
@@ -1,6 +1,7 @@
 #include "draw/draw_context.h"
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv20_context.h"
 #include "nv20_state.h"
@@ -13,6 +14,7 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
                     unsigned indexSize,
                     unsigned prim, unsigned start, unsigned count)
 {
+	struct pipe_screen *pscreen = pipe->screen;
 	struct nv20_context *nv20 = nv20_context( pipe );
 	struct draw_context *draw = nv20->draw;
 	unsigned i;
@@ -25,17 +27,17 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv20->vtxbuf[i].buffer) {
 			void *buf
-				= pipe->winsys->buffer_map(pipe->winsys,
-						nv20->vtxbuf[i].buffer,
-						PIPE_BUFFER_USAGE_CPU_READ);
+				= pipe_buffer_map(pscreen,
+						  nv20->vtxbuf[i].buffer,
+						  PIPE_BUFFER_USAGE_CPU_READ);
 			draw_set_mapped_vertex_buffer(draw, i, buf);
 		}
 	}
 	/* Map index buffer, if present */
 	if (indexBuffer) {
 		void *mapped_indexes
-			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
-					PIPE_BUFFER_USAGE_CPU_READ);
+			= pipe_buffer_map(pscreen, indexBuffer,
+					  PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
 	}
 	else {
@@ -55,12 +57,12 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
 	 */
 	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
 		if (nv20->vtxbuf[i].buffer) {
-			pipe->winsys->buffer_unmap(pipe->winsys, nv20->vtxbuf[i].buffer);
+			pipe_buffer_unmap(pscreen, nv20->vtxbuf[i].buffer);
 			draw_set_mapped_vertex_buffer(draw, i, NULL);
 		}
 	}
 	if (indexBuffer) {
-		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		pipe_buffer_unmap(pscreen, indexBuffer);
 		draw_set_mapped_element_buffer(draw, 0, NULL);
 	}
 
diff --git a/src/gallium/drivers/nv20/nv20_vertprog.c b/src/gallium/drivers/nv20/nv20_vertprog.c
index 5db0e807ff..c1e588902b 100644
--- a/src/gallium/drivers/nv20/nv20_vertprog.c
+++ b/src/gallium/drivers/nv20/nv20_vertprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -645,8 +646,8 @@ out_err:
 static boolean
 nv20_vertprog_validate(struct nv20_context *nv20)
 { 
+	struct pipe_screen *pscreen = nv20->pipe.screen;
 	struct nouveau_winsys *nvws = nv20->nvws;
-	struct pipe_winsys *ws = nv20->pipe.winsys;
 	struct nouveau_grobj *rankine = nv20->screen->rankine;
 	struct nv20_vertex_program *vp;
 	struct pipe_buffer *constbuf;
@@ -749,8 +750,8 @@ nv20_vertprog_validate(struct nv20_context *nv20)
 		float *map = NULL;
 
 		if (constbuf) {
-			map = ws->buffer_map(ws, constbuf,
-					     PIPE_BUFFER_USAGE_CPU_READ);
+			map = pipe_buffer_map(pscreen, constbuf,
+					      PIPE_BUFFER_USAGE_CPU_READ);
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
@@ -770,9 +771,8 @@ nv20_vertprog_validate(struct nv20_context *nv20)
 			OUT_RINGp ((uint32_t *)vpd->value, 4);
 		}
 
-		if (constbuf) {
-			ws->buffer_unmap(ws, constbuf);
-		}
+		if (constbuf)
+			pipe_buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c
index bdfe1425d2..1d1c556fb1 100644
--- a/src/gallium/drivers/nv30/nv30_fragprog.c
+++ b/src/gallium/drivers/nv30/nv30_fragprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -798,12 +799,12 @@ static void
 nv30_fragprog_upload(struct nv30_context *nv30,
 		     struct nv30_fragment_program *fp)
 {
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	const uint32_t le = 1;
 	uint32_t *map;
 	int i;
 
-	map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+	map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
 
 #if 0
 	for (i = 0; i < fp->insn_len; i++) {
@@ -825,7 +826,7 @@ nv30_fragprog_upload(struct nv30_context *nv30,
 		}
 	}
 
-	ws->buffer_unmap(ws, fp->buffer);
+	pipe_buffer_unmap(pscreen, fp->buffer);
 }
 
 static boolean
@@ -834,8 +835,7 @@ nv30_fragprog_validate(struct nv30_context *nv30)
 	struct nv30_fragment_program *fp = nv30->fragprog;
 	struct pipe_buffer *constbuf =
 		nv30->constbuf[PIPE_SHADER_FRAGMENT];
-	struct pipe_screen *screen = nv30->pipe.screen;
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	struct nouveau_stateobj *so;
 	boolean new_consts = FALSE;
 	int i;
@@ -850,14 +850,15 @@ nv30_fragprog_validate(struct nv30_context *nv30)
 		return FALSE;
 	}
 
-	fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4);
+	fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4);
 	nv30_fragprog_upload(nv30, fp);
 
 	so = so_new(8, 1);
 	so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1);
-	so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-		  NV34TCL_FP_ACTIVE_PROGRAM_DMA0, NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
+	so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM |
+		      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
+		      NOUVEAU_BO_OR, NV34TCL_FP_ACTIVE_PROGRAM_DMA0,
+		      NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
 	so_method(so, nv30->screen->rankine, NV34TCL_FP_CONTROL, 1);
 	so_data  (so, fp->fp_control);
 	so_method(so, nv30->screen->rankine, NV34TCL_FP_REG_CONTROL, 1);
@@ -871,7 +872,8 @@ update_constants:
 	if (fp->nr_consts) {
 		float *map;
 		
-		map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen, constbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		for (i = 0; i < fp->nr_consts; i++) {
 			struct nv30_fragment_program_data *fpd = &fp->consts[i];
 			uint32_t *p = &fp->insn[fpd->offset];
@@ -882,7 +884,7 @@ update_constants:
 			memcpy(p, cb, 4 * sizeof(float));
 			new_consts = TRUE;
 		}
-		ws->buffer_unmap(ws, constbuf);
+		pipe_buffer_unmap(pscreen, constbuf);
 
 		if (new_consts)
 			nv30_fragprog_upload(nv30, fp);
diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c
index 8b6ab992d1..822e1d8def 100644
--- a/src/gallium/drivers/nv30/nv30_fragtex.c
+++ b/src/gallium/drivers/nv30/nv30_fragtex.c
@@ -61,6 +61,7 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit)
 	struct nv30_sampler_state *ps = nv30->tex_sampler[unit];
 	struct nv30_miptree *nv30mt = nv30->tex_miptree[unit];
 	struct pipe_texture *pt = &nv30mt->base;
+	struct nouveau_bo *bo = nouveau_bo(nv30mt->buffer);
 	struct nv30_texture_format *tf;
 	struct nouveau_stateobj *so;
 	uint32_t txf, txs , txp;
@@ -106,9 +107,9 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit)
 
 	so = so_new(16, 2);
 	so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8);
-	so_reloc (so, nv30mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
-	so_reloc (so, nv30mt->buffer, txf, tex_flags | NOUVEAU_BO_OR,
-		  NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
+	so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR,
+		      NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
 	so_data  (so, ps->wrap);
 	so_data  (so, NV34TCL_TX_ENABLE_ENABLE | ps->en);
 	so_data  (so, txs);
diff --git a/src/gallium/drivers/nv30/nv30_query.c b/src/gallium/drivers/nv30/nv30_query.c
index 2f974cf5c4..1d1c8a484e 100644
--- a/src/gallium/drivers/nv30/nv30_query.c
+++ b/src/gallium/drivers/nv30/nv30_query.c
@@ -29,11 +29,10 @@ nv30_query_create(struct pipe_context *pipe, unsigned query_type)
 static void
 nv30_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
 {
-	struct nv30_context *nv30 = nv30_context(pipe);
 	struct nv30_query *q = nv30_query(pq);
 
 	if (q->object)
-		nv30->nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	FREE(q);
 }
 
@@ -54,9 +53,9 @@ nv30_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
 		pipe->get_query_result(pipe, pq, 1, &tmp);
 	}
 
-	if (nv30->nvws->res_alloc(nv30->screen->query_heap, 1, NULL, &q->object))
+	if (nouveau_resource_alloc(nv30->screen->query_heap, 1, NULL, &q->object))
 		assert(0);
-	nv30->nvws->notifier_reset(nv30->screen->query, q->object->start);
+	nouveau_notifier_reset(nv30->screen->query, q->object->start);
 
 	BEGIN_RING(rankine, NV34TCL_QUERY_RESET, 1);
 	OUT_RING  (1);
@@ -84,27 +83,27 @@ nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
 	struct nv30_query *q = nv30_query(pq);
-	struct nouveau_winsys *nvws = nv30->nvws;
 
 	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 
 	if (!q->ready) {
 		unsigned status;
 
-		status = nvws->notifier_status(nv30->screen->query,
-					       q->object->start);
+		status = nouveau_notifier_status(nv30->screen->query,
+						 q->object->start);
 		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
 			if (wait == FALSE)
 				return FALSE;
-			nvws->notifier_wait(nv30->screen->query, q->object->start,
-					    NV_NOTIFY_STATE_STATUS_COMPLETED,
-					    0);
+
+			nouveau_notifier_wait_status(nv30->screen->query,
+					q->object->start,
+					NV_NOTIFY_STATE_STATUS_COMPLETED, 0);
 		}
 
-		q->result = nvws->notifier_retval(nv30->screen->query,
-						  q->object->start);
+		q->result = nouveau_notifier_return_val(nv30->screen->query,
+							q->object->start);
 		q->ready = TRUE;
-		nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	}
 
 	*result = q->result;
diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c
index d395c5e1b7..c8b40784b0 100644
--- a/src/gallium/drivers/nv30/nv30_screen.c
+++ b/src/gallium/drivers/nv30/nv30_screen.c
@@ -1,5 +1,7 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
+#include "pipe/p_state.h"
+
+#include "nouveau/nouveau_screen.h"
 
 #include "nv30_context.h"
 #include "nv30_screen.h"
@@ -8,23 +10,6 @@
 #define NV34TCL_CHIPSET_3X_MASK 0x00000010
 #define NV35TCL_CHIPSET_3X_MASK 0x000001e0
 
-static const char *
-nv30_screen_get_name(struct pipe_screen *pscreen)
-{
-	struct nv30_screen *screen = nv30_screen(pscreen);
-	struct nouveau_device *dev = screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv30_screen_get_vendor(struct pipe_screen *pscreen)
-{
-	return "nouveau";
-}
-
 static int
 nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 {
@@ -61,6 +46,10 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 0;
 	case NOUVEAU_CAP_HW_VTXBUF:
 	case NOUVEAU_CAP_HW_IDXBUF:
 		return 1;
@@ -100,7 +89,15 @@ nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (tex_usage) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -139,45 +136,57 @@ static void
 nv30_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv30_screen *screen = nv30_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->res_free(&screen->vp_exec_heap);
-	nvws->res_free(&screen->vp_data_heap);
-	nvws->res_free(&screen->query_heap);
-	nvws->notifier_free(&screen->query);
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->rankine);
+	nouveau_resource_free(&screen->vp_exec_heap);
+	nouveau_resource_free(&screen->vp_data_heap);
+	nouveau_resource_free(&screen->query_heap);
+	nouveau_notifier_free(&screen->query);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->rankine);
 
 	FREE(pscreen);
 }
 
 struct pipe_screen *
-nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
 	unsigned rankine_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret, i;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv30_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv30_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv30_screen_destroy;
+	pscreen->get_param = nv30_screen_get_param;
+	pscreen->get_paramf = nv30_screen_get_paramf;
+	pscreen->is_format_supported = nv30_screen_surface_format_supported;
+
+	nv30_screen_init_miptree_functions(pscreen);
+	nv30_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	switch (chipset & 0xf0) {
+	switch (dev->chipset & 0xf0) {
 	case 0x30:
-		if (NV30TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+		if (NV30TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
 			rankine_class = 0x0397;
 		else
-		if (NV34TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+		if (NV34TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
 			rankine_class = 0x0697;
 		else
-		if (NV35TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+		if (NV35TCL_CHIPSET_3X_MASK & (1 << (dev->chipset & 0x0f)))
 			rankine_class = 0x0497;
 		break;
 	default:
@@ -185,43 +194,49 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	}
 
 	if (!rankine_class) {
-		NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, rankine_class, &screen->rankine);
+	ret = nouveau_grobj_alloc(chan, 0xbeef3097, rankine_class,
+				  &screen->rankine);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->rankine, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv30_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv30_screen_destroy(&screen->pipe);
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Query objects */
-	ret = nvws->notifier_alloc(nvws, 32, &screen->query);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
-		nv30_screen_destroy(&screen->pipe);
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	ret = nvws->res_init(&screen->query_heap, 0, 32);
+	ret = nouveau_resource_init(&screen->query_heap, 0, 32);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
-		nv30_screen_destroy(&screen->pipe);
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Vtxprog resources */
-	if (nvws->res_init(&screen->vp_exec_heap, 0, 256) ||
-	    nvws->res_init(&screen->vp_data_heap, 0, 256)) {
-		nv30_screen_destroy(&screen->pipe);
+	if (nouveau_resource_init(&screen->vp_exec_heap, 0, 256) ||
+	    nouveau_resource_init(&screen->vp_data_heap, 0, 256)) {
+		nv30_screen_destroy(pscreen);
 		return NULL;
 	}
 
@@ -230,23 +245,23 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1);
 	so_data  (so, screen->sync->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_COLOR1, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_COLOR0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_VTXBUF0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 /*	so_method(so, screen->rankine, NV34TCL_DMA_FENCE, 2);
 	so_data  (so, 0);
 	so_data  (so, screen->query->handle);*/
 	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY7, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY8, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 
 	for (i=1; i<8; i++) {
 		so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1);
@@ -301,23 +316,9 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->rankine, 0x1e94, 1);
 	so_data  (so, 0x13);
 
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref(NULL, &so);
-	nvws->push_flush(nvws, 0, NULL);
-
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv30_screen_destroy;
-
-	screen->pipe.get_name = nv30_screen_get_name;
-	screen->pipe.get_vendor = nv30_screen_get_vendor;
-	screen->pipe.get_param = nv30_screen_get_param;
-	screen->pipe.get_paramf = nv30_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv30_screen_surface_format_supported;
-
-	nv30_screen_init_miptree_functions(&screen->pipe);
-	nv30_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
+	nouveau_pushbuf_flush(chan, 0);
 
-	return &screen->pipe;
+	return pscreen;
 }
diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nv30/nv30_screen.h
index 8e36883975..5fbd998b53 100644
--- a/src/gallium/drivers/nv30/nv30_screen.h
+++ b/src/gallium/drivers/nv30/nv30_screen.h
@@ -1,11 +1,12 @@
 #ifndef __NV30_SCREEN_H__
 #define __NV30_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
+
 #include "nv04/nv04_surface_2d.h"
 
 struct nv30_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c
index c18be20a32..621b8846c8 100644
--- a/src/gallium/drivers/nv30/nv30_state_emit.c
+++ b/src/gallium/drivers/nv30/nv30_state_emit.c
@@ -38,6 +38,7 @@ nv30_state_do_validate(struct nv30_context *nv30,
 void
 nv30_state_emit(struct nv30_context *nv30)
 {
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 	struct nv30_state *state = &nv30->state;
 	struct nv30_screen *screen = nv30->screen;
 	unsigned i, samplers;
@@ -57,23 +58,23 @@ nv30_state_emit(struct nv30_context *nv30)
 			continue;
 		so_ref (state->hw[i], &nv30->screen->state[i]);
 		if (state->hw[i])
-			so_emit(nv30->nvws, nv30->screen->state[i]);
+			so_emit(chan, nv30->screen->state[i]);
 		states &= ~(1ULL << i);
 	}
 
 	state->dirty = 0;
 
-	so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FB]);
+	so_emit_reloc_markers(chan, state->hw[NV30_STATE_FB]);
 	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
 		if (!(samplers & (1 << i)))
 			continue;
-		so_emit_reloc_markers(nv30->nvws,
+		so_emit_reloc_markers(chan,
 				      state->hw[NV30_STATE_FRAGTEX0+i]);
 		samplers &= ~(1ULL << i);
 	}
-	so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FRAGPROG]);
+	so_emit_reloc_markers(chan, state->hw[NV30_STATE_FRAGPROG]);
 	if (state->hw[NV30_STATE_VTXBUF] /*&& nv30->render_mode == HW*/)
-		so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_VTXBUF]);
+		so_emit_reloc_markers(chan, state->hw[NV30_STATE_VTXBUF]);
 }
 
 boolean
diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c
index fdc1cade90..44b6a74715 100644
--- a/src/gallium/drivers/nv30/nv30_state_fb.c
+++ b/src/gallium/drivers/nv30/nv30_state_fb.c
@@ -5,6 +5,8 @@ static boolean
 nv30_state_framebuffer_validate(struct nv30_context *nv30)
 {
 	struct pipe_framebuffer_state *fb = &nv30->framebuffer;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
 	struct nv04_surface *rt[2], *zeta = NULL;
 	uint32_t rt_enable, rt_format;
 	int i, colour_format = 0, zeta_format = 0;
@@ -63,6 +65,7 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30)
 		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
 		break;
 	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z24X8_UNORM:
 	case 0:
 		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
 		break;
@@ -79,56 +82,53 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30)
 		}
 
 		nv30mt = (struct nv30_miptree *)rt[0]->base.texture;
-		so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR0, 1);
-		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
-			  nv30->nvws->channel->vram->handle,
-			  nv30->nvws->channel->gart->handle);
-		so_method(so, nv30->screen->rankine, NV34TCL_COLOR0_PITCH, 2);
+		so_method(so, rankine, NV34TCL_DMA_COLOR0, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, rankine, NV34TCL_COLOR0_PITCH, 2);
 		so_data  (so, pitch);
-		so_reloc (so, nv30mt->buffer, rt[0]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), rt[0]->base.offset,
+			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
 	}
 
 	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
 		nv30mt = (struct nv30_miptree *)rt[1]->base.texture;
-		so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR1, 1);
-		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
-			  nv30->nvws->channel->vram->handle,
-			  nv30->nvws->channel->gart->handle);
-		so_method(so, nv30->screen->rankine, NV34TCL_COLOR1_OFFSET, 2);
-		so_reloc (so, nv30mt->buffer, rt[1]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, rankine, NV34TCL_DMA_COLOR1, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, rankine, NV34TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), rt[1]->base.offset,
+			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
 		so_data  (so, rt[1]->pitch);
 	}
 
 	if (zeta_format) {
 		nv30mt = (struct nv30_miptree *)zeta->base.texture;
-		so_method(so, nv30->screen->rankine, NV34TCL_DMA_ZETA, 1);
-		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
-			  nv30->nvws->channel->vram->handle,
-			  nv30->nvws->channel->gart->handle);
-		so_method(so, nv30->screen->rankine, NV34TCL_ZETA_OFFSET, 1);
-		so_reloc (so, nv30mt->buffer, zeta->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, rankine, NV34TCL_DMA_ZETA, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, rankine, NV34TCL_ZETA_OFFSET, 1);
+		so_reloc (so, nouveau_bo(nv30mt->buffer), zeta->base.offset,
+			      rt_flags | NOUVEAU_BO_LOW, 0, 0);
 		/* TODO: allocate LMA depth buffer */
 	}
 
-	so_method(so, nv30->screen->rankine, NV34TCL_RT_ENABLE, 1);
+	so_method(so, rankine, NV34TCL_RT_ENABLE, 1);
 	so_data  (so, rt_enable);
-	so_method(so, nv30->screen->rankine, NV34TCL_RT_HORIZ, 3);
+	so_method(so, rankine, NV34TCL_RT_HORIZ, 3);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
 	so_data  (so, rt_format);
-	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_HORIZ, 2);
+	so_method(so, rankine, NV34TCL_VIEWPORT_HORIZ, 2);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
-	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_method(so, rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
 	so_data  (so, ((w - 1) << 16) | 0);
 	so_data  (so, ((h - 1) << 16) | 0);
-	so_method(so, nv30->screen->rankine, 0x1d88, 1);
+	so_method(so, rankine, 0x1d88, 1);
 	so_data  (so, (1 << 12) | h);
 	/* Wonder why this is needed, context should all be set to zero on init */
-	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
+	so_method(so, rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
 	so_data  (so, 0);
 
 	so_ref(so, &nv30->state.hw[NV30_STATE_FB]);
diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c
index 990a876382..189656ec81 100644
--- a/src/gallium/drivers/nv30/nv30_vbo.c
+++ b/src/gallium/drivers/nv30/nv30_vbo.c
@@ -1,5 +1,6 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv30_context.h"
 #include "nv30_state.h"
@@ -70,7 +71,7 @@ static boolean
 nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib,
 		    unsigned ib_size)
 {
-	struct pipe_screen *pscreen = &nv30->screen->pipe;
+	struct pipe_screen *pscreen = &nv30->screen->base.base;
 	unsigned type;
 
 	if (!ib) {
@@ -108,7 +109,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 		       int attrib, struct pipe_vertex_element *ve,
 		       struct pipe_vertex_buffer *vb)
 {
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	struct nouveau_grobj *rankine = nv30->screen->rankine;
 	unsigned type, ncomp;
 	void *map;
@@ -116,7 +117,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 	if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp))
 		return FALSE;
 
-	map  = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+	map  = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
 	map += vb->buffer_offset + ve->src_offset;
 
 	switch (type) {
@@ -148,18 +149,17 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
 			so_data  (so, fui(v[0]));
 			break;
 		default:
-			ws->buffer_unmap(ws, vb->buffer);
+			pipe_buffer_unmap(pscreen, vb->buffer);
 			return FALSE;
 		}
 	}
 		break;
 	default:
-		ws->buffer_unmap(ws, vb->buffer);
+		pipe_buffer_unmap(pscreen, vb->buffer);
 		return FALSE;
 	}
 
-	ws->buffer_unmap(ws, vb->buffer);
-
+	pipe_buffer_unmap(pscreen, vb->buffer);
 	return TRUE;
 }
 
@@ -168,7 +168,7 @@ nv30_draw_arrays(struct pipe_context *pipe,
 		 unsigned mode, unsigned start, unsigned count)
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 	unsigned restart = 0;
 
 	nv30_vbo_set_idxbuf(nv30, NULL, 0);
@@ -228,7 +228,7 @@ static INLINE void
 nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 
 	while (count) {
 		uint8_t *elts = (uint8_t *)ib + start;
@@ -277,7 +277,7 @@ static INLINE void
 nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 
 	while (count) {
 		uint16_t *elts = (uint16_t *)ib + start;
@@ -326,7 +326,7 @@ static INLINE void
 nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 
 	while (count) {
 		uint32_t *elts = (uint32_t *)ib + start;
@@ -368,10 +368,10 @@ nv30_draw_elements_inline(struct pipe_context *pipe,
 			  unsigned mode, unsigned start, unsigned count)
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 	void *map;
 
-	map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ);
+	map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ);
 	if (!ib) {
 		NOUVEAU_ERR("failed mapping ib\n");
 		return FALSE;
@@ -392,7 +392,7 @@ nv30_draw_elements_inline(struct pipe_context *pipe,
 		break;
 	}
 
-	ws->buffer_unmap(ws, ib);
+	pipe_buffer_unmap(pscreen, ib);
 	return TRUE;
 }
 
@@ -401,7 +401,7 @@ nv30_draw_elements_vbo(struct pipe_context *pipe,
 		       unsigned mode, unsigned start, unsigned count)
 {
 	struct nv30_context *nv30 = nv30_context(pipe);
-	struct nouveau_channel *chan = nv30->nvws->channel;
+	struct nouveau_channel *chan = nv30->screen->base.channel;
 	unsigned restart = 0;
 
 	while (count) {
@@ -521,18 +521,20 @@ nv30_vbo_validate(struct nv30_context *nv30)
 			return FALSE;
 		}
 
-		so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset,
-			 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-			 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
+		so_reloc(vtxbuf, nouveau_bo(vb->buffer), vb->buffer_offset +
+				 ve->src_offset, vb_flags | NOUVEAU_BO_LOW |
+				 NOUVEAU_BO_OR, 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
 		so_data (vtxfmt, ((vb->stride << NV34TCL_VTXFMT_STRIDE_SHIFT) |
 				  (ncomp << NV34TCL_VTXFMT_SIZE_SHIFT) | type));
 	}
 
 	if (ib) {
+		struct nouveau_bo *bo = nouveau_bo(ib);
+
 		so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2);
-		so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR,
-			  0, NV34TCL_IDXBUF_FORMAT_DMA1);
+		so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
+				  0, NV34TCL_IDXBUF_FORMAT_DMA1);
 	}
 
 	so_method(vtxbuf, rankine, 0x1710, 1);
diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c
index eaf543b8f7..c7514efcfe 100644
--- a/src/gallium/drivers/nv30/nv30_vertprog.c
+++ b/src/gallium/drivers/nv30/nv30_vertprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -645,8 +646,7 @@ out_err:
 static boolean
 nv30_vertprog_validate(struct nv30_context *nv30)
 { 
-	struct nouveau_winsys *nvws = nv30->nvws;
-	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct pipe_screen *pscreen = nv30->pipe.screen;
 	struct nouveau_grobj *rankine = nv30->screen->rankine;
 	struct nv30_vertex_program *vp;
 	struct pipe_buffer *constbuf;
@@ -669,15 +669,15 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 		struct nouveau_stateobj *so;
 		uint vplen = vp->nr_insns;
 
-		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+		if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
 			while (heap->next && heap->size < vplen) {
 				struct nv30_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->exec);
+				nouveau_resource_free(&evict->exec);
 			}
 
-			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+			if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec))
 				assert(0);
 		}
 
@@ -694,15 +694,16 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 	if (vp->nr_consts && !vp->data) {
 		struct nouveau_resource *heap = nv30->screen->vp_data_heap;
 
-		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+		if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
 			while (heap->next && heap->size < vp->nr_consts) {
 				struct nv30_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->data);
+				nouveau_resource_free(&evict->data);
 			}
 
-			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+			if (nouveau_resource_alloc(heap, vp->nr_consts, vp,
+						   &vp->data))
 				assert(0);
 		}
 
@@ -750,8 +751,8 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 		float *map = NULL;
 
 		if (constbuf) {
-			map = ws->buffer_map(ws, constbuf,
-					     PIPE_BUFFER_USAGE_CPU_READ);
+			map = pipe_buffer_map(pscreen, constbuf,
+					      PIPE_BUFFER_USAGE_CPU_READ);
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
@@ -771,9 +772,8 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 			OUT_RINGp ((uint32_t *)vpd->value, 4);
 		}
 
-		if (constbuf) {
-			ws->buffer_unmap(ws, constbuf);
-		}
+		if (constbuf)
+			pipe_buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
@@ -804,8 +804,6 @@ nv30_vertprog_validate(struct nv30_context *nv30)
 void
 nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp)
 {
-	struct nouveau_winsys *nvws = nv30->screen->nvws;
-
 	vp->translated = FALSE;
 
 	if (vp->nr_insns) {
@@ -820,9 +818,9 @@ nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp)
 		vp->nr_consts = 0;
 	}
 
-	nvws->res_free(&vp->exec);
+	nouveau_resource_free(&vp->exec);
 	vp->exec_start = 0;
-	nvws->res_free(&vp->data);
+	nouveau_resource_free(&vp->data);
 	vp->data_start = 0;
 	vp->data_start_min = 0;
 
diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c
index c83ff91d7e..b2f19ecb69 100644
--- a/src/gallium/drivers/nv40/nv40_draw.c
+++ b/src/gallium/drivers/nv40/nv40_draw.c
@@ -1,4 +1,5 @@
 #include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
 
 #include "util/u_pack_color.h"
 
@@ -81,7 +82,7 @@ nv40_render_prim(struct draw_stage *stage, struct prim_header *prim,
 {
 	struct nv40_render_stage *rs = nv40_render_stage(stage);
 	struct nv40_context *nv40 = rs->nv40;
-	struct nouveau_pushbuf *pb = nv40->nvws->channel->pushbuf;
+	struct nouveau_pushbuf *pb = nv40->screen->base.channel->pushbuf;
 	unsigned i;
 
 	/* Ensure there's room for 4xfloat32 + potentially 3 begin/end */
@@ -231,7 +232,7 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
 			 unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 	unsigned i;
 	void *map;
 
@@ -241,13 +242,14 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
 	nv40_state_emit(nv40);
 
 	for (i = 0; i < nv40->vtxbuf_nr; i++) {
-		map = ws->buffer_map(ws, nv40->vtxbuf[i].buffer,
+		map = pipe_buffer_map(pscreen, nv40->vtxbuf[i].buffer,
                                       PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_vertex_buffer(nv40->draw, i, map);
 	}
 
 	if (idxbuf) {
-		map = ws->buffer_map(ws, idxbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen, idxbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_element_buffer(nv40->draw, idxbuf_size, map);
 	} else {
 		draw_set_mapped_element_buffer(nv40->draw, 0, NULL);
@@ -256,21 +258,22 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
 	if (nv40->constbuf[PIPE_SHADER_VERTEX]) {
 		const unsigned nr = nv40->constbuf_nr[PIPE_SHADER_VERTEX];
 
-		map = ws->buffer_map(ws, nv40->constbuf[PIPE_SHADER_VERTEX],
-				     PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen,
+				      nv40->constbuf[PIPE_SHADER_VERTEX],
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		draw_set_mapped_constant_buffer(nv40->draw, map, nr);
 	}
 
 	draw_arrays(nv40->draw, mode, start, count);
 
 	for (i = 0; i < nv40->vtxbuf_nr; i++)
-		ws->buffer_unmap(ws, nv40->vtxbuf[i].buffer);
+		pipe_buffer_unmap(pscreen, nv40->vtxbuf[i].buffer);
 
 	if (idxbuf)
-		ws->buffer_unmap(ws, idxbuf);
+		pipe_buffer_unmap(pscreen, idxbuf);
 
 	if (nv40->constbuf[PIPE_SHADER_VERTEX])
-		ws->buffer_unmap(ws, nv40->constbuf[PIPE_SHADER_VERTEX]);
+		pipe_buffer_unmap(pscreen, nv40->constbuf[PIPE_SHADER_VERTEX]);
 
 	draw_flush(nv40->draw);
 	pipe->flush(pipe, 0, NULL);
diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c
index 16e40889ec..680976da56 100644
--- a/src/gallium/drivers/nv40/nv40_fragprog.c
+++ b/src/gallium/drivers/nv40/nv40_fragprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -881,12 +882,12 @@ static void
 nv40_fragprog_upload(struct nv40_context *nv40,
 		     struct nv40_fragment_program *fp)
 {
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	const uint32_t le = 1;
 	uint32_t *map;
 	int i;
 
-	map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+	map = pipe_buffer_map(pscreen, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
 
 #if 0
 	for (i = 0; i < fp->insn_len; i++) {
@@ -908,7 +909,7 @@ nv40_fragprog_upload(struct nv40_context *nv40,
 		}
 	}
 
-	ws->buffer_unmap(ws, fp->buffer);
+	pipe_buffer_unmap(pscreen, fp->buffer);
 }
 
 static boolean
@@ -917,8 +918,7 @@ nv40_fragprog_validate(struct nv40_context *nv40)
 	struct nv40_fragment_program *fp = nv40->fragprog;
 	struct pipe_buffer *constbuf =
 		nv40->constbuf[PIPE_SHADER_FRAGMENT];
-	struct pipe_screen *screen = nv40->pipe.screen;
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	struct nouveau_stateobj *so;
 	boolean new_consts = FALSE;
 	int i;
@@ -933,14 +933,15 @@ nv40_fragprog_validate(struct nv40_context *nv40)
 		return FALSE;
 	}
 
-	fp->buffer = screen->buffer_create(screen, 0x100, 0, fp->insn_len * 4);
+	fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4);
 	nv40_fragprog_upload(nv40, fp);
 
 	so = so_new(4, 1);
 	so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1);
-	so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
-		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-		  NV40TCL_FP_ADDRESS_DMA0, NV40TCL_FP_ADDRESS_DMA1);
+	so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM |
+		      NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
+		      NOUVEAU_BO_OR, NV40TCL_FP_ADDRESS_DMA0,
+		      NV40TCL_FP_ADDRESS_DMA1);
 	so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1);
 	so_data  (so, fp->fp_control);
 	so_ref(so, &fp->so);
@@ -950,7 +951,8 @@ update_constants:
 	if (fp->nr_consts) {
 		float *map;
 		
-		map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		map = pipe_buffer_map(pscreen, constbuf,
+				      PIPE_BUFFER_USAGE_CPU_READ);
 		for (i = 0; i < fp->nr_consts; i++) {
 			struct nv40_fragment_program_data *fpd = &fp->consts[i];
 			uint32_t *p = &fp->insn[fpd->offset];
@@ -961,7 +963,7 @@ update_constants:
 			memcpy(p, cb, 4 * sizeof(float));
 			new_consts = TRUE;
 		}
-		ws->buffer_unmap(ws, constbuf);
+		pipe_buffer_unmap(pscreen, constbuf);
 
 		if (new_consts)
 			nv40_fragprog_upload(nv40, fp);
diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c
index eb3002dc05..f6cdf31dfe 100644
--- a/src/gallium/drivers/nv40/nv40_fragtex.c
+++ b/src/gallium/drivers/nv40/nv40_fragtex.c
@@ -62,6 +62,7 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit)
 {
 	struct nv40_sampler_state *ps = nv40->tex_sampler[unit];
 	struct nv40_miptree *nv40mt = nv40->tex_miptree[unit];
+	struct nouveau_bo *bo = nouveau_bo(nv40mt->buffer);
 	struct pipe_texture *pt = &nv40mt->base;
 	struct nv40_texture_format *tf;
 	struct nouveau_stateobj *so;
@@ -108,9 +109,9 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit)
 
 	so = so_new(16, 2);
 	so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8);
-	so_reloc (so, nv40mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
-	so_reloc (so, nv40mt->buffer, txf, tex_flags | NOUVEAU_BO_OR,
-		  NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1);
+	so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR,
+		      NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1);
 	so_data  (so, ps->wrap);
 	so_data  (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en);
 	so_data  (so, txs);
diff --git a/src/gallium/drivers/nv40/nv40_query.c b/src/gallium/drivers/nv40/nv40_query.c
index 9b9a43f49d..7874aedd42 100644
--- a/src/gallium/drivers/nv40/nv40_query.c
+++ b/src/gallium/drivers/nv40/nv40_query.c
@@ -29,11 +29,10 @@ nv40_query_create(struct pipe_context *pipe, unsigned query_type)
 static void
 nv40_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
 {
-	struct nv40_context *nv40 = nv40_context(pipe);
 	struct nv40_query *q = nv40_query(pq);
 
 	if (q->object)
-		nv40->nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	FREE(q);
 }
 
@@ -54,9 +53,9 @@ nv40_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
 		pipe->get_query_result(pipe, pq, 1, &tmp);
 	}
 
-	if (nv40->nvws->res_alloc(nv40->screen->query_heap, 1, NULL, &q->object))
+	if (nouveau_resource_alloc(nv40->screen->query_heap, 1, NULL, &q->object))
 		assert(0);
-	nv40->nvws->notifier_reset(nv40->screen->query, q->object->start);
+	nouveau_notifier_reset(nv40->screen->query, q->object->start);
 
 	BEGIN_RING(curie, NV40TCL_QUERY_RESET, 1);
 	OUT_RING  (1);
@@ -84,27 +83,27 @@ nv40_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
 	struct nv40_query *q = nv40_query(pq);
-	struct nouveau_winsys *nvws = nv40->nvws;
 
 	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 
 	if (!q->ready) {
 		unsigned status;
 
-		status = nvws->notifier_status(nv40->screen->query,
-					       q->object->start);
+		status = nouveau_notifier_status(nv40->screen->query,
+						 q->object->start);
 		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
 			if (wait == FALSE)
 				return FALSE;
-			nvws->notifier_wait(nv40->screen->query, q->object->start,
-					    NV_NOTIFY_STATE_STATUS_COMPLETED,
-					    0);
+			nouveau_notifier_wait_status(nv40->screen->query,
+					      q->object->start,
+					      NV_NOTIFY_STATE_STATUS_COMPLETED,
+					      0);
 		}
 
-		q->result = nvws->notifier_retval(nv40->screen->query,
-						  q->object->start);
+		q->result = nouveau_notifier_return_val(nv40->screen->query,
+							q->object->start);
 		q->ready = TRUE;
-		nvws->res_free(&q->object);
+		nouveau_resource_free(&q->object);
 	}
 
 	*result = q->result;
diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c
index 0d4baefaea..5d2a4216c5 100644
--- a/src/gallium/drivers/nv40/nv40_screen.c
+++ b/src/gallium/drivers/nv40/nv40_screen.c
@@ -1,5 +1,4 @@
 #include "pipe/p_screen.h"
-#include "util/u_simple_screen.h"
 
 #include "nv40_context.h"
 #include "nv40_screen.h"
@@ -8,23 +7,6 @@
 #define NV4X_GRCLASS4497_CHIPSETS 0x00005450
 #define NV6X_GRCLASS4497_CHIPSETS 0x00000088
 
-static const char *
-nv40_screen_get_name(struct pipe_screen *pscreen)
-{
-	struct nv40_screen *screen = nv40_screen(pscreen);
-	struct nouveau_device *dev = screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv40_screen_get_vendor(struct pipe_screen *pscreen)
-{
-	return "nouveau";
-}
-
 static int
 nv40_screen_get_param(struct pipe_screen *pscreen, int param)
 {
@@ -62,6 +44,10 @@ nv40_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0; /* We have 4 - but unsupported currently */
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 1;
 	case NOUVEAU_CAP_HW_VTXBUF:
 		return 1;
 	case NOUVEAU_CAP_HW_IDXBUF:
@@ -104,7 +90,15 @@ nv40_screen_surface_format_supported(struct pipe_screen *pscreen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM: 
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -148,88 +142,107 @@ static void
 nv40_screen_destroy(struct pipe_screen *pscreen)
 {
 	struct nv40_screen *screen = nv40_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
 
-	nvws->res_free(&screen->vp_exec_heap);
-	nvws->res_free(&screen->vp_data_heap);
-	nvws->res_free(&screen->query_heap);
-	nvws->notifier_free(&screen->query);
-	nvws->notifier_free(&screen->sync);
-	nvws->grobj_free(&screen->curie);
+	nouveau_resource_free(&screen->vp_exec_heap);
+	nouveau_resource_free(&screen->vp_data_heap);
+	nouveau_resource_free(&screen->query_heap);
+	nouveau_notifier_free(&screen->query);
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->curie);
+
+	nouveau_screen_fini(&screen->base);
 
 	FREE(pscreen);
 }
 
 struct pipe_screen *
-nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
 	unsigned curie_class = 0;
-	unsigned chipset = nvws->channel->device->chipset;
 	int ret;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
 
-	/* 2D engine setup */
-	screen->eng2d = nv04_surface_2d_init(nvws);
-	screen->eng2d->buf = nv40_surface_buffer;
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv40_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv40_screen_destroy;
+	pscreen->get_param = nv40_screen_get_param;
+	pscreen->get_paramf = nv40_screen_get_paramf;
+	pscreen->is_format_supported = nv40_screen_surface_format_supported;
+
+	nv40_screen_init_miptree_functions(pscreen);
+	nv40_screen_init_transfer_functions(pscreen);
 
 	/* 3D object */
-	switch (chipset & 0xf0) {
+	switch (dev->chipset & 0xf0) {
 	case 0x40:
-		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (chipset & 0x0f)))
+		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (dev->chipset & 0x0f)))
 			curie_class = NV40TCL;
 		else
-		if (NV4X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f)))
+		if (NV4X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
 			curie_class = NV44TCL;
 		break;
 	case 0x60:
-		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f)))
+		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (dev->chipset & 0x0f)))
 			curie_class = NV44TCL;
 		break;
 	}
 
 	if (!curie_class) {
-		NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", chipset);
+		NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", dev->chipset);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, curie_class, &screen->curie);
+	ret = nouveau_grobj_alloc(chan, 0xbeef3097, curie_class, &screen->curie);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
 		return FALSE;
 	}
+	BIND_RING(chan, screen->curie, 7);
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(&screen->base);
+	screen->eng2d->buf = nv40_surface_buffer;
 
 	/* Notifier for sync purposes */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv40_screen_destroy(&screen->pipe);
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Query objects */
-	ret = nvws->notifier_alloc(nvws, 32, &screen->query);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0302, 32, &screen->query);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
-		nv40_screen_destroy(&screen->pipe);
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	ret = nvws->res_init(&screen->query_heap, 0, 32);
+	nouveau_resource_init(&screen->query_heap, 0, 32);
 	if (ret) {
 		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
-		nv40_screen_destroy(&screen->pipe);
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	/* Vtxprog resources */
-	if (nvws->res_init(&screen->vp_exec_heap, 0, 512) ||
-	    nvws->res_init(&screen->vp_data_heap, 0, 256)) {
-		nv40_screen_destroy(&screen->pipe);
+	if (nouveau_resource_init(&screen->vp_exec_heap, 0, 512) ||
+	    nouveau_resource_init(&screen->vp_data_heap, 0, 256)) {
+		nv40_screen_destroy(pscreen);
 		return NULL;
 	}
 
@@ -238,25 +251,25 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1);
 	so_data  (so, screen->sync->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_COLOR1, 1);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_COLOR0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_VTXBUF0, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->gart->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->gart->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_FENCE, 2);
 	so_data  (so, 0);
 	so_data  (so, screen->query->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_UNK01AC, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->curie, NV40TCL_DMA_COLOR2, 2);
-	so_data  (so, nvws->channel->vram->handle);
-	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 
 	so_method(so, screen->curie, 0x1ea4, 3);
 	so_data  (so, 0x00000010);
@@ -281,24 +294,10 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->curie, 0x1e94, 1);
 	so_data  (so, 0x00000001);
 
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref(NULL, &so);
-	nvws->push_flush(nvws, 0, NULL);
-
-	screen->pipe.winsys = ws;
-	screen->pipe.destroy = nv40_screen_destroy;
-
-	screen->pipe.get_name = nv40_screen_get_name;
-	screen->pipe.get_vendor = nv40_screen_get_vendor;
-	screen->pipe.get_param = nv40_screen_get_param;
-	screen->pipe.get_paramf = nv40_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv40_screen_surface_format_supported;
-
-	nv40_screen_init_miptree_functions(&screen->pipe);
-	nv40_screen_init_transfer_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
+	nouveau_pushbuf_flush(chan, 0);
 
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h
index 7b503bd207..57b4c8fc46 100644
--- a/src/gallium/drivers/nv40/nv40_screen.h
+++ b/src/gallium/drivers/nv40/nv40_screen.h
@@ -1,11 +1,11 @@
 #ifndef __NV40_SCREEN_H__
 #define __NV40_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 #include "nv04/nv04_surface_2d.h"
 
 struct nv40_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c
index 10aae29832..198692965d 100644
--- a/src/gallium/drivers/nv40/nv40_state_emit.c
+++ b/src/gallium/drivers/nv40/nv40_state_emit.c
@@ -54,6 +54,7 @@ nv40_state_do_validate(struct nv40_context *nv40,
 void
 nv40_state_emit(struct nv40_context *nv40)
 {
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 	struct nv40_state *state = &nv40->state;
 	struct nv40_screen *screen = nv40->screen;
 	unsigned i, samplers;
@@ -73,7 +74,7 @@ nv40_state_emit(struct nv40_context *nv40)
 			continue;
 		so_ref (state->hw[i], &nv40->screen->state[i]);
 		if (state->hw[i])
-			so_emit(nv40->nvws, nv40->screen->state[i]);
+			so_emit(chan, nv40->screen->state[i]);
 		states &= ~(1ULL << i);
 	}
 
@@ -87,17 +88,17 @@ nv40_state_emit(struct nv40_context *nv40)
 
 	state->dirty = 0;
 
-	so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FB]);
+	so_emit_reloc_markers(chan, state->hw[NV40_STATE_FB]);
 	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
 		if (!(samplers & (1 << i)))
 			continue;
-		so_emit_reloc_markers(nv40->nvws,
+		so_emit_reloc_markers(chan,
 				      state->hw[NV40_STATE_FRAGTEX0+i]);
 		samplers &= ~(1ULL << i);
 	}
-	so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FRAGPROG]);
+	so_emit_reloc_markers(chan, state->hw[NV40_STATE_FRAGPROG]);
 	if (state->hw[NV40_STATE_VTXBUF] && nv40->render_mode == HW)
-		so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_VTXBUF]);
+		so_emit_reloc_markers(chan, state->hw[NV40_STATE_VTXBUF]);
 }
 
 boolean
diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c
index be618a306b..c2f739157a 100644
--- a/src/gallium/drivers/nv40/nv40_state_fb.c
+++ b/src/gallium/drivers/nv40/nv40_state_fb.c
@@ -2,15 +2,19 @@
 #include "nouveau/nouveau_util.h"
 
 static struct pipe_buffer *
-nv40_surface_buffer(struct pipe_surface *surface)
+nv40_do_surface_buffer(struct pipe_surface *surface)
 {
 	struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture;
 	return mt->buffer;
 }
 
+#define nv40_surface_buffer(ps) nouveau_bo(nv40_do_surface_buffer(ps))
+
 static boolean
 nv40_state_framebuffer_validate(struct nv40_context *nv40)
 {
+	struct nouveau_channel *chan = nv40->screen->base.channel;
+	struct nouveau_grobj *curie = nv40->screen->curie;
 	struct pipe_framebuffer_state *fb = &nv40->framebuffer;
 	struct nv04_surface *rt[4], *zeta;
 	uint32_t rt_enable, rt_format;
@@ -69,6 +73,7 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40)
 		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z16;
 		break;
 	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z24X8_UNORM:
 	case 0:
 		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z24S8;
 		break;
@@ -77,76 +82,80 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40)
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR0, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR0_PITCH, 2);
+		so_method(so, curie, NV40TCL_DMA_COLOR0, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR0_PITCH, 2);
 		so_data  (so, rt[0]->pitch);
-		so_reloc (so, nv40_surface_buffer(&rt[0]->base), rt[0]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (so, nv40_surface_buffer(&rt[0]->base),
+			      rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR1, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR1_OFFSET, 2);
-		so_reloc (so, nv40_surface_buffer(&rt[1]->base), rt[1]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, curie, NV40TCL_DMA_COLOR1, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nv40_surface_buffer(&rt[1]->base),
+			      rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
 		so_data  (so, rt[1]->pitch);
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR2, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR2_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[2]->base), rt[2]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR2_PITCH, 1);
+		so_method(so, curie, NV40TCL_DMA_COLOR2, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR2_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[2]->base),
+			      rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
+		so_method(so, curie, NV40TCL_COLOR2_PITCH, 1);
 		so_data  (so, rt[2]->pitch);
 	}
 
 	if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR3, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR3_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&rt[3]->base), rt[3]->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_COLOR3_PITCH, 1);
+		so_method(so, curie, NV40TCL_DMA_COLOR3, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_COLOR3_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(&rt[3]->base),
+			      rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+			      0, 0);
+		so_method(so, curie, NV40TCL_COLOR3_PITCH, 1);
 		so_data  (so, rt[3]->pitch);
 	}
 
 	if (zeta_format) {
-		so_method(so, nv40->screen->curie, NV40TCL_DMA_ZETA, 1);
-		so_reloc (so, nv40_surface_buffer(&zeta->base), 0, rt_flags | NOUVEAU_BO_OR,
-			  nv40->nvws->channel->vram->handle,
-			  nv40->nvws->channel->gart->handle);
-		so_method(so, nv40->screen->curie, NV40TCL_ZETA_OFFSET, 1);
-		so_reloc (so, nv40_surface_buffer(&zeta->base), zeta->base.offset, rt_flags |
-			  NOUVEAU_BO_LOW, 0, 0);
-		so_method(so, nv40->screen->curie, NV40TCL_ZETA_PITCH, 1);
+		so_method(so, curie, NV40TCL_DMA_ZETA, 1);
+		so_reloc (so, nv40_surface_buffer(&zeta->base), 0,
+			      rt_flags | NOUVEAU_BO_OR,
+			      chan->vram->handle, chan->gart->handle);
+		so_method(so, curie, NV40TCL_ZETA_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(&zeta->base),
+			      zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, curie, NV40TCL_ZETA_PITCH, 1);
 		so_data  (so, zeta->pitch);
 	}
 
-	so_method(so, nv40->screen->curie, NV40TCL_RT_ENABLE, 1);
+	so_method(so, curie, NV40TCL_RT_ENABLE, 1);
 	so_data  (so, rt_enable);
-	so_method(so, nv40->screen->curie, NV40TCL_RT_HORIZ, 3);
+	so_method(so, curie, NV40TCL_RT_HORIZ, 3);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
 	so_data  (so, rt_format);
-	so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_HORIZ, 2);
+	so_method(so, curie, NV40TCL_VIEWPORT_HORIZ, 2);
 	so_data  (so, (w << 16) | 0);
 	so_data  (so, (h << 16) | 0);
-	so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_method(so, curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2);
 	so_data  (so, ((w - 1) << 16) | 0);
 	so_data  (so, ((h - 1) << 16) | 0);
-	so_method(so, nv40->screen->curie, 0x1d88, 1);
+	so_method(so, curie, 0x1d88, 1);
 	so_data  (so, (1 << 12) | h);
 
 	so_ref(so, &nv40->state.hw[NV40_STATE_FB]);
diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c
index 1a849da32d..a596547974 100644
--- a/src/gallium/drivers/nv40/nv40_surface.c
+++ b/src/gallium/drivers/nv40/nv40_surface.c
@@ -26,12 +26,13 @@
  * 
  **************************************************************************/
 
-#include "nv40_context.h"
 #include "pipe/p_defines.h"
-#include "pipe/internal/p_winsys_screen.h"
 #include "pipe/p_inlines.h"
+
 #include "util/u_tile.h"
 
+#include "nv40_context.h"
+
 static void
 nv40_surface_copy(struct pipe_context *pipe,
 		  struct pipe_surface *dest, unsigned destx, unsigned desty,
diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c
index ce45055fe8..6d92ac3db9 100644
--- a/src/gallium/drivers/nv40/nv40_transfer.c
+++ b/src/gallium/drivers/nv40/nv40_transfer.c
@@ -136,7 +136,7 @@ nv40_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 }
 
 static void
-nv40_transfer_del(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
+nv40_transfer_del(struct pipe_transfer *ptx)
 {
 	struct nv40_transfer *tx = (struct nv40_transfer *)ptx;
 
diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c
index f3518b2e4f..b2753b8e2e 100644
--- a/src/gallium/drivers/nv40/nv40_vbo.c
+++ b/src/gallium/drivers/nv40/nv40_vbo.c
@@ -1,5 +1,6 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv40_context.h"
 #include "nv40_state.h"
@@ -70,7 +71,7 @@ static boolean
 nv40_vbo_set_idxbuf(struct nv40_context *nv40, struct pipe_buffer *ib,
 		    unsigned ib_size)
 {
-	struct pipe_screen *pscreen = &nv40->screen->pipe;
+	struct pipe_screen *pscreen = &nv40->screen->base.base;
 	unsigned type;
 
 	if (!ib) {
@@ -108,7 +109,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
 		       int attrib, struct pipe_vertex_element *ve,
 		       struct pipe_vertex_buffer *vb)
 {
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	struct nouveau_grobj *curie = nv40->screen->curie;
 	unsigned type, ncomp;
 	void *map;
@@ -116,7 +117,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
 	if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp))
 		return FALSE;
 
-	map  = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+	map  = pipe_buffer_map(pscreen, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
 	map += vb->buffer_offset + ve->src_offset;
 
 	switch (type) {
@@ -148,17 +149,17 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
 			so_data  (so, fui(v[0]));
 			break;
 		default:
-			ws->buffer_unmap(ws, vb->buffer);
+			pipe_buffer_unmap(pscreen, vb->buffer);
 			return FALSE;
 		}
 	}
 		break;
 	default:
-		ws->buffer_unmap(ws, vb->buffer);
+		pipe_buffer_unmap(pscreen, vb->buffer);
 		return FALSE;
 	}
 
-	ws->buffer_unmap(ws, vb->buffer);
+	pipe_buffer_unmap(pscreen, vb->buffer);
 
 	return TRUE;
 }
@@ -168,7 +169,7 @@ nv40_draw_arrays(struct pipe_context *pipe,
 		 unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 	unsigned restart;
 
 	nv40_vbo_set_idxbuf(nv40, NULL, 0);
@@ -227,7 +228,7 @@ static INLINE void
 nv40_draw_elements_u08(struct nv40_context *nv40, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 
 	while (count) {
 		uint8_t *elts = (uint8_t *)ib + start;
@@ -276,7 +277,7 @@ static INLINE void
 nv40_draw_elements_u16(struct nv40_context *nv40, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 
 	while (count) {
 		uint16_t *elts = (uint16_t *)ib + start;
@@ -325,7 +326,7 @@ static INLINE void
 nv40_draw_elements_u32(struct nv40_context *nv40, void *ib,
 		       unsigned mode, unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 
 	while (count) {
 		uint32_t *elts = (uint32_t *)ib + start;
@@ -367,10 +368,10 @@ nv40_draw_elements_inline(struct pipe_context *pipe,
 			  unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct pipe_winsys *ws = pipe->winsys;
+	struct pipe_screen *pscreen = pipe->screen;
 	void *map;
 
-	map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ);
+	map = pipe_buffer_map(pscreen, ib, PIPE_BUFFER_USAGE_CPU_READ);
 	if (!ib) {
 		NOUVEAU_ERR("failed mapping ib\n");
 		return FALSE;
@@ -391,7 +392,7 @@ nv40_draw_elements_inline(struct pipe_context *pipe,
 		break;
 	}
 
-	ws->buffer_unmap(ws, ib);
+	pipe_buffer_unmap(pscreen, ib);
 	return TRUE;
 }
 
@@ -400,7 +401,7 @@ nv40_draw_elements_vbo(struct pipe_context *pipe,
 		       unsigned mode, unsigned start, unsigned count)
 {
 	struct nv40_context *nv40 = nv40_context(pipe);
-	struct nouveau_channel *chan = nv40->nvws->channel;
+	struct nouveau_channel *chan = nv40->screen->base.channel;
 	unsigned restart;
 
 	while (count) {
@@ -519,17 +520,20 @@ nv40_vbo_validate(struct nv40_context *nv40)
 			return FALSE;
 		}
 
-		so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset,
-			 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
-			 0, NV40TCL_VTXBUF_ADDRESS_DMA1);
+		so_reloc(vtxbuf, nouveau_bo(vb->buffer),
+				 vb->buffer_offset + ve->src_offset,
+				 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+				 0, NV40TCL_VTXBUF_ADDRESS_DMA1);
 		so_data (vtxfmt, ((vb->stride << NV40TCL_VTXFMT_STRIDE_SHIFT) |
 				  (ncomp << NV40TCL_VTXFMT_SIZE_SHIFT) | type));
 	}
 
 	if (ib) {
+		struct nouveau_bo *bo = nouveau_bo(ib);
+
 		so_method(vtxbuf, curie, NV40TCL_IDXBUF_ADDRESS, 2);
-		so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
-		so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR,
+		so_reloc (vtxbuf, bo, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (vtxbuf, bo, ib_format, vb_flags | NOUVEAU_BO_OR,
 			  0, NV40TCL_IDXBUF_FORMAT_DMA1);
 	}
 
diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c
index 7df9a4d326..e75e8d3f42 100644
--- a/src/gallium/drivers/nv40/nv40_vertprog.c
+++ b/src/gallium/drivers/nv40/nv40_vertprog.c
@@ -1,6 +1,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "pipe/p_shader_tokens.h"
 #include "tgsi/tgsi_parse.h"
@@ -855,8 +856,7 @@ out_err:
 static boolean
 nv40_vertprog_validate(struct nv40_context *nv40)
 { 
-	struct nouveau_winsys *nvws = nv40->nvws;
-	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct pipe_screen *pscreen = nv40->pipe.screen;
 	struct nouveau_grobj *curie = nv40->screen->curie;
 	struct nv40_vertex_program *vp;
 	struct pipe_buffer *constbuf;
@@ -895,15 +895,15 @@ check_gpu_resources:
 		struct nouveau_stateobj *so;
 		uint vplen = vp->nr_insns;
 
-		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+		if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
 			while (heap->next && heap->size < vplen) {
 				struct nv40_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->exec);
+				nouveau_resource_free(&evict->exec);
 			}
 
-			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+			if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec))
 				assert(0);
 		}
 
@@ -925,15 +925,15 @@ check_gpu_resources:
 	if (vp->nr_consts && !vp->data) {
 		struct nouveau_resource *heap = nv40->screen->vp_data_heap;
 
-		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+		if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
 			while (heap->next && heap->size < vp->nr_consts) {
 				struct nv40_vertex_program *evict;
 				
 				evict = heap->next->priv;
-				nvws->res_free(&evict->data);
+				nouveau_resource_free(&evict->data);
 			}
 
-			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+			if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data))
 				assert(0);
 		}
 
@@ -981,8 +981,8 @@ check_gpu_resources:
 		float *map = NULL;
 
 		if (constbuf) {
-			map = ws->buffer_map(ws, constbuf,
-					     PIPE_BUFFER_USAGE_CPU_READ);
+			map = pipe_buffer_map(pscreen, constbuf,
+					      PIPE_BUFFER_USAGE_CPU_READ);
 		}
 
 		for (i = 0; i < vp->nr_consts; i++) {
@@ -1003,7 +1003,7 @@ check_gpu_resources:
 		}
 
 		if (constbuf)
-			ws->buffer_unmap(ws, constbuf);
+			pscreen->buffer_unmap(pscreen, constbuf);
 	}
 
 	/* Upload vtxprog */
@@ -1035,8 +1035,6 @@ check_gpu_resources:
 void
 nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
 {
-	struct nouveau_winsys *nvws = nv40->screen->nvws;
-
 	vp->translated = FALSE;
 
 	if (vp->nr_insns) {
@@ -1051,9 +1049,9 @@ nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
 		vp->nr_consts = 0;
 	}
 
-	nvws->res_free(&vp->exec);
+	nouveau_resource_free(&vp->exec);
 	vp->exec_start = 0;
-	nvws->res_free(&vp->data);
+	nouveau_resource_free(&vp->data);
 	vp->data_start = 0;
 	vp->data_start_min = 0;
 
diff --git a/src/gallium/drivers/nv50/nv50_clear.c b/src/gallium/drivers/nv50/nv50_clear.c
index 33427a15a5..e0b2d2880b 100644
--- a/src/gallium/drivers/nv50/nv50_clear.c
+++ b/src/gallium/drivers/nv50/nv50_clear.c
@@ -31,7 +31,7 @@ nv50_clear(struct pipe_context *pipe, unsigned buffers,
 	   const float *rgba, double depth, unsigned stencil)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct pipe_framebuffer_state *fb = &nv50->framebuffer;
 	unsigned mode = 0, i;
diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c
index a511f655c1..e02afc4be9 100644
--- a/src/gallium/drivers/nv50/nv50_context.c
+++ b/src/gallium/drivers/nv50/nv50_context.c
@@ -33,7 +33,7 @@ nv50_flush(struct pipe_context *pipe, unsigned flags,
 {
 	struct nv50_context *nv50 = (struct nv50_context *)pipe;
 	
-	FIRE_RING(nv50->screen->nvws->channel);
+	FIRE_RING(nv50->screen->base.channel);
 }
 
 static void
diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
index 7b67a75439..9b8cc4d37d 100644
--- a/src/gallium/drivers/nv50/nv50_context.h
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -63,6 +63,11 @@ struct nv50_rasterizer_stateobj {
 	struct nouveau_stateobj *so;
 };
 
+struct nv50_sampler_stateobj {
+	bool normalized;
+	unsigned tsc[8];
+};
+
 struct nv50_miptree_level {
 	int *image_offset;
 	unsigned pitch;
@@ -70,7 +75,8 @@ struct nv50_miptree_level {
 
 struct nv50_miptree {
 	struct pipe_texture base;
-	struct pipe_buffer *buffer;
+
+	struct nouveau_bo *bo;
 
 	struct nv50_miptree_level level[PIPE_MAX_TEXTURE_LEVELS];
 	int image_nr;
@@ -93,13 +99,6 @@ nv50_surface(struct pipe_surface *pt)
 	return (struct nv50_surface *)pt;
 }
 
-static INLINE struct pipe_buffer *
-nv50_surface_buffer(struct pipe_surface *surface)
-{
-	struct nv50_miptree *mt = (struct nv50_miptree *)surface->texture;
-	return mt->buffer;
-}
-
 struct nv50_state {
 	unsigned dirty;
 
@@ -115,6 +114,7 @@ struct nv50_state {
 	unsigned viewport_bypass;
 	struct nouveau_stateobj *tsc_upload;
 	struct nouveau_stateobj *tic_upload;
+	unsigned miptree_nr;
 	struct nouveau_stateobj *vertprog;
 	struct nouveau_stateobj *fragprog;
 	struct nouveau_stateobj *vtxfmt;
@@ -147,7 +147,7 @@ struct nv50_context {
 	unsigned vtxbuf_nr;
 	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
 	unsigned vtxelt_nr;
-	unsigned *sampler[PIPE_MAX_SAMPLERS];
+	struct nv50_sampler_stateobj *sampler[PIPE_MAX_SAMPLERS];
 	unsigned sampler_nr;
 	struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS];
 	unsigned miptree_nr;
diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c
index f79a7ca86c..22465e0227 100644
--- a/src/gallium/drivers/nv50/nv50_miptree.c
+++ b/src/gallium/drivers/nv50/nv50_miptree.c
@@ -29,26 +29,36 @@
 static struct pipe_texture *
 nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 {
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
 	struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree);
 	struct pipe_texture *pt = &mt->base;
-	unsigned usage, width = tmp->width[0], height = tmp->height[0];
+	unsigned width = tmp->width[0], height = tmp->height[0];
 	unsigned depth = tmp->depth[0];
-	int i, l;
+	uint32_t tile_mode, tile_flags, tile_h;
+	int ret, i, l;
 
 	mt->base = *tmp;
 	pipe_reference_init(&mt->base.reference, 1);
 	mt->base.screen = pscreen;
 
-	usage = PIPE_BUFFER_USAGE_PIXEL;
 	switch (pt->format) {
+	case PIPE_FORMAT_Z24X8_UNORM:
 	case PIPE_FORMAT_Z24S8_UNORM:
 	case PIPE_FORMAT_Z16_UNORM:
-		usage |= NOUVEAU_BUFFER_USAGE_ZETA;
+		tile_flags = 0x2800;
 		break;
 	default:
+		tile_flags = 0x7000;
 		break;
 	}
 
+	if      (pt->height[0] > 32) tile_mode = 4;
+	else if (pt->height[0] > 16) tile_mode = 3;
+	else if (pt->height[0] >  8) tile_mode = 2;
+	else if (pt->height[0] >  4) tile_mode = 1;
+	else                         tile_mode = 0;
+	tile_h = 1 << (tile_mode + 2);
+
 	switch (pt->target) {
 	case PIPE_TEXTURE_3D:
 		mt->image_nr = pt->depth[0];
@@ -85,7 +95,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 
 			size  = align(pt->width[l], 8) * pt->block.size;
 			size  = align(size, 64);
-			size *= align(pt->height[l], 8) * pt->block.size;
+			size *= align(pt->height[l], tile_h) * pt->block.size;
 
 			lvl->image_offset[i] = mt->total_size;
 
@@ -93,12 +103,13 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 		}
 	}
 
-	mt->buffer = pscreen->buffer_create(pscreen, 256, usage, mt->total_size);
-	if (!mt->buffer) {
+	ret = nouveau_bo_new_tile(dev, NOUVEAU_BO_VRAM, 256, mt->total_size,
+				  tile_mode, tile_flags, &mt->bo);
+	if (ret) {
 		FREE(mt);
 		return NULL;
 	}
-
+			     
 	return &mt->base;
 }
 
@@ -106,6 +117,7 @@ static struct pipe_texture *
 nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
 		     const unsigned *stride, struct pipe_buffer *pb)
 {
+	struct nouveau_bo *bo = nouveau_bo(pb);
 	struct nv50_miptree *mt;
 
 	/* Only supports 2D, non-mipmapped textures for the moment */
@@ -124,7 +136,7 @@ nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
 	mt->level[0].pitch = *stride;
 	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
 
-	pipe_buffer_reference(&mt->buffer, pb);
+	nouveau_bo_ref(bo, &mt->bo);
 	return &mt->base;
 }
 
@@ -133,7 +145,7 @@ nv50_miptree_destroy(struct pipe_texture *pt)
 {
 	struct nv50_miptree *mt = nv50_miptree(pt);
 
-        pipe_buffer_reference(&mt->buffer, NULL);
+	nouveau_bo_ref(NULL, &mt->bo);
         FREE(mt);
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
index 2d15868ae8..5f7d06dbec 100644
--- a/src/gallium/drivers/nv50/nv50_program.c
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -85,6 +85,9 @@ struct nv50_reg {
 
 	int hw;
 	int neg;
+
+	int rhw; /* result hw for FP outputs, or interpolant index */
+	int acc; /* instruction where this reg is last read (first insn == 1) */
 };
 
 struct nv50_pc {
@@ -108,12 +111,23 @@ struct nv50_pc {
 
 	struct nv50_reg *temp_temp[16];
 	unsigned temp_temp_nr;
+
+	unsigned interp_mode[32];
+	/* perspective interpolation registers */
+	struct nv50_reg *iv_p;
+	struct nv50_reg *iv_c;
+
+	/* current instruction and total number of insns */
+	unsigned insn_cur;
+	unsigned insn_nr;
+
+	boolean allow32;
 };
 
 static void
 alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg)
 {
-	int i;
+	int i = 0;
 
 	if (reg->type == P_RESULT) {
 		if (pc->p->cfg.high_result < (reg->hw + 1))
@@ -131,7 +145,22 @@ alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg)
 		return;
 	}
 
-	for (i = 0; i < NV50_SU_MAX_TEMP; i++) {
+	if (reg->rhw != -1) {
+		/* try to allocate temporary with index rhw first */
+		if (!(pc->r_temp[reg->rhw])) {
+			pc->r_temp[reg->rhw] = reg;
+			reg->hw = reg->rhw;
+			if (pc->p->cfg.high_temp < (reg->rhw + 1))
+				pc->p->cfg.high_temp = reg->rhw + 1;
+			return;
+		}
+		/* make sure we don't get things like $r0 needs to go
+		 * in $r1 and $r1 in $r0
+		 */
+		i = pc->result_nr * 4;
+	}
+
+	for (; i < NV50_SU_MAX_TEMP; i++) {
 		if (!(pc->r_temp[i])) {
 			pc->r_temp[i] = reg;
 			reg->hw = i;
@@ -159,6 +188,7 @@ alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst)
 			r->type = P_TEMP;
 			r->index = -1;
 			r->hw = i;
+			r->rhw = -1;
 			pc->r_temp[i] = r;
 			return r;
 		}
@@ -168,6 +198,38 @@ alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst)
 	return NULL;
 }
 
+/* Assign the hw of the discarded temporary register src
+ * to the tgsi register dst and free src.
+ */
+static void
+assimilate_temp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	assert(src->index == -1 && src->hw != -1);
+
+	if (dst->hw != -1)
+		pc->r_temp[dst->hw] = NULL;
+	pc->r_temp[src->hw] = dst;
+	dst->hw = src->hw;
+
+	FREE(src);
+}
+
+/* release the hardware resource held by r */
+static void
+release_hw(struct nv50_pc *pc, struct nv50_reg *r)
+{
+	assert(r->type == P_TEMP);
+	if (r->hw == -1)
+		return;
+
+	assert(pc->r_temp[r->hw] == r);
+	pc->r_temp[r->hw] = NULL;
+
+	r->acc = 0;
+	if (r->index == -1)
+		FREE(r);
+}
+
 static void
 free_temp(struct nv50_pc *pc, struct nv50_reg *r)
 {
@@ -250,7 +312,13 @@ alloc_immd(struct nv50_pc *pc, float f)
 	struct nv50_reg *r = CALLOC_STRUCT(nv50_reg);
 	unsigned hw;
 
-	hw = ctor_immd(pc, f, 0, 0, 0) * 4;
+	for (hw = 0; hw < pc->immd_nr * 4; hw++)
+		if (pc->immd_buf[hw] == f)
+			break;
+
+	if (hw == pc->immd_nr * 4)
+		hw = ctor_immd(pc, f, -f, 0.5 * f, 0) * 4;
+
 	r->type = P_IMMD;
 	r->hw = hw;
 	r->index = -1;
@@ -341,7 +409,8 @@ set_dst(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_program_exec *e)
 static INLINE void
 set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e)
 {
-	unsigned val = fui(pc->immd_buf[imm->hw]); /* XXX */
+	float f = pc->immd_buf[imm->hw];
+	unsigned val = fui(imm->neg ? -f : f);
 
 	set_long(pc, e);
 	/*XXX: can't be predicated - bits overlap.. catch cases where both
@@ -354,20 +423,35 @@ set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e)
 	e->inst[1] |= (val >> 6) << 2;
 }
 
+
+#define INTERP_LINEAR		0
+#define INTERP_FLAT			1
+#define INTERP_PERSPECTIVE	2
+#define INTERP_CENTROID		4
+
+/* interpolant index has been stored in dst->rhw */
 static void
-emit_interp(struct nv50_pc *pc, struct nv50_reg *dst,
-	    struct nv50_reg *src, struct nv50_reg *iv)
+emit_interp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *iv,
+		unsigned mode)
 {
+	assert(dst->rhw != -1);
 	struct nv50_program_exec *e = exec(pc);
 
 	e->inst[0] |= 0x80000000;
 	set_dst(pc, dst, e);
-	alloc_reg(pc, src);
-	e->inst[0] |= (src->hw << 16);
-	if (iv) {
-		e->inst[0] |= (1 << 25);
-		alloc_reg(pc, iv);
-		e->inst[0] |= (iv->hw << 9);
+	e->inst[0] |= (dst->rhw << 16);
+
+	if (mode & INTERP_FLAT) {
+		e->inst[0] |= (1 << 8);
+	} else {
+		if (mode & INTERP_PERSPECTIVE) {
+			e->inst[0] |= (1 << 25);
+			alloc_reg(pc, iv);
+			e->inst[0] |= (iv->hw << 9);
+		}
+
+		if (mode & INTERP_CENTROID)
+			e->inst[0] |= (1 << 24);
 	}
 
 	emit(pc, e);
@@ -378,22 +462,12 @@ set_data(struct nv50_pc *pc, struct nv50_reg *src, unsigned m, unsigned s,
 	 struct nv50_program_exec *e)
 {
 	set_long(pc, e);
-#if 1
-	e->inst[1] |= (1 << 22);
-#else
-	if (src->type == P_IMMD) {
-		e->inst[1] |= (NV50_CB_PMISC << 22);
-	} else {
-		if (pc->p->type == PIPE_SHADER_VERTEX)
-			e->inst[1] |= (NV50_CB_PVP << 22);
-		else
-			e->inst[1] |= (NV50_CB_PFP << 22);
-	}
-#endif
 
 	e->param.index = src->hw;
 	e->param.shift = s;
 	e->param.mask = m << (s % 32);
+
+	e->inst[1] |= (((src->type == P_IMMD) ? 0 : 1) << 22);
 }
 
 static void
@@ -405,12 +479,11 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 
 	set_dst(pc, dst, e);
 
-	if (0 && dst->type != P_RESULT && src->type == P_IMMD) {
+	if (pc->allow32 && dst->type != P_RESULT && src->type == P_IMMD) {
 		set_immd(pc, src, e);
 		/*XXX: 32-bit, but steals part of "half" reg space - need to
 		 *     catch and handle this case if/when we do half-regs
 		 */
-		e->inst[0] |= 0x00008000;
 	} else
 	if (src->type == P_IMMD || src->type == P_CONST) {
 		set_long(pc, e);
@@ -426,18 +499,25 @@ emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 		e->inst[0] |= (src->hw << 9);
 	}
 
-	/* We really should support "half" instructions here at some point,
-	 * but I don't feel confident enough about them yet.
-	 */
-	set_long(pc, e);
 	if (is_long(e) && !is_immd(e)) {
 		e->inst[1] |= 0x04000000; /* 32-bit */
-		e->inst[1] |= 0x0003c000; /* "subsubop" 0xf == mov */
-	}
+		e->inst[1] |= 0x0000c000; /* "subsubop" 0x3 */
+		if (!(e->inst[1] & 0x20000000))
+			e->inst[1] |= 0x00030000; /* "subsubop" 0xf */
+	} else
+		e->inst[0] |= 0x00008000;
 
 	emit(pc, e);
 }
 
+static INLINE void
+emit_mov_immdval(struct nv50_pc *pc, struct nv50_reg *dst, float f)
+{
+	struct nv50_reg *imm = alloc_immd(pc, f);
+	emit_mov(pc, dst, imm);
+	FREE(imm);
+}
+
 static boolean
 check_swap_src_0_1(struct nv50_pc *pc,
 		   struct nv50_reg **s0, struct nv50_reg **s1)
@@ -541,12 +621,26 @@ emit_mul(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	struct nv50_program_exec *e = exec(pc);
 
 	e->inst[0] |= 0xc0000000;
-	set_long(pc, e);
+
+	if (!pc->allow32)
+		set_long(pc, e);
 
 	check_swap_src_0_1(pc, &src0, &src1);
 	set_dst(pc, dst, e);
 	set_src_0(pc, src0, e);
-	set_src_1(pc, src1, e);
+	if (src1->type == P_IMMD && !is_long(e)) {
+		if (src0->neg)
+			e->inst[0] |= 0x00008000;
+		set_immd(pc, src1, e);
+	} else {
+		set_src_1(pc, src1, e);
+		if (src0->neg ^ src1->neg) {
+			if (is_long(e))
+				e->inst[1] |= 0x08000000;
+			else
+				e->inst[0] |= 0x00008000;
+		}
+	}
 
 	emit(pc, e);
 }
@@ -560,11 +654,20 @@ emit_add(struct nv50_pc *pc, struct nv50_reg *dst,
 	e->inst[0] |= 0xb0000000;
 
 	check_swap_src_0_1(pc, &src0, &src1);
+
+	if (!pc->allow32 || src0->neg || src1->neg) {
+		set_long(pc, e);
+		e->inst[1] |= (src0->neg << 26) | (src1->neg << 27);
+	}
+
 	set_dst(pc, dst, e);
 	set_src_0(pc, src0, e);
-	if (is_long(e))
+	if (src1->type == P_CONST || src1->type == P_ATTR || is_long(e))
 		set_src_2(pc, src1, e);
 	else
+	if (src1->type == P_IMMD)
+		set_immd(pc, src1, e);
+	else
 		set_src_1(pc, src1, e);
 
 	emit(pc, e);
@@ -588,25 +691,13 @@ emit_minmax(struct nv50_pc *pc, unsigned sub, struct nv50_reg *dst,
 	emit(pc, e);
 }
 
-static void
+static INLINE void
 emit_sub(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	 struct nv50_reg *src1)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] |= 0xb0000000;
-
-	set_long(pc, e);
-	if (check_swap_src_0_1(pc, &src0, &src1))
-		e->inst[1] |= 0x04000000;
-	else
-		e->inst[1] |= 0x08000000;
-
-	set_dst(pc, dst, e);
-	set_src_0(pc, src0, e);
-	set_src_2(pc, src1, e);
-
-	emit(pc, e);
+	src1->neg ^= 1;
+	emit_add(pc, dst, src0, src1);
+	src1->neg ^= 1;
 }
 
 static void
@@ -623,26 +714,21 @@ emit_mad(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	set_src_1(pc, src1, e);
 	set_src_2(pc, src2, e);
 
+	if (src0->neg ^ src1->neg)
+		e->inst[1] |= 0x04000000;
+	if (src2->neg)
+		e->inst[1] |= 0x08000000;
+
 	emit(pc, e);
 }
 
-static void
+static INLINE void
 emit_msb(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
 	 struct nv50_reg *src1, struct nv50_reg *src2)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] |= 0xe0000000;
-	set_long(pc, e);
-	e->inst[1] |= 0x08000000; /* src0 * src1 - src2 */
-
-	check_swap_src_0_1(pc, &src0, &src1);
-	set_dst(pc, dst, e);
-	set_src_0(pc, src0, e);
-	set_src_1(pc, src1, e);
-	set_src_2(pc, src2, e);
-
-	emit(pc, e);
+	src2->neg ^= 1;
+	emit_mad(pc, dst, src0, src1, src2);
+	src2->neg ^= 1;
 }
 
 static void
@@ -693,6 +779,48 @@ emit_precossin(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 	emit(pc, e);
 }
 
+#define CVTOP_RN	0x01
+#define CVTOP_FLOOR	0x03
+#define CVTOP_CEIL	0x05
+#define CVTOP_TRUNC	0x07
+#define CVTOP_SAT	0x08
+#define CVTOP_ABS	0x10
+
+#define CVT_F32_F32 0xc4
+#define CVT_F32_S32 0x44
+#define CVT_F32_U32 0x64
+#define CVT_S32_F32 0x8c
+#define CVT_S32_S32 0x0c
+#define CVT_F32_F32_ROP 0xcc
+
+static void
+emit_cvt(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src,
+	 int wp, unsigned cop, unsigned fmt)
+{
+	struct nv50_program_exec *e;
+
+	e = exec(pc);
+	set_long(pc, e);
+
+	e->inst[0] |= 0xa0000000;
+	e->inst[1] |= 0x00004000;
+	e->inst[1] |= (cop << 16);
+	e->inst[1] |= (fmt << 24);
+	set_src_0(pc, src, e);
+
+	if (wp >= 0)
+		set_pred_wr(pc, 1, wp, e);
+
+	if (dst)
+		set_dst(pc, dst, e);
+	else {
+		e->inst[0] |= 0x000001fc;
+		e->inst[1] |= 0x00000008;
+	}
+
+	emit(pc, e);
+}
+
 static void
 emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst,
 	 struct nv50_reg *src0, struct nv50_reg *src1)
@@ -736,22 +864,10 @@ emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst,
 		free_temp(pc, dst);
 }
 
-static void
+static INLINE void
 emit_flr(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] = 0xa0000000; /* cvt */
-	set_long(pc, e);
-	e->inst[1] |= (6 << 29); /* cvt */
-	e->inst[1] |= 0x08000000; /* integer mode */
-	e->inst[1] |= 0x04000000; /* 32 bit */
-	e->inst[1] |= ((0x1 << 3)) << 14; /* .rn */
-	e->inst[1] |= (1 << 14); /* src .f32 */
-	set_dst(pc, dst, e);
-	set_src_0(pc, src, e);
-
-	emit(pc, e);
+	emit_cvt(pc, dst, src, -1, CVTOP_FLOOR, CVT_F32_F32_ROP);
 }
 
 static void
@@ -768,21 +884,10 @@ emit_pow(struct nv50_pc *pc, struct nv50_reg *dst,
 	free_temp(pc, temp);
 }
 
-static void
+static INLINE void
 emit_abs(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
 {
-	struct nv50_program_exec *e = exec(pc);
-
-	e->inst[0] = 0xa0000000; /* cvt */
-	set_long(pc, e);
-	e->inst[1] |= (6 << 29); /* cvt */
-	e->inst[1] |= 0x04000000; /* 32 bit */
-	e->inst[1] |= (1 << 14); /* src .f32 */
-	e->inst[1] |= ((1 << 6) << 14); /* .abs */
-	set_dst(pc, dst, e);
-	set_src_0(pc, src, e);
-
-	emit(pc, e);
+	emit_cvt(pc, dst, src, -1, CVTOP_ABS, CVT_F32_F32);
 }
 
 static void
@@ -794,18 +899,12 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
 	struct nv50_reg *neg128 = alloc_immd(pc, -127.999999);
 	struct nv50_reg *pos128 = alloc_immd(pc,  127.999999);
 	struct nv50_reg *tmp[4];
+	boolean allow32 = pc->allow32;
 
-	if (mask & (1 << 0))
-		emit_mov(pc, dst[0], one);
-
-	if (mask & (1 << 3))
-		emit_mov(pc, dst[3], one);
+	pc->allow32 = FALSE;
 
 	if (mask & (3 << 1)) {
-		if (mask & (1 << 1))
-			tmp[0] = dst[1];
-		else
-			tmp[0] = temp_temp(pc);
+		tmp[0] = alloc_temp(pc, NULL);
 		emit_minmax(pc, 4, tmp[0], src[0], zero);
 	}
 
@@ -823,6 +922,26 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
 		emit_mov(pc, dst[2], zero);
 		set_pred(pc, 3, 0, pc->p->exec_tail);
 	}
+
+	if (mask & (1 << 1))
+		assimilate_temp(pc, dst[1], tmp[0]);
+	else
+	if (mask & (1 << 2))
+		free_temp(pc, tmp[0]);
+
+	pc->allow32 = allow32;
+
+	/* do this last, in case src[i,j] == dst[0,3] */
+	if (mask & (1 << 0))
+		emit_mov(pc, dst[0], one);
+
+	if (mask & (1 << 3))
+		emit_mov(pc, dst[3], one);
+
+	FREE(pos128);
+	FREE(neg128);
+	FREE(zero);
+	FREE(one);
 }
 
 static void
@@ -853,6 +972,8 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src)
 	e->inst[1] = 0xc4014788;
 	set_src_0(pc, src, e);
 	set_pred_wr(pc, 1, r_pred, e);
+	if (src->neg)
+		e->inst[1] |= 0x20000000;
 	emit(pc, e);
 
 	/* This is probably KILP */
@@ -863,6 +984,180 @@ emit_kil(struct nv50_pc *pc, struct nv50_reg *src)
 	emit(pc, e);
 }
 
+static void
+emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
+	 struct nv50_reg **src, unsigned unit, unsigned type, boolean proj)
+{
+	struct nv50_reg *temp, *t[4];
+	struct nv50_program_exec *e;
+
+	unsigned c, mode, dim;
+
+	switch (type) {
+	case TGSI_TEXTURE_1D:
+		dim = 1;
+		break;
+	case TGSI_TEXTURE_UNKNOWN:
+	case TGSI_TEXTURE_2D:
+	case TGSI_TEXTURE_SHADOW1D: /* XXX: x, z */
+	case TGSI_TEXTURE_RECT:
+		dim = 2;
+		break;
+	case TGSI_TEXTURE_3D:
+	case TGSI_TEXTURE_CUBE:
+	case TGSI_TEXTURE_SHADOW2D:
+	case TGSI_TEXTURE_SHADOWRECT: /* XXX */
+		dim = 3;
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	alloc_temp4(pc, t, 0);
+
+	if (proj) {
+		if (src[0]->type == P_TEMP && src[0]->rhw != -1) {
+			mode = pc->interp_mode[src[0]->index];
+
+			t[3]->rhw = src[3]->rhw;
+			emit_interp(pc, t[3], NULL, (mode & INTERP_CENTROID));
+			emit_flop(pc, 0, t[3], t[3]);
+
+			for (c = 0; c < dim; c++) {
+				t[c]->rhw = src[c]->rhw;
+				emit_interp(pc, t[c], t[3],
+					    (mode | INTERP_PERSPECTIVE));
+			}
+		} else {
+			emit_flop(pc, 0, t[3], src[3]);
+			for (c = 0; c < dim; c++)
+				emit_mul(pc, t[c], src[c], t[3]);
+
+			/* XXX: for some reason the blob sometimes uses MAD:
+			 * emit_mad(pc, t[c], src[0][c], t[3], t[3])
+			 * pc->p->exec_tail->inst[1] |= 0x080fc000;
+			 */
+		}
+	} else {
+		if (type == TGSI_TEXTURE_CUBE) {
+			temp = temp_temp(pc);
+			emit_minmax(pc, 4, temp, src[0], src[1]);
+			emit_minmax(pc, 4, temp, temp, src[2]);
+			emit_flop(pc, 0, temp, temp);
+			for (c = 0; c < 3; c++)
+				emit_mul(pc, t[c], src[c], temp);
+		} else {
+			for (c = 0; c < dim; c++)
+				emit_mov(pc, t[c], src[c]);
+		}
+	}
+
+	e = exec(pc);
+	set_long(pc, e);
+	e->inst[0] |= 0xf0000000;
+	e->inst[1] |= 0x00000004;
+	set_dst(pc, t[0], e);
+	e->inst[0] |= (unit << 9);
+
+	if (dim == 2)
+		e->inst[0] |= 0x00400000;
+	else
+	if (dim == 3)
+		e->inst[0] |= 0x00800000;
+
+	e->inst[0] |= (mask & 0x3) << 25;
+	e->inst[1] |= (mask & 0xc) << 12;
+
+	emit(pc, e);
+
+#if 1
+	if (mask & 1) emit_mov(pc, dst[0], t[0]);
+	if (mask & 2) emit_mov(pc, dst[1], t[1]);
+	if (mask & 4) emit_mov(pc, dst[2], t[2]);
+	if (mask & 8) emit_mov(pc, dst[3], t[3]);
+
+	free_temp4(pc, t);
+#else
+	/* XXX: if p.e. MUL is used directly after TEX, it would still use
+	 * the texture coordinates, not the fetched values: latency ? */
+
+	for (c = 0; c < 4; c++) {
+		if (mask & (1 << c))
+			assimilate_temp(pc, dst[c], t[c]);
+		else
+			free_temp(pc, t[c]);
+	}
+#endif
+}
+
+static void
+convert_to_long(struct nv50_pc *pc, struct nv50_program_exec *e)
+{
+	unsigned q = 0, m = ~0;
+
+	assert(!is_long(e));
+
+	switch (e->inst[0] >> 28) {
+	case 0x1:
+		/* MOV */
+		q = 0x0403c000;
+		m = 0xffff7fff;
+		break;
+	case 0x8:
+		/* INTERP */
+		m = ~0x02000000;
+		if (e->inst[0] & 0x02000000)
+			q = 0x00020000;
+		break;
+	case 0x9:
+		/* RCP */
+		break;
+	case 0xB:
+		/* ADD */
+		m = ~(127 << 16);
+		q = ((e->inst[0] & (~m)) >> 2);
+		break;
+	case 0xC:
+		/* MUL */
+		m = ~0x00008000;
+		q = ((e->inst[0] & (~m)) << 12);
+		break;
+	case 0xE:
+		/* MAD (if src2 == dst) */
+		q = ((e->inst[0] & 0x1fc) << 12);
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	set_long(pc, e);
+	pc->p->exec_size++;
+
+	e->inst[0] &= m;
+	e->inst[1] |= q;
+}
+
+static boolean
+negate_supported(const struct tgsi_full_instruction *insn, int i)
+{
+	switch (insn->Instruction.Opcode) {
+	case TGSI_OPCODE_DP3:
+	case TGSI_OPCODE_DP4:
+	case TGSI_OPCODE_MUL:
+	case TGSI_OPCODE_KIL:
+	case TGSI_OPCODE_ADD:
+	case TGSI_OPCODE_SUB:
+	case TGSI_OPCODE_MAD:
+		return TRUE;
+	case TGSI_OPCODE_POW:
+		return (i == 1) ? TRUE : FALSE;
+	default:
+		return FALSE;
+	}
+}
+
 static struct nv50_reg *
 tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst)
 {
@@ -881,11 +1176,14 @@ tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst)
 }
 
 static struct nv50_reg *
-tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
+tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src,
+	 boolean neg)
 {
 	struct nv50_reg *r = NULL;
 	struct nv50_reg *temp;
-	unsigned c;
+	unsigned sgn, c;
+
+	sgn = tgsi_util_get_full_src_register_sign_mode(src, chan);
 
 	c = tgsi_util_get_full_src_register_extswizzle(src, chan);
 	switch (c) {
@@ -915,16 +1213,17 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
 		break;
 	case TGSI_EXTSWIZZLE_ZERO:
 		r = alloc_immd(pc, 0.0);
-		break;
+		return r;
 	case TGSI_EXTSWIZZLE_ONE:
-		r = alloc_immd(pc, 1.0);
-		break;
+		if (sgn == TGSI_UTIL_SIGN_TOGGLE || sgn == TGSI_UTIL_SIGN_SET)
+			return alloc_immd(pc, -1.0);
+		return alloc_immd(pc, 1.0);
 	default:
 		assert(0);
 		break;
 	}
 
-	switch (tgsi_util_get_full_src_register_sign_mode(src, chan)) {
+	switch (sgn) {
 	case TGSI_UTIL_SIGN_KEEP:
 		break;
 	case TGSI_UTIL_SIGN_CLEAR:
@@ -933,14 +1232,21 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
 		r = temp;
 		break;
 	case TGSI_UTIL_SIGN_TOGGLE:
-		temp = temp_temp(pc);
-		emit_neg(pc, temp, r);
-		r = temp;
+		if (neg)
+			r->neg = 1;
+		else {
+			temp = temp_temp(pc);
+			emit_neg(pc, temp, r);
+			r = temp;
+		}
 		break;
 	case TGSI_UTIL_SIGN_SET:
 		temp = temp_temp(pc);
 		emit_abs(pc, temp, r);
-		emit_neg(pc, temp, r);
+		if (neg)
+			temp->neg = 1;
+		else
+			emit_neg(pc, temp, temp);
 		r = temp;
 		break;
 	default:
@@ -951,12 +1257,40 @@ tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
 	return r;
 }
 
+/* returns TRUE if instruction can overwrite sources before they're read */
+static boolean
+direct2dest_op(const struct tgsi_full_instruction *insn)
+{
+	if (insn->Instruction.Saturate)
+		return FALSE;
+
+	switch (insn->Instruction.Opcode) {
+	case TGSI_OPCODE_COS:
+	case TGSI_OPCODE_DP3:
+	case TGSI_OPCODE_DP4:
+	case TGSI_OPCODE_DPH:
+	case TGSI_OPCODE_KIL:
+	case TGSI_OPCODE_LIT:
+	case TGSI_OPCODE_POW:
+	case TGSI_OPCODE_RCP:
+	case TGSI_OPCODE_RSQ:
+	case TGSI_OPCODE_SCS:
+	case TGSI_OPCODE_SIN:
+	case TGSI_OPCODE_TEX:
+	case TGSI_OPCODE_TXP:
+		return FALSE;
+	default:
+		return TRUE;
+	}
+}
+
 static boolean
 nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 {
 	const struct tgsi_full_instruction *inst = &tok->FullInstruction;
 	struct nv50_reg *rdst[4], *dst[4], *src[3][4], *temp;
 	unsigned mask, sat, unit;
+	boolean assimilate = FALSE;
 	int i, c;
 
 	mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
@@ -967,6 +1301,10 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			dst[c] = tgsi_dst(pc, c, &inst->FullDstRegisters[0]);
 		else
 			dst[c] = NULL;
+		rdst[c] = NULL;
+		src[0][c] = NULL;
+		src[1][c] = NULL;
+		src[2][c] = NULL;
 	}
 
 	for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
@@ -976,7 +1314,8 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			unit = fs->SrcRegister.Index;
 
 		for (c = 0; c < 4; c++)
-			src[i][c] = tgsi_src(pc, c, fs);
+			src[i][c] = tgsi_src(pc, c, fs,
+					     negate_supported(inst, i));
 	}
 
 	if (sat) {
@@ -984,6 +1323,25 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			rdst[c] = dst[c];
 			dst[c] = temp_temp(pc);
 		}
+	} else
+	if (direct2dest_op(inst)) {
+		for (c = 0; c < 4; c++) {
+			if (!dst[c] || dst[c]->type != P_TEMP)
+				continue;
+
+			for (i = c + 1; i < 4; i++) {
+				if (dst[c] == src[0][i] ||
+				    dst[c] == src[1][i] ||
+				    dst[c] == src[2][i])
+					break;
+			}
+			if (i == 4)
+				continue;
+
+			assimilate = TRUE;
+			rdst[c] = dst[c];
+			dst[c] = alloc_temp(pc, NULL);
+		}
 	}
 
 	switch (inst->Instruction.Opcode) {
@@ -1002,7 +1360,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_COS:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_precossin(pc, temp, src[0][0]);
 		emit_flop(pc, 5, temp, temp);
 		for (c = 0; c < 4; c++) {
@@ -1012,7 +1370,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_DP3:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_mul(pc, temp, src[0][0], src[1][0]);
 		emit_mad(pc, temp, src[0][1], src[1][1], temp);
 		emit_mad(pc, temp, src[0][2], src[1][2], temp);
@@ -1021,10 +1379,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_DP4:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_mul(pc, temp, src[0][0], src[1][0]);
 		emit_mad(pc, temp, src[0][1], src[1][1], temp);
 		emit_mad(pc, temp, src[0][2], src[1][2], temp);
@@ -1034,10 +1391,9 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_DPH:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_mul(pc, temp, src[0][0], src[1][0]);
 		emit_mad(pc, temp, src[0][1], src[1][1], temp);
 		emit_mad(pc, temp, src[0][2], src[1][2], temp);
@@ -1047,7 +1403,6 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_DST:
 	{
@@ -1064,7 +1419,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 	}
 		break;
 	case TGSI_OPCODE_EX2:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_preex2(pc, temp, src[0][0]);
 		emit_flop(pc, 6, temp, temp);
 		for (c = 0; c < 4; c++) {
@@ -1072,7 +1427,6 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_FLR:
 		for (c = 0; c < 4; c++) {
@@ -1082,26 +1436,26 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_FRC:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_flr(pc, temp, src[0][c]);
 			emit_sub(pc, dst[c], src[0][c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_KIL:
 		emit_kil(pc, src[0][0]);
 		emit_kil(pc, src[0][1]);
 		emit_kil(pc, src[0][2]);
 		emit_kil(pc, src[0][3]);
+		pc->p->cfg.fp.regs[2] |= 0x00100000;
 		break;
 	case TGSI_OPCODE_LIT:
 		emit_lit(pc, &dst[0], mask, &src[0][0]);
 		break;
 	case TGSI_OPCODE_LG2:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_flop(pc, 3, temp, src[0][0]);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
@@ -1110,15 +1464,12 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_LRP:
+		temp = temp_temp(pc);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
-			/*XXX: we can do better than this */
-			temp = alloc_temp(pc, NULL);
-			emit_neg(pc, temp, src[0][c]);
-			emit_mad(pc, temp, temp, src[2][c], src[2][c]);
-			emit_mad(pc, dst[c], src[0][c], src[1][c], temp);
-			free_temp(pc, temp);
+			emit_sub(pc, temp, src[1][c], src[2][c]);
+			emit_mad(pc, dst[c], temp, src[0][c], src[2][c]);
 		}
 		break;
 	case TGSI_OPCODE_MAD:
@@ -1157,36 +1508,39 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_POW:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_pow(pc, temp, src[0][0], src[1][0]);
 		for (c = 0; c < 4; c++) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_mov(pc, dst[c], temp);
 		}
-		free_temp(pc, temp);
 		break;
 	case TGSI_OPCODE_RCP:
-		for (c = 0; c < 4; c++) {
+		for (c = 3; c >= 0; c--) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_flop(pc, 0, dst[c], src[0][0]);
 		}
 		break;
 	case TGSI_OPCODE_RSQ:
-		for (c = 0; c < 4; c++) {
+		for (c = 3; c >= 0; c--) {
 			if (!(mask & (1 << c)))
 				continue;
 			emit_flop(pc, 2, dst[c], src[0][0]);
 		}
 		break;
 	case TGSI_OPCODE_SCS:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_precossin(pc, temp, src[0][0]);
 		if (mask & (1 << 0))
 			emit_flop(pc, 5, dst[0], temp);
 		if (mask & (1 << 1))
 			emit_flop(pc, 4, dst[1], temp);
+		if (mask & (1 << 2))
+			emit_mov_immdval(pc, dst[2], 0.0);
+		if (mask & (1 << 3))
+			emit_mov_immdval(pc, dst[3], 1.0);
 		break;
 	case TGSI_OPCODE_SGE:
 		for (c = 0; c < 4; c++) {
@@ -1196,7 +1550,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_SIN:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		emit_precossin(pc, temp, src[0][0]);
 		emit_flop(pc, 4, temp, temp);
 		for (c = 0; c < 4; c++) {
@@ -1220,33 +1574,15 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 		}
 		break;
 	case TGSI_OPCODE_TEX:
+		emit_tex(pc, dst, mask, src[0], unit,
+			 inst->InstructionExtTexture.Texture, FALSE);
+		break;
 	case TGSI_OPCODE_TXP:
-	{
-		struct nv50_reg *t[4];
-		struct nv50_program_exec *e;
-
-		alloc_temp4(pc, t, 0);
-		emit_mov(pc, t[0], src[0][0]);
-		emit_mov(pc, t[1], src[0][1]);
-
-		e = exec(pc);
-		e->inst[0] = 0xf6400000;
-		e->inst[0] |= (unit << 9);
-		set_long(pc, e);
-		e->inst[1] |= 0x0000c004;
-		set_dst(pc, t[0], e);
-		emit(pc, e);
-
-		if (mask & (1 << 0)) emit_mov(pc, dst[0], t[0]);
-		if (mask & (1 << 1)) emit_mov(pc, dst[1], t[1]);
-		if (mask & (1 << 2)) emit_mov(pc, dst[2], t[2]);
-		if (mask & (1 << 3)) emit_mov(pc, dst[3], t[3]);
-
-		free_temp4(pc, t);
-	}
+		emit_tex(pc, dst, mask, src[0], unit,
+			 inst->InstructionExtTexture.Texture, TRUE);
 		break;
 	case TGSI_OPCODE_XPD:
-		temp = alloc_temp(pc, NULL);
+		temp = temp_temp(pc);
 		if (mask & (1 << 0)) {
 			emit_mul(pc, temp, src[0][2], src[1][1]);
 			emit_msb(pc, dst[0], src[0][1], src[1][2], temp);
@@ -1259,7 +1595,8 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 			emit_mul(pc, temp, src[0][1], src[1][0]);
 			emit_msb(pc, dst[2], src[0][0], src[1][1], temp);
 		}
-		free_temp(pc, temp);
+		if (mask & (1 << 3))
+			emit_mov_immdval(pc, dst[3], 1.0);
 		break;
 	case TGSI_OPCODE_END:
 		break;
@@ -1270,21 +1607,26 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 
 	if (sat) {
 		for (c = 0; c < 4; c++) {
-			struct nv50_program_exec *e;
-
 			if (!(mask & (1 << c)))
 				continue;
-			e = exec(pc);
+			emit_cvt(pc, rdst[c], dst[c], -1, CVTOP_SAT,
+				 CVT_F32_F32);
+		}
+	} else if (assimilate) {
+		for (c = 0; c < 4; c++)
+			if (rdst[c])
+				assimilate_temp(pc, rdst[c], dst[c]);
+	}
 
-			e->inst[0] = 0xa0000000; /* cvt */
-			set_long(pc, e);
-			e->inst[1] |= (6 << 29); /* cvt */
-			e->inst[1] |= 0x04000000; /* 32 bit */
-			e->inst[1] |= (1 << 14); /* src .f32 */
-			e->inst[1] |= ((1 << 5) << 14); /* .sat */
-			set_dst(pc, rdst[c], e);
-			set_src_0(pc, dst[c], e);
-			emit(pc, e);
+	for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+		for (c = 0; c < 4; c++) {
+			if (!src[i][c])
+				continue;
+			if (src[i][c]->index == -1 && src[i][c]->type == P_IMMD)
+				FREE(src[i][c]);
+			else
+			if (src[i][c]->acc == pc->insn_cur)
+				release_hw(pc, src[i][c]);
 		}
 	}
 
@@ -1292,12 +1634,169 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
 	return TRUE;
 }
 
+/* Adjust a bitmask that indicates what components of a source are used,
+ * we use this in tx_prep so we only load interpolants that are needed.
+ */
+static void
+insn_adjust_mask(const struct tgsi_full_instruction *insn, unsigned *mask)
+{
+	const struct tgsi_instruction_ext_texture *tex;
+
+	switch (insn->Instruction.Opcode) {
+	case TGSI_OPCODE_DP3:
+		*mask = 0x7;
+		break;
+	case TGSI_OPCODE_DP4:
+	case TGSI_OPCODE_DPH:
+		*mask = 0xF;
+		break;
+	case TGSI_OPCODE_LIT:
+		*mask = 0xB;
+		break;
+	case TGSI_OPCODE_RCP:
+	case TGSI_OPCODE_RSQ:
+		*mask = 0x1;
+		break;
+	case TGSI_OPCODE_TEX:
+	case TGSI_OPCODE_TXP:
+		assert(insn->Instruction.Extended);
+		tex = &insn->InstructionExtTexture;
+
+		*mask = 0x7;
+		if (tex->Texture == TGSI_TEXTURE_1D)
+			*mask = 0x1;
+		else
+		if (tex->Texture == TGSI_TEXTURE_2D)
+			*mask = 0x3;
+
+		if (insn->Instruction.Opcode == TGSI_OPCODE_TXP)
+			*mask |= 0x8;
+		break;
+	default:
+		break;
+	}
+}
+
+static void
+prep_inspect_insn(struct nv50_pc *pc, const union tgsi_full_token *tok,
+		  unsigned *r_usage[2])
+{
+	const struct tgsi_full_instruction *insn;
+	const struct tgsi_full_src_register *src;
+	const struct tgsi_dst_register *dst;
+
+	unsigned i, c, k, n, mask, *acc_p;
+
+	insn = &tok->FullInstruction;
+	dst = &insn->FullDstRegisters[0].DstRegister;
+	mask = dst->WriteMask;
+
+	if (!r_usage[0])
+		r_usage[0] = CALLOC(pc->temp_nr * 4, sizeof(unsigned));
+	if (!r_usage[1])
+		r_usage[1] = CALLOC(pc->attr_nr * 4, sizeof(unsigned));
+
+	if (dst->File == TGSI_FILE_TEMPORARY) {
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			r_usage[0][dst->Index * 4 + c] = pc->insn_nr;
+		}
+	}
+
+	for (i = 0; i < insn->Instruction.NumSrcRegs; i++) {
+		src = &insn->FullSrcRegisters[i];
+
+		switch (src->SrcRegister.File) {
+		case TGSI_FILE_TEMPORARY:
+			acc_p = r_usage[0];
+			break;
+		case TGSI_FILE_INPUT:
+			acc_p = r_usage[1];
+			break;
+		default:
+			continue;
+		}
+
+		insn_adjust_mask(insn, &mask);
+
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+
+			k = tgsi_util_get_full_src_register_extswizzle(src, c);
+			switch (k) {
+			case TGSI_EXTSWIZZLE_X:
+			case TGSI_EXTSWIZZLE_Y:
+			case TGSI_EXTSWIZZLE_Z:
+			case TGSI_EXTSWIZZLE_W:
+				n = src->SrcRegister.Index * 4 + k;
+				acc_p[n] = pc->insn_nr;
+				break;
+			default:
+				break;
+			}
+		}
+	}
+}
+
+static unsigned
+load_fp_attrib(struct nv50_pc *pc, int i, unsigned *acc, int *mid,
+	       int *aid, int *p_oid)
+{
+	struct nv50_reg *iv;
+	int oid, c, n;
+	unsigned mask = 0;
+
+	iv = (pc->interp_mode[i] & INTERP_CENTROID) ? pc->iv_c : pc->iv_p;
+
+	for (c = 0, n = i * 4; c < 4; c++, n++) {
+		oid = (*p_oid)++;
+		pc->attr[n].type = P_TEMP;
+		pc->attr[n].index = i;
+
+		if (pc->attr[n].acc == acc[n])
+			continue;
+		mask |= (1 << c);
+
+		pc->attr[n].acc = acc[n];
+		pc->attr[n].rhw = pc->attr[n].hw = -1;
+		alloc_reg(pc, &pc->attr[n]);
+
+		pc->attr[n].rhw = (*aid)++;
+		emit_interp(pc, &pc->attr[n], iv, pc->interp_mode[i]);
+
+		pc->p->cfg.fp.map[(*mid) / 4] |= oid << (8 * ((*mid) % 4));
+		(*mid)++;
+		pc->p->cfg.fp.regs[1] += 0x00010001;
+	}
+
+	return mask;
+}
+
 static boolean
 nv50_program_tx_prep(struct nv50_pc *pc)
 {
 	struct tgsi_parse_context p;
 	boolean ret = FALSE;
 	unsigned i, c;
+	unsigned fcol, bcol, fcrd, depr;
+
+	/* count (centroid) perspective interpolations */
+	unsigned centroid_loads = 0;
+	unsigned perspect_loads = 0;
+
+	/* track register access for temps and attrs */
+	unsigned *r_usage[2];
+	r_usage[0] = NULL;
+	r_usage[1] = NULL;
+
+	depr = fcol = bcol = fcrd = 0xffff;
+
+	if (pc->p->type == PIPE_SHADER_FRAGMENT) {
+		pc->p->cfg.fp.regs[0] = 0x01000404;
+		pc->p->cfg.fp.regs[1] = 0x00000400;
+	}
 
 	tgsi_parse_init(&p, pc->p->pipe.tokens);
 	while (!tgsi_parse_end_of_tokens(&p)) {
@@ -1319,9 +1818,10 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 		case TGSI_TOKEN_TYPE_DECLARATION:
 		{
 			const struct tgsi_full_declaration *d;
-			unsigned last;
+			unsigned last, first, mode;
 
 			d = &p.FullToken.FullDeclaration;
+			first = d->DeclarationRange.First;
 			last = d->DeclarationRange.Last;
 
 			switch (d->Declaration.File) {
@@ -1332,10 +1832,69 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 			case TGSI_FILE_OUTPUT:
 				if (pc->result_nr < (last + 1))
 					pc->result_nr = last + 1;
+
+				if (!d->Declaration.Semantic)
+					break;
+
+				switch (d->Semantic.SemanticName) {
+				case TGSI_SEMANTIC_POSITION:
+					depr = first;
+					pc->p->cfg.fp.regs[2] |= 0x00000100;
+					pc->p->cfg.fp.regs[3] |= 0x00000011;
+					break;
+				default:
+					break;
+				}
+
 				break;
 			case TGSI_FILE_INPUT:
+			{
 				if (pc->attr_nr < (last + 1))
 					pc->attr_nr = last + 1;
+
+				if (pc->p->type != PIPE_SHADER_FRAGMENT)
+					break;
+
+				switch (d->Declaration.Interpolate) {
+				case TGSI_INTERPOLATE_CONSTANT:
+					mode = INTERP_FLAT;
+					break;
+				case TGSI_INTERPOLATE_PERSPECTIVE:
+					mode = INTERP_PERSPECTIVE;
+					break;
+				default:
+					mode = INTERP_LINEAR;
+					break;
+				}
+
+				if (d->Declaration.Semantic) {
+					switch (d->Semantic.SemanticName) {
+					case TGSI_SEMANTIC_POSITION:
+						fcrd = first;
+						break;
+					case TGSI_SEMANTIC_COLOR:
+						fcol = first;
+						mode = INTERP_PERSPECTIVE;
+						break;
+					case TGSI_SEMANTIC_BCOLOR:
+						bcol = first;
+						mode = INTERP_PERSPECTIVE;
+						break;
+					}
+				}
+
+				if (d->Declaration.Centroid) {
+					mode |= INTERP_CENTROID;
+					if (mode & INTERP_PERSPECTIVE)
+						centroid_loads++;
+				} else
+				if (mode & INTERP_PERSPECTIVE)
+					perspect_loads++;
+
+				assert(last < 32);
+				for (i = first; i <= last; i++)
+					pc->interp_mode[i] = mode;
+			}
 				break;
 			case TGSI_FILE_CONSTANT:
 				if (pc->param_nr < (last + 1))
@@ -1351,6 +1910,8 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 		}
 			break;
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			pc->insn_nr++;
+			prep_inspect_insn(pc, tok, r_usage);
 			break;
 		default:
 			break;
@@ -1366,56 +1927,95 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 			for (c = 0; c < 4; c++) {
 				pc->temp[i*4+c].type = P_TEMP;
 				pc->temp[i*4+c].hw = -1;
+				pc->temp[i*4+c].rhw = -1;
 				pc->temp[i*4+c].index = i;
+				pc->temp[i*4+c].acc = r_usage[0][i*4+c];
 			}
 		}
 	}
 
 	if (pc->attr_nr) {
-		struct nv50_reg *iv = NULL;
-		int aid = 0;
+		int oid = 4, mid = 4, aid = 0;
+		/* oid = VP output id
+		 * aid = FP attribute/interpolant id
+		 * mid = VP output mapping field ID
+		 */
 
 		pc->attr = CALLOC(pc->attr_nr * 4, sizeof(struct nv50_reg));
 		if (!pc->attr)
 			goto out_err;
 
 		if (pc->p->type == PIPE_SHADER_FRAGMENT) {
-			iv = alloc_temp(pc, NULL);
-			emit_interp(pc, iv, iv, NULL);
-			emit_flop(pc, 0, iv, iv);
-			aid++;
-		}
+			/* position should be loaded first */
+			if (fcrd != 0xffff) {
+				unsigned mask;
+				mid = 0;
+				mask = load_fp_attrib(pc, fcrd, r_usage[1],
+						      &mid, &aid, &oid);
+				oid = 0;
+				pc->p->cfg.fp.regs[1] |= (mask << 24);
+				pc->p->cfg.fp.map[0] = 0x04040404 * fcrd;
+			}
+			pc->p->cfg.fp.map[0] += 0x03020100;
 
-		for (i = 0; i < pc->attr_nr; i++) {
-			struct nv50_reg *a = &pc->attr[i*4];
+			/* should do MAD fcrd.xy, fcrd, SOME_CONST, fcrd */
 
-			for (c = 0; c < 4; c++) {
-				if (pc->p->type == PIPE_SHADER_FRAGMENT) {
-					struct nv50_reg *at =
-						alloc_temp(pc, NULL);
-					pc->attr[i*4+c].type = at->type;
-					pc->attr[i*4+c].hw = at->hw;
-					pc->attr[i*4+c].index = at->index;
+			if (perspect_loads) {
+				pc->iv_p = alloc_temp(pc, NULL);
+
+				if (!(pc->p->cfg.fp.regs[1] & 0x08000000)) {
+					pc->p->cfg.fp.regs[1] |= 0x08000000;
+					pc->iv_p->rhw = aid++;
+					emit_interp(pc, pc->iv_p, NULL,
+						    INTERP_LINEAR);
+					emit_flop(pc, 0, pc->iv_p, pc->iv_p);
 				} else {
-					pc->p->cfg.vp.attr[aid/32] |=
-						(1 << (aid % 32));
-					pc->attr[i*4+c].type = P_ATTR;
-					pc->attr[i*4+c].hw = aid++;
-					pc->attr[i*4+c].index = i;
+					pc->iv_p->rhw = aid - 1;
+					emit_flop(pc, 0, pc->iv_p,
+						  &pc->attr[fcrd * 4 + 3]);
 				}
 			}
 
-			if (pc->p->type != PIPE_SHADER_FRAGMENT)
-				continue;
+			if (centroid_loads) {
+				pc->iv_c = alloc_temp(pc, NULL);
+				pc->iv_c->rhw = pc->iv_p ? aid - 1 : aid++;
+				emit_interp(pc, pc->iv_c, NULL,
+					    INTERP_CENTROID);
+				emit_flop(pc, 0, pc->iv_c, pc->iv_c);
+				pc->p->cfg.fp.regs[1] |= 0x08000000;
+			}
 
-			emit_interp(pc, &a[0], &a[0], iv);
-			emit_interp(pc, &a[1], &a[1], iv);
-			emit_interp(pc, &a[2], &a[2], iv);
-			emit_interp(pc, &a[3], &a[3], iv);
-		}
+			for (c = 0; c < 4; c++) {
+				/* I don't know what these values do, but
+				 * let's set them like the blob does:
+				 */
+				if (fcol != 0xffff && r_usage[1][fcol * 4 + c])
+					pc->p->cfg.fp.regs[0] += 0x00010000;
+				if (bcol != 0xffff && r_usage[1][bcol * 4 + c])
+					pc->p->cfg.fp.regs[0] += 0x00010000;
+			}
 
-		if (iv)
-			free_temp(pc, iv);
+			for (i = 0; i < pc->attr_nr; i++)
+				load_fp_attrib(pc, i, r_usage[1],
+					       &mid, &aid, &oid);
+
+			if (pc->iv_p)
+				free_temp(pc, pc->iv_p);
+			if (pc->iv_c)
+				free_temp(pc, pc->iv_c);
+
+			pc->p->cfg.fp.high_map = (mid / 4);
+			pc->p->cfg.fp.high_map += ((mid % 4) ? 1 : 0);
+		} else {
+			/* vertex program */
+			for (i = 0; i < pc->attr_nr * 4; i++) {
+				pc->p->cfg.vp.attr[aid / 32] |=
+					(1 << (aid % 32));
+				pc->attr[i].type = P_ATTR;
+				pc->attr[i].hw = aid++;
+				pc->attr[i].index = i / 4;
+			}
+		}
 	}
 
 	if (pc->result_nr) {
@@ -1430,12 +2030,20 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 				if (pc->p->type == PIPE_SHADER_FRAGMENT) {
 					pc->result[i*4+c].type = P_TEMP;
 					pc->result[i*4+c].hw = -1;
+					pc->result[i*4+c].rhw = (i == depr) ?
+						-1 : rid++;
 				} else {
 					pc->result[i*4+c].type = P_RESULT;
 					pc->result[i*4+c].hw = rid++;
 				}
 				pc->result[i*4+c].index = i;
 			}
+
+			if (pc->p->type == PIPE_SHADER_FRAGMENT &&
+			    depr != 0xffff) {
+				pc->result[depr * 4 + 2].rhw =
+					(pc->result_nr - 1) * 4;
+			}
 		}
 	}
 
@@ -1456,7 +2064,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 	}
 
 	if (pc->immd_nr) {
-		int rid = pc->param_nr * 4;
+		int rid = 0;
 
 		pc->immd = CALLOC(pc->immd_nr * 4, sizeof(struct nv50_reg));
 		if (!pc->immd)
@@ -1473,15 +2081,38 @@ nv50_program_tx_prep(struct nv50_pc *pc)
 
 	ret = TRUE;
 out_err:
+	if (r_usage[0])
+		FREE(r_usage[0]);
+	if (r_usage[1])
+		FREE(r_usage[1]);
+
 	tgsi_parse_free(&p);
 	return ret;
 }
 
+static void
+free_nv50_pc(struct nv50_pc *pc)
+{
+	if (pc->immd)
+		FREE(pc->immd);
+	if (pc->param)
+		FREE(pc->param);
+	if (pc->result)
+		FREE(pc->result);
+	if (pc->attr)
+		FREE(pc->attr);
+	if (pc->temp)
+		FREE(pc->temp);
+
+	FREE(pc);
+}
+
 static boolean
 nv50_program_tx(struct nv50_program *p)
 {
 	struct tgsi_parse_context parse;
 	struct nv50_pc *pc;
+	unsigned k;
 	boolean ret;
 
 	pc = CALLOC_STRUCT(nv50_pc);
@@ -1498,10 +2129,16 @@ nv50_program_tx(struct nv50_program *p)
 	while (!tgsi_parse_end_of_tokens(&parse)) {
 		const union tgsi_full_token *tok = &parse.FullToken;
 
+		/* don't allow half insn/immd on first and last instruction */
+		pc->allow32 = TRUE;
+		if (pc->insn_cur == 0 || pc->insn_cur + 2 == pc->insn_nr)
+			pc->allow32 = FALSE;
+
 		tgsi_parse_token(&parse);
 
 		switch (tok->Token.Type) {
 		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			++pc->insn_cur;
 			ret = nv50_program_tx_insn(pc, tok);
 			if (ret == FALSE)
 				goto out_err;
@@ -1515,8 +2152,40 @@ nv50_program_tx(struct nv50_program *p)
 		struct nv50_reg out;
 
 		out.type = P_TEMP;
-		for (out.hw = 0; out.hw < pc->result_nr * 4; out.hw++)
-			emit_mov(pc, &out, &pc->result[out.hw]);
+		for (k = 0; k < pc->result_nr * 4; k++) {
+			if (pc->result[k].rhw == -1)
+				continue;
+			if (pc->result[k].hw != pc->result[k].rhw) {
+				out.hw = pc->result[k].rhw;
+				emit_mov(pc, &out, &pc->result[k]);
+			}
+			if (pc->p->cfg.high_result < (pc->result[k].rhw + 1))
+				pc->p->cfg.high_result = pc->result[k].rhw + 1;
+		}
+	}
+
+	/* look for single half instructions and make them long */
+	struct nv50_program_exec *e, *e_prev;
+
+	for (k = 0, e = pc->p->exec_head, e_prev = NULL; e; e = e->next) {
+		if (!is_long(e))
+			k++;
+
+		if (!e->next || is_long(e->next)) {
+			if (k & 1)
+				convert_to_long(pc, e);
+			k = 0;
+		}
+
+		if (e->next)
+			e_prev = e;
+	}
+
+	if (!is_long(pc->p->exec_tail)) {
+		/* this may occur if moving FP results */
+		assert(e_prev && !is_long(e_prev));
+		convert_to_long(pc, e_prev);
+		convert_to_long(pc, pc->p->exec_tail);
 	}
 
 	assert(is_long(pc->p->exec_tail) && !is_immd(pc->p->exec_head));
@@ -1530,6 +2199,7 @@ out_err:
 	tgsi_parse_free(&parse);
 
 out_cleanup:
+	free_nv50_pc(pc);
 	return ret;
 }
 
@@ -1543,16 +2213,16 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *p)
 
 static void
 nv50_program_upload_data(struct nv50_context *nv50, float *map,
-			 unsigned start, unsigned count)
+			unsigned start, unsigned count, unsigned cbuf)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	while (count) {
 		unsigned nr = count > 2047 ? 2047 : count;
 
 		BEGIN_RING(chan, tesla, 0x00000f00, 1);
-		OUT_RING  (chan, (NV50_CB_PMISC << 0) | (start << 8));
+		OUT_RING  (chan, (cbuf << 0) | (start << 8));
 		BEGIN_RING(chan, tesla, 0x40000f04, nr);
 		OUT_RINGp (chan, map, nr);
 
@@ -1565,70 +2235,93 @@ nv50_program_upload_data(struct nv50_context *nv50, float *map,
 static void
 nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
 {
-	struct nouveau_winsys *nvws = nv50->screen->nvws;
-	struct pipe_winsys *ws = nv50->pipe.winsys;
-	unsigned nr = p->param_nr + p->immd_nr;
+	struct pipe_screen *pscreen = nv50->pipe.screen;
 
-	if (!p->data && nr) {
-		struct nouveau_resource *heap = nv50->screen->vp_data_heap;
+	if (!p->data[0] && p->immd_nr) {
+		struct nouveau_resource *heap = nv50->screen->immd_heap[0];
 
-		if (nvws->res_alloc(heap, nr, p, &p->data)) {
-			while (heap->next && heap->size < nr) {
+		if (nouveau_resource_alloc(heap, p->immd_nr, p, &p->data[0])) {
+			while (heap->next && heap->size < p->immd_nr) {
 				struct nv50_program *evict = heap->next->priv;
-				nvws->res_free(&evict->data);
+				nouveau_resource_free(&evict->data[0]);
 			}
 
-			if (nvws->res_alloc(heap, nr, p, &p->data))
+			if (nouveau_resource_alloc(heap, p->immd_nr, p,
+						   &p->data[0]))
 				assert(0);
 		}
+
+		/* immediates only need to be uploaded again when freed */
+		nv50_program_upload_data(nv50, p->immd, p->data[0]->start,
+					 p->immd_nr, NV50_CB_PMISC);
 	}
 
-	if (p->param_nr) {
-		float *map = ws->buffer_map(ws, nv50->constbuf[p->type],
-					    PIPE_BUFFER_USAGE_CPU_READ);
-		nv50_program_upload_data(nv50, map, p->data->start,
-					 p->param_nr);
-		ws->buffer_unmap(ws, nv50->constbuf[p->type]);
+	if (!p->data[1] && p->param_nr) {
+		struct nouveau_resource *heap =
+			nv50->screen->parm_heap[p->type];
+
+		if (nouveau_resource_alloc(heap, p->param_nr, p, &p->data[1])) {
+			while (heap->next && heap->size < p->param_nr) {
+				struct nv50_program *evict = heap->next->priv;
+				nouveau_resource_free(&evict->data[1]);
+			}
+
+			if (nouveau_resource_alloc(heap, p->param_nr, p,
+						   &p->data[1]))
+				assert(0);
+		}
 	}
 
-	if (p->immd_nr) {
-		nv50_program_upload_data(nv50, p->immd,
-					 p->data->start + p->param_nr,
-					 p->immd_nr);
+	if (p->param_nr) {
+		unsigned cbuf = NV50_CB_PVP;
+		float *map = pipe_buffer_map(pscreen, nv50->constbuf[p->type],
+					     PIPE_BUFFER_USAGE_CPU_READ);
+		if (p->type == PIPE_SHADER_FRAGMENT)
+			cbuf = NV50_CB_PFP;
+		nv50_program_upload_data(nv50, map, p->data[1]->start,
+					 p->param_nr, cbuf);
+		pipe_buffer_unmap(pscreen, nv50->constbuf[p->type]);
 	}
 }
 
 static void
 nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	struct pipe_screen *screen = nv50->pipe.screen;
 	struct nv50_program_exec *e;
 	struct nouveau_stateobj *so;
 	const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR;
 	unsigned start, count, *up, *ptr;
 	boolean upload = FALSE;
 
-	if (!p->buffer) {
-		p->buffer = screen->buffer_create(screen, 0x100, 0, p->exec_size * 4);
+	if (!p->bo) {
+		nouveau_bo_new(chan->device, NOUVEAU_BO_VRAM, 0x100,
+			       p->exec_size * 4, &p->bo);
 		upload = TRUE;
 	}
 
-	if (p->data && p->data->start != p->data_start) {
+	if ((p->data[0] && p->data[0]->start != p->data_start[0]) ||
+		(p->data[1] && p->data[1]->start != p->data_start[1])) {
 		for (e = p->exec_head; e; e = e->next) {
-			unsigned ei, ci;
+			unsigned ei, ci, bs;
 
 			if (e->param.index < 0)
 				continue;
+			bs = (e->inst[1] >> 22) & 0x07;
+			assert(bs < 2);
 			ei = e->param.shift >> 5;
-			ci = e->param.index + p->data->start;
+			ci = e->param.index + p->data[bs]->start;
 
 			e->inst[ei] &= ~e->param.mask;
 			e->inst[ei] |= (ci << e->param.shift);
 		}
 
-		p->data_start = p->data->start;
+		if (p->data[0])
+			p->data_start[0] = p->data[0]->start;
+		if (p->data[1])
+			p->data_start[1] = p->data[1]->start;
+
 		upload = TRUE;
 	}
 
@@ -1637,13 +2330,11 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 
 #ifdef NV50_PROGRAM_DUMP
 	NOUVEAU_ERR("-------\n");
-	up = ptr = MALLOC(p->exec_size * 4);
 	for (e = p->exec_head; e; e = e->next) {
 		NOUVEAU_ERR("0x%08x\n", e->inst[0]);
 		if (is_long(e))
 			NOUVEAU_ERR("0x%08x\n", e->inst[1]);
 	}
-
 #endif
 
 	up = ptr = MALLOC(p->exec_size * 4);
@@ -1655,20 +2346,20 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
 
 	so = so_new(4,2);
 	so_method(so, nv50->screen->tesla, 0x1280, 3);
-	so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_LOW, 0, 0);
 	so_data  (so, (NV50_CB_PUPLOAD << 16) | 0x0800); //(p->exec_size * 4));
 
 	start = 0; count = p->exec_size;
 	while (count) {
-		struct nouveau_winsys *nvws = nv50->screen->nvws;
+		struct nouveau_channel *chan = nv50->screen->base.channel;
 		unsigned nr;
 
-		so_emit(nvws, so);
+		so_emit(chan, so);
 
 		nr = MIN2(count, 2047);
-		nr = MIN2(nvws->channel->pushbuf->remaining, nr);
-		if (nvws->channel->pushbuf->remaining < (nr + 3)) {
+		nr = MIN2(chan->pushbuf->remaining, nr);
+		if (chan->pushbuf->remaining < (nr + 3)) {
 			FIRE_RING(chan);
 			continue;
 		}
@@ -1704,10 +2395,10 @@ nv50_vertprog_validate(struct nv50_context *nv50)
 
 	so = so_new(13, 2);
 	so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_LOW, 0, 0);
 	so_method(so, tesla, 0x1650, 2);
 	so_data  (so, p->cfg.vp.attr[0]);
 	so_data  (so, p->cfg.vp.attr[1]);
@@ -1728,6 +2419,7 @@ nv50_fragprog_validate(struct nv50_context *nv50)
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_program *p = nv50->fragprog;
 	struct nouveau_stateobj *so;
+	unsigned i;
 
 	if (!p->translated) {
 		nv50_program_validate(nv50, p);
@@ -1740,22 +2432,27 @@ nv50_fragprog_validate(struct nv50_context *nv50)
 
 	so = so_new(64, 2);
 	so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
-		  NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		      NOUVEAU_BO_LOW, 0, 0);
 	so_method(so, tesla, 0x1904, 4);
-	so_data  (so, 0x00040404); /* p: 0x01000404 */
+	so_data  (so, p->cfg.fp.regs[0]); /* 0x01000404 / 0x00040404 */
 	so_data  (so, 0x00000004);
 	so_data  (so, 0x00000000);
 	so_data  (so, 0x00000000);
-	so_method(so, tesla, 0x16bc, 3); /*XXX: fixme */
-	so_data  (so, 0x03020100);
-	so_data  (so, 0x07060504);
-	so_data  (so, 0x0b0a0908);
+	so_method(so, tesla, 0x16bc, p->cfg.fp.high_map);
+	for (i = 0; i < p->cfg.fp.high_map; i++)
+		so_data(so, p->cfg.fp.map[i]);
 	so_method(so, tesla, 0x1988, 2);
-	so_data  (so, 0x08080408); //0x08040404); /* p: 0x0f000401 */
+	so_data  (so, p->cfg.fp.regs[1]); /* 0x08040404 / 0x0f000401 */
 	so_data  (so, p->cfg.high_temp);
+	so_method(so, tesla, 0x1298, 1);
+	so_data  (so, p->cfg.high_result);
+	so_method(so, tesla, 0x19a8, 1);
+	so_data  (so, p->cfg.fp.regs[2]);
+	so_method(so, tesla, 0x196c, 1);
+	so_data  (so, p->cfg.fp.regs[3]);
 	so_method(so, tesla, 0x1414, 1);
 	so_data  (so, 0); /* program start offset */
 	so_ref(so, &nv50->state.fragprog);
@@ -1765,8 +2462,6 @@ nv50_fragprog_validate(struct nv50_context *nv50)
 void
 nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p)
 {
-	struct pipe_screen *pscreen = nv50->pipe.screen;
-
 	while (p->exec_head) {
 		struct nv50_program_exec *e = p->exec_head;
 
@@ -1776,10 +2471,10 @@ nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p)
 	p->exec_tail = NULL;
 	p->exec_size = 0;
 
-	if (p->buffer)
-		pipe_buffer_reference(&p->buffer, NULL);
+	nouveau_bo_ref(NULL, &p->bo);
 
-	nv50->screen->nvws->res_free(&p->data);
+	nouveau_resource_free(&p->data[0]);
+	nouveau_resource_free(&p->data[1]);
 
 	p->translated = 0;
 }
diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h
index 78deed6a38..096e0476aa 100644
--- a/src/gallium/drivers/nv50/nv50_program.h
+++ b/src/gallium/drivers/nv50/nv50_program.h
@@ -24,10 +24,10 @@ struct nv50_program {
 	struct nv50_program_exec *exec_head;
 	struct nv50_program_exec *exec_tail;
 	unsigned exec_size;
-	struct nouveau_resource *data;
-	unsigned data_start;
+	struct nouveau_resource *data[2];
+	unsigned data_start[2];
 
-	struct pipe_buffer *buffer;
+	struct nouveau_bo *bo;
 
 	float *immd;
 	unsigned immd_nr;
@@ -39,6 +39,11 @@ struct nv50_program {
 		struct {
 			unsigned attr[2];
 		} vp;
+		struct {
+			unsigned regs[4];
+			unsigned map[5];
+			unsigned high_map;
+		} fp;
 	} cfg;
 };
 
diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c
index 35cebdbdc3..940e04365f 100644
--- a/src/gallium/drivers/nv50/nv50_query.c
+++ b/src/gallium/drivers/nv50/nv50_query.c
@@ -26,7 +26,7 @@
 #include "nv50_context.h"
 
 struct nv50_query {
-	struct pipe_buffer *buffer;
+	struct nouveau_bo *bo;
 	unsigned type;
 	boolean ready;
 	uint64_t result;
@@ -41,14 +41,16 @@ nv50_query(struct pipe_query *pipe)
 static struct pipe_query *
 nv50_query_create(struct pipe_context *pipe, unsigned type)
 {
-	struct pipe_screen *screen = pipe->screen;
+	struct nouveau_device *dev = nouveau_screen(pipe->screen)->device;
 	struct nv50_query *q = CALLOC_STRUCT(nv50_query);
+	int ret;
 
 	assert (q->type == PIPE_QUERY_OCCLUSION_COUNTER);
 	q->type = type;
 
-	q->buffer = screen->buffer_create(screen, 256, 0, 16);
-	if (!q->buffer) {
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM | NOUVEAU_BO_MAP, 256,
+			     16, &q->bo);
+	if (ret) {
 		FREE(q);
 		return NULL;
 	}
@@ -62,7 +64,7 @@ nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
 	struct nv50_query *q = nv50_query(pq);
 
 	if (q) {
-		pipe_buffer_reference(&q->buffer, NULL);
+		nouveau_bo_ref(NULL, &q->bo);
 		FREE(q);
 	}
 }
@@ -71,7 +73,7 @@ static void
 nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_query *q = nv50_query(pq);
 
@@ -87,15 +89,14 @@ static void
 nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->base.channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nv50_query *q = nv50_query(pq);
-	struct nouveau_bo *bo = nv50->screen->nvws->get_bo(q->buffer);
 
 	WAIT_RING (chan, 5);
 	BEGIN_RING(chan, tesla, 0x1b00, 4);
-	OUT_RELOCh(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
-	OUT_RELOCl(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 	OUT_RING  (chan, 0x00000000);
 	OUT_RING  (chan, 0x0100f002);
 	FIRE_RING (chan);
@@ -105,7 +106,6 @@ static boolean
 nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 		  boolean wait, uint64_t *result)
 {
-	struct pipe_winsys *ws = pipe->winsys;
 	struct nv50_query *q = nv50_query(pq);
 
 	/*XXX: Want to be able to return FALSE here instead of blocking
@@ -113,11 +113,10 @@ nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq,
 	 */
 
 	if (!q->ready) {
-		uint32_t *map = ws->buffer_map(ws, q->buffer,
-					       PIPE_BUFFER_USAGE_CPU_READ);
-		q->result = map[1];
+		nouveau_bo_map(q->bo, NOUVEAU_BO_RD);
+		q->result = ((uint32_t *)q->bo->map)[1];
 		q->ready = TRUE;
-		ws->buffer_unmap(ws, q->buffer);
+		nouveau_bo_unmap(q->bo);
 	}
 
 	*result = q->result;
diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c
index 2980564594..ce8f906b15 100644
--- a/src/gallium/drivers/nv50/nv50_screen.c
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -22,8 +22,6 @@
 
 #include "pipe/p_screen.h"
 
-#include "util/u_simple_screen.h"
-
 #include "nv50_context.h"
 #include "nv50_screen.h"
 
@@ -39,7 +37,15 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 		switch (format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 		case PIPE_FORMAT_R5G6B5_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else
+	if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) {
+		switch (format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 		case PIPE_FORMAT_Z16_UNORM:
 			return TRUE;
 		default:
@@ -68,23 +74,6 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen,
 	return FALSE;
 }
 
-static const char *
-nv50_screen_get_name(struct pipe_screen *pscreen)
-{
-	struct nv50_screen *screen = nv50_screen(pscreen);
-	struct nouveau_device *dev = screen->nvws->channel->device;
-	static char buffer[128];
-
-	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
-	return buffer;
-}
-
-static const char *
-nv50_screen_get_vendor(struct pipe_screen *pscreen)
-{
-	return "nouveau";
-}
-
 static int
 nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 {
@@ -120,6 +109,10 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param)
 		return 1;
 	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
 		return 0;
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+		return 0;
+	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+		return 1;
 	case NOUVEAU_CAP_HW_VTXBUF:
 		return 1;
 	case NOUVEAU_CAP_HW_IDXBUF:
@@ -153,37 +146,64 @@ nv50_screen_get_paramf(struct pipe_screen *pscreen, int param)
 static void
 nv50_screen_destroy(struct pipe_screen *pscreen)
 {
-	FREE(pscreen);
+	struct nv50_screen *screen = nv50_screen(pscreen);
+
+	nouveau_notifier_free(&screen->sync);
+	nouveau_grobj_free(&screen->tesla);
+	nouveau_grobj_free(&screen->eng2d);
+	nouveau_grobj_free(&screen->m2mf);
+	nouveau_screen_fini(&screen->base);
+	FREE(screen);
 }
 
 struct pipe_screen *
-nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 {
 	struct nv50_screen *screen = CALLOC_STRUCT(nv50_screen);
+	struct nouveau_channel *chan;
+	struct pipe_screen *pscreen;
 	struct nouveau_stateobj *so;
-	unsigned tesla_class = 0, ret;
-	unsigned chipset = nvws->channel->device->chipset;
-	int i;
+	unsigned chipset = dev->chipset;
+	unsigned tesla_class = 0;
+	int ret, i;
 
 	if (!screen)
 		return NULL;
-	screen->nvws = nvws;
+	pscreen = &screen->base.base;
+
+	ret = nouveau_screen_init(&screen->base, dev);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+	chan = screen->base.channel;
+
+	pscreen->winsys = ws;
+	pscreen->destroy = nv50_screen_destroy;
+	pscreen->get_param = nv50_screen_get_param;
+	pscreen->get_paramf = nv50_screen_get_paramf;
+	pscreen->is_format_supported = nv50_screen_is_format_supported;
+
+	nv50_screen_init_miptree_functions(pscreen);
+	nv50_transfer_init_screen_functions(pscreen);
 
 	/* DMA engine object */
-	ret = nvws->grobj_alloc(nvws, 0x5039, &screen->m2mf);
+	ret = nouveau_grobj_alloc(chan, 0xbeef5039, 0x5039, &screen->m2mf);
 	if (ret) {
 		NOUVEAU_ERR("Error creating M2MF object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BIND_RING(chan, screen->m2mf, 1);
 
 	/* 2D object */
-	ret = nvws->grobj_alloc(nvws, NV50_2D, &screen->eng2d);
+	ret = nouveau_grobj_alloc(chan, 0xbeef502d, 0x502d, &screen->eng2d);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 2D object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BIND_RING(chan, screen->eng2d, 2);
 
 	/* 3D object */
 	switch (chipset & 0xf0) {
@@ -199,70 +219,55 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 		break;
 	default:
 		NOUVEAU_ERR("Not a known NV50 chipset: NV%02x\n", chipset);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
 	if (tesla_class == 0) {
 		NOUVEAU_ERR("Unknown G8x chipset: NV%02x\n", chipset);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
-	ret = nvws->grobj_alloc(nvws, tesla_class, &screen->tesla);
+	ret = nouveau_grobj_alloc(chan, 0xbeef5097, tesla_class, &screen->tesla);
 	if (ret) {
 		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
+	BIND_RING(chan, screen->tesla, 3);
 
 	/* Sync notifier */
-	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync);
 	if (ret) {
 		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
-		nv50_screen_destroy(&screen->pipe);
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
-        /* Setup the pipe */
-	screen->pipe.winsys = ws;
-
-	screen->pipe.destroy = nv50_screen_destroy;
-
-	screen->pipe.get_name = nv50_screen_get_name;
-	screen->pipe.get_vendor = nv50_screen_get_vendor;
-	screen->pipe.get_param = nv50_screen_get_param;
-	screen->pipe.get_paramf = nv50_screen_get_paramf;
-
-	screen->pipe.is_format_supported = nv50_screen_is_format_supported;
-
-	nv50_screen_init_miptree_functions(&screen->pipe);
-	nv50_transfer_init_screen_functions(&screen->pipe);
-	u_simple_screen_init(&screen->pipe);
-
 	/* Static M2MF init */
 	so = so_new(32, 0);
 	so_method(so, screen->m2mf, 0x0180, 3);
 	so_data  (so, screen->sync->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_emit(nvws, so);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_emit(chan, so);
 	so_ref (NULL, &so);
 
 	/* Static 2D init */
 	so = so_new(64, 0);
 	so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4);
 	so_data  (so, screen->sync->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
-	so_data  (so, screen->nvws->channel->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
+	so_data  (so, chan->vram->handle);
 	so_method(so, screen->eng2d, NV50_2D_OPERATION, 1);
 	so_data  (so, NV50_2D_OPERATION_SRCCOPY);
 	so_method(so, screen->eng2d, 0x0290, 1);
 	so_data  (so, 0);
 	so_method(so, screen->eng2d, 0x0888, 1);
 	so_data  (so, 1);
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref(NULL, &so);
 
 	/* Static tesla init */
@@ -275,11 +280,11 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->tesla, NV50TCL_DMA_UNK0(0),
 				     NV50TCL_DMA_UNK0__SIZE);
 	for (i = 0; i < NV50TCL_DMA_UNK0__SIZE; i++)
-		so_data(so, nvws->channel->vram->handle);
+		so_data(so, chan->vram->handle);
 	so_method(so, screen->tesla, NV50TCL_DMA_UNK1(0),
 				     NV50TCL_DMA_UNK1__SIZE);
 	for (i = 0; i < NV50TCL_DMA_UNK1__SIZE; i++)
-		so_data(so, nvws->channel->vram->handle);
+		so_data(so, chan->vram->handle);
 	so_method(so, screen->tesla, 0x121c, 1);
 	so_data  (so, 1);
 
@@ -290,27 +295,81 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 	so_method(so, screen->tesla, 0x16b8, 1);
 	so_data  (so, 8);
 
-	/* Shared constant buffer */
-	screen->constbuf = screen->pipe.buffer_create(&screen->pipe, 0, 0, 128 * 4 * 4);
-	if (nvws->res_init(&screen->vp_data_heap, 0, 128)) {
-		NOUVEAU_ERR("Error initialising constant buffer\n");
-		nv50_screen_destroy(&screen->pipe);
+	/* constant buffers for immediates and VP/FP parameters */
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 128*4*4,
+			     &screen->constbuf_misc[0]);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+
+	for (i = 0; i < 2; i++) {
+		ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 128*4*4,
+				     &screen->constbuf_parm[i]);
+		if (ret) {
+			nv50_screen_destroy(pscreen);
+			return NULL;
+		}
+	}
+
+	if (nouveau_resource_init(&screen->immd_heap[0], 0, 128) ||
+		nouveau_resource_init(&screen->parm_heap[0], 0, 128) ||
+		nouveau_resource_init(&screen->parm_heap[1], 0, 128))
+	{
+		NOUVEAU_ERR("Error initialising constant buffers.\n");
+		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
 
+	/*
+	// map constant buffers:
+	//  B = buffer ID (maybe more than 1 byte)
+	//  N = CB index used in shader instruction
+	//  P = program type (0 = VP, 2 = GP, 3 = FP)
+	so_method(so, screen->tesla, 0x1694, 1);
+	so_data  (so, 0x000BBNP1);
+	*/
+
+	so_method(so, screen->tesla, 0x1280, 3);
+	so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_PMISC << 16) | 0x00000800);
+	so_method(so, screen->tesla, 0x1694, 1);
+	so_data  (so, 0x00000001 | (NV50_CB_PMISC << 12));
+	so_method(so, screen->tesla, 0x1694, 1);
+	so_data  (so, 0x00000031 | (NV50_CB_PMISC << 12));
+
 	so_method(so, screen->tesla, 0x1280, 3);
-	so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM |
+	so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-	so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM |
+	so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
-	so_data  (so, (NV50_CB_PMISC << 16) | 0x00001000);
+	so_data  (so, (NV50_CB_PVP << 16) | 0x00000800);
+	so_method(so, screen->tesla, 0x1694, 1);
+	so_data  (so, 0x00000101 | (NV50_CB_PVP << 12));
+
+	so_method(so, screen->tesla, 0x1280, 3);
+	so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_PFP << 16) | 0x00000800);
+	so_method(so, screen->tesla, 0x1694, 1);
+	so_data  (so, 0x00000131 | (NV50_CB_PFP << 12));
 
 	/* Texture sampler/image unit setup - we abuse the constant buffer
 	 * upload mechanism for the moment to upload data to the tex config
 	 * blocks.  At some point we *may* want to go the NVIDIA way of doing
 	 * things?
 	 */
-	screen->tic = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4);
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 32*8*4, &screen->tic);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+
 	so_method(so, screen->tesla, 0x1280, 3);
 	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
@@ -324,7 +383,12 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
 	so_data  (so, 0x00000800);
 
-	screen->tsc = screen->pipe.buffer_create(&screen->pipe, 0, 0, 32 * 8 * 4);
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, 32*8*4, &screen->tsc);
+	if (ret) {
+		nv50_screen_destroy(pscreen);
+		return NULL;
+	}
+
 	so_method(so, screen->tesla, 0x1280, 3);
 	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
 		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
@@ -352,14 +416,12 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
 
 	so_method(so, screen->tesla, 0x1234, 1);
 	so_data  (so, 1);
-	so_method(so, screen->tesla, 0x1458, 1);
-	so_data  (so, 1);
 
-	so_emit(nvws, so);
+	so_emit(chan, so);
 	so_ref (so, &screen->static_init);
 	so_ref (NULL, &so);
-	nvws->push_flush(nvws, 0, NULL);
+	nouveau_pushbuf_flush(chan, 0);
 
-	return &screen->pipe;
+	return pscreen;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h
index db567aaac8..61e24a5b57 100644
--- a/src/gallium/drivers/nv50/nv50_screen.h
+++ b/src/gallium/drivers/nv50/nv50_screen.h
@@ -1,10 +1,10 @@
 #ifndef __NV50_SCREEN_H__
 #define __NV50_SCREEN_H__
 
-#include "pipe/p_screen.h"
+#include "nouveau/nouveau_screen.h"
 
 struct nv50_screen {
-	struct pipe_screen pipe;
+	struct nouveau_screen base;
 
 	struct nouveau_winsys *nvws;
 
@@ -15,11 +15,14 @@ struct nv50_screen {
 	struct nouveau_grobj *m2mf;
 	struct nouveau_notifier *sync;
 
-	struct pipe_buffer *constbuf;
-	struct nouveau_resource *vp_data_heap;
+	struct nouveau_bo *constbuf_misc[1];
+	struct nouveau_bo *constbuf_parm[2];
 
-	struct pipe_buffer *tic;
-	struct pipe_buffer *tsc;
+	struct nouveau_resource *immd_heap[1];
+	struct nouveau_resource *parm_heap[2];
+
+	struct nouveau_bo *tic;
+	struct nouveau_bo *tsc;
 
 	struct nouveau_stateobj *static_init;
 };
diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c
index ba852194cd..116866a8e7 100644
--- a/src/gallium/drivers/nv50/nv50_state.c
+++ b/src/gallium/drivers/nv50/nv50_state.c
@@ -136,9 +136,11 @@ static void *
 nv50_sampler_state_create(struct pipe_context *pipe,
 			  const struct pipe_sampler_state *cso)
 {
-	unsigned *tsc = CALLOC(8, sizeof(unsigned));
+	struct nv50_sampler_stateobj *sso = CALLOC(1, sizeof(*sso));
+	unsigned *tsc = sso->tsc;
+	float limit;
 
-	tsc[0] = (0x00024000 |
+	tsc[0] = (0x00026000 |
 		  (wrap_mode(cso->wrap_s) << 0) |
 		  (wrap_mode(cso->wrap_t) << 3) |
 		  (wrap_mode(cso->wrap_r) << 6));
@@ -202,7 +204,14 @@ nv50_sampler_state_create(struct pipe_context *pipe,
 		tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7);
 	}
 
-	return (void *)tsc;
+	limit = CLAMP(cso->lod_bias, -16.0, 15.0);
+	tsc[1] |= ((int)(limit * 256.0) & 0x1fff) << 11;
+
+	tsc[2] |= ((int)CLAMP(cso->max_lod, 0.0, 15.0) << 20) |
+		  ((int)CLAMP(cso->min_lod, 0.0, 15.0) << 8);
+
+	sso->normalized = cso->normalized_coords;
+	return (void *)sso;
 }
 
 static void
diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c
index c13d3de1cb..d313e9de4f 100644
--- a/src/gallium/drivers/nv50/nv50_state_validate.c
+++ b/src/gallium/drivers/nv50/nv50_state_validate.c
@@ -32,6 +32,9 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 	unsigned i, w, h, gw = 0;
 
 	for (i = 0; i < fb->nr_cbufs; i++) {
+		struct pipe_texture *pt = fb->cbufs[i]->texture;
+		struct nouveau_bo *bo = nv50_miptree(pt)->bo;
+
 		if (!gw) {
 			w = fb->cbufs[i]->width;
 			h = fb->cbufs[i]->height;
@@ -46,12 +49,10 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 		so_data  (so, fb->cbufs[i]->height);
 
 		so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5);
-		so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
-			  NOUVEAU_BO_RDWR, 0, 0);
-		so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
-			  NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->cbufs[i]->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0);
 		switch (fb->cbufs[i]->format) {
 		case PIPE_FORMAT_A8R8G8B8_UNORM:
 			so_data(so, 0xcf);
@@ -65,7 +66,7 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 			so_data(so, 0xe6);
 			break;
 		}
-		so_data(so, 0x00000000);
+		so_data(so, bo->tile_mode << 4);
 		so_data(so, 0x00000000);
 
 		so_method(so, tesla, 0x1224, 1);
@@ -73,6 +74,9 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 	}
 
 	if (fb->zsbuf) {
+		struct pipe_texture *pt = fb->zsbuf->texture;
+		struct nouveau_bo *bo = nv50_miptree(pt)->bo;
+
 		if (!gw) {
 			w = fb->zsbuf->width;
 			h = fb->zsbuf->height;
@@ -83,14 +87,13 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 		}
 
 		so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5);
-		so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
-			  NOUVEAU_BO_RDWR, 0, 0);
-		so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset,
-			  NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
-			  NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_HIGH | NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM |
+			      NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0);
 		switch (fb->zsbuf->format) {
 		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z24X8_UNORM:
 			so_data(so, 0x16);
 			break;
 		case PIPE_FORMAT_Z16_UNORM:
@@ -102,7 +105,7 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 			so_data(so, 0x16);
 			break;
 		}
-		so_data(so, 0x00000000);
+		so_data(so, bo->tile_mode << 4);
 		so_data(so, 0x00000000);
 
 		so_method(so, tesla, 0x1538, 1);
@@ -131,7 +134,7 @@ static void
 nv50_state_emit(struct nv50_context *nv50)
 {
 	struct nv50_screen *screen = nv50->screen;
-	struct nouveau_winsys *nvws = screen->nvws;
+	struct nouveau_channel *chan = screen->base.channel;
 
 	if (nv50->pctx_id != screen->cur_pctx) {
 		nv50->state.dirty |= 0xffffffff;
@@ -139,40 +142,40 @@ nv50_state_emit(struct nv50_context *nv50)
 	}
 
 	if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER)
-		so_emit(nvws, nv50->state.fb);
+		so_emit(chan, nv50->state.fb);
 	if (nv50->state.dirty & NV50_NEW_BLEND)
-		so_emit(nvws, nv50->state.blend);
+		so_emit(chan, nv50->state.blend);
 	if (nv50->state.dirty & NV50_NEW_ZSA)
-		so_emit(nvws, nv50->state.zsa);
+		so_emit(chan, nv50->state.zsa);
 	if (nv50->state.dirty & NV50_NEW_VERTPROG)
-		so_emit(nvws, nv50->state.vertprog);
+		so_emit(chan, nv50->state.vertprog);
 	if (nv50->state.dirty & NV50_NEW_FRAGPROG)
-		so_emit(nvws, nv50->state.fragprog);
+		so_emit(chan, nv50->state.fragprog);
 	if (nv50->state.dirty & NV50_NEW_RASTERIZER)
-		so_emit(nvws, nv50->state.rast);
+		so_emit(chan, nv50->state.rast);
 	if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR)
-		so_emit(nvws, nv50->state.blend_colour);
+		so_emit(chan, nv50->state.blend_colour);
 	if (nv50->state.dirty & NV50_NEW_STIPPLE)
-		so_emit(nvws, nv50->state.stipple);
+		so_emit(chan, nv50->state.stipple);
 	if (nv50->state.dirty & NV50_NEW_SCISSOR)
-		so_emit(nvws, nv50->state.scissor);
+		so_emit(chan, nv50->state.scissor);
 	if (nv50->state.dirty & NV50_NEW_VIEWPORT)
-		so_emit(nvws, nv50->state.viewport);
+		so_emit(chan, nv50->state.viewport);
 	if (nv50->state.dirty & NV50_NEW_SAMPLER)
-		so_emit(nvws, nv50->state.tsc_upload);
+		so_emit(chan, nv50->state.tsc_upload);
 	if (nv50->state.dirty & NV50_NEW_TEXTURE)
-		so_emit(nvws, nv50->state.tic_upload);
+		so_emit(chan, nv50->state.tic_upload);
 	if (nv50->state.dirty & NV50_NEW_ARRAYS) {
-		so_emit(nvws, nv50->state.vtxfmt);
-		so_emit(nvws, nv50->state.vtxbuf);
+		so_emit(chan, nv50->state.vtxfmt);
+		so_emit(chan, nv50->state.vtxbuf);
 	}
 	nv50->state.dirty = 0;
 
-	so_emit_reloc_markers(nvws, nv50->state.fb);
-	so_emit_reloc_markers(nvws, nv50->state.vertprog);
-	so_emit_reloc_markers(nvws, nv50->state.fragprog);
-	so_emit_reloc_markers(nvws, nv50->state.vtxbuf);
-	so_emit_reloc_markers(nvws, nv50->screen->static_init);
+	so_emit_reloc_markers(chan, nv50->state.fb);
+	so_emit_reloc_markers(chan, nv50->state.vertprog);
+	so_emit_reloc_markers(chan, nv50->state.fragprog);
+	so_emit_reloc_markers(chan, nv50->state.vtxbuf);
+	so_emit_reloc_markers(chan, nv50->screen->static_init);
 }
 
 boolean
@@ -293,12 +296,12 @@ viewport_uptodate:
 		so_data  (so, NV50_CB_TSC);
 		so_method(so, tesla, 0x40000f04, nv50->sampler_nr * 8);
 		for (i = 0; i < nv50->sampler_nr; i++)
-			so_datap (so, nv50->sampler[i], 8);
+			so_datap (so, nv50->sampler[i]->tsc, 8);
 		so_ref(so, &nv50->state.tsc_upload);
 		so_ref(NULL, &so);
 	}
 
-	if (nv50->dirty & NV50_NEW_TEXTURE)
+	if (nv50->dirty & (NV50_NEW_TEXTURE | NV50_NEW_SAMPLER))
 		nv50_tex_validate(nv50);
 
 	if (nv50->dirty & NV50_NEW_ARRAYS)
diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c
index 0cc5168144..3da9d6e728 100644
--- a/src/gallium/drivers/nv50/nv50_surface.c
+++ b/src/gallium/drivers/nv50/nv50_surface.c
@@ -35,7 +35,6 @@ nv50_format(enum pipe_format format)
 {
 	switch (format) {
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
-	case PIPE_FORMAT_Z24S8_UNORM:
 		return NV50_2D_DST_FORMAT_32BPP;
 	case PIPE_FORMAT_X8R8G8B8_UNORM:
 		return NV50_2D_DST_FORMAT_24BPP;
@@ -52,21 +51,17 @@ static int
 nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
 {
 	struct nv50_miptree *mt = nv50_miptree(ps->texture);
-	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_channel *chan = screen->eng2d->channel;
 	struct nouveau_grobj *eng2d = screen->eng2d;
-	struct nouveau_bo *bo;
+	struct nouveau_bo *bo = nv50_miptree(ps->texture)->bo;
  	int format, mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT;
  	int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD);
- 
-	bo = screen->nvws->get_bo(nv50_miptree(ps->texture)->buffer);
-	if (!bo)
-		return 1;
 
  	format = nv50_format(ps->format);
  	if (format < 0)
  		return 1;
   
- 	if (!bo->tiled) {
+ 	if (!bo->tile_flags) {
  		BEGIN_RING(chan, eng2d, mthd, 2);
  		OUT_RING  (chan, format);
  		OUT_RING  (chan, 1);
@@ -80,7 +75,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
  		BEGIN_RING(chan, eng2d, mthd, 5);
  		OUT_RING  (chan, format);
  		OUT_RING  (chan, 0);
- 		OUT_RING  (chan, 0);
+ 		OUT_RING  (chan, bo->tile_mode << 4);
  		OUT_RING  (chan, 1);
  		OUT_RING  (chan, 0);
  		BEGIN_RING(chan, eng2d, mthd + 0x18, 4);
@@ -108,7 +103,7 @@ nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst,
 		     int dx, int dy, struct pipe_surface *src, int sx, int sy,
 		     int w, int h)
 {
-	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_channel *chan = screen->eng2d->channel;
 	struct nouveau_grobj *eng2d = screen->eng2d;
 	int ret;
 
@@ -165,7 +160,7 @@ nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
 {
 	struct nv50_context *nv50 = (struct nv50_context *)pipe;
 	struct nv50_screen *screen = nv50->screen;
-	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_channel *chan = screen->eng2d->channel;
 	struct nouveau_grobj *eng2d = screen->eng2d;
 	int format, ret;
 
diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c
index 223c8a3a45..ff40c2ad81 100644
--- a/src/gallium/drivers/nv50/nv50_tex.c
+++ b/src/gallium/drivers/nv50/nv50_tex.c
@@ -26,7 +26,8 @@
 #include "nouveau/nouveau_stateobj.h"
 
 static int
-nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt)
+nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so,
+		   struct nv50_miptree *mt, int unit)
 {
 	switch (mt->base.format) {
 	case PIPE_FORMAT_A8R8G8B8_UNORM:
@@ -117,15 +118,18 @@ nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt)
 		return 1;
 	}
 
-	so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
+	so_reloc(so, mt->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
 		     NOUVEAU_BO_RD, 0, 0);
-	so_data (so, 0xd0005000);
+	if (nv50->sampler[unit]->normalized)
+		so_data (so, 0xd0005000 | mt->bo->tile_mode << 22);
+	else
+		so_data (so, 0x5001d000 | mt->bo->tile_mode << 22);
 	so_data (so, 0x00300000);
 	so_data (so, mt->base.width[0]);
-	so_data (so, (mt->base.depth[0] << 16) | mt->base.height[0]);
+	so_data (so, (mt->base.last_level << 28) |
+		     (mt->base.depth[0] << 16) | mt->base.height[0]);
 	so_data (so, 0x03000000);
-	so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
-		     NOUVEAU_BO_RD, 0, 0);
+	so_data (so, mt->base.last_level << 4);
 
 	return 0;
 }
@@ -135,23 +139,35 @@ nv50_tex_validate(struct nv50_context *nv50)
 {
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 	struct nouveau_stateobj *so;
-	int unit;
+	int unit, push;
+
+	push  = nv50->miptree_nr * 9 + 2;
+	push += MAX2(nv50->miptree_nr, nv50->state.miptree_nr) * 2;
 
-	so = so_new(nv50->miptree_nr * 8 + 3, nv50->miptree_nr * 2);
+	so = so_new(push, nv50->miptree_nr * 2);
 	so_method(so, tesla, 0x0f00, 1);
 	so_data  (so, NV50_CB_TIC);
-	so_method(so, tesla, 0x40000f04, nv50->miptree_nr * 8);
 	for (unit = 0; unit < nv50->miptree_nr; unit++) {
 		struct nv50_miptree *mt = nv50->miptree[unit];
 
-		if (nv50_tex_construct(so, mt)) {
+		so_method(so, tesla, 0x40000f04, 8);
+		if (nv50_tex_construct(nv50, so, mt, unit)) {
 			NOUVEAU_ERR("failed tex validate\n");
 			so_ref(NULL, &so);
 			return;
 		}
+
+		so_method(so, tesla, 0x1458, 1);
+		so_data  (so, (unit << 9) | (unit << 1) | 1);
+	}
+
+	for (; unit < nv50->state.miptree_nr; unit++) {
+		so_method(so, tesla, 0x1458, 1);
+		so_data  (so, (unit << 1) | 0);
 	}
 
 	so_ref(so, &nv50->state.tic_upload);
 	so_ref(NULL, &so);
+	nv50->state.miptree_nr = nv50->miptree_nr;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c
index 747195b4f6..d0b7f0bef4 100644
--- a/src/gallium/drivers/nv50/nv50_transfer.c
+++ b/src/gallium/drivers/nv50/nv50_transfer.c
@@ -6,8 +6,8 @@
 
 struct nv50_transfer {
 	struct pipe_transfer base;
-	struct pipe_buffer *buffer;
-	struct nv50_miptree_level *level;
+	struct nouveau_bo *bo;
+	unsigned level_offset;
 	int level_pitch;
 	int level_width;
 	int level_height;
@@ -16,51 +16,48 @@ struct nv50_transfer {
 };
 
 static void
-nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src,
-			int src_pitch, int sx, int sy, int sw, int sh,
-			struct pipe_buffer *dst, int dst_pitch, int dx, int dy,
+nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo,
+			unsigned src_offset, int src_pitch, int sx, int sy,
+			int sw, int sh, struct nouveau_bo *dst_bo,
+			unsigned dst_offset, int dst_pitch, int dx, int dy,
 			int dw, int dh, int cpp, int width, int height,
 			unsigned src_reloc, unsigned dst_reloc)
 {
 	struct nv50_screen *screen = nv50_screen(pscreen);
-	struct nouveau_winsys *nvws = screen->nvws;
-	struct nouveau_channel *chan = nvws->channel;
+	struct nouveau_channel *chan = screen->m2mf->channel;
 	struct nouveau_grobj *m2mf = screen->m2mf;
-	struct nouveau_bo *src_bo = nvws->get_bo(src);
-	struct nouveau_bo *dst_bo = nvws->get_bo(dst);
-	unsigned src_offset = 0, dst_offset = 0;
 
 	src_reloc |= NOUVEAU_BO_RD;
 	dst_reloc |= NOUVEAU_BO_WR;
 
 	WAIT_RING (chan, 14);
 
-	if (!src_bo->tiled) {
+	if (!src_bo->tile_flags) {
 		BEGIN_RING(chan, m2mf, 0x0200, 1);
 		OUT_RING  (chan, 1);
 		BEGIN_RING(chan, m2mf, 0x0314, 1);
 		OUT_RING  (chan, src_pitch);
-		src_offset = (sy * src_pitch) + (sx * cpp);
+		src_offset += (sy * src_pitch) + (sx * cpp);
 	} else {
 		BEGIN_RING(chan, m2mf, 0x0200, 6);
 		OUT_RING  (chan, 0);
-		OUT_RING  (chan, 0);
+		OUT_RING  (chan, src_bo->tile_mode << 4);
 		OUT_RING  (chan, sw * cpp);
 		OUT_RING  (chan, sh);
 		OUT_RING  (chan, 1);
 		OUT_RING  (chan, 0);
 	}
 
-	if (!dst_bo->tiled) {
+	if (!dst_bo->tile_flags) {
 		BEGIN_RING(chan, m2mf, 0x021c, 1);
 		OUT_RING  (chan, 1);
 		BEGIN_RING(chan, m2mf, 0x0318, 1);
 		OUT_RING  (chan, dst_pitch);
-		dst_offset = (dy * dst_pitch) + (dx * cpp);
+		dst_offset += (dy * dst_pitch) + (dx * cpp);
 	} else {
 		BEGIN_RING(chan, m2mf, 0x021c, 6);
 		OUT_RING  (chan, 0);
-		OUT_RING  (chan, 0);
+		OUT_RING  (chan, dst_bo->tile_mode << 4);
 		OUT_RING  (chan, dw * cpp);
 		OUT_RING  (chan, dh);
 		OUT_RING  (chan, 1);
@@ -77,13 +74,13 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct pipe_buffer *src,
 		BEGIN_RING(chan, m2mf, 0x030c, 2);
 		OUT_RELOCl(chan, src_bo, src_offset, src_reloc);
 		OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc);
-		if (src_bo->tiled) {
+		if (src_bo->tile_flags) {
 			BEGIN_RING(chan, m2mf, 0x0218, 1);
 			OUT_RING  (chan, (dy << 16) | sx);
 		} else {
 			src_offset += (line_count * src_pitch);
 		}
-		if (dst_bo->tiled) {
+		if (dst_bo->tile_flags) {
 			BEGIN_RING(chan, m2mf, 0x0234, 1);
 			OUT_RING  (chan, (sy << 16) | dx);
 		} else {
@@ -108,10 +105,12 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 		  enum pipe_transfer_usage usage,
 		  unsigned x, unsigned y, unsigned w, unsigned h)
 {
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
 	struct nv50_miptree *mt = nv50_miptree(pt);
 	struct nv50_miptree_level *lvl = &mt->level[level];
 	struct nv50_transfer *tx;
 	unsigned image = 0;
+	int ret;
 
 	if (pt->target == PIPE_TEXTURE_CUBE)
 		image = face;
@@ -133,20 +132,24 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
 	tx->base.stride = (w * pt->block.size);
 	tx->base.usage = usage;
 
-	tx->level = lvl;
 	tx->level_pitch = lvl->pitch;
 	tx->level_width = mt->base.width[level];
 	tx->level_height = mt->base.height[level];
+	tx->level_offset = lvl->image_offset[image];
 	tx->level_x = x;
 	tx->level_y = y;
-	tx->buffer =
-		pipe_buffer_create(pscreen, 0, NOUVEAU_BUFFER_USAGE_TRANSFER,
-				   w * tx->base.block.size * h);
+	ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0,
+			     w * pt->block.size * h, &tx->bo);
+	if (ret) {
+		FREE(tx);
+		return NULL;
+	}
 
 	if (usage != PIPE_TRANSFER_WRITE) {
-		nv50_transfer_rect_m2mf(pscreen, mt->buffer, tx->level_pitch,
-					x, y, tx->level_width, tx->level_height,
-					tx->buffer, tx->base.stride, 0, 0,
+		nv50_transfer_rect_m2mf(pscreen, mt->bo, tx->level_offset,
+					tx->level_pitch, x, y, tx->level_width,
+					tx->level_height, tx->bo, 0,
+					tx->base.stride, 0, 0,
 					tx->base.width, tx->base.height,
 					tx->base.block.size, w, h,
 					NOUVEAU_BO_VRAM | NOUVEAU_BO_GART,
@@ -164,17 +167,18 @@ nv50_transfer_del(struct pipe_transfer *ptx)
 
 	if (ptx->usage != PIPE_TRANSFER_READ) {
 		struct pipe_screen *pscreen = ptx->texture->screen;
-		nv50_transfer_rect_m2mf(pscreen, tx->buffer, tx->base.stride,
+		nv50_transfer_rect_m2mf(pscreen, tx->bo, 0, tx->base.stride,
 					0, 0, tx->base.width, tx->base.height,
-					mt->buffer, tx->level_pitch,
-					tx->level_x, tx->level_y,
-					tx->level_width, tx->level_height,
-					tx->base.block.size, tx->base.width,
-					tx->base.height, NOUVEAU_BO_GART,
-					NOUVEAU_BO_VRAM | NOUVEAU_BO_GART);
+					mt->bo, tx->level_offset,
+					tx->level_pitch, tx->level_x,
+					tx->level_y, tx->level_width,
+					tx->level_height, tx->base.block.size,
+					tx->base.width, tx->base.height,
+					NOUVEAU_BO_GART, NOUVEAU_BO_VRAM |
+					NOUVEAU_BO_GART);
 	}
 
-	pipe_buffer_reference(&tx->buffer, NULL);
+	nouveau_bo_ref(NULL, &tx->bo);
 	pipe_texture_reference(&ptx->texture, NULL);
 	FREE(ptx);
 }
@@ -184,13 +188,17 @@ nv50_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 {
 	struct nv50_transfer *tx = (struct nv50_transfer *)ptx;
 	unsigned flags = 0;
+	int ret;
 
 	if (ptx->usage & PIPE_TRANSFER_WRITE)
-		flags |= PIPE_BUFFER_USAGE_CPU_WRITE;
+		flags |= NOUVEAU_BO_WR;
 	if (ptx->usage & PIPE_TRANSFER_READ)
-		flags |= PIPE_BUFFER_USAGE_CPU_READ;
+		flags |= NOUVEAU_BO_RD;
 
-	return pipe_buffer_map(pscreen, tx->buffer, flags);
+	ret = nouveau_bo_map(tx->bo, flags);
+	if (ret)
+		return NULL;
+	return tx->bo->map;
 }
 
 static void
@@ -198,7 +206,7 @@ nv50_transfer_unmap(struct pipe_screen *pscreen, struct pipe_transfer *ptx)
 {
 	struct nv50_transfer *tx = (struct nv50_transfer *)ptx;
 
-	pipe_buffer_unmap(pscreen, tx->buffer);
+	nouveau_bo_unmap(tx->bo);
 }
 
 void
diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c
index 0749c90691..f81929f238 100644
--- a/src/gallium/drivers/nv50/nv50_vbo.c
+++ b/src/gallium/drivers/nv50/nv50_vbo.c
@@ -22,6 +22,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
 
 #include "nv50_context.h"
 
@@ -53,7 +54,7 @@ nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start,
 		 unsigned count)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	nv50_state_validate(nv50);
@@ -83,7 +84,7 @@ static INLINE void
 nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map,
 			      unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	map += start;
@@ -112,7 +113,7 @@ static INLINE void
 nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map,
 			      unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	map += start;
@@ -141,7 +142,7 @@ static INLINE void
 nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint8_t *map,
 			      unsigned start, unsigned count)
 {
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
 
 	map += start;
@@ -163,10 +164,12 @@ nv50_draw_elements(struct pipe_context *pipe,
 		   unsigned mode, unsigned start, unsigned count)
 {
 	struct nv50_context *nv50 = nv50_context(pipe);
-	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_channel *chan = nv50->screen->tesla->channel;
 	struct nouveau_grobj *tesla = nv50->screen->tesla;
-	struct pipe_winsys *ws = pipe->winsys;
-	void *map = ws->buffer_map(ws, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ);
+	struct pipe_screen *pscreen = pipe->screen;
+	void *map;
+	
+	map = pipe_buffer_map(pscreen, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ);
 
 	nv50_state_validate(nv50);
 
@@ -193,6 +196,7 @@ nv50_draw_elements(struct pipe_context *pipe,
 	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
 	OUT_RING  (chan, 0);
 
+	pipe_buffer_unmap(pscreen, indexBuffer);
 	pipe->flush(pipe, 0, NULL);
 	return TRUE;
 }
@@ -212,6 +216,7 @@ nv50_vbo_validate(struct nv50_context *nv50)
 		struct pipe_vertex_element *ve = &nv50->vtxelt[i];
 		struct pipe_vertex_buffer *vb =
 			&nv50->vtxbuf[ve->vertex_buffer_index];
+		struct nouveau_bo *bo = nouveau_bo(vb->buffer);
 
 		switch (ve->src_format) {
 		case PIPE_FORMAT_R32G32B32A32_FLOAT:
@@ -240,10 +245,10 @@ nv50_vbo_validate(struct nv50_context *nv50)
 
 		so_method(vtxbuf, tesla, 0x900 + (i * 16), 3);
 		so_data  (vtxbuf, 0x20000000 | vb->stride);
-		so_reloc (vtxbuf, vb->buffer, vb->buffer_offset +
+		so_reloc (vtxbuf, bo, vb->buffer_offset +
 			  ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
 			  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
-		so_reloc (vtxbuf, vb->buffer, vb->buffer_offset +
+		so_reloc (vtxbuf, bo, vb->buffer_offset +
 			  ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
 			  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
 	}
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
index e44f9b9dfc..faceec9842 100644
--- a/src/gallium/drivers/r300/Makefile
+++ b/src/gallium/drivers/r300/Makefile
@@ -4,20 +4,22 @@ include $(TOP)/configs/current
 LIBNAME = r300
 
 C_SOURCES = \
+	r3xx_fs.c \
+	r5xx_fs.c \
 	r300_chipset.c \
 	r300_clear.c \
 	r300_context.c \
 	r300_debug.c \
 	r300_emit.c \
 	r300_flush.c \
+	r300_fs.c \
 	r300_query.c \
 	r300_render.c \
 	r300_screen.c \
 	r300_state.c \
 	r300_state_derived.c \
 	r300_state_invariant.c \
-	r300_state_shader.c \
-	r300_state_tcl.c \
+	r300_vs.c \
 	r300_surface.c \
 	r300_texture.c
 
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
index 182ed2d459..493d7b28bc 100644
--- a/src/gallium/drivers/r300/SConscript
+++ b/src/gallium/drivers/r300/SConscript
@@ -5,20 +5,22 @@ env = env.Clone()
 r300 = env.ConvenienceLibrary(
     target = 'r300',
     source = [
+        'r3xx_fs.c',
+        'r5xx_fs.c',
         'r300_chipset.c',
         'r300_clear.c',
         'r300_context.c',
         'r300_debug.c',
         'r300_emit.c',
         'r300_flush.c',
+        'r300_fs.c',
         'r300_query.c',
         'r300_render.c',
         'r300_screen.c',
         'r300_state.c',
         'r300_state_derived.c',
         'r300_state_invariant.c',
-        'r300_state_shader.c',
-        'r300_state_tcl.c',
+        'r300_vs.c',
         'r300_surface.c',
         'r300_texture.c',
     ])
diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c
index 9d95ad918c..00fae8d26f 100644
--- a/src/gallium/drivers/r300/r300_chipset.c
+++ b/src/gallium/drivers/r300/r300_chipset.c
@@ -34,7 +34,6 @@ void r300_parse_chipset(struct r300_capabilities* caps)
     caps->is_r500 = FALSE;
     caps->num_vert_fpus = 4;
 
-
     /* Note: These are not ordered by PCI ID. I leave that task to GCC,
      * which will perform the ordering while collating jump tables. Instead,
      * I've tried to group them according to capabilities and age. */
@@ -150,6 +149,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->num_vert_fpus = 6;
             break;
 
+        case 0x4B48:
         case 0x4B49:
         case 0x4B4A:
         case 0x4B4B:
@@ -349,7 +349,4 @@ void r300_parse_chipset(struct r300_capabilities* caps)
                 caps->pci_id);
             break;
     }
-
-    /* XXX SW TCL is broken so no forcing it off right now
-    caps->has_tcl = FALSE; */
 }
diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h
index 21eebeae60..5b2e1f0568 100644
--- a/src/gallium/drivers/r300/r300_chipset.h
+++ b/src/gallium/drivers/r300/r300_chipset.h
@@ -34,8 +34,6 @@ struct r300_capabilities {
     int family;
     /* The number of vertex floating-point units */
     int num_vert_fpus;
-    /* The number of fragment pipes */
-    int num_frag_pipes;
     /* Whether or not TCL is physically present */
     boolean has_tcl;
     /* Whether or not this is an RV515 or newer; R500s have many differences
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 6bdf544a05..233a32b53c 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -34,10 +34,6 @@ static boolean r300_draw_range_elements(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
     int i;
 
-    if (r300->dirty_state) {
-        r300_emit_dirty_state(r300);
-    }
-
     for (i = 0; i < r300->vertex_buffer_count; i++) {
         void* buf = pipe_buffer_map(pipe->screen,
                                     r300->vertex_buffers[i].buffer,
@@ -133,7 +129,6 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     if (!r300)
         return NULL;
 
-    /* XXX this could be refactored now? */
     r300->winsys = r300_winsys;
 
     r300->context.winsys = (struct pipe_winsys*)r300_winsys;
@@ -150,8 +145,15 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->context.is_texture_referenced = r300_is_texture_referenced;
     r300->context.is_buffer_referenced = r300_is_buffer_referenced;
 
+    /* Create a Draw. This is used for vert collation and SW TCL. */
     r300->draw = draw_create();
+    /* Enable our renderer. */
     draw_set_rasterize_stage(r300->draw, r300_draw_stage(r300));
+    /* Disable Draw's clipping if TCL is present. */
+    draw_set_driver_clipping(r300->draw, r300_screen(screen)->caps->has_tcl);
+    /* Force Draw to never do viewport transform, since (again) we can do
+     * transform in hardware, always. */
+    draw_set_viewport_state(r300->draw, &r300_viewport_identity);
 
     r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state);
     r300->rs_block = CALLOC_STRUCT(r300_rs_block);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 6f62998b35..d891fd6265 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -63,6 +63,11 @@ struct r300_rs_state {
     /* Draw-specific rasterizer state */
     struct pipe_rasterizer_state rs;
 
+    /* Whether or not to enable the VTE. This is referenced at the very
+     * last moment during emission of VTE state, to decide whether or not
+     * the VTE should be used for transformation. */
+    boolean enable_vte;
+
     uint32_t vap_control_status;    /* R300_VAP_CNTL_STATUS: 0x2140 */
     uint32_t point_size;            /* R300_GA_POINT_SIZE: 0x421c */
     uint32_t point_minmax;          /* R300_GA_POINT_MINMAX: 0x4230 */
@@ -112,23 +117,24 @@ struct r300_viewport_state {
     uint32_t vte_control; /* R300_VAP_VTE_CNTL:      0x20b0 */
 };
 
-#define R300_NEW_BLEND           0x0000001
-#define R300_NEW_BLEND_COLOR     0x0000002
-#define R300_NEW_CONSTANTS       0x0000004
-#define R300_NEW_DSA             0x0000008
-#define R300_NEW_FRAMEBUFFERS    0x0000010
-#define R300_NEW_FRAGMENT_SHADER 0x0000020
-#define R300_NEW_RASTERIZER      0x0000040
-#define R300_NEW_RS_BLOCK        0x0000080
-#define R300_NEW_SAMPLER         0x0000100
-#define R300_ANY_NEW_SAMPLERS    0x000ff00
-#define R300_NEW_SCISSOR         0x0010000
-#define R300_NEW_TEXTURE         0x0020000
-#define R300_ANY_NEW_TEXTURES    0x1fe0000
-#define R300_NEW_VERTEX_FORMAT   0x2000000
-#define R300_NEW_VERTEX_SHADER   0x4000000
-#define R300_NEW_VIEWPORT        0x8000000
-#define R300_NEW_KITCHEN_SINK    0xfffffff
+#define R300_NEW_BLEND           0x00000001
+#define R300_NEW_BLEND_COLOR     0x00000002
+#define R300_NEW_CLIP            0x00000004
+#define R300_NEW_CONSTANTS       0x00000008
+#define R300_NEW_DSA             0x00000010
+#define R300_NEW_FRAMEBUFFERS    0x00000020
+#define R300_NEW_FRAGMENT_SHADER 0x00000040
+#define R300_NEW_RASTERIZER      0x00000080
+#define R300_NEW_RS_BLOCK        0x00000100
+#define R300_NEW_SAMPLER         0x00000200
+#define R300_ANY_NEW_SAMPLERS    0x0001fe00
+#define R300_NEW_SCISSOR         0x00020000
+#define R300_NEW_TEXTURE         0x00040000
+#define R300_ANY_NEW_TEXTURES    0x03fc0000
+#define R300_NEW_VERTEX_FORMAT   0x04000000
+#define R300_NEW_VERTEX_SHADER   0x08000000
+#define R300_NEW_VIEWPORT        0x10000000
+#define R300_NEW_KITCHEN_SINK    0x1fffffff
 
 /* The next several objects are not pure Radeon state; they inherit from
  * various Gallium classes. */
@@ -136,14 +142,14 @@ struct r300_viewport_state {
 struct r300_constant_buffer {
     /* Buffer of constants */
     /* XXX first number should be raised */
-    float constants[8][4];
+    float constants[32][4];
     /* Number of user-defined constants */
-    int user_count;
+    unsigned user_count;
     /* Total number of constants */
-    int count;
+    unsigned count;
 };
 
-struct r3xx_fragment_shader {
+struct r300_fragment_shader {
     /* Parent class */
     struct pipe_shader_state state;
     struct tgsi_shader_info info;
@@ -153,11 +159,15 @@ struct r3xx_fragment_shader {
 
     /* Pixel stack size */
     int stack_size;
+
+    /* Are there immediates in this shader?
+     * If not, we can heavily optimize recompilation. */
+    boolean uses_imms;
 };
 
-struct r300_fragment_shader {
+struct r3xx_fragment_shader {
     /* Parent class */
-    struct r3xx_fragment_shader shader;
+    struct r300_fragment_shader shader;
 
     /* Number of ALU instructions */
     int alu_instruction_count;
@@ -180,9 +190,9 @@ struct r300_fragment_shader {
     } instructions[64]; /* XXX magic num */
 };
 
-struct r500_fragment_shader {
+struct r5xx_fragment_shader {
     /* Parent class */
-    struct r3xx_fragment_shader shader;
+    struct r300_fragment_shader shader;
 
     /* Number of used instructions */
     int instruction_count;
@@ -207,7 +217,7 @@ struct r300_texture {
 
     /* Stride (pitch?) of this texture in bytes */
     unsigned stride;
-    
+
     /* Total size of this texture, in bytes. */
     unsigned size;
 
@@ -243,6 +253,10 @@ struct r300_vertex_shader {
     /* Has this shader been translated yet? */
     boolean translated;
 
+    /* Are there immediates in this shader?
+     * If not, we can heavily optimize recompilation. */
+    boolean uses_imms;
+
     /* Number of used instructions */
     int instruction_count;
 
@@ -255,6 +269,11 @@ struct r300_vertex_shader {
     } instructions[128]; /*< XXX magic number */
 };
 
+static struct pipe_viewport_state r300_viewport_identity = {
+    .scale = {1.0, 1.0, 1.0, 1.0},
+    .translate = {0.0, 0.0, 0.0, 0.0},
+};
+
 struct r300_context {
     /* Parent class */
     struct pipe_context context;
@@ -264,17 +283,24 @@ struct r300_context {
     /* Draw module. Used mostly for SW TCL. */
     struct draw_context* draw;
 
+    /* Vertex buffer for rendering. */
+    struct pipe_buffer* vbo;
+    /* Offset into the VBO. */
+    size_t vbo_offset;
+
     /* Various CSO state objects. */
     /* Blend state. */
     struct r300_blend_state* blend_state;
     /* Blend color state. */
     struct r300_blend_color_state* blend_color_state;
+    /* User clip planes. */
+    struct pipe_clip_state clip_state;
     /* Shader constants. */
     struct r300_constant_buffer shader_constants[PIPE_SHADER_TYPES];
     /* Depth, stencil, and alpha state. */
     struct r300_dsa_state* dsa_state;
     /* Fragment shader. */
-    struct r3xx_fragment_shader* fs;
+    struct r300_fragment_shader* fs;
     /* Framebuffer state. We currently don't need our own version of this. */
     struct pipe_framebuffer_state framebuffer_state;
     /* Rasterizer state. */
@@ -289,7 +315,7 @@ struct r300_context {
     /* Texture states. */
     struct r300_texture* textures[8];
     int texture_count;
-    /* Vertex buffers. */
+    /* Vertex buffers for Gallium. */
     struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
     int vertex_buffer_count;
     /* Vertex information. */
@@ -305,7 +331,8 @@ struct r300_context {
 };
 
 /* Convenience cast wrapper. */
-static struct r300_context* r300_context(struct pipe_context* context) {
+static INLINE struct r300_context* r300_context(struct pipe_context* context)
+{
     return (struct r300_context*)context;
 }
 
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 82a3942248..71b142c0db 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -34,6 +34,7 @@
 
 #define MAX_CS_SIZE 64 * 1024 / 4
 
+#define VERY_VERBOSE_CS 0
 #define VERY_VERBOSE_REGISTERS 0
 
 /* XXX stolen from radeon_drm.h */
@@ -56,8 +57,10 @@
 
 #define BEGIN_CS(size) do { \
     CHECK_CS(size); \
-    debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \
-        size, __FUNCTION__, __FILE__, __LINE__); \
+    if (VERY_VERBOSE_CS) { \
+        debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \
+                size, __FUNCTION__, __FILE__, __LINE__); \
+    } \
     cs_winsys->begin_cs(cs_winsys, (size), \
             __FILE__, __FUNCTION__, __LINE__); \
     cs_count = size; \
@@ -93,8 +96,9 @@
 } while (0)
 
 #define OUT_CS_RELOC(bo, offset, rd, wd, flags) do { \
-    debug_printf("r300: writing relocation for buffer %p, offset %d\n", \
-        bo, offset); \
+    debug_printf("r300: writing relocation for buffer %p, offset %d, " \
+            "domains (%d, %d, %d)\n", \
+        bo, offset, rd, wd, flags); \
     assert(bo); \
     OUT_CS(offset); \
     cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \
@@ -102,16 +106,20 @@
 } while (0)
 
 #define END_CS do { \
-    debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, __FILE__, \
-        __LINE__); \
+    if (VERY_VERBOSE_CS) { \
+        debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, \
+                __FILE__, __LINE__); \
+    } \
     if (cs_count != 0) \
         debug_printf("r300: Warning: cs_count off by %d\n", cs_count); \
     cs_winsys->end_cs(cs_winsys, __FILE__, __FUNCTION__, __LINE__); \
 } while (0)
 
 #define FLUSH_CS do { \
-    debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, __FILE__, \
-        __LINE__); \
+    if (VERY_VERBOSE_CS) { \
+        debug_printf("r300: FLUSH_CS in %s (%s:%d)\n\n", __FUNCTION__, \
+                __FILE__, __LINE__); \
+    } \
     cs_winsys->flush_cs(cs_winsys); \
 } while (0)
 
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
index dd63136c9d..c83e8526cf 100644
--- a/src/gallium/drivers/r300/r300_debug.c
+++ b/src/gallium/drivers/r300/r300_debug.c
@@ -22,7 +22,7 @@
 
 #include "r300_debug.h"
 
-static void r300_dump_fs(struct r300_fragment_shader* fs)
+void r3xx_dump_fs(struct r3xx_fragment_shader* fs)
 {
     int i;
 
@@ -30,82 +30,7 @@ static void r300_dump_fs(struct r300_fragment_shader* fs)
     }
 }
 
-static char* r500_fs_swiz[] = {
-    " R",
-    " G",
-    " B",
-    " A",
-    " 0",
-    ".5",
-    " 1",
-    " U",
-};
-
-static char* r500_fs_op_rgb[] = {
-    "MAD",
-    "DP3",
-    "DP4",
-    "D2A",
-    "MIN",
-    "MAX",
-    "---",
-    "CND",
-    "CMP",
-    "FRC",
-    "SOP",
-    "MDH",
-    "MDV",
-};
-
-static char* r500_fs_op_alpha[] = {
-    "MAD",
-    " DP",
-    "MIN",
-    "MAX",
-    "---",
-    "CND",
-    "CMP",
-    "FRC",
-    "EX2",
-    "LN2",
-    "RCP",
-    "RSQ",
-    "SIN",
-    "COS",
-    "MDH",
-    "MDV",
-};
-
-static char* r500_fs_mask[] = {
-    "NONE",
-    "R   ",
-    " G  ",
-    "RG  ",
-    "  B ",
-    "R B ",
-    " GB ",
-    "RGB ",
-    "   A",
-    "R  A",
-    " G A",
-    "RG A",
-    "  BA",
-    "R BA",
-    " GBA",
-    "RGBA",
-};
-
-static char* r500_fs_tex[] = {
-    "    NOP",
-    "     LD",
-    "TEXKILL",
-    "   PROJ",
-    "LODBIAS",
-    "    LOD",
-    "   DXDY",
-};
-
-void r500_fs_dump(struct r500_fragment_shader* fs)
+void r5xx_fs_dump(struct r5xx_fragment_shader* fs)
 {
     int i;
     uint32_t inst;
@@ -133,8 +58,8 @@ void r500_fs_dump(struct r500_fragment_shader* fs)
                 inst & R500_INST_NOP ? "NOP" : "",
                 inst & R500_INST_ALU_WAIT ? "ALU_WAIT" : "");
         debug_printf("wmask: %s omask: %s\n",
-                r500_fs_mask[(inst >> 11) & 0xf],
-                r500_fs_mask[(inst >> 15) & 0xf]);
+                r5xx_fs_mask[(inst >> 11) & 0xf],
+                r5xx_fs_mask[(inst >> 15) & 0xf]);
         switch (inst & 0x3) {
             case R500_INST_TYPE_ALU:
             case R500_INST_TYPE_OUT:
@@ -160,36 +85,36 @@ void r500_fs_dump(struct r500_fragment_shader* fs)
                 debug_printf("    3: RGB_INST   0x%08x:", inst);
                 debug_printf("rgb_A_src:%d %s/%s/%s %d "
                         "rgb_B_src:%d %s/%s/%s %d\n",
-                        inst & 0x3, r500_fs_swiz[(inst >> 2) & 0x7],
-                        r500_fs_swiz[(inst >> 5) & 0x7],
-                        r500_fs_swiz[(inst >> 8) & 0x7],
+                        inst & 0x3, r5xx_fs_swiz[(inst >> 2) & 0x7],
+                        r5xx_fs_swiz[(inst >> 5) & 0x7],
+                        r5xx_fs_swiz[(inst >> 8) & 0x7],
                         (inst >> 11) & 0x3, (inst >> 13) & 0x3,
-                        r500_fs_swiz[(inst >> 15) & 0x7],
-                        r500_fs_swiz[(inst >> 18) & 0x7],
-                        r500_fs_swiz[(inst >> 21) & 0x7],
+                        r5xx_fs_swiz[(inst >> 15) & 0x7],
+                        r5xx_fs_swiz[(inst >> 18) & 0x7],
+                        r5xx_fs_swiz[(inst >> 21) & 0x7],
                         (inst >> 24) & 0x3);
 
                 inst = fs->instructions[i].inst4;
                 debug_printf("    4: ALPHA_INST 0x%08x:", inst);
                 debug_printf("%s dest:%d%s alp_A_src:%d %s %d "
                         "alp_B_src:%d %s %d w:%d\n",
-                        r500_fs_op_alpha[inst & 0xf], (inst >> 4) & 0x7f,
+                        r5xx_fs_op_alpha[inst & 0xf], (inst >> 4) & 0x7f,
                         inst & (1<<11) ? "(rel)":"", (inst >> 12) & 0x3,
-                        r500_fs_swiz[(inst >> 14) & 0x7], (inst >> 17) & 0x3,
-                        (inst >> 19) & 0x3, r500_fs_swiz[(inst >> 21) & 0x7],
+                        r5xx_fs_swiz[(inst >> 14) & 0x7], (inst >> 17) & 0x3,
+                        (inst >> 19) & 0x3, r5xx_fs_swiz[(inst >> 21) & 0x7],
                         (inst >> 24) & 0x3, (inst >> 31) & 0x1);
 
                 inst = fs->instructions[i].inst5;
                 debug_printf("    5: RGBA_INST  0x%08x:", inst);
                 debug_printf("%s dest:%d%s rgb_C_src:%d %s/%s/%s %d "
                         "alp_C_src:%d %s %d\n",
-                        r500_fs_op_rgb[inst & 0xf], (inst >> 4) & 0x7f,
+                        r5xx_fs_op_rgb[inst & 0xf], (inst >> 4) & 0x7f,
                         inst & (1 << 11) ? "(rel)":"", (inst >> 12) & 0x3,
-                        r500_fs_swiz[(inst >> 14) & 0x7],
-                        r500_fs_swiz[(inst >> 17) & 0x7],
-                        r500_fs_swiz[(inst >> 20) & 0x7],
+                        r5xx_fs_swiz[(inst >> 14) & 0x7],
+                        r5xx_fs_swiz[(inst >> 17) & 0x7],
+                        r5xx_fs_swiz[(inst >> 20) & 0x7],
                         (inst >> 23) & 0x3, (inst >> 25) & 0x3,
-                        r500_fs_swiz[(inst >> 27) & 0x7], (inst >> 30) & 0x3);
+                        r5xx_fs_swiz[(inst >> 27) & 0x7], (inst >> 30) & 0x3);
                 break;
             case R500_INST_TYPE_FC:
                 /* XXX don't even bother yet */
@@ -199,7 +124,7 @@ void r500_fs_dump(struct r500_fragment_shader* fs)
                 debug_printf("    1: TEX_INST   0x%08x: id: %d "
                         "op:%s, %s, %s %s\n",
                         inst, (inst >> 16) & 0xf,
-                        r500_fs_tex[(inst >> 22) & 0x7],
+                        r5xx_fs_tex[(inst >> 22) & 0x7],
                         (inst & (1 << 25)) ? "ACQ" : "",
                         (inst & (1 << 26)) ? "IGNUNC" : "",
                         (inst & (1 << 27)) ? "UNSCALED" : "SCALED");
@@ -208,15 +133,15 @@ void r500_fs_dump(struct r500_fragment_shader* fs)
                 debug_printf("    2: TEX_ADDR   0x%08x: "
                         "src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n",
                         inst, inst & 0x7f, inst & (1 << 7) ? "(rel)" : "",
-                        r500_fs_swiz[(inst >> 8) & 0x3],
-                        r500_fs_swiz[(inst >> 10) & 0x3],
-                        r500_fs_swiz[(inst >> 12) & 0x3],
-                        r500_fs_swiz[(inst >> 14) & 0x3],
+                        r5xx_fs_swiz[(inst >> 8) & 0x3],
+                        r5xx_fs_swiz[(inst >> 10) & 0x3],
+                        r5xx_fs_swiz[(inst >> 12) & 0x3],
+                        r5xx_fs_swiz[(inst >> 14) & 0x3],
                         (inst >> 16) & 0x7f, inst & (1 << 23) ? "(rel)" : "",
-                        r500_fs_swiz[(inst >> 24) & 0x3],
-                        r500_fs_swiz[(inst >> 26) & 0x3],
-                        r500_fs_swiz[(inst >> 28) & 0x3],
-                        r500_fs_swiz[(inst >> 30) & 0x3]);
+                        r5xx_fs_swiz[(inst >> 24) & 0x3],
+                        r5xx_fs_swiz[(inst >> 26) & 0x3],
+                        r5xx_fs_swiz[(inst >> 28) & 0x3],
+                        r5xx_fs_swiz[(inst >> 30) & 0x3]);
                 
                 inst = fs->instructions[i].inst3;
                 debug_printf("    3: TEX_DXDY   0x%08x\n", inst);
@@ -225,14 +150,49 @@ void r500_fs_dump(struct r500_fragment_shader* fs)
     }
 }
 
+static void r300_vs_op_dump(uint32_t op)
+{
+    debug_printf(" dst: %d%s op: ",
+            (op >> 13) & 0x7f, r300_vs_dst_debug[(op >> 8) & 0x7]);
+    if (op & 0x80) {
+        if (op & 0x1) {
+            debug_printf("PVS_MACRO_OP_2CLK_M2X_ADD\n");
+        } else {
+            debug_printf("   PVS_MACRO_OP_2CLK_MADD\n");
+        }
+    } else if (op & 0x40) {
+        debug_printf("%s\n", r300_vs_me_ops[op & 0x1f]);
+    } else {
+        debug_printf("%s\n", r300_vs_ve_ops[op & 0x1f]);
+    }
+}
+
+void r300_vs_src_dump(uint32_t src)
+{
+    debug_printf(" reg: %d%s swiz: %s%s/%s%s/%s%s/%s%s\n",
+            (src >> 5) & 0x7f, r300_vs_src_debug[src & 0x3],
+            src & (1 << 25) ? "-" : " ",
+            r300_vs_swiz_debug[(src >> 13) & 0x7],
+            src & (1 << 26) ? "-" : " ",
+            r300_vs_swiz_debug[(src >> 16) & 0x7],
+            src & (1 << 27) ? "-" : " ",
+            r300_vs_swiz_debug[(src >> 19) & 0x7],
+            src & (1 << 28) ? "-" : " ",
+            r300_vs_swiz_debug[(src >> 22) & 0x7]);
+}
+
 void r300_vs_dump(struct r300_vertex_shader* vs)
 {
     int i;
 
     for (i = 0; i < vs->instruction_count; i++) {
-        debug_printf("inst0: 0x%x\n", vs->instructions[i].inst0);
-        debug_printf("inst1: 0x%x\n", vs->instructions[i].inst1);
-        debug_printf("inst2: 0x%x\n", vs->instructions[i].inst2);
-        debug_printf("inst3: 0x%x\n", vs->instructions[i].inst3);
+        debug_printf("%d: op: 0x%08x", i, vs->instructions[i].inst0);
+        r300_vs_op_dump(vs->instructions[i].inst0);
+        debug_printf(" src0: 0x%08x", vs->instructions[i].inst1);
+        r300_vs_src_dump(vs->instructions[i].inst1);
+        debug_printf(" src1: 0x%08x", vs->instructions[i].inst2);
+        r300_vs_src_dump(vs->instructions[i].inst2);
+        debug_printf(" src2: 0x%08x", vs->instructions[i].inst3);
+        r300_vs_src_dump(vs->instructions[i].inst3);
     }
 }
diff --git a/src/gallium/drivers/r300/r300_debug.h b/src/gallium/drivers/r300/r300_debug.h
index a1f873656d..6b58c1e250 100644
--- a/src/gallium/drivers/r300/r300_debug.h
+++ b/src/gallium/drivers/r300/r300_debug.h
@@ -24,10 +24,187 @@
 #define R300_DEBUG_H
 
 #include "r300_reg.h"
-#include "r300_state_shader.h"
-#include "r300_state_tcl.h"
+#include "r300_fs.h"
+#include "r300_vs.h"
 
-void r500_fs_dump(struct r500_fragment_shader* fs);
+static char* r5xx_fs_swiz[] = {
+    " R",
+    " G",
+    " B",
+    " A",
+    " 0",
+    ".5",
+    " 1",
+    " U",
+};
+
+static char* r5xx_fs_op_rgb[] = {
+    "MAD",
+    "DP3",
+    "DP4",
+    "D2A",
+    "MIN",
+    "MAX",
+    "---",
+    "CND",
+    "CMP",
+    "FRC",
+    "SOP",
+    "MDH",
+    "MDV",
+};
+
+static char* r5xx_fs_op_alpha[] = {
+    "MAD",
+    " DP",
+    "MIN",
+    "MAX",
+    "---",
+    "CND",
+    "CMP",
+    "FRC",
+    "EX2",
+    "LN2",
+    "RCP",
+    "RSQ",
+    "SIN",
+    "COS",
+    "MDH",
+    "MDV",
+};
+
+static char* r5xx_fs_mask[] = {
+    "NONE",
+    "R   ",
+    " G  ",
+    "RG  ",
+    "  B ",
+    "R B ",
+    " GB ",
+    "RGB ",
+    "   A",
+    "R  A",
+    " G A",
+    "RG A",
+    "  BA",
+    "R BA",
+    " GBA",
+    "RGBA",
+};
+
+static char* r5xx_fs_tex[] = {
+    "    NOP",
+    "     LD",
+    "TEXKILL",
+    "   PROJ",
+    "LODBIAS",
+    "    LOD",
+    "   DXDY",
+};
+
+static char* r300_vs_ve_ops[] = {
+    /* R300 vector ops */
+    "                 VE_NO_OP",
+    "           VE_DOT_PRODUCT",
+    "              VE_MULTIPLY",
+    "                   VE_ADD",
+    "          VE_MULTIPLY_ADD",
+    "       VE_DISTANCE_FACTOR",
+    "              VE_FRACTION",
+    "               VE_MAXIMUM",
+    "               VE_MINIMUM",
+    "VE_SET_GREATER_THAN_EQUAL",
+    "         VE_SET_LESS_THAN",
+    "        VE_MULTIPLYX2_ADD",
+    "        VE_MULTIPLY_CLAMP",
+    "            VE_FLT2FIX_DX",
+    "        VE_FLT2FIX_DX_RND",
+    /* R500 vector ops */
+    "      VE_PRED_SET_EQ_PUSH",
+    "      VE_PRED_SET_GT_PUSH",
+    "     VE_PRED_SET_GTE_PUSH",
+    "     VE_PRED_SET_NEQ_PUSH",
+    "         VE_COND_WRITE_EQ",
+    "         VE_COND_WRITE_GT",
+    "        VE_COND_WRITE_GTE",
+    "        VE_COND_WRITE_NEQ",
+    "      VE_SET_GREATER_THAN",
+    "             VE_SET_EQUAL",
+    "         VE_SET_NOT_EQUAL",
+    "               (reserved)",
+    "               (reserved)",
+    "               (reserved)",
+};
+
+static char* r300_vs_me_ops[] = {
+    /* R300 math ops */
+    "                 ME_NO_OP",
+    "          ME_EXP_BASE2_DX",
+    "          ME_LOG_BASE2_DX",
+    "          ME_EXP_BASEE_FF",
+    "        ME_LIGHT_COEFF_DX",
+    "         ME_POWER_FUNC_FF",
+    "              ME_RECIP_DX",
+    "              ME_RECIP_FF",
+    "         ME_RECIP_SQRT_DX",
+    "         ME_RECIP_SQRT_FF",
+    "              ME_MULTIPLY",
+    "     ME_EXP_BASE2_FULL_DX",
+    "     ME_LOG_BASE2_FULL_DX",
+    " ME_POWER_FUNC_FF_CLAMP_B",
+    "ME_POWER_FUNC_FF_CLAMP_B1",
+    "ME_POWER_FUNC_FF_CLAMP_01",
+    "                   ME_SIN",
+    "                   ME_COS",
+    /* R500 math ops */
+    "        ME_LOG_BASE2_IEEE",
+    "            ME_RECIP_IEEE",
+    "       ME_RECIP_SQRT_IEEE",
+    "           ME_PRED_SET_EQ",
+    "           ME_PRED_SET_GT",
+    "          ME_PRED_SET_GTE",
+    "          ME_PRED_SET_NEQ",
+    "          ME_PRED_SET_CLR",
+    "          ME_PRED_SET_INV",
+    "          ME_PRED_SET_POP",
+    "      ME_PRED_SET_RESTORE",
+    "               (reserved)",
+    "               (reserved)",
+    "               (reserved)",
+};
+
+/* XXX refactor to avoid clashing symbols */
+static char* r300_vs_src_debug[] = {
+    "t",
+    "i",
+    "c",
+    "a",
+};
+
+static char* r300_vs_dst_debug[] = {
+    "t",
+    "a0",
+    "o",
+    "ox",
+    "a",
+    "i",
+    "u",
+    "u",
+};
+
+static char* r300_vs_swiz_debug[] = {
+    "X",
+    "Y",
+    "Z",
+    "W",
+    "0",
+    "1",
+    "U",
+    "U",
+};
+
+void r5xx_fs_dump(struct r5xx_fragment_shader* fs);
+void r3xx_dump_fs(struct r3xx_fragment_shader* fs);
 
 void r300_vs_dump(struct r300_vertex_shader* vs);
 
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 01bac5f759..7ba56cdc1d 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -56,6 +56,36 @@ void r300_emit_blend_color_state(struct r300_context* r300,
     }
 }
 
+void r300_emit_clip_state(struct r300_context* r300,
+                          struct pipe_clip_state* clip)
+{
+    int i;
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
+    CS_LOCALS(r300);
+
+    if (r300screen->caps->has_tcl) {
+        BEGIN_CS(5 + (6 * 4));
+        OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
+                (r300screen->caps->is_r500 ?
+                 R500_PVS_UCP_START : R300_PVS_UCP_START));
+        OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, 6 * 4);
+        for (i = 0; i < 6; i++) {
+            OUT_CS_32F(clip->ucp[i][0]);
+            OUT_CS_32F(clip->ucp[i][1]);
+            OUT_CS_32F(clip->ucp[i][2]);
+            OUT_CS_32F(clip->ucp[i][3]);
+        }
+        OUT_CS_REG(R300_VAP_CLIP_CNTL, ((1 << clip->nr) - 1) |
+                R300_PS_UCP_MODE_CLIP_AS_TRIFAN);
+        END_CS;
+    } else {
+        BEGIN_CS(2);
+        OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE);
+        END_CS;
+    }
+
+}
+
 void r300_emit_dsa_state(struct r300_context* r300,
                            struct r300_dsa_state* dsa)
 {
@@ -80,7 +110,7 @@ void r300_emit_dsa_state(struct r300_context* r300,
 }
 
 void r300_emit_fragment_shader(struct r300_context* r300,
-                               struct r300_fragment_shader* fs)
+                               struct r3xx_fragment_shader* fs)
 {
     int i;
     CS_LOCALS(r300);
@@ -112,7 +142,7 @@ void r300_emit_fragment_shader(struct r300_context* r300,
 }
 
 void r500_emit_fragment_shader(struct r300_context* r300,
-                               struct r500_fragment_shader* fs)
+                               struct r5xx_fragment_shader* fs)
 {
     int i;
     struct r300_constant_buffer* constants =
@@ -163,6 +193,7 @@ void r300_emit_fb_state(struct r300_context* r300,
     BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4);
     for (i = 0; i < fb->nr_cbufs; i++) {
         tex = (struct r300_texture*)fb->cbufs[i]->texture;
+        assert(tex && tex->buffer && "cbuf is marked, but NULL!");
         pixpitch = tex->stride / tex->tex.block.size;
 
         OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
@@ -177,7 +208,8 @@ void r300_emit_fb_state(struct r300_context* r300,
 
     if (fb->zsbuf) {
         tex = (struct r300_texture*)fb->zsbuf->texture;
-        pixpitch = (tex->stride / tex->tex.block.size);
+        assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
+        pixpitch = tex->stride / tex->tex.block.size;
 
         OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1);
         OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
@@ -234,7 +266,7 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->ip[i]);
-        debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]);
+        /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */
     }
 
     OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
@@ -248,27 +280,15 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->inst[i]);
-        debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]);
+        /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */
     }
 
-    debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
-            rs->inst_count);
+    /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
+     *        rs->inst_count); */
 
     END_CS;
 }
 
-void r300_emit_sampler(struct r300_context* r300,
-                       struct r300_sampler_state* sampler, unsigned offset)
-{
-    CS_LOCALS(r300);
-
-    BEGIN_CS(6);
-    OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0);
-    OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
-    OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
-    END_CS;
-}
-
 void r300_emit_scissor_state(struct r300_context* r300,
                              struct r300_scissor_state* scissor)
 {
@@ -282,11 +302,17 @@ void r300_emit_scissor_state(struct r300_context* r300,
 }
 
 void r300_emit_texture(struct r300_context* r300,
-                       struct r300_texture* tex, unsigned offset)
+                       struct r300_sampler_state* sampler,
+                       struct r300_texture* tex,
+                       unsigned offset)
 {
     CS_LOCALS(r300);
 
-    BEGIN_CS(10);
+    BEGIN_CS(16);
+    OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0);
+    OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
+    OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
+
     OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0);
     OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1);
     OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2);
@@ -296,6 +322,30 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
+void r300_emit_vertex_buffer(struct r300_context* r300)
+{
+    CS_LOCALS(r300);
+
+    debug_printf("r300: Preparing vertex buffer %p for render, "
+            "vertex size %d\n", r300->vbo,
+            r300->vertex_info.vinfo.size);
+    /* Set the pointer to our vertex buffer. The emitted values are this:
+     * PACKET3 [3D_LOAD_VBPNTR]
+     * COUNT   [1]
+     * FORMAT  [size | stride << 8]
+     * OFFSET  [offset into BO]
+     * VBPNTR  [relocated BO]
+     */
+    BEGIN_CS(7);
+    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
+    OUT_CS(1);
+    OUT_CS(r300->vertex_info.vinfo.size |
+            (r300->vertex_info.vinfo.size << 8));
+    OUT_CS(r300->vbo_offset);
+    OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
+    END_CS;
+}
+
 void r300_emit_vertex_format_state(struct r300_context* r300)
 {
     int i;
@@ -310,22 +360,22 @@ void r300_emit_vertex_format_state(struct r300_context* r300)
     OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
     OUT_CS(r300->vertex_info.vinfo.hwfmt[2]);
     OUT_CS(r300->vertex_info.vinfo.hwfmt[3]);
-    for (i = 0; i < 4; i++) {
-        debug_printf("hwfmt%d: 0x%08x\n", i,
-                r300->vertex_info.vinfo.hwfmt[i]);
-    }
+    /* for (i = 0; i < 4; i++) {
+     *    debug_printf("hwfmt%d: 0x%08x\n", i,
+     *            r300->vertex_info.vinfo.hwfmt[i]);
+     * } */
 
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8);
     for (i = 0; i < 8; i++) {
         OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]);
-        debug_printf("prog_stream_cntl%d: 0x%08x\n", i,
-                r300->vertex_info.vap_prog_stream_cntl[i]);
+        /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i,
+         *        r300->vertex_info.vap_prog_stream_cntl[i]); */
     }
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8);
     for (i = 0; i < 8; i++) {
         OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]);
-        debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i,
-                r300->vertex_info.vap_prog_stream_cntl_ext[i]);
+        /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i,
+         *        r300->vertex_info.vap_prog_stream_cntl_ext[i]); */
     }
     END_CS;
 }
@@ -346,17 +396,22 @@ void r300_emit_vertex_shader(struct r300_context* r300,
     }
 
     if (constants->count) {
-        BEGIN_CS(16 + (vs->instruction_count * 4) + (constants->count * 4));
+        BEGIN_CS(14 + (vs->instruction_count * 4) + (constants->count * 4));
     } else {
-        BEGIN_CS(13 + (vs->instruction_count * 4) + (constants->count * 4));
+        BEGIN_CS(11 + (vs->instruction_count * 4));
     }
 
-    OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, R300_PVS_FIRST_INST(0) |
+    /* R300_VAP_PVS_CODE_CNTL_0
+     * R300_VAP_PVS_CONST_CNTL
+     * R300_VAP_PVS_CODE_CNTL_1
+     * See the r5xx docs for instructions on how to use these.
+     * XXX these could be optimized to select better values... */
+    OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3);
+    OUT_CS(R300_PVS_FIRST_INST(0) |
+            R300_PVS_XYZW_VALID_INST(vs->instruction_count - 1) |
             R300_PVS_LAST_INST(vs->instruction_count - 1));
-    OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, vs->instruction_count - 1);
-
-    /* XXX */
-    OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x0);
+    OUT_CS(R300_PVS_MAX_CONST_ADDR(constants->count - 1));
+    OUT_CS(vs->instruction_count - 1);
 
     OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0);
     OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, vs->instruction_count * 4);
@@ -386,7 +441,6 @@ void r300_emit_vertex_shader(struct r300_context* r300,
             R300_PVS_VF_MAX_VTX_NUM(12));
     OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
     END_CS;
-
 }
 
 void r300_emit_viewport_state(struct r300_context* r300,
@@ -403,7 +457,11 @@ void r300_emit_viewport_state(struct r300_context* r300,
     OUT_CS_32F(viewport->zscale);
     OUT_CS_32F(viewport->zoffset);
 
-    OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
+    if (r300->rs_state->enable_vte) {
+        OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
+    } else {
+        OUT_CS_REG(R300_VAP_VTE_CNTL, 0);
+    }
     END_CS;
 }
 
@@ -421,23 +479,66 @@ void r300_flush_textures(struct r300_context* r300)
 void r300_emit_dirty_state(struct r300_context* r300)
 {
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
-    int i;
-    int dirty_tex = 0;
+    struct r300_texture* tex;
+    int i, dirty_tex = 0;
+    boolean invalid = FALSE;
 
-    if (!(r300->dirty_hw)) {
+    if (!(r300->dirty_state)) {
         return;
     }
 
     r300_update_derived_state(r300);
 
     /* XXX check size */
-    struct r300_texture* fb_tex =
-        (struct r300_texture*)r300->framebuffer_state.cbufs[0];
-    r300->winsys->add_buffer(r300->winsys, fb_tex->buffer,
-            0, RADEON_GEM_DOMAIN_VRAM);
+validate:
+    /* Color buffers... */
+    for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) {
+        tex = (struct r300_texture*)r300->framebuffer_state.cbufs[i]->texture;
+        assert(tex && tex->buffer && "cbuf is marked, but NULL!");
+        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                    0, RADEON_GEM_DOMAIN_VRAM)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    }
+    /* ...depth buffer... */
+    if (r300->framebuffer_state.zsbuf) {
+        tex = (struct r300_texture*)r300->framebuffer_state.zsbuf->texture;
+        assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
+        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                    0, RADEON_GEM_DOMAIN_VRAM)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    }
+    /* ...textures... */
+    for (i = 0; i < r300->texture_count; i++) {
+        tex = r300->textures[i];
+        assert(tex && tex->buffer && "texture is marked, but NULL!");
+        if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                    RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    }
+    /* ...and vertex buffer. */
+    if (r300->vbo) {
+        if (!r300->winsys->add_buffer(r300->winsys, r300->vbo,
+                    RADEON_GEM_DOMAIN_GTT, 0)) {
+            r300->context.flush(&r300->context, 0, NULL);
+            goto validate;
+        }
+    } else {
+        debug_printf("No VBO while emitting dirty state!\n");
+    }
     if (r300->winsys->validate(r300->winsys)) {
-        /* XXX */
         r300->context.flush(&r300->context, 0, NULL);
+        if (invalid) {
+            /* Well, hell. */
+            exit(1);
+        }
+        invalid = TRUE;
+        goto validate;
     }
 
     if (r300->dirty_state & R300_NEW_BLEND) {
@@ -450,6 +551,11 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300->dirty_state &= ~R300_NEW_BLEND_COLOR;
     }
 
+    if (r300->dirty_state & R300_NEW_CLIP) {
+        r300_emit_clip_state(r300, &r300->clip_state);
+        r300->dirty_state &= ~R300_NEW_CLIP;
+    }
+
     if (r300->dirty_state & R300_NEW_DSA) {
         r300_emit_dsa_state(r300, r300->dsa_state);
         r300->dirty_state &= ~R300_NEW_DSA;
@@ -458,10 +564,10 @@ void r300_emit_dirty_state(struct r300_context* r300)
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) {
         if (r300screen->caps->is_r500) {
             r500_emit_fragment_shader(r300,
-                (struct r500_fragment_shader*)r300->fs);
+                (struct r5xx_fragment_shader*)r300->fs);
         } else {
             r300_emit_fragment_shader(r300,
-                (struct r300_fragment_shader*)r300->fs);
+                (struct r3xx_fragment_shader*)r300->fs);
         }
         r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER;
     }
@@ -481,29 +587,27 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300->dirty_state &= ~R300_NEW_RS_BLOCK;
     }
 
-    if (r300->dirty_state & R300_ANY_NEW_SAMPLERS) {
-        for (i = 0; i < r300->sampler_count; i++) {
-            if (r300->dirty_state & (R300_NEW_SAMPLER << i)) {
-                r300_emit_sampler(r300, r300->sampler_states[i], i);
-                r300->dirty_state &= ~(R300_NEW_SAMPLER << i);
-                dirty_tex++;
-            }
-        }
-    }
-
     if (r300->dirty_state & R300_NEW_SCISSOR) {
         r300_emit_scissor_state(r300, r300->scissor_state);
         r300->dirty_state &= ~R300_NEW_SCISSOR;
     }
 
-    if (r300->dirty_state & R300_ANY_NEW_TEXTURES) {
-        for (i = 0; i < r300->texture_count; i++) {
-            if (r300->dirty_state & (R300_NEW_TEXTURE << i)) {
-                r300_emit_texture(r300, r300->textures[i], i);
-                r300->dirty_state &= ~(R300_NEW_TEXTURE << i);
+    /* Samplers and textures are tracked separately but emitted together. */
+    if (r300->dirty_state &
+            (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) {
+        for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) {
+            if (r300->dirty_state &
+                    ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) {
+                r300_emit_texture(r300,
+                        r300->sampler_states[i],
+                        r300->textures[i],
+                        i);
+                r300->dirty_state &=
+                    ~((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i));
                 dirty_tex++;
             }
         }
+        r300->dirty_state &= ~(R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES);
     }
 
     if (r300->dirty_state & R300_NEW_VIEWPORT) {
@@ -519,4 +623,18 @@ void r300_emit_dirty_state(struct r300_context* r300)
         r300_emit_vertex_format_state(r300);
         r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT;
     }
+
+    if (r300->dirty_state & R300_NEW_VERTEX_SHADER) {
+        r300_emit_vertex_shader(r300, r300->vs);
+        r300->dirty_state &= ~R300_NEW_VERTEX_SHADER;
+    }
+
+    /* XXX
+    assert(r300->dirty_state == 0);
+    */
+
+    /* Finally, emit the VBO. */
+    r300_emit_vertex_buffer(r300);
+
+    r300->dirty_hw++;
 }
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 31dbc7ab85..fda26f3948 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -36,14 +36,17 @@ void r300_emit_blend_state(struct r300_context* r300,
 void r300_emit_blend_color_state(struct r300_context* r300,
                                  struct r300_blend_color_state* bc);
 
+void r300_emit_clip_state(struct r300_context* r300,
+                          struct pipe_clip_state* clip);
+
 void r300_emit_dsa_state(struct r300_context* r300,
                          struct r300_dsa_state* dsa);
 
 void r300_emit_fragment_shader(struct r300_context* r300,
-                               struct r300_fragment_shader* fs);
+                               struct r3xx_fragment_shader* fs);
 
 void r500_emit_fragment_shader(struct r300_context* r300,
-                               struct r500_fragment_shader* fs);
+                               struct r5xx_fragment_shader* fs);
 
 void r300_emit_fb_state(struct r300_context* r300,
                         struct pipe_framebuffer_state* fb);
@@ -53,14 +56,15 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs);
 void r300_emit_rs_block_state(struct r300_context* r300,
                               struct r300_rs_block* rs);
 
-void r300_emit_sampler(struct r300_context* r300,
-                       struct r300_sampler_state* sampler, unsigned offset);
-
 void r300_emit_scissor_state(struct r300_context* r300,
                              struct r300_scissor_state* scissor);
 
 void r300_emit_texture(struct r300_context* r300,
-                       struct r300_texture* tex, unsigned offset);
+                       struct r300_sampler_state* sampler,
+                       struct r300_texture* tex,
+                       unsigned offset);
+
+void r300_emit_vertex_buffer(struct r300_context* r300);
 
 void r300_emit_vertex_format_state(struct r300_context* r300);
 
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index 89a5f2b20c..0dff1c6f4f 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -29,7 +29,11 @@ static void r300_flush(struct pipe_context* pipe,
     struct r300_context* r300 = r300_context(pipe);
     CS_LOCALS(r300);
 
-    draw_flush(r300->draw);
+    /* We probably need to flush Draw, but we may have been called from
+     * within Draw. This feels kludgy, but it might be the best thing. */
+    if (!r300->draw->flushing) {
+        draw_flush(r300->draw);
+    }
 
     if (r300->dirty_hw) {
         FLUSH_CS;
diff --git a/src/gallium/drivers/r300/r300_flush.h b/src/gallium/drivers/r300/r300_flush.h
index a1b224b39c..9a83d89daa 100644
--- a/src/gallium/drivers/r300/r300_flush.h
+++ b/src/gallium/drivers/r300/r300_flush.h
@@ -23,6 +23,8 @@
 #ifndef R300_FLUSH_H
 #define R300_FLUSH_H
 
+#include "draw/draw_private.h"
+
 #include "pipe/p_context.h"
 
 #include "r300_context.h"
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
new file mode 100644
index 0000000000..4b304306d0
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -0,0 +1,109 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_fs.h"
+
+void r300_translate_fragment_shader(struct r300_context* r300,
+                                    struct r300_fragment_shader* fs)
+{
+    struct tgsi_parse_context parser;
+    int i;
+    boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
+    struct r300_constant_buffer* consts =
+        &r300->shader_constants[PIPE_SHADER_FRAGMENT];
+
+    struct r300_fs_asm* assembler = CALLOC_STRUCT(r300_fs_asm);
+    if (assembler == NULL) {
+        return;
+    }
+    /* Setup starting offset for immediates. */
+    assembler->imm_offset = consts->user_count;
+    /* Enable depth writes, if needed. */
+    assembler->writes_depth = fs->info.writes_z;
+
+    /* Make sure we start at the beginning of the shader. */
+    if (is_r500) {
+        ((struct r5xx_fragment_shader*)fs)->instruction_count = 0;
+    }
+
+    tgsi_parse_init(&parser, fs->state.tokens);
+
+    while (!tgsi_parse_end_of_tokens(&parser)) {
+        tgsi_parse_token(&parser);
+
+        /* This is seriously the lamest way to create fragment programs ever.
+         * I blame TGSI. */
+        switch (parser.FullToken.Token.Type) {
+            case TGSI_TOKEN_TYPE_DECLARATION:
+                /* Allocated registers sitting at the beginning
+                 * of the program. */
+                r300_fs_declare(assembler, &parser.FullToken.FullDeclaration);
+                break;
+            case TGSI_TOKEN_TYPE_IMMEDIATE:
+                debug_printf("r300: Emitting immediate to constant buffer, "
+                        "position %d\n",
+                        assembler->imm_offset + assembler->imm_count);
+                /* I am not amused by the length of these. */
+                for (i = 0; i < 4; i++) {
+                    consts->constants[assembler->imm_offset +
+                        assembler->imm_count][i] =
+                        parser.FullToken.FullImmediate.u.ImmediateFloat32[i]
+                        .Float;
+                }
+                assembler->imm_count++;
+                break;
+            case TGSI_TOKEN_TYPE_INSTRUCTION:
+                if (is_r500) {
+                    r5xx_fs_instruction((struct r5xx_fragment_shader*)fs,
+                            assembler, &parser.FullToken.FullInstruction);
+                } else {
+                    r3xx_fs_instruction((struct r3xx_fragment_shader*)fs,
+                            assembler, &parser.FullToken.FullInstruction);
+                }
+                break;
+        }
+    }
+
+    debug_printf("r300: fs: %d texs and %d colors, first free reg is %d\n",
+            assembler->tex_count, assembler->color_count,
+            assembler->tex_count + assembler->color_count);
+
+    consts->count = consts->user_count + assembler->imm_count;
+    fs->uses_imms = assembler->imm_count;
+    debug_printf("r300: fs: %d total constants, "
+            "%d from user and %d from immediates\n", consts->count,
+            consts->user_count, assembler->imm_count);
+    r3xx_fs_finalize(fs, assembler);
+    if (is_r500) {
+        r5xx_fs_finalize((struct r5xx_fragment_shader*)fs, assembler);
+    }
+
+    tgsi_dump(fs->state.tokens, 0);
+    /* XXX finish r300 dumper too */
+    if (is_r500) {
+        r5xx_fs_dump((struct r5xx_fragment_shader*)fs);
+    }
+
+    tgsi_parse_free(&parser);
+    FREE(assembler);
+}
diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h
new file mode 100644
index 0000000000..18deb7a05e
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_fs.h
@@ -0,0 +1,36 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_FS_H
+#define R300_FS_H
+
+#include "tgsi/tgsi_dump.h"
+
+#include "r300_context.h"
+#include "r3xx_fs.h"
+#include "r5xx_fs.h"
+
+void r300_translate_fragment_shader(struct r300_context* r300,
+                                    struct r300_fragment_shader* fs);
+
+    #endif /* R300_FS_H */
diff --git a/src/gallium/drivers/r300/r300_fs_inlines.h b/src/gallium/drivers/r300/r300_fs_inlines.h
new file mode 100644
index 0000000000..be4be9465e
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_fs_inlines.h
@@ -0,0 +1,158 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_FS_INLINES_H
+#define R300_FS_INLINES_H
+
+#include "tgsi/tgsi_parse.h"
+
+#include "r300_context.h"
+#include "r300_debug.h"
+#include "r300_reg.h"
+#include "r300_screen.h"
+#include "r300_shader_inlines.h"
+
+/* Temporary struct used to hold assembly state while putting together
+ * fragment programs. */
+struct r300_fs_asm {
+    /* Pipe context. */
+    struct r300_context* r300;
+    /* Number of colors. */
+    unsigned color_count;
+    /* Number of texcoords. */
+    unsigned tex_count;
+    /* Offset for temporary registers. Inputs and temporaries have no
+     * distinguishing markings, so inputs start at 0 and the first usable
+     * temporary register is after all inputs. */
+    unsigned temp_offset;
+    /* Number of requested temporary registers. */
+    unsigned temp_count;
+    /* Offset for immediate constants. Neither R300 nor R500 can do four
+     * inline constants per source, so instead we copy immediates into the
+     * constant buffer. */
+    unsigned imm_offset;
+    /* Number of immediate constants. */
+    unsigned imm_count;
+    /* Are depth writes enabled? */
+    boolean writes_depth;
+    /* Depth write offset. This is the TGSI output that corresponds to
+     * depth writes. */
+    unsigned depth_output;
+};
+
+static INLINE void r300_fs_declare(struct r300_fs_asm* assembler,
+                            struct tgsi_full_declaration* decl)
+{
+    switch (decl->Declaration.File) {
+        case TGSI_FILE_INPUT:
+            switch (decl->Semantic.SemanticName) {
+                case TGSI_SEMANTIC_COLOR:
+                    assembler->color_count++;
+                    break;
+                case TGSI_SEMANTIC_FOG:
+                case TGSI_SEMANTIC_GENERIC:
+                    assembler->tex_count++;
+                    break;
+                default:
+                    debug_printf("r300: fs: Bad semantic declaration %d\n",
+                        decl->Semantic.SemanticName);
+                    break;
+            }
+            break;
+        case TGSI_FILE_OUTPUT:
+            /* Depth write. Mark the position of the output so we can
+             * identify it later. */
+            if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) {
+                assembler->depth_output = decl->DeclarationRange.First;
+            }
+            break;
+        case TGSI_FILE_CONSTANT:
+            break;
+        case TGSI_FILE_TEMPORARY:
+            assembler->temp_count++;
+            break;
+        default:
+            debug_printf("r300: fs: Bad file %d\n", decl->Declaration.File);
+            break;
+    }
+
+    assembler->temp_offset = assembler->color_count + assembler->tex_count;
+}
+
+static INLINE unsigned r300_fs_src(struct r300_fs_asm* assembler,
+                                   struct tgsi_src_register* src)
+{
+    switch (src->File) {
+        case TGSI_FILE_NULL:
+            return 0;
+        case TGSI_FILE_INPUT:
+            /* XXX may be wrong */
+            return src->Index;
+            break;
+        case TGSI_FILE_TEMPORARY:
+            return src->Index + assembler->temp_offset;
+            break;
+        case TGSI_FILE_IMMEDIATE:
+            return (src->Index + assembler->imm_offset) | (1 << 8);
+            break;
+        case TGSI_FILE_CONSTANT:
+            /* XXX magic */
+            return src->Index | (1 << 8);
+            break;
+        default:
+            debug_printf("r300: fs: Unimplemented src %d\n", src->File);
+            break;
+    }
+    return 0;
+}
+
+static INLINE unsigned r300_fs_dst(struct r300_fs_asm* assembler,
+                                   struct tgsi_dst_register* dst)
+{
+    switch (dst->File) {
+        case TGSI_FILE_NULL:
+            /* This happens during KIL instructions. */
+            return 0;
+            break;
+        case TGSI_FILE_OUTPUT:
+            return 0;
+            break;
+        case TGSI_FILE_TEMPORARY:
+            return dst->Index + assembler->temp_offset;
+            break;
+        default:
+            debug_printf("r300: fs: Unimplemented dst %d\n", dst->File);
+            break;
+    }
+    return 0;
+}
+
+static INLINE boolean r300_fs_is_depr(struct r300_fs_asm* assembler,
+                                      struct tgsi_dst_register* dst)
+{
+    return (assembler->writes_depth &&
+            (dst->File == TGSI_FILE_OUTPUT) &&
+            (dst->Index == assembler->depth_output));
+}
+
+#endif /* R300_FS_INLINES_H */
diff --git a/src/gallium/drivers/r300/r300_query.h b/src/gallium/drivers/r300/r300_query.h
index 4f447ea45b..6a7646087a 100644
--- a/src/gallium/drivers/r300/r300_query.h
+++ b/src/gallium/drivers/r300/r300_query.h
@@ -27,6 +27,8 @@
 #include "r300_cs.h"
 #include "r300_reg.h"
 
+struct r300_context;
+
 struct r300_query {
     /* The kind of query. Currently only OQ is supported. */
     unsigned type;
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index 660816e1da..6825d99870 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -511,11 +511,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_PVS_XYZW_VALID_INST_SHIFT    10
 #       define R300_PVS_LAST_INST_SHIFT          20
 #       define R300_PVS_FIRST_INST(x)            ((x) << 0)
+#       define R300_PVS_XYZW_VALID_INST(x)       ((x) << 10)
 #       define R300_PVS_LAST_INST(x)             ((x) << 20)
 /* Addresses are relative the the vertex program parameters area. */
 #define R300_VAP_PVS_CONST_CNTL             0x22D4
 #       define R300_PVS_CONST_BASE_OFFSET_SHIFT  0
 #       define R300_PVS_MAX_CONST_ADDR_SHIFT     16
+#       define R300_PVS_MAX_CONST_ADDR(x)        ((x) << 16)
 #define R300_VAP_PVS_CODE_CNTL_1	    0x22D8
 #       define R300_PVS_LAST_VTX_SRC_INST_SHIFT  0
 #define R300_VAP_PVS_FLOW_CNTL_OPC          0x22DC
@@ -1062,8 +1064,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
         R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | \
         R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \
         R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | \
-        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT | \
-        R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST )
+        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT )
 
 #       define R300_SHADE_MODEL_SMOOTH ( \
         R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | \
@@ -1073,8 +1074,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
         R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | \
         R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \
         R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | \
-        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \
-        R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST )
+        R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD )
 
 /* Specifies red & green components of fill color -- S312 format -- Backwards comp. */
 #define R300_GA_SOLID_RG                         0x427c
@@ -1478,6 +1478,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_TX_PITCH_EN                  (1 << 31)
 #       define R300_TX_WIDTH(x)                  ((x) << 0)
 #       define R300_TX_HEIGHT(x)                 ((x) << 11)
+#       define R300_TX_NUM_LEVELS(x)             ((x) << 26)
 
 #define R300_TX_FORMAT1_0                   0x44C0
 	/* The interpretation of the format word by Wladimir van der Laan */
@@ -1485,9 +1486,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 	   They are given meanings as R, G, B and Alpha by the swizzle
 	   specification */
 #	define R300_TX_FORMAT_X8		    0x0
-#	define R500_TX_FORMAT_X1		    0x0 // bit set in format 2
 #	define R300_TX_FORMAT_X16		    0x1
-#	define R500_TX_FORMAT_X1_REV		    0x0 // bit set in format 2
 #	define R300_TX_FORMAT_Y4X4		    0x2
 #	define R300_TX_FORMAT_Y8X8		    0x3
 #	define R300_TX_FORMAT_Y16X16		    0x4
@@ -1504,31 +1503,29 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_DXT1	    	    0xF
 #	define R300_TX_FORMAT_DXT3	    	    0x10
 #	define R300_TX_FORMAT_DXT5	    	    0x11
-#	define R300_TX_FORMAT_D3DMFT_CxV8U8	    0x12     /* no swizzle */
-#	define R300_TX_FORMAT_A8R8G8B8	    	    0x13     /* no swizzle */
-#	define R300_TX_FORMAT_B8G8_B8G8	    	    0x14     /* no swizzle */
-#	define R300_TX_FORMAT_G8R8_G8B8	    	    0x15     /* no swizzle */
-
-	/* These two values are wrong, but they're the only values that
-	 * produce any even vaguely correct results.  Can r300 only do 16-bit
-	 * depth textures?
-	 */
-#	define R300_TX_FORMAT_X24_Y8	    	    0x1e
-#	define R300_TX_FORMAT_X32	    	    0x1e
-
-	/* 0x16 - some 16 bit green format.. ?? */
+#	define R300_TX_FORMAT_Y8           	    0x12
+#	define R300_TX_FORMAT_AVYU444 	    	    0x13
+#	define R300_TX_FORMAT_VYUY422  	    	    0x14
+#	define R300_TX_FORMAT_YVYU422  	    	    0x15
+#	define R300_TX_FORMAT_16_MPEG  	    	    0x16
+#	define R300_TX_FORMAT_16_16_MPEG    	    0x17
+#	define R300_TX_FORMAT_16F     	    	    0x18
+#	define R300_TX_FORMAT_16F_16F 	    	    0x19
+#	define R300_TX_FORMAT_16F_16F_16F_16F  	    0x1A
+#	define R300_TX_FORMAT_32F     	    	    0x1B
+#	define R300_TX_FORMAT_32F_32F 	    	    0x1C
+#	define R300_TX_FORMAT_32F_32F_32F_32F  	    0x1D
+#       define R300_TX_FORMAT_W24_FP                0x1E
+
+#       define R300_TX_FORMAT_SIGNED_W             (1 << 5)
+#       define R300_TX_FORMAT_SIGNED_Z             (1 << 6)
+#       define R300_TX_FORMAT_SIGNED_Y             (1 << 7)
+#       define R300_TX_FORMAT_SIGNED_X             (1 << 8)
+#       define R300_TX_FORMAT_SIGNED               (0xf << 5)
+
 #	define R300_TX_FORMAT_3D		   (1 << 25)
 #	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
 
-	/* gap */
-	/* Floating point formats */
-	/* Note - hardware supports both 16 and 32 bit floating point */
-#	define R300_TX_FORMAT_FL_I16	    	    0x18
-#	define R300_TX_FORMAT_FL_I16A16	    	    0x19
-#	define R300_TX_FORMAT_FL_R16G16B16A16	    0x1A
-#	define R300_TX_FORMAT_FL_I32	    	    0x1B
-#	define R300_TX_FORMAT_FL_I32A32	    	    0x1C
-#	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
 	/* alpha modes, convenience mostly */
 	/* if you have alpha, pick constant appropriate to the
 	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
@@ -1569,7 +1566,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_CONST_Z		(4<<5)
 #	define R300_TX_FORMAT_CONST_W		(8<<5)
 
-#	define R300_TX_FORMAT_YUV_MODE		0x00800000
+#       define R300_TX_FORMAT_GAMMA               (1 << 21)
+#       define R300_TX_FORMAT_YUV_TO_RGB          (1 << 22)
 
 #define R300_TX_FORMAT2_0		    0x4500 /* obvious missing in gap */
 #       define R300_TX_PITCHMASK_SHIFT           0
@@ -3040,6 +3038,7 @@ enum {
 #   define R500_INST_RGB_WMASK_R			(1 << 11)
 #   define R500_INST_RGB_WMASK_G			(1 << 12)
 #   define R500_INST_RGB_WMASK_B			(1 << 13)
+#   define R500_INST_RGB_WMASK_RGB			(7 << 11)
 #   define R500_INST_ALPHA_WMASK			(1 << 14)
 #   define R500_INST_RGB_OMASK_R			(1 << 15)
 #   define R500_INST_RGB_OMASK_G			(1 << 16)
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index cbd84d7c56..cd458d019a 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -45,11 +45,7 @@ struct r300_render {
 
     /* VBO */
     struct pipe_buffer* vbo;
-    size_t vbo_size;
-    size_t vbo_offset;
-    void* vbo_map;
     size_t vbo_alloc_size;
-    size_t vbo_max_used;
 };
 
 static INLINE struct r300_render*
@@ -78,24 +74,21 @@ static boolean r300_render_allocate_vertices(struct vbuf_render* render,
     struct pipe_screen* screen = r300->context.screen;
     size_t size = (size_t)vertex_size * (size_t)count;
 
-    if (r300render->vbo) {
+    if (r300render->vbo && (size > r300render->vbo_alloc_size)) {
         pipe_buffer_reference(&r300render->vbo, NULL);
     }
+    
+    if (!r300render->vbo) {
+        r300render->vbo = pipe_buffer_create(screen,
+                                             64,
+                                             PIPE_BUFFER_USAGE_VERTEX,
+                                             size);
+    }
 
-    r300render->vbo_size = MAX2(size, r300render->vbo_alloc_size);
-    r300render->vbo_offset = 0;
-    r300render->vbo = pipe_buffer_create(screen,
-                                         64,
-                                         PIPE_BUFFER_USAGE_VERTEX,
-                                         r300render->vbo_size);
-
+    r300render->vbo_alloc_size = MAX2(size, r300render->vbo_alloc_size);
     r300render->vertex_size = vertex_size;
 
-    if (r300render->vbo) {
-        return TRUE;
-    } else {
-        return FALSE;
-    }
+    return (r300render->vbo) ? TRUE : FALSE;
 }
 
 static void* r300_render_map_vertices(struct vbuf_render* render)
@@ -103,10 +96,8 @@ static void* r300_render_map_vertices(struct vbuf_render* render)
     struct r300_render* r300render = r300_render(render);
     struct pipe_screen* screen = r300render->r300->context.screen;
 
-    r300render->vbo_map = pipe_buffer_map(screen, r300render->vbo,
-                                          PIPE_BUFFER_USAGE_CPU_WRITE);
-
-    return (unsigned char*)r300render->vbo_map + r300render->vbo_offset;
+    return (unsigned char*)pipe_buffer_map(screen, r300render->vbo,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE);
 }
 
 static void r300_render_unmap_vertices(struct vbuf_render* render,
@@ -116,9 +107,6 @@ static void r300_render_unmap_vertices(struct vbuf_render* render,
     struct r300_render* r300render = r300_render(render);
     struct pipe_screen* screen = r300render->r300->context.screen;
 
-    r300render->vbo_max_used = MAX2(r300render->vbo_max_used,
-             r300render->vertex_size * (max + 1));
-
     pipe_buffer_unmap(screen, r300render->vbo);
 }
 
@@ -180,27 +168,9 @@ static void prepare_render(struct r300_render* render, unsigned count)
 
     CS_LOCALS(r300);
 
-    /* Make sure that all possible state is emitted. */
-    r300_emit_dirty_state(r300);
+    r300->vbo = render->vbo;
 
-    debug_printf("r300: Preparing vertex buffer %p for render, "
-            "vertex size %d, vertex count %d\n", render->vbo,
-            r300->vertex_info.vinfo.size, count);
-    /* Set the pointer to our vertex buffer. The emitted values are this:
-     * PACKET3 [3D_LOAD_VBPNTR]
-     * COUNT   [1]
-     * FORMAT  [size | stride << 8]
-     * OFFSET  [0]
-     * VBPNTR  [relocated BO]
-     */
-    BEGIN_CS(7);
-    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
-    OUT_CS(1);
-    OUT_CS(r300->vertex_info.vinfo.size |
-            (r300->vertex_info.vinfo.size << 8));
-    OUT_CS(render->vbo_offset);
-    OUT_CS_RELOC(render->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
-    END_CS;
+    r300_emit_dirty_state(r300);
 }
 
 static void r300_render_draw_arrays(struct vbuf_render* render,
@@ -212,8 +182,6 @@ static void r300_render_draw_arrays(struct vbuf_render* render,
 
     CS_LOCALS(r300);
 
-    r300render->vbo_offset = start;
-
     prepare_render(r300render, count);
 
     debug_printf("r300: Doing vbuf render, count %d\n", count);
@@ -248,13 +216,14 @@ static void r300_render_draw(struct vbuf_render* render,
         return;
     }
 
+/*
     index_map = pipe_buffer_map(screen, index_buffer,
                                 PIPE_BUFFER_USAGE_CPU_WRITE);
     memcpy(index_map, indices, count);
     pipe_buffer_unmap(screen, index_buffer);
 
     debug_printf("r300: Doing indexbuf render, count %d\n", count);
-/*
+
     BEGIN_CS(8);
     OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0);
     OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
@@ -264,13 +233,15 @@ static void r300_render_draw(struct vbuf_render* render,
     OUT_CS_INDEX_RELOC(index_buffer, 0, count, RADEON_GEM_DOMAIN_GTT, 0, 0);
     END_CS; */
 
-    BEGIN_CS(2 + count);
-    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, count);
+    BEGIN_CS(2 + (count+1)/2);
+    OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, (count+1)/2);
     OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) |
-           r300render->hwprim | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
-    for (i = 0; i < count; i++) {
-        index = indices[i];
-        OUT_CS(index);
+           r300render->hwprim);
+    for (i = 0; i < count-1; i += 2) {
+        OUT_CS(indices[i+1] << 16 | indices[i]);
+    }
+    if (count % 2) {
+        OUT_CS(indices[count-1]);
     }
     END_CS;
 }
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index d2c5998c26..da1d5ffe2f 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -87,23 +87,24 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
             } else {
                 return 0;
             }
-            return 0;
         case PIPE_CAP_GLSL:
-            /* IN THEORY */
-            return 0;
+            if (r300screen->caps->is_r500) {
+                return 1;
+            } else {
+                return 0;
+            }
         case PIPE_CAP_S3TC:
-            /* IN THEORY */
-            return 0;
+            return 1;
         case PIPE_CAP_ANISOTROPIC_FILTER:
-            /* IN THEORY */
-            return 0;
+            return 1;
         case PIPE_CAP_POINT_SPRITE:
             /* IN THEORY */
             return 0;
         case PIPE_CAP_MAX_RENDER_TARGETS:
             return 4;
         case PIPE_CAP_OCCLUSION_QUERY:
-            return 1;
+            /* IN THEORY */
+            return 0;
         case PIPE_CAP_TEXTURE_SHADOW_MAP:
             /* IN THEORY */
             return 0;
@@ -143,6 +144,9 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
         case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
             /* XXX guessing (what a terrible guess) */
             return 2;
+        case PIPE_CAP_TGSI_CONT_SUPPORTED:
+            /* XXX */
+            return 0;
         default:
             debug_printf("r300: Implementation error: Bad param %d\n",
                 param);
@@ -152,17 +156,20 @@ static int r300_get_param(struct pipe_screen* pscreen, int param)
 
 static float r300_get_paramf(struct pipe_screen* pscreen, int param)
 {
+    struct r300_screen* r300screen = r300_screen(pscreen);
+
     switch (param) {
         case PIPE_CAP_MAX_LINE_WIDTH:
         case PIPE_CAP_MAX_LINE_WIDTH_AA:
-            /* XXX this is the biggest thing that will fit in that register.
-            * Perhaps the actual rendering limits are less? */
-            return 10922.0f;
         case PIPE_CAP_MAX_POINT_WIDTH:
         case PIPE_CAP_MAX_POINT_WIDTH_AA:
-            /* XXX this is the biggest thing that will fit in that register.
-             * Perhaps the actual rendering limits are less? */
-            return 10922.0f;
+            /* The maximum dimensions of the colorbuffer are our practical
+             * rendering limits. 2048 pixels should be enough for anybody. */
+            if (r300screen->caps->is_r500) {
+                return 4096.0f;
+            } else {
+                return 2048.0f;
+            }
         case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
             return 16.0f;
         case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
@@ -174,21 +181,58 @@ static float r300_get_paramf(struct pipe_screen* pscreen, int param)
     }
 }
 
-static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500)
+static boolean check_tex_2d_format(enum pipe_format format, uint32_t usage,
+                                   boolean is_r500)
 {
     switch (format) {
+        /* Supported formats. */
         /* Colorbuffer */
         case PIPE_FORMAT_A4R4G4B4_UNORM:
         case PIPE_FORMAT_R5G6B5_UNORM:
         case PIPE_FORMAT_A1R5G5B5_UNORM:
-        case PIPE_FORMAT_A8R8G8B8_UNORM:
+            return usage &
+                (PIPE_TEXTURE_USAGE_RENDER_TARGET |
+                 PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                 PIPE_TEXTURE_USAGE_PRIMARY);
+
+        /* Texture */
+        case PIPE_FORMAT_A8R8G8B8_SRGB:
+        case PIPE_FORMAT_R8G8B8A8_SRGB:
+        case PIPE_FORMAT_DXT1_RGB:
+        case PIPE_FORMAT_DXT1_RGBA:
+        case PIPE_FORMAT_DXT3_RGBA:
+        case PIPE_FORMAT_DXT5_RGBA:
+        case PIPE_FORMAT_YCBCR:
+            return usage & PIPE_TEXTURE_USAGE_SAMPLER;
+
         /* Colorbuffer or texture */
+        case PIPE_FORMAT_A8R8G8B8_UNORM:
+        case PIPE_FORMAT_R8G8B8A8_UNORM:
         case PIPE_FORMAT_I8_UNORM:
+            return usage &
+                (PIPE_TEXTURE_USAGE_RENDER_TARGET |
+                 PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                 PIPE_TEXTURE_USAGE_PRIMARY |
+                 PIPE_TEXTURE_USAGE_SAMPLER);
+
         /* Z buffer */
         case PIPE_FORMAT_Z16_UNORM:
-        /* Z buffer with stencil */
+            return usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+
+        /* Z buffer with stencil or texture */
         case PIPE_FORMAT_Z24S8_UNORM:
-            return TRUE;
+            return usage &
+                (PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
+                 PIPE_TEXTURE_USAGE_SAMPLER);
+
+        /* Definitely unsupported formats. */
+        /* Non-usable Z buffer/stencil formats. */
+        case PIPE_FORMAT_Z24X8_UNORM:
+        case PIPE_FORMAT_S8Z24_UNORM:
+        case PIPE_FORMAT_X8Z24_UNORM:
+            debug_printf("r300: Note: Got unsupported format: %s in %s\n",
+                pf_name(format), __FUNCTION__);
+            return FALSE;
 
         /* XXX These don't even exist
         case PIPE_FORMAT_A32R32G32B32:
@@ -211,7 +255,8 @@ static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500)
             return FALSE; */
 
         default:
-            debug_printf("r300: Warning: Got unsupported format: %s in %s\n",
+            /* Unknown format... */
+            debug_printf("r300: Warning: Got unknown format: %s in %s\n",
                 pf_name(format), __FUNCTION__);
             break;
     }
@@ -228,11 +273,18 @@ static boolean r300_is_format_supported(struct pipe_screen* pscreen,
 {
     switch (target) {
         case PIPE_TEXTURE_2D:
-            return check_tex_2d_format(format,
+            return check_tex_2d_format(format, tex_usage,
                 r300_screen(pscreen)->caps->is_r500);
+        case PIPE_TEXTURE_1D:
+        case PIPE_TEXTURE_3D:
+        case PIPE_TEXTURE_CUBE:
+            debug_printf("r300: Implementation error: Unsupported format "
+                    "target: %d\n", target);
+            break;
         default:
-            debug_printf("r300: Warning: Got unknown format target: %d\n",
-                format);
+            debug_printf("r300: Fatal: This is not a format target: %d\n",
+                target);
+            assert(0);
             break;
     }
 
@@ -337,7 +389,6 @@ struct pipe_screen* r300_create_screen(struct r300_winsys* r300_winsys)
         return NULL;
 
     caps->pci_id = r300_winsys->pci_id;
-    caps->num_frag_pipes = r300_winsys->gb_pipes;
 
     r300_parse_chipset(caps);
 
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index 3f52dbc3be..2a0e41fbc3 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -49,7 +49,7 @@ struct r300_transfer {
 };
 
 /* Convenience cast wrapper. */
-static struct r300_screen* r300_screen(struct pipe_screen* screen) {
+static INLINE struct r300_screen* r300_screen(struct pipe_screen* screen) {
     return (struct r300_screen*)screen;
 }
 
diff --git a/src/gallium/drivers/r300/r300_shader_inlines.h b/src/gallium/drivers/r300/r300_shader_inlines.h
new file mode 100644
index 0000000000..a04f45b03e
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_shader_inlines.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_SHADER_INLINES_H
+#define R300_SHADER_INLINES_H
+
+/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're
+ * not using enough of it. */
+static const struct tgsi_full_src_register r300_constant_zero = {
+    .SrcRegister.Extended = TRUE,
+    .SrcRegister.File = TGSI_FILE_NULL,
+    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO,
+    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO,
+    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO,
+    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO,
+};
+
+static const struct tgsi_full_src_register r300_constant_one = {
+    .SrcRegister.Extended = TRUE,
+    .SrcRegister.File = TGSI_FILE_NULL,
+    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE,
+    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE,
+    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE,
+    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE,
+};
+
+#endif /* R300_SHADER_INLINES_H */
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 184a23c9e6..68da0aa4cb 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -24,12 +24,14 @@
 #include "util/u_pack_color.h"
 
 #include "util/u_debug.h"
+
+#include "pipe/p_config.h"
 #include "pipe/internal/p_winsys_screen.h"
 
 #include "r300_context.h"
 #include "r300_reg.h"
 #include "r300_state_inlines.h"
-#include "r300_state_shader.h"
+#include "r300_fs.h"
 
 /* r300_state: Functions used to intialize state context by translating
  * Gallium state objects into semi-native r300 state objects. */
@@ -62,8 +64,6 @@ static void* r300_create_blend_state(struct pipe_context* pipe,
     }
 
     /* PIPE_LOGICOP_* don't need to be translated, fortunately. */
-    /* XXX are logicops still allowed if blending's disabled?
-     * Does Gallium take care of it for us? */
     if (state->logicop_enable) {
         blend->rop = R300_RB3D_ROPCNTL_ROP_ENABLE |
                 (state->logicop_func) << R300_RB3D_ROPCNTL_ROP_SHIFT;
@@ -121,9 +121,14 @@ static void r300_set_clip_state(struct pipe_context* pipe,
                                 const struct pipe_clip_state* state)
 {
     struct r300_context* r300 = r300_context(pipe);
-    /* XXX Draw */
-    draw_flush(r300->draw);
-    draw_set_clip_state(r300->draw, state);
+
+    if (r300_screen(pipe->screen)->caps->has_tcl) {
+        r300->clip_state = *state;
+        r300->dirty_state |= R300_NEW_CLIP;
+    } else {
+        draw_flush(r300->draw);
+        draw_set_clip_state(r300->draw, state);
+    }
 }
 
 static void
@@ -153,10 +158,12 @@ static void
 
     /* If the number of constants have changed, invalidate the shader. */
     if (r300->shader_constants[shader].user_count != i) {
-        if (shader == PIPE_SHADER_FRAGMENT && r300->fs) {
+        if (shader == PIPE_SHADER_FRAGMENT && r300->fs &&
+                r300->fs->uses_imms) {
             r300->fs->translated = FALSE;
             r300_translate_fragment_shader(r300, r300->fs);
-        } else if (shader == PIPE_SHADER_VERTEX && r300->vs) {
+        } else if (shader == PIPE_SHADER_VERTEX && r300->vs &&
+                r300->vs->uses_imms) {
             r300->vs->translated = FALSE;
             r300_translate_vertex_shader(r300, r300->vs);
         }
@@ -257,6 +264,7 @@ static void r300_set_edgeflags(struct pipe_context* pipe,
                                const unsigned* bitfield)
 {
     /* XXX you know it's bad when i915 has this blank too */
+    /* XXX and even worse, I have no idea WTF the bitfield is */
 }
 
 static void
@@ -277,18 +285,17 @@ static void* r300_create_fs_state(struct pipe_context* pipe,
                                   const struct pipe_shader_state* shader)
 {
     struct r300_context* r300 = r300_context(pipe);
-    struct r3xx_fragment_shader* fs = NULL;
+    struct r300_fragment_shader* fs = NULL;
 
     if (r300_screen(r300->context.screen)->caps->is_r500) {
-        fs =
-            (struct r3xx_fragment_shader*)CALLOC_STRUCT(r500_fragment_shader);
+        fs = (struct r300_fragment_shader*)CALLOC_STRUCT(r5xx_fragment_shader);
     } else {
-        fs =
-            (struct r3xx_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader);
+        fs = (struct r300_fragment_shader*)CALLOC_STRUCT(r3xx_fragment_shader);
     }
 
     /* Copy state directly into shader. */
     fs->state = *shader;
+    fs->state.tokens = tgsi_dup_tokens(shader->tokens);
 
     tgsi_scan_shader(shader->tokens, &fs->info);
 
@@ -299,7 +306,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe,
 static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
 {
     struct r300_context* r300 = r300_context(pipe);
-    struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader;
+    struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;
 
     if (fs == NULL) {
         r300->fs = NULL;
@@ -317,13 +324,15 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
 /* Delete fragment shader state. */
 static void r300_delete_fs_state(struct pipe_context* pipe, void* shader)
 {
+    struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;
+    FREE(fs->state.tokens);
     FREE(shader);
 }
 
 static void r300_set_polygon_stipple(struct pipe_context* pipe,
                                      const struct pipe_poly_stipple* state)
 {
-    /* XXX */
+    /* XXX no idea how to set this up, but not terribly important */
 }
 
 /* Create a new rasterizer state based on the CSO rasterizer state.
@@ -341,14 +350,21 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
     /* Copy rasterizer state for Draw. */
     rs->rs = *state;
 
+    rs->enable_vte = !state->bypass_vs_clip_and_viewport;
+
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+    rs->vap_control_status = R300_VC_NO_SWAP;
+#else
+    rs->vap_control_status = R300_VC_32BIT_SWAP;
+#endif
+
     /* If bypassing TCL, or if no TCL engine is present, turn off the HW TCL.
      * Else, enable HW TCL and force Draw's TCL off. */
     if (state->bypass_vs_clip_and_viewport ||
             !r300_screen(pipe->screen)->caps->has_tcl) {
-        rs->vap_control_status = R300_VAP_TCL_BYPASS;
+        rs->vap_control_status |= R300_VAP_TCL_BYPASS;
     } else {
         rs->rs.bypass_vs_clip_and_viewport = TRUE;
-        rs->vap_control_status = 0;
     }
 
     rs->point_size = pack_float_16_6x(state->point_size) |
@@ -412,6 +428,10 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
         rs->color_control = R300_SHADE_MODEL_SMOOTH;
     }
 
+    if (!state->flatshade_first) {
+        rs->color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST;
+    }
+
     return (void*)rs;
 }
 
@@ -555,40 +575,35 @@ static void r300_set_viewport_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    draw_flush(r300->draw);
-
-    if (r300_screen(r300->context.screen)->caps->has_tcl) {
-        /* Do the transform in HW. */
-        r300->viewport_state->vte_control = R300_VTX_W0_FMT;
+    /* Do the transform in HW. */
+    r300->viewport_state->vte_control = R300_VTX_W0_FMT;
 
-        if (state->scale[0] != 1.0f) {
-            r300->viewport_state->xscale = state->scale[0];
-            r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA;
-        }
-        if (state->scale[1] != 1.0f) {
-            r300->viewport_state->yscale = state->scale[1];
-            r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA;
-        }
-        if (state->scale[2] != 1.0f) {
-            r300->viewport_state->zscale = state->scale[2];
-            r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA;
-        }
-        if (state->translate[0] != 0.0f) {
-            r300->viewport_state->xoffset = state->translate[0];
-            r300->viewport_state->vte_control |= R300_VPORT_X_OFFSET_ENA;
-        }
-        if (state->translate[1] != 0.0f) {
-            r300->viewport_state->yoffset = state->translate[1];
-            r300->viewport_state->vte_control |= R300_VPORT_Y_OFFSET_ENA;
-        }
-        if (state->translate[2] != 0.0f) {
-            r300->viewport_state->zoffset = state->translate[2];
-            r300->viewport_state->vte_control |= R300_VPORT_Z_OFFSET_ENA;
-        }
-    } else {
-        r300->viewport_state->vte_control = 0;
-        /* Have Draw do the actual transform. */
-        draw_set_viewport_state(r300->draw, state);
+    if (state->scale[0] != 1.0f) {
+        assert(state->scale[0] != 0.0f);
+        r300->viewport_state->xscale = state->scale[0];
+        r300->viewport_state->vte_control |= R300_VPORT_X_SCALE_ENA;
+    }
+    if (state->scale[1] != 1.0f) {
+        assert(state->scale[1] != 0.0f);
+        r300->viewport_state->yscale = state->scale[1];
+        r300->viewport_state->vte_control |= R300_VPORT_Y_SCALE_ENA;
+    }
+    if (state->scale[2] != 1.0f) {
+        assert(state->scale[2] != 0.0f);
+        r300->viewport_state->zscale = state->scale[2];
+        r300->viewport_state->vte_control |= R300_VPORT_Z_SCALE_ENA;
+    }
+    if (state->translate[0] != 0.0f) {
+        r300->viewport_state->xoffset = state->translate[0];
+        r300->viewport_state->vte_control |= R300_VPORT_X_OFFSET_ENA;
+    }
+    if (state->translate[1] != 0.0f) {
+        r300->viewport_state->yoffset = state->translate[1];
+        r300->viewport_state->vte_control |= R300_VPORT_Y_OFFSET_ENA;
+    }
+    if (state->translate[2] != 0.0f) {
+        r300->viewport_state->zoffset = state->translate[2];
+        r300->viewport_state->vte_control |= R300_VPORT_Z_OFFSET_ENA;
     }
 
     r300->dirty_state |= R300_NEW_VIEWPORT;
@@ -628,6 +643,7 @@ static void* r300_create_vs_state(struct pipe_context* pipe,
         struct r300_vertex_shader* vs = CALLOC_STRUCT(r300_vertex_shader);
         /* Copy state directly into shader. */
         vs->state = *shader;
+        vs->state.tokens = tgsi_dup_tokens(shader->tokens);
 
         tgsi_scan_shader(shader->tokens, &vs->info);
 
@@ -673,6 +689,7 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader)
         struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
 
         draw_delete_vertex_shader(r300->draw, vs->draw);
+        FREE(vs->state.tokens);
         FREE(shader);
     } else {
         draw_delete_vertex_shader(r300->draw,
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index c4c9784a00..2477b30822 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -64,6 +64,7 @@ static void r300_vs_tab_routes(struct r300_context* r300,
                     break;
                 case TGSI_SEMANTIC_FOG:
                     fog = TRUE;
+                    /* Fall through */
                 case TGSI_SEMANTIC_GENERIC:
                     texs++;
                     break;
@@ -103,6 +104,9 @@ static void r300_vs_tab_routes(struct r300_context* r300,
         }
     }
 
+    /* XXX magic */
+    assert(texs <= 8);
+
     /* Do the actual vertex_info setup.
      *
      * vertex_info has four uints of hardware-specific data in it.
@@ -140,21 +144,32 @@ static void r300_vs_tab_routes(struct r300_context* r300,
         vinfo->hwfmt[2] |= (R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i);
     }
 
-    for (i = 0; i < texs; i++) {
+    /* Init i right here, increment it if fog is enabled.
+     * This gets around a double-increment problem. */
+    i = 0;
+
+    if (fog) {
+        i++;
         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE,
-            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i));
+            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0));
         vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i);
         vinfo->hwfmt[3] |= (4 << (3 * i));
     }
 
-    if (fog) {
-        i++;
+    for (i; i < texs; i++) {
         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE,
-            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_FOG, 0));
+            draw_find_vs_output(r300->draw, TGSI_SEMANTIC_GENERIC, i));
         vinfo->hwfmt[1] |= (R300_INPUT_CNTL_TC0 << i);
         vinfo->hwfmt[3] |= (4 << (3 * i));
     }
 
+    /* Handle the case where the vertex shader will be generating some of
+     * the attribs based on its inputs. */
+    if (r300screen->caps->has_tcl &&
+            info->num_inputs < info->num_outputs) {
+        vinfo->num_attribs = info->num_inputs;
+    }
+
     draw_compute_vertex_size(vinfo);
 }
 
@@ -162,26 +177,40 @@ static void r300_vs_tab_routes(struct r300_context* r300,
 static void r300_vertex_psc(struct r300_context* r300,
                             struct r300_vertex_format* vformat)
 {
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
     struct vertex_info* vinfo = &vformat->vinfo;
     int* tab = vformat->vs_tab;
     uint32_t temp;
-    int i;
+    int i, attrib_count;
 
-    debug_printf("r300: attrib count: %d\n", vinfo->num_attribs);
-    for (i = 0; i < vinfo->num_attribs; i++) {
-        debug_printf("r300: attrib: offset %d, interp %d, size %d,"
-               " tab %d\n", vinfo->attrib[i].src_index,
-               vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit,
-               tab[i]);
+    /* Vertex shaders have no semantics on their inputs,
+     * so PSC should just route stuff based on their info,
+     * and not on attrib information. */
+    if (r300screen->caps->has_tcl) {
+        attrib_count = r300->vs->info.num_inputs;
+        debug_printf("r300: routing %d attribs in psc for vs\n",
+                attrib_count);
+    } else {
+        attrib_count = vinfo->num_attribs;
+        debug_printf("r300: attrib count: %d\n", attrib_count);
+        for (i = 0; i < attrib_count; i++) {
+            debug_printf("r300: attrib: offset %d, interp %d, size %d,"
+                   " tab %d\n", vinfo->attrib[i].src_index,
+                   vinfo->attrib[i].interp_mode, vinfo->attrib[i].emit,
+                   tab[i]);
+        }
     }
 
-    for (i = 0; i < vinfo->num_attribs; i++) {
+    for (i = 0; i < attrib_count; i++) {
         /* Make sure we have a proper destination for our attribute */
         assert(tab[i] != -1);
 
         /* Add the attribute to the PSC table. */
-        temp = translate_vertex_data_type(vinfo->attrib[i].emit) |
-            (tab[i] << R300_DST_VEC_LOC_SHIFT);
+        temp = r300screen->caps->has_tcl ?
+            R300_DATA_TYPE_FLOAT_4 :
+            translate_vertex_data_type(vinfo->attrib[i].emit);
+        temp |= tab[i] << R300_DST_VEC_LOC_SHIFT;
+
         if (i & 1) {
             vformat->vap_prog_stream_cntl[i >> 1] &= 0x0000ffff;
             vformat->vap_prog_stream_cntl[i >> 1] |= temp << 16;
@@ -206,7 +235,6 @@ static void r300_vertex_psc(struct r300_context* r300,
 /* Update the vertex format. */
 static void r300_update_vertex_format(struct r300_context* r300)
 {
-    struct r300_screen* r300screen = r300_screen(r300->context.screen);
     struct r300_vertex_format vformat;
     int i;
 
diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h
index 91b93fc367..22c8e199ae 100644
--- a/src/gallium/drivers/r300/r300_state_inlines.h
+++ b/src/gallium/drivers/r300/r300_state_inlines.h
@@ -353,25 +353,6 @@ static INLINE uint32_t r300_translate_out_fmt(enum pipe_format format)
 
 /* Non-CSO state. (For now.) */
 
-static INLINE uint32_t r300_translate_gb_pipes(int pipe_count)
-{
-    switch (pipe_count) {
-        case 1:
-            return R300_GB_TILE_PIPE_COUNT_RV300;
-            break;
-        case 2:
-            return R300_GB_TILE_PIPE_COUNT_R300;
-            break;
-        case 3:
-            return R300_GB_TILE_PIPE_COUNT_R420_3P;
-            break;
-        case 4:
-            return R300_GB_TILE_PIPE_COUNT_R420;
-            break;
-    }
-    return 0;
-}
-
 static INLINE uint32_t translate_vertex_data_type(int type) {
     switch (type) {
         case EMIT_1F:
diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c
index 8bd9b41bd7..9f534b8ce3 100644
--- a/src/gallium/drivers/r300/r300_state_invariant.c
+++ b/src/gallium/drivers/r300/r300_state_invariant.c
@@ -34,36 +34,22 @@ void r300_emit_invariant_state(struct r300_context* r300)
     struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
     CS_LOCALS(r300);
 
-    BEGIN_CS(30 + (caps->has_tcl ? 2: 0));
+    BEGIN_CS(22 + (caps->has_tcl ? 2: 0));
 
     /*** Graphics Backend (GB) ***/
     /* Various GB enables */
     OUT_CS_REG(R300_GB_ENABLE, 0x0);
-    /* Subpixel multisampling for AA */
-    OUT_CS_REG(R300_GB_MSPOS0, 0x66666666);
-    OUT_CS_REG(R300_GB_MSPOS1, 0x66666666);
-    /* GB tile config and pipe setup */
-    OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_DISABLE |
-        r300_translate_gb_pipes(caps->num_frag_pipes));
+    /* Subpixel multisampling for AA
+     * These are commented out because glisse's CS checker doesn't like them.
+     * I presume these will be re-enabled later.
+     * OUT_CS_REG(R300_GB_MSPOS0, 0x66666666);
+     * OUT_CS_REG(R300_GB_MSPOS1, 0x6666666);
+     */
     /* Source of fog depth */
     OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W);
     /* AA enable */
     OUT_CS_REG(R300_GB_AA_CONFIG, 0x0);
 
-    /*** Geometry Assembly (GA) ***/
-    /* GA errata fixes. */
-    if (caps->is_r500) {
-        OUT_CS_REG(R300_GA_ENHANCE,
-                R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL |
-                R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE |
-                R500_GA_ENHANCE_REG_READWRITE_ENABLE |
-                R500_GA_ENHANCE_REG_NOSTALL_ENABLE);
-    } else {
-        OUT_CS_REG(R300_GA_ENHANCE,
-                R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL |
-                R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE);
-    }
-
     /*** Fog (FG) ***/
     OUT_CS_REG(R300_FG_FOG_BLEND, 0x0);
     OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x0);
@@ -86,7 +72,7 @@ void r300_emit_invariant_state(struct r300_context* r300)
     END_CS;
 
     /* XXX unsorted stuff from surface_fill */
-    BEGIN_CS(79 + (caps->has_tcl ? 7 : 0));
+    BEGIN_CS(75 + (caps->has_tcl ? 5 : 0));
     /* Flush PVS. */
     OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
 
@@ -94,30 +80,26 @@ void r300_emit_invariant_state(struct r300_context* r300)
         R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
         R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
         R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT);
-    /* XXX endian */
     if (caps->has_tcl) {
-        OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP);
-        OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE |
-            R300_PS_UCP_MODE_CLIP_AS_TRIFAN);
         OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4);
         OUT_CS_32F(1.0);
         OUT_CS_32F(1.0);
         OUT_CS_32F(1.0);
         OUT_CS_32F(1.0);
-    } else {
-        OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP |
-                R300_VAP_TCL_BYPASS);
     }
     /* XXX point tex stuffing */
     OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1);
     OUT_CS_32F(0.0);
     OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1);
     OUT_CS_32F(1.0);
+    /* XXX line tex stuffing */
+    OUT_CS_REG_SEQ(R300_GA_LINE_S0, 1);
+    OUT_CS_32F(0.0);
+    OUT_CS_REG_SEQ(R300_GA_LINE_S1, 1);
+    OUT_CS_32F(1.0);
     OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 |
         (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT));
     /* XXX this big chunk should be refactored into rs_state */
-    OUT_CS_REG(R300_GA_LINE_S0, 0x00000000);
-    OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000);
     OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000);
     OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000);
     OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000);
@@ -144,8 +126,6 @@ void r300_emit_invariant_state(struct r300_context* r300)
     OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1);
     OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405);
     OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F);
-    /* Vertex size. */
-    OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8);
 
     /* XXX */
     OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa);
diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c
index 4dd5b8af99..fdabe4d9cf 100644
--- a/src/gallium/drivers/r300/r300_surface.c
+++ b/src/gallium/drivers/r300/r300_surface.c
@@ -23,30 +23,36 @@
 
 #include "r300_surface.h"
 
-static void r300_surface_setup(struct pipe_context* pipe,
-                               struct pipe_surface* dest,
+static void r300_surface_setup(struct r300_context* r300,
+                               struct r300_texture* dest,
                                unsigned x, unsigned y,
                                unsigned w, unsigned h)
 {
-    struct r300_context* r300 = r300_context(pipe);
-    struct r300_capabilities* caps = r300_screen(pipe->screen)->caps;
-    struct r300_texture* tex = (struct r300_texture*)dest->texture;
-    unsigned pixpitch = tex->stride / tex->tex.block.size;
+    struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
+    unsigned pixpitch = dest->stride / dest->tex.block.size;
     CS_LOCALS(r300);
 
-    /* Make sure our target BO is okay. */
-    r300->winsys->add_buffer(r300->winsys, tex->buffer,
-            0, RADEON_GEM_DOMAIN_VRAM);
-    if (r300->winsys->validate(r300->winsys)) {
-        r300->context.flush(&r300->context, 0, NULL);
-    }
-
     r300_emit_blend_state(r300, &blend_clear_state);
     r300_emit_blend_color_state(r300, &blend_color_clear_state);
     r300_emit_dsa_state(r300, &dsa_clear_state);
     r300_emit_rs_state(r300, &rs_clear_state);
 
-    BEGIN_CS(15);
+    BEGIN_CS(24);
+
+    /* Viewport setup */
+    OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
+    OUT_CS_32F((float)w);
+    OUT_CS_32F((float)x);
+    OUT_CS_32F((float)h);
+    OUT_CS_32F((float)y);
+    OUT_CS_32F(1.0);
+    OUT_CS_32F(0.0);
+
+    OUT_CS_REG(R300_VAP_VTE_CNTL, R300_VPORT_X_SCALE_ENA |
+            R300_VPORT_X_OFFSET_ENA |
+            R300_VPORT_Y_SCALE_ENA |
+            R300_VPORT_Y_OFFSET_ENA |
+            R300_VTX_XY_FMT | R300_VTX_Z_FMT);
 
     /* Pixel scissors. */
     OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2);
@@ -71,9 +77,9 @@ static void r300_surface_setup(struct pipe_context* pipe,
 
     /* Setup colorbuffer. */
     OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1);
-    OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+    OUT_CS_RELOC(dest->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
     OUT_CS_REG(R300_RB3D_COLORPITCH0, pixpitch |
-        r300_translate_colorformat(tex->tex.format));
+        r300_translate_colorformat(dest->tex.format));
     OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0xf);
 
     END_CS;
@@ -93,6 +99,7 @@ static void r300_surface_fill(struct pipe_context* pipe,
     struct r300_capabilities* caps = r300_screen(pipe->screen)->caps;
     struct r300_texture* tex = (struct r300_texture*)dest->texture;
     unsigned pixpitch = tex->stride / tex->tex.block.size;
+    boolean invalid = FALSE;
     CS_LOCALS(r300);
 
     a = (float)((color >> 24) & 0xff) / 255.0f;
@@ -105,12 +112,29 @@ static void r300_surface_fill(struct pipe_context* pipe,
 
     /* Fallback? */
     if (FALSE) {
+fallback:
         debug_printf("r300: Falling back on surface clear...");
         util_surface_fill(pipe, dest, x, y, w, h, color);
         return;
     }
 
-    r300_surface_setup(r300, dest, x, y, w, h);
+    /* Make sure our target BO is okay. */
+validate:
+    if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
+                0, RADEON_GEM_DOMAIN_VRAM)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    if (r300->winsys->validate(r300->winsys)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        if (invalid) {
+            goto fallback;
+        }
+        invalid = TRUE;
+        goto validate;
+    }
+
+    r300_surface_setup(r300, tex, x, y, w, h);
 
     /* Vertex shader setup */
     if (caps->has_tcl) {
@@ -127,14 +151,14 @@ static void r300_surface_fill(struct pipe_context* pipe,
 
     /* Fragment shader setup */
     if (caps->is_r500) {
-        r500_emit_fragment_shader(r300, &r500_passthrough_fragment_shader);
-        r300_emit_rs_block_state(r300, &r500_rs_block_clear_state);
+        r500_emit_fragment_shader(r300, &r5xx_passthrough_fragment_shader);
+        r300_emit_rs_block_state(r300, &r5xx_rs_block_clear_state);
     } else {
-        r300_emit_fragment_shader(r300, &r300_passthrough_fragment_shader);
-        r300_emit_rs_block_state(r300, &r300_rs_block_clear_state);
+        r300_emit_fragment_shader(r300, &r3xx_passthrough_fragment_shader);
+        r300_emit_rs_block_state(r300, &r3xx_rs_block_clear_state);
     }
 
-    BEGIN_CS(31);
+    BEGIN_CS(26);
 
     /* VAP stream control, mapping from input memory to PVS/RS memory */
     if (caps->has_tcl) {
@@ -161,27 +185,21 @@ static void r300_surface_fill(struct pipe_context* pipe,
     /* Disable textures */
     OUT_CS_REG(R300_TX_ENABLE, 0x0);
 
-    /* Viewport setup */
-    OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
-    OUT_CS_32F(1.0);
-    OUT_CS_32F((float)x);
-    OUT_CS_32F(1.0);
-    OUT_CS_32F((float)y);
-    OUT_CS_32F(1.0);
-    OUT_CS_32F(0.0);
-
     /* The size of the point we're about to draw, in sixths of pixels */
     OUT_CS_REG(R300_GA_POINT_SIZE,
-        ((h * 6) & R300_POINTSIZE_Y_MASK) |
+        ((h * 6)  & R300_POINTSIZE_Y_MASK) |
         ((w * 6) << R300_POINTSIZE_X_SHIFT));
 
+    /* Vertex size. */
+    OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8);
+
     /* Packet3 with our point vertex */
     OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 8);
     OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
             (1 << R300_PRIM_NUM_VERTICES_SHIFT));
     /* Position */
-    OUT_CS_32F(w / 2.0);
-    OUT_CS_32F(h / 2.0);
+    OUT_CS_32F(0.5);
+    OUT_CS_32F(0.5);
     OUT_CS_32F(1.0);
     OUT_CS_32F(1.0);
     /* Color */
@@ -190,11 +208,7 @@ static void r300_surface_fill(struct pipe_context* pipe,
     OUT_CS_32F(b);
     OUT_CS_32F(a);
 
-    /* XXX figure out why this is 0xA and not 0x2 */
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA);
-    /* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
-        R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
-        R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */
 
     END_CS;
 
@@ -213,27 +227,55 @@ static void r300_surface_copy(struct pipe_context* pipe,
     struct r300_texture* srctex = (struct r300_texture*)src->texture;
     struct r300_texture* desttex = (struct r300_texture*)dest->texture;
     unsigned pixpitch = srctex->stride / srctex->tex.block.size;
+    boolean invalid = FALSE;
+    float fsrcx = srcx, fsrcy = srcy, fdestx = destx, fdesty = desty;
     CS_LOCALS(r300);
 
     debug_printf("r300: Copying surface %p at (%d,%d) to %p at (%d, %d),"
         " dimensions %dx%d (pixel pitch %d)\n",
         src, srcx, srcy, dest, destx, desty, w, h, pixpitch);
 
-    if ((srctex == desttex) &&
+    if ((srctex->buffer == desttex->buffer) &&
             ((destx < srcx + w) || (srcx < destx + w)) &&
             ((desty < srcy + h) || (srcy < desty + h))) {
+fallback:
         debug_printf("r300: Falling back on surface_copy\n");
         util_surface_copy(pipe, FALSE, dest, destx, desty, src,
                 srcx, srcy, w, h);
     }
 
-    r300_emit_sampler(r300, &r300_sampler_copy_state, 0);
-    r300_emit_texture(r300, srctex, 0);
+    /* Add our target BOs to the list. */
+validate:
+    if (!r300->winsys->add_buffer(r300->winsys, srctex->buffer,
+                RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    if (!r300->winsys->add_buffer(r300->winsys, desttex->buffer,
+                0, RADEON_GEM_DOMAIN_VRAM)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
+    if (r300->winsys->validate(r300->winsys)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        if (invalid) {
+            goto fallback;
+        }
+        invalid = TRUE;
+        goto validate;
+    }
+
+    r300_surface_setup(r300, desttex, destx, desty, w, h);
+
+    /* Setup the texture. */
+    r300_emit_texture(r300, &r300_sampler_copy_state, srctex, 0);
+
+    /* Flush and enable. */
     r300_flush_textures(r300);
 
     /* Vertex shader setup */
     if (caps->has_tcl) {
-        r300_emit_vertex_shader(r300, &r300_texture_vertex_shader);
+        r300_emit_vertex_shader(r300, &r300_passthrough_vertex_shader);
     } else {
         BEGIN_CS(4);
         OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VAP_TCL_BYPASS);
@@ -246,13 +288,14 @@ static void r300_surface_copy(struct pipe_context* pipe,
 
     /* Fragment shader setup */
     if (caps->is_r500) {
-        r500_emit_fragment_shader(r300, &r500_texture_fragment_shader);
-        r300_emit_rs_block_state(r300, &r500_rs_block_copy_state);
+        r500_emit_fragment_shader(r300, &r5xx_texture_fragment_shader);
+        r300_emit_rs_block_state(r300, &r5xx_rs_block_copy_state);
     } else {
-        r300_emit_fragment_shader(r300, &r300_texture_fragment_shader);
-        r300_emit_rs_block_state(r300, &r300_rs_block_copy_state);
+        r300_emit_fragment_shader(r300, &r3xx_texture_fragment_shader);
+        r300_emit_rs_block_state(r300, &r3xx_rs_block_copy_state);
     }
 
+    BEGIN_CS(30);
     /* VAP stream control, mapping from input memory to PVS/RS memory */
     if (caps->has_tcl) {
         OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0,
@@ -275,33 +318,38 @@ static void r300_surface_copy(struct pipe_context* pipe,
     /* Two components of texture 0 */
     OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x2);
 
+    /* Vertex size. */
+    OUT_CS_REG(R300_VAP_VTX_SIZE, 0x4);
+
     /* Packet3 with our texcoords */
-    OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 8);
+    OUT_CS_PKT3(R200_3D_DRAW_IMMD_2, 16);
     OUT_CS(R300_PRIM_TYPE_QUADS | R300_PRIM_WALK_RING |
             (4 << R300_PRIM_NUM_VERTICES_SHIFT));
     /* (x    , y    ) */
-    OUT_CS_32F((float)destx);
-    OUT_CS_32F((float)desty);
-    OUT_CS_32F((float)srcx);
-    OUT_CS_32F((float)srcy);
+    OUT_CS_32F(fdestx / dest->width);
+    OUT_CS_32F(fdesty / dest->height);
+    OUT_CS_32F(fsrcx  / src->width);
+    OUT_CS_32F(fsrcy  / src->height);
     /* (x    , y + h) */
-    OUT_CS_32F((float)destx);
-    OUT_CS_32F((float)(desty + h));
-    OUT_CS_32F((float)srcx);
-    OUT_CS_32F((float)(srcy + h));
+    OUT_CS_32F(fdestx / dest->width);
+    OUT_CS_32F((fdesty + h) / dest->height);
+    OUT_CS_32F(fsrcx  / src->width);
+    OUT_CS_32F((fsrcy  + h) / src->height);
     /* (x + w, y + h) */
-    OUT_CS_32F((float)(destx + w));
-    OUT_CS_32F((float)(desty + h));
-    OUT_CS_32F((float)(srcx + w));
-    OUT_CS_32F((float)(srcy + h));
+    OUT_CS_32F((fdestx + w) / dest->width);
+    OUT_CS_32F((fdesty + h) / dest->height);
+    OUT_CS_32F((fsrcx  + w) / src->width);
+    OUT_CS_32F((fsrcy  + h) / src->height);
     /* (x + w, y    ) */
-    OUT_CS_32F((float)(destx + w));
-    OUT_CS_32F((float)desty);
-    OUT_CS_32F((float)(srcx + w));
-    OUT_CS_32F((float)srcy);
+    OUT_CS_32F((fdestx + w) / dest->width);
+    OUT_CS_32F(fdesty / dest->height);
+    OUT_CS_32F((fsrcx  + w) / src->width);
+    OUT_CS_32F(fsrcy  / src->height);
 
     OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA);
 
+    END_CS;
+
     r300->dirty_hw++;
 }
 
diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h
index 894def07aa..d01f0b143f 100644
--- a/src/gallium/drivers/r300/r300_surface.h
+++ b/src/gallium/drivers/r300/r300_surface.h
@@ -31,8 +31,8 @@
 #include "r300_context.h"
 #include "r300_cs.h"
 #include "r300_emit.h"
-#include "r300_state_shader.h"
-#include "r300_state_tcl.h"
+#include "r300_fs.h"
+#include "r300_vs.h"
 #include "r300_state_inlines.h"
 
 static struct r300_blend_state blend_clear_state = {
@@ -72,7 +72,7 @@ static struct r300_rs_state rs_clear_state = {
     .color_control = R300_SHADE_MODEL_FLAT,
 };
 
-static struct r300_rs_block r300_rs_block_clear_state = {
+static struct r300_rs_block r3xx_rs_block_clear_state = {
     .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) |
         R500_RS_SEL_T(R300_RS_SEL_K0) |
         R500_RS_SEL_R(R300_RS_SEL_K0) |
@@ -82,7 +82,7 @@ static struct r300_rs_block r300_rs_block_clear_state = {
     .inst_count = 0,
 };
 
-static struct r300_rs_block r500_rs_block_clear_state = {
+static struct r300_rs_block r5xx_rs_block_clear_state = {
     .ip[0] = R500_RS_SEL_S(R500_RS_IP_PTR_K0) |
         R500_RS_SEL_T(R500_RS_IP_PTR_K0) |
         R500_RS_SEL_R(R500_RS_IP_PTR_K0) |
@@ -94,24 +94,24 @@ static struct r300_rs_block r500_rs_block_clear_state = {
 
 /* The following state is used for surface_copy only. */
 
-static struct r300_rs_block r300_rs_block_copy_state = {
+static struct r300_rs_block r3xx_rs_block_copy_state = {
     .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) |
         R500_RS_SEL_T(R300_RS_SEL_K0) |
         R500_RS_SEL_R(R300_RS_SEL_K0) |
         R500_RS_SEL_Q(R300_RS_SEL_K1),
     .inst[0] = R300_RS_INST_COL_CN_WRITE,
     .count = R300_IT_COUNT(2) | R300_IC_COUNT(0) | R300_HIRES_EN,
-    .inst_count = R300_RS_TX_OFFSET(6),
+    .inst_count = R300_RS_TX_OFFSET(0),
 };
 
-static struct r300_rs_block r500_rs_block_copy_state = {
+static struct r300_rs_block r5xx_rs_block_copy_state = {
     .ip[0] = R500_RS_SEL_S(0) |
         R500_RS_SEL_T(1) |
         R500_RS_SEL_R(R500_RS_IP_PTR_K0) |
         R500_RS_SEL_Q(R500_RS_IP_PTR_K1),
     .inst[0] = R500_RS_INST_TEX_CN_WRITE,
     .count = R300_IT_COUNT(2) | R300_IC_COUNT(0) | R300_HIRES_EN,
-    .inst_count = R300_RS_TX_OFFSET(6),
+    .inst_count = R300_RS_TX_OFFSET(0),
 };
 
 static struct r300_sampler_state r300_sampler_copy_state = {
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index fe91f4e184..11c7858d42 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -22,6 +22,8 @@
 
 #include "r300_texture.h"
 
+/* XXX maths need to go to util */
+
 static int minify(int i)
 {
     return MAX2(1, i >> 1);
@@ -30,25 +32,29 @@ static int minify(int i)
 static void r300_setup_texture_state(struct r300_texture* tex,
                                      unsigned width,
                                      unsigned height,
-                                     unsigned pitch)
+                                     unsigned pitch,
+                                     unsigned levels)
 {
     struct r300_texture_state* state = &tex->state;
 
     state->format0 = R300_TX_WIDTH((width - 1) & 0x7ff) |
-        R300_TX_HEIGHT((height - 1) & 0x7ff) | R300_TX_PITCH_EN;
+        R300_TX_HEIGHT((height - 1) & 0x7ff) |
+        R300_TX_NUM_LEVELS(levels) |
+        R300_TX_PITCH_EN;
 
     /* XXX */
-    state->format1 = R300_TX_FORMAT_A8R8G8B8;
+    state->format1 = r300_translate_texformat(tex->tex.format);
 
     state->format2 = pitch - 1;
 
-    /* XXX
+    /* Assume (somewhat foolishly) that oversized textures will
+     * not be permitted by the state tracker. */
     if (width > 2048) {
-        state->pitch |= R300_TXWIDTH_11;
+        state->format2 |= R500_TXWIDTH_BIT11;
     }
     if (height > 2048) {
-        state->pitch |= R300_TXHEIGHT_11;
-    } */
+        state->format2 |= R500_TXHEIGHT_BIT11;
+    }
 }
 
 static void r300_setup_miptree(struct r300_texture* tex)
@@ -67,14 +73,18 @@ static void r300_setup_miptree(struct r300_texture* tex)
         base->nblocksx[i] = pf_get_nblocksx(&base->block, base->width[i]);
         base->nblocksy[i] = pf_get_nblocksy(&base->block, base->width[i]);
 
-        /* Radeons enjoy things in multiples of 32. */
-        /* XXX this can be 32 when POT */
-        stride = (base->nblocksx[i] * base->block.size + 63) & ~63;
+        /* Radeons enjoy things in multiples of 64.
+         *
+         * XXX
+         * POT, uncompressed, unmippmapped textures can be aligned to 32,
+         * instead of 64. */
+        stride = align(base->nblocksx[i] * base->block.size, 64);
         size = stride * base->nblocksy[i] * base->depth[i];
 
-        tex->offset[i] = (tex->size + 63) & ~63;
+        tex->offset[i] = align(tex->size, 64);
         tex->size = tex->offset[i] + size;
 
+        /* Save stride of first level to the texture. */
         if (i == 0) {
             tex->stride = stride;
         }
@@ -86,8 +96,6 @@ static struct pipe_texture*
     r300_texture_create(struct pipe_screen* screen,
                         const struct pipe_texture* template)
 {
-    /* XXX struct r300_screen* r300screen = r300_screen(screen); */
-
     struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
 
     if (!tex) {
@@ -100,9 +108,8 @@ static struct pipe_texture*
 
     r300_setup_miptree(tex);
 
-    /* XXX */
-    r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0],
-            tex->tex.width[0]);
+    r300_setup_texture_state(tex, template->width[0], template->height[0],
+            template->width[0], template->last_level);
 
     tex->buffer = screen->buffer_create(screen, 64,
                                         PIPE_BUFFER_USAGE_PIXEL,
@@ -166,6 +173,7 @@ static struct pipe_texture*
 {
     struct r300_texture* tex;
 
+    /* XXX we should start doing mips now... */
     if (base->target != PIPE_TEXTURE_2D ||
         base->last_level != 0 ||
         base->depth[0] != 1) {
@@ -183,8 +191,9 @@ static struct pipe_texture*
 
     tex->stride = *stride;
 
+    /* XXX */
     r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0],
-            tex->stride);
+            tex->stride, 0);
 
     pipe_buffer_reference(&tex->buffer, buffer);
 
diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h
index 98fb5c9a08..3b56f0307c 100644
--- a/src/gallium/drivers/r300/r300_texture.h
+++ b/src/gallium/drivers/r300/r300_texture.h
@@ -32,6 +32,52 @@
 
 void r300_init_screen_texture_functions(struct pipe_screen* screen);
 
+/* Note the signature of R300_EASY_TX_FORMAT(A, R, G, B, FORMAT)... */
+static INLINE uint32_t r300_translate_texformat(enum pipe_format format)
+{
+    switch (format) {
+        /* X8 */
+        case PIPE_FORMAT_I8_UNORM:
+            return R300_EASY_TX_FORMAT(X, X, X, X, X8);
+        /* W8Z8Y8X8 */
+        case PIPE_FORMAT_A8R8G8B8_UNORM:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+        case PIPE_FORMAT_R8G8B8A8_UNORM:
+            return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8);
+        case PIPE_FORMAT_A8R8G8B8_SRGB:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8) |
+                R300_TX_FORMAT_GAMMA;
+        case PIPE_FORMAT_R8G8B8A8_SRGB:
+            return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) |
+                R300_TX_FORMAT_GAMMA;
+        /* DXT1 */
+        case PIPE_FORMAT_DXT1_RGB:
+            return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1);
+        case PIPE_FORMAT_DXT1_RGBA:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1);
+        /* DXT3 */
+        case PIPE_FORMAT_DXT3_RGBA:
+            return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3);
+        /* DXT5 */
+        case PIPE_FORMAT_DXT5_RGBA:
+            return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5);
+        /* YVYU422 */
+        case PIPE_FORMAT_YCBCR:
+            return R300_EASY_TX_FORMAT(X, Y, Z, ONE, YVYU422) |
+                R300_TX_FORMAT_YUV_TO_RGB;
+        /* W24_FP */
+        case PIPE_FORMAT_Z24S8_UNORM:
+            return R300_EASY_TX_FORMAT(X, X, X, X, W24_FP);
+        default:
+            debug_printf("r300: Implementation error: "
+                "Got unsupported texture format %s in %s\n",
+                pf_name(format), __FUNCTION__);
+            assert(0);
+            break;
+    }
+    return 0;
+}
+
 #ifndef R300_WINSYS_H
 
 boolean r300_get_texture_buffer(struct pipe_texture* texture,
diff --git a/src/gallium/drivers/r300/r300_state_tcl.c b/src/gallium/drivers/r300/r300_vs.c
index d84912de48..f87435f9f0 100644
--- a/src/gallium/drivers/r300/r300_state_tcl.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -20,7 +20,7 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
-#include "r300_state_tcl.h"
+#include "r300_vs.h"
 
 static void r300_vs_declare(struct r300_vs_asm* assembler,
                             struct tgsi_full_declaration* decl)
@@ -71,16 +71,13 @@ static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler,
 {
     switch (src->File) {
         case TGSI_FILE_NULL:
-            /* Probably a zero or one swizzle */
-            return R300_PVS_SRC_REG_INPUT;
-            break;
         case TGSI_FILE_INPUT:
+            /* Probably a zero or one swizzle */
             return R300_PVS_SRC_REG_INPUT;
-            break;
         case TGSI_FILE_TEMPORARY:
             return R300_PVS_SRC_REG_TEMPORARY;
-            break;
         case TGSI_FILE_CONSTANT:
+        case TGSI_FILE_IMMEDIATE:
             return R300_PVS_SRC_REG_CONSTANT;
         default:
             debug_printf("r300: vs: Unimplemented src type %d\n", src->File);
@@ -89,16 +86,32 @@ static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler,
     return 0;
 }
 
+static INLINE unsigned r300_vs_src(struct r300_vs_asm* assembler,
+                                   struct tgsi_src_register* src)
+{
+    switch (src->File) {
+        case TGSI_FILE_NULL:
+        case TGSI_FILE_INPUT:
+        case TGSI_FILE_TEMPORARY:
+        case TGSI_FILE_CONSTANT:
+            return src->Index;
+        case TGSI_FILE_IMMEDIATE:
+            return src->Index + assembler->imm_offset;
+        default:
+            debug_printf("r300: vs: Unimplemented src type %d\n", src->File);
+            break;
+    }
+    return 0;
+}
+
 static INLINE unsigned r300_vs_dst_type(struct r300_vs_asm* assembler,
                                         struct tgsi_dst_register* dst)
 {
     switch (dst->File) {
         case TGSI_FILE_TEMPORARY:
             return R300_PVS_DST_REG_TEMPORARY;
-            break;
         case TGSI_FILE_OUTPUT:
             return R300_PVS_DST_REG_OUT;
-            break;
         default:
             debug_printf("r300: vs: Unimplemented dst type %d\n", dst->File);
             break;
@@ -112,10 +125,8 @@ static INLINE unsigned r300_vs_dst(struct r300_vs_asm* assembler,
     switch (dst->File) {
         case TGSI_FILE_TEMPORARY:
             return dst->Index;
-            break;
         case TGSI_FILE_OUTPUT:
             return assembler->tab[dst->Index];
-            break;
         default:
             debug_printf("r300: vs: Unimplemented dst %d\n", dst->File);
             break;
@@ -133,8 +144,15 @@ static uint32_t r300_vs_op(unsigned op)
             return R300_VE_MULTIPLY;
         case TGSI_OPCODE_ADD:
         case TGSI_OPCODE_MOV:
+        case TGSI_OPCODE_SUB:
         case TGSI_OPCODE_SWZ:
             return R300_VE_ADD;
+        case TGSI_OPCODE_MAX:
+            return R300_VE_MAXIMUM;
+        case TGSI_OPCODE_SLT:
+            return R300_VE_SET_LESS_THAN;
+        case TGSI_OPCODE_RSQ:
+            return R300_PVS_DST_MATH_INST | R300_ME_RECIP_DX;
         case TGSI_OPCODE_MAD:
             return R300_PVS_DST_MACRO_INST | R300_PVS_MACRO_OP_2CLK_MADD;
         default:
@@ -146,51 +164,78 @@ static uint32_t r300_vs_op(unsigned op)
 static uint32_t r300_vs_swiz(struct tgsi_full_src_register* reg)
 {
     if (reg->SrcRegister.Extended) {
-        return reg->SrcRegisterExtSwz.ExtSwizzleX |
+        return (reg->SrcRegister.Negate ? (0xf << 12) : 0) |
+            reg->SrcRegisterExtSwz.ExtSwizzleX |
             (reg->SrcRegisterExtSwz.ExtSwizzleY << 3) |
             (reg->SrcRegisterExtSwz.ExtSwizzleZ << 6) |
             (reg->SrcRegisterExtSwz.ExtSwizzleW << 9);
     } else {
-        return reg->SrcRegister.SwizzleX |
+        return (reg->SrcRegister.Negate ? (0xf << 12) : 0) |
+            reg->SrcRegister.SwizzleX |
             (reg->SrcRegister.SwizzleY << 3) |
             (reg->SrcRegister.SwizzleZ << 6) |
             (reg->SrcRegister.SwizzleW << 9);
     }
 }
 
+/* XXX icky icky icky icky */
+static uint32_t r300_vs_scalar_swiz(struct tgsi_full_src_register* reg)
+{
+    if (reg->SrcRegister.Extended) {
+        return (reg->SrcRegister.Negate ? (0xf << 12) : 0) |
+            reg->SrcRegisterExtSwz.ExtSwizzleX |
+            (reg->SrcRegisterExtSwz.ExtSwizzleX << 3) |
+            (reg->SrcRegisterExtSwz.ExtSwizzleX << 6) |
+            (reg->SrcRegisterExtSwz.ExtSwizzleX << 9);
+    } else {
+        return (reg->SrcRegister.Negate ? (0xf << 12) : 0) |
+            reg->SrcRegister.SwizzleX |
+            (reg->SrcRegister.SwizzleX << 3) |
+            (reg->SrcRegister.SwizzleX << 6) |
+            (reg->SrcRegister.SwizzleX << 9);
+    }
+}
+
+/* XXX scalar stupidity */
 static void r300_vs_emit_inst(struct r300_vertex_shader* vs,
                               struct r300_vs_asm* assembler,
                               struct tgsi_full_src_register* src,
                               struct tgsi_full_dst_register* dst,
                               unsigned op,
-                              unsigned count)
+                              unsigned count,
+                              boolean is_scalar)
 {
     int i = vs->instruction_count;
     vs->instructions[i].inst0 = R300_PVS_DST_OPCODE(r300_vs_op(op)) |
         R300_PVS_DST_REG_TYPE(r300_vs_dst_type(assembler, &dst->DstRegister)) |
         R300_PVS_DST_OFFSET(r300_vs_dst(assembler, &dst->DstRegister)) |
-        R300_PVS_DST_WE_XYZW;
+        R300_PVS_DST_WE(dst->DstRegister.WriteMask);
     switch (count) {
         case 3:
             vs->instructions[i].inst3 =
                 R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
                             &src[2].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[2].SrcRegister.Index) |
+                R300_PVS_SRC_OFFSET(r300_vs_src(assembler,
+                            &src[2].SrcRegister)) |
                 R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[2]));
             /* Fall through */
         case 2:
             vs->instructions[i].inst2 =
                 R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
                             &src[1].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[1].SrcRegister.Index) |
+                R300_PVS_SRC_OFFSET(r300_vs_src(assembler,
+                            &src[1].SrcRegister)) |
                 R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[1]));
             /* Fall through */
         case 1:
             vs->instructions[i].inst1 =
                 R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler,
                             &src[0].SrcRegister)) |
-                R300_PVS_SRC_OFFSET(src[0].SrcRegister.Index) |
-                R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[0]));
+                R300_PVS_SRC_OFFSET(r300_vs_src(assembler,
+                            &src[0].SrcRegister)) |
+                /* XXX the icky, it burns */
+                R300_PVS_SRC_SWIZZLE(is_scalar ? r300_vs_scalar_swiz(&src[0])
+                        : r300_vs_swiz(&src[0]));
             break;
     }
     vs->instruction_count++;
@@ -201,11 +246,22 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs,
                                 struct tgsi_full_instruction* inst)
 {
     switch (inst->Instruction.Opcode) {
+        case TGSI_OPCODE_RSQ:
+            r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], inst->Instruction.Opcode,
+                    1, TRUE);
+            break;
+        case TGSI_OPCODE_SUB:
+            inst->FullSrcRegisters[1].SrcRegister.Negate =
+                !inst->FullSrcRegisters[1].SrcRegister.Negate;
+            /* Fall through */
         case TGSI_OPCODE_ADD:
         case TGSI_OPCODE_MUL:
+        case TGSI_OPCODE_MAX:
+        case TGSI_OPCODE_SLT:
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
+                    2, FALSE);
             break;
         case TGSI_OPCODE_DP3:
             /* Set alpha swizzle to zero for src0 and src1 */
@@ -235,19 +291,19 @@ static void r300_vs_instruction(struct r300_vertex_shader* vs,
         case TGSI_OPCODE_DP4:
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
+                    2, FALSE);
             break;
         case TGSI_OPCODE_MOV:
         case TGSI_OPCODE_SWZ:
             inst->FullSrcRegisters[1] = r300_constant_zero;
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    2);
+                    2, FALSE);
             break;
         case TGSI_OPCODE_MAD:
             r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode,
-                    3);
+                    3, FALSE);
             break;
         case TGSI_OPCODE_END:
             break;
@@ -278,6 +334,8 @@ static void r300_vs_init(struct r300_vertex_shader* vs,
                 break;
         }
     }
+
+    vs->instruction_count = 0;
 }
 
 void r300_translate_vertex_shader(struct r300_context* r300,
@@ -337,6 +395,7 @@ void r300_translate_vertex_shader(struct r300_context* r300,
             assembler->tex_count + assembler->color_count);
 
     consts->count = consts->user_count + assembler->imm_count;
+    vs->uses_imms = assembler->imm_count;
     debug_printf("r300: vs: %d total constants, "
             "%d from user and %d from immediates\n", consts->count,
             consts->user_count, assembler->imm_count);
@@ -344,7 +403,7 @@ void r300_translate_vertex_shader(struct r300_context* r300,
     debug_printf("r300: vs: tab: %d %d %d %d\n", assembler->tab[0],
             assembler->tab[1], assembler->tab[2], assembler->tab[3]);
 
-    tgsi_dump(vs->state.tokens);
+    tgsi_dump(vs->state.tokens, 0);
     /* XXX finish r300 vertex shader dumper */
     r300_vs_dump(vs);
 
diff --git a/src/gallium/drivers/r300/r300_state_tcl.h b/src/gallium/drivers/r300/r300_vs.h
index e2e1357d43..165d717812 100644
--- a/src/gallium/drivers/r300/r300_state_tcl.h
+++ b/src/gallium/drivers/r300/r300_vs.h
@@ -20,21 +20,27 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
-#ifndef R300_STATE_TCL_H
-#define R300_STATE_TCL_H
+#ifndef R300_VS_H
+#define R300_VS_H
 
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_dump.h"
 
 #include "r300_context.h"
 #include "r300_debug.h"
 #include "r300_reg.h"
 #include "r300_screen.h"
+#include "r300_shader_inlines.h"
 
 /* XXX get these to r300_reg */
 #define R300_PVS_DST_OPCODE(x)   ((x) << 0)
 #   define R300_VE_DOT_PRODUCT            1
 #   define R300_VE_MULTIPLY               2
 #   define R300_VE_ADD                    3
+#   define R300_VE_MAXIMUM                7
+#   define R300_VE_SET_LESS_THAN          10
+#define R300_PVS_DST_MATH_INST     (1 << 6)
+#   define R300_ME_RECIP_DX               6
 #define R300_PVS_DST_MACRO_INST    (1 << 7)
 #   define R300_PVS_MACRO_OP_2CLK_MADD    0
 #define R300_PVS_DST_REG_TYPE(x) ((x) << 8)
@@ -72,15 +78,13 @@
     ((R300_PVS_SRC_SELECT_FORCE_1 | (R300_PVS_SRC_SELECT_FORCE_1 << 3) | \
      (R300_PVS_SRC_SELECT_FORCE_1 << 6) | \
       (R300_PVS_SRC_SELECT_FORCE_1 << 9)) << 13)
-
-static const struct tgsi_full_src_register r300_constant_zero = {
-    .SrcRegister.Extended = TRUE,
-    .SrcRegister.File = TGSI_FILE_NULL,
-    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO,
-};
+#define R300_PVS_MODIFIER_X        (1 << 25)
+#define R300_PVS_MODIFIER_Y        (1 << 26)
+#define R300_PVS_MODIFIER_Z        (1 << 27)
+#define R300_PVS_MODIFIER_W        (1 << 28)
+#define R300_PVS_NEGATE_XYZW \
+    (R300_PVS_MODIFIER_X | R300_PVS_MODIFIER_Y | \
+     R300_PVS_MODIFIER_Z | R300_PVS_MODIFIER_W)
 
 /* Temporary struct used to hold assembly state while putting together
  * fragment programs. */
@@ -150,4 +154,4 @@ static struct r300_vertex_shader r300_texture_vertex_shader = {
 void r300_translate_vertex_shader(struct r300_context* r300,
                                   struct r300_vertex_shader* vs);
 
-#endif /* R300_STATE_TCL_H */
+#endif /* R300_VS_H */
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index a833bb0399..d2893c3b9d 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -45,9 +45,6 @@ struct r300_winsys {
     /* PCI ID */
     uint32_t pci_id;
 
-    /* GB pipe count */
-    uint32_t gb_pipes;
-
     /* GART size. */
     uint32_t gart_size;
 
@@ -55,10 +52,10 @@ struct r300_winsys {
     uint32_t vram_size;
 
     /* Add a pipe_buffer to the list of buffer objects to validate. */
-    void (*add_buffer)(struct r300_winsys* winsys,
-                       struct pipe_buffer* pbuffer,
-                       uint32_t rd,
-                       uint32_t wd);
+    boolean (*add_buffer)(struct r300_winsys* winsys,
+                          struct pipe_buffer* pbuffer,
+                          uint32_t rd,
+                          uint32_t wd);
 
     /* Revalidate all currently setup pipe_buffers.
      * Returns TRUE if a flush is required. */
diff --git a/src/gallium/drivers/r300/r3xx_fs.c b/src/gallium/drivers/r300/r3xx_fs.c
new file mode 100644
index 0000000000..6e05d76977
--- /dev/null
+++ b/src/gallium/drivers/r300/r3xx_fs.c
@@ -0,0 +1,96 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r3xx_fs.h"
+
+static INLINE uint32_t r3xx_rgb_op(unsigned op)
+{
+    switch (op) {
+        case TGSI_OPCODE_MOV:
+            return R300_ALU_OUTC_CMP;
+        default:
+            return 0;
+    }
+}
+
+static INLINE uint32_t r3xx_alpha_op(unsigned op)
+{
+    switch (op) {
+        case TGSI_OPCODE_MOV:
+            return R300_ALU_OUTA_CMP;
+        default:
+            return 0;
+    }
+}
+
+static INLINE void r3xx_emit_maths(struct r3xx_fragment_shader* fs,
+                                   struct r300_fs_asm* assembler,
+                                   struct tgsi_full_src_register* src,
+                                   struct tgsi_full_dst_register* dst,
+                                   unsigned op,
+                                   unsigned count)
+{
+    int i = fs->alu_instruction_count;
+
+    fs->instructions[i].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
+        r3xx_rgb_op(op);
+    fs->instructions[i].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
+        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ;
+    fs->instructions[i].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
+        r3xx_alpha_op(op);
+    fs->instructions[i].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
+        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT;
+
+    fs->alu_instruction_count++;
+}
+
+void r3xx_fs_finalize(struct r300_fragment_shader* fs,
+                      struct r300_fs_asm* assembler)
+{
+    fs->stack_size = assembler->temp_count + assembler->temp_offset + 1;
+}
+
+void r3xx_fs_instruction(struct r3xx_fragment_shader* fs,
+                         struct r300_fs_asm* assembler,
+                         struct tgsi_full_instruction* inst)
+{
+    switch (inst->Instruction.Opcode) {
+        case TGSI_OPCODE_MOV:
+            /* src0 -> src1 and src2 forced to zero */
+            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
+            inst->FullSrcRegisters[2] = r300_constant_zero;
+            r3xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
+            break;
+        case TGSI_OPCODE_END:
+            break;
+        default:
+            debug_printf("r300: fs: Bad opcode %d\n",
+                    inst->Instruction.Opcode);
+            break;
+    }
+}
diff --git a/src/gallium/drivers/r300/r3xx_fs.h b/src/gallium/drivers/r300/r3xx_fs.h
new file mode 100644
index 0000000000..3da39ec252
--- /dev/null
+++ b/src/gallium/drivers/r300/r3xx_fs.h
@@ -0,0 +1,76 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R3XX_FS_H
+#define R3XX_FS_H
+
+#include "r300_fs_inlines.h"
+
+static struct r3xx_fragment_shader r3xx_passthrough_fragment_shader = {
+    .alu_instruction_count = 1,
+    .tex_instruction_count = 0,
+    .indirections = 0,
+    .shader.stack_size = 1,
+
+    .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
+        R300_ALU_OUTC_CMP,
+    .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
+        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
+    .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
+        R300_ALU_OUTA_CMP,
+    .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
+        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
+};
+
+static struct r3xx_fragment_shader r3xx_texture_fragment_shader = {
+    .alu_instruction_count = 1,
+    .tex_instruction_count = 0,
+    .indirections = 0,
+    .shader.stack_size = 1,
+
+    .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
+        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
+        R300_ALU_OUTC_CMP,
+    .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
+        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
+    .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
+        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
+        R300_ALU_OUTA_CMP,
+    .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
+        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
+};
+
+void r3xx_fs_finalize(struct r300_fragment_shader* fs,
+                      struct r300_fs_asm* assembler);
+
+void r3xx_fs_instruction(struct r3xx_fragment_shader* fs,
+                         struct r300_fs_asm* assembler,
+                         struct tgsi_full_instruction* inst);
+
+#endif /* R3XX_FS_H */
diff --git a/src/gallium/drivers/r300/r300_state_shader.c b/src/gallium/drivers/r300/r5xx_fs.c
index 1b02239ee7..99d826278c 100644
--- a/src/gallium/drivers/r300/r300_state_shader.c
+++ b/src/gallium/drivers/r300/r5xx_fs.c
@@ -1,5 +1,6 @@
 /*
  * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -20,107 +21,9 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
-#include "r300_state_shader.h"
+#include "r5xx_fs.h"
 
-static void r300_copy_passthrough_shader(struct r300_fragment_shader* fs)
-{
-    struct r300_fragment_shader* pt = &r300_passthrough_fragment_shader;
-    fs->shader.stack_size = pt->shader.stack_size;
-    fs->alu_instruction_count = pt->alu_instruction_count;
-    fs->tex_instruction_count = pt->tex_instruction_count;
-    fs->indirections = pt->indirections;
-    fs->instructions[0] = pt->instructions[0];
-}
-
-static void r500_copy_passthrough_shader(struct r500_fragment_shader* fs)
-{
-    struct r500_fragment_shader* pt = &r500_passthrough_fragment_shader;
-    fs->shader.stack_size = pt->shader.stack_size;
-    fs->instruction_count = pt->instruction_count;
-    fs->instructions[0] = pt->instructions[0];
-}
-
-static void r300_fs_declare(struct r300_fs_asm* assembler,
-                            struct tgsi_full_declaration* decl)
-{
-    switch (decl->Declaration.File) {
-        case TGSI_FILE_INPUT:
-            switch (decl->Semantic.SemanticName) {
-                case TGSI_SEMANTIC_COLOR:
-                    assembler->color_count++;
-                    break;
-                case TGSI_SEMANTIC_GENERIC:
-                    assembler->tex_count++;
-                    break;
-                default:
-                    debug_printf("r300: fs: Bad semantic declaration %d\n",
-                        decl->Semantic.SemanticName);
-                    break;
-            }
-            break;
-        case TGSI_FILE_OUTPUT:
-        case TGSI_FILE_CONSTANT:
-            break;
-        case TGSI_FILE_TEMPORARY:
-            assembler->temp_count++;
-            break;
-        default:
-            debug_printf("r300: fs: Bad file %d\n", decl->Declaration.File);
-            break;
-    }
-
-    assembler->temp_offset = assembler->color_count + assembler->tex_count;
-}
-
-static INLINE unsigned r300_fs_src(struct r300_fs_asm* assembler,
-                                   struct tgsi_src_register* src)
-{
-    switch (src->File) {
-        case TGSI_FILE_NULL:
-            return 0;
-        case TGSI_FILE_INPUT:
-            /* XXX may be wrong */
-            return src->Index;
-            break;
-        case TGSI_FILE_TEMPORARY:
-            return src->Index + assembler->temp_offset;
-            break;
-        case TGSI_FILE_IMMEDIATE:
-            return (src->Index + assembler->imm_offset) | (1 << 8);
-            break;
-        case TGSI_FILE_CONSTANT:
-            /* XXX magic */
-            return src->Index | (1 << 8);
-            break;
-        default:
-            debug_printf("r300: fs: Unimplemented src %d\n", src->File);
-            break;
-    }
-    return 0;
-}
-
-static INLINE unsigned r300_fs_dst(struct r300_fs_asm* assembler,
-                                   struct tgsi_dst_register* dst)
-{
-    switch (dst->File) {
-        case TGSI_FILE_NULL:
-            /* This happens during KIL instructions. */
-            return 0;
-            break;
-        case TGSI_FILE_OUTPUT:
-            return 0;
-            break;
-        case TGSI_FILE_TEMPORARY:
-            return dst->Index + assembler->temp_offset;
-            break;
-        default:
-            debug_printf("r300: fs: Unimplemented dst %d\n", dst->File);
-            break;
-    }
-    return 0;
-}
-
-static INLINE unsigned r500_fix_swiz(unsigned s)
+static INLINE unsigned r5xx_fix_swiz(unsigned s)
 {
     /* For historical reasons, the swizzle values x, y, z, w, and 0 are
      * equivalent to the actual machine code, but 1 is not. Thus, we just
@@ -132,13 +35,13 @@ static INLINE unsigned r500_fix_swiz(unsigned s)
     }
 }
 
-static uint32_t r500_rgba_swiz(struct tgsi_full_src_register* reg)
+static uint32_t r5xx_rgba_swiz(struct tgsi_full_src_register* reg)
 {
     if (reg->SrcRegister.Extended) {
-        return r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleX) |
-            (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleY) << 3) |
-            (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleZ) << 6) |
-            (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleW) << 9);
+        return r5xx_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleX) |
+            (r5xx_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleY) << 3) |
+            (r5xx_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleZ) << 6) |
+            (r5xx_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleW) << 9);
     } else {
         return reg->SrcRegister.SwizzleX |
             (reg->SrcRegister.SwizzleY << 3) |
@@ -147,7 +50,7 @@ static uint32_t r500_rgba_swiz(struct tgsi_full_src_register* reg)
     }
 }
 
-static uint32_t r500_strq_swiz(struct tgsi_full_src_register* reg)
+static uint32_t r5xx_strq_swiz(struct tgsi_full_src_register* reg)
 {
     return reg->SrcRegister.SwizzleX |
         (reg->SrcRegister.SwizzleY << 2) |
@@ -155,50 +58,36 @@ static uint32_t r500_strq_swiz(struct tgsi_full_src_register* reg)
         (reg->SrcRegister.SwizzleW << 6);
 }
 
-static INLINE uint32_t r500_rgb_swiz(struct tgsi_full_src_register* reg)
+static INLINE uint32_t r5xx_rgb_swiz(struct tgsi_full_src_register* reg)
 {
     /* Only the first 9 bits... */
-    return (r500_rgba_swiz(reg) & 0x1ff) |
+    return (r5xx_rgba_swiz(reg) & 0x1ff) |
         (reg->SrcRegister.Negate ? (1 << 9) : 0) |
         (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0);
 }
 
-static INLINE uint32_t r500_alpha_swiz(struct tgsi_full_src_register* reg)
+static INLINE uint32_t r5xx_alpha_swiz(struct tgsi_full_src_register* reg)
 {
     /* Only the last 3 bits... */
-    return (r500_rgba_swiz(reg) >> 9) |
+    return (r5xx_rgba_swiz(reg) >> 9) |
         (reg->SrcRegister.Negate ? (1 << 9) : 0) |
         (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0);
 }
 
-static INLINE uint32_t r300_rgb_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_MOV:
-            return R300_ALU_OUTC_CMP;
-        default:
-            return 0;
-    }
-}
-
-static INLINE uint32_t r300_alpha_op(unsigned op)
-{
-    switch (op) {
-        case TGSI_OPCODE_MOV:
-            return R300_ALU_OUTA_CMP;
-        default:
-            return 0;
-    }
-}
-
-static INLINE uint32_t r500_rgba_op(unsigned op)
+static INLINE uint32_t r5xx_rgba_op(unsigned op)
 {
     switch (op) {
+        case TGSI_OPCODE_COS:
         case TGSI_OPCODE_EX2:
         case TGSI_OPCODE_LG2:
         case TGSI_OPCODE_RCP:
         case TGSI_OPCODE_RSQ:
+        case TGSI_OPCODE_SIN:
             return R500_ALU_RGBA_OP_SOP;
+        case TGSI_OPCODE_DDX:
+            return R500_ALU_RGBA_OP_MDH;
+        case TGSI_OPCODE_DDY:
+            return R500_ALU_RGBA_OP_MDV;
         case TGSI_OPCODE_FRC:
             return R500_ALU_RGBA_OP_FRC;
         case TGSI_OPCODE_DP3:
@@ -221,9 +110,11 @@ static INLINE uint32_t r500_rgba_op(unsigned op)
     }
 }
 
-static INLINE uint32_t r500_alpha_op(unsigned op)
+static INLINE uint32_t r5xx_alpha_op(unsigned op)
 {
     switch (op) {
+        case TGSI_OPCODE_COS:
+            return R500_ALPHA_OP_COS;
         case TGSI_OPCODE_EX2:
             return R500_ALPHA_OP_EX2;
         case TGSI_OPCODE_LG2:
@@ -234,6 +125,12 @@ static INLINE uint32_t r500_alpha_op(unsigned op)
             return R500_ALPHA_OP_RSQ;
         case TGSI_OPCODE_FRC:
             return R500_ALPHA_OP_FRC;
+        case TGSI_OPCODE_SIN:
+            return R500_ALPHA_OP_SIN;
+        case TGSI_OPCODE_DDX:
+            return R500_ALPHA_OP_MDH;
+        case TGSI_OPCODE_DDY:
+            return R500_ALPHA_OP_MDV;
         case TGSI_OPCODE_DP3:
         case TGSI_OPCODE_DP4:
         case TGSI_OPCODE_DPH:
@@ -253,7 +150,7 @@ static INLINE uint32_t r500_alpha_op(unsigned op)
     }
 }
 
-static INLINE uint32_t r500_tex_op(unsigned op)
+static INLINE uint32_t r5xx_tex_op(unsigned op)
 {
     switch (op) {
         case TGSI_OPCODE_KIL:
@@ -269,64 +166,35 @@ static INLINE uint32_t r500_tex_op(unsigned op)
     }
 }
 
-static INLINE void r300_emit_maths(struct r300_fragment_shader* fs,
+/* Setup an ALU operation. */
+static INLINE void r5xx_emit_maths(struct r5xx_fragment_shader* fs,
                                    struct r300_fs_asm* assembler,
                                    struct tgsi_full_src_register* src,
                                    struct tgsi_full_dst_register* dst,
                                    unsigned op,
                                    unsigned count)
 {
-    int i = fs->alu_instruction_count;
-
-    fs->instructions[i].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
-        r300_rgb_op(op);
-    fs->instructions[i].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
-        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ;
-    fs->instructions[i].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
-        r300_alpha_op(op);
-    fs->instructions[i].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
-        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT;
-
-    fs->alu_instruction_count++;
-}
-
-/* Setup an ALU operation. */
-static INLINE void r500_emit_alu(struct r500_fragment_shader* fs,
-                                 struct r300_fs_asm* assembler,
-                                 struct tgsi_full_dst_register* dst)
-{
     int i = fs->instruction_count;
 
     if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
-        fs->instructions[i].inst0 = R500_INST_TYPE_OUT |
-        R500_ALU_OMASK(dst->DstRegister.WriteMask);
+        fs->instructions[i].inst0 = R500_INST_TYPE_OUT;
+        if (r300_fs_is_depr(assembler, dst)) {
+            fs->instructions[i].inst4 = R500_W_OMASK;
+        } else {
+            fs->instructions[i].inst0 |=
+                R500_ALU_OMASK(dst->DstRegister.WriteMask);
+        }
     } else {
         fs->instructions[i].inst0 = R500_INST_TYPE_ALU |
-        R500_ALU_WMASK(dst->DstRegister.WriteMask);
+            R500_ALU_WMASK(dst->DstRegister.WriteMask);
     }
 
     fs->instructions[i].inst0 |= R500_INST_TEX_SEM_WAIT;
 
-    fs->instructions[i].inst4 =
+    fs->instructions[i].inst4 |=
         R500_ALPHA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister));
     fs->instructions[i].inst5 =
         R500_ALU_RGBA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister));
-}
-
-static INLINE void r500_emit_maths(struct r500_fragment_shader* fs,
-                                   struct r300_fs_asm* assembler,
-                                   struct tgsi_full_src_register* src,
-                                   struct tgsi_full_dst_register* dst,
-                                   unsigned op,
-                                   unsigned count)
-{
-    int i = fs->instruction_count;
-
-    r500_emit_alu(fs, assembler, dst);
 
     switch (count) {
         case 3:
@@ -336,9 +204,9 @@ static INLINE void r500_emit_maths(struct r500_fragment_shader* fs,
                 R500_ALPHA_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister));
             fs->instructions[i].inst5 |=
                 R500_ALU_RGBA_SEL_C_SRC2 |
-                R500_SWIZ_RGBA_C(r500_rgb_swiz(&src[2])) |
+                R500_SWIZ_RGBA_C(r5xx_rgb_swiz(&src[2])) |
                 R500_ALU_RGBA_ALPHA_SEL_C_SRC2 |
-                R500_SWIZ_ALPHA_C(r500_alpha_swiz(&src[2]));
+                R500_SWIZ_ALPHA_C(r5xx_alpha_swiz(&src[2]));
         case 2:
             fs->instructions[i].inst1 |=
                 R500_RGB_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister));
@@ -346,10 +214,10 @@ static INLINE void r500_emit_maths(struct r500_fragment_shader* fs,
                 R500_ALPHA_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister));
             fs->instructions[i].inst3 =
                 R500_ALU_RGB_SEL_B_SRC1 |
-                R500_SWIZ_RGB_B(r500_rgb_swiz(&src[1]));
+                R500_SWIZ_RGB_B(r5xx_rgb_swiz(&src[1]));
             fs->instructions[i].inst4 |=
-                R500_SWIZ_ALPHA_B(r500_alpha_swiz(&src[1])) |
-                R500_ALPHA_SEL_B_SRC1;
+                R500_ALPHA_SEL_B_SRC1 |
+                R500_SWIZ_ALPHA_B(r5xx_alpha_swiz(&src[1]));
         case 1:
         case 0:
         default:
@@ -359,20 +227,20 @@ static INLINE void r500_emit_maths(struct r500_fragment_shader* fs,
                 R500_ALPHA_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister));
             fs->instructions[i].inst3 |=
                 R500_ALU_RGB_SEL_A_SRC0 |
-                R500_SWIZ_RGB_A(r500_rgb_swiz(&src[0]));
+                R500_SWIZ_RGB_A(r5xx_rgb_swiz(&src[0]));
             fs->instructions[i].inst4 |=
-                R500_SWIZ_ALPHA_A(r500_alpha_swiz(&src[0])) |
-                R500_ALPHA_SEL_A_SRC0;
+                R500_ALPHA_SEL_A_SRC0 |
+                R500_SWIZ_ALPHA_A(r5xx_alpha_swiz(&src[0]));
             break;
     }
 
-    fs->instructions[i].inst4 |= r500_alpha_op(op);
-    fs->instructions[i].inst5 |= r500_rgba_op(op);
+    fs->instructions[i].inst4 |= r5xx_alpha_op(op);
+    fs->instructions[i].inst5 |= r5xx_rgba_op(op);
 
     fs->instruction_count++;
 }
 
-static INLINE void r500_emit_tex(struct r500_fragment_shader* fs,
+static INLINE void r5xx_emit_tex(struct r5xx_fragment_shader* fs,
                                  struct r300_fs_asm* assembler,
                                  struct tgsi_full_src_register* src,
                                  struct tgsi_full_dst_register* dst,
@@ -385,10 +253,10 @@ static INLINE void r500_emit_tex(struct r500_fragment_shader* fs,
         R500_INST_TEX_SEM_WAIT;
     fs->instructions[i].inst1 = R500_TEX_ID(0) |
         R500_TEX_SEM_ACQUIRE | //R500_TEX_IGNORE_UNCOVERED |
-        r500_tex_op(op);
+        r5xx_tex_op(op);
     fs->instructions[i].inst2 =
         R500_TEX_SRC_ADDR(r300_fs_src(assembler, &src->SrcRegister)) |
-        R500_SWIZ_TEX_STRQ(r500_strq_swiz(src)) |
+        R500_SWIZ_TEX_STRQ(r5xx_strq_swiz(src)) |
         R500_TEX_DST_ADDR(r300_fs_dst(assembler, &dst->DstRegister)) |
         R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G |
         R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A;
@@ -405,42 +273,32 @@ static INLINE void r500_emit_tex(struct r500_fragment_shader* fs,
         src[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
 
         src[1] = src[0];
-        src[2] = r500_constant_zero;
-        r500_emit_maths(fs, assembler, src, dst, TGSI_OPCODE_MOV, 3);
+        src[2] = r300_constant_zero;
+        r5xx_emit_maths(fs, assembler, src, dst, TGSI_OPCODE_MOV, 3);
     } else {
         fs->instruction_count++;
     }
 }
 
-static void r300_fs_instruction(struct r300_fragment_shader* fs,
-                                struct r300_fs_asm* assembler,
-                                struct tgsi_full_instruction* inst)
+void r5xx_fs_finalize(struct r5xx_fragment_shader* fs,
+                      struct r300_fs_asm* assembler)
 {
-    switch (inst->Instruction.Opcode) {
-        case TGSI_OPCODE_MOV:
-            /* src0 -> src1 and src2 forced to zero */
-            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[2] = r500_constant_zero;
-            r300_emit_maths(fs, assembler, inst->FullSrcRegisters,
-                    &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
-            break;
-        case TGSI_OPCODE_END:
-            break;
-        default:
-            debug_printf("r300: fs: Bad opcode %d\n",
-                    inst->Instruction.Opcode);
-            break;
-    }
+    /* XXX should this just go with OPCODE_END? */
+    fs->instructions[fs->instruction_count - 1].inst0 |=
+        R500_INST_LAST;
 }
 
-static void r500_fs_instruction(struct r500_fragment_shader* fs,
-                                struct r300_fs_asm* assembler,
-                                struct tgsi_full_instruction* inst)
+void r5xx_fs_instruction(struct r5xx_fragment_shader* fs,
+                         struct r300_fs_asm* assembler,
+                         struct tgsi_full_instruction* inst)
 {
     /* Switch between opcodes. When possible, prefer using the official
      * AMD/ATI names for opcodes, please, as it facilitates using the
      * documentation. */
     switch (inst->Instruction.Opcode) {
+        /* XXX trig needs extra prep */
+        case TGSI_OPCODE_COS:
+        case TGSI_OPCODE_SIN:
         /* The simple scalar ops. */
         case TGSI_OPCODE_EX2:
         case TGSI_OPCODE_LG2:
@@ -452,8 +310,10 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs,
             inst->FullSrcRegisters[0].SrcRegister.SwizzleW =
                 inst->FullSrcRegisters[0].SrcRegister.SwizzleX;
             /* Fall through */
+        case TGSI_OPCODE_DDX:
+        case TGSI_OPCODE_DDY:
         case TGSI_OPCODE_FRC:
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 1);
             break;
 
@@ -474,7 +334,7 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs,
             /* Fall through */
         case TGSI_OPCODE_DP3:
         case TGSI_OPCODE_DP4:
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 2);
             break;
 
@@ -484,7 +344,7 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs,
             inst->FullSrcRegisters[3] = inst->FullSrcRegisters[2];
             inst->FullSrcRegisters[2] = inst->FullSrcRegisters[0];
             inst->FullSrcRegisters[0] = inst->FullSrcRegisters[3];
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
             break;
 
@@ -498,18 +358,18 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs,
             /* Force src0 to one, move all registers over */
             inst->FullSrcRegisters[2] = inst->FullSrcRegisters[1];
             inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[0] = r500_constant_one;
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            inst->FullSrcRegisters[0] = r300_constant_one;
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
             break;
         case TGSI_OPCODE_MUL:
             /* Force our src2 to zero */
-            inst->FullSrcRegisters[2] = r500_constant_zero;
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            inst->FullSrcRegisters[2] = r300_constant_zero;
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
             break;
         case TGSI_OPCODE_MAD:
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
             break;
 
@@ -522,17 +382,71 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs,
         case TGSI_OPCODE_SWZ:
             /* src0 -> src1 and src2 forced to zero */
             inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0];
-            inst->FullSrcRegisters[2] = r500_constant_zero;
-            r500_emit_maths(fs, assembler, inst->FullSrcRegisters,
+            inst->FullSrcRegisters[2] = r300_constant_zero;
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3);
             break;
 
+        /* The compound and hybrid insts. */
+        case TGSI_OPCODE_LRP:
+            /* LRP DST A, B, C -> MAD TMP -A, C, C; MAD DST A, B, TMP */
+            inst->FullSrcRegisters[3] = inst->FullSrcRegisters[1];
+            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[2];
+            inst->FullSrcRegisters[0].SrcRegister.Negate =
+                !(inst->FullSrcRegisters[0].SrcRegister.Negate);
+            inst->FullDstRegisters[1] = inst->FullDstRegisters[0];
+            inst->FullDstRegisters[0].DstRegister.Index =
+                assembler->temp_count;
+            inst->FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], TGSI_OPCODE_MAD, 3);
+            inst->FullSrcRegisters[2].SrcRegister.Index =
+                assembler->temp_count;
+            inst->FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
+            inst->FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+            inst->FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+            inst->FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z;
+            inst->FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
+            inst->FullSrcRegisters[1] = inst->FullSrcRegisters[3];
+            inst->FullSrcRegisters[0].SrcRegister.Negate =
+                !(inst->FullSrcRegisters[0].SrcRegister.Negate);
+            inst->FullDstRegisters[0] = inst->FullDstRegisters[1];
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], TGSI_OPCODE_MAD, 3);
+            break;
+        case TGSI_OPCODE_POW:
+            /* POW DST A, B -> LG2 TMP A; MUL TMP TMP, B; EX2 DST TMP */
+            inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW =
+                inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX;
+            inst->FullSrcRegisters[0].SrcRegister.SwizzleW =
+                inst->FullSrcRegisters[0].SrcRegister.SwizzleX;
+            inst->FullDstRegisters[1] = inst->FullDstRegisters[0];
+            inst->FullDstRegisters[0].DstRegister.Index =
+                assembler->temp_count;
+            inst->FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], TGSI_OPCODE_LG2, 1);
+            inst->FullSrcRegisters[0].SrcRegister.Index =
+                assembler->temp_count;
+            inst->FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+            inst->FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+            inst->FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+            inst->FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z;
+            inst->FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
+            inst->FullSrcRegisters[2] = r300_constant_zero;
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], TGSI_OPCODE_MUL, 3);
+            inst->FullDstRegisters[0] = inst->FullDstRegisters[1];
+            r5xx_emit_maths(fs, assembler, inst->FullSrcRegisters,
+                    &inst->FullDstRegisters[0], TGSI_OPCODE_EX2, 1);
+            break;
+
         /* The texture instruction set. */
         case TGSI_OPCODE_KIL:
         case TGSI_OPCODE_TEX:
         case TGSI_OPCODE_TXB:
         case TGSI_OPCODE_TXP:
-            r500_emit_tex(fs, assembler, &inst->FullSrcRegisters[0],
+            r5xx_emit_tex(fs, assembler, &inst->FullSrcRegisters[0],
                     &inst->FullDstRegisters[0], inst->Instruction.Opcode);
             break;
 
@@ -551,99 +465,3 @@ static void r500_fs_instruction(struct r500_fragment_shader* fs,
             R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP;
     }
 }
-
-static void r300_fs_finalize(struct r3xx_fragment_shader* fs,
-                             struct r300_fs_asm* assembler)
-{
-    fs->stack_size = assembler->temp_count + assembler->temp_offset;
-}
-
-static void r500_fs_finalize(struct r500_fragment_shader* fs,
-                             struct r300_fs_asm* assembler)
-{
-    /* XXX should this just go with OPCODE_END? */
-    fs->instructions[fs->instruction_count - 1].inst0 |=
-        R500_INST_LAST;
-}
-
-void r300_translate_fragment_shader(struct r300_context* r300,
-                                    struct r3xx_fragment_shader* fs)
-{
-    struct tgsi_parse_context parser;
-    int i;
-    boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500;
-    struct r300_constant_buffer* consts =
-        &r300->shader_constants[PIPE_SHADER_FRAGMENT];
-
-    struct r300_fs_asm* assembler = CALLOC_STRUCT(r300_fs_asm);
-    if (assembler == NULL) {
-        return;
-    }
-    /* Setup starting offset for immediates. */
-    assembler->imm_offset = consts->user_count;
-
-    /* Make sure we start at the beginning of the shader. */
-    if (is_r500) {
-        ((struct r500_fragment_shader*)fs)->instruction_count = 0;
-    }
-
-    tgsi_parse_init(&parser, fs->state.tokens);
-
-    while (!tgsi_parse_end_of_tokens(&parser)) {
-        tgsi_parse_token(&parser);
-
-        /* This is seriously the lamest way to create fragment programs ever.
-         * I blame TGSI. */
-        switch (parser.FullToken.Token.Type) {
-            case TGSI_TOKEN_TYPE_DECLARATION:
-                /* Allocated registers sitting at the beginning
-                 * of the program. */
-                r300_fs_declare(assembler, &parser.FullToken.FullDeclaration);
-                break;
-            case TGSI_TOKEN_TYPE_IMMEDIATE:
-                debug_printf("r300: Emitting immediate to constant buffer, "
-                        "position %d\n",
-                        assembler->imm_offset + assembler->imm_count);
-                /* I am not amused by the length of these. */
-                for (i = 0; i < 4; i++) {
-                    consts->constants[assembler->imm_offset +
-                        assembler->imm_count][i] =
-                        parser.FullToken.FullImmediate.u.ImmediateFloat32[i]
-                        .Float;
-                }
-                assembler->imm_count++;
-                break;
-            case TGSI_TOKEN_TYPE_INSTRUCTION:
-                if (is_r500) {
-                    r500_fs_instruction((struct r500_fragment_shader*)fs,
-                            assembler, &parser.FullToken.FullInstruction);
-                } else {
-                    r300_fs_instruction((struct r300_fragment_shader*)fs,
-                            assembler, &parser.FullToken.FullInstruction);
-                }
-                break;
-        }
-    }
-
-    debug_printf("r300: fs: %d texs and %d colors, first free reg is %d\n",
-            assembler->tex_count, assembler->color_count,
-            assembler->tex_count + assembler->color_count);
-
-    consts->count = consts->user_count + assembler->imm_count;
-    debug_printf("r300: fs: %d total constants, "
-            "%d from user and %d from immediates\n", consts->count,
-            consts->user_count, assembler->imm_count);
-    r300_fs_finalize(fs, assembler);
-    if (is_r500) {
-        r500_fs_finalize((struct r500_fragment_shader*)fs, assembler);
-    }
-
-    tgsi_dump(fs->state.tokens);
-    /* XXX finish r300 dumper too */
-    if (is_r500) {
-        r500_fs_dump((struct r500_fragment_shader*)fs);
-    }
-
-    tgsi_parse_free(&parser);
-    FREE(assembler);
-}
diff --git a/src/gallium/drivers/r300/r300_state_shader.h b/src/gallium/drivers/r300/r5xx_fs.h
index 185fdd90f0..629e587be4 100644
--- a/src/gallium/drivers/r300/r300_state_shader.h
+++ b/src/gallium/drivers/r300/r5xx_fs.h
@@ -1,5 +1,6 @@
 /*
  * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *                Joakim Sindholt <opensource@zhasha.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -20,15 +21,10 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
-#ifndef R300_STATE_SHADER_H
-#define R300_STATE_SHADER_H
+#ifndef R5XX_FS_H
+#define R5XX_FS_H
 
-#include "tgsi/tgsi_parse.h"
-
-#include "r300_context.h"
-#include "r300_debug.h"
-#include "r300_reg.h"
-#include "r300_screen.h"
+#include "r300_fs_inlines.h"
 
 /* XXX this all should find its way back to r300_reg */
 /* Swizzle tools */
@@ -57,82 +53,9 @@
 #define R500_TEX_WMASK(x) ((x) << 11)
 #define R500_ALU_WMASK(x) ((x) << 11)
 #define R500_ALU_OMASK(x) ((x) << 15)
+#define R500_W_OMASK (1 << 31)
 
-/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're
- * not using enough of it. */
-static const struct tgsi_full_src_register r500_constant_zero = {
-    .SrcRegister.Extended = TRUE,
-    .SrcRegister.File = TGSI_FILE_NULL,
-    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO,
-    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO,
-};
-
-static const struct tgsi_full_src_register r500_constant_one = {
-    .SrcRegister.Extended = TRUE,
-    .SrcRegister.File = TGSI_FILE_NULL,
-    .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE,
-    .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE,
-    .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE,
-    .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE,
-};
-
-/* Temporary struct used to hold assembly state while putting together
- * fragment programs. */
-struct r300_fs_asm {
-    /* Pipe context. */
-    struct r300_context* r300;
-    /* Number of colors. */
-    unsigned color_count;
-    /* Number of texcoords. */
-    unsigned tex_count;
-    /* Offset for temporary registers. Inputs and temporaries have no
-     * distinguishing markings, so inputs start at 0 and the first usable
-     * temporary register is after all inputs. */
-    unsigned temp_offset;
-    /* Number of requested temporary registers. */
-    unsigned temp_count;
-    /* Offset for immediate constants. Neither R300 nor R500 can do four
-     * inline constants per source, so instead we copy immediates into the
-     * constant buffer. */
-    unsigned imm_offset;
-    /* Number of immediate constants. */
-    unsigned imm_count;
-};
-
-void r300_translate_fragment_shader(struct r300_context* r300,
-                           struct r3xx_fragment_shader* fs);
-
-static struct r300_fragment_shader r300_passthrough_fragment_shader = {
-    /* XXX This is the emission code. TODO: decode
-    OUT_CS_REG(R300_US_CONFIG, 0);
-    OUT_CS_REG(R300_US_CODE_OFFSET, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000);
-*/
-    .alu_instruction_count = 1,
-    .tex_instruction_count = 0,
-    .indirections = 0,
-    .shader.stack_size = 1,
-
-    .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
-        R300_ALU_OUTC_CMP,
-    .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
-        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
-    .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
-        R300_ALU_OUTA_CMP,
-    .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
-        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
-};
-
-static struct r500_fragment_shader r500_passthrough_fragment_shader = {
+static struct r5xx_fragment_shader r5xx_passthrough_fragment_shader = {
     .shader.stack_size = 0,
     .instruction_count = 1,
     .instructions[0].inst0 = R500_INST_TYPE_OUT |
@@ -158,40 +81,12 @@ static struct r500_fragment_shader r500_passthrough_fragment_shader = {
         R500_ALU_RGBA_A_SWIZ_0,
 };
 
-static struct r300_fragment_shader r300_texture_fragment_shader = {
-    /* XXX This is the emission code. TODO: decode
-    OUT_CS_REG(R300_US_CONFIG, 0);
-    OUT_CS_REG(R300_US_CODE_OFFSET, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000);
-*/
-    .alu_instruction_count = 1,
-    .tex_instruction_count = 0,
-    .indirections = 0,
-    .shader.stack_size = 1,
-
-    .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) |
-        R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) |
-        R300_ALU_OUTC_CMP,
-    .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) |
-        R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ,
-    .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) |
-        R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) |
-        R300_ALU_OUTA_CMP,
-    .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) |
-        R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT,
-};
-
-static struct r500_fragment_shader r500_texture_fragment_shader = {
+static struct r5xx_fragment_shader r5xx_texture_fragment_shader = {
     .shader.stack_size = 1,
     .instruction_count = 2,
     .instructions[0].inst0 = R500_INST_TYPE_TEX |
         R500_INST_TEX_SEM_WAIT |
-        R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK |
+        R500_INST_RGB_WMASK_RGB | R500_INST_ALPHA_WMASK |
         R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP,
     .instructions[0].inst1 = R500_TEX_ID(0) | R500_TEX_INST_LD |
         R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED,
@@ -227,4 +122,11 @@ static struct r500_fragment_shader r500_texture_fragment_shader = {
         R500_ALU_RGBA_A_SWIZ_0,
 };
 
-#endif /* R300_STATE_SHADER_H */
+void r5xx_fs_finalize(struct r5xx_fragment_shader* fs,
+                      struct r300_fs_asm* assembler);
+
+void r5xx_fs_instruction(struct r5xx_fragment_shader* fs,
+                         struct r300_fs_asm* assembler,
+                         struct tgsi_full_instruction* inst);
+
+#endif /* R5XX_FS_H */
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index dffc15a4f1..7888c2f644 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -92,6 +92,7 @@ struct softpipe_context {
     * queries.
     */
    uint64_t occlusion_count;
+   unsigned active_query_count;
 
    /** Mapped vertex buffers */
    ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS];
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index ba2766ff13..d4045816d0 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -35,6 +35,7 @@
 #include "pipe/p_context.h"
 #include "pipe/internal/p_winsys_screen.h"
 #include "pipe/p_inlines.h"
+#include "util/u_prim.h"
 
 #include "sp_context.h"
 #include "sp_state.h"
@@ -65,6 +66,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp)
                                    size);
 }
 
+
 static void
 softpipe_unmap_constant_buffers(struct softpipe_context *sp)
 {
@@ -86,20 +88,6 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp)
 }
 
 
-static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = {
-   PIPE_PRIM_POINTS,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_LINES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES,
-   PIPE_PRIM_TRIANGLES
-};
-
-
 boolean
 softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
                      unsigned start, unsigned count)
@@ -108,15 +96,11 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
 }
 
 
-
 /**
  * Draw vertex arrays, with optional indexing.
  * Basically, map the vertex buffers (and drawing surfaces), then hand off
  * the drawing to the 'draw' module.
- *
- * XXX should the element buffer be specified/bound with a separate function?
  */
-
 boolean
 softpipe_draw_range_elements(struct pipe_context *pipe,
                              struct pipe_buffer *indexBuffer,
@@ -129,7 +113,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
    struct draw_context *draw = sp->draw;
    unsigned i;
 
-   sp->reduced_api_prim = reduced_prim[mode];
+   sp->reduced_api_prim = u_reduced_prim(mode);
 
    if (sp->dirty)
       softpipe_update_derived( sp );
@@ -147,6 +131,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
                                     PIPE_BUFFER_USAGE_CPU_READ);
       draw_set_mapped_vertex_buffer(draw, i, buf);
    }
+
    /* Map index buffer, if present */
    if (indexBuffer) {
       void *mapped_indexes
@@ -159,10 +144,10 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
    }
    else {
       /* no index/element buffer */
-      draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL);
+      draw_set_mapped_element_buffer_range(draw, 0, start,
+                                           start + count - 1, NULL);
    }
 
-
    /* draw! */
    draw_arrays(draw, mode, start, count);
 
@@ -187,6 +172,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe,
    return TRUE;
 }
 
+
 boolean
 softpipe_draw_elements(struct pipe_context *pipe,
                        struct pipe_buffer *indexBuffer,
@@ -200,11 +186,9 @@ softpipe_draw_elements(struct pipe_context *pipe,
 }
 
 
-
 void
 softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags)
 {
    struct softpipe_context *sp = softpipe_context(pipe);
    draw_set_edgeflags(sp->draw, edgeflags);
 }
-
diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c
index 31c3ca21c5..f4fa0905d7 100644
--- a/src/gallium/drivers/softpipe/sp_fs_sse.c
+++ b/src/gallium/drivers/softpipe/sp_fs_sse.c
@@ -45,17 +45,6 @@
 
 #include "rtasm/rtasm_x86sse.h"
 
-/* Surely this should be defined somewhere in a tgsi header:
- */
-typedef void (PIPE_CDECL *codegen_function)(
-   const struct tgsi_exec_vector *input,
-   struct tgsi_exec_vector *output,
-   const float (*constant)[4],
-   struct tgsi_exec_vector *temporary,
-   const struct tgsi_interp_coef *coef,
-   float (*immediates)[4]
-   //, const struct tgsi_exec_vector *quadPos
- );
 
 
 /**
@@ -65,7 +54,7 @@ struct sp_sse_fragment_shader
 {
    struct sp_fragment_shader base;
    struct x86_function sse2_program;
-   codegen_function func;
+   tgsi_sse2_fs_function func;
    float immediates[TGSI_EXEC_NUM_IMMEDIATES][4];
 };
 
@@ -83,6 +72,7 @@ fs_sse_prepare( const struct sp_fragment_shader *base,
 		struct tgsi_exec_machine *machine,
 		struct tgsi_sampler **samplers )
 {
+   machine->Samplers = samplers;
 }
 
 
@@ -107,12 +97,10 @@ fs_sse_run( const struct sp_fragment_shader *base,
    tgsi_set_kill_mask(machine, 0x0);
    tgsi_set_exec_mask(machine, 1, 1, 1, 1);
 
-   shader->func( machine->Inputs,
-		 machine->Outputs,
+   shader->func( machine,
 		 machine->Consts,
-		 machine->Temps,
-		 machine->InterpCoefs,
-                 shader->immediates
+                 (const float (*)[4])shader->immediates,
+		 machine->InterpCoefs
 		 //	 , &machine->QuadPos
       );
 
@@ -151,7 +139,7 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe,
       return NULL;
    }
 
-   shader->func = (codegen_function) x86_get_func( &shader->sse2_program );
+   shader->func = (tgsi_sse2_fs_function) x86_get_func( &shader->sse2_program );
    if (!shader->func) {
       x86_release_func( &shader->sse2_program );
       FREE(shader);
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
index 06725fd09b..42021789ea 100644
--- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -238,65 +238,117 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr)
 
    case PIPE_PRIM_TRIANGLES:
       for (i = 2; i < nr; i += 3) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_STRIP:
       for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i+(i&1)-2], stride),
-                    get_vert(vertex_buffer, indices[i-(i&1)-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i+(i&1)-1], stride),
+                       get_vert(vertex_buffer, indices[i-(i&1)], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i+(i&1)-2], stride),
+                       get_vert(vertex_buffer, indices[i-(i&1)-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_FAN:
       for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[0], stride),
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUADS:
       for (i = 3; i < nr; i += 4) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-3], stride),
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 3; i < nr; i += 2) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-3], stride),
-                    get_vert(vertex_buffer, indices[i-2], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[i-1], stride),
-                    get_vert(vertex_buffer, indices[i-3], stride),
-                    get_vert(vertex_buffer, indices[i-0], stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride));
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-2], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, indices[i-1], stride),
+                       get_vert(vertex_buffer, indices[i-3], stride),
+                       get_vert(vertex_buffer, indices[i-0], stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_POLYGON:
+      /* Almost same as tri fan but the _first_ vertex specifies the flat
+       * shading color.  Note that the first polygon vertex is passed as
+       * the last triangle vertex here.
+       * flatshade_first state makes no difference.
+       */
       for (i = 2; i < nr; i += 1) {
          setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, indices[0-1], stride),
                     get_vert(vertex_buffer, indices[i-0], stride),
-                    get_vert(vertex_buffer, indices[0], stride));
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[0], stride) );
       }
       break;
 
@@ -368,58 +420,104 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
       }
       break;
 
-
    case PIPE_PRIM_TRIANGLES:
       for (i = 2; i < nr; i += 3) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-2, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_STRIP:
-      for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i+(i&1)-2, stride),
-                    get_vert(vertex_buffer, i-(i&1)-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+      for (i = 2; i < nr; i++) {
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i+(i&1)-1, stride),
+                       get_vert(vertex_buffer, i-(i&1), stride),
+                       get_vert(vertex_buffer, i-2, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i+(i&1)-2, stride),
+                       get_vert(vertex_buffer, i-(i&1)-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_TRIANGLE_FAN:
       for (i = 2; i < nr; i += 1) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, 0, stride),
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, 0, stride),
+                       get_vert(vertex_buffer, i-1, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, 0, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
    case PIPE_PRIM_QUADS:
       for (i = 3; i < nr; i += 4) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-3, stride),
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-0, stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
+
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 3; i < nr; i += 2) {
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-3, stride),
-                    get_vert(vertex_buffer, i-2, stride),
-                    get_vert(vertex_buffer, i-0, stride));
-
-         setup_tri( setup_ctx,
-                    get_vert(vertex_buffer, i-1, stride),
-                    get_vert(vertex_buffer, i-3, stride),
-                    get_vert(vertex_buffer, i-0, stride));
+         if (softpipe->rasterizer->flatshade_first) {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride),
+                       get_vert(vertex_buffer, i-3, stride) );
+         }
+         else {
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-2, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+            setup_tri( setup_ctx,
+                       get_vert(vertex_buffer, i-1, stride),
+                       get_vert(vertex_buffer, i-3, stride),
+                       get_vert(vertex_buffer, i-0, stride) );
+         }
       }
       break;
 
@@ -427,12 +525,13 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
       /* Almost same as tri fan but the _first_ vertex specifies the flat
        * shading color.  Note that the first polygon vertex is passed as
        * the last triangle vertex here.
+       * flatshade_first state makes no difference.
        */
       for (i = 2; i < nr; i += 1) {
          setup_tri( setup_ctx,
                     get_vert(vertex_buffer, i-1, stride),
                     get_vert(vertex_buffer, i-0, stride),
-                    get_vert(vertex_buffer, 0, stride));
+                    get_vert(vertex_buffer, 0, stride) );
       }
       break;
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index ca637a1d6a..28f8d1a60e 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -52,7 +52,7 @@
 struct quad_shade_stage
 {
    struct quad_stage stage;  /**< base class */
-   struct tgsi_exec_machine machine;
+   struct tgsi_exec_machine *machine;
    struct tgsi_exec_vector *inputs, *outputs;
 };
 
@@ -73,7 +73,7 @@ shade_quad(struct quad_stage *qs, struct quad_header *quad)
 {
    struct quad_shade_stage *qss = quad_shade_stage( qs );
    struct softpipe_context *softpipe = qs->softpipe;
-   struct tgsi_exec_machine *machine = &qss->machine;
+   struct tgsi_exec_machine *machine = qss->machine;
    boolean z_written;
    
    /* Consts do not require 16 byte alignment. */
@@ -146,7 +146,7 @@ shade_begin(struct quad_stage *qs)
    struct softpipe_context *softpipe = qs->softpipe;
 
    softpipe->fs->prepare( softpipe->fs, 
-			  &qss->machine,
+			  qss->machine,
 			  (struct tgsi_sampler **)
                              softpipe->tgsi.frag_samplers_list );
 
@@ -159,9 +159,8 @@ shade_destroy(struct quad_stage *qs)
 {
    struct quad_shade_stage *qss = (struct quad_shade_stage *) qs;
 
-   tgsi_exec_machine_free_data(&qss->machine);
-   FREE( qss->inputs );
-   FREE( qss->outputs );
+   tgsi_exec_machine_destroy(qss->machine);
+
    FREE( qs );
 }
 
@@ -170,19 +169,24 @@ struct quad_stage *
 sp_quad_shade_stage( struct softpipe_context *softpipe )
 {
    struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage);
-
-   /* allocate storage for program inputs/outputs, aligned to 16 bytes */
-   qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16);
-   qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16);
-   qss->machine.Inputs = align16(qss->inputs);
-   qss->machine.Outputs = align16(qss->outputs);
+   if (!qss)
+      goto fail;
 
    qss->stage.softpipe = softpipe;
    qss->stage.begin = shade_begin;
    qss->stage.run = shade_quad;
    qss->stage.destroy = shade_destroy;
 
-   tgsi_exec_machine_init( &qss->machine );
+   qss->machine = tgsi_exec_machine_create();
+   if (!qss->machine)
+      goto fail;
 
    return &qss->stage;
+
+fail:
+   if (qss && qss->machine)
+      tgsi_exec_machine_destroy(qss->machine);
+
+   FREE(qss);
+   return NULL;
 }
diff --git a/src/gallium/drivers/softpipe/sp_quad_pipe.c b/src/gallium/drivers/softpipe/sp_quad_pipe.c
index 892ef87ee9..b5f69b7426 100644
--- a/src/gallium/drivers/softpipe/sp_quad_pipe.c
+++ b/src/gallium/drivers/softpipe/sp_quad_pipe.c
@@ -80,7 +80,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp)
          sp_push_quad_first( sp, sp->quad[i].blend, i );
       }
 
-      if (sp->depth_stencil->depth.occlusion_count) {
+      if (sp->active_query_count) {
          sp_push_quad_first( sp, sp->quad[i].occlusion, i );
       }
 
diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c
index 93dab236d6..379cf4ad06 100644
--- a/src/gallium/drivers/softpipe/sp_query.c
+++ b/src/gallium/drivers/softpipe/sp_query.c
@@ -34,6 +34,7 @@
 #include "util/u_memory.h"
 #include "sp_context.h"
 #include "sp_query.h"
+#include "sp_state.h"
 
 struct softpipe_query {
    uint64_t start;
@@ -69,6 +70,8 @@ softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
    struct softpipe_query *sq = softpipe_query(q);
    
    sq->start = softpipe->occlusion_count;
+   softpipe->active_query_count++;
+   softpipe->dirty |= SP_NEW_QUERY;
 }
 
 
@@ -78,7 +81,9 @@ softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
    struct softpipe_context *softpipe = softpipe_context( pipe );
    struct softpipe_query *sq = softpipe_query(q);
 
+   softpipe->active_query_count--;
    sq->end = softpipe->occlusion_count;
+   softpipe->dirty |= SP_NEW_QUERY;
 }
 
 
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index cc39d33ede..6178c4ac7e 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -82,13 +82,15 @@ softpipe_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
       return 1;
    case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-      return 12; /* max 2Kx2K */
+      return 13; /* max 4Kx4K */
    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
       return 8;  /* max 128x128x128 */
    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-      return 12; /* max 2Kx2K */
+      return 13; /* max 4Kx4K */
    case PIPE_CAP_TGSI_CONT_SUPPORTED:
       return 1;
+   case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+      return 1;
    default:
       return 0;
    }
@@ -108,7 +110,7 @@ softpipe_get_paramf(struct pipe_screen *screen, int param)
    case PIPE_CAP_MAX_POINT_WIDTH_AA:
       return 255.0; /* arbitrary */
    case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
-      return 0.0;
+      return 16.0; /* not actually signficant at this time */
    case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
       return 16.0; /* arbitrary */
    default:
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
index 1eb23cd6b6..de3ae3c369 100644
--- a/src/gallium/drivers/softpipe/sp_setup.c
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -507,6 +507,8 @@ static void print_vertex(const struct setup_context *setup,
 #endif
 
 /**
+ * Sort the vertices from top to bottom order, setting up the triangle
+ * edge fields (ebot, emaj, etop).
  * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise
  */
 static boolean setup_sort_vertices( struct setup_context *setup,
@@ -1051,7 +1053,10 @@ setup_line_coefficients(struct setup_context *setup,
    float area;
 
    /* use setup->vmin, vmax to point to vertices */
-   setup->vprovoke = v1;
+   if (softpipe->rasterizer->flatshade_first)
+      setup->vprovoke = v0;
+   else
+      setup->vprovoke = v1;
    setup->vmin = v0;
    setup->vmax = v1;
 
diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
index 5de358dae9..a1d3bade27 100644
--- a/src/gallium/drivers/softpipe/sp_tex_sample.c
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
@@ -705,15 +705,18 @@ get_texel(const struct tgsi_sampler *tgsi_sampler,
  * Compare texcoord 'p' (aka R) against texture value 'rgba[0]'
  * When we sampled the depth texture, the depth value was put into all
  * RGBA channels.  We look at the red channel here.
+ * \param rgba  quad of (depth) texel values
+ * \param p  texture 'P' components for four pixels in quad
+ * \param j  which pixel in the quad to test [0..3]
  */
 static INLINE void
-shadow_compare(uint compare_func,
+shadow_compare(const struct pipe_sampler_state *sampler,
                float rgba[NUM_CHANNELS][QUAD_SIZE],
                const float p[QUAD_SIZE],
                uint j)
 {
    int k;
-   switch (compare_func) {
+   switch (sampler->compare_func) {
    case PIPE_FUNC_LESS:
       k = p[j] < rgba[0][j];
       break;
@@ -751,6 +754,78 @@ shadow_compare(uint compare_func,
 
 
 /**
+ * As above, but do four z/texture comparisons.
+ */
+static INLINE void
+shadow_compare4(const struct pipe_sampler_state *sampler,
+                float rgba[NUM_CHANNELS][QUAD_SIZE],
+                const float p[QUAD_SIZE])
+{
+   int j, k0, k1, k2, k3;
+   float val;
+
+   /* compare four texcoords vs. four texture samples */
+   switch (sampler->compare_func) {
+   case PIPE_FUNC_LESS:
+      k0 = p[0] < rgba[0][0];
+      k1 = p[1] < rgba[0][1];
+      k2 = p[2] < rgba[0][2];
+      k3 = p[3] < rgba[0][3];
+      break;
+   case PIPE_FUNC_LEQUAL:
+      k0 = p[0] <= rgba[0][0];
+      k1 = p[1] <= rgba[0][1];
+      k2 = p[2] <= rgba[0][2];
+      k3 = p[3] <= rgba[0][3];
+      break;
+   case PIPE_FUNC_GREATER:
+      k0 = p[0] > rgba[0][0];
+      k1 = p[1] > rgba[0][1];
+      k2 = p[2] > rgba[0][2];
+      k3 = p[3] > rgba[0][3];
+      break;
+   case PIPE_FUNC_GEQUAL:
+      k0 = p[0] >= rgba[0][0];
+      k1 = p[1] >= rgba[0][1];
+      k2 = p[2] >= rgba[0][2];
+      k3 = p[3] >= rgba[0][3];
+      break;
+   case PIPE_FUNC_EQUAL:
+      k0 = p[0] == rgba[0][0];
+      k1 = p[1] == rgba[0][1];
+      k2 = p[2] == rgba[0][2];
+      k3 = p[3] == rgba[0][3];
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      k0 = p[0] != rgba[0][0];
+      k1 = p[1] != rgba[0][1];
+      k2 = p[2] != rgba[0][2];
+      k3 = p[3] != rgba[0][3];
+      break;
+   case PIPE_FUNC_ALWAYS:
+      k0 = k1 = k2 = k3 = 1;
+      break;
+   case PIPE_FUNC_NEVER:
+      k0 = k1 = k2 = k3 = 0;
+      break;
+   default:
+      k0 = k1 = k2 = k3 = 0;
+      assert(0);
+      break;
+   }
+
+   /* convert four pass/fail values to an intensity in [0,1] */
+   val = 0.25F * (k0 + k1 + k2 + k3);
+
+   /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */
+   for (j = 0; j < 4; j++) {
+      rgba[0][j] = rgba[1][j] = rgba[2][j] = val;
+      rgba[3][j] = 1.0F;
+   }
+}
+
+
+/**
  * Common code for sampling 1D/2D/cube textures.
  * Could probably extend for 3D...
  */
@@ -769,7 +844,6 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
    const uint unit = samp->unit;
    const struct pipe_texture *texture = sp->texture[unit];
    const struct pipe_sampler_state *sampler = sp->sampler[unit];
-   const uint compare_func = sampler->compare_func;
    unsigned level0, level1, j, imgFilter;
    int width, height;
    float levelBlend;
@@ -794,7 +868,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
          for (j = 0; j < QUAD_SIZE; j++) {
             get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, rgba, p, j);
+               shadow_compare(sampler, rgba, p, j);
             }
 
             if (level0 != level1) {
@@ -806,7 +880,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
                get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0,
                          rgba2, j);
                if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
-                  shadow_compare(compare_func, rgba2, p, j);
+                  shadow_compare(sampler, rgba2, p, j);
                }
 
                for (c = 0; c < NUM_CHANNELS; c++) {
@@ -833,10 +907,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
             get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2);
             get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, tx, p, 0);
-               shadow_compare(compare_func, tx, p, 1);
-               shadow_compare(compare_func, tx, p, 2);
-               shadow_compare(compare_func, tx, p, 3);
+               shadow_compare4(sampler, tx, p);
             }
 
             /* interpolate R, G, B, A */
@@ -858,10 +929,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
                get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2);
                get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3);
                if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
-                  shadow_compare(compare_func, tx, p, 0);
-                  shadow_compare(compare_func, tx, p, 1);
-                  shadow_compare(compare_func, tx, p, 2);
-                  shadow_compare(compare_func, tx, p, 3);
+                  shadow_compare4(sampler, tx, p);
                }
 
                /* interpolate R, G, B, A */
@@ -1076,7 +1144,6 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
    const struct pipe_texture *texture = sp->texture[unit];
    const struct pipe_sampler_state *sampler = sp->sampler[unit];
    const uint face = 0;
-   const uint compare_func = sampler->compare_func;
    unsigned level0, level1, j, imgFilter;
    int width, height;
    float levelBlend;
@@ -1101,7 +1168,7 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
          for (j = 0; j < QUAD_SIZE; j++) {
             get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, rgba, p, j);
+               shadow_compare(sampler, rgba, p, j);
             }
          }
       }
@@ -1121,10 +1188,7 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
             get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2);
             get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3);
             if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
-               shadow_compare(compare_func, tx, p, 0);
-               shadow_compare(compare_func, tx, p, 1);
-               shadow_compare(compare_func, tx, p, 2);
-               shadow_compare(compare_func, tx, p, 3);
+               shadow_compare4(sampler, tx, p);
             }
             for (c = 0; c < 4; c++) {
                rgba[c][j] = lerp_2d(xw[j], yw[j],
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 9e19745889..7a533dad9f 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -42,6 +42,7 @@
 #include "sp_texture.h"
 #include "sp_tile_cache.h"
 #include "sp_screen.h"
+#include "sp_winsys.h"
 
 
 /* Simple, maximally packed layout.
diff --git a/src/gallium/drivers/trace/Makefile b/src/gallium/drivers/trace/Makefile
index e087db169a..dd6831c70a 100644
--- a/src/gallium/drivers/trace/Makefile
+++ b/src/gallium/drivers/trace/Makefile
@@ -7,8 +7,11 @@ C_SOURCES = \
 	tr_buffer.c \
 	tr_context.c \
 	tr_dump.c \
+	tr_dump_state.c \
 	tr_screen.c \
 	tr_state.c \
+	tr_rbug.c \
+	tr_drm.c \
 	tr_texture.c
 
 include ../../Makefile.template
diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README
index 73dce20372..1000c31e49 100644
--- a/src/gallium/drivers/trace/README
+++ b/src/gallium/drivers/trace/README
@@ -3,7 +3,8 @@
 
 = About =
 
-This directory contains a Gallium3D pipe driver which traces all incoming calls.
+This directory contains a Gallium3D debugger pipe driver.
+It can traces all incoming calls and/or provide remote debugging functionality.
 
 
 = Build Instructions =
@@ -23,7 +24,9 @@ ensure the right libGL.so is being picked by doing
 
  ldd progs/trivial/tri 
 
-and then try running
+== Traceing ==
+
+For traceing then do
 
  export XMESA_TRACE=y
  GALLIUM_TRACE=tri.trace progs/trivial/tri
@@ -32,6 +35,16 @@ which should create a tri.trace file, which is an XML file. You can view copying
 trace.xsl to the same directory, and opening with a XSLT capable browser such as 
 Firefox or Internet Explorer.
 
+== Remote debugging ==
+
+For remote debugging
+
+ export XMESA_TRACE=y
+ GALLIUM_RBUG=true progs/trivial/tri
+
+which should open gallium remote debugging session. While the program is running
+you can launch the small remote debugging application from progs/rbug. More
+information is in that directory.
 
 = Integrating =
 
@@ -62,3 +75,4 @@ trace_screen with real_screen when creating them.
 
 --
 Jose Fonseca <jrfonseca@tungstengraphics.com>
+Jakob Bornecrantz <jakob@vmware.com>
diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript
index 4215215d1a..e635fed77d 100644
--- a/src/gallium/drivers/trace/SConscript
+++ b/src/gallium/drivers/trace/SConscript
@@ -8,8 +8,10 @@ trace = env.ConvenienceLibrary(
         'tr_buffer.c',
         'tr_context.c',
         'tr_dump.c',
+        'tr_dump_state.c',
         'tr_screen.c',
         'tr_state.c',
+        'tr_rbug.c',
         'tr_texture.c',
     ])
 
diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c
index 47280459a7..4ab718f233 100644
--- a/src/gallium/drivers/trace/tr_context.c
+++ b/src/gallium/drivers/trace/tr_context.c
@@ -31,11 +31,11 @@
 #include "pipe/p_screen.h"
 
 #include "tr_dump.h"
+#include "tr_dump_state.h"
 #include "tr_state.h"
 #include "tr_buffer.h"
 #include "tr_screen.h"
 #include "tr_texture.h"
-#include "tr_context.h"
 
 
 static INLINE struct pipe_buffer *
@@ -113,6 +113,65 @@ trace_context_set_edgeflags(struct pipe_context *_pipe,
 }
 
 
+static INLINE void
+trace_context_draw_block(struct trace_context *tr_ctx, int flag)
+{
+   int k;
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+
+   if (tr_ctx->draw_blocker & flag) {
+      tr_ctx->draw_blocked |= flag;
+   } else if ((tr_ctx->draw_rule.blocker & flag) &&
+              (tr_ctx->draw_blocker & 4)) {
+      boolean block = FALSE;
+      debug_printf("%s (%lu %lu) (%lu %lu) (%lu %u) (%lu %u)\n", __FUNCTION__,
+					tr_ctx->draw_rule.fs, tr_ctx->curr.fs,
+					tr_ctx->draw_rule.vs, tr_ctx->curr.vs,
+					tr_ctx->draw_rule.surf, 0,
+					tr_ctx->draw_rule.tex, 0);
+      if (tr_ctx->draw_rule.fs &&
+          tr_ctx->draw_rule.fs == tr_ctx->curr.fs)
+         block = TRUE;
+      if (tr_ctx->draw_rule.vs &&
+          tr_ctx->draw_rule.vs == tr_ctx->curr.vs)
+         block = TRUE;
+      if (tr_ctx->draw_rule.surf &&
+          tr_ctx->draw_rule.surf == tr_ctx->curr.zsbuf)
+            block = TRUE;
+      if (tr_ctx->draw_rule.surf)
+         for (k = 0; k < tr_ctx->curr.nr_cbufs; k++)
+            if (tr_ctx->draw_rule.surf == tr_ctx->curr.cbufs[k])
+               block = TRUE;
+      if (tr_ctx->draw_rule.tex)
+         for (k = 0; k < tr_ctx->curr.num_texs; k++)
+            if (tr_ctx->draw_rule.tex == tr_ctx->curr.tex[k])
+               block = TRUE;
+
+      if (block)
+         tr_ctx->draw_blocked |= (flag | 4);
+   }
+
+   if (tr_ctx->draw_blocked)
+      trace_rbug_notify_draw_blocked(tr_ctx);
+
+   /* wait for rbug to clear the blocked flag */
+   while (tr_ctx->draw_blocked & flag) {
+      tr_ctx->draw_blocked |= flag;
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+      pipe_condvar_wait(tr_ctx->draw_cond, tr_ctx->draw_mutex);
+#else
+      pipe_mutex_unlock(tr_ctx->draw_mutex);
+#ifdef PIPE_SUBSYSTEM_WINDOWS_USER
+      Sleep(1);
+#endif
+      pipe_mutex_lock(tr_ctx->draw_mutex);
+#endif
+   }
+
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+}
+
 static INLINE boolean
 trace_context_draw_arrays(struct pipe_context *_pipe,
                           unsigned mode, unsigned start, unsigned count)
@@ -121,6 +180,11 @@ trace_context_draw_arrays(struct pipe_context *_pipe,
    struct pipe_context *pipe = tr_ctx->pipe;
    boolean result;
 
+   if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled)
+      return 0;
+
+   trace_context_draw_block(tr_ctx, 1);
+
    trace_dump_call_begin("pipe_context", "draw_arrays");
 
    trace_dump_arg(ptr, pipe);
@@ -134,6 +198,8 @@ trace_context_draw_arrays(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   trace_context_draw_block(tr_ctx, 2);
+
    return result;
 }
 
@@ -150,6 +216,11 @@ trace_context_draw_elements(struct pipe_context *_pipe,
    struct pipe_buffer *indexBuffer = tr_buf->buffer;
    boolean result;
 
+   if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled)
+      return 0;
+
+   trace_context_draw_block(tr_ctx, 1);
+
    trace_screen_user_buffer_update(_pipe->screen, indexBuffer);
 
    trace_dump_call_begin("pipe_context", "draw_elements");
@@ -167,6 +238,8 @@ trace_context_draw_elements(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   trace_context_draw_block(tr_ctx, 2);
+
    return result;
 }
 
@@ -187,6 +260,11 @@ trace_context_draw_range_elements(struct pipe_context *_pipe,
    struct pipe_buffer *indexBuffer = tr_buf->buffer;
    boolean result;
 
+   if (tr_ctx->curr.fs->disabled || tr_ctx->curr.vs->disabled)
+      return 0;
+
+   trace_context_draw_block(tr_ctx, 1);
+
    trace_screen_user_buffer_update(_pipe->screen, indexBuffer);
 
    trace_dump_call_begin("pipe_context", "draw_range_elements");
@@ -209,6 +287,8 @@ trace_context_draw_range_elements(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   trace_context_draw_block(tr_ctx, 2);
+
    return result;
 }
 
@@ -573,23 +653,32 @@ trace_context_create_fs_state(struct pipe_context *_pipe,
 
    trace_dump_call_end();
 
+   result = trace_shader_create(tr_ctx, state, result, TRACE_SHADER_FRAGMENT);
+
    return result;
 }
 
 
 static INLINE void
 trace_context_bind_fs_state(struct pipe_context *_pipe,
-                            void *state)
+                            void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr ? tr_shdr->state : NULL;
 
    trace_dump_call_begin("pipe_context", "bind_fs_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
 
-   pipe->bind_fs_state(pipe, state);;
+   tr_ctx->curr.fs = tr_shdr;
+
+   if (tr_shdr && tr_shdr->replaced)
+      state = tr_shdr->replaced;
+
+   pipe->bind_fs_state(pipe, state);
 
    trace_dump_call_end();
 }
@@ -597,19 +686,23 @@ trace_context_bind_fs_state(struct pipe_context *_pipe,
 
 static INLINE void
 trace_context_delete_fs_state(struct pipe_context *_pipe,
-                              void *state)
+                              void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr->state;
 
    trace_dump_call_begin("pipe_context", "delete_fs_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
 
-   pipe->delete_fs_state(pipe, state);;
+   pipe->delete_fs_state(pipe, state);
 
    trace_dump_call_end();
+
+   trace_shader_destroy(tr_ctx, tr_shdr);
 }
 
 
@@ -626,28 +719,37 @@ trace_context_create_vs_state(struct pipe_context *_pipe,
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(shader_state, state);
 
-   result = pipe->create_vs_state(pipe, state);;
+   result = pipe->create_vs_state(pipe, state);
 
    trace_dump_ret(ptr, result);
 
    trace_dump_call_end();
 
+   result = trace_shader_create(tr_ctx, state, result, TRACE_SHADER_VERTEX);
+
    return result;
 }
 
 
 static INLINE void
 trace_context_bind_vs_state(struct pipe_context *_pipe,
-                            void *state)
+                            void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr ? tr_shdr->state : NULL;
 
    trace_dump_call_begin("pipe_context", "bind_vs_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
 
+   tr_ctx->curr.vs = tr_shdr;
+
+   if (tr_shdr && tr_shdr->replaced)
+      state = tr_shdr->replaced;
+
    pipe->bind_vs_state(pipe, state);;
 
    trace_dump_call_end();
@@ -656,10 +758,12 @@ trace_context_bind_vs_state(struct pipe_context *_pipe,
 
 static INLINE void
 trace_context_delete_vs_state(struct pipe_context *_pipe,
-                              void *state)
+                              void *_state)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_shader *tr_shdr = trace_shader(_state);
    struct pipe_context *pipe = tr_ctx->pipe;
+   void *state = tr_shdr->state;
 
    trace_dump_call_begin("pipe_context", "delete_vs_state");
 
@@ -669,6 +773,8 @@ trace_context_delete_vs_state(struct pipe_context *_pipe,
    pipe->delete_vs_state(pipe, state);;
 
    trace_dump_call_end();
+
+   trace_shader_destroy(tr_ctx, tr_shdr);
 }
 
 
@@ -747,6 +853,19 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe,
    struct pipe_framebuffer_state unwrapped_state;
    unsigned i;
 
+   {
+      tr_ctx->curr.nr_cbufs = state->nr_cbufs;
+      for (i = 0; i < state->nr_cbufs; i++)
+         if (state->cbufs[i])
+            tr_ctx->curr.cbufs[i] = trace_texture(state->cbufs[i]->texture);
+         else
+            tr_ctx->curr.cbufs[i] = NULL;
+      if (state->zsbuf)
+         tr_ctx->curr.zsbuf = trace_texture(state->zsbuf->texture);
+      else
+         tr_ctx->curr.zsbuf = NULL;
+   }
+
    /* Unwrap the input state */
    memcpy(&unwrapped_state, state, sizeof(unwrapped_state));
    for(i = 0; i < state->nr_cbufs; ++i)
@@ -827,12 +946,17 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe,
                                    struct pipe_texture **textures)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
+   struct trace_texture *tr_tex;
    struct pipe_context *pipe = tr_ctx->pipe;
    struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
    unsigned i;
 
-   for(i = 0; i < num_textures; ++i)
-      unwrapped_textures[i] = trace_texture_unwrap(tr_ctx, textures[i]);
+   tr_ctx->curr.num_texs = num_textures;
+   for(i = 0; i < num_textures; ++i) {
+      tr_tex = trace_texture(textures[i]);
+      tr_ctx->curr.tex[i] = tr_tex;
+      unwrapped_textures[i] = tr_tex ? tr_tex->texture : NULL;
+   }
    textures = unwrapped_textures;
 
    trace_dump_call_begin("pipe_context", "set_sampler_textures");
@@ -1078,6 +1202,12 @@ trace_is_buffer_referenced( struct pipe_context *_pipe,
    return referenced;
 }
 
+static const struct debug_named_value rbug_blocker_flags[] = {
+   {"before", 1},
+   {"after", 2},
+   {NULL, 0},
+};
+
 struct pipe_context *
 trace_context_create(struct pipe_screen *_screen,
                      struct pipe_context *pipe)
@@ -1089,7 +1219,7 @@ trace_context_create(struct pipe_screen *_screen,
    if(!pipe)
       goto error1;
 
-   if(!trace_dump_trace_enabled())
+   if(!trace_enabled())
       goto error1;
 
    tr_scr = trace_screen(_screen);
@@ -1099,6 +1229,14 @@ trace_context_create(struct pipe_screen *_screen,
    if(!tr_ctx)
       goto error1;
 
+   tr_ctx->draw_blocker = debug_get_flags_option("RBUG_BLOCK",
+                                                 rbug_blocker_flags,
+                                                 0);
+   pipe_mutex_init(tr_ctx->draw_mutex);
+   pipe_condvar_init(tr_ctx->draw_cond);
+   pipe_mutex_init(tr_ctx->list_mutex);
+   make_empty_list(&tr_ctx->shaders);
+
    tr_ctx->base.winsys = _screen->winsys;
    tr_ctx->base.screen = _screen;
    tr_ctx->base.destroy = trace_context_destroy;
diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h
index 2512a0a232..6febe4b411 100644
--- a/src/gallium/drivers/trace/tr_context.h
+++ b/src/gallium/drivers/trace/tr_context.h
@@ -33,6 +33,7 @@
 #include "util/u_debug.h"
 #include "pipe/p_context.h"
 
+#include "tr_screen.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -45,7 +46,41 @@ struct trace_context
 
    struct pipe_context *pipe;
 
+   /* current state */
+   struct {
+      struct trace_shader *fs;
+      struct trace_shader *vs;
+
+      struct trace_texture *tex[PIPE_MAX_SAMPLERS];
+      unsigned num_texs;
+
+      unsigned nr_cbufs;
+      struct trace_texture *cbufs[PIPE_MAX_COLOR_BUFS];
+      struct trace_texture *zsbuf;
+   } curr;
+
+   struct {
+      struct trace_shader *fs;
+      struct trace_shader *vs;
+
+      struct trace_texture *tex;
+      struct trace_texture *surf;
+
+      int blocker;
+   } draw_rule;
+   unsigned draw_num_rules;
+   pipe_condvar draw_cond;
+   pipe_mutex draw_mutex;
+   int draw_blocker;
+   int draw_blocked;
+
+   /* for list on screen */
    struct tr_list list;
+
+   /* list of state objects */
+   pipe_mutex list_mutex;
+   unsigned num_shaders;
+   struct tr_list shaders;
 };
 
 
@@ -62,6 +97,9 @@ struct pipe_context *
 trace_context_create(struct pipe_screen *screen,
                      struct pipe_context *pipe);
 
+void
+trace_rbug_notify_draw_blocked(struct trace_context *tr_ctx);
+
 
 #ifdef __cplusplus
 }
diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c
new file mode 100644
index 0000000000..98ac75e3fa
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_drm.c
@@ -0,0 +1,211 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "state_tracker/drm_api.h"
+
+#include "util/u_memory.h"
+#include "trace/tr_drm.h"
+#include "trace/tr_screen.h"
+#include "trace/tr_context.h"
+#include "trace/tr_buffer.h"
+#include "trace/tr_texture.h"
+
+struct trace_drm_api
+{
+   struct drm_api base;
+
+   struct drm_api *api;
+};
+
+static INLINE struct trace_drm_api *
+trace_drm_api(struct drm_api *_api)
+{
+   return (struct trace_drm_api *)_api;
+}
+
+static struct pipe_screen *
+trace_drm_create_screen(struct drm_api *_api, int fd,
+                           struct drm_create_screen_arg *arg)
+{
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct drm_api *api = tr_api->api;
+   struct pipe_screen *screen;
+
+   /* TODO trace call */
+
+   if (arg && arg->mode != DRM_CREATE_NORMAL)
+      return NULL;
+
+   screen = api->create_screen(api, fd, arg);
+
+   return trace_screen_create(screen);
+};
+
+static struct pipe_context *
+trace_drm_create_context(struct drm_api *_api,
+                            struct pipe_screen *_screen)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct drm_api *api = tr_api->api;
+   struct pipe_context *pipe;
+
+   /* TODO trace call */
+
+   pipe = api->create_context(api, screen);
+
+   pipe = trace_context_create(_screen, pipe);
+
+   return pipe;
+};
+
+static boolean
+trace_drm_buffer_from_texture(struct drm_api *_api,
+                                 struct pipe_texture *_texture,
+                                 struct pipe_buffer **_buffer,
+                                 unsigned *stride)
+{
+   struct trace_texture *tr_texture = trace_texture(_texture);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_texture *texture = tr_texture->texture;
+   struct drm_api *api = tr_api->api;
+   struct pipe_buffer *buffer = NULL;
+   boolean result;
+
+   /* TODO trace call */
+
+   result = api->buffer_from_texture(api, texture, &buffer, stride);
+
+   if (result && _buffer)
+      buffer = trace_buffer_create(trace_screen(texture->screen), buffer);
+
+   if (_buffer)
+      *_buffer = buffer;
+   else
+      pipe_buffer_reference(&buffer, NULL);
+
+   return result;
+}
+
+static struct pipe_buffer *
+trace_drm_buffer_from_handle(struct drm_api *_api,
+                                struct pipe_screen *_screen,
+                                const char *name,
+                                unsigned handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct drm_api *api = tr_api->api;
+   struct pipe_buffer *result;
+
+   /* TODO trace call */
+
+   result = api->buffer_from_handle(api, screen, name, handle);
+
+   result = trace_buffer_create(trace_screen(_screen), result);
+
+   return result;
+}
+
+static boolean
+trace_drm_handle_from_buffer(struct drm_api *_api,
+                                struct pipe_screen *_screen,
+                                struct pipe_buffer *_buffer,
+                                unsigned *handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_buffer *tr_buffer = trace_buffer(_buffer);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct pipe_buffer *buffer = tr_buffer->buffer;
+   struct drm_api *api = tr_api->api;
+
+   /* TODO trace call */
+
+   return api->handle_from_buffer(api, screen, buffer, handle);
+}
+
+static boolean
+trace_drm_global_handle_from_buffer(struct drm_api *_api,
+                                       struct pipe_screen *_screen,
+                                       struct pipe_buffer *_buffer,
+                                       unsigned *handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_buffer *tr_buffer = trace_buffer(_buffer);
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct pipe_buffer *buffer = tr_buffer->buffer;
+   struct drm_api *api = tr_api->api;
+
+   /* TODO trace call */
+
+   return api->global_handle_from_buffer(api, screen, buffer, handle);
+}
+
+static void
+trace_drm_destroy(struct drm_api *_api)
+{
+   struct trace_drm_api *tr_api = trace_drm_api(_api);
+   struct drm_api *api = tr_api->api;
+   api->destroy(api);
+
+   free(tr_api);
+}
+
+struct drm_api *
+trace_drm_create(struct drm_api *api)
+{
+   struct trace_drm_api *tr_api;
+
+   if (!api)
+      goto error;
+
+   if (!trace_enabled())
+      goto error;
+
+   tr_api = CALLOC_STRUCT(trace_drm_api);
+
+   if (!tr_api)
+      goto error;
+
+   tr_api->base.create_screen = trace_drm_create_screen;
+   tr_api->base.create_context = trace_drm_create_context;
+   tr_api->base.buffer_from_texture = trace_drm_buffer_from_texture;
+   tr_api->base.buffer_from_handle = trace_drm_buffer_from_handle;
+   tr_api->base.handle_from_buffer = trace_drm_handle_from_buffer;
+   tr_api->base.global_handle_from_buffer = trace_drm_global_handle_from_buffer;
+   tr_api->base.destroy = trace_drm_destroy;
+   tr_api->api = api;
+
+   return &tr_api->base;
+
+error:
+   return api;
+}
diff --git a/src/gallium/drivers/trace/tr_drm.h b/src/gallium/drivers/trace/tr_drm.h
new file mode 100644
index 0000000000..845c66a32a
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_drm.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_DRM_H
+#define TR_DRM_H
+
+struct drm_api;
+
+struct drm_api* trace_drm_create(struct drm_api *api);
+
+#endif /* ID_DRM_H */
diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c
new file mode 100644
index 0000000000..bcf6751af4
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_dump_state.c
@@ -0,0 +1,549 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "tr_dump.h"
+#include "tr_dump_state.h"
+
+
+void trace_dump_format(enum pipe_format format)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_enum(pf_name(format) );
+}
+
+
+void trace_dump_block(const struct pipe_format_block *block)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_struct_begin("pipe_format_block");
+   trace_dump_member(uint, block, size);
+   trace_dump_member(uint, block, width);
+   trace_dump_member(uint, block, height);
+   trace_dump_struct_end();
+}
+
+
+static void trace_dump_reference(const struct pipe_reference *reference)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_struct_begin("pipe_reference");
+   trace_dump_member(int, &reference->count, count);
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_template(const struct pipe_texture *templat)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!templat) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_texture");
+
+   trace_dump_member(int, templat, target);
+   trace_dump_member(format, templat, format);
+
+   trace_dump_member_begin("width");
+   trace_dump_array(uint, templat->width, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("height");
+   trace_dump_array(uint, templat->height, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("depth");
+   trace_dump_array(uint, templat->depth, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("block");
+   trace_dump_block(&templat->block);
+   trace_dump_member_end();
+
+   trace_dump_member(uint, templat, last_level);
+   trace_dump_member(uint, templat, tex_usage);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_rasterizer_state");
+
+   trace_dump_member(bool, state, flatshade);
+   trace_dump_member(bool, state, light_twoside);
+   trace_dump_member(uint, state, front_winding);
+   trace_dump_member(uint, state, cull_mode);
+   trace_dump_member(uint, state, fill_cw);
+   trace_dump_member(uint, state, fill_ccw);
+   trace_dump_member(bool, state, offset_cw);
+   trace_dump_member(bool, state, offset_ccw);
+   trace_dump_member(bool, state, scissor);
+   trace_dump_member(bool, state, poly_smooth);
+   trace_dump_member(bool, state, poly_stipple_enable);
+   trace_dump_member(bool, state, point_smooth);
+   trace_dump_member(bool, state, point_sprite);
+   trace_dump_member(bool, state, point_size_per_vertex);
+   trace_dump_member(bool, state, multisample);
+   trace_dump_member(bool, state, line_smooth);
+   trace_dump_member(bool, state, line_stipple_enable);
+   trace_dump_member(uint, state, line_stipple_factor);
+   trace_dump_member(uint, state, line_stipple_pattern);
+   trace_dump_member(bool, state, line_last_pixel);
+   trace_dump_member(bool, state, bypass_vs_clip_and_viewport);
+   trace_dump_member(bool, state, flatshade_first);
+   trace_dump_member(bool, state, gl_rasterization_rules);
+
+   trace_dump_member(float, state, line_width);
+   trace_dump_member(float, state, point_size);
+   trace_dump_member(float, state, point_size_min);
+   trace_dump_member(float, state, point_size_max);
+   trace_dump_member(float, state, offset_units);
+   trace_dump_member(float, state, offset_scale);
+
+   trace_dump_member_array(uint, state, sprite_coord_mode);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_poly_stipple(const struct pipe_poly_stipple *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_poly_stipple");
+
+   trace_dump_member_begin("stipple");
+   trace_dump_array(uint,
+                    state->stipple,
+                    Elements(state->stipple));
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_viewport_state(const struct pipe_viewport_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_viewport_state");
+
+   trace_dump_member_array(float, state, scale);
+   trace_dump_member_array(float, state, translate);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_scissor_state(const struct pipe_scissor_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_scissor_state");
+
+   trace_dump_member(uint, state, minx);
+   trace_dump_member(uint, state, miny);
+   trace_dump_member(uint, state, maxx);
+   trace_dump_member(uint, state, maxy);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_clip_state(const struct pipe_clip_state *state)
+{
+   unsigned i;
+
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_clip_state");
+
+   trace_dump_member_begin("ucp");
+   trace_dump_array_begin();
+   for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) {
+      trace_dump_elem_begin();
+      trace_dump_array(float, state->ucp[i], 4);
+      trace_dump_elem_end();
+   }
+   trace_dump_array_end();
+   trace_dump_member_end();
+
+   trace_dump_member(uint, state, nr);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_constant_buffer(const struct pipe_constant_buffer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_constant_buffer");
+
+   trace_dump_member(buffer_ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_shader_state(const struct pipe_shader_state *state)
+{
+   static char str[8192];
+
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   tgsi_dump_str(state->tokens, 0, str, sizeof(str));
+
+   trace_dump_struct_begin("pipe_shader_state");
+
+   trace_dump_member_begin("tokens");
+   trace_dump_string(str);
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state)
+{
+   unsigned i;
+
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_depth_stencil_alpha_state");
+
+   trace_dump_member_begin("depth");
+   trace_dump_struct_begin("pipe_depth_state");
+   trace_dump_member(bool, &state->depth, enabled);
+   trace_dump_member(bool, &state->depth, writemask);
+   trace_dump_member(uint, &state->depth, func);
+   trace_dump_struct_end();
+   trace_dump_member_end();
+
+   trace_dump_member_begin("stencil");
+   trace_dump_array_begin();
+   for(i = 0; i < Elements(state->stencil); ++i) {
+      trace_dump_elem_begin();
+      trace_dump_struct_begin("pipe_stencil_state");
+      trace_dump_member(bool, &state->stencil[i], enabled);
+      trace_dump_member(uint, &state->stencil[i], func);
+      trace_dump_member(uint, &state->stencil[i], fail_op);
+      trace_dump_member(uint, &state->stencil[i], zpass_op);
+      trace_dump_member(uint, &state->stencil[i], zfail_op);
+      trace_dump_member(uint, &state->stencil[i], ref_value);
+      trace_dump_member(uint, &state->stencil[i], valuemask);
+      trace_dump_member(uint, &state->stencil[i], writemask);
+      trace_dump_struct_end();
+      trace_dump_elem_end();
+   }
+   trace_dump_array_end();
+   trace_dump_member_end();
+
+   trace_dump_member_begin("alpha");
+   trace_dump_struct_begin("pipe_alpha_state");
+   trace_dump_member(bool, &state->alpha, enabled);
+   trace_dump_member(uint, &state->alpha, func);
+   trace_dump_member(float, &state->alpha, ref_value);
+   trace_dump_struct_end();
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_blend_state(const struct pipe_blend_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_blend_state");
+
+   trace_dump_member(bool, state, blend_enable);
+
+   trace_dump_member(uint, state, rgb_func);
+   trace_dump_member(uint, state, rgb_src_factor);
+   trace_dump_member(uint, state, rgb_dst_factor);
+
+   trace_dump_member(uint, state, alpha_func);
+   trace_dump_member(uint, state, alpha_src_factor);
+   trace_dump_member(uint, state, alpha_dst_factor);
+
+   trace_dump_member(bool, state, logicop_enable);
+   trace_dump_member(uint, state, logicop_func);
+
+   trace_dump_member(uint, state, colormask);
+   trace_dump_member(bool, state, dither);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_blend_color(const struct pipe_blend_color *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_blend_color");
+
+   trace_dump_member_array(float, state, color);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   trace_dump_struct_begin("pipe_framebuffer_state");
+
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+   trace_dump_member(uint, state, nr_cbufs);
+   trace_dump_member_array(ptr, state, cbufs);
+   trace_dump_member(ptr, state, zsbuf);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_sampler_state(const struct pipe_sampler_state *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_sampler_state");
+
+   trace_dump_member(uint, state, wrap_s);
+   trace_dump_member(uint, state, wrap_t);
+   trace_dump_member(uint, state, wrap_r);
+   trace_dump_member(uint, state, min_img_filter);
+   trace_dump_member(uint, state, min_mip_filter);
+   trace_dump_member(uint, state, mag_img_filter);
+   trace_dump_member(bool, state, compare_mode);
+   trace_dump_member(uint, state, compare_func);
+   trace_dump_member(bool, state, normalized_coords);
+   trace_dump_member(uint, state, prefilter);
+   trace_dump_member(float, state, lod_bias);
+   trace_dump_member(float, state, min_lod);
+   trace_dump_member(float, state, max_lod);
+   trace_dump_member_array(float, state, border_color);
+   trace_dump_member(float, state, max_anisotropy);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_surface(const struct pipe_surface *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_surface");
+
+   trace_dump_reference(&state->reference);
+
+   trace_dump_member(format, state, format);
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+
+   trace_dump_member(uint, state, layout);
+   trace_dump_member(uint, state, offset);
+   trace_dump_member(uint, state, usage);
+
+   trace_dump_member(ptr, state, texture);
+   trace_dump_member(uint, state, face);
+   trace_dump_member(uint, state, level);
+   trace_dump_member(uint, state, zslice);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_transfer(const struct pipe_transfer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_transfer");
+
+   trace_dump_member(format, state, format);
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+
+   trace_dump_member_begin("block");
+   trace_dump_block(&state->block);
+   trace_dump_member_end();
+
+   trace_dump_member(uint, state, nblocksx);
+   trace_dump_member(uint, state, nblocksy);
+   trace_dump_member(uint, state, stride);
+   trace_dump_member(uint, state, usage);
+
+   trace_dump_member(ptr, state, texture);
+   trace_dump_member(uint, state, face);
+   trace_dump_member(uint, state, level);
+   trace_dump_member(uint, state, zslice);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_vertex_buffer");
+
+   trace_dump_member(uint, state, stride);
+   trace_dump_member(uint, state, max_index);
+   trace_dump_member(uint, state, buffer_offset);
+   trace_dump_member(buffer_ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_vertex_element(const struct pipe_vertex_element *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_vertex_element");
+
+   trace_dump_member(uint, state, src_offset);
+
+   trace_dump_member(uint, state, vertex_buffer_index);
+   trace_dump_member(uint, state, nr_components);
+
+   trace_dump_member(format, state, src_format);
+
+   trace_dump_struct_end();
+}
diff --git a/src/gallium/drivers/trace/tr_dump_state.h b/src/gallium/drivers/trace/tr_dump_state.h
new file mode 100644
index 0000000000..05b821adb6
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_dump_state.h
@@ -0,0 +1,78 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_DUMP_STATE_H_
+#define TR_DUMP_STATE_H_
+
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+
+
+void trace_dump_format(enum pipe_format format);
+
+void trace_dump_block(const struct pipe_format_block *block);
+
+void trace_dump_template(const struct pipe_texture *templat);
+
+
+void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state);
+
+void trace_dump_poly_stipple(const struct pipe_poly_stipple *state);
+
+void trace_dump_viewport_state(const struct pipe_viewport_state *state);
+
+void trace_dump_scissor_state(const struct pipe_scissor_state *state);
+
+void trace_dump_clip_state(const struct pipe_clip_state *state);
+
+void trace_dump_constant_buffer(const struct pipe_constant_buffer *state);
+
+void trace_dump_token(const struct tgsi_token *token);
+
+void trace_dump_shader_state(const struct pipe_shader_state *state);
+
+void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state);
+
+void trace_dump_blend_state(const struct pipe_blend_state *state);
+
+void trace_dump_blend_color(const struct pipe_blend_color *state);
+
+void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state);
+
+void trace_dump_sampler_state(const struct pipe_sampler_state *state);
+
+void trace_dump_surface(const struct pipe_surface *state);
+
+void trace_dump_transfer(const struct pipe_transfer *state);
+
+void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state);
+
+void trace_dump_vertex_element(const struct pipe_vertex_element *state);
+
+
+#endif /* TR_STATE_H */
diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c
new file mode 100644
index 0000000000..e85ac15edc
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_rbug.c
@@ -0,0 +1,864 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "util/u_string.h"
+#include "util/u_memory.h"
+#include "util/u_simple_list.h"
+#include "util/u_network.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "tr_dump.h"
+#include "tr_state.h"
+#include "tr_buffer.h"
+#include "tr_texture.h"
+
+#include "rbug/rbug.h"
+
+#include <errno.h>
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+#  define sleep Sleep
+#elif defined(PIPE_OS_LINUX)
+void usleep(int);
+#  define sleep usleep
+#else
+#  warning "No socket implementation"
+#endif
+
+#define U642VOID(x) ((void *)(unsigned long)(x))
+#define VOID2U64(x) ((uint64_t)(unsigned long)(x))
+
+struct trace_rbug
+{
+   struct trace_screen *tr_scr;
+   struct rbug_connection *con;
+   pipe_thread thread;
+   boolean running;
+};
+
+PIPE_THREAD_ROUTINE(trace_rbug_thread, void_tr_rbug);
+
+
+/**********************************************************
+ * Helper functions
+ */
+
+
+static struct trace_context *
+trace_rbug_get_context_locked(struct trace_screen *tr_scr, rbug_context_t ctx)
+{
+   struct trace_context *tr_ctx = NULL;
+   struct tr_list *ptr;
+
+   foreach(ptr, &tr_scr->contexts) {
+      tr_ctx = (struct trace_context *)((char*)ptr - offsetof(struct trace_context, list));
+      if (ctx == VOID2U64(tr_ctx))
+         break;
+      tr_ctx = NULL;
+   }
+
+   return tr_ctx;
+}
+
+static struct trace_shader *
+trace_rbug_get_shader_locked(struct trace_context *tr_ctx, rbug_shader_t shdr)
+{
+   struct trace_shader *tr_shdr = NULL;
+   struct tr_list *ptr;
+
+   foreach(ptr, &tr_ctx->shaders) {
+      tr_shdr = (struct trace_shader *)((char*)ptr - offsetof(struct trace_shader, list));
+      if (shdr == VOID2U64(tr_shdr))
+         break;
+      tr_shdr = NULL;
+   }
+
+   return tr_shdr;
+}
+
+static void *
+trace_shader_create_locked(struct pipe_context *pipe,
+                           struct trace_shader *tr_shdr,
+                           struct tgsi_token *tokens)
+{
+   void *state = NULL;
+   struct pipe_shader_state pss = { 0 };
+   pss.tokens = tokens;
+
+   if (tr_shdr->type == TRACE_SHADER_FRAGMENT) {
+      state = pipe->create_fs_state(pipe, &pss);
+   } else if (tr_shdr->type == TRACE_SHADER_VERTEX) {
+      state = pipe->create_vs_state(pipe, &pss);
+   } else
+      assert(0);
+
+   return state;
+}
+
+static void
+trace_shader_bind_locked(struct pipe_context *pipe,
+                         struct trace_shader *tr_shdr,
+                         void *state)
+{
+   if (tr_shdr->type == TRACE_SHADER_FRAGMENT) {
+      pipe->bind_fs_state(pipe, state);
+   } else if (tr_shdr->type == TRACE_SHADER_VERTEX) {
+      pipe->bind_vs_state(pipe, state);
+   } else
+      assert(0);
+}
+
+static void
+trace_shader_delete_locked(struct pipe_context *pipe,
+                           struct trace_shader *tr_shdr,
+                           void *state)
+{
+   if (tr_shdr->type == TRACE_SHADER_FRAGMENT) {
+      pipe->delete_fs_state(pipe, state);
+   } else if (tr_shdr->type == TRACE_SHADER_VERTEX) {
+      pipe->delete_vs_state(pipe, state);
+   } else
+      assert(0);
+}
+
+/************************************************
+ * Request handler functions
+ */
+
+
+static int
+trace_rbug_texture_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_texture *tr_tex = NULL;
+   struct tr_list *ptr;
+   rbug_texture_t *texs;
+   int i = 0;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   texs = MALLOC(tr_scr->num_textures * sizeof(rbug_texture_t));
+   foreach(ptr, &tr_scr->textures) {
+      tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list));
+      texs[i++] = VOID2U64(tr_tex);
+   }
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   rbug_send_texture_list_reply(tr_rbug->con, serial, texs, i, NULL);
+   FREE(texs);
+
+   return 0;
+}
+
+static int
+trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_texture *tr_tex;
+   struct rbug_proto_texture_info *gpti = (struct rbug_proto_texture_info *)header;
+   struct tr_list *ptr;
+   struct pipe_texture *t;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   foreach(ptr, &tr_scr->textures) {
+      tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list));
+      if (gpti->texture == VOID2U64(tr_tex))
+         break;
+      tr_tex = NULL;
+   }
+
+   if (!tr_tex) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   t = tr_tex->texture;
+   rbug_send_texture_info_reply(tr_rbug->con, serial,
+                               t->target, t->format,
+                               t->width, t->last_level + 1,
+                               t->height, t->last_level + 1,
+                               t->depth, t->last_level + 1,
+                               t->block.width, t->block.height, t->block.size,
+                               t->last_level,
+                               t->nr_samples,
+                               t->tex_usage,
+                               NULL);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_texture_read(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_texture_read *gptr = (struct rbug_proto_texture_read *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_texture *tr_tex;
+   struct tr_list *ptr;
+
+   struct pipe_screen *screen = tr_scr->screen;
+   struct pipe_texture *tex;
+   struct pipe_transfer *t;
+
+   void *map;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   foreach(ptr, &tr_scr->textures) {
+      tr_tex = (struct trace_texture *)((char*)ptr - offsetof(struct trace_texture, list));
+      if (gptr->texture == VOID2U64(tr_tex))
+         break;
+      tr_tex = NULL;
+   }
+
+   if (!tr_tex) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   tex = tr_tex->texture;
+   t = screen->get_tex_transfer(tr_scr->screen, tex,
+                                gptr->face, gptr->level, gptr->zslice,
+                                PIPE_TRANSFER_READ,
+                                gptr->x, gptr->y, gptr->w, gptr->h);
+
+   map = screen->transfer_map(screen, t);
+
+   rbug_send_texture_read_reply(tr_rbug->con, serial,
+                                t->format,
+                                t->block.width, t->block.height, t->block.size,
+                                (uint8_t*)map, t->stride * t->nblocksy,
+                                t->stride,
+                                NULL);
+
+   screen->transfer_unmap(screen, t);
+   screen->tex_transfer_destroy(t);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct tr_list *ptr;
+   struct trace_context *tr_ctx = NULL;
+   rbug_context_t *ctxs;
+   int i = 0;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   ctxs = MALLOC(tr_scr->num_contexts * sizeof(rbug_context_t));
+   foreach(ptr, &tr_scr->contexts) {
+      tr_ctx = (struct trace_context *)((char*)ptr - offsetof(struct trace_context, list));
+      ctxs[i++] = VOID2U64(tr_ctx);
+   }
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   rbug_send_context_list_reply(tr_rbug->con, serial, ctxs, i, NULL);
+   FREE(ctxs);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_info *info = (struct rbug_proto_context_info *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   rbug_texture_t cbufs[PIPE_MAX_COLOR_BUFS];
+   rbug_texture_t texs[PIPE_MAX_SAMPLERS];
+   int i;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, info->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* protect the pipe context */
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   trace_dump_call_lock();
+
+   for (i = 0; i < tr_ctx->curr.nr_cbufs; i++)
+      cbufs[i] = VOID2U64(tr_ctx->curr.cbufs[i]);
+
+   for (i = 0; i < tr_ctx->curr.num_texs; i++)
+      texs[i] = VOID2U64(tr_ctx->curr.tex[i]);
+
+   rbug_send_context_info_reply(tr_rbug->con, serial,
+                                VOID2U64(tr_ctx->curr.vs), VOID2U64(tr_ctx->curr.fs),
+                                texs, tr_ctx->curr.num_texs,
+                                cbufs, tr_ctx->curr.nr_cbufs,
+                                VOID2U64(tr_ctx->curr.zsbuf),
+                                tr_ctx->draw_blocker, tr_ctx->draw_blocked, NULL);
+
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_block(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_block *block = (struct rbug_proto_context_draw_block *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, block->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   tr_ctx->draw_blocker |= block->block;
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_step(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_step *step = (struct rbug_proto_context_draw_step *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, step->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   if (tr_ctx->draw_blocked & RBUG_BLOCK_RULE) {
+      if (step->step & RBUG_BLOCK_RULE)
+         tr_ctx->draw_blocked &= ~RBUG_BLOCK_MASK;
+   } else {
+      tr_ctx->draw_blocked &= ~step->step;
+   }
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+   pipe_condvar_broadcast(tr_ctx->draw_cond);
+#endif
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_unblock(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_unblock *unblock = (struct rbug_proto_context_draw_unblock *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, unblock->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   if (tr_ctx->draw_blocked & RBUG_BLOCK_RULE) {
+      if (unblock->unblock & RBUG_BLOCK_RULE)
+         tr_ctx->draw_blocked &= ~RBUG_BLOCK_MASK;
+   } else {
+      tr_ctx->draw_blocked &= ~unblock->unblock;
+   }
+   tr_ctx->draw_blocker &= ~unblock->unblock;
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+   pipe_condvar_broadcast(tr_ctx->draw_cond);
+#endif
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_draw_rule(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_draw_rule *rule = (struct rbug_proto_context_draw_rule *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, rule->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->draw_mutex);
+   tr_ctx->draw_rule.vs = U642VOID(rule->vertex);
+   tr_ctx->draw_rule.fs = U642VOID(rule->fragment);
+   tr_ctx->draw_rule.tex = U642VOID(rule->texture);
+   tr_ctx->draw_rule.surf = U642VOID(rule->surface);
+   tr_ctx->draw_rule.blocker = rule->block;
+   tr_ctx->draw_blocker |= RBUG_BLOCK_RULE;
+   pipe_mutex_unlock(tr_ctx->draw_mutex);
+
+#ifdef PIPE_THREAD_HAVE_CONDVAR
+   pipe_condvar_broadcast(tr_ctx->draw_cond);
+#endif
+
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_context_flush(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_context_flush *flush = (struct rbug_proto_context_flush *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, flush->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* protect the pipe context */
+   trace_dump_call_lock();
+
+   tr_ctx->pipe->flush(tr_ctx->pipe, flush->flags, NULL);
+
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_list(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_shader_list *list = (struct rbug_proto_shader_list *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+   struct tr_list *ptr;
+   rbug_shader_t *shdrs;
+   int i = 0;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, list->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+   shdrs = MALLOC(tr_ctx->num_shaders * sizeof(rbug_shader_t));
+   foreach(ptr, &tr_ctx->shaders) {
+      tr_shdr = (struct trace_shader *)((char*)ptr - offsetof(struct trace_shader, list));
+      shdrs[i++] = VOID2U64(tr_shdr);
+   }
+
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   rbug_send_shader_list_reply(tr_rbug->con, serial, shdrs, i, NULL);
+   FREE(shdrs);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_info(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   struct rbug_proto_shader_info *info = (struct rbug_proto_shader_info *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+   unsigned original_len;
+   unsigned replaced_len;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, info->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+
+   tr_shdr = trace_rbug_get_shader_locked(tr_ctx, info->shader);
+
+   if (!tr_shdr) {
+      pipe_mutex_unlock(tr_ctx->list_mutex);
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* just in case */
+   assert(sizeof(struct tgsi_token) == 4);
+
+   original_len = tgsi_num_tokens(tr_shdr->tokens);
+   if (tr_shdr->replaced_tokens)
+      replaced_len = tgsi_num_tokens(tr_shdr->replaced_tokens);
+   else
+      replaced_len = 0;
+
+   rbug_send_shader_info_reply(tr_rbug->con, serial,
+                               (uint32_t*)tr_shdr->tokens, original_len,
+                               (uint32_t*)tr_shdr->replaced_tokens, replaced_len,
+                               tr_shdr->disabled,
+                               NULL);
+
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_disable(struct trace_rbug *tr_rbug, struct rbug_header *header)
+{
+   struct rbug_proto_shader_disable *dis = (struct rbug_proto_shader_disable *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, dis->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+
+   tr_shdr = trace_rbug_get_shader_locked(tr_ctx, dis->shader);
+
+   if (!tr_shdr) {
+      pipe_mutex_unlock(tr_ctx->list_mutex);
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   tr_shdr->disabled = dis->disable;
+
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+}
+
+static int
+trace_rbug_shader_replace(struct trace_rbug *tr_rbug, struct rbug_header *header)
+{
+   struct rbug_proto_shader_replace *rep = (struct rbug_proto_shader_replace *)header;
+
+   struct trace_screen *tr_scr = tr_rbug->tr_scr;
+   struct trace_context *tr_ctx = NULL;
+   struct trace_shader *tr_shdr = NULL;
+   struct pipe_context *pipe = NULL;
+   void *state;
+
+   pipe_mutex_lock(tr_scr->list_mutex);
+   tr_ctx = trace_rbug_get_context_locked(tr_scr, rep->context);
+
+   if (!tr_ctx) {
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   pipe_mutex_lock(tr_ctx->list_mutex);
+
+   tr_shdr = trace_rbug_get_shader_locked(tr_ctx, rep->shader);
+
+   if (!tr_shdr) {
+      pipe_mutex_unlock(tr_ctx->list_mutex);
+      pipe_mutex_unlock(tr_scr->list_mutex);
+      return -ESRCH;
+   }
+
+   /* protect the pipe context */
+   trace_dump_call_lock();
+
+   pipe = tr_ctx->pipe;
+
+   /* remove old replaced shader */
+   if (tr_shdr->replaced) {
+      if (tr_ctx->curr.fs == tr_shdr || tr_ctx->curr.vs == tr_shdr)
+         trace_shader_bind_locked(pipe, tr_shdr, tr_shdr->state);
+
+      FREE(tr_shdr->replaced_tokens);
+      trace_shader_delete_locked(pipe, tr_shdr, tr_shdr->replaced);
+      tr_shdr->replaced = NULL;
+      tr_shdr->replaced_tokens = NULL;
+   }
+
+   /* empty inputs means restore old which we did above */
+   if (rep->tokens_len == 0)
+      goto out;
+
+   tr_shdr->replaced_tokens = tgsi_dup_tokens((struct tgsi_token *)rep->tokens);
+   if (!tr_shdr->replaced_tokens)
+      goto err;
+
+   state = trace_shader_create_locked(pipe, tr_shdr, tr_shdr->replaced_tokens);
+   if (!state)
+      goto err;
+
+   /* bind new shader if the shader is currently a bound */
+   if (tr_ctx->curr.fs == tr_shdr || tr_ctx->curr.vs == tr_shdr)
+      trace_shader_bind_locked(pipe, tr_shdr, state);
+
+   /* save state */
+   tr_shdr->replaced = state;
+
+out:
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+
+   return 0;
+
+err:
+   FREE(tr_shdr->replaced_tokens);
+   tr_shdr->replaced = NULL;
+   tr_shdr->replaced_tokens = NULL;
+
+   trace_dump_call_unlock();
+   pipe_mutex_unlock(tr_ctx->list_mutex);
+   pipe_mutex_unlock(tr_scr->list_mutex);
+   return -EINVAL;
+}
+
+static boolean
+trace_rbug_header(struct trace_rbug *tr_rbug, struct rbug_header *header, uint32_t serial)
+{
+   int ret = 0;
+
+   switch(header->opcode) {
+      case RBUG_OP_PING:
+         rbug_send_ping_reply(tr_rbug->con, serial, NULL);
+         break;
+      case RBUG_OP_TEXTURE_LIST:
+         ret = trace_rbug_texture_list(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_TEXTURE_INFO:
+         ret = trace_rbug_texture_info(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_TEXTURE_READ:
+         ret = trace_rbug_texture_read(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_LIST:
+         ret = trace_rbug_context_list(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_INFO:
+         ret = trace_rbug_context_info(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_BLOCK:
+         ret = trace_rbug_context_draw_block(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_STEP:
+         ret = trace_rbug_context_draw_step(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_UNBLOCK:
+         ret = trace_rbug_context_draw_unblock(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_DRAW_RULE:
+         ret = trace_rbug_context_draw_rule(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_CONTEXT_FLUSH:
+         ret = trace_rbug_context_flush(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_SHADER_LIST:
+         ret = trace_rbug_shader_list(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_SHADER_INFO:
+         ret = trace_rbug_shader_info(tr_rbug, header, serial);
+         break;
+      case RBUG_OP_SHADER_DISABLE:
+         ret = trace_rbug_shader_disable(tr_rbug, header);
+         break;
+      case RBUG_OP_SHADER_REPLACE:
+         ret = trace_rbug_shader_replace(tr_rbug, header);
+         break;
+      default:
+         debug_printf("%s - unsupported opcode %u\n", __FUNCTION__, header->opcode);
+         ret = -ENOSYS;
+         break;
+   }
+   rbug_free_header(header);
+
+   if (ret)
+      rbug_send_error_reply(tr_rbug->con, serial, ret, NULL);
+
+   return TRUE;
+}
+
+static void
+trace_rbug_con(struct trace_rbug *tr_rbug)
+{
+   struct rbug_header *header;
+   uint32_t serial;
+
+   debug_printf("%s - connection received\n", __FUNCTION__);
+
+   while(tr_rbug->running) {
+      header = rbug_get_message(tr_rbug->con, &serial);
+      if (!header)
+         break;
+
+      if (!trace_rbug_header(tr_rbug, header, serial))
+         break;
+   }
+
+   debug_printf("%s - connection closed\n", __FUNCTION__);
+
+   rbug_disconnect(tr_rbug->con);
+   tr_rbug->con = NULL;
+}
+
+PIPE_THREAD_ROUTINE(trace_rbug_thread, void_tr_rbug)
+{
+   struct trace_rbug *tr_rbug = void_tr_rbug;
+   uint16_t port = 13370;
+   int s = -1;
+   int c;
+
+   u_socket_init();
+
+   for (;port <= 13379 && s < 0; port++)
+      s = u_socket_listen_on_port(port);
+
+   if (s < 0) {
+      debug_printf("trace_rbug - failed to listen\n");
+      return NULL;
+   }
+
+   u_socket_block(s, false);
+
+   debug_printf("trace_rbug - remote debugging listening on port %u\n", --port);
+
+   while(tr_rbug->running) {
+      sleep(1);
+
+      c = u_socket_accept(s);
+      if (c < 0)
+         continue;
+
+      u_socket_block(c, true);
+      tr_rbug->con = rbug_from_socket(c);
+
+      trace_rbug_con(tr_rbug);
+
+      u_socket_close(c);
+   }
+
+   u_socket_close(s);
+
+   u_socket_stop();
+
+   return NULL;
+}
+
+/**********************************************************
+ *
+ */
+
+struct trace_rbug *
+trace_rbug_start(struct trace_screen *tr_scr)
+{
+   struct trace_rbug *tr_rbug = CALLOC_STRUCT(trace_rbug);
+   if (!tr_rbug)
+      return NULL;
+
+   tr_rbug->tr_scr = tr_scr;
+   tr_rbug->running = TRUE;
+   tr_rbug->thread = pipe_thread_create(trace_rbug_thread, tr_rbug);
+
+   return tr_rbug;
+}
+
+void
+trace_rbug_stop(struct trace_rbug *tr_rbug)
+{
+   if (!tr_rbug)
+      return;
+
+   tr_rbug->running = false;
+   pipe_thread_wait(tr_rbug->thread);
+
+   FREE(tr_rbug);
+
+   return;
+}
+
+void
+trace_rbug_notify_draw_blocked(struct trace_context *tr_ctx)
+{
+   struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen);
+   struct trace_rbug *tr_rbug = tr_scr->rbug;
+
+   if (tr_rbug && tr_rbug->con)
+      rbug_send_context_draw_blocked(tr_rbug->con,
+                                     VOID2U64(tr_ctx), tr_ctx->draw_blocked, NULL);
+}
diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c
index 12a8535342..5b1e26a52d 100644
--- a/src/gallium/drivers/trace/tr_screen.c
+++ b/src/gallium/drivers/trace/tr_screen.c
@@ -30,13 +30,16 @@
 
 #include "tr_buffer.h"
 #include "tr_dump.h"
-#include "tr_state.h"
+#include "tr_dump_state.h"
 #include "tr_texture.h"
 #include "tr_screen.h"
 
 #include "pipe/p_inlines.h"
 
 
+static boolean trace = FALSE;
+static boolean rbug = FALSE;
+
 static const char *
 trace_screen_get_name(struct pipe_screen *_screen)
 {
@@ -212,10 +215,12 @@ static struct pipe_texture *
 trace_screen_texture_blanket(struct pipe_screen *_screen,
                              const struct pipe_texture *templat,
                              const unsigned *ppitch,
-                             struct pipe_buffer *buffer)
+                             struct pipe_buffer *_buffer)
 {
    struct trace_screen *tr_scr = trace_screen(_screen);
+   struct trace_buffer *tr_buf = trace_buffer(_buffer);
    struct pipe_screen *screen = tr_scr->screen;
+   struct pipe_buffer *buffer = tr_buf->buffer;
    unsigned pitch = *ppitch;
    struct pipe_texture *result;
 
@@ -818,18 +823,41 @@ trace_screen_destroy(struct pipe_screen *_screen)
    struct pipe_screen *screen = tr_scr->screen;
 
    trace_dump_call_begin("pipe_screen", "destroy");
-
    trace_dump_arg(ptr, screen);
-
-   screen->destroy(screen);
-
    trace_dump_call_end();
-
    trace_dump_trace_end();
 
+   if (tr_scr->rbug)
+      trace_rbug_stop(tr_scr->rbug);
+
+   screen->destroy(screen);
+
    FREE(tr_scr);
 }
 
+boolean
+trace_enabled(void)
+{
+   static boolean firstrun = TRUE;
+
+   if (!firstrun)
+      return trace;
+   firstrun = FALSE;
+
+   trace_dump_init();
+
+   if(trace_dump_trace_begin()) {
+      trace_dumping_start();
+      trace = TRUE;
+   }
+
+   if (debug_get_bool_option("GALLIUM_RBUG", FALSE)) {
+      trace = TRUE;
+      rbug = TRUE;
+   }
+
+   return trace;
+}
 
 struct pipe_screen *
 trace_screen_create(struct pipe_screen *screen)
@@ -840,13 +868,9 @@ trace_screen_create(struct pipe_screen *screen)
    if(!screen)
       goto error1;
 
-   trace_dump_init();
-
-   if(!trace_dump_trace_begin())
+   if (!trace_enabled())
       goto error1;
 
-   trace_dumping_start();
-
    trace_dump_call_begin("", "pipe_screen_create");
 
    tr_scr = CALLOC_STRUCT(trace_screen);
@@ -904,6 +928,9 @@ trace_screen_create(struct pipe_screen *screen)
    trace_dump_ret(ptr, screen);
    trace_dump_call_end();
 
+   if (rbug)
+      tr_scr->rbug = trace_rbug_start(tr_scr);
+
    return &tr_scr->base;
 
 #if 0
diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h
index 59f254166d..dba8cd7c65 100644
--- a/src/gallium/drivers/trace/tr_screen.h
+++ b/src/gallium/drivers/trace/tr_screen.h
@@ -57,6 +57,9 @@ struct trace_screen
 
    struct pipe_screen *screen;
 
+   /* remote debugger */
+   struct trace_rbug *rbug;
+
    pipe_mutex list_mutex;
    int num_buffers;
    int num_contexts;
@@ -71,14 +74,34 @@ struct trace_screen
 };
 
 
+/*
+ * tr_rbug.c
+ */
+
+
+struct trace_rbug;
+
+struct trace_rbug *
+trace_rbug_start(struct trace_screen *tr_scr);
+
+void
+trace_rbug_stop(struct trace_rbug *tr_rbug);
+
+
+/*
+ * tr_screen.c
+ */
+
+
+boolean
+trace_enabled(void);
+
 struct trace_screen *
 trace_screen(struct pipe_screen *screen);
 
-
 struct pipe_screen *
 trace_screen_create(struct pipe_screen *screen);
 
-
 void
 trace_screen_user_buffer_update(struct pipe_screen *screen,
                                 struct pipe_buffer *buffer);
@@ -99,6 +122,7 @@ trace_screen_user_buffer_update(struct pipe_screen *screen,
       pipe_mutex_unlock(tr_scr->list_mutex);             \
    } while (0)
 
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c
index a9570c1aeb..d8c11640bf 100644
--- a/src/gallium/drivers/trace/tr_state.c
+++ b/src/gallium/drivers/trace/tr_state.c
@@ -1,491 +1,66 @@
-/**************************************************************************
- *
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+/*
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
  *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
  *
- **************************************************************************/
-
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
 
-#include "pipe/p_compiler.h"
-#include "util/u_memory.h"
-#include "tgsi/tgsi_dump.h"
-
-#include "tr_dump.h"
 #include "tr_state.h"
 
+#include "util/u_memory.h"
+#include "util/u_simple_list.h"
 
-void trace_dump_format(enum pipe_format format)
-{
-   trace_dump_enum(pf_name(format) );
-}
-
-
-void trace_dump_block(const struct pipe_format_block *block)
-{
-   trace_dump_struct_begin("pipe_format_block");
-   trace_dump_member(uint, block, size);
-   trace_dump_member(uint, block, width);
-   trace_dump_member(uint, block, height);
-   trace_dump_struct_end();
-}
-
-
-static void trace_dump_reference(const struct pipe_reference *reference)
-{
-   trace_dump_struct_begin("pipe_reference");
-   trace_dump_member(int, &reference->count, count);
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_template(const struct pipe_texture *templat)
-{
-   if(!templat) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_texture");
-
-   trace_dump_member(int, templat, target);
-   trace_dump_member(format, templat, format);
-
-   trace_dump_member_begin("width");
-   trace_dump_array(uint, templat->width, 1);
-   trace_dump_member_end();
-
-   trace_dump_member_begin("height");
-   trace_dump_array(uint, templat->height, 1);
-   trace_dump_member_end();
-
-   trace_dump_member_begin("depth");
-   trace_dump_array(uint, templat->depth, 1);
-   trace_dump_member_end();
-
-   trace_dump_member_begin("block");
-   trace_dump_block(&templat->block);
-   trace_dump_member_end();
-
-   trace_dump_member(uint, templat, last_level);
-   trace_dump_member(uint, templat, tex_usage);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_rasterizer_state");
-
-   trace_dump_member(bool, state, flatshade);
-   trace_dump_member(bool, state, light_twoside);
-   trace_dump_member(uint, state, front_winding);
-   trace_dump_member(uint, state, cull_mode);
-   trace_dump_member(uint, state, fill_cw);
-   trace_dump_member(uint, state, fill_ccw);
-   trace_dump_member(bool, state, offset_cw);
-   trace_dump_member(bool, state, offset_ccw);
-   trace_dump_member(bool, state, scissor);
-   trace_dump_member(bool, state, poly_smooth);
-   trace_dump_member(bool, state, poly_stipple_enable);
-   trace_dump_member(bool, state, point_smooth);
-   trace_dump_member(bool, state, point_sprite);
-   trace_dump_member(bool, state, point_size_per_vertex);
-   trace_dump_member(bool, state, multisample);
-   trace_dump_member(bool, state, line_smooth);
-   trace_dump_member(bool, state, line_stipple_enable);
-   trace_dump_member(uint, state, line_stipple_factor);
-   trace_dump_member(uint, state, line_stipple_pattern);
-   trace_dump_member(bool, state, line_last_pixel);
-   trace_dump_member(bool, state, bypass_vs_clip_and_viewport);
-   trace_dump_member(bool, state, flatshade_first);
-   trace_dump_member(bool, state, gl_rasterization_rules);
-
-   trace_dump_member(float, state, line_width);
-   trace_dump_member(float, state, point_size);
-   trace_dump_member(float, state, point_size_min);
-   trace_dump_member(float, state, point_size_max);
-   trace_dump_member(float, state, offset_units);
-   trace_dump_member(float, state, offset_scale);
-
-   trace_dump_member_array(uint, state, sprite_coord_mode);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_poly_stipple(const struct pipe_poly_stipple *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_poly_stipple");
-
-   trace_dump_member_begin("stipple");
-   trace_dump_array(uint,
-                    state->stipple,
-                    Elements(state->stipple));
-   trace_dump_member_end();
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_viewport_state(const struct pipe_viewport_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_viewport_state");
-
-   trace_dump_member_array(float, state, scale);
-   trace_dump_member_array(float, state, translate);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_scissor_state(const struct pipe_scissor_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_scissor_state");
-
-   trace_dump_member(uint, state, minx);
-   trace_dump_member(uint, state, miny);
-   trace_dump_member(uint, state, maxx);
-   trace_dump_member(uint, state, maxy);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_clip_state(const struct pipe_clip_state *state)
-{
-   unsigned i;
-
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_clip_state");
-
-   trace_dump_member_begin("ucp");
-   trace_dump_array_begin();
-   for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) {
-      trace_dump_elem_begin();
-      trace_dump_array(float, state->ucp[i], 4);
-      trace_dump_elem_end();
-   }
-   trace_dump_array_end();
-   trace_dump_member_end();
-
-   trace_dump_member(uint, state, nr);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_constant_buffer(const struct pipe_constant_buffer *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_constant_buffer");
-
-   trace_dump_member(buffer_ptr, state, buffer);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_shader_state(const struct pipe_shader_state *state)
-{
-   static char str[8192];
-
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   tgsi_dump_str(state->tokens, 0, str, sizeof(str));
-
-   trace_dump_struct_begin("pipe_shader_state");
-
-   trace_dump_member_begin("tokens");
-   trace_dump_string(str);
-   trace_dump_member_end();
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state)
-{
-   unsigned i;
-
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_depth_stencil_alpha_state");
-
-   trace_dump_member_begin("depth");
-   trace_dump_struct_begin("pipe_depth_state");
-   trace_dump_member(bool, &state->depth, enabled);
-   trace_dump_member(bool, &state->depth, writemask);
-   trace_dump_member(uint, &state->depth, func);
-   trace_dump_member(bool, &state->depth, occlusion_count);
-   trace_dump_struct_end();
-   trace_dump_member_end();
-
-   trace_dump_member_begin("stencil");
-   trace_dump_array_begin();
-   for(i = 0; i < Elements(state->stencil); ++i) {
-      trace_dump_elem_begin();
-      trace_dump_struct_begin("pipe_stencil_state");
-      trace_dump_member(bool, &state->stencil[i], enabled);
-      trace_dump_member(uint, &state->stencil[i], func);
-      trace_dump_member(uint, &state->stencil[i], fail_op);
-      trace_dump_member(uint, &state->stencil[i], zpass_op);
-      trace_dump_member(uint, &state->stencil[i], zfail_op);
-      trace_dump_member(uint, &state->stencil[i], ref_value);
-      trace_dump_member(uint, &state->stencil[i], valuemask);
-      trace_dump_member(uint, &state->stencil[i], writemask);
-      trace_dump_struct_end();
-      trace_dump_elem_end();
-   }
-   trace_dump_array_end();
-   trace_dump_member_end();
-
-   trace_dump_member_begin("alpha");
-   trace_dump_struct_begin("pipe_alpha_state");
-   trace_dump_member(bool, &state->alpha, enabled);
-   trace_dump_member(uint, &state->alpha, func);
-   trace_dump_member(float, &state->alpha, ref_value);
-   trace_dump_struct_end();
-   trace_dump_member_end();
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_blend_state(const struct pipe_blend_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_blend_state");
-
-   trace_dump_member(bool, state, blend_enable);
-
-   trace_dump_member(uint, state, rgb_func);
-   trace_dump_member(uint, state, rgb_src_factor);
-   trace_dump_member(uint, state, rgb_dst_factor);
-
-   trace_dump_member(uint, state, alpha_func);
-   trace_dump_member(uint, state, alpha_src_factor);
-   trace_dump_member(uint, state, alpha_dst_factor);
-
-   trace_dump_member(bool, state, logicop_enable);
-   trace_dump_member(uint, state, logicop_func);
-
-   trace_dump_member(uint, state, colormask);
-   trace_dump_member(bool, state, dither);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_blend_color(const struct pipe_blend_color *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_blend_color");
-
-   trace_dump_member_array(float, state, color);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state)
-{
-   trace_dump_struct_begin("pipe_framebuffer_state");
-
-   trace_dump_member(uint, state, width);
-   trace_dump_member(uint, state, height);
-   trace_dump_member(uint, state, nr_cbufs);
-   trace_dump_member_array(ptr, state, cbufs);
-   trace_dump_member(ptr, state, zsbuf);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_sampler_state(const struct pipe_sampler_state *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_sampler_state");
-
-   trace_dump_member(uint, state, wrap_s);
-   trace_dump_member(uint, state, wrap_t);
-   trace_dump_member(uint, state, wrap_r);
-   trace_dump_member(uint, state, min_img_filter);
-   trace_dump_member(uint, state, min_mip_filter);
-   trace_dump_member(uint, state, mag_img_filter);
-   trace_dump_member(bool, state, compare_mode);
-   trace_dump_member(uint, state, compare_func);
-   trace_dump_member(bool, state, normalized_coords);
-   trace_dump_member(uint, state, prefilter);
-   trace_dump_member(float, state, shadow_ambient);
-   trace_dump_member(float, state, lod_bias);
-   trace_dump_member(float, state, min_lod);
-   trace_dump_member(float, state, max_lod);
-   trace_dump_member_array(float, state, border_color);
-   trace_dump_member(float, state, max_anisotropy);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_surface(const struct pipe_surface *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
-
-   trace_dump_struct_begin("pipe_surface");
-
-   trace_dump_reference(&state->reference);
-
-   trace_dump_member(format, state, format);
-   trace_dump_member(uint, state, width);
-   trace_dump_member(uint, state, height);
-
-   trace_dump_member(uint, state, layout);
-   trace_dump_member(uint, state, offset);
-   trace_dump_member(uint, state, usage);
-
-   trace_dump_member(ptr, state, texture);
-   trace_dump_member(uint, state, face);
-   trace_dump_member(uint, state, level);
-   trace_dump_member(uint, state, zslice);
-
-   trace_dump_struct_end();
-}
-
+#include "tgsi/tgsi_parse.h"
 
-void trace_dump_transfer(const struct pipe_transfer *state)
+struct trace_shader * trace_shader_create(struct trace_context *tr_ctx,
+                                          const struct pipe_shader_state *state,
+                                          void *result,
+                                          enum trace_shader_type type)
 {
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
+   struct trace_shader *tr_shdr = CALLOC_STRUCT(trace_shader);
 
-   trace_dump_struct_begin("pipe_transfer");
+   tr_shdr->state = result;
+   tr_shdr->type = type;
+   tr_shdr->tokens = tgsi_dup_tokens(state->tokens);
 
-   trace_dump_member(format, state, format);
-   trace_dump_member(uint, state, width);
-   trace_dump_member(uint, state, height);
+   /* works on context as well */
+   trace_screen_add_to_list(tr_ctx, shaders, tr_shdr);
 
-   trace_dump_member_begin("block");
-   trace_dump_block(&state->block);
-   trace_dump_member_end();
-
-   trace_dump_member(uint, state, nblocksx);
-   trace_dump_member(uint, state, nblocksy);
-   trace_dump_member(uint, state, stride);
-   trace_dump_member(uint, state, usage);
-
-   trace_dump_member(ptr, state, texture);
-   trace_dump_member(uint, state, face);
-   trace_dump_member(uint, state, level);
-   trace_dump_member(uint, state, zslice);
-
-   trace_dump_struct_end();
+   return tr_shdr;
 }
 
-
-void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state)
+void trace_shader_destroy(struct trace_context *tr_ctx,
+                          struct trace_shader *tr_shdr)
 {
-   if(!state) {
-      trace_dump_null();
-      return;
-   }
+   trace_screen_remove_from_list(tr_ctx, shaders, tr_shdr);
 
-   trace_dump_struct_begin("pipe_vertex_buffer");
-
-   trace_dump_member(uint, state, stride);
-   trace_dump_member(uint, state, max_index);
-   trace_dump_member(uint, state, buffer_offset);
-   trace_dump_member(buffer_ptr, state, buffer);
-
-   trace_dump_struct_end();
-}
-
-
-void trace_dump_vertex_element(const struct pipe_vertex_element *state)
-{
-   if(!state) {
-      trace_dump_null();
-      return;
+   if (tr_shdr->replaced) {
+      if (tr_shdr->type == TRACE_SHADER_FRAGMENT)
+         tr_ctx->pipe->delete_fs_state(tr_ctx->pipe, tr_shdr->replaced);
+      else if (tr_shdr->type == TRACE_SHADER_VERTEX)
+         tr_ctx->pipe->delete_vs_state(tr_ctx->pipe, tr_shdr->replaced);
+      else
+         assert(0);
    }
 
-   trace_dump_struct_begin("pipe_vertex_element");
-
-   trace_dump_member(uint, state, src_offset);
-
-   trace_dump_member(uint, state, vertex_buffer_index);
-   trace_dump_member(uint, state, nr_components);
-
-   trace_dump_member(format, state, src_format);
-
-   trace_dump_struct_end();
+   FREE(tr_shdr->replaced_tokens);
+   FREE(tr_shdr->tokens);
+   FREE(tr_shdr);
 }
diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h
index 513ed0ac98..1c16042ee5 100644
--- a/src/gallium/drivers/trace/tr_state.h
+++ b/src/gallium/drivers/trace/tr_state.h
@@ -1,78 +1,68 @@
-/**************************************************************************
- *
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+/*
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
  *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
  *
- **************************************************************************/
-
-#ifndef TR_STATE_H
-#define TR_STATE_H
-
-#include "pipe/p_format.h"
-#include "pipe/p_state.h"
-#include "pipe/p_shader_tokens.h"
-
-
-void trace_dump_format(enum pipe_format format);
-
-void trace_dump_block(const struct pipe_format_block *block);
-
-void trace_dump_template(const struct pipe_texture *templat);
-
-
-void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state);
-
-void trace_dump_poly_stipple(const struct pipe_poly_stipple *state);
-
-void trace_dump_viewport_state(const struct pipe_viewport_state *state);
-
-void trace_dump_scissor_state(const struct pipe_scissor_state *state);
-
-void trace_dump_clip_state(const struct pipe_clip_state *state);
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
 
-void trace_dump_constant_buffer(const struct pipe_constant_buffer *state);
+#ifndef TR_STATE_H_
+#define TR_STATE_H_
 
-void trace_dump_token(const struct tgsi_token *token);
+#include "tr_context.h"
 
-void trace_dump_shader_state(const struct pipe_shader_state *state);
+struct tgsi_token;
 
-void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state);
+enum trace_shader_type {
+   TRACE_SHADER_FRAGMENT = 0,
+   TRACE_SHADER_VERTEX   = 1,
+   TRACE_SHADER_GEOMETRY = 2,
+};
 
-void trace_dump_blend_state(const struct pipe_blend_state *state);
+struct trace_shader
+{
+   struct tr_list list;
 
-void trace_dump_blend_color(const struct pipe_blend_color *state);
+   enum trace_shader_type type;
 
-void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state);
+   void *state;
+   void *replaced;
 
-void trace_dump_sampler_state(const struct pipe_sampler_state *state);
+   struct tgsi_token *tokens;
+   struct tgsi_token *replaced_tokens;
 
-void trace_dump_surface(const struct pipe_surface *state);
+   boolean disabled;
+};
 
-void trace_dump_transfer(const struct pipe_transfer *state);
 
-void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state);
+static INLINE struct trace_shader *
+trace_shader(void *state)
+{
+   return (struct trace_shader *)state;
+}
 
-void trace_dump_vertex_element(const struct pipe_vertex_element *state);
+struct trace_shader * trace_shader_create(struct trace_context *tr_ctx,
+                                          const struct pipe_shader_state *state,
+                                          void *result,
+                                          enum trace_shader_type type);
 
+void trace_shader_destroy(struct trace_context *tr_ctx,
+                          struct trace_shader *tr_shdr);
 
-#endif /* TR_STATE_H */
+#endif
diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h
index 63238ea46e..4152d6ac36 100644
--- a/src/gallium/include/pipe/p_config.h
+++ b/src/gallium/include/pipe/p_config.h
@@ -102,6 +102,19 @@
 
 
 /*
+ * Endian detection.
+ */
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+#define PIPE_ARCH_LITTLE_ENDIAN
+#elif defined(PIPE_ARCH_PPC) || defined(PIPE_ARCH_PPC_64)
+#define PIPE_ARCH_BIG_ENDIAN
+#else
+#define PIPE_ARCH_UNKNOWN_ENDIAN
+#endif
+
+
+/*
  * Operating system family.
  * 
  * See subsystem below for a more fine-grained distinction.
@@ -119,6 +132,10 @@
 #define PIPE_OS_SOLARIS
 #endif
 
+#if defined(__APPLE__)
+#define PIPE_OS_APPLE
+#endif
+
 #if defined(_WIN32) || defined(WIN32)
 #define PIPE_OS_WINDOWS
 #endif
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 5c16884c16..bc4bc70759 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -132,6 +132,7 @@ enum pipe_texture_target {
 #define PIPE_TEX_FACE_NEG_Y 3
 #define PIPE_TEX_FACE_POS_Z 4
 #define PIPE_TEX_FACE_NEG_Z 5
+#define PIPE_TEX_FACE_MAX   6
 
 #define PIPE_TEX_WRAP_REPEAT                   0
 #define PIPE_TEX_WRAP_CLAMP                    1
@@ -158,14 +159,6 @@ enum pipe_texture_target {
 #define PIPE_TEX_COMPARE_NONE          0
 #define PIPE_TEX_COMPARE_R_TO_TEXTURE  1
 
-#define PIPE_TEX_FACE_POS_X   0
-#define PIPE_TEX_FACE_NEG_X   1
-#define PIPE_TEX_FACE_POS_Y   2
-#define PIPE_TEX_FACE_NEG_Y   3
-#define PIPE_TEX_FACE_POS_Z   4
-#define PIPE_TEX_FACE_NEG_Z   5
-#define PIPE_TEX_FACE_MAX     6
-
 #define PIPE_TEXTURE_USAGE_RENDER_TARGET   0x1
 #define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2 /* ie a backbuffer */
 #define PIPE_TEXTURE_USAGE_PRIMARY         0x4 /* ie a frontbuffer */
@@ -281,9 +274,8 @@ enum pipe_transfer_usage {
 
 
 /**
- * Implementation capabilities/limits
- * Passed to pipe->get_param()
- * XXX this will need some fine tuning...
+ * Implementation capabilities/limits which are queried through
+ * pipe_screen::get_param() and pipe_screen::get_paramf().
  */
 #define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1
 #define PIPE_CAP_NPOT_TEXTURES           2
@@ -312,6 +304,7 @@ enum pipe_transfer_usage {
 #define PIPE_CAP_TEXTURE_MIRROR_REPEAT   25
 #define PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS 26
 #define PIPE_CAP_TGSI_CONT_SUPPORTED     27
+#define PIPE_CAP_BLEND_EQUATION_SEPARATE 28
 
 
 /**
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 4b590bdc90..626bedb35a 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -75,11 +75,11 @@ struct pipe_surface;
  */
 struct pipe_buffer
 {
-   struct pipe_reference reference;
-   struct pipe_screen *screen;
-   unsigned alignment;
-   unsigned usage;
-   unsigned size;
+   struct pipe_reference  reference;
+   unsigned               size;
+   struct pipe_screen    *screen;
+   unsigned               alignment;
+   unsigned               usage;
 };
 
 
@@ -265,7 +265,6 @@ struct pipe_sampler_state
    unsigned compare_func:3;      /**< PIPE_FUNC_x */
    unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */
    unsigned prefilter:4;         /**< Wierd sampling state exposed by some api's */
-   float shadow_ambient;         /**< shadow test fail color/intensity */
    float lod_bias;               /**< LOD/lambda bias */
    float min_lod, max_lod;       /**< LOD clamp range, after bias */
    float border_color[4];
@@ -287,10 +286,10 @@ struct pipe_surface
    unsigned offset;              /**< offset from start of buffer, in bytes */
    unsigned usage;               /**< PIPE_BUFFER_USAGE_*  */
 
+   unsigned zslice;
    struct pipe_texture *texture; /**< texture into which this is a view  */
    unsigned face;
    unsigned level;
-   unsigned zslice;
 };
 
 
diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h
index ce8d79549a..96e8e08744 100644
--- a/src/gallium/include/pipe/p_thread.h
+++ b/src/gallium/include/pipe/p_thread.h
@@ -44,6 +44,8 @@
 #include <pthread.h> /* POSIX threads headers */
 #include <stdio.h> /* for perror() */
 
+#define PIPE_THREAD_HAVE_CONDVAR
+
 typedef pthread_t pipe_thread;
 
 #define PIPE_THREAD_ROUTINE( name, param ) \
diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h
index 5790b2f6c7..7a38b508fb 100644
--- a/src/gallium/include/state_tracker/drm_api.h
+++ b/src/gallium/include/state_tracker/drm_api.h
@@ -32,33 +32,37 @@ struct drm_api
 	 * Special buffer functions
 	 */
 	/*@{*/
-	struct pipe_screen*  (*create_screen)(int drm_fd,
-					      struct drm_create_screen_arg *arg);
-	struct pipe_context* (*create_context)(struct pipe_screen *screen);
+	struct pipe_screen*  (*create_screen)(struct drm_api *api, int drm_fd,
+	                                      struct drm_create_screen_arg *arg);
+	struct pipe_context* (*create_context)(struct drm_api *api,
+	                                       struct pipe_screen *screen);
 	/*@}*/
 
 	/**
 	 * Special buffer functions
 	 */
 	/*@{*/
-	boolean (*buffer_from_texture)(struct pipe_texture *texture,
-                                       struct pipe_buffer **buffer,
-                                       unsigned *stride);
-	struct pipe_buffer* (*buffer_from_handle)(struct pipe_screen *screen,
+	boolean (*buffer_from_texture)(struct drm_api *api,
+	                               struct pipe_texture *texture,
+	                               struct pipe_buffer **buffer,
+	                               unsigned *stride);
+	struct pipe_buffer* (*buffer_from_handle)(struct drm_api *api,
+	                                          struct pipe_screen *screen,
                                                   const char *name,
                                                   unsigned handle);
-	boolean (*handle_from_buffer)(struct pipe_screen *screen,
-                                      struct pipe_buffer *buffer,
-                                      unsigned *handle);
-	boolean (*global_handle_from_buffer)(struct pipe_screen *screen,
-                                             struct pipe_buffer *buffer,
-                                             unsigned *handle);
+	boolean (*handle_from_buffer)(struct drm_api *api,
+	                              struct pipe_screen *screen,
+	                              struct pipe_buffer *buffer,
+	                              unsigned *handle);
+	boolean (*global_handle_from_buffer)(struct drm_api *api,
+	                                     struct pipe_screen *screen,
+	                                     struct pipe_buffer *buffer,
+	                                     unsigned *handle);
 	/*@}*/
+
+	void (*destroy)(struct drm_api *api);
 };
 
-/**
- * A driver needs to export this symbol
- */
-extern struct drm_api drm_api_hooks;
+extern struct drm_api * drm_api_create(void);
 
 #endif
diff --git a/src/gallium/state_trackers/dri/dri_context.c b/src/gallium/state_trackers/dri/dri_context.c
index 45eaec4ed3..6c617197ec 100644
--- a/src/gallium/state_trackers/dri/dri_context.c
+++ b/src/gallium/state_trackers/dri/dri_context.c
@@ -69,7 +69,7 @@ dri_create_context(const __GLcontextModes * visual,
    driParseConfigFiles(&ctx->optionCache,
 		       &screen->optionCache, sPriv->myNum, "dri");
 
-   ctx->pipe = drm_api_hooks.create_context(screen->pipe_screen);
+   ctx->pipe = screen->api->create_context(screen->api, screen->pipe_screen);
 
    if (ctx->pipe == NULL)
       goto fail;
diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c
index 5846390d02..1d91fbb89f 100644
--- a/src/gallium/state_trackers/dri/dri_drawable.c
+++ b/src/gallium/state_trackers/dri/dri_drawable.c
@@ -53,7 +53,8 @@ dri_copy_to_front(__DRIdrawablePrivate * dPriv,
 }
 
 static struct pipe_surface *
-dri_surface_from_handle(struct pipe_screen *screen,
+dri_surface_from_handle(struct drm_api *api,
+			struct pipe_screen *screen,
 			unsigned handle,
 			enum pipe_format format,
 			unsigned width, unsigned height, unsigned pitch)
@@ -63,7 +64,7 @@ dri_surface_from_handle(struct pipe_screen *screen,
    struct pipe_texture templat;
    struct pipe_buffer *buf = NULL;
 
-   buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle);
+   buf = api->buffer_from_handle(api, screen, "dri2 buffer", handle);
    if (!buf)
       return NULL;
 
@@ -100,12 +101,14 @@ dri_surface_from_handle(struct pipe_screen *screen,
 void
 dri_get_buffers(__DRIdrawablePrivate * dPriv)
 {
+
    struct dri_drawable *drawable = dri_drawable(dPriv);
    struct pipe_surface *surface = NULL;
    struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen;
    __DRIbuffer *buffers = NULL;
    __DRIscreen *dri_screen = drawable->sPriv;
    __DRIdrawable *dri_drawable = drawable->dPriv;
+   struct drm_api *api = ((struct dri_screen*)(dri_screen->private))->api;
    boolean have_depth = FALSE;
    int i, count;
 
@@ -138,6 +141,18 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv)
    dri_drawable->pBackClipRects[0].x2 = dri_drawable->w;
    dri_drawable->pBackClipRects[0].y2 = dri_drawable->h;
 
+   if (drawable->old_num == count &&
+       drawable->old_w == dri_drawable->w &&
+       drawable->old_h == dri_drawable->h &&
+       memcmp(drawable->old, buffers, sizeof(__DRIbuffer) * count) == 0) {
+       return;
+   } else {
+      drawable->old_num = count;
+      drawable->old_w = dri_drawable->w;
+      drawable->old_h = dri_drawable->h;
+      memcpy(drawable->old, buffers, sizeof(__DRIbuffer) * count);
+   }
+
    for (i = 0; i < count; i++) {
       enum pipe_format format = 0;
       int index = 0;
@@ -145,29 +160,28 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv)
       switch (buffers[i].attachment) {
       case __DRI_BUFFER_FRONT_LEFT:
 	 index = ST_SURFACE_FRONT_LEFT;
-	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 format = drawable->color_format;
 	 break;
       case __DRI_BUFFER_FAKE_FRONT_LEFT:
 	 index = ST_SURFACE_FRONT_LEFT;
-	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 format = drawable->color_format;
 	 break;
       case __DRI_BUFFER_BACK_LEFT:
 	 index = ST_SURFACE_BACK_LEFT;
-	 format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	 format = drawable->color_format;
 	 break;
       case __DRI_BUFFER_DEPTH:
 	 index = ST_SURFACE_DEPTH;
-	 format = PIPE_FORMAT_Z24S8_UNORM;
+	 format = drawable->depth_format;
 	 break;
       case __DRI_BUFFER_STENCIL:
 	 index = ST_SURFACE_DEPTH;
-	 format = PIPE_FORMAT_Z24S8_UNORM;
+	 format = drawable->stencil_format;
 	 break;
       case __DRI_BUFFER_ACCUM:
       default:
 	 assert(0);
       }
-      assert(buffers[i].cpp == 4);
 
       if (index == ST_SURFACE_DEPTH) {
 	 if (have_depth)
@@ -176,7 +190,8 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv)
 	    have_depth = TRUE;
       }
 
-      surface = dri_surface_from_handle(screen,
+      surface = dri_surface_from_handle(api,
+					screen,
 					buffers[i].name,
 					format,
 					dri_drawable->w,
@@ -206,10 +221,8 @@ dri_create_buffer(__DRIscreenPrivate * sPriv,
 		  __DRIdrawablePrivate * dPriv,
 		  const __GLcontextModes * visual, boolean isPixmap)
 {
-   enum pipe_format colorFormat, depthFormat, stencilFormat;
    struct dri_screen *screen = sPriv->private;
    struct dri_drawable *drawable = NULL;
-   struct pipe_screen *pscreen = screen->pipe_screen;
    int i;
 
    if (isPixmap)
@@ -219,39 +232,52 @@ dri_create_buffer(__DRIscreenPrivate * sPriv,
    if (drawable == NULL)
       goto fail;
 
-   /* XXX: todo: use the pipe_screen queries to figure out which
-    * render targets are supportable.
-    */
-   assert(visual->redBits == 8);
-   assert(visual->depthBits == 24 || visual->depthBits == 0);
-   assert(visual->stencilBits == 8 || visual->stencilBits == 0);
-
-   colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
+   drawable->color_format = (visual->redBits == 8) ?
+      PIPE_FORMAT_A8R8G8B8_UNORM : PIPE_FORMAT_R5G6B5_UNORM;
+
+   debug_printf("Red bits is %d\n", visual->redBits);
+
+   switch(visual->depthBits) {
+   default:
+   case 0:
+      debug_printf("Depth buffer 0.\n");
+      drawable->depth_format = PIPE_FORMAT_NONE;
+      break;
+   case 16:
+      debug_printf("Depth buffer 16.\n");
+      drawable->depth_format = PIPE_FORMAT_Z16_UNORM;
+      break;
+   case 24:
+      if (visual->stencilBits == 0) {
+         debug_printf("Depth buffer 24. Stencil 0.\n");
+	 drawable->depth_format = (screen->d_depth_bits_last) ?
+	    PIPE_FORMAT_X8Z24_UNORM:
+	    PIPE_FORMAT_Z24X8_UNORM;
+      } else {
+         debug_printf("Combined depth stencil 24 / 8.\n");
+	 drawable->depth_format = (screen->sd_depth_bits_last) ?
+	    PIPE_FORMAT_S8Z24_UNORM:
+	    PIPE_FORMAT_Z24S8_UNORM;
+      }
+      break;
+   }
 
-   if (visual->depthBits) {
-      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
-				       PIPE_TEXTURE_2D,
-				       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
-	 depthFormat = PIPE_FORMAT_Z24S8_UNORM;
-      else
-	 depthFormat = PIPE_FORMAT_S8Z24_UNORM;
-   } else
-      depthFormat = PIPE_FORMAT_NONE;
-
-   if (visual->stencilBits) {
-      if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM,
-				       PIPE_TEXTURE_2D,
-				       PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0))
-	 stencilFormat = PIPE_FORMAT_Z24S8_UNORM;
-      else
-	 stencilFormat = PIPE_FORMAT_S8Z24_UNORM;
-   } else
-      stencilFormat = PIPE_FORMAT_NONE;
+   switch(visual->stencilBits) {
+   default:
+   case 0:
+      drawable->stencil_format = PIPE_FORMAT_NONE;
+      break;
+   case 8:
+      drawable->stencil_format = (screen->sd_depth_bits_last) ?
+	 PIPE_FORMAT_S8Z24_UNORM:
+         PIPE_FORMAT_Z24S8_UNORM;
+      break;
+   }
 
    drawable->stfb = st_create_framebuffer(visual,
-					  colorFormat,
-					  depthFormat,
-					  stencilFormat,
+					  drawable->color_format,
+					  drawable->depth_format,
+					  drawable->stencil_format,
 					  dPriv->w,
 					  dPriv->h, (void *)drawable);
    if (drawable->stfb == NULL)
diff --git a/src/gallium/state_trackers/dri/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h
index 78a66624aa..2fbd5f1eb7 100644
--- a/src/gallium/state_trackers/dri/dri_drawable.h
+++ b/src/gallium/state_trackers/dri/dri_drawable.h
@@ -46,6 +46,11 @@ struct dri_drawable
    unsigned attachments[8];
    unsigned num_attachments;
 
+   __DRIbuffer old[8];
+   unsigned old_num;
+   unsigned old_w;
+   unsigned old_h;
+
    /* gallium */
    struct st_framebuffer *stfb;
    struct pipe_fence_handle *swap_fences[DRI_SWAP_FENCES_MAX];
@@ -53,6 +58,10 @@ struct dri_drawable
    unsigned int tail;
    unsigned int desired_fences;
    unsigned int cur_fences;
+
+   enum pipe_format color_format;
+   enum pipe_format depth_format;
+   enum pipe_format stencil_format;
 };
 
 static INLINE struct dri_drawable *
diff --git a/src/gallium/state_trackers/dri/dri_extensions.c b/src/gallium/state_trackers/dri/dri_extensions.c
index 0c59d42d5c..7c04c2b970 100644
--- a/src/gallium/state_trackers/dri/dri_extensions.c
+++ b/src/gallium/state_trackers/dri/dri_extensions.c
@@ -36,9 +36,11 @@
 #define need_GL_ARB_multisample
 #define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
 #define need_GL_ARB_texture_compression
 #define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
 #define need_GL_ARB_window_pos
 #define need_GL_EXT_blend_color
 #define need_GL_EXT_blend_equation_separate
@@ -50,16 +52,23 @@
 #define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_secondary_color
 #define need_GL_NV_vertex_program
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
 #include "extension_helper.h"
 
 /**
  * Extension strings exported by the driver.
  */
 const struct dri_extension card_extensions[] = {
+   {"GL_ARB_fragment_shader", NULL},
    {"GL_ARB_multisample", GL_ARB_multisample_functions},
    {"GL_ARB_multitexture", NULL},
    {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions},
+   {"GL_ARB_pixel_buffer_object", NULL},
    {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
+   {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions },
+   {"GL_ARB_shading_language_120", GL_VERSION_2_1_functions },
+   {"GL_ARB_shader_objects", GL_ARB_shader_objects_functions},
    {"GL_ARB_texture_border_clamp", NULL},
    {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
    {"GL_ARB_texture_cube_map", NULL},
@@ -69,12 +78,11 @@ const struct dri_extension card_extensions[] = {
    {"GL_ARB_texture_mirrored_repeat", NULL},
    {"GL_ARB_texture_rectangle", NULL},
    {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
-   {"GL_ARB_pixel_buffer_object", NULL},
+   {"GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions},
    {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
    {"GL_ARB_window_pos", GL_ARB_window_pos_functions},
    {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
-   {"GL_EXT_blend_equation_separate",
-    GL_EXT_blend_equation_separate_functions},
+   {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
    {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
    {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
    {"GL_EXT_blend_subtract", NULL},
diff --git a/src/gallium/state_trackers/dri/dri_screen.c b/src/gallium/state_trackers/dri/dri_screen.c
index d3392ee690..5f78b7264a 100644
--- a/src/gallium/state_trackers/dri/dri_screen.c
+++ b/src/gallium/state_trackers/dri/dri_screen.c
@@ -69,39 +69,65 @@ PUBLIC const char __driConfigOptions[] =
 struct dri1_api *__dri1_api_hooks = NULL;
 
 static const __DRIconfig **
-dri_fill_in_modes(__DRIscreenPrivate * psp,
-		  unsigned pixel_bits, unsigned depth_bits,
-		  unsigned stencil_bits, GLboolean have_back_buffer)
+dri_fill_in_modes(struct dri_screen *screen,
+		  unsigned pixel_bits)
 {
    __DRIconfig **configs;
-   __GLcontextModes *m;
    unsigned num_modes;
-   uint8_t depth_bits_array[3];
-   uint8_t stencil_bits_array[3];
+   uint8_t depth_bits_array[4];
+   uint8_t stencil_bits_array[4];
    uint8_t msaa_samples_array[1];
    unsigned depth_buffer_factor;
    unsigned back_buffer_factor;
    unsigned msaa_samples_factor;
    GLenum fb_format;
    GLenum fb_type;
-   int i;
+   struct pipe_screen *p_screen = screen->pipe_screen;
 
    static const GLenum back_buffer_modes[] = {
       GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
    };
 
-   /* TODO probe the hardware of what is supports */
    depth_bits_array[0] = 0;
-   depth_bits_array[1] = 24;
-   depth_bits_array[2] = 24;
-
-   stencil_bits_array[0] = 0;	       /* no depth or stencil */
-   stencil_bits_array[1] = 0;	       /* z24x8 */
-   stencil_bits_array[2] = 8;	       /* z24s8 */
+   stencil_bits_array[0] = 0;
+   depth_buffer_factor = 1;
+
+   if (p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z16_UNORM,
+				     PIPE_TEXTURE_2D,
+				     PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) {
+      depth_bits_array[depth_buffer_factor] = 16;
+      stencil_bits_array[depth_buffer_factor++] = 0;
+   }
+   if (p_screen->is_format_supported(p_screen, PIPE_FORMAT_X8Z24_UNORM,
+				     PIPE_TEXTURE_2D,
+				     PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) {
+      depth_bits_array[depth_buffer_factor] = 24;
+      stencil_bits_array[depth_buffer_factor++] = 0;
+      screen->d_depth_bits_last = TRUE;
+   } else if (p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z24X8_UNORM,
+					    PIPE_TEXTURE_2D,
+					    PIPE_TEXTURE_USAGE_DEPTH_STENCIL,
+					    0)) {
+      depth_bits_array[depth_buffer_factor] = 24;
+      stencil_bits_array[depth_buffer_factor++] = 0;
+      screen->d_depth_bits_last = FALSE;
+   }
+   if (p_screen->is_format_supported(p_screen, PIPE_FORMAT_S8Z24_UNORM,
+				     PIPE_TEXTURE_2D,
+				     PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) {
+      depth_bits_array[depth_buffer_factor] = 24;
+      stencil_bits_array[depth_buffer_factor++] = 8;
+      screen->sd_depth_bits_last = TRUE;
+   } else if (p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z24S8_UNORM,
+					    PIPE_TEXTURE_2D,
+					    PIPE_TEXTURE_USAGE_DEPTH_STENCIL,
+					    0)) {
+      depth_bits_array[depth_buffer_factor] = 24;
+      stencil_bits_array[depth_buffer_factor++] = 8;
+      screen->sd_depth_bits_last = FALSE;
+   }
 
    msaa_samples_array[0] = 0;
-
-   depth_buffer_factor = 3;
    back_buffer_factor = 3;
    msaa_samples_factor = 1;
 
@@ -109,9 +135,19 @@ dri_fill_in_modes(__DRIscreenPrivate * psp,
       depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4;
 
    if (pixel_bits == 16) {
+      if (!p_screen->is_format_supported(p_screen,
+					 PIPE_FORMAT_R5G6B5_UNORM,
+					 PIPE_TEXTURE_2D,
+					 PIPE_TEXTURE_USAGE_RENDER_TARGET, 0))
+	 return NULL;
       fb_format = GL_RGB;
       fb_type = GL_UNSIGNED_SHORT_5_6_5;
    } else {
+      if (!p_screen->is_format_supported(p_screen,
+					 PIPE_FORMAT_A8R8G8B8_UNORM,
+					 PIPE_TEXTURE_2D,
+					 PIPE_TEXTURE_USAGE_RENDER_TARGET, 0))
+	 return NULL;
       fb_format = GL_BGRA;
       fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
    }
@@ -126,13 +162,6 @@ dri_fill_in_modes(__DRIscreenPrivate * psp,
       return NULL;
    }
 
-   for (i = 0; configs[i]; i++) {
-      m = &configs[i]->modes;
-      if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
-	 m->visualRating = GLX_SLOW_CONFIG;
-      }
-   }
-
    return (const const __DRIconfig **)configs;
 }
 
@@ -170,6 +199,7 @@ dri_init_screen(__DRIscreenPrivate * sPriv)
    if (!screen)
       return NULL;
 
+   screen->api = drm_api_create();
    screen->sPriv = sPriv;
    screen->fd = sPriv->fd;
    screen->drmLock = (drmLock *) & sPriv->pSAREA->lock;
@@ -187,7 +217,7 @@ dri_init_screen(__DRIscreenPrivate * sPriv)
    dri_copy_version(&arg.drm_version, &sPriv->drm_version);
    arg.api = NULL;
 
-   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg.base);
+   screen->pipe_screen = screen->api->create_screen(screen->api, screen->fd, &arg.base);
 
    if (!screen->pipe_screen || !arg.api) {
       debug_printf("%s: failed to create dri1 screen\n", __FUNCTION__);
@@ -200,7 +230,13 @@ dri_init_screen(__DRIscreenPrivate * sPriv)
    driParseOptionInfo(&screen->optionCache,
 		      __driConfigOptions, __driNConfigOptions);
 
-   configs = dri_fill_in_modes(sPriv, sPriv->fbBPP, 24, 8, 1);
+   /**
+    * FIXME: If the driver supports format conversion swapbuffer blits, we might
+    * want to support other color bit depths than the server is currently
+    * using.
+    */
+
+   configs = dri_fill_in_modes(screen, sPriv->fbBPP);
    if (!configs)
       goto out_no_configs;
 
@@ -230,13 +266,14 @@ dri_init_screen2(__DRIscreenPrivate * sPriv)
    if (!screen)
       goto fail;
 
+   screen->api = drm_api_create();
    screen->sPriv = sPriv;
    screen->fd = sPriv->fd;
    sPriv->private = (void *)screen;
    sPriv->extensions = dri_screen_extensions;
    arg.mode = DRM_CREATE_NORMAL;
 
-   screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg);
+   screen->pipe_screen = screen->api->create_screen(screen->api, screen->fd, &arg);
    if (!screen->pipe_screen) {
       debug_printf("%s: failed to create pipe_screen\n", __FUNCTION__);
       goto fail;
@@ -248,7 +285,7 @@ dri_init_screen2(__DRIscreenPrivate * sPriv)
    driParseOptionInfo(&screen->optionCache,
 		      __driConfigOptions, __driNConfigOptions);
 
-   return dri_fill_in_modes(sPriv, 4 * 8, 24, 8, 1);
+   return dri_fill_in_modes(screen, 32);
  fail:
    return NULL;
 }
diff --git a/src/gallium/state_trackers/dri/dri_screen.h b/src/gallium/state_trackers/dri/dri_screen.h
index 100d9e50e0..f3335bb09f 100644
--- a/src/gallium/state_trackers/dri/dri_screen.h
+++ b/src/gallium/state_trackers/dri/dri_screen.h
@@ -60,8 +60,11 @@ struct dri_screen
    drmLock *drmLock;
 
    /* gallium */
+   struct drm_api *api;
    struct pipe_winsys *pipe_winsys;
    struct pipe_screen *pipe_screen;
+   boolean d_depth_bits_last;
+   boolean sd_depth_bits_last;
 };
 
 /** cast wrapper */
diff --git a/src/gallium/state_trackers/egl/Makefile b/src/gallium/state_trackers/egl/Makefile
index 692a3c8b76..e825aa718b 100644
--- a/src/gallium/state_trackers/egl/Makefile
+++ b/src/gallium/state_trackers/egl/Makefile
@@ -1,29 +1,19 @@
-TARGET     = libegldrm.a
-CFILES     = $(wildcard ./*.c)
-OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
-GALLIUMDIR = ../..
-TOP        = ../../../..
+TOP = ../../../..
+include $(TOP)/configs/current
 
-include ${TOP}/configs/current
+LIBNAME = egldrm
 
-CFLAGS := \
-          -I${GALLIUMDIR}/include \
-          -I${GALLIUMDIR}/auxiliary \
-          -I${TOP}/src/mesa/drivers/dri/common \
-          -I${TOP}/src/mesa \
-          -I$(TOP)/include \
-          -I$(TOP)/src/egl/main \
-          ${LIBDRM_CFLAGS} \
-          ${CFLAGS}
+LIBRARY_INCLUDES = \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/include \
+	-I$(TOP)/src/egl/main \
+	$(shell pkg-config --cflags-only-I libdrm)
 
-#############################################
 
-.PHONY	= all clean
+C_SOURCES = $(wildcard ./*.c)
 
-all: ${TARGET}
 
-${TARGET}: ${OBJECTS}
-	ar rcs $@ $^
-
-clean:
-	rm -rf ${OBJECTS} ${TARGET}
+include ../../Makefile.template
diff --git a/src/gallium/state_trackers/egl/egl_context.c b/src/gallium/state_trackers/egl/egl_context.c
index 36548fae26..e2da2180f7 100644
--- a/src/gallium/state_trackers/egl/egl_context.c
+++ b/src/gallium/state_trackers/egl/egl_context.c
@@ -113,9 +113,9 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext
 	if (!ctx)
 		goto err_c;
 
-	_eglInitContext(drv, dpy, &ctx->base, config, attrib_list);
+	_eglInitContext(drv, &ctx->base, conf, attrib_list);
 
-	ctx->pipe = drm_api_hooks.create_context(dev->screen);
+	ctx->pipe = dev->api->create_context(dev->api, dev->screen);
 	if (!ctx->pipe)
 		goto err_pipe;
 
@@ -129,8 +129,8 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext
 	if (!ctx->st)
 		goto err_gl;
 
-	/* generate handle and insert into hash table */
-	_eglSaveContext(&ctx->base);
+        /* link to display */
+        _eglLinkContext(&ctx->base, _eglLookupDisplay(dpy));
 	assert(_eglGetContextHandle(&ctx->base));
 
 	return _eglGetContextHandle(&ctx->base);
@@ -147,10 +147,8 @@ EGLBoolean
 drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
 {
 	struct drm_context *c = lookup_drm_context(context);
-	_eglRemoveContext(&c->base);
-	if (c->base.IsBound) {
-		c->base.DeletePending = EGL_TRUE;
-	} else {
+        _eglUnlinkContext(&c->base);
+	if (!c->base.IsBound) {
 		st_destroy_context(c->st);
 		c->pipe->destroy(c->pipe);
 		free(c);
diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c
index 489aa8d9af..86f2ea97e5 100644
--- a/src/gallium/state_trackers/egl/egl_surface.c
+++ b/src/gallium/state_trackers/egl/egl_surface.c
@@ -132,7 +132,7 @@ drm_create_texture(_EGLDriver *drv,
 	scrn->front.width = w;
 	scrn->front.height = h;
 	scrn->front.pitch = pitch;
-	drm_api_hooks.handle_from_buffer(screen, scrn->buffer, &scrn->front.handle);
+	dev->api->handle_from_buffer(dev->api, screen, scrn->buffer, &scrn->front.handle);
 	if (0)
 		goto err_handle;
 
@@ -232,7 +232,7 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 	if (!surf)
 		goto err;
 
-	if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list))
+	if (!_eglInitSurface(drv, &surf->base, EGL_PBUFFER_BIT, conf, attrib_list))
 		goto err_surf;
 
 	surf->w = width;
@@ -245,7 +245,7 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 	                                    (void*)surf);
 	drm_visual_modes_destroy(visual);
 
-	_eglSaveSurface(&surf->base);
+        _eglLinkSurface(&surf->base, _eglLookupDisplay(dpy));
 	return surf->base.Handle;
 
 err_surf:
@@ -364,10 +364,9 @@ EGLBoolean
 drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
 	struct drm_surface *surf = lookup_drm_surface(surface);
-	_eglRemoveSurface(&surf->base);
-	if (surf->base.IsBound) {
-		surf->base.DeletePending = EGL_TRUE;
-	} else {
+	_eglUnlinkSurface(&surf->base);
+
+	if (!surf->base.IsBound) {
 		if (surf->screen)
 			drm_takedown_shown_screen(drv, surf->screen);
 		st_unreference_framebuffer(surf->stfb);
diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c
index 8e62008461..521c91d895 100644
--- a/src/gallium/state_trackers/egl/egl_tracker.c
+++ b/src/gallium/state_trackers/egl/egl_tracker.c
@@ -35,6 +35,8 @@ _eglMain(_EGLDisplay *dpy, const char *args)
 		return NULL;
 	}
 
+	drm->api = drm_api_create();
+
 	/* First fill in the dispatch table with defaults */
 	_eglInitDriverFallbacks(&drm->base);
 	/* then plug in our Drm-specific functions */
@@ -146,7 +148,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 	dev->drmFD = fd;
 	drm_get_device_id(dev);
 
-	dev->screen = drm_api_hooks.create_screen(dev->drmFD, NULL);
+	dev->screen = dev->api->create_screen(dev->api, dev->drmFD, NULL);
 	if (!dev->screen)
 		goto err_screen;
 	dev->winsys = dev->screen->winsys;
@@ -234,6 +236,7 @@ drm_terminate(_EGLDriver *drv, EGLDisplay dpy)
 
 	dev->screen->destroy(dev->screen);
 	dev->winsys = NULL;
+	dev->api->destroy(dev->api);
 
 	drmClose(dev->drmFD);
 
diff --git a/src/gallium/state_trackers/egl/egl_tracker.h b/src/gallium/state_trackers/egl/egl_tracker.h
index ce2717de63..3b8836720d 100644
--- a/src/gallium/state_trackers/egl/egl_tracker.h
+++ b/src/gallium/state_trackers/egl/egl_tracker.h
@@ -38,6 +38,7 @@ struct drm_device
 	 * pipe
 	 */
 
+	struct drm_api *api;
 	struct pipe_winsys *winsys;
 	struct pipe_screen *screen;
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
index 16d4f1e32c..643e68e6a6 100644
--- a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
@@ -669,7 +669,6 @@ static int vlInit
 	sampler.compare_func = PIPE_FUNC_ALWAYS;
 	sampler.normalized_coords = 1;
 	/*sampler.prefilter = ;*/
-	/*sampler.shadow_ambient = ;*/
 	/*sampler.lod_bias = ;*/
 	/*sampler.min_lod = ;*/
 	/*sampler.max_lod = ;*/
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
index eb8270ecad..79318fe24b 100644
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
@@ -1074,7 +1074,6 @@ static int vlInit
 		sampler.compare_func = PIPE_FUNC_ALWAYS;
 		sampler.normalized_coords = 1;
 		/*sampler.prefilter = ;*/
-		/*sampler.shadow_ambient = ;*/
 		/*sampler.lod_bias = ;*/
 		sampler.min_lod = 0;
 		/*sampler.max_lod = ;*/
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.c b/src/gallium/state_trackers/glx/xlib/fakeglx.c
index 6df4c7d693..23777c76f6 100644
--- a/src/gallium/state_trackers/glx/xlib/fakeglx.c
+++ b/src/gallium/state_trackers/glx/xlib/fakeglx.c
@@ -97,6 +97,9 @@ struct fake_glx_context {
 
 
 
+#define DEFAULT_DIRECT GL_TRUE
+
+
 /**********************************************************************/
 /***                       GLX Visual Code                          ***/
 /**********************************************************************/
@@ -1059,7 +1062,7 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.isDirect = DEFAULT_DIRECT;
    glxCtx->glxContext.currentDpy = dpy;
    glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
 
@@ -1296,9 +1299,9 @@ Fake_glXDestroyContext( Display *dpy, GLXContext ctx )
 static Bool
 Fake_glXIsDirect( Display *dpy, GLXContext ctx )
 {
-   (void) dpy;
+   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
    (void) ctx;
-   return False;
+   return glxCtx->glxContext.isDirect;
 }
 
 
@@ -2055,7 +2058,7 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config,
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.isDirect = DEFAULT_DIRECT;
    glxCtx->glxContext.currentDpy = dpy;
    glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
 
@@ -2277,7 +2280,7 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re
       return NULL;
    }
 
-   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.isDirect = DEFAULT_DIRECT;
    glxCtx->glxContext.currentDpy = dpy;
    glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
 
diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c
index a3d1651653..79c2230588 100644
--- a/src/gallium/state_trackers/glx/xlib/xm_api.c
+++ b/src/gallium/state_trackers/glx/xlib/xm_api.c
@@ -1100,26 +1100,19 @@ XMesaContext XMesaGetCurrentContext( void )
 
 
 
-
-
-
-/*
- * Copy the back buffer to the front buffer.  If there's no back buffer
- * this is a no-op.
+/**
+ * Swap front and back color buffers and have winsys display front buffer.
+ * If there's no front color buffer no swap actually occurs.
  */
 PUBLIC
 void XMesaSwapBuffers( XMesaBuffer b )
 {
-   struct pipe_surface *surf;
+   struct pipe_surface *frontLeftSurf;
 
-   /* If we're swapping the buffer associated with the current context
-    * we have to flush any pending rendering commands first.
-    */
-   st_notify_swapbuffers(b->stfb);
+   st_swapbuffers(b->stfb, &frontLeftSurf, NULL);
 
-   st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT, &surf);
-   if (surf) {
-      driver.display_surface(b, surf);
+   if (frontLeftSurf) {
+      driver.display_surface(b, frontLeftSurf);
    }
 
    xmesa_check_and_update_buffer_size(NULL, b);
diff --git a/src/gallium/state_trackers/vega/Makefile b/src/gallium/state_trackers/vega/Makefile
new file mode 100644
index 0000000000..b8c805b06c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/Makefile
@@ -0,0 +1,128 @@
+# src/mesa/Makefile
+
+TOP = ../../../..
+include $(TOP)/configs/current
+GALLIUM = $(TOP)
+
+### Lists of source files, included by Makefiles
+
+VG_SOURCES = \
+           api_context.c   \
+           api_filters.c   \
+           api_images.c    \
+           api_masks.c     \
+           api_misc.c      \
+           api_paint.c     \
+           api_params.c    \
+           api_path.c      \
+           api_text.c      \
+           api_transform.c \
+           vgu.c        \
+           vg_context.c \
+           vg_state.c   \
+           vg_tracker.c \
+           vg_translate.c \
+           polygon.c    \
+           bezier.c     \
+           path.c       \
+           paint.c      \
+           arc.c \
+           image.c \
+           renderer.c \
+           stroker.c \
+           mask.c \
+           shader.c \
+           shaders_cache.c
+
+
+### All the core C sources
+
+ALL_SOURCES = \
+        $(VG_SOURCES)
+
+
+### Object files
+VG_OBJECTS = \
+	$(VG_SOURCES:.c=.o)
+
+### Include directories
+
+INCLUDE_DIRS = \
+	-I$(TOP)/include \
+	-I$(GALLIUM)/include \
+	-I$(GALLIUM)/src/gallium/include \
+	-I$(GALLIUM)/src/gallium/auxiliary
+
+VG_LIB = OpenVG
+VG_LIB_NAME = lib$(VG_LIB).so
+
+VG_MAJOR = 1
+VG_MINOR = 0
+VG_TINY = 0
+
+GALLIUM_LIBS = \
+	$(GALLIUM)/src/gallium/auxiliary/pipebuffer/libpipebuffer.a \
+	$(GALLIUM)/src/gallium/auxiliary/sct/libsct.a \
+	$(GALLIUM)/src/gallium/auxiliary/draw/libdraw.a \
+	$(GALLIUM)/src/gallium/auxiliary/rtasm/librtasm.a \
+	$(GALLIUM)/src/gallium/auxiliary/translate/libtranslate.a \
+	$(GALLIUM)/src/gallium/auxiliary/cso_cache/libcso_cache.a \
+	$(GALLIUM)/src/gallium/auxiliary/util/libutil.a \
+	$(GALLIUM)/src/gallium/auxiliary/tgsi/libtgsi.a
+
+.SUFFIXES : .cpp
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+default: depend subdirs $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME)
+
+# Make the OpenVG library
+$(TOP)/$(LIB_DIR)/$(VG_LIB_NAME): $(VG_OBJECTS) $(GALLIUM_LIBS)
+	$(TOP)/bin/mklib -o $(VG_LIB) \
+		-major $(VG_MAJOR) \
+		-minor $(VG_MINOR) \
+		-patch $(VG_TINY) \
+		-install $(TOP)/$(LIB_DIR) \
+		$(VG_OBJECTS) $(GALLIUM_LIBS) \
+		-Wl,--whole-archive $(LIBS) -Wl,--no-whole-archive $(SYS_LIBS)
+
+######################################################################
+# Generic stuff
+
+depend: $(ALL_SOURCES)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend  # workaround oops on gutsy?!?
+	@ touch depend
+	@ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \
+		> /dev/null 2>/dev/null
+
+
+subdirs:
+
+install: default
+	$(INSTALL) -d $(INSTALL_DIR)/include/VG
+	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
+	$(INSTALL) -m 644 $(TOP)/include/VG/*.h $(INSTALL_DIR)/include/VG
+	@if [ -e $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME) ]; then \
+		$(INSTALL) $(TOP)/$(LIB_DIR)/libOpenVG* $(INSTALL_DIR)/$(LIB_DIR); \
+	fi
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` $(TOP)/include/VG/*.h
+
+clean:
+	-rm -f *.o
+	-rm -f */*.o
+	-rm -f */*/*.o
+	-rm -f depend depend.bak
+
+include depend
diff --git a/src/gallium/state_trackers/vega/api_consts.h b/src/gallium/state_trackers/vega/api_consts.h
new file mode 100644
index 0000000000..e1b48d4a46
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_consts.h
@@ -0,0 +1,56 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef API_CONSTS_H
+#define API_CONSTS_H
+
+/*must be at least 32*/
+#define VEGA_MAX_SCISSOR_RECTS 32
+
+/*must be at least 16*/
+#define VEGA_MAX_DASH_COUNT 32
+
+/*must be at least 7*/
+#define VEGA_MAX_KERNEL_SIZE 7
+
+/*must be at least 15*/
+#define VEGA_MAX_SEPARABLE_KERNEL_SIZE 15
+
+/*must be at least 32*/
+#define VEGA_MAX_COLOR_RAMP_STOPS 256
+
+#define VEGA_MAX_IMAGE_WIDTH 2048
+
+#define VEGA_MAX_IMAGE_HEIGHT 2048
+
+#define VEGA_MAX_IMAGE_PIXELS (2048*2048)
+
+#define VEGA_MAX_IMAGE_BYTES (2048*2048 * 4)
+
+/*must be at least 128*/
+#define VEGA_MAX_GAUSSIAN_STD_DEVIATION 128
+
+#endif
diff --git a/src/gallium/state_trackers/vega/api_context.c b/src/gallium/state_trackers/vega/api_context.c
new file mode 100644
index 0000000000..47db102dd2
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_context.c
@@ -0,0 +1,75 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+
+VGErrorCode vgGetError(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGErrorCode error = VG_NO_CONTEXT_ERROR;
+
+   if (!ctx)
+      return error;
+
+   error = ctx->_error;
+   ctx->_error = VG_NO_ERROR;
+
+   return error;
+}
+
+void vgFlush(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe;
+
+   if (!ctx)
+      return;
+
+   pipe = ctx->pipe;
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+}
+
+void vgFinish(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_fence_handle *fence = NULL;
+   struct pipe_context *pipe;
+
+   if (!ctx)
+      return;
+
+   pipe = ctx->pipe;
+
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, &fence);
+
+   pipe->screen->fence_finish(pipe->screen, fence, 0);
+   pipe->screen->fence_reference(pipe->screen, &fence, NULL);
+}
diff --git a/src/gallium/state_trackers/vega/api_filters.c b/src/gallium/state_trackers/vega/api_filters.c
new file mode 100644
index 0000000000..862cbb03c4
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_filters.c
@@ -0,0 +1,805 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "image.h"
+#include "renderer.h"
+#include "shaders_cache.h"
+#include "st_inlines.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_memory.h"
+
+
+#include "asm_filters.h"
+
+
+struct filter_info {
+   struct vg_image *dst;
+   struct vg_image *src;
+   struct vg_shader * (*setup_shader)(struct vg_context *, void *);
+   void *user_data;
+   const void *const_buffer;
+   VGint const_buffer_len;
+   VGTilingMode tiling_mode;
+   struct pipe_texture *extra_texture;
+};
+
+static INLINE struct pipe_texture *create_texture_1d(struct vg_context *ctx,
+                                                     const VGuint *color_data,
+                                                     const VGint color_data_len)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *tex = 0;
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_1D;
+   templ.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   templ.last_level = 0;
+   templ.width[0] = color_data_len;
+   templ.height[0] = 1;
+   templ.depth[0] = 1;
+   pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   tex = screen->texture_create(screen, &templ);
+
+   { /* upload color_data */
+      struct pipe_transfer *transfer =
+         screen->get_tex_transfer(screen, tex,
+                                  0, 0, 0,
+                                  PIPE_TRANSFER_READ_WRITE ,
+                                  0, 0, tex->width[0], tex->height[0]);
+      void *map = screen->transfer_map(screen, transfer);
+      memcpy(map, color_data, sizeof(VGint)*color_data_len);
+      screen->transfer_unmap(screen, transfer);
+      screen->tex_transfer_destroy(transfer);
+   }
+
+   return tex;
+}
+
+static INLINE struct pipe_surface * setup_framebuffer(struct vg_image *dst)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_framebuffer_state fb;
+   struct pipe_surface *dst_surf = pipe->screen->get_tex_surface(
+      pipe->screen, dst->texture, 0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width  = dst->x + dst_surf->width;
+   fb.height = dst->y + dst_surf->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst_surf;
+   {
+      VGint i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(ctx->cso_context, &fb);
+
+   return dst_surf;
+}
+
+static void setup_viewport(struct vg_image *dst)
+{
+   struct vg_context *ctx = vg_current_context();
+   vg_set_viewport(ctx, VEGA_Y0_TOP);
+}
+
+static void setup_blend()
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_blend_state blend;
+   memset(&blend, 0, sizeof(blend));
+   blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   if (ctx->state.vg.filter_channel_mask & VG_RED)
+      blend.colormask |= PIPE_MASK_R;
+   if (ctx->state.vg.filter_channel_mask & VG_GREEN)
+      blend.colormask |= PIPE_MASK_G;
+   if (ctx->state.vg.filter_channel_mask & VG_BLUE)
+      blend.colormask |= PIPE_MASK_B;
+   if (ctx->state.vg.filter_channel_mask & VG_ALPHA)
+      blend.colormask |= PIPE_MASK_A;
+   blend.blend_enable = 1;
+   cso_set_blend(ctx->cso_context, &blend);
+}
+
+static void setup_constant_buffer(struct vg_context *ctx, const void *buffer,
+                                  VGint param_bytes)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_constant_buffer *cbuf = &ctx->filter.buffer;
+
+   /* We always need to get a new buffer, to keep the drivers simple and
+    * avoid gratuitous rendering synchronization. */
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   cbuf->buffer = pipe_buffer_create(pipe->screen, 16,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+
+   if (cbuf->buffer) {
+      st_no_flush_pipe_buffer_write(ctx, cbuf->buffer,
+                                    0, param_bytes, buffer);
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+}
+
+static void setup_samplers(struct vg_context *ctx, struct filter_info *info)
+{
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   struct pipe_sampler_state sampler[3];
+   int num_samplers = 0;
+   int num_textures = 0;
+
+   samplers[0] = NULL;
+   samplers[1] = NULL;
+   samplers[2] = NULL;
+   samplers[3] = NULL;
+   textures[0] = NULL;
+   textures[1] = NULL;
+   textures[2] = NULL;
+   textures[3] = NULL;
+
+   memset(&sampler[0], 0, sizeof(struct pipe_sampler_state));
+   sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler[0].wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler[0].min_img_filter = PIPE_TEX_MIPFILTER_LINEAR;
+   sampler[0].mag_img_filter = PIPE_TEX_MIPFILTER_LINEAR;
+   sampler[0].normalized_coords = 1;
+
+   switch(info->tiling_mode) {
+   case VG_TILE_FILL:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      memcpy(sampler[0].border_color,
+             ctx->state.vg.tile_fill_color,
+             sizeof(VGfloat) * 4);
+      break;
+   case VG_TILE_PAD:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      break;
+   case VG_TILE_REPEAT:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_REPEAT;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_REPEAT;
+      break;
+   case VG_TILE_REFLECT:
+      sampler[0].wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      sampler[0].wrap_t = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      break;
+   default:
+      debug_assert(!"Unknown tiling mode");
+   }
+
+   samplers[0] = &sampler[0];
+   textures[0] = info->src->texture;
+   ++num_samplers;
+   ++num_textures;
+
+   if (info->extra_texture) {
+      memcpy(&sampler[1], &sampler[0], sizeof(struct pipe_sampler_state));
+      samplers[1] = &sampler[1];
+      textures[1] = info->extra_texture;
+      ++num_samplers;
+      ++num_textures;
+   }
+
+
+   cso_set_samplers(ctx->cso_context, num_samplers, (const struct pipe_sampler_state **)samplers);
+   cso_set_sampler_textures(ctx->cso_context, num_textures, textures);
+}
+
+static struct vg_shader * setup_color_matrix(struct vg_context *ctx, void *user_data)
+{
+   struct vg_shader *shader =
+      shader_create_from_text(ctx->pipe, color_matrix_asm, 200,
+         PIPE_SHADER_FRAGMENT);
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+static struct vg_shader * setup_convolution(struct vg_context *ctx, void *user_data)
+{
+   char buffer[1024];
+   VGint num_consts = (VGint)(long)(user_data);
+   struct vg_shader *shader;
+
+   snprintf(buffer, 1023, convolution_asm, num_consts, num_consts / 2 + 1);
+
+   shader = shader_create_from_text(ctx->pipe, buffer, 200,
+                                    PIPE_SHADER_FRAGMENT);
+
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+static struct vg_shader * setup_lookup(struct vg_context *ctx, void *user_data)
+{
+   struct vg_shader *shader =
+      shader_create_from_text(ctx->pipe, lookup_asm,
+                              200, PIPE_SHADER_FRAGMENT);
+
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+
+static struct vg_shader * setup_lookup_single(struct vg_context *ctx, void *user_data)
+{
+   char buffer[1024];
+   VGImageChannel channel = (VGImageChannel)(user_data);
+   struct vg_shader *shader;
+
+   switch(channel) {
+   case VG_RED:
+      snprintf(buffer, 1023, lookup_single_asm, "xxxx");
+      break;
+   case VG_GREEN:
+      snprintf(buffer, 1023, lookup_single_asm, "yyyy");
+      break;
+   case VG_BLUE:
+      snprintf(buffer, 1023, lookup_single_asm, "zzzz");
+      break;
+   case VG_ALPHA:
+      snprintf(buffer, 1023, lookup_single_asm, "wwww");
+      break;
+   default:
+      debug_assert(!"Unknown color channel");
+   }
+
+   shader = shader_create_from_text(ctx->pipe, buffer, 200,
+                                    PIPE_SHADER_FRAGMENT);
+
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->driver);
+   return shader;
+}
+
+static void execute_filter(struct vg_context *ctx,
+                           struct filter_info *info)
+{
+   struct pipe_surface *dst_surf;
+   struct vg_shader *shader;
+
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+   cso_save_blend(ctx->cso_context);
+   cso_save_samplers(ctx->cso_context);
+   cso_save_sampler_textures(ctx->cso_context);
+
+   dst_surf = setup_framebuffer(info->dst);
+   setup_viewport(info->dst);
+   setup_blend();
+   setup_constant_buffer(ctx, info->const_buffer, info->const_buffer_len);
+   shader = info->setup_shader(ctx, info->user_data);
+   setup_samplers(ctx, info);
+
+   renderer_draw_texture(ctx->renderer,
+                         info->src->texture,
+                         info->dst->x, info->dst->y,
+                         info->dst->x + info->dst->width,
+                         info->dst->y + info->dst->height,
+                         info->dst->x, info->dst->y,
+                         info->dst->x + info->dst->width,
+                         info->dst->y + info->dst->height);
+
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+   cso_restore_blend(ctx->cso_context);
+   cso_restore_samplers(ctx->cso_context);
+   cso_restore_sampler_textures(ctx->cso_context);
+
+   vg_shader_destroy(ctx, shader);
+
+   pipe_surface_reference(&dst_surf, NULL);
+}
+
+void vgColorMatrix(VGImage dst, VGImage src,
+                   const VGfloat * matrix)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   struct filter_info info;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!matrix || !is_aligned(matrix)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_color_matrix;
+   info.user_data = NULL;
+   info.const_buffer = matrix;
+   info.const_buffer_len = 20 * sizeof(VGfloat);
+   info.tiling_mode = VG_TILE_PAD;
+   info.extra_texture = 0;
+   execute_filter(ctx, &info);
+}
+
+static VGfloat texture_offset(VGfloat width, VGint kernelSize, VGint current, VGint shift)
+{
+   VGfloat diff = current - shift;
+
+   return diff / width;
+}
+
+void vgConvolve(VGImage dst, VGImage src,
+                VGint kernelWidth, VGint kernelHeight,
+                VGint shiftX, VGint shiftY,
+                const VGshort * kernel,
+                VGfloat scale,
+                VGfloat bias,
+                VGTilingMode tilingMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat *buffer;
+   VGint buffer_len;
+   VGint i, j;
+   VGint idx = 0;
+   struct vg_image *d, *s;
+   VGint kernel_size = kernelWidth * kernelHeight;
+   struct filter_info info;
+   const VGint max_kernel_size = vgGeti(VG_MAX_KERNEL_SIZE);
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (kernelWidth <= 0 || kernelHeight <= 0 ||
+      kernelWidth > max_kernel_size || kernelHeight > max_kernel_size) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!kernel || !is_aligned_to(kernel, 2)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (tilingMode < VG_TILE_FILL ||
+       tilingMode > VG_TILE_REFLECT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+
+   buffer_len = 8 + 2 * 4 * kernel_size;
+   buffer = (VGfloat*)malloc(buffer_len * sizeof(VGfloat));
+
+   buffer[0] = 0.f;
+   buffer[1] = 1.f;
+   buffer[2] = 2.f; /*unused*/
+   buffer[3] = 4.f; /*unused*/
+
+   buffer[4] = kernelWidth * kernelHeight;
+   buffer[5] = scale;
+   buffer[6] = bias;
+   buffer[7] = 0.f;
+
+   idx = 8;
+   for (j = 0; j < kernelHeight; ++j) {
+      for (i = 0; i < kernelWidth; ++i) {
+         VGint index = j * kernelWidth + i;
+         VGfloat x, y;
+
+         x = texture_offset(s->width, kernelWidth, i, shiftX);
+         y = texture_offset(s->height, kernelHeight, j, shiftY);
+
+         buffer[idx + index*4 + 0] = x;
+         buffer[idx + index*4 + 1] = y;
+         buffer[idx + index*4 + 2] = 0.f;
+         buffer[idx + index*4 + 3] = 0.f;
+      }
+   }
+   idx += kernel_size * 4;
+
+   for (j = 0; j < kernelHeight; ++j) {
+      for (i = 0; i < kernelWidth; ++i) {
+         /* transpose the kernel */
+         VGint index = j * kernelWidth + i;
+         VGint kindex = (kernelWidth - i - 1) * kernelHeight + (kernelHeight - j - 1);
+         buffer[idx + index*4 + 0] = kernel[kindex];
+         buffer[idx + index*4 + 1] = kernel[kindex];
+         buffer[idx + index*4 + 2] = kernel[kindex];
+         buffer[idx + index*4 + 3] = kernel[kindex];
+      }
+   }
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_convolution;
+   info.user_data = (void*)(long)(buffer_len/4);
+   info.const_buffer = buffer;
+   info.const_buffer_len = buffer_len * sizeof(VGfloat);
+   info.tiling_mode = tilingMode;
+   info.extra_texture = 0;
+   execute_filter(ctx, &info);
+
+   free(buffer);
+}
+
+void vgSeparableConvolve(VGImage dst, VGImage src,
+                         VGint kernelWidth,
+                         VGint kernelHeight,
+                         VGint shiftX, VGint shiftY,
+                         const VGshort * kernelX,
+                         const VGshort * kernelY,
+                         VGfloat scale,
+                         VGfloat bias,
+                         VGTilingMode tilingMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGshort *kernel;
+   VGint i, j, idx = 0;
+   const VGint max_kernel_size = vgGeti(VG_MAX_SEPARABLE_KERNEL_SIZE);
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (kernelWidth <= 0 || kernelHeight <= 0 ||
+       kernelWidth > max_kernel_size || kernelHeight > max_kernel_size) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!kernelX || !kernelY ||
+       !is_aligned_to(kernelX, 2) || !is_aligned_to(kernelY, 2)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (tilingMode < VG_TILE_FILL ||
+       tilingMode > VG_TILE_REFLECT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   kernel = malloc(sizeof(VGshort)*kernelWidth*kernelHeight);
+   for (i = 0; i < kernelWidth; ++i) {
+      for (j = 0; j < kernelHeight; ++j) {
+         kernel[idx] = kernelX[i] * kernelY[j];
+         ++idx;
+      }
+   }
+   vgConvolve(dst, src, kernelWidth, kernelHeight, shiftX, shiftY,
+              kernel, scale, bias, tilingMode);
+   free(kernel);
+}
+
+static INLINE VGfloat compute_gaussian_componenet(VGfloat x, VGfloat y,
+                                                  VGfloat stdDeviationX,
+                                                  VGfloat stdDeviationY)
+{
+   VGfloat mult = 1 / ( 2 * M_PI * stdDeviationX * stdDeviationY);
+   VGfloat e = exp( - ( pow(x, 2)/(2*pow(stdDeviationX, 2)) +
+                        pow(y, 2)/(2*pow(stdDeviationY, 2)) ) );
+   return mult * e;
+}
+
+static INLINE VGint compute_kernel_size(VGfloat deviation)
+{
+   VGint size = ceil(2.146 * deviation);
+   if (size > 11)
+      return 11;
+   return size;
+}
+
+static void compute_gaussian_kernel(VGfloat *kernel,
+                                    VGint width, VGint height,
+                                    VGfloat stdDeviationX,
+                                    VGfloat stdDeviationY)
+{
+   VGint i, j;
+   VGfloat scale = 0.0f;
+
+   for (j = 0; j < height; ++j) {
+      for (i = 0; i < width; ++i) {
+         VGint idx =  (height - j -1) * width + (width - i -1);
+         kernel[idx] = compute_gaussian_componenet(i-(ceil(width/2))-1,
+                                                   j-ceil(height/2)-1,
+                                                   stdDeviationX, stdDeviationY);
+         scale += kernel[idx];
+      }
+   }
+
+   for (j = 0; j < height; ++j) {
+      for (i = 0; i < width; ++i) {
+         VGint idx = j * width + i;
+         kernel[idx] /= scale;
+      }
+   }
+}
+
+void vgGaussianBlur(VGImage dst, VGImage src,
+                    VGfloat stdDeviationX,
+                    VGfloat stdDeviationY,
+                    VGTilingMode tilingMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   VGfloat *buffer, *kernel;
+   VGint kernel_width, kernel_height, kernel_size;
+   VGint buffer_len;
+   VGint idx, i, j;
+   struct filter_info info;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (stdDeviationX <= 0 || stdDeviationY <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (tilingMode < VG_TILE_FILL ||
+       tilingMode > VG_TILE_REFLECT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   kernel_width = compute_kernel_size(stdDeviationX);
+   kernel_height = compute_kernel_size(stdDeviationY);
+   kernel_size = kernel_width * kernel_height;
+   kernel = malloc(sizeof(VGfloat)*kernel_size);
+   compute_gaussian_kernel(kernel, kernel_width, kernel_height,
+                           stdDeviationX, stdDeviationY);
+
+   buffer_len = 8 + 2 * 4 * kernel_size;
+   buffer = (VGfloat*)malloc(buffer_len * sizeof(VGfloat));
+
+   buffer[0] = 0.f;
+   buffer[1] = 1.f;
+   buffer[2] = 2.f; /*unused*/
+   buffer[3] = 4.f; /*unused*/
+
+   buffer[4] = kernel_width * kernel_height;
+   buffer[5] = 1.f;/*scale*/
+   buffer[6] = 0.f;/*bias*/
+   buffer[7] = 0.f;
+
+   idx = 8;
+   for (j = 0; j < kernel_height; ++j) {
+      for (i = 0; i < kernel_width; ++i) {
+         VGint index = j * kernel_width + i;
+         VGfloat x, y;
+
+         x = texture_offset(s->width, kernel_width, i, kernel_width/2);
+         y = texture_offset(s->height, kernel_height, j, kernel_height/2);
+
+         buffer[idx + index*4 + 0] = x;
+         buffer[idx + index*4 + 1] = y;
+         buffer[idx + index*4 + 2] = 0.f;
+         buffer[idx + index*4 + 3] = 0.f;
+      }
+   }
+   idx += kernel_size * 4;
+
+   for (j = 0; j < kernel_height; ++j) {
+      for (i = 0; i < kernel_width; ++i) {
+         /* transpose the kernel */
+         VGint index = j * kernel_width + i;
+         VGint kindex = (kernel_width - i - 1) * kernel_height + (kernel_height - j - 1);
+         buffer[idx + index*4 + 0] = kernel[kindex];
+         buffer[idx + index*4 + 1] = kernel[kindex];
+         buffer[idx + index*4 + 2] = kernel[kindex];
+         buffer[idx + index*4 + 3] = kernel[kindex];
+      }
+   }
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_convolution;
+   info.user_data = (void*)(long)(buffer_len/4);
+   info.const_buffer = buffer;
+   info.const_buffer_len = buffer_len * sizeof(VGfloat);
+   info.tiling_mode = tilingMode;
+   info.extra_texture = 0;
+   execute_filter(ctx, &info);
+
+   free(buffer);
+   free(kernel);
+}
+
+void vgLookup(VGImage dst, VGImage src,
+              const VGubyte * redLUT,
+              const VGubyte * greenLUT,
+              const VGubyte * blueLUT,
+              const VGubyte * alphaLUT,
+              VGboolean outputLinear,
+              VGboolean outputPremultiplied)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   VGuint color_data[256];
+   VGint i;
+   struct pipe_texture *lut_texture;
+   VGfloat buffer[4];
+   struct filter_info info;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!redLUT || !greenLUT || !blueLUT || !alphaLUT) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   for (i = 0; i < 256; ++i) {
+      color_data[i] = blueLUT[i] << 24 | greenLUT[i] << 16 |
+                      redLUT[i]  <<  8 | alphaLUT[i];
+   }
+   lut_texture = create_texture_1d(ctx, color_data, 255);
+
+   buffer[0] = 0.f;
+   buffer[1] = 0.f;
+   buffer[2] = 1.f;
+   buffer[3] = 1.f;
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_lookup;
+   info.user_data = NULL;
+   info.const_buffer = buffer;
+   info.const_buffer_len = 4 * sizeof(VGfloat);
+   info.tiling_mode = VG_TILE_PAD;
+   info.extra_texture = lut_texture;
+
+   execute_filter(ctx, &info);
+
+   pipe_texture_reference(&lut_texture, NULL);
+}
+
+void vgLookupSingle(VGImage dst, VGImage src,
+                    const VGuint * lookupTable,
+                    VGImageChannel sourceChannel,
+                    VGboolean outputLinear,
+                    VGboolean outputPremultiplied)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *d, *s;
+   struct pipe_texture *lut_texture;
+   VGfloat buffer[4];
+   struct filter_info info;
+   VGuint color_data[256];
+   VGint i;
+
+   if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!lookupTable || !is_aligned(lookupTable)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (sourceChannel != VG_RED && sourceChannel != VG_GREEN &&
+       sourceChannel != VG_BLUE && sourceChannel != VG_ALPHA) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   d = (struct vg_image*)dst;
+   s = (struct vg_image*)src;
+
+   if (vg_image_overlaps(d, s)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   for (i = 0; i < 256; ++i) {
+      VGuint rgba = lookupTable[i];
+      VGubyte blue, green, red, alpha;
+      red   = (rgba & 0xff000000)>>24;
+      green = (rgba & 0x00ff0000)>>16;
+      blue  = (rgba & 0x0000ff00)>> 8;
+      alpha = (rgba & 0x000000ff)>> 0;
+      color_data[i] = blue << 24 | green << 16 |
+                      red  <<  8 | alpha;
+   }
+   lut_texture = create_texture_1d(ctx, color_data, 256);
+
+   buffer[0] = 0.f;
+   buffer[1] = 0.f;
+   buffer[2] = 1.f;
+   buffer[3] = 1.f;
+
+   info.dst = d;
+   info.src = s;
+   info.setup_shader = &setup_lookup_single;
+   info.user_data = (void*)sourceChannel;
+   info.const_buffer = buffer;
+   info.const_buffer_len = 4 * sizeof(VGfloat);
+   info.tiling_mode = VG_TILE_PAD;
+   info.extra_texture = lut_texture;
+
+   execute_filter(ctx, &info);
+
+   pipe_texture_reference(&lut_texture, NULL);
+}
diff --git a/src/gallium/state_trackers/vega/api_images.c b/src/gallium/state_trackers/vega/api_images.c
new file mode 100644
index 0000000000..c437553bc2
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_images.c
@@ -0,0 +1,489 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "image.h"
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "vg_translate.h"
+#include "api_consts.h"
+#include "image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_blit.h"
+#include "util/u_tile.h"
+#include "util/u_memory.h"
+
+static INLINE VGboolean supported_image_format(VGImageFormat format)
+{
+   switch(format) {
+   case VG_sRGBX_8888:
+   case VG_sRGBA_8888:
+   case VG_sRGBA_8888_PRE:
+   case VG_sRGB_565:
+   case VG_sRGBA_5551:
+   case VG_sRGBA_4444:
+   case VG_sL_8:
+   case VG_lRGBX_8888:
+   case VG_lRGBA_8888:
+   case VG_lRGBA_8888_PRE:
+   case VG_lL_8:
+   case VG_A_8:
+   case VG_BW_1:
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1:
+   case VG_A_4:
+#endif
+   case VG_sXRGB_8888:
+   case VG_sARGB_8888:
+   case VG_sARGB_8888_PRE:
+   case VG_sARGB_1555:
+   case VG_sARGB_4444:
+   case VG_lXRGB_8888:
+   case VG_lARGB_8888:
+   case VG_lARGB_8888_PRE:
+   case VG_sBGRX_8888:
+   case VG_sBGRA_8888:
+   case VG_sBGRA_8888_PRE:
+   case VG_sBGR_565:
+   case VG_sBGRA_5551:
+   case VG_sBGRA_4444:
+   case VG_lBGRX_8888:
+   case VG_lBGRA_8888:
+   case VG_lBGRA_8888_PRE:
+   case VG_sXBGR_8888:
+   case VG_sABGR_8888:
+   case VG_sABGR_8888_PRE:
+   case VG_sABGR_1555:
+   case VG_sABGR_4444:
+   case VG_lXBGR_8888:
+   case VG_lABGR_8888:
+   case VG_lABGR_8888_PRE:
+      return VG_TRUE;
+   default:
+      return VG_FALSE;
+   }
+   return VG_FALSE;
+}
+
+VGImage vgCreateImage(VGImageFormat format,
+                      VGint width, VGint height,
+                      VGbitfield allowedQuality)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (!supported_image_format(format)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width > vgGeti(VG_MAX_IMAGE_WIDTH) ||
+       height > vgGeti(VG_MAX_IMAGE_HEIGHT)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width * height > vgGeti(VG_MAX_IMAGE_PIXELS)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   if (!(allowedQuality & ((VG_IMAGE_QUALITY_NONANTIALIASED |
+                           VG_IMAGE_QUALITY_FASTER |
+                           VG_IMAGE_QUALITY_BETTER)))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGImage)image_create(format, width, height);
+}
+
+void vgDestroyImage(VGImage image)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img = (struct vg_image *)image;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)image, VG_OBJECT_IMAGE)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   image_destroy(img);
+}
+
+void vgClearImage(VGImage image,
+                  VGint x, VGint y,
+                  VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   img = (struct vg_image*)image;
+
+   if (x + width < 0 || y + height < 0)
+      return;
+
+   image_clear(img, x, y, width, height);
+
+}
+
+void vgImageSubData(VGImage image,
+                    const void * data,
+                    VGint dataStride,
+                    VGImageFormat dataFormat,
+                    VGint x, VGint y,
+                    VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0 || !data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   img = (struct vg_image*)(image);
+   image_sub_data(img, data, dataStride, dataFormat,
+                  x, y, width, height);
+}
+
+void vgGetImageSubData(VGImage image,
+                       void * data,
+                       VGint dataStride,
+                       VGImageFormat dataFormat,
+                       VGint x, VGint y,
+                       VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0 || !data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   img = (struct vg_image*)image;
+   image_get_sub_data(img, data, dataStride, dataFormat,
+                      x, y, width, height);
+}
+
+VGImage vgChildImage(VGImage parent,
+                     VGint x, VGint y,
+                     VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *p;
+
+   if (parent == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void*)parent) ||
+       !vg_object_is_valid((void*)parent, VG_OBJECT_IMAGE)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (width <= 0 || height <= 0 || x < 0 || y < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   p = (struct vg_image *)parent;
+   if (x > p->width  || y > p->height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (x + width > p->width  || y + height > p->height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGImage)image_child_image(p, x, y, width, height);
+}
+
+VGImage vgGetParent(VGImage image)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   img = (struct vg_image*)image;
+   if (img->parent)
+      return (VGImage)img->parent;
+   else
+      return image;
+}
+
+void vgCopyImage(VGImage dst, VGint dx, VGint dy,
+                 VGImage src, VGint sx, VGint sy,
+                 VGint width, VGint height,
+                 VGboolean dither)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (src == VG_INVALID_HANDLE || dst == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   vg_validate_state(ctx);
+   image_copy((struct vg_image*)dst, dx, dy,
+              (struct vg_image*)src, sx, sy,
+              width, height, dither);
+}
+
+void vgDrawImage(VGImage image)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (!ctx)
+      return;
+
+   if (image == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+   image_draw((struct vg_image*)image);
+}
+
+void vgSetPixels(VGint dx, VGint dy,
+                 VGImage src, VGint sx, VGint sy,
+                 VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   vg_validate_state(ctx);
+
+   if (src == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   image_set_pixels(dx, dy, (struct vg_image*)src, sx, sy, width,
+                    height);
+}
+
+void vgGetPixels(VGImage dst, VGint dx, VGint dy,
+                 VGint sx, VGint sy,
+                 VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *img;
+
+   if (dst == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   img = (struct vg_image*)dst;
+
+   image_get_pixels(img, dx, dy,
+                    sx, sy, width, height);
+}
+
+void vgWritePixels(const void * data, VGint dataStride,
+                   VGImageFormat dataFormat,
+                   VGint dx, VGint dy,
+                   VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (!data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+   {
+      struct vg_image *img = image_create(dataFormat, width, height);
+      image_sub_data(img, data, dataStride, dataFormat, 0, 0,
+                     width, height);
+#if 0
+      struct matrix *matrix = &ctx->state.vg.image_user_to_surface_matrix;
+      matrix_translate(matrix, dx, dy);
+      image_draw(img);
+      matrix_translate(matrix, -dx, -dy);
+#else
+      /* this looks like a better approach */
+      image_set_pixels(dx, dy, img, 0, 0, width, height);
+#endif
+      image_destroy(img);
+   }
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+}
+
+void vgReadPixels(void * data, VGint dataStride,
+                  VGImageFormat dataFormat,
+                  VGint sx, VGint sy,
+                  VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->strb;
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint y = (fb->height - sy) - 1, yStep = -1;
+   VGint i;
+   VGubyte *dst = (VGubyte *)data;
+   VGint xoffset = 0, yoffset = 0;
+
+   if (!supported_image_format(dataFormat)) {
+      vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR);
+      return;
+   }
+   if (!data || !is_aligned(data)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   if (sx < 0) {
+      xoffset = -sx;
+      xoffset *= _vega_size_for_format(dataFormat);
+      width += sx;
+      sx = 0;
+   }
+   if (sy < 0) {
+      yoffset = -sy;
+      height += sy;
+      sy = 0;
+      y = (fb->height - sy) - 1;
+      yoffset *= dataStride;
+   }
+
+   {
+      struct pipe_transfer *transfer;
+
+      transfer = screen->get_tex_transfer(screen, strb->texture,  0, 0, 0,
+                                          PIPE_TRANSFER_READ,
+                                          0, 0, width, height);
+
+      /* Do a row at a time to flip image data vertically */
+      for (i = 0; i < height; i++) {
+#if 0
+         debug_printf("%d-%d  == %d\n", sy, height, y);
+#endif
+         pipe_get_tile_rgba(transfer, sx, y, width, 1, df);
+         y += yStep;
+         _vega_pack_rgba_span_float(ctx, width, temp, dataFormat,
+                                    dst + yoffset + xoffset);
+         dst += dataStride;
+      }
+
+      screen->tex_transfer_destroy(transfer);
+   }
+}
+
+void vgCopyPixels(VGint dx, VGint dy,
+                  VGint sx, VGint sy,
+                  VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+   struct st_renderbuffer *strb = ctx->draw_buffer->strb;
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   /* do nothing if we copy from outside the fb */
+   if (dx >= (VGint)fb->width || dy >= (VGint)fb->height ||
+       sx >= (VGint)fb->width || sy >= (VGint)fb->height)
+      return;
+
+   vg_validate_state(ctx);
+   /* make sure rendering has completed */
+   vgFinish();
+
+   vg_copy_surface(ctx, strb->surface, dx, dy,
+                   strb->surface, sx, sy, width, height);
+}
diff --git a/src/gallium/state_trackers/vega/api_masks.c b/src/gallium/state_trackers/vega/api_masks.c
new file mode 100644
index 0000000000..4f9f3dae17
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_masks.c
@@ -0,0 +1,373 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "mask.h"
+#include "renderer.h"
+
+#include "vg_context.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h" /* for winsys->update_buffer */
+
+#include "util/u_pack_color.h"
+#include "util/u_draw_quad.h"
+#include "util/u_memory.h"
+
+
+#define DISABLE_1_1_MASKING 1
+
+/**
+ * Draw a screen-aligned quadrilateral.
+ * Coords are window coords with y=0=bottom.  These coords will be transformed
+ * by the vertex shader and viewport transform.
+ */
+static void
+draw_clear_quad(struct vg_context *st,
+                float x0, float y0, float x1, float y1, float z,
+                const VGfloat color[4])
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_buffer *buf;
+   VGuint i;
+
+   /* positions */
+   st->clear.vertices[0][0][0] = x0;
+   st->clear.vertices[0][0][1] = y0;
+
+   st->clear.vertices[1][0][0] = x1;
+   st->clear.vertices[1][0][1] = y0;
+
+   st->clear.vertices[2][0][0] = x1;
+   st->clear.vertices[2][0][1] = y1;
+
+   st->clear.vertices[3][0][0] = x0;
+   st->clear.vertices[3][0][1] = y1;
+
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->clear.vertices[i][0][2] = z;
+      st->clear.vertices[i][0][3] = 1.0;
+      st->clear.vertices[i][1][0] = color[0];
+      st->clear.vertices[i][1][1] = color[1];
+      st->clear.vertices[i][1][2] = color[2];
+      st->clear.vertices[i][1][3] = color[3];
+   }
+
+
+   /* put vertex data into vbuf */
+   buf =  pipe_user_buffer_create(pipe->screen,
+                                  st->clear.vertices,
+                                  sizeof(st->clear.vertices));
+
+
+   /* draw */
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference(&buf, NULL);
+   }
+}
+
+/**
+ * Do vgClear by drawing a quadrilateral.
+ */
+static void
+clear_with_quad(struct vg_context *st, float x0, float y0,
+                float width, float height, const VGfloat clear_color[4])
+{
+   VGfloat x1, y1;
+
+   vg_validate_state(st);
+
+   x1 = x0 + width;
+   y1 = y0 + height;
+
+   /*
+     printf("%s %f,%f %f,%f\n", __FUNCTION__,
+     x0, y0,
+     x1, y1);
+   */
+
+   if (st->pipe->winsys && st->pipe->winsys->update_buffer)
+      st->pipe->winsys->update_buffer( st->pipe->winsys,
+                                       st->pipe->priv );
+
+   cso_save_blend(st->cso_context);
+   cso_save_rasterizer(st->cso_context);
+   cso_save_fragment_shader(st->cso_context);
+   cso_save_vertex_shader(st->cso_context);
+
+   /* blend state: RGBA masking */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask |= PIPE_MASK_R;
+      blend.colormask |= PIPE_MASK_G;
+      blend.colormask |= PIPE_MASK_B;
+      blend.colormask |= PIPE_MASK_A;
+      cso_set_blend(st->cso_context, &blend);
+   }
+
+   cso_set_rasterizer(st->cso_context, &st->clear.raster);
+
+   cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
+   cso_set_vertex_shader_handle(st->cso_context, vg_clear_vs(st));
+
+   /* draw quad matching scissor rect (XXX verify coord round-off) */
+   draw_clear_quad(st, x0, y0, x1, y1, 0, clear_color);
+
+   /* Restore pipe state */
+   cso_restore_blend(st->cso_context);
+   cso_restore_rasterizer(st->cso_context);
+   cso_restore_fragment_shader(st->cso_context);
+   cso_restore_vertex_shader(st->cso_context);
+}
+
+
+void vgMask(VGHandle mask, VGMaskOperation operation,
+            VGint x, VGint y,
+            VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (width <=0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (operation < VG_CLEAR_MASK || operation > VG_SUBTRACT_MASK) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+
+   vg_validate_state(ctx);
+
+   if (operation == VG_CLEAR_MASK) {
+      mask_fill(x, y, width, height, 0.f);
+   } else if (operation == VG_FILL_MASK) {
+      mask_fill(x, y, width, height, 1.f);
+   } else if (vg_object_is_valid((void*)mask, VG_OBJECT_IMAGE)) {
+      struct vg_image *image = (struct vg_image *)mask;
+      mask_using_image(image, operation, x, y, width, height);
+   } else if (vg_object_is_valid((void*)mask, VG_OBJECT_MASK)) {
+#if DISABLE_1_1_MASKING
+      return;
+#else
+      struct vg_mask_layer *layer = (struct vg_mask_layer *)mask;
+      mask_using_layer(layer, operation, x, y, width, height);
+#endif
+   } else {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+   }
+}
+
+void vgClear(VGint x, VGint y,
+             VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state *fb;
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   vg_validate_state(ctx);
+#if 0
+   debug_printf("Clear [%d, %d, %d, %d] with [%f, %f, %f, %f]\n",
+                x, y, width, height,
+                ctx->state.vg.clear_color[0],
+                ctx->state.vg.clear_color[1],
+                ctx->state.vg.clear_color[2],
+                ctx->state.vg.clear_color[3]);
+#endif
+
+   fb = &ctx->state.g3d.fb;
+   /* check for a whole surface clear */
+   if (!ctx->state.vg.scissoring &&
+       (x == 0 && y == 0 && width == fb->width && height == fb->height)) {
+      ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_COLOR | PIPE_CLEAR_DEPTHSTENCIL,
+                       ctx->state.vg.clear_color, 1., 0);
+   } else {
+      clear_with_quad(ctx, x, y, width, height, ctx->state.vg.clear_color);
+   }
+}
+
+
+#ifdef OPENVG_VERSION_1_1
+
+
+void vgRenderToMask(VGPath path,
+                    VGbitfield paintModes,
+                    VGMaskOperation operation)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!paintModes || (paintModes&(~(VG_STROKE_PATH|VG_FILL_PATH)))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (operation < VG_CLEAR_MASK ||
+       operation > VG_SUBTRACT_MASK) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)path, VG_OBJECT_PATH)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+#if DISABLE_1_1_MASKING
+   return;
+#endif
+
+   vg_validate_state(ctx);
+
+   mask_render_to((struct path *)path, paintModes, operation);
+}
+
+VGMaskLayer vgCreateMaskLayer(VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (width <= 0 || height <= 0 ||
+       width > vgGeti(VG_MAX_IMAGE_WIDTH) ||
+       height > vgGeti(VG_MAX_IMAGE_HEIGHT)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGMaskLayer)mask_layer_create(width, height);
+}
+
+void vgDestroyMaskLayer(VGMaskLayer maskLayer)
+{
+   struct vg_mask_layer *mask = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (maskLayer == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   mask = (struct vg_mask_layer *)maskLayer;
+   mask_layer_destroy(mask);
+}
+
+void vgFillMaskLayer(VGMaskLayer maskLayer,
+                     VGint x, VGint y,
+                     VGint width, VGint height,
+                     VGfloat value)
+{
+   struct vg_mask_layer *mask = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (maskLayer == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (value < 0 || value > 1) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (x < 0 || y < 0 || (x + width) < 0 || (y + height) < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   mask = (struct vg_mask_layer*)maskLayer;
+
+   if (x + width > mask_layer_width(mask) ||
+       y + height > mask_layer_height(mask)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+#if DISABLE_1_1_MASKING
+   return;
+#endif
+   mask_layer_fill(mask, x, y, width, height, value);
+}
+
+void vgCopyMask(VGMaskLayer maskLayer,
+                VGint sx, VGint sy,
+                VGint dx, VGint dy,
+                VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_mask_layer *mask = 0;
+
+   if (maskLayer == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+#if DISABLE_1_1_MASKING
+   return;
+#endif
+
+   mask = (struct vg_mask_layer*)maskLayer;
+   mask_copy(mask, sx, sy, dx, dy, width, height);
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/api_misc.c b/src/gallium/state_trackers/vega/api_misc.c
new file mode 100644
index 0000000000..78ba0bc110
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_misc.c
@@ -0,0 +1,83 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+/* Hardware Queries */
+VGHardwareQueryResult vgHardwareQuery(VGHardwareQueryType key,
+                                      VGint setting)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (key < VG_IMAGE_FORMAT_QUERY ||
+       key > VG_PATH_DATATYPE_QUERY) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_HARDWARE_UNACCELERATED;
+   }
+
+   if (key == VG_IMAGE_FORMAT_QUERY) {
+      if (setting < VG_sRGBX_8888 ||
+          setting > VG_lABGR_8888_PRE) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return VG_HARDWARE_UNACCELERATED;
+      }
+   } else if (key == VG_PATH_DATATYPE_QUERY) {
+      if (setting < VG_PATH_DATATYPE_S_8 ||
+          setting > VG_PATH_DATATYPE_F) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return VG_HARDWARE_UNACCELERATED;
+      }
+   }
+   /* we're supposed to accelerate everything */
+   return VG_HARDWARE_ACCELERATED;
+}
+
+/* Renderer and Extension Information */
+const VGubyte *vgGetString(VGStringID name)
+{
+   struct vg_context *ctx = vg_current_context();
+   static const VGubyte *vendor = (VGubyte *)"Tungsten Graphics, Inc";
+   static const VGubyte *renderer = (VGubyte *)"Vega OpenVG 1.0";
+   static const VGubyte *version = (VGubyte *)"1.0";
+
+   if (!ctx)
+      return NULL;
+
+   switch(name) {
+   case VG_VENDOR:
+      return vendor;
+   case VG_RENDERER:
+      return renderer;
+   case VG_VERSION:
+      return version;
+   case VG_EXTENSIONS:
+      return NULL;
+   default:
+      return NULL;
+   }
+}
diff --git a/src/gallium/state_trackers/vega/api_paint.c b/src/gallium/state_trackers/vega/api_paint.c
new file mode 100644
index 0000000000..dd3ac5bdb0
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_paint.c
@@ -0,0 +1,166 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "paint.h"
+#include "image.h"
+
+VGPaint vgCreatePaint(void)
+{
+   return (VGPaint) paint_create(vg_current_context());
+}
+
+void vgDestroyPaint(VGPaint p)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_paint *paint;
+
+   if (p == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   paint = (struct vg_paint *)p;
+   paint_destroy(paint);
+}
+
+void vgSetPaint(VGPaint paint, VGbitfield paintModes)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (paint == VG_INVALID_HANDLE) {
+      /* restore the default */
+      paint = (VGPaint)ctx->default_paint;
+   } else if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!(paintModes & ((VG_FILL_PATH|VG_STROKE_PATH)))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (paintModes & VG_FILL_PATH) {
+      ctx->state.vg.fill_paint = (struct vg_paint *)paint;
+   }
+   if (paintModes & VG_STROKE_PATH) {
+      ctx->state.vg.stroke_paint = (struct vg_paint *)paint;
+   }
+}
+
+VGPaint vgGetPaint(VGPaintMode paintMode)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGPaint paint = VG_INVALID_HANDLE;
+
+   if (paintMode < VG_STROKE_PATH || paintMode > VG_FILL_PATH) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   if (paintMode == VG_FILL_PATH)
+      paint = (VGPaint)ctx->state.vg.fill_paint;
+   else if (paintMode == VG_STROKE_PATH)
+      paint = (VGPaint)ctx->state.vg.stroke_paint;
+
+   if (paint == (VGPaint)ctx->default_paint)
+      paint = VG_INVALID_HANDLE;
+
+   return paint;
+}
+
+void vgSetColor(VGPaint paint, VGuint rgba)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (paint == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   {
+      struct vg_paint *p = (struct vg_paint *)paint;
+      paint_set_colori(p, rgba);
+   }
+}
+
+VGuint vgGetColor(VGPaint paint)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_paint *p;
+   VGuint rgba = 0;
+
+   if (paint == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return rgba;
+   }
+
+   if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return rgba;
+   }
+   p = (struct vg_paint *)paint;
+
+   return paint_colori(p);
+}
+
+void vgPaintPattern(VGPaint paint, VGImage pattern)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (paint == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_PAINT, (void *)paint)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (pattern == VG_INVALID_HANDLE) {
+      paint_set_type((struct vg_paint*)paint, VG_PAINT_TYPE_COLOR);
+      return;
+   }
+
+   if (!vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void *)pattern)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+
+   if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT) ||
+       !vg_object_is_valid((void*)pattern, VG_OBJECT_IMAGE)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   paint_set_pattern((struct vg_paint*)paint,
+                     (struct vg_image*)pattern);
+}
+
diff --git a/src/gallium/state_trackers/vega/api_params.c b/src/gallium/state_trackers/vega/api_params.c
new file mode 100644
index 0000000000..db77fd9cb0
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_params.c
@@ -0,0 +1,1673 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "paint.h"
+#include "path.h"
+#include "image.h"
+#include "matrix.h"
+#include "api_consts.h"
+
+#include "pipe/p_compiler.h"
+#include "util/u_pointer.h"
+#include "util/u_math.h"
+
+#include <math.h>
+
+static INLINE struct vg_state *current_state()
+{
+   struct vg_context *ctx = vg_current_context();
+   if (!ctx)
+      return 0;
+   else
+      return &ctx->state.vg;
+}
+
+static INLINE VGboolean count_in_bounds(VGParamType type, VGint count)
+{
+   if (count < 0)
+      return VG_FALSE;
+
+   if (type == VG_SCISSOR_RECTS)
+      return (!(count % 4) && (count >= 0 || count <= VEGA_MAX_SCISSOR_RECTS * 4));
+   else if (type == VG_STROKE_DASH_PATTERN) {
+      return count <= VEGA_MAX_DASH_COUNT;
+   } else {
+      VGint real_count = vgGetVectorSize(type);
+      return count == real_count;
+   }
+}
+
+void vgSetf (VGParamType type, VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      vgSeti(type, floor(value));
+      return;
+      break;
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = value;
+      state->stroke.line_width.i = float_to_int_floor(*((VGuint*)(&value)));
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = value;
+      state->stroke.miter_limit.i = float_to_int_floor(*((VGuint*)(&value)));
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = value;
+      state->stroke.dash_phase.i = float_to_int_floor(*((VGuint*)(&value)));
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+}
+
+void vgSeti (VGParamType type, VGint value)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+      if (value < VG_MATRIX_PATH_USER_TO_SURFACE ||
+#ifdef OPENVG_VERSION_1_1
+          value > VG_MATRIX_GLYPH_USER_TO_SURFACE)
+#else
+          value > VG_MATRIX_STROKE_PAINT_TO_USER)
+#endif
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->matrix_mode = value;
+      break;
+   case VG_FILL_RULE:
+      if (value < VG_EVEN_ODD ||
+          value > VG_NON_ZERO)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->fill_rule = value;
+      break;
+   case VG_IMAGE_QUALITY:
+      state->image_quality = value;
+      break;
+   case VG_RENDERING_QUALITY:
+      if (value < VG_RENDERING_QUALITY_NONANTIALIASED ||
+          value > VG_RENDERING_QUALITY_BETTER)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->rendering_quality = value;
+      break;
+   case VG_BLEND_MODE:
+      if (value < VG_BLEND_SRC ||
+          value > VG_BLEND_ADDITIVE)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else {
+         ctx->state.dirty |= BLEND_DIRTY;
+         state->blend_mode = value;
+      }
+      break;
+   case VG_IMAGE_MODE:
+      if (value < VG_DRAW_IMAGE_NORMAL ||
+          value > VG_DRAW_IMAGE_STENCIL)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->image_mode = value;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+      state->color_transform = value;
+#endif
+      break;
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = value;
+      state->stroke.line_width.i = value;
+      break;
+   case VG_STROKE_CAP_STYLE:
+      if (value < VG_CAP_BUTT ||
+          value > VG_CAP_SQUARE)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->stroke.cap_style = value;
+      break;
+   case VG_STROKE_JOIN_STYLE:
+      if (value < VG_JOIN_MITER ||
+          value > VG_JOIN_BEVEL)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->stroke.join_style = value;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = value;
+      state->stroke.miter_limit.i = value;
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = value;
+      state->stroke.dash_phase.i = value;
+      break;
+   case VG_STROKE_DASH_PHASE_RESET:
+      state->stroke.dash_phase_reset = value;
+      break;
+   case VG_MASKING:
+      state->masking = value;
+      break;
+   case VG_SCISSORING:
+      state->scissoring = value;
+      ctx->state.dirty |= DEPTH_STENCIL_DIRTY;
+      break;
+   case VG_PIXEL_LAYOUT:
+      if (value < VG_PIXEL_LAYOUT_UNKNOWN ||
+          value > VG_PIXEL_LAYOUT_BGR_HORIZONTAL)
+         error = VG_ILLEGAL_ARGUMENT_ERROR;
+      else
+         state->pixel_layout = value;
+      break;
+   case VG_SCREEN_LAYOUT:
+      /* read only ignore */
+      break;
+   case VG_FILTER_FORMAT_LINEAR:
+      state->filter_format_linear = value;
+      break;
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+      state->filter_format_premultiplied = value;
+      break;
+   case VG_FILTER_CHANNEL_MASK:
+      state->filter_channel_mask = value;
+      break;
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      /* read only ignore */
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+}
+
+void vgSetfv(VGParamType type, VGint count,
+             const VGfloat * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+
+   if ((count && !values) || !count_in_bounds(type, count) || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+      vgSeti(type, floor(values[0]));
+      return;
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      VGuint *x = (VGuint*)values;
+      for (i = 0; i < count; ++i) {
+         state->scissor_rects[i].f =  values[i];
+         state->scissor_rects[i].i =  float_to_int_floor(x[i]);
+      }
+      state->scissor_rects_num = count / 4;
+      ctx->state.dirty |= DEPTH_STENCIL_DIRTY;
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         state->color_transform_values[i] =  values[i];
+      }
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = values[0];
+      state->stroke.line_width.i = float_to_int_floor(*((VGuint*)(values)));
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = values[0];
+      state->stroke.miter_limit.i = float_to_int_floor(*((VGuint*)(values)));
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      int i;
+      for (i = 0; i < count; ++i) {
+         state->stroke.dash_pattern[i].f = values[i];
+         state->stroke.dash_pattern[i].i =
+            float_to_int_floor(*((VGuint*)(values + i)));
+      }
+      state->stroke.dash_pattern_num = count;
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = values[0];
+      state->stroke.dash_phase.i = float_to_int_floor(*((VGuint*)(values)));
+      break;
+   case VG_TILE_FILL_COLOR:
+      state->tile_fill_color[0] = values[0];
+      state->tile_fill_color[1] = values[1];
+      state->tile_fill_color[2] = values[2];
+      state->tile_fill_color[3] = values[3];
+
+      state->tile_fill_colori[0] = float_to_int_floor(*((VGuint*)(values + 0)));
+      state->tile_fill_colori[1] = float_to_int_floor(*((VGuint*)(values + 1)));
+      state->tile_fill_colori[2] = float_to_int_floor(*((VGuint*)(values + 2)));
+      state->tile_fill_colori[3] = float_to_int_floor(*((VGuint*)(values + 3)));
+      break;
+   case VG_CLEAR_COLOR:
+      state->clear_color[0] = values[0];
+      state->clear_color[1] = values[1];
+      state->clear_color[2] = values[2];
+      state->clear_color[3] = values[3];
+
+      state->clear_colori[0] = float_to_int_floor(*((VGuint*)(values + 0)));
+      state->clear_colori[1] = float_to_int_floor(*((VGuint*)(values + 1)));
+      state->clear_colori[2] = float_to_int_floor(*((VGuint*)(values + 2)));
+      state->clear_colori[3] = float_to_int_floor(*((VGuint*)(values + 3)));
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      state->glyph_origin[0].f = values[0];
+      state->glyph_origin[1].f = values[1];
+
+      state->glyph_origin[0].i = float_to_int_floor(*((VGuint*)(values + 0)));
+      state->glyph_origin[1].i = float_to_int_floor(*((VGuint*)(values + 1)));
+      break;
+#endif
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+}
+
+void vgSetiv(VGParamType type, VGint count,
+             const VGint * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_state *state = current_state();
+
+   if ((count && !values) || !count_in_bounds(type, count) || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+      vgSeti(type, values[0]);
+      return;
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         state->scissor_rects[i].i =  values[i];
+         state->scissor_rects[i].f =  values[i];
+      }
+      state->scissor_rects_num = count / 4;
+      ctx->state.dirty |= DEPTH_STENCIL_DIRTY;
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         state->color_transform_values[i] =  values[i];
+      }
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      state->stroke.line_width.f = values[0];
+      state->stroke.line_width.i = values[0];
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      state->stroke.miter_limit.f = values[0];
+      state->stroke.miter_limit.i = values[0];
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      int i;
+      for (i = 0; i < count; ++i) {
+         state->stroke.dash_pattern[i].f = values[i];
+         state->stroke.dash_pattern[i].i = values[i];
+      }
+      state->stroke.dash_pattern_num = count;
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      state->stroke.dash_phase.f = values[0];
+      state->stroke.dash_phase.i = values[0];
+      break;
+   case VG_TILE_FILL_COLOR:
+      state->tile_fill_color[0] = values[0];
+      state->tile_fill_color[1] = values[1];
+      state->tile_fill_color[2] = values[2];
+      state->tile_fill_color[3] = values[3];
+
+      state->tile_fill_colori[0] = values[0];
+      state->tile_fill_colori[1] = values[1];
+      state->tile_fill_colori[2] = values[2];
+      state->tile_fill_colori[3] = values[3];
+      break;
+   case VG_CLEAR_COLOR:
+      state->clear_color[0] = values[0];
+      state->clear_color[1] = values[1];
+      state->clear_color[2] = values[2];
+      state->clear_color[3] = values[3];
+
+      state->clear_colori[0] = values[0];
+      state->clear_colori[1] = values[1];
+      state->clear_colori[2] = values[2];
+      state->clear_colori[3] = values[3];
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      state->glyph_origin[0].f = values[0];
+      state->glyph_origin[1].f = values[1];
+      state->glyph_origin[0].i = values[0];
+      state->glyph_origin[1].i = values[1];
+      break;
+#endif
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+   case VG_MAX_FLOAT:
+      break;
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+VGfloat vgGetf(VGParamType type)
+{
+   struct vg_context *ctx = vg_current_context();
+   const struct vg_state *state = current_state();
+   VGErrorCode error = VG_NO_ERROR;
+   VGfloat value = 0.0f;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+      return vgGeti(type);
+      break;
+   case VG_STROKE_LINE_WIDTH:
+      value = state->stroke.line_width.f;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      value = state->stroke.miter_limit.f;
+      break;
+   case VG_STROKE_DASH_PHASE:
+      value = state->stroke.dash_phase.f;
+      break;
+
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      return vgGeti(type);
+      break;
+   case VG_MAX_FLOAT:
+      value = 1e+10;/*must be at least 1e+10*/
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+   return value;
+}
+
+VGint vgGeti(VGParamType type)
+{
+   const struct vg_state *state = current_state();
+   struct vg_context *ctx = vg_current_context();
+   VGErrorCode error = VG_NO_ERROR;
+   VGint value = 0;
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+      value = state->matrix_mode;
+      break;
+   case VG_FILL_RULE:
+      value = state->fill_rule;
+      break;
+   case VG_IMAGE_QUALITY:
+      value = state->image_quality;
+      break;
+   case VG_RENDERING_QUALITY:
+      value = state->rendering_quality;
+      break;
+   case VG_BLEND_MODE:
+      value = state->blend_mode;
+      break;
+   case VG_IMAGE_MODE:
+      value = state->image_mode;
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+      value = state->color_transform;
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      value = state->stroke.line_width.i;
+      break;
+   case VG_STROKE_CAP_STYLE:
+      value = state->stroke.cap_style;
+      break;
+   case VG_STROKE_JOIN_STYLE:
+      value = state->stroke.join_style;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      value = state->stroke.miter_limit.i;
+      break;
+   case VG_STROKE_DASH_PHASE:
+      value = state->stroke.dash_phase.i;
+      break;
+   case VG_STROKE_DASH_PHASE_RESET:
+      value = state->stroke.dash_phase_reset;
+      break;
+   case VG_MASKING:
+      value = state->masking;
+      break;
+   case VG_SCISSORING:
+      value = state->scissoring;
+      break;
+   case VG_PIXEL_LAYOUT:
+      value = state->pixel_layout;
+      break;
+   case VG_SCREEN_LAYOUT:
+      value = state->screen_layout;
+      break;
+   case VG_FILTER_FORMAT_LINEAR:
+      value = state->filter_format_linear;
+      break;
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+      value = state->filter_format_premultiplied;
+      break;
+   case VG_FILTER_CHANNEL_MASK:
+      value = state->filter_channel_mask;
+      break;
+
+   case VG_MAX_SCISSOR_RECTS:
+      value = 32; /*must be at least 32*/
+      break;
+   case VG_MAX_DASH_COUNT:
+      value = 16; /*must be at least 16*/
+      break;
+   case VG_MAX_KERNEL_SIZE:
+      value = 7; /*must be at least 7*/
+      break;
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+      value = 15; /*must be at least 15*/
+      break;
+   case VG_MAX_COLOR_RAMP_STOPS:
+      value = 256; /*must be at least 32*/
+      break;
+   case VG_MAX_IMAGE_WIDTH:
+      value = 2048;
+      break;
+   case VG_MAX_IMAGE_HEIGHT:
+      value = 2048;
+      break;
+   case VG_MAX_IMAGE_PIXELS:
+      value = 2048*2048;
+      break;
+   case VG_MAX_IMAGE_BYTES:
+      value = 2048*2048 * 4;
+      break;
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      value = 128; /*must be at least 128*/
+      break;
+
+   case VG_MAX_FLOAT: {
+      VGfloat val = vgGetf(type);
+      value = float_to_int_floor(*((VGuint*)&val));
+   }
+      break;
+   default:
+      error = VG_ILLEGAL_ARGUMENT_ERROR;
+      break;
+   }
+   vg_set_error(ctx, error);
+   return value;
+}
+
+VGint vgGetVectorSize(VGParamType type)
+{
+   struct vg_context *ctx = vg_current_context();
+   const struct vg_state *state = current_state();
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+      return 1;
+   case VG_SCISSOR_RECTS:
+      return state->scissor_rects_num * 4;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+      return 1;
+   case VG_COLOR_TRANSFORM_VALUES:
+      return 8;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_MITER_LIMIT:
+      return 1;
+   case VG_STROKE_DASH_PATTERN:
+      return state->stroke.dash_pattern_num;
+   case VG_STROKE_DASH_PHASE:
+      return 1;
+   case VG_STROKE_DASH_PHASE_RESET:
+      return 1;
+   case VG_TILE_FILL_COLOR:
+      return 4;
+   case VG_CLEAR_COLOR:
+      return 4;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      return 2;
+#endif
+   case VG_MASKING:
+      return 1;
+   case VG_SCISSORING:
+      return 1;
+   case VG_PIXEL_LAYOUT:
+      return 1;
+   case VG_SCREEN_LAYOUT:
+      return 1;
+   case VG_FILTER_FORMAT_LINEAR:
+      return 1;
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+      return 1;
+   case VG_FILTER_CHANNEL_MASK:
+      return 1;
+
+   case VG_MAX_COLOR_RAMP_STOPS:
+      return 1;
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_FLOAT:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      return 1;
+   default:
+      if (ctx)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return 0;
+   }
+}
+
+void vgGetfv(VGParamType type, VGint count,
+             VGfloat * values)
+{
+   const struct vg_state *state = current_state();
+   struct vg_context *ctx = vg_current_context();
+   VGint real_count = vgGetVectorSize(type);
+
+   if (!values || count <= 0 || count > real_count || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      values[0] = vgGeti(type);
+      break;
+   case VG_MAX_FLOAT:
+      values[0] = vgGetf(type);
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->scissor_rects[i].f;
+      }
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      memcpy(values, state->color_transform_values,
+             sizeof(VGfloat) * count);
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      values[0] = state->stroke.line_width.f;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      values[0] = state->stroke.miter_limit.f;
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->stroke.dash_pattern[i].f;
+      }
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      values[0] = state->stroke.dash_phase.f;
+      break;
+   case VG_TILE_FILL_COLOR:
+      values[0] = state->tile_fill_color[0];
+      values[1] = state->tile_fill_color[1];
+      values[2] = state->tile_fill_color[2];
+      values[3] = state->tile_fill_color[3];
+      break;
+   case VG_CLEAR_COLOR:
+      values[0] = state->clear_color[0];
+      values[1] = state->clear_color[1];
+      values[2] = state->clear_color[2];
+      values[3] = state->clear_color[3];
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      values[0] = state->glyph_origin[0].f;
+      values[1] = state->glyph_origin[1].f;
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgGetiv(VGParamType type, VGint count,
+             VGint * values)
+{
+   const struct vg_state *state = current_state();
+   struct vg_context *ctx = vg_current_context();
+   VGint real_count = vgGetVectorSize(type);
+
+   if (!values || count <= 0 || count > real_count || !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(type) {
+   case VG_MATRIX_MODE:
+   case VG_FILL_RULE:
+   case VG_IMAGE_QUALITY:
+   case VG_RENDERING_QUALITY:
+   case VG_BLEND_MODE:
+   case VG_IMAGE_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM:
+#endif
+   case VG_STROKE_CAP_STYLE:
+   case VG_STROKE_JOIN_STYLE:
+   case VG_STROKE_DASH_PHASE_RESET:
+   case VG_MASKING:
+   case VG_SCISSORING:
+   case VG_PIXEL_LAYOUT:
+   case VG_SCREEN_LAYOUT:
+   case VG_FILTER_FORMAT_LINEAR:
+   case VG_FILTER_FORMAT_PREMULTIPLIED:
+   case VG_FILTER_CHANNEL_MASK:
+   case VG_MAX_SCISSOR_RECTS:
+   case VG_MAX_DASH_COUNT:
+   case VG_MAX_KERNEL_SIZE:
+   case VG_MAX_SEPARABLE_KERNEL_SIZE:
+   case VG_MAX_COLOR_RAMP_STOPS:
+   case VG_MAX_IMAGE_WIDTH:
+   case VG_MAX_IMAGE_HEIGHT:
+   case VG_MAX_IMAGE_PIXELS:
+   case VG_MAX_IMAGE_BYTES:
+   case VG_MAX_GAUSSIAN_STD_DEVIATION:
+      values[0] = vgGeti(type);
+      break;
+   case VG_MAX_FLOAT: {
+      VGfloat val = vgGetf(type);
+      values[0] = float_to_int_floor(*((VGuint*)&val));
+   }
+      break;
+   case VG_SCISSOR_RECTS: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->scissor_rects[i].i;
+      }
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_COLOR_TRANSFORM_VALUES: {
+      VGint i;
+      VGuint *x = (VGuint*)state->color_transform_values;
+      for (i = 0; i < count; ++i) {
+         values[i] = float_to_int_floor(x[i]);
+      }
+   }
+      break;
+#endif
+   case VG_STROKE_LINE_WIDTH:
+      values[0] = state->stroke.line_width.i;
+      break;
+   case VG_STROKE_MITER_LIMIT:
+      values[0] = state->stroke.miter_limit.i;
+      break;
+   case VG_STROKE_DASH_PATTERN: {
+      VGint i;
+      for (i = 0; i < count; ++i) {
+         values[i] = state->stroke.dash_pattern[i].i;
+      }
+   }
+      break;
+   case VG_STROKE_DASH_PHASE:
+      values[0] = state->stroke.dash_phase.i;
+      break;
+   case VG_TILE_FILL_COLOR:
+      values[0] = state->tile_fill_colori[0];
+      values[1] = state->tile_fill_colori[1];
+      values[2] = state->tile_fill_colori[2];
+      values[3] = state->tile_fill_colori[3];
+      break;
+   case VG_CLEAR_COLOR:
+      values[0] = state->clear_colori[0];
+      values[1] = state->clear_colori[1];
+      values[2] = state->clear_colori[2];
+      values[3] = state->clear_colori[3];
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_GLYPH_ORIGIN:
+      values[0] = state->glyph_origin[0].i;
+      values[1] = state->glyph_origin[1].i;
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgSetParameterf(VGHandle object,
+                     VGint paramType,
+                     VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      vgSetParameteri(object, paramType, floor(value));
+      return;
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      /* it's an error if paramType refers to a vector parameter */
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      paint_set_color_ramp_premultiplied(p, value);
+   }
+      break;
+
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      /* read only don't produce an error */
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgSetParameteri(VGHandle object,
+                     VGint paramType,
+                     VGint value)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+      if (value < VG_PAINT_TYPE_COLOR ||
+          value > VG_PAINT_TYPE_PATTERN)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         paint_set_type(paint, value);
+      }
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      /* it's an error if paramType refers to a vector parameter */
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+      if (value < VG_COLOR_RAMP_SPREAD_PAD ||
+          value > VG_COLOR_RAMP_SPREAD_REFLECT)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         paint_set_spread_mode(paint, value);
+      }
+      break;
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      paint_set_color_ramp_premultiplied(p, value);
+   }
+      break;
+   case VG_PAINT_PATTERN_TILING_MODE:
+      if (value < VG_TILE_FILL ||
+          value > VG_TILE_REFLECT)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         paint_set_pattern_tiling(paint, value);
+      }
+      break;
+
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      /* read only don't produce an error */
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+void vgSetParameterfv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      const VGfloat * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (count < 0 || count < real_count ||
+       (values == NULL && count != 0) ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      if (count != 1)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else
+         vgSetParameterf(object, paramType, values[0]);
+      return;
+      break;
+   case VG_PAINT_COLOR: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_color(paint, values);
+      }
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      if (count && count < 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         count = MIN2(count, VEGA_MAX_COLOR_RAMP_STOPS);
+         paint_set_ramp_stops(paint, values, count);
+         {
+            VGint stopsi[VEGA_MAX_COLOR_RAMP_STOPS];
+            int i = 0;
+            for (i = 0; i < count; ++i) {
+               stopsi[i] = float_to_int_floor(*((VGuint*)(values + i)));
+            }
+            paint_set_ramp_stopsi(paint, stopsi, count);
+         }
+      }
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_linear_gradient(paint, values);
+         {
+            VGint vals[4];
+            vals[0] = FLT_TO_INT(values[0]);
+            vals[1] = FLT_TO_INT(values[1]);
+            vals[2] = FLT_TO_INT(values[2]);
+            vals[3] = FLT_TO_INT(values[3]);
+            paint_set_linear_gradienti(paint, vals);
+         }
+      }
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      if (count != 5)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_radial_gradient(paint, values);
+         {
+            VGint vals[5];
+            vals[0] = FLT_TO_INT(values[0]);
+            vals[1] = FLT_TO_INT(values[1]);
+            vals[2] = FLT_TO_INT(values[2]);
+            vals[3] = FLT_TO_INT(values[3]);
+            vals[4] = FLT_TO_INT(values[4]);
+            paint_set_radial_gradienti(paint, vals);
+         }
+      }
+   }
+      break;
+
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      /* read only don't produce an error */
+      break;
+#endif
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+void vgSetParameteriv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      const VGint * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (count < 0 || count < real_count ||
+       (values == NULL && count != 0) ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      if (count != 1)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else
+         vgSetParameteri(object, paramType, values[0]);
+      return;
+      break;
+   case VG_PAINT_COLOR: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         struct vg_paint *paint = (struct vg_paint *)object;
+         paint_set_coloriv(paint, values);
+      }
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      if ((count % 5))
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         VGfloat *vals = 0;
+         int i;
+         struct vg_paint *paint = (struct vg_paint *)object;
+         if (count) {
+            vals = malloc(sizeof(VGfloat)*count);
+            for (i = 0; i < count; ++i)
+               vals[i] = values[i];
+         }
+
+         paint_set_ramp_stopsi(paint, values, count);
+         paint_set_ramp_stops(paint, vals, count);
+         free(vals);
+      }
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      if (count != 4)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         VGfloat vals[4];
+         struct vg_paint *paint = (struct vg_paint *)object;
+         vals[0] = values[0];
+         vals[1] = values[1];
+         vals[2] = values[2];
+         vals[3] = values[3];
+         paint_set_linear_gradient(paint, vals);
+         paint_set_linear_gradienti(paint, values);
+      }
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      if (count != 5)
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      else {
+         VGfloat vals[5];
+         struct vg_paint *paint = (struct vg_paint *)object;
+         vals[0] = values[0];
+         vals[1] = values[1];
+         vals[2] = values[2];
+         vals[3] = values[3];
+         vals[4] = values[4];
+         paint_set_radial_gradient(paint, vals);
+         paint_set_radial_gradienti(paint, values);
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE:
+   case VG_PATH_FORMAT:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      /* read only don't produce an error */
+      break;
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+VGint vgGetParameterVectorSize(VGHandle object,
+                               VGint paramType)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      return 1;
+   case VG_PAINT_COLOR:
+      return 4;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_num_ramp_stops(p);
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT:
+      return 4;
+   case VG_PAINT_RADIAL_GRADIENT:
+      return 5;
+
+
+   case VG_PATH_FORMAT:
+   case VG_PATH_DATATYPE:
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      return 1;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+      return 1;
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      return 1;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+   return 0;
+}
+
+
+VGfloat vgGetParameterf(VGHandle object,
+                        VGint paramType)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+      return vgGetParameteri(object, paramType);
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+
+   case VG_PATH_FORMAT:
+      return VG_PATH_FORMAT_STANDARD;
+   case VG_PATH_SCALE: {
+      struct path *p = (struct path*)object;
+      return path_scale(p);
+   }
+   case VG_PATH_BIAS: {
+      struct path *p = (struct path*)object;
+      return path_bias(p);
+   }
+   case VG_PATH_DATATYPE:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      return vgGetParameteri(object, paramType);
+      break;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS: 
+      return vgGetParameteri(object, paramType);
+      break;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+   return 0;
+}
+
+VGint vgGetParameteri(VGHandle object,
+                      VGint paramType)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE: {
+         struct vg_paint *paint = (struct vg_paint *)ptr;
+         return paint_type(paint);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_spread_mode(p);
+   }
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_color_ramp_premultiplied(p);
+   }
+      break;
+   case VG_PAINT_PATTERN_TILING_MODE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      return paint_pattern_tiling(p);
+   }
+      break;
+   case VG_PAINT_COLOR:
+   case VG_PAINT_COLOR_RAMP_STOPS:
+   case VG_PAINT_LINEAR_GRADIENT:
+   case VG_PAINT_RADIAL_GRADIENT:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+
+   case VG_PATH_FORMAT:
+      return VG_PATH_FORMAT_STANDARD;
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+      return vgGetParameterf(object, paramType);
+   case VG_PATH_DATATYPE: {
+      struct path *p = (struct path*)object;
+      return path_datatype(p);
+   }
+   case VG_PATH_NUM_SEGMENTS: {
+      struct path *p = (struct path*)object;
+      return path_num_segments(p);
+   }
+   case VG_PATH_NUM_COORDS: {
+      struct path *p = (struct path*)object;
+      return path_num_coords(p);
+   }
+      break;
+
+   case VG_IMAGE_FORMAT: {
+      struct vg_image *img = (struct vg_image*)object;
+      return img->format;
+   }
+      break;
+   case VG_IMAGE_WIDTH: {
+      struct vg_image *img = (struct vg_image*)object;
+      return img->width;
+   }
+      break;
+   case VG_IMAGE_HEIGHT: {
+      struct vg_image *img = (struct vg_image*)object;
+      return img->height;
+   }
+      break;
+
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS: {
+      return 1;
+   }
+      break;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+   return 0;
+}
+
+void vgGetParameterfv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      VGfloat * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!values || count <= 0 || count > real_count ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      values[0] = paint_type(p);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      values[0] = paint_spread_mode(p);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: {
+      struct vg_paint *p = (struct vg_paint *)object;
+      values[0] = paint_color_ramp_premultiplied(p);
+   }
+      break;
+   case VG_PAINT_PATTERN_TILING_MODE: {
+      values[0] = vgGetParameterf(object, paramType);
+   }
+      break;
+   case VG_PAINT_COLOR: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_get_color(paint, values);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_ramp_stops(paint, values, count);
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_linear_gradient(paint, values);
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_radial_gradient(paint, values);
+   }
+      break;
+
+   case VG_PATH_FORMAT:
+   case VG_PATH_DATATYPE:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+      values[0] = vgGetParameterf(object, paramType);
+      break;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+#endif
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
+
+void vgGetParameteriv(VGHandle object,
+                      VGint paramType,
+                      VGint count,
+                      VGint * values)
+{
+   struct vg_context *ctx = vg_current_context();
+   void *ptr = (void*)object;
+   VGint real_count = vgGetParameterVectorSize(object, paramType);
+
+   if (!ptr || object == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!values || count <= 0 || count > real_count ||
+       !is_aligned(values)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   switch(paramType) {
+   case VG_PAINT_TYPE:
+   case VG_PAINT_COLOR_RAMP_SPREAD_MODE:
+   case VG_PAINT_COLOR_RAMP_PREMULTIPLIED:
+   case VG_PAINT_PATTERN_TILING_MODE:
+#ifdef OPENVG_VERSION_1_1
+   case VG_FONT_NUM_GLYPHS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+#endif
+   case VG_PAINT_COLOR: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_get_coloriv(paint, values);
+   }
+      break;
+   case VG_PAINT_COLOR_RAMP_STOPS: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_ramp_stopsi(paint, values, count);
+   }
+      break;
+   case VG_PAINT_LINEAR_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_linear_gradienti(paint, values);
+   }
+      break;
+   case VG_PAINT_RADIAL_GRADIENT: {
+      struct vg_paint *paint = (struct vg_paint *)object;
+      paint_radial_gradienti(paint, values);
+   }
+      break;
+
+   case VG_PATH_SCALE:
+   case VG_PATH_BIAS:
+      values[0] = vgGetParameterf(object, paramType);
+      break;
+   case VG_PATH_FORMAT:
+   case VG_PATH_DATATYPE:
+   case VG_PATH_NUM_SEGMENTS:
+   case VG_PATH_NUM_COORDS:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+
+   case VG_IMAGE_FORMAT:
+   case VG_IMAGE_WIDTH:
+   case VG_IMAGE_HEIGHT:
+      values[0] = vgGetParameteri(object, paramType);
+      break;
+
+   default:
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      break;
+   }
+}
diff --git a/src/gallium/state_trackers/vega/api_path.c b/src/gallium/state_trackers/vega/api_path.c
new file mode 100644
index 0000000000..a6b7a2bb93
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_path.c
@@ -0,0 +1,488 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+#include "path.h"
+#include "polygon.h"
+#include "paint.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_draw_quad.h"
+
+VGPath vgCreatePath(VGint pathFormat,
+                    VGPathDatatype datatype,
+                    VGfloat scale, VGfloat bias,
+                    VGint segmentCapacityHint,
+                    VGint coordCapacityHint,
+                    VGbitfield capabilities)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (pathFormat != VG_PATH_FORMAT_STANDARD) {
+      vg_set_error(ctx, VG_UNSUPPORTED_PATH_FORMAT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (datatype < VG_PATH_DATATYPE_S_8 ||
+       datatype > VG_PATH_DATATYPE_F) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+   if (!scale) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   return (VGPath)path_create(datatype, scale, bias,
+                              segmentCapacityHint, coordCapacityHint,
+                              capabilities);
+}
+
+void vgClearPath(VGPath path, VGbitfield capabilities)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   p = (struct path *)path;
+   path_clear(p, capabilities);
+}
+
+void vgDestroyPath(VGPath p)
+{
+   struct path *path = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (p == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   path = (struct path *)p;
+   path_destroy(path);
+}
+
+void vgRemovePathCapabilities(VGPath path,
+                              VGbitfield capabilities)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGbitfield current;
+   struct path *p;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+   current = path_capabilities(p);
+   path_set_capabilities(p, (current &
+                             (~(capabilities & VG_PATH_CAPABILITY_ALL))));
+}
+
+VGbitfield vgGetPathCapabilities(VGPath path)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return 0;
+   }
+   p = (struct path*)path;
+   return path_capabilities(p);
+}
+
+void vgAppendPath(VGPath dstPath, VGPath srcPath)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *src, *dst;
+
+   if (dstPath == VG_INVALID_HANDLE || srcPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   src = (struct path *)srcPath;
+   dst = (struct path *)dstPath;
+
+   if (!(path_capabilities(src) & VG_PATH_CAPABILITY_APPEND_FROM) ||
+       !(path_capabilities(dst) & VG_PATH_CAPABILITY_APPEND_TO)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+   path_append_path(dst, src);
+}
+
+void vgAppendPathData(VGPath dstPath,
+                      VGint numSegments,
+                      const VGubyte * pathSegments,
+                      const void * pathData)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGint i;
+
+   if (dstPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!pathSegments) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   for (i = 0; i < numSegments; ++i) {
+      if (pathSegments[i] < VG_CLOSE_PATH ||
+          pathSegments[i] > VG_LCWARC_TO_REL) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return;
+      }
+   }
+
+   p = (struct path*)dstPath;
+
+   if (!pathData || !is_aligned_to(pathData, path_datatype_size(p))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!(path_capabilities(p)&VG_PATH_CAPABILITY_APPEND_TO)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+   path_append_data(p, numSegments, pathSegments, pathData);
+}
+
+void vgModifyPathCoords(VGPath dstPath,
+                        VGint startIndex,
+                        VGint numSegments,
+                        const void * pathData)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (dstPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (startIndex < 0 || numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path *)dstPath;
+
+   if (!pathData || !is_aligned_to(pathData, path_datatype_size(p))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (startIndex + numSegments > path_num_segments(p)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!(path_capabilities(p)&VG_PATH_CAPABILITY_MODIFY)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+   path_modify_coords(p, startIndex, numSegments, pathData);
+}
+
+void vgTransformPath(VGPath dstPath, VGPath srcPath)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *src = 0, *dst = 0;
+
+   if (dstPath == VG_INVALID_HANDLE || srcPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   src = (struct path *)srcPath;
+   dst = (struct path *)dstPath;
+
+   if (!(path_capabilities(src) & VG_PATH_CAPABILITY_TRANSFORM_FROM) ||
+       !(path_capabilities(dst) & VG_PATH_CAPABILITY_TRANSFORM_TO)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+   path_transform(dst, src);
+}
+
+VGboolean vgInterpolatePath(VGPath dstPath,
+                            VGPath startPath,
+                            VGPath endPath,
+                            VGfloat amount)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *start = 0, *dst = 0, *end = 0;
+
+   if (dstPath == VG_INVALID_HANDLE ||
+       startPath == VG_INVALID_HANDLE ||
+       endPath == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return VG_FALSE;
+   }
+   dst = (struct path *)dstPath;
+   start = (struct path *)startPath;
+   end = (struct path *)endPath;
+
+   if (!(path_capabilities(dst) & VG_PATH_CAPABILITY_INTERPOLATE_TO) ||
+       !(path_capabilities(start) & VG_PATH_CAPABILITY_INTERPOLATE_FROM) ||
+       !(path_capabilities(end) & VG_PATH_CAPABILITY_INTERPOLATE_FROM)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return VG_FALSE;
+   }
+
+   return path_interpolate(dst,
+                           start, end, amount);
+}
+
+VGfloat vgPathLength(VGPath path,
+                     VGint startSegment,
+                     VGint numSegments)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return -1;
+   }
+   if (startSegment < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return -1;
+   }
+   if (numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return -1;
+   }
+   p = (struct path*)path;
+
+   if (!(path_capabilities(p) & VG_PATH_CAPABILITY_PATH_LENGTH)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return -1;
+   }
+   if (startSegment + numSegments > path_num_segments(p)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return -1;
+   }
+
+   return path_length(p, startSegment, numSegments);
+}
+
+void vgPointAlongPath(VGPath path,
+                      VGint startSegment,
+                      VGint numSegments,
+                      VGfloat distance,
+                      VGfloat * x, VGfloat * y,
+                      VGfloat * tangentX,
+                      VGfloat * tangentY)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (startSegment < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (numSegments <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!is_aligned(x) || !is_aligned(y) ||
+       !is_aligned(tangentX) || !is_aligned(tangentY)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+
+   caps = path_capabilities(p);
+   if (!(caps & VG_PATH_CAPABILITY_POINT_ALONG_PATH) ||
+       !(caps & VG_PATH_CAPABILITY_TANGENT_ALONG_PATH)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+   if (startSegment + numSegments > path_num_segments(p)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   {
+      VGfloat point[2], normal[2];
+      path_point(p, startSegment, numSegments, distance,
+                 point, normal);
+      if (x)
+         *x = point[0];
+      if (y)
+         *y = point[1];
+      if (tangentX)
+         *tangentX = -normal[1];
+      if (tangentY)
+         *tangentY = normal[0];
+   }
+}
+
+void vgPathBounds(VGPath path,
+                  VGfloat * minX,
+                  VGfloat * minY,
+                  VGfloat * width,
+                  VGfloat * height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!minX || !minY || !width || !height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!is_aligned(minX) || !is_aligned(minY) ||
+       !is_aligned(width) || !is_aligned(height)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+
+   caps = path_capabilities(p);
+   if (!(caps & VG_PATH_CAPABILITY_PATH_BOUNDS)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+   path_bounding_rect(p, minX, minY, width, height);
+}
+
+void vgPathTransformedBounds(VGPath path,
+                             VGfloat * minX,
+                             VGfloat * minY,
+                             VGfloat * width,
+                             VGfloat * height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct path *p = 0;
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!minX || !minY || !width || !height) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (!is_aligned(minX) || !is_aligned(minY) ||
+       !is_aligned(width) || !is_aligned(height)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   p = (struct path*)path;
+
+   caps = path_capabilities(p);
+   if (!(caps & VG_PATH_CAPABILITY_PATH_TRANSFORMED_BOUNDS)) {
+      vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR);
+      return;
+   }
+
+#if 0
+   /* faster, but seems to have precision problems... */
+   path_bounding_rect(p, minX, minY, width, height);
+   if (*width > 0 && *height > 0) {
+      VGfloat pts[] = {*minX,          *minY,
+                       *minX + *width, *minY,
+                       *minX + *width, *minY + *height,
+                       *minX,          *minY + *height};
+      struct matrix *matrix = &ctx->state.vg.path_user_to_surface_matrix;
+      VGfloat maxX, maxY;
+      matrix_map_point(matrix, pts[0], pts[1], pts + 0, pts + 1);
+      matrix_map_point(matrix, pts[2], pts[3], pts + 2, pts + 3);
+      matrix_map_point(matrix, pts[4], pts[5], pts + 4, pts + 5);
+      matrix_map_point(matrix, pts[6], pts[7], pts + 6, pts + 7);
+      *minX = MIN2(pts[0], MIN2(pts[2], MIN2(pts[4], pts[6])));
+      *minY = MIN2(pts[1], MIN2(pts[3], MIN2(pts[5], pts[7])));
+      maxX = MAX2(pts[0], MAX2(pts[2], MAX2(pts[4], pts[6])));
+      maxY = MAX2(pts[1], MAX2(pts[3], MAX2(pts[5], pts[7])));
+      *width  = maxX - *minX;
+      *height = maxY - *minY;
+   }
+#else
+   {
+      struct path *dst = path_create(VG_PATH_DATATYPE_F, 1.0, 0,
+                                     0, 0, VG_PATH_CAPABILITY_ALL);
+      path_transform(dst, p);
+      path_bounding_rect(dst, minX, minY, width, height);
+      path_destroy(dst);
+   }
+#endif
+}
+
+
+void vgDrawPath(VGPath path, VGbitfield paintModes)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (path == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   if (!(paintModes & (VG_STROKE_PATH | VG_FILL_PATH))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (path_is_empty((struct path*)path))
+      return;
+   path_render((struct path*)path, paintModes);
+}
+
diff --git a/src/gallium/state_trackers/vega/api_text.c b/src/gallium/state_trackers/vega/api_text.c
new file mode 100644
index 0000000000..d8411cf3e8
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_text.c
@@ -0,0 +1,258 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+#include "util/u_memory.h"
+
+#ifdef OPENVG_VERSION_1_1
+
+struct vg_font {
+   struct vg_object base;
+
+   VGint glyph_indices[200];
+   VGint num_glyphs;
+};
+
+VGFont vgCreateFont(VGint glyphCapacityHint)
+{
+   struct vg_font *font = 0;
+   struct vg_context *ctx = vg_current_context();
+
+   if (glyphCapacityHint < 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return VG_INVALID_HANDLE;
+   }
+
+   font = CALLOC_STRUCT(vg_font);
+   vg_init_object(&font->base, ctx, VG_OBJECT_FONT);
+   vg_context_add_object(ctx, VG_OBJECT_FONT, font);
+   return (VGFont)font;
+}
+
+void vgDestroyFont(VGFont f)
+{
+   struct vg_font *font = (struct vg_font *)f;
+   struct vg_context *ctx = vg_current_context();
+
+   if (f == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   vg_context_remove_object(ctx, VG_OBJECT_FONT, font);
+   /*free(font);*/
+}
+
+void vgSetGlyphToPath(VGFont font,
+                      VGuint glyphIndex,
+                      VGPath path,
+                      VGboolean isHinted,
+                      VGfloat glyphOrigin [2],
+                      VGfloat escapement[2])
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_object *pathObj;
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_FONT, (void *)font)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!glyphOrigin || !escapement ||
+       !is_aligned(glyphOrigin) || !is_aligned(escapement)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (path != VG_INVALID_HANDLE &&
+       !vg_context_is_object_valid(ctx, VG_OBJECT_PATH, (void *)path)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   pathObj = (struct vg_object*)path;
+   if (pathObj && pathObj->type != VG_OBJECT_PATH) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+
+   f = (struct vg_font*)font;
+   f->glyph_indices[f->num_glyphs] = glyphIndex;
+   ++f->num_glyphs;
+}
+
+void vgSetGlyphToImage(VGFont font,
+                       VGuint glyphIndex,
+                       VGImage image,
+                       VGfloat glyphOrigin [2],
+                       VGfloat escapement[2])
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_object *img_obj;
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE ||
+       !vg_context_is_object_valid(ctx, VG_OBJECT_FONT, (void *)font)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (!glyphOrigin || !escapement ||
+       !is_aligned(glyphOrigin) || !is_aligned(escapement)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (image != VG_INVALID_HANDLE &&
+       !vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void *)image)) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   img_obj = (struct vg_object*)image;
+   if (img_obj && img_obj->type != VG_OBJECT_IMAGE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   f = (struct vg_font*)font;
+   f->glyph_indices[f->num_glyphs] = glyphIndex;
+   ++f->num_glyphs;
+}
+
+static INLINE VGboolean font_contains_glyph(struct vg_font *font,
+                                            VGuint glyph_index)
+{
+   VGint i;
+   for (i = 0; i < font->num_glyphs; ++i) {
+      if (font->glyph_indices[i] == glyph_index) {
+         return VG_TRUE;
+      }
+   }
+   return VG_FALSE;
+}
+
+void vgClearGlyph(VGFont font,
+                  VGuint glyphIndex)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_font *f;
+   VGint i;
+
+   if (font == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (glyphIndex <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   f = (struct vg_font*)font;
+   if (!font_contains_glyph(f, glyphIndex)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   for (i = 0; i < f->num_glyphs; ++i) {
+      if (f->glyph_indices[i] == glyphIndex) {
+         /*FIXME*/
+         f->glyph_indices[f->num_glyphs] = 0;
+         --f->num_glyphs;
+         return;
+      }
+   }
+}
+
+void vgDrawGlyph(VGFont font,
+                 VGuint glyphIndex,
+                 VGbitfield paintModes,
+                 VGboolean allowAutoHinting)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (glyphIndex <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (paintModes & (~(VG_STROKE_PATH|VG_FILL_PATH))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   f = (struct vg_font*)font;
+   if (!font_contains_glyph(f, glyphIndex)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+}
+
+void vgDrawGlyphs(VGFont font,
+                  VGint glyphCount,
+                  VGuint *glyphIndices,
+                  VGfloat *adjustments_x,
+                  VGfloat *adjustments_y,
+                  VGbitfield paintModes,
+                  VGboolean allowAutoHinting)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGint i;
+   struct vg_font *f;
+
+   if (font == VG_INVALID_HANDLE) {
+      vg_set_error(ctx, VG_BAD_HANDLE_ERROR);
+      return;
+   }
+   if (glyphCount <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!glyphIndices || !is_aligned(glyphIndices)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (!adjustments_x || !is_aligned(adjustments_x) ||
+       !adjustments_y || !is_aligned(adjustments_y)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   if (paintModes & (~(VG_STROKE_PATH|VG_FILL_PATH))) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   f = (struct vg_font*)font;
+   for (i = 0; i < glyphCount; ++i) {
+      VGuint glyph_index = glyphIndices[i];
+      if (!font_contains_glyph(f, glyph_index)) {
+         vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+         return;
+      }
+   }
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/api_transform.c b/src/gallium/state_trackers/vega/api_transform.c
new file mode 100644
index 0000000000..763a5ec415
--- /dev/null
+++ b/src/gallium/state_trackers/vega/api_transform.c
@@ -0,0 +1,128 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+
+#include "vg_context.h"
+
+#include "matrix.h"
+
+void vgLoadIdentity(void)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct  matrix *mat = vg_state_matrix(&ctx->state.vg);
+   matrix_load_identity(mat);
+}
+
+void vgLoadMatrix(const VGfloat * m)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct  matrix *mat;
+
+   if (!ctx)
+      return;
+
+   if (!m || !is_aligned(m)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   mat = vg_state_matrix(&ctx->state.vg);
+   matrix_init(mat, m);
+   if (!matrix_is_affine(mat)) {
+      if (ctx->state.vg.matrix_mode != VG_MATRIX_IMAGE_USER_TO_SURFACE) {
+         matrix_make_affine(mat);
+      }
+   }
+}
+
+void vgGetMatrix(VGfloat * m)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *mat;
+
+   if (!ctx)
+      return;
+
+   if (!m || !is_aligned(m)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   mat = vg_state_matrix(&ctx->state.vg);
+   memcpy(m, mat->m, sizeof(VGfloat)*9);
+}
+
+void vgMultMatrix(const VGfloat * m)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst, src;
+
+   if (!ctx)
+      return;
+
+   if (!m || !is_aligned(m)) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   matrix_init(&src, m);
+   dst = vg_state_matrix(&ctx->state.vg);
+   if (!matrix_is_affine(&src)) {
+      if (ctx->state.vg.matrix_mode != VG_MATRIX_IMAGE_USER_TO_SURFACE) {
+         matrix_make_affine(&src);
+      }
+   }
+   matrix_mult(dst, &src);
+
+}
+
+void vgTranslate(VGfloat tx, VGfloat ty)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_translate(dst, tx, ty);
+}
+
+void vgScale(VGfloat sx, VGfloat sy)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_scale(dst, sx, sy);
+}
+
+void vgShear(VGfloat shx, VGfloat shy)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_shear(dst, shx, shy);
+}
+
+void vgRotate(VGfloat angle)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *dst = vg_state_matrix(&ctx->state.vg);
+   matrix_rotate(dst, angle);
+}
diff --git a/src/gallium/state_trackers/vega/arc.c b/src/gallium/state_trackers/vega/arc.c
new file mode 100644
index 0000000000..e74c7f0334
--- /dev/null
+++ b/src/gallium/state_trackers/vega/arc.c
@@ -0,0 +1,708 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "arc.h"
+
+#include "matrix.h"
+#include "bezier.h"
+#include "polygon.h"
+#include "stroker.h"
+#include "path.h"
+
+#include "util/u_debug.h"
+
+#include <math.h>
+
+#ifndef M_PI
+#define M_PI 3.14159265358979323846
+#endif
+
+#define DEBUG_ARCS 0
+
+static const VGfloat two_pi = M_PI * 2;
+
+
+static const double coeffs3Low[2][4][4] = {
+   {
+      {  3.85268,   -21.229,      -0.330434,    0.0127842  },
+      { -1.61486,     0.706564,    0.225945,    0.263682   },
+      { -0.910164,    0.388383,    0.00551445,  0.00671814 },
+      { -0.630184,    0.192402,    0.0098871,   0.0102527  }
+   },
+   {
+      { -0.162211,    9.94329,     0.13723,     0.0124084  },
+      { -0.253135,    0.00187735,  0.0230286,   0.01264    },
+      { -0.0695069,  -0.0437594,   0.0120636,   0.0163087  },
+      { -0.0328856,  -0.00926032, -0.00173573,  0.00527385 }
+   }
+};
+
+/* coefficients for error estimation
+   while using cubic Bézier curves for approximation
+   1/4 <= b/a <= 1 */
+static const double coeffs3High[2][4][4] = {
+   {
+      {  0.0899116, -19.2349,     -4.11711,     0.183362   },
+      {  0.138148,   -1.45804,     1.32044,     1.38474    },
+      {  0.230903,   -0.450262,    0.219963,    0.414038   },
+      {  0.0590565,  -0.101062,    0.0430592,   0.0204699  }
+   },
+   {
+      {  0.0164649,   9.89394,     0.0919496,   0.00760802 },
+      {  0.0191603,  -0.0322058,   0.0134667,  -0.0825018  },
+      {  0.0156192,  -0.017535,    0.00326508, -0.228157   },
+      { -0.0236752,   0.0405821,  -0.0173086,   0.176187   }
+   }
+};
+
+/* safety factor to convert the "best" error approximation
+   into a "max bound" error */
+static const double safety3[] = {
+   0.001, 4.98, 0.207, 0.0067
+};
+
+/* The code below is from the OpenVG 1.1 Spec
+ * Section 18.4 */
+
+/* Given: Points (x0, y0) and (x1, y1)
+ * Return: TRUE if a solution exists, FALSE otherwise
+ *         Circle centers are written to (cx0, cy0) and (cx1, cy1)
+ */
+static VGboolean
+find_unit_circles(double x0, double y0, double x1, double y1,
+                  double *cx0, double *cy0,
+                  double *cx1, double *cy1)
+{
+   /* Compute differences and averages */
+   double dx = x0 - x1;
+   double dy = y0 - y1;
+   double xm = (x0 + x1)/2;
+   double ym = (y0 + y1)/2;
+   double dsq, disc, s, sdx, sdy;
+
+   /* Solve for intersecting unit circles */
+   dsq = dx*dx + dy*dy;
+   if (dsq == 0.0) return VG_FALSE; /* Points are coincident */
+   disc = 1.0/dsq - 1.0/4.0;
+
+   /* the precision we care about here is around float so if we're
+    * around the float defined zero then make it official to avoid
+    * precision problems later on */
+   if (floatIsZero(disc))
+      disc = 0.0;
+
+   if (disc < 0.0) return VG_FALSE; /* Points are too far apart */
+   s = sqrt(disc);
+   sdx = s*dx;
+   sdy = s*dy;
+   *cx0 = xm + sdy;
+   *cy0 = ym - sdx;
+   *cx1 = xm - sdy;
+   *cy1 = ym + sdx;
+   return VG_TRUE;
+}
+
+
+/* Given:  Ellipse parameters rh, rv, rot (in degrees),
+ *         endpoints (x0, y0) and (x1, y1)
+ * Return: TRUE if a solution exists, FALSE otherwise
+ *         Ellipse centers are written to (cx0, cy0) and (cx1, cy1)
+ */
+static VGboolean
+find_ellipses(double rh, double rv, double rot,
+              double x0, double y0, double x1, double y1,
+              double *cx0, double *cy0, double *cx1, double *cy1)
+{
+   double COS, SIN, x0p, y0p, x1p, y1p, pcx0, pcy0, pcx1, pcy1;
+   /* Convert rotation angle from degrees to radians */
+   rot *= M_PI/180.0;
+   /* Pre-compute rotation matrix entries */
+   COS = cos(rot); SIN = sin(rot);
+   /* Transform (x0, y0) and (x1, y1) into unit space */
+   /* using (inverse) rotate, followed by (inverse) scale   */
+   x0p = (x0*COS + y0*SIN)/rh;
+   y0p = (-x0*SIN + y0*COS)/rv;
+   x1p = (x1*COS + y1*SIN)/rh;
+   y1p = (-x1*SIN + y1*COS)/rv;
+   if (!find_unit_circles(x0p, y0p, x1p, y1p,
+                          &pcx0, &pcy0, &pcx1, &pcy1)) {
+      return VG_FALSE;
+   }
+   /* Transform back to original coordinate space */
+   /* using (forward) scale followed by (forward) rotate */
+   pcx0 *= rh; pcy0 *= rv;
+   pcx1 *= rh; pcy1 *= rv;
+   *cx0 = pcx0*COS - pcy0*SIN;
+   *cy0 = pcx0*SIN + pcy0*COS;
+   *cx1 = pcx1*COS - pcy1*SIN;
+   *cy1 = pcx1*SIN + pcy1*COS;
+   return VG_TRUE;
+}
+
+static INLINE VGboolean
+try_to_fix_radii(struct arc *arc)
+{
+   double COS, SIN, rot, x0p, y0p, x1p, y1p;
+   double dx, dy, dsq, scale;
+
+   /* Convert rotation angle from degrees to radians */
+   rot = DEGREES_TO_RADIANS(arc->theta);
+
+   /* Pre-compute rotation matrix entries */
+   COS = cos(rot); SIN = sin(rot);
+
+   /* Transform (x0, y0) and (x1, y1) into unit space */
+   /* using (inverse) rotate, followed by (inverse) scale   */
+   x0p = (arc->x1*COS + arc->y1*SIN)/arc->a;
+   y0p = (-arc->x1*SIN + arc->y1*COS)/arc->b;
+   x1p = (arc->x2*COS + arc->y2*SIN)/arc->a;
+   y1p = (-arc->x2*SIN + arc->y2*COS)/arc->b;
+   /* Compute differences and averages */
+   dx = x0p - x1p;
+   dy = y0p - y1p;
+
+   dsq = dx*dx + dy*dy;
+#if 0
+   if (dsq <= 0.001) {
+      debug_printf("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAaaaaa\n");
+   }
+#endif
+   scale = 1/(2/sqrt(dsq));
+   arc->a *= scale;
+   arc->b *= scale;
+   return VG_TRUE;
+}
+
+static INLINE double vector_normalize(double *v)
+{
+   double sq = v[0] * v[0] + v[1] * v[1];
+   return sqrt(sq);
+}
+static INLINE double vector_orientation(double *v)
+{
+   double norm = vector_normalize(v);
+   double cosa = v[0] / norm;
+   double sina = v[1] / norm;
+   return (sina>=0 ? acos(cosa) : 2*M_PI - acos(cosa));
+}
+static INLINE double vector_dot(double *v0,
+                                double *v1)
+{
+   return v0[0] * v1[0] + v0[1] * v1[1];
+}
+
+static INLINE double vector_angles(double *v0,
+                                   double *v1)
+{
+   double dot = vector_dot(v0, v1);
+   double norm0 = vector_normalize(v0);
+   double norm1 = vector_normalize(v1);
+
+   return acos(dot / (norm0 * norm1));
+}
+
+static VGboolean find_angles(struct arc *arc)
+{
+   double vec0[2], vec1[2];
+   double lambda1, lambda2;
+   double angle;
+   struct matrix matrix;
+
+   if (floatIsZero(arc->a) || floatIsZero(arc->b)) {
+      return VG_FALSE;
+   }
+   /* map the points to an identity circle */
+   matrix_load_identity(&matrix);
+   matrix_scale(&matrix, 1.f, arc->a/arc->b);
+   matrix_rotate(&matrix, -arc->theta);
+   matrix_map_point(&matrix,
+                    arc->x1, arc->y1,
+                    &arc->x1, &arc->y1);
+   matrix_map_point(&matrix,
+                    arc->x2, arc->y2,
+                    &arc->x2, &arc->y2);
+   matrix_map_point(&matrix,
+                    arc->cx, arc->cy,
+                    &arc->cx, &arc->cy);
+
+#if DEBUG_ARCS
+   debug_printf("Matrix 3 [%f, %f, %f| %f, %f, %f| %f, %f, %f]\n",
+                matrix.m[0], matrix.m[1], matrix.m[2],
+                matrix.m[3], matrix.m[4], matrix.m[5],
+                matrix.m[6], matrix.m[7], matrix.m[8]);
+   debug_printf("Endpoints [%f, %f], [%f, %f]\n",
+                arc->x1, arc->y1, arc->x2, arc->y2);
+#endif
+
+   vec0[0] = arc->x1 - arc->cx;
+   vec0[1] = arc->y1 - arc->cy;
+   vec1[0] = arc->x2 - arc->cx;
+   vec1[1] = arc->y2 - arc->cy;
+
+#if DEBUG_ARCS
+   debug_printf("Vec is [%f, %f], [%f, %f], [%f, %f]\n",
+                vec0[0], vec0[1], vec1[0], vec1[1], arc->cx, arc->cy);
+#endif
+
+   lambda1 = vector_orientation(vec0);
+
+   if (isnan(lambda1))
+      lambda1 = 0.f;
+
+   if (arc->type == VG_SCWARC_TO ||
+       arc->type == VG_SCCWARC_TO)
+      angle = vector_angles(vec0, vec1);
+   else if (arc->type == VG_LCWARC_TO ||
+            arc->type == VG_LCCWARC_TO) {
+      angle = 2*M_PI - vector_angles(vec0, vec1);
+   } else
+      abort();
+
+   if (isnan(angle))
+      angle = M_PI;
+
+
+   if (arc->type == VG_SCWARC_TO ||
+       arc->type == VG_LCWARC_TO)
+      lambda2 = lambda1 - angle;
+   else
+      lambda2 = lambda1 + angle;
+
+#if DEBUG_ARCS
+   debug_printf("Angle is %f and (%f, %f)\n", angle, lambda1, lambda2);
+#endif
+
+#if 0
+   arc->eta1 = atan2(sin(lambda1) / arc->b,
+                     cos(lambda1) / arc->a);
+   arc->eta2 = atan2(sin(lambda2) / arc->b,
+                     cos(lambda2) / arc->a);
+
+   /* make sure we have eta1 <= eta2 <= eta1 + 2 PI */
+   arc->eta2 -= two_pi * floor((arc->eta2 - arc->eta1) / two_pi);
+
+   /* the preceding correction fails if we have exactly et2 - eta1 = 2 PI
+      it reduces the interval to zero length */
+   if ((lambda2 - lambda1 > M_PI) && (arc->eta2 - arc->eta1 < M_PI)) {
+      arc->eta2 += 2 * M_PI;
+   }
+#else
+   arc->eta1 = lambda1;
+   arc->eta2 = lambda2;
+#endif
+
+   return VG_TRUE;
+}
+
+#if DEBUG_ARCS
+static void check_endpoints(struct arc *arc)
+{
+   double x1, y1, x2, y2;
+
+   double a_cos_eta1 = arc->a * cos(arc->eta1);
+   double b_sin_eta1 = arc->b * sin(arc->eta1);
+   x1 = arc->cx + a_cos_eta1 * arc->cos_theta -
+        b_sin_eta1 * arc->sin_theta;
+   y1 = arc->cy + a_cos_eta1 * arc->sin_theta +
+        b_sin_eta1 * arc->cos_theta;
+
+   double a_cos_eta2 = arc->a * cos(arc->eta2);
+   double b_sin_eta2 = arc->b * sin(arc->eta2);
+   x2 = arc->cx + a_cos_eta2 * arc->cos_theta -
+        b_sin_eta2 * arc->sin_theta;
+   y2 = arc->cy + a_cos_eta2 * arc->sin_theta +
+        b_sin_eta2 * arc->cos_theta;
+
+   debug_printf("Computed (%f, %f), (%f, %f)\n",
+                x1, y1, x2, y2);
+   debug_printf("Real     (%f, %f), (%f, %f)\n",
+                arc->x1, arc->y1,
+                arc->x2, arc->y2);
+}
+#endif
+
+void arc_init(struct arc *arc,
+              VGPathSegment type,
+              VGfloat x1, VGfloat y1,
+              VGfloat x2, VGfloat y2,
+              VGfloat rh, VGfloat rv,
+              VGfloat rot)
+{
+   assert(type == VG_SCCWARC_TO ||
+          type == VG_SCWARC_TO ||
+          type == VG_LCCWARC_TO ||
+          type == VG_LCWARC_TO);
+   arc->type = type;
+   arc->x1  = x1;
+   arc->y1  = y1;
+   arc->x2  = x2;
+   arc->y2  = y2;
+   arc->a   = rh;
+   arc->b   = rv;
+   arc->theta = rot;
+   arc->cos_theta = cos(arc->theta);
+   arc->sin_theta = sin(arc->theta);
+   {
+      double cx0, cy0, cx1, cy1;
+      double cx, cy;
+      arc->is_valid =  find_ellipses(rh, rv, rot, x1, y1, x2, y2,
+                                     &cx0, &cy0, &cx1, &cy1);
+
+      if (!arc->is_valid && try_to_fix_radii(arc)) {
+         rh = arc->a;
+         rv = arc->b;
+         arc->is_valid =
+            find_ellipses(rh, rv, rot, x1, y1, x2, y2,
+                          &cx0, &cy0, &cx1, &cy1);
+      }
+
+      if (type == VG_SCWARC_TO ||
+          type == VG_LCCWARC_TO) {
+         cx = cx1;
+         cy = cy1;
+      } else {
+         cx = cx0;
+         cy = cy0;
+      }
+#if DEBUG_ARCS
+      debug_printf("Centers are : (%f, %f) , (%f, %f). Real (%f, %f)\n",
+                   cx0, cy0, cx1, cy1, cx, cy);
+#endif
+      arc->cx = cx;
+      arc->cy = cy;
+      if (arc->is_valid) {
+         arc->is_valid = find_angles(arc);
+#if DEBUG_ARCS
+         check_endpoints(arc);
+#endif
+         /* remap a few points. find_angles requires
+          * rot in angles, the rest of the code
+          * will need them in radians. and find_angles
+          * modifies the center to match an identity
+          * circle so lets reset it */
+         arc->theta = DEGREES_TO_RADIANS(rot);
+         arc->cos_theta = cos(arc->theta);
+         arc->sin_theta = sin(arc->theta);
+         arc->cx = cx;
+         arc->cy = cy;
+      }
+   }
+}
+
+static INLINE double rational_function(double x, const double *c)
+{
+   return (x * (x * c[0] + c[1]) + c[2]) / (x + c[3]);
+}
+
+static double estimate_error(struct arc *arc,
+                             double etaA, double etaB)
+{
+   double eta  = 0.5 * (etaA + etaB);
+
+   double x    = arc->b / arc->a;
+   double dEta = etaB - etaA;
+   double cos2 = cos(2 * eta);
+   double cos4 = cos(4 * eta);
+   double cos6 = cos(6 * eta);
+   double c0, c1;
+
+   /* select the right coeficients set according to degree and b/a */
+   const double (*coeffs)[4][4];
+   const double *safety;
+   coeffs = (x < 0.25) ? coeffs3Low : coeffs3High;
+   safety = safety3;
+
+   c0 = rational_function(x, coeffs[0][0])
+        + cos2 * rational_function(x, coeffs[0][1])
+        + cos4 * rational_function(x, coeffs[0][2])
+        + cos6 * rational_function(x, coeffs[0][3]);
+
+   c1 = rational_function(x, coeffs[1][0])
+        + cos2 * rational_function(x, coeffs[1][1])
+        + cos4 * rational_function(x, coeffs[1][2])
+        + cos6 * rational_function(x, coeffs[1][3]);
+
+   return rational_function(x, safety) * arc->a * exp(c0 + c1 * dEta);
+}
+
+struct arc_cb {
+   void (*move)(struct arc_cb *cb, VGfloat x, VGfloat y);
+   void (*point)(struct arc_cb *cb, VGfloat x, VGfloat y);
+   void (*bezier)(struct arc_cb *cb, struct bezier *bezier);
+
+   void *user_data;
+};
+
+static void cb_null_move(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+}
+
+static void polygon_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct polygon *poly = (struct polygon*)cb->user_data;
+   polygon_vertex_append(poly, x, y);
+}
+
+static void polygon_bezier(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct polygon *poly = (struct polygon*)cb->user_data;
+   bezier_add_to_polygon(bezier, poly);
+}
+
+static void stroke_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_line_to(stroker, x, y);
+}
+
+static void stroke_curve(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_curve_to(stroker,
+                    bezier->x2, bezier->y2,
+                    bezier->x3, bezier->y3,
+                    bezier->x4, bezier->y4);
+}
+
+static void stroke_emit_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_emit_line_to(stroker, x, y);
+}
+
+static void stroke_emit_curve(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct stroker *stroker = (struct stroker*)cb->user_data;
+   stroker_emit_curve_to(stroker,
+                         bezier->x2, bezier->y2,
+                         bezier->x3, bezier->y3,
+                         bezier->x4, bezier->y4);
+}
+
+static void arc_path_move(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct path *path = (struct path*)cb->user_data;
+   path_move_to(path, x, y);
+}
+
+static void arc_path_point(struct arc_cb *cb, VGfloat x, VGfloat y)
+{
+   struct path *path = (struct path*)cb->user_data;
+   path_line_to(path, x, y);
+}
+
+static void arc_path_bezier(struct arc_cb *cb, struct bezier *bezier)
+{
+   struct path *path = (struct path*)cb->user_data;
+   path_cubic_to(path,
+                 bezier->x2, bezier->y2,
+                 bezier->x3, bezier->y3,
+                 bezier->x4, bezier->y4);
+}
+
+static INLINE int num_beziers_needed(struct arc *arc)
+{
+   double threshold = 0.05;
+   VGboolean found = VG_FALSE;
+   int n = 1;
+   double min_eta, max_eta;
+
+   min_eta = MIN2(arc->eta1, arc->eta2);
+   max_eta = MAX2(arc->eta1, arc->eta2);
+
+   while ((! found) && (n < 1024)) {
+      double d_eta = (max_eta - min_eta) / n;
+      if (d_eta <= 0.5 * M_PI) {
+         double eta_b = min_eta;
+         found = VG_TRUE;
+         for (int i = 0; found && (i < n); ++i) {
+            double etaA = eta_b;
+            eta_b += d_eta;
+            found = (estimate_error(arc, etaA, eta_b) <= threshold);
+         }
+      }
+      n = n << 1;
+   }
+
+   return n;
+}
+
+static void arc_to_beziers(struct arc *arc,
+                           struct arc_cb cb,
+                           struct matrix *matrix)
+{
+   int n = 1;
+   double d_eta, eta_b, cos_eta_b,
+      sin_eta_b, a_cos_eta_b, b_sin_eta_b, a_sin_eta_b,
+      b_cos_eta_b, x_b, y_b, x_b_dot, y_b_dot, lx, ly;
+   double t, alpha;
+
+   { /* always move to the start of the arc */
+      VGfloat x = arc->x1;
+      VGfloat y = arc->y1;
+      matrix_map_point(matrix, x, y, &x, &y);
+      cb.move(&cb, x, y);
+   }
+
+   if (!arc->is_valid) {
+      VGfloat x = arc->x2;
+      VGfloat y = arc->y2;
+      matrix_map_point(matrix, x, y, &x, &y);
+      cb.point(&cb, x, y);
+      return;
+   }
+
+   /* find the number of Bézier curves needed */
+   n = num_beziers_needed(arc);
+
+   d_eta = (arc->eta2 - arc->eta1) / n;
+   eta_b = arc->eta1;
+
+   cos_eta_b  = cos(eta_b);
+   sin_eta_b  = sin(eta_b);
+   a_cos_eta_b = arc->a * cos_eta_b;
+   b_sin_eta_b = arc->b * sin_eta_b;
+   a_sin_eta_b = arc->a * sin_eta_b;
+   b_cos_eta_b = arc->b * cos_eta_b;
+   x_b       = arc->cx + a_cos_eta_b * arc->cos_theta -
+               b_sin_eta_b * arc->sin_theta;
+   y_b       = arc->cy + a_cos_eta_b * arc->sin_theta +
+               b_sin_eta_b * arc->cos_theta;
+   x_b_dot    = -a_sin_eta_b * arc->cos_theta -
+                b_cos_eta_b * arc->sin_theta;
+   y_b_dot    = -a_sin_eta_b * arc->sin_theta +
+                b_cos_eta_b * arc->cos_theta;
+
+   {
+      VGfloat x = x_b, y = y_b;
+      matrix_map_point(matrix, x, y, &x, &y);
+      cb.point(&cb, x, y);
+   }
+   lx = x_b;
+   ly = y_b;
+
+   t     = tan(0.5 * d_eta);
+   alpha = sin(d_eta) * (sqrt(4 + 3 * t * t) - 1) / 3;
+
+   for (int i = 0; i < n; ++i) {
+      struct bezier bezier;
+      double xA    = x_b;
+      double yA    = y_b;
+      double xADot = x_b_dot;
+      double yADot = y_b_dot;
+
+      eta_b    += d_eta;
+      cos_eta_b  = cos(eta_b);
+      sin_eta_b  = sin(eta_b);
+      a_cos_eta_b = arc->a * cos_eta_b;
+      b_sin_eta_b = arc->b * sin_eta_b;
+      a_sin_eta_b = arc->a * sin_eta_b;
+      b_cos_eta_b = arc->b * cos_eta_b;
+      x_b       = arc->cx + a_cos_eta_b * arc->cos_theta -
+                  b_sin_eta_b * arc->sin_theta;
+      y_b       = arc->cy + a_cos_eta_b * arc->sin_theta +
+                  b_sin_eta_b * arc->cos_theta;
+      x_b_dot    = -a_sin_eta_b * arc->cos_theta -
+                   b_cos_eta_b * arc->sin_theta;
+      y_b_dot    = -a_sin_eta_b * arc->sin_theta +
+                   b_cos_eta_b * arc->cos_theta;
+
+      bezier_init(&bezier,
+                  lx, ly,
+                  (float) (xA + alpha * xADot), (float) (yA + alpha * yADot),
+                  (float) (x_b - alpha * x_b_dot), (float) (y_b - alpha * y_b_dot),
+                  (float) x_b,                   (float) y_b);
+#if 0
+      debug_printf("%d) Bezier (%f, %f), (%f, %f), (%f, %f), (%f, %f)\n",
+                   i,
+                   bezier.x1, bezier.y1,
+                   bezier.x2, bezier.y2,
+                   bezier.x3, bezier.y3,
+                   bezier.x4, bezier.y4);
+#endif
+      bezier_transform(&bezier, matrix);
+      cb.bezier(&cb, &bezier);
+      lx = x_b;
+      ly = y_b;
+   }
+}
+
+
+void arc_add_to_polygon(struct arc *arc,
+                        struct polygon *poly,
+                        struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = cb_null_move;
+   cb.point = polygon_point;
+   cb.bezier = polygon_bezier;
+   cb.user_data = poly;
+
+   arc_to_beziers(arc, cb, matrix);
+}
+
+void arc_stroke_cb(struct arc *arc,
+                   struct stroker *stroke,
+                   struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = cb_null_move;
+   cb.point = stroke_point;
+   cb.bezier = stroke_curve;
+   cb.user_data = stroke;
+
+   arc_to_beziers(arc, cb, matrix);
+}
+
+void arc_stroker_emit(struct arc *arc,
+                      struct stroker *stroker,
+                      struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = cb_null_move;
+   cb.point = stroke_emit_point;
+   cb.bezier = stroke_emit_curve;
+   cb.user_data = stroker;
+
+   arc_to_beziers(arc, cb, matrix);
+}
+
+void arc_to_path(struct arc *arc,
+                 struct path *path,
+                 struct matrix *matrix)
+{
+   struct arc_cb cb;
+
+   cb.move = arc_path_move;
+   cb.point = arc_path_point;
+   cb.bezier = arc_path_bezier;
+   cb.user_data = path;
+
+   arc_to_beziers(arc, cb, matrix);
+}
diff --git a/src/gallium/state_trackers/vega/arc.h b/src/gallium/state_trackers/vega/arc.h
new file mode 100644
index 0000000000..3205cd5021
--- /dev/null
+++ b/src/gallium/state_trackers/vega/arc.h
@@ -0,0 +1,80 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ARC_H
+#define ARC_H
+
+#include "VG/openvg.h"
+
+struct polygon;
+struct matrix;
+struct stroker;
+struct path;
+
+struct arc {
+   VGPathSegment type;
+
+   VGfloat cx, cy;
+
+   VGfloat a, b;
+
+   VGfloat theta;
+   VGfloat cos_theta, sin_theta;
+
+   VGfloat eta1;
+   VGfloat eta2;
+
+   VGfloat x1, y1, x2, y2;
+
+   VGboolean is_valid;
+};
+
+void arc_init(struct arc *arc,
+              VGPathSegment type,
+              VGfloat x1, VGfloat y1,
+              VGfloat x2, VGfloat y2,
+              VGfloat rh, VGfloat rv,
+              VGfloat rot);
+
+void arc_add_to_polygon(struct arc *arc,
+                        struct polygon *poly,
+                        struct matrix *matrix);
+
+
+void arc_to_path(struct arc *arc,
+                 struct path *p,
+                 struct matrix *matrix);
+
+void arc_stroke_cb(struct arc *arc,
+                   struct stroker *stroke,
+                   struct matrix *matrix);
+
+void arc_stroker_emit(struct arc *arc,
+                      struct stroker *stroke,
+                      struct matrix *matrix);
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/asm_fill.h b/src/gallium/state_trackers/vega/asm_fill.h
new file mode 100644
index 0000000000..2f394ad6c5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/asm_fill.h
@@ -0,0 +1,246 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ASM_FILL_H
+#define ASM_FILL_H
+
+static const char solid_fill_asm[] =
+   "MOV %s, CONST[0]\n";
+
+
+static const char linear_grad_asm[] =
+   "MOV TEMP[0].xy, IN[0]\n"
+   "MOV TEMP[0].z, CONST[1].yyyy\n"
+   "DP3 TEMP[1], CONST[2], TEMP[0]\n"
+   "DP3 TEMP[2], CONST[3], TEMP[0]\n"
+   "DP3 TEMP[3], CONST[4], TEMP[0]\n"
+   "RCP TEMP[3], TEMP[3]\n"
+   "MUL TEMP[1], TEMP[1], TEMP[3]\n"
+   "MUL TEMP[2], TEMP[2], TEMP[3]\n"
+   "MOV TEMP[4].x, TEMP[1]\n"
+   "MOV TEMP[4].y, TEMP[2]\n"
+   "MUL TEMP[0], CONST[0].yyyy, TEMP[4].yyyy\n"
+   "MAD TEMP[1], CONST[0].xxxx, TEMP[4].xxxx, TEMP[0]\n"
+   "MUL TEMP[2], TEMP[1], CONST[0].zzzz\n"
+   "TEX %s, TEMP[2], SAMP[0], 1D\n";
+
+static const char radial_grad_asm[] =
+   "MOV TEMP[0].xy, IN[0]\n"
+   "MOV TEMP[0].z, CONST[1].yyyy\n"
+   "DP3 TEMP[1], CONST[2], TEMP[0]\n"
+   "DP3 TEMP[2], CONST[3], TEMP[0]\n"
+   "DP3 TEMP[3], CONST[4], TEMP[0]\n"
+   "RCP TEMP[3], TEMP[3]\n"
+   "MUL TEMP[1], TEMP[1], TEMP[3]\n"
+   "MUL TEMP[2], TEMP[2], TEMP[3]\n"
+   "MOV TEMP[5].x, TEMP[1]\n"
+   "MOV TEMP[5].y, TEMP[2]\n"
+   "MUL TEMP[0], CONST[0].yyyy, TEMP[5].yyyy\n"
+   "MAD TEMP[1], CONST[0].xxxx, TEMP[5].xxxx, TEMP[0]\n"
+   "ADD TEMP[1], TEMP[1], TEMP[1]\n"
+   "MUL TEMP[3], TEMP[5].yyyy, TEMP[5].yyyy\n"
+   "MAD TEMP[4], TEMP[5].xxxx, TEMP[5].xxxx, TEMP[3]\n"
+   "MOV TEMP[4], -TEMP[4]\n"
+   "MUL TEMP[2], CONST[0].zzzz, TEMP[4]\n"
+   "MUL TEMP[0], CONST[1].wwww, TEMP[2]\n"
+   "MUL TEMP[3], TEMP[1], TEMP[1]\n"
+   "SUB TEMP[2], TEMP[3], TEMP[0]\n"
+   "RSQ TEMP[2], |TEMP[2]|\n"
+   "RCP TEMP[2], TEMP[2]\n"
+   "SUB TEMP[1], TEMP[2], TEMP[1]\n"
+   "ADD TEMP[0], CONST[0].zzzz, CONST[0].zzzz\n"
+   "RCP TEMP[0], TEMP[0]\n"
+   "MUL TEMP[2], TEMP[1], TEMP[0]\n"
+   "TEX %s, TEMP[2], SAMP[0], 1D\n";
+
+static const char pattern_asm[] =
+   "MOV TEMP[0].xy, IN[0]\n"
+   "MOV TEMP[0].z, CONST[1].yyyy\n"
+   "DP3 TEMP[1], CONST[2], TEMP[0]\n"
+   "DP3 TEMP[2], CONST[3], TEMP[0]\n"
+   "DP3 TEMP[3], CONST[4], TEMP[0]\n"
+   "RCP TEMP[3], TEMP[3]\n"
+   "MUL TEMP[1], TEMP[1], TEMP[3]\n"
+   "MUL TEMP[2], TEMP[2], TEMP[3]\n"
+   "MOV TEMP[4].x, TEMP[1]\n"
+   "MOV TEMP[4].y, TEMP[2]\n"
+   "RCP TEMP[0], CONST[1].zwzw\n"
+   "MOV TEMP[1], TEMP[4]\n"
+   "MUL TEMP[1].x, TEMP[1], TEMP[0]\n"
+   "MUL TEMP[1].y, TEMP[1], TEMP[0]\n"
+   "TEX %s, TEMP[1], SAMP[0], 2D\n";
+
+
+static const char mask_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[1], 2D\n"
+   "MUL TEMP[0].w, TEMP[0].wwww, TEMP[1].wwww\n"
+   "MOV %s, TEMP[0]\n";
+
+
+static const char image_normal_asm[] =
+   "TEX %s, IN[1], SAMP[3], 2D\n";
+
+static const char image_multiply_asm[] =
+   "TEX TEMP[1], IN[1], SAMP[3], 2D\n"
+   "MUL %s, TEMP[0], TEMP[1]\n";
+
+static const char image_stencil_asm[] =
+   "TEX TEMP[1], IN[1], SAMP[3], 2D\n"
+   "MUL %s, TEMP[0], TEMP[1]\n";
+
+
+#define EXTENDED_BLEND_OVER                     \
+   "SUB TEMP[3], CONST[1].yyyy, TEMP[1].wwww\n" \
+   "SUB TEMP[4], CONST[1].yyyy, TEMP[0].wwww\n" \
+   "MUL TEMP[3], TEMP[0], TEMP[3]\n"            \
+   "MUL TEMP[4], TEMP[1], TEMP[4]\n"            \
+   "ADD TEMP[3], TEMP[3], TEMP[4]\n"
+
+static const char blend_multiply_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   EXTENDED_BLEND_OVER
+   "MUL TEMP[4], TEMP[0], TEMP[1]\n"
+   "ADD TEMP[1], TEMP[4], TEMP[3]\n"/*result.rgb*/
+   "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n"
+   "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n"
+   "SUB TEMP[1].w, TEMP[3], TEMP[2]\n"
+   "MOV %s, TEMP[1]\n";
+#if 1
+static const char blend_screen_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   "ADD TEMP[3], TEMP[0], TEMP[1]\n"
+   "MUL TEMP[2], TEMP[0], TEMP[1]\n"
+   "SUB %s, TEMP[3], TEMP[2]\n";
+#else
+static const char blend_screen_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   "MOV %s, TEMP[1]\n";
+#endif
+
+static const char blend_darken_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   EXTENDED_BLEND_OVER
+   "MUL TEMP[4], TEMP[0], TEMP[1].wwww\n"
+   "MUL TEMP[5], TEMP[1], TEMP[0].wwww\n"
+   "MIN TEMP[4], TEMP[4], TEMP[5]\n"
+   "ADD TEMP[1], TEMP[3], TEMP[4]\n"
+   "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n"
+   "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n"
+   "SUB TEMP[1].w, TEMP[3], TEMP[2]\n"
+   "MOV %s, TEMP[1]\n";
+
+static const char blend_lighten_asm[] =
+   "TEX TEMP[1], IN[0], SAMP[2], 2D\n"
+   EXTENDED_BLEND_OVER
+   "MUL TEMP[4], TEMP[0], TEMP[1].wwww\n"
+   "MUL TEMP[5], TEMP[1], TEMP[0].wwww\n"
+   "MAX TEMP[4], TEMP[4], TEMP[5]\n"
+   "ADD TEMP[1], TEMP[3], TEMP[4]\n"
+   "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n"
+   "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n"
+   "SUB TEMP[1].w, TEMP[3], TEMP[2]\n"
+   "MOV %s, TEMP[1]\n";
+
+
+static const char premultiply_asm[] =
+   "MUL TEMP[0].xyz, TEMP[0], TEMP[0].wwww\n";
+
+static const char unpremultiply_asm[] =
+   "TEX TEMP[0], IN[0], SAMP[1], 2D\n";
+
+
+static const char color_bw_asm[] =
+   "ADD TEMP[1], CONST[1].yyyy, CONST[1].yyyy\n"
+   "RCP TEMP[2], TEMP[1]\n"
+   "ADD TEMP[1], CONST[1].yyyy, TEMP[2]\n"
+   "ADD TEMP[2].x, TEMP[0].xxxx, TEMP[0].yyyy\n"
+   "ADD TEMP[2].x, TEMP[0].zzzz, TEMP[0].xxxx\n"
+   "SGE TEMP[0].xyz, TEMP[2].xxxx, TEMP[1]\n"
+   "SGE TEMP[0].w, TEMP[0].wwww, TEMP[2].yyyy\n"
+   "MOV %s, TEMP[0]\n";
+
+
+struct shader_asm_info {
+   VGint id;
+   VGint num_tokens;
+   const char * txt;
+
+   VGboolean needs_position;
+
+   VGint start_const;
+   VGint num_consts;
+
+   VGint start_sampler;
+   VGint num_samplers;
+
+   VGint start_temp;
+   VGint num_temps;
+};
+
+
+static const struct shader_asm_info shaders_asm[] = {
+   /* fills */
+   {VEGA_SOLID_FILL_SHADER,       40,  solid_fill_asm,
+    VG_FALSE, 0, 1, 0, 0, 0, 0},
+   {VEGA_LINEAR_GRADIENT_SHADER, 200,  linear_grad_asm,
+    VG_TRUE,  0, 5, 0, 1, 0, 5},
+   {VEGA_RADIAL_GRADIENT_SHADER, 200,  radial_grad_asm,
+    VG_TRUE,  0, 5, 0, 1, 0, 6},
+   {VEGA_PATTERN_SHADER,         100,      pattern_asm,
+    VG_TRUE,  1, 4, 0, 1, 0, 5},
+
+   /* image draw modes */
+   {VEGA_IMAGE_NORMAL_SHADER,    200, image_normal_asm,
+    VG_TRUE,  0, 0, 3, 1, 0, 0},
+   {VEGA_IMAGE_MULTIPLY_SHADER,  200, image_multiply_asm,
+    VG_TRUE,  0, 0, 3, 1, 0, 2},
+   {VEGA_IMAGE_STENCIL_SHADER,   200, image_stencil_asm,
+    VG_TRUE,  0, 0, 3, 1, 0, 2},
+
+   {VEGA_MASK_SHADER,            100,         mask_asm,
+    VG_TRUE,  0, 0, 1, 1, 0, 2},
+
+   /* extra blend modes */
+   {VEGA_BLEND_MULTIPLY_SHADER,  200, blend_multiply_asm,
+    VG_TRUE,  1, 1, 2, 1, 0, 5},
+   {VEGA_BLEND_SCREEN_SHADER,    200, blend_screen_asm,
+    VG_TRUE,  0, 0, 2, 1, 0, 4},
+   {VEGA_BLEND_DARKEN_SHADER,    200, blend_darken_asm,
+    VG_TRUE,  1, 1, 2, 1, 0, 6},
+   {VEGA_BLEND_LIGHTEN_SHADER,   200, blend_lighten_asm,
+    VG_TRUE,  1, 1, 2, 1, 0, 6},
+
+   /* premultiply */
+   {VEGA_PREMULTIPLY_SHADER,   100, premultiply_asm,
+    VG_FALSE,  0, 0, 0, 0, 0, 1},
+   {VEGA_UNPREMULTIPLY_SHADER,   100, unpremultiply_asm,
+    VG_FALSE,  0, 0, 0, 0, 0, 1},
+
+   /* color transform to black and white */
+   {VEGA_BW_SHADER,   150, color_bw_asm,
+    VG_FALSE,  1, 1, 0, 0, 0, 3},
+};
+#endif
diff --git a/src/gallium/state_trackers/vega/asm_filters.h b/src/gallium/state_trackers/vega/asm_filters.h
new file mode 100644
index 0000000000..49807b9ab4
--- /dev/null
+++ b/src/gallium/state_trackers/vega/asm_filters.h
@@ -0,0 +1,117 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ASM_FILTERS_H
+#define ASM_FILTERS_H
+
+static const char color_matrix_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL CONST[0..4], CONSTANT\n"
+   "DCL TEMP[0..4], CONSTANT\n"
+   "DCL SAMP[0], CONSTANT\n"
+   "TEX TEMP[0], IN[0], SAMP[0], 2D\n"
+   "MOV TEMP[1], TEMP[0].xxxx\n"
+   "MOV TEMP[2], TEMP[0].yyyy\n"
+   "MOV TEMP[3], TEMP[0].zzzz\n"
+   "MOV TEMP[4], TEMP[0].wwww\n"
+   "MUL TEMP[1], TEMP[1], CONST[0]\n"
+   "MUL TEMP[2], TEMP[2], CONST[1]\n"
+   "MUL TEMP[3], TEMP[3], CONST[2]\n"
+   "MUL TEMP[4], TEMP[4], CONST[3]\n"
+   "ADD TEMP[0], TEMP[1], CONST[4]\n"
+   "ADD TEMP[0], TEMP[0], TEMP[2]\n"
+   "ADD TEMP[0], TEMP[0], TEMP[3]\n"
+   "ADD TEMP[0], TEMP[0], TEMP[4]\n"
+   "MOV OUT[0], TEMP[0]\n"
+   "END\n";
+
+static const char convolution_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL TEMP[0..4], CONSTANT\n"
+   "DCL ADDR[0], CONSTANT\n"
+   "DCL CONST[0..%d], CONSTANT\n"
+   "DCL SAMP[0], CONSTANT\n"
+   "0: MOV TEMP[0], CONST[0].xxxx\n"
+   "1: MOV TEMP[1], CONST[0].xxxx\n"
+   "2: BGNLOOP2 :14\n"
+   "3: SGE TEMP[0].z, TEMP[0].yyyy, CONST[1].xxxx\n"
+   "4: IF TEMP[0].zzzz :7\n"
+   "5: BRK\n"
+   "6: ENDIF\n"
+   "7: ARL ADDR[0].x, TEMP[0].yyyy\n"
+   "8: MOV TEMP[3], CONST[ADDR[0]+2]\n"
+   "9: ADD TEMP[4].xy, IN[0], TEMP[3]\n"
+   "10: TEX TEMP[2], TEMP[4], SAMP[0], 2D\n"
+   "11: MOV TEMP[3], CONST[ADDR[0]+%d]\n"
+   "12: MAD TEMP[1], TEMP[2], TEMP[3], TEMP[1]\n"
+   "13: ADD TEMP[0].y, TEMP[0].yyyy, CONST[0].yyyy\n"
+   "14: ENDLOOP2 :2\n"
+   "15: MAD OUT[0], TEMP[1], CONST[1].yyyy, CONST[1].zzzz\n"
+   "16: END\n";
+
+
+static const char lookup_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL TEMP[0..2], CONSTANT\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "TEX TEMP[0], IN[0], SAMP[0], 2D\n"
+   "MOV TEMP[1], TEMP[0]\n"
+   /* do red */
+   "TEX TEMP[2], TEMP[1].xxxx, SAMP[1], 1D\n"
+   "MOV TEMP[0].x, TEMP[2].xxxx\n"
+   /* do blue */
+   "TEX TEMP[2], TEMP[1].yyyy, SAMP[1], 1D\n"
+   "MOV TEMP[0].y, TEMP[2].yyyy\n"
+   /* do green */
+   "TEX TEMP[2], TEMP[1].zzzz, SAMP[1], 1D\n"
+   "MOV TEMP[0].z, TEMP[2].zzzz\n"
+   /* do alpha */
+   "TEX TEMP[2], TEMP[1].wwww, SAMP[1], 1D\n"
+   "MOV TEMP[0].w, TEMP[2].wwww\n"
+   "MOV OUT[0], TEMP[0]\n"
+   "END\n";
+
+
+static const char lookup_single_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL TEMP[0..2], CONSTANT\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "TEX TEMP[0], IN[0], SAMP[0], 2D\n"
+   "TEX TEMP[1], TEMP[0].%s, SAMP[1], 1D\n"
+   "MOV OUT[0], TEMP[1]\n"
+   "END\n";
+
+#endif
diff --git a/src/gallium/state_trackers/vega/asm_util.h b/src/gallium/state_trackers/vega/asm_util.h
new file mode 100644
index 0000000000..218e1d166d
--- /dev/null
+++ b/src/gallium/state_trackers/vega/asm_util.h
@@ -0,0 +1,136 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ASM_UTIL_H
+#define ASM_UTIL_H
+
+
+static const char pass_through_depth_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], POSITION, LINEAR\n"
+   "DCL OUT[0].z, POSITION, CONSTANT\n"
+   "0: MOV OUT[0].z, IN[0].zzzz\n"
+   "1: END\n";
+
+
+
+/* μnew = μmask */
+static const char set_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL SAMP[0], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX OUT[0], IN[0], SAMP[0], 2D\n"/*umask*/
+   "1: END\n";
+
+/* μnew = 1 – (1 – μmask)*(1 – μprev) */
+static const char union_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL IN[1], POSITION, LINEAR\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "DCL TEMP[0..3], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/
+   "1: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/
+   "2: SUB TEMP[2], CONST[0], TEMP[0]\n"
+   "3: SUB TEMP[3], CONST[0], TEMP[1]\n"
+   "4: MUL TEMP[0].w, TEMP[2].wwww, TEMP[3].wwww\n"
+   "5: SUB OUT[0], CONST[0], TEMP[0]\n"
+   "6: END\n";
+
+/* μnew = μmask *μprev */
+static const char intersect_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL IN[1], POSITION, LINEAR\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "DCL TEMP[0..1], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/
+   "1: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/
+   "2: MUL OUT[0], TEMP[0].wwww, TEMP[1].wwww\n"
+   "3: END\n";
+
+/* μnew = μprev*(1 – μmask) */
+static const char subtract_mask_asm[] =
+   "FRAG1.1\n"
+   "DCL IN[0], GENERIC[0], PERSPECTIVE\n"
+   "DCL IN[1], POSITION, LINEAR\n"
+   "DCL CONST[0], CONSTANT\n"
+   "DCL SAMP[0..1], CONSTANT\n"
+   "DCL TEMP[0..2], CONSTANT\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "0: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/
+   "1: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/
+   "2: SUB TEMP[2], CONST[0], TEMP[1]\n"
+   "3: MUL OUT[0], TEMP[2].wwww, TEMP[0].wwww\n"
+   "4: END\n";
+
+
+static const char vs_plain_asm[] =
+   "VERT1.1\n"
+   "DCL IN[0]\n"
+   "DCL OUT[0], POSITION\n"
+   "DCL TEMP[0]\n"
+   "DCL CONST[0..1]\n"
+   "0: MUL TEMP[0], IN[0], CONST[0]\n"
+   "1: ADD TEMP[0], TEMP[0], CONST[1]\n"
+   "2: MOV OUT[0], TEMP[0]\n"
+   "3: END\n";
+
+static const char vs_clear_asm[] =
+   "VERT1.1\n"
+   "DCL IN[0]\n"
+   "DCL IN[1]\n"
+   "DCL OUT[0], POSITION\n"
+   "DCL OUT[1], COLOR\n"
+   "DCL TEMP[0]\n"
+   "DCL CONST[0..1]\n"
+   "0: MUL TEMP[0], IN[0], CONST[0]\n"
+   "1: ADD TEMP[0], TEMP[0], CONST[1]\n"
+   "2: MOV OUT[0], TEMP[0]\n"
+   "3: MOV OUT[1], IN[1]\n"
+   "4: END\n";
+
+
+static const char vs_texture_asm[] =
+   "VERT1.1\n"
+   "DCL IN[0]\n"
+   "DCL IN[1]\n"
+   "DCL OUT[0], POSITION\n"
+   "DCL OUT[1], GENERIC\n"
+   "DCL TEMP[0]\n"
+   "DCL CONST[0..1]\n"
+   "0: MUL TEMP[0], IN[0], CONST[0]\n"
+   "1: ADD TEMP[0], TEMP[0], CONST[1]\n"
+   "2: MOV OUT[0], TEMP[0]\n"
+   "3: MOV OUT[1], IN[1]\n"
+   "4: END\n";
+
+#endif
diff --git a/src/gallium/state_trackers/vega/bezier.c b/src/gallium/state_trackers/vega/bezier.c
new file mode 100644
index 0000000000..39a7ade016
--- /dev/null
+++ b/src/gallium/state_trackers/vega/bezier.c
@@ -0,0 +1,704 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "bezier.h"
+
+#include "matrix.h"
+#include "polygon.h"
+
+#include "pipe/p_compiler.h"
+#include "util/u_debug.h"
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <assert.h>
+#include <math.h>
+
+static const float flatness = 0.5;
+
+
+static INLINE void split_left(struct bezier *bez, VGfloat t, struct bezier* left)
+{
+    left->x1 = bez->x1;
+    left->y1 = bez->y1;
+
+    left->x2 = bez->x1 + t * (bez->x2 - bez->x1);
+    left->y2 = bez->y1 + t * (bez->y2 - bez->y1);
+
+    left->x3 = bez->x2 + t * (bez->x3 - bez->x2);
+    left->y3 = bez->y2 + t * (bez->y3 - bez->y2);
+
+    bez->x3 = bez->x3 + t * (bez->x4 - bez->x3);
+    bez->y3 = bez->y3 + t * (bez->y4 - bez->y3);
+
+    bez->x2 = left->x3 + t * (bez->x3 - left->x3);
+    bez->y2 = left->y3 + t * (bez->y3 - left->y3);
+
+    left->x3 = left->x2 + t * (left->x3 - left->x2);
+    left->y3 = left->y2 + t * (left->y3 - left->y2);
+
+    left->x4 = bez->x1 = left->x3 + t * (bez->x2 - left->x3);
+    left->y4 = bez->y1 = left->y3 + t * (bez->y2 - left->y3);
+}
+
+static INLINE void split(struct bezier *bez,
+                         struct bezier *first_half,
+                         struct bezier *second_half)
+{
+   double c         = (bez->x2 + bez->x3) * 0.5;
+   first_half->x2  = (bez->x1 + bez->x2) * 0.5;
+   second_half->x3 = (bez->x3 + bez->x4) * 0.5;
+   first_half->x1  = bez->x1;
+   second_half->x4 = bez->x4;
+   first_half->x3  = (first_half->x2 + c) * 0.5;
+   second_half->x2 = (second_half->x3 + c) * 0.5;
+   first_half->x4  = second_half->x1 =
+                     (first_half->x3 + second_half->x2) * 0.5;
+
+   c = (bez->y2 + bez->y3) / 2;
+   first_half->y2  = (bez->y1 + bez->y2) * 0.5;
+   second_half->y3 = (bez->y3 + bez->y4) * 0.5;
+   first_half->y1  = bez->y1;
+   second_half->y4 = bez->y4;
+   first_half->y3  = (first_half->y2 + c) * 0.5;
+   second_half->y2 = (second_half->y3 + c) * 0.5;
+   first_half->y4  = second_half->y1 =
+                     (first_half->y3 + second_half->y2) * 0.5;
+}
+
+struct polygon * bezier_to_polygon(struct bezier *bez)
+{
+   struct polygon *poly = polygon_create(64);
+   polygon_vertex_append(poly, bez->x1, bez->y1);
+   bezier_add_to_polygon(bez, poly);
+   return poly;
+}
+
+void bezier_add_to_polygon(const struct bezier *bez,
+                           struct polygon *poly)
+{
+   struct bezier beziers[32];
+   struct bezier *b;
+
+   beziers[0] = *bez;
+   b = beziers;
+
+   while (b >= beziers) {
+      double y4y1 = b->y4 - b->y1;
+      double x4x1 = b->x4 - b->x1;
+      double l = ABS(x4x1) + ABS(y4y1);
+      double d;
+      if (l > 1.f) {
+         d = ABS((x4x1)*(b->y1 - b->y2) - (y4y1)*(b->x1 - b->x2))
+             + ABS((x4x1)*(b->y1 - b->y3) - (y4y1)*(b->x1 - b->x3));
+      } else {
+         d = ABS(b->x1 - b->x2) + ABS(b->y1 - b->y2) +
+             ABS(b->x1 - b->x3) + ABS(b->y1 - b->y3);
+         l = 1.;
+      }
+      if (d < flatness*l || b == beziers + 31) {
+         /* good enough, we pop it off and add the endpoint */
+         polygon_vertex_append(poly, b->x4, b->y4);
+         --b;
+      } else {
+         /* split, second half of the bezier goes lower into the stack */
+         split(b, b+1, b);
+         ++b;
+      }
+   }
+}
+
+static void add_if_close(struct bezier *bez, VGfloat *length, VGfloat error)
+{
+   struct bezier left, right;     /* bez poly splits */
+   VGfloat len = 0.0;        /* arc length */
+   VGfloat chord;            /* chord length */
+
+   len = len + line_length(bez->x1, bez->y1, bez->x2, bez->y2);
+   len = len + line_length(bez->x2, bez->y2, bez->x3, bez->y3);
+   len = len + line_length(bez->x3, bez->y3, bez->x4, bez->y4);
+
+   chord = line_length(bez->x1, bez->y1, bez->x4, bez->y4);
+
+   if ((len-chord) > error) {
+      split(bez, &left, &right);                 /* split in two */
+      add_if_close(&left, length, error);       /* try left side */
+      add_if_close(&right, length, error);      /* try right side */
+      return;
+   }
+
+   *length = *length + len;
+
+   return;
+}
+
+float bezier_length(struct bezier *bez, float error)
+{
+   VGfloat length = 0.f;
+
+   add_if_close(bez, &length, error);
+   return length;
+}
+
+void bezier_init(struct bezier *bez,
+                 float x1, float y1,
+                 float x2, float y2,
+                 float x3, float y3,
+                 float x4, float y4)
+{
+   bez->x1 = x1;
+   bez->y1 = y1;
+   bez->x2 = x2;
+   bez->y2 = y2;
+   bez->x3 = x3;
+   bez->y3 = y3;
+   bez->x4 = x4;
+   bez->y4 = y4;
+#if 0
+   debug_printf("bezier in [%f, %f, %f, %f, %f, %f]\n",
+                x1, y1, x2, y2, x3, y3, x4, y4);
+#endif
+}
+
+
+static INLINE void bezier_init2v(struct bezier *bez,
+                                 float *pt1,
+                                 float *pt2,
+                                 float *pt3,
+                                 float *pt4)
+{
+   bez->x1 = pt1[0];
+   bez->y1 = pt1[1];
+
+   bez->x2 = pt2[0];
+   bez->y2 = pt2[1];
+
+   bez->x3 = pt3[0];
+   bez->y3 = pt3[1];
+
+   bez->x4 = pt4[0];
+   bez->y4 = pt4[1];
+}
+
+
+void bezier_transform(struct bezier *bez,
+                      struct matrix *matrix)
+{
+   assert(matrix_is_affine(matrix));
+   matrix_map_point(matrix, bez->x1, bez->y1, &bez->x1, &bez->y1);
+   matrix_map_point(matrix, bez->x2, bez->y2, &bez->x2, &bez->y2);
+   matrix_map_point(matrix, bez->x3, bez->y3, &bez->x3, &bez->y3);
+   matrix_map_point(matrix, bez->x4, bez->y4, &bez->x4, &bez->y4);
+}
+
+static INLINE void bezier_point_at(const struct bezier *bez, float t, float *pt)
+{
+   float a, b, c, d;
+   float m_t;
+   m_t = 1. - t;
+   b = m_t * m_t;
+   c = t * t;
+   d = c * t;
+   a = b * m_t;
+   b *= 3. * t;
+   c *= 3. * m_t;
+   pt[0] = a*bez->x1 + b*bez->x2 + c*bez->x3 + d*bez->x4;
+   pt[1] = a*bez->y1 + b*bez->y2 + c*bez->y3 + d*bez->y4;
+}
+
+static INLINE void bezier_normal_at(const struct bezier *bez, float t, float *norm)
+{
+   float m_t = 1. - t;
+   float a = m_t * m_t;
+   float b = t * m_t;
+   float c = t * t;
+
+   norm[0] =  (bez->y2-bez->y1) * a + (bez->y3-bez->y2) * b + (bez->y4-bez->y3) * c;
+   norm[1] = -(bez->x2-bez->x1) * a - (bez->x3-bez->x2) * b - (bez->x4-bez->x3) * c;
+}
+
+enum shift_result {
+   Ok,
+   Discard,
+   Split,
+   Circle
+};
+
+static enum shift_result good_offset(const struct bezier *b1,
+                                     const struct bezier *b2,
+                                     float offset, float threshold)
+{
+   const float o2 = offset*offset;
+   const float max_dist_line = threshold*offset*offset;
+   const float max_dist_normal = threshold*offset;
+   const float spacing = 0.25;
+   for (float i = spacing; i < 0.99; i += spacing) {
+      float p1[2],p2[2], d, l;
+      float normal[2];
+      bezier_point_at(b1, i, p1);
+      bezier_point_at(b2, i, p2);
+      d = (p1[0] - p2[0])*(p1[0] - p2[0]) + (p1[1] - p2[1])*(p1[1] - p2[1]);
+      if (ABS(d - o2) > max_dist_line)
+         return Split;
+
+      bezier_normal_at(b1, i, normal);
+      l = ABS(normal[0]) + ABS(normal[1]);
+      if (l != 0.) {
+         d = ABS(normal[0]*(p1[1] - p2[1]) - normal[1]*(p1[0] - p2[0]) ) / l;
+         if (d > max_dist_normal)
+            return Split;
+      }
+   }
+   return Ok;
+}
+
+static INLINE void shift_line_by_normal(float *l, float offset)
+{
+   float norm[4];
+   float tx, ty;
+
+   line_normal(l, norm);
+   line_normalize(norm);
+
+   tx = (norm[2] - norm[0]) * offset;
+   ty = (norm[3] - norm[1]) * offset;
+   l[0] += tx; l[1] += ty;
+   l[2] += tx; l[3] += ty;
+}
+
+static INLINE VGboolean is_bezier_line(float (*points)[2], int count)
+{
+   float dx13 = points[2][0] - points[0][0];
+   float dy13 = points[2][1] - points[0][1];
+
+   float dx12 = points[1][0] - points[0][0];
+   float dy12 = points[1][1] - points[0][1];
+
+   debug_assert(count > 2);
+
+   if (count == 3) {
+      return floatsEqual(dx12 * dy13, dx13 * dy12);
+   } else if (count == 4) {
+      float dx14 = points[3][0] - points[0][0];
+      float dy14 = points[3][1] - points[0][1];
+
+      return (floatsEqual(dx12 * dy13, dx13 * dy12) &&
+              floatsEqual(dx12 * dy14, dx14 * dy12));
+   }
+
+   return VG_FALSE;
+}
+
+static INLINE void compute_pt_normal(float *pt1, float *pt2, float *res)
+{
+   float line[4];
+   float normal[4];
+   line[0] = 0.f; line[1] = 0.f;
+   line[2] = pt2[0] - pt1[0];
+   line[3] = pt2[1] - pt1[1];
+   line_normal(line, normal);
+   line_normalize(normal);
+
+   res[0] = normal[2];
+   res[1] = normal[3];
+}
+
+static enum shift_result shift(const struct bezier *orig,
+                               struct bezier *shifted,
+                               float offset, float threshold)
+{
+   int map[4];
+   VGboolean p1_p2_equal = (orig->x1 == orig->x2 && orig->y1 == orig->y2);
+   VGboolean p2_p3_equal = (orig->x2 == orig->x3 && orig->y2 == orig->y3);
+   VGboolean p3_p4_equal = (orig->x3 == orig->x4 && orig->y3 == orig->y4);
+
+   float points[4][2];
+   int np = 0;
+   float bounds[4];
+   float points_shifted[4][2];
+   float prev_normal[2];
+
+   points[np][0] = orig->x1;
+   points[np][1] = orig->y1;
+   map[0] = 0;
+   ++np;
+   if (!p1_p2_equal) {
+      points[np][0] = orig->x2;
+      points[np][1] = orig->y2;
+      ++np;
+   }
+   map[1] = np - 1;
+   if (!p2_p3_equal) {
+      points[np][0] = orig->x3;
+      points[np][1] = orig->y3;
+      ++np;
+   }
+   map[2] = np - 1;
+   if (!p3_p4_equal) {
+      points[np][0] = orig->x4;
+      points[np][1] = orig->y4;
+      ++np;
+   }
+   map[3] = np - 1;
+   if (np == 1)
+      return Discard;
+
+   /* We need to specialcase lines of 3 or 4 points due to numerical
+      instability in intersection code below */
+   if (np > 2 && is_bezier_line(points, np)) {
+      float l[4] = { points[0][0], points[0][1],
+                     points[np-1][0], points[np-1][1] };
+      float ctrl1[2], ctrl2[2];
+      if (floatsEqual(points[0][0], points[np-1][0]) &&
+          floatsEqual(points[0][1], points[np-1][1]))
+         return Discard;
+
+      shift_line_by_normal(l, offset);
+      line_point_at(l, 0.33, ctrl1);
+      line_point_at(l, 0.66, ctrl2);
+      bezier_init(shifted, l[0], l[1],
+                  ctrl1[0], ctrl1[1], ctrl2[0], ctrl2[1],
+                  l[2], l[3]);
+      return Ok;
+   }
+
+   bezier_bounds(orig, bounds);
+   if (np == 4 && bounds[2] < .1*offset && bounds[3] < .1*offset) {
+      float l = (orig->x1 - orig->x2)*(orig->x1 - orig->x2) +
+                (orig->y1 - orig->y2)*(orig->y1 - orig->y1) *
+                (orig->x3 - orig->x4)*(orig->x3 - orig->x4) +
+                (orig->y3 - orig->y4)*(orig->y3 - orig->y4);
+      float dot = (orig->x1 - orig->x2)*(orig->x3 - orig->x4) +
+                  (orig->y1 - orig->y2)*(orig->y3 - orig->y4);
+      if (dot < 0 && dot*dot < 0.8*l)
+         /* the points are close and reverse dirction. Approximate the whole
+            thing by a semi circle */
+         return Circle;
+   }
+
+   compute_pt_normal(points[0], points[1], prev_normal);
+
+   points_shifted[0][0] = points[0][0] + offset * prev_normal[0];
+   points_shifted[0][1] = points[0][1] + offset * prev_normal[1];
+
+   for (int i = 1; i < np - 1; ++i) {
+      float normal_sum[2], r;
+      float next_normal[2];
+      compute_pt_normal(points[i], points[i + 1], next_normal);
+
+      normal_sum[0] = prev_normal[0] + next_normal[0];
+      normal_sum[1] = prev_normal[1] + next_normal[1];
+
+      r = 1.0 + prev_normal[0] * next_normal[0]
+          + prev_normal[1] * next_normal[1];
+
+      if (floatsEqual(r + 1, 1)) {
+         points_shifted[i][0] = points[i][0] + offset * prev_normal[0];
+         points_shifted[i][1] = points[i][1] + offset * prev_normal[1];
+      } else {
+         float k = offset / r;
+         points_shifted[i][0] = points[i][0] + k * normal_sum[0];
+         points_shifted[i][1] = points[i][1] + k * normal_sum[1];
+      }
+
+      prev_normal[0] = next_normal[0];
+      prev_normal[1] = next_normal[1];
+   }
+
+   points_shifted[np - 1][0] = points[np - 1][0] + offset * prev_normal[0];
+   points_shifted[np - 1][1] = points[np - 1][1] + offset * prev_normal[1];
+
+   bezier_init2v(shifted,
+                 points_shifted[map[0]], points_shifted[map[1]],
+                 points_shifted[map[2]], points_shifted[map[3]]);
+
+   return good_offset(orig, shifted, offset, threshold);
+}
+
+static VGboolean make_circle(const struct bezier *b, float offset, struct bezier *o)
+{
+   float normals[3][2];
+   float dist;
+   float angles[2];
+   float sign = 1.f;
+   int i;
+   float circle[3][2];
+
+   normals[0][0] = b->y2 - b->y1;
+   normals[0][1] = b->x1 - b->x2;
+   dist = sqrt(normals[0][0]*normals[0][0] + normals[0][1]*normals[0][1]);
+   if (floatsEqual(dist + 1, 1.f))
+      return VG_FALSE;
+   normals[0][0] /= dist;
+   normals[0][1] /= dist;
+
+   normals[2][0] = b->y4 - b->y3;
+   normals[2][1] = b->x3 - b->x4;
+   dist = sqrt(normals[2][0]*normals[2][0] + normals[2][1]*normals[2][1]);
+   if (floatsEqual(dist + 1, 1.f))
+      return VG_FALSE;
+   normals[2][0] /= dist;
+   normals[2][1] /= dist;
+
+   normals[1][0] = b->x1 - b->x2 - b->x3 + b->x4;
+   normals[1][1] = b->y1 - b->y2 - b->y3 + b->y4;
+   dist = -1*sqrt(normals[1][0]*normals[1][0] + normals[1][1]*normals[1][1]);
+   normals[1][0] /= dist;
+   normals[1][1] /= dist;
+
+   for (i = 0; i < 2; ++i) {
+      float cos_a = normals[i][0]*normals[i+1][0] + normals[i][1]*normals[i+1][1];
+      if (cos_a > 1.)
+         cos_a = 1.;
+      if (cos_a < -1.)
+         cos_a = -1;
+      angles[i] = acos(cos_a)/M_PI;
+   }
+
+   if (angles[0] + angles[1] > 1.) {
+      /* more than 180 degrees */
+      normals[1][0] = -normals[1][0];
+      normals[1][1] = -normals[1][1];
+      angles[0] = 1. - angles[0];
+      angles[1] = 1. - angles[1];
+      sign = -1.;
+   }
+
+   circle[0][0] = b->x1 + normals[0][0]*offset;
+   circle[0][1] = b->y1 + normals[0][1]*offset;
+
+   circle[1][0] = 0.5*(b->x1 + b->x4) + normals[1][0]*offset;
+   circle[1][1] = 0.5*(b->y1 + b->y4) + normals[1][1]*offset;
+
+   circle[2][0] = b->x4 + normals[2][0]*offset;
+   circle[2][1] = b->y4 + normals[2][1]*offset;
+
+   for (i = 0; i < 2; ++i) {
+      float kappa = 2.*KAPPA * sign * offset * angles[i];
+
+      o->x1 = circle[i][0];
+      o->y1 = circle[i][1];
+      o->x2 = circle[i][0] - normals[i][1]*kappa;
+      o->y2 = circle[i][1] + normals[i][0]*kappa;
+      o->x3 = circle[i+1][0] + normals[i+1][1]*kappa;
+      o->y3 = circle[i+1][1] - normals[i+1][0]*kappa;
+      o->x4 = circle[i+1][0];
+      o->y4 = circle[i+1][1];
+
+      ++o;
+   }
+   return VG_TRUE;
+}
+
+int bezier_translate_by_normal(struct bezier *bez,
+                               struct bezier *curves,
+                               int max_curves,
+                               float normal_len,
+                               float threshold)
+{
+   struct bezier beziers[10];
+   struct bezier *b, *o;
+
+   /* fixme: this should really be floatsEqual */
+   if (bez->x1 == bez->x2 && bez->x1 == bez->x3 && bez->x1 == bez->x4 &&
+       bez->y1 == bez->y2 && bez->y1 == bez->y3 && bez->y1 == bez->y4)
+      return 0;
+
+   --max_curves;
+redo:
+   beziers[0] = *bez;
+   b = beziers;
+   o = curves;
+
+   while (b >= beziers) {
+      int stack_segments = b - beziers + 1;
+      enum shift_result res;
+      if ((stack_segments == 10) || (o - curves == max_curves - stack_segments)) {
+         threshold *= 1.5;
+         if (threshold > 2.)
+            goto give_up;
+         goto redo;
+      }
+      res = shift(b, o, normal_len, threshold);
+      if (res == Discard) {
+         --b;
+      } else if (res == Ok) {
+         ++o;
+         --b;
+         continue;
+      } else if (res == Circle && max_curves - (o - curves) >= 2) {
+         /* add semi circle */
+         if (make_circle(b, normal_len, o))
+            o += 2;
+         --b;
+      } else {
+         split(b, b+1, b);
+         ++b;
+      }
+   }
+
+give_up:
+   while (b >= beziers) {
+      enum shift_result res = shift(b, o, normal_len, threshold);
+
+      /* if res isn't Ok or Split then *o is undefined */
+      if (res == Ok || res == Split)
+         ++o;
+
+      --b;
+   }
+
+   debug_assert(o - curves <= max_curves);
+   return o - curves;
+}
+
+void bezier_bounds(const struct bezier *bez,
+                   float *bounds/*x/y/width/height*/)
+{
+   float xmin = bez->x1;
+   float xmax = bez->x1;
+   float ymin = bez->y1;
+   float ymax = bez->y1;
+
+   if (bez->x2 < xmin)
+      xmin = bez->x2;
+   else if (bez->x2 > xmax)
+      xmax = bez->x2;
+   if (bez->x3 < xmin)
+      xmin = bez->x3;
+   else if (bez->x3 > xmax)
+      xmax = bez->x3;
+   if (bez->x4 < xmin)
+      xmin = bez->x4;
+   else if (bez->x4 > xmax)
+      xmax = bez->x4;
+
+   if (bez->y2 < ymin)
+      ymin = bez->y2;
+   else if (bez->y2 > ymax)
+      ymax = bez->y2;
+   if (bez->y3 < ymin)
+      ymin = bez->y3;
+   else if (bez->y3 > ymax)
+      ymax = bez->y3;
+   if (bez->y4 < ymin)
+      ymin = bez->y4;
+   else if (bez->y4 > ymax)
+      ymax = bez->y4;
+
+   bounds[0] = xmin; /* x */
+   bounds[1] = ymin; /* y */
+   bounds[2] = xmax - xmin; /* width */
+   bounds[3] = ymax - ymin; /* height */
+}
+
+void bezier_start_tangent(const struct bezier *bez,
+                          float *tangent)
+{
+   tangent[0] = bez->x1;
+   tangent[1] = bez->y1;
+   tangent[2] = bez->x2;
+   tangent[3] = bez->y2;
+
+   if (null_line(tangent)) {
+      tangent[0] = bez->x1;
+      tangent[1] = bez->y1;
+      tangent[2] = bez->x3;
+      tangent[3] = bez->y3;
+   }
+   if (null_line(tangent)) {
+      tangent[0] = bez->x1;
+      tangent[1] = bez->y1;
+      tangent[2] = bez->x4;
+      tangent[3] = bez->y4;
+   }
+}
+
+
+static INLINE VGfloat bezier_t_at_length(struct bezier *bez,
+                                         VGfloat at_length,
+                                         VGfloat error)
+{
+   VGfloat len = bezier_length(bez, error);
+   VGfloat t   = 1.0;
+   VGfloat last_bigger = 1.;
+
+   if (at_length > len || floatsEqual(at_length, len))
+      return t;
+
+   if (floatIsZero(at_length))
+      return 0.f;
+
+   t *= 0.5;
+   while (1) {
+      struct bezier right = *bez;
+      struct bezier left;
+      VGfloat tmp_len;
+      split_left(&right, t, &left);
+      tmp_len = bezier_length(&left, error);
+      if (ABS(tmp_len - at_length) < error)
+         break;
+
+      if (tmp_len < at_length) {
+         t += (last_bigger - t)*.5;
+      } else {
+         last_bigger = t;
+         t -= t*.5;
+      }
+   }
+   return t;
+}
+
+void bezier_point_at_length(struct bezier *bez,
+                            float length,
+                            float *point,
+                            float *normal)
+{
+   /* ~0.000001 seems to be required to pass G2080x tests */
+   VGfloat t = bezier_t_at_length(bez, length, 0.000001);
+   bezier_point_at(bez, t, point);
+   bezier_normal_at(bez, t, normal);
+   vector_unit(normal);
+}
+
+void bezier_point_at_t(struct bezier *bez, float t,
+                       float *point, float *normal)
+{
+   bezier_point_at(bez, t, point);
+   bezier_normal_at(bez, t, normal);
+   vector_unit(normal);
+}
+
+void bezier_exact_bounds(const struct bezier *bez,
+                         float *bounds/*x/y/width/height*/)
+{
+   struct polygon *poly = polygon_create(64);
+   polygon_vertex_append(poly, bez->x1, bez->y1);
+   bezier_add_to_polygon(bez, poly);
+   polygon_bounding_rect(poly, bounds);
+   polygon_destroy(poly);
+}
+
diff --git a/src/gallium/state_trackers/vega/bezier.h b/src/gallium/state_trackers/vega/bezier.h
new file mode 100644
index 0000000000..e06666051c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/bezier.h
@@ -0,0 +1,81 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef BEZIER_H
+#define BEZIER_H
+
+struct polygon;
+struct matrix;
+
+struct bezier {
+    float x1, y1;
+    float x2, y2;
+    float x3, y3;
+    float x4, y4;
+};
+
+
+#define BEZIER_DEFAULT_ERROR 0.01
+
+/* kappa as being l of a circle with r = 1, we can emulate any
+ * circle of radius r by using the formula
+ * l = r . kappa
+ * More at:
+ * http://www.whizkidtech.redprince.net/bezier/circle/ */
+#define KAPPA 0.5522847498
+
+void bezier_init(struct bezier *bez,
+                 float x1, float y1,
+                 float x2, float y2,
+                 float x3, float y3,
+                 float x4, float y4);
+
+struct polygon *bezier_to_polygon(struct bezier *bez);
+void bezier_add_to_polygon(const struct bezier *bez,
+                           struct polygon *poly);
+float bezier_length(struct bezier *bez, float error);
+void bezier_transform(struct bezier *bez,
+                      struct matrix *mat);
+
+int bezier_translate_by_normal(struct bezier *b,
+                               struct bezier *curves,
+                               int max_curves,
+                               float normal_len,
+                               float threshold);
+void bezier_bounds(const struct bezier *bez,
+                   float *bounds/*x/y/width/height*/);
+void bezier_exact_bounds(const struct bezier *bez,
+                         float *bounds/*x/y/width/height*/);
+
+void bezier_start_tangent(const struct bezier *bez,
+                          float *tangent);
+
+void bezier_point_at_length(struct bezier *bez, float length,
+                            float *point, float *normal);
+void bezier_point_at_t(struct bezier *bez, float t,
+                       float *point, float *normal);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/image.c b/src/gallium/state_trackers/vega/image.c
new file mode 100644
index 0000000000..9a722980d5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/image.c
@@ -0,0 +1,654 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "image.h"
+
+#include "vg_translate.h"
+#include "vg_context.h"
+#include "matrix.h"
+#include "renderer.h"
+#include "util_array.h"
+#include "api_consts.h"
+#include "shaders_cache.h"
+#include "shader.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_blit.h"
+#include "util/u_tile.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+static enum pipe_format vg_format_to_pipe(VGImageFormat format)
+{
+   switch(format) {
+   case VG_sRGB_565:
+      return PIPE_FORMAT_R5G6B5_UNORM;
+   case VG_sRGBA_5551:
+      return PIPE_FORMAT_A1R5G5B5_UNORM;
+   case VG_sRGBA_4444:
+      return PIPE_FORMAT_A4R4G4B4_UNORM;
+   case VG_sL_8:
+   case VG_lL_8:
+      return PIPE_FORMAT_L8_UNORM;
+   case VG_BW_1:
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   case VG_A_8:
+      return PIPE_FORMAT_A8_UNORM;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1:
+   case VG_A_4:
+      return PIPE_FORMAT_A8_UNORM;
+#endif
+   default:
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   }
+}
+
+static INLINE void vg_sync_size(VGfloat *src_loc, VGfloat *dst_loc)
+{
+   src_loc[2] = MIN2(src_loc[2], dst_loc[2]);
+   src_loc[3] = MIN2(src_loc[3], dst_loc[3]);
+   dst_loc[2] = src_loc[2];
+   dst_loc[3] = src_loc[3];
+}
+
+
+static void vg_copy_texture(struct vg_context *ctx,
+                            struct pipe_texture *dst, VGint dx, VGint dy,
+                            struct pipe_texture *src, VGint sx, VGint sy,
+                            VGint width, VGint height)
+{
+   VGfloat dst_loc[4], src_loc[4];
+   VGfloat dst_bounds[4], src_bounds[4];
+   VGfloat src_shift[4], dst_shift[4], shift[4];
+
+   dst_loc[0] = dx;
+   dst_loc[1] = dy;
+   dst_loc[2] = width;
+   dst_loc[3] = height;
+   dst_bounds[0] = 0.f;
+   dst_bounds[1] = 0.f;
+   dst_bounds[2] = dst->width[0];
+   dst_bounds[3] = dst->height[0];
+
+   src_loc[0] = sx;
+   src_loc[1] = sy;
+   src_loc[2] = width;
+   src_loc[3] = height;
+   src_bounds[0] = 0.f;
+   src_bounds[1] = 0.f;
+   src_bounds[2] = src->width[0];
+   src_bounds[3] = src->height[0];
+
+   vg_bound_rect(src_loc, src_bounds, src_shift);
+   vg_bound_rect(dst_loc, dst_bounds, dst_shift);
+   shift[0] = src_shift[0] - dst_shift[0];
+   shift[1] = src_shift[1] - dst_shift[1];
+
+   if (shift[0] < 0)
+      vg_shift_rectx(src_loc, src_bounds, -shift[0]);
+   else
+      vg_shift_rectx(dst_loc, dst_bounds, shift[0]);
+
+   if (shift[1] < 0)
+      vg_shift_recty(src_loc, src_bounds, -shift[1]);
+   else
+      vg_shift_recty(dst_loc, dst_bounds, shift[1]);
+
+   vg_sync_size(src_loc, dst_loc);
+
+   if (src_loc[2] >= 0 && src_loc[3] >= 0 &&
+       dst_loc[2] >= 0 && dst_loc[3] >= 0) {
+      renderer_copy_texture(ctx->renderer,
+                            src,
+                            src_loc[0],
+                            src_loc[1] + src_loc[3],
+                            src_loc[0] + src_loc[2],
+                            src_loc[1],
+                            dst,
+                            dst_loc[0],
+                            dst_loc[1] + dst_loc[3],
+                            dst_loc[0] + dst_loc[2],
+                            dst_loc[1]);
+   }
+
+}
+
+void vg_copy_surface(struct vg_context *ctx,
+                     struct pipe_surface *dst, VGint dx, VGint dy,
+                     struct pipe_surface *src, VGint sx, VGint sy,
+                     VGint width, VGint height)
+{
+   VGfloat dst_loc[4], src_loc[4];
+   VGfloat dst_bounds[4], src_bounds[4];
+   VGfloat src_shift[4], dst_shift[4], shift[4];
+
+   dst_loc[0] = dx;
+   dst_loc[1] = dy;
+   dst_loc[2] = width;
+   dst_loc[3] = height;
+   dst_bounds[0] = 0.f;
+   dst_bounds[1] = 0.f;
+   dst_bounds[2] = dst->width;
+   dst_bounds[3] = dst->height;
+
+   src_loc[0] = sx;
+   src_loc[1] = sy;
+   src_loc[2] = width;
+   src_loc[3] = height;
+   src_bounds[0] = 0.f;
+   src_bounds[1] = 0.f;
+   src_bounds[2] = src->width;
+   src_bounds[3] = src->height;
+
+   vg_bound_rect(src_loc, src_bounds, src_shift);
+   vg_bound_rect(dst_loc, dst_bounds, dst_shift);
+   shift[0] = src_shift[0] - dst_shift[0];
+   shift[1] = src_shift[1] - dst_shift[1];
+
+   if (shift[0] < 0)
+      vg_shift_rectx(src_loc, src_bounds, -shift[0]);
+   else
+      vg_shift_rectx(dst_loc, dst_bounds, shift[0]);
+
+   if (shift[1] < 0)
+      vg_shift_recty(src_loc, src_bounds, -shift[1]);
+   else
+      vg_shift_recty(dst_loc, dst_bounds, shift[1]);
+
+   vg_sync_size(src_loc, dst_loc);
+
+   if (src_loc[2] > 0 && src_loc[3] > 0 &&
+       dst_loc[2] > 0 && dst_loc[3] > 0) {
+      if (src == dst)
+         renderer_copy_surface(ctx->renderer,
+                               src,
+                               src_loc[0],
+                               src->height - (src_loc[1] + src_loc[3]),
+                               src_loc[0] + src_loc[2],
+                               src->height - src_loc[1],
+                               dst,
+                               dst_loc[0],
+                               dst->height - (dst_loc[1] + dst_loc[3]),
+                               dst_loc[0] + dst_loc[2],
+                               dst->height - dst_loc[1],
+                               0, 0);
+      else
+         renderer_copy_surface(ctx->renderer,
+                               src,
+                               src_loc[0],
+                               src->height - src_loc[1],
+                               src_loc[0] + src_loc[2],
+                               src->height - (src_loc[1] + src_loc[3]),
+                               dst,
+                               dst_loc[0],
+                               dst->height - (dst_loc[1] + dst_loc[3]),
+                               dst_loc[0] + dst_loc[2],
+                               dst->height - dst_loc[1],
+                               0, 0);
+   }
+
+}
+
+static struct pipe_texture *image_texture(struct vg_image *img)
+{
+   struct pipe_texture *tex = img->texture;
+   return tex;
+}
+
+
+static void image_cleari(struct vg_image *img, VGint clear_colori,
+                         VGint x, VGint y, VGint width, VGint height)
+{
+   VGint *clearbuf;
+   VGint i;
+   VGfloat dwidth, dheight;
+
+   clearbuf = malloc(sizeof(VGint)*width*height);
+   for (i = 0; i < width*height; ++i)
+      clearbuf[i] = clear_colori;
+
+   dwidth = MIN2(width, img->width);
+   dheight = MIN2(height, img->height);
+
+   image_sub_data(img, clearbuf, width * sizeof(VGint),
+                  VG_sRGBA_8888,
+                  x, y, dwidth, dheight);
+   free(clearbuf);
+}
+
+struct vg_image * image_create(VGImageFormat format,
+                               VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *image = CALLOC_STRUCT(vg_image);
+   enum pipe_format pformat = vg_format_to_pipe(format);
+   struct pipe_texture pt, *newtex;
+   struct pipe_screen *screen = ctx->pipe->screen;
+
+   vg_init_object(&image->base, ctx, VG_OBJECT_IMAGE);
+
+   image->format = format;
+   image->width = width;
+   image->height = height;
+
+   image->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.normalized_coords = 1;
+
+   assert(screen->is_format_supported(screen, pformat, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   memset(&pt, 0, sizeof(pt));
+   pt.target = PIPE_TEXTURE_2D;
+   pt.format = pformat;
+   pf_get_block(pformat, &pt.block);
+   pt.last_level = 0;
+   pt.width[0] = width;
+   pt.height[0] = height;
+   pt.depth[0] = 1;
+   pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   newtex = screen->texture_create(screen, &pt);
+
+   debug_assert(newtex);
+
+   image->texture = newtex;
+
+   vg_context_add_object(ctx, VG_OBJECT_IMAGE, image);
+
+   image_cleari(image, 0, 0, 0, image->width, image->height);
+   return image;
+}
+
+void image_destroy(struct vg_image *img)
+{
+   struct vg_context *ctx = vg_current_context();
+   vg_context_remove_object(ctx, VG_OBJECT_IMAGE, img);
+
+
+   if (img->parent) {
+      /* remove img from the parent child array */
+      int idx;
+      struct vg_image **array =
+         (struct vg_image **)img->parent->children_array->data;
+
+      for (idx = 0; idx < img->parent->children_array->num_elements; ++idx) {
+         struct vg_image *child = array[idx];
+         if (child == img) {
+            break;
+         }
+      }
+      debug_assert(idx < img->parent->children_array->num_elements);
+      array_remove_element(img->parent->children_array, idx);
+   }
+
+   if (img->children_array && img->children_array->num_elements) {
+      /* reparent the children */
+      VGint i;
+      struct vg_image *parent = img->parent;
+      struct vg_image **children =
+         (struct vg_image **)img->children_array->data;
+      if (!parent) {
+         VGint min_x = children[0]->x;
+         parent = children[0];
+
+         for (i = 1; i < img->children_array->num_elements; ++i) {
+            struct vg_image *child = children[i];
+            if (child->x < min_x) {
+               parent = child;
+            }
+         }
+      }
+
+      for (i = 0; i < img->children_array->num_elements; ++i) {
+         struct vg_image *child = children[i];
+         if (child != parent) {
+            child->parent = parent;
+            if (!parent->children_array) {
+               parent->children_array = array_create(
+                  sizeof(struct vg_image*));
+            }
+            array_append_data(parent->children_array,
+                              &child, 1);
+         } else
+            child->parent = NULL;
+      }
+      array_destroy(img->children_array);
+   }
+
+   pipe_texture_reference(&img->texture, NULL);
+   free(img);
+}
+
+void image_clear(struct vg_image *img,
+                 VGint x, VGint y, VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat *clear_colorf = ctx->state.vg.clear_color;
+   VGubyte r, g, b ,a;
+   VGint clear_colori;
+   /* FIXME: this is very nasty */
+   r = float_to_ubyte(clear_colorf[0]);
+   g = float_to_ubyte(clear_colorf[1]);
+   b = float_to_ubyte(clear_colorf[2]);
+   a = float_to_ubyte(clear_colorf[3]);
+   clear_colori = r << 24 | g << 16 | b << 8 | a;
+   image_cleari(img, clear_colori, x, y, width, height);
+}
+
+void image_sub_data(struct vg_image *image,
+                    const void * data,
+                    VGint dataStride,
+                    VGImageFormat dataFormat,
+                    VGint x, VGint y,
+                    VGint width, VGint height)
+{
+   const VGint yStep = 1;
+   VGubyte *src = (VGubyte *)data;
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint i;
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_screen *screen = ctx->pipe->screen;
+   struct pipe_texture *texture = image_texture(image);
+   VGint xoffset = 0, yoffset = 0;
+
+   if (x < 0) {
+      xoffset -= x;
+      width += x;
+      x = 0;
+   }
+   if (y < 0) {
+      yoffset -= y;
+      height += y;
+      y = 0;
+   }
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (x > image->width || y > image->width) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+
+   if (x + width > image->width) {
+      width = image->width - x;
+   }
+
+   if (y + height > image->height) {
+      height = image->height - y;
+   }
+
+   { /* upload color_data */
+      struct pipe_transfer *transfer = screen->get_tex_transfer(
+         screen, texture, 0, 0, 0,
+         PIPE_TRANSFER_WRITE, 0, 0, texture->width[0], texture->height[0]);
+      src += (dataStride * yoffset);
+      for (i = 0; i < height; i++) {
+         _vega_unpack_float_span_rgba(ctx, width, xoffset, src, dataFormat, temp);
+         pipe_put_tile_rgba(transfer, x+image->x, y+image->y, width, 1, df);
+         y += yStep;
+         src += dataStride;
+      }
+      screen->tex_transfer_destroy(transfer);
+   }
+}
+
+void image_get_sub_data(struct vg_image * image,
+                        void * data,
+                        VGint dataStride,
+                        VGImageFormat dataFormat,
+                        VGint sx, VGint sy,
+                        VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint y = 0, yStep = 1;
+   VGint i;
+   VGubyte *dst = (VGubyte *)data;
+
+   {
+      struct pipe_transfer *transfer =
+         screen->get_tex_transfer(screen,
+                                  image->texture,  0, 0, 0,
+                                  PIPE_TRANSFER_READ,
+                                  0, 0,
+                                  image->x + image->width,
+                                  image->y + image->height);
+      /* Do a row at a time to flip image data vertically */
+      for (i = 0; i < height; i++) {
+#if 0
+         debug_printf("%d-%d  == %d\n", sy, height, y);
+#endif
+         pipe_get_tile_rgba(transfer, sx+image->x, y, width, 1, df);
+         y += yStep;
+         _vega_pack_rgba_span_float(ctx, width, temp, dataFormat, dst);
+         dst += dataStride;
+      }
+
+      screen->tex_transfer_destroy(transfer);
+   }
+}
+
+struct vg_image * image_child_image(struct vg_image *parent,
+                                    VGint x, VGint y,
+                                    VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_image *image = CALLOC_STRUCT(vg_image);
+
+   vg_init_object(&image->base, ctx, VG_OBJECT_IMAGE);
+
+   image->x = parent->x + x;
+   image->y = parent->y + y;
+   image->width = width;
+   image->height = height;
+   image->parent = parent;
+   image->texture = 0;
+   pipe_texture_reference(&image->texture,
+                          parent->texture);
+
+   image->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   image->sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   image->sampler.normalized_coords = 1;
+
+   if (!parent->children_array)
+      parent->children_array = array_create(
+         sizeof(struct vg_image*));
+
+   array_append_data(parent->children_array,
+                     &image, 1);
+
+   vg_context_add_object(ctx, VG_OBJECT_IMAGE, image);
+
+   return image;
+}
+
+void image_copy(struct vg_image *dst, VGint dx, VGint dy,
+                struct vg_image *src, VGint sx, VGint sy,
+                VGint width, VGint height,
+                VGboolean dither)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (width <= 0 || height <= 0) {
+      vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR);
+      return;
+   }
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   vg_copy_texture(ctx, dst->texture, dst->x + dx, dst->y + dy,
+                   src->texture, src->x + sx, src->y + sy, width, height);
+}
+
+void image_draw(struct vg_image *img)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat x1, y1;
+   VGfloat x2, y2;
+   VGfloat x3, y3;
+   VGfloat x4, y4;
+   struct matrix *matrix;
+
+   x1 = 0;
+   y1 = 0;
+   x2 = img->width;
+   y2 = 0;
+   x3 = img->width;
+   y3 = img->height;
+   x4 = 0;
+   y4 = img->height;
+
+   matrix = &ctx->state.vg.image_user_to_surface_matrix;
+
+   matrix_map_point(matrix, x1, y1, &x1, &y1);
+   matrix_map_point(matrix, x2, y2, &x2, &y2);
+   matrix_map_point(matrix, x3, y3, &x3, &y3);
+   matrix_map_point(matrix, x4, y4, &x4, &y4);
+
+   shader_set_drawing_image(ctx->shader, VG_TRUE);
+   shader_set_paint(ctx->shader, ctx->state.vg.fill_paint);
+   shader_set_image(ctx->shader, img);
+   shader_bind(ctx->shader);
+
+   renderer_texture_quad(ctx->renderer, image_texture(img),
+                         img->x, img->y, img->x + img->width, img->y + img->height,
+                         x1, y1, x2, y2, x3, y3, x4, y4);
+}
+
+void image_set_pixels(VGint dx, VGint dy,
+                      struct vg_image *src, VGint sx, VGint sy,
+                      VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surf;
+   struct st_renderbuffer *strb = ctx->draw_buffer->strb;
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   surf = screen->get_tex_surface(screen, image_texture(src),  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_GPU_READ);
+
+   vg_copy_surface(ctx, strb->surface, dx, dy,
+                   surf, sx+src->x, sy+src->y, width, height);
+
+   screen->tex_surface_destroy(surf);
+}
+
+void image_get_pixels(struct vg_image *dst, VGint dx, VGint dy,
+                      VGint sx, VGint sy,
+                      VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surf;
+   struct st_renderbuffer *strb = ctx->draw_buffer->strb;
+
+   /* flip the y coordinates */
+   /*dy = dst->height - dy - height;*/
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   surf = screen->get_tex_surface(screen, image_texture(dst),  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_GPU_WRITE |
+                                  PIPE_BUFFER_USAGE_GPU_READ);
+   vg_copy_surface(ctx, surf, dst->x + dx, dst->y + dy,
+                   strb->surface, sx, sy, width, height);
+
+   pipe_surface_reference(&surf, NULL);
+}
+
+
+VGboolean vg_image_overlaps(struct vg_image *dst,
+                            struct vg_image *src)
+{
+   if (dst == src || dst->parent == src ||
+       dst == src->parent)
+      return VG_TRUE;
+   if (dst->parent && dst->parent == src->parent) {
+      VGfloat left1 = dst->x;
+      VGfloat left2 = src->x;
+      VGfloat right1 = dst->x + dst->width;
+      VGfloat right2 = src->x + src->width;
+      VGfloat bottom1 = dst->y;
+      VGfloat bottom2 = src->y;
+      VGfloat top1 = dst->y + dst->height;
+      VGfloat top2 = src->y + src->height;
+
+      return !(left2 > right1 || right2 < left1 ||
+               top2 > bottom1 || bottom2 < top1);
+   }
+   return VG_FALSE;
+}
+
+VGint image_bind_samplers(struct vg_image *img, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures)
+{
+   img->sampler.min_img_filter = image_sampler_filter(img->base.ctx);
+   img->sampler.mag_img_filter = image_sampler_filter(img->base.ctx);
+   samplers[3] = &img->sampler;
+   textures[3] = img->texture;
+   return 1;
+}
+
+VGint image_sampler_filter(struct vg_context *ctx)
+{
+    switch(ctx->state.vg.image_quality) {
+    case VG_IMAGE_QUALITY_NONANTIALIASED:
+       return PIPE_TEX_FILTER_NEAREST;
+       break;
+    case VG_IMAGE_QUALITY_FASTER:
+       return PIPE_TEX_FILTER_NEAREST;
+       break;
+    case VG_IMAGE_QUALITY_BETTER:
+       /*return PIPE_TEX_FILTER_ANISO;*/
+       return PIPE_TEX_FILTER_LINEAR;
+       break;
+    default:
+       debug_printf("Unknown image quality");
+    }
+    return PIPE_TEX_FILTER_NEAREST;
+}
diff --git a/src/gallium/state_trackers/vega/image.h b/src/gallium/state_trackers/vega/image.h
new file mode 100644
index 0000000000..78e17cffa6
--- /dev/null
+++ b/src/gallium/state_trackers/vega/image.h
@@ -0,0 +1,104 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef IMAGES_H
+#define IMAGES_H
+
+#include "vg_context.h"
+#include "pipe/p_state.h"
+
+struct pipe_texture;
+struct array;
+struct vg_context;
+struct pipe_surface;
+
+struct vg_image {
+   struct vg_object base;
+   VGImageFormat format;
+   VGint x, y;
+   VGint width, height;
+
+   struct vg_image *parent;
+
+   struct pipe_texture *texture;
+   struct pipe_sampler_state sampler;
+
+   struct array *children_array;
+};
+
+struct vg_image *image_create(VGImageFormat format,
+                              VGint width, VGint height);
+void image_destroy(struct vg_image *img);
+
+void image_clear(struct vg_image *img,
+                 VGint x, VGint y, VGint width, VGint height);
+
+void image_sub_data(struct vg_image *image,
+                    const void * data,
+                    VGint dataStride,
+                    VGImageFormat dataFormat,
+                    VGint x, VGint y,
+                    VGint width, VGint height);
+
+void image_get_sub_data(struct vg_image * image,
+                        void * data,
+                        VGint dataStride,
+                        VGImageFormat dataFormat,
+                        VGint x, VGint y,
+                        VGint width, VGint height);
+
+struct vg_image *image_child_image(struct vg_image *parent,
+                                   VGint x, VGint y,
+                                   VGint width, VGint height);
+
+void image_copy(struct vg_image *dst, VGint dx, VGint dy,
+                struct vg_image *src, VGint sx, VGint sy,
+                VGint width, VGint height,
+                VGboolean dither);
+
+void image_draw(struct vg_image *img);
+
+void image_set_pixels(VGint dx, VGint dy,
+                      struct vg_image *src, VGint sx, VGint sy,
+                      VGint width, VGint height);
+void image_get_pixels(struct vg_image *dst, VGint dx, VGint dy,
+                      VGint sx, VGint sy,
+                      VGint width, VGint height);
+
+VGint image_bind_samplers(struct vg_image *dst, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures);
+
+VGboolean vg_image_overlaps(struct vg_image *dst,
+                            struct vg_image *src);
+
+VGint image_sampler_filter(struct vg_context *ctx);
+
+void vg_copy_surface(struct vg_context *ctx,
+                     struct pipe_surface *dst, VGint dx, VGint dy,
+                     struct pipe_surface *src, VGint sx, VGint sy,
+                     VGint width, VGint height);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/mask.c b/src/gallium/state_trackers/vega/mask.c
new file mode 100644
index 0000000000..24650a37d5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/mask.c
@@ -0,0 +1,690 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "mask.h"
+
+#include "path.h"
+#include "image.h"
+#include "shaders_cache.h"
+#include "renderer.h"
+#include "asm_util.h"
+#include "st_inlines.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+
+struct vg_mask_layer {
+   struct vg_object base;
+
+   VGint width;
+   VGint height;
+
+   struct pipe_texture *texture;
+};
+
+static INLINE struct pipe_surface *
+alpha_mask_surface(struct vg_context *ctx, int usage)
+{
+   struct pipe_screen *screen = ctx->pipe->screen;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   return screen->get_tex_surface(screen,
+                                  stfb->alpha_mask,
+                                  0, 0, 0,
+                                  usage);
+}
+
+static INLINE VGboolean
+intersect_rectangles(VGint dwidth, VGint dheight,
+                     VGint swidth, VGint sheight,
+                     VGint tx, VGint ty,
+                     VGint twidth, VGint theight,
+                     VGint *offsets,
+                     VGint *location)
+{
+   if (tx + twidth <= 0 || tx >= dwidth)
+      return VG_FALSE;
+   if (ty + theight <= 0 || ty >= dheight)
+      return VG_FALSE;
+
+   offsets[0] = 0;
+   offsets[1] = 0;
+   location[0] = tx;
+   location[1] = ty;
+
+   if (tx < 0) {
+      offsets[0] -= tx;
+      location[0] = 0;
+
+      location[2] = MIN2(tx + swidth, MIN2(dwidth, tx + twidth));
+      offsets[2] = location[2];
+   } else {
+      offsets[2] = MIN2(twidth, MIN2(dwidth - tx, swidth ));
+      location[2] = offsets[2];
+   }
+
+   if (ty < 0) {
+      offsets[1] -= ty;
+      location[1] = 0;
+
+      location[3] = MIN2(ty + sheight, MIN2(dheight, ty + theight));
+      offsets[3] = location[3];
+   } else {
+      offsets[3] = MIN2(theight, MIN2(dheight - ty, sheight));
+      location[3] = offsets[3];
+   }
+
+   return VG_TRUE;
+}
+
+#if DEBUG_MASKS
+static void read_alpha_mask(void * data, VGint dataStride,
+                            VGImageFormat dataFormat,
+                            VGint sx, VGint sy,
+                            VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->alpha_mask;
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+
+   VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4];
+   VGfloat *df = (VGfloat*)temp;
+   VGint y = (fb->height - sy) - 1, yStep = -1;
+   VGint i;
+   VGubyte *dst = (VGubyte *)data;
+   VGint xoffset = 0, yoffset = 0;
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   if (sx < 0) {
+      xoffset = -sx;
+      xoffset *= _vega_size_for_format(dataFormat);
+      width += sx;
+      sx = 0;
+   }
+   if (sy < 0) {
+      yoffset = -sy;
+      height += sy;
+      sy = 0;
+      y = (fb->height - sy) - 1;
+      yoffset *= dataStride;
+   }
+
+   {
+      struct pipe_surface *surf;
+
+      surf = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_READ);
+
+      /* Do a row at a time to flip image data vertically */
+      for (i = 0; i < height; i++) {
+#if 0
+         debug_printf("%d-%d  == %d\n", sy, height, y);
+#endif
+         pipe_get_tile_rgba(surf, sx, y, width, 1, df);
+         y += yStep;
+         _vega_pack_rgba_span_float(ctx, width, temp, dataFormat,
+                                    dst + yoffset + xoffset);
+         dst += dataStride;
+      }
+
+      pipe_surface_reference(&surf, NULL);
+   }
+}
+
+void save_alpha_to_file(const char *filename)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+   VGint *data;
+   int i, j;
+
+   data = malloc(sizeof(int) * fb->width * fb->height);
+   read_alpha_mask(data, fb->width * sizeof(int),
+                   VG_sRGBA_8888,
+                   0, 0, fb->width, fb->height);
+   fprintf(stderr, "/*---------- start */\n");
+   fprintf(stderr, "const int image_width = %d;\n",
+           fb->width);
+   fprintf(stderr, "const int image_height = %d;\n",
+           fb->height);
+   fprintf(stderr, "const int image_data = {\n");
+   for (i = 0; i < fb->height; ++i) {
+      for (j = 0; j < fb->width; ++j) {
+         int rgba = data[i * fb->height + j];
+         int argb = 0;
+         argb = (rgba >> 8);
+         argb |= ((rgba & 0xff) << 24);
+         fprintf(stderr, "0x%x, ", argb);
+      }
+      fprintf(stderr, "\n");
+   }
+   fprintf(stderr, "};\n");
+   fprintf(stderr, "/*---------- end */\n");
+}
+#endif
+
+static void setup_mask_framebuffer(struct pipe_surface *surf,
+                                   VGint surf_width, VGint surf_height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_framebuffer_state fb;
+
+   memset(&fb, 0, sizeof(fb));
+   fb.width = surf_width;
+   fb.height = surf_height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = surf;
+   {
+      VGint i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(ctx->cso_context, &fb);
+}
+
+
+/* setup shader constants */
+static void setup_mask_operation(VGMaskOperation operation)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf;
+   const VGint param_bytes = 4 * sizeof(VGfloat);
+   const VGfloat ones[4] = {1.f, 1.f, 1.f, 1.f};
+   void *shader = 0;
+
+   /* We always need to get a new buffer, to keep the drivers simple and
+    * avoid gratuitous rendering synchronization.
+    */
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 1,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+   if (cbuf->buffer) {
+      st_no_flush_pipe_buffer_write(ctx, cbuf->buffer,
+                                    0, param_bytes, ones);
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+   switch (operation) {
+   case VG_UNION_MASK: {
+      if (!ctx->mask.union_fs) {
+         ctx->mask.union_fs = shader_create_from_text(ctx->pipe,
+                                                      union_mask_asm,
+                                                      200,
+                                                      PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.union_fs->driver;
+   }
+      break;
+   case VG_INTERSECT_MASK: {
+      if (!ctx->mask.intersect_fs) {
+         ctx->mask.intersect_fs = shader_create_from_text(ctx->pipe,
+                                                          intersect_mask_asm,
+                                                          200,
+                                                          PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.intersect_fs->driver;
+   }
+      break;
+   case VG_SUBTRACT_MASK: {
+      if (!ctx->mask.subtract_fs) {
+         ctx->mask.subtract_fs = shader_create_from_text(ctx->pipe,
+                                                         subtract_mask_asm,
+                                                         200,
+                                                         PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.subtract_fs->driver;
+   }
+      break;
+   case VG_SET_MASK: {
+      if (!ctx->mask.set_fs) {
+         ctx->mask.set_fs = shader_create_from_text(ctx->pipe,
+                                                    set_mask_asm,
+                                                    200,
+                                                    PIPE_SHADER_FRAGMENT);
+      }
+      shader = ctx->mask.set_fs->driver;
+   }
+      break;
+   default:
+         assert(0);
+      break;
+   }
+   cso_set_fragment_shader_handle(ctx->cso_context, shader);
+}
+
+static void setup_mask_samplers(struct pipe_texture *umask)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+   struct pipe_texture *uprev = NULL;
+   struct pipe_sampler_state sampler;
+
+   uprev = fb_buffers->blend_texture;
+   sampler = ctx->mask.sampler;
+   sampler.normalized_coords = 1;
+
+   samplers[0] = NULL;
+   samplers[1] = NULL;
+   samplers[2] = NULL;
+   textures[0] = NULL;
+   textures[1] = NULL;
+   textures[2] = NULL;
+
+   samplers[0] = &sampler;
+   samplers[1] = &ctx->mask.sampler;
+
+   textures[0] = umask;
+   textures[1] = uprev;
+
+   cso_set_samplers(ctx->cso_context, 2,
+                    (const struct pipe_sampler_state **)samplers);
+   cso_set_sampler_textures(ctx->cso_context, 2, textures);
+}
+
+
+/* setup shader constants */
+static void setup_mask_fill(const VGfloat color[4])
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf;
+   const VGint param_bytes = 4 * sizeof(VGfloat);
+
+   /* We always need to get a new buffer, to keep the drivers simple and
+    * avoid gratuitous rendering synchronization.
+    */
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 1,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+   if (cbuf->buffer) {
+      st_no_flush_pipe_buffer_write(ctx, cbuf->buffer, 0, param_bytes, color);
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+   cso_set_fragment_shader_handle(ctx->cso_context,
+                                  shaders_cache_fill(ctx->sc,
+                                                     VEGA_SOLID_FILL_SHADER));
+}
+
+static void setup_mask_viewport()
+{
+   struct vg_context *ctx = vg_current_context();
+   vg_set_viewport(ctx, VEGA_Y0_TOP);
+}
+
+static void setup_mask_blend()
+{
+   struct vg_context *ctx = vg_current_context();
+
+   struct pipe_blend_state blend;
+
+   memset(&blend, 0, sizeof(struct pipe_blend_state));
+   blend.blend_enable = 1;
+   blend.colormask |= PIPE_MASK_R;
+   blend.colormask |= PIPE_MASK_G;
+   blend.colormask |= PIPE_MASK_B;
+   blend.colormask |= PIPE_MASK_A;
+   blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+
+   cso_set_blend(ctx->cso_context, &blend);
+}
+
+
+static void surface_fill(struct pipe_surface *surf,
+                         int surf_width, int surf_height,
+                         int x, int y, int width, int height,
+                         const VGfloat color[4])
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (x < 0) {
+      width += x;
+      x = 0;
+   }
+   if (y < 0) {
+      height += y;
+      y = 0;
+   }
+
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_blend(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+
+   setup_mask_blend();
+   setup_mask_fill(color);
+   setup_mask_framebuffer(surf, surf_width, surf_height);
+   setup_mask_viewport();
+
+   renderer_draw_quad(ctx->renderer, x, y,
+                      x + width, y + height, 0.0f/*depth should be disabled*/);
+
+
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe,
+                    PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME,
+                    NULL);
+
+#if DEBUG_MASKS
+   save_alpha_to_file(0);
+#endif
+
+   cso_restore_blend(ctx->cso_context);
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+}
+
+
+static void mask_using_texture(struct pipe_texture *texture,
+                               VGMaskOperation operation,
+                               VGint x, VGint y,
+                               VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct pipe_surface *surface =
+      alpha_mask_surface(ctx, PIPE_BUFFER_USAGE_GPU_WRITE);
+   VGint offsets[4], loc[4];
+
+   if (!surface)
+      return;
+   if (!intersect_rectangles(surface->width, surface->height,
+                             texture->width[0], texture->height[0],
+                             x, y, width, height,
+                             offsets, loc))
+      return;
+#if 0
+   debug_printf("Offset = [%d, %d, %d, %d]\n", offsets[0],
+                offsets[1], offsets[2], offsets[3]);
+   debug_printf("Locati = [%d, %d, %d, %d]\n", loc[0],
+                loc[1], loc[2], loc[3]);
+#endif
+
+   /* prepare our blend surface */
+   vg_prepare_blend_surface_from_mask(ctx);
+
+   cso_save_samplers(ctx->cso_context);
+   cso_save_sampler_textures(ctx->cso_context);
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_blend(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+
+   setup_mask_samplers(texture);
+   setup_mask_blend();
+   setup_mask_operation(operation);
+   setup_mask_framebuffer(surface, surface->width, surface->height);
+   setup_mask_viewport();
+
+   /* render the quad to propagate the rendering from stencil */
+   renderer_draw_texture(ctx->renderer, texture,
+                         offsets[0], offsets[1],
+                         offsets[0] + offsets[2], offsets[1] + offsets[3],
+                         loc[0], loc[1], loc[0] + loc[2], loc[1] + loc[3]);
+
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   cso_restore_blend(ctx->cso_context);
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_samplers(ctx->cso_context);
+   cso_restore_sampler_textures(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+
+   pipe_surface_reference(&surface, NULL);
+}
+
+
+#ifdef OPENVG_VERSION_1_1
+
+struct vg_mask_layer * mask_layer_create(VGint width, VGint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct vg_mask_layer *mask = 0;
+
+   mask = CALLOC_STRUCT(vg_mask_layer);
+   vg_init_object(&mask->base, ctx, VG_OBJECT_MASK);
+   mask->width = width;
+   mask->height = height;
+
+   {
+      struct pipe_texture pt;
+      struct pipe_screen *screen = ctx->pipe->screen;
+
+      memset(&pt, 0, sizeof(pt));
+      pt.target = PIPE_TEXTURE_2D;
+      pt.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+      pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &pt.block);
+      pt.last_level = 0;
+      pt.width[0] = width;
+      pt.height[0] = height;
+      pt.depth[0] = 1;
+      pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+      pt.compressed = 0;
+
+      mask->texture = screen->texture_create(screen, &pt);
+   }
+
+   vg_context_add_object(ctx, VG_OBJECT_MASK, mask);
+
+   return mask;
+}
+
+void mask_layer_destroy(struct vg_mask_layer *layer)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   vg_context_remove_object(ctx, VG_OBJECT_MASK, layer);
+   pipe_texture_release(&layer->texture);
+   free(layer);
+}
+
+void mask_layer_fill(struct vg_mask_layer *layer,
+                     VGint x, VGint y,
+                     VGint width, VGint height,
+                     VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat alpha_color[4] = {0, 0, 0, 0};
+   struct pipe_surface *surface;
+
+   alpha_color[3] = value;
+
+   surface = ctx->pipe->screen->get_tex_surface(
+      ctx->pipe->screen, layer->texture,
+      0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   surface_fill(surface,
+                layer->width, layer->height,
+                x, y, width, height, alpha_color);
+
+   ctx->pipe->screen->tex_surface_release(ctx->pipe->screen, &surface);
+}
+
+void mask_copy(struct vg_mask_layer *layer,
+               VGint sx, VGint sy,
+               VGint dx, VGint dy,
+               VGint width, VGint height)
+{
+    struct vg_context *ctx = vg_current_context();
+    struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+
+    renderer_copy_texture(ctx->renderer,
+                          layer->texture,
+                          sx, sy,
+                          sx + width, sy + height,
+                          fb_buffers->alpha_mask,
+                          dx, dy,
+                          dx + width, dy + height);
+}
+
+static void mask_layer_render_to(struct vg_mask_layer *layer,
+                                 struct path *path,
+                                 VGbitfield paint_modes)
+{
+   struct vg_context *ctx = vg_current_context();
+   const VGfloat fill_color[4] = {1.f, 1.f, 1.f, 1.f};
+   struct pipe_screen *screen = ctx->pipe->screen;
+   struct pipe_surface *surface;
+
+   surface = screen->get_tex_surface(screen, layer->texture,  0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   cso_save_framebuffer(ctx->cso_context);
+   cso_save_fragment_shader(ctx->cso_context);
+   cso_save_viewport(ctx->cso_context);
+
+   setup_mask_blend();
+   setup_mask_fill(fill_color);
+   setup_mask_framebuffer(surface, layer->width, layer->height);
+   setup_mask_viewport();
+
+   if (paint_modes & VG_FILL_PATH) {
+      struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix;
+      path_fill(path, mat);
+   }
+
+   if (paint_modes & VG_STROKE_PATH){
+      path_stroke(path);
+   }
+
+
+   /* make sure rendering has completed */
+   ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   cso_restore_framebuffer(ctx->cso_context);
+   cso_restore_fragment_shader(ctx->cso_context);
+   cso_restore_viewport(ctx->cso_context);
+   ctx->state.dirty |= BLEND_DIRTY;
+
+   screen->tex_surface_release(ctx->pipe->screen, &surface);
+}
+
+void mask_render_to(struct path *path,
+                    VGbitfield paint_modes,
+                    VGMaskOperation operation)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+   struct vg_mask_layer *temp_layer;
+   VGint width, height;
+
+   width = fb_buffers->alpha_mask->width[0];
+   height = fb_buffers->alpha_mask->width[0];
+
+   temp_layer = mask_layer_create(width, height);
+
+   mask_layer_render_to(temp_layer, path, paint_modes);
+
+   mask_using_layer(temp_layer, 0, 0, width, height,
+                    operation);
+
+   mask_layer_destroy(temp_layer);
+}
+
+void mask_using_layer(struct vg_mask_layer *layer,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height)
+{
+   mask_using_texture(layer->texture, operation,
+                      x, y, width, height);
+}
+
+VGint mask_layer_width(struct vg_mask_layer *layer)
+{
+   return layer->width;
+}
+
+VGint mask_layer_height(struct vg_mask_layer *layer)
+{
+   return layer->height;
+}
+
+
+#endif
+
+void mask_using_image(struct vg_image *image,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height)
+{
+   mask_using_texture(image->texture, operation,
+                      x, y, width, height);
+}
+
+void mask_fill(VGint x, VGint y, VGint width, VGint height,
+               VGfloat value)
+{
+   struct vg_context *ctx = vg_current_context();
+   VGfloat alpha_color[4] = {.0f, .0f, .0f, value};
+   struct pipe_surface *surf = alpha_mask_surface(
+      ctx, PIPE_BUFFER_USAGE_GPU_WRITE);
+
+#if DEBUG_MASKS
+   debug_printf("mask_fill(%d, %d, %d, %d) with  rgba(%f, %f, %f, %f)\n",
+                x, y, width, height,
+                alpha_color[0], alpha_color[1],
+                alpha_color[2], alpha_color[3]);
+   debug_printf("XXX %f  === %f \n",
+                alpha_color[3], value);
+#endif
+
+   surface_fill(surf, surf->width, surf->height,
+                x, y, width, height, alpha_color);
+
+   pipe_surface_reference(&surf, NULL);
+}
+
+VGint mask_bind_samplers(struct pipe_sampler_state **samplers,
+                         struct pipe_texture **textures)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   if (ctx->state.vg.masking) {
+      struct st_framebuffer *fb_buffers = ctx->draw_buffer;
+
+      samplers[1] = &ctx->mask.sampler;
+      textures[1] = fb_buffers->alpha_mask;
+      return 1;
+   } else
+      return 0;
+}
diff --git a/src/gallium/state_trackers/vega/mask.h b/src/gallium/state_trackers/vega/mask.h
new file mode 100644
index 0000000000..5eaaede0e3
--- /dev/null
+++ b/src/gallium/state_trackers/vega/mask.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef MASK_H
+#define MASK_H
+
+#include "vg_context.h"
+
+struct path;
+struct vg_image;
+struct pipe_texture;
+
+struct vg_mask_layer *mask_layer_create(VGint width, VGint height);
+void mask_layer_destroy(struct vg_mask_layer *layer);
+void mask_layer_fill(struct vg_mask_layer *layer,
+                     VGint x, VGint y,
+                     VGint width, VGint height,
+                     VGfloat value);
+VGint mask_layer_width(struct vg_mask_layer *layer);
+VGint mask_layer_height(struct vg_mask_layer *layer);
+void mask_copy(struct vg_mask_layer *layer,
+               VGint sx, VGint sy,
+               VGint dx, VGint dy,
+               VGint width, VGint height);
+
+void mask_render_to(struct path *path,
+                    VGbitfield paint_modes,
+                    VGMaskOperation operation);
+
+void mask_using_layer(struct vg_mask_layer *layer,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height);
+void mask_using_image(struct vg_image *image,
+                      VGMaskOperation operation,
+                      VGint x, VGint y,
+                      VGint width, VGint height);
+void mask_fill(VGint x, VGint y,
+               VGint width, VGint height,
+               VGfloat value);
+
+VGint mask_bind_samplers(struct pipe_sampler_state **samplers,
+                         struct pipe_texture **textures);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/matrix.h b/src/gallium/state_trackers/vega/matrix.h
new file mode 100644
index 0000000000..4c207f912a
--- /dev/null
+++ b/src/gallium/state_trackers/vega/matrix.h
@@ -0,0 +1,462 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef MATRIX_H
+#define MATRIX_H
+
+#include "VG/openvg.h"
+
+#include "pipe/p_compiler.h"
+#include "util/u_math.h"
+
+#include <stdio.h>
+#include <math.h>
+
+#define floatsEqual(x, y) (fabs(x - y) <= 0.00001f * MIN2(fabs(x), fabs(y)))
+#define floatIsZero(x) (floatsEqual((x) + 1, 1))
+#define ABS(x) (fabsf(x))
+
+#define DEGREES_TO_RADIANS(d) (0.0174532925199 * (d))
+#define FLT_TO_INT(flt) float_to_int_floor(((VGuint*)&(flt))[0])
+
+static INLINE VGint float_to_int_floor(VGuint bits)
+{
+   int sign = (bits >> 31) ? -1 : 1;
+   int exp  = ((bits >> 23) & 255) - 127;
+   int mant = bits & 0x007fffff;
+   int sh   = 23 - exp;
+
+   /* abs(value) >= 2^31 -> clamp. */
+
+   if (exp >= 31)
+      return (VGint)((sign < 0) ? 0x80000000u : 0x7fffffffu);
+
+   /* abs(value) < 1 -> return -1 or 0. */
+
+   if (exp < 0)
+      return (sign < 0 && (exp > -127 || mant != 0)) ? -1 : 0;
+
+   /* abs(value) >= 2^23 -> shift left. */
+
+   mant |= 0x00800000;
+   if (sh <= 0)
+      return sign * (mant << -sh);
+
+   /* Negative -> add a rounding term. */
+
+   if (sign < 0)
+      mant += (1 << sh) - 1;
+
+   /* Shift right to obtain the result. */
+
+   return sign * (mant >> sh);
+}
+
+
+struct matrix {
+   VGfloat m[9];
+};
+
+static INLINE void matrix_init(struct matrix *mat,
+                               const VGfloat *val)
+{
+   memcpy(mat->m, val, sizeof(VGfloat) * 9);
+}
+
+static INLINE void matrix_inits(struct matrix *mat,
+                                VGfloat m11, VGfloat m12, VGfloat m13,
+                                VGfloat m21, VGfloat m22, VGfloat m23,
+                                VGfloat m31, VGfloat m32, VGfloat m33)
+{
+   mat->m[0] = m11; mat->m[1] = m12; mat->m[2] = m13;
+   mat->m[3] = m21; mat->m[4] = m22; mat->m[5] = m23;
+   mat->m[6] = m31; mat->m[7] = m32; mat->m[8] = m33;
+}
+
+static INLINE void matrix_load_identity(struct matrix *matrix)
+{
+   static const VGfloat identity[9] = {1.f, 0.f, 0.f,
+                                       0.f, 1.f, 0.f,
+                                       0.f, 0.f, 1.f};
+   memcpy(matrix->m, identity, sizeof(identity));
+}
+
+static INLINE VGboolean matrix_is_identity(struct matrix *matrix)
+{
+   return floatsEqual(matrix->m[0], 1) && floatIsZero(matrix->m[1]) &&
+      floatIsZero(matrix->m[2]) &&
+      floatIsZero(matrix->m[3]) && floatsEqual(matrix->m[4], 1) &&
+      floatIsZero(matrix->m[5]) &&
+      floatIsZero(matrix->m[6]) && floatIsZero(matrix->m[7]) &&
+      floatIsZero(matrix->m[8]);
+}
+
+static INLINE VGboolean matrix_is_affine(struct matrix *matrix)
+{
+   return floatIsZero(matrix->m[2]) && floatIsZero(matrix->m[5])
+      && floatsEqual(matrix->m[8], 1);
+}
+
+
+static INLINE void matrix_make_affine(struct matrix *matrix)
+{
+   matrix->m[2] = 0.f;
+   matrix->m[5] = 0.f;
+   matrix->m[8] = 1.f;
+}
+
+static INLINE void matrix_mult(struct matrix *dst,
+                               struct matrix *src)
+{
+   VGfloat m11 = dst->m[0]*src->m[0] + dst->m[3]*src->m[1] + dst->m[6]*src->m[2];
+   VGfloat m12 = dst->m[0]*src->m[3] + dst->m[3]*src->m[4] + dst->m[6]*src->m[5];
+   VGfloat m13 = dst->m[0]*src->m[6] + dst->m[3]*src->m[7] + dst->m[6]*src->m[8];
+
+   VGfloat m21 = dst->m[1]*src->m[0] + dst->m[4]*src->m[1] + dst->m[7]*src->m[2];
+   VGfloat m22 = dst->m[1]*src->m[3] + dst->m[4]*src->m[4] + dst->m[7]*src->m[5];
+   VGfloat m23 = dst->m[1]*src->m[6] + dst->m[4]*src->m[7] + dst->m[7]*src->m[8];
+
+   VGfloat m31 = dst->m[2]*src->m[0] + dst->m[5]*src->m[1] + dst->m[8]*src->m[2];
+   VGfloat m32 = dst->m[2]*src->m[3] + dst->m[5]*src->m[4] + dst->m[8]*src->m[5];
+   VGfloat m33 = dst->m[2]*src->m[6] + dst->m[5]*src->m[7] + dst->m[8]*src->m[8];
+
+   dst->m[0] = m11; dst->m[1] = m21; dst->m[2] = m31;
+   dst->m[3] = m12; dst->m[4] = m22; dst->m[5] = m32;
+   dst->m[6] = m13; dst->m[7] = m23; dst->m[8] = m33;
+}
+
+
+static INLINE void matrix_map_point(struct matrix *mat,
+                                    VGfloat x, VGfloat y,
+                                    VGfloat *out_x, VGfloat *out_y)
+{
+   /* to be able to do matrix_map_point(m, x, y, &x, &y) use
+    * temporaries */
+   VGfloat tmp_x = x, tmp_y = y;
+
+   *out_x = mat->m[0]*tmp_x + mat->m[3]*tmp_y + mat->m[6];
+   *out_y = mat->m[1]*tmp_x + mat->m[4]*tmp_y + mat->m[7];
+   if (!matrix_is_affine(mat)) {
+      VGfloat w = 1/(mat->m[2]*tmp_x + mat->m[5]*tmp_y + mat->m[8]);
+      *out_x *= w;
+      *out_y *= w;
+   }
+}
+
+static INLINE void matrix_translate(struct matrix *dst,
+                                    VGfloat tx, VGfloat ty)
+{
+   if (!matrix_is_affine(dst)) {
+      struct matrix trans_matrix;
+      matrix_load_identity(&trans_matrix);
+      trans_matrix.m[6] = tx;
+      trans_matrix.m[7] = ty;
+      matrix_mult(dst, &trans_matrix);
+   } else {
+      dst->m[6] += tx*dst->m[0] + ty*dst->m[3];
+      dst->m[7] += ty*dst->m[4] + tx*dst->m[1];
+   }
+}
+
+static INLINE void matrix_scale(struct matrix *dst,
+                                VGfloat sx, VGfloat sy)
+{
+   if (!matrix_is_affine(dst)) {
+      struct matrix scale_matrix;
+      matrix_load_identity(&scale_matrix);
+      scale_matrix.m[0] = sx;
+      scale_matrix.m[4] = sy;
+      matrix_mult(dst, &scale_matrix);
+   } else {
+      dst->m[0] *= sx; dst->m[1] *= sx;
+      dst->m[3] *= sy; dst->m[4] *= sy;
+   }
+}
+
+static INLINE void matrix_shear(struct matrix *dst,
+                                VGfloat shx, VGfloat shy)
+{
+   struct matrix shear_matrix;
+   matrix_load_identity(&shear_matrix);
+   shear_matrix.m[1] = shy;
+   shear_matrix.m[3] = shx;
+   matrix_mult(dst, &shear_matrix);
+}
+
+static INLINE void matrix_rotate(struct matrix *dst,
+                                 VGfloat angle)
+{
+   struct matrix mat;
+   float sin_val = 0;
+   float cos_val = 0;
+
+
+   if (floatsEqual(angle, 90) || floatsEqual(angle, -270))
+      sin_val = 1.f;
+   else if (floatsEqual(angle, 270) || floatsEqual(angle, -90))
+      sin_val = -1.f;
+   else if (floatsEqual(angle, 180))
+      cos_val = -1.f;
+   else {
+      float radians = DEGREES_TO_RADIANS(angle);
+      sin_val = sin(radians);
+      cos_val = cos(radians);
+   }
+
+   if (!matrix_is_affine(dst)) {
+      matrix_load_identity(&mat);
+      mat.m[0] =  cos_val;   mat.m[1] =  sin_val;
+      mat.m[3] = -sin_val;   mat.m[4] =  cos_val;
+
+      matrix_mult(dst, &mat);
+   } else  {
+      VGfloat m11 =  cos_val*dst->m[0] + sin_val*dst->m[3];
+      VGfloat m12 =  cos_val*dst->m[1] + sin_val*dst->m[4];
+      VGfloat m21 = -sin_val*dst->m[0] + cos_val*dst->m[3];
+      VGfloat m22 = -sin_val*dst->m[1] + cos_val*dst->m[4];
+      dst->m[0] = m11; dst->m[1] = m12;
+      dst->m[3] = m21; dst->m[4] = m22;
+   }
+}
+
+
+static INLINE VGfloat matrix_determinant(struct matrix *mat)
+{
+   return mat->m[0]*(mat->m[8]*mat->m[4]-mat->m[7]*mat->m[5]) -
+      mat->m[3]*(mat->m[8]*mat->m[1]-mat->m[7]*mat->m[2])+
+      mat->m[6]*(mat->m[5]*mat->m[1]-mat->m[4]*mat->m[2]);
+}
+
+
+static INLINE void matrix_adjoint(struct matrix *mat)
+{
+    VGfloat h[9];
+    h[0] = mat->m[4]*mat->m[8] - mat->m[5]*mat->m[7];
+    h[3] = mat->m[5]*mat->m[6] - mat->m[3]*mat->m[8];
+    h[6] = mat->m[3]*mat->m[7] - mat->m[4]*mat->m[6];
+    h[1] = mat->m[2]*mat->m[7] - mat->m[1]*mat->m[8];
+    h[4] = mat->m[0]*mat->m[8] - mat->m[2]*mat->m[6];
+    h[7] = mat->m[1]*mat->m[6] - mat->m[0]*mat->m[7];
+    h[2] = mat->m[1]*mat->m[5] - mat->m[2]*mat->m[4];
+    h[5] = mat->m[2]*mat->m[3] - mat->m[0]*mat->m[5];
+    h[8] = mat->m[0]*mat->m[4] - mat->m[1]*mat->m[3];
+
+
+    memcpy(mat->m, h, sizeof(VGfloat) * 9);
+}
+
+static INLINE void matrix_divs(struct matrix *mat,
+                               VGfloat s)
+{
+   mat->m[0] /= s;
+   mat->m[1] /= s;
+   mat->m[2] /= s;
+   mat->m[3] /= s;
+   mat->m[4] /= s;
+   mat->m[5] /= s;
+   mat->m[6] /= s;
+   mat->m[7] /= s;
+   mat->m[8] /= s;
+}
+
+static INLINE VGboolean matrix_invert(struct matrix *mat)
+{
+   VGfloat det = matrix_determinant(mat);
+
+   if (floatIsZero(det))
+      return VG_FALSE;
+
+   matrix_adjoint(mat);
+   matrix_divs(mat, det);
+   return VG_TRUE;
+}
+
+static INLINE VGboolean matrix_is_invertible(struct matrix *mat)
+{
+   return !floatIsZero(matrix_determinant(mat));
+}
+
+
+static INLINE VGboolean matrix_square_to_quad(VGfloat dx0, VGfloat dy0,
+                                              VGfloat dx1, VGfloat dy1,
+                                              VGfloat dx3, VGfloat dy3,
+                                              VGfloat dx2, VGfloat dy2,
+                                              struct matrix *mat)
+{
+   VGfloat ax  = dx0 - dx1 + dx2 - dx3;
+   VGfloat ay  = dy0 - dy1 + dy2 - dy3;
+
+   if (floatIsZero(ax) && floatIsZero(ay)) {
+      /* affine case */
+      matrix_inits(mat,
+                   dx1 - dx0, dy1 - dy0,  0,
+                   dx2 - dx1, dy2 - dy1,  0,
+                         dx0,       dy0,  1);
+   } else {
+      VGfloat a, b, c, d, e, f, g, h;
+      VGfloat ax1 = dx1 - dx2;
+      VGfloat ax2 = dx3 - dx2;
+      VGfloat ay1 = dy1 - dy2;
+      VGfloat ay2 = dy3 - dy2;
+
+      /* determinants */
+      VGfloat gtop    =  ax  * ay2 - ax2 * ay;
+      VGfloat htop    =  ax1 * ay  - ax  * ay1;
+      VGfloat bottom  =  ax1 * ay2 - ax2 * ay1;
+
+      if (!bottom)
+         return VG_FALSE;
+
+      g = gtop / bottom;
+      h = htop / bottom;
+
+      a = dx1 - dx0 + g * dx1;
+      b = dx3 - dx0 + h * dx3;
+      c = dx0;
+      d = dy1 - dy0 + g * dy1;
+      e = dy3 - dy0 + h * dy3;
+      f = dy0;
+
+      matrix_inits(mat,
+                   a,  d,   g,
+                   b,  e,   h,
+                   c,  f, 1.f);
+   }
+
+   return VG_TRUE;
+}
+
+static INLINE VGboolean matrix_quad_to_square(VGfloat sx0, VGfloat sy0,
+                                              VGfloat sx1, VGfloat sy1,
+                                              VGfloat sx2, VGfloat sy2,
+                                              VGfloat sx3, VGfloat sy3,
+                                              struct matrix *mat)
+{
+   if (!matrix_square_to_quad(sx0, sy0, sx1, sy1,
+                              sx2, sy2, sx3, sy3,
+                              mat))
+      return VG_FALSE;
+
+    return matrix_invert(mat);
+}
+
+
+static INLINE VGboolean matrix_quad_to_quad(VGfloat dx0, VGfloat dy0,
+                                            VGfloat dx1, VGfloat dy1,
+                                            VGfloat dx2, VGfloat dy2,
+                                            VGfloat dx3, VGfloat dy3,
+                                            VGfloat sx0, VGfloat sy0,
+                                            VGfloat sx1, VGfloat sy1,
+                                            VGfloat sx2, VGfloat sy2,
+                                            VGfloat sx3, VGfloat sy3,
+                                            struct matrix *mat)
+{
+   struct matrix sqr_to_qd;
+
+   if (!matrix_square_to_quad(dx0, dy0, dx1, dy1,
+                              dx2, dy2, dx3, dy3,
+                              mat))
+      return VG_FALSE;
+
+   if (!matrix_quad_to_square(sx0, sy0, sx1, sy1,
+                              sx2, sy2, sx3, sy3,
+                              &sqr_to_qd))
+      return VG_FALSE;
+
+   matrix_mult(mat, &sqr_to_qd);
+
+   return VG_TRUE;
+}
+
+
+static INLINE VGboolean null_line(const VGfloat *l)
+{
+   return floatsEqual(l[0], l[2]) && floatsEqual(l[1], l[3]);
+}
+
+static INLINE void line_normal(float *l, float *norm)
+{
+   norm[0] = l[0];
+   norm[1] = l[1];
+
+   norm[2] = l[0] + (l[3] - l[1]);
+   norm[3] = l[1] - (l[2] - l[0]);
+}
+
+static INLINE void line_normalize(float *l)
+{
+   float x = l[2] - l[0];
+   float y = l[3] - l[1];
+   float len = sqrt(x*x + y*y);
+   l[2] = l[0] + x/len;
+   l[3] = l[1] + y/len;
+}
+
+static INLINE VGfloat line_length(VGfloat x1, VGfloat y1,
+                                  VGfloat x2, VGfloat y2)
+{
+   VGfloat x = x2 - x1;
+   VGfloat y = y2 - y1;
+   return sqrt(x*x + y*y);
+}
+
+static INLINE VGfloat line_lengthv(const VGfloat *l)
+{
+   VGfloat x = l[2] - l[0];
+   VGfloat y = l[3] - l[1];
+   return sqrt(x*x + y*y);
+}
+
+
+static INLINE void line_point_at(float *l, float t, float *pt)
+{
+   float dx = l[2] - l[0];
+   float dy = l[3] - l[1];
+
+   pt[0] = l[0] + dx * t;
+   pt[1] = l[1] + dy * t;
+}
+
+static INLINE void vector_unit(float *vec)
+{
+   float len = sqrt(vec[0] * vec[0] + vec[1] * vec[1]);
+   vec[0] /= len;
+   vec[1] /= len;
+}
+
+static INLINE void line_normal_vector(float *line, float *vec)
+{
+   VGfloat normal[4];
+
+   line_normal(line, normal);
+
+   vec[0] = normal[2] - normal[0];
+   vec[1] = normal[3] - normal[1];
+
+   vector_unit(vec);
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c
new file mode 100644
index 0000000000..04a6ba9cdc
--- /dev/null
+++ b/src/gallium/state_trackers/vega/paint.c
@@ -0,0 +1,699 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "paint.h"
+
+#include "shaders_cache.h"
+#include "matrix.h"
+#include "image.h"
+#include "st_inlines.h"
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_inlines.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "cso_cache/cso_context.h"
+
+struct vg_paint {
+   struct vg_object base;
+
+   VGPaintType type;
+
+   struct {
+      VGfloat color[4];
+      VGint colori[4];
+   } solid;
+
+   struct {
+      VGColorRampSpreadMode spread;
+      VGuint color_data[1024];
+      struct {
+         VGfloat  coords[4];
+         VGint  coordsi[4];
+      } linear;
+      struct {
+         VGfloat vals[5];
+         VGint valsi[5];
+      } radial;
+      struct pipe_texture *texture;
+      struct pipe_sampler_state sampler;
+
+      VGfloat *ramp_stops;
+      VGint *ramp_stopsi;
+      VGint    num_stops;
+
+      VGboolean color_ramps_premultiplied;
+   } gradient;
+
+   struct {
+      struct pipe_texture *texture;
+      VGTilingMode tiling_mode;
+      struct pipe_sampler_state sampler;
+   } pattern;
+
+   struct pipe_constant_buffer cbuf;
+   struct pipe_shader_state fs_state;
+   void *fs;
+};
+
+static INLINE VGuint mix_pixels(VGuint p1, VGuint a, VGuint p2, VGuint b)
+{
+   VGuint t = (p1 & 0xff00ff) * a + (p2 & 0xff00ff) * b;
+   t >>= 8; t &= 0xff00ff;
+
+   p1 = ((p1 >> 8) & 0xff00ff) * a + ((p2 >> 8) & 0xff00ff) * b;
+   p1 &= 0xff00ff00; p1 |= t;
+
+   return p1;
+}
+
+static INLINE VGuint float4_to_argb(const VGfloat *clr)
+{
+   return float_to_ubyte(clr[3]) << 24 |
+      float_to_ubyte(clr[0]) << 16 |
+      float_to_ubyte(clr[1]) << 8 |
+      float_to_ubyte(clr[2]) << 0;
+}
+
+static INLINE void create_gradient_data(const VGfloat *ramp_stops,
+                                        VGint num,
+                                        VGuint *data,
+                                        VGint size)
+{
+   VGint i;
+   VGint pos = 0;
+   VGfloat fpos = 0, incr = 1.f / size;
+   VGuint last_color;
+
+   while (fpos < ramp_stops[0]) {
+      data[pos] = float4_to_argb(ramp_stops + 1);
+      fpos += incr;
+      ++pos;
+   }
+
+   for (i = 0; i < num - 1; ++i) {
+      VGint rcur  = 5 * i;
+      VGint rnext = 5 * (i + 1);
+      VGfloat delta = 1.f/(ramp_stops[rnext] - ramp_stops[rcur]);
+      while (fpos < ramp_stops[rnext] && pos < size) {
+         VGint dist = 256 * ((fpos - ramp_stops[rcur]) * delta);
+         VGint idist = 256 - dist;
+         VGuint current_color = float4_to_argb(ramp_stops + rcur + 1);
+         VGuint next_color = float4_to_argb(ramp_stops + rnext + 1);
+         data[pos] = mix_pixels(current_color, idist,
+                                next_color, dist);
+         fpos += incr;
+         ++pos;
+      }
+   }
+
+   last_color = float4_to_argb(ramp_stops + ((num - 1) * 5 + 1));
+   while (pos < size) {
+      data[pos] = last_color;
+      ++pos;
+   }
+   data[size-1] = last_color;
+}
+
+static INLINE struct pipe_texture *create_gradient_texture(struct vg_paint *p)
+{
+   struct pipe_context *pipe = p->base.ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *tex = 0;
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_1D;
+   templ.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   templ.last_level = 0;
+   templ.width[0] = 1024;
+   templ.height[0] = 1;
+   templ.depth[0] = 1;
+   pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   tex = screen->texture_create(screen, &templ);
+
+   { /* upload color_data */
+      struct pipe_transfer *transfer =
+         st_no_flush_get_tex_transfer(p->base.ctx, tex, 0, 0, 0,
+                                      PIPE_TRANSFER_WRITE, 0, 0, 1024, 1);
+      void *map = screen->transfer_map(screen, transfer);
+      memcpy(map, p->gradient.color_data, sizeof(VGint)*1024);
+      screen->transfer_unmap(screen, transfer);
+      screen->tex_transfer_destroy(transfer);
+   }
+
+   return tex;
+}
+
+struct vg_paint * paint_create(struct vg_context *ctx)
+{
+   struct vg_paint *paint = CALLOC_STRUCT(vg_paint);
+   const VGfloat default_color[] = {0.0f, 0.0f, 0.0f, 1.0f};
+   const VGfloat def_ling[] = {0.0f, 0.0f, 1.0f, 0.0f};
+   const VGfloat def_radg[] = {0.0f, 0.0f, 0.0f, 0.0f, 1.0f};
+   vg_init_object(&paint->base, ctx, VG_OBJECT_PAINT);
+   vg_context_add_object(ctx, VG_OBJECT_PAINT, paint);
+
+   paint->type = VG_PAINT_TYPE_COLOR;
+   memcpy(paint->solid.color, default_color,
+          4 * sizeof(VGfloat));
+   paint->gradient.spread = VG_COLOR_RAMP_SPREAD_PAD;
+   memcpy(paint->gradient.linear.coords, def_ling,
+          4 * sizeof(VGfloat));
+   memcpy(paint->gradient.radial.vals, def_radg,
+          5 * sizeof(VGfloat));
+
+   paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   paint->gradient.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   paint->gradient.sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   paint->gradient.sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   paint->gradient.sampler.normalized_coords = 1;
+
+   memcpy(&paint->pattern.sampler,
+          &paint->gradient.sampler,
+          sizeof(struct pipe_sampler_state));
+
+   return paint;
+}
+
+void paint_destroy(struct vg_paint *paint)
+{
+   struct vg_context *ctx = paint->base.ctx;
+   if (paint->pattern.texture)
+      pipe_texture_reference(&paint->pattern.texture, NULL);
+   if (ctx)
+      vg_context_remove_object(ctx, VG_OBJECT_PAINT, paint);
+
+   free(paint->gradient.ramp_stopsi);
+   free(paint->gradient.ramp_stops);
+   free(paint);
+}
+
+void paint_set_color(struct vg_paint *paint,
+                     const VGfloat *color)
+{
+   paint->solid.color[0] = color[0];
+   paint->solid.color[1] = color[1];
+   paint->solid.color[2] = color[2];
+   paint->solid.color[3] = color[3];
+
+   paint->solid.colori[0] = FLT_TO_INT(color[0]);
+   paint->solid.colori[1] = FLT_TO_INT(color[1]);
+   paint->solid.colori[2] = FLT_TO_INT(color[2]);
+   paint->solid.colori[3] = FLT_TO_INT(color[3]);
+}
+
+static INLINE void paint_color_buffer(struct vg_paint *paint, void *buffer)
+{
+   VGfloat *map = (VGfloat*)buffer;
+   memcpy(buffer, paint->solid.color, 4 * sizeof(VGfloat));
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = 2.f;
+   map[7] = 4.f;
+}
+
+static INLINE void paint_linear_gradient_buffer(struct vg_paint *paint, void *buffer)
+{
+   struct vg_context *ctx = paint->base.ctx;
+   VGfloat *map = (VGfloat*)buffer;
+
+   map[0] = paint->gradient.linear.coords[2] - paint->gradient.linear.coords[0];
+   map[1] = paint->gradient.linear.coords[3] - paint->gradient.linear.coords[1];
+   map[2] = 1.f / (map[0] * map[0] + map[1] * map[1]);
+   map[3] = 1.f;
+
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = 2.f;
+   map[7] = 4.f;
+   {
+      struct matrix mat;
+      struct matrix inv;
+      matrix_load_identity(&mat);
+      matrix_translate(&mat, -paint->gradient.linear.coords[0], -paint->gradient.linear.coords[1]);
+      memcpy(&inv, &ctx->state.vg.fill_paint_to_user_matrix,
+             sizeof(struct matrix));
+      matrix_invert(&inv);
+      matrix_mult(&inv, &mat);
+      memcpy(&mat, &inv,
+             sizeof(struct matrix));
+
+      map[8]  = mat.m[0]; map[9]  = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f;
+      map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f;
+      map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f;
+   }
+#if 0
+   debug_printf("Coords  (%f, %f, %f, %f)\n",
+                map[0], map[1], map[2], map[3]);
+#endif
+}
+
+
+static INLINE void paint_radial_gradient_buffer(struct vg_paint *paint, void *buffer)
+{
+   VGfloat *radialCoords = paint->gradient.radial.vals;
+   struct vg_context *ctx = paint->base.ctx;
+
+   VGfloat *map = (VGfloat*)buffer;
+
+   map[0] = radialCoords[0] - radialCoords[2];
+   map[1] = radialCoords[1] - radialCoords[3];
+   map[2] = -map[0] * map[0] - map[1] * map[1] +
+            radialCoords[4] * radialCoords[4];
+   map[3] = 1.f;
+
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = 2.f;
+   map[7] = 4.f;
+
+   {
+      struct matrix mat;
+      struct matrix inv;
+      matrix_load_identity(&mat);
+      matrix_translate(&mat, -radialCoords[2], -radialCoords[3]);
+      memcpy(&inv, &ctx->state.vg.fill_paint_to_user_matrix,
+             sizeof(struct matrix));
+      matrix_invert(&inv);
+      matrix_mult(&inv, &mat);
+      memcpy(&mat, &inv,
+             sizeof(struct matrix));
+
+      map[8]  = mat.m[0]; map[9]  = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f;
+      map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f;
+      map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f;
+   }
+
+#if 0
+   debug_printf("Coords  (%f, %f, %f, %f)\n",
+                map[0], map[1], map[2], map[3]);
+#endif
+}
+
+
+static INLINE void  paint_pattern_buffer(struct vg_paint *paint, void *buffer)
+{
+   struct vg_context *ctx = paint->base.ctx;
+
+   VGfloat *map = (VGfloat *)buffer;
+   memcpy(map, paint->solid.color, 4 * sizeof(VGfloat));
+
+   map[4] = 0.f;
+   map[5] = 1.f;
+   map[6] = paint->pattern.texture->width[0];
+   map[7] = paint->pattern.texture->height[0];
+   {
+      struct matrix mat;
+      memcpy(&mat, &ctx->state.vg.fill_paint_to_user_matrix,
+             sizeof(struct matrix));
+      matrix_invert(&mat);
+      {
+         struct matrix pm;
+         memcpy(&pm, &ctx->state.vg.path_user_to_surface_matrix,
+                sizeof(struct matrix));
+         matrix_invert(&pm);
+         matrix_mult(&pm, &mat);
+         memcpy(&mat, &pm, sizeof(struct matrix));
+      }
+
+      map[8]  = mat.m[0]; map[9]  = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f;
+      map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f;
+      map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f;
+   }
+}
+
+void paint_set_type(struct vg_paint *paint, VGPaintType type)
+{
+   paint->type = type;
+}
+
+void paint_set_ramp_stops(struct vg_paint *paint, const VGfloat *stops,
+                          int num)
+{
+   const VGfloat default_stops[] = {0.0f, 0.0f, 0.0f, 0.0f, 1.0f,
+                                    1.0f, 1.0f, 1.0f, 1.0f, 1.0f};
+   VGint i;
+   const VGint num_stops = num / 5;
+   VGfloat last_coord;
+
+   paint->gradient.num_stops = num;
+   if (num) {
+      free(paint->gradient.ramp_stops);
+      paint->gradient.ramp_stops = malloc(sizeof(VGfloat)*num);
+      memcpy(paint->gradient.ramp_stops, stops, sizeof(VGfloat)*num);
+   } else
+      return;
+
+   /* stops must be in increasing order. the last stop is 1.0. if the
+    * first one is bigger than 1 then the whole sequence is invalid*/
+   if (stops[0] > 1) {
+      stops = default_stops;
+      num = 10;
+   }
+   last_coord = stops[0];
+   for (i = 1; i < num_stops; ++i) {
+      VGint idx = 5 * i;
+      VGfloat coord = stops[idx];
+      if (!floatsEqual(last_coord, coord) && coord < last_coord) {
+         stops = default_stops;
+         num = 10;
+         break;
+      }
+      last_coord = coord;
+   }
+
+   create_gradient_data(stops, num / 5, paint->gradient.color_data,
+                        1024);
+
+   if (paint->gradient.texture) {
+      pipe_texture_reference(&paint->gradient.texture, NULL);
+      paint->gradient.texture = 0;
+   }
+
+   paint->gradient.texture = create_gradient_texture(paint);
+}
+
+void paint_set_colori(struct vg_paint *p,
+                      VGuint rgba)
+{
+   p->solid.color[0] = ((rgba >> 24) & 0xff) / 255.f;
+   p->solid.color[1] = ((rgba >> 16) & 0xff) / 255.f;
+   p->solid.color[2] = ((rgba >>  8) & 0xff) / 255.f;
+   p->solid.color[3] = ((rgba >>  0) & 0xff) / 255.f;
+}
+
+VGuint paint_colori(struct vg_paint *p)
+{
+#define F2B(f) (float_to_ubyte(f))
+
+   return ((F2B(p->solid.color[0]) << 24) |
+           (F2B(p->solid.color[1]) << 16) |
+           (F2B(p->solid.color[2]) << 8)  |
+           (F2B(p->solid.color[3]) << 0));
+#undef F2B
+}
+
+void paint_set_linear_gradient(struct vg_paint *paint,
+                               const VGfloat *coords)
+{
+   memcpy(paint->gradient.linear.coords, coords, sizeof(VGfloat) * 4);
+}
+
+void paint_set_spread_mode(struct vg_paint *paint,
+                           VGint mode)
+{
+   paint->gradient.spread = mode;
+   switch(mode) {
+   case VG_COLOR_RAMP_SPREAD_PAD:
+      paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      break;
+   case VG_COLOR_RAMP_SPREAD_REPEAT:
+      paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+      break;
+   case VG_COLOR_RAMP_SPREAD_REFLECT:
+      paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      break;
+   }
+}
+
+VGColorRampSpreadMode paint_spread_mode(struct vg_paint *paint)
+{
+   return paint->gradient.spread;
+}
+
+void paint_set_radial_gradient(struct vg_paint *paint,
+                               const VGfloat *values)
+{
+   memcpy(paint->gradient.radial.vals, values, sizeof(VGfloat) * 5);
+}
+
+void paint_set_pattern(struct vg_paint *paint,
+                       struct vg_image *img)
+{
+   if (paint->pattern.texture)
+      pipe_texture_reference(&paint->pattern.texture, NULL);
+
+   paint->pattern.texture = 0;
+   pipe_texture_reference(&paint->pattern.texture,
+                          img->texture);
+}
+
+void paint_set_pattern_tiling(struct vg_paint *paint,
+                              VGTilingMode mode)
+{
+   paint->pattern.tiling_mode = mode;
+
+   switch(mode) {
+   case VG_TILE_FILL:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+      break;
+   case VG_TILE_PAD:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      break;
+   case VG_TILE_REPEAT:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_REPEAT;
+      break;
+   case VG_TILE_REFLECT:
+      paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_MIRROR_REPEAT;
+      break;
+   default:
+      debug_assert("!Unknown tiling mode");
+   }
+}
+
+void paint_get_color(struct vg_paint *paint,
+                     VGfloat *color)
+{
+   color[0] = paint->solid.color[0];
+   color[1] = paint->solid.color[1];
+   color[2] = paint->solid.color[2];
+   color[3] = paint->solid.color[3];
+}
+
+void paint_ramp_stops(struct vg_paint *paint, VGfloat *stops,
+                      int num)
+{
+   memcpy(stops, paint->gradient.ramp_stops, sizeof(VGfloat)*num);
+}
+
+void paint_linear_gradient(struct vg_paint *paint,
+                           VGfloat *coords)
+{
+   memcpy(coords, paint->gradient.linear.coords, sizeof(VGfloat)*4);
+}
+
+void paint_radial_gradient(struct vg_paint *paint,
+                           VGfloat *coords)
+{
+   memcpy(coords, paint->gradient.radial.vals, sizeof(VGfloat)*5);
+}
+
+int paint_num_ramp_stops(struct vg_paint *paint)
+{
+   return paint->gradient.num_stops;
+}
+
+VGPaintType paint_type(struct vg_paint *paint)
+{
+   return paint->type;
+}
+
+void paint_set_coloriv(struct vg_paint *paint,
+                      const VGint *color)
+{
+   paint->solid.color[0] = color[0];
+   paint->solid.color[1] = color[1];
+   paint->solid.color[2] = color[2];
+   paint->solid.color[3] = color[3];
+
+   paint->solid.colori[0] = color[0];
+   paint->solid.colori[1] = color[1];
+   paint->solid.colori[2] = color[2];
+   paint->solid.colori[3] = color[3];
+}
+
+void paint_get_coloriv(struct vg_paint *paint,
+                      VGint *color)
+{
+   color[0] = paint->solid.colori[0];
+   color[1] = paint->solid.colori[1];
+   color[2] = paint->solid.colori[2];
+   color[3] = paint->solid.colori[3];
+}
+
+void paint_set_color_ramp_premultiplied(struct vg_paint *paint,
+                                        VGboolean set)
+{
+   paint->gradient.color_ramps_premultiplied = set;
+}
+
+VGboolean paint_color_ramp_premultiplied(struct vg_paint *paint)
+{
+   return paint->gradient.color_ramps_premultiplied;
+}
+
+void paint_set_ramp_stopsi(struct vg_paint *paint, const VGint *stops,
+                           int num)
+{
+   if (num) {
+      free(paint->gradient.ramp_stopsi);
+      paint->gradient.ramp_stopsi = malloc(sizeof(VGint)*num);
+      memcpy(paint->gradient.ramp_stopsi, stops, sizeof(VGint)*num);
+   }
+}
+
+void paint_ramp_stopsi(struct vg_paint *paint, VGint *stops,
+                       int num)
+{
+   memcpy(stops, paint->gradient.ramp_stopsi, sizeof(VGint)*num);
+}
+
+void paint_set_linear_gradienti(struct vg_paint *paint,
+                                const VGint *coords)
+{
+   memcpy(paint->gradient.linear.coordsi, coords, sizeof(VGint) * 4);
+}
+
+void paint_linear_gradienti(struct vg_paint *paint,
+                            VGint *coords)
+{
+   memcpy(coords, paint->gradient.linear.coordsi, sizeof(VGint)*4);
+}
+
+void paint_set_radial_gradienti(struct vg_paint *paint,
+                                const VGint *values)
+{
+   memcpy(paint->gradient.radial.valsi, values, sizeof(VGint) * 5);
+}
+
+void paint_radial_gradienti(struct vg_paint *paint,
+                            VGint *coords)
+{
+   memcpy(coords, paint->gradient.radial.valsi, sizeof(VGint)*5);
+}
+
+VGTilingMode paint_pattern_tiling(struct vg_paint *paint)
+{
+   return paint->pattern.tiling_mode;
+}
+
+VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures)
+{
+   struct vg_context *ctx = vg_current_context();
+
+   switch(paint->type) {
+   case VG_PAINT_TYPE_LINEAR_GRADIENT:
+   case VG_PAINT_TYPE_RADIAL_GRADIENT: {
+      if (paint->gradient.texture) {
+         paint->gradient.sampler.min_img_filter = image_sampler_filter(ctx);
+         paint->gradient.sampler.mag_img_filter = image_sampler_filter(ctx);
+         samplers[0] = &paint->gradient.sampler;
+         textures[0] = paint->gradient.texture;
+         return 1;
+      }
+   }
+      break;
+   case VG_PAINT_TYPE_PATTERN: {
+      memcpy(paint->pattern.sampler.border_color,
+             ctx->state.vg.tile_fill_color,
+             sizeof(VGfloat) * 4);
+      paint->pattern.sampler.min_img_filter = image_sampler_filter(ctx);
+      paint->pattern.sampler.mag_img_filter = image_sampler_filter(ctx);
+      samplers[0] = &paint->pattern.sampler;
+      textures[0] = paint->pattern.texture;
+      return 1;
+   }
+      break;
+   default:
+      samplers[0] = &paint->pattern.sampler; /* dummy */
+      textures[0] = 0;
+      return 0;
+      break;
+   }
+   return 0;
+}
+
+void paint_resolve_type(struct vg_paint *paint)
+{
+   if (paint->type == VG_PAINT_TYPE_PATTERN &&
+       !paint->pattern.texture) {
+      paint->type = VG_PAINT_TYPE_COLOR;
+   }
+}
+
+VGint paint_constant_buffer_size(struct vg_paint *paint)
+{
+   switch(paint->type) {
+   case VG_PAINT_TYPE_COLOR:
+      return 8 * sizeof(VGfloat);/*4 color + 4 constants (0.f,1.f,2.f,4.f)*/
+      break;
+   case VG_PAINT_TYPE_LINEAR_GRADIENT:
+      return 20 * sizeof(VGfloat);
+      break;
+   case VG_PAINT_TYPE_RADIAL_GRADIENT:
+      return 20 * sizeof(VGfloat);
+      break;
+   case VG_PAINT_TYPE_PATTERN:
+      return 20 * sizeof(VGfloat);
+      break;
+   default:
+      debug_printf("Uknown paint type: %d\n", paint->type);
+   }
+
+   return 0;
+}
+
+void paint_fill_constant_buffer(struct vg_paint *paint,
+                                void *buffer)
+{
+   switch(paint->type) {
+   case VG_PAINT_TYPE_COLOR:
+      paint_color_buffer(paint, buffer);
+      break;
+   case VG_PAINT_TYPE_LINEAR_GRADIENT:
+      paint_linear_gradient_buffer(paint, buffer);
+      break;
+   case VG_PAINT_TYPE_RADIAL_GRADIENT:
+      paint_radial_gradient_buffer(paint, buffer);
+      break;
+   case VG_PAINT_TYPE_PATTERN:
+      paint_pattern_buffer(paint, buffer);
+      break;
+
+   default:
+      abort();
+   }
+}
diff --git a/src/gallium/state_trackers/vega/paint.h b/src/gallium/state_trackers/vega/paint.h
new file mode 100644
index 0000000000..999b5c167c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/paint.h
@@ -0,0 +1,118 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef PAINT_H
+#define PAINT_H
+
+#include "vg_context.h"
+
+#include "VG/openvg.h"
+#include "pipe/p_state.h"
+
+struct vg_paint;
+struct vg_image;
+struct pipe_sampler_state;
+struct pipe_texture;
+
+struct vg_paint *paint_create(struct vg_context *ctx);
+void paint_destroy(struct vg_paint *paint);
+
+void paint_set_color(struct vg_paint *paint,
+                     const VGfloat *color);
+void paint_get_color(struct vg_paint *paint,
+                     VGfloat *color);
+
+void paint_set_coloriv(struct vg_paint *paint,
+                      const VGint *color);
+void paint_get_coloriv(struct vg_paint *paint,
+                      VGint *color);
+
+void paint_set_colori(struct vg_paint *paint,
+                      VGuint rgba);
+
+VGuint paint_colori(struct vg_paint *paint);
+
+void paint_set_type(struct vg_paint *paint, VGPaintType type);
+VGPaintType paint_type(struct vg_paint *paint);
+void paint_resolve_type(struct vg_paint *paint);
+
+void paint_set_linear_gradient(struct vg_paint *paint,
+                               const VGfloat *coords);
+void paint_linear_gradient(struct vg_paint *paint,
+                           VGfloat *coords);
+void paint_set_linear_gradienti(struct vg_paint *paint,
+                               const VGint *coords);
+void paint_linear_gradienti(struct vg_paint *paint,
+                           VGint *coords);
+
+
+void paint_set_radial_gradient(struct vg_paint *paint,
+                               const VGfloat *values);
+void paint_radial_gradient(struct vg_paint *paint,
+                           VGfloat *coords);
+void paint_set_radial_gradienti(struct vg_paint *paint,
+                                const VGint *values);
+void paint_radial_gradienti(struct vg_paint *paint,
+                            VGint *coords);
+
+
+void paint_set_ramp_stops(struct vg_paint *paint, const VGfloat *stops,
+                          int num);
+void paint_ramp_stops(struct vg_paint *paint, VGfloat *stops,
+                      int num);
+
+void paint_set_ramp_stopsi(struct vg_paint *paint, const VGint *stops,
+                           int num);
+void paint_ramp_stopsi(struct vg_paint *paint, VGint *stops,
+                       int num);
+
+int paint_num_ramp_stops(struct vg_paint *paint);
+
+void paint_set_spread_mode(struct vg_paint *paint,
+                           VGint mode);
+VGColorRampSpreadMode paint_spread_mode(struct vg_paint *paint);
+
+
+void paint_set_pattern(struct vg_paint *paint,
+                       struct vg_image *img);
+void paint_set_pattern_tiling(struct vg_paint *paint,
+                              VGTilingMode mode);
+VGTilingMode paint_pattern_tiling(struct vg_paint *paint);
+
+void paint_set_color_ramp_premultiplied(struct vg_paint *paint,
+                                        VGboolean set);
+VGboolean paint_color_ramp_premultiplied(struct vg_paint *paint);
+
+
+VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **samplers,
+                          struct pipe_texture **textures);
+
+VGint paint_constant_buffer_size(struct vg_paint *paint);
+void paint_fill_constant_buffer(struct vg_paint *paint,
+                                void *buffer);
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/path.c b/src/gallium/state_trackers/vega/path.c
new file mode 100644
index 0000000000..d04f9d9ae6
--- /dev/null
+++ b/src/gallium/state_trackers/vega/path.c
@@ -0,0 +1,2034 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "path.h"
+
+#include "stroker.h"
+#include "polygon.h"
+#include "bezier.h"
+#include "matrix.h"
+#include "vg_context.h"
+#include "util_array.h"
+#include "arc.h"
+#include "path_utils.h"
+#include "paint.h"
+#include "shader.h"
+
+#include "util/u_memory.h"
+
+#include <assert.h>
+
+#define DEBUG_PATH 0
+
+struct path {
+   struct vg_object base;
+   VGbitfield caps;
+   VGboolean dirty;
+   VGboolean dirty_stroke;
+
+   VGPathDatatype datatype;
+
+   VGfloat scale;
+   VGfloat bias;
+
+   VGint num_segments;
+
+   struct array * segments;
+   struct array * control_points;
+
+   struct {
+      struct polygon_array polygon_array;
+      struct matrix matrix;
+   } fill_polys;
+
+   struct {
+      struct path *path;
+      struct matrix matrix;
+      VGfloat stroke_width;
+      VGfloat miter_limit;
+      VGCapStyle cap_style;
+      VGJoinStyle join_style;
+   } stroked;
+};
+
+
+static INLINE void data_at(void **data,
+                           struct path *p,
+                           VGint start, VGint count,
+                           VGfloat *out)
+{
+   VGPathDatatype dt = p->datatype;
+   VGint i;
+   VGint end = start + count;
+   VGfloat *itr = out;
+
+   switch(dt) {
+   case VG_PATH_DATATYPE_S_8: {
+      VGbyte **bdata = (VGbyte **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*bdata)[i];
+         ++itr;
+      }
+      *bdata += count;
+   }
+      break;
+   case VG_PATH_DATATYPE_S_16: {
+      VGshort **bdata = (VGshort **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*bdata)[i];
+         ++itr;
+      }
+      *bdata += count;
+   }
+      break;
+   case VG_PATH_DATATYPE_S_32: {
+      VGint **bdata = (VGint **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*bdata)[i];
+         ++itr;
+      }
+      *bdata += count;
+   }
+      break;
+   case VG_PATH_DATATYPE_F: {
+      VGfloat **fdata = (VGfloat **)data;
+      for (i = start; i < end; ++i) {
+         *itr = (*fdata)[i];
+         ++itr;
+      }
+      *fdata += count;
+   }
+      break;
+   default:
+      debug_assert(!"Unknown path datatype!");
+   }
+}
+
+
+void vg_float_to_datatype(VGPathDatatype datatype,
+                          VGubyte *common_data,
+                          const VGfloat *data,
+                          VGint num_coords)
+{
+   VGint i;
+   switch(datatype) {
+   case VG_PATH_DATATYPE_S_8: {
+      for (i = 0; i < num_coords; ++i) {
+         common_data[i] = (VGubyte)data[i];
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE_S_16: {
+      VGshort *buf = (VGshort*)common_data;
+      for (i = 0; i < num_coords; ++i) {
+         buf[i] = (VGshort)data[i];
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE_S_32: {
+      VGint *buf = (VGint*)common_data;
+      for (i = 0; i < num_coords; ++i) {
+         buf[i] = (VGint)data[i];
+      }
+   }
+      break;
+   case VG_PATH_DATATYPE_F: {
+      memcpy(common_data, data, sizeof(VGfloat) * num_coords);
+   }
+      break;
+   default:
+      debug_assert(!"Unknown path datatype!");
+   }
+}
+
+static void coords_adjust_by_scale_bias(struct path *p,
+                                        void *pdata, VGint num_coords,
+                                        VGfloat scale, VGfloat bias,
+                                        VGPathDatatype datatype)
+{
+   VGfloat data[8];
+   void *coords = (VGfloat *)pdata;
+   VGubyte *common_data = (VGubyte *)pdata;
+   VGint size_dst = size_for_datatype(datatype);
+   VGint i;
+
+   for (i = 0; i < num_coords; ++i) {
+      data_at(&coords, p, 0, 1, data);
+      data[0] = data[0] * scale + bias;
+      vg_float_to_datatype(datatype, common_data, data, 1);
+      common_data += size_dst;
+   }
+}
+
+struct path * path_create(VGPathDatatype dt, VGfloat scale, VGfloat bias,
+                          VGint segmentCapacityHint,
+                          VGint coordCapacityHint,
+                          VGbitfield capabilities)
+{
+   struct path *path = CALLOC_STRUCT(path);
+
+   vg_init_object(&path->base, vg_current_context(), VG_OBJECT_PATH);
+   path->caps = capabilities & VG_PATH_CAPABILITY_ALL;
+   vg_context_add_object(vg_current_context(), VG_OBJECT_PATH, path);
+
+   path->datatype = dt;
+   path->scale = scale;
+   path->bias = bias;
+
+   path->segments = array_create(size_for_datatype(VG_PATH_DATATYPE_S_8));
+   path->control_points = array_create(size_for_datatype(dt));
+
+   path->dirty = VG_TRUE;
+   path->dirty_stroke = VG_TRUE;
+
+   return path;
+}
+
+void path_destroy(struct path *p)
+{
+   vg_context_remove_object(vg_current_context(), VG_OBJECT_PATH, p);
+
+   array_destroy(p->segments);
+   array_destroy(p->control_points);
+   array_destroy(p->fill_polys.polygon_array.array);
+
+   if (p->stroked.path)
+      path_destroy(p->stroked.path);
+
+   free(p);
+}
+
+VGbitfield path_capabilities(struct path *p)
+{
+   return p->caps;
+}
+
+void path_set_capabilities(struct path *p, VGbitfield bf)
+{
+   p->caps = (bf & VG_PATH_CAPABILITY_ALL);
+}
+
+void path_append_data(struct path *p,
+                      VGint numSegments,
+                      const VGubyte * pathSegments,
+                      const void * pathData)
+{
+   VGint old_segments = p->num_segments;
+   VGint num_new_coords = num_elements_for_segments(pathSegments, numSegments);
+   array_append_data(p->segments, pathSegments, numSegments);
+   array_append_data(p->control_points, pathData, num_new_coords);
+
+   p->num_segments += numSegments;
+   if (!floatsEqual(p->scale, 1.f) || !floatsEqual(p->bias, 0.f)) {
+      VGubyte *coords = (VGubyte*)p->control_points->data;
+      coords_adjust_by_scale_bias(p,
+                                  coords + old_segments * p->control_points->datatype_size,
+                                  num_new_coords,
+                                  p->scale, p->bias, p->datatype);
+   }
+   p->dirty = VG_TRUE;
+   p->dirty_stroke = VG_TRUE;
+}
+
+VGint path_num_segments(struct path *p)
+{
+   return p->num_segments;
+}
+
+static INLINE void map_if_relative(VGfloat ox, VGfloat oy,
+                                   VGboolean relative,
+                                   VGfloat *x, VGfloat *y)
+{
+   if (relative) {
+      if (x)
+         *x += ox;
+      if (y)
+         *y += oy;
+   }
+}
+
+static INLINE void close_polygon(struct polygon *current,
+                                 VGfloat sx, VGfloat sy,
+                                 VGfloat ox, VGfloat oy,
+                                 struct  matrix *matrix)
+{
+   if (!floatsEqual(sx, ox) ||
+       !floatsEqual(sy, oy)) {
+      VGfloat x0 = sx;
+      VGfloat y0 = sy;
+      matrix_map_point(matrix, x0, y0, &x0, &y0);
+      polygon_vertex_append(current, x0, y0);
+   }
+}
+
+static void convert_path(struct path *p,
+                          VGPathDatatype to,
+                          void *dst,
+                          VGint num_coords)
+{
+   VGfloat data[8];
+   void *coords = (VGfloat *)p->control_points->data;
+   VGubyte *common_data = (VGubyte *)dst;
+   VGint size_dst = size_for_datatype(to);
+   VGint i;
+
+   for (i = 0; i < num_coords; ++i) {
+      data_at(&coords, p, 0, 1, data);
+      vg_float_to_datatype(to, common_data, data, 1);
+      common_data += size_dst;
+   }
+}
+
+
+static void polygon_array_calculate_bounds( struct polygon_array *polyarray )
+{
+   struct array *polys = polyarray->array;
+   VGfloat min_x, max_x;
+   VGfloat min_y, max_y;
+   VGfloat bounds[4];
+   unsigned i;
+
+   assert(polys);
+   assert(polys->num_elements);
+   polygon_bounding_rect((((struct polygon**)polys->data)[0]), bounds);
+   min_x = bounds[0];
+   min_y = bounds[1];
+   max_x = bounds[0] + bounds[2];
+   max_y = bounds[1] + bounds[3];
+   for (i = 1; i < polys->num_elements; ++i) {
+      struct polygon *p = (((struct polygon**)polys->data)[i]);
+      polygon_bounding_rect(p, bounds);
+      min_x = MIN2(min_x, bounds[0]);
+      min_y = MIN2(min_y, bounds[1]);
+      max_x = MAX2(max_x, bounds[0] + bounds[2]);
+      max_y = MAX2(max_y, bounds[1] + bounds[3]);
+   }
+
+   polyarray->min_x = min_x;
+   polyarray->min_y = min_y;
+   polyarray->max_x = max_x;
+   polyarray->max_y = max_y;
+}
+
+
+static struct polygon_array * path_get_fill_polygons(struct path *p, struct matrix *matrix)
+{
+   VGint i;
+   struct polygon *current = 0;
+   VGfloat sx, sy, px, py, ox, oy;
+   VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+   VGfloat data[8];
+   void *coords = (VGfloat *)p->control_points->data;
+   struct array *array;
+
+   if (p->fill_polys.polygon_array.array)
+   {
+      if (memcmp( &p->fill_polys.matrix,
+                  matrix,
+                  sizeof *matrix ) == 0 && p->dirty == VG_FALSE)
+      {
+         return &p->fill_polys.polygon_array;
+      }
+      else {
+         array_destroy( p->fill_polys.polygon_array.array );
+         p->fill_polys.polygon_array.array = NULL;
+      }
+   }
+
+   array = array_create(sizeof(struct array*));
+
+   sx = sy = px = py = ox = oy = 0.f;
+
+   current = polygon_create(32);
+
+   for (i = 0; i < p->num_segments; ++i) {
+      VGubyte segment = ((VGubyte*)(p->segments->data))[i];
+      VGint command = SEGMENT_COMMAND(segment);
+      VGboolean relative = SEGMENT_ABS_REL(segment);
+
+      switch(command) {
+      case VG_CLOSE_PATH:
+         close_polygon(current, sx, sy, ox, oy, matrix);
+         ox = sx;
+         oy = sy;
+         break;
+      case VG_MOVE_TO:
+         if (current && polygon_vertex_count(current) > 0) {
+            /* add polygon */
+            close_polygon(current, sx, sy, ox, oy, matrix);
+            array_append_data(array, &current, 1);
+            current = polygon_create(32);
+         }
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         sx = x0;
+         sy = y0;
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_LINE_TO:
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_HLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = data[0];
+         y0 = oy;
+         map_if_relative(ox, oy, relative, &x0, 0);
+         ox = x0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_VLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = ox;
+         y0 = data[0];
+         map_if_relative(ox, oy, relative, 0, &y0);
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         polygon_vertex_append(current, x0, y0);
+         break;
+      case VG_CUBIC_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 6, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x2 = data[2];
+         y2 = data[3];
+         x3 = data[4];
+         y3 = data[5];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                       x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_QUAD_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                       x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_SQUAD_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 2, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x3 = data[0];
+         y3 = data[1];
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                     x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_SCUBIC_TO: {
+         struct bezier bezier;
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x2 = data[0];
+         y2 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         bezier_init(&bezier, x0, y0, x1, y1,
+                              x2, y2, x3, y3);
+         bezier_add_to_polygon(&bezier, current);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         VGfloat rh, rv, rot;
+         struct arc arc;
+
+         data_at(&coords, p, 0, 5, data);
+         x0  = ox;
+         y0  = oy;
+         rh  = data[0];
+         rv  = data[1];
+         rot = data[2];
+         x1  = data[3];
+         y1  = data[4];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+#if 0
+         debug_printf("------- ARC (%f, %f), (%f, %f) %f, %f, %f\n",
+                      x0, y0, x1, y1, rh, rv, rot);
+#endif
+         arc_init(&arc, command, x0, y0, x1, y1,
+                  rh, rv, rot);
+         arc_add_to_polygon(&arc, current,
+                            matrix);
+         ox = x1;
+         oy = y1;
+         px = x1;
+         py = y1;
+      }
+         break;
+      default:
+         abort();
+         assert(!"Unknown segment!");
+      }
+   }
+   if (current) {
+      if (polygon_vertex_count(current) > 0) {
+         close_polygon(current, sx, sy, ox, oy, matrix);
+         array_append_data(array, &current, 1);
+      } else
+         polygon_destroy(current);
+   }
+
+   p->fill_polys.polygon_array.array = array;
+   p->fill_polys.matrix = *matrix;
+
+   polygon_array_calculate_bounds( &p->fill_polys.polygon_array );
+
+   p->dirty = VG_FALSE;
+
+   return &p->fill_polys.polygon_array;
+}
+
+VGbyte path_datatype_size(struct path *p)
+{
+   return size_for_datatype(p->datatype);
+}
+
+VGPathDatatype path_datatype(struct path *p)
+{
+   return p->datatype;
+}
+
+VGfloat path_scale(struct path *p)
+{
+   return p->scale;
+}
+
+VGfloat path_bias(struct path *p)
+{
+   return p->bias;
+}
+
+VGint path_num_coords(struct path *p)
+{
+   return num_elements_for_segments((VGubyte*)p->segments->data,
+                                    p->num_segments);
+}
+
+void path_modify_coords(struct path *p,
+                        VGint startIndex,
+                        VGint numSegments,
+                        const void * pathData)
+{
+   VGubyte *segments = (VGubyte*)(p->segments->data);
+   VGint count = num_elements_for_segments(&segments[startIndex], numSegments);
+   VGint start_cp = num_elements_for_segments(segments, startIndex);
+
+   array_change_data(p->control_points, pathData, start_cp, count);
+   coords_adjust_by_scale_bias(p,
+                               ((VGubyte*)p->control_points->data) +
+                               (startIndex * p->control_points->datatype_size),
+                               path_num_coords(p),
+                               p->scale, p->bias, p->datatype);
+   p->dirty = VG_TRUE;
+   p->dirty_stroke = VG_TRUE;
+}
+
+void path_for_each_segment(struct path *path,
+                           path_for_each_cb cb,
+                           void *user_data)
+{
+   VGint i;
+   struct path_for_each_data p;
+   VGfloat data[8];
+   void *coords = (VGfloat *)path->control_points->data;
+
+   p.coords = data;
+   p.sx = p.sy = p.px = p.py = p.ox = p.oy = 0.f;
+   p.user_data = user_data;
+
+   for (i = 0; i < path->num_segments; ++i) {
+      VGint command;
+      VGboolean relative;
+
+      p.segment = ((VGubyte*)(path->segments->data))[i];
+      command = SEGMENT_COMMAND(p.segment);
+      relative = SEGMENT_ABS_REL(p.segment);
+
+      switch(command) {
+      case VG_CLOSE_PATH:
+         cb(path, &p);
+         break;
+      case VG_MOVE_TO:
+         data_at(&coords, path, 0, 2, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         cb(path, &p);
+         p.sx = data[0];
+         p.sy = data[1];
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_LINE_TO:
+         data_at(&coords, path, 0, 2, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         cb(path, &p);
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_HLINE_TO:
+         data_at(&coords, path, 0, 1, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], 0);
+         p.segment = VG_LINE_TO;
+         data[1] = p.oy;
+         cb(path, &p);
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_VLINE_TO:
+         data_at(&coords, path, 0, 1, data);
+         map_if_relative(p.ox, p.oy, relative, 0, &data[0]);
+         p.segment = VG_LINE_TO;
+         data[1] = data[0];
+         data[0] = p.ox;
+         cb(path, &p);
+         p.ox = data[0];
+         p.oy = data[1];
+         p.px = data[0];
+         p.py = data[1];
+         break;
+      case VG_CUBIC_TO: {
+         data_at(&coords, path, 0, 6, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]);
+         map_if_relative(p.ox, p.oy, relative, &data[4], &data[5]);
+         cb(path, &p);
+         p.px = data[2];
+         p.py = data[3];
+         p.ox = data[4];
+         p.oy = data[5];
+      }
+         break;
+      case VG_QUAD_TO: {
+         data_at(&coords, path, 0, 4, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]);
+         cb(path, &p);
+         p.px = data[0];
+         p.py = data[1];
+         p.ox = data[2];
+         p.oy = data[3];
+      }
+         break;
+      case VG_SQUAD_TO: {
+         data_at(&coords, path, 0, 2, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         cb(path, &p);
+         p.px = 2*p.ox-p.px;
+         p.py = 2*p.oy-p.py;
+         p.ox = data[2];
+         p.oy = data[3];
+      }
+         break;
+      case VG_SCUBIC_TO: {
+         data_at(&coords, path, 0, 4, data);
+         map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]);
+         map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]);
+         cb(path, &p);
+         p.px = data[0];
+         p.py = data[1];
+         p.ox = data[2];
+         p.oy = data[3];
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         data_at(&coords, path, 0, 5, data);
+         map_if_relative(p.ox, p.oy, relative, &data[3], &data[4]);
+#if 0
+         debug_printf("------- ARC (%f, %f), (%f, %f) %f, %f, %f\n",
+                      p.ox, p.oy, data[3], data[4], data[0], data[1], data[2]);
+#endif
+         cb(path, &p);
+         p.ox = data[3];
+         p.oy = data[4];
+         p.px = data[3];
+         p.py = data[4];
+      }
+         break;
+      default:
+         abort();
+         assert(!"Unknown segment!");
+      }
+   }
+}
+
+struct transform_data {
+   struct array *segments;
+   struct array *coords;
+
+   struct matrix *matrix;
+
+   VGPathDatatype datatype;
+};
+
+static VGboolean transform_cb(struct path *p,
+                              struct path_for_each_data *pd)
+{
+   struct transform_data *td = (struct transform_data *)pd->user_data;
+   VGint num_coords = num_elements_for_segments(&pd->segment, 1);
+   VGubyte segment = SEGMENT_COMMAND(pd->segment);/* abs bit is 0 */
+   VGfloat data[8];
+   VGubyte common_data[sizeof(VGfloat)*8];
+
+   memcpy(data, pd->coords, sizeof(VGfloat) * num_coords);
+
+   switch(segment) {
+   case VG_CLOSE_PATH:
+      break;
+   case VG_MOVE_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      break;
+   case VG_LINE_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      break;
+   case VG_HLINE_TO:
+   case VG_VLINE_TO:
+      assert(0);
+      break;
+   case VG_QUAD_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      matrix_map_point(td->matrix,
+                       data[2], data[3], &data[2], &data[3]);
+      break;
+   case VG_CUBIC_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      matrix_map_point(td->matrix,
+                       data[2], data[3], &data[2], &data[3]);
+      matrix_map_point(td->matrix,
+                       data[4], data[5], &data[4], &data[5]);
+      break;
+   case VG_SQUAD_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      break;
+   case VG_SCUBIC_TO:
+      matrix_map_point(td->matrix,
+                       data[0], data[1], &data[0], &data[1]);
+      matrix_map_point(td->matrix,
+                       data[2], data[3], &data[2], &data[3]);
+      break;
+   case VG_SCCWARC_TO:
+   case VG_SCWARC_TO:
+   case VG_LCCWARC_TO:
+   case VG_LCWARC_TO: {
+      struct arc arc;
+      struct path *path = path_create(td->datatype,
+                                      1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+      arc_init(&arc, segment,
+               pd->ox, pd->oy, data[3], data[4],
+               data[0], data[1], data[2]);
+
+      arc_to_path(&arc, path, td->matrix);
+
+      num_coords = path_num_coords(path);
+
+      array_append_data(td->segments, path->segments->data,
+                        path->num_segments);
+      array_append_data(td->coords, path->control_points->data,
+                        num_coords);
+      path_destroy(path);
+
+      return VG_TRUE;
+   }
+      break;
+   default:
+      break;
+   }
+
+   vg_float_to_datatype(td->datatype, common_data, data, num_coords);
+
+   array_append_data(td->segments, &segment, 1);
+   array_append_data(td->coords, common_data, num_coords);
+   return VG_TRUE;
+}
+
+void path_transform(struct path *dst, struct path *src)
+{
+   struct transform_data data;
+   struct vg_context *ctx = dst->base.ctx;
+
+   data.segments =  dst->segments;
+   data.coords   =  dst->control_points;
+   data.matrix   = &ctx->state.vg.path_user_to_surface_matrix;
+   data.datatype = dst->datatype;
+
+   path_for_each_segment(src, transform_cb, (void*)&data);
+
+   dst->num_segments = dst->segments->num_elements;
+   dst->dirty = VG_TRUE;
+   dst->dirty_stroke = VG_TRUE;
+}
+
+void path_append_path(struct path *dst,
+                      struct path *src)
+{
+   VGint num_coords = path_num_coords(src);
+   void *dst_data = malloc(size_for_datatype(dst->datatype) * num_coords);
+   array_append_data(dst->segments,
+                     src->segments->data,
+                     src->num_segments);
+   convert_path(src, dst->datatype,
+                dst_data, num_coords);
+   array_append_data(dst->control_points,
+                     dst_data,
+                     num_coords);
+   free(dst_data);
+
+   dst->num_segments += src->num_segments;
+   dst->dirty = VG_TRUE;
+   dst->dirty_stroke = VG_TRUE;
+}
+
+static INLINE VGboolean is_segment_arc(VGubyte segment)
+{
+   VGubyte scommand = SEGMENT_COMMAND(segment);
+   return (scommand == VG_SCCWARC_TO ||
+           scommand == VG_SCWARC_TO ||
+           scommand == VG_LCCWARC_TO ||
+           scommand == VG_LCWARC_TO);
+}
+
+struct path_iter_data {
+   struct path *path;
+   VGubyte segment;
+   void *coords;
+   VGfloat px, py, ox, oy, sx, sy;
+};
+static INLINE VGubyte normalize_coords(struct path_iter_data *pd,
+                                       VGint *num_coords,
+                                       VGfloat *data)
+{
+   VGint command = SEGMENT_COMMAND(pd->segment);
+   VGboolean relative = SEGMENT_ABS_REL(pd->segment);
+
+   switch(command) {
+   case VG_CLOSE_PATH:
+      *num_coords = 0;
+      pd->ox = pd->sx;
+      pd->oy = pd->sy;
+      return VG_CLOSE_PATH;
+      break;
+   case VG_MOVE_TO:
+      data_at(&pd->coords, pd->path, 0, 2, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]);
+      pd->sx = data[0];
+      pd->sy = data[1];
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_MOVE_TO_ABS;
+      break;
+   case VG_LINE_TO:
+      data_at(&pd->coords, pd->path, 0, 2, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]);
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_LINE_TO_ABS;
+      break;
+   case VG_HLINE_TO:
+      data_at(&pd->coords, pd->path, 0, 1, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], 0);
+      data[1] = pd->oy;
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_LINE_TO_ABS;
+      break;
+   case VG_VLINE_TO:
+      data_at(&pd->coords, pd->path, 0, 1, data);
+      map_if_relative(pd->ox, pd->oy, relative, 0, &data[0]);
+      data[1] = data[0];
+      data[0] = pd->ox;
+      pd->ox = data[0];
+      pd->oy = data[1];
+      pd->px = data[0];
+      pd->py = data[1];
+      *num_coords = 2;
+      return VG_LINE_TO_ABS;
+      break;
+   case VG_CUBIC_TO: {
+      data_at(&pd->coords, pd->path, 0, 6, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]);
+      map_if_relative(pd->ox, pd->oy, relative, &data[2], &data[3]);
+      map_if_relative(pd->ox, pd->oy, relative, &data[4], &data[5]);
+      pd->px = data[2];
+      pd->py = data[3];
+      pd->ox = data[4];
+      pd->oy = data[5];
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_QUAD_TO: {
+      VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+      data_at(&pd->coords, pd->path, 0, 4, data);
+      x0 = pd->ox;
+      y0 = pd->oy;
+      x1 = data[0];
+      y1 = data[1];
+      x3 = data[2];
+      y3 = data[3];
+      map_if_relative(pd->ox, pd->oy, relative, &x1, &y1);
+      map_if_relative(pd->ox, pd->oy, relative, &x3, &y3);
+      pd->px = x1;
+      pd->py = y1;
+      { /* form a cubic out of it */
+         x2 = (x3 + 2*x1) / 3.f;
+         y2 = (y3 + 2*y1) / 3.f;
+         x1 = (x0 + 2*x1) / 3.f;
+         y1 = (y0 + 2*y1) / 3.f;
+      }
+      pd->ox = x3;
+      pd->oy = y3;
+      data[0] = x1;
+      data[1] = y1;
+      data[2] = x2;
+      data[3] = y2;
+      data[4] = x3;
+      data[5] = y3;
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_SQUAD_TO: {
+      VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+      data_at(&pd->coords, pd->path, 0, 2, data);
+      x0 = pd->ox;
+      y0 = pd->oy;
+      x1 = 2 * pd->ox - pd->px;
+      y1 = 2 * pd->oy - pd->py;
+      x3 = data[0];
+      y3 = data[1];
+      map_if_relative(pd->ox, pd->oy, relative, &x3, &y3);
+      pd->px = x1;
+      pd->py = y1;
+      { /* form a cubic out of it */
+         x2 = (x3 + 2*x1) / 3.f;
+         y2 = (y3 + 2*y1) / 3.f;
+         x1 = (x0 + 2*x1) / 3.f;
+         y1 = (y0 + 2*y1) / 3.f;
+      }
+      pd->ox = x3;
+      pd->oy = y3;
+      data[0] = x1;
+      data[1] = y1;
+      data[2] = x2;
+      data[3] = y2;
+      data[4] = x3;
+      data[5] = y3;
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_SCUBIC_TO: {
+      VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+      data_at(&pd->coords, pd->path, 0, 4, data);
+      x0 = pd->ox;
+      y0 = pd->oy;
+      x1 = 2*pd->ox-pd->px;
+      y1 = 2*pd->oy-pd->py;
+      x2 = data[0];
+      y2 = data[1];
+      x3 = data[2];
+      y3 = data[3];
+      map_if_relative(pd->ox, pd->oy, relative, &x2, &y2);
+      map_if_relative(pd->ox, pd->oy, relative, &x3, &y3);
+      pd->ox = x3;
+      pd->oy = y3;
+      pd->px = x2;
+      pd->py = y2;
+      data[0] = x1;
+      data[1] = y1;
+      data[2] = x2;
+      data[3] = y2;
+      data[4] = x3;
+      data[5] = y3;
+      *num_coords = 6;
+      return VG_CUBIC_TO_ABS;
+   }
+      break;
+   case VG_SCCWARC_TO:
+   case VG_SCWARC_TO:
+   case VG_LCCWARC_TO:
+   case VG_LCWARC_TO: {
+      data_at(&pd->coords, pd->path, 0, 5, data);
+      map_if_relative(pd->ox, pd->oy, relative, &data[3], &data[4]);
+      pd->ox = data[3];
+      pd->oy = data[4];
+      pd->px = data[3];
+      pd->py = data[4];
+      *num_coords = 5;
+      return command | VG_ABSOLUTE;
+   }
+      break;
+   default:
+      abort();
+      assert(!"Unknown segment!");
+   }
+}
+
+static void linearly_interpolate(VGfloat *result,
+                                 const VGfloat *start,
+                                 const VGfloat *end,
+                                 VGfloat amount,
+                                 VGint number)
+{
+   VGint i;
+   for (i = 0; i < number; ++i) {
+      result[i] = start[i] + (end[i] - start[i]) * amount;
+   }
+}
+
+VGboolean path_interpolate(struct path *dst,
+                           struct path *start, struct path *end,
+                           VGfloat amount)
+{
+   /* temporary path that we can discard if it will turn
+    * out that start is not compatible with end */
+   struct path *res_path = path_create(dst->datatype,
+                                       1.0, 0.0,
+                                       0, 0, dst->caps);
+   VGint i;
+   VGfloat start_coords[8];
+   VGfloat end_coords[8];
+   VGfloat results[8];
+   VGubyte common_data[sizeof(VGfloat)*8];
+   struct path_iter_data start_iter, end_iter;
+
+   memset(&start_iter, 0, sizeof(struct path_iter_data));
+   memset(&end_iter, 0, sizeof(struct path_iter_data));
+
+   start_iter.path = start;
+   start_iter.coords = start->control_points->data;
+   end_iter.path = end;
+   end_iter.coords = end->control_points->data;
+
+   for (i = 0; i < start->num_segments; ++i) {
+      VGubyte segment;
+      VGubyte ssegment, esegment;
+      VGint snum_coords, enum_coords;
+      start_iter.segment = ((VGubyte*)(start->segments->data))[i];
+      end_iter.segment = ((VGubyte*)(end->segments->data))[i];
+
+      ssegment = normalize_coords(&start_iter, &snum_coords,
+                                  start_coords);
+      esegment = normalize_coords(&end_iter, &enum_coords,
+                                  end_coords);
+
+      if (is_segment_arc(ssegment)) {
+         if (!is_segment_arc(esegment)) {
+            path_destroy(res_path);
+            return VG_FALSE;
+         }
+         if (amount > 0.5)
+            segment = esegment;
+         else
+            segment = ssegment;
+      } else if (is_segment_arc(esegment)) {
+         path_destroy(res_path);
+         return VG_FALSE;
+      }
+      else if (ssegment != esegment) {
+         path_destroy(res_path);
+         return VG_FALSE;
+      }
+      else
+         segment = ssegment;
+
+      linearly_interpolate(results, start_coords, end_coords,
+                           amount, snum_coords);
+      vg_float_to_datatype(dst->datatype, common_data, results, snum_coords);
+      path_append_data(res_path, 1, &segment, common_data);
+   }
+
+   path_append_path(dst, res_path);
+   path_destroy(res_path);
+
+   dst->dirty = VG_TRUE;
+   dst->dirty_stroke = VG_TRUE;
+
+   return VG_TRUE;
+}
+
+void path_clear(struct path *p, VGbitfield capabilities)
+{
+   path_set_capabilities(p, capabilities);
+   array_destroy(p->segments);
+   array_destroy(p->control_points);
+   p->segments = array_create(size_for_datatype(VG_PATH_DATATYPE_S_8));
+   p->control_points = array_create(size_for_datatype(p->datatype));
+   p->num_segments = 0;
+   p->dirty = VG_TRUE;
+   p->dirty_stroke = VG_TRUE;
+}
+
+struct path * path_create_stroke(struct path *p,
+                                 struct matrix *matrix)
+{
+   VGint i;
+   VGfloat sx, sy, px, py, ox, oy;
+   VGfloat x0, y0, x1, y1, x2, y2, x3, y3;
+   VGfloat data[8];
+   void *coords = (VGfloat *)p->control_points->data;
+   int dashed = (p->base.ctx->state.vg.stroke.dash_pattern_num ? 1 : 0);
+   struct dash_stroker stroker;
+   struct vg_state *vg_state = &p->base.ctx->state.vg;
+
+   if (p->stroked.path)
+   {
+      /* ### compare the dash patterns to see if we can cache them.
+       *     for now we simply always bail out if the path is dashed.
+       */
+      if (memcmp( &p->stroked.matrix,
+                  matrix,
+                  sizeof *matrix ) == 0 &&
+          !dashed && !p->dirty_stroke &&
+          floatsEqual(p->stroked.stroke_width, vg_state->stroke.line_width.f) &&
+          floatsEqual(p->stroked.miter_limit, vg_state->stroke.miter_limit.f) &&
+          p->stroked.cap_style == vg_state->stroke.cap_style &&
+          p->stroked.join_style == vg_state->stroke.join_style)
+      {
+         return p->stroked.path;
+      }
+      else {
+         path_destroy( p->stroked.path );
+         p->stroked.path = NULL;
+      }
+   }
+
+
+   sx = sy = px = py = ox = oy = 0.f;
+
+   if (dashed)
+      dash_stroker_init((struct stroker *)&stroker, vg_state);
+   else
+      stroker_init((struct stroker *)&stroker, vg_state);
+
+   stroker_begin((struct stroker *)&stroker);
+
+   for (i = 0; i < p->num_segments; ++i) {
+      VGubyte segment = ((VGubyte*)(p->segments->data))[i];
+      VGint command = SEGMENT_COMMAND(segment);
+      VGboolean relative = SEGMENT_ABS_REL(segment);
+
+      switch(command) {
+      case VG_CLOSE_PATH: {
+            VGfloat x0 = sx;
+            VGfloat y0 = sy;
+            matrix_map_point(matrix, x0, y0, &x0, &y0);
+            stroker_line_to((struct stroker *)&stroker, x0, y0);
+      }
+         break;
+      case VG_MOVE_TO:
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         sx = x0;
+         sy = y0;
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_move_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_LINE_TO:
+         data_at(&coords, p, 0, 2, data);
+         x0 = data[0];
+         y0 = data[1];
+         map_if_relative(ox, oy, relative, &x0, &y0);
+         ox = x0;
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_line_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_HLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = data[0];
+         y0 = oy;
+         map_if_relative(ox, oy, relative, &x0, 0);
+         ox = x0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_line_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_VLINE_TO:
+         data_at(&coords, p, 0, 1, data);
+         x0 = ox;
+         y0 = data[0];
+         map_if_relative(ox, oy, relative, 0, &y0);
+         oy = y0;
+         px = x0;
+         py = y0;
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         stroker_line_to((struct stroker *)&stroker, x0, y0);
+         break;
+      case VG_CUBIC_TO: {
+         data_at(&coords, p, 0, 6, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x2 = data[2];
+         y2 = data[3];
+         x3 = data[4];
+         y3 = data[5];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_QUAD_TO: {
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = data[0];
+         y1 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_SQUAD_TO: {
+         data_at(&coords, p, 0, 2, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x3 = data[0];
+         y3 = data[1];
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         px = x1;
+         py = y1;
+         { /* form a cubic out of it */
+            x2 = (x3 + 2*x1) / 3.f;
+            y2 = (y3 + 2*y1) / 3.f;
+            x1 = (x0 + 2*x1) / 3.f;
+            y1 = (y0 + 2*y1) / 3.f;
+         }
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_SCUBIC_TO: {
+         data_at(&coords, p, 0, 4, data);
+         x0 = ox;
+         y0 = oy;
+         x1 = 2*ox-px;
+         y1 = 2*oy-py;
+         x2 = data[0];
+         y2 = data[1];
+         x3 = data[2];
+         y3 = data[3];
+         map_if_relative(ox, oy, relative, &x2, &y2);
+         map_if_relative(ox, oy, relative, &x3, &y3);
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy) &&
+             floatsEqual(x1, x2) && floatsEqual(y1, y2) &&
+             floatsEqual(x2, x3) && floatsEqual(y2, y3)) {
+            /*ignore the empty segment */
+            continue;
+         } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x3;
+            oy = y3;
+            matrix_map_point(matrix, x3, y3, &x3, &y3);
+            stroker_line_to((struct stroker *)&stroker, x3, y3);
+            continue;
+         }
+         ox = x3;
+         oy = y3;
+         px = x2;
+         py = y2;
+         assert(matrix_is_affine(matrix));
+         matrix_map_point(matrix, x0, y0, &x0, &y0);
+         matrix_map_point(matrix, x1, y1, &x1, &y1);
+         matrix_map_point(matrix, x2, y2, &x2, &y2);
+         matrix_map_point(matrix, x3, y3, &x3, &y3);
+         stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         VGfloat rh, rv, rot;
+         struct arc arc;
+
+         data_at(&coords, p, 0, 5, data);
+         x0  = ox;
+         y0  = oy;
+         rh  = data[0];
+         rv  = data[1];
+         rot = data[2];
+         x1  = data[3];
+         y1  = data[4];
+         map_if_relative(ox, oy, relative, &x1, &y1);
+         if (floatsEqual(x1, ox) && floatsEqual(y1, oy)) {
+            /* if dup vertex, emit a line */
+            ox = x1;
+            oy = y1;
+            matrix_map_point(matrix, x1, y1, &x1, &y1);
+            stroker_line_to((struct stroker *)&stroker, x1, y1);
+            continue;
+         }
+         arc_init(&arc, command, x0, y0, x1, y1,
+                  rh, rv, rot);
+         arc_stroke_cb(&arc, (struct stroker *)&stroker,
+                       matrix);
+         ox = x1;
+         oy = y1;
+         px = x1;
+         py = y1;
+      }
+         break;
+      default:
+         abort();
+         assert(!"Unknown segment!");
+      }
+   }
+
+   stroker_end((struct stroker *)&stroker);
+
+   if (dashed)
+      dash_stroker_cleanup((struct dash_stroker *)&stroker);
+   else
+      stroker_cleanup((struct stroker *)&stroker);
+
+   p->stroked.path = stroker.base.path;
+   p->stroked.matrix = *matrix;
+   p->dirty_stroke = VG_FALSE;
+   p->stroked.stroke_width = vg_state->stroke.line_width.f;
+   p->stroked.miter_limit = vg_state->stroke.miter_limit.f;
+   p->stroked.cap_style = vg_state->stroke.cap_style;
+   p->stroked.join_style = vg_state->stroke.join_style;
+
+   return stroker.base.path;
+}
+
+void path_render(struct path *p, VGbitfield paintModes)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix;
+
+   vg_validate_state(ctx);
+
+   shader_set_drawing_image(ctx->shader, VG_FALSE);
+   shader_set_image(ctx->shader, 0);
+#if 0
+   fprintf(stderr, "Matrix(11=%f 12=%f 13=%f 21=%f 22=%f 23=%f 31=%f 32=%f 33=%f)\n",
+           mat->m[0], mat->m[1], mat->m[2],
+           mat->m[3], mat->m[4], mat->m[5],
+           mat->m[6], mat->m[7], mat->m[8]);
+#endif
+   if (paintModes & VG_FILL_PATH) {
+      /* First the fill */
+      shader_set_paint(ctx->shader, ctx->state.vg.fill_paint);
+      shader_bind(ctx->shader);
+      path_fill(p, mat);
+   }
+
+   if (paintModes & VG_STROKE_PATH){
+      /* 8.7.5: "line width less than or equal to 0 prevents stroking from
+       *  taking place."*/
+      if (ctx->state.vg.stroke.line_width.f <= 0)
+         return;
+      shader_set_paint(ctx->shader, ctx->state.vg.stroke_paint);
+      shader_bind(ctx->shader);
+      path_stroke(p);
+   }
+}
+
+void path_fill(struct path *p, struct matrix *mat)
+{
+   struct vg_context *ctx = vg_current_context();
+   {
+      struct polygon_array *polygon_array = path_get_fill_polygons(p, mat);
+      struct array *polys = polygon_array->array;
+
+      if (!polygon_array || !polys || !polys->num_elements) {
+         return;
+      }
+      polygon_array_fill(polygon_array, ctx);
+   }
+}
+
+void path_stroke(struct path *p)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix;
+   VGFillRule old_fill = ctx->state.vg.fill_rule;
+   struct matrix identity;
+   struct path *stroke;
+
+   matrix_load_identity(&identity);
+   stroke = path_create_stroke(p, &identity);
+   if (stroke && !path_is_empty(stroke)) {
+      ctx->state.vg.fill_rule = VG_NON_ZERO;
+
+      path_fill(stroke, mat);
+
+      ctx->state.vg.fill_rule = old_fill;
+   }
+}
+
+void path_move_to(struct path *p, float x, float y)
+{
+   VGubyte segment = VG_MOVE_TO_ABS;
+   VGubyte common_data[sizeof(VGfloat) * 2];
+   VGfloat data[2] = {x, y};
+
+   vg_float_to_datatype(p->datatype, common_data, data, 2);
+   path_append_data(p, 1, &segment, common_data);
+}
+
+void path_line_to(struct path *p, float x, float y)
+{
+   VGubyte segment = VG_LINE_TO_ABS;
+   VGubyte common_data[sizeof(VGfloat) * 2];
+   VGfloat data[2] = {x, y};
+
+   vg_float_to_datatype(p->datatype, common_data, data, 2);
+
+   path_append_data(p, 1, &segment, common_data);
+}
+
+void path_cubic_to(struct path *p, float px1, float py1,
+                   float px2, float py2,
+                   float x, float y)
+{
+   VGubyte segment = VG_CUBIC_TO_ABS;
+   VGubyte common_data[sizeof(VGfloat) * 6];
+   VGfloat data[6];
+
+   data[0] = px1; data[1] = py1;
+   data[2] = px2; data[3] = py2;
+   data[4] = x;   data[5] = y;
+
+   vg_float_to_datatype(p->datatype, common_data, data, 6);
+
+   path_append_data(p, 1, &segment, common_data);
+}
+
+static INLINE void line_bounds(VGfloat *line /*x1,y1,x2,y2*/,
+                               VGfloat *bounds)
+{
+   bounds[0] = MIN2(line[0], line[2]);
+   bounds[1] = MIN2(line[1], line[3]);
+   bounds[2] = MAX2(line[0], line[2]) - bounds[0];
+   bounds[3] = MAX2(line[1], line[3]) - bounds[1];
+}
+
+static INLINE void unite_bounds(VGfloat *bounds,
+                                VGfloat *el)
+{
+   VGfloat cx1, cy1, cx2, cy2;
+   VGfloat nx1, ny1, nx2, ny2;
+
+   cx1 = bounds[0];
+   cy1 = bounds[1];
+   cx2 = bounds[0] + bounds[2];
+   cy2 = bounds[1] + bounds[3];
+
+   nx1 = el[0];
+   ny1 = el[1];
+   nx2 = el[0] + el[2];
+   ny2 = el[1] + el[3];
+
+   bounds[0] = MIN2(cx1, nx1);
+   bounds[1] = MIN2(cy1, ny1);
+   bounds[2] = MAX2(cx2, nx2) - bounds[0];
+   bounds[3] = MAX2(cy2, ny2) - bounds[1];
+}
+
+static INLINE void set_bounds(VGfloat *bounds,
+                              VGfloat *element_bounds,
+                              VGboolean *initialized)
+{
+   if (!(*initialized)) {
+      memcpy(bounds, element_bounds, 4 * sizeof(VGfloat));
+      *initialized = VG_TRUE;
+   } else
+      unite_bounds(bounds, element_bounds);
+}
+
+void path_bounding_rect(struct path *p, float *x, float *y,
+                        float *w, float *h)
+{
+   VGint i;
+   VGfloat coords[8];
+   struct path_iter_data iter;
+   VGint num_coords;
+   VGfloat bounds[4];
+   VGfloat element_bounds[4];
+   VGfloat ox, oy;
+   VGboolean bounds_inited = VG_FALSE;
+
+   memset(&iter, 0, sizeof(struct path_iter_data));
+   memset(&bounds, 0, sizeof(bounds));
+
+   if (!p->num_segments) {
+      bounds[2] = -1;
+      bounds[3] = -1;
+   }
+
+
+   iter.path = p;
+   iter.coords = p->control_points->data;
+
+   for (i = 0; i < p->num_segments; ++i) {
+      VGubyte segment;
+      iter.segment = ((VGubyte*)(p->segments->data))[i];
+
+      ox = iter.ox;
+      oy = iter.oy;
+
+      segment = normalize_coords(&iter, &num_coords, coords);
+
+      switch(segment) {
+      case VG_CLOSE_PATH:
+      case VG_MOVE_TO_ABS:
+         break;
+      case VG_LINE_TO_ABS: {
+         VGfloat line[4] = {ox, oy, coords[0], coords[1]};
+         line_bounds(line, element_bounds);
+         set_bounds(bounds, element_bounds, &bounds_inited);
+      }
+         break;
+      case VG_CUBIC_TO_ABS: {
+         struct bezier bezier;
+         bezier_init(&bezier, ox, oy,
+                     coords[0], coords[1],
+                     coords[2], coords[3],
+                     coords[4], coords[5]);
+         bezier_exact_bounds(&bezier, element_bounds);
+         set_bounds(bounds, element_bounds, &bounds_inited);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         struct arc arc;
+         struct matrix identity;
+         struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                         1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+         matrix_load_identity(&identity);
+         arc_init(&arc, segment,
+                  ox, oy, coords[3], coords[4],
+                  coords[0], coords[1], coords[2]);
+
+         arc_to_path(&arc, path, &identity);
+
+         path_bounding_rect(path, element_bounds + 0, element_bounds + 1,
+                            element_bounds + 2, element_bounds + 3);
+         set_bounds(bounds, element_bounds, &bounds_inited);
+      }
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   *x = bounds[0];
+   *y = bounds[1];
+   *w = bounds[2];
+   *h = bounds[3];
+}
+
+float path_length(struct path *p, int start_segment, int num_segments)
+{
+   VGint i;
+   VGfloat coords[8];
+   struct path_iter_data iter;
+   VGint num_coords;
+   VGfloat length = 0;
+   VGfloat ox, oy;
+   VGboolean in_range = VG_FALSE;
+
+   memset(&iter, 0, sizeof(struct path_iter_data));
+
+   iter.path = p;
+   iter.coords = p->control_points->data;
+
+   for (i = 0; i < (start_segment + num_segments); ++i) {
+      VGubyte segment;
+
+      iter.segment = ((VGubyte*)(p->segments->data))[i];
+
+      ox = iter.ox;
+      oy = iter.oy;
+
+      segment = normalize_coords(&iter, &num_coords, coords);
+
+      in_range = (i >= start_segment) && i <= (start_segment + num_segments);
+      if (!in_range)
+         continue;
+
+      switch(segment) {
+      case VG_MOVE_TO_ABS:
+         break;
+      case VG_CLOSE_PATH: {
+         VGfloat line[4] = {ox, oy, iter.sx, iter.sy};
+         length += line_lengthv(line);
+      }
+         break;
+      case VG_LINE_TO_ABS: {
+         VGfloat line[4] = {ox, oy, coords[0], coords[1]};
+         length += line_lengthv(line);
+      }
+         break;
+      case VG_CUBIC_TO_ABS: {
+         struct bezier bezier;
+         bezier_init(&bezier, ox, oy,
+                     coords[0], coords[1],
+                     coords[2], coords[3],
+                     coords[4], coords[5]);
+         length += bezier_length(&bezier, BEZIER_DEFAULT_ERROR);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         struct arc arc;
+         struct matrix identity;
+         struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                         1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+         matrix_load_identity(&identity);
+         arc_init(&arc, segment,
+                  ox, oy, coords[3], coords[4],
+                  coords[0], coords[1], coords[2]);
+
+         arc_to_path(&arc, path, &identity);
+
+         length += path_length(path, 0, path_num_segments(path));
+      }
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   return length;
+}
+
+static INLINE VGboolean point_on_current_segment(VGfloat distance,
+                                                 VGfloat length,
+                                                 VGfloat segment_length)
+{
+   return
+      (((floatIsZero(distance) || distance < 0) && floatIsZero(length)) ||
+       ((distance > length || floatsEqual(distance, length)) &&
+        (floatsEqual(distance, length + segment_length) ||
+         distance < (length + segment_length))));
+}
+
+static VGboolean path_point_segment(struct path_iter_data iter,
+                                    struct path_iter_data prev_iter,
+                                    VGfloat coords[8],
+                                    VGfloat distance,
+                                    VGfloat length, VGfloat *current_length,
+                                    VGfloat *point, VGfloat *normal)
+{
+   switch (iter.segment) {
+   case VG_MOVE_TO_ABS:
+      break;
+   case VG_CLOSE_PATH: {
+      VGfloat line[4] = {prev_iter.ox, prev_iter.oy, iter.sx, iter.sy};
+      VGboolean on_current_segment = VG_FALSE;
+      *current_length = line_lengthv(line);
+      on_current_segment = point_on_current_segment(distance,
+                                                    length,
+                                                    *current_length);
+      if (on_current_segment) {
+         VGfloat at = (distance - length) / line_lengthv(line);
+         line_normal_vector(line, normal);
+         line_point_at(line, at, point);
+         return VG_TRUE;
+      }
+   }
+      break;
+   case VG_LINE_TO_ABS: {
+      VGfloat line[4] = {prev_iter.ox, prev_iter.oy, coords[0], coords[1]};
+      VGboolean on_current_segment = VG_FALSE;
+      *current_length = line_lengthv(line);
+      on_current_segment = point_on_current_segment(distance,
+                                                    length,
+                                                    *current_length);
+      if (on_current_segment) {
+         VGfloat at = (distance - length) / line_lengthv(line);
+         line_normal_vector(line, normal);
+         line_point_at(line, at, point);
+         return VG_TRUE;
+      }
+   }
+      break;
+   case VG_CUBIC_TO_ABS: {
+      struct bezier bezier;
+      bezier_init(&bezier, prev_iter.ox, prev_iter.oy,
+                  coords[0], coords[1],
+                  coords[2], coords[3],
+                  coords[4], coords[5]);
+      *current_length = bezier_length(&bezier, BEZIER_DEFAULT_ERROR);
+      if (point_on_current_segment(distance, length, *current_length)) {
+         bezier_point_at_length(&bezier, distance - length,
+                                point, normal);
+         return VG_TRUE;
+      }
+   }
+      break;
+   case VG_SCCWARC_TO:
+   case VG_SCWARC_TO:
+   case VG_LCCWARC_TO:
+   case VG_LCWARC_TO: {
+      struct arc arc;
+      struct matrix identity;
+      struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                      1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+      matrix_load_identity(&identity);
+      arc_init(&arc, iter.segment,
+               prev_iter.ox, prev_iter.oy, coords[3], coords[4],
+               coords[0], coords[1], coords[2]);
+
+      arc_to_path(&arc, path, &identity);
+
+      *current_length = path_length(path, 0, path_num_segments(path));
+      if (point_on_current_segment(distance, length, *current_length)) {
+         path_point(path, 0, path_num_segments(path),
+                    distance - length, point, normal);
+         return VG_TRUE;
+      }
+   }
+      break;
+   default:
+      assert(0);
+   }
+   return VG_FALSE;
+}
+
+void path_point(struct path *p, VGint start_segment, VGint num_segments,
+                VGfloat distance, VGfloat *point, VGfloat *normal)
+{
+   VGint i;
+   VGfloat coords[8];
+   struct path_iter_data iter, prev_iter;
+   VGint num_coords;
+   VGfloat length = 0;
+   VGfloat current_length = 0;
+
+   memset(&iter, 0, sizeof(struct path_iter_data));
+   memset(&prev_iter, 0, sizeof(struct path_iter_data));
+
+   point[0] = 0;
+   point[1] = 0;
+
+   normal[0] = 0;
+   normal[1] = -1;
+
+   iter.path = p;
+   iter.coords = p->control_points->data;
+   if (distance < 0)
+      distance = 0;
+
+   for (i = 0; i < (start_segment + num_segments); ++i) {
+      VGboolean outside_range = (i < start_segment ||
+                                 i >= (start_segment + num_segments));
+
+      prev_iter = iter;
+
+      iter.segment = ((VGubyte*)(p->segments->data))[i];
+      iter.segment = normalize_coords(&iter, &num_coords, coords);
+
+      if (outside_range)
+         continue;
+
+      if (path_point_segment(iter, prev_iter, coords,
+                             distance, length, &current_length,
+                             point, normal))
+         return;
+
+      length += current_length;
+   }
+
+   /*
+    *OpenVG 1.0 - 8.6.11 vgPointAlongPath
+    *
+    * If distance is greater than or equal to the path length
+    *(i.e., the value returned by vgPathLength when called with the same
+    *startSegment and numSegments parameters), the visual ending point of
+    *the path is used.
+    */
+   {
+      switch (iter.segment) {
+      case VG_MOVE_TO_ABS:
+         break;
+      case VG_CLOSE_PATH: {
+         VGfloat line[4] = {prev_iter.ox, prev_iter.oy, iter.sx, iter.sy};
+         line_normal_vector(line, normal);
+         line_point_at(line, 1.f, point);
+      }
+         break;
+      case VG_LINE_TO_ABS: {
+         VGfloat line[4] = {prev_iter.ox, prev_iter.oy, coords[0], coords[1]};
+         line_normal_vector(line, normal);
+         line_point_at(line, 1.f, point);
+      }
+         break;
+      case VG_CUBIC_TO_ABS: {
+         struct bezier bezier;
+         bezier_init(&bezier, prev_iter.ox, prev_iter.oy,
+                     coords[0], coords[1],
+                     coords[2], coords[3],
+                     coords[4], coords[5]);
+         bezier_point_at_t(&bezier, 1.f, point, normal);
+      }
+         break;
+      case VG_SCCWARC_TO:
+      case VG_SCWARC_TO:
+      case VG_LCCWARC_TO:
+      case VG_LCWARC_TO: {
+         struct arc arc;
+         struct matrix identity;
+         struct path *path = path_create(VG_PATH_DATATYPE_F,
+                                         1, 0, 0, 0, VG_PATH_CAPABILITY_ALL);
+
+         matrix_load_identity(&identity);
+         arc_init(&arc, iter.segment,
+                  prev_iter.ox, prev_iter.oy, coords[3], coords[4],
+                  coords[0], coords[1], coords[2]);
+
+         arc_to_path(&arc, path, &identity);
+
+         path_point(path, 0, path_num_segments(path),
+                    /* to make sure we're bigger than len * 2 it */
+                    2 * path_length(path, 0, path_num_segments(path)),
+                    point, normal);
+      }
+         break;
+      default:
+         assert(0);
+      }
+   }
+}
+
+VGboolean path_is_empty(struct path *p)
+{
+   return p->segments->num_elements == 0;
+}
diff --git a/src/gallium/state_trackers/vega/path.h b/src/gallium/state_trackers/vega/path.h
new file mode 100644
index 0000000000..e34538b736
--- /dev/null
+++ b/src/gallium/state_trackers/vega/path.h
@@ -0,0 +1,126 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef _PATH_H
+#define _PATH_H
+
+#include "VG/openvg.h"
+
+struct path;
+struct polygon;
+struct matrix;
+
+enum fill_rule {
+   ODD_EVEN_FILL,
+   WINDING_FILL
+};
+
+
+struct path_for_each_data {
+   VGubyte segment;
+   /* all coords are absolute, even if segment is relative */
+   const VGfloat *coords;
+   VGfloat sx, sy, ox, oy, px, py;
+   void *user_data;
+};
+
+typedef VGboolean (*path_for_each_cb)(struct path *p,
+                                      struct path_for_each_data *data);
+
+
+struct path *path_create(VGPathDatatype dt, VGfloat scale, VGfloat bias,
+                         VGint segmentCapacityHint,
+                         VGint coordCapacityHint,
+                         VGbitfield capabilities);
+void path_destroy(struct path *p);
+
+VGbitfield path_capabilities(struct path *p);
+void path_set_capabilities(struct path *p, VGbitfield bf);
+
+void path_append_data(struct path *p,
+                      VGint numSegments,
+                      const VGubyte * pathSegments,
+                      const void * pathData);
+
+void path_append_path(struct path *dst,
+                      struct path *src);
+
+VGint path_num_segments(struct path *p);
+
+void path_bounding_rect(struct path *p, float *x, float *y,
+                        float *w, float *h);
+float path_length(struct path *p, int start_segment, int num_segments);
+
+void path_set_fill_rule(enum fill_rule fill);
+enum fill_rule path_fill_rule(enum fill_rule fill);
+
+VGboolean path_is_empty(struct path *p);
+
+VGbyte path_datatype_size(struct path *p);
+
+VGPathDatatype path_datatype(struct path *p);
+VGfloat path_scale(struct path *p);
+VGfloat path_bias(struct path *p);
+VGint path_num_coords(struct path *p);
+
+void path_modify_coords(struct path *p,
+                        VGint startIndex,
+                        VGint numSegments,
+                        const void * pathData);
+
+struct path *path_create_stroke(struct path *p,
+                                struct matrix *m);
+
+void path_for_each_segment(struct path *path,
+                           path_for_each_cb cb,
+                           void *user_data);
+
+void path_transform(struct path *dst, struct path *src);
+VGboolean path_interpolate(struct path *dst,
+                           struct path *start, struct path *end,
+                           VGfloat amount);
+
+void path_clear(struct path *p, VGbitfield capabilities);
+void path_render(struct path *p, VGbitfield paintModes);
+void path_fill(struct path *p, struct matrix *mat);
+void path_stroke(struct path *p);
+
+void path_move_to(struct path *p, float x, float y);
+void path_line_to(struct path *p, float x, float y);
+void path_cubic_to(struct path *p, float px1, float py1,
+                   float px2, float py2,
+                   float x, float y);
+
+void path_point(struct path *p, VGint startSegment, VGint numSegments,
+                VGfloat distance, VGfloat *point, VGfloat *normal);
+
+
+
+void vg_float_to_datatype(VGPathDatatype datatype,
+                          VGubyte *common_data,
+                          const VGfloat *data,
+                          VGint num_coords);
+#endif
diff --git a/src/gallium/state_trackers/vega/path_utils.h b/src/gallium/state_trackers/vega/path_utils.h
new file mode 100644
index 0000000000..c2b3221dc5
--- /dev/null
+++ b/src/gallium/state_trackers/vega/path_utils.h
@@ -0,0 +1,109 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef PATH_UTILS_H
+#define PATH_UTILS_H
+
+#include "VG/openvg.h"
+
+#define SEGMENT_COMMAND(command) /* Extract segment type */     \
+    ((command) & 0x1e)
+#define SEGMENT_ABS_REL(command) /* Extract absolute/relative bit */ \
+    ((command) & 0x1)
+
+static INLINE VGint size_for_datatype(VGPathDatatype datatype)
+{
+   switch(datatype) {
+   case VG_PATH_DATATYPE_S_8:
+      return 1;
+   case VG_PATH_DATATYPE_S_16:
+      return 2;
+   case VG_PATH_DATATYPE_S_32:
+      return 4;
+   case VG_PATH_DATATYPE_F:
+      return 4;
+   default:
+      assert(!"unknown datatype");
+   }
+   return 0;
+}
+
+static INLINE VGint num_elements_for_segments(const VGubyte *segments,
+                                              VGint num_segments)
+{
+   VGint i;
+   VGint count = 0;
+
+   for (i = 0; i < num_segments; ++i) {
+      VGubyte segment = segments[i];
+      VGint command = SEGMENT_COMMAND(segment);
+      switch(command) {
+      case VG_CLOSE_PATH:
+         break;
+      case VG_MOVE_TO:
+         count += 2;
+         break;
+      case VG_LINE_TO:
+         count += 2;
+         break;
+      case VG_HLINE_TO:
+         count += 1;
+         break;
+      case VG_VLINE_TO:
+         count += 1;
+         break;
+      case VG_QUAD_TO:
+         count += 4;
+         break;
+      case VG_CUBIC_TO:
+         count += 6;
+         break;
+      case VG_SQUAD_TO:
+         count += 2;
+         break;
+      case VG_SCUBIC_TO:
+         count += 4;
+         break;
+      case VG_SCCWARC_TO:
+         count += 5;
+         break;
+      case VG_SCWARC_TO:
+         count += 5;
+         break;
+      case VG_LCCWARC_TO:
+         count += 5;
+         break;
+      case VG_LCWARC_TO:
+         count += 5;
+         break;
+      default:
+         assert(!"Unknown segment!");
+      }
+   }
+   return count;
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/polygon.c b/src/gallium/state_trackers/vega/polygon.c
new file mode 100644
index 0000000000..b6d282d803
--- /dev/null
+++ b/src/gallium/state_trackers/vega/polygon.c
@@ -0,0 +1,550 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "polygon.h"
+
+#include "matrix.h" /*for floatsEqual*/
+#include "vg_context.h"
+#include "vg_state.h"
+#include "paint.h"
+#include "renderer.h"
+#include "util_array.h"
+#include "VG/openvg.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+
+#include "util/u_draw_quad.h"
+#include "util/u_math.h"
+
+#include <string.h>
+#include <stdlib.h>
+
+#define DEBUG_POLYGON 0
+
+#define COMPONENTS 2
+
+struct polygon
+{
+   VGfloat *data;
+   VGint    size;
+
+   VGint    num_verts;
+
+   VGboolean dirty;
+   struct pipe_buffer *vbuf;
+   struct pipe_screen *screen;
+};
+
+static float *ptr_to_vertex(float *data, int idx)
+{
+   return data + (idx * COMPONENTS);
+}
+
+#if 0
+static void polygon_print(struct polygon *poly)
+{
+   int i;
+   float *vert;
+   debug_printf("Polygon %p, size = %d\n", poly, poly->num_verts);
+   for (i = 0; i < poly->num_verts; ++i) {
+      vert = ptr_to_vertex(poly->data, i);
+      debug_printf("%f, %f,  ", vert[0], vert[1]);
+   }
+   debug_printf("\nend\n");
+}
+#endif
+
+
+struct polygon * polygon_create(int size)
+{
+   struct polygon *poly = (struct polygon*)malloc(sizeof(struct polygon));
+
+   poly->data = malloc(sizeof(float) * COMPONENTS * size);
+   poly->size = size;
+   poly->num_verts = 0;
+   poly->dirty = VG_TRUE;
+   poly->vbuf = NULL;
+
+   return poly;
+}
+
+struct polygon * polygon_create_from_data(float *data, int size)
+{
+   struct polygon *poly = polygon_create(size);
+
+   memcpy(poly->data, data, sizeof(float) * COMPONENTS * size);
+   poly->num_verts = size;
+   poly->dirty = VG_TRUE;
+   poly->vbuf = NULL;
+
+   return poly;
+}
+
+void polygon_destroy(struct polygon *poly)
+{
+   if (poly->vbuf)
+      pipe_buffer_reference(&poly->vbuf, NULL);
+
+   free(poly->data);
+   free(poly);
+}
+
+void polygon_resize(struct polygon *poly, int new_size)
+{
+   float *data = (float*)malloc(sizeof(float) * COMPONENTS * new_size);
+   int size = MIN2(sizeof(float) * COMPONENTS * new_size,
+                   sizeof(float) * COMPONENTS * poly->size);
+   memcpy(data, poly->data, size);
+   free(poly->data);
+   poly->data = data;
+   poly->size = new_size;
+   poly->dirty = VG_TRUE;
+}
+
+int polygon_size(struct polygon *poly)
+{
+   return poly->size;
+}
+
+int polygon_vertex_count(struct polygon *poly)
+{
+   return poly->num_verts;
+}
+
+float * polygon_data(struct polygon *poly)
+{
+   return poly->data;
+}
+
+void polygon_vertex_append(struct polygon *p,
+                           float x, float y)
+{
+   float *vert;
+#if DEBUG_POLYGON
+   debug_printf("Append vertex [%f, %f]\n", x, y);
+#endif
+   if (p->num_verts >= p->size) {
+      polygon_resize(p, p->size * 2);
+   }
+
+   vert = ptr_to_vertex(p->data, p->num_verts);
+   vert[0] = x;
+   vert[1] = y;
+   ++p->num_verts;
+   p->dirty = VG_TRUE;
+}
+
+void polygon_set_vertex(struct polygon *p, int idx,
+                        float x, float y)
+{
+   float *vert;
+   if (idx >= p->num_verts) {
+      /*fixme: error reporting*/
+      abort();
+      return;
+   }
+
+   vert = ptr_to_vertex(p->data, idx);
+   vert[0] = x;
+   vert[1] = y;
+   p->dirty = VG_TRUE;
+}
+
+void polygon_vertex(struct polygon *p, int idx,
+                    float *vertex)
+{
+   float *vert;
+   if (idx >= p->num_verts) {
+      /*fixme: error reporting*/
+      abort();
+      return;
+   }
+
+   vert = ptr_to_vertex(p->data, idx);
+   vertex[0] = vert[0];
+   vertex[1] = vert[1];
+}
+
+void polygon_bounding_rect(struct polygon *p,
+                           float *rect)
+{
+   int i;
+   float minx, miny, maxx, maxy;
+   float *vert = ptr_to_vertex(p->data, 0);
+   minx = vert[0];
+   maxx = vert[0];
+   miny = vert[1];
+   maxy = vert[1];
+
+   for (i = 1; i < p->num_verts; ++i) {
+      vert = ptr_to_vertex(p->data, i);
+      minx = MIN2(vert[0], minx);
+      miny = MIN2(vert[1], miny);
+
+      maxx = MAX2(vert[0], maxx);
+      maxy = MAX2(vert[1], maxy);
+   }
+
+   rect[0] = minx;
+   rect[1] = miny;
+   rect[2] = maxx - minx;
+   rect[3] = maxy - miny;
+}
+
+int polygon_contains_point(struct polygon *p,
+                           float x, float y)
+{
+   return 0;
+}
+
+void polygon_append_polygon(struct polygon *dst,
+                            struct polygon *src)
+{
+   if (dst->num_verts + src->num_verts >= dst->size) {
+      polygon_resize(dst, dst->num_verts + src->num_verts * 1.5);
+   }
+   memcpy(ptr_to_vertex(dst->data, dst->num_verts),
+          src->data, src->num_verts * COMPONENTS * sizeof(VGfloat));
+   dst->num_verts += src->num_verts;
+}
+
+VGboolean polygon_is_closed(struct polygon *p)
+{
+   VGfloat start[2], end[2];
+
+   polygon_vertex(p, 0, start);
+   polygon_vertex(p, p->num_verts - 1, end);
+
+   return floatsEqual(start[0], end[0]) && floatsEqual(start[1], end[1]);
+}
+
+static void set_blend_for_fill(struct pipe_blend_state *blend)
+{
+   memset(blend, 0, sizeof(struct pipe_blend_state));
+   blend->colormask = 0; /*disable colorwrites*/
+
+   blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   blend->rgb_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+   blend->alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+}
+
+static void draw_polygon(struct vg_context *ctx,
+                         struct polygon *poly)
+{
+   int vert_size;
+   struct pipe_context *pipe;
+   struct pipe_vertex_buffer vbuffer;
+   struct pipe_vertex_element velement;
+
+   vert_size = poly->num_verts * COMPONENTS * sizeof(float);
+
+   /*polygon_print(poly);*/
+
+   pipe = ctx->pipe;
+
+   if (poly->vbuf == NULL || poly->dirty) {
+      if (poly->vbuf) {
+         pipe_buffer_reference(&poly->vbuf,
+                               NULL);
+      }
+      poly->screen = pipe->screen;
+      poly->vbuf= pipe_user_buffer_create(poly->screen,
+                                          poly->data,
+                                          vert_size);
+      poly->dirty = VG_FALSE;
+   }
+
+
+   /* tell pipe about the vertex buffer */
+   memset(&vbuffer, 0, sizeof(vbuffer));
+   vbuffer.buffer = poly->vbuf;
+   vbuffer.stride = COMPONENTS * sizeof(float);  /* vertex size */
+   vbuffer.buffer_offset = 0;
+   vbuffer.max_index = poly->num_verts - 1;
+   pipe->set_vertex_buffers(pipe, 1, &vbuffer);
+
+   /* tell pipe about the vertex attributes */
+   velement.src_offset = 0;
+   velement.vertex_buffer_index = 0;
+   velement.src_format = PIPE_FORMAT_R32G32_FLOAT;
+   velement.nr_components = COMPONENTS;
+   pipe->set_vertex_elements(pipe, 1, &velement);
+
+   /* draw */
+   pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 
+                     0, poly->num_verts);
+}
+
+void polygon_fill(struct polygon *poly, struct vg_context *ctx)
+{
+   struct pipe_depth_stencil_alpha_state dsa;
+   struct pipe_blend_state blend;
+   VGfloat bounds[4];
+   VGfloat min_x, min_y, max_x, max_y;
+   assert(poly);
+   polygon_bounding_rect(poly, bounds);
+   min_x = bounds[0];
+   min_y = bounds[1];
+   max_x = bounds[0] + bounds[2];
+   max_y = bounds[1] + bounds[3];
+
+#if DEBUG_POLYGON
+   debug_printf("Poly bounds are [%f, %f], [%f, %f]\n",
+                min_x, min_y, max_x, max_y);
+#endif
+
+   set_blend_for_fill(&blend);
+
+   memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+
+   cso_save_blend(ctx->cso_context);
+   cso_save_depth_stencil_alpha(ctx->cso_context);
+
+   dsa.stencil[0].enabled = 1;
+   if (ctx->state.vg.fill_rule == VG_EVEN_ODD) {
+      dsa.stencil[0].writemask = 1;
+      dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INVERT;
+      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+      dsa.stencil[0].ref_value = 0;
+      dsa.stencil[0].valuemask = ~0;
+
+      cso_set_blend(ctx->cso_context, &blend);
+      cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+      draw_polygon(ctx, poly);
+   } else if (ctx->state.vg.fill_rule == VG_NON_ZERO) {
+      struct pipe_screen *screen = ctx->pipe->screen;
+
+      if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) {
+         /* front */
+         dsa.stencil[0].writemask = ~0;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         /* back */
+         dsa.stencil[1].enabled = 1;
+         dsa.stencil[1].writemask = ~0;
+         dsa.stencil[1].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         dsa.stencil[1].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[1].ref_value = 0;
+         dsa.stencil[1].valuemask = ~0;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         draw_polygon(ctx, poly);
+      } else {
+         struct pipe_rasterizer_state raster;
+
+         memcpy(&raster, &ctx->state.g3d.rasterizer, sizeof(struct pipe_rasterizer_state));
+
+         cso_save_rasterizer(ctx->cso_context);
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         draw_polygon(ctx, poly);
+
+         raster.cull_mode = raster.front_winding;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         draw_polygon(ctx, poly);
+
+         cso_restore_rasterizer(ctx->cso_context);
+      }
+   }
+
+   /* restore color writes */
+   cso_restore_blend(ctx->cso_context);
+   /* setup stencil ops */
+   dsa.stencil[0].func = PIPE_FUNC_NOTEQUAL;
+   dsa.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].ref_value = 0;
+   dsa.stencil[0].valuemask = dsa.stencil[0].writemask;
+   dsa.stencil[1].enabled = 0;
+   memcpy(&dsa.depth, &ctx->state.g3d.dsa.depth,
+          sizeof(struct pipe_depth_state));
+   cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+
+   /* render the quad to propagate the rendering from stencil */
+   renderer_draw_quad(ctx->renderer, min_x, min_y,
+                      max_x, max_y, 0.0f/*depth should be disabled*/);
+
+   cso_restore_depth_stencil_alpha(ctx->cso_context);
+}
+
+void polygon_array_fill(struct polygon_array *polyarray, struct vg_context *ctx)
+{
+   struct array *polys = polyarray->array;
+   struct pipe_depth_stencil_alpha_state dsa;
+   struct pipe_blend_state blend;
+   VGfloat min_x = polyarray->min_x;
+   VGfloat min_y = polyarray->min_y;
+   VGfloat max_x = polyarray->max_x;
+   VGfloat max_y = polyarray->max_y;
+   VGint i;
+
+
+#if DEBUG_POLYGON
+   debug_printf("%s: Poly bounds are [%f, %f], [%f, %f]\n",
+                __FUNCTION__,
+                min_x, min_y, max_x, max_y);
+#endif
+
+   set_blend_for_fill(&blend);
+
+   memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+
+   cso_save_blend(ctx->cso_context);
+   cso_save_depth_stencil_alpha(ctx->cso_context);
+
+   dsa.stencil[0].enabled = 1;
+   if (ctx->state.vg.fill_rule == VG_EVEN_ODD) {
+      dsa.stencil[0].writemask = 1;
+      dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INVERT;
+      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+      dsa.stencil[0].ref_value = 0;
+      dsa.stencil[0].valuemask = ~0;
+
+      cso_set_blend(ctx->cso_context, &blend);
+      cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+      for (i = 0; i < polys->num_elements; ++i) {
+         struct polygon *poly = (((struct polygon**)polys->data)[i]);
+         draw_polygon(ctx, poly);
+      }
+   } else if (ctx->state.vg.fill_rule == VG_NON_ZERO) {
+      struct pipe_screen *screen = ctx->pipe->screen;
+
+      if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) {
+         /* front */
+         dsa.stencil[0].writemask = ~0;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         /* back */
+         dsa.stencil[1].enabled = 1;
+         dsa.stencil[1].writemask = ~0;
+         dsa.stencil[1].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[1].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         dsa.stencil[1].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[1].ref_value = 0;
+         dsa.stencil[1].valuemask = ~0;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         for (i = 0; i < polys->num_elements; ++i) {
+            struct polygon *poly = (((struct polygon**)polys->data)[i]);
+            draw_polygon(ctx, poly);
+         }
+      } else {
+         struct pipe_rasterizer_state raster;
+
+         memcpy(&raster, &ctx->state.g3d.rasterizer, sizeof(struct pipe_rasterizer_state));
+
+         cso_save_rasterizer(ctx->cso_context);
+         dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+         dsa.stencil[0].ref_value = 0;
+         dsa.stencil[0].valuemask = ~0;
+
+         raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP;
+
+         cso_set_blend(ctx->cso_context, &blend);
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         for (i = 0; i < polys->num_elements; ++i) {
+            struct polygon *poly = (((struct polygon**)polys->data)[i]);
+            draw_polygon(ctx, poly);
+         }
+
+         raster.cull_mode = raster.front_winding;
+         dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP;
+         dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_DECR_WRAP;
+         cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+         cso_set_rasterizer(ctx->cso_context, &raster);
+         for (i = 0; i < polys->num_elements; ++i) {
+            struct polygon *poly = (((struct polygon**)polys->data)[i]);
+            draw_polygon(ctx, poly);
+         }
+
+         cso_restore_rasterizer(ctx->cso_context);
+      }
+   }
+
+   /* restore color writes */
+   cso_restore_blend(ctx->cso_context);
+   /* setup stencil ops */
+   dsa.stencil[0].func = PIPE_FUNC_NOTEQUAL;
+   dsa.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+   dsa.stencil[0].ref_value = 0;
+   dsa.stencil[0].valuemask = dsa.stencil[0].writemask;
+   dsa.stencil[1].enabled = 0;
+   memcpy(&dsa.depth, &ctx->state.g3d.dsa.depth,
+          sizeof(struct pipe_depth_state));
+   cso_set_depth_stencil_alpha(ctx->cso_context, &dsa);
+
+   /* render the quad to propagate the rendering from stencil */
+   renderer_draw_quad(ctx->renderer, min_x, min_y,
+                      max_x, max_y, 0.0f/*depth should be disabled*/);
+
+   cso_restore_depth_stencil_alpha(ctx->cso_context);
+}
diff --git a/src/gallium/state_trackers/vega/polygon.h b/src/gallium/state_trackers/vega/polygon.h
new file mode 100644
index 0000000000..22672b728e
--- /dev/null
+++ b/src/gallium/state_trackers/vega/polygon.h
@@ -0,0 +1,75 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef POLYGON_H
+#define POLYGON_H
+
+#include "VG/openvg.h"
+
+struct polygon;
+struct vg_context;
+struct vg_paint;
+struct array;
+
+struct polygon *polygon_create(int size);
+struct polygon *polygon_create_from_data(float *data, int size);
+void            polygon_destroy(struct polygon *poly);
+
+void            polygon_resize(struct polygon *poly, int new_size);
+int             polygon_size(struct polygon *poly);
+
+int             polygon_vertex_count(struct polygon *poly);
+float *         polygon_data(struct polygon *poly);
+
+void            polygon_vertex_append(struct polygon *p,
+                                      float x, float y);
+void            polygon_append_polygon(struct polygon *dst,
+                                       struct polygon *src);
+void            polygon_set_vertex(struct polygon *p, int idx,
+                                   float x, float y);
+void            polygon_vertex(struct polygon *p, int idx,
+                               float *vertex);
+
+void            polygon_bounding_rect(struct polygon *p,
+                                      float *rect);
+int             polygon_contains_point(struct polygon *p,
+                                       float x, float y);
+
+VGboolean       polygon_is_closed(struct polygon *p);
+
+void polygon_fill(struct polygon *p, struct vg_context *pipe);
+
+/* TODO: make a file/module around this struct
+ */
+struct polygon_array {
+   struct array *array;
+   VGfloat min_x, max_x;
+   VGfloat min_y, max_y;
+};
+
+void polygon_array_fill(struct polygon_array *polyarray, struct vg_context *ctx);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c
new file mode 100644
index 0000000000..f7c5f2f0cd
--- /dev/null
+++ b/src/gallium/state_trackers/vega/renderer.c
@@ -0,0 +1,592 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "renderer.h"
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_draw_quad.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_memory.h"
+
+#include "cso_cache/cso_context.h"
+
+struct renderer {
+   struct pipe_context *pipe;
+   struct vg_context *owner;
+
+   struct cso_context *cso;
+
+   void *fs;
+
+   VGfloat vertices[4][2][4];
+};
+
+static void setup_shaders(struct renderer *ctx)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   /* fragment shader */
+   ctx->fs = util_make_fragment_tex_shader(pipe);
+}
+
+static struct pipe_buffer *
+setup_vertex_data(struct renderer *ctx,
+                  float x0, float y0, float x1, float y1, float z)
+{
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = 0.0f; /*s*/
+   ctx->vertices[0][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = 1.0f; /*s*/
+   ctx->vertices[1][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = 1.0f;
+   ctx->vertices[2][1][1] = 1.0f;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = 0.0f;
+   ctx->vertices[3][1][1] = 1.0f;
+
+   return pipe_user_buffer_create( ctx->pipe->screen,
+                                   ctx->vertices,
+                                   sizeof(ctx->vertices) );
+}
+
+static struct pipe_buffer *
+setup_vertex_data_tex(struct renderer *ctx,
+                      float x0, float y0, float x1, float y1,
+                      float s0, float t0, float s1, float t1,
+                      float z)
+{
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = s0; /*s*/
+   ctx->vertices[0][1][1] = t0; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = s1; /*s*/
+   ctx->vertices[1][1][1] = t0; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = s1;
+   ctx->vertices[2][1][1] = t1;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = s0;
+   ctx->vertices[3][1][1] = t1;
+
+   return pipe_user_buffer_create( ctx->pipe->screen,
+                                   ctx->vertices,
+                                   sizeof(ctx->vertices) );
+}
+
+
+static struct pipe_buffer *
+setup_vertex_data_qtex(struct renderer *ctx,
+                       float x0, float y0, float x1, float y1,
+                       float x2, float y2, float x3, float y3,
+                       float s0, float t0, float s1, float t1,
+                       float z)
+{
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = s0; /*s*/
+   ctx->vertices[0][1][1] = t0; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y1;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = s1; /*s*/
+   ctx->vertices[1][1][1] = t0; /*t*/
+
+   ctx->vertices[2][0][0] = x2;
+   ctx->vertices[2][0][1] = y2;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = s1;
+   ctx->vertices[2][1][1] = t1;
+
+   ctx->vertices[3][0][0] = x3;
+   ctx->vertices[3][0][1] = y3;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = s0;
+   ctx->vertices[3][1][1] = t1;
+
+   return pipe_user_buffer_create( ctx->pipe->screen,
+                                   ctx->vertices,
+                                   sizeof(ctx->vertices) );
+}
+
+struct renderer * renderer_create(struct vg_context *owner)
+{
+   VGint i;
+   struct renderer *renderer = CALLOC_STRUCT(renderer);
+
+   if (!renderer)
+      return NULL;
+
+   renderer->owner = owner;
+   renderer->pipe = owner->pipe;
+   renderer->cso = owner->cso_context;
+
+   setup_shaders(renderer);
+
+   /* init vertex data that doesn't change */
+   for (i = 0; i < 4; i++) {
+      renderer->vertices[i][0][3] = 1.0f; /* w */
+      renderer->vertices[i][1][2] = 0.0f; /* r */
+      renderer->vertices[i][1][3] = 1.0f; /* q */
+   }
+
+   return renderer;
+}
+
+void renderer_destroy(struct renderer *ctx)
+{
+#if 0
+   if (ctx->fs) {
+      cso_delete_fragment_shader(ctx->cso, ctx->fs);
+      ctx->fs = NULL;
+   }
+#endif
+   free(ctx);
+}
+
+void renderer_draw_quad(struct renderer *r,
+                        VGfloat x1, VGfloat y1,
+                        VGfloat x2, VGfloat y2,
+                        VGfloat depth)
+{
+   struct pipe_buffer *buf;
+
+   buf = setup_vertex_data(r, x1, y1, x2, y2, depth);
+
+   if (buf) {
+      util_draw_vertex_buffer(r->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+}
+
+void renderer_draw_texture(struct renderer *r,
+                           struct pipe_texture *tex,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_buffer *buf;
+   VGfloat s0, t0, s1, t1;
+
+   assert(tex->width[0] != 0);
+   assert(tex->height[0] != 0);
+
+   s0 = x1offset / tex->width[0];
+   s1 = x2offset / tex->width[0];
+   t0 = y1offset / tex->height[0];
+   t1 = y2offset / tex->height[0];
+
+   cso_save_vertex_shader(r->cso);
+   /* shaders */
+   cso_set_vertex_shader_handle(r->cso, vg_texture_vs(r->owner));
+
+   /* draw quad */
+   buf = setup_vertex_data_tex(r, x1, y1, x2, y2,
+                               s0, t0, s1, t1, 0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+
+   cso_restore_vertex_shader(r->cso);
+}
+
+void renderer_copy_texture(struct renderer *ctx,
+                           struct pipe_texture *src,
+                           VGfloat sx1, VGfloat sy1,
+                           VGfloat sx2, VGfloat sy2,
+                           struct pipe_texture *dst,
+                           VGfloat dx1, VGfloat dy1,
+                           VGfloat dx2, VGfloat dy2)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_buffer *buf;
+   struct pipe_surface *dst_surf = screen->get_tex_surface(
+      screen, dst, 0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+   struct pipe_framebuffer_state fb;
+   float s0, t0, s1, t1;
+
+   assert(src->width[0] != 0);
+   assert(src->height[0] != 0);
+   assert(dst->width[0] != 0);
+   assert(dst->height[0] != 0);
+
+#if 0
+   debug_printf("copy texture [%f, %f, %f, %f], [%f, %f, %f, %f]\n",
+                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2);
+#endif
+
+#if 1
+   s0 = sx1 / src->width[0];
+   s1 = sx2 / src->width[0];
+   t0 = sy1 / src->height[0];
+   t1 = sy2 / src->height[0];
+#else
+   s0 = 0;
+   s1 = 1;
+   t0 = 0;
+   t1 = 1;
+#endif
+
+   assert(screen->is_format_supported(screen, dst_surf->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+
+   cso_save_viewport(ctx->cso);
+
+
+   /* set misc state we care about */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask = PIPE_MASK_RGBA;
+      cso_set_blend(ctx->cso, &blend);
+   }
+
+   /* sampler */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+      cso_single_sampler(ctx->cso, 0, &sampler);
+      cso_single_sampler_done(ctx->cso);
+   }
+
+   vg_set_viewport(ctx->owner, VEGA_Y0_TOP);
+
+   /* texture */
+   cso_set_sampler_textures(ctx->cso, 1, &src);
+
+   /* shaders */
+   cso_set_vertex_shader_handle(ctx->cso, vg_texture_vs(ctx->owner));
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width = dst_surf->width;
+   fb.height = dst_surf->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst_surf;
+   {
+      VGint i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(ctx->cso, &fb);
+
+   /* draw quad */
+   buf = setup_vertex_data_tex(ctx,
+                         dx1, dy1,
+                         dx2, dy2,
+                         s0, t0, s1, t1,
+                         0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(ctx->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+
+   pipe_surface_reference(&dst_surf, NULL);
+}
+
+void renderer_copy_surface(struct renderer *ctx,
+                           struct pipe_surface *src,
+                           int srcX0, int srcY0,
+                           int srcX1, int srcY1,
+                           struct pipe_surface *dst,
+                           int dstX0, int dstY0,
+                           int dstX1, int dstY1,
+                           float z, unsigned filter)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_buffer *buf;
+   struct pipe_texture texTemp, *tex;
+   struct pipe_surface *texSurf;
+   struct pipe_framebuffer_state fb;
+   struct st_framebuffer *stfb = ctx->owner->draw_buffer;
+   const int srcW = abs(srcX1 - srcX0);
+   const int srcH = abs(srcY1 - srcY0);
+   const int srcLeft = MIN2(srcX0, srcX1);
+   const int srcTop = MIN2(srcY0, srcY1);
+
+   assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
+          filter == PIPE_TEX_MIPFILTER_LINEAR);
+
+   if (srcLeft != srcX0) {
+      /* left-right flip */
+      int tmp = dstX0;
+      dstX0 = dstX1;
+      dstX1 = tmp;
+   }
+
+   if (srcTop != srcY0) {
+      /* up-down flip */
+      int tmp = dstY0;
+      dstY0 = dstY1;
+      dstY1 = tmp;
+   }
+
+   assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+
+   /*
+    * XXX for now we're always creating a temporary texture.
+    * Strictly speaking that's not always needed.
+    */
+
+   /* create temp texture */
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = src->format;
+   texTemp.last_level = 0;
+   texTemp.width[0] = srcW;
+   texTemp.height[0] = srcH;
+   texTemp.depth[0] = 1;
+   pf_get_block(src->format, &texTemp.block);
+
+   tex = screen->texture_create(screen, &texTemp);
+   if (!tex)
+      return;
+
+   texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* load temp texture */
+   pipe->surface_copy(pipe,
+                      texSurf, 0, 0,   /* dest */
+                      src, srcLeft, srcTop, /* src */
+                      srcW, srcH);     /* size */
+
+   /* free the surface, update the texture if necessary.*/
+   screen->tex_surface_destroy(texSurf);
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+   cso_save_viewport(ctx->cso);
+
+   /* set misc state we care about */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask = PIPE_MASK_RGBA;
+      cso_set_blend(ctx->cso, &blend);
+   }
+
+   vg_set_viewport(ctx->owner, VEGA_Y0_TOP);
+
+   /* sampler */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+      cso_single_sampler(ctx->cso, 0, &sampler);
+      cso_single_sampler_done(ctx->cso);
+   }
+
+   /* texture */
+   cso_set_sampler_textures(ctx->cso, 1, &tex);
+
+   /* shaders */
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+   cso_set_vertex_shader_handle(ctx->cso, vg_texture_vs(ctx->owner));
+
+   /* drawing dest */
+   if (stfb->strb->surface != dst) {
+      memset(&fb, 0, sizeof(fb));
+      fb.width = dst->width;
+      fb.height = dst->height;
+      fb.nr_cbufs = 1;
+      fb.cbufs[0] = dst;
+      fb.zsbuf = stfb->dsrb->surface;
+      cso_set_framebuffer(ctx->cso, &fb);
+   }
+
+   /* draw quad */
+   buf = setup_vertex_data(ctx,
+                           (float) dstX0, (float) dstY0,
+                           (float) dstX1, (float) dstY1, z);
+
+   if (buf) {
+      util_draw_vertex_buffer(ctx->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference( &buf,
+                             NULL );
+   }
+
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+
+   pipe_texture_reference(&tex, NULL);
+}
+
+void renderer_texture_quad(struct renderer *r,
+                           struct pipe_texture *tex,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2,
+                           VGfloat x3, VGfloat y3,
+                           VGfloat x4, VGfloat y4)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_buffer *buf;
+   VGfloat s0, t0, s1, t1;
+
+   assert(tex->width[0] != 0);
+   assert(tex->height[0] != 0);
+
+   s0 = x1offset / tex->width[0];
+   s1 = x2offset / tex->width[0];
+   t0 = y1offset / tex->height[0];
+   t1 = y2offset / tex->height[0];
+
+   cso_save_vertex_shader(r->cso);
+   /* shaders */
+   cso_set_vertex_shader_handle(r->cso, vg_texture_vs(r->owner));
+
+   /* draw quad */
+   buf = setup_vertex_data_qtex(r, x1, y1, x2, y2, x3, y3, x4, y4,
+                          s0, t0, s1, t1, 0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference(&buf,
+                            NULL);
+   }
+
+   cso_restore_vertex_shader(r->cso);
+}
diff --git a/src/gallium/state_trackers/vega/renderer.h b/src/gallium/state_trackers/vega/renderer.h
new file mode 100644
index 0000000000..990cd32c31
--- /dev/null
+++ b/src/gallium/state_trackers/vega/renderer.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef RENDERER_H
+#define RENDERER_H
+
+#include "VG/openvg.h"
+
+struct renderer;
+
+struct vg_context;
+struct pipe_texture;
+struct pipe_surface;
+
+struct renderer *renderer_create(struct vg_context *owner);
+void renderer_destroy(struct renderer *);
+
+void renderer_draw_quad(struct renderer *,
+                        VGfloat x1, VGfloat y1,
+                        VGfloat x2, VGfloat y2,
+                        VGfloat depth);
+void renderer_draw_texture(struct renderer *,
+                           struct pipe_texture *texture,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2);
+void renderer_texture_quad(struct renderer *,
+                           struct pipe_texture *texture,
+                           VGfloat x1offset, VGfloat y1offset,
+                           VGfloat x2offset, VGfloat y2offset,
+                           VGfloat x1, VGfloat y1,
+                           VGfloat x2, VGfloat y2,
+                           VGfloat x3, VGfloat y3,
+                           VGfloat x4, VGfloat y4);
+void renderer_copy_texture(struct renderer *r,
+                           struct pipe_texture *src,
+                           VGfloat sx1, VGfloat sy1,
+                           VGfloat sx2, VGfloat sy2,
+                           struct pipe_texture *dst,
+                           VGfloat dx1, VGfloat dy1,
+                           VGfloat dx2, VGfloat dy2);
+void renderer_copy_surface(struct renderer *r,
+                           struct pipe_surface *src,
+                           int sx1, int sy1,
+                           int sx2, int sy2,
+                           struct pipe_surface *dst,
+                           int dx1, int dy1,
+                           int dx2, int dy2,
+                           float z, unsigned filter);
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/shader.c b/src/gallium/state_trackers/vega/shader.c
new file mode 100644
index 0000000000..d9074a377b
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shader.c
@@ -0,0 +1,310 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "shader.h"
+
+#include "vg_context.h"
+#include "shaders_cache.h"
+#include "paint.h"
+#include "mask.h"
+#include "image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+
+#define MAX_CONSTANTS 20
+
+struct shader {
+   struct vg_context *context;
+
+   VGboolean masking;
+   struct vg_paint *paint;
+   struct vg_image *image;
+
+   VGboolean drawing_image;
+   VGImageMode image_mode;
+
+   float constants[MAX_CONSTANTS];
+   struct pipe_constant_buffer cbuf;
+   struct pipe_shader_state fs_state;
+   void *fs;
+};
+
+struct shader * shader_create(struct vg_context *ctx)
+{
+   struct shader *shader = 0;
+
+   shader = CALLOC_STRUCT(shader);
+   shader->context = ctx;
+
+   return shader;
+}
+
+void shader_destroy(struct shader *shader)
+{
+   free(shader);
+}
+
+void shader_set_masking(struct shader *shader, VGboolean set)
+{
+   shader->masking = set;
+}
+
+VGboolean shader_is_masking(struct shader *shader)
+{
+   return shader->masking;
+}
+
+void shader_set_paint(struct shader *shader, struct vg_paint *paint)
+{
+   shader->paint = paint;
+}
+
+struct vg_paint * shader_paint(struct shader *shader)
+{
+   return shader->paint;
+}
+
+
+static void setup_constant_buffer(struct shader *shader)
+{
+   struct vg_context *ctx = shader->context;
+   struct pipe_context *pipe = shader->context->pipe;
+   struct pipe_constant_buffer *cbuf = &shader->cbuf;
+   VGint param_bytes = paint_constant_buffer_size(shader->paint);
+   float temp_buf[MAX_CONSTANTS];
+
+   assert(param_bytes <= sizeof(temp_buf));
+   paint_fill_constant_buffer(shader->paint, temp_buf);
+
+   if (cbuf->buffer == NULL ||
+       memcmp(temp_buf, shader->constants, param_bytes) != 0)
+   {
+      pipe_buffer_reference(&cbuf->buffer, NULL);
+
+      memcpy(shader->constants, temp_buf, param_bytes);
+      cbuf->buffer = pipe_user_buffer_create(pipe->screen,
+                                             &shader->constants,
+                                             sizeof(shader->constants));
+   }
+
+   ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+}
+
+static VGint blend_bind_samplers(struct vg_context *ctx,
+                                 struct pipe_sampler_state **samplers,
+                                 struct pipe_texture **textures)
+{
+   VGBlendMode bmode = ctx->state.vg.blend_mode;
+
+   if (bmode == VG_BLEND_MULTIPLY ||
+       bmode == VG_BLEND_SCREEN ||
+       bmode == VG_BLEND_DARKEN ||
+       bmode == VG_BLEND_LIGHTEN) {
+      struct st_framebuffer *stfb = ctx->draw_buffer;
+
+      vg_prepare_blend_surface(ctx);
+
+      samplers[2] = &ctx->blend_sampler;
+      textures[2] = stfb->blend_texture;
+
+      if (!samplers[0] || !textures[0]) {
+         samplers[1] = samplers[2];
+         textures[1] = textures[2];
+      }
+      if (!samplers[1] || !textures[1]) {
+         samplers[1] = samplers[0];
+         textures[1] = textures[0];
+      }
+
+      return 1;
+   }
+   return 0;
+}
+
+static void setup_samplers(struct shader *shader)
+{
+   struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   struct vg_context *ctx = shader->context;
+   /* a little wonky: we use the num as a boolean that just says
+    * whether any sampler/textures have been set. the actual numbering
+    * for samplers is always the same:
+    * 0 - paint sampler/texture for gradient/pattern
+    * 1 - mask sampler/texture
+    * 2 - blend sampler/texture
+    * 3 - image sampler/texture
+    * */
+   VGint num = 0;
+
+   samplers[0] = NULL;
+   samplers[1] = NULL;
+   samplers[2] = NULL;
+   samplers[3] = NULL;
+   textures[0] = NULL;
+   textures[1] = NULL;
+   textures[2] = NULL;
+   textures[3] = NULL;
+
+   num += paint_bind_samplers(shader->paint, samplers, textures);
+   num += mask_bind_samplers(samplers, textures);
+   num += blend_bind_samplers(ctx, samplers, textures);
+   if (shader->drawing_image && shader->image)
+      num += image_bind_samplers(shader->image, samplers, textures);
+
+   if (num) {
+      cso_set_samplers(ctx->cso_context, 4, (const struct pipe_sampler_state **)samplers);
+      cso_set_sampler_textures(ctx->cso_context, 4, textures);
+   }
+}
+
+static INLINE VGboolean is_format_bw(struct shader *shader)
+{
+#if 0
+   struct vg_context *ctx = shader->context;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+#endif
+
+   if (shader->drawing_image && shader->image) {
+      if (shader->image->format == VG_BW_1)
+         return VG_TRUE;
+   }
+
+   return VG_FALSE;
+}
+
+static void setup_shader_program(struct shader *shader)
+{
+   struct vg_context *ctx = shader->context;
+   VGint shader_id = 0;
+   VGBlendMode blend_mode = ctx->state.vg.blend_mode;
+   VGboolean black_white = is_format_bw(shader);
+
+   /* 1st stage: fill */
+   if (!shader->drawing_image ||
+       (shader->image_mode == VG_DRAW_IMAGE_MULTIPLY || shader->image_mode == VG_DRAW_IMAGE_STENCIL)) {
+      switch(paint_type(shader->paint)) {
+      case VG_PAINT_TYPE_COLOR:
+         shader_id |= VEGA_SOLID_FILL_SHADER;
+         break;
+      case VG_PAINT_TYPE_LINEAR_GRADIENT:
+         shader_id |= VEGA_LINEAR_GRADIENT_SHADER;
+         break;
+      case VG_PAINT_TYPE_RADIAL_GRADIENT:
+         shader_id |= VEGA_RADIAL_GRADIENT_SHADER;
+         break;
+      case VG_PAINT_TYPE_PATTERN:
+         shader_id |= VEGA_PATTERN_SHADER;
+         break;
+
+      default:
+         abort();
+      }
+   }
+
+   /* second stage image */
+   if (shader->drawing_image) {
+      switch(shader->image_mode) {
+      case VG_DRAW_IMAGE_NORMAL:
+         shader_id |= VEGA_IMAGE_NORMAL_SHADER;
+         break;
+      case VG_DRAW_IMAGE_MULTIPLY:
+         shader_id |= VEGA_IMAGE_MULTIPLY_SHADER;
+         break;
+      case VG_DRAW_IMAGE_STENCIL:
+         shader_id |= VEGA_IMAGE_STENCIL_SHADER;
+         break;
+      default:
+         debug_printf("Unknown image mode!");
+      }
+   }
+
+   if (shader->masking)
+      shader_id |= VEGA_MASK_SHADER;
+
+   switch(blend_mode) {
+   case VG_BLEND_MULTIPLY:
+      shader_id |= VEGA_BLEND_MULTIPLY_SHADER;
+      break;
+   case VG_BLEND_SCREEN:
+      shader_id |= VEGA_BLEND_SCREEN_SHADER;
+      break;
+   case VG_BLEND_DARKEN:
+      shader_id |= VEGA_BLEND_DARKEN_SHADER;
+      break;
+   case VG_BLEND_LIGHTEN:
+      shader_id |= VEGA_BLEND_LIGHTEN_SHADER;
+      break;
+   default:
+      /* handled by pipe_blend_state */
+      break;
+   }
+
+   if (black_white)
+      shader_id |= VEGA_BW_SHADER;
+
+   shader->fs = shaders_cache_fill(ctx->sc, shader_id);
+   cso_set_fragment_shader_handle(ctx->cso_context, shader->fs);
+}
+
+
+void shader_bind(struct shader *shader)
+{
+   /* first resolve the real paint type */
+   paint_resolve_type(shader->paint);
+
+   setup_constant_buffer(shader);
+   setup_samplers(shader);
+   setup_shader_program(shader);
+}
+
+void shader_set_image_mode(struct shader *shader, VGImageMode image_mode)
+{
+   shader->image_mode = image_mode;
+}
+
+VGImageMode shader_image_mode(struct shader *shader)
+{
+   return shader->image_mode;
+}
+
+void shader_set_drawing_image(struct shader *shader, VGboolean drawing_image)
+{
+   shader->drawing_image = drawing_image;
+}
+
+VGboolean shader_drawing_image(struct shader *shader)
+{
+   return shader->drawing_image;
+}
+
+void shader_set_image(struct shader *shader, struct vg_image *img)
+{
+   shader->image = img;
+}
diff --git a/src/gallium/state_trackers/vega/shader.h b/src/gallium/state_trackers/vega/shader.h
new file mode 100644
index 0000000000..847eee6a31
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shader.h
@@ -0,0 +1,56 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef SHADER_H
+#define SHADER_H
+
+#include "VG/openvg.h"
+
+struct shader;
+struct vg_paint;
+struct vg_context;
+struct vg_image;
+
+struct shader *shader_create(struct vg_context *context);
+void shader_destroy(struct shader *shader);
+
+void shader_set_masking(struct shader *shader, VGboolean set);
+VGboolean shader_is_masking(struct shader *shader);
+
+void shader_set_paint(struct shader *shader, struct vg_paint *paint);
+struct vg_paint *shader_paint(struct shader *shader);
+
+void shader_set_image_mode(struct shader *shader, VGImageMode image_mode);
+VGImageMode shader_image_mode(struct shader *shader);
+
+void shader_set_drawing_image(struct shader *shader, VGboolean drawing_image);
+VGboolean shader_drawing_image(struct shader *shader);
+
+void shader_set_image(struct shader *shader, struct vg_image *img);
+
+void shader_bind(struct shader *shader);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/shaders_cache.c b/src/gallium/state_trackers/vega/shaders_cache.c
new file mode 100644
index 0000000000..fd0831fab1
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shaders_cache.c
@@ -0,0 +1,439 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "shaders_cache.h"
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_text.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "util/u_debug.h"
+#include "cso_cache/cso_hash.h"
+#include "cso_cache/cso_context.h"
+
+#include "VG/openvg.h"
+
+#include "asm_fill.h"
+
+/* Essentially we construct an ubber-shader based on the state
+ * of the pipeline. The stages are:
+ * 1) Fill (mandatory, solid color/gradient/pattern/image draw)
+ * 2) Image composition (image mode multiply and stencil)
+ * 3) Mask
+ * 4) Extended blend (multiply/screen/darken/lighten)
+ * 5) Premultiply/Unpremultiply
+ * 6) Color transform (to black and white)
+ */
+#define SHADER_STAGES 6
+
+struct cached_shader {
+   void *driver_shader;
+   struct pipe_shader_state state;
+};
+
+struct shaders_cache {
+   struct vg_context *pipe;
+
+   struct cso_hash *hash;
+};
+
+
+static INLINE struct tgsi_token *tokens_from_assembly(const char *txt, int num_tokens)
+{
+   struct tgsi_token *tokens;
+
+   tokens = (struct tgsi_token *) MALLOC(num_tokens * sizeof(tokens[0]));
+
+   tgsi_text_translate(txt, tokens, num_tokens);
+
+#if DEBUG_SHADERS
+   tgsi_dump(tokens, 0);
+#endif
+
+   return tokens;
+}
+
+#define ALL_FILLS (VEGA_SOLID_FILL_SHADER | \
+   VEGA_LINEAR_GRADIENT_SHADER | \
+   VEGA_RADIAL_GRADIENT_SHADER | \
+   VEGA_PATTERN_SHADER         | \
+   VEGA_IMAGE_NORMAL_SHADER)
+
+
+/*
+static const char max_shader_preamble[] =
+   "FRAG1.1\n"
+   "DCL IN[0], POSITION, LINEAR\n"
+   "DCL IN[1], GENERIC[0], PERSPECTIVE\n"
+   "DCL OUT[0], COLOR, CONSTANT\n"
+   "DCL CONST[0..9], CONSTANT\n"
+   "DCL TEMP[0..9], CONSTANT\n"
+   "DCL SAMP[0..9], CONSTANT\n";
+
+   max_shader_preamble strlen == 175
+*/
+#define MAX_PREAMBLE 175
+
+static INLINE VGint range_min(VGint min, VGint current)
+{
+   if (min < 0)
+      min = current;
+   else
+      min = MIN2(min, current);
+   return min;
+}
+
+static INLINE VGint range_max(VGint max, VGint current)
+{
+   return MAX2(max, current);
+}
+
+static void
+create_preamble(char *txt,
+                const struct shader_asm_info *shaders[SHADER_STAGES],
+                int num_shaders)
+{
+   VGboolean declare_input = VG_FALSE;
+   VGint start_const   = -1, end_const   = 0;
+   VGint start_temp    = -1, end_temp    = 0;
+   VGint start_sampler = -1, end_sampler = 0;
+   VGint i;
+   VGint num_consts, num_temps, num_samplers;
+
+   for (i = 0; i < num_shaders; ++i) {
+      if (shaders[i]->num_consts)
+         start_const = range_min(start_const, shaders[i]->start_const);
+      if (shaders[i]->num_temps)
+         start_temp = range_min(start_temp, shaders[i]->start_temp);
+      if (shaders[i]->num_samplers)
+         start_sampler = range_min(start_sampler, shaders[i]->start_sampler);
+
+      end_const = range_max(end_const, shaders[i]->start_const +
+                            shaders[i]->num_consts);
+      end_temp = range_max(end_temp, shaders[i]->start_temp +
+                            shaders[i]->num_temps);
+      end_sampler = range_max(end_sampler, shaders[i]->start_sampler +
+                            shaders[i]->num_samplers);
+      if (shaders[i]->needs_position)
+         declare_input = VG_TRUE;
+   }
+   /* if they're still unitialized, initialize them */
+   if (start_const < 0)
+      start_const = 0;
+   if (start_temp < 0)
+      start_temp = 0;
+   if (start_sampler < 0)
+      start_sampler = 0;
+
+   num_consts   = end_const   - start_const;
+   num_temps    = end_temp    - start_temp;
+   num_samplers = end_sampler - start_sampler;
+   /* end exclusive */
+   --end_const;
+   --end_temp;
+   --end_sampler;
+
+   sprintf(txt, "FRAG1.1\n");
+
+   if (declare_input) {
+      sprintf(txt + strlen(txt), "DCL IN[0], POSITION, LINEAR\n");
+      sprintf(txt + strlen(txt), "DCL IN[1], GENERIC[0], PERSPECTIVE\n");
+   }
+
+   /* we always have a color output */
+   sprintf(txt + strlen(txt), "DCL OUT[0], COLOR, CONSTANT\n");
+
+   if (num_consts > 1)
+      sprintf(txt + strlen(txt), "DCL CONST[%d..%d], CONSTANT\n", start_const, end_const);
+   else if (num_consts == 1)
+      sprintf(txt + strlen(txt), "DCL CONST[%d], CONSTANT\n", start_const);
+
+   if (num_temps > 1)
+      sprintf(txt + strlen(txt), "DCL TEMP[%d..%d], CONSTANT\n", start_temp, end_temp);
+   else if (num_temps > 1)
+      sprintf(txt + strlen(txt), "DCL TEMP[%d], CONSTANT\n", start_temp);
+
+   if (num_samplers > 1)
+      sprintf(txt + strlen(txt), "DCL SAMP[%d..%d], CONSTANT\n", start_sampler, end_sampler);
+   else if (num_samplers == 1)
+      sprintf(txt + strlen(txt), "DCL SAMP[%d], CONSTANT\n", start_sampler);
+}
+
+static void *
+combine_shaders(const struct shader_asm_info *shaders[SHADER_STAGES], int num_shaders,
+                struct pipe_context *pipe,
+                struct pipe_shader_state *shader)
+{
+   char *combined_txt;
+   int combined_len = MAX_PREAMBLE;
+   int combined_tokens = 0;
+   int i = 0;
+   int current_shader = 0;
+   int current_len;
+
+   for (i = 0; i < num_shaders; ++i) {
+      combined_len += strlen(shaders[i]->txt);
+      combined_tokens += shaders[i]->num_tokens;
+   }
+   /* add for the %s->TEMP[0] substitutions */
+   combined_len += num_shaders * 7 /*TEMP[0]*/ + 4 /*"END\n"*/;
+
+   combined_txt = (char*)malloc(combined_len);
+   combined_txt[0] = '\0';
+
+   create_preamble(combined_txt, shaders, num_shaders);
+
+   while (current_shader < num_shaders) {
+      const char temp[] = "TEMP[0]";
+      const char out[] = "OUT[0]";
+      const char *subst = temp;
+
+      current_len = strlen(combined_txt);
+
+      /* if the last shader then output */
+      if (current_shader + 1 == num_shaders)
+         subst = out;
+
+      snprintf(combined_txt + current_len,
+               combined_len - current_len,
+               shaders[current_shader]->txt,
+               subst);
+      ++current_shader;
+   }
+
+
+   current_len = strlen(combined_txt);
+   snprintf(combined_txt + current_len,
+            combined_len - current_len,
+            "END\n");
+
+   debug_printf("Combined shader is : \n%s\n",
+                 combined_txt);
+
+   shader->tokens = tokens_from_assembly(
+            combined_txt, combined_tokens);
+
+   free(combined_txt);
+
+   return pipe->create_fs_state(pipe, shader);
+}
+
+static void *
+create_shader(struct pipe_context *pipe,
+              int id,
+              struct pipe_shader_state *shader)
+{
+   int idx = 0;
+   const struct shader_asm_info * shaders[SHADER_STAGES];
+
+   /* the shader has to have a fill */
+   debug_assert(id & ALL_FILLS);
+
+   /* first stage */
+   if (id & VEGA_SOLID_FILL_SHADER) {
+      debug_assert(idx == 0);
+      shaders[idx] = &shaders_asm[0];
+      debug_assert(shaders_asm[0].id == VEGA_SOLID_FILL_SHADER);
+      ++idx;
+   }
+   if ((id & VEGA_LINEAR_GRADIENT_SHADER)) {
+      debug_assert(idx == 0);
+      shaders[idx] = &shaders_asm[1];
+      debug_assert(shaders_asm[1].id == VEGA_LINEAR_GRADIENT_SHADER);
+      ++idx;
+   }
+   if ((id & VEGA_RADIAL_GRADIENT_SHADER)) {
+      debug_assert(idx == 0);
+      shaders[idx] = &shaders_asm[2];
+      debug_assert(shaders_asm[2].id == VEGA_RADIAL_GRADIENT_SHADER);
+      ++idx;
+   }
+   if ((id & VEGA_PATTERN_SHADER)) {
+      debug_assert(idx == 0);
+      debug_assert(shaders_asm[3].id == VEGA_PATTERN_SHADER);
+      shaders[idx] = &shaders_asm[3];
+      ++idx;
+   }
+   if ((id & VEGA_IMAGE_NORMAL_SHADER)) {
+      debug_assert(idx == 0);
+      debug_assert(shaders_asm[4].id == VEGA_IMAGE_NORMAL_SHADER);
+      shaders[idx] = &shaders_asm[4];
+      ++idx;
+   }
+
+   /* second stage */
+   if ((id & VEGA_IMAGE_MULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[5].id == VEGA_IMAGE_MULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[5];
+      ++idx;
+   } else if ((id & VEGA_IMAGE_STENCIL_SHADER)) {
+      debug_assert(shaders_asm[6].id == VEGA_IMAGE_STENCIL_SHADER);
+      shaders[idx] = &shaders_asm[6];
+      ++idx;
+   }
+
+   /* third stage */
+   if ((id & VEGA_MASK_SHADER)) {
+      debug_assert(idx == 1);
+      debug_assert(shaders_asm[7].id == VEGA_MASK_SHADER);
+      shaders[idx] = &shaders_asm[7];
+      ++idx;
+   }
+
+   /* fourth stage */
+   if ((id & VEGA_BLEND_MULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[8].id == VEGA_BLEND_MULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[8];
+      ++idx;
+   } else if ((id & VEGA_BLEND_SCREEN_SHADER)) {
+      debug_assert(shaders_asm[9].id == VEGA_BLEND_SCREEN_SHADER);
+      shaders[idx] = &shaders_asm[9];
+      ++idx;
+   } else if ((id & VEGA_BLEND_DARKEN_SHADER)) {
+      debug_assert(shaders_asm[10].id == VEGA_BLEND_DARKEN_SHADER);
+      shaders[idx] = &shaders_asm[10];
+      ++idx;
+   } else if ((id & VEGA_BLEND_LIGHTEN_SHADER)) {
+      debug_assert(shaders_asm[11].id == VEGA_BLEND_LIGHTEN_SHADER);
+      shaders[idx] = &shaders_asm[11];
+      ++idx;
+   }
+
+   /* fifth stage */
+   if ((id & VEGA_PREMULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[12].id == VEGA_PREMULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[12];
+      ++idx;
+   } else if ((id & VEGA_UNPREMULTIPLY_SHADER)) {
+      debug_assert(shaders_asm[13].id == VEGA_UNPREMULTIPLY_SHADER);
+      shaders[idx] = &shaders_asm[13];
+      ++idx;
+   }
+
+   /* sixth stage */
+   if ((id & VEGA_BW_SHADER)) {
+      debug_assert(shaders_asm[14].id == VEGA_BW_SHADER);
+      shaders[idx] = &shaders_asm[14];
+      ++idx;
+   }
+
+   return combine_shaders(shaders, idx, pipe, shader);
+}
+
+/*************************************************/
+
+struct shaders_cache * shaders_cache_create(struct vg_context *vg)
+{
+   struct shaders_cache *sc = CALLOC_STRUCT(shaders_cache);
+
+   sc->pipe = vg;
+   sc->hash = cso_hash_create();
+
+   return sc;
+}
+
+void shaders_cache_destroy(struct shaders_cache *sc)
+{
+   struct cso_hash_iter iter = cso_hash_first_node(sc->hash);
+
+   while (!cso_hash_iter_is_null(iter)) {
+      struct cached_shader *cached =
+         (struct cached_shader *)cso_hash_iter_data(iter);
+      cso_delete_fragment_shader(sc->pipe->cso_context,
+                                 cached->driver_shader);
+      iter = cso_hash_erase(sc->hash, iter);
+   }
+
+   cso_hash_delete(sc->hash);
+   free(sc);
+}
+
+void * shaders_cache_fill(struct shaders_cache *sc,
+                          int shader_key)
+{
+   VGint key = shader_key;
+   struct cached_shader *cached;
+   struct cso_hash_iter iter = cso_hash_find(sc->hash, key);
+
+   if (cso_hash_iter_is_null(iter)) {
+      cached = CALLOC_STRUCT(cached_shader);
+      cached->driver_shader = create_shader(sc->pipe->pipe, key, &cached->state);
+
+      cso_hash_insert(sc->hash, key, cached);
+
+      return cached->driver_shader;
+   }
+
+   cached = (struct cached_shader *)cso_hash_iter_data(iter);
+
+   assert(cached->driver_shader);
+   return cached->driver_shader;
+}
+
+struct vg_shader * shader_create_from_text(struct pipe_context *pipe,
+                                           const char *txt, int num_tokens,
+                                           int type)
+{
+   struct vg_shader *shader = (struct vg_shader *)malloc(
+      sizeof(struct vg_shader));
+   struct tgsi_token *tokens = tokens_from_assembly(txt, num_tokens);
+   struct pipe_shader_state state;
+
+   debug_assert(type == PIPE_SHADER_VERTEX ||
+                type == PIPE_SHADER_FRAGMENT);
+
+   state.tokens = tokens;
+   shader->type = type;
+   shader->tokens = tokens;
+
+   if (type == PIPE_SHADER_FRAGMENT)
+      shader->driver = pipe->create_fs_state(pipe, &state);
+   else
+      shader->driver = pipe->create_vs_state(pipe, &state);
+   return shader;
+}
+
+void vg_shader_destroy(struct vg_context *ctx, struct vg_shader *shader)
+{
+   if (shader->type == PIPE_SHADER_FRAGMENT)
+      cso_delete_fragment_shader(ctx->cso_context, shader->driver);
+   else
+      cso_delete_vertex_shader(ctx->cso_context, shader->driver);
+   free(shader->tokens);
+   free(shader);
+}
diff --git a/src/gallium/state_trackers/vega/shaders_cache.h b/src/gallium/state_trackers/vega/shaders_cache.h
new file mode 100644
index 0000000000..feca58b61a
--- /dev/null
+++ b/src/gallium/state_trackers/vega/shaders_cache.h
@@ -0,0 +1,77 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef SHADERS_CACHE_H
+#define SHADERS_CACHE_H
+
+
+struct vg_context;
+struct pipe_context;
+struct tgsi_token;
+struct shaders_cache;
+
+enum VegaShaderType {
+   VEGA_SOLID_FILL_SHADER         = 1 <<  0,
+   VEGA_LINEAR_GRADIENT_SHADER    = 1 <<  1,
+   VEGA_RADIAL_GRADIENT_SHADER    = 1 <<  2,
+   VEGA_PATTERN_SHADER            = 1 <<  3,
+   VEGA_IMAGE_NORMAL_SHADER       = 1 <<  4,
+   VEGA_IMAGE_MULTIPLY_SHADER     = 1 <<  5,
+   VEGA_IMAGE_STENCIL_SHADER      = 1 <<  6,
+
+   VEGA_MASK_SHADER               = 1 <<  7,
+
+   VEGA_BLEND_MULTIPLY_SHADER     = 1 <<  8,
+   VEGA_BLEND_SCREEN_SHADER       = 1 <<  9,
+   VEGA_BLEND_DARKEN_SHADER       = 1 << 10,
+   VEGA_BLEND_LIGHTEN_SHADER      = 1 << 11,
+
+   VEGA_PREMULTIPLY_SHADER        = 1 << 12,
+   VEGA_UNPREMULTIPLY_SHADER      = 1 << 13,
+
+   VEGA_BW_SHADER                 = 1 << 14
+};
+
+struct vg_shader {
+   void *driver;
+   struct tgsi_token *tokens;
+   int type;/* PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT */
+};
+
+struct shaders_cache *shaders_cache_create(struct vg_context *pipe);
+void shaders_cache_destroy(struct shaders_cache *sc);
+void *shaders_cache_fill(struct shaders_cache *sc,
+                         int shader_key);
+
+struct vg_shader *shader_create_from_text(struct pipe_context *pipe,
+                                          const char *txt, int num_tokens,
+                                          int type);
+
+void vg_shader_destroy(struct vg_context *ctx, struct vg_shader *shader);
+
+
+
+#endif
diff --git a/src/gallium/state_trackers/vega/st_inlines.h b/src/gallium/state_trackers/vega/st_inlines.h
new file mode 100644
index 0000000000..1f331dfcdb
--- /dev/null
+++ b/src/gallium/state_trackers/vega/st_inlines.h
@@ -0,0 +1,159 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Functions for checking if buffers/textures are referenced when we need
+ * to read/write from/to them.  Flush when needed.
+ */
+
+#ifndef ST_INLINES_H
+#define ST_INLINES_H
+
+#include "vg_context.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_state.h"
+
+static INLINE struct pipe_transfer *
+st_cond_flush_get_tex_transfer(struct vg_context *st,
+			       struct pipe_texture *pt,
+			       unsigned int face,
+			       unsigned int level,
+			       unsigned int zslice,
+			       enum pipe_transfer_usage usage,
+			       unsigned int x, unsigned int y,
+			       unsigned int w, unsigned int h)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_context *pipe = st->pipe;
+   unsigned referenced =
+      pipe->is_texture_referenced(pipe, pt, face, level);
+
+   if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+		      usage == PIPE_TRANSFER_WRITE ||
+		      usage == PIPE_TRANSFER_READ_WRITE))
+      vgFlush();
+
+   return screen->get_tex_transfer(screen, pt, face, level, zslice, usage,
+				   x, y, w, h);
+}
+
+static INLINE struct pipe_transfer *
+st_no_flush_get_tex_transfer(struct vg_context *st,
+			     struct pipe_texture *pt,
+			     unsigned int face,
+			     unsigned int level,
+			     unsigned int zslice,
+			     enum pipe_transfer_usage usage,
+			     unsigned int x, unsigned int y,
+			     unsigned int w, unsigned int h)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+
+   return screen->get_tex_transfer(screen, pt, face, level,
+				   zslice, usage, x, y, w, h);
+}
+
+static INLINE void *
+st_cond_flush_pipe_buffer_map(struct vg_context *st,
+			      struct pipe_buffer *buf,
+			      unsigned int map_flags)
+{
+   struct pipe_context *pipe = st->pipe;
+   unsigned int referenced = pipe->is_buffer_referenced(pipe, buf);
+
+   if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+		      (map_flags & PIPE_BUFFER_USAGE_CPU_WRITE)))
+      vgFlush();
+
+   return pipe_buffer_map(pipe->screen, buf, map_flags);
+}
+
+static INLINE void *
+st_no_flush_pipe_buffer_map(struct vg_context *st,
+			    struct pipe_buffer *buf,
+			    unsigned int map_flags)
+{
+   return pipe_buffer_map(st->pipe->screen, buf, map_flags);
+}
+
+
+static INLINE void
+st_cond_flush_pipe_buffer_write(struct vg_context *st,
+				struct pipe_buffer *buf,
+				unsigned int offset,
+				unsigned int size,
+				const void * data)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (pipe->is_buffer_referenced(pipe, buf))
+      vgFlush();
+
+   pipe_buffer_write(pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_no_flush_pipe_buffer_write(struct vg_context *st,
+			      struct pipe_buffer *buf,
+			      unsigned int offset,
+			      unsigned int size,
+			      const void * data)
+{
+   pipe_buffer_write(st->pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_cond_flush_pipe_buffer_read(struct vg_context *st,
+			       struct pipe_buffer *buf,
+			       unsigned int offset,
+			       unsigned int size,
+			       void * data)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (pipe->is_buffer_referenced(pipe, buf) & PIPE_REFERENCED_FOR_WRITE)
+      vgFlush();
+
+   pipe_buffer_read(pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_no_flush_pipe_buffer_read(struct vg_context *st,
+			     struct pipe_buffer *buf,
+			     unsigned int offset,
+			     unsigned int size,
+			     void * data)
+{
+   pipe_buffer_read(st->pipe->screen, buf, offset, size, data);
+}
+
+#endif
+
diff --git a/src/gallium/state_trackers/vega/stroker.c b/src/gallium/state_trackers/vega/stroker.c
new file mode 100644
index 0000000000..1b92d2b5c6
--- /dev/null
+++ b/src/gallium/state_trackers/vega/stroker.c
@@ -0,0 +1,1349 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "stroker.h"
+
+#include "path.h"
+#include "vg_state.h"
+#include "util_array.h"
+#include "arc.h"
+#include "bezier.h"
+#include "matrix.h"
+#include "path_utils.h"
+#include "polygon.h"
+
+#include "math.h"
+
+#ifndef M_2PI
+#define M_2PI 6.28318530717958647692528676655900576
+#endif
+
+#define STROKE_SEGMENTS 0
+#define STROKE_DEBUG 0
+#define DEBUG_EMITS 0
+
+static const VGfloat curve_threshold = 0.25f;
+
+static const VGfloat zero_coords[] = {0.f, 0.f};
+
+enum intersection_type {
+   NoIntersections,
+   BoundedIntersection,
+   UnboundedIntersection,
+};
+
+enum line_join_mode {
+   FlatJoin,
+   SquareJoin,
+   MiterJoin,
+   RoundJoin,
+   RoundCap
+};
+
+struct stroke_iterator {
+   void (*next)(struct stroke_iterator *);
+   VGboolean (*has_next)(struct stroke_iterator *);
+
+   VGPathCommand (*current_command)(struct stroke_iterator *it);
+   void (*current_coords)(struct stroke_iterator *it, VGfloat *coords);
+
+   VGint position;
+   VGint coord_position;
+
+   const VGubyte *cmds;
+   const VGfloat *coords;
+   VGint num_commands;
+   VGint num_coords;
+
+   struct polygon *curve_poly;
+   VGint curve_index;
+};
+
+static VGPathCommand stroke_itr_command(struct stroke_iterator *itr)
+{
+   return itr->current_command(itr);
+}
+
+static void stroke_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   itr->current_coords(itr, coords);
+}
+
+static void stroke_fw_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   if (itr->position >= itr->num_commands)
+      return;
+   switch (stroke_itr_command(itr)) {
+   case VG_MOVE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_LINE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_CUBIC_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      coords[2] = itr->coords[itr->coord_position + 2];
+      coords[3] = itr->coords[itr->coord_position + 3];
+      coords[4] = itr->coords[itr->coord_position + 4];
+      coords[5] = itr->coords[itr->coord_position + 5];
+      break;
+   default:
+      debug_assert(!"invalid command!\n");
+   }
+}
+
+
+static void stroke_bw_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   if (itr->position >= itr->num_commands)
+      return;
+   switch (stroke_itr_command(itr)) {
+   case VG_MOVE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_LINE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_CUBIC_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position + 4];
+      coords[1] = itr->coords[itr->coord_position + 5];
+      coords[2] = itr->coords[itr->coord_position + 2];
+      coords[3] = itr->coords[itr->coord_position + 3];
+      coords[4] = itr->coords[itr->coord_position + 0];
+      coords[5] = itr->coords[itr->coord_position + 1];
+      break;
+   default:
+      debug_assert(!"invalid command!\n");
+   }
+}
+
+
+static VGPathCommand stroke_fw_current_command(struct stroke_iterator *it)
+{
+   return it->cmds[it->position];
+}
+
+static VGPathCommand stroke_bw_current_command(struct stroke_iterator *it)
+{
+   VGPathCommand prev_cmd;
+   if (it->position == it->num_commands  -1)
+      return VG_MOVE_TO_ABS;
+
+   prev_cmd = it->cmds[it->position + 1];
+   return prev_cmd;
+}
+
+static VGboolean stroke_fw_has_next(struct stroke_iterator *itr)
+{
+   return itr->position < (itr->num_commands - 1);
+}
+
+static VGboolean stroke_bw_has_next(struct stroke_iterator *itr)
+{
+   return itr->position > 0;
+}
+
+static void stroke_fw_next(struct stroke_iterator *itr)
+{
+   VGubyte cmd;
+   debug_assert(stroke_fw_has_next(itr));
+
+   cmd = stroke_itr_command(itr);
+
+   itr->coord_position += num_elements_for_segments(&cmd, 1);
+   ++itr->position;
+}
+
+static void stroke_bw_next(struct stroke_iterator *itr)
+{
+   VGubyte cmd;
+   debug_assert(stroke_bw_has_next(itr));
+
+   --itr->position;
+   cmd = stroke_itr_command(itr);
+
+   itr->coord_position -= num_elements_for_segments(&cmd, 1);
+}
+
+static void stroke_itr_common_init(struct stroke_iterator *itr,
+                                   struct array *cmds,
+                                   struct array *coords)
+{
+   itr->cmds = (VGubyte*)cmds->data;
+   itr->num_commands = cmds->num_elements;
+
+   itr->coords = (VGfloat*)coords->data;
+   itr->num_coords = coords->num_elements;
+}
+
+static void stroke_forward_iterator(struct stroke_iterator *itr,
+                                    struct array *cmds,
+                                    struct array *coords)
+{
+   stroke_itr_common_init(itr, cmds, coords);
+   itr->position = 0;
+   itr->coord_position = 0;
+
+   itr->next = stroke_fw_next;
+   itr->has_next = stroke_fw_has_next;
+   itr->current_command = stroke_fw_current_command;
+   itr->current_coords = stroke_fw_itr_coords;
+}
+
+static void stroke_backward_iterator(struct stroke_iterator *itr,
+                                     struct array *cmds,
+                                     struct array *coords)
+{
+   VGubyte cmd;
+   stroke_itr_common_init(itr, cmds, coords);
+   itr->position = itr->num_commands - 1;
+
+   cmd = stroke_bw_current_command(itr);
+   itr->coord_position = itr->num_coords -
+                         num_elements_for_segments(&cmd, 1);
+
+   itr->next = stroke_bw_next;
+   itr->has_next = stroke_bw_has_next;
+   itr->current_command = stroke_bw_current_command;
+   itr->current_coords = stroke_bw_itr_coords;
+}
+
+
+
+static void stroke_flat_next(struct stroke_iterator *itr)
+{
+   VGubyte cmd;
+
+   if (itr->curve_index >= 0) {
+      ++itr->curve_index;
+      if (itr->curve_index >= polygon_vertex_count(itr->curve_poly)) {
+         itr->curve_index = -1;
+         polygon_destroy(itr->curve_poly);
+         itr->curve_poly = 0;
+      } else
+         return;
+   }
+   debug_assert(stroke_fw_has_next(itr));
+
+   cmd = itr->cmds[itr->position];
+   itr->coord_position += num_elements_for_segments(&cmd, 1);
+   ++itr->position;
+
+   cmd = itr->cmds[itr->position];
+
+   if (cmd == VG_CUBIC_TO_ABS) {
+      struct bezier bezier;
+      VGfloat bez[8];
+
+      bez[0] = itr->coords[itr->coord_position - 2];
+      bez[1] = itr->coords[itr->coord_position - 1];
+      bez[2] = itr->coords[itr->coord_position];
+      bez[3] = itr->coords[itr->coord_position + 1];
+      bez[4] = itr->coords[itr->coord_position + 2];
+      bez[5] = itr->coords[itr->coord_position + 3];
+      bez[6] = itr->coords[itr->coord_position + 4];
+      bez[7] = itr->coords[itr->coord_position + 5];
+
+      bezier_init(&bezier,
+                  bez[0], bez[1],
+                  bez[2], bez[3],
+                  bez[4], bez[5],
+                  bez[6], bez[7]);
+      /* skip the first one, it's the same as the prev point */
+      itr->curve_index = 1;
+      if (itr->curve_poly) {
+         polygon_destroy(itr->curve_poly);
+         itr->curve_poly = 0;
+      }
+      itr->curve_poly = bezier_to_polygon(&bezier);
+   }
+}
+
+static VGboolean stroke_flat_has_next(struct stroke_iterator *itr)
+{
+   return  (itr->curve_index >= 0 &&
+            itr->curve_index < (polygon_vertex_count(itr->curve_poly)-1))
+            || itr->position < (itr->num_commands - 1);
+}
+
+static VGPathCommand stroke_flat_current_command(struct stroke_iterator *it)
+{
+   if (it->cmds[it->position] == VG_CUBIC_TO_ABS) {
+      return VG_LINE_TO_ABS;
+   }
+   return it->cmds[it->position];
+}
+
+static void stroke_flat_itr_coords(struct stroke_iterator *itr, VGfloat *coords)
+{
+   if (itr->curve_index <= -1 && itr->position >= itr->num_commands)
+      return;
+
+   if (itr->curve_index >= 0) {
+      polygon_vertex(itr->curve_poly, itr->curve_index,
+                     coords);
+      return;
+   }
+
+   switch (stroke_itr_command(itr)) {
+   case VG_MOVE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_LINE_TO_ABS:
+      coords[0] = itr->coords[itr->coord_position];
+      coords[1] = itr->coords[itr->coord_position + 1];
+      break;
+   case VG_CUBIC_TO_ABS:
+   default:
+      debug_assert(!"invalid command!\n");
+   }
+}
+
+static void stroke_flat_iterator(struct stroke_iterator *itr,
+                                 struct array *cmds,
+                                 struct array *coords)
+{
+   stroke_itr_common_init(itr, cmds, coords);
+   itr->position = 0;
+   itr->coord_position = 0;
+
+   itr->next = stroke_flat_next;
+   itr->has_next = stroke_flat_has_next;
+   itr->current_command = stroke_flat_current_command;
+   itr->current_coords = stroke_flat_itr_coords;
+   itr->curve_index = -1;
+   itr->curve_poly = 0;
+}
+
+
+static INLINE VGboolean finite_coords4(const VGfloat *c)
+{
+   return
+      isfinite(c[0]) && isfinite(c[1]) &&
+      isfinite(c[2]) && isfinite(c[3]);
+}
+
+/* from Graphics Gems II */
+#define SAME_SIGNS(a, b) ((a) * (b) >= 0)
+static VGboolean do_lines_intersect(VGfloat x1, VGfloat y1, VGfloat x2, VGfloat y2,
+                                    VGfloat x3, VGfloat y3, VGfloat x4, VGfloat y4)
+{
+   VGfloat a1, a2, b1, b2, c1, c2; /* Coefficients of line eqns */
+   VGfloat r1, r2, r3, r4;         /* 'sign' values */
+
+   a1 = y2 - y1;
+   b1 = x1 - x2;
+   c1 = x2 * y1 - x1 * y2;
+
+   r3 = a1 * x3 + b1 * y3 + c1;
+   r4 = a1 * x4 + b1 * y4 + c1;
+
+   if (r3 != 0 && r4 != 0 && SAME_SIGNS(r3, r4))
+      return VG_FALSE;
+
+   a2 = y4 - y3;
+   b2 = x3 - x4;
+   c2 = x4 * y3 - x3 * y4;
+
+   r1 = a2 * x1 + b2 * y1 + c2;
+   r2 = a2 * x2 + b2 * y2 + c2;
+
+   if (r1 != 0 && r2 != 0 && SAME_SIGNS(r1, r2))
+      return VG_FALSE;
+
+   return VG_TRUE;
+}
+
+static INLINE VGfloat line_dx(const VGfloat *l)
+{
+   return l[2] - l[0];
+}
+
+static INLINE VGfloat line_dy(const VGfloat *l)
+{
+   return l[3] - l[1];
+}
+
+static INLINE VGfloat line_angle(const VGfloat *l)
+{
+   const VGfloat dx = line_dx(l);
+   const VGfloat dy = line_dy(l);
+
+   const VGfloat theta = atan2(-dy, dx) * 360.0 / M_2PI;
+
+   const VGfloat theta_normalized = theta < 0 ? theta + 360 : theta;
+
+   if (floatsEqual(theta_normalized, 360.f))
+      return 0;
+   else
+      return theta_normalized;
+}
+
+static INLINE void line_set_length(VGfloat *l, VGfloat len)
+{
+   VGfloat uv[] = {l[0], l[1], l[2], l[3]};
+   if (null_line(l))
+      return;
+   line_normalize(uv);
+   l[2] = l[0] + line_dx(uv) * len;
+   l[3] = l[1] + line_dy(uv) * len;
+}
+
+static INLINE void line_translate(VGfloat *l, VGfloat x, VGfloat y)
+{
+   l[0] += x;
+   l[1] += y;
+   l[2] += x;
+   l[3] += y;
+}
+
+static INLINE VGfloat line_angle_to(const VGfloat *l1,
+                                    const VGfloat *l2)
+{
+   VGfloat a1, a2, delta, delta_normalized;
+   if (null_line(l1) || null_line(l1))
+      return 0;
+
+   a1 = line_angle(l1);
+   a2 = line_angle(l2);
+
+   delta = a2 - a1;
+   delta_normalized = delta < 0 ? delta + 360 : delta;
+
+   if (floatsEqual(delta, 360.f))
+      return 0;
+   else
+      return delta_normalized;
+}
+
+static INLINE VGfloat line_angles(const VGfloat *l1,
+                                  const VGfloat *l2)
+{
+   VGfloat cos_line, rad = 0;
+
+   if (null_line(l1) || null_line(l2))
+      return 0;
+
+   cos_line = (line_dx(l1)*line_dx(l2) + line_dy(l1)*line_dy(l2)) /
+              (line_lengthv(l1)*line_lengthv(l2));
+   rad = 0;
+
+   if (cos_line >= -1.0 && cos_line <= 1.0)
+      rad = acos(cos_line);
+   return rad * 360 / M_2PI;
+}
+
+
+static INLINE VGfloat adapted_angle_on_x(const VGfloat *line)
+{
+   const VGfloat identity[] = {0, 0, 1, 0};
+   VGfloat angle = line_angles(line, identity);
+   if (line_dy(line) > 0)
+      angle = 360 - angle;
+   return angle;
+}
+
+static enum intersection_type line_intersect(const VGfloat *l1,
+                                             const VGfloat *l2,
+                                             float *intersection_point)
+{
+   VGfloat isect[2];
+   enum intersection_type type;
+   VGboolean dx_zero, ldx_zero;
+
+   if (null_line(l1) || null_line(l2) ||
+       !finite_coords4(l1) || !finite_coords4(l2))
+      return NoIntersections;
+
+   type = do_lines_intersect(l1[0], l1[1], l1[2], l1[3], l2[0], l2[1], l2[2], l2[3])
+          ? BoundedIntersection : UnboundedIntersection;
+
+   dx_zero  = floatsEqual(line_dx(l1) + 1, 1);
+   ldx_zero = floatsEqual(line_dx(l2) + 1, 1);
+
+   /* one of the lines is vertical */
+   if (dx_zero && ldx_zero) {
+      type = NoIntersections;
+   } else if (dx_zero) {
+      VGfloat la = line_dy(l2) / line_dx(l2);
+      isect[0] = l1[0];
+      isect[1] = la * l1[0] + l2[1] - la * l2[0];
+   } else if (ldx_zero) {
+      VGfloat ta = line_dy(l1) / line_dx(l1);
+      isect[0] = l2[0];
+      isect[1] = ta * l2[0] + l1[1] - ta*l1[0];
+   } else {
+      VGfloat x;
+      VGfloat ta = line_dy(l1) / line_dx(l1);
+      VGfloat la = line_dy(l2) / line_dx(l2);
+      if (ta == la)
+         return NoIntersections;
+
+      x = ( - l2[1] + la * l2[0] + l1[1] - ta * l1[0] ) / (la - ta);
+      isect[0] = x;
+      isect[1] = ta*(x - l1[0]) + l1[1];
+   }
+   if (intersection_point) {
+      intersection_point[0] = isect[0];
+      intersection_point[1] = isect[1];
+   }
+   return type;
+}
+
+static INLINE enum line_join_mode stroker_join_mode(struct stroker *s)
+{
+   switch(s->join_style) {
+   case VG_JOIN_MITER:
+      return MiterJoin;
+   case VG_JOIN_ROUND:
+      return RoundJoin;
+   case VG_JOIN_BEVEL:
+      return FlatJoin;
+   default:
+      return FlatJoin;
+   }
+}
+
+static INLINE enum line_join_mode stroker_cap_mode(struct stroker *s)
+{
+   switch(s->cap_style) {
+   case VG_CAP_BUTT:
+      return FlatJoin;
+   case VG_CAP_ROUND:
+      return RoundCap;
+   case VG_CAP_SQUARE:
+      return SquareJoin;
+   default:
+      return FlatJoin;
+   }
+}
+
+void stroker_emit_move_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGubyte cmds = VG_MOVE_TO_ABS;
+   VGfloat coords[2] = {x, y};
+#if DEBUG_EMITS
+   debug_printf("emit move %f, %f\n", x, y);
+#endif
+   stroker->back2_x = stroker->back1_x;
+   stroker->back2_y = stroker->back1_y;
+   stroker->back1_x = x;
+   stroker->back1_y = y;
+
+   path_append_data(stroker->path,
+                    1,
+                    &cmds, &coords);
+}
+
+void stroker_emit_line_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGubyte cmds = VG_LINE_TO_ABS;
+   VGfloat coords[2] = {x, y};
+#if DEBUG_EMITS
+   debug_printf("emit line %f, %f\n", x, y);
+#endif
+   stroker->back2_x = stroker->back1_x;
+   stroker->back2_y = stroker->back1_y;
+   stroker->back1_x = x;
+   stroker->back1_y = y;
+   path_append_data(stroker->path,
+                    1,
+                    &cmds, &coords);
+}
+
+void stroker_emit_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                                  VGfloat px2, VGfloat py2,
+                                  VGfloat x, VGfloat y)
+{
+   VGubyte cmds = VG_CUBIC_TO_ABS;
+   VGfloat coords[6] = {px1, py1, px2, py2, x, y};
+#if DEBUG_EMITS
+   debug_printf("emit curve %f, %f, %f, %f, %f, %f\n", px1, py1,
+                px2, py2, x, y);
+#endif
+
+   if (px2 == x && py2 == y) {
+      if (px1 == x && py1 == y) {
+         stroker->back2_x = stroker->back1_x;
+         stroker->back2_y = stroker->back1_y;
+      } else {
+         stroker->back2_x = px1;
+         stroker->back2_y = py1;
+      }
+   } else {
+      stroker->back2_x = px2;
+      stroker->back2_y = py2;
+   }
+   stroker->back1_x = x;
+   stroker->back1_y = y;
+
+   path_append_data(stroker->path,
+                    1,
+                    &cmds, &coords);
+}
+
+static INLINE void create_round_join(struct stroker *stroker,
+                                     VGfloat x1, VGfloat y1,
+                                     VGfloat x2, VGfloat y2,
+                                     VGfloat width, VGfloat height)
+{
+   struct arc arc;
+   struct matrix matrix;
+
+   matrix_load_identity(&matrix);
+
+   /*stroker_emit_line_to(stroker, nx, ny);*/
+
+   arc_init(&arc, VG_SCCWARC_TO_ABS,
+            x1, y1, x2, y2, width/2, height/2, 0);
+   arc_stroker_emit(&arc, stroker, &matrix);
+}
+
+
+static void create_joins(struct stroker *stroker,
+                         VGfloat focal_x, VGfloat focal_y,
+                         const VGfloat *next_line, enum line_join_mode join)
+{
+#if DEBUG_EMITS
+   debug_printf("create_joins: focal=[%f, %f], next_line=[%f, %f,%f, %f]\n",
+                focal_x, focal_y,
+                next_line[0], next_line[1], next_line[2], next_line[3]);
+#endif
+   /* if we're alredy connected do nothing */
+   if (floatsEqual(stroker->back1_x, next_line[0]) &&
+       floatsEqual(stroker->back1_y, next_line[1]))
+      return;
+
+   if (join == FlatJoin) {
+      stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+   } else {
+      VGfloat prev_line[] = {stroker->back2_x, stroker->back2_y,
+                             stroker->back1_x, stroker->back1_y};
+
+      VGfloat isect[2];
+      enum intersection_type type = line_intersect(prev_line, next_line, isect);
+
+      if (join == SquareJoin) {
+         VGfloat offset = stroker->stroke_width / 2;
+         VGfloat l1[4] = {prev_line[0],
+                          prev_line[1],
+                          prev_line[2],
+                          prev_line[3]};
+         VGfloat l2[4] = {next_line[2],
+                          next_line[3],
+                          next_line[0],
+                          next_line[1]};
+
+         line_translate(l1, line_dx(l1), line_dy(l1));
+         line_set_length(l1, offset);
+
+         line_translate(l2, line_dx(l2), line_dy(l2));
+         line_set_length(l2, offset);
+
+         stroker_emit_line_to(stroker, l1[2], l1[3]);
+         stroker_emit_line_to(stroker, l2[2], l2[3]);
+         stroker_emit_line_to(stroker, l2[0], l2[1]);
+      } else if (join == RoundJoin) {
+         VGfloat offset = stroker->stroke_width / 2;
+         VGfloat short_cut[4] = {prev_line[2], prev_line[3],
+                                 next_line[0], next_line[1]};
+         VGfloat angle = line_angles(prev_line, short_cut);
+
+         if (type == BoundedIntersection ||
+             (angle > 90 && !floatsEqual(angle, 90.f))) {
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+            return;
+         }
+         create_round_join(stroker, prev_line[2], prev_line[3],
+                           next_line[0], next_line[1],
+                           offset * 2, offset * 2);
+
+         stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+      } else if (join == RoundCap) {
+         VGfloat offset = stroker->stroke_width / 2;
+         VGfloat l1[4] = { prev_line[0], prev_line[1],
+                           prev_line[2], prev_line[3] };
+         VGfloat l2[4] = {focal_x, focal_y,
+                          prev_line[2], prev_line[3]};
+
+         line_translate(l1, line_dx(l1), line_dy(l1));
+         line_set_length(l1, KAPPA * offset);
+
+         /* normal between prev_line and focal */
+         line_translate(l2, -line_dy(l2), line_dx(l2));
+         line_set_length(l2, KAPPA * offset);
+
+         stroker_emit_curve_to(stroker, l1[2], l1[3],
+                               l2[2], l2[3],
+                               l2[0], l2[1]);
+
+         l2[0] = l2[0];
+         l2[1] = l2[1];
+         l2[2] = l2[0] - line_dx(l2);
+         l2[3] = l2[1] - line_dy(l2);
+
+         line_translate(l1, next_line[0] - l1[0], next_line[1] - l1[1]);
+
+         stroker_emit_curve_to(stroker,
+                               l2[2], l2[3],
+                               l1[2], l1[3],
+                               l1[0], l1[1]);
+      } else if (join == MiterJoin) {
+         VGfloat miter_line[4] = {stroker->back1_x, stroker->back1_y,
+                                  isect[0], isect[1]};
+         VGfloat sl = (stroker->stroke_width * stroker->miter_limit);
+         VGfloat inside_line[4] = {prev_line[2], prev_line[3],
+                                   next_line[0], next_line[1]};
+         VGfloat angle = line_angle_to(inside_line, prev_line);
+
+         if (type == BoundedIntersection ||
+             (angle > 90 && !floatsEqual(angle, 90.f))) {
+            /*
+            debug_printf("f = %f, nl = %f, pl = %f, is = %f\n",
+                         focal_x, next_line[0],
+                         prev_line[2], isect[0]);*/
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+            return;
+         }
+
+         if (type == NoIntersections || line_lengthv(miter_line) > sl) {
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+         } else {
+            stroker_emit_line_to(stroker, isect[0], isect[1]);
+            stroker_emit_line_to(stroker, next_line[0], next_line[1]);
+         }
+      } else {
+         debug_assert(!"create_joins bad join style");
+      }
+   }
+}
+
+static void stroker_add_segment(struct stroker *stroker,
+                                VGPathCommand cmd,
+                                const VGfloat *coords,
+                                VGint num_coords)
+{
+   /* skip duplicated points */
+   if (stroker->segments->num_elements &&
+       stroker->last_cmd == cmd) {
+      VGfloat *data = stroker->control_points->data;
+      data += stroker->control_points->num_elements;
+      data -= num_coords;
+      switch (cmd) {
+      case VG_MOVE_TO_ABS:
+         if (floatsEqual(coords[0], data[0]) &&
+             floatsEqual(coords[1], data[1]))
+            return;
+         break;
+      case VG_LINE_TO_ABS:
+         if (floatsEqual(coords[0], data[0]) &&
+             floatsEqual(coords[1], data[1]))
+            return;
+         break;
+      case VG_CUBIC_TO_ABS:
+         if (floatsEqual(coords[0], data[0]) &&
+             floatsEqual(coords[1], data[1]) &&
+             floatsEqual(coords[2], data[2]) &&
+             floatsEqual(coords[3], data[3]) &&
+             floatsEqual(coords[4], data[4]) &&
+             floatsEqual(coords[5], data[5]))
+            return;
+         break;
+      default:
+         debug_assert(!"Invalid stroke segment");
+      }
+   } else if (stroker->last_cmd == VG_CUBIC_TO_ABS &&
+              cmd == VG_LINE_TO_ABS) {
+      VGfloat *data = stroker->control_points->data;
+      data += stroker->control_points->num_elements;
+      data -= 2;
+      if (floatsEqual(coords[0], data[0]) &&
+          floatsEqual(coords[1], data[1]))
+         return;
+   }
+   stroker->last_cmd = cmd;
+   array_append_data(stroker->segments, &cmd, 1);
+   array_append_data(stroker->control_points, coords, num_coords);
+}
+
+void stroker_move_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGfloat coords[2] = {x, y};
+#if STROKE_SEGMENTS
+   debug_printf("stroker_move_to(%f, %f)\n", x, y);
+#endif
+
+   if (stroker->segments->num_elements > 0)
+      stroker->process_subpath(stroker);
+
+   array_reset(stroker->segments);
+   array_reset(stroker->control_points);
+
+   stroker_add_segment(stroker, VG_MOVE_TO_ABS, coords, 2);
+}
+
+void stroker_line_to(struct stroker *stroker, VGfloat x, VGfloat y)
+{
+   VGfloat coords[] = {x, y};
+
+#if STROKE_SEGMENTS
+   debug_printf("stroker_line_to(%f, %f)\n", x, y);
+#endif
+   if (!stroker->segments->num_elements)
+      stroker_add_segment(stroker, VG_MOVE_TO_ABS, zero_coords, 2);
+
+   stroker_add_segment(stroker, VG_LINE_TO_ABS, coords, 2);
+}
+
+void stroker_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                      VGfloat px2, VGfloat py2,
+                      VGfloat x, VGfloat y)
+{
+   VGfloat coords[] = {px1, py1,
+                       px2, py2,
+                       x, y};
+#if STROKE_SEGMENTS
+   debug_printf("stroker_curve_to(%f, %f, %f, %f, %f, %f)\n",
+                px1, py1, px2, py2, x, y);
+#endif
+   if (!stroker->segments->num_elements)
+      stroker_add_segment(stroker, VG_MOVE_TO_ABS, zero_coords, 2);
+
+   stroker_add_segment(stroker, VG_CUBIC_TO_ABS, coords, 6);
+}
+
+static INLINE VGboolean is_segment_null(VGPathCommand cmd,
+                                        VGfloat *coords,
+                                        VGfloat *res)
+{
+   switch(cmd) {
+   case VG_MOVE_TO_ABS:
+   case VG_LINE_TO_ABS:
+      return floatsEqual(coords[0], res[0]) &&
+         floatsEqual(coords[1], res[1]);
+      break;
+   case VG_CUBIC_TO_ABS:
+      return floatsEqual(coords[0], res[0]) &&
+         floatsEqual(coords[1], res[1]) &&
+         floatsEqual(coords[2], res[0]) &&
+         floatsEqual(coords[3], res[1]) &&
+         floatsEqual(coords[4], res[0]) &&
+         floatsEqual(coords[5], res[1]);
+      break;
+   default:
+      assert(0);
+   }
+   return VG_FALSE;
+}
+
+static VGboolean vg_stroke_outline(struct stroke_iterator *it,
+                                struct stroker *stroker,
+                                VGboolean cap_first,
+                                VGfloat *start_tangent)
+{
+   const int MAX_OFFSET = 16;
+   struct bezier offset_curves[MAX_OFFSET];
+   VGPathCommand first_element;
+   VGfloat start[2], prev[2];
+   VGboolean first = VG_TRUE;
+   VGfloat offset;
+
+   first_element = stroke_itr_command(it);
+   if (first_element != VG_MOVE_TO_ABS) {
+      stroker_emit_move_to(stroker, 0.f, 0.f);
+      prev[0] = 0.f;
+      prev[1] = 0.f;
+   }
+   stroke_itr_coords(it, start);
+#if STROKE_DEBUG
+   debug_printf(" -> (side) [%.2f, %.2f]\n",
+                start[0],
+                start[1]);
+#endif
+
+   prev[0] = start[0];
+   prev[1] = start[1];
+
+   offset = stroker->stroke_width / 2;
+
+   if (!it->has_next(it)) {
+      /* single point */
+
+      return VG_TRUE;
+   }
+
+   while (it->has_next(it)) {
+      VGPathCommand cmd;
+      VGfloat coords[8];
+
+      it->next(it);
+      cmd = stroke_itr_command(it);
+      stroke_itr_coords(it, coords);
+
+      if (cmd == VG_LINE_TO_ABS) {
+         VGfloat line[4] = {prev[0], prev[1], coords[0], coords[1]};
+         VGfloat normal[4];
+         line_normal(line, normal);
+
+#if STROKE_DEBUG
+         debug_printf("\n ---> (side) lineto [%.2f, %.2f]\n", coords[0], coords[1]);
+#endif
+         line_set_length(normal, offset);
+         line_translate(line, line_dx(normal), line_dy(normal));
+
+         /* if we are starting a new subpath, move to correct starting point */
+         if (first) {
+            if (cap_first)
+               create_joins(stroker, prev[0], prev[1], line,
+                            stroker_cap_mode(stroker));
+            else
+               stroker_emit_move_to(stroker, line[0], line[1]);
+            memcpy(start_tangent, line,
+                   sizeof(VGfloat) * 4);
+            first = VG_FALSE;
+         } else {
+            create_joins(stroker, prev[0], prev[1], line,
+                         stroker_join_mode(stroker));
+         }
+
+         /* add the stroke for this line */
+         stroker_emit_line_to(stroker, line[2], line[3]);
+         prev[0] = coords[0];
+         prev[1] = coords[1];
+      } else if (cmd == VG_CUBIC_TO_ABS) {
+#if STROKE_DEBUG
+         debug_printf("\n ---> (side) cubicTo [%.2f, %.2f]\n",
+                coords[4],
+                coords[5]);
+#endif
+         struct bezier bezier;
+         int count;
+
+         bezier_init(&bezier,
+                     prev[0], prev[1], coords[0], coords[1],
+                     coords[2], coords[3], coords[4], coords[5]);
+
+         count = bezier_translate_by_normal(&bezier,
+                                            offset_curves,
+                                            MAX_OFFSET,
+                                            offset,
+                                            curve_threshold);
+
+         if (count) {
+            /* if we are starting a new subpath, move to correct starting point */
+            VGfloat tangent[4];
+            VGint i;
+
+            bezier_start_tangent(&bezier, tangent);
+            line_translate(tangent,
+                           offset_curves[0].x1 - bezier.x1,
+                           offset_curves[0].y1 - bezier.y1);
+            if (first) {
+               VGfloat pt[2] = {offset_curves[0].x1,
+                                offset_curves[0].y1};
+
+               if (cap_first) {
+                  create_joins(stroker, prev[0], prev[1], tangent,
+                               stroker_cap_mode(stroker));
+               } else {
+                  stroker_emit_move_to(stroker, pt[0], pt[1]);
+               }
+               start_tangent[0] = tangent[0];
+               start_tangent[1] = tangent[1];
+               start_tangent[2] = tangent[2];
+               start_tangent[3] = tangent[3];
+               first = VG_FALSE;
+            } else {
+               create_joins(stroker, prev[0], prev[1], tangent,
+                            stroker_join_mode(stroker));
+            }
+
+            /* add these beziers */
+            for (i = 0; i < count; ++i) {
+               struct bezier *bez = &offset_curves[i];
+               stroker_emit_curve_to(stroker,
+                                     bez->x2, bez->y2,
+                                     bez->x3, bez->y3,
+                                     bez->x4, bez->y4);
+            }
+         }
+
+         prev[0] = coords[4];
+         prev[1] = coords[5];
+      }
+   }
+
+   if (floatsEqual(start[0], prev[0]) &&
+       floatsEqual(start[1], prev[1])) {
+      /* closed subpath, join first and last point */
+#if STROKE_DEBUG
+      debug_printf("\n stroker: closed subpath\n");
+#endif
+      create_joins(stroker, prev[0], prev[1], start_tangent,
+                   stroker_join_mode(stroker));
+      return VG_TRUE;
+   } else {
+#if STROKE_DEBUG
+      debug_printf("\n stroker: open subpath\n");
+#endif
+      return VG_FALSE;
+   }
+}
+
+static void stroker_process_subpath(struct stroker *stroker)
+{
+   VGboolean fwclosed, bwclosed;
+   VGfloat fw_start_tangent[4], bw_start_tangent[4];
+   struct stroke_iterator fwit;
+   struct stroke_iterator bwit;
+   debug_assert(stroker->segments->num_elements > 0);
+
+   memset(fw_start_tangent, 0,
+          sizeof(VGfloat)*4);
+   memset(bw_start_tangent, 0,
+          sizeof(VGfloat)*4);
+
+   stroke_forward_iterator(&fwit, stroker->segments,
+                           stroker->control_points);
+   stroke_backward_iterator(&bwit, stroker->segments,
+                            stroker->control_points);
+
+   debug_assert(fwit.cmds[0] == VG_MOVE_TO_ABS);
+
+   fwclosed = vg_stroke_outline(&fwit, stroker, VG_FALSE, fw_start_tangent);
+   bwclosed = vg_stroke_outline(&bwit, stroker, !fwclosed, bw_start_tangent);
+
+   if (!bwclosed)
+      create_joins(stroker,
+                   fwit.coords[0], fwit.coords[1], fw_start_tangent,
+                   stroker_cap_mode(stroker));
+   else {
+      /* hack to handle the requirement of the VG spec that says that strokes
+       * of len==0 that have butt cap or round cap still need
+       * to be rendered. (8.7.4 Stroke Generation) */
+      if (stroker->segments->num_elements <= 3) {
+         VGPathCommand cmd;
+         VGfloat data[8], coords[8];
+         struct stroke_iterator *it = &fwit;
+
+         stroke_forward_iterator(it, stroker->segments,
+                                 stroker->control_points);
+         cmd = stroke_itr_command(it);
+         stroke_itr_coords(it, coords);
+         if (cmd != VG_MOVE_TO_ABS) {
+            memset(data, 0, sizeof(VGfloat) * 8);
+            if (!is_segment_null(cmd, coords, data))
+               return;
+         } else {
+            data[0] = coords[0];
+            data[1] = coords[1];
+         }
+         while (it->has_next(it)) {
+            it->next(it);
+            cmd = stroke_itr_command(it);
+            stroke_itr_coords(it, coords);
+            if (!is_segment_null(cmd, coords, data))
+               return;
+         }
+         /* generate the square/round cap */
+         if (stroker->cap_style == VG_CAP_SQUARE) {
+            VGfloat offset = stroker->stroke_width / 2;
+            stroker_emit_move_to(stroker, data[0] - offset,
+                                 data[1] - offset);
+            stroker_emit_line_to(stroker, data[0] + offset,
+                                 data[1] - offset);
+            stroker_emit_line_to(stroker, data[0] + offset,
+                                 data[1] + offset);
+            stroker_emit_line_to(stroker, data[0] - offset,
+                                 data[1] + offset);
+            stroker_emit_line_to(stroker, data[0] - offset,
+                                 data[1] - offset);
+         } else if (stroker->cap_style == VG_CAP_ROUND) {
+            VGfloat offset = stroker->stroke_width / 2;
+            VGfloat cx = data[0], cy = data[1];
+            { /*circle */
+               struct arc arc;
+               struct matrix matrix;
+               matrix_load_identity(&matrix);
+
+               stroker_emit_move_to(stroker, cx + offset, cy);
+               arc_init(&arc, VG_SCCWARC_TO_ABS,
+                        cx + offset, cy,
+                        cx - offset, cy,
+                        offset, offset, 0);
+               arc_stroker_emit(&arc, stroker, &matrix);
+               arc_init(&arc, VG_SCCWARC_TO_ABS,
+                         cx - offset, cy,
+                         cx + offset, cy,
+                         offset, offset, 0);
+               arc_stroker_emit(&arc, stroker, &matrix);
+            }
+         }
+      }
+   }
+}
+
+static INLINE VGfloat dash_pattern(struct dash_stroker *stroker,
+                                   VGint idx)
+{
+   if (stroker->dash_pattern[idx] < 0)
+      return 0.f;
+   return stroker->dash_pattern[idx];
+}
+
+static void dash_stroker_process_subpath(struct stroker *str)
+{
+   struct dash_stroker *stroker = (struct dash_stroker *)str;
+   VGfloat sum_length = 0;
+   VGint i;
+   VGint idash = 0;
+   VGfloat pos = 0;
+   VGfloat elen = 0;
+   VGfloat doffset = stroker->dash_phase;
+   VGfloat estart = 0;
+   VGfloat estop = 0;
+   VGfloat cline[4];
+   struct stroke_iterator it;
+   VGfloat prev[2];
+   VGfloat move_to_pos[2];
+   VGfloat line_to_pos[2];
+
+   VGboolean has_move_to = VG_FALSE;
+
+   stroke_flat_iterator(&it, stroker->base.segments,
+                        stroker->base.control_points);
+
+   stroke_itr_coords(&it, prev);
+   move_to_pos[0] = prev[0];
+   move_to_pos[1] = prev[1];
+
+   debug_assert(stroker->dash_pattern_num > 0);
+
+   for (i = 0; i < stroker->dash_pattern_num; ++i) {
+      sum_length += dash_pattern(stroker, i);
+   }
+
+   if (floatIsZero(sum_length)) {
+      return;
+   }
+
+   doffset -= floorf(doffset / sum_length) * sum_length;
+
+   while (!floatIsZero(doffset) && doffset >= dash_pattern(stroker, idash)) {
+      doffset -= dash_pattern(stroker, idash);
+      idash = (idash + 1) % stroker->dash_pattern_num;
+   }
+
+   while (it.has_next(&it)) {
+      VGPathCommand cmd;
+      VGfloat coords[8];
+      VGboolean done;
+
+      it.next(&it);
+      cmd = stroke_itr_command(&it);
+      stroke_itr_coords(&it, coords);
+
+      debug_assert(cmd == VG_LINE_TO_ABS);
+      cline[0] = prev[0];
+      cline[1] = prev[1];
+      cline[2] = coords[0];
+      cline[3] = coords[1];
+
+      elen = line_lengthv(cline);
+
+      estop = estart + elen;
+
+      done = pos >= estop;
+      while (!done) {
+         VGfloat p2[2];
+
+         VGint idash_incr = 0;
+         VGboolean has_offset = doffset > 0;
+         VGfloat dpos = pos + dash_pattern(stroker, idash) - doffset - estart;
+
+         debug_assert(dpos >= 0);
+
+         if (dpos > elen) { /* dash extends this line */
+            doffset = dash_pattern(stroker, idash) - (dpos - elen);
+            pos = estop;
+            done = VG_TRUE;
+            p2[0] = cline[2];
+            p2[1] = cline[3];
+         } else { /* Dash is on this line */
+            line_point_at(cline, dpos/elen, p2);
+            pos = dpos + estart;
+            done = pos >= estop;
+            idash_incr = 1;
+            doffset = 0;
+         }
+
+         if (idash % 2 == 0) {
+            line_to_pos[0] = p2[0];
+            line_to_pos[1] = p2[1];
+
+            if (!has_offset || !has_move_to) {
+               stroker_move_to(&stroker->stroker, move_to_pos[0], move_to_pos[1]);
+               has_move_to = VG_TRUE;
+            }
+            stroker_line_to(&stroker->stroker, line_to_pos[0], line_to_pos[1]);
+         } else {
+            move_to_pos[0] = p2[0];
+            move_to_pos[1] = p2[1];
+         }
+
+         idash = (idash + idash_incr) % stroker->dash_pattern_num;
+      }
+
+      estart = estop;
+      prev[0] = coords[0];
+      prev[1] = coords[1];
+   }
+
+   if (it.curve_poly) {
+      polygon_destroy(it.curve_poly);
+      it.curve_poly = 0;
+   }
+
+   stroker->base.path = stroker->stroker.path;
+}
+
+static void default_begin(struct stroker *stroker)
+{
+   array_reset(stroker->segments);
+   array_reset(stroker->control_points);
+}
+
+static void default_end(struct stroker *stroker)
+{
+   if (stroker->segments->num_elements > 0)
+      stroker->process_subpath(stroker);
+}
+
+
+static void dash_stroker_begin(struct stroker *stroker)
+{
+   struct dash_stroker *dasher =
+      (struct dash_stroker *)stroker;
+
+   default_begin(&dasher->stroker);
+   default_begin(stroker);
+}
+
+static void dash_stroker_end(struct stroker *stroker)
+{
+   struct dash_stroker *dasher =
+      (struct dash_stroker *)stroker;
+
+   default_end(stroker);
+   default_end(&dasher->stroker);
+}
+
+void stroker_init(struct stroker *stroker,
+                  struct vg_state *state)
+{
+   stroker->stroke_width = state->stroke.line_width.f;
+   stroker->miter_limit = state->stroke.miter_limit.f;
+   stroker->cap_style = state->stroke.cap_style;
+   stroker->join_style = state->stroke.join_style;
+
+   stroker->begin = default_begin;
+   stroker->process_subpath = stroker_process_subpath;
+   stroker->end = default_end;
+
+   stroker->segments = array_create(sizeof(VGubyte));
+   stroker->control_points = array_create(sizeof(VGfloat));
+
+   stroker->back1_x = 0;
+   stroker->back1_y = 0;
+   stroker->back2_x = 0;
+   stroker->back2_y = 0;
+
+   stroker->path = path_create(VG_PATH_DATATYPE_F, 1.0f, 0.0f,
+                               0, 0, VG_PATH_CAPABILITY_ALL);
+
+   stroker->last_cmd = VG_CLOSE_PATH;
+}
+
+void dash_stroker_init(struct stroker *str,
+                       struct vg_state *state)
+{
+   struct dash_stroker *stroker = (struct dash_stroker *)str;
+   int i;
+
+   stroker_init(str, state);
+   stroker_init(&stroker->stroker, state);
+
+   {
+      int real_num = state->stroke.dash_pattern_num;
+      if (real_num % 2)/* if odd, ignore the last one */
+         --real_num;
+      for (i = 0; i < real_num; ++i)
+         stroker->dash_pattern[i] = state->stroke.dash_pattern[i].f;
+      stroker->dash_pattern_num = real_num;
+   }
+
+   stroker->dash_phase = state->stroke.dash_phase.f;
+   stroker->dash_phase_reset = state->stroke.dash_phase_reset;
+
+   stroker->base.begin = dash_stroker_begin;
+   stroker->base.process_subpath = dash_stroker_process_subpath;
+   stroker->base.end = dash_stroker_end;
+   path_destroy(stroker->base.path);
+   stroker->base.path = NULL;
+}
+
+void stroker_begin(struct stroker *stroker)
+{
+   stroker->begin(stroker);
+}
+
+void stroker_end(struct stroker *stroker)
+{
+   stroker->end(stroker);
+}
+
+void stroker_cleanup(struct stroker *stroker)
+{
+   array_destroy(stroker->segments);
+   array_destroy(stroker->control_points);
+}
+
+void dash_stroker_cleanup(struct dash_stroker *stroker)
+{
+   /* if stroker->base.path is null means we never
+    * processed a valid path so delete the temp one
+    * we already created */
+   if (!stroker->base.path)
+      path_destroy(stroker->stroker.path);
+   stroker_cleanup(&stroker->stroker);
+   stroker_cleanup((struct stroker*)stroker);
+}
diff --git a/src/gallium/state_trackers/vega/stroker.h b/src/gallium/state_trackers/vega/stroker.h
new file mode 100644
index 0000000000..36543cd923
--- /dev/null
+++ b/src/gallium/state_trackers/vega/stroker.h
@@ -0,0 +1,89 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STROKER_H
+#define STROKER_H
+
+#include "VG/openvg.h"
+#include "api_consts.h"
+
+struct path;
+struct vg_state;
+struct array;
+
+struct stroker {
+   void (*begin)(struct stroker *stroker);
+   void (*process_subpath)(struct stroker *stroker);
+   void (*end)(struct stroker *stroker);
+
+   struct array *segments;
+   struct array *control_points;
+   struct path *path;
+
+   VGfloat back1_x, back1_y;
+   VGfloat back2_x, back2_y;
+
+   VGfloat     stroke_width;
+   VGfloat     miter_limit;
+   VGCapStyle  cap_style;
+   VGJoinStyle join_style;
+
+   VGPathCommand last_cmd;
+};
+
+struct dash_stroker {
+   struct stroker base;
+
+   struct stroker stroker;
+
+   VGfloat dash_pattern[VEGA_MAX_DASH_COUNT];
+   VGint dash_pattern_num;
+   VGfloat dash_phase;
+   VGboolean dash_phase_reset;
+};
+
+void stroker_init(struct stroker *stroker,
+                  struct vg_state *state);
+void dash_stroker_init(struct stroker *stroker,
+                       struct vg_state *state);
+void dash_stroker_cleanup(struct dash_stroker *stroker);
+void stroker_cleanup(struct stroker *stroker);
+
+void stroker_begin(struct stroker *stroker);
+void stroker_move_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_line_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                      VGfloat px2, VGfloat py2,
+                      VGfloat x, VGfloat y);
+void stroker_end(struct stroker *stroker);
+
+void stroker_emit_move_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_emit_line_to(struct stroker *stroker, VGfloat x, VGfloat y);
+void stroker_emit_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1,
+                           VGfloat px2, VGfloat py2,
+                           VGfloat x, VGfloat y);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/util_array.h b/src/gallium/state_trackers/vega/util_array.h
new file mode 100644
index 0000000000..4343dfe36e
--- /dev/null
+++ b/src/gallium/state_trackers/vega/util_array.h
@@ -0,0 +1,122 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef UTIL_ARRAY_H
+#define UTIL_ARRAY_H
+
+#include "util/u_memory.h"
+
+#define DEFAULT_ARRAY_SIZE 256
+
+struct array {
+   VGint          datatype_size;
+   void          *data;
+   VGint          size;
+   VGint          num_elements;
+};
+
+static INLINE struct array *array_create(VGint datatype_size)
+{
+   struct array *array = CALLOC_STRUCT(array);
+   array->datatype_size = datatype_size;
+
+   array->size = DEFAULT_ARRAY_SIZE;
+   array->data = malloc(array->size * array->datatype_size);
+
+   return array;
+}
+
+
+static INLINE struct array *array_create_size(VGint datatype_size, VGint size)
+{
+   struct array *array = CALLOC_STRUCT(array);
+   array->datatype_size = datatype_size;
+
+   array->size = size;
+   array->data = malloc(array->size * array->datatype_size);
+
+   return array;
+}
+
+static INLINE void array_destroy(struct array *array)
+{
+   if (array)
+      free(array->data);
+   free(array);
+}
+
+static INLINE void array_resize(struct array *array, int num)
+{
+   VGint size = array->datatype_size * num;
+   void *data = malloc(size);
+   memcpy(data, array->data, array->size * array->datatype_size);
+   free(array->data);
+   array->data = data;
+   array->size = num;
+   array->num_elements = (array->num_elements > num) ? num :
+                         array->num_elements;
+}
+
+static INLINE void array_append_data(struct array *array,
+                              const void *data, int num_elements)
+{
+   VGbyte *adata;
+
+   while (array->num_elements + num_elements > array->size) {
+      array_resize(array, (array->num_elements + num_elements) * 1.5);
+   }
+   adata = (VGbyte *)array->data;
+   memcpy(adata + (array->num_elements * array->datatype_size), data,
+          num_elements * array->datatype_size);
+   array->num_elements += num_elements;
+}
+
+static INLINE void array_change_data(struct array *array,
+                              const void *data,
+                              int start_idx,
+                              int num_elements)
+{
+   VGbyte *adata = (VGbyte *)array->data;
+   memcpy(adata + (start_idx * array->datatype_size), data,
+          num_elements * array->datatype_size);
+}
+
+static INLINE void array_remove_element(struct array *array,
+                                        int idx)
+{
+   VGbyte *adata = (VGbyte *)array->data;
+   memmove(adata + (idx * array->datatype_size),
+           adata + ((idx + 1) * array->datatype_size),
+           (array->num_elements - idx - 1) * array->datatype_size);
+   --array->num_elements;
+}
+
+static INLINE void array_reset(struct array *array)
+{
+   array->num_elements = 0;
+}
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_context.c b/src/gallium/state_trackers/vega/vg_context.c
new file mode 100644
index 0000000000..e0ff02f3a9
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_context.c
@@ -0,0 +1,543 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_context.h"
+
+#include "paint.h"
+#include "renderer.h"
+#include "shaders_cache.h"
+#include "shader.h"
+#include "asm_util.h"
+#include "st_inlines.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "cso_cache/cso_context.h"
+
+#include "util/u_simple_shaders.h"
+#include "util/u_memory.h"
+#include "util/u_blit.h"
+
+struct vg_context *_vg_context = 0;
+
+struct vg_context * vg_current_context(void)
+{
+   return _vg_context;
+}
+
+static void init_clear(struct vg_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   /* rasterizer state: bypass clipping */
+   memset(&st->clear.raster, 0, sizeof(st->clear.raster));
+   st->clear.raster.gl_rasterization_rules = 1;
+
+   /* fragment shader state: color pass-through program */
+   st->clear.fs =
+      util_make_fragment_passthrough_shader(pipe);
+}
+void vg_set_current_context(struct vg_context *ctx)
+{
+   _vg_context = ctx;
+}
+
+struct vg_context * vg_create_context(struct pipe_context *pipe,
+                                      const void *visual,
+                                      struct vg_context *share)
+{
+   struct vg_context *ctx;
+
+   ctx = CALLOC_STRUCT(vg_context);
+
+   ctx->pipe = pipe;
+
+   vg_init_state(&ctx->state.vg);
+   ctx->state.dirty = ALL_DIRTY;
+
+   ctx->cso_context = cso_create_context(pipe);
+
+   init_clear(ctx);
+
+   ctx->default_paint = paint_create(ctx);
+   ctx->state.vg.stroke_paint = ctx->default_paint;
+   ctx->state.vg.fill_paint = ctx->default_paint;
+
+
+   ctx->mask.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->mask.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->mask.sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   ctx->mask.sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->mask.sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->mask.sampler.normalized_coords = 0;
+
+   ctx->blend_sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->blend_sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->blend_sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   ctx->blend_sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->blend_sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   ctx->blend_sampler.normalized_coords = 0;
+
+   vg_set_error(ctx, VG_NO_ERROR);
+
+   ctx->owned_objects[VG_OBJECT_PAINT] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_IMAGE] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_MASK] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_FONT] = cso_hash_create();
+   ctx->owned_objects[VG_OBJECT_PATH] = cso_hash_create();
+
+   ctx->renderer = renderer_create(ctx);
+   ctx->sc = shaders_cache_create(ctx);
+   ctx->shader = shader_create(ctx);
+
+   ctx->blit = util_create_blit(ctx->pipe, ctx->cso_context);
+
+   return ctx;
+}
+
+void vg_destroy_context(struct vg_context *ctx)
+{
+   struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf;
+   struct pipe_constant_buffer *vsbuf = &ctx->vs_const_buffer;
+
+   util_destroy_blit(ctx->blit);
+   renderer_destroy(ctx->renderer);
+   shaders_cache_destroy(ctx->sc);
+   shader_destroy(ctx->shader);
+   paint_destroy(ctx->default_paint);
+
+   if (cbuf && cbuf->buffer)
+      pipe_buffer_reference(&cbuf->buffer, NULL);
+
+   if (vsbuf && vsbuf->buffer)
+      pipe_buffer_reference(&vsbuf->buffer, NULL);
+
+   if (ctx->clear.fs) {
+      cso_delete_fragment_shader(ctx->cso_context, ctx->clear.fs);
+      ctx->clear.fs = NULL;
+   }
+
+   if (ctx->plain_vs) {
+      vg_shader_destroy(ctx, ctx->plain_vs);
+      ctx->plain_vs = NULL;
+   }
+   if (ctx->clear_vs) {
+      vg_shader_destroy(ctx, ctx->clear_vs);
+      ctx->clear_vs = NULL;
+   }
+   if (ctx->texture_vs) {
+      vg_shader_destroy(ctx, ctx->texture_vs);
+      ctx->texture_vs = NULL;
+   }
+
+   if (ctx->pass_through_depth_fs)
+      vg_shader_destroy(ctx, ctx->pass_through_depth_fs);
+   if (ctx->mask.union_fs)
+      vg_shader_destroy(ctx, ctx->mask.union_fs);
+   if (ctx->mask.intersect_fs)
+      vg_shader_destroy(ctx, ctx->mask.intersect_fs);
+   if (ctx->mask.subtract_fs)
+      vg_shader_destroy(ctx, ctx->mask.subtract_fs);
+   if (ctx->mask.set_fs)
+      vg_shader_destroy(ctx, ctx->mask.set_fs);
+
+   cso_release_all(ctx->cso_context);
+   cso_destroy_context(ctx->cso_context);
+
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_PAINT]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_IMAGE]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_MASK]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_FONT]);
+   cso_hash_delete(ctx->owned_objects[VG_OBJECT_PATH]);
+
+   free(ctx);
+}
+
+void vg_init_object(struct vg_object *obj, struct vg_context *ctx, enum vg_object_type type)
+{
+   obj->type = type;
+   obj->ctx = ctx;
+}
+
+VGboolean vg_context_is_object_valid(struct vg_context *ctx,
+                                enum vg_object_type type,
+                                void *ptr)
+{
+    if (ctx) {
+       struct cso_hash *hash = ctx->owned_objects[type];
+       if (!hash)
+          return VG_FALSE;
+       return cso_hash_contains(hash, (unsigned)(long)ptr);
+    }
+    return VG_FALSE;
+}
+
+void vg_context_add_object(struct vg_context *ctx,
+                           enum vg_object_type type,
+                           void *ptr)
+{
+    if (ctx) {
+       struct cso_hash *hash = ctx->owned_objects[type];
+       if (!hash)
+          return;
+       cso_hash_insert(hash, (unsigned)(long)ptr, ptr);
+    }
+}
+
+void vg_context_remove_object(struct vg_context *ctx,
+                              enum vg_object_type type,
+                              void *ptr)
+{
+   if (ctx) {
+      struct cso_hash *hash = ctx->owned_objects[type];
+      if (!hash)
+         return;
+      cso_hash_take(hash, (unsigned)(long)ptr);
+   }
+}
+
+static void update_clip_state(struct vg_context *ctx)
+{
+   struct pipe_depth_stencil_alpha_state *dsa = &ctx->state.g3d.dsa;
+   struct vg_state *state =  &ctx->state.vg;
+
+   memset(dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+
+   if (state->scissoring) {
+      struct pipe_blend_state *blend = &ctx->state.g3d.blend;
+      struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+      dsa->depth.writemask = 1;/*glDepthMask(TRUE);*/
+      dsa->depth.func = PIPE_FUNC_ALWAYS;
+      dsa->depth.enabled = 1;
+
+      cso_save_blend(ctx->cso_context);
+      cso_save_fragment_shader(ctx->cso_context);
+      /* set a passthrough shader */
+      if (!ctx->pass_through_depth_fs)
+         ctx->pass_through_depth_fs = shader_create_from_text(ctx->pipe,
+                                                              pass_through_depth_asm,
+                                                              40,
+                                                              PIPE_SHADER_FRAGMENT);
+      cso_set_fragment_shader_handle(ctx->cso_context,
+                                     ctx->pass_through_depth_fs->driver);
+      cso_set_depth_stencil_alpha(ctx->cso_context, dsa);
+
+      ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_DEPTHSTENCIL, NULL, 1.0, 0);
+
+      /* disable color writes */
+      blend->colormask = 0; /*disable colorwrites*/
+      cso_set_blend(ctx->cso_context, blend);
+
+      /* enable scissoring */
+      for (int i = 0; i < state->scissor_rects_num; ++i) {
+         const float x      = state->scissor_rects[i * 4 + 0].f;
+         const float y      = state->scissor_rects[i * 4 + 1].f;
+         const float width  = state->scissor_rects[i * 4 + 2].f;
+         const float height = state->scissor_rects[i * 4 + 3].f;
+         VGfloat minx, miny, maxx, maxy;
+
+         minx = 0;
+         miny = 0;
+         maxx = fb->width;
+         maxy = fb->height;
+
+         if (x > minx)
+            minx = x;
+         if (y > miny)
+            miny = y;
+
+         if (x + width < maxx)
+            maxx = x + width;
+         if (y + height < maxy)
+            maxy = y + height;
+
+         /* check for null space */
+         if (minx >= maxx || miny >= maxy)
+            minx = miny = maxx = maxy = 0;
+
+         /*glClear(GL_DEPTH_BUFFER_BIT);*/
+         renderer_draw_quad(ctx->renderer, minx, miny, maxx, maxy, 0.0f);
+      }
+
+      blend->colormask = 1; /*enable colorwrites*/
+      cso_restore_blend(ctx->cso_context);
+      cso_restore_fragment_shader(ctx->cso_context);
+
+      dsa->depth.enabled = 1; /* glEnable(GL_DEPTH_TEST); */
+      dsa->depth.writemask = 0;/*glDepthMask(FALSE);*/
+      dsa->depth.func = PIPE_FUNC_GEQUAL;
+   }
+}
+
+void vg_validate_state(struct vg_context *ctx)
+{
+   if ((ctx->state.dirty & BLEND_DIRTY)) {
+      struct pipe_blend_state *blend = &ctx->state.g3d.blend;
+      memset(blend, 0, sizeof(struct pipe_blend_state));
+      blend->blend_enable = 1;
+      blend->colormask |= PIPE_MASK_R;
+      blend->colormask |= PIPE_MASK_G;
+      blend->colormask |= PIPE_MASK_B;
+      blend->colormask |= PIPE_MASK_A;
+
+      switch (ctx->state.vg.blend_mode) {
+      case VG_BLEND_SRC:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+         break;
+      case VG_BLEND_SRC_OVER:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_SRC_ALPHA;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+         break;
+      case VG_BLEND_DST_OVER:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_INV_DST_ALPHA;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_INV_DST_ALPHA;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_DST_ALPHA;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_DST_ALPHA;
+         break;
+      case VG_BLEND_SRC_IN:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_DST_ALPHA;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_DST_ALPHA;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+         break;
+      case VG_BLEND_DST_IN:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ZERO;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_SRC_ALPHA;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_SRC_ALPHA;
+         break;
+      case VG_BLEND_MULTIPLY:
+      case VG_BLEND_SCREEN:
+      case VG_BLEND_DARKEN:
+      case VG_BLEND_LIGHTEN:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+         break;
+      case VG_BLEND_ADDITIVE:
+         blend->rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor   = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ONE;
+         break;
+      default:
+         assert(!"not implemented blend mode");
+      }
+      cso_set_blend(ctx->cso_context, &ctx->state.g3d.blend);
+   }
+   if ((ctx->state.dirty & RASTERIZER_DIRTY)) {
+      struct pipe_rasterizer_state *raster = &ctx->state.g3d.rasterizer;
+      memset(raster, 0, sizeof(struct pipe_rasterizer_state));
+      raster->gl_rasterization_rules = 1;
+      cso_set_rasterizer(ctx->cso_context, &ctx->state.g3d.rasterizer);
+   }
+   if ((ctx->state.dirty & VIEWPORT_DIRTY)) {
+      struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+      const VGint param_bytes = 8 * sizeof(VGfloat);
+      VGfloat vs_consts[8] = {
+         2.f/fb->width, 2.f/fb->height, 1, 1,
+         -1, -1, 0, 0
+      };
+      struct pipe_constant_buffer *cbuf = &ctx->vs_const_buffer;
+
+      vg_set_viewport(ctx, VEGA_Y0_BOTTOM);
+
+      pipe_buffer_reference(&cbuf->buffer, NULL);
+      cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 16,
+                                        PIPE_BUFFER_USAGE_CONSTANT,
+                                        param_bytes);
+
+      if (cbuf->buffer) {
+         st_no_flush_pipe_buffer_write(ctx, cbuf->buffer,
+                                       0, param_bytes, vs_consts);
+      }
+      ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_VERTEX, 0, cbuf);
+   }
+   if ((ctx->state.dirty & VS_DIRTY)) {
+      cso_set_vertex_shader_handle(ctx->cso_context,
+                                   vg_plain_vs(ctx));
+   }
+
+   /* must be last because it renders to the depth buffer*/
+   if ((ctx->state.dirty & DEPTH_STENCIL_DIRTY)) {
+      update_clip_state(ctx);
+      cso_set_depth_stencil_alpha(ctx->cso_context, &ctx->state.g3d.dsa);
+   }
+
+   shader_set_masking(ctx->shader, ctx->state.vg.masking);
+   shader_set_image_mode(ctx->shader, ctx->state.vg.image_mode);
+
+   ctx->state.dirty = NONE_DIRTY;
+}
+
+VGboolean vg_object_is_valid(void *ptr, enum vg_object_type type)
+{
+   struct vg_object *obj = ptr;
+   if (ptr && is_aligned(obj) && obj->type == type)
+      return VG_TRUE;
+   else
+      return VG_FALSE;
+}
+
+void vg_set_error(struct vg_context *ctx,
+                  VGErrorCode code)
+{
+   /*vgGetError returns the oldest error code provided by
+    * an API call on the current context since the previous
+    * call to vgGetError on that context (or since the creation
+    of the context).*/
+   if (ctx->_error == VG_NO_ERROR)
+      ctx->_error = code;
+}
+
+void vg_prepare_blend_surface(struct vg_context *ctx)
+{
+   struct pipe_surface *dest_surface = NULL;
+   struct pipe_context *pipe = ctx->pipe;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->strb;
+
+   /* first finish all pending rendering */
+   vgFinish();
+
+   dest_surface = pipe->screen->get_tex_surface(pipe->screen,
+                                                stfb->blend_texture,
+                                                0, 0, 0,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+   /* flip it, because we want to use it as a sampler */
+   util_blit_pixels_tex(ctx->blit,
+                        strb->texture,
+                        0, strb->height,
+                        strb->width, 0,
+                        dest_surface,
+                        0, 0,
+                        strb->width, strb->height,
+                        0.0, PIPE_TEX_MIPFILTER_NEAREST);
+
+   if (dest_surface)
+      pipe_surface_reference(&dest_surface, NULL);
+
+   /* make sure it's complete */
+   vgFinish();
+}
+
+
+void vg_prepare_blend_surface_from_mask(struct vg_context *ctx)
+{
+   struct pipe_surface *dest_surface = NULL;
+   struct pipe_context *pipe = ctx->pipe;
+   struct st_framebuffer *stfb = ctx->draw_buffer;
+   struct st_renderbuffer *strb = stfb->strb;
+
+   vg_validate_state(ctx);
+
+   /* first finish all pending rendering */
+   vgFinish();
+
+   dest_surface = pipe->screen->get_tex_surface(pipe->screen,
+                                                stfb->blend_texture,
+                                                0, 0, 0,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* flip it, because we want to use it as a sampler */
+   util_blit_pixels_tex(ctx->blit,
+                        stfb->alpha_mask,
+                        0, strb->height,
+                        strb->width, 0,
+                        dest_surface,
+                        0, 0,
+                        strb->width, strb->height,
+                        0.0, PIPE_TEX_MIPFILTER_NEAREST);
+
+   /* make sure it's complete */
+   vgFinish();
+
+   if (dest_surface)
+      pipe_surface_reference(&dest_surface, NULL);
+}
+
+void * vg_plain_vs(struct vg_context *ctx)
+{
+   if (!ctx->plain_vs) {
+      ctx->plain_vs = shader_create_from_text(ctx->pipe,
+                                              vs_plain_asm,
+                                              200,
+                                              PIPE_SHADER_VERTEX);
+   }
+
+   return ctx->plain_vs->driver;
+}
+
+
+void * vg_clear_vs(struct vg_context *ctx)
+{
+   if (!ctx->clear_vs) {
+      ctx->clear_vs = shader_create_from_text(ctx->pipe,
+                                              vs_clear_asm,
+                                              200,
+                                              PIPE_SHADER_VERTEX);
+   }
+
+   return ctx->clear_vs->driver;
+}
+
+void * vg_texture_vs(struct vg_context *ctx)
+{
+   if (!ctx->texture_vs) {
+      ctx->texture_vs = shader_create_from_text(ctx->pipe,
+                                                vs_texture_asm,
+                                                200,
+                                                PIPE_SHADER_VERTEX);
+   }
+
+   return ctx->texture_vs->driver;
+}
+
+void vg_set_viewport(struct vg_context *ctx, VegaOrientation orientation)
+{
+   struct pipe_viewport_state viewport;
+   struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb;
+   VGfloat y_scale = (orientation == VEGA_Y0_BOTTOM) ? -2.f : 2.f;
+
+   viewport.scale[0] =  fb->width / 2.f;
+   viewport.scale[1] =  fb->height / y_scale;
+   viewport.scale[2] =  1.0;
+   viewport.scale[3] =  1.0;
+   viewport.translate[0] = fb->width / 2.f;
+   viewport.translate[1] = fb->height / 2.f;
+   viewport.translate[2] = 0.0;
+   viewport.translate[3] = 0.0;
+
+   cso_set_viewport(ctx->cso_context, &viewport);
+}
diff --git a/src/gallium/state_trackers/vega/vg_context.h b/src/gallium/state_trackers/vega/vg_context.h
new file mode 100644
index 0000000000..ccc8889c8c
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_context.h
@@ -0,0 +1,292 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_CONTEXT_H
+#define VG_CONTEXT_H
+
+#include "vg_state.h"
+
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "util/u_pointer.h"
+#include "util/u_math.h"
+
+#include "cso_cache/cso_hash.h"
+#include "cso_cache/cso_context.h"
+
+struct renderer;
+struct shaders_cache;
+struct shader;
+struct vg_shader;
+
+struct st_renderbuffer {
+   enum pipe_format   format;
+   struct pipe_surface *surface;
+   struct pipe_texture *texture;
+   VGint width, height;
+};
+
+struct st_framebuffer {
+   VGint init_width, init_height;
+   struct st_renderbuffer *strb;
+   struct st_renderbuffer *dsrb;
+
+   struct pipe_texture *alpha_mask;
+
+   struct pipe_texture *blend_texture;
+
+   void *privateData;
+};
+
+enum vg_object_type {
+   VG_OBJECT_UNKNOWN = 0,
+   VG_OBJECT_PAINT,
+   VG_OBJECT_IMAGE,
+   VG_OBJECT_MASK,
+   VG_OBJECT_FONT,
+   VG_OBJECT_PATH,
+
+   VG_OBJECT_LAST
+};
+enum dirty_state {
+   NONE_DIRTY          = 0<<0,
+   BLEND_DIRTY         = 1<<1,
+   RASTERIZER_DIRTY    = 1<<2,
+   VIEWPORT_DIRTY      = 1<<3,
+   VS_DIRTY            = 1<<4,
+   DEPTH_STENCIL_DIRTY = 1<<5,
+   ALL_DIRTY           = BLEND_DIRTY | RASTERIZER_DIRTY |
+   VIEWPORT_DIRTY | VS_DIRTY | DEPTH_STENCIL_DIRTY
+};
+
+struct vg_context
+{
+   struct pipe_context *pipe;
+
+   struct {
+      struct vg_state vg;
+      struct {
+         struct pipe_blend_state blend;
+         struct pipe_rasterizer_state rasterizer;
+         struct pipe_shader_state vs_state;
+         struct pipe_depth_stencil_alpha_state dsa;
+         struct pipe_framebuffer_state fb;
+      } g3d;
+      VGbitfield dirty;
+   } state;
+
+   VGErrorCode _error;
+
+   struct st_framebuffer *draw_buffer;
+
+   struct cso_hash *owned_objects[VG_OBJECT_LAST];
+
+   struct {
+      struct pipe_shader_state vert_shader;
+      struct pipe_shader_state frag_shader;
+      struct pipe_rasterizer_state raster;
+      void *fs;
+      float vertices[4][2][4];  /**< vertex pos + color */
+   } clear;
+
+   struct {
+      struct pipe_constant_buffer cbuf;
+      struct pipe_sampler_state sampler;
+
+      struct vg_shader *union_fs;
+      struct vg_shader *intersect_fs;
+      struct vg_shader *subtract_fs;
+      struct vg_shader *set_fs;
+   } mask;
+
+   struct vg_shader *pass_through_depth_fs;
+
+   struct cso_context *cso_context;
+
+   struct pipe_buffer *stencil_quad;
+   VGfloat stencil_vertices[4][2][4];
+
+   struct renderer *renderer;
+   struct shaders_cache *sc;
+   struct shader *shader;
+
+   struct pipe_sampler_state blend_sampler;
+   struct {
+      struct pipe_constant_buffer buffer;
+      void *color_matrix_fs;
+   } filter;
+   struct vg_paint *default_paint;
+
+   struct blit_state *blit;
+
+   struct vg_shader *plain_vs;
+   struct vg_shader *clear_vs;
+   struct vg_shader *texture_vs;
+   struct pipe_constant_buffer vs_const_buffer;
+};
+
+struct vg_object {
+   enum vg_object_type type;
+   struct vg_context *ctx;
+};
+void vg_init_object(struct vg_object *obj, struct vg_context *ctx, enum vg_object_type type);
+VGboolean vg_object_is_valid(void *ptr, enum vg_object_type type);
+
+struct vg_context *vg_create_context(struct pipe_context *pipe,
+                                     const void *visual,
+                                     struct vg_context *share);
+void vg_destroy_context(struct vg_context *ctx);
+struct vg_context *vg_current_context(void);
+void vg_set_current_context(struct vg_context *ctx);
+
+VGboolean vg_context_is_object_valid(struct vg_context *ctx,
+                                     enum vg_object_type type,
+                                     void *ptr);
+void vg_context_add_object(struct vg_context *ctx,
+                           enum vg_object_type type,
+                           void *ptr);
+void vg_context_remove_object(struct vg_context *ctx,
+                              enum vg_object_type type,
+                              void *ptr);
+
+void vg_validate_state(struct vg_context *ctx);
+
+void vg_set_error(struct vg_context *ctx,
+                  VGErrorCode code);
+
+void vg_prepare_blend_surface(struct vg_context *ctx);
+void vg_prepare_blend_surface_from_mask(struct vg_context *ctx);
+
+
+static INLINE VGboolean is_aligned_to(const void *ptr, VGbyte alignment)
+{
+   void *aligned = align_pointer(ptr, alignment);
+   return (ptr == aligned) ? VG_TRUE : VG_FALSE;
+}
+
+static INLINE VGboolean is_aligned(const void *ptr)
+{
+   return is_aligned_to(ptr, 4);
+}
+
+static INLINE void vg_shift_rectx(VGfloat coords[4],
+                                 const VGfloat *bounds,
+                                 const VGfloat shift)
+{
+   coords[0] += shift;
+   coords[2] -= shift;
+   if (bounds) {
+      coords[2] = MIN2(coords[2], bounds[2]);
+      /* bound x/y + width/height */
+      if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
+         coords[2] = (bounds[0] + bounds[2]) - coords[0];
+      }
+   }
+}
+
+static INLINE void vg_shift_recty(VGfloat coords[4],
+                                 const VGfloat *bounds,
+                                 const VGfloat shift)
+{
+   coords[1] += shift;
+   coords[3] -= shift;
+   if (bounds) {
+      coords[3] = MIN2(coords[3], bounds[3]);
+      if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
+         coords[3] = (bounds[1] + bounds[3]) - coords[1];
+      }
+   }
+}
+
+static INLINE void vg_bound_rect(VGfloat coords[4],
+                                 const VGfloat bounds[4],
+                                 VGfloat shift[4])
+{
+   /* if outside the bounds */
+   if (coords[0] > (bounds[0] + bounds[2]) ||
+       coords[1] > (bounds[1] + bounds[3]) ||
+       (coords[0] + coords[2]) < bounds[0] ||
+       (coords[1] + coords[3]) < bounds[1]) {
+      coords[0] = 0.f;
+      coords[1] = 0.f;
+      coords[2] = 0.f;
+      coords[3] = 0.f;
+      shift[0] = 0.f;
+      shift[1] = 0.f;
+      return;
+   }
+
+   /* bound x */
+   if (coords[0] < bounds[0]) {
+      shift[0] = bounds[0] - coords[0];
+      coords[2] -= shift[0];
+      coords[0] = bounds[0];
+   } else
+      shift[0] = 0.f;
+
+   /* bound y */
+   if (coords[1] < bounds[1]) {
+      shift[1] = bounds[1] - coords[1];
+      coords[3] -= shift[1];
+      coords[1] = bounds[1];
+   } else
+      shift[1] = 0.f;
+
+   shift[2] = bounds[2] - coords[2];
+   shift[3] = bounds[3] - coords[3];
+   /* bound width/height */
+   coords[2] = MIN2(coords[2], bounds[2]);
+   coords[3] = MIN2(coords[3], bounds[3]);
+
+   /* bound x/y + width/height */
+   if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
+      coords[2] = (bounds[0] + bounds[2]) - coords[0];
+   }
+   if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
+      coords[3] = (bounds[1] + bounds[3]) - coords[1];
+   }
+
+   /* if outside the bounds */
+   if ((coords[0] + coords[2]) < bounds[0] ||
+       (coords[1] + coords[3]) < bounds[1]) {
+      coords[0] = 0.f;
+      coords[1] = 0.f;
+      coords[2] = 0.f;
+      coords[3] = 0.f;
+      return;
+   }
+}
+
+void *vg_plain_vs(struct vg_context *ctx);
+void *vg_clear_vs(struct vg_context *ctx);
+void *vg_texture_vs(struct vg_context *ctx);
+typedef enum {
+   VEGA_Y0_TOP,
+   VEGA_Y0_BOTTOM
+} VegaOrientation;
+void vg_set_viewport(struct vg_context *ctx, VegaOrientation orientation);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_state.c b/src/gallium/state_trackers/vega/vg_state.c
new file mode 100644
index 0000000000..6f6bfdaf7a
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_state.c
@@ -0,0 +1,124 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_state.h"
+
+#include <string.h>
+
+void vg_init_state(struct vg_state *state)
+{
+   state->matrix_mode = VG_MATRIX_PATH_USER_TO_SURFACE;
+   state->fill_rule = VG_EVEN_ODD;
+   state->image_quality = VG_IMAGE_QUALITY_FASTER;
+   state->rendering_quality = VG_RENDERING_QUALITY_BETTER;
+   state->blend_mode = VG_BLEND_SRC_OVER;
+   state->image_mode = VG_DRAW_IMAGE_NORMAL;
+
+   memset(state->scissor_rects, 0, sizeof(state->scissor_rects));
+   state->scissor_rects_num = 0;
+
+   state->color_transform = VG_FALSE;
+   state->color_transform_values[0] = 1.0f;
+   state->color_transform_values[1] = 1.0f;
+   state->color_transform_values[2] = 1.0f;
+   state->color_transform_values[3] = 1.0f;
+   state->color_transform_values[4] = 0.0f;
+   state->color_transform_values[5] = 0.0f;
+   state->color_transform_values[6] = 0.0f;
+   state->color_transform_values[7] = 0.0f;
+
+   /* Stroke parameters */
+   state->stroke.line_width.f       = 1.0f;
+   state->stroke.line_width.i       = 1;
+   state->stroke.cap_style        = VG_CAP_BUTT;
+   state->stroke.join_style       = VG_JOIN_MITER;
+   state->stroke.miter_limit.f      = 4.0f;
+   state->stroke.miter_limit.i      = 4;
+   state->stroke.dash_pattern_num = 0;
+   state->stroke.dash_phase.f       = 0.0f;
+   state->stroke.dash_phase.i       = 0;
+   state->stroke.dash_phase_reset = VG_FALSE;
+
+   /* Edge fill color for VG_TILE_FILL tiling mode */
+   state->tile_fill_color[0] = 0.0f;
+   state->tile_fill_color[1] = 0.0f;
+   state->tile_fill_color[2] = 0.0f;
+   state->tile_fill_color[3] = 0.0f;
+
+   /* Color for vgClear */
+   state->clear_color[0] = 0.0f;
+   state->clear_color[1] = 0.0f;
+   state->clear_color[2] = 0.0f;
+   state->clear_color[3] = 0.0f;
+
+   /* Glyph origin */
+   state->glyph_origin[0].f = 0.0f;
+   state->glyph_origin[1].f = 0.0f;
+   state->glyph_origin[0].i = 0;
+   state->glyph_origin[1].i = 0;
+
+   /* Enable/disable alpha masking and scissoring */
+   state->masking = VG_FALSE;
+   state->scissoring = VG_FALSE;
+
+   /* Pixel layout information */
+   state->pixel_layout = VG_PIXEL_LAYOUT_UNKNOWN;
+   state->screen_layout = VG_PIXEL_LAYOUT_UNKNOWN;
+
+   /* Source format selection for image filters */
+   state->filter_format_linear = VG_FALSE;
+   state->filter_format_premultiplied = VG_FALSE;
+
+   /* Destination write enable mask for image filters */
+   state->filter_channel_mask = (VG_RED | VG_GREEN | VG_BLUE | VG_ALPHA);
+
+   matrix_load_identity(&state->path_user_to_surface_matrix);
+   matrix_load_identity(&state->image_user_to_surface_matrix);
+   matrix_load_identity(&state->fill_paint_to_user_matrix);
+   matrix_load_identity(&state->stroke_paint_to_user_matrix);
+   matrix_load_identity(&state->glyph_user_to_surface_matrix);
+}
+
+struct matrix *vg_state_matrix(struct vg_state *state)
+{
+    switch(state->matrix_mode) {
+    case VG_MATRIX_PATH_USER_TO_SURFACE:
+       return &state->path_user_to_surface_matrix;
+    case VG_MATRIX_IMAGE_USER_TO_SURFACE:
+       return &state->image_user_to_surface_matrix;
+    case VG_MATRIX_FILL_PAINT_TO_USER:
+       return &state->fill_paint_to_user_matrix;
+    case VG_MATRIX_STROKE_PAINT_TO_USER:
+       return &state->stroke_paint_to_user_matrix;
+#ifdef OPENVG_VERSION_1_1
+    case VG_MATRIX_GLYPH_USER_TO_SURFACE:
+       return &state->glyph_user_to_surface_matrix;
+#endif
+    default:
+       break;
+    }
+    return NULL;
+}
diff --git a/src/gallium/state_trackers/vega/vg_state.h b/src/gallium/state_trackers/vega/vg_state.h
new file mode 100644
index 0000000000..ed90689f91
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_state.h
@@ -0,0 +1,109 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_STATE_H
+#define VG_STATE_H
+
+#include "VG/openvg.h"
+
+#include "api_consts.h"
+#include "matrix.h"
+
+struct vg_value
+{
+   VGfloat f;
+   VGint   i;
+};
+
+struct vg_state {
+   /* Mode settings */
+   VGMatrixMode matrix_mode;
+   VGFillRule fill_rule;
+   VGImageQuality image_quality;
+   VGRenderingQuality rendering_quality;
+   VGBlendMode blend_mode;
+   VGImageMode image_mode;
+
+   /* Scissoring rectangles */
+   struct vg_value  scissor_rects[32*4];
+   VGint  scissor_rects_num;
+
+   /* Color Transformation */
+   VGboolean color_transform;
+   VGfloat color_transform_values[8];
+
+   /* Stroke parameters */
+   struct {
+      struct vg_value line_width;
+      VGCapStyle cap_style;
+      VGJoinStyle join_style;
+      struct vg_value miter_limit;
+      struct vg_value dash_pattern[VEGA_MAX_DASH_COUNT];
+      VGint   dash_pattern_num;
+      struct vg_value dash_phase;
+      VGboolean dash_phase_reset;
+   } stroke;
+
+   /* Edge fill color for VG_TILE_FILL tiling mode */
+   VGfloat tile_fill_color[4];
+   VGint tile_fill_colori[4];
+
+   /* Color for vgClear */
+   VGfloat clear_color[4];
+   VGint clear_colori[4];
+
+   /* Glyph origin */
+   struct vg_value glyph_origin[2];
+
+   /* Enable/disable alpha masking and scissoring */
+   VGboolean masking;
+   VGboolean scissoring;
+
+   /* Pixel layout information */
+   VGPixelLayout pixel_layout;
+   VGPixelLayout screen_layout;
+
+   /* Source format selection for image filters */
+   VGboolean filter_format_linear;
+   VGboolean filter_format_premultiplied;
+
+   /* Destination write enable mask for image filters */
+   VGbitfield filter_channel_mask;
+
+   struct matrix path_user_to_surface_matrix;
+   struct matrix image_user_to_surface_matrix;
+   struct matrix fill_paint_to_user_matrix;
+   struct matrix stroke_paint_to_user_matrix;
+   struct matrix glyph_user_to_surface_matrix;
+
+   struct vg_paint *stroke_paint;
+   struct vg_paint *fill_paint;
+};
+
+void vg_init_state(struct vg_state *state);
+struct matrix * vg_state_matrix(struct vg_state *state);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_tracker.c b/src/gallium/state_trackers/vega/vg_tracker.c
new file mode 100644
index 0000000000..c262ce08fa
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_tracker.c
@@ -0,0 +1,406 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_context.h"
+#include "vg_tracker.h"
+#include "mask.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+static struct pipe_texture *
+create_texture(struct pipe_context *pipe, enum pipe_format format,
+               VGint width, VGint height)
+{
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+
+   if (format != PIPE_FORMAT_NONE) {
+      templ.format = format;
+   }
+   else {
+      templ.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   }
+
+   templ.target = PIPE_TEXTURE_2D;
+   pf_get_block(templ.format, &templ.block);
+   templ.width[0] = width;
+   templ.height[0] = height;
+   templ.depth[0] = 1;
+   templ.last_level = 0;
+
+   if (pf_get_component_bits(format, PIPE_FORMAT_COMP_S)) {
+      templ.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   } else {
+      templ.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                         PIPE_TEXTURE_USAGE_RENDER_TARGET |
+                         PIPE_TEXTURE_USAGE_SAMPLER);
+   }
+
+   return pipe->screen->texture_create(pipe->screen, &templ);
+}
+
+/**
+ * Allocate a renderbuffer for a an on-screen window (not a user-created
+ * renderbuffer).  The window system code determines the format.
+ */
+static struct st_renderbuffer *
+st_new_renderbuffer_fb(enum pipe_format format)
+{
+   struct st_renderbuffer *strb;
+
+   strb = CALLOC_STRUCT(st_renderbuffer);
+   if (!strb) {
+      /*_vega_error(NULL, VG_OUT_OF_MEMORY, "creating renderbuffer");*/
+      return NULL;
+   }
+
+   strb->format = format;
+
+   return strb;
+}
+
+
+/**
+ * This is called to allocate the original drawing surface, and
+ * during window resize.
+ */
+static VGboolean
+st_renderbuffer_alloc_storage(struct vg_context * ctx,
+                              struct st_renderbuffer *strb,
+                              VGuint width, VGuint height)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   unsigned surface_usage;
+
+   /* Free the old surface and texture
+    */
+   pipe_surface_reference(&strb->surface, NULL);
+   pipe_texture_reference(&strb->texture, NULL);
+
+
+   /* Probably need dedicated flags for surface usage too:
+    */
+   surface_usage = (PIPE_BUFFER_USAGE_GPU_READ  |
+                    PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   strb->texture = create_texture(pipe, strb->format,
+                                  width, height);
+
+   if (!strb->texture)
+      return FALSE;
+
+   strb->surface = pipe->screen->get_tex_surface(pipe->screen,
+                                                 strb->texture,
+                                                 0, 0, 0,
+                                                 surface_usage);
+   strb->width = width;
+   strb->height = height;
+
+   assert(strb->surface->width == width);
+   assert(strb->surface->height == height);
+
+   return strb->surface != NULL;
+}
+
+struct vg_context * st_create_context(struct pipe_context *pipe,
+                                      const void *visual,
+                                      struct vg_context *share)
+{
+   struct vg_context *ctx = vg_create_context(pipe, visual, share);
+   /*debug_printf("--------- CREATE CONTEXT %p\n", ctx);*/
+   return ctx;
+}
+
+void st_destroy_context(struct vg_context *st)
+{
+   /*debug_printf("--------- DESTROY CONTEXT %p\n", st);*/
+   vg_destroy_context(st);
+}
+
+void st_copy_context_state(struct vg_context *dst, struct vg_context *src,
+                           uint mask)
+{
+   fprintf(stderr, "FIXME: %s\n", __FUNCTION__);
+}
+
+void st_get_framebuffer_dimensions(struct st_framebuffer *stfb,
+				   uint *width,
+				   uint *height)
+{
+   *width = stfb->strb->width;
+   *height = stfb->strb->height;
+}
+
+struct st_framebuffer * st_create_framebuffer(const void *visual,
+                                              enum pipe_format colorFormat,
+                                              enum pipe_format depthFormat,
+                                              enum pipe_format stencilFormat,
+                                              uint width, uint height,
+                                              void *privateData)
+{
+   struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer);
+   if (stfb) {
+      struct st_renderbuffer *rb =
+         st_new_renderbuffer_fb(colorFormat);
+      stfb->strb = rb;
+#if 0
+      if (doubleBuffer) {
+         struct st_renderbuffer *rb =
+            st_new_renderbuffer_fb(colorFormat);
+      }
+#endif
+
+      /* we want to combine the depth/stencil */
+      if (stencilFormat == depthFormat)
+         stfb->dsrb = st_new_renderbuffer_fb(stencilFormat);
+      else
+         stfb->dsrb = st_new_renderbuffer_fb(PIPE_FORMAT_S8Z24_UNORM);
+
+      /*### currently we always allocate it but it's possible it's
+        not necessary if EGL_ALPHA_MASK_SIZE was 0
+      */
+      stfb->alpha_mask = 0;
+
+      stfb->init_width = width;
+      stfb->init_height = height;
+      stfb->privateData = privateData;
+   }
+
+   return stfb;
+}
+
+static void setup_new_alpha_mask(struct vg_context *ctx,
+                                 struct st_framebuffer *stfb,
+                                 uint width, uint height)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_texture *old_texture = stfb->alpha_mask;
+
+   /*
+     we use PIPE_FORMAT_A8R8G8B8_UNORM because we want to render to
+     this texture and use it as a sampler, so while this wastes some
+     space it makes both of those a lot simpler
+   */
+   stfb->alpha_mask =
+      create_texture(pipe, PIPE_FORMAT_A8R8G8B8_UNORM, width, height);
+
+   if (!stfb->alpha_mask) {
+      if (old_texture)
+         pipe_texture_reference(&old_texture, NULL);
+      return;
+   }
+
+   vg_validate_state(ctx);
+
+   /* alpha mask starts with 1.f alpha */
+   mask_fill(0, 0, width, height, 1.f);
+
+   /* if we had an old surface copy it over */
+   if (old_texture) {
+      struct pipe_surface *surface = pipe->screen->get_tex_surface(
+         pipe->screen,
+         stfb->alpha_mask,
+         0, 0, 0,
+         PIPE_BUFFER_USAGE_GPU_WRITE);
+      struct pipe_surface *old_surface = pipe->screen->get_tex_surface(
+         pipe->screen,
+         old_texture,
+         0, 0, 0,
+         PIPE_BUFFER_USAGE_GPU_READ);
+      pipe->surface_copy(pipe,
+                         surface,
+                         0, 0,
+                         old_surface,
+                         0, 0,
+                         MIN2(old_surface->width, width),
+                         MIN2(old_surface->height, height));
+      if (surface)
+         pipe_surface_reference(&surface, NULL);
+      if (old_surface)
+         pipe_surface_reference(&old_surface, NULL);
+   }
+
+   /* Free the old texture
+    */
+   if (old_texture)
+      pipe_texture_reference(&old_texture, NULL);
+}
+
+void st_resize_framebuffer(struct st_framebuffer *stfb,
+                           uint width, uint height)
+{
+   struct vg_context *ctx = vg_current_context();
+   struct st_renderbuffer *strb = stfb->strb;
+   struct pipe_framebuffer_state *state;
+
+   if (!ctx)
+      return;
+
+   state = &ctx->state.g3d.fb;
+
+   /* If this is a noop, exit early and don't do the clear, etc below.
+    */
+   if (strb->width == width &&
+       strb->height == height &&
+       state->zsbuf)
+      return;
+
+   if (strb->width != width || strb->height != height)
+      st_renderbuffer_alloc_storage(ctx, strb,
+                                 width, height);
+
+   if (stfb->dsrb->width != width || stfb->dsrb->height != height)
+      st_renderbuffer_alloc_storage(ctx, stfb->dsrb,
+                                 width, height);
+
+   {
+      VGuint i;
+
+      memset(state, 0, sizeof(struct pipe_framebuffer_state));
+
+      state->width  = width;
+      state->height = height;
+
+      state->nr_cbufs = 1;
+      state->cbufs[0] = strb->surface;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         state->cbufs[i] = 0;
+
+      state->zsbuf = stfb->dsrb->surface;
+
+      cso_set_framebuffer(ctx->cso_context, state);
+   }
+
+   ctx->state.dirty |= VIEWPORT_DIRTY;
+   ctx->state.dirty |= DEPTH_STENCIL_DIRTY;/*to reset the scissors*/
+
+   ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_DEPTHSTENCIL,
+                    NULL, 0.0, 0);
+
+   /* we need all the other state already set */
+
+   setup_new_alpha_mask(ctx, stfb, width, height);
+
+   pipe_texture_reference( &stfb->blend_texture, NULL );
+   stfb->blend_texture = create_texture(ctx->pipe, PIPE_FORMAT_A8R8G8B8_UNORM,
+                                        width, height);
+}
+
+void st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                                uint surfIndex, struct pipe_surface *surf)
+{
+   struct st_renderbuffer *rb = stfb->strb;
+
+   /* unreference existing surfaces */
+   pipe_surface_reference( &rb->surface, NULL );
+   pipe_texture_reference( &rb->texture, NULL );
+
+   /* reference new ones */
+   pipe_surface_reference( &rb->surface, surf );
+   pipe_texture_reference( &rb->texture, surf->texture );
+
+   rb->width  = surf->width;
+   rb->height = surf->height;
+}
+
+int st_get_framebuffer_surface(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_surface **surf)
+{
+   struct st_renderbuffer *rb = stfb->strb;
+   *surf = rb->surface;
+   return VG_TRUE;
+}
+
+int st_get_framebuffer_texture(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_texture **tex)
+{
+   struct st_renderbuffer *rb = stfb->strb;
+   *tex = rb->texture;
+   return VG_TRUE;
+}
+
+void * st_framebuffer_private(struct st_framebuffer *stfb)
+{
+   return stfb->privateData;
+}
+
+void st_unreference_framebuffer(struct st_framebuffer *stfb)
+{
+   /* FIXME */
+}
+
+void st_make_current(struct vg_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read)
+{
+   vg_set_current_context(st);
+   if (st) {
+      st->draw_buffer = draw;
+   }
+}
+
+void st_flush(struct vg_context *st, uint pipeFlushFlags,
+              struct pipe_fence_handle **fence)
+{
+   st->pipe->flush(st->pipe, pipeFlushFlags, fence);
+}
+
+void st_finish(struct vg_context *st)
+{
+   struct pipe_fence_handle *fence = NULL;
+
+   st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence);
+
+   st->pipe->screen->fence_finish(st->pipe->screen, fence, 0);
+   st->pipe->screen->fence_reference(st->pipe->screen, &fence, NULL);
+}
+
+void st_notify_swapbuffers(struct st_framebuffer *stfb)
+{
+   struct vg_context *ctx = vg_current_context();
+   if (ctx && ctx->draw_buffer == stfb) {
+      st_flush(ctx,
+	       PIPE_FLUSH_RENDER_CACHE | 
+	       PIPE_FLUSH_SWAPBUFFERS |
+	       PIPE_FLUSH_FRAME,
+               NULL);
+   }
+}
+
+void st_notify_swapbuffers_complete(struct st_framebuffer *stfb)
+{
+}
+
+int
+st_set_teximage(struct pipe_texture *pt, int target)
+{
+   return 0;
+}
diff --git a/src/gallium/state_trackers/vega/vg_tracker.h b/src/gallium/state_trackers/vega/vg_tracker.h
new file mode 100644
index 0000000000..805c58ccc7
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_tracker.h
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_TRACKER_H
+#define VG_TRACKER_H
+
+#include "VG/openvg.h"
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+#define ST_SURFACE_FRONT_LEFT   0
+#define ST_SURFACE_BACK_LEFT    1
+#define ST_SURFACE_FRONT_RIGHT  2
+#define ST_SURFACE_BACK_RIGHT   3
+#define ST_SURFACE_DEPTH        8
+
+struct vg_context;
+struct st_framebuffer;
+struct pipe_context;
+struct pipe_fence_handle;
+struct pipe_surface;
+
+
+struct vg_context *st_create_context(struct pipe_context *pipe,
+                                     const void *visual,
+                                     struct vg_context *share);
+
+void st_destroy_context( struct vg_context *st );
+
+void st_copy_context_state(struct vg_context *dst, struct vg_context *src,
+                           uint mask);
+
+struct st_framebuffer *st_create_framebuffer(const void *visual,
+                                             enum pipe_format colorFormat,
+                                             enum pipe_format depthFormat,
+                                             enum pipe_format stencilFormat,
+                                             uint width, uint height,
+                                             void *privateData);
+
+void st_resize_framebuffer(struct st_framebuffer *stfb,
+                           uint width, uint height);
+
+void st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                                uint surfIndex, struct pipe_surface *surf);
+
+void st_get_framebuffer_dimensions( struct st_framebuffer *stfb,
+				    uint *width, uint *height);
+
+int st_set_teximage(struct pipe_texture *pt, int target);
+
+int st_get_framebuffer_surface(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_surface **surf);
+
+int st_get_framebuffer_texture(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_texture **tex);
+
+void *st_framebuffer_private(struct st_framebuffer *stfb);
+
+void st_unreference_framebuffer(struct st_framebuffer *stfb);
+
+void st_make_current(struct vg_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read);
+
+void st_flush(struct vg_context *st, uint pipeFlushFlags,
+               struct pipe_fence_handle **fence);
+void st_finish(struct vg_context *st);
+
+void st_notify_swapbuffers(struct st_framebuffer *stfb);
+void st_notify_swapbuffers_complete(struct st_framebuffer *stfb);
+
+
+/** Generic function type */
+typedef void (*st_proc)();
+
+st_proc st_get_proc_address(const char *procname);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vg_translate.c b/src/gallium/state_trackers/vega/vg_translate.c
new file mode 100644
index 0000000000..00e0764706
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_translate.c
@@ -0,0 +1,1030 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "vg_translate.h"
+
+#include "pipe/p_format.h"
+#include "util/u_pack_color.h"
+
+void _vega_pack_rgba_span_float(struct vg_context *ctx,
+                                VGuint n, VGfloat rgba[][4],
+                                VGImageFormat dstFormat,
+                                void *dstAddr)
+{
+   VGint i;
+
+   switch (dstFormat) {
+   case VG_sRGBX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 255;
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sRGB_565: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         r = (r / 255.0) * 32;
+         g = (g / 255.0) * 32;
+         b = (b / 255.0) * 32;
+
+         dst[i] = b | g << 5 | r << 11;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_5551: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         r = (r / 255.0) * 32;
+         g = (g / 255.0) * 32;
+         b = (b / 255.0) * 32;
+         a = (a / 255.0);
+
+         dst[i] =  a | b << 1 | g << 6 | r << 11;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_4444: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         r = (r / 255.0) * 16;
+         g = (g / 255.0) * 16;
+         b = (b / 255.0) * 16;
+         a = (a / 255.0) * 16;
+
+         dst[i] =  a | b << 4 | g << 8 | r << 12;
+      }
+      return;
+   }
+      break;
+   case VG_sL_8: {
+      VGubyte *dst = (VGubyte *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         dst[i] =  a;
+      }
+      return;
+   }
+      break;
+   case VG_lRGBX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 255;
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lRGBA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+   case VG_lRGBA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = r << 24 | g << 16 | b << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lL_8: {
+      VGubyte *dst = (VGubyte *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a;
+      }
+      return;
+   }
+      break;
+   case VG_A_8: {
+      VGubyte *dst = (VGubyte *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         dst[i] = a;
+      }
+      return;
+   }
+      break;
+   case VG_BW_1: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         VGubyte res;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         res = (r + g + b + a)/4;
+         dst[i] =   (res & (128));
+      }
+      return;
+   }
+      break;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         dst[i] =   (a & (128));
+      }
+      return;
+   }
+      break;
+   case VG_A_4: {
+      VGshort *dst = (VGshort *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b, a;
+         VGubyte res;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+
+         res = a/4;
+         dst[i] =   (res & (128));
+      }
+      return;
+   }
+      break;
+#endif
+   case VG_sXRGB_8888:
+      break;
+   case VG_sARGB_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_8888_PRE:  {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_1555:
+      break;
+   case VG_sARGB_4444:
+      break;
+   case VG_lXRGB_8888:
+      break;
+   case VG_lARGB_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_lARGB_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | r << 16 | g << 8 | b;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 0xff;
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sBGR_565:
+      break;
+   case VG_sBGRA_5551:
+      break;
+   case VG_sBGRA_4444:
+      break;
+   case VG_lBGRX_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = 0xff;
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lBGRA_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_lBGRA_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = b << 24 | g << 16 | r << 8 | a;
+      }
+      return;
+   }
+      break;
+   case VG_sXBGR_8888:
+      break;
+   case VG_sABGR_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_1555:
+      break;
+   case VG_sABGR_4444:
+      break;
+   case VG_lXBGR_8888:
+      break;
+   case VG_lABGR_8888: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   case VG_lABGR_8888_PRE: {
+      VGint *dst = (VGint *)dstAddr;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = float_to_ubyte(rgba[i][0]);
+         g = float_to_ubyte(rgba[i][1]);
+         b = float_to_ubyte(rgba[i][2]);
+         a = float_to_ubyte(rgba[i][3]);
+         dst[i] = a << 24 | b << 16 | g << 8 | r;
+      }
+      return;
+   }
+      break;
+   default:
+      assert(!"Unknown ReadPixels format");
+      break;
+   }
+   assert(!"Not implemented ReadPixels format");
+}
+
+void _vega_unpack_float_span_rgba(struct vg_context *ctx,
+                                  VGuint n,
+                                  VGuint offset,
+                                  const void * data,
+                                  VGImageFormat dataFormat,
+                                  VGfloat rgba[][4])
+{
+   VGint i;
+
+   switch (dataFormat) {
+   case VG_sRGBX_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sRGBA_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sRGBA_8888_PRE: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sRGB_565: {
+      VGshort *src = (VGshort *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGfloat clr[4];
+         clr[0] = ((*src >> 10) & 31)/31.;
+         clr[1] = ((*src >>  5) & 95)/95.;
+         clr[2] = ((*src >>  0) & 31)/31.;
+         clr[3] = 1.f;
+
+         util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sRGBA_5551: {
+      VGshort *src = (VGshort *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGfloat clr[4];
+         clr[0] = ((*src >> 10) & 31)/31.;
+         clr[1] = ((*src >>  5) & 31)/31.;
+         clr[2] = ((*src >>  1) & 31)/31.;
+         clr[3] = ((*src >>  0) & 1)/1.;
+
+         util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sRGBA_4444: {
+      VGshort *src = (VGshort *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGfloat clr[4];
+         clr[0] = ((*src >> 12) & 15)/15.;
+         clr[1] = ((*src >>  8) & 15)/15.;
+         clr[2] = ((*src >>  4) & 15)/15.;
+         clr[3] = ((*src >>  0) & 15)/15.;
+
+         util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_sL_8: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_lRGBX_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_lRGBA_8888: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lRGBA_8888_PRE: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         r = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         b = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lL_8: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                         rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_A_8: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+   }
+      return;
+   case VG_BW_1: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; i += 8) {
+         VGfloat clr[4];
+         VGint j;
+         for (j = 0; j < 8 && j < n ; ++j) {
+            VGint shift = j;
+            clr[0] = (((*src) & (1<<shift)) >> shift);
+            clr[1] = clr[0];
+            clr[2] = clr[0];
+            clr[3] = 1.f;
+
+            util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i+j]);
+         }
+         ++src;
+      }
+   }
+      return;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset;
+      for (i = 0; i < n; i += 8) {
+         VGfloat clr[4];
+         VGint j;
+         for (j = 0; j < 8 && j < n ; ++j) {
+            VGint shift = j;
+            clr[0] = 0.f;
+            clr[1] = 0.f;
+            clr[2] = 0.f;
+            clr[3] = (((*src) & (1<<shift)) >> shift);
+
+            util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i+j]);
+         }
+         ++src;
+      }
+   }
+      return;
+   case VG_A_4: {
+      VGubyte *src = (VGubyte *)data;
+      src += offset/2;
+      for (i = 0; i < n; i += 2) {
+         VGfloat clr[4];
+         VGint j;
+         for (j = 0; j < n && j < 2; ++j) {
+            VGint bitter, shift;
+            if (j == 0) {
+               bitter = 0x0f;
+               shift = 0;
+            } else {
+               bitter = 0xf0;
+               shift = 4;
+            }
+            clr[0] = 0.f;
+            clr[1] = 0.f;
+            clr[2] = 0.f;
+            clr[3] = ((*src) & (bitter)) >> shift;
+
+            util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i +j]);
+         }
+         ++src;
+      }
+   }
+      return;
+#endif
+   case VG_sXRGB_8888:
+      break;
+   case VG_sARGB_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_8888_PRE: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sARGB_1555:
+      break;
+   case VG_sARGB_4444:
+      break;
+   case VG_lXRGB_8888:
+      break;
+   case VG_lARGB_8888: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lARGB_8888_PRE: {
+      VGint *src = (VGint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         r = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         b = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRX_8888:
+      break;
+   case VG_sBGRA_8888:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sBGRA_8888_PRE:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sBGR_565:
+      break;
+   case VG_sBGRA_5551:
+      break;
+   case VG_sBGRA_4444:
+      break;
+   case VG_lBGRX_8888:
+      break;
+   case VG_lBGRA_8888:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lBGRA_8888_PRE:  {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         b = (*src >> 24) & 0xff;
+         g = (*src >> 16) & 0xff;
+         r = (*src >>  8) & 0xff;
+         a = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sXBGR_8888:
+      break;
+   case VG_sABGR_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_8888_PRE: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_sABGR_1555:
+      break;
+   case VG_sABGR_4444:
+      break;
+   case VG_lXBGR_8888:
+      break;
+   case VG_lABGR_8888: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   case VG_lABGR_8888_PRE: {
+      VGuint *src = (VGuint *)data;
+      src += offset;
+      for (i = 0; i < n; ++i) {
+         VGubyte r, g, b ,a;
+         a = (*src >> 24) & 0xff;
+         b = (*src >> 16) & 0xff;
+         g = (*src >>  8) & 0xff;
+         r = (*src >>  0) & 0xff;
+
+         util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT,
+                            rgba[i]);
+         ++src;
+      }
+      return;
+   }
+      break;
+   default:
+      assert(!"Unknown ReadPixels format");
+      break;
+   }
+   assert(!"Not implemented ReadPixels format");
+}
+
+VGint _vega_size_for_format(VGImageFormat dataFormat)
+{
+   switch (dataFormat) {
+   case VG_sRGBX_8888:
+   case VG_sRGBA_8888:
+   case VG_sRGBA_8888_PRE:
+      return 4;
+   case VG_sRGB_565:
+   case VG_sRGBA_5551:
+   case VG_sRGBA_4444:
+      return 2;
+   case VG_sL_8:
+      return 1;
+   case VG_lRGBX_8888:
+   case VG_lRGBA_8888:
+   case VG_lRGBA_8888_PRE:
+      return 4;
+   case VG_lL_8:
+      return 1;
+   case VG_A_8:
+      return 1;
+   case VG_BW_1:
+      return 1;
+#ifdef OPENVG_VERSION_1_1
+   case VG_A_1:
+      break;
+   case VG_A_4:
+      break;
+#endif
+   case VG_sXRGB_8888:
+   case VG_sARGB_8888:
+   case VG_sARGB_8888_PRE:
+      return 4;
+   case VG_sARGB_1555:
+   case VG_sARGB_4444:
+      return 2;
+   case VG_lXRGB_8888:
+   case VG_lARGB_8888:
+   case VG_lARGB_8888_PRE:
+   case VG_sBGRX_8888:
+   case VG_sBGRA_8888:
+   case VG_sBGRA_8888_PRE:
+      return 4;
+   case VG_sBGR_565:
+   case VG_sBGRA_5551:
+   case VG_sBGRA_4444:
+      return 2;
+   case VG_lBGRX_8888:
+   case VG_lBGRA_8888:
+   case VG_lBGRA_8888_PRE:
+   case VG_sXBGR_8888:
+   case VG_sABGR_8888:
+   case VG_sABGR_8888_PRE:
+      return 4;
+   case VG_sABGR_1555:
+   case VG_sABGR_4444:
+      return 2;
+   case VG_lXBGR_8888:
+   case VG_lABGR_8888:
+   case VG_lABGR_8888_PRE:
+      return 4;
+   default:
+      assert(!"Unknown ReadPixels format");
+      break;
+   }
+   assert(!"Not implemented ReadPixels format");
+   return 0;
+}
diff --git a/src/gallium/state_trackers/vega/vg_translate.h b/src/gallium/state_trackers/vega/vg_translate.h
new file mode 100644
index 0000000000..70815bacbc
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vg_translate.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef VG_TRANSLATE_H
+#define VG_TRANSLATE_H
+
+#include "VG/openvg.h"
+#include "vg_context.h"
+
+/*FIXME: we really should be using translate module
+ * but pipe_format can't express some of the VG formats
+ * (the premultiplied ones) so currently it won't work */
+
+void _vega_pack_rgba_span_float(struct vg_context *ctx,
+                                VGuint n, VGfloat rgba[][4],
+                                VGImageFormat dstFormat,
+                                void *dstAddr);
+void _vega_unpack_float_span_rgba(struct vg_context *ctx,
+                                  VGuint n,
+                                  VGuint offset,
+                                  const void * data,
+                                  VGImageFormat dataFormat,
+                                  VGfloat rgba[][4]);
+VGint _vega_size_for_format(VGImageFormat format);
+
+#endif
diff --git a/src/gallium/state_trackers/vega/vgu.c b/src/gallium/state_trackers/vega/vgu.c
new file mode 100644
index 0000000000..7dc51c5599
--- /dev/null
+++ b/src/gallium/state_trackers/vega/vgu.c
@@ -0,0 +1,450 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "VG/openvg.h"
+#include "VG/vgu.h"
+
+#include "matrix.h"
+#include "path.h"
+
+#include "util/u_debug.h"
+#include "util/u_pointer.h"
+
+#include <math.h>
+#include <assert.h>
+
+static VGboolean is_aligned_to(const void *ptr, VGbyte alignment)
+{
+   void *aligned = align_pointer(ptr, alignment);
+   return (ptr == aligned) ? VG_TRUE : VG_FALSE;
+}
+
+static VGboolean is_aligned(const void *ptr)
+{
+   return is_aligned_to(ptr, 4);
+}
+
+static void vgu_append_float_coords(VGPath path,
+                                    const VGubyte *cmds,
+                                    VGint num_cmds,
+                                    const VGfloat *coords,
+                                    VGint num_coords)
+{
+   VGubyte common_data[40 * sizeof(VGfloat)];
+   struct path *p = (struct path *)path;
+
+   vg_float_to_datatype(path_datatype(p), common_data, coords, num_coords);
+   vgAppendPathData(path, num_cmds, cmds, common_data);
+}
+
+VGUErrorCode vguLine(VGPath path,
+                     VGfloat x0, VGfloat y0,
+                     VGfloat x1, VGfloat y1)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS};
+   VGfloat coords[4];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+
+   coords[0] = x0;
+   coords[1] = y0;
+   coords[2] = x1;
+   coords[3] = y1;
+
+   vgu_append_float_coords(path, cmds, 2, coords, 4);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguPolygon(VGPath path,
+                        const VGfloat * points,
+                        VGint count,
+                        VGboolean closed)
+{
+   VGubyte *cmds;
+   VGfloat *coords;
+   VGbitfield caps;
+   VGint i;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+
+   if (!points || count <= 0 || !is_aligned(points)) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+
+   cmds   = malloc(sizeof(VGubyte) * count + 1);
+   coords = malloc(sizeof(VGfloat) * count * 2);
+
+   cmds[0] = VG_MOVE_TO_ABS;
+   coords[0] = points[0];
+   coords[1] = points[1];
+   for (i = 1; i < count; ++i) {
+      cmds[i] = VG_LINE_TO_ABS;
+      coords[2*i + 0] = points[2*i + 0];
+      coords[2*i + 1] = points[2*i + 1];
+   }
+
+   if (closed) {
+      cmds[i] = VG_CLOSE_PATH;
+      ++i;
+   }
+
+   vgu_append_float_coords(path, cmds, i, coords, 2*i);
+
+   free(cmds);
+   free(coords);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode  vguRect(VGPath path,
+                      VGfloat x, VGfloat y,
+                      VGfloat width, VGfloat height)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_HLINE_TO_REL,
+                                  VG_VLINE_TO_REL,
+                                  VG_HLINE_TO_REL,
+                                  VG_CLOSE_PATH
+   };
+   VGfloat coords[5];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   coords[0] =  x;
+   coords[1] =  y;
+   coords[2] =  width;
+   coords[3] =  height;
+   coords[4] = -width;
+
+   vgu_append_float_coords(path, cmds, 5, coords, 5);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguRoundRect(VGPath path,
+                          VGfloat x, VGfloat y,
+                          VGfloat width,
+                          VGfloat height,
+                          VGfloat arcWidth,
+                          VGfloat arcHeight)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_HLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_VLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_HLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_VLINE_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_CLOSE_PATH
+   };
+   VGfloat c[26];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   c[0] =  x + arcWidth/2; c[1] =  y;
+
+   c[2] = width - arcWidth;
+
+   c[3] = arcWidth/2; c[4] = arcHeight/2; c[5] = 0;
+   c[6] = arcWidth/2; c[7] = arcHeight/2;
+
+   c[8] = height - arcHeight;
+
+   c[9] = arcWidth/2; c[10] = arcHeight/2; c[11] = 0;
+   c[12] = -arcWidth/2; c[13] = arcHeight/2;
+
+   c[14] = -(width - arcWidth);
+
+   c[15] = arcWidth/2; c[16] = arcHeight/2; c[17] = 0;
+   c[18] = -arcWidth/2; c[19] = -arcHeight/2;
+
+   c[20] = -(height - arcHeight);
+
+   c[21] = arcWidth/2; c[22] = arcHeight/2; c[23] = 0;
+   c[24] = arcWidth/2; c[25] = -arcHeight/2;
+
+   vgu_append_float_coords(path, cmds, 10, c, 26);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguEllipse(VGPath path,
+                        VGfloat cx, VGfloat cy,
+                        VGfloat width,
+                        VGfloat height)
+{
+   static const VGubyte cmds[] = {VG_MOVE_TO_ABS,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_SCCWARC_TO_REL,
+                                  VG_CLOSE_PATH
+   };
+   VGfloat coords[12];
+   VGbitfield caps;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   coords[0] = cx + width/2; coords[1] = cy;
+
+   coords[2] = width/2; coords[3] = height/2; coords[4] = 0;
+   coords[5] = -width; coords[6] = 0;
+
+   coords[7] = width/2; coords[8] = height/2; coords[9] = 0;
+   coords[10] = width; coords[11] = 0;
+
+   vgu_append_float_coords(path, cmds, 4, coords, 11);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguArc(VGPath path,
+                    VGfloat x, VGfloat y,
+                    VGfloat width, VGfloat height,
+                    VGfloat startAngle,
+                    VGfloat angleExtent,
+                    VGUArcType arcType)
+{
+   VGubyte cmds[11];
+   VGfloat coords[40];
+   VGbitfield caps;
+   VGfloat last = startAngle + angleExtent;
+   VGint i, c = 0;
+
+   if (path == VG_INVALID_HANDLE) {
+      return VGU_BAD_HANDLE_ERROR;
+   }
+   caps = vgGetPathCapabilities(path);
+   if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) {
+      return VGU_PATH_CAPABILITY_ERROR;
+   }
+   if (width <= 0 || height <= 0) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+   if (arcType != VGU_ARC_OPEN &&
+       arcType != VGU_ARC_CHORD &&
+       arcType != VGU_ARC_PIE) {
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+   }
+
+   cmds[c] = VG_MOVE_TO_ABS; ++c;
+   coords[0] = x+cos(DEGREES_TO_RADIANS(startAngle))*width/2;
+   coords[1] = y+sin(DEGREES_TO_RADIANS(startAngle))*height/2;
+#ifdef DEBUG_VGUARC
+   debug_printf("start [%f, %f]\n", coords[0], coords[1]);
+#endif
+   i = 2;
+   if (angleExtent > 0) {
+      VGfloat angle = startAngle + 180;
+      while (angle < last) {
+         cmds[c] = VG_SCCWARC_TO_ABS; ++c;
+         coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0;
+         coords[i+3] = x + cos(DEGREES_TO_RADIANS(angle))*width/2;
+         coords[i+4] = y + sin(DEGREES_TO_RADIANS(angle))*height/2;
+#ifdef DEBUG_VGUARC
+         debug_printf("1 [%f, %f]\n", coords[i+3],
+                      coords[i+4]);
+#endif
+         i += 5;
+         angle += 180;
+      }
+      cmds[c] = VG_SCCWARC_TO_ABS; ++c;
+      coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0;
+      coords[i+3] = x+cos(DEGREES_TO_RADIANS(last))*width/2;
+      coords[i+4] = y+sin(DEGREES_TO_RADIANS(last))*height/2;
+#ifdef DEBUG_VGUARC
+      debug_printf("2 [%f, %f]\n", coords[i+3],
+                   coords[i+4]);
+#endif
+      i += 5;
+   } else {
+      VGfloat angle = startAngle - 180;
+      while (angle > last) {
+         cmds[c] = VG_SCWARC_TO_ABS; ++c;
+         coords[i] =  width/2; coords[i+1] = height/2; coords[i+2] = 0;
+         coords[i+3] = x + cos(DEGREES_TO_RADIANS(angle)) * width/2;
+         coords[i+4] = y + sin(DEGREES_TO_RADIANS(angle)) * height/2;
+#ifdef DEBUG_VGUARC
+         debug_printf("3 [%f, %f]\n", coords[i+3],
+                      coords[i+4]);
+#endif
+         angle -= 180;
+         i += 5;
+      }
+      cmds[c] = VG_SCWARC_TO_ABS; ++c;
+      coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0;
+      coords[i+3] = x + cos(DEGREES_TO_RADIANS(last)) * width/2;
+      coords[i+4] = y + sin(DEGREES_TO_RADIANS(last)) * height/2;
+#ifdef DEBUG_VGUARC
+      debug_printf("4 [%f, %f]\n", coords[i+3],
+                   coords[i+4]);
+#endif
+      i += 5;
+   }
+
+   if (arcType == VGU_ARC_PIE) {
+      cmds[c] = VG_LINE_TO_ABS; ++c;
+      coords[i] = x; coords[i + 1] = y;
+      i += 2;
+   }
+   if (arcType == VGU_ARC_PIE || arcType == VGU_ARC_CHORD) {
+      cmds[c] = VG_CLOSE_PATH;
+      ++c;
+   }
+
+   assert(c < 11);
+
+   vgu_append_float_coords(path, cmds, c, coords, i);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguComputeWarpQuadToSquare(VGfloat sx0, VGfloat sy0,
+                                        VGfloat sx1, VGfloat sy1,
+                                        VGfloat sx2, VGfloat sy2,
+                                        VGfloat sx3, VGfloat sy3,
+                                        VGfloat * matrix)
+{
+   struct matrix mat;
+
+   if (!matrix || !is_aligned(matrix))
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+
+   if (!matrix_quad_to_square(sx0, sy0,
+                              sx1, sy1,
+                              sx2, sy2,
+                              sx3, sy3,
+                              &mat))
+      return VGU_BAD_WARP_ERROR;
+
+   if (!matrix_is_invertible(&mat))
+      return VGU_BAD_WARP_ERROR;
+
+   memcpy(matrix, mat.m, sizeof(VGfloat) * 9);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguComputeWarpSquareToQuad(VGfloat dx0, VGfloat dy0,
+                                        VGfloat dx1, VGfloat dy1,
+                                        VGfloat dx2, VGfloat dy2,
+                                        VGfloat dx3, VGfloat dy3,
+                                        VGfloat * matrix)
+{
+   struct matrix mat;
+
+   if (!matrix || !is_aligned(matrix))
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+
+   if (!matrix_square_to_quad(dx0, dy0,
+                              dx1, dy1,
+                              dx2, dy2,
+                              dx3, dy3,
+                              &mat))
+      return VGU_BAD_WARP_ERROR;
+
+   if (!matrix_is_invertible(&mat))
+      return VGU_BAD_WARP_ERROR;
+
+   memcpy(matrix, mat.m, sizeof(VGfloat) * 9);
+
+   return VGU_NO_ERROR;
+}
+
+VGUErrorCode vguComputeWarpQuadToQuad(VGfloat dx0, VGfloat dy0,
+                                      VGfloat dx1, VGfloat dy1,
+                                      VGfloat dx2, VGfloat dy2,
+                                      VGfloat dx3, VGfloat dy3,
+                                      VGfloat sx0, VGfloat sy0,
+                                      VGfloat sx1, VGfloat sy1,
+                                      VGfloat sx2, VGfloat sy2,
+                                      VGfloat sx3, VGfloat sy3,
+                                      VGfloat * matrix)
+{
+   struct matrix mat;
+
+   if (!matrix || !is_aligned(matrix))
+      return VGU_ILLEGAL_ARGUMENT_ERROR;
+
+   if (!matrix_quad_to_quad(dx0, dy0,
+                            dx1, dy1,
+                            dx2, dy2,
+                            dx3, dy3,
+                            sx0, sy0,
+                            sx1, sy1,
+                            sx2, sy2,
+                            sx3, sy3,
+                            &mat))
+      return VGU_BAD_WARP_ERROR;
+
+   memcpy(matrix, mat.m, sizeof(VGfloat) * 9);
+
+   return VGU_NO_ERROR;
+}
diff --git a/src/gallium/state_trackers/xorg/Makefile b/src/gallium/state_trackers/xorg/Makefile
index a00ea3e2a4..27a1990724 100644
--- a/src/gallium/state_trackers/xorg/Makefile
+++ b/src/gallium/state_trackers/xorg/Makefile
@@ -1,29 +1,18 @@
-TARGET     = libxorgtracker.a
-CFILES     = $(wildcard ./*.c)
-OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
-GALLIUMDIR = ../..
-TOP        = ../../../..
-
+TOP = ../../../..
 include $(TOP)/configs/current
 
-CFLAGS = -DHAVE_CONFIG_H \
-         -g -Wall -Wimplicit-function-declaration -fPIC \
-         $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \
-         -I$(GALLIUMDIR)/include \
-         -I$(GALLIUMDIR)/auxiliary \
-         -I$(TOP)/src/mesa/drivers/dri/common \
-         -I$(TOP)/src/mesa \
-         -I$(TOP)/include \
-         -I$(TOP)/src/egl/main
-
-#############################################
-
-.PHONY	= all clean
+LIBNAME = xorgtracker
 
-all: $(TARGET)
+LIBRARY_INCLUDES = \
+	-DHAVE_CONFIG_H \
+	$(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/include \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-I$(TOP)/src/mesa/main
 
-$(TARGET): $(OBJECTS)
-	ar rcs $(TARGET) $(OBJECTS)
+C_SOURCES = $(wildcard ./*.c)
 
-clean:
-	rm -rf $(OBJECTS) $(TARGET)
+include ../../Makefile.template
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c
new file mode 100644
index 0000000000..255ae91f09
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_composite.c
@@ -0,0 +1,27 @@
+#include "xorg_composite.h"
+
+
+boolean xorg_composite_accelerated(int op,
+                                   PicturePtr pSrcPicture,
+                                   PicturePtr pMaskPicture,
+                                   PicturePtr pDstPicture)
+{
+   return FALSE;
+}
+
+boolean xorg_composite_bind_state(struct exa_context *exa,
+                                  int op,
+                                  PicturePtr pSrcPicture,
+                                  PicturePtr pMaskPicture,
+                                  PicturePtr pDstPicture)
+{
+   return FALSE;
+}
+
+void xorg_composite(struct exa_context *exa,
+                    struct exa_pixmap_priv *dst,
+                    int srcX, int srcY, int maskX, int maskY,
+                    int dstX, int dstY, int width, int height)
+{
+}
+
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.h b/src/gallium/state_trackers/xorg/xorg_composite.h
new file mode 100644
index 0000000000..a52e0e6dff
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_composite.h
@@ -0,0 +1,22 @@
+#ifndef XORG_COMPOSITE_H
+#define XORG_COMPOSITE_H
+
+#include "xorg_exa.h"
+
+boolean xorg_composite_accelerated(int op,
+                                   PicturePtr pSrcPicture,
+                                   PicturePtr pMaskPicture,
+                                   PicturePtr pDstPicture);
+
+boolean xorg_composite_bind_state(struct exa_context *exa,
+                                  int op,
+                                  PicturePtr pSrcPicture,
+                                  PicturePtr pMaskPicture,
+                                  PicturePtr pDstPicture);
+
+void xorg_composite(struct exa_context *exa,
+                    struct exa_pixmap_priv *dst,
+                    int srcX, int srcY, int maskX, int maskY,
+                    int dstX, int dstY, int width, int height);
+
+#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c
index 7304113a65..2235961b03 100644
--- a/src/gallium/state_trackers/xorg/xorg_crtc.c
+++ b/src/gallium/state_trackers/xorg/xorg_crtc.c
@@ -171,7 +171,6 @@ crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data)
 static void
 crtc_destroy(xf86CrtcPtr crtc)
 {
-    modesettingPtr ms = modesettingPTR(crtc->scrn);
     struct crtc_private *crtcp = crtc->driver_private;
 
     if (crtcp->cursor_buf)
@@ -194,9 +193,10 @@ crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image)
 					       PIPE_BUFFER_USAGE_CPU_WRITE |
 					       PIPE_BUFFER_USAGE_GPU_READ,
 					       64*64*4);
-	drm_api_hooks.handle_from_buffer(ms->screen,
-					 crtcp->cursor_buf,
-					 &crtcp->cursor_handle);
+	ms->api->handle_from_buffer(ms->api,
+				    ms->screen,
+				    crtcp->cursor_buf,
+				    &crtcp->cursor_handle);
     }
 
     ptr = pipe_buffer_map(ms->screen, crtcp->cursor_buf, PIPE_BUFFER_USAGE_CPU_WRITE);
diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c
index 401bd39dac..ae3338ffef 100644
--- a/src/gallium/state_trackers/xorg/xorg_dri2.c
+++ b/src/gallium/state_trackers/xorg/xorg_dri2.c
@@ -108,8 +108,8 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
 	    tex = ms->screen->texture_create(ms->screen, &template);
 	}
 
-	drm_api_hooks.buffer_from_texture(tex, &buf, &stride);
-	drm_api_hooks.global_handle_from_buffer(ms->screen, buf, &handle);
+	ms->api->buffer_from_texture(ms->api, tex, &buf, &stride);
+	ms->api->global_handle_from_buffer(ms->api, ms->screen, buf, &handle);
 
 	buffers[i].name = handle;
 	buffers[i].attachment = attachments[i];
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index 45e831f0c2..e01e5294b1 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -300,6 +300,7 @@ PreInit(ScrnInfoPtr pScrn, int flags)
 	    ms->PciInfo->dev, ms->PciInfo->func
 	);
 
+    ms->api = drm_api_create();
     ms->fd = drmOpen(NULL, BusID);
 
     if (ms->fd < 0)
@@ -476,7 +477,7 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
     }
 
     if (!ms->screen) {
-	ms->screen = drm_api_hooks.create_screen(ms->fd, NULL);
+	ms->screen = ms->api->create_screen(ms->api, ms->fd, NULL);
 
 	if (!ms->screen) {
 	    FatalError("Could not init pipe_screen\n");
@@ -678,6 +679,8 @@ CloseScreen(int scrnIndex, ScreenPtr pScreen)
     if (ms->exa)
 	xorg_exa_close(pScrn);
 
+	ms->api->destroy(ms->api);
+	ms->api = NULL;
     drmClose(ms->fd);
     ms->fd = -1;
 
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index 7913174354..2c4291aa4e 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -28,9 +28,14 @@
  *
  */
 
-#include "xorg-server.h"
-#include "xf86.h"
+#include "xorg_exa.h"
 #include "xorg_tracker.h"
+#include "xorg_composite.h"
+
+#include <xorg-server.h>
+#include <xf86.h>
+#include <picturestr.h>
+#include <picture.h>
 
 #include "pipe/p_format.h"
 #include "pipe/p_context.h"
@@ -39,24 +44,6 @@
 
 #include "util/u_rect.h"
 
-struct exa_entity
-{
-    ExaDriverPtr pExa;
-    struct pipe_context *ctx;
-    struct pipe_screen *scrn;
-};
-
-struct PixmapPriv
-{
-    int flags;
-    struct pipe_texture *tex;
-    unsigned int color;
-    struct pipe_surface *src_surf; /* for copies */
-
-    struct pipe_transfer *map_transfer;
-    unsigned map_count;
-};
-
 /*
  * Helper functions
  */
@@ -113,8 +100,8 @@ ExaPrepareAccess(PixmapPtr pPix, int index)
     ScreenPtr pScreen = pPix->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv;
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv;
 
     priv = exaGetPixmapDriverPrivate(pPix);
 
@@ -145,8 +132,8 @@ ExaFinishAccess(PixmapPtr pPix, int index)
     ScreenPtr pScreen = pPix->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv;
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv;
     priv = exaGetPixmapDriverPrivate(pPix);
 
     if (!priv)
@@ -168,8 +155,8 @@ ExaDone(PixmapPtr pPixmap)
 {
     ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct exa_entity *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
 
     if (!priv)
 	return;
@@ -190,8 +177,8 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg)
 {
     ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct exa_entity *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
 
     if (1)
         return FALSE;
@@ -221,8 +208,8 @@ ExaSolid(PixmapPtr pPixmap, int x0, int y0, int x1, int y1)
 {
     ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
     struct pipe_surface *surf =
 	exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0,
 				   PIPE_BUFFER_USAGE_GPU_READ |
@@ -240,9 +227,9 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir,
 {
     ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
-    struct PixmapPriv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
+    struct exa_pixmap_priv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap);
 
     if (1)
         return FALSE;
@@ -279,8 +266,8 @@ ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY,
 {
     ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
+    struct exa_context *exa = ms->exa;
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDstPixmap);
     struct pipe_surface *surf =
 	exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0,
 				   PIPE_BUFFER_USAGE_GPU_READ |
@@ -296,13 +283,25 @@ ExaPrepareComposite(int op, PicturePtr pSrcPicture,
 		    PicturePtr pMaskPicture, PicturePtr pDstPicture,
 		    PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst)
 {
-    return FALSE;
+   ScrnInfoPtr pScrn = xf86Screens[pDst->drawable.pScreen->myNum];
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa = ms->exa;
+
+   return xorg_composite_bind_state(exa, op, pSrcPicture, pMaskPicture,
+                                    pDstPicture);
 }
 
 static void
 ExaComposite(PixmapPtr pDst, int srcX, int srcY, int maskX, int maskY,
 	     int dstX, int dstY, int width, int height)
 {
+   ScrnInfoPtr pScrn = xf86Screens[pDst->drawable.pScreen->myNum];
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa = ms->exa;
+   struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDst);
+
+   xorg_composite(exa, priv, srcX, srcY, maskX, maskY,
+                  dstX, dstY, width, height);
 }
 
 static Bool
@@ -310,15 +309,18 @@ ExaCheckComposite(int op,
 		  PicturePtr pSrcPicture, PicturePtr pMaskPicture,
 		  PicturePtr pDstPicture)
 {
-    return FALSE;
+   return xorg_composite_accelerated(op,
+                                     pSrcPicture,
+                                     pMaskPicture,
+                                     pDstPicture);
 }
 
 static void *
 ExaCreatePixmap(ScreenPtr pScreen, int size, int align)
 {
-    struct PixmapPriv *priv;
+    struct exa_pixmap_priv *priv;
 
-    priv = xcalloc(1, sizeof(struct PixmapPriv));
+    priv = xcalloc(1, sizeof(struct exa_pixmap_priv));
     if (!priv)
 	return NULL;
 
@@ -328,7 +330,7 @@ ExaCreatePixmap(ScreenPtr pScreen, int size, int align)
 static void
 ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv)
 {
-    struct PixmapPriv *priv = (struct PixmapPriv *)dPriv;
+    struct exa_pixmap_priv *priv = (struct exa_pixmap_priv *)dPriv;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
 
@@ -344,7 +346,7 @@ ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv)
 static Bool
 ExaPixmapIsOffscreen(PixmapPtr pPixmap)
 {
-    struct PixmapPriv *priv;
+    struct exa_pixmap_priv *priv;
 
     priv = exaGetPixmapDriverPrivate(pPixmap);
 
@@ -363,7 +365,7 @@ xorg_exa_get_pixmap_handle(PixmapPtr pPixmap)
     ScreenPtr pScreen = pPixmap->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct PixmapPriv *priv;
+    struct exa_pixmap_priv *priv;
     struct pipe_buffer *buffer = NULL;
     unsigned handle;
     unsigned stride;
@@ -380,8 +382,8 @@ xorg_exa_get_pixmap_handle(PixmapPtr pPixmap)
 	return 0;
     }
 
-    drm_api_hooks.buffer_from_texture(priv->tex, &buffer, &stride);
-    drm_api_hooks.handle_from_buffer(ms->screen, buffer, &handle);
+    ms->api->buffer_from_texture(ms->api, priv->tex, &buffer, &stride);
+    ms->api->handle_from_buffer(ms->api, ms->screen, buffer, &handle);
     pipe_buffer_reference(&buffer, NULL);
     return handle;
 }
@@ -393,9 +395,9 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 {
     ScreenPtr pScreen = pPixmap->drawable.pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
+    struct exa_context *exa = ms->exa;
 
     if (!priv)
 	return FALSE;
@@ -456,7 +458,7 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 struct pipe_texture *
 xorg_exa_get_texture(PixmapPtr pPixmap)
 {
-    struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
     struct pipe_texture *tex = NULL;
     pipe_texture_reference(&tex, priv->tex);
     return tex;
@@ -466,7 +468,7 @@ void
 xorg_exa_close(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa = ms->exa;
+    struct exa_context *exa = ms->exa;
 
     if (exa->ctx)
 	exa->ctx->destroy(exa->ctx);
@@ -480,10 +482,10 @@ void *
 xorg_exa_init(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_entity *exa;
+    struct exa_context *exa;
     ExaDriverPtr pExa;
 
-    exa = xcalloc(1, sizeof(struct exa_entity));
+    exa = xcalloc(1, sizeof(struct exa_context));
     if (!exa)
 	return NULL;
 
@@ -527,7 +529,7 @@ xorg_exa_init(ScrnInfoPtr pScrn)
     }
 
     exa->scrn = ms->screen;
-    exa->ctx = drm_api_hooks.create_context(exa->scrn);
+    exa->ctx = ms->api->create_context(ms->api, exa->scrn);
     /* Share context with DRI */
     ms->ctx = exa->ctx;
 
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.h b/src/gallium/state_trackers/xorg/xorg_exa.h
new file mode 100644
index 0000000000..650997aec6
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_exa.h
@@ -0,0 +1,27 @@
+#ifndef XORG_EXA_H
+#define XORG_EXA_H
+
+#include "xorg_tracker.h"
+
+struct exa_context
+{
+    ExaDriverPtr pExa;
+    struct pipe_context *ctx;
+    struct pipe_screen *scrn;
+};
+
+
+struct exa_pixmap_priv
+{
+    int flags;
+    struct pipe_texture *tex;
+    unsigned int color;
+    struct pipe_surface *src_surf; /* for copies */
+
+    struct pipe_transfer *map_transfer;
+    unsigned map_count;
+};
+
+
+
+#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h
index 82c3890dfb..b4742bdbf5 100644
--- a/src/gallium/state_trackers/xorg/xorg_tracker.h
+++ b/src/gallium/state_trackers/xorg/xorg_tracker.h
@@ -31,11 +31,14 @@
 #ifndef _XORG_TRACKER_H_
 #define _XORG_TRACKER_H_
 
+#include <stdint.h>
 #include <errno.h>
 #include <drm.h>
 #include <xf86drm.h>
 #include <xf86drmMode.h>
-#include "exa.h"
+#include <xorg-server.h>
+#include <xf86.h>
+#include <exa.h>
 
 #include "pipe/p_screen.h"
 #include "state_tracker/drm_api.h"
@@ -75,6 +78,7 @@ typedef struct _modesettingRec
     CreateScreenResourcesProcPtr createScreenResources;
 
     /* gallium */
+    struct drm_api *api;
     struct pipe_screen *screen;
     struct pipe_context *ctx;
 
diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template
index 9f92cb4207..985e5a861f 100644
--- a/src/gallium/winsys/drm/Makefile.template
+++ b/src/gallium/winsys/drm/Makefile.template
@@ -118,7 +118,7 @@ clean:
 
 install: $(LIBNAME)
 	$(INSTALL) -d $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
 
 
 include depend
diff --git a/src/gallium/winsys/drm/intel/SConscript b/src/gallium/winsys/drm/intel/SConscript
new file mode 100644
index 0000000000..50d7b75ed6
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/SConscript
@@ -0,0 +1,7 @@
+Import('*')
+
+SConscript(['gem/SConscript',])
+
+if 'mesa' in env['statetrackers']:
+
+    SConscript(['dri/SConscript'])
diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile
index ac0891a646..5e212b62a4 100644
--- a/src/gallium/winsys/drm/intel/dri/Makefile
+++ b/src/gallium/winsys/drm/intel/dri/Makefile
@@ -6,7 +6,9 @@ LIBNAME = i915_dri.so
 PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/identity/libidentity.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a
 
 
diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript
new file mode 100644
index 0000000000..e14e96e32f
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/SConscript
@@ -0,0 +1,16 @@
+Import('*')
+
+env = drienv.Clone()
+
+drivers = [
+    softpipe,
+    i915simple,
+    trace,
+    inteldrm
+]
+
+env.SharedLibrary(
+    target ='i915_dri.so',
+    source = COMMON_GALLIUM_SOURCES,
+    LIBS = drivers + mesa + auxiliaries + env['LIBS'],
+)
diff --git a/src/gallium/winsys/drm/intel/gem/SConscript b/src/gallium/winsys/drm/intel/gem/SConscript
new file mode 100644
index 0000000000..ea8a2e55f6
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/SConscript
@@ -0,0 +1,17 @@
+Import('*')
+
+env = drienv.Clone()
+
+inteldrm_sources = [
+    'intel_be_api.c',
+    'intel_be_batchbuffer.c',
+    'intel_be_context.c',
+    'intel_be_device.c',
+]
+
+inteldrm = env.ConvenienceLibrary(
+    target ='inteldrm',
+    source = inteldrm_sources,
+)
+
+Export('inteldrm')
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.c b/src/gallium/winsys/drm/intel/gem/intel_be_api.c
index f4ef7c2d88..dd5f814da3 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_api.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.c
@@ -1,15 +1,37 @@
 
 #include "intel_be_api.h"
 #include "i915simple/i915_winsys.h"
+#include "identity/id_drm.h"
+#include "trace/tr_drm.h"
 
-struct drm_api drm_api_hooks =
+static void destroy(struct drm_api *api)
+{
+
+}
+
+struct drm_api intel_be_drm_api =
 {
 	/* intel_be_context.c */
 	.create_context = intel_be_create_context,
 	/* intel_be_device.c */
 	.create_screen = intel_be_create_screen,
-	.buffer_from_texture = i915_get_texture_buffer,
+	.buffer_from_texture = intel_be_get_texture_buffer,
 	.buffer_from_handle = intel_be_buffer_from_handle,
 	.handle_from_buffer = intel_be_handle_from_buffer,
 	.global_handle_from_buffer = intel_be_global_handle_from_buffer,
+	.destroy = destroy,
 };
+
+struct drm_api *
+drm_api_create()
+{
+#ifdef DEBUG
+#if 0
+	return identity_drm_create(&intel_be_drm_api);
+#else
+	return trace_drm_create(&intel_be_drm_api);
+#endif
+#else
+	return &intel_be_drm_api;
+#endif
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h
index 1c622f3b97..f286b62eb8 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_api.h
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.h
@@ -8,8 +8,11 @@
 
 #include "intel_be_device.h"
 
-struct pipe_screen *intel_be_create_screen(int drmFD,
+extern struct drm_api intel_be_drm_api;
+
+struct pipe_screen *intel_be_create_screen(struct drm_api *api, int drmFD,
 					   struct drm_create_screen_arg *arg);
-struct pipe_context *intel_be_create_context(struct pipe_screen *screen);
+struct pipe_context *intel_be_create_context(struct drm_api *api,
+					     struct pipe_screen *screen);
 
 #endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
index fe0b138fbe..db84f9af51 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
@@ -97,7 +97,7 @@ intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *de
 }
 
 struct pipe_context *
-intel_be_create_context(struct pipe_screen *screen)
+intel_be_create_context(struct drm_api *api, struct pipe_screen *screen)
 {
 	struct intel_be_context *intel;
 	struct pipe_context *pipe;
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
index 907ac86637..e3630f5d12 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
@@ -142,8 +142,27 @@ err:
 	return NULL;
 }
 
+boolean
+intel_be_get_texture_buffer(struct drm_api *api,
+                            struct pipe_texture *texture,
+                            struct pipe_buffer **buffer,
+                            unsigned *stride)
+{
+	struct intel_be_device *dev;
+
+	if (!texture)
+		return FALSE;
+
+	dev = intel_be_device(texture->screen->winsys);
+	if (dev->softpipe)
+		return softpipe_get_texture_buffer(texture, buffer, stride);
+	else
+		return i915_get_texture_buffer(texture, buffer, stride);
+}
+
 struct pipe_buffer *
-intel_be_buffer_from_handle(struct pipe_screen *screen,
+intel_be_buffer_from_handle(struct drm_api *api,
+                            struct pipe_screen *screen,
                             const char* name, unsigned handle)
 {
 	struct intel_be_device *dev = intel_be_device(screen->winsys);
@@ -174,7 +193,8 @@ err:
 }
 
 boolean
-intel_be_handle_from_buffer(struct pipe_screen *screen,
+intel_be_handle_from_buffer(struct drm_api *api,
+                            struct pipe_screen *screen,
                             struct pipe_buffer *buffer,
                             unsigned *handle)
 {
@@ -186,7 +206,8 @@ intel_be_handle_from_buffer(struct pipe_screen *screen,
 }
 
 boolean
-intel_be_global_handle_from_buffer(struct pipe_screen *screen,
+intel_be_global_handle_from_buffer(struct drm_api *api,
+                                   struct pipe_screen *screen,
 				   struct pipe_buffer *buffer,
 				   unsigned *handle)
 {
@@ -296,6 +317,7 @@ intel_be_get_device_id(unsigned int *device_id)
 {
    char path[512];
    FILE *file;
+   void *shutup_gcc;
 
    /*
     * FIXME: Fix this up to use a drm ioctl or whatever.
@@ -307,13 +329,14 @@ intel_be_get_device_id(unsigned int *device_id)
       return;
    }
 
-   fgets(path, sizeof(path), file);
+   shutup_gcc = fgets(path, sizeof(path), file);
    sscanf(path, "%x", device_id);
    fclose(file);
 }
 
 struct pipe_screen *
-intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg)
+intel_be_create_screen(struct drm_api *api, int drmFD,
+		       struct drm_create_screen_arg *arg)
 {
 	struct intel_be_device *dev;
 	struct pipe_screen *screen;
@@ -339,7 +362,6 @@ intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg)
 
 	if (dev->softpipe) {
 		screen = softpipe_create_screen(&dev->base);
-		drm_api_hooks.buffer_from_texture = softpipe_get_texture_buffer;
 	} else
 		screen = i915_create_screen(&dev->base, deviceID);
 
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
index b32637ece2..56d95bd7fe 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
@@ -8,6 +8,8 @@
 #include "drm.h"
 #include "intel_bufmgr.h"
 
+struct drm_api;
+
 /*
  * Device
  */
@@ -50,13 +52,23 @@ struct intel_be_buffer {
 	unsigned flink;
 };
 
+/*
+ * Wrapper for driver get_texture_buffer functions.
+ */
+boolean
+intel_be_get_texture_buffer(struct drm_api *api,
+                            struct pipe_texture *texture,
+                            struct pipe_buffer **buffer,
+                            unsigned *stride);
+
 /**
  * Create a be buffer from a drm bo handle.
  *
  * Takes a reference.
  */
 struct pipe_buffer *
-intel_be_buffer_from_handle(struct pipe_screen *screen,
+intel_be_buffer_from_handle(struct drm_api *api,
+                            struct pipe_screen *screen,
                             const char* name, unsigned handle);
 
 /**
@@ -65,7 +77,8 @@ intel_be_buffer_from_handle(struct pipe_screen *screen,
  * If buffer is destroyed handle may become invalid.
  */
 boolean
-intel_be_handle_from_buffer(struct pipe_screen *screen,
+intel_be_handle_from_buffer(struct drm_api *api,
+                            struct pipe_screen *screen,
                             struct pipe_buffer *buffer,
                             unsigned *handle);
 
@@ -75,7 +88,8 @@ intel_be_handle_from_buffer(struct pipe_screen *screen,
  * If buffer is destroyed handle may become invalid.
  */
 boolean
-intel_be_global_handle_from_buffer(struct pipe_screen *screen,
+intel_be_global_handle_from_buffer(struct drm_api *api,
+                                   struct pipe_screen *screen,
                                    struct pipe_buffer *buffer,
                                    unsigned *handle);
 
diff --git a/src/gallium/winsys/drm/intel/xorg/Makefile b/src/gallium/winsys/drm/intel/xorg/Makefile
index b1b6b9362b..d51cca8d21 100644
--- a/src/gallium/winsys/drm/intel/xorg/Makefile
+++ b/src/gallium/winsys/drm/intel/xorg/Makefile
@@ -1,36 +1,38 @@
 TARGET     = modesetting_drv.so
 CFILES     = $(wildcard ./*.c)
 OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
-GALLIUMDIR = ../../../..
 TOP        = ../../../../../..
 
-include ${TOP}/configs/current
+include $(TOP)/configs/current
 
-CFLAGS = -DHAVE_CONFIG_H \
-         -g -Wall -Wimplicit-function-declaration -fPIC \
-         $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \
-	 -I../gem \
-         -I${GALLIUMDIR}/include \
-         -I${GALLIUMDIR}/drivers \
-         -I${GALLIUMDIR}/auxiliary \
-         -I${TOP}/src/mesa \
-         -I$(TOP)/include \
-         -I$(TOP)/src/egl/main
+INCLUDES = \
+	$(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \
+	-I../gem \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/include \
+	-I$(TOP)/src/egl/main
 
 LIBS = \
-	$(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \
-	$(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \
+	$(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a \
+	$(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \
 	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
 	$(GALLIUM_AUXILIARIES)
 
+DRIVER_DEFINES = \
+	-DHAVE_CONFIG_H
+
+
 #############################################
 
 
 
 all default: $(TARGET)
 
-$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a $(LIBS)
+$(TARGET): $(OBJECTS) Makefile $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a $(LIBS)
 	$(TOP)/bin/mklib -noprefix -o $@ \
 	$(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel
 
@@ -39,6 +41,16 @@ clean:
 
 install:
 	$(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
+
+
+##############################################
+
+
+.c.o:
+	$(CC) -c $(CFLAGS) $(INCLUDES) $(DRIVER_DEFINES) $< -o $@
+
+
+##############################################
 
 .PHONY	= all clean install
diff --git a/src/gallium/winsys/drm/nouveau/Makefile b/src/gallium/winsys/drm/nouveau/Makefile
index f8c8135854..6c9cbef26d 100644
--- a/src/gallium/winsys/drm/nouveau/Makefile
+++ b/src/gallium/winsys/drm/nouveau/Makefile
@@ -2,7 +2,7 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
-SUBDIRS = drm dri dri2
+SUBDIRS = drm $(GALLIUM_STATE_TRACKERS_DIRS)
 
 default install clean:
 	@for dir in $(SUBDIRS) ; do \
diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile
index f7db6201fe..0937f68c34 100644
--- a/src/gallium/winsys/drm/nouveau/dri/Makefile
+++ b/src/gallium/winsys/drm/nouveau/dri/Makefile
@@ -3,32 +3,25 @@ include $(TOP)/configs/current
 
 LIBNAME = nouveau_dri.so
 
-MINIGLX_SOURCES =
-
 PIPE_DRIVERS = \
+	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
 	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
 	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
 	$(TOP)/src/gallium/drivers/nv20/libnv20.a \
 	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
 	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
-	$(TOP)/src/gallium/drivers/nv50/libnv50.a
-	
-DRIVER_SOURCES = \
-	nouveau_context.c \
-	nouveau_screen.c \
-	nouveau_swapbuffers.c \
-	nouveau_lock.c
+	$(TOP)/src/gallium/drivers/nv50/libnv50.a \
+	$(TOP)/src/gallium/drivers/nouveau/libnouveau.a
+
+DRIVER_SOURCES =
 
 C_SOURCES = \
 	$(COMMON_GALLIUM_SOURCES) \
 	$(DRIVER_SOURCES)
 
-ASM_SOURCES = 
+include ../../Makefile.template
 
-DRIVER_DEFINES = $(shell pkg-config libdrm_nouveau --cflags)
 DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
 
-include ../../Makefile.template
-
 symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c
deleted file mode 100644
index deb6ffcff1..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.c
+++ /dev/null
@@ -1,118 +0,0 @@
-#include <main/glheader.h>
-#include <glapi/glthread.h>
-#include <GL/internal/glcore.h>
-#include <utils.h>
-
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_context.h>
-#include <state_tracker/drm_api.h>
-#include <pipe/p_defines.h>
-#include <pipe/p_context.h>
-#include <pipe/p_screen.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-
-#include "nouveau_drmif.h"
-
-GLboolean
-nouveau_context_create(const __GLcontextModes *glVis,
-		       __DRIcontextPrivate *driContextPriv,
-		       void *sharedContextPrivate)
-{
-	__DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv;
-	struct nouveau_screen  *nv_screen = driScrnPriv->private;
-	struct nouveau_context *nv;
-	struct pipe_context *pipe;
-	struct st_context *st_share = NULL;
-
-	if (sharedContextPrivate)
-		st_share = ((struct nouveau_context *)sharedContextPrivate)->st;
-
-	nv = CALLOC_STRUCT(nouveau_context);
-	if (!nv)
-		return GL_FALSE;
-
-	{
-		struct nouveau_device_priv *nvdev =
-			nouveau_device(nv_screen->device);
-
-		nvdev->ctx  = driContextPriv->hHWContext;
-		nvdev->lock = (drmLock *)&driScrnPriv->pSAREA->lock;
-	}
-
-	pipe = drm_api_hooks.create_context(nv_screen->pscreen);
-	if (!pipe) {
-		FREE(nv);
-		return GL_FALSE;
-	}
-	pipe->priv = nv;
-
-	driContextPriv->driverPrivate = nv;
-	nv->dri_screen = driScrnPriv;
-
-	driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache,
-			    nv->dri_screen->myNum, "nouveau");
-
-	nv->st = st_create_context(pipe, glVis, st_share);
-	return GL_TRUE;
-}
-
-void
-nouveau_context_destroy(__DRIcontextPrivate *driContextPriv)
-{
-	struct nouveau_context *nv = driContextPriv->driverPrivate;
-
-	assert(nv);
-
-	st_finish(nv->st);
-	st_destroy_context(nv->st);
-
-	FREE(nv);
-}
-
-GLboolean
-nouveau_context_bind(__DRIcontextPrivate *driContextPriv,
-		     __DRIdrawablePrivate *driDrawPriv,
-		     __DRIdrawablePrivate *driReadPriv)
-{
-	struct nouveau_context *nv;
-	struct nouveau_framebuffer *draw, *read;
-
-	if (!driContextPriv) {
-		st_make_current(NULL, NULL, NULL);
-		return GL_TRUE;
-	}
-
-	nv = driContextPriv->driverPrivate;
-	draw = driDrawPriv->driverPrivate;
-	read = driReadPriv->driverPrivate;
-
-	st_make_current(nv->st, draw->stfb, read->stfb);
-
-	if ((nv->dri_drawable != driDrawPriv) ||
-	    (nv->last_stamp != driDrawPriv->lastStamp)) {
-		nv->dri_drawable = driDrawPriv;
-		st_resize_framebuffer(draw->stfb, driDrawPriv->w,
-				      driDrawPriv->h);
-		nv->last_stamp = driDrawPriv->lastStamp;
-	}
-
-	if (driDrawPriv != driReadPriv) {
-		st_resize_framebuffer(read->stfb, driReadPriv->w,
-				      driReadPriv->h);
-	}
-
-	return GL_TRUE;
-}
-
-GLboolean
-nouveau_context_unbind(__DRIcontextPrivate *driContextPriv)
-{
-	struct nouveau_context *nv = driContextPriv->driverPrivate;
-	(void)nv;
-
-	st_flush(nv->st, 0, NULL);
-	return GL_TRUE;
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h
deleted file mode 100644
index 2779b092e6..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_context.h
+++ /dev/null
@@ -1,53 +0,0 @@
-#ifndef __NOUVEAU_CONTEXT_DRI_H__
-#define __NOUVEAU_CONTEXT_DRI_H__
-
-#include <dri_util.h>
-#include <xmlconfig.h>
-
-#include "nouveau/nouveau_winsys.h"
-
-#define NOUVEAU_ERR(fmt, args...) debug_printf("%s: "fmt, __func__, ##args)
-
-struct nouveau_framebuffer {
-	struct st_framebuffer *stfb;
-};
-
-struct nouveau_context {
-	struct st_context *st;
-
-	/* DRI stuff */
-	__DRIscreenPrivate    *dri_screen;
-	__DRIdrawablePrivate  *dri_drawable;
-	unsigned int           last_stamp;
-	driOptionCache         dri_option_cache;
-	drm_context_t          drm_context;
-	drmLock                drm_lock;
-	int                    locked;
-};
-
-extern GLboolean nouveau_context_create(const __GLcontextModes *,
-					__DRIcontextPrivate *, void *);
-extern void nouveau_context_destroy(__DRIcontextPrivate *);
-extern GLboolean nouveau_context_bind(__DRIcontextPrivate *,
-				      __DRIdrawablePrivate *draw,
-				      __DRIdrawablePrivate *read);
-extern GLboolean nouveau_context_unbind(__DRIcontextPrivate *);
-
-extern void nouveau_contended_lock(struct nouveau_context *nv);
-extern void LOCK_HARDWARE(struct nouveau_context *nv);
-extern void UNLOCK_HARDWARE(struct nouveau_context *nv);
-
-#ifdef DEBUG
-extern int __nouveau_debug;
-
-#define DEBUG_BO (1 << 0)
-
-#define DBG(flag, ...) do {                   \
-	if (__nouveau_debug & (DEBUG_##flag)) \
-		NOUVEAU_ERR(__VA_ARGS__);     \
-} while(0)
-#else
-#define DBG(flag, ...)
-#endif
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c
deleted file mode 100644
index 4e9b76a909..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.c
+++ /dev/null
@@ -1,330 +0,0 @@
-#include <utils.h>
-#include <vblank.h>
-#include <xmlpool.h>
-
-#include <pipe/p_context.h>
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_cb_fbo.h>
-#include <state_tracker/drm_api.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_swapbuffers.h"
-#include "nouveau_dri.h"
-
-#include "nouveau_drm.h"
-#include "nouveau_drmif.h"
-
-#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12
-#error nouveau_drm.h version does not match expected version
-#endif
-
-/* Extension stuff, enabling of extensions handled by Gallium's GL state
- * tracker.  But, we still need to define the entry points we want.
- */
-#define need_GL_ARB_fragment_program
-#define need_GL_ARB_multisample
-#define need_GL_ARB_occlusion_query
-#define need_GL_ARB_point_parameters
-#define need_GL_ARB_shader_objects
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_program
-#define need_GL_ARB_vertex_shader
-#define need_GL_ARB_vertex_buffer_object
-#define need_GL_EXT_compiled_vertex_array
-#define need_GL_EXT_fog_coord
-#define need_GL_EXT_secondary_color
-#define need_GL_EXT_framebuffer_object
-#define need_GL_VERSION_2_0
-#define need_GL_VERSION_2_1
-#include "extension_helper.h"
-
-const struct dri_extension card_extensions[] =
-{
-	{ "GL_ARB_multisample", GL_ARB_multisample_functions },
-	{ "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions },
-	{ "GL_ARB_point_parameters", GL_ARB_point_parameters_functions },
-	{ "GL_ARB_shader_objects", GL_ARB_shader_objects_functions },
-	{ "GL_ARB_shading_language_100", GL_VERSION_2_0_functions },
-	{ "GL_ARB_shading_language_120", GL_VERSION_2_1_functions },
-	{ "GL_ARB_texture_compression", GL_ARB_texture_compression_functions },
-	{ "GL_ARB_vertex_program", GL_ARB_vertex_program_functions },
-	{ "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions },
-	{ "GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions },
-	{ "GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions },
-	{ "GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
-	{ "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
-	{ "GL_EXT_secondary_color", GL_EXT_secondary_color_functions },
-	{ NULL, 0 }
-};
-
-PUBLIC const char __driConfigOptions[] =
-DRI_CONF_BEGIN
-DRI_CONF_END;
-static const GLuint __driNConfigOptions = 0;
-
-extern const struct dri_extension common_extensions[];
-extern const struct dri_extension nv40_extensions[];
-
-static GLboolean
-nouveau_create_buffer(__DRIscreenPrivate * driScrnPriv,
-		      __DRIdrawablePrivate * driDrawPriv,
-		      const __GLcontextModes *glVis, GLboolean pixmapBuffer)
-{
-	struct nouveau_framebuffer *nvfb;
-	enum pipe_format colour, depth, stencil;
-
-	if (pixmapBuffer)
-		return GL_FALSE;
-
-	nvfb = CALLOC_STRUCT(nouveau_framebuffer);
-	if (!nvfb)
-		return GL_FALSE;
-
-	if (glVis->redBits == 5)
-		colour = PIPE_FORMAT_R5G6B5_UNORM;
-	else
-		colour = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-	if (glVis->depthBits == 16)
-		depth = PIPE_FORMAT_Z16_UNORM;
-	else if (glVis->depthBits == 24)
-		depth = PIPE_FORMAT_Z24S8_UNORM;
-	else
-		depth = PIPE_FORMAT_NONE;
-
-	if (glVis->stencilBits == 8)
-		stencil = PIPE_FORMAT_Z24S8_UNORM;
-	else
-		stencil = PIPE_FORMAT_NONE;
-
-	nvfb->stfb = st_create_framebuffer(glVis, colour, depth, stencil,
-					   driDrawPriv->w, driDrawPriv->h,
-					   (void*)nvfb);
-	if (!nvfb->stfb) {
-		free(nvfb);
-		return  GL_FALSE;
-	}
-
-	driDrawPriv->driverPrivate = (void *)nvfb;
-	return GL_TRUE;
-}
-
-static void
-nouveau_destroy_buffer(__DRIdrawablePrivate * driDrawPriv)
-{
-	struct nouveau_framebuffer *nvfb;
-	
-	nvfb = (struct nouveau_framebuffer *)driDrawPriv->driverPrivate;
-	st_unreference_framebuffer(nvfb->stfb);
-	free(nvfb);
-}
-
-static __DRIconfig **
-nouveau_fill_in_modes(__DRIscreenPrivate *psp,
-		      unsigned pixel_bits, unsigned depth_bits,
-		      unsigned stencil_bits, GLboolean have_back_buffer)
-{
-	__DRIconfig **configs;
-	unsigned depth_buffer_factor;
-	unsigned back_buffer_factor;
-	GLenum fb_format;
-	GLenum fb_type;
-
-	static const GLenum back_buffer_modes[] = {
-		GLX_NONE, GLX_SWAP_UNDEFINED_OML,
-	};
-
-	uint8_t depth_bits_array[3];
-	uint8_t stencil_bits_array[3];
-	uint8_t msaa_samples_array[1];
-
-	depth_bits_array[0] = 0;
-	depth_bits_array[1] = depth_bits;
-	depth_bits_array[2] = depth_bits;
-
-	/* Just like with the accumulation buffer, always provide some modes
-	 * with a stencil buffer.  It will be a sw fallback, but some apps won't
-	 * care about that.
-	 */
-	stencil_bits_array[0] = 0;
-	stencil_bits_array[1] = 0;
-	if (depth_bits == 24)
-		stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
-	stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
-
-	msaa_samples_array[0] = 0;
-
-	depth_buffer_factor =
-		((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
-	back_buffer_factor = (have_back_buffer) ? 3 : 1;
-
-	if (pixel_bits == 16) {
-		fb_format = GL_RGB;
-		fb_type = GL_UNSIGNED_SHORT_5_6_5;
-	}
-	else {
-		fb_format = GL_BGRA;
-		fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
-	}
-
-	configs = driCreateConfigs(fb_format, fb_type,
-				   depth_bits_array, stencil_bits_array,
-				   depth_buffer_factor, back_buffer_modes,
-				   back_buffer_factor, msaa_samples_array, 1);
-	if (configs == NULL) {
-	 fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
-			 __func__, __LINE__);
-		return NULL;
-	}
-
-	return configs;
-}
-
-static struct pipe_surface *
-dri_surface_from_handle(struct pipe_screen *screen,
-                        unsigned handle,
-                        enum pipe_format format,
-                        unsigned width,
-                        unsigned height,
-                        unsigned pitch)
-{
-   struct pipe_surface *surface = NULL;
-   struct pipe_texture *texture = NULL;
-   struct pipe_texture templat;
-   struct pipe_buffer *buf = NULL;
-
-   buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle);
-   if (!buf)
-      return NULL;
-
-   memset(&templat, 0, sizeof(templat));
-   templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY |
-                       NOUVEAU_TEXTURE_USAGE_LINEAR;
-   templat.target = PIPE_TEXTURE_2D;
-   templat.last_level = 0;
-   templat.depth[0] = 1;
-   templat.format = format;
-   templat.width[0] = width;
-   templat.height[0] = height;
-   pf_get_block(templat.format, &templat.block);
-
-   texture = screen->texture_blanket(screen,
-                                     &templat,
-                                     &pitch,
-                                     buf);
-
-   /* we don't need the buffer from this point on */
-   pipe_buffer_reference(&buf, NULL);
-
-   if (!texture)
-      return NULL;
-
-   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
-                                     PIPE_BUFFER_USAGE_GPU_READ |
-                                     PIPE_BUFFER_USAGE_GPU_WRITE);
-
-   /* we don't need the texture from this point on */
-   pipe_texture_reference(&texture, NULL);
-   return surface;
-}
-
-static const __DRIconfig **
-nouveau_screen_create(__DRIscreenPrivate *psp)
-{
-	struct nouveau_dri *nv_dri = psp->pDevPriv;
-	struct nouveau_screen *nv_screen;
-	static const __DRIversion ddx_expected =
-		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-	static const __DRIversion dri_expected = { 4, 0, 0 };
-	static const __DRIversion drm_expected =
-		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-
-	if (!driCheckDriDdxDrmVersions2("nouveau",
-					&psp->dri_version, &dri_expected,
-					&psp->ddx_version, &ddx_expected,
-					&psp->drm_version, &drm_expected)) {
-		return NULL;
-	}
-
-	if (drm_expected.patch != psp->drm_version.patch) {
-		fprintf(stderr, "Incompatible DRM patch level.\n"
-				"Expected: %d\n" "Current : %d\n",
-			drm_expected.patch, psp->drm_version.patch);
-		return NULL;
-	}
-
-	driInitExtensions(NULL, card_extensions, GL_FALSE);
-
-	if (psp->devPrivSize != sizeof(struct nouveau_dri)) {
-		NOUVEAU_ERR("DRI struct mismatch between DDX/DRI\n");
-		return NULL;
-	}
-
-	nv_screen = CALLOC_STRUCT(nouveau_screen);
-	if (!nv_screen)
-		return NULL;
-
-	nouveau_device_open_existing(&nv_screen->device, 0, psp->fd, 0);
-
-	nv_screen->pscreen = drm_api_hooks.create_screen(psp->fd, NULL);
-	if (!nv_screen->pscreen) {
-		FREE(nv_screen);
-		return NULL;
-	}
-	nv_screen->pscreen->flush_frontbuffer = nouveau_flush_frontbuffer;
-
-	{
-		enum pipe_format format;
-
-		if (nv_dri->bpp == 16)
-			format = PIPE_FORMAT_R5G6B5_UNORM;
-		else
-			format = PIPE_FORMAT_A8R8G8B8_UNORM;
-
-		nv_screen->fb = dri_surface_from_handle(nv_screen->pscreen,
-							nv_dri->front_offset,
-							format,
-							nv_dri->width,
-							nv_dri->height,
-							nv_dri->front_pitch *
-							nv_dri->bpp / 8);
-	}
-						
-	driParseOptionInfo(&nv_screen->option_cache,
-			   __driConfigOptions, __driNConfigOptions);
-
-	nv_screen->driScrnPriv = psp;
-	psp->private = (void *)nv_screen;
-
-	return (const __DRIconfig **)
-		nouveau_fill_in_modes(psp, nv_dri->bpp,
-				      (nv_dri->bpp == 16) ? 16 : 24,
-				      (nv_dri->bpp == 16) ? 0 : 8, 1);
-}
-
-static void
-nouveau_screen_destroy(__DRIscreenPrivate *driScrnPriv)
-{
-	struct nouveau_screen *nv_screen = driScrnPriv->private;
-
-	driScrnPriv->private = NULL;
-	FREE(nv_screen);
-}
-
-const struct __DriverAPIRec
-driDriverAPI = {
-	.InitScreen	= nouveau_screen_create,
-	.DestroyScreen	= nouveau_screen_destroy,
-	.CreateContext	= nouveau_context_create,
-	.DestroyContext	= nouveau_context_destroy,
-	.CreateBuffer	= nouveau_create_buffer,
-	.DestroyBuffer	= nouveau_destroy_buffer,
-	.SwapBuffers	= nouveau_swap_buffers,
-	.MakeCurrent	= nouveau_context_bind,
-	.UnbindContext	= nouveau_context_unbind,
-	.CopySubBuffer	= nouveau_copy_sub_buffer,
-
-	.InitScreen2	= NULL, /* one day, I promise! */
-};
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h
deleted file mode 100644
index ac078f3c63..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen.h
+++ /dev/null
@@ -1,16 +0,0 @@
-#ifndef __NOUVEAU_SCREEN_DRI_H__
-#define __NOUVEAU_SCREEN_DRI_H__
-
-#include "xmlconfig.h"
-
-struct nouveau_screen {
-	__DRIscreenPrivate *driScrnPriv;
-	driOptionCache      option_cache;
-
-	struct nouveau_device *device;
-
-	struct pipe_screen *pscreen;
-	struct pipe_surface *fb;
-};
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
deleted file mode 100644
index 9c841a0b2d..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
+++ /dev/null
@@ -1,115 +0,0 @@
-#include <main/glheader.h>
-#include <glapi/glthread.h>
-#include <GL/internal/glcore.h>
-
-#include <pipe/p_context.h>
-#include <state_tracker/st_public.h>
-#include <state_tracker/st_context.h>
-#include <state_tracker/st_cb_fbo.h>
-
-#include "nouveau_context.h"
-#include "nouveau_screen.h"
-#include "nouveau_swapbuffers.h"
-
-#include "nouveau_pushbuf.h"
-
-void
-nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf,
-		    const drm_clip_rect_t *rect)
-{
-	struct nouveau_context *nv = dPriv->driContextPriv->driverPrivate;
-	struct nouveau_screen *nv_screen = nv->dri_screen->private;
-	struct pipe_context *pipe = nv->st->pipe;
-	drm_clip_rect_t *pbox;
-	int nbox, i;
-
-	LOCK_HARDWARE(nv);
-	if (!dPriv->numClipRects) {
-		UNLOCK_HARDWARE(nv);
-		return;
-	}
-	pbox = dPriv->pClipRects;
-	nbox = dPriv->numClipRects;
-
-	for (i = 0; i < nbox; i++, pbox++) {
-		int sx, sy, dx, dy, w, h;
-
-		sx = pbox->x1 - dPriv->x;
-		sy = pbox->y1 - dPriv->y;
-		dx = pbox->x1;
-		dy = pbox->y1;
-		w  = pbox->x2 - pbox->x1;
-		h  = pbox->y2 - pbox->y1;
-
-		pipe->surface_copy(pipe, nv_screen->fb, dx, dy, surf,
-				   sx, sy, w, h);
-	}
-
-	pipe->flush(pipe, 0, NULL);
-	UNLOCK_HARDWARE(nv);
-
-	if (nv->last_stamp != dPriv->lastStamp) {
-		struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-		st_resize_framebuffer(nvfb->stfb, dPriv->w, dPriv->h);
-		nv->last_stamp = dPriv->lastStamp;
-	}
-}
-
-void
-nouveau_copy_sub_buffer(__DRIdrawablePrivate *dPriv, int x, int y, int w, int h)
-{
-	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-	struct pipe_surface *surf;
-
-	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
-	if (surf) {
-		drm_clip_rect_t rect;
-		rect.x1 = x;
-		rect.y1 = y;
-		rect.x2 = x + w;
-		rect.y2 = y + h;
-
-		st_notify_swapbuffers(nvfb->stfb);
-		nouveau_copy_buffer(dPriv, surf, &rect);
-	}
-}
-
-void
-nouveau_swap_buffers(__DRIdrawablePrivate *dPriv)
-{
-	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
-	struct pipe_surface *surf;
-
-	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
-	if (surf) {
-		st_notify_swapbuffers(nvfb->stfb);
-		nouveau_copy_buffer(dPriv, surf, NULL);
-	}
-}
-
-void
-nouveau_flush_frontbuffer(struct pipe_screen *pscreen, struct pipe_surface *ps,
-			  void *context_private)
-{
-	struct nouveau_context *nv = context_private;
-	__DRIdrawablePrivate *dPriv = nv->dri_drawable;
-
-	nouveau_copy_buffer(dPriv, ps, NULL);
-}
-
-void
-nouveau_contended_lock(struct nouveau_context *nv)
-{
-	struct nouveau_context *nv_sub = (struct nouveau_context*)nv;
-	__DRIdrawablePrivate *dPriv = nv_sub->dri_drawable;
-	__DRIscreenPrivate *sPriv = nv_sub->dri_screen;
-
-	/* If the window moved, may need to set a new cliprect now.
-	 *
-	 * NOTE: This releases and regains the hw lock, so all state
-	 * checking must be done *after* this call:
-	 */
-	if (dPriv)
-		DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
deleted file mode 100644
index 4ca9cc2283..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
+++ /dev/null
@@ -1,11 +0,0 @@
-#ifndef __NOUVEAU_SWAPBUFFERS_H__
-#define __NOUVEAU_SWAPBUFFERS_H__
-
-void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *,
-			 const drm_clip_rect_t *);
-void nouveau_copy_sub_buffer(__DRIdrawablePrivate *, int x, int y, int w, int h);
-void nouveau_swap_buffers(__DRIdrawablePrivate *);
-void nouveau_flush_frontbuffer(struct pipe_screen *, struct pipe_surface *,
-			       void *context_private);
-
-#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri2/Makefile b/src/gallium/winsys/drm/nouveau/dri2/Makefile
deleted file mode 100644
index 377a80d518..0000000000
--- a/src/gallium/winsys/drm/nouveau/dri2/Makefile
+++ /dev/null
@@ -1,26 +0,0 @@
-TOP = ../../../../../..
-include $(TOP)/configs/current
-
-LIBNAME = nouveau_dri2.so
-
-PIPE_DRIVERS = \
-	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
-	$(TOP)/src/gallium/winsys/drm/nouveau/drm/libnouveaudrm.a \
-	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
-	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
-	$(TOP)/src/gallium/drivers/nv20/libnv20.a \
-	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
-	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
-	$(TOP)/src/gallium/drivers/nv50/libnv50.a
-
-DRIVER_SOURCES =
-
-C_SOURCES = \
-	$(COMMON_GALLIUM_SOURCES) \
-	$(DRIVER_SOURCES)
-
-include ../../Makefile.template
-
-DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
-
-symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/drm/Makefile b/src/gallium/winsys/drm/nouveau/drm/Makefile
index 2da78d8690..54c3b26c75 100644
--- a/src/gallium/winsys/drm/nouveau/drm/Makefile
+++ b/src/gallium/winsys/drm/nouveau/drm/Makefile
@@ -3,9 +3,7 @@ include $(TOP)/configs/current
 
 LIBNAME = nouveaudrm
 
-C_SOURCES = nouveau_drm_api.c \
-	    nouveau_winsys_pipe.c \
-	    nouveau_winsys.c
+C_SOURCES = nouveau_drm_api.c
 
 LIBRARY_INCLUDES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-I)
 LIBRARY_DEFINES = $(shell pkg-config libdrm libdrm_nouveau --cflags-only-other)
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h
index 1207c2d609..1207c2d609 100644
--- a/src/gallium/winsys/drm/nouveau/dri/nouveau_dri.h
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_dri.h
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index a558fda140..3167ccfdc1 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -1,20 +1,84 @@
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
 #include "util/u_memory.h"
 
 #include "nouveau_drm_api.h"
-#include "nouveau_winsys_pipe.h"
 
 #include "nouveau_drmif.h"
 #include "nouveau_channel.h"
 #include "nouveau_bo.h"
 
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_screen.h"
+
+static struct pipe_surface *
+dri_surface_from_handle(struct drm_api *api, struct pipe_screen *screen,
+                        unsigned handle,
+                        enum pipe_format format,
+                        unsigned width,
+                        unsigned height,
+                        unsigned pitch)
+{
+   struct pipe_surface *surface = NULL;
+   struct pipe_texture *texture = NULL;
+   struct pipe_texture templat;
+   struct pipe_buffer *buf = NULL;
+
+   buf = api->buffer_from_handle(api, screen, "front buffer", handle);
+   if (!buf)
+      return NULL;
+
+   memset(&templat, 0, sizeof(templat));
+   templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY |
+                       NOUVEAU_TEXTURE_USAGE_LINEAR;
+   templat.target = PIPE_TEXTURE_2D;
+   templat.last_level = 0;
+   templat.depth[0] = 1;
+   templat.format = format;
+   templat.width[0] = width;
+   templat.height[0] = height;
+   pf_get_block(templat.format, &templat.block);
+
+   texture = screen->texture_blanket(screen,
+                                     &templat,
+                                     &pitch,
+                                     buf);
+
+   /* we don't need the buffer from this point on */
+   pipe_buffer_reference(&buf, NULL);
+
+   if (!texture)
+      return NULL;
+
+   surface = screen->get_tex_surface(screen, texture, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_READ |
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* we don't need the texture from this point on */
+   pipe_texture_reference(&texture, NULL);
+   return surface;
+}
+
+static struct pipe_surface *
+nouveau_dri1_front_surface(struct pipe_context *pipe)
+{
+	return nouveau_winsys_screen(pipe->screen)->front;
+}
+
+static struct dri1_api nouveau_dri1_api = {
+	nouveau_dri1_front_surface,
+};
+
 static struct pipe_screen *
-nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg)
+nouveau_drm_create_screen(struct drm_api *api, int fd,
+			  struct drm_create_screen_arg *arg)
 {
-	struct pipe_winsys *ws;
+	struct dri1_create_screen_arg *dri1 = (void *)arg;
 	struct nouveau_winsys *nvws;
+	struct pipe_winsys *ws;
 	struct nouveau_device *dev = NULL;
 	struct pipe_screen *(*init)(struct pipe_winsys *,
-				    struct nouveau_winsys *);
+				    struct nouveau_device *);
 	int ret;
 
 	ret = nouveau_device_open_existing(&dev, 0, fd, 0);
@@ -49,33 +113,53 @@ nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg)
 		return NULL;
 	}
 
-	ws = nouveau_pipe_winsys_new(dev);
-	if (!ws) {
+	nvws = CALLOC_STRUCT(nouveau_winsys);
+	if (!nvws) {
 		nouveau_device_close(&dev);
 		return NULL;
 	}
+	ws = &nvws->base;
 
-	nvws = nouveau_winsys_new(ws);
-	if (!nvws) {
+	nvws->pscreen = init(ws, dev);
+	if (!nvws->pscreen) {
 		ws->destroy(ws);
 		return NULL;
 	}
 
-	nouveau_pipe_winsys(ws)->pscreen = init(ws, nvws);
-	if (!nouveau_pipe_winsys(ws)->pscreen) {
-		ws->destroy(ws);
-		return NULL;
+	if (arg->mode == DRM_CREATE_DRI1) {
+		struct nouveau_dri *nvdri = dri1->ddx_info;
+		enum pipe_format format;
+
+		if (nvdri->bpp == 16)
+			format = PIPE_FORMAT_R5G6B5_UNORM;
+		else
+			format = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+		nvws->front = dri_surface_from_handle(api, nvws->pscreen,
+						      nvdri->front_offset,
+						      format, nvdri->width,
+						      nvdri->height,
+						      nvdri->front_pitch *
+						      (nvdri->bpp / 8));
+		if (!nvws->front) {
+			debug_printf("%s: error referencing front buffer\n",
+				     __func__);
+			ws->destroy(ws);
+			return NULL;
+		}
+
+		dri1->api = &nouveau_dri1_api;
 	}
 
-	return nouveau_pipe_winsys(ws)->pscreen;
+	return nvws->pscreen;
 }
 
 static struct pipe_context *
-nouveau_drm_create_context(struct pipe_screen *pscreen)
+nouveau_drm_create_context(struct drm_api *api, struct pipe_screen *pscreen)
 {
-	struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen);
+	struct nouveau_winsys *nvws = nouveau_winsys_screen(pscreen);
 	struct pipe_context *(*init)(struct pipe_screen *, unsigned);
-	unsigned chipset = nvpws->channel->device->chipset;
+	unsigned chipset = nouveau_screen(pscreen)->device->chipset;
 	int i;
 
 	switch (chipset & 0xf0) {
@@ -106,81 +190,80 @@ nouveau_drm_create_context(struct pipe_screen *pscreen)
 	}
 
 	/* Find a free slot for a pipe context, allocate a new one if needed */
-	for (i = 0; i < nvpws->nr_pctx; i++) {
-		if (nvpws->pctx[i] == NULL)
+	for (i = 0; i < nvws->nr_pctx; i++) {
+		if (nvws->pctx[i] == NULL)
 			break;
 	}
 
-	if (i == nvpws->nr_pctx) {
-		nvpws->nr_pctx++;
-		nvpws->pctx = realloc(nvpws->pctx,
-				      sizeof(*nvpws->pctx) * nvpws->nr_pctx);
+	if (i == nvws->nr_pctx) {
+		nvws->nr_pctx++;
+		nvws->pctx = realloc(nvws->pctx,
+				      sizeof(*nvws->pctx) * nvws->nr_pctx);
 	}
 
-	nvpws->pctx[i] = init(pscreen, i);
-	return nvpws->pctx[i];
+	nvws->pctx[i] = init(pscreen, i);
+	return nvws->pctx[i];
 }
 
 static boolean
-nouveau_drm_pb_from_pt(struct pipe_texture *pt, struct pipe_buffer **ppb,
-		       unsigned *stride)
+nouveau_drm_pb_from_pt(struct drm_api *api, struct pipe_texture *pt,
+		       struct pipe_buffer **ppb, unsigned *stride)
 {
 	return false;
 }
 
 static struct pipe_buffer *
-nouveau_drm_pb_from_handle(struct pipe_screen *pscreen, const char *name,
-			   unsigned handle)
+nouveau_drm_pb_from_handle(struct drm_api *api, struct pipe_screen *pscreen,
+			   const char *name, unsigned handle)
 {
-	struct nouveau_pipe_winsys *nvpws = nouveau_screen(pscreen);
-	struct nouveau_device *dev = nvpws->channel->device;
-	struct nouveau_pipe_buffer *nvpb;
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct pipe_buffer *pb;
 	int ret;
 
-	nvpb = CALLOC_STRUCT(nouveau_pipe_buffer);
-	if (!nvpb)
+	pb = CALLOC(1, sizeof(struct pipe_buffer) + sizeof(struct nouveau_bo*));
+	if (!pb)
 		return NULL;
 
-	ret = nouveau_bo_handle_ref(dev, handle, &nvpb->bo);
+	ret = nouveau_bo_handle_ref(dev, handle, (struct nouveau_bo**)(pb+1));
 	if (ret) {
 		debug_printf("%s: ref name 0x%08x failed with %d\n",
 			     __func__, handle, ret);
-		FREE(nvpb);
+		FREE(pb);
 		return NULL;
 	}
 
-	pipe_reference_init(&nvpb->base.reference, 1);
-	nvpb->base.screen = pscreen;
-	nvpb->base.alignment = 0;
-	nvpb->base.usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE |
-			   PIPE_BUFFER_USAGE_CPU_READ_WRITE;
-	nvpb->base.size = nvpb->bo->size;
-	return &nvpb->base;
+	pipe_reference_init(&pb->reference, 1);
+	pb->screen = pscreen;
+	pb->alignment = 0;
+	pb->usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE |
+		    PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+	pb->size = nouveau_bo(pb)->size;
+	return pb;
 }
 
 static boolean
-nouveau_drm_handle_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb,
-			   unsigned *handle)
+nouveau_drm_handle_from_pb(struct drm_api *api, struct pipe_screen *pscreen,
+			   struct pipe_buffer *pb, unsigned *handle)
 {
-	struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb);
+	struct nouveau_bo *bo = nouveau_bo(pb);
 
-	if (!nvpb)
+	if (!bo)
 		return FALSE;
 
-	*handle = nvpb->bo->handle;
+	*handle = bo->handle;
 	return TRUE;
 }
 
 static boolean
-nouveau_drm_name_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb,
-			 unsigned *handle)
+nouveau_drm_name_from_pb(struct drm_api *api, struct pipe_screen *pscreen,
+			 struct pipe_buffer *pb, unsigned *handle)
 {
-	struct nouveau_pipe_buffer *nvpb = nouveau_pipe_buffer(pb);
+	struct nouveau_bo *bo = nouveau_bo(pb);
 
-	if (!nvpb)
+	if (!bo)
 		return FALSE;
 
-	return nouveau_bo_handle_get(nvpb->bo, handle) == 0;
+	return nouveau_bo_handle_get(bo, handle) == 0;
 }
 
 struct drm_api drm_api_hooks = {
@@ -192,3 +275,8 @@ struct drm_api drm_api_hooks = {
 	.global_handle_from_buffer = nouveau_drm_name_from_pb,
 };
 
+struct drm_api *
+drm_api_create() {
+	return &drm_api_hooks;
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
index 2782c83c0e..e61e0e0957 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.h
@@ -1,5 +1,34 @@
 #ifndef __NOUVEAU_DRM_API_H__
 #define __NOUVEAU_DRM_API_H__
+
 #include "state_tracker/drm_api.h"
+#include "state_tracker/dri1_api.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nouveau_dri.h"
+
+struct nouveau_winsys {
+	struct pipe_winsys base;
+
+	struct pipe_screen *pscreen;
+
+	unsigned nr_pctx;
+	struct pipe_context **pctx;
+
+	struct pipe_surface *front;
+};
+
+static INLINE struct nouveau_winsys *
+nouveau_winsys(struct pipe_winsys *ws)
+{
+	return (struct nouveau_winsys *)ws;
+}
+
+static INLINE struct nouveau_winsys *
+nouveau_winsys_screen(struct pipe_screen *pscreen)
+{
+	return nouveau_winsys(pscreen->winsys);
+}
 
 #endif
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c
deleted file mode 100644
index e3175fd775..0000000000
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys.c
+++ /dev/null
@@ -1,94 +0,0 @@
-#include "util/u_memory.h"
-
-#include "nouveau_winsys_pipe.h"
-
-static int
-nouveau_pipe_notifier_alloc(struct nouveau_winsys *nvws, int count,
-			    struct nouveau_notifier **notify)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws);
-
-	return nouveau_notifier_alloc(nvpws->channel, nvpws->next_handle++,
-				      count, notify);
-}
-
-static int
-nouveau_pipe_grobj_alloc(struct nouveau_winsys *nvws, int grclass,
-			 struct nouveau_grobj **grobj)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(nvws->ws);
-	struct nouveau_channel *chan = nvpws->channel;
-	int ret;
-
-	ret = nouveau_grobj_alloc(chan, nvpws->next_handle++, grclass, grobj);
-	if (ret)
-		return ret;
-
-	BEGIN_RING(chan, *grobj, 0x0000, 1);
-	OUT_RING  (chan, (*grobj)->handle);
-	(*grobj)->bound = NOUVEAU_GROBJ_BOUND_EXPLICIT;
-	return 0;
-}
-
-static int
-nouveau_pipe_push_reloc(struct nouveau_winsys *nvws, void *ptr,
-			struct pipe_buffer *buf, uint32_t data,
-			uint32_t flags, uint32_t vor, uint32_t tor)
-{
-	struct nouveau_bo *bo = nouveau_pipe_buffer(buf)->bo;
-
-	return nouveau_pushbuf_emit_reloc(nvws->channel, ptr, bo,
-					  data, flags, vor, tor);
-}
-
-static int
-nouveau_pipe_push_flush(struct nouveau_winsys *nvws, unsigned size,
-			struct pipe_fence_handle **fence)
-{
-	if (fence)
-		*fence = NULL;
-
-	return nouveau_pushbuf_flush(nvws->channel, size);
-}
-
-static struct nouveau_bo *
-nouveau_pipe_get_bo(struct pipe_buffer *pb)
-{
-	return nouveau_pipe_buffer(pb)->bo;
-}
-
-struct nouveau_winsys *
-nouveau_winsys_new(struct pipe_winsys *ws)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct nouveau_winsys *nvws;
-
-	nvws = CALLOC_STRUCT(nouveau_winsys);
-	if (!nvws)
-		return NULL;
-
-	nvws->ws		= ws;
-	nvws->channel		= nvpws->channel;
-
-	nvws->res_init		= nouveau_resource_init;
-	nvws->res_alloc		= nouveau_resource_alloc;
-	nvws->res_free		= nouveau_resource_free;
-
-	nvws->push_reloc        = nouveau_pipe_push_reloc;
-	nvws->push_flush	= nouveau_pipe_push_flush;
-
-	nvws->grobj_alloc	= nouveau_pipe_grobj_alloc;
-	nvws->grobj_free	= nouveau_grobj_free;
-
-	nvws->notifier_alloc	= nouveau_pipe_notifier_alloc;
-	nvws->notifier_free	= nouveau_notifier_free;
-	nvws->notifier_reset	= nouveau_notifier_reset;
-	nvws->notifier_status	= nouveau_notifier_status;
-	nvws->notifier_retval	= nouveau_notifier_return_val;
-	nvws->notifier_wait	= nouveau_notifier_wait_status;
-
-	nvws->get_bo		= nouveau_pipe_get_bo;
-
-	return nvws;
-}
-
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c
deleted file mode 100644
index 9e03a9f5db..0000000000
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.c
+++ /dev/null
@@ -1,204 +0,0 @@
-#include "pipe/internal/p_winsys_screen.h"
-#include <pipe/p_defines.h>
-#include <pipe/p_inlines.h>
-#include <util/u_memory.h>
-
-#include "nouveau_winsys_pipe.h"
-
-#include "nouveau_drmif.h"
-#include "nouveau_bo.h"
-
-static const char *
-nouveau_get_name(struct pipe_winsys *pws)
-{
-	return "Nouveau/DRI";
-}
-
-static uint32_t
-nouveau_flags_from_usage(struct pipe_winsys *ws, unsigned usage)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct pipe_screen *pscreen = nvpws->pscreen;
-	uint32_t flags = NOUVEAU_BO_LOCAL;
-
-	if (usage & NOUVEAU_BUFFER_USAGE_TRANSFER)
-		flags |= NOUVEAU_BO_GART;
-
-	if (usage & PIPE_BUFFER_USAGE_PIXEL) {
-		if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE)
-			flags |= NOUVEAU_BO_GART;
-		if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE))
-			flags |= NOUVEAU_BO_VRAM;
-
-		switch (nvpws->channel->device->chipset & 0xf0) {
-		case 0x50:
-		case 0x80:
-		case 0x90:
-			flags |= NOUVEAU_BO_TILED;
-			if (usage & NOUVEAU_BUFFER_USAGE_ZETA)
-				flags |= NOUVEAU_BO_ZTILE;
-			break;
-		default:
-			break;
-		}
-	}
-
-	if (usage & PIPE_BUFFER_USAGE_VERTEX) {
-		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF))
-			flags |= NOUVEAU_BO_GART;
-	}
-
-	if (usage & PIPE_BUFFER_USAGE_INDEX) {
-		if (pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF))
-			flags |= NOUVEAU_BO_GART;
-	}
-
-	return flags;
-}
-
-static struct pipe_buffer *
-nouveau_pipe_bo_create(struct pipe_winsys *ws, unsigned alignment,
-		       unsigned usage, unsigned size)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct nouveau_device *dev = nvpws->channel->device;
-	struct nouveau_pipe_buffer *nvbuf;
-	uint32_t flags;
-
-	nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer);
-	if (!nvbuf)
-		return NULL;
-	pipe_reference_init(&nvbuf->base.reference, 1);
-	nvbuf->base.alignment = alignment;
-	nvbuf->base.usage = usage;
-	nvbuf->base.size = size;
-
-	flags = nouveau_flags_from_usage(ws, usage);
-	if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) {
-		FREE(nvbuf);
-		return NULL;
-	}
-
-	return &nvbuf->base;
-}
-
-static struct pipe_buffer *
-nouveau_pipe_bo_user_create(struct pipe_winsys *ws, void *ptr, unsigned bytes)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-	struct nouveau_device *dev = nvpws->channel->device;
-	struct nouveau_pipe_buffer *nvbuf;
-
-	nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer);
-	if (!nvbuf)
-		return NULL;
-	pipe_reference_init(&nvbuf->base.reference, 1);
-	nvbuf->base.size = bytes;
-
-	if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) {
-		FREE(nvbuf);
-		return NULL;
-	}
-
-	return &nvbuf->base;
-}
-
-static void
-nouveau_pipe_bo_del(struct pipe_buffer *buf)
-{
-	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
-
-	nouveau_bo_ref(NULL, &nvbuf->bo);
-	FREE(nvbuf);
-}
-
-static void *
-nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf,
-		    unsigned flags)
-{
-	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
-	uint32_t map_flags = 0;
-
-	if (flags & PIPE_BUFFER_USAGE_CPU_READ)
-		map_flags |= NOUVEAU_BO_RD;
-	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
-		map_flags |= NOUVEAU_BO_WR;
-
-	if (nouveau_bo_map(nvbuf->bo, map_flags))
-		return NULL;
-	return nvbuf->bo->map;
-}
-
-static void
-nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
-{
-	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
-
-	nouveau_bo_unmap(nvbuf->bo);
-}
-
-static void
-nouveau_pipe_fence_reference(struct pipe_winsys *ws,
-			     struct pipe_fence_handle **ptr,
-			     struct pipe_fence_handle *pfence)
-{
-	*ptr = pfence;
-}
-
-static int
-nouveau_pipe_fence_signalled(struct pipe_winsys *ws,
-			     struct pipe_fence_handle *pfence, unsigned flag)
-{
-	return 0;
-}
-
-static int
-nouveau_pipe_fence_finish(struct pipe_winsys *ws,
-			  struct pipe_fence_handle *pfence, unsigned flag)
-{
-	return 0;
-}
-
-static void
-nouveau_destroy(struct pipe_winsys *ws)
-{
-	struct nouveau_pipe_winsys *nvpws = nouveau_pipe_winsys(ws);
-
-	nouveau_device_close(&nvpws->channel->device);
-	FREE(nvpws);
-}
-
-struct pipe_winsys *
-nouveau_pipe_winsys_new(struct nouveau_device *dev)
-{
-	struct nouveau_pipe_winsys *nvpws;
-	int ret;
-
-	nvpws = CALLOC_STRUCT(nouveau_pipe_winsys);
-	if (!nvpws)
-		return NULL;
-
-	ret = nouveau_channel_alloc(dev, 0xbeef0201, 0xbeef0202,
-				    &nvpws->channel);
-	if (ret) {
-		debug_printf("%s: error opening GPU channel: %d\n",
-			     __func__, ret);
-		FREE(nvpws);
-		return NULL;
-	}
-	nvpws->next_handle = 0x77000000;
-
-	nvpws->base.buffer_create = nouveau_pipe_bo_create;
-	nvpws->base.buffer_destroy = nouveau_pipe_bo_del;
-	nvpws->base.user_buffer_create = nouveau_pipe_bo_user_create;
-	nvpws->base.buffer_map = nouveau_pipe_bo_map;
-	nvpws->base.buffer_unmap = nouveau_pipe_bo_unmap;
-
-	nvpws->base.fence_reference = nouveau_pipe_fence_reference;
-	nvpws->base.fence_signalled = nouveau_pipe_fence_signalled;
-	nvpws->base.fence_finish = nouveau_pipe_fence_finish;
-
-	nvpws->base.get_name = nouveau_get_name;
-	nvpws->base.destroy = nouveau_destroy;
-	return &nvpws->base;
-}
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
deleted file mode 100644
index 10e1e269e8..0000000000
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_winsys_pipe.h
+++ /dev/null
@@ -1,52 +0,0 @@
-#ifndef NOUVEAU_PIPE_WINSYS_H
-#define NOUVEAU_PIPE_WINSYS_H
-
-#include "pipe/internal/p_winsys_screen.h"
-#include "pipe/p_context.h"
-
-#include "nouveau/nouveau_winsys.h"
-
-#include "nouveau_device.h"
-
-struct nouveau_pipe_buffer {
-	struct pipe_buffer base;
-	struct nouveau_bo *bo;
-};
-
-static INLINE struct nouveau_pipe_buffer *
-nouveau_pipe_buffer(struct pipe_buffer *buf)
-{
-	return (struct nouveau_pipe_buffer *)buf;
-}
-
-struct nouveau_pipe_winsys {
-	struct pipe_winsys base;
-
-	struct pipe_screen *pscreen;
-
-	struct nouveau_channel *channel;
-	uint32_t next_handle;
-
-	unsigned nr_pctx;
-	struct pipe_context **pctx;
-};
-
-static INLINE struct nouveau_pipe_winsys *
-nouveau_pipe_winsys(struct pipe_winsys *ws)
-{
-	return (struct nouveau_pipe_winsys *)ws;
-}
-
-static INLINE struct nouveau_pipe_winsys *
-nouveau_screen(struct pipe_screen *pscreen)
-{
-	return nouveau_pipe_winsys(pscreen->winsys);
-}
-
-struct pipe_winsys *
-nouveau_pipe_winsys_new(struct nouveau_device *);
-
-struct nouveau_winsys *
-nouveau_winsys_new(struct pipe_winsys *ws);
-
-#endif
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index a15487352b..684a487f24 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -1,8 +1,8 @@
-/* 
+/*
  * Copyright © 2008 Jérôme Glisse
  *             2009 Corbin Simpson
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -10,14 +10,14 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  * The above copyright notice and this permission notice (including the
@@ -93,6 +93,29 @@ static struct pipe_buffer *radeon_buffer_user_create(struct pipe_winsys *ws,
     return &radeon_buffer->base;
 }
 
+static struct pipe_buffer *radeon_surface_buffer_create(struct pipe_winsys *ws,
+                                                        unsigned width,
+                                                        unsigned height,
+                                                        enum pipe_format format,
+                                                        unsigned usage,
+                                                        unsigned *stride)
+{
+    struct pipe_format_block block;
+    unsigned nblocksx, nblocksy, size;
+
+    pf_get_block(format, &block);
+
+    nblocksx = pf_get_nblocksx(&block, width);
+    nblocksy = pf_get_nblocksy(&block, height);
+
+    /* Radeons enjoy things in multiples of 32. */
+    /* XXX this can be 32 when POT */
+    *stride = (nblocksx * block.size + 63) & ~63;
+    size = *stride * nblocksy;
+
+    return radeon_buffer_create(ws, 64, usage, size);
+}
+
 static void radeon_buffer_del(struct pipe_buffer *buffer)
 {
     struct radeon_pipe_buffer *radeon_buffer =
@@ -118,6 +141,8 @@ static void *radeon_buffer_map(struct pipe_winsys *ws,
         write = 1;
     }
 
+    radeon_bo_wait(radeon_buffer->bo);
+
     if (radeon_bo_map(radeon_buffer->bo, write))
         return NULL;
     return radeon_buffer->bo->ptr;
@@ -175,15 +200,17 @@ struct radeon_winsys* radeon_pipe_winsys(int fd)
         return NULL;
     }
 
+    radeon_ws->priv->fd = fd;
     radeon_ws->priv->bom = radeon_bo_manager_gem_ctor(fd);
 
     radeon_ws->base.flush_frontbuffer = radeon_flush_frontbuffer;
 
     radeon_ws->base.buffer_create = radeon_buffer_create;
-    radeon_ws->base.buffer_destroy = radeon_buffer_del;
     radeon_ws->base.user_buffer_create = radeon_buffer_user_create;
+    radeon_ws->base.surface_buffer_create = radeon_surface_buffer_create;
     radeon_ws->base.buffer_map = radeon_buffer_map;
     radeon_ws->base.buffer_unmap = radeon_buffer_unmap;
+    radeon_ws->base.buffer_destroy = radeon_buffer_del;
 
     radeon_ws->base.fence_reference = radeon_fence_reference;
     radeon_ws->base.fence_signalled = radeon_fence_signalled;
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
index ca8bbb3c11..8c8b61fa10 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h
@@ -1,7 +1,7 @@
-/* 
+/*
  * Copyright © 2008 Jérôme Glisse
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -9,14 +9,14 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  * The above copyright notice and this permission notice (including the
@@ -48,11 +48,16 @@
 struct radeon_pipe_buffer {
     struct pipe_buffer  base;
     struct radeon_bo    *bo;
+    boolean flinked;
+    uint32_t flink;
 };
 
 #define RADEON_MAX_BOS 24
 
 struct radeon_winsys_priv {
+    /* DRM FD */
+    int fd;
+
     /* Radeon BO manager. */
     struct radeon_bo_manager* bom;
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
index 428d3f65a1..da2010184a 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
@@ -1,7 +1,7 @@
-/* 
+/*
  * Copyright © 2009 Corbin Simpson
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
@@ -9,14 +9,14 @@
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
  * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  * The above copyright notice and this permission notice (including the
@@ -29,9 +29,11 @@
  */
 
 #include "radeon_drm.h"
+#include "trace/tr_drm.h"
 
 /* Create a pipe_screen. */
-struct pipe_screen* radeon_create_screen(int drmFB,
+struct pipe_screen* radeon_create_screen(struct drm_api* api,
+                                         int drmFB,
 					 struct drm_create_screen_arg *arg)
 {
     struct radeon_winsys* winsys = radeon_pipe_winsys(drmFB);
@@ -46,7 +48,8 @@ struct pipe_screen* radeon_create_screen(int drmFB,
 }
 
 /* Create a pipe_context. */
-struct pipe_context* radeon_create_context(struct pipe_screen* screen)
+struct pipe_context* radeon_create_context(struct drm_api* api,
+                                           struct pipe_screen* screen)
 {
     if (getenv("RADEON_SOFTPIPE")) {
         return radeon_create_softpipe(screen->winsys);
@@ -55,16 +58,19 @@ struct pipe_context* radeon_create_context(struct pipe_screen* screen)
     }
 }
 
-boolean radeon_buffer_from_texture(struct pipe_texture* texture,
+boolean radeon_buffer_from_texture(struct drm_api* api,
+                                   struct pipe_texture* texture,
                                    struct pipe_buffer** buffer,
                                    unsigned* stride)
 {
-    return FALSE;
+    /* XXX fix this */
+    return r300_get_texture_buffer(texture, buffer, stride);
 }
 
 /* Create a buffer from a handle. */
 /* XXX what's up with name? */
-struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
+struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api,
+                                              struct pipe_screen* screen,
                                               const char* name,
                                               unsigned handle)
 {
@@ -91,7 +97,8 @@ struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
     return &radeon_buffer->base;
 }
 
-boolean radeon_handle_from_buffer(struct pipe_screen* screen,
+boolean radeon_handle_from_buffer(struct drm_api* api,
+                                  struct pipe_screen* screen,
                                   struct pipe_buffer* buffer,
                                   unsigned* handle)
 {
@@ -101,23 +108,50 @@ boolean radeon_handle_from_buffer(struct pipe_screen* screen,
     return TRUE;
 }
 
-boolean radeon_global_handle_from_buffer(struct pipe_screen* screen,
+boolean radeon_global_handle_from_buffer(struct drm_api* api,
+                                         struct pipe_screen* screen,
                                          struct pipe_buffer* buffer,
                                          unsigned* handle)
 {
-    /* XXX WTF is the difference here? global? */
+    int retval, fd;
+    struct drm_gem_flink flink;
     struct radeon_pipe_buffer* radeon_buffer =
         (struct radeon_pipe_buffer*)buffer;
-    *handle = radeon_buffer->bo->handle;
+
+    if (!radeon_buffer->flinked) {
+        fd = ((struct radeon_winsys*)screen->winsys)->priv->fd;
+
+        flink.handle = radeon_buffer->bo->handle;
+
+        retval = ioctl(fd, DRM_IOCTL_GEM_FLINK, &flink);
+        if (retval) {
+            debug_printf("radeon: DRM_IOCTL_GEM_FLINK failed, error %d\n",
+                    retval);
+            return FALSE;
+        }
+
+        radeon_buffer->flink = flink.name;
+        radeon_buffer->flinked = TRUE;
+    }
+
+    *handle = radeon_buffer->flink;
     return TRUE;
 }
 
 struct drm_api drm_api_hooks = {
     .create_screen = radeon_create_screen,
     .create_context = radeon_create_context,
-    /* XXX fix this */
-    .buffer_from_texture = r300_get_texture_buffer,
+    .buffer_from_texture = radeon_buffer_from_texture,
     .buffer_from_handle = radeon_buffer_from_handle,
     .handle_from_buffer = radeon_handle_from_buffer,
     .global_handle_from_buffer = radeon_global_handle_from_buffer,
 };
+
+struct drm_api* drm_api_create()
+{
+#ifdef DEBUG
+    return trace_drm_create(&drm_api_hooks);
+#else
+    return &drm_api_hooks;
+#endif
+}
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
index 049f9984db..8560f71db6 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
@@ -40,25 +40,32 @@
 #include "radeon_r300.h"
 #include "radeon_winsys_softpipe.h"
 
-struct pipe_screen* radeon_create_screen(int drmFB,
+struct pipe_screen* radeon_create_screen(struct drm_api* api,
+                                         int drmFB,
 					 struct drm_create_screen_arg *arg);
 
-struct pipe_context* radeon_create_context(struct pipe_screen* screen);
+struct pipe_context* radeon_create_context(struct drm_api* api,
+                                           struct pipe_screen* screen);
 
-boolean radeon_buffer_from_texture(struct pipe_texture* texture,
+boolean radeon_buffer_from_texture(struct drm_api* api,
+                                   struct pipe_texture* texture,
                                    struct pipe_buffer** buffer,
                                    unsigned* stride);
 
-struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen,
+struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api,
+                                              struct pipe_screen* screen,
                                               const char* name,
                                               unsigned handle);
 
-boolean radeon_handle_from_buffer(struct pipe_screen* screen,
+boolean radeon_handle_from_buffer(struct drm_api* api,
+                                  struct pipe_screen* screen,
                                   struct pipe_buffer* buffer,
                                   unsigned* handle);
 
-boolean radeon_global_handle_from_buffer(struct pipe_screen* screen,
+boolean radeon_global_handle_from_buffer(struct drm_api* api,
+                                         struct pipe_screen* screen,
                                          struct pipe_buffer* buffer,
                                          unsigned* handle);
 
+void radeon_destroy_drm_api(struct drm_api* api);
 #endif
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
index da233203d7..8c5f756ddf 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c
@@ -22,10 +22,10 @@
 
 #include "radeon_r300.h"
 
-static void radeon_r300_add_buffer(struct r300_winsys* winsys,
-                                   struct pipe_buffer* pbuffer,
-                                   uint32_t rd,
-                                   uint32_t wd)
+static boolean radeon_r300_add_buffer(struct r300_winsys* winsys,
+                                      struct pipe_buffer* pbuffer,
+                                      uint32_t rd,
+                                      uint32_t wd)
 {
     int i;
     struct radeon_winsys_priv* priv =
@@ -35,24 +35,30 @@ static void radeon_r300_add_buffer(struct r300_winsys* winsys,
 
     /* Check to see if this BO is already in line for validation;
      * find a slot for it otherwise. */
-    for (i = 0; i < RADEON_MAX_BOS; i++) {
+    for (i = 0; i < priv->bo_count; i++) {
         if (sc[i].bo == bo) {
-            return;
-        } else if (sc[i].bo == NULL) {
-            sc[i].bo = bo;
-            sc[i].read_domains = rd;
-            sc[i].write_domain = wd;
-            priv->bo_count = i + 1;
-            return;
+            sc[i].read_domains |= rd;
+            sc[i].write_domain |= wd;
+            return TRUE;
         }
     }
 
-    assert(FALSE && "Oh God too many BOs!");
+    if (priv->bo_count >= RADEON_MAX_BOS) {
+        /* Dohoho. Not falling for that one again. Request a flush. */
+        return FALSE;
+    }
+
+    sc[priv->bo_count].bo = bo;
+    sc[priv->bo_count].read_domains = rd;
+    sc[priv->bo_count].write_domain = wd;
+    priv->bo_count++;
+
+    return TRUE;
 }
 
 static boolean radeon_r300_validate(struct r300_winsys* winsys)
 {
-    int retval;
+    int retval, i;
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
     struct radeon_cs_space_check* sc = priv->sc;
@@ -62,12 +68,23 @@ static boolean radeon_r300_validate(struct r300_winsys* winsys)
     if (retval == RADEON_CS_SPACE_OP_TO_BIG) {
         /* We might as well HCF, since this is not going to fit in the card,
          * period. */
+        /* XXX just drop it on the floor instead */
 	exit(1);
     } else if (retval == RADEON_CS_SPACE_FLUSH) {
         /* We must flush before more rendering can commence. */
         return TRUE;
     }
 
+    /* XXX should probably be its own function */
+    for (i = 0; i < priv->bo_count; i++) {
+        if (sc[i].read_domains && sc[i].write_domain) {
+            /* Cute, cute. We need to flush first. */
+            debug_printf("radeon: BO %p can't be read and written; "
+                    "requesting flush.\n", sc[i].bo);
+            return TRUE;
+        }
+    }
+
     /* Things are fine, we can proceed as normal. */
     return FALSE;
 }
@@ -108,9 +125,15 @@ static void radeon_r300_write_cs_reloc(struct r300_winsys* winsys,
 {
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
+    int retval = 0;
 
-    radeon_cs_write_reloc(priv->cs,
+    retval = radeon_cs_write_reloc(priv->cs,
             ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags);
+
+    if (retval) {
+        debug_printf("radeon: Relocation of %p (%d, %d, %d) failed!\n",
+                pbuffer, rd, wd, flags);
+    }
 }
 
 static void radeon_r300_end_cs(struct r300_winsys* winsys,
@@ -128,57 +151,63 @@ static void radeon_r300_flush_cs(struct r300_winsys* winsys)
 {
     struct radeon_winsys_priv* priv =
         (struct radeon_winsys_priv*)winsys->radeon_winsys;
-    int retval = 0;
+    struct radeon_cs_space_check* sc = priv->sc;
+    int retval = 1;
 
+    /* Emit the CS. */
     retval = radeon_cs_emit(priv->cs);
     if (retval) {
         debug_printf("radeon: Bad CS, dumping...\n");
         radeon_cs_print(priv->cs, stderr);
     }
     radeon_cs_erase(priv->cs);
+
+    /* Clean out BOs. */
+    memset(sc, 0, sizeof(struct radeon_cs_space_check) * RADEON_MAX_BOS);
+    priv->bo_count = 0;
 }
 
 /* Helper function to do the ioctls needed for setup and init. */
 static void do_ioctls(struct r300_winsys* winsys, int fd)
 {
-    struct drm_radeon_gem_info info;
-    drm_radeon_getparam_t gp;
-    int target;
+    struct drm_radeon_gem_info gem_info = {0};
+    drm_radeon_getparam_t gp = {0};
+    struct drm_radeon_info info = {0};
+    int target = 0;
     int retval;
 
+    info.value = &target;
     gp.value = &target;
 
-    /* First, get the number of pixel pipes */
-    gp.param = RADEON_PARAM_NUM_GB_PIPES;
-    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+    /* First, get PCI ID */
+    info.request = RADEON_INFO_DEVICE_ID;
+    retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info));
     if (retval) {
-        fprintf(stderr, "%s: Failed to get GB pipe count, error number %d\n",
+        fprintf(stderr, "%s: New ioctl for PCI ID failed "
+                "(error number %d), trying classic ioctl...\n",
                 __FUNCTION__, retval);
-        exit(1);
-    }
-    winsys->gb_pipes = target;
-
-    /* Then, get PCI ID */
-    gp.param = RADEON_PARAM_DEVICE_ID;
-    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
-    if (retval) {
-        fprintf(stderr, "%s: Failed to get PCI ID, error number %d\n",
-                __FUNCTION__, retval);
-        exit(1);
+        gp.param = RADEON_PARAM_DEVICE_ID;
+        retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp,
+                sizeof(gp));
+        if (retval) {
+            fprintf(stderr, "%s: Failed to get PCI ID, "
+                    "error number %d\n", __FUNCTION__, retval);
+            exit(1);
+        }
     }
     winsys->pci_id = target;
 
-    /* Finally, retrieve MM info */
+    /* Then, retrieve MM info */
     retval = drmCommandWriteRead(fd, DRM_RADEON_GEM_INFO,
-            &info, sizeof(info));
+            &gem_info, sizeof(gem_info));
     if (retval) {
         fprintf(stderr, "%s: Failed to get MM info, error number %d\n",
                 __FUNCTION__, retval);
         exit(1);
     }
-    winsys->gart_size = info.gart_size;
+    winsys->gart_size = gem_info.gart_size;
     /* XXX */
-    winsys->vram_size = info.vram_visible;
+    winsys->vram_size = gem_info.vram_visible;
 }
 
 struct r300_winsys*
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.h b/src/gallium/winsys/drm/radeon/core/radeon_r300.h
index 5c373cd084..a2e0e58248 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_r300.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.h
@@ -31,5 +31,18 @@
 
 #include "radeon_buffer.h"
 
+/* protect us from bonghits */
+#ifndef RADEON_INFO_DEVICE_ID
+#define RADEON_INFO_DEVICE_ID 0
+#endif
+#ifndef DRM_RADEON_INFO
+#define DRM_RADEON_INFO 0x1
+struct drm_radeon_info {
+	uint32_t		request;
+	uint32_t		pad;
+	uint64_t		value;
+};
+#endif
+
 struct r300_winsys*
 radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys);
diff --git a/src/gallium/winsys/drm/radeon/dri/Makefile b/src/gallium/winsys/drm/radeon/dri/Makefile
index c218ee9d01..a9889444de 100644
--- a/src/gallium/winsys/drm/radeon/dri/Makefile
+++ b/src/gallium/winsys/drm/radeon/dri/Makefile
@@ -10,6 +10,7 @@ PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/libdridrm.a \
 	$(TOP)/src/gallium/winsys/drm/radeon/core/libradeonwinsys.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/r300/libr300.a
 
 C_SOURCES = \
diff --git a/src/gallium/winsys/drm/radeon/egl/Makefile b/src/gallium/winsys/drm/radeon/egl/Makefile
index d989b3aa93..6a1448d1b9 100644
--- a/src/gallium/winsys/drm/radeon/egl/Makefile
+++ b/src/gallium/winsys/drm/radeon/egl/Makefile
@@ -8,6 +8,7 @@ PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/egl/libegldrm.a \
 	$(GALLIUMDIR)/winsys/drm/radeon/core/libradeonwinsys.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
 	$(TOP)/src/gallium/drivers/r300/libr300.a
 
 DRIVER_SOURCES =
diff --git a/src/gallium/winsys/drm/radeon/xorg/Makefile b/src/gallium/winsys/drm/radeon/xorg/Makefile
index 6ffd4a3a54..0241625f69 100644
--- a/src/gallium/winsys/drm/radeon/xorg/Makefile
+++ b/src/gallium/winsys/drm/radeon/xorg/Makefile
@@ -37,6 +37,6 @@ clean:
 
 install:
 	$(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR)
 
 .PHONY	= all clean install
diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile
index 8646ee3b52..a33a50ec22 100644
--- a/src/gallium/winsys/egl_xlib/Makefile
+++ b/src/gallium/winsys/egl_xlib/Makefile
@@ -74,7 +74,7 @@ depend: $(ALL_SOURCES)
 install: default
 	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
 	@if [ -e $(TOP)/$(LIB_DIR) ]; then \
-		$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \
+		$(MINSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \
 	fi
 
 
diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c
index b52f427e4a..e1ddcae97b 100644
--- a/src/gallium/winsys/egl_xlib/egl_xlib.c
+++ b/src/gallium/winsys/egl_xlib/egl_xlib.c
@@ -345,7 +345,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       return EGL_NO_CONTEXT;
 
    /* let EGL lib init the common stuff */
-   if (!_eglInitContext(drv, dpy, &ctx->Base, config, attrib_list)) {
+   if (!_eglInitContext(drv, &ctx->Base, conf, attrib_list)) {
       free(ctx);
       return EGL_NO_CONTEXT;
    }
@@ -370,7 +370,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       return EGL_NO_CONTEXT;
    }
 
-   _eglSaveContext(&ctx->Base);
+   _eglLinkContext(&ctx->Base, _eglLookupDisplay(dpy));
 
    return _eglGetContextHandle(&ctx->Base);
 }
@@ -381,10 +381,8 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
 {
    struct xlib_egl_context *context = lookup_context(ctx);
    if (context) {
-      if (context->Base.IsBound) {
-         context->Base.DeletePending = EGL_TRUE;
-      }
-      else {
+      _eglUnlinkContext(&context->Base);
+      if (!context->Base.IsBound) {
          /* API-dependent clean-up */
          switch (context->Base.ClientAPI) {
          case EGL_OPENGL_ES_API:
@@ -491,13 +489,13 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
       return EGL_NO_SURFACE;
 
    /* Let EGL lib init the common stuff */
-   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_WINDOW_BIT,
-                        config, attrib_list)) {
+   if (!_eglInitSurface(drv, &surf->Base, EGL_WINDOW_BIT,
+                        conf, attrib_list)) {
       free(surf);
       return EGL_NO_SURFACE;
    }
 
-   _eglSaveSurface(&surf->Base);
+   _eglLinkSurface(&surf->Base, disp);
 
    /*
     * Now init the Xlib and gallium stuff
@@ -534,11 +532,8 @@ xlib_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    struct xlib_egl_surface *surf = lookup_surface(surface);
    if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
-      if (surf->Base.IsBound) {
-         surf->Base.DeletePending = EGL_TRUE;
-      }
-      else {
+      _eglUnlinkSurface(&surf->Base);
+      if (!surf->Base.IsBound) {
          XFreeGC(surf->Dpy, surf->Gc);
          st_unreference_framebuffer(surf->Framebuffer);
          free(surf);
@@ -588,7 +583,9 @@ find_supported_apis(void)
    EGLint mask = 0;
    void *handle;
 
-   handle = dlopen(NULL, 0);
+   handle = dlopen(NULL, RTLD_LAZY | RTLD_LOCAL);
+   if(!handle)
+      return mask;
 
    if (dlsym(handle, "st_api_OpenGL_ES1"))
       mask |= EGL_OPENGL_ES_BIT;
diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript
index aabab95f3a..86eb9ef55e 100644
--- a/src/gallium/winsys/gdi/SConscript
+++ b/src/gallium/winsys/gdi/SConscript
@@ -15,6 +15,7 @@ if env['platform'] == 'windows':
 		'gdi32',
 		'user32',
 		'kernel32',
+		'ws2_32',
 	])
 
 	sources = [
diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile
index 04309e67ee..522f6dc5ae 100644
--- a/src/gallium/winsys/xlib/Makefile
+++ b/src/gallium/winsys/xlib/Makefile
@@ -90,7 +90,7 @@ install: default
 	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
 	$(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL
 	@if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \
-		$(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \
+		$(MINSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \
 	fi
 
 
diff --git a/src/glu/Makefile b/src/glu/Makefile
index 5c26ead1bb..b268265976 100644
--- a/src/glu/Makefile
+++ b/src/glu/Makefile
@@ -30,7 +30,7 @@ glu.pc: glu.pc.in
 install: glu.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GLU_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GLU_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glu.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 clean:
diff --git a/src/glut/fbdev/Makefile b/src/glut/fbdev/Makefile
index 199d8c390a..c150ea88dc 100644
--- a/src/glut/fbdev/Makefile
+++ b/src/glut/fbdev/Makefile
@@ -72,7 +72,7 @@ install:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_INC_DIR)/GL
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 $(TOP)/include/GL/glut.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
 
 # Run 'make -f Makefile.solo dep' to update the dependencies if you change
 # what's included by any source file.
diff --git a/src/glut/glx/Makefile b/src/glut/glx/Makefile
index 1b072906c7..6889cd4b40 100644
--- a/src/glut/glx/Makefile
+++ b/src/glut/glx/Makefile
@@ -117,7 +117,7 @@ install: glut.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 644 $(TOP)/include/GL/glut.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GLUT_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GLUT_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glut.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 
diff --git a/src/glut/mini/Makefile b/src/glut/mini/Makefile
index 0e42436133..b82a758d2c 100644
--- a/src/glut/mini/Makefile
+++ b/src/glut/mini/Makefile
@@ -91,7 +91,7 @@ install: glut.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 644 $(TOP)/include/GL/glut.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/libglut* $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glut.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 
diff --git a/src/glw/Makefile b/src/glw/Makefile
index d88d773313..1fb3d3c320 100644
--- a/src/glw/Makefile
+++ b/src/glw/Makefile
@@ -43,7 +43,7 @@ install: glw.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 644 *.h $(DESTDIR)$(INSTALL_INC_DIR)/GL
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GLW_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GLW_LIB_GLOB) $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 glw.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 clean:
diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c
index fb31898db2..f4865aecb1 100644
--- a/src/glx/x11/dri2_glx.c
+++ b/src/glx/x11/dri2_glx.c
@@ -80,6 +80,8 @@ struct __GLXDRIdrawablePrivateRec {
     int have_fake_front;
 };
 
+static void dri2WaitX(__GLXDRIdrawable *pdraw);
+
 static void dri2DestroyContext(__GLXDRIcontext *context,
 			      __GLXscreenConfigs *psc, Display *dpy)
 {
@@ -215,6 +217,11 @@ static void dri2CopySubBuffer(__GLXDRIdrawable *pdraw,
     DRI2CopyRegion(pdraw->psc->dpy, pdraw->drawable, region,
 		   DRI2BufferFrontLeft, DRI2BufferBackLeft);
     XFixesDestroyRegion(pdraw->psc->dpy, region);
+
+    /* Refresh the fake front (if present) after we just damaged the real
+     * front.
+     */
+    dri2WaitX(pdraw);
 }
 
 static void dri2SwapBuffers(__GLXDRIdrawable *pdraw)
diff --git a/src/glx/x11/indirect.c b/src/glx/x11/indirect.c
index aea117ec32..e0cafd43bc 100644
--- a/src/glx/x11/indirect.c
+++ b/src/glx/x11/indirect.c
@@ -5201,10 +5201,9 @@ glDeleteTexturesEXT(GLsizei n, const GLuint * textures)
 #ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_DeleteTextures(GET_DISPATCH(), (n, textures));
-    } else {
-#else
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
@@ -5270,13 +5269,12 @@ glGenTexturesEXT(GLsizei n, GLuint * textures)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GenTextures(GET_DISPATCH(), (n, textures));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 4;
@@ -5336,13 +5334,12 @@ glIsTextureEXT(GLuint texture)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         return CALL_IsTexture(GET_DISPATCH(), (texture));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         GLboolean retval = (GLboolean) 0;
@@ -5653,13 +5650,12 @@ glGetColorTableEXT(GLenum target, GLenum format, GLenum type, GLvoid * table)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetColorTable(GET_DISPATCH(), (target, format, type, table));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -5730,14 +5726,13 @@ glGetColorTableParameterfvEXT(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetColorTableParameterfv(GET_DISPATCH(),
                                       (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -5804,14 +5799,13 @@ glGetColorTableParameterivEXT(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetColorTableParameteriv(GET_DISPATCH(),
                                       (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6131,14 +6125,13 @@ gl_dispatch_stub_356(GLenum target, GLenum format, GLenum type,
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetConvolutionFilter(GET_DISPATCH(),
                                   (target, format, type, image));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -6210,14 +6203,13 @@ gl_dispatch_stub_357(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetConvolutionParameterfv(GET_DISPATCH(),
                                        (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6284,14 +6276,13 @@ gl_dispatch_stub_358(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetConvolutionParameteriv(GET_DISPATCH(),
                                        (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6365,14 +6356,13 @@ gl_dispatch_stub_361(GLenum target, GLboolean reset, GLenum format,
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetHistogram(GET_DISPATCH(),
                           (target, reset, format, type, values));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -6443,13 +6433,12 @@ gl_dispatch_stub_362(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetHistogramParameterfv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6515,13 +6504,12 @@ gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetHistogramParameteriv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6591,13 +6579,12 @@ gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format,
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetMinmax(GET_DISPATCH(), (target, reset, format, type, values));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
@@ -6666,13 +6653,12 @@ gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetMinmaxParameterfv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
@@ -6735,13 +6721,12 @@ gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params)
 {
     __GLXcontext *const gc = __glXGetCurrentContext();
 
-#ifdef GLX_DIRECT_RENDERING    
+#ifdef GLX_DIRECT_RENDERING
     if (gc->driContext) {
         CALL_GetMinmaxParameteriv(GET_DISPATCH(), (target, pname, params));
-    } else {
-#else 
-    {
+    } else
 #endif
+    {
         __GLXcontext *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index bb18bee8ea..8300b30144 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -129,14 +129,14 @@ install-headers:
 install-libgl: default gl.pc install-headers
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(GL_LIB_GLOB) \
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(GL_LIB_GLOB) \
 		$(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 gl.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 install-osmesa: default osmesa.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
-	$(INSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_GLOB) \
+	$(MINSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_GLOB) \
 		$(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 osmesa.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c
index 6a98c29a3d..3b397fef7d 100644
--- a/src/mesa/drivers/common/driverfuncs.c
+++ b/src/mesa/drivers/common/driverfuncs.c
@@ -202,16 +202,11 @@ _mesa_init_driver_functions(struct dd_function_table *driver)
    driver->GetIntegerv = NULL;
    driver->GetPointerv = NULL;
    
-#if FEATURE_ARB_vertex_buffer_object
-   driver->NewBufferObject = _mesa_new_buffer_object;
-   driver->DeleteBuffer = _mesa_delete_buffer_object;
-   driver->BindBuffer = NULL;
-   driver->BufferData = _mesa_buffer_data;
-   driver->BufferSubData = _mesa_buffer_subdata;
-   driver->GetBufferSubData = _mesa_buffer_get_subdata;
-   driver->MapBuffer = _mesa_buffer_map;
-   driver->UnmapBuffer = _mesa_buffer_unmap;
-#endif
+   /* buffer objects */
+   _mesa_init_buffer_object_functions(driver);
+
+   /* query objects */
+   _mesa_init_query_object_functions(driver);
 
 #if FEATURE_EXT_framebuffer_object
    driver->NewFramebuffer = _mesa_new_framebuffer;
@@ -225,14 +220,6 @@ _mesa_init_driver_functions(struct dd_function_table *driver)
    driver->BlitFramebuffer = _swrast_BlitFramebuffer;
 #endif
 
-   /* query objects */
-   driver->NewQueryObject = _mesa_new_query_object;
-   driver->DeleteQuery = _mesa_delete_query;
-   driver->BeginQuery = _mesa_begin_query;
-   driver->EndQuery = _mesa_end_query;
-   driver->WaitQuery = _mesa_wait_query;
-   driver->CheckQuery = _mesa_check_query;
-
    /* APPLE_vertex_array_object */
    driver->NewArrayObject = _mesa_new_array_object;
    driver->DeleteArrayObject = _mesa_delete_array_object;
diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template
index 5c01d233c1..18dbeba24a 100644
--- a/src/mesa/drivers/dri/Makefile.template
+++ b/src/mesa/drivers/dri/Makefile.template
@@ -11,7 +11,8 @@ COMMON_GALLIUM_SOURCES = \
 COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
         ../../common/driverfuncs.c \
         ../common/texmem.c \
-        ../common/drirenderbuffer.c
+        ../common/drirenderbuffer.c \
+	../common/dri_metaops.c
 
 ifeq ($(WINDOW_SYSTEM),dri)
 WINOBJ=
@@ -92,7 +93,7 @@ clean:
 
 install: $(LIBNAME)
 	$(INSTALL) -d $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
-	$(INSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
+	$(MINSTALL) -m 755 $(LIBNAME) $(DESTDIR)$(DRI_DRIVER_INSTALL_DIR)
 
 
 -include depend
diff --git a/src/mesa/drivers/dri/common/dri_metaops.c b/src/mesa/drivers/dri/common/dri_metaops.c
new file mode 100644
index 0000000000..fe183c2e87
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_metaops.c
@@ -0,0 +1,541 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "main/arrayobj.h"
+#include "main/attrib.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/enable.h"
+#include "main/matrix.h"
+#include "main/macros.h"
+#include "main/polygon.h"
+#include "main/shaders.h"
+#include "main/stencil.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "main/viewport.h"
+#include "shader/arbprogram.h"
+#include "shader/program.h"
+#include "dri_metaops.h"
+
+void
+meta_set_passthrough_transform(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   meta->saved_vp_x = ctx->Viewport.X;
+   meta->saved_vp_y = ctx->Viewport.Y;
+   meta->saved_vp_width = ctx->Viewport.Width;
+   meta->saved_vp_height = ctx->Viewport.Height;
+   meta->saved_matrix_mode = ctx->Transform.MatrixMode;
+
+   meta->internal_viewport_call = GL_TRUE;
+   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
+   meta->internal_viewport_call = GL_FALSE;
+
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
+
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+}
+
+void
+meta_restore_transform(struct dri_metaops *meta)
+{
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PopMatrix();
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PopMatrix();
+
+   _mesa_MatrixMode(meta->saved_matrix_mode);
+
+   meta->internal_viewport_call = GL_TRUE;
+   _mesa_Viewport(meta->saved_vp_x, meta->saved_vp_y,
+		  meta->saved_vp_width, meta->saved_vp_height);
+   meta->internal_viewport_call = GL_FALSE;
+}
+
+
+/**
+ * Set up a vertex program to pass through the position and first texcoord
+ * for pixel path.
+ */
+void
+meta_set_passthrough_vertex_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   static const char *vp =
+      "!!ARBvp1.0\n"
+      "TEMP vertexClip;\n"
+      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
+      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
+      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
+      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
+      "MOV result.position, vertexClip;\n"
+      "MOV result.texcoord[0], vertex.texcoord[0];\n"
+      "MOV result.color, vertex.color;\n"
+      "END\n";
+
+   assert(meta->saved_vp == NULL);
+
+   _mesa_reference_vertprog(ctx, &meta->saved_vp,
+			    ctx->VertexProgram.Current);
+   if (meta->passthrough_vp == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(vp), (const GLubyte *)vp);
+      _mesa_reference_vertprog(ctx, &meta->passthrough_vp,
+			       ctx->VertexProgram.Current);
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    meta->passthrough_vp);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &meta->passthrough_vp->Base);
+
+   meta->saved_vp_enable = ctx->VertexProgram.Enabled;
+   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous vertex program after
+ * meta_set_passthrough_vertex_program()
+ */
+void
+meta_restore_vertex_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    meta->saved_vp);
+   _mesa_reference_vertprog(ctx, &meta->saved_vp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &ctx->VertexProgram.Current->Base);
+
+   if (!meta->saved_vp_enable)
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
+ * program object.
+ */
+void
+meta_set_fragment_program(struct dri_metaops *meta,
+			  struct gl_fragment_program **prog,
+			  const char *prog_string)
+{
+   GLcontext *ctx = meta->ctx;
+   assert(meta->saved_fp == NULL);
+
+   _mesa_reference_fragprog(ctx, &meta->saved_fp,
+			    ctx->FragmentProgram.Current);
+   if (*prog == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(prog_string), (const GLubyte *)prog_string);
+      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
+      /* Note that DeletePrograms unbinds the program on us */
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
+
+   meta->saved_fp_enable = ctx->FragmentProgram.Enabled;
+   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous fragment program after
+ * meta_set_fragment_program()
+ */
+void
+meta_restore_fragment_program(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
+			    meta->saved_fp);
+   _mesa_reference_fragprog(ctx, &meta->saved_fp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
+			   &ctx->FragmentProgram.Current->Base);
+
+   if (!meta->saved_fp_enable)
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+static const float default_texcoords[4][2] = { { 0.0, 0.0 },
+					       { 1.0, 0.0 },
+					       { 1.0, 1.0 },
+					       { 0.0, 1.0 } };
+
+void
+meta_set_default_texrect(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   struct gl_client_array *old_texcoord_array;
+
+   meta->saved_active_texture = ctx->Texture.CurrentUnit;
+   if (meta->saved_array_vbo == NULL) {
+      _mesa_reference_buffer_object(ctx, &meta->saved_array_vbo,
+				    ctx->Array.ArrayBufferObj);
+   }
+
+   old_texcoord_array = &ctx->Array.ArrayObj->TexCoord[0];
+   meta->saved_texcoord_type = old_texcoord_array->Type;
+   meta->saved_texcoord_size = old_texcoord_array->Size;
+   meta->saved_texcoord_stride = old_texcoord_array->Stride;
+   meta->saved_texcoord_enable = old_texcoord_array->Enabled;
+   meta->saved_texcoord_ptr = old_texcoord_array->Ptr;
+   _mesa_reference_buffer_object(ctx, &meta->saved_texcoord_vbo,
+				 old_texcoord_array->BufferObj);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
+
+   if (meta->texcoord_vbo == NULL) {
+      GLuint vbo_name;
+
+      _mesa_GenBuffersARB(1, &vbo_name);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, vbo_name);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(default_texcoords),
+			  default_texcoords, GL_STATIC_DRAW_ARB);
+      _mesa_reference_buffer_object(ctx, &meta->texcoord_vbo,
+				    ctx->Array.ArrayBufferObj);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->texcoord_vbo->Name);
+   }
+   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), NULL);
+
+   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+}
+
+void
+meta_restore_texcoords(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+
+   /* Restore the old TexCoordPointer */
+   if (meta->saved_texcoord_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->saved_texcoord_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &meta->saved_texcoord_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+
+   _mesa_TexCoordPointer(meta->saved_texcoord_size,
+			 meta->saved_texcoord_type,
+			 meta->saved_texcoord_stride,
+			 meta->saved_texcoord_ptr);
+   if (!meta->saved_texcoord_enable)
+      _mesa_Disable(GL_TEXTURE_COORD_ARRAY);
+
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0 +
+				meta->saved_active_texture);
+
+   if (meta->saved_array_vbo) {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+			  meta->saved_array_vbo->Name);
+      _mesa_reference_buffer_object(ctx, &meta->saved_array_vbo, NULL);
+   } else {
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   }
+}
+
+
+/**
+ * Perform glClear where mask contains only color, depth, and/or stencil.
+ *
+ * The implementation is based on calling into Mesa to set GL state and
+ * performing normal triangle rendering.  The intent of this path is to
+ * have as generic a path as possible, so that any driver could make use of
+ * it.
+ */
+
+/**
+ * Per-context one-time init of things for intl_clear_tris().
+ * Basically set up a private array object for vertex/color arrays.
+ */
+static void
+meta_init_clear(struct dri_metaops *meta)
+{
+   GLcontext *ctx = meta->ctx;
+   struct gl_array_object *arraySave = NULL;
+   const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name;
+   const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name;
+
+   /* create new array object */
+   meta->clear.arrayObj = _mesa_new_array_object(ctx, ~0);
+
+   /* save current array object, bind new one */
+   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, meta->clear.arrayObj);
+
+   /* one-time setup of vertex arrays (pos, color) */
+   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
+   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
+   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), meta->clear.color);
+   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), meta->clear.vertices);
+   _mesa_Enable(GL_COLOR_ARRAY);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+
+   /* restore original array object */
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
+   _mesa_reference_array_object(ctx, &arraySave, NULL);
+
+   /* restore original buffer objects */
+   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer);
+   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer);
+}
+
+
+
+/**
+ * Perform glClear where mask contains only color, depth, and/or stencil.
+ *
+ * The implementation is based on calling into Mesa to set GL state and
+ * performing normal triangle rendering.  The intent of this path is to
+ * have as generic a path as possible, so that any driver could make use of
+ * it.
+ */
+void
+meta_clear_tris(struct dri_metaops *meta, GLbitfield mask)
+{
+   GLcontext *ctx = meta->ctx;
+   GLfloat dst_z;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   int i;
+   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
+   GLuint saved_shader_program = 0;
+   unsigned int saved_active_texture;
+   struct gl_array_object *arraySave = NULL;
+
+   if (!meta->clear.arrayObj)
+      meta_init_clear(meta);
+
+   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
+		    BUFFER_BIT_STENCIL)) == 0);
+
+   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
+		    GL_DEPTH_BUFFER_BIT |
+		    GL_ENABLE_BIT |
+		    GL_POLYGON_BIT |
+		    GL_STENCIL_BUFFER_BIT |
+		    GL_TRANSFORM_BIT |
+		    GL_CURRENT_BIT |
+		    GL_VIEWPORT_BIT);
+   saved_active_texture = ctx->Texture.CurrentUnit;
+
+   /* Disable existing GL state we don't want to apply to a clear. */
+   _mesa_Disable(GL_ALPHA_TEST);
+   _mesa_Disable(GL_BLEND);
+   _mesa_Disable(GL_CULL_FACE);
+   _mesa_Disable(GL_FOG);
+   _mesa_Disable(GL_POLYGON_SMOOTH);
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
+   _mesa_Disable(GL_LIGHTING);
+   _mesa_Disable(GL_CLIP_PLANE0);
+   _mesa_Disable(GL_CLIP_PLANE1);
+   _mesa_Disable(GL_CLIP_PLANE2);
+   _mesa_Disable(GL_CLIP_PLANE3);
+   _mesa_Disable(GL_CLIP_PLANE4);
+   _mesa_Disable(GL_CLIP_PLANE5);
+   _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
+   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
+      saved_fp_enable = GL_TRUE;
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
+      saved_vp_enable = GL_TRUE;
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
+      saved_shader_program = ctx->Shader.CurrentProgram->Name;
+      _mesa_UseProgramObjectARB(0);
+   }
+
+   if (ctx->Texture._EnabledUnits != 0) {
+      int i;
+
+      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
+	 _mesa_Disable(GL_TEXTURE_1D);
+	 _mesa_Disable(GL_TEXTURE_2D);
+	 _mesa_Disable(GL_TEXTURE_3D);
+	 if (ctx->Extensions.ARB_texture_cube_map)
+	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
+	 if (ctx->Extensions.NV_texture_rectangle)
+	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
+	 if (ctx->Extensions.MESA_texture_array) {
+	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
+	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
+	 }
+      }
+   }
+
+   /* save current array object, bind our private one */
+   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, meta->clear.arrayObj);
+
+   meta_set_passthrough_transform(meta);
+
+   for (i = 0; i < 4; i++) {
+      COPY_4FV(meta->clear.color[i], ctx->Color.ClearColor);
+   }
+
+   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
+   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
+
+   /* The ClearDepth value is unaffected by DepthRange, so do a default
+    * mapping.
+    */
+   _mesa_DepthRange(0.0, 1.0);
+
+   /* Prepare the vertices, which are the same regardless of which buffer we're
+    * drawing to.
+    */
+   meta->clear.vertices[0][0] = fb->_Xmin;
+   meta->clear.vertices[0][1] = fb->_Ymin;
+   meta->clear.vertices[0][2] = dst_z;
+   meta->clear.vertices[1][0] = fb->_Xmax;
+   meta->clear.vertices[1][1] = fb->_Ymin;
+   meta->clear.vertices[1][2] = dst_z;
+   meta->clear.vertices[2][0] = fb->_Xmax;
+   meta->clear.vertices[2][1] = fb->_Ymax;
+   meta->clear.vertices[2][2] = dst_z;
+   meta->clear.vertices[3][0] = fb->_Xmin;
+   meta->clear.vertices[3][1] = fb->_Ymax;
+   meta->clear.vertices[3][2] = dst_z;
+
+   while (mask != 0) {
+      GLuint this_mask = 0;
+      GLuint color_bit;
+
+      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
+      if (color_bit != 0)
+	 this_mask |= (1 << (color_bit - 1));
+
+      /* Clear depth/stencil in the same pass as color. */
+      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
+
+      /* Select the current color buffer and use the color write mask if
+       * we have one, otherwise don't write any color channels.
+       */
+      if (this_mask & BUFFER_BIT_FRONT_LEFT)
+	 _mesa_DrawBuffer(GL_FRONT_LEFT);
+      else if (this_mask & BUFFER_BIT_BACK_LEFT)
+	 _mesa_DrawBuffer(GL_BACK_LEFT);
+      else if (color_bit != 0)
+	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
+			  (color_bit - BUFFER_COLOR0 - 1));
+      else
+	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+
+      /* Control writing of the depth clear value to depth. */
+      if (this_mask & BUFFER_BIT_DEPTH) {
+	 _mesa_DepthFunc(GL_ALWAYS);
+	 _mesa_Enable(GL_DEPTH_TEST);
+      } else {
+	 _mesa_Disable(GL_DEPTH_TEST);
+	 _mesa_DepthMask(GL_FALSE);
+      }
+
+      /* Control writing of the stencil clear value to stencil. */
+      if (this_mask & BUFFER_BIT_STENCIL) {
+	 _mesa_Enable(GL_STENCIL_TEST);
+	 _mesa_StencilOpSeparate(GL_FRONT_AND_BACK,
+				 GL_REPLACE, GL_REPLACE, GL_REPLACE);
+	 _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS,
+				   ctx->Stencil.Clear & 0x7fffffff,
+				   ctx->Stencil.WriteMask[0]);
+      } else {
+	 _mesa_Disable(GL_STENCIL_TEST);
+      }
+
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+      mask &= ~this_mask;
+   }
+
+   meta_restore_transform(meta);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
+   if (saved_fp_enable)
+      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+   if (saved_vp_enable)
+      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+
+   if (saved_shader_program)
+      _mesa_UseProgramObjectARB(saved_shader_program);
+
+   _mesa_PopAttrib();
+
+   /* restore current array object */
+   ctx->NewState |= _NEW_ARRAY;
+   ctx->Array.NewState |= _NEW_ARRAY_ALL;
+   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
+   _mesa_reference_array_object(ctx, &arraySave, NULL);
+}
+
+void meta_init_metaops(GLcontext *ctx, struct dri_metaops *meta)
+{
+   meta->ctx = ctx;
+}
+
+void meta_destroy_metaops(struct dri_metaops *meta)
+{
+  if (meta->clear.arrayObj)
+    _mesa_delete_array_object(meta->ctx, meta->clear.arrayObj);
+
+}
diff --git a/src/mesa/drivers/dri/common/dri_metaops.h b/src/mesa/drivers/dri/common/dri_metaops.h
new file mode 100644
index 0000000000..bb4079d535
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_metaops.h
@@ -0,0 +1,99 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+#ifndef DRI_METAOPS_H
+#define DRI_METAOPS_H
+
+#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
+			      BUFFER_BIT_FRONT_LEFT |			\
+			      BUFFER_BIT_COLOR0 |			\
+			      BUFFER_BIT_COLOR1 |			\
+			      BUFFER_BIT_COLOR2 |			\
+			      BUFFER_BIT_COLOR3 |			\
+			      BUFFER_BIT_COLOR4 |			\
+			      BUFFER_BIT_COLOR5 |			\
+			      BUFFER_BIT_COLOR6 |			\
+			      BUFFER_BIT_COLOR7)
+
+struct dri_meta_clear {
+    struct gl_array_object *arrayObj;
+    GLfloat vertices[4][3];
+    GLfloat color[4][4];
+};
+
+struct dri_metaops {
+    GLcontext *ctx;
+    GLboolean internal_viewport_call;
+    struct gl_fragment_program *bitmap_fp;
+    struct gl_vertex_program *passthrough_vp;
+    struct gl_buffer_object *texcoord_vbo;
+    
+    struct gl_fragment_program *saved_fp;
+    GLboolean saved_fp_enable;
+    struct gl_vertex_program *saved_vp;
+    GLboolean saved_vp_enable;
+
+    struct gl_fragment_program *tex2d_fp;
+    
+    GLboolean saved_texcoord_enable;
+    struct gl_buffer_object *saved_array_vbo, *saved_texcoord_vbo;
+    GLenum saved_texcoord_type;
+    GLsizei saved_texcoord_size, saved_texcoord_stride;
+    const void *saved_texcoord_ptr;
+    int saved_active_texture;
+
+    GLint saved_vp_x, saved_vp_y;
+    GLsizei saved_vp_width, saved_vp_height;
+    GLenum saved_matrix_mode;
+    
+    struct dri_meta_clear clear;
+};
+
+
+void meta_set_passthrough_transform(struct dri_metaops *meta);
+
+void meta_restore_transform(struct dri_metaops *meta);
+
+void meta_set_passthrough_vertex_program(struct dri_metaops *meta);
+
+void meta_restore_vertex_program(struct dri_metaops *meta);
+
+void meta_set_fragment_program(struct dri_metaops *meta,
+			  struct gl_fragment_program **prog,
+			  const char *prog_string);
+
+void meta_restore_fragment_program(struct dri_metaops *meta);
+
+void meta_set_default_texrect(struct dri_metaops *meta);
+
+void meta_restore_texcoords(struct dri_metaops *meta);
+void meta_clear_tris(struct dri_metaops *meta, GLbitfield mask);
+
+void meta_init_metaops(GLcontext *ctx, struct dri_metaops *meta);
+void meta_destroy_metaops(struct dri_metaops *meta);
+#endif
+
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index 576494940f..1d940603fa 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -1,4 +1,3 @@
-/* $XFree86: xc/lib/GL/dri/dri_util.c,v 1.7 2003/04/28 17:01:25 dawes Exp $ */
 /**
  * \file dri_util.c
  * DRI utility functions.
@@ -37,6 +36,9 @@
 typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRIdrawable *drawable, int32_t *numerator, int32_t *denominator);
 #endif
 
+static void dri_get_drawable(__DRIdrawable *pdp);
+static void dri_put_drawable(__DRIdrawable *pdp);
+
 /**
  * This is just a token extension used to signal that the driver
  * supports setting a read drawable.
@@ -130,7 +132,7 @@ static int driUnbindContext(__DRIcontext *pcp)
 	return GL_FALSE;
     }
 
-    pdp->refcount--;
+    dri_put_drawable(pdp);
 
     if (prp != pdp) {
         if (prp->refcount == 0) {
@@ -138,7 +140,7 @@ static int driUnbindContext(__DRIcontext *pcp)
 	    return GL_FALSE;
 	}
 
-	prp->refcount--;
+    	dri_put_drawable(prp);
     }
 
 
@@ -174,10 +176,10 @@ static int driBindContext(__DRIcontext *pcp,
 	pcp->driReadablePriv = prp;
 	if (pdp) {
 	    pdp->driContextPriv = pcp;
-	    pdp->refcount++;
+    	    dri_get_drawable(pdp);
 	}
 	if ( prp && pdp != prp ) {
-	    prp->refcount++;
+    	    dri_get_drawable(prp);
 	}
     }
 
@@ -434,7 +436,7 @@ driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config,
 
     pdp->loaderPrivate = data;
     pdp->hHWDrawable = hwDrawable;
-    pdp->refcount = 0;
+    pdp->refcount = 1;
     pdp->pStamp = NULL;
     pdp->lastStamp = 0;
     pdp->index = 0;
@@ -487,12 +489,19 @@ dri2CreateNewDrawable(__DRIscreen *screen,
     return pdraw;
 }
 
-
-static void
-driDestroyDrawable(__DRIdrawable *pdp)
+static void dri_get_drawable(__DRIdrawable *pdp)
+{
+    pdp->refcount++;
+}
+	
+static void dri_put_drawable(__DRIdrawable *pdp)
 {
     __DRIscreenPrivate *psp;
 
+    pdp->refcount--;
+    if (pdp->refcount)
+	return;
+
     if (pdp) {
 	psp = pdp->driScreenPriv;
         (*psp->DriverAPI.DestroyBuffer)(pdp);
@@ -508,6 +517,12 @@ driDestroyDrawable(__DRIdrawable *pdp)
     }
 }
 
+static void
+driDestroyDrawable(__DRIdrawable *pdp)
+{
+    dri_put_drawable(pdp);
+}
+
 /*@}*/
 
 
diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h
index 8dcaaee307..e308fd2831 100644
--- a/src/mesa/drivers/dri/common/extension_helper.h
+++ b/src/mesa/drivers/dri/common/extension_helper.h
@@ -406,9 +406,10 @@ static const char UniformMatrix4fvARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_APPLE_vertex_array_object)
+#if defined(need_GL_ARB_vertex_array_object) || defined(need_GL_APPLE_vertex_array_object)
 static const char DeleteVertexArraysAPPLE_names[] = 
     "ip\0" /* Parameter signature */
+    "glDeleteVertexArrays\0"
     "glDeleteVertexArraysAPPLE\0"
     "";
 #endif
@@ -941,6 +942,13 @@ static const char WeightivARB_names[] =
     "";
 #endif
 
+#if defined(need_GL_SGIX_instruments)
+static const char PollInstrumentsSGIX_names[] = 
+    "p\0" /* Parameter signature */
+    "glPollInstrumentsSGIX\0"
+    "";
+#endif
+
 #if defined(need_GL_SUN_global_alpha)
 static const char GlobalAlphaFactordSUN_names[] = 
     "d\0" /* Parameter signature */
@@ -1763,6 +1771,13 @@ static const char DeleteFencesNV_names[] =
     "";
 #endif
 
+#if defined(need_GL_SGIX_polynomial_ffd)
+static const char DeformationMap3dSGIX_names[] = 
+    "iddiiddiiddiip\0" /* Parameter signature */
+    "glDeformationMap3dSGIX\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_2_0)
 static const char IsShader_names[] = 
     "i\0" /* Parameter signature */
@@ -2107,6 +2122,13 @@ static const char Tangent3fvEXT_names[] =
     "";
 #endif
 
+#if defined(need_GL_ARB_vertex_array_object)
+static const char GenVertexArrays_names[] = 
+    "ip\0" /* Parameter signature */
+    "glGenVertexArrays\0"
+    "";
+#endif
+
 #if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char BindFramebufferEXT_names[] = 
     "ii\0" /* Parameter signature */
@@ -2333,10 +2355,10 @@ static const char GetCombinerStageParameterfvNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_coordinate_frame)
-static const char Binormal3fEXT_names[] = 
-    "fff\0" /* Parameter signature */
-    "glBinormal3fEXT\0"
+#if defined(need_GL_ARB_vertex_array_object)
+static const char BindVertexArray_names[] = 
+    "i\0" /* Parameter signature */
+    "glBindVertexArray\0"
     "";
 #endif
 
@@ -2766,10 +2788,10 @@ static const char Uniform4fARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_IBM_multimode_draw_arrays)
-static const char MultiModeDrawArraysIBM_names[] = 
-    "pppii\0" /* Parameter signature */
-    "glMultiModeDrawArraysIBM\0"
+#if defined(need_GL_ARB_map_buffer_range)
+static const char FlushMappedBufferRange_names[] = 
+    "iii\0" /* Parameter signature */
+    "glFlushMappedBufferRange\0"
     "";
 #endif
 
@@ -3396,10 +3418,11 @@ static const char GetProgramParameterdvNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_SGIX_instruments)
-static const char PollInstrumentsSGIX_names[] = 
-    "p\0" /* Parameter signature */
-    "glPollInstrumentsSGIX\0"
+#if defined(need_GL_ARB_vertex_array_object) || defined(need_GL_APPLE_vertex_array_object)
+static const char IsVertexArrayAPPLE_names[] = 
+    "i\0" /* Parameter signature */
+    "glIsVertexArray\0"
+    "glIsVertexArrayAPPLE\0"
     "";
 #endif
 
@@ -3897,6 +3920,13 @@ static const char VertexAttribs4dvNV_names[] =
     "";
 #endif
 
+#if defined(need_GL_IBM_multimode_draw_arrays)
+static const char MultiModeDrawArraysIBM_names[] = 
+    "pppii\0" /* Parameter signature */
+    "glMultiModeDrawArraysIBM\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program)
 static const char VertexAttrib4dARB_names[] = 
     "idddd\0" /* Parameter signature */
@@ -3926,6 +3956,13 @@ static const char VertexWeightfEXT_names[] =
     "";
 #endif
 
+#if defined(need_GL_EXT_coordinate_frame)
+static const char Binormal3fEXT_names[] = 
+    "fff\0" /* Parameter signature */
+    "glBinormal3fEXT\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
 static const char FogCoordfvEXT_names[] = 
     "p\0" /* Parameter signature */
@@ -4041,10 +4078,10 @@ static const char BlendFuncSeparateEXT_names[] =
     "";
 #endif
 
-#if defined(need_GL_APPLE_vertex_array_object)
-static const char IsVertexArrayAPPLE_names[] = 
-    "i\0" /* Parameter signature */
-    "glIsVertexArrayAPPLE\0"
+#if defined(need_GL_ARB_map_buffer_range)
+static const char MapBufferRange_names[] = 
+    "iiii\0" /* Parameter signature */
+    "glMapBufferRange\0"
     "";
 #endif
 
@@ -4289,6 +4326,13 @@ static const char SpriteParameterivSGIX_names[] =
     "";
 #endif
 
+#if defined(need_GL_EXT_provoking_vertex)
+static const char ProvokingVertexEXT_names[] = 
+    "i\0" /* Parameter signature */
+    "glProvokingVertexEXT\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_3)
 static const char MultiTexCoord1fARB_names[] = 
     "if\0" /* Parameter signature */
@@ -4396,6 +4440,13 @@ static const char WindowPos3ivMESA_names[] =
     "";
 #endif
 
+#if defined(need_GL_ARB_copy_buffer)
+static const char CopyBufferSubData_names[] = 
+    "iiiii\0" /* Parameter signature */
+    "glCopyBufferSubData\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
 static const char IsBufferARB_names[] = 
     "i\0" /* Parameter signature */
@@ -4560,13 +4611,6 @@ static const char Minmax_names[] =
     "";
 #endif
 
-#if defined(need_GL_SGIX_polynomial_ffd)
-static const char DeformationMap3dSGIX_names[] = 
-    "iddiiddiiddiip\0" /* Parameter signature */
-    "glDeformationMap3dSGIX\0"
-    "";
-#endif
-
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
 static const char FogCoorddvEXT_names[] = 
     "p\0" /* Parameter signature */
@@ -4933,8 +4977,15 @@ static const struct dri_extension_function GL_3DFX_tbuffer_functions[] = {
 static const struct dri_extension_function GL_APPLE_vertex_array_object_functions[] = {
     { DeleteVertexArraysAPPLE_names, DeleteVertexArraysAPPLE_remap_index, -1 },
     { GenVertexArraysAPPLE_names, GenVertexArraysAPPLE_remap_index, -1 },
-    { BindVertexArrayAPPLE_names, BindVertexArrayAPPLE_remap_index, -1 },
     { IsVertexArrayAPPLE_names, IsVertexArrayAPPLE_remap_index, -1 },
+    { BindVertexArrayAPPLE_names, BindVertexArrayAPPLE_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
+#if defined(need_GL_ARB_copy_buffer)
+static const struct dri_extension_function GL_ARB_copy_buffer_functions[] = {
+    { CopyBufferSubData_names, CopyBufferSubData_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -4972,6 +5023,14 @@ static const struct dri_extension_function GL_ARB_framebuffer_object_functions[]
 };
 #endif
 
+#if defined(need_GL_ARB_map_buffer_range)
+static const struct dri_extension_function GL_ARB_map_buffer_range_functions[] = {
+    { FlushMappedBufferRange_names, FlushMappedBufferRange_remap_index, -1 },
+    { MapBufferRange_names, MapBufferRange_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_ARB_matrix_palette)
 static const struct dri_extension_function GL_ARB_matrix_palette_functions[] = {
     { MatrixIndexusvARB_names, MatrixIndexusvARB_remap_index, -1 },
@@ -5080,6 +5139,16 @@ static const struct dri_extension_function GL_ARB_transpose_matrix_functions[] =
 };
 #endif
 
+#if defined(need_GL_ARB_vertex_array_object)
+static const struct dri_extension_function GL_ARB_vertex_array_object_functions[] = {
+    { DeleteVertexArraysAPPLE_names, DeleteVertexArraysAPPLE_remap_index, -1 },
+    { GenVertexArrays_names, GenVertexArrays_remap_index, -1 },
+    { BindVertexArray_names, BindVertexArray_remap_index, -1 },
+    { IsVertexArrayAPPLE_names, IsVertexArrayAPPLE_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_ARB_vertex_blend)
 static const struct dri_extension_function GL_ARB_vertex_blend_functions[] = {
     { WeightubvARB_names, WeightubvARB_remap_index, -1 },
@@ -5333,7 +5402,6 @@ static const struct dri_extension_function GL_EXT_coordinate_frame_functions[] =
     { Binormal3ivEXT_names, Binormal3ivEXT_remap_index, -1 },
     { Tangent3sEXT_names, Tangent3sEXT_remap_index, -1 },
     { Tangent3fvEXT_names, Tangent3fvEXT_remap_index, -1 },
-    { Binormal3fEXT_names, Binormal3fEXT_remap_index, -1 },
     { Tangent3dvEXT_names, Tangent3dvEXT_remap_index, -1 },
     { Binormal3bvEXT_names, Binormal3bvEXT_remap_index, -1 },
     { Binormal3dEXT_names, Binormal3dEXT_remap_index, -1 },
@@ -5342,6 +5410,7 @@ static const struct dri_extension_function GL_EXT_coordinate_frame_functions[] =
     { Tangent3ivEXT_names, Tangent3ivEXT_remap_index, -1 },
     { Tangent3dEXT_names, Tangent3dEXT_remap_index, -1 },
     { Binormal3svEXT_names, Binormal3svEXT_remap_index, -1 },
+    { Binormal3fEXT_names, Binormal3fEXT_remap_index, -1 },
     { Binormal3dvEXT_names, Binormal3dvEXT_remap_index, -1 },
     { Tangent3iEXT_names, Tangent3iEXT_remap_index, -1 },
     { Tangent3bvEXT_names, Tangent3bvEXT_remap_index, -1 },
@@ -5527,6 +5596,13 @@ static const struct dri_extension_function GL_EXT_polygon_offset_functions[] = {
 };
 #endif
 
+#if defined(need_GL_EXT_provoking_vertex)
+static const struct dri_extension_function GL_EXT_provoking_vertex_functions[] = {
+    { ProvokingVertexEXT_names, ProvokingVertexEXT_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_EXT_secondary_color)
 static const struct dri_extension_function GL_EXT_secondary_color_functions[] = {
     { SecondaryColor3iEXT_names, SecondaryColor3iEXT_remap_index, -1 },
@@ -6024,11 +6100,11 @@ static const struct dri_extension_function GL_SGIX_igloo_interface_functions[] =
 #if defined(need_GL_SGIX_instruments)
 static const struct dri_extension_function GL_SGIX_instruments_functions[] = {
     { ReadInstrumentsSGIX_names, ReadInstrumentsSGIX_remap_index, -1 },
+    { PollInstrumentsSGIX_names, PollInstrumentsSGIX_remap_index, -1 },
     { GetInstrumentsSGIX_names, GetInstrumentsSGIX_remap_index, -1 },
     { StartInstrumentsSGIX_names, StartInstrumentsSGIX_remap_index, -1 },
     { StopInstrumentsSGIX_names, StopInstrumentsSGIX_remap_index, -1 },
     { InstrumentsBufferSGIX_names, InstrumentsBufferSGIX_remap_index, -1 },
-    { PollInstrumentsSGIX_names, PollInstrumentsSGIX_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -6055,9 +6131,9 @@ static const struct dri_extension_function GL_SGIX_pixel_texture_functions[] = {
 #if defined(need_GL_SGIX_polynomial_ffd)
 static const struct dri_extension_function GL_SGIX_polynomial_ffd_functions[] = {
     { LoadIdentityDeformationMapSGIX_names, LoadIdentityDeformationMapSGIX_remap_index, -1 },
+    { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 },
     { DeformSGIX_names, DeformSGIX_remap_index, -1 },
     { DeformationMap3fSGIX_names, DeformationMap3fSGIX_remap_index, -1 },
-    { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
diff --git a/src/mesa/drivers/dri/common/xmlpool.h b/src/mesa/drivers/dri/common/xmlpool.h
index 7fbc6e800d..587517ea10 100644
--- a/src/mesa/drivers/dri/common/xmlpool.h
+++ b/src/mesa/drivers/dri/common/xmlpool.h
@@ -60,6 +60,10 @@
 #define DRI_CONF_OPT_BEGIN(name,type,def) \
 "<option name=\""#name"\" type=\""#type"\" default=\""#def"\">\n"
 
+/** \brief Begin an option definition with qouted default value */
+#define DRI_CONF_OPT_BEGIN_Q(name,type,def) \
+"<option name=\""#name"\" type=\""#type"\" default="#def">\n"
+
 /** \brief Begin an option definition with restrictions on valid values */
 #define DRI_CONF_OPT_BEGIN_V(name,type,def,valid) \
 "<option name=\""#name"\" type=\""#type"\" default=\""#def"\" valid=\""valid"\">\n"
diff --git a/src/mesa/drivers/dri/fb/fb_egl.c b/src/mesa/drivers/dri/fb/fb_egl.c
index 35c268441c..dee67feb5a 100644
--- a/src/mesa/drivers/dri/fb/fb_egl.c
+++ b/src/mesa/drivers/dri/fb/fb_egl.c
@@ -472,8 +472,8 @@ fbCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext sh
    c->Base.DrawSurface = EGL_NO_SURFACE;
    c->Base.ReadSurface = EGL_NO_SURFACE;
 
-   /* generate handle and insert into hash table */
-   _eglSaveContext(&c->Base);
+   /* link to display */
+   _eglLinkContext(&c->Base, disp);
    assert(c->Base.Handle);
 
    /* Init default driver functions then plug in our FBdev-specific functions
@@ -604,13 +604,9 @@ static EGLBoolean
 fbDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    fbSurface *fs = Lookup_fbSurface(surface);
-   _eglRemoveSurface(&fs->Base);
-   if (fs->Base.IsBound) {
-      fs->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   _eglUnlinkSurface(&fs->Base);
+   if (!fs->Base.IsBound)
       free(fs);
-   }
    return EGL_TRUE;
 }
 
@@ -619,13 +615,9 @@ static EGLBoolean
 fbDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
 {
    fbContext *fc = Lookup_fbContext(context);
-   _eglRemoveContext(&fc->Base);
-   if (fc->Base.IsBound) {
-      fc->Base.DeletePending = EGL_TRUE;
-   }
-   else {
+   _eglUnlinkContext(&fc->Base);
+   if (!fc->Base.IsBound)
       free(fc);
-   }
    return EGL_TRUE;
 }
 
@@ -688,7 +680,7 @@ fbCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    surface->mesa_framebuffer = _mesa_create_framebuffer(&vis);
    if (!surface->mesa_framebuffer) {
       free(surface);
-      _eglRemoveSurface(&surface->Base);
+      _eglUnlinkSurface(&surface->Base);
       return EGL_NO_SURFACE;
    }
 
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index 9f4bd1699f..beaf9a4b12 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -19,6 +19,7 @@ DRIVER_SOURCES = \
 	intel_batchbuffer.c \
 	intel_clear.c \
 	intel_extensions.c \
+	intel_generatemipmap.c \
 	intel_mipmap_tree.c \
 	intel_tex_layout.c \
 	intel_tex_image.c \
diff --git a/src/mesa/drivers/dri/i915/i830_reg.h b/src/mesa/drivers/dri/i915/i830_reg.h
index d210c2d08e..db16871001 100644
--- a/src/mesa/drivers/dri/i915/i830_reg.h
+++ b/src/mesa/drivers/dri/i915/i830_reg.h
@@ -48,19 +48,6 @@
 #define AA_LINE_ENABLE			((1<<1) | 1)
 #define AA_LINE_DISABLE			(1<<1)
 
-#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
-/* Dword 1 */
-#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
-#define BUF_3D_ID_DEPTH 	(0x7<<24)
-#define BUF_3D_USE_FENCE	(1<<23)
-#define BUF_3D_TILED_SURFACE	(1<<22)
-#define BUF_3D_TILE_WALK_X	0
-#define BUF_3D_TILE_WALK_Y	(1<<21)
-#define BUF_3D_PITCH(x)         (((x)/4)<<2)
-/* Dword 2 */
-#define BUF_3D_ADDR(x)		((x) & ~0x3)
-
-
 #define _3DSTATE_COLOR_FACTOR_CMD	(CMD_3D | (0x1d<<24) | (0x1<<16))
 
 #define _3DSTATE_COLOR_FACTOR_N_CMD(stage)	(CMD_3D | (0x1d<<24) | \
diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c
index 753c25b57e..6f998fa6f7 100644
--- a/src/mesa/drivers/dri/i915/i830_texstate.c
+++ b/src/mesa/drivers/dri/i915/i830_texstate.c
@@ -174,14 +174,16 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    state[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 |
                                (LOAD_TEXTURE_MAP0 << unit) | 4);
 
-/*    state[I830_TEXREG_TM0S0] = (TM0S0_USE_FENCE | */
-/* 			       t->intel.TextureOffset); */
-
-
    state[I830_TEXREG_TM0S1] =
       (((firstImage->Height - 1) << TM0S1_HEIGHT_SHIFT) |
        ((firstImage->Width - 1) << TM0S1_WIDTH_SHIFT) | format);
 
+   if (intelObj->mt->region->tiling != I915_TILING_NONE) {
+      state[I830_TEXREG_TM0S1] |= TM0S1_TILED_SURFACE;
+      if (intelObj->mt->region->tiling == I915_TILING_Y)
+	 state[I830_TEXREG_TM0S1] |= TM0S1_TILE_WALK;
+   }
+
    state[I830_TEXREG_TM0S2] =
       ((((pitch / 4) - 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK);
 
diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c
index 3bf02de61f..9c6f891dd3 100644
--- a/src/mesa/drivers/dri/i915/i830_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i830_vtbl.c
@@ -552,7 +552,7 @@ i830_emit_state(struct intel_context *intel)
          if (state->tex_buffer[i]) {
             OUT_RELOC(state->tex_buffer[i],
 		      I915_GEM_DOMAIN_SAMPLER, 0,
-                      state->tex_offset[i] | TM0S0_USE_FENCE);
+                      state->tex_offset[i]);
          }
 	 else if (state == &i830->meta) {
 	    assert(i == 0);
@@ -634,21 +634,11 @@ i830_state_draw_region(struct intel_context *intel,
    /*
     * Set stride/cpp values
     */
-   if (color_region) {
-      state->Buffer[I830_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I830_DESTREG_CBUFADDR1] =
-         (BUF_3D_ID_COLOR_BACK |
-          BUF_3D_PITCH(color_region->pitch * color_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I830_DESTREG_CBUFADDR0],
+				color_region, BUF_3D_ID_COLOR_BACK);
 
-   if (depth_region) {
-      state->Buffer[I830_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I830_DESTREG_DBUFADDR1] =
-         (BUF_3D_ID_DEPTH |
-          BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I830_DESTREG_DBUFADDR0],
+				depth_region, BUF_3D_ID_DEPTH);
 
    /*
     * Compute/set I830_DESTREG_DV1 value
@@ -718,26 +708,6 @@ i830_set_draw_region(struct intel_context *intel,
    i830_state_draw_region(intel, &i830->state, color_regions[0], depth_region);
 }
 
-#if 0
-static void
-i830_update_color_z_regions(intelContextPtr intel,
-                            const intelRegion * colorRegion,
-                            const intelRegion * depthRegion)
-{
-   i830ContextPtr i830 = I830_CONTEXT(intel);
-
-   i830->state.Buffer[I830_DESTREG_CBUFADDR1] =
-      (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(colorRegion->pitch) |
-       BUF_3D_USE_FENCE);
-   i830->state.Buffer[I830_DESTREG_CBUFADDR2] = colorRegion->offset;
-
-   i830->state.Buffer[I830_DESTREG_DBUFADDR1] =
-      (BUF_3D_ID_DEPTH | BUF_3D_PITCH(depthRegion->pitch) | BUF_3D_USE_FENCE);
-   i830->state.Buffer[I830_DESTREG_DBUFADDR2] = depthRegion->offset;
-}
-#endif
-
-
 /* This isn't really handled at the moment.
  */
 static void
@@ -767,12 +737,6 @@ i830_assert_not_dirty( struct intel_context *intel )
    assert(!get_dirty(state));
 }
 
-static void
-i830_note_unlock( struct intel_context *intel )
-{
-    /* nothing */
-}
-
 void
 i830InitVtbl(struct i830_context *i830)
 {
@@ -787,6 +751,5 @@ i830InitVtbl(struct i830_context *i830)
    i830->intel.vtbl.render_start = i830_render_start;
    i830->intel.vtbl.render_prevalidate = i830_render_prevalidate;
    i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty;
-   i830->intel.vtbl.note_unlock = i830_note_unlock; 
    i830->intel.vtbl.finish_batch = intel_finish_vb;
 }
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index 1f9f363df9..367d2a3b64 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -73,7 +73,7 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state)
          p->params_uptodate = 0;
    }
 
-   if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM))
+   if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS))
       i915_update_fog(ctx);
 }
 
diff --git a/src/mesa/drivers/dri/i915/i915_reg.h b/src/mesa/drivers/dri/i915/i915_reg.h
index 8891e11c6f..84db58ea95 100644
--- a/src/mesa/drivers/dri/i915/i915_reg.h
+++ b/src/mesa/drivers/dri/i915/i915_reg.h
@@ -93,20 +93,6 @@
 
 /* 3DSTATE_BIN_CONTROL p141 */
 
-/* p143 */
-#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
-/* Dword 1 */
-#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
-#define BUF_3D_ID_DEPTH 	(0x7<<24)
-#define BUF_3D_USE_FENCE	(1<<23)
-#define BUF_3D_TILED_SURFACE	(1<<22)
-#define BUF_3D_TILE_WALK_X	0
-#define BUF_3D_TILE_WALK_Y	(1<<21)
-#define BUF_3D_PITCH(x)         (((x)/4)<<2)
-/* Dword 2 */
-#define BUF_3D_ADDR(x)		((x) & ~0x3)
-
-
 /* 3DSTATE_CHROMA_KEY */
 
 /* 3DSTATE_CLEAR_PARAMETERS, p150 */
@@ -155,6 +141,7 @@
 /* p161 */
 #define _3DSTATE_DST_BUF_VARS_CMD	(CMD_3D | (0x1d<<24) | (0x85<<16))
 /* Dword 1 */
+#define CLASSIC_EARLY_DEPTH		(1<<31)
 #define TEX_DEFAULT_COLOR_OGL           (0<<30)
 #define TEX_DEFAULT_COLOR_D3D           (1<<30)
 #define ZR_EARLY_DEPTH                  (1<<29)
diff --git a/src/mesa/drivers/dri/i915/i915_tex_layout.c b/src/mesa/drivers/dri/i915/i915_tex_layout.c
index 7cc1c096e4..d9588e5b56 100644
--- a/src/mesa/drivers/dri/i915/i915_tex_layout.c
+++ b/src/mesa/drivers/dri/i915/i915_tex_layout.c
@@ -55,6 +55,17 @@ static GLint step_offsets[6][2] = {
    [FACE_NEG_Z] = {-1, 1},
 };
 
+
+static GLint bottom_offsets[6] = {
+   [FACE_POS_X] = 16 + 0 * 8,
+   [FACE_POS_Y] = 16 + 1 * 8,
+   [FACE_POS_Z] = 16 + 2 * 8,
+   [FACE_NEG_X] = 16 + 3 * 8,
+   [FACE_NEG_Y] = 16 + 4 * 8,
+   [FACE_NEG_Z] = 16 + 5 * 8,
+};
+
+
 /**
  * Cube texture map layout for i830M-GM915.
  *
@@ -101,7 +112,8 @@ static GLint step_offsets[6][2] = {
  */
 static void
 i915_miptree_layout_cube(struct intel_context *intel,
-			 struct intel_mipmap_tree * mt)
+			 struct intel_mipmap_tree * mt,
+			 uint32_t tiling)
 {
    const GLuint dim = mt->width0;
    GLuint face;
@@ -111,7 +123,7 @@ i915_miptree_layout_cube(struct intel_context *intel,
    assert(lvlWidth == lvlHeight); /* cubemap images are square */
 
    /* double pitch for cube layouts */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, dim * 2);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, dim * 2);
    mt->total_height = dim * 4;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
@@ -145,7 +157,8 @@ i915_miptree_layout_cube(struct intel_context *intel,
 
 static void
 i915_miptree_layout_3d(struct intel_context *intel,
-		       struct intel_mipmap_tree * mt)
+		       struct intel_mipmap_tree * mt,
+		       uint32_t tiling)
 {
    GLuint width = mt->width0;
    GLuint height = mt->height0;
@@ -154,7 +167,7 @@ i915_miptree_layout_3d(struct intel_context *intel,
    GLint level;
 
    /* Calculate the size of a single slice. */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
 
    /* XXX: hardware expects/requires 9 levels at minimum. */
    for (level = mt->first_level; level <= MAX2(8, mt->last_level); level++) {
@@ -189,14 +202,15 @@ i915_miptree_layout_3d(struct intel_context *intel,
 
 static void
 i915_miptree_layout_2d(struct intel_context *intel,
-		       struct intel_mipmap_tree * mt)
+		       struct intel_mipmap_tree * mt,
+		       uint32_t tiling)
 {
    GLuint width = mt->width0;
    GLuint height = mt->height0;
    GLuint img_height;
    GLint level;
 
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
    mt->total_height = 0;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
@@ -217,19 +231,20 @@ i915_miptree_layout_2d(struct intel_context *intel,
 }
 
 GLboolean
-i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
+i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt,
+		    uint32_t tiling)
 {
    switch (mt->target) {
    case GL_TEXTURE_CUBE_MAP:
-      i915_miptree_layout_cube(intel, mt);
+      i915_miptree_layout_cube(intel, mt, tiling);
       break;
    case GL_TEXTURE_3D:
-      i915_miptree_layout_3d(intel, mt);
+      i915_miptree_layout_3d(intel, mt, tiling);
       break;
    case GL_TEXTURE_1D:
    case GL_TEXTURE_2D:
    case GL_TEXTURE_RECTANGLE_ARB:
-      i915_miptree_layout_2d(intel, mt);
+      i915_miptree_layout_2d(intel, mt, tiling);
       break;
    default:
       _mesa_problem(NULL, "Unexpected tex target in i915_miptree_layout()");
@@ -297,7 +312,7 @@ i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
  * +---+   +---+   +---+   +---+   +---+   +---+
  *
  * The bottom row continues with the remaining 2x2 then the 1x1 mip contents
- * in order, with each of them aligned to a 4x4 block boundary.  Thus, for
+ * in order, with each of them aligned to a 8x8 block boundary.  Thus, for
  * 32x32 cube maps and smaller, the bottom row layout is going to dictate the
  * pitch of the tree.  For a tree with 4x4 images, the pitch is at least
  * 14 * 8 = 112 texels, for 2x2 it is at least 12 * 8 texels, and for 1x1
@@ -306,7 +321,8 @@ i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
 
 static void
 i945_miptree_layout_cube(struct intel_context *intel,
-			 struct intel_mipmap_tree * mt)
+			 struct intel_mipmap_tree * mt,
+			 uint32_t tiling)
 {
    const GLuint dim = mt->width0;
    GLuint face;
@@ -320,9 +336,9 @@ i945_miptree_layout_cube(struct intel_context *intel,
     * or the final row of 4x4, 2x2 and 1x1 faces below this.
     */
    if (dim > 32)
-      mt->pitch = intel_miptree_pitch_align (intel, mt, dim * 2);
+      mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, dim * 2);
    else
-      mt->pitch = intel_miptree_pitch_align (intel, mt, 14 * 8);
+      mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, 14 * 8);
 
    if (dim >= 4)
       mt->total_height = dim * 4 + 4;
@@ -375,10 +391,11 @@ i945_miptree_layout_cube(struct intel_context *intel,
 	       x = (face - 4) * 8;
 	       break;
 	    }
+	    break;
 
 	 case 2:
 	    y = mt->total_height - 4;
-	    x = 16 + face * 8;
+	    x = bottom_offsets[face];
 	    break;
 
 	 case 1:
@@ -396,7 +413,8 @@ i945_miptree_layout_cube(struct intel_context *intel,
 
 static void
 i945_miptree_layout_3d(struct intel_context *intel,
-		       struct intel_mipmap_tree * mt)
+		       struct intel_mipmap_tree * mt,
+		       uint32_t tiling)
 {
    GLuint width = mt->width0;
    GLuint height = mt->height0;
@@ -405,7 +423,7 @@ i945_miptree_layout_3d(struct intel_context *intel,
    GLuint pack_y_pitch;
    GLuint level;
 
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
    mt->total_height = 0;
 
    pack_y_pitch = MAX2(mt->height0, 2);
@@ -450,22 +468,23 @@ i945_miptree_layout_3d(struct intel_context *intel,
 }
 
 GLboolean
-i945_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt)
+i945_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt,
+		    uint32_t tiling)
 {
    switch (mt->target) {
    case GL_TEXTURE_CUBE_MAP:
       if (mt->compressed)
-	 i945_miptree_layout_cube(intel, mt);
+	 i945_miptree_layout_cube(intel, mt, tiling);
       else
-	 i915_miptree_layout_cube(intel, mt);
+	 i915_miptree_layout_cube(intel, mt, tiling);
       break;
    case GL_TEXTURE_3D:
-      i945_miptree_layout_3d(intel, mt);
+      i945_miptree_layout_3d(intel, mt, tiling);
       break;
    case GL_TEXTURE_1D:
    case GL_TEXTURE_2D:
    case GL_TEXTURE_RECTANGLE_ARB:
-      i945_miptree_layout_2d(intel, mt);
+      i945_miptree_layout_2d(intel, mt, tiling);
       break;
    default:
       _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()");
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index a37dd7f4fb..32d4b30cf9 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -185,8 +185,13 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 
    state[I915_TEXREG_MS3] =
       (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) |
-       ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format |
-       MS3_USE_FENCE_REGS);
+       ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format);
+
+   if (intelObj->mt->region->tiling != I915_TILING_NONE) {
+      state[I915_TEXREG_MS3] |= MS3_TILED_SURFACE;
+      if (intelObj->mt->region->tiling == I915_TILING_Y)
+	 state[I915_TEXREG_MS3] |= MS3_TILE_WALK;
+   }
 
    state[I915_TEXREG_MS4] =
      ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK |
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index 115004616f..fe1be93a6d 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -42,6 +42,7 @@
 #include "intel_regions.h"
 #include "intel_tris.h"
 #include "intel_fbo.h"
+#include "intel_chipset.h"
 
 #include "i915_reg.h"
 #include "i915_context.h"
@@ -529,6 +530,23 @@ i915_destroy_context(struct intel_context *intel)
    _tnl_free_vertices(&intel->ctx);
 }
 
+void
+i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
+			     uint32_t buffer_id)
+{
+   state[0] = _3DSTATE_BUF_INFO_CMD;
+   state[1] = buffer_id;
+
+   if (region != NULL) {
+      state[1] |= BUF_3D_PITCH(region->pitch * region->cpp);
+
+      if (region->tiling != I915_TILING_NONE) {
+	 state[1] |= BUF_3D_TILED_SURFACE;
+	 if (region->tiling == I915_TILING_Y)
+	    state[1] |= BUF_3D_TILE_WALK_Y;
+      }
+   }
+}
 
 /**
  * Set the drawing regions for the color and depth/stencil buffers.
@@ -562,21 +580,11 @@ i915_state_draw_region(struct intel_context *intel,
    /*
     * Set stride/cpp values
     */
-   if (color_region) {
-      state->Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I915_DESTREG_CBUFADDR1] =
-         (BUF_3D_ID_COLOR_BACK |
-          BUF_3D_PITCH(color_region->pitch * color_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I915_DESTREG_CBUFADDR0],
+				color_region, BUF_3D_ID_COLOR_BACK);
 
-   if (depth_region) {
-      state->Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
-      state->Buffer[I915_DESTREG_DBUFADDR1] =
-         (BUF_3D_ID_DEPTH |
-          BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) |
-          BUF_3D_USE_FENCE);
-   }
+   i915_set_buf_info_for_region(&state->Buffer[I915_DESTREG_DBUFADDR0],
+				depth_region, BUF_3D_ID_DEPTH);
 
    /*
     * Compute/set I915_DESTREG_DV1 value
@@ -604,6 +612,14 @@ i915_state_draw_region(struct intel_context *intel,
       }
    }
 
+   /* This isn't quite safe, thus being hidden behind an option.  When changing
+    * the value of this bit, the pipeline needs to be MI_FLUSHed.  And it
+    * can only be set when a depth buffer is already defined.
+    */
+   if (IS_945(intel->intelScreen->deviceID) && intel->use_early_z &&
+       depth_region->tiling != I915_TILING_NONE)
+      value |= CLASSIC_EARLY_DEPTH;
+
    if (depth_region && depth_region->cpp == 4) {
       value |= DEPTH_FRMT_24_FIXED_8_OTHER;
    }
@@ -676,13 +692,6 @@ i915_assert_not_dirty( struct intel_context *intel )
    assert(!dirty);
 }
 
-static void
-i915_note_unlock( struct intel_context *intel )
-{
-    /* nothing */
-}
-
-
 void
 i915InitVtbl(struct i915_context *i915)
 {
@@ -697,6 +706,5 @@ i915InitVtbl(struct i915_context *i915)
    i915->intel.vtbl.update_texture_state = i915UpdateTextureState;
    i915->intel.vtbl.flush_cmd = i915_flush_cmd;
    i915->intel.vtbl.assert_not_dirty = i915_assert_not_dirty;
-   i915->intel.vtbl.note_unlock = i915_note_unlock; 
    i915->intel.vtbl.finish_batch = intel_finish_vb;
 }
diff --git a/src/mesa/drivers/dri/i915/intel_generatemipmap.c b/src/mesa/drivers/dri/i915/intel_generatemipmap.c
new file mode 120000
index 0000000000..4c6b37ada0
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_generatemipmap.c
@@ -0,0 +1 @@
+../intel/intel_generatemipmap.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_pixel_read.c b/src/mesa/drivers/dri/i915/intel_pixel_read.c
index 56087aacd4..cc4589f4d4 100644..120000
--- a/src/mesa/drivers/dri/i915/intel_pixel_read.c
+++ b/src/mesa/drivers/dri/i915/intel_pixel_read.c
@@ -1,306 +1 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include "main/glheader.h"
-#include "main/enums.h"
-#include "main/mtypes.h"
-#include "main/macros.h"
-#include "main/image.h"
-#include "main/bufferobj.h"
-#include "swrast/swrast.h"
-
-#include "intel_screen.h"
-#include "intel_context.h"
-#include "intel_batchbuffer.h"
-#include "intel_blit.h"
-#include "intel_buffers.h"
-#include "intel_regions.h"
-#include "intel_pixel.h"
-#include "intel_buffer_objects.h"
-
-/* For many applications, the new ability to pull the source buffers
- * back out of the GTT and then do the packing/conversion operations
- * in software will be as much of an improvement as trying to get the
- * blitter and/or texture engine to do the work. 
- *
- * This step is gated on private backbuffers.
- * 
- * Obviously the frontbuffer can't be pulled back, so that is either
- * an argument for blit/texture readpixels, or for blitting to a
- * temporary and then pulling that back.
- *
- * When the destination is a pbo, however, it's not clear if it is
- * ever going to be pulled to main memory (though the access param
- * will be a good hint).  So it sounds like we do want to be able to
- * choose between blit/texture implementation on the gpu and pullback
- * and cpu-based copying.
- *
- * Unless you can magically turn client memory into a PBO for the
- * duration of this call, there will be a cpu-based copying step in
- * any case.
- */
-
-
-static GLboolean
-do_texture_readpixels(GLcontext * ctx,
-                      GLint x, GLint y, GLsizei width, GLsizei height,
-                      GLenum format, GLenum type,
-                      const struct gl_pixelstore_attrib *pack,
-                      struct intel_region *dest_region)
-{
-#if 0
-   struct intel_context *intel = intel_context(ctx);
-   intelScreenPrivate *screen = intel->intelScreen;
-   GLint pitch = pack->RowLength ? pack->RowLength : width;
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   int textureFormat;
-   GLenum glTextureFormat;
-   int destFormat, depthFormat, destPitch;
-   drm_clip_rect_t tmp;
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-
-   if (ctx->_ImageTransferState ||
-       pack->SwapBytes || pack->LsbFirst || !pack->Invert) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   intel->vtbl.meta_texrect_source(intel, intel_readbuf_region(intel));
-
-   if (!intel->vtbl.meta_render_dest(intel, dest_region, type, format)) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         fprintf(stderr, "%s: couldn't set dest %s/%s\n",
-                 __FUNCTION__,
-                 _mesa_lookup_enum_by_nr(type),
-                 _mesa_lookup_enum_by_nr(format));
-      return GL_FALSE;
-   }
-
-   LOCK_HARDWARE(intel);
-
-   if (intel->driDrawable->numClipRects) {
-      intel->vtbl.install_meta_state(intel);
-      intel->vtbl.meta_no_depth_write(intel);
-      intel->vtbl.meta_no_stencil_write(intel);
-
-      if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) {
-         UNLOCK_HARDWARE(intel);
-         SET_STATE(i830, state);
-         if (INTEL_DEBUG & DEBUG_PIXEL)
-            fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__);
-         return GL_TRUE;
-      }
-
-      y = dPriv->h - y - height;
-      x += dPriv->x;
-      y += dPriv->y;
-
-
-      /* Set the frontbuffer up as a large rectangular texture.
-       */
-      intel->vtbl.meta_tex_rect_source(intel, src_region, textureFormat);
-
-
-      intel->vtbl.meta_texture_blend_replace(i830, glTextureFormat);
-
-
-      /* Set the 3d engine to draw into the destination region:
-       */
-
-      intel->vtbl.meta_draw_region(intel, dest_region);
-      intel->vtbl.meta_draw_format(intel, destFormat, depthFormat);     /* ?? */
-
-
-      /* Draw a single quad, no cliprects:
-       */
-      intel->vtbl.meta_disable_cliprects(intel);
-
-      intel->vtbl.draw_quad(intel,
-                            0, width, 0, height,
-                            0x00ff00ff, x, x + width, y, y + height);
-
-      intel->vtbl.leave_meta_state(intel);
-   }
-   UNLOCK_HARDWARE(intel);
-
-   intel_region_wait_fence(ctx, dest_region);   /* required by GL */
-   return GL_TRUE;
-#endif
-
-   return GL_FALSE;
-}
-
-
-
-
-static GLboolean
-do_blit_readpixels(GLcontext * ctx,
-                   GLint x, GLint y, GLsizei width, GLsizei height,
-                   GLenum format, GLenum type,
-                   const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_region *src = intel_readbuf_region(intel);
-   struct intel_buffer_object *dst = intel_buffer_object(pack->BufferObj);
-   GLuint dst_offset;
-   GLuint rowLength;
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      _mesa_printf("%s\n", __FUNCTION__);
-
-   if (!src)
-      return GL_FALSE;
-
-   if (dst) {
-      /* XXX This validation should be done by core mesa:
-       */
-      if (!_mesa_validate_pbo_access(2, pack, width, height, 1,
-                                     format, type, pixels)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels");
-         return GL_TRUE;
-      }
-   }
-   else {
-      /* PBO only for now:
-       */
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         _mesa_printf("%s - not PBO\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-
-   if (ctx->_ImageTransferState ||
-       !intel_check_blit_format(src, format, type)) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         _mesa_printf("%s - bad format for blit\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         _mesa_printf("%s: bad packing params\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   if (pack->RowLength > 0)
-      rowLength = pack->RowLength;
-   else
-      rowLength = width;
-
-   if (pack->Invert) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         _mesa_printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-   else {
-      rowLength = -rowLength;
-   }
-
-   /* XXX 64-bit cast? */
-   dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height,
-                                             format, type, 0, 0, 0);
-
-
-   /* Although the blits go on the command buffer, need to do this and
-    * fire with lock held to guarentee cliprects are correct.
-    */
-   intelFlush(&intel->ctx);
-   LOCK_HARDWARE(intel);
-
-   if (intel->driDrawable->numClipRects) {
-      GLboolean all = (width * height * src->cpp == dst->Base.Size &&
-                       x == 0 && dst_offset == 0);
-
-      dri_bo *dst_buffer = intel_bufferobj_buffer(intel, dst,
-						  all ? INTEL_WRITE_FULL :
-						  INTEL_WRITE_PART);
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-      int nbox = dPriv->numClipRects;
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t rect;
-      drm_clip_rect_t src_rect;
-      int i;
-
-      src_rect.x1 = dPriv->x + x;
-      src_rect.y1 = dPriv->y + dPriv->h - (y + height);
-      src_rect.x2 = src_rect.x1 + width;
-      src_rect.y2 = src_rect.y1 + height;
-
-
-
-      for (i = 0; i < nbox; i++) {
-         if (!intel_intersect_cliprects(&rect, &src_rect, &box[i]))
-            continue;
-
-         intelEmitCopyBlit(intel,
-                           src->cpp,
-                           src->pitch, src->buffer, 0, src->tiling,
-                           rowLength, dst_buffer, dst_offset, GL_FALSE,
-                           rect.x1,
-                           rect.y1,
-                           rect.x1 - src_rect.x1,
-                           rect.y2 - src_rect.y2,
-                           rect.x2 - rect.x1, rect.y2 - rect.y1,
-			   GL_COPY);
-      }
-   }
-   UNLOCK_HARDWARE(intel);
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      _mesa_printf("%s - DONE\n", __FUNCTION__);
-
-   return GL_TRUE;
-}
-
-void
-intelReadPixels(GLcontext * ctx,
-                GLint x, GLint y, GLsizei width, GLsizei height,
-                GLenum format, GLenum type,
-                const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
-{
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   intelFlush(ctx);
-
-   if (do_blit_readpixels
-       (ctx, x, y, width, height, format, type, pack, pixels))
-      return;
-
-   if (do_texture_readpixels
-       (ctx, x, y, width, height, format, type, pack, pixels))
-      return;
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
-
-   _swrast_ReadPixels(ctx, x, y, width, height, format, type, pack, pixels);
-}
+../intel/intel_pixel_read.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index 1d39278cbf..a905455342 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -1255,11 +1255,9 @@ intel_meta_draw_poly(struct intel_context *intel,
 {
    union fi *vb;
    GLint i;
-   GLboolean was_locked = intel->locked;
    unsigned int saved_vertex_size = intel->vertex_size;
 
-   if (!was_locked)
-       LOCK_HARDWARE(intel);
+   LOCK_HARDWARE(intel);
 
    intel->vertex_size = 6;
 
@@ -1283,8 +1281,7 @@ intel_meta_draw_poly(struct intel_context *intel,
 
    intel->vertex_size = saved_vertex_size;
 
-   if (!was_locked)
-       UNLOCK_HARDWARE(intel);
+   UNLOCK_HARDWARE(intel);
 }
 
 static void
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 2934414d99..00a42111da 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -14,6 +14,7 @@ DRIVER_SOURCES = \
 	intel_decode.c \
 	intel_extensions.c \
 	intel_fbo.c \
+	intel_generatemipmap.c \
 	intel_mipmap_tree.c \
 	intel_regions.c \
 	intel_screen.c \
@@ -22,6 +23,7 @@ DRIVER_SOURCES = \
 	intel_pixel_bitmap.c \
 	intel_pixel_copy.c \
 	intel_pixel_draw.c \
+	intel_pixel_read.c \
 	intel_state.c \
 	intel_swapbuffers.c \
 	intel_tex.c \
@@ -69,6 +71,7 @@ DRIVER_SOURCES = \
 	brw_vs_constval.c \
 	brw_vs_emit.c \
 	brw_vs_state.c \
+	brw_vs_surface_state.c \
 	brw_vtbl.c \
 	brw_wm.c \
 	brw_wm_debug.c \
diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c
index 5cffcebde4..54d30a3f42 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.c
+++ b/src/mesa/drivers/dri/i965/brw_clip.c
@@ -65,21 +65,31 @@ static void compile_clip_prog( struct brw_context *brw,
    c.func.single_program_flow = 1;
 
    c.key = *key;
-
+   c.need_ff_sync = BRW_IS_IGDNG(brw);
 
    /* Need to locate the two positions present in vertex + header.
     * These are currently hardcoded:
     */
    c.header_position_offset = ATTR_SIZE;
 
-   for (i = 0, delta = REG_SIZE; i < VERT_RESULT_MAX; i++)
+   if (BRW_IS_IGDNG(brw))
+       delta = 3 * REG_SIZE;
+   else
+       delta = REG_SIZE;
+
+   for (i = 0; i < VERT_RESULT_MAX; i++)
       if (c.key.attrs & (1<<i)) {
 	 c.offset[i] = delta;
 	 delta += ATTR_SIZE;
       }
 
    c.nr_attrs = brw_count_bits(c.key.attrs);
-   c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+   
+   if (BRW_IS_IGDNG(brw))
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 3;  /* are vertices packed, or reg-aligned? */
+   else
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+
    c.nr_bytes = c.nr_regs * REG_SIZE;
 
    c.prog_data.clip_mode = c.key.clip_mode; /* XXX */
@@ -148,7 +158,11 @@ static void upload_clip_prog(struct brw_context *brw)
    key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
    /* _NEW_TRANSFORM */
    key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled);
-   key.clip_mode = BRW_CLIPMODE_NORMAL;
+
+   if (BRW_IS_IGDNG(brw))
+       key.clip_mode = BRW_CLIPMODE_KERNEL_CLIP;
+   else
+       key.clip_mode = BRW_CLIPMODE_NORMAL;
 
    /* _NEW_POLYGON */
    if (key.primitive == GL_TRIANGLES) {
diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h
index e06747864b..12e8548df1 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.h
+++ b/src/mesa/drivers/dri/i965/brw_clip.h
@@ -117,6 +117,7 @@ struct brw_clip_compile {
 
    GLuint header_position_offset;
    GLuint offset[VERT_ATTRIB_MAX];
+   GLboolean need_ff_sync;
 };
 
 #define ATTR_SIZE  (4*4)
@@ -171,5 +172,5 @@ struct brw_reg get_tmp( struct brw_clip_compile *c );
 
 void brw_clip_project_position(struct brw_clip_compile *c,
              struct brw_reg pos );
-
+void brw_clip_ff_sync(struct brw_clip_compile *c);
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_clip_line.c b/src/mesa/drivers/dri/i965/brw_clip_line.c
index d830e49e50..9abd0642aa 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_line.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_line.c
@@ -130,7 +130,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
    struct brw_instruction *plane_loop;
    struct brw_instruction *plane_active;
    struct brw_instruction *is_negative;
-   struct brw_instruction *is_neg2;
+   struct brw_instruction *is_neg2 = NULL;
    struct brw_instruction *not_culled;
    struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD);
 
@@ -148,7 +148,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
    brw_clip_init_clipmask(c);
 
    /* -ve rhw workaround */
-   if (!BRW_IS_G4X(p->brw)) {
+   if (BRW_IS_965(p->brw)) {
       brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
       brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2),
               brw_imm_ud(1<<20));
@@ -185,7 +185,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
               * Both can be negative on GM965/G965 due to RHW workaround
               * if so, this object should be rejected.
               */
-             if (!BRW_IS_G4X(p->brw)) {
+             if (BRW_IS_965(p->brw)) {
                  brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_LE, c->reg.dp0, brw_imm_f(0.0));
                  is_neg2 = brw_IF(p, BRW_EXECUTE_1);
                  {
@@ -210,7 +210,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
 
              /* If both are positive, do nothing */
              /* Only on GM965/G965 */
-             if (!BRW_IS_G4X(p->brw)) {
+             if (BRW_IS_965(p->brw)) {
                  brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0));
                  is_neg2 = brw_IF(p, BRW_EXECUTE_1);
              }
@@ -225,7 +225,7 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
                  brw_set_predicate_control(p, BRW_PREDICATE_NONE);
              }
 
-             if (!BRW_IS_G4X(p->brw)) {
+             if (BRW_IS_965(p->brw)) {
                  brw_ENDIF(p, is_neg2);
              }
          }
@@ -246,6 +246,8 @@ static void clip_and_emit_line( struct brw_clip_compile *c )
 
    brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1);
    brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0));
+   if (c->need_ff_sync)
+	   brw_clip_ff_sync(c);      
    not_culled = brw_IF(p, BRW_EXECUTE_1);
    {
       brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, GL_FALSE);
diff --git a/src/mesa/drivers/dri/i965/brw_clip_point.c b/src/mesa/drivers/dri/i965/brw_clip_point.c
index d17b199b89..9738299168 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_point.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_point.c
@@ -50,5 +50,7 @@ void brw_emit_point_clip( struct brw_clip_compile *c )
    /* Send an empty message to kill the thread:
     */
    brw_clip_tri_alloc_regs(c, 0);
+   if (c->need_ff_sync)
+	   brw_clip_ff_sync(c);      
    brw_clip_kill_thread(c);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c
index 9b0d7eab7b..5762c9577c 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_state.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_state.c
@@ -95,7 +95,14 @@ clip_unit_create_from_key(struct brw_context *brw,
        * even number.
        */
       assert(key->nr_urb_entries % 2 == 0);
-      clip.thread4.max_threads = 2 - 1;
+      
+      /* Although up to 16 concurrent Clip threads are allowed on IGDNG, 
+       * only 2 threads can output VUEs at a time.
+       */
+      if (BRW_IS_IGDNG(brw))
+         clip.thread4.max_threads = 16 - 1;        
+      else
+         clip.thread4.max_threads = 2 - 1;
    } else {
       assert(key->nr_urb_entries >= 5);
       clip.thread4.max_threads = 1 - 1;
diff --git a/src/mesa/drivers/dri/i965/brw_clip_tri.c b/src/mesa/drivers/dri/i965/brw_clip_tri.c
index 7fd37bd05f..4c2d655fb1 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_tri.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_tri.c
@@ -77,6 +77,10 @@ void brw_clip_tri_alloc_regs( struct brw_clip_compile *c,
    if (c->nr_attrs & 1) {
       for (j = 0; j < 3; j++) {
 	 GLuint delta = c->nr_attrs*16 + 32;
+
+         if (BRW_IS_IGDNG(c->func.brw))
+             delta = c->nr_attrs * 16 + 32 * 3;
+
 	 brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0));
       }
    }
@@ -562,7 +566,7 @@ void brw_emit_tri_clip( struct brw_clip_compile *c )
 
    /* if -ve rhw workaround bit is set, 
       do cliptest */
-   if (!BRW_IS_G4X(p->brw)) {
+   if (BRW_IS_965(p->brw)) {
       brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
       brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), 
               brw_imm_ud(1<<20));
@@ -579,11 +583,14 @@ void brw_emit_tri_clip( struct brw_clip_compile *c )
    if (c->key.do_flat_shading) 
       brw_clip_tri_flat_shade(c); 
       
-   if (c->key.clip_mode == BRW_CLIPMODE_NORMAL)
+   if ((c->key.clip_mode == BRW_CLIPMODE_NORMAL) ||
+       (c->key.clip_mode == BRW_CLIPMODE_KERNEL_CLIP))
       do_clip_tri(c);
    else 
       maybe_do_clip_tri(c);
-      
+
+   if (c->need_ff_sync)
+	   brw_clip_ff_sync(c);      
    brw_clip_tri_emit_polygon(c);
 
    /* Send an empty message to kill the thread:
diff --git a/src/mesa/drivers/dri/i965/brw_clip_unfilled.c b/src/mesa/drivers/dri/i965/brw_clip_unfilled.c
index d7ca517927..26950383c1 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_unfilled.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_unfilled.c
@@ -496,6 +496,8 @@ void brw_emit_unfilled_clip( struct brw_clip_compile *c )
    }
    brw_ENDIF(p, do_clip);
    
+   if (c->need_ff_sync)
+	   brw_clip_ff_sync(c);      
    emit_unfilled_primitives(c);
    brw_clip_kill_thread(c);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_clip_util.c b/src/mesa/drivers/dri/i965/brw_clip_util.c
index 9d3b0be694..e09efc07ed 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_util.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_util.c
@@ -140,6 +140,10 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
 
    /* Just copy the vertex header:
     */
+   /*
+    * After CLIP stage, only first 256 bits of the VUE are read
+    * back on IGDNG, so needn't change it
+    */
    brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1);
       
    /* Iterate over each attribute (could be done in pairs?)
@@ -147,6 +151,9 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
    for (i = 0; i < c->nr_attrs; i++) {
       GLuint delta = i*16 + 32;
 
+      if (BRW_IS_IGDNG(p->brw))
+          delta = i * 16 + 32 * 3;
+
       if (delta == c->offset[VERT_RESULT_EDGE]) {
 	 if (force_edgeflag) 
 	    brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1));
@@ -177,6 +184,10 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
 
    if (i & 1) {
       GLuint delta = i*16 + 32;
+
+      if (BRW_IS_IGDNG(p->brw))
+          delta = i * 16 + 32 * 3;
+
       brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0));
    }
 
@@ -343,3 +354,19 @@ void brw_clip_init_clipmask( struct brw_clip_compile *c )
    }
 }
 
+void brw_clip_ff_sync(struct brw_clip_compile *c)
+{
+	struct brw_compile *p = &c->func;
+	brw_ff_sync(p, 
+				c->reg.R0,
+				0,
+				c->reg.R0,
+				1,	
+				1,		/* used */
+				1,  	/* msg length */
+				1,		/* response length */
+				0,		/* eot */
+				1,		/* write compelete */
+				0,		/* urb offset */
+				BRW_URB_SWIZZLE_NONE);
+}
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 577497bf6b..57ddf75413 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -129,8 +129,8 @@ struct brw_context;
 #define BRW_NEW_PRIMITIVE               0x40
 #define BRW_NEW_CONTEXT                 0x80
 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
-#define BRW_NEW_INPUT_VARYING           0x200
 #define BRW_NEW_PSP                     0x800
+#define BRW_NEW_WM_SURFACES		0x1000
 #define BRW_NEW_FENCE                   0x2000
 #define BRW_NEW_INDICES			0x4000
 #define BRW_NEW_VERTICES		0x8000
@@ -401,7 +401,6 @@ struct brw_vertex_element {
 
 
 struct brw_vertex_info {
-   GLuint varying;  /* varying:1[VERT_ATTRIB_MAX] */
    GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */
 };
 
@@ -450,8 +449,6 @@ struct brw_context
 
    struct {
       struct brw_state_flags dirty;
-      struct brw_tracked_state **atoms;
-      GLuint nr_atoms;
 
       GLuint nr_color_regions;
       struct intel_region *color_regions[MAX_DRAW_BUFFERS];
@@ -471,7 +468,8 @@ struct brw_context
       int validated_bo_count;
    } state;
 
-   struct brw_cache cache;
+   struct brw_cache cache;  /** non-surface items */
+   struct brw_cache surface_cache;  /* surface items */
    struct brw_cached_batch_item *cached_batch_items;
 
    struct {
@@ -555,11 +553,6 @@ struct brw_context
       GLuint vs_size;
       GLuint total_size;
 
-      /* Dynamic tracker which changes to reflect the state referenced
-       * by active fp and vp program parameters:
-       */
-      struct brw_tracked_state tracked_state;
-
       dri_bo *curbe_bo;
       /** Offset within curbe_bo of space for current curbe entry */
       GLuint curbe_offset;
diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c
index 9197fede2d..a1a6c53d0e 100644
--- a/src/mesa/drivers/dri/i965/brw_curbe.c
+++ b/src/mesa/drivers/dri/i965/brw_curbe.c
@@ -36,6 +36,7 @@
 #include "main/macros.h"
 #include "main/enums.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
@@ -188,13 +189,6 @@ static void prepare_constant_buffer(struct brw_context *brw)
    GLfloat *buf;
    GLuint i;
 
-   /* Update our own dependency flags.  This works because this
-    * function will also be called whenever fp or vp changes.
-    */
-   brw->curbe.tracked_state.dirty.mesa = (_NEW_TRANSFORM|_NEW_PROJECTION);
-   brw->curbe.tracked_state.dirty.mesa |= vp->program.Base.Parameters->StateFlags;
-   brw->curbe.tracked_state.dirty.mesa |= fp->program.Base.Parameters->StateFlags;
-
    if (sz == 0) {
       if (brw->curbe.last_buf) {
 	 free(brw->curbe.last_buf);
@@ -335,78 +329,11 @@ static void prepare_constant_buffer(struct brw_context *brw)
     */
 }
 
-
-/**
- * Copy Mesa program parameters into given constant buffer.
- */
-static void
-update_constant_buffer(struct brw_context *brw,
-                       const struct gl_program_parameter_list *params,
-                       dri_bo *const_buffer)
-{
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
-
-   /* copy Mesa program constants into the buffer */
-   if (const_buffer && size > 0) {
-
-      assert(const_buffer);
-      assert(const_buffer->size >= size);
-
-      if (intel->intelScreen->kernel_exec_fencing) {
-         drm_intel_gem_bo_map_gtt(const_buffer);
-         memcpy(const_buffer->virtual, params->ParameterValues, size);
-         drm_intel_gem_bo_unmap_gtt(const_buffer);
-      }
-      else {
-         dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
-      }
-
-      if (0) {
-         int i;
-         for (i = 0; i < params->NumParameters; i++) {
-            float *p = params->ParameterValues[i];
-            printf("%d: %f %f %f %f\n", i, p[0], p[1], p[2], p[3]);
-         }
-      }
-   }
-}
-
-
-/** Copy current vertex program's parameters into the constant buffer */
-static void
-update_vertex_constant_buffer(struct brw_context *brw)
-{
-   struct brw_vertex_program *vp =
-      (struct brw_vertex_program *) brw->vertex_program;
-   if (0) {
-      printf("update VS constants in buffer %p\n", vp->const_buffer);
-      printf("program %u\n", vp->program.Base.Id);
-   }
-   if (vp->use_const_buffer)
-      update_constant_buffer(brw, vp->program.Base.Parameters, vp->const_buffer);
-}
-
-
-/** Copy current fragment program's parameters into the constant buffer */
-static void
-update_fragment_constant_buffer(struct brw_context *brw)
-{
-   struct brw_fragment_program *fp =
-      (struct brw_fragment_program *) brw->fragment_program;
-   if (fp->use_const_buffer)
-      update_constant_buffer(brw, fp->program.Base.Parameters, fp->const_buffer);
-}
-
-
 static void emit_constant_buffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    GLuint sz = brw->curbe.total_size;
 
-   update_vertex_constant_buffer(brw);
-   update_fragment_constant_buffer(brw);
-
    BEGIN_BATCH(2, IGNORE_CLIPRECTS);
    if (sz == 0) {
       OUT_BATCH((CMD_CONST_BUFFER << 16) | (2 - 2));
@@ -428,7 +355,7 @@ static void emit_constant_buffer(struct brw_context *brw)
  */
 const struct brw_tracked_state brw_constant_buffer = {
    .dirty = {
-      .mesa = (_NEW_TRANSFORM|_NEW_PROJECTION),      /* plus fp and vp flags */
+      .mesa = _NEW_PROGRAM_CONSTANTS,
       .brw  = (BRW_NEW_FRAGMENT_PROGRAM |
 	       BRW_NEW_VERTEX_PROGRAM |
 	       BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */
diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 98fc909c2a..d166250b4f 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -139,6 +139,7 @@
 #define BRW_CLIPMODE_CLIP_NON_REJECTED   2
 #define BRW_CLIPMODE_REJECT_ALL          3
 #define BRW_CLIPMODE_ACCEPT_ALL          4
+#define BRW_CLIPMODE_KERNEL_CLIP         5
 
 #define BRW_CLIP_NDCSPACE     0
 #define BRW_CLIP_SCREENSPACE  1
@@ -670,6 +671,25 @@
 #define BRW_SAMPLER_MESSAGE_SIMD8_LD                  3
 #define BRW_SAMPLER_MESSAGE_SIMD16_LD                 3
 
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG            0
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_IGDNG          0
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG           0
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG       1
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_BIAS_IGDNG     1
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG      1
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_LOD_IGDNG        2
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD_IGDNG      2
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD_IGDNG       2
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG    3
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE_IGDNG  3
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG   3
+
+/* for IGDNG only */
+#define BRW_SAMPLER_SIMD_MODE_SIMD4X2                   0
+#define BRW_SAMPLER_SIMD_MODE_SIMD8                     1
+#define BRW_SAMPLER_SIMD_MODE_SIMD16                    2
+#define BRW_SAMPLER_SIMD_MODE_SIMD32_64                 3
+
 #define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW   0
 #define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH  1
 #define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS     2
@@ -819,8 +839,11 @@
 #include "intel_chipset.h"
 
 #define BRW_IS_G4X(brw)         (IS_G4X((brw)->intel.intelScreen->deviceID))
-#define CMD_PIPELINE_SELECT(brw)        (BRW_IS_G4X(brw) ? CMD_PIPELINE_SELECT_GM45 : CMD_PIPELINE_SELECT_965)
-#define CMD_VF_STATISTICS(brw)          (BRW_IS_G4X(brw) ? CMD_VF_STATISTICS_GM45 : CMD_VF_STATISTICS_965)
-#define URB_SIZES(brw)                  (BRW_IS_G4X(brw) ? 384 : 256)  /* 512 bit units */
+#define BRW_IS_IGDNG(brw)         (IS_IGDNG((brw)->intel.intelScreen->deviceID))
+#define BRW_IS_965(brw)         (!(BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)))
+#define CMD_PIPELINE_SELECT(brw)        ((BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? CMD_PIPELINE_SELECT_GM45 : CMD_PIPELINE_SELECT_965)
+#define CMD_VF_STATISTICS(brw)          ((BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? CMD_VF_STATISTICS_GM45 : CMD_VF_STATISTICS_965)
+#define URB_SIZES(brw)                  (BRW_IS_IGDNG(brw) ? 1024 : \
+                                         (BRW_IS_G4X(brw) ? 384 : 256))  /* 512 bit units */
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 5342622a73..5152c3f3a5 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -187,19 +187,13 @@ static void brw_merge_inputs( struct brw_context *brw,
       brw->vb.inputs[i].glarray = arrays[i];
 
       if (arrays[i]->StrideB != 0)
-	 brw->vb.info.varying |= 1 << i;
-
 	 brw->vb.info.sizes[i/16] |= (brw->vb.inputs[i].glarray->Size - 1) <<
 	    ((i%16) * 2);
    }
 
-   /* Raise statechanges if input sizes and varying have changed: 
-    */
+   /* Raise statechanges if input sizes have changed. */
    if (memcmp(brw->vb.info.sizes, old.sizes, sizeof(old.sizes)) != 0)
       brw->state.dirty.brw |= BRW_NEW_INPUT_DIMENSIONS;
-
-   if (brw->vb.info.varying != old.varying)
-      brw->state.dirty.brw |= BRW_NEW_INPUT_VARYING;
 }
 
 /* XXX: could split the primitive list to fallback only on the
@@ -416,6 +410,8 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
  out:
    UNLOCK_HARDWARE(intel);
 
+   brw_state_cache_check_size(brw);
+
    if (warn)
       fprintf(stderr, "i965: Single primitive emit potentially exceeded "
 	      "available aperture space\n");
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index b91b20bec6..c29f1dd5c0 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -343,7 +343,7 @@ static void brw_prepare_vertices(struct brw_context *brw)
 {
    GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = intel_context(ctx);
-   GLuint tmp = brw->vs.prog_data->inputs_read; 
+   GLbitfield vs_inputs = brw->vs.prog_data->inputs_read; 
    GLuint i;
    const unsigned char *ptr = NULL;
    GLuint interleave = 0;
@@ -362,11 +362,11 @@ static void brw_prepare_vertices(struct brw_context *brw)
       _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
 
    /* Accumulate the list of enabled arrays. */
-   while (tmp) {
-      GLuint i = _mesa_ffsll(tmp)-1;
+   while (vs_inputs) {
+      GLuint i = _mesa_ffsll(vs_inputs) - 1;
       struct brw_vertex_element *input = &brw->vb.inputs[i];
 
-      tmp &= ~(1<<i);
+      vs_inputs &= ~(1 << i);
       enabled[nr_enabled++] = input;
    }
 
@@ -477,17 +477,17 @@ static void brw_emit_vertices(struct brw_context *brw)
 {
    GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = intel_context(ctx);
-   GLuint tmp = brw->vs.prog_data->inputs_read;
+   GLbitfield vs_inputs = brw->vs.prog_data->inputs_read;
    struct brw_vertex_element *enabled[VERT_ATTRIB_MAX];
    GLuint i;
    GLuint nr_enabled = 0;
 
   /* Accumulate the list of enabled arrays. */
-   while (tmp) {
-      i = _mesa_ffsll(tmp)-1;
+   while (vs_inputs) {
+      i = _mesa_ffsll(vs_inputs) - 1;
       struct brw_vertex_element *input = &brw->vb.inputs[i];
 
-      tmp &= ~(1<<i);
+      vs_inputs &= ~(1 << i);
       enabled[nr_enabled++] = input;
    }
 
@@ -512,7 +512,19 @@ static void brw_emit_vertices(struct brw_context *brw)
       OUT_RELOC(input->bo,
 		I915_GEM_DOMAIN_VERTEX, 0,
 		input->offset);
-      OUT_BATCH(brw->vb.max_index);
+      if (BRW_IS_IGDNG(brw)) {
+          if (input->stride) {
+              OUT_RELOC(input->bo,
+                        I915_GEM_DOMAIN_VERTEX, 0,
+                        input->offset + input->stride * input->count);
+          } else {
+              assert(input->count == 1);
+              OUT_RELOC(input->bo,
+                        I915_GEM_DOMAIN_VERTEX, 0,
+                        input->offset + input->element_size);
+          }
+      } else
+          OUT_BATCH(brw->vb.max_index);
       OUT_BATCH(0); /* Instance data step rate */
    }
    ADVANCE_BATCH();
@@ -542,11 +554,18 @@ static void brw_emit_vertices(struct brw_context *brw)
 		BRW_VE0_VALID |
 		(format << BRW_VE0_FORMAT_SHIFT) |
 		(0 << BRW_VE0_SRC_OFFSET_SHIFT));
-      OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
-		(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
-		(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
-		(comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
-		((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
+
+      if (BRW_IS_IGDNG(brw))
+          OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
+                    (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
+                    (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
+                    (comp3 << BRW_VE1_COMPONENT_3_SHIFT));
+      else
+          OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
+                    (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
+                    (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
+                    (comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
+                    ((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
    }
    ADVANCE_BATCH();
 }
@@ -635,7 +654,7 @@ static void brw_emit_indices(struct brw_context *brw)
    if (index_buffer == NULL)
       return;
 
-   ib_size = get_size(index_buffer->type) * index_buffer->count;
+   ib_size = get_size(index_buffer->type) * index_buffer->count - 1;
 
    /* Emit the indexbuffer packet:
     */
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 003332f78c..30603bdd0e 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -97,7 +97,7 @@ struct brw_glsl_call;
 
 
 #define BRW_EU_MAX_INSN_STACK 5
-#define BRW_EU_MAX_INSN 4000
+#define BRW_EU_MAX_INSN 10000
 
 struct brw_compile {
    struct brw_instruction store[BRW_EU_MAX_INSN];
@@ -171,9 +171,9 @@ static INLINE struct brw_reg brw_reg( GLuint file,
 {
    struct brw_reg reg;
    if (type == BRW_GENERAL_REGISTER_FILE)
-      assert(nr < 128);
+      assert(nr < BRW_MAX_GRF);
    else if (type == BRW_MESSAGE_REGISTER_FILE)
-      assert(nr < 9);
+      assert(nr < BRW_MAX_MRF);
    else if (type == BRW_ARCHITECTURE_REGISTER_FILE)
       assert(nr <= BRW_ARF_IP);
 
@@ -538,6 +538,7 @@ static INLINE struct brw_reg brw_mask_reg( GLuint subnr )
 
 static INLINE struct brw_reg brw_message_reg( GLuint nr )
 {
+   assert(nr < BRW_MAX_MRF);
    return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE,
 		       nr,
 		       0);
@@ -815,6 +816,19 @@ void brw_urb_WRITE(struct brw_compile *p,
 		   GLuint offset,
 		   GLuint swizzle);
 
+void brw_ff_sync(struct brw_compile *p,
+		   struct brw_reg dest,
+		   GLuint msg_reg_nr,
+		   struct brw_reg src0,
+		   GLboolean allocate,
+		   GLboolean used,
+		   GLuint msg_length,
+		   GLuint response_length,
+		   GLboolean eot,
+		   GLboolean writes_complete,
+		   GLuint offset,
+		   GLuint swizzle);
+
 void brw_fb_WRITE(struct brw_compile *p,
 		   struct brw_reg dest,
 		   GLuint msg_reg_nr,
@@ -834,7 +848,9 @@ void brw_SAMPLE(struct brw_compile *p,
 		GLuint msg_type,
 		GLuint response_length,
 		GLuint msg_length,
-		GLboolean eot);
+		GLboolean eot,
+		GLuint header_present,
+		GLuint simd_mode);
 
 void brw_math_16( struct brw_compile *p,
 		  struct brw_reg dest,
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 2a147fb8c3..2412014248 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -241,7 +241,8 @@ void brw_set_src1( struct brw_instruction *insn,
 
 
 
-static void brw_set_math_message( struct brw_instruction *insn,
+static void brw_set_math_message( struct brw_context *brw,
+				  struct brw_instruction *insn,
 				  GLuint msg_length,
 				  GLuint response_length,
 				  GLuint function,
@@ -252,18 +253,35 @@ static void brw_set_math_message( struct brw_instruction *insn,
 {
    brw_set_src1(insn, brw_imm_d(0));
 
-   insn->bits3.math.function = function;
-   insn->bits3.math.int_type = integer_type;
-   insn->bits3.math.precision = low_precision;
-   insn->bits3.math.saturate = saturate;
-   insn->bits3.math.data_type = dataType;
-   insn->bits3.math.response_length = response_length;
-   insn->bits3.math.msg_length = msg_length;
-   insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH;
-   insn->bits3.math.end_of_thread = 0;
+   if (BRW_IS_IGDNG(brw)) {
+       insn->bits3.math_igdng.function = function;
+       insn->bits3.math_igdng.int_type = integer_type;
+       insn->bits3.math_igdng.precision = low_precision;
+       insn->bits3.math_igdng.saturate = saturate;
+       insn->bits3.math_igdng.data_type = dataType;
+       insn->bits3.math_igdng.snapshot = 0;
+       insn->bits3.math_igdng.header_present = 0;
+       insn->bits3.math_igdng.response_length = response_length;
+       insn->bits3.math_igdng.msg_length = msg_length;
+       insn->bits3.math_igdng.end_of_thread = 0;
+       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_MATH;
+       insn->bits2.send_igdng.end_of_thread = 0;
+   } else {
+       insn->bits3.math.function = function;
+       insn->bits3.math.int_type = integer_type;
+       insn->bits3.math.precision = low_precision;
+       insn->bits3.math.saturate = saturate;
+       insn->bits3.math.data_type = dataType;
+       insn->bits3.math.response_length = response_length;
+       insn->bits3.math.msg_length = msg_length;
+       insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH;
+       insn->bits3.math.end_of_thread = 0;
+   }
 }
 
-static void brw_set_urb_message( struct brw_instruction *insn,
+
+static void brw_set_ff_sync_message( struct brw_context *brw,
+				 struct brw_instruction *insn,
 				 GLboolean allocate,
 				 GLboolean used,
 				 GLuint msg_length,
@@ -273,21 +291,64 @@ static void brw_set_urb_message( struct brw_instruction *insn,
 				 GLuint offset,
 				 GLuint swizzle_control )
 {
-   brw_set_src1(insn, brw_imm_d(0));
+	brw_set_src1(insn, brw_imm_d(0));
+
+	insn->bits3.urb_igdng.opcode = 1;
+	insn->bits3.urb_igdng.offset = offset;
+	insn->bits3.urb_igdng.swizzle_control = swizzle_control;
+	insn->bits3.urb_igdng.allocate = allocate;
+	insn->bits3.urb_igdng.used = used;
+	insn->bits3.urb_igdng.complete = complete;
+	insn->bits3.urb_igdng.header_present = 1;
+	insn->bits3.urb_igdng.response_length = response_length;
+	insn->bits3.urb_igdng.msg_length = msg_length;
+	insn->bits3.urb_igdng.end_of_thread = end_of_thread;
+	insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_URB;
+	insn->bits2.send_igdng.end_of_thread = end_of_thread;
+}
 
-   insn->bits3.urb.opcode = 0;	/* ? */
-   insn->bits3.urb.offset = offset;
-   insn->bits3.urb.swizzle_control = swizzle_control;
-   insn->bits3.urb.allocate = allocate;
-   insn->bits3.urb.used = used;	/* ? */
-   insn->bits3.urb.complete = complete;
-   insn->bits3.urb.response_length = response_length;
-   insn->bits3.urb.msg_length = msg_length;
-   insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB;
-   insn->bits3.urb.end_of_thread = end_of_thread;
+static void brw_set_urb_message( struct brw_context *brw,
+				 struct brw_instruction *insn,
+				 GLboolean allocate,
+				 GLboolean used,
+				 GLuint msg_length,
+				 GLuint response_length,
+				 GLboolean end_of_thread,
+				 GLboolean complete,
+				 GLuint offset,
+				 GLuint swizzle_control )
+{
+    brw_set_src1(insn, brw_imm_d(0));
+
+    if (BRW_IS_IGDNG(brw)) {
+        insn->bits3.urb_igdng.opcode = 0;	/* ? */
+        insn->bits3.urb_igdng.offset = offset;
+        insn->bits3.urb_igdng.swizzle_control = swizzle_control;
+        insn->bits3.urb_igdng.allocate = allocate;
+        insn->bits3.urb_igdng.used = used;	/* ? */
+        insn->bits3.urb_igdng.complete = complete;
+        insn->bits3.urb_igdng.header_present = 1;
+        insn->bits3.urb_igdng.response_length = response_length;
+        insn->bits3.urb_igdng.msg_length = msg_length;
+        insn->bits3.urb_igdng.end_of_thread = end_of_thread;
+        insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_URB;
+        insn->bits2.send_igdng.end_of_thread = end_of_thread;
+    } else {
+        insn->bits3.urb.opcode = 0;	/* ? */
+        insn->bits3.urb.offset = offset;
+        insn->bits3.urb.swizzle_control = swizzle_control;
+        insn->bits3.urb.allocate = allocate;
+        insn->bits3.urb.used = used;	/* ? */
+        insn->bits3.urb.complete = complete;
+        insn->bits3.urb.response_length = response_length;
+        insn->bits3.urb.msg_length = msg_length;
+        insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB;
+        insn->bits3.urb.end_of_thread = end_of_thread;
+    }
 }
 
-static void brw_set_dp_write_message( struct brw_instruction *insn,
+static void brw_set_dp_write_message( struct brw_context *brw,
+				      struct brw_instruction *insn,
 				      GLuint binding_table_index,
 				      GLuint msg_control,
 				      GLuint msg_type,
@@ -298,18 +359,33 @@ static void brw_set_dp_write_message( struct brw_instruction *insn,
 {
    brw_set_src1(insn, brw_imm_d(0));
 
-   insn->bits3.dp_write.binding_table_index = binding_table_index;
-   insn->bits3.dp_write.msg_control = msg_control;
-   insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear;
-   insn->bits3.dp_write.msg_type = msg_type;
-   insn->bits3.dp_write.send_commit_msg = 0;
-   insn->bits3.dp_write.response_length = response_length;
-   insn->bits3.dp_write.msg_length = msg_length;
-   insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
-   insn->bits3.urb.end_of_thread = end_of_thread;
+   if (BRW_IS_IGDNG(brw)) {
+       insn->bits3.dp_write_igdng.binding_table_index = binding_table_index;
+       insn->bits3.dp_write_igdng.msg_control = msg_control;
+       insn->bits3.dp_write_igdng.pixel_scoreboard_clear = pixel_scoreboard_clear;
+       insn->bits3.dp_write_igdng.msg_type = msg_type;
+       insn->bits3.dp_write_igdng.send_commit_msg = 0;
+       insn->bits3.dp_write_igdng.header_present = 1;
+       insn->bits3.dp_write_igdng.response_length = response_length;
+       insn->bits3.dp_write_igdng.msg_length = msg_length;
+       insn->bits3.dp_write_igdng.end_of_thread = end_of_thread;
+       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
+       insn->bits2.send_igdng.end_of_thread = end_of_thread;
+   } else {
+       insn->bits3.dp_write.binding_table_index = binding_table_index;
+       insn->bits3.dp_write.msg_control = msg_control;
+       insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear;
+       insn->bits3.dp_write.msg_type = msg_type;
+       insn->bits3.dp_write.send_commit_msg = 0;
+       insn->bits3.dp_write.response_length = response_length;
+       insn->bits3.dp_write.msg_length = msg_length;
+       insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
+       insn->bits3.dp_write.end_of_thread = end_of_thread;
+   }
 }
 
-static void brw_set_dp_read_message( struct brw_instruction *insn,
+static void brw_set_dp_read_message( struct brw_context *brw,
+				      struct brw_instruction *insn,
 				      GLuint binding_table_index,
 				      GLuint msg_control,
 				      GLuint msg_type,
@@ -320,15 +396,29 @@ static void brw_set_dp_read_message( struct brw_instruction *insn,
 {
    brw_set_src1(insn, brw_imm_d(0));
 
-   insn->bits3.dp_read.binding_table_index = binding_table_index; /*0:7*/
-   insn->bits3.dp_read.msg_control = msg_control;  /*8:11*/
-   insn->bits3.dp_read.msg_type = msg_type;  /*12:13*/
-   insn->bits3.dp_read.target_cache = target_cache;  /*14:15*/
-   insn->bits3.dp_read.response_length = response_length;  /*16:19*/
-   insn->bits3.dp_read.msg_length = msg_length;  /*20:23*/
-   insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; /*24:27*/
-   insn->bits3.dp_read.pad1 = 0;  /*28:30*/
-   insn->bits3.dp_read.end_of_thread = end_of_thread;  /*31*/
+   if (BRW_IS_IGDNG(brw)) {
+       insn->bits3.dp_read_igdng.binding_table_index = binding_table_index;
+       insn->bits3.dp_read_igdng.msg_control = msg_control;
+       insn->bits3.dp_read_igdng.msg_type = msg_type;
+       insn->bits3.dp_read_igdng.target_cache = target_cache;
+       insn->bits3.dp_read_igdng.header_present = 1;
+       insn->bits3.dp_read_igdng.response_length = response_length;
+       insn->bits3.dp_read_igdng.msg_length = msg_length;
+       insn->bits3.dp_read_igdng.pad1 = 0;
+       insn->bits3.dp_read_igdng.end_of_thread = end_of_thread;
+       insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_DATAPORT_READ;
+       insn->bits2.send_igdng.end_of_thread = end_of_thread;
+   } else {
+       insn->bits3.dp_read.binding_table_index = binding_table_index; /*0:7*/
+       insn->bits3.dp_read.msg_control = msg_control;  /*8:11*/
+       insn->bits3.dp_read.msg_type = msg_type;  /*12:13*/
+       insn->bits3.dp_read.target_cache = target_cache;  /*14:15*/
+       insn->bits3.dp_read.response_length = response_length;  /*16:19*/
+       insn->bits3.dp_read.msg_length = msg_length;  /*20:23*/
+       insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; /*24:27*/
+       insn->bits3.dp_read.pad1 = 0;  /*28:30*/
+       insn->bits3.dp_read.end_of_thread = end_of_thread;  /*31*/
+   }
 }
 
 static void brw_set_sampler_message(struct brw_context *brw,
@@ -338,11 +428,25 @@ static void brw_set_sampler_message(struct brw_context *brw,
                                     GLuint msg_type,
                                     GLuint response_length,
                                     GLuint msg_length,
-                                    GLboolean eot)
+                                    GLboolean eot,
+                                    GLuint header_present,
+                                    GLuint simd_mode)
 {
+   assert(eot == 0);
    brw_set_src1(insn, brw_imm_d(0));
 
-   if (BRW_IS_G4X(brw)) {
+   if (BRW_IS_IGDNG(brw)) {
+      insn->bits3.sampler_igdng.binding_table_index = binding_table_index;
+      insn->bits3.sampler_igdng.sampler = sampler;
+      insn->bits3.sampler_igdng.msg_type = msg_type;
+      insn->bits3.sampler_igdng.simd_mode = simd_mode;
+      insn->bits3.sampler_igdng.header_present = header_present;
+      insn->bits3.sampler_igdng.response_length = response_length;
+      insn->bits3.sampler_igdng.msg_length = msg_length;
+      insn->bits3.sampler_igdng.end_of_thread = eot;
+      insn->bits2.send_igdng.sfid = BRW_MESSAGE_TARGET_SAMPLER;
+      insn->bits2.send_igdng.end_of_thread = eot;
+   } else if (BRW_IS_G4X(brw)) {
       insn->bits3.sampler_g4x.binding_table_index = binding_table_index;
       insn->bits3.sampler_g4x.sampler = sampler;
       insn->bits3.sampler_g4x.msg_type = msg_type;
@@ -484,6 +588,10 @@ struct brw_instruction *brw_JMPI(struct brw_compile *p,
 {
    struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1);
 
+   insn->header.execution_size = 1;
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.mask_control = BRW_MASK_DISABLE;
+
    p->current->header.predicate_control = BRW_PREDICATE_NONE;
 
    return insn;
@@ -540,6 +648,10 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
 				 struct brw_instruction *if_insn)
 {
    struct brw_instruction *insn;
+   GLuint br = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+      br = 2;
 
    if (p->single_program_flow) {
       insn = next_insn(p, BRW_OPCODE_ADD);
@@ -566,7 +678,7 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
    } else {
       assert(if_insn->header.opcode == BRW_OPCODE_IF);
 
-      if_insn->bits3.if_else.jump_count = insn - if_insn;
+      if_insn->bits3.if_else.jump_count = br * (insn - if_insn);
       if_insn->bits3.if_else.pop_count = 1;
       if_insn->bits3.if_else.pad0 = 0;
    }
@@ -577,6 +689,11 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
 void brw_ENDIF(struct brw_compile *p, 
 	       struct brw_instruction *patch_insn)
 {
+   GLuint br = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+      br = 2; 
+ 
    if (p->single_program_flow) {
       /* In single program flow mode, there's no need to execute an ENDIF,
        * since we don't need to do any stack operations, and if we're executing
@@ -608,11 +725,11 @@ void brw_ENDIF(struct brw_compile *p,
 	 /* Automagically turn it into an IFF:
 	  */
 	 patch_insn->header.opcode = BRW_OPCODE_IFF;
-	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1);
 	 patch_insn->bits3.if_else.pop_count = 0;
 	 patch_insn->bits3.if_else.pad0 = 0;
       } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) {
-	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.jump_count = br * (insn - patch_insn + 1);
 	 patch_insn->bits3.if_else.pop_count = 1;
 	 patch_insn->bits3.if_else.pad0 = 0;
       } else {
@@ -686,6 +803,10 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p,
                                   struct brw_instruction *do_insn)
 {
    struct brw_instruction *insn;
+   GLuint br = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+      br = 2;
 
    if (p->single_program_flow)
       insn = next_insn(p, BRW_OPCODE_ADD);
@@ -706,7 +827,7 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p,
       insn->header.execution_size = do_insn->header.execution_size;
 
       assert(do_insn->header.opcode == BRW_OPCODE_DO);
-      insn->bits3.if_else.jump_count = do_insn - insn + 1;
+      insn->bits3.if_else.jump_count = br * (do_insn - insn + 1);
       insn->bits3.if_else.pop_count = 0;
       insn->bits3.if_else.pad0 = 0;
    }
@@ -725,11 +846,15 @@ void brw_land_fwd_jump(struct brw_compile *p,
 		       struct brw_instruction *jmp_insn)
 {
    struct brw_instruction *landing = &p->store[p->nr_insn];
+   GLuint jmpi = 1;
+
+   if (BRW_IS_IGDNG(p->brw))
+       jmpi = 2;
 
    assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI);
    assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE);
 
-   jmp_insn->bits3.ud = (landing - jmp_insn) - 1; 
+   jmp_insn->bits3.ud = jmpi * ((landing - jmp_insn) - 1);
 }
 
 
@@ -794,7 +919,8 @@ void brw_math( struct brw_compile *p,
 
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src);
-   brw_set_math_message(insn, 
+   brw_set_math_message(p->brw,
+			insn, 
 			msg_length, response_length, 
 			function,
 			BRW_MATH_INTEGER_UNSIGNED,
@@ -830,7 +956,8 @@ void brw_math_16( struct brw_compile *p,
 
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src);
-   brw_set_math_message(insn, 
+   brw_set_math_message(p->brw,
+			insn, 
 			msg_length, response_length, 
 			function,
 			BRW_MATH_INTEGER_UNSIGNED,
@@ -846,7 +973,8 @@ void brw_math_16( struct brw_compile *p,
 
    brw_set_dest(insn, offset(dest,1));
    brw_set_src0(insn, src);
-   brw_set_math_message(insn, 
+   brw_set_math_message(p->brw, 
+			insn, 
 			msg_length, response_length, 
 			function,
 			BRW_MATH_INTEGER_UNSIGNED,
@@ -893,7 +1021,8 @@ void brw_dp_WRITE_16( struct brw_compile *p,
       brw_set_dest(insn, dest);
       brw_set_src0(insn, src);
 
-      brw_set_dp_write_message(insn,
+      brw_set_dp_write_message(p->brw,
+			       insn,
 			       255, /* binding table index (255=stateless) */
 			       BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */
 			       BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */
@@ -938,7 +1067,8 @@ void brw_dp_READ_16( struct brw_compile *p,
       brw_set_dest(insn, dest);	/* UW? */
       brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW));
 
-      brw_set_dp_read_message(insn,
+      brw_set_dp_read_message(p->brw,
+			      insn,
 			      255, /* binding table index (255=stateless) */
 			      3,  /* msg_control (3 means 4 Owords) */
 			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
@@ -995,7 +1125,8 @@ void brw_dp_READ_4( struct brw_compile *p,
       brw_set_dest(insn, dest);
       brw_set_src0(insn, brw_null_reg());
 
-      brw_set_dp_read_message(insn,
+      brw_set_dp_read_message(p->brw,
+			      insn,
 			      bind_table_index,
 			      0,  /* msg_control (0 means 1 Oword) */
 			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
@@ -1066,7 +1197,8 @@ void brw_dp_READ_4_vs(struct brw_compile *p,
       brw_set_dest(insn, dest);
       brw_set_src0(insn, brw_null_reg());
 
-      brw_set_dp_read_message(insn,
+      brw_set_dp_read_message(p->brw,
+			      insn,
 			      bind_table_index,
 			      oword,  /* 0 = lower Oword, 1 = upper Oword */
 			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
@@ -1096,7 +1228,8 @@ void brw_fb_WRITE(struct brw_compile *p,
   
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src0);
-   brw_set_dp_write_message(insn,
+   brw_set_dp_write_message(p->brw,
+			    insn,
 			    binding_table_index,
 			    BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */
 			    BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */
@@ -1122,7 +1255,9 @@ void brw_SAMPLE(struct brw_compile *p,
 		GLuint msg_type,
 		GLuint response_length,
 		GLuint msg_length,
-		GLboolean eot)
+		GLboolean eot,
+		GLuint header_present,
+		GLuint simd_mode)
 {
    GLboolean need_stall = 0;
    
@@ -1197,7 +1332,9 @@ void brw_SAMPLE(struct brw_compile *p,
 			      msg_type,
 			      response_length, 
 			      msg_length,
-			      eot);
+			      eot,
+			      header_present,
+			      simd_mode);
    }
 
    if (need_stall) {
@@ -1232,7 +1369,7 @@ void brw_urb_WRITE(struct brw_compile *p,
 {
    struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
 
-   assert(msg_length < 16);
+   assert(msg_length < BRW_MAX_MRF);
 
    brw_set_dest(insn, dest);
    brw_set_src0(insn, src0);
@@ -1240,7 +1377,8 @@ void brw_urb_WRITE(struct brw_compile *p,
 
    insn->header.destreg__conditonalmod = msg_reg_nr;
 
-   brw_set_urb_message(insn,
+   brw_set_urb_message(p->brw,
+		       insn,
 		       allocate,
 		       used,
 		       msg_length,
@@ -1251,3 +1389,37 @@ void brw_urb_WRITE(struct brw_compile *p,
 		       swizzle);
 }
 
+void brw_ff_sync(struct brw_compile *p,
+		   struct brw_reg dest,
+		   GLuint msg_reg_nr,
+		   struct brw_reg src0,
+		   GLboolean allocate,
+		   GLboolean used,
+		   GLuint msg_length,
+		   GLuint response_length,
+		   GLboolean eot,
+		   GLboolean writes_complete,
+		   GLuint offset,
+		   GLuint swizzle)
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+   assert(msg_length < 16);
+
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src0);
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->header.destreg__conditonalmod = msg_reg_nr;
+
+   brw_set_ff_sync_message(p->brw,
+		       insn,
+		       allocate,
+		       used,
+		       msg_length,
+		       response_length, 
+		       eot, 
+		       writes_complete, 
+		       offset,
+		       swizzle);
+}
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 299357409c..d27c6c24ca 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -37,6 +37,9 @@
 #include "tnl/tnl.h"
 #include "brw_context.h"
 #include "brw_fallback.h"
+#include "intel_chipset.h"
+#include "intel_fbo.h"
+#include "intel_regions.h"
 
 #include "glapi/glapi.h"
 
@@ -44,6 +47,7 @@
 
 static GLboolean do_check_fallback(struct brw_context *brw)
 {
+   struct intel_context *intel = &brw->intel;
    GLcontext *ctx = &brw->intel.ctx;
    GLuint i;
 
@@ -81,6 +85,33 @@ static GLboolean do_check_fallback(struct brw_context *brw)
       return GL_TRUE;
    }
 
+   /* _NEW_BUFFERS */
+   if (IS_965(intel->intelScreen->deviceID) &&
+       !IS_G4X(intel->intelScreen->deviceID)) {
+      for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
+	 struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
+	 struct intel_renderbuffer *irb = intel_renderbuffer(rb);
+
+	 /* The original gen4 hardware couldn't set up WM surfaces pointing
+	  * at an offset within a tile, which can happen when rendering to
+	  * anything but the base level of a texture or the +X face/0 depth.
+	  * This was fixed with the 4 Series hardware.
+	  *
+	  * For these original chips, you would have to make the depth and
+	  * color destination surfaces include information on the texture
+	  * type, LOD, face, and various limits to use them as a destination.
+	  * I would have done this, but there's also a nasty requirement that
+	  * the depth and the color surfaces all be of the same LOD, which
+	  * may be a worse requirement than this alignment.  (Also, we may
+	  * want to just demote the texture to untiled, instead).
+	  */
+	 if (irb->region && irb->region->tiling != I915_TILING_NONE &&
+	     (irb->region->draw_offset & 4095)) {
+	    DBG("FALLBACK: non-tile-aligned destination for tiled FBO\n");
+	    return GL_TRUE;
+	 }
+      }
+   }
 
    return GL_FALSE;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c
index a8b74a0afe..48c2b9a41c 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.c
+++ b/src/mesa/drivers/dri/i965/brw_gs.c
@@ -54,12 +54,17 @@ static void compile_gs_prog( struct brw_context *brw,
    memset(&c, 0, sizeof(c));
    
    c.key = *key;
-
+   c.need_ff_sync = BRW_IS_IGDNG(brw);
    /* Need to locate the two positions present in vertex + header.
     * These are currently hardcoded:
     */
    c.nr_attrs = brw_count_bits(c.key.attrs);
-   c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+
+   if (BRW_IS_IGDNG(brw))
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 3;  /* are vertices packed, or reg-aligned? */
+   else
+       c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+
    c.nr_bytes = c.nr_regs * REG_SIZE;
 
    
diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h
index 18a4537c32..bbb991ea2e 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.h
+++ b/src/mesa/drivers/dri/i965/brw_gs.h
@@ -62,6 +62,7 @@ struct brw_gs_compile {
    GLuint nr_attrs;
    GLuint nr_regs;
    GLuint nr_bytes;
+   GLboolean need_ff_sync;
 };
 
 #define ATTR_SIZE  (4*4)
diff --git a/src/mesa/drivers/dri/i965/brw_gs_emit.c b/src/mesa/drivers/dri/i965/brw_gs_emit.c
index 22e0d25c2e..980eac7646 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_emit.c
@@ -101,6 +101,23 @@ static void brw_gs_emit_vue(struct brw_gs_compile *c,
 		 BRW_URB_SWIZZLE_NONE);
 }
 
+void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim)
+{
+	struct brw_compile *p = &c->func;
+	brw_MOV(p, get_element_ud(c->reg.R0, 1), brw_imm_ud(num_prim));
+	brw_ff_sync(p, 
+				c->reg.R0,
+				0,
+				c->reg.R0,
+				1,	
+				1,		/* used */
+				1,  	/* msg length */
+				1,		/* response length */
+				0,		/* eot */
+				1,		/* write compelete */
+				0,		/* urb offset */
+				BRW_URB_SWIZZLE_NONE);
+}
 
 
 void brw_gs_quads( struct brw_gs_compile *c )
@@ -110,6 +127,8 @@ void brw_gs_quads( struct brw_gs_compile *c )
    /* Use polygons for correct edgeflag behaviour. Note that vertex 3
     * is the PV for quads, but vertex 0 for polygons:
     */
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);    
    brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2));
    brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); 
@@ -120,6 +139,8 @@ void brw_gs_quad_strip( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 4);
    
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2));
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); 
@@ -129,6 +150,9 @@ void brw_gs_quad_strip( struct brw_gs_compile *c )
 void brw_gs_tris( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 3);
+
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2));
    brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END));
@@ -137,6 +161,9 @@ void brw_gs_tris( struct brw_gs_compile *c )
 void brw_gs_lines( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 2);
+
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START));
    brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END));
 }
@@ -144,6 +171,9 @@ void brw_gs_lines( struct brw_gs_compile *c )
 void brw_gs_points( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 1);
+
+   if (c->need_ff_sync)
+	   brw_gs_ff_sync(c, 1);      
    brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END));
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c
index 27023cf034..a761c03153 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_state.c
@@ -95,6 +95,9 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
 
    gs.thread4.max_threads = 0; /* Hardware requirement */
 
+   if (BRW_IS_IGDNG(brw))
+      gs.thread4.rendering_enable = 1;
+
    if (INTEL_DEBUG & DEBUG_STATS)
       gs.thread4.stats_enable = 1;
 
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 9bc5c35139..85a7706404 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -118,7 +118,10 @@ static void upload_binding_table_pointers(struct brw_context *brw)
 
    BEGIN_BATCH(6, IGNORE_CLIPRECTS);
    OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2));
-   OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
+   if (brw->vs.bind_bo != NULL)
+      OUT_RELOC(brw->vs.bind_bo, I915_GEM_DOMAIN_SAMPLER, 0, 0); /* vs */
+   else
+      OUT_BATCH(0);
    OUT_BATCH(0); /* gs */
    OUT_BATCH(0); /* clip */
    OUT_BATCH(0); /* sf */
@@ -208,7 +211,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 {
    struct intel_context *intel = &brw->intel;
    struct intel_region *region = brw->state.depth_region;
-   unsigned int len = BRW_IS_G4X(brw) ? 6 : 5;
+   unsigned int len = (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? 6 : 5;
 
    if (region == NULL) {
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
@@ -219,7 +222,7 @@ static void emit_depthbuffer(struct brw_context *brw)
       OUT_BATCH(0);
       OUT_BATCH(0);
 
-      if (BRW_IS_G4X(brw))
+      if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))
          OUT_BATCH(0);
 
       ADVANCE_BATCH();
@@ -241,6 +244,8 @@ static void emit_depthbuffer(struct brw_context *brw)
 	 return;
       }
 
+      assert(region->tiling != I915_TILING_X);
+
       BEGIN_BATCH(len, IGNORE_CLIPRECTS);
       OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2));
       OUT_BATCH(((region->pitch * region->cpp) - 1) |
@@ -256,7 +261,7 @@ static void emit_depthbuffer(struct brw_context *brw)
 		((region->height - 1) << 19));
       OUT_BATCH(0);
 
-      if (BRW_IS_G4X(brw))
+      if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))
          OUT_BATCH(0);
 
       ADVANCE_BATCH();
@@ -369,7 +374,7 @@ static void upload_aa_line_parameters(struct brw_context *brw)
 {
    struct brw_aa_line_parameters balp;
    
-   if (!BRW_IS_G4X(brw))
+   if (BRW_IS_965(brw))
       return;
 
    /* use legacy aa line coverage computation */
@@ -506,14 +511,27 @@ static void upload_state_base_address( struct brw_context *brw )
    /* Output the structure (brw_state_base_address) directly to the
     * batchbuffer, so we can emit relocations inline.
     */
-   BEGIN_BATCH(6, IGNORE_CLIPRECTS);
-   OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
-   OUT_BATCH(1); /* General state base address */
-   OUT_BATCH(1); /* Surface state base address */
-   OUT_BATCH(1); /* Indirect object base address */
-   OUT_BATCH(1); /* General state upper bound */
-   OUT_BATCH(1); /* Indirect object upper bound */
-   ADVANCE_BATCH();
+   if (BRW_IS_IGDNG(brw)) {
+       BEGIN_BATCH(8, IGNORE_CLIPRECTS);
+       OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (8 - 2));
+       OUT_BATCH(1); /* General state base address */
+       OUT_BATCH(1); /* Surface state base address */
+       OUT_BATCH(1); /* Indirect object base address */
+       OUT_BATCH(1); /* Instruction base address */
+       OUT_BATCH(1); /* General state upper bound */
+       OUT_BATCH(1); /* Indirect object upper bound */
+       OUT_BATCH(1); /* Instruction access upper bound */
+       ADVANCE_BATCH();
+   } else {
+       BEGIN_BATCH(6, IGNORE_CLIPRECTS);
+       OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
+       OUT_BATCH(1); /* General state base address */
+       OUT_BATCH(1); /* Surface state base address */
+       OUT_BATCH(1); /* Indirect object base address */
+       OUT_BATCH(1); /* General state upper bound */
+       OUT_BATCH(1); /* Indirect object upper bound */
+       ADVANCE_BATCH();
+   }
 }
 
 const struct brw_tracked_state brw_state_base_address = {
diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c
index c3c85978f4..e1c2c7777b 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.c
+++ b/src/mesa/drivers/dri/i965/brw_sf.c
@@ -166,6 +166,9 @@ static void upload_sf_prog(struct brw_context *brw)
    key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
    key.do_twoside_color = (ctx->Light.Enabled && ctx->Light.Model.TwoSide);
 
+   /* _NEW_HINT */
+   key.linear_color = (ctx->Hint.PerspectiveCorrection == GL_FASTEST);
+
    /* _NEW_POLYGON */
    if (key.do_twoside_color) {
       /* If we're rendering to a FBO, we have to invert the polygon
@@ -188,7 +191,7 @@ static void upload_sf_prog(struct brw_context *brw)
 
 const struct brw_tracked_state brw_sf_prog = {
    .dirty = {
-      .mesa  = (_NEW_LIGHT|_NEW_POLYGON|_NEW_POINT),
+      .mesa  = (_NEW_HINT | _NEW_LIGHT | _NEW_POLYGON | _NEW_POINT),
       .brw   = (BRW_NEW_REDUCED_PRIMITIVE),
       .cache = CACHE_NEW_VS_PROG
    },
diff --git a/src/mesa/drivers/dri/i965/brw_sf.h b/src/mesa/drivers/dri/i965/brw_sf.h
index 1c0fb70fe0..6426b6df9f 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.h
+++ b/src/mesa/drivers/dri/i965/brw_sf.h
@@ -51,7 +51,8 @@ struct brw_sf_prog_key {
    GLuint do_flat_shading:1;
    GLuint frontface_ccw:1;
    GLuint do_point_sprite:1;
-   GLuint pad:10;
+   GLuint linear_color:1;  /**< linear interp vs. perspective interp */
+   GLuint pad:25;
    GLenum SpriteOrigin;
 };
 
diff --git a/src/mesa/drivers/dri/i965/brw_sf_emit.c b/src/mesa/drivers/dri/i965/brw_sf_emit.c
index 862835f157..ca8f97f9f9 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_emit.c
@@ -151,6 +151,8 @@ static void do_flatshade_triangle( struct brw_sf_compile *c )
    struct brw_compile *p = &c->func;
    struct brw_reg ip = brw_ip_reg();
    GLuint nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS);
+   GLuint jmpi = 1;
+
    if (!nr)
       return;
 
@@ -159,18 +161,21 @@ static void do_flatshade_triangle( struct brw_sf_compile *c )
    if (c->key.primitive == SF_UNFILLED_TRIS)
       return;
 
+   if (BRW_IS_IGDNG(p->brw))
+       jmpi = 2;
+
    brw_push_insn_state(p);
    
-   brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr*2+1));
+   brw_MUL(p, c->pv, c->pv, brw_imm_d(jmpi*(nr*2+1)));
    brw_JMPI(p, ip, ip, c->pv);
 
    copy_colors(c, c->vert[1], c->vert[0]);
    copy_colors(c, c->vert[2], c->vert[0]);
-   brw_JMPI(p, ip, ip, brw_imm_ud(nr*4+1));
+   brw_JMPI(p, ip, ip, brw_imm_d(jmpi*(nr*4+1)));
 
    copy_colors(c, c->vert[0], c->vert[1]);
    copy_colors(c, c->vert[2], c->vert[1]);
-   brw_JMPI(p, ip, ip, brw_imm_ud(nr*2));
+   brw_JMPI(p, ip, ip, brw_imm_d(jmpi*nr*2));
 
    copy_colors(c, c->vert[0], c->vert[2]);
    copy_colors(c, c->vert[1], c->vert[2]);
@@ -184,7 +189,8 @@ static void do_flatshade_line( struct brw_sf_compile *c )
    struct brw_compile *p = &c->func;
    struct brw_reg ip = brw_ip_reg();
    GLuint nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS);
-   
+   GLuint jmpi = 1;
+
    if (!nr)
       return;
 
@@ -193,13 +199,16 @@ static void do_flatshade_line( struct brw_sf_compile *c )
    if (c->key.primitive == SF_UNFILLED_TRIS)
       return;
 
+   if (BRW_IS_IGDNG(p->brw))
+       jmpi = 2;
+
    brw_push_insn_state(p);
    
-   brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr+1));
+   brw_MUL(p, c->pv, c->pv, brw_imm_d(jmpi*(nr+1)));
    brw_JMPI(p, ip, ip, c->pv);
    copy_colors(c, c->vert[1], c->vert[0]);
 
-   brw_JMPI(p, ip, ip, brw_imm_ud(nr));
+   brw_JMPI(p, ip, ip, brw_imm_ud(jmpi*nr));
    copy_colors(c, c->vert[0], c->vert[1]);
 
    brw_pop_insn_state(p);
@@ -218,7 +227,7 @@ static void alloc_regs( struct brw_sf_compile *c )
 
    /* Values computed by fixed function unit:
     */
-   c->pv  = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_UD);
+   c->pv  = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_D);
    c->det = brw_vec1_grf(1, 2);
    c->dx0 = brw_vec1_grf(1, 3);
    c->dx2 = brw_vec1_grf(1, 4);
@@ -295,9 +304,6 @@ static void invert_det( struct brw_sf_compile *c)
 
 }
 
-#define NON_PERPECTIVE_ATTRS  (FRAG_BIT_WPOS | \
-                               FRAG_BIT_COL0 | \
-			       FRAG_BIT_COL1)
 
 static GLboolean calculate_masks( struct brw_sf_compile *c,
 				  GLuint reg,
@@ -306,9 +312,16 @@ static GLboolean calculate_masks( struct brw_sf_compile *c,
 				  GLushort *pc_linear)
 {
    GLboolean is_last_attr = (reg == c->nr_setup_regs - 1);
-   GLuint persp_mask = c->key.attrs & ~NON_PERPECTIVE_ATTRS;
+   GLuint persp_mask;
    GLuint linear_mask;
 
+   if (c->key.do_flat_shading || c->key.linear_color)
+      persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS |
+                                    FRAG_BIT_COL0 |
+                                    FRAG_BIT_COL1);
+   else
+      persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS);
+
    if (c->key.do_flat_shading)
       linear_mask = c->key.attrs & ~(FRAG_BIT_COL0|FRAG_BIT_COL1);
    else
@@ -674,7 +687,7 @@ void brw_emit_anyprim_setup( struct brw_sf_compile *c )
 					       (1<<_3DPRIM_POLYGON) |
 					       (1<<_3DPRIM_RECTLIST) |
 					       (1<<_3DPRIM_TRIFAN_NOSTIPPLE)));
-   jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+   jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
    {
       saveflag = p->flag_value;
       brw_push_insn_state(p); 
@@ -695,7 +708,7 @@ void brw_emit_anyprim_setup( struct brw_sf_compile *c )
 					       (1<<_3DPRIM_LINESTRIP_CONT) |
 					       (1<<_3DPRIM_LINESTRIP_BF) |
 					       (1<<_3DPRIM_LINESTRIP_CONT_BF)));
-   jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+   jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
    {
       saveflag = p->flag_value;
       brw_push_insn_state(p); 
@@ -708,7 +721,7 @@ void brw_emit_anyprim_setup( struct brw_sf_compile *c )
 
    brw_set_conditionalmod(p, BRW_CONDITIONAL_Z);
    brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<<BRW_SPRITE_POINT_ENABLE));
-   jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+   jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
    {
       saveflag = p->flag_value;
       brw_push_insn_state(p); 
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index c99918724b..b5f6371c82 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -162,7 +162,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 {
    struct brw_sf_unit_state sf;
    dri_bo *bo;
-
+   int chipset_max_threads;
    memset(&sf, 0, sizeof(sf));
 
    sf.thread0.grf_reg_count = ALIGN(key->total_grf, 16) / 16 - 1;
@@ -171,13 +171,26 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
 
    sf.thread3.dispatch_grf_start_reg = 3;
-   sf.thread3.urb_entry_read_offset = 1;
+
+   if (BRW_IS_IGDNG(brw))
+       sf.thread3.urb_entry_read_offset = 3;
+   else
+       sf.thread3.urb_entry_read_offset = 1;
+
    sf.thread3.urb_entry_read_length = key->urb_entry_read_length;
 
    sf.thread4.nr_urb_entries = key->nr_urb_entries;
    sf.thread4.urb_entry_allocation_size = key->sfsize - 1;
-   /* Each SF thread produces 1 PUE, and there can be up to 24 threads */
-   sf.thread4.max_threads = MIN2(24, key->nr_urb_entries) - 1;
+
+   /* Each SF thread produces 1 PUE, and there can be up to 24(Pre-IGDNG) or 
+    * 48(IGDNG) threads 
+    */
+   if (BRW_IS_IGDNG(brw))
+      chipset_max_threads = 48;
+   else
+      chipset_max_threads = 24;
+
+   sf.thread4.max_threads = MIN2(chipset_max_threads, key->nr_urb_entries) - 1;
 
    if (INTEL_DEBUG & DEBUG_SINGLE_THREAD)
       sf.thread4.max_threads = 0;
@@ -233,7 +246,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    else if (sf.sf6.line_width <= 0x2)
        sf.sf6.line_width = 0;
 
-   /* _NEW_POINT */
+   /* _NEW_BUFFERS */
    key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
    if (!key->render_to_fbo) {
       /* Rendering to an OpenGL window */
@@ -263,6 +276,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    }
    /* XXX clamp max depends on AA vs. non-AA */
 
+   /* _NEW_POINT */
    sf.sf7.sprite_point = key->point_sprite;
    sf.sf7.point_size = CLAMP(rint(key->point_size), 1, 255) * (1<<3);
    sf.sf7.use_point_size_state = !key->point_attenuated;
@@ -328,7 +342,8 @@ const struct brw_tracked_state brw_sf_unit = {
       .mesa  = (_NEW_POLYGON | 
 		_NEW_LINE | 
 		_NEW_POINT | 
-		_NEW_SCISSOR),
+		_NEW_SCISSOR |
+		_NEW_BUFFERS),
       .brw   = BRW_NEW_URB_FENCE,
       .cache = (CACHE_NEW_SF_VP |
 		CACHE_NEW_SF_PROG)
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index 81b0a45998..bf9f6cae55 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -72,11 +72,13 @@ const struct brw_tracked_state brw_sf_vp;
 const struct brw_tracked_state brw_state_base_address;
 const struct brw_tracked_state brw_urb_fence;
 const struct brw_tracked_state brw_vertex_state;
+const struct brw_tracked_state brw_vs_surfaces;
 const struct brw_tracked_state brw_vs_prog;
 const struct brw_tracked_state brw_vs_unit;
 const struct brw_tracked_state brw_wm_input_sizes;
 const struct brw_tracked_state brw_wm_prog;
 const struct brw_tracked_state brw_wm_samplers;
+const struct brw_tracked_state brw_wm_constant_surface;
 const struct brw_tracked_state brw_wm_surfaces;
 const struct brw_tracked_state brw_wm_unit;
 
@@ -91,6 +93,20 @@ const struct brw_tracked_state brw_drawing_rect;
 const struct brw_tracked_state brw_indices;
 const struct brw_tracked_state brw_vertices;
 
+/**
+ * Use same key for WM and VS surfaces.
+ */
+struct brw_surface_key {
+   GLenum target, depthmode;
+   dri_bo *bo;
+   GLint format, internal_format;
+   GLint first_level, last_level;
+   GLint width, height, depth;
+   GLint pitch, cpp;
+   uint32_t tiling;
+   GLuint offset;
+};
+
 /***********************************************************************
  * brw_state.c
  */
@@ -135,8 +151,8 @@ dri_bo *brw_search_cache( struct brw_cache *cache,
 			  void *aux_return);
 void brw_state_cache_check_size( struct brw_context *brw );
 
-void brw_init_cache( struct brw_context *brw );
-void brw_destroy_cache( struct brw_context *brw );
+void brw_init_caches( struct brw_context *brw );
+void brw_destroy_caches( struct brw_context *brw );
 
 /***********************************************************************
  * brw_state_batch.c
@@ -150,4 +166,9 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw,
 void brw_destroy_batch_cache( struct brw_context *brw );
 void brw_clear_batch_cache_flush( struct brw_context *brw );
 
+/* brw_wm_surface_state.c */
+dri_bo *
+brw_create_constant_surface( struct brw_context *brw,
+                             struct brw_surface_key *key );
+
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index d5b5166406..e40d7a0416 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -56,9 +56,9 @@
  * incorrect program is run for the other instance.
  */
 
+#include "main/imports.h"
 #include "brw_state.h"
 #include "intel_batchbuffer.h"
-#include "main/imports.h"
 
 /* XXX: Fixme - have to include these to get the sizes of the prog_key
  * structs:
@@ -69,8 +69,10 @@
 #include "brw_sf.h"
 #include "brw_gs.h"
 
-static GLuint hash_key( const void *key, GLuint key_size,
-			dri_bo **reloc_bufs, GLuint nr_reloc_bufs)
+
+static GLuint
+hash_key(const void *key, GLuint key_size,
+         dri_bo **reloc_bufs, GLuint nr_reloc_bufs)
 {
    GLuint *ikey = (GLuint *)key;
    GLuint hash = 0, i;
@@ -95,6 +97,7 @@ static GLuint hash_key( const void *key, GLuint key_size,
    return hash;
 }
 
+
 /**
  * Marks a new buffer as being chosen for the given cache id.
  */
@@ -111,6 +114,7 @@ update_cache_last(struct brw_cache *cache, enum brw_cache_id cache_id,
    cache->brw->state.dirty.cache |= 1 << cache_id;
 }
 
+
 static struct brw_cache_item *
 search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
 	     GLuint hash, const void *key, GLuint key_size,
@@ -143,7 +147,8 @@ search_cache(struct brw_cache *cache, enum brw_cache_id cache_id,
 }
 
 
-static void rehash( struct brw_cache *cache )
+static void
+rehash(struct brw_cache *cache)
 {
    struct brw_cache_item **items;
    struct brw_cache_item *c, *next;
@@ -164,15 +169,17 @@ static void rehash( struct brw_cache *cache )
    cache->size = size;
 }
 
+
 /**
  * Returns the buffer object matching cache_id and key, or NULL.
  */
-dri_bo *brw_search_cache( struct brw_cache *cache,
-			  enum brw_cache_id cache_id,
-			  const void *key,
-			  GLuint key_size,
-			  dri_bo **reloc_bufs, GLuint nr_reloc_bufs,
-			  void *aux_return )
+dri_bo *
+brw_search_cache(struct brw_cache *cache,
+                 enum brw_cache_id cache_id,
+                 const void *key,
+                 GLuint key_size,
+                 dri_bo **reloc_bufs, GLuint nr_reloc_bufs,
+                 void *aux_return)
 {
    struct brw_cache_item *item;
    GLuint hash = hash_key(key, key_size, reloc_bufs, nr_reloc_bufs);
@@ -192,6 +199,7 @@ dri_bo *brw_search_cache( struct brw_cache *cache,
    return item->bo;
 }
 
+
 dri_bo *
 brw_upload_cache( struct brw_cache *cache,
 		  enum brw_cache_id cache_id,
@@ -265,7 +273,9 @@ brw_upload_cache( struct brw_cache *cache,
    return bo;
 }
 
-/* This doesn't really work with aux data.  Use search/upload instead
+
+/**
+ * This doesn't really work with aux data.  Use search/upload instead
  */
 dri_bo *
 brw_cache_data_sz(struct brw_cache *cache,
@@ -296,6 +306,7 @@ brw_cache_data_sz(struct brw_cache *cache,
    return bo;
 }
 
+
 /**
  * Wrapper around brw_cache_data_sz using the cache_id's canonical key size.
  *
@@ -319,21 +330,22 @@ enum pool_type {
    DW_GENERAL_STATE
 };
 
+
 static void
-brw_init_cache_id( struct brw_context *brw,
-		const char *name,
-		enum brw_cache_id id,
-		GLuint key_size,
-		GLuint aux_size)
+brw_init_cache_id(struct brw_cache *cache,
+                  const char *name,
+                  enum brw_cache_id id,
+                  GLuint key_size,
+                  GLuint aux_size)
 {
-   struct brw_cache *cache = &brw->cache;
-
    cache->name[id] = strdup(name);
    cache->key_size[id] = key_size;
    cache->aux_size[id] = aux_size;
 }
 
-void brw_init_cache( struct brw_context *brw )
+
+static void
+brw_init_non_surface_cache(struct brw_context *brw)
 {
    struct brw_cache *cache = &brw->cache;
 
@@ -342,114 +354,136 @@ void brw_init_cache( struct brw_context *brw )
    cache->size = 7;
    cache->n_items = 0;
    cache->items = (struct brw_cache_item **)
-      _mesa_calloc(cache->size * 
-		   sizeof(struct brw_cache_item));
+      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CC_VP",
 		     BRW_CC_VP,
 		     sizeof(struct brw_cc_viewport),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CC_UNIT",
 		     BRW_CC_UNIT,
 		     sizeof(struct brw_cc_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "WM_PROG",
 		     BRW_WM_PROG,
 		     sizeof(struct brw_wm_prog_key),
 		     sizeof(struct brw_wm_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SAMPLER_DEFAULT_COLOR",
 		     BRW_SAMPLER_DEFAULT_COLOR,
 		     sizeof(struct brw_sampler_default_color),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SAMPLER",
 		     BRW_SAMPLER,
 		     0,		/* variable key/data size */
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "WM_UNIT",
 		     BRW_WM_UNIT,
 		     sizeof(struct brw_wm_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_PROG",
 		     BRW_SF_PROG,
 		     sizeof(struct brw_sf_prog_key),
 		     sizeof(struct brw_sf_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_VP",
 		     BRW_SF_VP,
 		     sizeof(struct brw_sf_viewport),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SF_UNIT",
 		     BRW_SF_UNIT,
 		     sizeof(struct brw_sf_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "VS_UNIT",
 		     BRW_VS_UNIT,
 		     sizeof(struct brw_vs_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "VS_PROG",
 		     BRW_VS_PROG,
 		     sizeof(struct brw_vs_prog_key),
 		     sizeof(struct brw_vs_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CLIP_UNIT",
 		     BRW_CLIP_UNIT,
 		     sizeof(struct brw_clip_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "CLIP_PROG",
 		     BRW_CLIP_PROG,
 		     sizeof(struct brw_clip_prog_key),
 		     sizeof(struct brw_clip_prog_data));
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "GS_UNIT",
 		     BRW_GS_UNIT,
 		     sizeof(struct brw_gs_unit_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "GS_PROG",
 		     BRW_GS_PROG,
 		     sizeof(struct brw_gs_prog_key),
 		     sizeof(struct brw_gs_prog_data));
+}
+
+
+static void
+brw_init_surface_cache(struct brw_context *brw)
+{
+   struct brw_cache *cache = &brw->surface_cache;
+
+   cache->brw = brw;
 
-   brw_init_cache_id(brw,
+   cache->size = 7;
+   cache->n_items = 0;
+   cache->items = (struct brw_cache_item **)
+      _mesa_calloc(cache->size * sizeof(struct brw_cache_item));
+
+   brw_init_cache_id(cache,
 		     "SS_SURFACE",
 		     BRW_SS_SURFACE,
 		     sizeof(struct brw_surface_state),
 		     0);
 
-   brw_init_cache_id(brw,
+   brw_init_cache_id(cache,
 		     "SS_SURF_BIND",
 		     BRW_SS_SURF_BIND,
 		     0,
 		     0);
 }
 
+
+void
+brw_init_caches(struct brw_context *brw)
+{
+   brw_init_non_surface_cache(brw);
+   brw_init_surface_cache(brw);
+}
+
+
 static void
-brw_clear_cache( struct brw_context *brw )
+brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    struct brw_cache_item *c, *next;
    GLuint i;
@@ -457,8 +491,8 @@ brw_clear_cache( struct brw_context *brw )
    if (INTEL_DEBUG & DEBUG_STATE)
       _mesa_printf("%s\n", __FUNCTION__);
 
-   for (i = 0; i < brw->cache.size; i++) {
-      for (c = brw->cache.items[i]; c; c = next) {
+   for (i = 0; i < cache->size; i++) {
+      for (c = cache->items[i]; c; c = next) {
 	 int j;
 
 	 next = c->next;
@@ -468,10 +502,10 @@ brw_clear_cache( struct brw_context *brw )
 	 free((void *)c->key);
 	 free(c);
       }
-      brw->cache.items[i] = NULL;
+      cache->items[i] = NULL;
    }
 
-   brw->cache.n_items = 0;
+   cache->n_items = 0;
 
    if (brw->curbe.last_buf) {
       _mesa_free(brw->curbe.last_buf);
@@ -483,25 +517,46 @@ brw_clear_cache( struct brw_context *brw )
    brw->state.dirty.cache |= ~0;
 }
 
-void brw_state_cache_check_size( struct brw_context *brw )
+
+void
+brw_state_cache_check_size(struct brw_context *brw)
 {
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+
    /* un-tuned guess.  We've got around 20 state objects for a total of around
     * 32k, so 1000 of them is around 1.5MB.
     */
    if (brw->cache.n_items > 1000)
-      brw_clear_cache(brw);
+      brw_clear_cache(brw, &brw->cache);
+
+   if (brw->surface_cache.n_items > 1000)
+      brw_clear_cache(brw, &brw->surface_cache);
 }
 
-void brw_destroy_cache( struct brw_context *brw )
+
+static void
+brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
-   brw_clear_cache(brw);
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s\n", __FUNCTION__);
+
+   brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
-      dri_bo_unreference(brw->cache.last_bo[i]);
-      free(brw->cache.name[i]);
+      dri_bo_unreference(cache->last_bo[i]);
+      free(cache->name[i]);
    }
-   free(brw->cache.items);
-   brw->cache.items = NULL;
-   brw->cache.size = 0;
+   free(cache->items);
+   cache->items = NULL;
+   cache->size = 0;
+}
+
+
+void
+brw_destroy_caches(struct brw_context *brw)
+{
+   brw_destroy_cache(brw, &brw->cache);
+   brw_destroy_cache(brw, &brw->surface_cache);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_state_dump.c b/src/mesa/drivers/dri/i965/brw_state_dump.c
index a713262269..e94fa7d2b4 100644
--- a/src/mesa/drivers/dri/i965/brw_state_dump.c
+++ b/src/mesa/drivers/dri/i965/brw_state_dump.c
@@ -126,6 +126,8 @@ static void dump_wm_surface_state(struct brw_context *brw)
 		surf->ss3.pitch + 1, surf->ss3.tiled_surface ? "" : "not ");
       state_out(name, surf, surfoff, 4, "mip base %d\n",
 		surf->ss4.min_lod);
+      state_out(name, surf, surfoff, 5, "x,y offset: %d,%d\n",
+		surf->ss5.x_offset, surf->ss5.y_offset);
 
       dri_bo_unmap(surf_bo);
    }
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 5de1450e61..38d9dd8991 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -59,11 +59,12 @@ const struct brw_tracked_state *atoms[] =
    &brw_curbe_offsets,
    &brw_recalculate_urb_fence,
 
-
    &brw_cc_vp,
    &brw_cc_unit,
 
-   &brw_wm_surfaces,		/* must do before samplers */
+   &brw_vs_surfaces,		/* must do before unit */
+   &brw_wm_constant_surface,	/* must do before wm surfaces/bind bo */
+   &brw_wm_surfaces,		/* must do before samplers and unit */
    &brw_wm_samplers,
 
    &brw_wm_unit,
@@ -88,54 +89,26 @@ const struct brw_tracked_state *atoms[] =
 
    &brw_line_stipple,
    &brw_aa_line_parameters,
-   /* Ordering of the commands below is documented as fixed.  
-    */
-#if 0
-   &brw_pipelined_state_pointers,
-   &brw_urb_fence,
-   &brw_constant_buffer_state,
-#else
+
    &brw_psp_urb_cbs,
-#endif
 
    &brw_drawing_rect,
    &brw_indices,
    &brw_vertices,
 
-   NULL,			/* brw_constant_buffer */
+   &brw_constant_buffer
 };
 
 
 void brw_init_state( struct brw_context *brw )
 {
-   GLuint i;
-
-   brw_init_cache(brw);
-
-   brw->state.atoms = _mesa_malloc(sizeof(atoms));
-   brw->state.nr_atoms = sizeof(atoms)/sizeof(*atoms);
-   _mesa_memcpy(brw->state.atoms, atoms, sizeof(atoms));
-
-   /* Patch in a pointer to the dynamic state atom:
-    */
-   for (i = 0; i < brw->state.nr_atoms; i++)
-      if (brw->state.atoms[i] == NULL)
-	 brw->state.atoms[i] = &brw->curbe.tracked_state;
-
-   _mesa_memcpy(&brw->curbe.tracked_state, 
-		&brw_constant_buffer,
-		sizeof(brw_constant_buffer));
+   brw_init_caches(brw);
 }
 
 
 void brw_destroy_state( struct brw_context *brw )
 {
-   if (brw->state.atoms) {
-      _mesa_free(brw->state.atoms);
-      brw->state.atoms = NULL;
-   }
-
-   brw_destroy_cache(brw);
+   brw_destroy_caches(brw);
    brw_destroy_batch_cache(brw);
 }
 
@@ -218,6 +191,7 @@ static struct dirty_bit_map mesa_bits[] = {
    DEFINE_BIT(_NEW_MULTISAMPLE),
    DEFINE_BIT(_NEW_TRACK_MATRIX),
    DEFINE_BIT(_NEW_PROGRAM),
+   DEFINE_BIT(_NEW_PROGRAM_CONSTANTS),
    {0, 0, 0}
 };
 
@@ -231,7 +205,6 @@ static struct dirty_bit_map brw_bits[] = {
    DEFINE_BIT(BRW_NEW_PRIMITIVE),
    DEFINE_BIT(BRW_NEW_CONTEXT),
    DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS),
-   DEFINE_BIT(BRW_NEW_INPUT_VARYING),
    DEFINE_BIT(BRW_NEW_PSP),
    DEFINE_BIT(BRW_NEW_FENCE),
    DEFINE_BIT(BRW_NEW_INDICES),
@@ -336,7 +309,7 @@ void brw_validate_state( struct brw_context *brw )
 
    /* do prepare stage for all atoms */
    for (i = 0; i < Elements(atoms); i++) {
-      const struct brw_tracked_state *atom = brw->state.atoms[i];
+      const struct brw_tracked_state *atom = atoms[i];
 
       if (brw->intel.Fallback)
          break;
@@ -367,8 +340,8 @@ void brw_upload_state(struct brw_context *brw)
       _mesa_memset(&examined, 0, sizeof(examined));
       prev = *state;
 
-      for (i = 0; i < brw->state.nr_atoms; i++) {	 
-	 const struct brw_tracked_state *atom = brw->state.atoms[i];
+      for (i = 0; i < Elements(atoms); i++) {	 
+	 const struct brw_tracked_state *atom = atoms[i];
 	 struct brw_state_flags generated;
 
 	 assert(atom->dirty.mesa ||
@@ -397,7 +370,7 @@ void brw_upload_state(struct brw_context *brw)
    }
    else {
       for (i = 0; i < Elements(atoms); i++) {	 
-	 const struct brw_tracked_state *atom = brw->state.atoms[i];
+	 const struct brw_tracked_state *atom = atoms[i];
 
 	 if (brw->intel.Fallback)
 	    break;
diff --git a/src/mesa/drivers/dri/i965/brw_structs.h b/src/mesa/drivers/dri/i965/brw_structs.h
index 89e2981203..8ba7eb27b3 100644
--- a/src/mesa/drivers/dri/i965/brw_structs.h
+++ b/src/mesa/drivers/dri/i965/brw_structs.h
@@ -33,6 +33,14 @@
 #ifndef BRW_STRUCTS_H
 #define BRW_STRUCTS_H
 
+
+/** Number of general purpose registers (VS, WM, etc) */
+#define BRW_MAX_GRF 128
+
+/** Number of message register file registers */
+#define BRW_MAX_MRF 16
+
+
 /* Command packets:
  */
 struct header 
@@ -815,7 +823,9 @@ struct brw_gs_unit_state
 
    struct
    {
-      GLuint pad0:10;
+      GLuint pad0:8;
+      GLuint rendering_enable:1; /* for IGDNG */
+      GLuint pad4:1;
       GLuint stats_enable:1; 
       GLuint nr_urb_entries:7; 
       GLuint pad1:1;
@@ -923,6 +933,28 @@ struct brw_wm_unit_state
    
    GLfloat global_depth_offset_constant;  
    GLfloat global_depth_offset_scale;   
+   
+   /* for IGDNG only */
+   struct {
+      GLuint pad0:1;
+      GLuint grf_reg_count_1:3; 
+      GLuint pad1:2;
+      GLuint kernel_start_pointer_1:26;
+   } wm8;       
+
+   struct {
+      GLuint pad0:1;
+      GLuint grf_reg_count_2:3; 
+      GLuint pad1:2;
+      GLuint kernel_start_pointer_2:26;
+   } wm9;       
+
+   struct {
+      GLuint pad0:1;
+      GLuint grf_reg_count_3:3; 
+      GLuint pad1:2;
+      GLuint kernel_start_pointer_3:26;
+   } wm10;       
 };
 
 struct brw_sampler_default_color {
@@ -1075,7 +1107,7 @@ struct brw_surface_state
       GLuint y_offset:4;
       GLuint pad0:1;
       GLuint x_offset:7;
-   } ss5;   /* NEW in Integrated Graphics Device */
+   } ss5;   /* New in G4X */
 
 };
 
@@ -1298,6 +1330,14 @@ struct brw_instruction
 	 GLuint pad1:6;
       } ia16;
 
+       struct 
+       {
+           GLuint pad:26;
+           GLuint end_of_thread:1;
+           GLuint pad1:1;
+           GLuint sfid:4;
+       } send_igdng;  /* for IGDNG only */
+
    } bits2;
 
    union
@@ -1385,6 +1425,21 @@ struct brw_instruction
       } math;
 
       struct {
+	 GLuint function:4;
+	 GLuint int_type:1;
+	 GLuint precision:1;
+	 GLuint saturate:1;
+	 GLuint data_type:1;
+	 GLuint snapshot:1;
+	 GLuint pad0:10;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } math_igdng;
+
+      struct {
 	 GLuint binding_table_index:8;
 	 GLuint sampler:4;
 	 GLuint return_format:2; 
@@ -1407,9 +1462,38 @@ struct brw_instruction
          GLuint end_of_thread:1;
       } sampler_g4x;
 
+      struct {
+	 GLuint binding_table_index:8;
+	 GLuint sampler:4;
+	 GLuint msg_type:4;
+	 GLuint simd_mode:2;
+	 GLuint pad0:1;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } sampler_igdng;
+
       struct brw_urb_immediate urb;
 
       struct {
+	 GLuint opcode:4;
+	 GLuint offset:6;
+	 GLuint swizzle_control:2; 
+	 GLuint pad:1;
+	 GLuint allocate:1;
+	 GLuint used:1;
+	 GLuint complete:1;
+	 GLuint pad0:3;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } urb_igdng;
+
+      struct {
 	 GLuint binding_table_index:8;
 	 GLuint msg_control:4;  
 	 GLuint msg_type:2;  
@@ -1423,6 +1507,19 @@ struct brw_instruction
 
       struct {
 	 GLuint binding_table_index:8;
+	 GLuint msg_control:3;  
+	 GLuint msg_type:3;  
+	 GLuint target_cache:2;    
+	 GLuint pad0:3;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } dp_read_igdng;
+
+      struct {
+	 GLuint binding_table_index:8;
 	 GLuint msg_control:3;
 	 GLuint pixel_scoreboard_clear:1;
 	 GLuint msg_type:3;    
@@ -1435,6 +1532,20 @@ struct brw_instruction
       } dp_write;
 
       struct {
+	 GLuint binding_table_index:8;
+	 GLuint msg_control:3;
+	 GLuint pixel_scoreboard_clear:1;
+	 GLuint msg_type:3;    
+	 GLuint send_commit_msg:1;
+	 GLuint pad0:3;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } dp_write_igdng;
+
+      struct {
 	 GLuint pad:16;
 	 GLuint response_length:4;
 	 GLuint msg_length:4;
@@ -1443,6 +1554,15 @@ struct brw_instruction
 	 GLuint end_of_thread:1;
       } generic;
 
+      struct {
+	 GLuint pad:19;
+	 GLuint header_present:1;
+	 GLuint response_length:5;
+	 GLuint msg_length:4;
+	 GLuint pad1:2;
+	 GLuint end_of_thread:1;
+      } generic_igdng;
+
       GLint d;
       GLuint ud;
    } bits3;
diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c
index 51a617fcb4..7f9b253534 100644
--- a/src/mesa/drivers/dri/i965/brw_tex_layout.c
+++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c
@@ -28,7 +28,6 @@
   * Authors:
   *   Keith Whitwell <keith@tungstengraphics.com>
   */
-        
 
 /* Code to layout images in a mipmap tree for i965.
  */
@@ -37,17 +36,90 @@
 #include "intel_tex_layout.h"
 #include "intel_context.h"
 #include "main/macros.h"
+#include "intel_chipset.h"
 
 #define FILE_DEBUG_FLAG DEBUG_MIPTREE
 
-GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_tree *mt )
+GLboolean brw_miptree_layout(struct intel_context *intel,
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling)
 {
-   /* XXX: these vary depending on image format: 
-    */
-/*    GLint align_w = 4; */
+   /* XXX: these vary depending on image format: */
+   /* GLint align_w = 4; */
 
    switch (mt->target) {
-   case GL_TEXTURE_CUBE_MAP: 
+   case GL_TEXTURE_CUBE_MAP:
+      if (IS_IGDNG(intel->intelScreen->deviceID)) {
+          GLuint align_h = 2, align_w = 4;
+          GLuint level;
+          GLuint x = 0;
+          GLuint y = 0;
+          GLuint width = mt->width0;
+          GLuint height = mt->height0;
+          GLuint qpitch = 0;
+          GLuint y_pitch = 0;
+
+          mt->pitch = mt->width0;
+          intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
+          y_pitch = ALIGN(height, align_h);
+
+          if (mt->compressed) {
+              mt->pitch = ALIGN(mt->width0, align_w);
+              qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * mt->pitch * mt->cpp;
+              mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * 6;
+          } else {
+              qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * mt->pitch * mt->cpp;
+              mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * 6;
+          }
+
+          if (mt->first_level != mt->last_level) {
+              GLuint mip1_width;
+
+              if (mt->compressed) {
+                  mip1_width = ALIGN(minify(mt->width0), align_w)
+                      + ALIGN(minify(minify(mt->width0)), align_w);
+              } else {
+                  mip1_width = ALIGN(minify(mt->width0), align_w)
+                      + minify(minify(mt->width0));
+              }
+
+              if (mip1_width > mt->pitch) {
+                  mt->pitch = mip1_width;
+              }
+          }
+
+          mt->pitch = intel_miptree_pitch_align(intel, mt, tiling, mt->pitch);
+
+          for (level = mt->first_level; level <= mt->last_level; level++) {
+              GLuint img_height;
+              GLuint nr_images = 6;
+              GLuint q = 0;
+
+              intel_miptree_set_level_info(mt, level, nr_images, x, y, width, 
+                                           height, 1);
+
+              for (q = 0; q < nr_images; q++)
+                  intel_miptree_set_image_offset_ex(mt, level, q, x, y, q * qpitch);
+
+              if (mt->compressed)
+                  img_height = MAX2(1, height/4);
+              else
+                  img_height = ALIGN(height, align_h);
+
+              if (level == mt->first_level + 1) {
+                  x += ALIGN(width, align_w);
+              }
+              else {
+                  y += img_height;
+              }
+
+              width  = minify(width);
+              height = minify(height);
+          }
+
+          break;
+      }
+
    case GL_TEXTURE_3D: {
       GLuint width  = mt->width0;
       GLuint height = mt->height0;
@@ -59,25 +131,25 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
       GLuint align_w = 4;
 
       mt->total_height = 0;
-      
+      intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
+
       if (mt->compressed) {
-          align_w = intel_compressed_alignment(mt->internal_format);
           mt->pitch = ALIGN(width, align_w);
           pack_y_pitch = (height + 3) / 4;
       } else {
-          mt->pitch = intel_miptree_pitch_align (intel, mt, mt->width0);
-          pack_y_pitch = ALIGN(mt->height0, align_h);
+	 mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
+	 pack_y_pitch = ALIGN(mt->height0, align_h);
       }
 
-      pack_x_pitch = mt->pitch;
+      pack_x_pitch = width;
       pack_x_nr = 1;
 
-      for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
+      for (level = mt->first_level ; level <= mt->last_level ; level++) {
 	 GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6;
 	 GLint x = 0;
 	 GLint y = 0;
 	 GLint q, j;
-	    
+
 	 intel_miptree_set_level_info(mt, level, nr_images,
 				      0, mt->total_height,
 				      width, height, depth);
@@ -89,7 +161,7 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
 	    }
 
 	    x = 0;
-	    y += pack_y_pitch;	    
+	    y += pack_y_pitch;
 	 }
 
 
@@ -98,40 +170,40 @@ GLboolean brw_miptree_layout( struct intel_context *intel, struct intel_mipmap_t
 	 height = minify(height);
 	 depth  = minify(depth);
 
-    if (mt->compressed) {
-        pack_y_pitch = (height + 3) / 4;
-        
-        if (pack_x_pitch > ALIGN(width, align_w)) {
-            pack_x_pitch = ALIGN(width, align_w);
-            pack_x_nr <<= 1;
-        }
-    } else {
-        if (pack_x_pitch > 4) {
-            pack_x_pitch >>= 1;
-            pack_x_nr <<= 1;
-            assert(pack_x_pitch * pack_x_nr <= mt->pitch);
-        }
-
-        if (pack_y_pitch > 2) {
-            pack_y_pitch >>= 1;
-            pack_y_pitch = ALIGN(pack_y_pitch, align_h);
-        }
-    }
+	 if (mt->compressed) {
+	    pack_y_pitch = (height + 3) / 4;
+
+	    if (pack_x_pitch > ALIGN(width, align_w)) {
+	       pack_x_pitch = ALIGN(width, align_w);
+	       pack_x_nr <<= 1;
+	    }
+	 } else {
+	    if (pack_x_pitch > 4) {
+	       pack_x_pitch >>= 1;
+	       pack_x_nr <<= 1;
+	       assert(pack_x_pitch * pack_x_nr <= mt->pitch);
+	    }
+
+	    if (pack_y_pitch > 2) {
+	       pack_y_pitch >>= 1;
+	       pack_y_pitch = ALIGN(pack_y_pitch, align_h);
+	    }
+	 }
 
       }
       break;
    }
 
    default:
-      i945_miptree_layout_2d(intel, mt);
+      i945_miptree_layout_2d(intel, mt, tiling);
       break;
    }
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, 
-		mt->pitch, 
+   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
+		mt->pitch,
 		mt->total_height,
 		mt->cpp,
 		mt->pitch * mt->total_height * mt->cpp );
-		
+
    return GL_TRUE;
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c
index 7673dd36eb..8c6f4355a6 100644
--- a/src/mesa/drivers/dri/i965/brw_urb.c
+++ b/src/mesa/drivers/dri/i965/brw_urb.c
@@ -143,7 +143,29 @@ static void recalculate_urb_fence( struct brw_context *brw )
       brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries;
       brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries;	
       brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries;	
-      
+
+      brw->urb.constrained = 0;
+
+      if (BRW_IS_IGDNG(brw)) {
+         brw->urb.nr_vs_entries = 128;
+         brw->urb.nr_sf_entries = 48;
+         if (check_urb_layout(brw)) {
+            goto done;
+         } else {
+            brw->urb.constrained = 1;
+            brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries;
+            brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries;
+         }
+      } else if (BRW_IS_G4X(brw)) {
+	 brw->urb.nr_vs_entries = 64;
+	 if (check_urb_layout(brw)) {
+	    goto done;
+	 } else {
+	    brw->urb.constrained = 1;
+	    brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries;
+	 }
+      }
+
       if (!check_urb_layout(brw)) {
 	 brw->urb.nr_vs_entries = limits[VS].min_nr_entries;	
 	 brw->urb.nr_gs_entries = limits[GS].min_nr_entries;	
@@ -169,9 +191,8 @@ static void recalculate_urb_fence( struct brw_context *brw )
 	 if (INTEL_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
 	    _mesa_printf("URB CONSTRAINED\n");
       }
-      else 
-	 brw->urb.constrained = 0;
 
+done:
       if (INTEL_DEBUG & DEBUG_URB)
 	 _mesa_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
 		      brw->urb.vs_start,
diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h
index 1e4f66091e..4a591365c9 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.h
+++ b/src/mesa/drivers/dri/i965/brw_vs.h
@@ -58,6 +58,7 @@ struct brw_vs_compile {
 
    GLuint first_output;
    GLuint nr_outputs;
+   GLuint first_overflow_output; /**< VERT_ATTRIB_x */
 
    GLuint first_tmp;
    GLuint last_tmp;
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index b69616d6e5..514f15d5e3 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -69,13 +69,18 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 {
    GLuint i, reg = 0, mrf;
 
-#if 0
-   if (c->vp->program.Base.Parameters->NumParameters >= 6)
-      c->vp->use_const_buffer = 1;
+   /* Determine whether to use a real constant buffer or use a block
+    * of GRF registers for constants.  The later is faster but only
+    * works if everything fits in the GRF.
+    * XXX this heuristic/check may need some fine tuning...
+    */
+   if (c->vp->program.Base.Parameters->NumParameters +
+       c->vp->program.Base.NumTemporaries + 20 > BRW_MAX_GRF)
+      c->vp->use_const_buffer = GL_TRUE;
    else
-#endif
       c->vp->use_const_buffer = GL_FALSE;
-   /*printf("use_const_buffer = %d\n", c->use_const_buffer);*/
+
+   /*printf("use_const_buffer = %d\n", c->vp->use_const_buffer);*/
 
    /* r0 -- reserved as usual
     */
@@ -124,15 +129,21 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
       }
    }
 
-   /* Allocate outputs: TODO: could organize the non-position outputs
-    * to go straight into message regs.
+   /* Allocate outputs.  The non-position outputs go straight into message regs.
     */
    c->nr_outputs = 0;
    c->first_output = reg;
-   mrf = 4;
+   c->first_overflow_output = 0;
+
+   if (BRW_IS_IGDNG(c->func.brw))
+       mrf = 8;
+   else
+       mrf = 4;
+
    for (i = 0; i < VERT_RESULT_MAX; i++) {
       if (c->prog_data.outputs_written & (1 << i)) {
 	 c->nr_outputs++;
+         assert(i < Elements(c->regs[PROGRAM_OUTPUT]));
 	 if (i == VERT_RESULT_HPOS) {
 	    c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
 	    reg++;
@@ -143,8 +154,17 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 	    mrf++;		/* just a placeholder?  XXX fix later stages & remove this */
 	 }
 	 else {
-	    c->regs[PROGRAM_OUTPUT][i] = brw_message_reg(mrf);
-	    mrf++;
+            if (mrf < 16) {
+               c->regs[PROGRAM_OUTPUT][i] = brw_message_reg(mrf);
+               mrf++;
+            }
+            else {
+               /* too many vertex results to fit in MRF, use GRF for overflow */
+               if (!c->first_overflow_output)
+                  c->first_overflow_output = i;
+               c->regs[PROGRAM_OUTPUT][i] = brw_vec8_grf(reg, 0);
+               reg++;
+            }
 	 }
       }
    }     
@@ -201,7 +221,11 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
     */
    c->prog_data.urb_read_length = (c->nr_inputs + 1) / 2;
 
-   c->prog_data.urb_entry_size = (c->nr_outputs + 2 + 3) / 4;
+   if (BRW_IS_IGDNG(c->func.brw))
+       c->prog_data.urb_entry_size = (c->nr_outputs + 6 + 3) / 4;
+   else
+       c->prog_data.urb_entry_size = (c->nr_outputs + 2 + 3) / 4;
+
    c->prog_data.total_grf = reg;
 
    if (INTEL_DEBUG & DEBUG_VS) {
@@ -709,10 +733,11 @@ get_constant(struct brw_vs_compile *c,
    struct brw_compile *p = &c->func;
    struct brw_reg const_reg;
    struct brw_reg const2_reg;
+   const GLboolean relAddr = src->RelAddr;
 
    assert(argIndex < 3);
 
-   if (c->current_const[argIndex].index != src->Index || src->RelAddr) {
+   if (c->current_const[argIndex].index != src->Index || relAddr) {
       struct brw_reg addrReg = c->regs[PROGRAM_ADDRESS][0];
 
       c->current_const[argIndex].index = src->Index;
@@ -725,13 +750,13 @@ get_constant(struct brw_vs_compile *c,
       brw_dp_READ_4_vs(p,
                        c->current_const[argIndex].reg,/* writeback dest */
                        0,                             /* oword */
-                       src->RelAddr,                  /* relative indexing? */
+                       relAddr,                       /* relative indexing? */
                        addrReg,                       /* address register */
                        16 * src->Index,               /* byte offset */
                        SURF_INDEX_VERT_CONST_BUFFER   /* binding table index */
                        );
 
-      if (src->RelAddr) {
+      if (relAddr) {
          /* second read */
          const2_reg = get_tmp(c);
 
@@ -742,7 +767,7 @@ get_constant(struct brw_vs_compile *c,
          brw_dp_READ_4_vs(p,
                           const2_reg,              /* writeback dest */
                           1,                       /* oword */
-                          src->RelAddr,            /* relative indexing? */
+                          relAddr,                 /* relative indexing? */
                           addrReg,                 /* address register */
                           16 * src->Index,         /* byte offset */
                           SURF_INDEX_VERT_CONST_BUFFER
@@ -752,7 +777,7 @@ get_constant(struct brw_vs_compile *c,
 
    const_reg = c->current_const[argIndex].reg;
 
-   if (src->RelAddr) {
+   if (relAddr) {
       /* merge the two Owords into the constant register */
       /* const_reg[7..4] = const2_reg[7..4] */
       brw_MOV(p,
@@ -1061,6 +1086,8 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    struct brw_reg m0 = brw_message_reg(0);
    struct brw_reg pos = c->regs[PROGRAM_OUTPUT][VERT_RESULT_HPOS];
    struct brw_reg ndc;
+   int eot;
+   GLuint len_vertext_header = 2;
 
    if (c->key.copy_edgeflag) {
       brw_MOV(p, 
@@ -1070,14 +1097,16 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 
    /* Build ndc coords */
    ndc = get_tmp(c);
+   /* ndc = 1.0 / pos.w */
    emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL);
+   /* ndc.xyz = pos * ndc */
    brw_MUL(p, brw_writemask(ndc, WRITEMASK_XYZ), pos, ndc);
 
    /* Update the header for point size, user clipping flags, and -ve rhw
     * workaround.
     */
    if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) ||
-       c->key.nr_userclip || !BRW_IS_G4X(p->brw))
+       c->key.nr_userclip || BRW_IS_965(p->brw))
    {
       struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
       GLuint i;
@@ -1108,7 +1137,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
        * Later, clipping will detect ucp[6] and ensure the primitive is
        * clipped against all fixed planes.
        */
-      if (!BRW_IS_G4X(p->brw)) {
+      if (BRW_IS_965(p->brw)) {
 	 brw_CMP(p,
 		 vec8(brw_null_reg()),
 		 BRW_CONDITIONAL_L,
@@ -1135,7 +1164,23 @@ static void emit_vertex_write( struct brw_vs_compile *c)
     */
    brw_set_access_mode(p, BRW_ALIGN_1);
    brw_MOV(p, offset(m0, 2), ndc);
-   brw_MOV(p, offset(m0, 3), pos);
+
+   if (BRW_IS_IGDNG(p->brw)) {
+       /* There are 20 DWs (D0-D19) in VUE vertex header on IGDNG */
+       brw_MOV(p, offset(m0, 3), pos); /* a portion of vertex header */
+       /* m4, m5 contain the distances from vertex to the user clip planeXXX. 
+        * Seems it is useless for us.
+        * m6 is used for aligning, so that the remainder of vertex element is 
+        * reg-aligned.
+        */
+       brw_MOV(p, offset(m0, 7), pos); /* the remainder of vertex element */
+       len_vertext_header = 6;
+   } else {
+       brw_MOV(p, offset(m0, 3), pos);
+       len_vertext_header = 2;
+   }
+
+   eot = (c->first_overflow_output == 0);
 
    brw_urb_WRITE(p, 
 		 brw_null_reg(), /* dest */
@@ -1143,12 +1188,43 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 		 c->r0,		/* src */
 		 0,		/* allocate */
 		 1,		/* used */
-		 c->nr_outputs + 3, /* msg len */
+		 MIN2(c->nr_outputs + 1 + len_vertext_header, (BRW_MAX_MRF-1)), /* msg len */
 		 0,		/* response len */
-		 1, 		/* eot */
+		 eot, 		/* eot */
 		 1, 		/* writes complete */
 		 0, 		/* urb destination offset */
 		 BRW_URB_SWIZZLE_INTERLEAVE);
+
+   if (c->first_overflow_output > 0) {
+      /* Not all of the vertex outputs/results fit into the MRF.
+       * Move the overflowed attributes from the GRF to the MRF and
+       * issue another brw_urb_WRITE().
+       */
+      /* XXX I'm not 100% sure about which MRF regs to use here.  Starting
+       * at mrf[4] atm...
+       */
+      GLuint i, mrf = 0;
+      for (i = c->first_overflow_output; i < VERT_RESULT_MAX; i++) {
+         if (c->prog_data.outputs_written & (1 << i)) {
+            /* move from GRF to MRF */
+            brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]);
+            mrf++;
+         }
+      }
+
+      brw_urb_WRITE(p,
+                    brw_null_reg(), /* dest */
+                    4,              /* starting mrf reg nr */
+                    c->r0,          /* src */
+                    0,              /* allocate */
+                    1,              /* used */
+                    mrf+1,          /* msg len */
+                    0,              /* response len */
+                    1,              /* eot */
+                    1,              /* writes complete */
+                    BRW_MAX_MRF-1,  /* urb destination offset */
+                    BRW_URB_SWIZZLE_INTERLEAVE);
+   }
 }
 
 
@@ -1177,15 +1253,15 @@ post_vs_emit( struct brw_vs_compile *c,
  */
 void brw_vs_emit(struct brw_vs_compile *c )
 {
-#define MAX_IFSN 32
+#define MAX_IF_DEPTH 32
+#define MAX_LOOP_DEPTH 32
    struct brw_compile *p = &c->func;
-   GLuint nr_insns = c->vp->program.Base.NumInstructions;
-   GLuint insn, if_insn = 0;
+   const GLuint nr_insns = c->vp->program.Base.NumInstructions;
+   GLuint insn, if_depth = 0, loop_depth = 0;
    GLuint end_offset = 0;
    struct brw_instruction *end_inst, *last_inst;
-   struct brw_instruction *if_inst[MAX_IFSN];
-   struct brw_indirect stack_index = brw_indirect(0, 0);   
-
+   struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH];
+   const struct brw_indirect stack_index = brw_indirect(0, 0);   
    GLuint index;
    GLuint file;
 
@@ -1219,7 +1295,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 
    for (insn = 0; insn < nr_insns; insn++) {
 
-      struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
+      const struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn];
       struct brw_reg args[3], dst;
       GLuint i;
       
@@ -1232,7 +1308,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
        */
       if (inst->Opcode != OPCODE_SWZ)
 	  for (i = 0; i < 3; i++) {
-	      struct prog_src_register *src = &inst->SrcReg[i];
+	      const struct prog_src_register *src = &inst->SrcReg[i];
 	      index = src->Index;
 	      file = src->File;	
 	      if (file == PROGRAM_OUTPUT && c->output_regs[index].used_in_src)
@@ -1376,16 +1452,51 @@ void brw_vs_emit(struct brw_vs_compile *c )
 	 emit_xpd(p, dst, args[0], args[1]);
 	 break;
       case OPCODE_IF:
-	 assert(if_insn < MAX_IFSN);
-         if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+	 assert(if_depth < MAX_IF_DEPTH);
+         if_inst[if_depth++] = brw_IF(p, BRW_EXECUTE_8);
 	 break;
       case OPCODE_ELSE:
-	 if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]);
+	 if_inst[if_depth-1] = brw_ELSE(p, if_inst[if_depth-1]);
 	 break;
       case OPCODE_ENDIF:
-         assert(if_insn > 0);
-	 brw_ENDIF(p, if_inst[--if_insn]);
+         assert(if_depth > 0);
+	 brw_ENDIF(p, if_inst[--if_depth]);
 	 break;			
+#if 0
+      case OPCODE_BGNLOOP:
+         loop_inst[loop_depth++] = brw_DO(p, BRW_EXECUTE_8);
+         break;
+      case OPCODE_BRK:
+         brw_BREAK(p);
+         brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+         break;
+      case OPCODE_CONT:
+         brw_CONT(p);
+         brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+         break;
+      case OPCODE_ENDLOOP: 
+         {
+            struct brw_instruction *inst0, *inst1;
+            loop_depth--;
+            inst0 = inst1 = brw_WHILE(p, loop_inst[loop_depth]);
+            /* patch all the BREAK/CONT instructions from last BEGINLOOP */
+            while (inst0 > loop_inst[loop_depth]) {
+               inst0--;
+               if (inst0->header.opcode == BRW_OPCODE_BREAK) {
+                  inst0->bits3.if_else.jump_count = inst1 - inst0 + 1;
+                  inst0->bits3.if_else.pop_count = 0;
+               }
+               else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
+                  inst0->bits3.if_else.jump_count = inst1 - inst0;
+                  inst0->bits3.if_else.pop_count = 0;
+               }
+            }
+         }
+         break;
+#else
+         (void) loop_inst;
+         (void) loop_depth;
+#endif
       case OPCODE_BRA:
          brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
          brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index 3d29538843..d790ab6555 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -97,7 +97,11 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
     * brw_urb_WRITE() results.
     */
    vs.thread1.single_program_flow = 0;
-   vs.thread1.binding_table_entry_count = key->nr_surfaces;
+
+   if (BRW_IS_IGDNG(brw))
+      vs.thread1.binding_table_entry_count = 0; /* hardware requirement */
+   else
+      vs.thread1.binding_table_entry_count = key->nr_surfaces;
 
    vs.thread3.urb_entry_read_length = key->urb_entry_read_length;
    vs.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
@@ -105,10 +109,16 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
    vs.thread3.urb_entry_read_offset = 0;
    vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
 
-   vs.thread4.nr_urb_entries = key->nr_urb_entries;
+   if (BRW_IS_IGDNG(brw))
+       vs.thread4.nr_urb_entries = key->nr_urb_entries >> 2;
+   else
+       vs.thread4.nr_urb_entries = key->nr_urb_entries;
+
    vs.thread4.urb_entry_allocation_size = key->urb_size - 1;
 
-   if (BRW_IS_G4X(brw))
+   if (BRW_IS_IGDNG(brw))
+      chipset_max_threads = 72;
+   else if (BRW_IS_G4X(brw))
       chipset_max_threads = 32;
    else
       chipset_max_threads = 16;
@@ -120,6 +130,8 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
 
    /* No samplers for ARB_vp programs:
     */
+   /* It has to be set to 0 for IGDNG
+    */
    vs.vs5.sampler_count = 0;
 
    if (INTEL_DEBUG & DEBUG_STATS)
diff --git a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c
new file mode 100644
index 0000000000..89f47522a1
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c
@@ -0,0 +1,226 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "main/mtypes.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "shader/prog_parameter.h"
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+/* Creates a new VS constant buffer reflecting the current VS program's
+ * constants, if needed by the VS program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_vs_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_vertex_program *vp =
+      (struct brw_vertex_program *) brw->vertex_program;
+   const struct gl_program_parameter_list *params = vp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_VERTEX_PROGRAM */
+   if (!vp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer",
+				     size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
+
+/**
+ * Update the surface state for a VS constant buffer.
+ *
+ * Sets brw->vs.surf_bo[surf] and brw->vp->const_buffer.
+ */
+static void
+brw_update_vs_constant_surface( GLcontext *ctx,
+                                GLuint surf)
+{
+   struct brw_context *brw = brw_context(ctx);
+   struct brw_surface_key key;
+   struct brw_vertex_program *vp =
+      (struct brw_vertex_program *) brw->vertex_program;
+   const struct gl_program_parameter_list *params = vp->program.Base.Parameters;
+
+   assert(surf == 0);
+
+   /* If we're in this state update atom, we need to update VS constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(vp->const_buffer);
+   vp->const_buffer = brw_vs_update_constant_buffer(brw);
+
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (vp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->vs.surf_bo[surf]);
+      brw->vs.surf_bo[surf] = NULL;
+      return;
+   }
+
+   memset(&key, 0, sizeof(key));
+
+   key.format = MESA_FORMAT_RGBA_FLOAT32;
+   key.internal_format = GL_RGBA;
+   key.bo = vp->const_buffer;
+   key.depthmode = GL_NONE;
+   key.pitch = params->NumParameters;
+   key.width = params->NumParameters;
+   key.height = 1;
+   key.depth = 1;
+   key.cpp = 16;
+
+   /*
+   printf("%s:\n", __FUNCTION__);
+   printf("  width %d  height %d  depth %d  cpp %d  pitch %d\n",
+          key.width, key.height, key.depth, key.cpp, key.pitch);
+   */
+
+   drm_intel_bo_unreference(brw->vs.surf_bo[surf]);
+   brw->vs.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
+   if (brw->vs.surf_bo[surf] == NULL) {
+      brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key);
+   }
+}
+
+
+/**
+ * Constructs the binding table for the VS surface state.
+ */
+static dri_bo *
+brw_vs_get_binding_table(struct brw_context *brw)
+{
+   dri_bo *bind_bo;
+
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
+			      NULL, 0,
+			      brw->vs.surf_bo, BRW_VS_MAX_SURF,
+			      NULL);
+
+   if (bind_bo == NULL) {
+      GLuint data_size = BRW_VS_MAX_SURF * sizeof(GLuint);
+      uint32_t *data = malloc(data_size);
+      int i;
+
+      for (i = 0; i < BRW_VS_MAX_SURF; i++)
+         if (brw->vs.surf_bo[i])
+            data[i] = brw->vs.surf_bo[i]->offset;
+         else
+            data[i] = 0;
+
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
+				  NULL, 0,
+				  brw->vs.surf_bo, BRW_VS_MAX_SURF,
+				  data, data_size,
+				  NULL, NULL);
+
+      /* Emit binding table relocations to surface state */
+      for (i = 0; i < BRW_VS_MAX_SURF; i++) {
+	 if (brw->vs.surf_bo[i] != NULL) {
+	    /* The presumed offsets were set in the data values for
+	     * brw_upload_cache.
+	     */
+	    drm_intel_bo_emit_reloc(bind_bo, i * 4,
+				    brw->vs.surf_bo[i], 0,
+				    I915_GEM_DOMAIN_INSTRUCTION, 0);
+	 }
+      }
+
+      free(data);
+   }
+
+   return bind_bo;
+}
+
+/**
+ * Vertex shader surfaces (constant buffer).
+ *
+ * This consumes the state updates for the constant buffer needing
+ * to be updated, and produces BRW_NEW_NR_VS_SURFACES for the VS unit and
+ * CACHE_NEW_SURF_BIND for the binding table upload.
+ */
+static void prepare_vs_surfaces(struct brw_context *brw )
+{
+   GLcontext *ctx = &brw->intel.ctx;
+   int i;
+   int nr_surfaces = 0;
+
+   brw_update_vs_constant_surface(ctx, SURF_INDEX_VERT_CONST_BUFFER);
+
+   for (i = 0; i < BRW_VS_MAX_SURF; i++) {
+      if (brw->vs.surf_bo[i] != NULL) {
+	 nr_surfaces = i + 1;
+      }
+   }
+
+   if (brw->vs.nr_surfaces != nr_surfaces) {
+      brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES;
+      brw->vs.nr_surfaces = nr_surfaces;
+   }
+
+   /* Note that we don't end up updating the bind_bo if we don't have a
+    * surface to be pointing at.  This should be relatively harmless, as it
+    * just slightly increases our working set size.
+    */
+   if (brw->vs.nr_surfaces != 0) {
+      dri_bo_unreference(brw->vs.bind_bo);
+      brw->vs.bind_bo = brw_vs_get_binding_table(brw);
+   }
+}
+
+const struct brw_tracked_state brw_vs_surfaces = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_VERTEX_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_vs_surfaces,
+};
+
+
+
diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c
index ba03afd6c1..ac11790151 100644
--- a/src/mesa/drivers/dri/i965/brw_vtbl.c
+++ b/src/mesa/drivers/dri/i965/brw_vtbl.c
@@ -177,14 +177,6 @@ static void brw_note_fence( struct intel_context *intel, GLuint fence )
    brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE;
 }
 
-
-static void brw_note_unlock( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-   brw_state_cache_check_size(brw);
-}
-
-
 /* called from intelWaitForIdle() and intelFlush()
  *
  * For now, just flush everything.  Could be smarter later.
@@ -194,7 +186,7 @@ static GLuint brw_flush_cmd( void )
    struct brw_mi_flush flush;
    flush.opcode = CMD_MI_FLUSH;
    flush.pad = 0;
-   flush.flags = BRW_FLUSH_READ_CACHE | BRW_FLUSH_STATE_CACHE;
+   flush.flags = BRW_FLUSH_STATE_CACHE;
    return *(GLuint *)&flush;
 }
 
@@ -215,7 +207,6 @@ void brwInitVtbl( struct brw_context *brw )
 
    brw->intel.vtbl.invalidate_state = brw_invalidate_state;
    brw->intel.vtbl.note_fence = brw_note_fence;
-   brw->intel.vtbl.note_unlock = brw_note_unlock;
    brw->intel.vtbl.new_batch = brw_new_batch;
    brw->intel.vtbl.finish_batch = brw_finish_batch;
    brw->intel.vtbl.destroy = brw_destroy_context;
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 8a3b7df9c7..14e05be4f6 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -272,6 +272,9 @@ static void brw_wm_populate_key( struct brw_context *brw,
    /* _NEW_LIGHT */
    key->flat_shade = (ctx->Light.ShadeModel == GL_FLAT);
 
+   /* _NEW_HINT */
+   key->linear_color = (ctx->Hint.PerspectiveCorrection == GL_FASTEST);
+
    /* _NEW_TEXTURE */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
       const struct gl_texture_unit *unit = &ctx->Texture.Unit[i];
@@ -351,6 +354,7 @@ const struct brw_tracked_state brw_wm_prog = {
    .dirty = {
       .mesa  = (_NEW_COLOR |
 		_NEW_DEPTH |
+                _NEW_HINT |
 		_NEW_STENCIL |
 		_NEW_POLYGON |
 		_NEW_LINE |
diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index 295fed851b..ba497432c6 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -63,6 +63,7 @@ struct brw_wm_prog_key {
    GLuint computes_depth:1;	/* could be derived from program string */
    GLuint source_depth_to_render_target:1;
    GLuint flat_shade:1;
+   GLuint linear_color:1;  /**< linear interpolation vs perspective interp */
    GLuint runtime_check_aads_emit:1;
    
    GLbitfield proj_attrib_mask; /**< one bit per fragment program attribute */
@@ -241,15 +242,20 @@ struct brw_wm_compile {
    GLuint max_wm_grf;
    GLuint last_scratch;
 
+   GLuint cur_inst;  /**< index of current instruction */
+
+   GLboolean out_of_regs;  /**< ran out of GRF registers? */
+
    /** Mapping from Mesa registers to hardware registers */
    struct {
       GLboolean inited;
       struct brw_reg reg;
    } wm_regs[PROGRAM_PAYLOAD+1][256][4];
 
+   GLboolean used_grf[BRW_WM_MAX_GRF];
+   GLuint first_free_grf;
    struct brw_reg stack;
    struct brw_reg emit_mask_reg;
-   GLuint reg_index;  /**< Index of next free GRF register */
    GLuint tmp_regs[BRW_WM_MAX_GRF];
    GLuint tmp_index;
    GLuint tmp_max;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index a870e75e52..9f82916c02 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -353,6 +353,19 @@ static void emit_mad( struct brw_compile *p,
    }
 }
 
+static void emit_trunc( struct brw_compile *p,
+		      const struct brw_reg *dst,
+		      GLuint mask,
+		      const struct brw_reg *arg0)
+{
+   GLuint i;
+
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 brw_RNDZ(p, dst[i], arg0[i]);
+      }
+   }
+}
 
 static void emit_lrp( struct brw_compile *p, 
 		      const struct brw_reg *dst,
@@ -701,6 +714,7 @@ static void emit_tex( struct brw_wm_compile *c,
    GLuint msgLength, responseLength;
    GLuint i, nr;
    GLuint emit;
+   GLuint msg_type;
 
    /* How many input regs are there?
     */
@@ -738,6 +752,18 @@ static void emit_tex( struct brw_wm_compile *c,
 
    responseLength = 8;		/* always */
 
+   if (BRW_IS_IGDNG(p->brw)) {
+       if (inst->tex_shadow)
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG;
+       else
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG;
+   } else {
+       if (inst->tex_shadow)
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE;
+       else
+           msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE;
+   }
+
    brw_SAMPLE(p, 
 	      retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW),
 	      1,
@@ -745,12 +771,12 @@ static void emit_tex( struct brw_wm_compile *c,
               SURF_INDEX_TEXTURE(inst->tex_unit),
 	      inst->tex_unit,	  /* sampler */
 	      inst->writemask,
-	      (inst->tex_shadow ? 
-	       BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE : 
-	       BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE),
+	      msg_type, 
 	      responseLength,
 	      msgLength,
-	      0);	
+	      0,	
+	      1,
+	      BRW_SAMPLER_SIMD_MODE_SIMD16);	
 }
 
 
@@ -762,7 +788,7 @@ static void emit_txb( struct brw_wm_compile *c,
 {
    struct brw_compile *p = &c->func;
    GLuint msgLength;
-
+   GLuint msg_type;
    /* Shadow ignored for txb.
     */
    switch (inst->tex_idx) {
@@ -787,6 +813,11 @@ static void emit_txb( struct brw_wm_compile *c,
    brw_MOV(p, brw_message_reg(8), arg[3]);
    msgLength = 9;
 
+   if (BRW_IS_IGDNG(p->brw))
+       msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG;
+   else
+       msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS;
+
    brw_SAMPLE(p, 
 	      retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW),
 	      1,
@@ -794,10 +825,12 @@ static void emit_txb( struct brw_wm_compile *c,
               SURF_INDEX_TEXTURE(inst->tex_unit),
 	      inst->tex_unit,	  /* sampler */
 	      inst->writemask,
-	      BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS,
+	      msg_type,
 	      8,		/* responseLength */
 	      msgLength,
-	      0);	
+	      0,	
+	      1,
+	      BRW_SAMPLER_SIMD_MODE_SIMD16);	
 }
 
 
@@ -1009,7 +1042,7 @@ static void emit_fb_write( struct brw_wm_compile *c,
 	      get_element_ud(brw_vec8_grf(1,0), 6), 
 	      brw_imm_ud(1<<26)); 
 
-      jmp = brw_JMPI(p, ip, ip, brw_imm_w(0));
+      jmp = brw_JMPI(p, ip, ip, brw_imm_d(0));
       {
 	 emit_aa(c, arg1, 2);
 	 fire_fb_write(c, 0, nr, target, eot);
@@ -1222,6 +1255,10 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 emit_dph(p, dst, dst_flags, args[0], args[1]);
 	 break;
 
+      case OPCODE_TRUNC:
+	 emit_trunc(p, dst, dst_flags, args[0]);
+	 break;
+
       case OPCODE_LRP:
 	 emit_lrp(p, dst, dst_flags, args[0], args[1], args[2]);
 	 break;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index 49aad281d7..b9e8dd2e96 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -354,13 +354,25 @@ static void emit_interp( struct brw_wm_compile *c,
 		 src_undef());
       }
       else {
-	 emit_op(c,
-		 WM_LINTERP,
-		 dst,
-		 0,
-		 interp,
-		 deltas,
-		 src_undef());
+         if (c->key.linear_color) {
+            emit_op(c,
+                    WM_LINTERP,
+                    dst,
+                    0,
+                    interp,
+                    deltas,
+                    src_undef());
+         }
+         else {
+            /* perspective-corrected color interpolation */
+            emit_op(c,
+                    WM_PINTERP,
+                    dst,
+                    0,
+                    interp,
+                    deltas,
+                    get_pixel_w(c));
+         }
       }
       break;
    case FRAG_ATTRIB_FOGC:
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index a907e1b788..19f777fe32 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -1,5 +1,7 @@
 #include "main/macros.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "shader/prog_optimize.h"
 #include "brw_context.h"
 #include "brw_eu.h"
 #include "brw_wm.h"
@@ -21,7 +23,6 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 	const struct prog_instruction *inst = &fp->Base.Instructions[i];
 	switch (inst->Opcode) {
 	    case OPCODE_IF:
-	    case OPCODE_TRUNC:
 	    case OPCODE_ENDIF:
 	    case OPCODE_CAL:
 	    case OPCODE_BRK:
@@ -42,6 +43,83 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 }
 
 
+
+static void
+reclaim_temps(struct brw_wm_compile *c);
+
+
+/** Mark GRF register as used. */
+static void
+prealloc_grf(struct brw_wm_compile *c, int r)
+{
+   c->used_grf[r] = GL_TRUE;
+}
+
+
+/** Mark given GRF register as not in use. */
+static void
+release_grf(struct brw_wm_compile *c, int r)
+{
+   /*assert(c->used_grf[r]);*/
+   c->used_grf[r] = GL_FALSE;
+   c->first_free_grf = MIN2(c->first_free_grf, r);
+}
+
+
+/** Return index of a free GRF, mark it as used. */
+static int
+alloc_grf(struct brw_wm_compile *c)
+{
+   GLuint r;
+   for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) {
+      if (!c->used_grf[r]) {
+         c->used_grf[r] = GL_TRUE;
+         c->first_free_grf = r + 1;  /* a guess */
+         return r;
+      }
+   }
+
+   /* no free temps, try to reclaim some */
+   reclaim_temps(c);
+   c->first_free_grf = 0;
+
+   /* try alloc again */
+   for (r = c->first_free_grf; r < BRW_WM_MAX_GRF; r++) {
+      if (!c->used_grf[r]) {
+         c->used_grf[r] = GL_TRUE;
+         c->first_free_grf = r + 1;  /* a guess */
+         return r;
+      }
+   }
+
+   for (r = 0; r < BRW_WM_MAX_GRF; r++) {
+      assert(c->used_grf[r]);
+   }
+
+   /* really, no free GRF regs found */
+   if (!c->out_of_regs) {
+      /* print warning once per compilation */
+      _mesa_warning(NULL, "i965: ran out of registers for fragment program");
+      c->out_of_regs = GL_TRUE;
+   }
+
+   return -1;
+}
+
+
+/** Return number of GRF registers used */
+static int
+num_grf_used(const struct brw_wm_compile *c)
+{
+   int r;
+   for (r = BRW_WM_MAX_GRF - 1; r >= 0; r--)
+      if (c->used_grf[r])
+         return r + 1;
+   return 0;
+}
+
+
+
 /**
  * Record the mapping of a Mesa register to a hardware register.
  */
@@ -68,11 +146,23 @@ static int get_scalar_dst_index(const struct prog_instruction *inst)
 static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
 {
     struct brw_reg reg;
-    if(c->tmp_index == c->tmp_max)
-	c->tmp_regs[ c->tmp_max++ ] = c->reg_index++;
-    
+
+    /* if we need to allocate another temp, grow the tmp_regs[] array */
+    if (c->tmp_index == c->tmp_max) {
+       int r = alloc_grf(c);
+       if (r < 0) {
+          /*printf("Out of temps in %s\n", __FUNCTION__);*/
+          r = 50; /* XXX random register! */
+       }
+       c->tmp_regs[ c->tmp_max++ ] = r;
+    }
+
+    /* form the GRF register */
     reg = brw_vec8_grf(c->tmp_regs[ c->tmp_index++ ], 0);
+    /*printf("alloc_temp %d\n", reg.nr);*/
+    assert(reg.nr < BRW_WM_MAX_GRF);
     return reg;
+
 }
 
 /**
@@ -130,35 +220,29 @@ get_reg(struct brw_wm_compile *c, int file, int index, int component,
 	    return brw_null_reg();
     }
 
+    assert(index < 256);
+    assert(component < 4);
+
     /* see if we've already allocated a HW register for this Mesa register */
     if (c->wm_regs[file][index][component].inited) {
-	/* yes, re-use */
-	reg = c->wm_regs[file][index][component].reg;
+       /* yes, re-use */
+       reg = c->wm_regs[file][index][component].reg;
     }
     else {
 	/* no, allocate new register */
-	reg = brw_vec8_grf(c->reg_index, 0);
-    }
+       int grf = alloc_grf(c);
+       /*printf("alloc grf %d for reg %d:%d.%d\n", grf, file, index, component);*/
+       if (grf < 0) {
+          /* totally out of temps */
+          grf = 51; /* XXX random register! */
+       }
 
-    /* if this is a new register allocation, record it in the table */
-    if (!c->wm_regs[file][index][component].inited) {
-	set_reg(c, file, index, component, reg);
-	c->reg_index++;
-    }
+       reg = brw_vec8_grf(grf, 0);
+       /*printf("Alloc new grf %d for %d.%d\n", reg.nr, index, component);*/
 
-    if (c->reg_index >= BRW_WM_MAX_GRF - 12) {
-	/* ran out of temporary registers! */
-#if 1
-        /* This is a big hack for now.
-         * Return bad register index, just don't hang the GPU.
-         */
-        _mesa_fprintf(stderr, "out of regs %d\n", c->reg_index);
-        c->reg_index = BRW_WM_MAX_GRF - 13;
-#else
-	return brw_null_reg();
-#endif
+       set_reg(c, file, index, component, reg);
     }
- 
+
     if (neg & (1 << component)) {
 	reg = negate(reg);
     }
@@ -168,6 +252,46 @@ get_reg(struct brw_wm_compile *c, int file, int index, int component,
 }
 
 
+
+/**
+ * This is called if we run out of GRF registers.  Examine the live intervals
+ * of temp regs in the program and free those which won't be used again.
+ */
+static void
+reclaim_temps(struct brw_wm_compile *c)
+{
+   GLint intBegin[MAX_PROGRAM_TEMPS];
+   GLint intEnd[MAX_PROGRAM_TEMPS];
+   int index;
+
+   /*printf("Reclaim temps:\n");*/
+
+   _mesa_find_temp_intervals(c->prog_instructions, c->nr_fp_insns,
+                             intBegin, intEnd);
+
+   for (index = 0; index < MAX_PROGRAM_TEMPS; index++) {
+      if (intEnd[index] != -1 && intEnd[index] < c->cur_inst) {
+         /* program temp[i] can be freed */
+         int component;
+         /*printf("  temp[%d] is dead\n", index);*/
+         for (component = 0; component < 4; component++) {
+            if (c->wm_regs[PROGRAM_TEMPORARY][index][component].inited) {
+               int r = c->wm_regs[PROGRAM_TEMPORARY][index][component].reg.nr;
+               release_grf(c, r);
+               /*
+               printf("  Reclaim temp %d, reg %d at inst %d\n",
+                      index, r, c->cur_inst);
+               */
+               c->wm_regs[PROGRAM_TEMPORARY][index][component].inited = GL_FALSE;
+            }
+         }
+      }
+   }
+}
+
+
+
+
 /**
  * Preallocate registers.  This sets up the Mesa to hardware register
  * mapping for certain registers, such as constants (uniforms/state vars)
@@ -179,6 +303,10 @@ static void prealloc_reg(struct brw_wm_compile *c)
     struct brw_reg reg;
     int urb_read_length = 0;
     GLuint inputs = FRAG_BIT_WPOS | c->fp_interp_emitted | c->fp_deriv_emitted;
+    GLuint reg_index = 0;
+
+    memset(c->used_grf, GL_FALSE, sizeof(c->used_grf));
+    c->first_free_grf = 0;
 
     for (i = 0; i < 4; i++) {
         if (i < c->key.nr_depth_regs) 
@@ -187,14 +315,20 @@ static void prealloc_reg(struct brw_wm_compile *c)
             reg = brw_vec8_grf(0, 0);
 	set_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, reg);
     }
-    c->reg_index += 2 * c->key.nr_depth_regs;
+    reg_index += 2 * c->key.nr_depth_regs;
 
     /* constants */
     {
-        const int nr_params = c->fp->program.Base.Parameters->NumParameters;
+        const GLuint nr_params = c->fp->program.Base.Parameters->NumParameters;
+        const GLuint nr_temps = c->fp->program.Base.NumTemporaries;
 
         /* use a real constant buffer, or just use a section of the GRF? */
-        c->fp->use_const_buffer = GL_FALSE; /* (nr_params > 8);*/
+        /* XXX this heuristic may need adjustment... */
+        if ((nr_params + nr_temps) * 4 + reg_index > 80)
+           c->fp->use_const_buffer = GL_TRUE;
+        else
+           c->fp->use_const_buffer = GL_FALSE;
+        /*printf("WM use_const_buffer = %d\n", c->fp->use_const_buffer);*/
 
         if (c->fp->use_const_buffer) {
            /* We'll use a real constant buffer and fetch constants from
@@ -216,7 +350,7 @@ static void prealloc_reg(struct brw_wm_compile *c)
            for (i = 0; i < nr_params; i++) {
               /* loop over XYZW channels */
               for (j = 0; j < 4; j++, index++) {
-                 reg = brw_vec1_grf(c->reg_index + index / 8, index % 8);
+                 reg = brw_vec1_grf(reg_index + index / 8, index % 8);
                  /* Save pointer to parameter/constant value.
                   * Constants will be copied in prepare_constant_buffer()
                   */
@@ -226,7 +360,7 @@ static void prealloc_reg(struct brw_wm_compile *c)
            }
            /* number of constant regs used (each reg is float[8]) */
            c->nr_creg = 2 * ((4 * nr_params + 15) / 16);
-           c->reg_index += c->nr_creg;
+           reg_index += c->nr_creg;
         }
     }
 
@@ -242,23 +376,31 @@ static void prealloc_reg(struct brw_wm_compile *c)
 	  fp_input = -1;
 
        if (fp_input >= 0 && inputs & (1 << fp_input)) {
-	  urb_read_length = c->reg_index;
-	  reg = brw_vec8_grf(c->reg_index, 0);
+	  urb_read_length = reg_index;
+	  reg = brw_vec8_grf(reg_index, 0);
 	  for (j = 0; j < 4; j++)
 	     set_reg(c, PROGRAM_PAYLOAD, fp_input, j, reg);
        }
        if (c->key.vp_outputs_written & (1 << i)) {
-	  c->reg_index += 2;
+	  reg_index += 2;
        }
     }
 
     c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2;
     c->prog_data.urb_read_length = urb_read_length;
     c->prog_data.curb_read_length = c->nr_creg;
-    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
-    c->reg_index++;
-    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
-    c->reg_index += 2;
+    c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0);
+    reg_index++;
+    c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg_index, 0);
+    reg_index += 2;
+
+    /* mark GRF regs [0..reg_index-1] as in-use */
+    for (i = 0; i < reg_index; i++)
+       prealloc_grf(c, i);
+
+    /* Don't use GRF 126, 127.  Using them seems to lead to GPU lock-ups */
+    prealloc_grf(c, 126);
+    prealloc_grf(c, 127);
 
     /* An instruction may reference up to three constants.
      * They'll be found in these registers.
@@ -267,12 +409,12 @@ static void prealloc_reg(struct brw_wm_compile *c)
     if (c->fp->use_const_buffer) {
        for (i = 0; i < 3; i++) {
           c->current_const[i].index = -1;
-          c->current_const[i].reg = alloc_tmp(c);
+          c->current_const[i].reg = brw_vec8_grf(alloc_grf(c), 0);
        }
     }
 #if 0
     printf("USE CONST BUFFER? %d\n", c->fp->use_const_buffer);
-    printf("AFTER PRE_ALLOC, reg_index = %d\n", c->reg_index);
+    printf("AFTER PRE_ALLOC, reg_index = %d\n", reg_index);
 #endif
 }
 
@@ -376,6 +518,14 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c,
     const GLuint nr = 1;
     const GLuint component = GET_SWZ(src->Swizzle, channel);
 
+    /* Extended swizzle terms */
+    if (component == SWIZZLE_ZERO) {
+       return brw_imm_f(0.0F);
+    }
+    else if (component == SWIZZLE_ONE) {
+       return brw_imm_f(1.0F);
+    }
+
     if (c->fp->use_const_buffer &&
         (src->File == PROGRAM_STATE_VAR ||
          src->File == PROGRAM_CONSTANT ||
@@ -673,27 +823,26 @@ static void emit_fb_write(struct brw_wm_compile *c,
     }
 
     if (c->key.dest_depth_reg) {
-        GLuint comp = c->key.dest_depth_reg / 2;
-        GLuint off = c->key.dest_depth_reg % 2;
+        const GLuint comp = c->key.dest_depth_reg / 2;
+        const GLuint off = c->key.dest_depth_reg % 2;
 
-        assert(comp == 1);
-        assert(off == 0);
-#if 0
-        /* XXX do we need this code?   comp always 1, off always 0, it seems */
         if (off != 0) {
+            /* XXX this code needs review/testing */
+            struct brw_reg arg1_0 = get_src_reg(c, inst, 1, comp);
+            struct brw_reg arg1_1 = get_src_reg(c, inst, 1, comp+1);
+
             brw_push_insn_state(p);
             brw_set_compression_control(p, BRW_COMPRESSION_NONE);
 
-            brw_MOV(p, brw_message_reg(nr), offset(arg1[comp],1));
+            brw_MOV(p, brw_message_reg(nr), offset(arg1_0, 1));
             /* 2nd half? */
-            brw_MOV(p, brw_message_reg(nr+1), arg1[comp+1]);
+            brw_MOV(p, brw_message_reg(nr+1), arg1_1);
             brw_pop_insn_state(p);
         }
         else
-#endif
         {
-           struct brw_reg src =  get_src_reg(c, inst, 1, 1);
-           brw_MOV(p, brw_message_reg(nr), src);
+            struct brw_reg src =  get_src_reg(c, inst, 1, 1);
+            brw_MOV(p, brw_message_reg(nr), src);
         }
         nr += 2;
    }
@@ -2476,6 +2625,7 @@ static void emit_txb(struct brw_wm_compile *c,
     struct brw_reg dst[4], src[4], payload_reg;
     GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
     GLuint i;
+    GLuint msg_type;
 
     payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0);
 
@@ -2504,6 +2654,14 @@ static void emit_txb(struct brw_wm_compile *c,
     }
     brw_MOV(p, brw_message_reg(5), src[3]);          /* bias */
     brw_MOV(p, brw_message_reg(6), brw_imm_f(0));    /* ref (unused?) */
+
+    if (BRW_IS_IGDNG(p->brw)) {
+        msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG;
+    } else {
+        /* Does it work well on SIMD8? */
+        msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS;
+    }
+
     brw_SAMPLE(p,
                retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW),  /* dest */
                1,                                           /* msg_reg_nr */
@@ -2511,10 +2669,12 @@ static void emit_txb(struct brw_wm_compile *c,
                SURF_INDEX_TEXTURE(unit),
                unit,                                        /* sampler */
                inst->DstReg.WriteMask,                      /* writemask */
-               BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS,      /* msg_type */
+               msg_type,                                    /* msg_type */
                4,                                           /* response_length */
                4,                                           /* msg_length */
-               0);                                          /* eot */
+               0,                                           /* eot */
+               1,
+               BRW_SAMPLER_SIMD_MODE_SIMD8);	
 }
 
 
@@ -2528,6 +2688,7 @@ static void emit_tex(struct brw_wm_compile *c,
     GLuint i, nr;
     GLuint emit;
     GLboolean shadow = (c->key.shadowtex_mask & (1<<unit)) ? 1 : 0;
+    GLuint msg_type;
 
     payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0);
 
@@ -2568,6 +2729,16 @@ static void emit_tex(struct brw_wm_compile *c,
        brw_MOV(p, brw_message_reg(6), src[2]);        /* ref value / R coord */
     }
 
+    if (BRW_IS_IGDNG(p->brw)) {
+        if (shadow)
+            msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG;
+        else
+            msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG;
+    } else {
+        /* Does it work for shadow on SIMD8 ? */
+        msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE;
+    }
+    
     brw_SAMPLE(p,
                retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */
                1,                                          /* msg_reg_nr */
@@ -2575,10 +2746,12 @@ static void emit_tex(struct brw_wm_compile *c,
                SURF_INDEX_TEXTURE(unit),
                unit,                                       /* sampler */
                inst->DstReg.WriteMask,                     /* writemask */
-               BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE,           /* msg_type */
+               msg_type,                                   /* msg_type */
                4,                                          /* response_length */
                shadow ? 6 : 4,                             /* msg_length */
-               0);                                         /* eot */
+               0,                                          /* eot */
+               1,
+               BRW_SAMPLER_SIMD_MODE_SIMD8);	
 
     if (shadow)
 	brw_MOV(p, dst[3], brw_imm_f(1.0));
@@ -2595,15 +2768,15 @@ static void post_wm_emit( struct brw_wm_compile *c )
 
 static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 {
-#define MAX_IFSN 32
+#define MAX_IF_DEPTH 32
 #define MAX_LOOP_DEPTH 32
-    struct brw_instruction *if_inst[MAX_IFSN], *loop_inst[MAX_LOOP_DEPTH];
-    struct brw_instruction *inst0, *inst1;
-    int i, if_insn = 0, loop_insn = 0;
+    struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH];
+    GLuint i, if_depth = 0, loop_depth = 0;
     struct brw_compile *p = &c->func;
     struct brw_indirect stack_index = brw_indirect(0, 0);
 
-    c->reg_index = 0;
+    c->out_of_regs = GL_FALSE;
+
     prealloc_reg(c);
     brw_set_compression_control(p, BRW_COMPRESSION_NONE);
     brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack));
@@ -2611,6 +2784,8 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
     for (i = 0; i < c->nr_fp_insns; i++) {
         const struct prog_instruction *inst = &c->prog_instructions[i];
 
+        c->cur_inst = i;
+
 #if 0
         _mesa_printf("Inst %d: ", i);
         _mesa_print_instruction(inst);
@@ -2770,15 +2945,15 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		emit_kil(c);
 		break;
 	    case OPCODE_IF:
-		assert(if_insn < MAX_IFSN);
-		if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+		assert(if_depth < MAX_IF_DEPTH);
+		if_inst[if_depth++] = brw_IF(p, BRW_EXECUTE_8);
 		break;
 	    case OPCODE_ELSE:
-		if_inst[if_insn-1]  = brw_ELSE(p, if_inst[if_insn-1]);
+		if_inst[if_depth-1]  = brw_ELSE(p, if_inst[if_depth-1]);
 		break;
 	    case OPCODE_ENDIF:
-		assert(if_insn > 0);
-		brw_ENDIF(p, if_inst[--if_insn]);
+		assert(if_depth > 0);
+		brw_ENDIF(p, if_inst[--if_depth]);
 		break;
 	    case OPCODE_BGNSUB:
 		brw_save_label(p, inst->Comment, p->nr_insn);
@@ -2812,7 +2987,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		break;
 	    case OPCODE_BGNLOOP:
                 /* XXX may need to invalidate the current_constant regs */
-		loop_inst[loop_insn++] = brw_DO(p, BRW_EXECUTE_8);
+		loop_inst[loop_depth++] = brw_DO(p, BRW_EXECUTE_8);
 		break;
 	    case OPCODE_BRK:
 		brw_BREAK(p);
@@ -2823,36 +2998,40 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 		brw_set_predicate_control(p, BRW_PREDICATE_NONE);
 		break;
 	    case OPCODE_ENDLOOP: 
-		loop_insn--;
-		inst0 = inst1 = brw_WHILE(p, loop_inst[loop_insn]);
-		/* patch all the BREAK instructions from
-		   last BEGINLOOP */
-		while (inst0 > loop_inst[loop_insn]) {
-		    inst0--;
-		    if (inst0->header.opcode == BRW_OPCODE_BREAK) {
-			inst0->bits3.if_else.jump_count = inst1 - inst0 + 1;
+               {
+                  struct brw_instruction *inst0, *inst1;
+                  GLuint br = 1;
+
+                  if (BRW_IS_IGDNG(brw))
+                     br = 2;
+ 
+                  loop_depth--;
+                  inst0 = inst1 = brw_WHILE(p, loop_inst[loop_depth]);
+                  /* patch all the BREAK/CONT instructions from last BEGINLOOP */
+                  while (inst0 > loop_inst[loop_depth]) {
+                     inst0--;
+                     if (inst0->header.opcode == BRW_OPCODE_BREAK) {
+			inst0->bits3.if_else.jump_count = br * (inst1 - inst0 + 1);
 			inst0->bits3.if_else.pop_count = 0;
-		    } else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
-                        inst0->bits3.if_else.jump_count = inst1 - inst0;
+                     }
+                     else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) {
+                        inst0->bits3.if_else.jump_count = br * (inst1 - inst0);
                         inst0->bits3.if_else.pop_count = 0;
-                    }
-		}
-		break;
+                     }
+                  }
+               }
+               break;
 	    default:
 		_mesa_printf("unsupported IR in fragment shader %d\n",
 			inst->Opcode);
 	}
+
 	if (inst->CondUpdate)
 	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
 	else
 	    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
     }
     post_wm_emit(c);
-
-    if (c->reg_index >= BRW_WM_MAX_GRF) {
-        _mesa_problem(NULL, "Ran out of registers in brw_wm_emit_glsl()");
-        /* XXX we need to do some proper error recovery here */
-    }
 }
 
 
@@ -2876,6 +3055,6 @@ void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c)
         brw_wm_print_program(c, "brw_wm_glsl_emit done");
     }
 
-    c->prog_data.total_grf = c->reg_index;
+    c->prog_data.total_grf = num_grf_used(c);
     c->prog_data.total_scratch = 0;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_iz.c b/src/mesa/drivers/dri/i965/brw_wm_iz.c
index bd60ac9b31..8fd067abe7 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_iz.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_iz.c
@@ -116,6 +116,10 @@ const struct {
  { C, 0, 1, 1, 1 } 
 };
 
+/**
+ * \param line_aa  AA_NEVER, AA_ALWAYS or AA_SOMETIMES
+ * \param lookup  bitmask of IZ_* flags
+ */
 void brw_wm_lookup_iz( GLuint line_aa,
 		       GLuint lookup,
 		       struct brw_wm_prog_key *key )
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
index ab9aa2f10d..3436a24717 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
@@ -159,6 +159,7 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       case OPCODE_FRC:
       case OPCODE_MOV:
       case OPCODE_SWZ:
+      case OPCODE_TRUNC:
 	 read0 = writemask;
 	 break;
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index 67b41173fb..39f8c6d522 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -71,7 +71,9 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
       key->max_threads = 1;
    else {
       /* WM maximum threads is number of EUs times number of threads per EU. */
-      if (BRW_IS_G4X(brw))
+      if (BRW_IS_IGDNG(brw))
+         key->max_threads = 12 * 6;
+      else if (BRW_IS_G4X(brw))
 	 key->max_threads = 10 * 5;
       else
 	 key->max_threads = 8 * 4;
@@ -141,7 +143,11 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
    wm.thread0.kernel_start_pointer = brw->wm.prog_bo->offset >> 6; /* reloc */
    wm.thread1.depth_coef_urb_read_offset = 1;
    wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
-   wm.thread1.binding_table_entry_count = key->nr_surfaces;
+
+   if (BRW_IS_IGDNG(brw))
+      wm.thread1.binding_table_entry_count = 0; /* hardware requirement */
+   else
+      wm.thread1.binding_table_entry_count = key->nr_surfaces;
 
    if (key->total_scratch != 0) {
       wm.thread2.scratch_space_base_pointer =
@@ -158,7 +164,11 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
    wm.thread3.const_urb_entry_read_length = key->curb_entry_read_length;
    wm.thread3.const_urb_entry_read_offset = key->curbe_offset * 2;
 
-   wm.wm4.sampler_count = (key->sampler_count + 1) / 4;
+   if (BRW_IS_IGDNG(brw)) 
+      wm.wm4.sampler_count = 0; /* hardware requirement */
+   else
+      wm.wm4.sampler_count = (key->sampler_count + 1) / 4;
+
    if (brw->wm.sampler_bo != NULL) {
       /* reloc */
       wm.wm4.sampler_state_pointer = brw->wm.sampler_bo->offset >> 5;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index 805df8a4af..096f74394e 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -176,22 +176,6 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format,
    }
 }
 
-
-/**
- * Use same key for WM and VS surfaces.
- */
-struct brw_surface_key {
-   GLenum target, depthmode;
-   dri_bo *bo;
-   GLint format, internal_format;
-   GLint first_level, last_level;
-   GLint width, height, depth;
-   GLint pitch, cpp;
-   uint32_t tiling;
-   GLuint offset;
-};
-
-
 static void
 brw_set_surface_tiling(struct brw_surface_state *surf, uint32_t tiling)
 {
@@ -268,7 +252,7 @@ brw_create_texture_surface( struct brw_context *brw,
       surf.ss0.cube_neg_z = 1;
    }
 
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
 			 key, sizeof(*key),
 			 &key->bo, key->bo ? 1 : 0,
 			 &surf, sizeof(surf),
@@ -321,10 +305,11 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
    key.tiling = intelObj->mt->region->tiling;
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                         &key, sizeof(key),
-                                         &key.bo, key.bo ? 1 : 0,
-                                         NULL);
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
+                                            &key, sizeof(key),
+                                            &key.bo, key.bo ? 1 : 0,
+                                            NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_texture_surface(brw, &key);
    }
@@ -336,7 +321,7 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
  * Create the constant buffer surface.  Vertex/fragment shader constants will be
  * read from this buffer with Data Port Read instructions/messages.
  */
-static dri_bo *
+dri_bo *
 brw_create_constant_surface( struct brw_context *brw,
                              struct brw_surface_key *key )
 {
@@ -362,7 +347,7 @@ brw_create_constant_surface( struct brw_context *brw,
    surf.ss3.pitch = (key->pitch * key->cpp) - 1; /* ignored?? */
    brw_set_surface_tiling(&surf, key->tiling); /* tiling now allowed */
  
-   bo = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+   bo = brw_upload_cache(&brw->surface_cache, BRW_SS_SURFACE,
 			 key, sizeof(*key),
 			 &key->bo, key->bo ? 1 : 0,
 			 &surf, sizeof(surf),
@@ -380,39 +365,70 @@ brw_create_constant_surface( struct brw_context *brw,
    return bo;
 }
 
+/* Creates a new WM constant buffer reflecting the current fragment program's
+ * constants, if needed by the fragment program.
+ *
+ * Otherwise, constants go through the CURBEs using the brw_constant_buffer
+ * state atom.
+ */
+static drm_intel_bo *
+brw_wm_update_constant_buffer(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params = fp->program.Base.Parameters;
+   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   drm_intel_bo *const_buffer;
+
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   if (!fp->use_const_buffer)
+      return NULL;
+
+   const_buffer = drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer",
+				     size, 64);
+
+   /* _NEW_PROGRAM_CONSTANTS */
+   dri_bo_subdata(const_buffer, 0, size, params->ParameterValues);
+
+   return const_buffer;
+}
 
 /**
  * Update the surface state for a WM constant buffer.
  * The constant buffer will be (re)allocated here if needed.
  */
-static dri_bo *
+static void
 brw_update_wm_constant_surface( GLcontext *ctx,
-                                GLuint surf,
-                                dri_bo *const_buffer,
-                                const struct gl_program_parameter_list *params)
+                                GLuint surf)
 {
    struct brw_context *brw = brw_context(ctx);
    struct brw_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   const struct gl_program_parameter_list *params =
+      fp->program.Base.Parameters;
 
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
-   }
+   /* If we're in this state update atom, we need to update WM constants, so
+    * free the old buffer and create a new one for the new contents.
+    */
+   dri_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "fp_const_buffer", size, 64);
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+      brw->wm.surf_bo[surf] = NULL;
+      return;
    }
 
    memset(&key, 0, sizeof(key));
 
    key.format = MESA_FORMAT_RGBA_FLOAT32;
    key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
+   key.bo = fp->const_buffer;
    key.depthmode = GL_NONE;
    key.pitch = params->NumParameters;
    key.width = params->NumParameters;
@@ -427,77 +443,59 @@ brw_update_wm_constant_surface( GLcontext *ctx,
    */
 
    dri_bo_unreference(brw->wm.surf_bo[surf]);
-   brw->wm.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
+   brw->wm.surf_bo[surf] = brw_search_cache(&brw->surface_cache,
+                                            BRW_SS_SURFACE,
                                             &key, sizeof(key),
                                             &key.bo, key.bo ? 1 : 0,
                                             NULL);
    if (brw->wm.surf_bo[surf] == NULL) {
       brw->wm.surf_bo[surf] = brw_create_constant_surface(brw, &key);
    }
-
-   return const_buffer;
+   brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
 }
 
-
 /**
- * Update the surface state for a VS constant buffer.
- * The constant buffer will be (re)allocated here if needed.
+ * Updates surface / buffer for fragment shader constant buffer, if
+ * one is required.
+ *
+ * This consumes the state updates for the constant buffer, and produces
+ * BRW_NEW_WM_SURFACES to get picked up by brw_prepare_wm_surfaces for
+ * inclusion in the binding table.
  */
-static dri_bo *
-brw_update_vs_constant_surface( GLcontext *ctx,
-                                GLuint surf,
-                                dri_bo *const_buffer,
-                                const struct gl_program_parameter_list *params)
+static void prepare_wm_constant_surface(struct brw_context *brw )
 {
-   struct brw_context *brw = brw_context(ctx);
-   struct brw_surface_key key;
-   struct intel_context *intel = &brw->intel;
-   const int size = params->NumParameters * 4 * sizeof(GLfloat);
-
-   assert(surf == 0);
-
-   /* free old const buffer if too small */
-   if (const_buffer && const_buffer->size < size) {
-      dri_bo_unreference(const_buffer);
-      const_buffer = NULL;
-   }
-
-   /* alloc new buffer if needed */
-   if (!const_buffer) {
-      const_buffer =
-         drm_intel_bo_alloc(intel->bufmgr, "vp_const_buffer", size, 64);
-   }
-
-   memset(&key, 0, sizeof(key));
-
-   key.format = MESA_FORMAT_RGBA_FLOAT32;
-   key.internal_format = GL_RGBA;
-   key.bo = const_buffer;
-   key.depthmode = GL_NONE;
-   key.pitch = params->NumParameters;
-   key.width = params->NumParameters;
-   key.height = 1;
-   key.depth = 1;
-   key.cpp = 16;
+   GLcontext *ctx = &brw->intel.ctx;
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *) brw->fragment_program;
+   GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
 
-   /*
-   printf("%s:\n", __FUNCTION__);
-   printf("  width %d  height %d  depth %d  cpp %d  pitch %d\n",
-          key.width, key.height, key.depth, key.cpp, key.pitch);
-   */
+   drm_intel_bo_unreference(fp->const_buffer);
+   fp->const_buffer = brw_wm_update_constant_buffer(brw);
 
-   dri_bo_unreference(brw->vs.surf_bo[surf]);
-   brw->vs.surf_bo[surf] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-                                            &key, sizeof(key),
-                                            &key.bo, key.bo ? 1 : 0,
-                                            NULL);
-   if (brw->vs.surf_bo[surf] == NULL) {
-      brw->vs.surf_bo[surf] = brw_create_constant_surface(brw, &key);
+   /* If there's no constant buffer, then no surface BO is needed to point at
+    * it.
+    */
+   if (fp->const_buffer == 0) {
+      if (brw->wm.surf_bo[surf] != NULL) {
+	 drm_intel_bo_unreference(brw->wm.surf_bo[surf]);
+	 brw->wm.surf_bo[surf] = NULL;
+	 brw->state.dirty.brw |= BRW_NEW_WM_SURFACES;
+      }
+      return;
    }
 
-   return const_buffer;
+   brw_update_wm_constant_surface(ctx, surf);
 }
 
+const struct brw_tracked_state brw_wm_constant_surface = {
+   .dirty = {
+      .mesa = (_NEW_PROGRAM_CONSTANTS),
+      .brw = (BRW_NEW_FRAGMENT_PROGRAM),
+      .cache = 0
+   },
+   .prepare = prepare_wm_constant_surface,
+};
+
 
 /**
  * Sets up a surface state structure to point at the given region.
@@ -507,7 +505,7 @@ brw_update_vs_constant_surface( GLcontext *ctx,
 static void
 brw_update_renderbuffer_surface(struct brw_context *brw,
 				struct gl_renderbuffer *rb,
-				unsigned int unit, GLboolean cached)
+				unsigned int unit)
 {
    GLcontext *ctx = &brw->intel.ctx;
    dri_bo *region_bo = NULL;
@@ -567,12 +565,11 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 		      ctx->Color.BlendEnabled);
 
    dri_bo_unreference(brw->wm.surf_bo[unit]);
-   brw->wm.surf_bo[unit] = NULL;
-   if (cached) 
-       brw->wm.surf_bo[unit] = brw_search_cache(&brw->cache, BRW_SS_SURFACE,
-	       &key, sizeof(key),
-	       &region_bo, 1,
-	       NULL);
+   brw->wm.surf_bo[unit] = brw_search_cache(&brw->surface_cache,
+					    BRW_SS_SURFACE,
+					    &key, sizeof(key),
+					    &region_bo, 1,
+					    NULL);
 
    if (brw->wm.surf_bo[unit] == NULL) {
       struct brw_surface_state surf;
@@ -581,7 +578,27 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 
       surf.ss0.surface_format = key.surface_format;
       surf.ss0.surface_type = key.surface_type;
-      surf.ss1.base_addr =  key.draw_offset;
+      if (key.tiling == I915_TILING_NONE) {
+	 surf.ss1.base_addr = key.draw_offset;
+      } else {
+	 uint32_t tile_offset = key.draw_offset % 4096;
+
+	 surf.ss1.base_addr = key.draw_offset - tile_offset;
+
+	 assert(BRW_IS_G4X(brw) || tile_offset == 0);
+	 if (BRW_IS_G4X(brw)) {
+	    if (key.tiling == I915_TILING_X) {
+	       /* Note that the low bits of these fields are missing, so
+		* there's the possibility of getting in trouble.
+		*/
+	       surf.ss5.x_offset = (tile_offset % 512) / key.cpp / 4;
+	       surf.ss5.y_offset = tile_offset / 512 / 2;
+	    } else {
+	       surf.ss5.x_offset = (tile_offset % 128) / key.cpp / 4;
+	       surf.ss5.y_offset = tile_offset / 128 / 2;
+	    }
+	 }
+      }
       if (region_bo != NULL)
 	 surf.ss1.base_addr += region_bo->offset; /* reloc */
 
@@ -598,7 +615,8 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
       surf.ss0.writedisable_alpha = !key.color_mask[3];
 
       /* Key size will never match key size for textures, so we're safe. */
-      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->cache, BRW_SS_SURFACE,
+      brw->wm.surf_bo[unit] = brw_upload_cache(&brw->surface_cache,
+                                               BRW_SS_SURFACE,
                                                &key, sizeof(key),
 					       &region_bo, 1,
 					       &surf, sizeof(surf),
@@ -611,7 +629,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw,
 	 drm_intel_bo_emit_reloc(brw->wm.surf_bo[unit],
 				 offsetof(struct brw_surface_state, ss1),
 				 region_bo,
-				 key.draw_offset,
+				 surf.ss1.base_addr,
 				 I915_GEM_DOMAIN_RENDER,
 				 I915_GEM_DOMAIN_RENDER);
       }
@@ -630,7 +648,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
 
    assert(brw->wm.nr_surfaces <= BRW_WM_MAX_SURF);
 
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
+   bind_bo = brw_search_cache(&brw->surface_cache, BRW_SS_SURF_BIND,
 			      NULL, 0,
 			      brw->wm.surf_bo, brw->wm.nr_surfaces,
 			      NULL);
@@ -646,7 +664,7 @@ brw_wm_get_binding_table(struct brw_context *brw)
          else
             data[i] = 0;
 
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
+      bind_bo = brw_upload_cache( &brw->surface_cache, BRW_SS_SURF_BIND,
 				  NULL, 0,
 				  brw->wm.surf_bo, brw->wm.nr_surfaces,
 				  data, data_size,
@@ -682,27 +700,17 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
          brw_update_renderbuffer_surface(brw,
 					 ctx->DrawBuffer->_ColorDrawBuffers[i],
-					 i,
-					 GL_FALSE);
+					 i);
       }
    } else {
-      brw_update_renderbuffer_surface(brw, NULL, 0, GL_TRUE);
+      brw_update_renderbuffer_surface(brw, NULL, 0);
    }
 
    old_nr_surfaces = brw->wm.nr_surfaces;
    brw->wm.nr_surfaces = MAX_DRAW_BUFFERS;
 
-   /* Update surface / buffer for fragment shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_FRAG_CONST_BUFFER;
-      struct brw_fragment_program *fp =
-         (struct brw_fragment_program *) brw->fragment_program;
-      fp->const_buffer =
-         brw_update_wm_constant_surface(ctx, surf, fp->const_buffer,
-                                     fp->program.Base.Parameters);
-
-      brw->wm.nr_surfaces = surf + 1;
-   }
+   if (brw->wm.surf_bo[SURF_INDEX_FRAG_CONST_BUFFER] != NULL)
+       brw->wm.nr_surfaces = SURF_INDEX_FRAG_CONST_BUFFER + 1;
 
    /* Update surfaces for textures */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
@@ -735,100 +743,16 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       brw->state.dirty.brw |= BRW_NEW_NR_WM_SURFACES;
 }
 
-
-/**
- * Constructs the binding table for the VS surface state.
- */
-static dri_bo *
-brw_vs_get_binding_table(struct brw_context *brw)
-{
-   dri_bo *bind_bo;
-
-   assert(brw->vs.nr_surfaces <= BRW_VS_MAX_SURF);
-
-   bind_bo = brw_search_cache(&brw->cache, BRW_SS_SURF_BIND,
-			      NULL, 0,
-			      brw->vs.surf_bo, brw->vs.nr_surfaces,
-			      NULL);
-
-   if (bind_bo == NULL) {
-      GLuint data_size = brw->vs.nr_surfaces * sizeof(GLuint);
-      uint32_t *data = malloc(data_size);
-      int i;
-
-      for (i = 0; i < brw->vs.nr_surfaces; i++)
-         if (brw->vs.surf_bo[i])
-            data[i] = brw->vs.surf_bo[i]->offset;
-         else
-            data[i] = 0;
-
-      bind_bo = brw_upload_cache( &brw->cache, BRW_SS_SURF_BIND,
-				  NULL, 0,
-				  brw->vs.surf_bo, brw->vs.nr_surfaces,
-				  data, data_size,
-				  NULL, NULL);
-
-      /* Emit binding table relocations to surface state */
-      for (i = 0; i < BRW_VS_MAX_SURF; i++) {
-	 if (brw->vs.surf_bo[i] != NULL) {
-	    dri_bo_emit_reloc(bind_bo,
-			      I915_GEM_DOMAIN_INSTRUCTION, 0,
-			      0,
-			      i * sizeof(GLuint),
-			      brw->vs.surf_bo[i]);
-	 }
-      }
-
-      free(data);
-   }
-
-   return bind_bo;
-}
-
-
-/**
- * Vertex shader surfaces.  Just constant buffer for now.  Could add vertex 
- * shader textures in the future.
- */
-static void prepare_vs_surfaces(struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   /* Update surface / buffer for vertex shader constant buffer */
-   {
-      const GLuint surf = SURF_INDEX_VERT_CONST_BUFFER;
-      struct brw_vertex_program *vp =
-         (struct brw_vertex_program *) brw->vertex_program;
-      vp->const_buffer =
-         brw_update_vs_constant_surface(ctx, surf, vp->const_buffer,
-                                        vp->program.Base.Parameters);
-
-      brw->vs.nr_surfaces = 1;
-   }
-
-   dri_bo_unreference(brw->vs.bind_bo);
-   brw->vs.bind_bo = brw_vs_get_binding_table(brw);
-
-   if (1)
-      brw->state.dirty.brw |= BRW_NEW_NR_VS_SURFACES;
-}
-
-
-static void
-prepare_surfaces(struct brw_context *brw)
-{
-   prepare_wm_surfaces(brw);
-   prepare_vs_surfaces(brw);
-}
-
-
 const struct brw_tracked_state brw_wm_surfaces = {
    .dirty = {
-      .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS | _NEW_PROGRAM,
-      .brw = BRW_NEW_CONTEXT,
+      .mesa = (_NEW_COLOR |
+               _NEW_TEXTURE |
+               _NEW_BUFFERS),
+      .brw = (BRW_NEW_CONTEXT |
+	      BRW_NEW_WM_SURFACES),
       .cache = 0
    },
-   .prepare = prepare_surfaces,
+   .prepare = prepare_wm_surfaces,
 };
 
 
diff --git a/src/mesa/drivers/dri/i965/intel_generatemipmap.c b/src/mesa/drivers/dri/i965/intel_generatemipmap.c
new file mode 120000
index 0000000000..4c6b37ada0
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_generatemipmap.c
@@ -0,0 +1 @@
+../intel/intel_generatemipmap.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/intel_pixel_read.c b/src/mesa/drivers/dri/i965/intel_pixel_read.c
new file mode 120000
index 0000000000..cc4589f4d4
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_pixel_read.c
@@ -0,0 +1 @@
+../intel/intel_pixel_read.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index 29dc05c518..0f87fc46a4 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -195,7 +195,6 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
 {
    struct intel_context *intel = batch->intel;
    GLuint used = batch->ptr - batch->map;
-   GLboolean was_locked = intel->locked;
 
    if (used == 0) {
       batch->cliprect_mode = IGNORE_CLIPRECTS;
@@ -243,13 +242,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
    /* TODO: Just pass the relocation list and dma buffer up to the
     * kernel.
     */
-   if (!was_locked)
-      LOCK_HARDWARE(intel);
-
+   LOCK_HARDWARE(intel);
    do_flush_locked(batch, used, GL_FALSE);
-
-   if (!was_locked)
-      UNLOCK_HARDWARE(intel);
+   UNLOCK_HARDWARE(intel);
 
    if (INTEL_DEBUG & DEBUG_SYNC) {
       fprintf(stderr, "waiting for idle\n");
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index 4919828131..2e95bd1013 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -108,6 +108,8 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
 	 CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
       }
 
+      assert(src->tiling != I915_TILING_Y);
+      assert(dst->tiling != I915_TILING_Y);
 #ifndef I915
       if (src->tiling != I915_TILING_NONE) {
 	 CMD |= XY_SRC_TILED;
@@ -175,66 +177,6 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
    UNLOCK_HARDWARE(intel);
 }
 
-
-
-
-void
-intelEmitFillBlit(struct intel_context *intel,
-		  GLuint cpp,
-		  GLshort dst_pitch,
-		  dri_bo *dst_buffer,
-		  GLuint dst_offset,
-		  uint32_t dst_tiling,
-		  GLshort x, GLshort y,
-		  GLshort w, GLshort h,
-		  GLuint color)
-{
-   GLuint BR13, CMD;
-   BATCH_LOCALS;
-
-   dst_pitch *= cpp;
-
-   switch (cpp) {
-   case 1:
-      BR13 = (0xF0 << 16);
-      CMD = XY_COLOR_BLT_CMD;
-      break;
-   case 2:
-      BR13 = (0xF0 << 16) | BR13_565;
-      CMD = XY_COLOR_BLT_CMD;
-      break;
-   case 4:
-      BR13 = (0xF0 << 16) | BR13_8888;
-      CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
-      break;
-   default:
-      return;
-   }
-#ifndef I915
-   if (dst_tiling != I915_TILING_NONE) {
-      CMD |= XY_DST_TILED;
-      dst_pitch /= 4;
-   }
-#endif
-
-   DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
-       __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h);
-
-   assert(w > 0);
-   assert(h > 0);
-
-   BEGIN_BATCH(6, NO_LOOP_CLIPRECTS);
-   OUT_BATCH(CMD);
-   OUT_BATCH(BR13 | dst_pitch);
-   OUT_BATCH((y << 16) | x);
-   OUT_BATCH(((y + h) << 16) | (x + w));
-   OUT_RELOC(dst_buffer,
-	     I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-	     dst_offset);
-   OUT_BATCH(color);
-   ADVANCE_BATCH();
-}
-
 static GLuint translate_raster_op(GLenum logicop)
 {
    switch(logicop) {
@@ -261,7 +203,7 @@ static GLuint translate_raster_op(GLenum logicop)
 
 /* Copy BitBlt
  */
-void
+GLboolean
 intelEmitCopyBlit(struct intel_context *intel,
 		  GLuint cpp,
 		  GLshort src_pitch,
@@ -283,6 +225,19 @@ intelEmitCopyBlit(struct intel_context *intel,
    dri_bo *aper_array[3];
    BATCH_LOCALS;
 
+   if (dst_tiling != I915_TILING_NONE) {
+      if (dst_offset & 4095)
+	 return GL_FALSE;
+      if (dst_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+   if (src_tiling != I915_TILING_NONE) {
+      if (src_offset & 4095)
+	 return GL_FALSE;
+      if (src_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+
    /* do space/cliprects check before going any further */
    do {
        aper_array[0] = intel->batch->buf;
@@ -297,12 +252,7 @@ intelEmitCopyBlit(struct intel_context *intel,
    } while (pass < 2);
 
    if (pass >= 2) {
-       GLboolean locked = GL_FALSE;       
-       if (!intel->locked) {
-           LOCK_HARDWARE(intel);
-           locked = GL_TRUE;
-       }
-
+       LOCK_HARDWARE(intel);
        dri_bo_map(dst_buffer, GL_TRUE);
        dri_bo_map(src_buffer, GL_FALSE);
        _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset,
@@ -316,11 +266,9 @@ intelEmitCopyBlit(struct intel_context *intel,
        
        dri_bo_unmap(src_buffer);
        dri_bo_unmap(dst_buffer);
-       
-       if (locked)
-           UNLOCK_HARDWARE(intel);
+       UNLOCK_HARDWARE(intel);
 
-       return;
+       return GL_TRUE;
    }
 
    intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS);
@@ -347,7 +295,7 @@ intelEmitCopyBlit(struct intel_context *intel,
       CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
       break;
    default:
-      return;
+      return GL_FALSE;
    }
 
 #ifndef I915
@@ -362,7 +310,7 @@ intelEmitCopyBlit(struct intel_context *intel,
 #endif
 
    if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
-      return;
+      return GL_TRUE;
    }
 
    assert(dst_x < dst_x2);
@@ -384,6 +332,8 @@ intelEmitCopyBlit(struct intel_context *intel,
    ADVANCE_BATCH();
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   return GL_TRUE;
 }
 
 
@@ -596,7 +546,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
    UNLOCK_HARDWARE(intel);
 }
 
-void
+GLboolean
 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLuint cpp,
 				  GLubyte *src_bits, GLuint src_size,
@@ -612,11 +562,18 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
    int dwords = ALIGN(src_size, 8) / 4;
    uint32_t opcode, br13, blit_cmd;
 
+   if (dst_tiling != I915_TILING_NONE) {
+      if (dst_offset & 4095)
+	 return GL_FALSE;
+      if (dst_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+
    assert( logic_op - GL_CLEAR >= 0 );
    assert( logic_op - GL_CLEAR < 0x10 );
 
    if (w < 0 || h < 0)
-      return;
+      return GL_TRUE;
 
    dst_pitch *= cpp;
 
@@ -673,4 +630,6 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 			   REFERENCES_CLIPRECTS );
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_blit.h b/src/mesa/drivers/dri/intel/intel_blit.h
index 52065b13ed..152fa3f17b 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.h
+++ b/src/mesa/drivers/dri/intel/intel_blit.h
@@ -35,7 +35,8 @@ extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv,
 
 extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask);
 
-extern void intelEmitCopyBlit(struct intel_context *intel,
+GLboolean
+intelEmitCopyBlit(struct intel_context *intel,
                               GLuint cpp,
                               GLshort src_pitch,
                               dri_bo *src_buffer,
@@ -50,16 +51,7 @@ extern void intelEmitCopyBlit(struct intel_context *intel,
                               GLshort w, GLshort h,
 			      GLenum logicop );
 
-extern void intelEmitFillBlit(struct intel_context *intel,
-                              GLuint cpp,
-                              GLshort dst_pitch,
-                              dri_bo *dst_buffer,
-                              GLuint dst_offset,
-			      uint32_t dst_tiling,
-                              GLshort x, GLshort y,
-                              GLshort w, GLshort h, GLuint color);
-
-void
+GLboolean
 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLuint cpp,
 				  GLubyte *src_bits, GLuint src_size,
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index 2e6b77824d..9600557f2c 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -35,6 +35,9 @@
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
 
+static GLboolean
+intel_bufferobj_unmap(GLcontext * ctx,
+                      GLenum target, struct gl_buffer_object *obj);
 
 /** Allocates a new dri_bo to store the data for the buffer object. */
 static void
@@ -100,7 +103,13 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
 
    assert(intel_obj);
-   assert(!obj->Pointer); /* Mesa should have unmapped it */
+
+   /* Buffer objects are automatically unmapped when deleting according
+    * to the spec, but Mesa doesn't do UnmapBuffer for us at context destroy
+    * (though it does if you call glDeleteBuffers)
+    */
+   if (obj->Pointer)
+      intel_bufferobj_unmap(ctx, 0, obj);
 
    _mesa_free(intel_obj->sys_buffer);
    if (intel_obj->region) {
@@ -234,6 +243,9 @@ intel_bufferobj_map(GLcontext * ctx,
       return obj->Pointer;
    }
 
+   if (!read_only)
+      intelFlush(ctx);
+
    if (intel_obj->region)
       intel_bufferobj_cow(intel, intel_obj);
 
@@ -294,14 +306,19 @@ intel_bufferobj_buffer(struct intel_context *intel,
    }
 
    if (intel_obj->buffer == NULL) {
+      void *sys_buffer = intel_obj->sys_buffer;
+
+      /* only one of buffer and sys_buffer could be non-NULL */
       intel_bufferobj_alloc_buffer(intel, intel_obj);
+      intel_obj->sys_buffer = NULL;
+
       intel_bufferobj_subdata(&intel->ctx,
 			      GL_ARRAY_BUFFER_ARB,
 			      0,
 			      intel_obj->Base.Size,
-			      intel_obj->sys_buffer,
+			      sys_buffer,
 			      &intel_obj->Base);
-      _mesa_free(intel_obj->sys_buffer);
+      _mesa_free(sys_buffer);
       intel_obj->sys_buffer = NULL;
    }
 
@@ -309,15 +326,13 @@ intel_bufferobj_buffer(struct intel_context *intel,
 }
 
 void
-intel_bufferobj_init(struct intel_context *intel)
+intelInitBufferObjectFuncs(struct dd_function_table *functions)
 {
-   GLcontext *ctx = &intel->ctx;
-
-   ctx->Driver.NewBufferObject = intel_bufferobj_alloc;
-   ctx->Driver.DeleteBuffer = intel_bufferobj_free;
-   ctx->Driver.BufferData = intel_bufferobj_data;
-   ctx->Driver.BufferSubData = intel_bufferobj_subdata;
-   ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata;
-   ctx->Driver.MapBuffer = intel_bufferobj_map;
-   ctx->Driver.UnmapBuffer = intel_bufferobj_unmap;
+   functions->NewBufferObject = intel_bufferobj_alloc;
+   functions->DeleteBuffer = intel_bufferobj_free;
+   functions->BufferData = intel_bufferobj_data;
+   functions->BufferSubData = intel_bufferobj_subdata;
+   functions->GetBufferSubData = intel_bufferobj_get_subdata;
+   functions->MapBuffer = intel_bufferobj_map;
+   functions->UnmapBuffer = intel_bufferobj_unmap;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.h b/src/mesa/drivers/dri/intel/intel_buffer_objects.h
index 0431015631..8164407f07 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.h
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.h
@@ -60,7 +60,7 @@ dri_bo *intel_bufferobj_buffer(struct intel_context *intel,
 
 /* Hook the bufferobject implementation into mesa: 
  */
-void intel_bufferobj_init(struct intel_context *intel);
+void intelInitBufferObjectFuncs(struct dd_function_table *functions);
 
 
 
@@ -72,10 +72,7 @@ void intel_bufferobj_init(struct intel_context *intel);
 static INLINE struct intel_buffer_object *
 intel_buffer_object(struct gl_buffer_object *obj)
 {
-   if (obj->Name)
-      return (struct intel_buffer_object *) obj;
-   else
-      return NULL;
+   return (struct intel_buffer_object *) obj;
 }
 
 /* Helpers for zerocopy image uploads.  See also intel_regions.h:
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index d2fad9e4ea..e7357e78c5 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -157,7 +157,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    /* Do this here, not core Mesa, since this function is called from
     * many places within the driver.
     */
-   if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+   if (ctx->NewState & _NEW_BUFFERS) {
       /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */
       _mesa_update_framebuffer(ctx);
       /* this updates the DrawBuffer's Width/Height if it's a FBO */
@@ -345,6 +345,23 @@ intelDrawBuffer(GLcontext * ctx, GLenum mode)
 static void
 intelReadBuffer(GLcontext * ctx, GLenum mode)
 {
+   if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) {
+      struct intel_context *const intel = intel_context(ctx);
+      const GLboolean was_front_buffer_reading =
+	intel->is_front_buffer_reading;
+
+      intel->is_front_buffer_reading = (mode == GL_FRONT_LEFT)
+	|| (mode == GL_FRONT);
+
+      /* If we weren't front-buffer reading before but we are now, make sure
+       * that the front-buffer has actually been allocated.
+       */
+      if (!was_front_buffer_reading && intel->is_front_buffer_reading) {
+	 intel_update_renderbuffers(intel->driContext,
+				    intel->driContext->driDrawablePriv);
+      }
+   }
+
    if (ctx->ReadBuffer == ctx->DrawBuffer) {
       /* This will update FBO completeness status.
        * A framebuffer will be incomplete if the GL_READ_BUFFER setting
diff --git a/src/mesa/drivers/dri/intel/intel_chipset.h b/src/mesa/drivers/dri/intel/intel_chipset.h
index 4593d90df3..3c38f1676c 100644
--- a/src/mesa/drivers/dri/intel/intel_chipset.h
+++ b/src/mesa/drivers/dri/intel/intel_chipset.h
@@ -67,13 +67,18 @@
 #define PCI_CHIP_G45_G                  0x2E22
 #define PCI_CHIP_G41_G                  0x2E32
 
+#define PCI_CHIP_ILD_G                  0x0042
+#define PCI_CHIP_ILM_G                  0x0046
+
 #define IS_MOBILE(devid)	(devid == PCI_CHIP_I855_GM || \
 				 devid == PCI_CHIP_I915_GM || \
 				 devid == PCI_CHIP_I945_GM || \
 				 devid == PCI_CHIP_I945_GME || \
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
-				 devid == PCI_CHIP_GM45_GM || IS_IGD(devid))
+				 devid == PCI_CHIP_GM45_GM || \
+				 IS_IGD(devid) || \
+				 devid == PCI_CHIP_ILM_G)
 
 #define IS_G45(devid)           (devid == PCI_CHIP_IGD_E_G || \
                                  devid == PCI_CHIP_Q45_G || \
@@ -82,6 +87,10 @@
 #define IS_GM45(devid)          (devid == PCI_CHIP_GM45_GM)
 #define IS_G4X(devid)		(IS_G45(devid) || IS_GM45(devid))
 
+#define IS_ILD(devid)           (devid == PCI_CHIP_ILD_G)
+#define IS_ILM(devid)           (devid == PCI_CHIP_ILM_G)
+#define IS_IGDNG(devid)           (IS_ILD(devid) || IS_ILM(devid))
+
 #define IS_915(devid)		(devid == PCI_CHIP_I915_G || \
 				 devid == PCI_CHIP_E7221_G || \
 				 devid == PCI_CHIP_I915_GM)
@@ -99,7 +108,8 @@
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
 				 devid == PCI_CHIP_I946_GZ || \
-				 IS_G4X(devid))
+				 IS_G4X(devid) || \
+				 IS_IGDNG(devid))
 
 #define IS_9XX(devid)		(IS_915(devid) || \
 				 IS_945(devid) || \
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
index 19f47632ac..cfddabd318 100644
--- a/src/mesa/drivers/dri/intel/intel_clear.c
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -53,240 +53,10 @@
 #include "intel_clear.h"
 #include "intel_fbo.h"
 #include "intel_pixel.h"
+#include "intel_regions.h"
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
 
-#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
-			      BUFFER_BIT_FRONT_LEFT |			\
-			      BUFFER_BIT_COLOR0 |			\
-			      BUFFER_BIT_COLOR1 |			\
-			      BUFFER_BIT_COLOR2 |			\
-			      BUFFER_BIT_COLOR3 |			\
-			      BUFFER_BIT_COLOR4 |			\
-			      BUFFER_BIT_COLOR5 |			\
-			      BUFFER_BIT_COLOR6 |			\
-			      BUFFER_BIT_COLOR7)
-
-
-/**
- * Per-context one-time init of things for intl_clear_tris().
- * Basically set up a private array object for vertex/color arrays.
- */
-static void
-init_clear(GLcontext *ctx)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct gl_array_object *arraySave = NULL;
-   const GLuint arrayBuffer = ctx->Array.ArrayBufferObj->Name;
-   const GLuint elementBuffer = ctx->Array.ElementArrayBufferObj->Name;
-
-   /* create new array object */
-   intel->clear.arrayObj = _mesa_new_array_object(ctx, ~0);
-
-   /* save current array object, bind new one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj);
-
-   /* one-time setup of vertex arrays (pos, color) */
-   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
-   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, 0);
-   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), intel->clear.color);
-   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), intel->clear.vertices);
-   _mesa_Enable(GL_COLOR_ARRAY);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-
-   /* restore original array object */
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
-
-   /* restore original buffer objects */
-   _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, arrayBuffer);
-   _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuffer);
-}
-
-
-
-/**
- * Perform glClear where mask contains only color, depth, and/or stencil.
- *
- * The implementation is based on calling into Mesa to set GL state and
- * performing normal triangle rendering.  The intent of this path is to
- * have as generic a path as possible, so that any driver could make use of
- * it.
- */
-void
-intel_clear_tris(GLcontext *ctx, GLbitfield mask)
-{
-   struct intel_context *intel = intel_context(ctx);
-   GLfloat dst_z;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   int i;
-   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
-   GLuint saved_shader_program = 0;
-   unsigned int saved_active_texture;
-   struct gl_array_object *arraySave = NULL;
-
-   if (!intel->clear.arrayObj)
-      init_clear(ctx);
-
-   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
-		    BUFFER_BIT_STENCIL)) == 0);
-
-   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
-		    GL_CURRENT_BIT |
-		    GL_DEPTH_BUFFER_BIT |
-		    GL_ENABLE_BIT |
-		    GL_POLYGON_BIT |
-		    GL_STENCIL_BUFFER_BIT |
-		    GL_TRANSFORM_BIT |
-		    GL_CURRENT_BIT);
-   saved_active_texture = ctx->Texture.CurrentUnit;
-
-   /* Disable existing GL state we don't want to apply to a clear. */
-   _mesa_Disable(GL_ALPHA_TEST);
-   _mesa_Disable(GL_BLEND);
-   _mesa_Disable(GL_CULL_FACE);
-   _mesa_Disable(GL_FOG);
-   _mesa_Disable(GL_POLYGON_SMOOTH);
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
-   _mesa_Disable(GL_LIGHTING);
-   _mesa_Disable(GL_CLIP_PLANE0);
-   _mesa_Disable(GL_CLIP_PLANE1);
-   _mesa_Disable(GL_CLIP_PLANE2);
-   _mesa_Disable(GL_CLIP_PLANE3);
-   _mesa_Disable(GL_CLIP_PLANE4);
-   _mesa_Disable(GL_CLIP_PLANE5);
-   _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL);
-   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
-      saved_fp_enable = GL_TRUE;
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
-      saved_vp_enable = GL_TRUE;
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
-      saved_shader_program = ctx->Shader.CurrentProgram->Name;
-      _mesa_UseProgramObjectARB(0);
-   }
-
-   if (ctx->Texture._EnabledUnits != 0) {
-      int i;
-
-      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
-	 _mesa_Disable(GL_TEXTURE_1D);
-	 _mesa_Disable(GL_TEXTURE_2D);
-	 _mesa_Disable(GL_TEXTURE_3D);
-	 if (ctx->Extensions.ARB_texture_cube_map)
-	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
-	 if (ctx->Extensions.NV_texture_rectangle)
-	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
-	 if (ctx->Extensions.MESA_texture_array) {
-	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
-	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
-	 }
-      }
-   }
-
-   /* save current array object, bind our private one */
-   _mesa_reference_array_object(ctx, &arraySave, ctx->Array.ArrayObj);
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, intel->clear.arrayObj);
-
-   intel_meta_set_passthrough_transform(intel);
-
-   for (i = 0; i < 4; i++) {
-      COPY_4FV(intel->clear.color[i], ctx->Color.ClearColor);
-   }
-
-   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
-   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
-
-   /* Prepare the vertices, which are the same regardless of which buffer we're
-    * drawing to.
-    */
-   intel->clear.vertices[0][0] = fb->_Xmin;
-   intel->clear.vertices[0][1] = fb->_Ymin;
-   intel->clear.vertices[0][2] = dst_z;
-   intel->clear.vertices[1][0] = fb->_Xmax;
-   intel->clear.vertices[1][1] = fb->_Ymin;
-   intel->clear.vertices[1][2] = dst_z;
-   intel->clear.vertices[2][0] = fb->_Xmax;
-   intel->clear.vertices[2][1] = fb->_Ymax;
-   intel->clear.vertices[2][2] = dst_z;
-   intel->clear.vertices[3][0] = fb->_Xmin;
-   intel->clear.vertices[3][1] = fb->_Ymax;
-   intel->clear.vertices[3][2] = dst_z;
-
-   while (mask != 0) {
-      GLuint this_mask = 0;
-      GLuint color_bit;
-
-      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
-      if (color_bit != 0)
-	 this_mask |= (1 << (color_bit - 1));
-
-      /* Clear depth/stencil in the same pass as color. */
-      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
-
-      /* Select the current color buffer and use the color write mask if
-       * we have one, otherwise don't write any color channels.
-       */
-      if (this_mask & BUFFER_BIT_FRONT_LEFT)
-	 _mesa_DrawBuffer(GL_FRONT_LEFT);
-      else if (this_mask & BUFFER_BIT_BACK_LEFT)
-	 _mesa_DrawBuffer(GL_BACK_LEFT);
-      else if (color_bit != 0)
-	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
-			  (color_bit - BUFFER_COLOR0 - 1));
-      else
-	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
-
-      /* Control writing of the depth clear value to depth. */
-      if (this_mask & BUFFER_BIT_DEPTH) {
-	 _mesa_DepthFunc(GL_ALWAYS);
-	 _mesa_Enable(GL_DEPTH_TEST);
-      } else {
-	 _mesa_Disable(GL_DEPTH_TEST);
-	 _mesa_DepthMask(GL_FALSE);
-      }
-
-      /* Control writing of the stencil clear value to stencil. */
-      if (this_mask & BUFFER_BIT_STENCIL) {
-	 _mesa_Enable(GL_STENCIL_TEST);
-	 _mesa_StencilOpSeparate(GL_FRONT_AND_BACK,
-				 GL_REPLACE, GL_REPLACE, GL_REPLACE);
-	 _mesa_StencilFuncSeparate(GL_FRONT_AND_BACK, GL_ALWAYS,
-				   ctx->Stencil.Clear,
-				   ctx->Stencil.WriteMask[0]);
-      } else {
-	 _mesa_Disable(GL_STENCIL_TEST);
-      }
-
-      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
-
-      mask &= ~this_mask;
-   }
-
-   intel_meta_restore_transform(intel);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
-   if (saved_fp_enable)
-      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-   if (saved_vp_enable)
-      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-
-   if (saved_shader_program)
-      _mesa_UseProgramObjectARB(saved_shader_program);
-
-   _mesa_PopAttrib();
-
-   /* restore current array object */
-   _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj, arraySave);
-   _mesa_reference_array_object(ctx, &arraySave, NULL);
-}
-
 static const char *buffer_names[] = {
    [BUFFER_FRONT_LEFT] = "front",
    [BUFFER_BACK_LEFT] = "back",
@@ -340,7 +110,7 @@ intelClear(GLcontext *ctx, GLbitfield mask)
          = intel_get_rb_region(fb, BUFFER_STENCIL);
       if (stencilRegion) {
          /* have hw stencil */
-         if (IS_965(intel->intelScreen->deviceID) ||
+         if (stencilRegion->tiling == I915_TILING_Y ||
 	     (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
 	    /* We have to use the 3D engine if we're clearing a partial mask
 	     * of the stencil buffer, or if we're on a 965 which has a tiled
@@ -357,9 +127,10 @@ intelClear(GLcontext *ctx, GLbitfield mask)
 
    /* HW depth */
    if (mask & BUFFER_BIT_DEPTH) {
+      const struct intel_region *irb = intel_get_rb_region(fb, BUFFER_DEPTH);
+
       /* clear depth with whatever method is used for stencil (see above) */
-      if (IS_965(intel->intelScreen->deviceID) ||
-	  tri_mask & BUFFER_BIT_STENCIL)
+      if (irb->tiling == I915_TILING_Y || tri_mask & BUFFER_BIT_STENCIL)
          tri_mask |= BUFFER_BIT_DEPTH;
       else
          blit_mask |= BUFFER_BIT_DEPTH;
@@ -411,7 +182,7 @@ intelClear(GLcontext *ctx, GLbitfield mask)
 	 }
 	 DBG("\n");
       }
-      intel_clear_tris(ctx, tri_mask);
+      meta_clear_tris(&intel->meta, tri_mask);
    }
 
    if (swrast_mask) {
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index cfd983d368..4abb525f78 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -161,6 +161,12 @@ intelGetString(GLcontext * ctx, GLenum name)
       case PCI_CHIP_G41_G:
          chipset = "Intel(R) G41";
          break;
+      case PCI_CHIP_ILD_G:
+         chipset = "Intel(R) IGDNG_D";
+         break;
+      case PCI_CHIP_ILM_G:
+         chipset = "Intel(R) IGDNG_M";
+         break;
       default:
          chipset = "Unknown Intel Chipset";
          break;
@@ -220,7 +226,9 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
       struct intel_renderbuffer *stencil_rb;
 
       i = 0;
-      if ((intel->is_front_buffer_rendering || !intel_fb->color_rb[1])
+      if ((intel->is_front_buffer_rendering ||
+	   intel->is_front_buffer_reading ||
+	   !intel_fb->color_rb[1])
 	   && intel_fb->color_rb[0]) {
 	 attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
 	 attachments[i++] = intel_bits_per_pixel(intel_fb->color_rb[0]);
@@ -396,7 +404,7 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
     if (!driContext->driScreenPriv->dri2.enabled)
 	return;
 
-    if (!intel->internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+    if (!intel->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
        /* If we're rendering to the fake front buffer, make sure all the pending
 	* drawing has landed on the real front buffer.  Otherwise when we
 	* eventually get to DRI2GetBuffersWithFormat the stale real front
@@ -561,10 +569,14 @@ intelInitDriverFunctions(struct dd_function_table *functions)
    functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D;
 
    intelInitTextureFuncs(functions);
+   intelInitTextureImageFuncs(functions);
+   intelInitTextureSubImageFuncs(functions);
+   intelInitTextureCopyImageFuncs(functions);
    intelInitStateFuncs(functions);
    intelInitClearFuncs(functions);
    intelInitBufferFuncs(functions);
    intelInitPixelFuncs(functions);
+   intelInitBufferObjectFuncs(functions);
 }
 
 
@@ -660,7 +672,15 @@ intelInitContext(struct intel_context *intel,
     */
    _mesa_init_point(ctx);
 
+   meta_init_metaops(ctx, &intel->meta);
    ctx->Const.MaxColorAttachments = 4;  /* XXX FBO: review this */
+   if (IS_965(intelScreen->deviceID)) {
+      if (MAX_WIDTH > 8192)
+	 ctx->Const.MaxRenderbufferSize = 8192;
+   } else {
+      if (MAX_WIDTH > 2048)
+	 ctx->Const.MaxRenderbufferSize = 2048;
+   }
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext(ctx);
@@ -718,7 +738,6 @@ intelInitContext(struct intel_context *intel,
 
    intel->batch = intel_batchbuffer_alloc(intel);
 
-   intel_bufferobj_init(intel);
    intel_fbo_init(intel);
 
    if (intel->ctx.Mesa_DXTn) {
@@ -728,6 +747,15 @@ intelInitContext(struct intel_context *intel,
    else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
    }
+   intel->use_texture_tiling = driQueryOptionb(&intel->optionCache,
+					       "texture_tiling");
+   if (intel->use_texture_tiling &&
+       !intel->intelScreen->kernel_exec_fencing) {
+      fprintf(stderr, "No kernel support for execution fencing, "
+	      "disabling texture tiling");
+      intel->use_texture_tiling = GL_FALSE;
+   }
+   intel->use_early_z = driQueryOptionb(&intel->optionCache, "early_z");
 
    intel->prim.primitive = ~0;
 
@@ -767,8 +795,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       INTEL_FIREVERTICES(intel);
 
-      if (intel->clear.arrayObj)
-         _mesa_delete_array_object(&intel->ctx, intel->clear.arrayObj);
+      meta_destroy_metaops(&intel->meta);
 
       intel->vtbl.destroy(intel);
 
@@ -789,13 +816,64 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       intel->prim.vb_bo = NULL;
 
       if (release_texture_heaps) {
-         /* This share group is about to go away, free our private
-          * texture object data.
+         /* Nothing is currently done here to free texture heaps;
+          * but we're not using the texture heap utilities, so I
+          * rather think we shouldn't.  I've taken a look, and can't
+          * find any private texture data hanging around anywhere, but
+          * I'm not yet certain there isn't any at all...
           */
-         if (INTEL_DEBUG & DEBUG_TEXTURE)
+         /* if (INTEL_DEBUG & DEBUG_TEXTURE)
             fprintf(stderr, "do something to free texture heaps\n");
+          */
       }
 
+      /* XXX In intelMakeCurrent() below, the context's static regions are 
+       * referenced inside the frame buffer; it's listed as a hack,
+       * with a comment of "XXX FBO temporary fix-ups!", but
+       * as long as it's there, we should release the regions here.
+       * The do/while loop around the block is used to allow the
+       * "continue" statements inside the block to exit the block,
+       * to avoid many layers of "if" constructs.
+       */
+      do {
+         __DRIdrawablePrivate * driDrawPriv = intel->driDrawable;
+         struct intel_framebuffer *intel_fb;
+         struct intel_renderbuffer *irbDepth, *irbStencil;
+         if (!driDrawPriv) {
+            /* We're already detached from the drawable; exit this block. */
+            continue;
+         }
+         intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate;
+         if (!intel_fb) {
+            /* The frame buffer is already gone; exit this block. */
+            continue;
+         }
+         irbDepth = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+         irbStencil = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+
+         /* If the regions of the frame buffer still match the regions
+          * of the context, release them.  If they've changed somehow,
+          * leave them alone.
+          */
+         if (intel_fb->color_rb[0] && intel_fb->color_rb[0]->region == intel->front_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[0], NULL);
+         }
+         if (intel_fb->color_rb[1] && intel_fb->color_rb[1]->region == intel->back_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[1], NULL);
+         }
+
+         if (irbDepth && irbDepth->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbDepth, NULL);
+         }
+         /* Usually, the stencil buffer is the same as the depth buffer;
+          * but they're handled separately in MakeCurrent, so we'll
+          * handle them separately here.
+          */
+         if (irbStencil && irbStencil->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbStencil, NULL);
+         }
+      } while (0);
+
       intel_region_release(&intel->front_region);
       intel_region_release(&intel->back_region);
       intel_region_release(&intel->depth_region);
@@ -804,6 +882,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       /* free the Mesa context */
       _mesa_free_context_data(&intel->ctx);
+
+      FREE(intel);
+      driContextPriv->driverPrivate = NULL;
    }
 }
 
@@ -832,7 +913,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
           if (driDrawPriv != driReadPriv)
               intel_update_renderbuffers(driContextPriv, driReadPriv);
       } else {
-          /* XXX FBO temporary fix-ups! */
+          /* XXX FBO temporary fix-ups!  These are released in 
+           * intelDextroyContext(), above.  Changes here should be
+           * reflected there.
+           */
           /* if the renderbuffers don't have regions, init them from the context */
          struct intel_renderbuffer *irbDepth
             = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
@@ -917,7 +1001,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
    int me = intel->hHWContext;
 
    drmGetLock(intel->driFd, intel->hHWContext, flags);
-   intel->locked = 1;
 
    if (INTEL_DEBUG & DEBUG_LOCK)
       _mesa_printf("%s - got contended lock\n", __progname);
@@ -974,9 +1057,12 @@ void LOCK_HARDWARE( struct intel_context *intel )
     struct intel_framebuffer *intel_fb = NULL;
     struct intel_renderbuffer *intel_rb = NULL;
 
-    _glthread_LOCK_MUTEX(lockMutex);
-    assert(!intel->locked);
-    intel->locked = 1;
+    intel->locked++;
+    if (intel->locked >= 2)
+       return;
+
+    if (!sPriv->dri2.enabled)
+       _glthread_LOCK_MUTEX(lockMutex);
 
     if (intel->driDrawable) {
        intel_fb = intel->driDrawable->driverPrivate;
@@ -1023,13 +1109,16 @@ void UNLOCK_HARDWARE( struct intel_context *intel )
 {
     __DRIscreen *sPriv = intel->driScreen;
 
-   intel->vtbl.note_unlock( intel );
-   intel->locked = 0;
+   intel->locked--;
+   if (intel->locked > 0)
+      return;
 
-   if (!sPriv->dri2.enabled)
-      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+   assert(intel->locked == 0);
 
-   _glthread_UNLOCK_MUTEX(lockMutex);
+   if (!sPriv->dri2.enabled) {
+      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+      _glthread_UNLOCK_MUTEX(lockMutex);
+   }
 
    if (INTEL_DEBUG & DEBUG_LOCK)
       _mesa_printf("%s - unlocked\n", __progname);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index b5cf7e6648..08bea88c95 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -33,6 +33,7 @@
 #include "main/mtypes.h"
 #include "main/mm.h"
 #include "texmem.h"
+#include "dri_metaops.h"
 #include "drm.h"
 #include "intel_bufmgr.h"
 
@@ -91,7 +92,6 @@ struct intel_context
       void (*new_batch) (struct intel_context * intel);
       void (*emit_invarient_state) (struct intel_context * intel);
       void (*note_fence) (struct intel_context *intel, GLuint fence);
-      void (*note_unlock) (struct intel_context *intel);
       void (*update_texture_state) (struct intel_context * intel);
 
       void (*render_start) (struct intel_context * intel);
@@ -158,19 +158,7 @@ struct intel_context
       void (*debug_batch)(struct intel_context *intel);
    } vtbl;
 
-   struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-      GLenum saved_matrix_mode;
-   } meta;
+   struct dri_metaops meta;
 
    GLint refcount;
    GLuint Fallback;
@@ -182,7 +170,6 @@ struct intel_context
    struct intel_region *front_region;
    struct intel_region *back_region;
    struct intel_region *depth_region;
-   GLboolean internal_viewport_call;
 
    /**
     * This value indicates that the kernel memory manager is being used
@@ -215,13 +202,6 @@ struct intel_context
    GLuint ClearColor565;
    GLuint ClearColor8888;
 
-   /* info for intel_clear_tris() */
-   struct
-   {
-      struct gl_array_object *arrayObj;
-      GLfloat vertices[4][3];
-      GLfloat color[4][4];
-   } clear;
 
    /* Offsets of fields within the current vertex:
     */
@@ -294,6 +274,17 @@ struct intel_context
     * easily.
     */
    GLboolean is_front_buffer_rendering;
+   /**
+    * Track whether front-buffer is the current read target.
+    *
+    * This is closely associated with is_front_buffer_rendering, but may
+    * be set separately.  The DRI2 fake front buffer must be referenced
+    * either way.
+    */
+   GLboolean is_front_buffer_reading;
+
+   GLboolean use_texture_tiling;
+   GLboolean use_early_z;
 
    drm_clip_rect_t fboRect;     /**< cliprect for FBO rendering */
 
@@ -549,6 +540,9 @@ void intel_viewport(GLcontext * ctx, GLint x, GLint y,
 void intel_update_renderbuffers(__DRIcontext *context,
 				__DRIdrawable *drawable);
 
+void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
+				  uint32_t buffer_id);
+
 /*======================================================================
  * Inline conversion functions.  
  * These are better-typed than the macros used previously:
diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c
index 9ec1b4ec2f..7742609d24 100644
--- a/src/mesa/drivers/dri/intel/intel_extensions.c
+++ b/src/mesa/drivers/dri/intel/intel_extensions.c
@@ -34,6 +34,7 @@
 #define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
 #define need_GL_ARB_shader_objects
+#define need_GL_ARB_vertex_array_object
 #define need_GL_ARB_vertex_program
 #define need_GL_ARB_vertex_shader
 #define need_GL_ARB_window_pos
@@ -45,9 +46,11 @@
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_framebuffer_blit
+#define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_point_parameters
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_stencil_two_side
+#define need_GL_APPLE_vertex_array_object
 #define need_GL_ATI_separate_stencil
 #define need_GL_ATI_envmap_bumpmap
 #define need_GL_NV_point_sprite
@@ -65,6 +68,7 @@
  * i965_dri.
  */
 static const struct dri_extension card_extensions[] = {
+   { "GL_ARB_half_float_pixel",           NULL },
    { "GL_ARB_multitexture",               NULL },
    { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
    { "GL_ARB_texture_border_clamp",       NULL },
@@ -75,6 +79,7 @@ static const struct dri_extension card_extensions[] = {
    { "GL_ARB_texture_env_dot3",           NULL },
    { "GL_ARB_texture_mirrored_repeat",    NULL },
    { "GL_ARB_texture_rectangle",          NULL },
+   { "GL_ARB_vertex_array_object",        GL_ARB_vertex_array_object_functions},
    { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
    { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
    { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
@@ -85,6 +90,7 @@ static const struct dri_extension card_extensions[] = {
    { "GL_EXT_blend_subtract",             NULL },
    { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
    { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+   { "GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions },
    { "GL_EXT_packed_depth_stencil",       NULL },
    { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
    { "GL_EXT_stencil_wrap",               NULL },
@@ -95,6 +101,7 @@ static const struct dri_extension card_extensions[] = {
    { "GL_EXT_texture_lod_bias",           NULL },
    { "GL_3DFX_texture_compression_FXT1",  NULL },
    { "GL_APPLE_client_storage",           NULL },
+   { "GL_APPLE_vertex_array_object",      GL_APPLE_vertex_array_object_functions},
    { "GL_MESA_pack_invert",               NULL },
    { "GL_MESA_ycbcr_texture",             NULL },
    { "GL_NV_blend_square",                NULL },
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c
index 30f58b1f44..666893596e 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.c
+++ b/src/mesa/drivers/dri/intel/intel_fbo.c
@@ -217,7 +217,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width,
 	  height, pitch);
 
-      irb->region = intel_region_alloc(intel, cpp, width, height, pitch,
+      irb->region = intel_region_alloc(intel, I915_TILING_NONE,
+				       cpp, width, height, pitch,
 				       GL_TRUE);
       if (!irb->region)
          return GL_FALSE;       /* out of memory? */
@@ -575,9 +576,10 @@ intel_render_texture(GLcontext * ctx,
 
    ASSERT(newImage);
 
-   if (newImage->Border != 0) {
-      /* Fallback on drawing to a texture with a border, which won't have a
-       * miptree.
+   intel_image = intel_texture_image(newImage);
+   if (!intel_image->mt) {
+      /* Fallback on drawing to a texture that doesn't have a miptree
+       * (has a border, width/height 0, etc.)
        */
       _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
       _mesa_render_texture(ctx, fb, att);
@@ -608,7 +610,6 @@ intel_render_texture(GLcontext * ctx,
        irb->Base.RefCount);
 
    /* point the renderbufer's region to the texture image region */
-   intel_image = intel_texture_image(newImage);
    if (irb->region != intel_image->mt->region) {
       if (irb->region)
 	 intel_region_release(&irb->region);
@@ -680,6 +681,11 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
       if (rb == NULL)
 	 continue;
 
+      if (irb == NULL) {
+	 fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+	 continue;
+      }
+
       switch (irb->texformat->MesaFormat) {
       case MESA_FORMAT_ARGB8888:
       case MESA_FORMAT_RGB565:
diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
new file mode 100644
index 0000000000..fe986092db
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
@@ -0,0 +1,283 @@
+/*
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/bufferobj.h"
+#include "main/teximage.h"
+#include "main/texenv.h"
+#include "main/texobj.h"
+#include "main/texstate.h"
+#include "main/texparam.h"
+#include "main/varray.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/buffers.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/depth.h"
+#include "main/hash.h"
+#include "main/mipmap.h"
+#include "main/blend.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_pixel.h"
+#include "intel_tex.h"
+#include "intel_mipmap_tree.h"
+
+static const char *intel_fp_tex2d =
+      "!!ARBfp1.0\n"
+      "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n"
+      "END\n";
+
+static GLboolean
+intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name,
+			    int level, int width, int height)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLfloat vertices[4][2];
+   GLint status;
+
+   /* Set to source from the previous level */
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1);
+
+   /* Set to draw into the current level */
+   _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT,
+				 GL_COLOR_ATTACHMENT0_EXT,
+				 GL_TEXTURE_2D,
+				 tex_name,
+				 level);
+   /* Choose to render to the color attachment. */
+   _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
+
+   status = _mesa_CheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT);
+   if (status != GL_FRAMEBUFFER_COMPLETE_EXT)
+      return GL_FALSE;
+
+   meta_set_passthrough_transform(&intel->meta);
+
+   /* XXX: Doing it right would involve setting up the transformation to do
+    * 0-1 mapping or something, and not changing the vertex data.
+    */
+   vertices[0][0] = 0;
+   vertices[0][1] = 0;
+   vertices[1][0] = width;
+   vertices[1][1] = 0;
+   vertices[2][0] = width;
+   vertices[2][1] = height;
+   vertices[3][0] = 0;
+   vertices[3][1] = height;
+
+   _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+   meta_set_default_texrect(&intel->meta);
+
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+
+   return GL_TRUE;
+}
+
+static GLboolean
+intel_generate_mipmap_2d(GLcontext *ctx,
+			 GLenum target,
+			 struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLint old_active_texture;
+   int level, max_levels, start_level, end_level;
+   GLuint fb_name;
+   GLboolean success = GL_FALSE;
+   struct gl_framebuffer *saved_fbo = NULL;
+
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
+		    GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT |
+		    GL_DEPTH_BUFFER_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+   old_active_texture = ctx->Texture.CurrentUnit;
+   _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer);
+
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_DEPTH_TEST);
+   _mesa_Disable(GL_STENCIL_TEST);
+   _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+   _mesa_DepthMask(GL_FALSE);
+
+   /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our
+    * minification.
+    */
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
+   _mesa_Enable(GL_TEXTURE_2D);
+   _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER,
+		       GL_LINEAR_MIPMAP_NEAREST);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+
+   /* Bind the new renderbuffer to the color attachment point. */
+   _mesa_GenFramebuffersEXT(1, &fb_name);
+   _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name);
+
+   meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp,
+			     intel_fp_tex2d);
+   meta_set_passthrough_vertex_program(&intel->meta);
+
+   max_levels = _mesa_max_texture_levels(ctx, texObj->Target);
+   start_level = texObj->BaseLevel;
+   end_level = texObj->MaxLevel;
+
+   /* Loop generating level+1 from level. */
+   for (level = start_level; level < end_level && level < max_levels - 1; level++) {
+      const struct gl_texture_image *srcImage;
+      int width, height;
+
+      srcImage = _mesa_select_tex_image(ctx, texObj, target, level);
+      if (srcImage->Border != 0)
+	 goto fail;
+
+      width = srcImage->Width / 2;
+      if (width < 1)
+	 width = 1;
+      height = srcImage->Height / 2;
+      if (height < 1)
+	 height = 1;
+
+      if (width == srcImage->Width &&
+	  height == srcImage->Height) {
+	 /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the
+	  * maximum texture level for the object, so break out when we've gone
+	  * over the edge.
+	  */
+	 break;
+      }
+
+      /* Make sure that there's space allocated for the target level.
+       * We could skip this if there's already space allocated and save some
+       * time.
+       */
+      _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat,
+		       width, height, 0,
+		       GL_RGBA, GL_UNSIGNED_INT, NULL);
+
+      if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1,
+				       width, height))
+	 goto fail;
+   }
+
+   success = GL_TRUE;
+
+fail:
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
+
+   _mesa_DeleteFramebuffersEXT(1, &fb_name);
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
+   if (saved_fbo)
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name);
+   _mesa_reference_framebuffer(&saved_fbo, NULL);
+   _mesa_PopClientAttrib();
+   _mesa_PopAttrib();
+
+   return success;
+}
+
+
+/**
+ * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
+ * level).
+ *
+ * The texture object's miptree must be mapped.
+ *
+ * It would be really nice if this was just called by Mesa whenever mipmaps
+ * needed to be regenerated, rather than us having to remember to do so in
+ * each texture image modification path.
+ *
+ * This function should also include an accelerated path.
+ */
+void
+intel_generate_mipmap(GLcontext *ctx, GLenum target,
+                      struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_texture_object *intelObj = intel_texture_object(texObj);
+   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+   int face, i;
+
+   /* HW path */
+   if (target == GL_TEXTURE_2D &&
+       ctx->Extensions.EXT_framebuffer_object &&
+       ctx->Extensions.ARB_fragment_program &&
+       ctx->Extensions.ARB_vertex_program) {
+      GLboolean success;
+
+      /* We'll be accessing this texture using GL entrypoints, which should
+       * be resilient against other access to this texture.
+       */
+      _mesa_unlock_texture(ctx, texObj);
+      success = intel_generate_mipmap_2d(ctx, target, texObj);
+      _mesa_lock_texture(ctx, texObj);
+
+      if (success)
+	 return;
+   }
+
+   /* SW path */
+   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
+   _mesa_generate_mipmap(ctx, target, texObj);
+   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
+
+   /* Update the level information in our private data in the new images, since
+    * it didn't get set as part of a normal TexImage path.
+    */
+   for (face = 0; face < nr_faces; face++) {
+      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
+         struct intel_texture_image *intelImage;
+
+	 intelImage = intel_texture_image(texObj->Image[face][i]);
+	 if (intelImage == NULL)
+	    break;
+
+	 intelImage->level = i;
+	 intelImage->face = face;
+	 /* Unreference the miptree to signal that the new Data is a bare
+	  * pointer from mesa.
+	  */
+	 intel_miptree_release(intel, &intelImage->mt);
+      }
+   }
+}
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index 6e1e034e53..c985da5aa2 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -57,14 +57,16 @@ intel_miptree_create_internal(struct intel_context *intel,
 			      GLuint last_level,
 			      GLuint width0,
 			      GLuint height0,
-			      GLuint depth0, GLuint cpp, GLuint compress_byte)
+			      GLuint depth0, GLuint cpp, GLuint compress_byte,
+			      uint32_t tiling)
 {
    GLboolean ok;
    struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1);
 
-   DBG("%s target %s format %s level %d..%d\n", __FUNCTION__,
+   DBG("%s target %s format %s level %d..%d <-- %p\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target),
-       _mesa_lookup_enum_by_nr(internal_format), first_level, last_level);
+       _mesa_lookup_enum_by_nr(internal_format), 
+       first_level, last_level, mt);
 
    mt->target = target_to_target(target);
    mt->internal_format = internal_format;
@@ -80,15 +82,16 @@ intel_miptree_create_internal(struct intel_context *intel,
 
 #ifdef I915
    if (IS_945(intel->intelScreen->deviceID))
-      ok = i945_miptree_layout(intel, mt);
+      ok = i945_miptree_layout(intel, mt, tiling);
    else
-      ok = i915_miptree_layout(intel, mt);
+      ok = i915_miptree_layout(intel, mt, tiling);
 #else
-   ok = brw_miptree_layout(intel, mt);
+   ok = brw_miptree_layout(intel, mt, tiling);
 #endif
 
    if (!ok) {
       free(mt);
+      DBG("%s not okay - returning NULL\n", __FUNCTION__);
       return NULL;
    }
 
@@ -98,6 +101,7 @@ intel_miptree_create_internal(struct intel_context *intel,
 struct intel_mipmap_tree *
 intel_miptree_create(struct intel_context *intel,
 		     GLenum target,
+		     GLenum base_format,
 		     GLenum internal_format,
 		     GLuint first_level,
 		     GLuint last_level,
@@ -107,10 +111,23 @@ intel_miptree_create(struct intel_context *intel,
 		     GLboolean expect_accelerated_upload)
 {
    struct intel_mipmap_tree *mt;
+   uint32_t tiling;
+
+   if (intel->use_texture_tiling && compress_byte == 0 &&
+       intel->intelScreen->kernel_exec_fencing) {
+      if (IS_965(intel->intelScreen->deviceID) &&
+	  (base_format == GL_DEPTH_COMPONENT ||
+	   base_format == GL_DEPTH_STENCIL_EXT))
+	 tiling = I915_TILING_Y;
+      else
+	 tiling = I915_TILING_X;
+   } else
+      tiling = I915_TILING_NONE;
 
    mt = intel_miptree_create_internal(intel, target, internal_format,
 				      first_level, last_level, width0,
-				      height0, depth0, cpp, compress_byte);
+				      height0, depth0, cpp, compress_byte,
+				      tiling);
    /*
     * pitch == 0 || height == 0  indicates the null texture
     */
@@ -118,6 +135,7 @@ intel_miptree_create(struct intel_context *intel,
       return NULL;
 
    mt->region = intel_region_alloc(intel,
+				   tiling,
 				   mt->cpp,
 				   mt->pitch,
 				   mt->total_height,
@@ -147,7 +165,8 @@ intel_miptree_create_for_region(struct intel_context *intel,
    mt = intel_miptree_create_internal(intel, target, internal_format,
 				      first_level, last_level,
 				      region->width, region->height, 1,
-				      region->cpp, compress_byte);
+				      region->cpp, compress_byte,
+				      I915_TILING_NONE);
    if (!mt)
       return mt;
 #if 0
@@ -185,6 +204,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
 
 int intel_miptree_pitch_align (struct intel_context *intel,
 			       struct intel_mipmap_tree *mt,
+			       uint32_t tiling,
 			       int pitch)
 {
 #ifdef I915
@@ -205,6 +225,11 @@ int intel_miptree_pitch_align (struct intel_context *intel,
 	 pitch_align = 4;
       }
 
+      if (tiling == I915_TILING_X)
+	 pitch_align = 512;
+      else if (tiling == I915_TILING_Y)
+	 pitch_align = 128;
+
       pitch = ALIGN(pitch * mt->cpp, pitch_align);
 
 #ifdef I915
@@ -328,23 +353,31 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
 }
 
 
-
 void
-intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
-			       GLuint level, GLuint img,
-			       GLuint x, GLuint y)
+intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt,
+                                  GLuint level, GLuint img,
+                                  GLuint x, GLuint y, 
+                                  GLuint offset)
 {
    if (img == 0 && level == 0)
       assert(x == 0 && y == 0);
 
    assert(img < mt->level[level].nr_images);
 
-   mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp;
+   mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp + offset;
 
    DBG("%s level %d img %d pos %d,%d image_offset %x\n",
        __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]);
 }
 
+void
+intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
+			       GLuint level, GLuint img,
+			       GLuint x, GLuint y)
+{
+    intel_miptree_set_image_offset_ex(mt, level, img, x, y, 0);
+}
+
 
 /* Although we use the image_offset[] array to store relative offsets
  * to cube faces, Mesa doesn't know anything about this and expects
@@ -454,11 +487,11 @@ intel_miptree_image_data(struct intel_context *intel,
 			0, 0,                             /* source x, y */
 			dst->level[level].width, height); /* width, height */
 
-      src += src_image_pitch * dst->cpp;
+      src = (char *)src + src_image_pitch * dst->cpp;
    }
 }
 
-extern GLuint intel_compressed_alignment(GLenum);
+extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *);
 /* Copy mipmap image between trees
  */
 void
@@ -475,20 +508,37 @@ intel_miptree_image_copy(struct intel_context *intel,
    const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level);
    const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level);
    GLuint i;
+   GLboolean success;
 
    if (dst->compressed) {
-       GLuint alignment = intel_compressed_alignment(dst->internal_format);
+       GLuint align_w, align_h;
+
+       intel_get_texture_alignment_unit(dst->internal_format, &align_w, &align_h);
        height = (height + 3) / 4;
-       width = ((width + alignment - 1) & ~(alignment - 1));
+       width = ALIGN(width, align_w);
    }
 
    for (i = 0; i < depth; i++) {
-      intel_region_copy(intel,
-                        dst->region, dst_offset + dst_depth_offset[i],
-                        0,
-                        0,
-                        src->region, src_offset + src_depth_offset[i],
-                        0, 0, width, height);
+      success = intel_region_copy(intel,
+				  dst->region, dst_offset + dst_depth_offset[i],
+				  0, 0,
+				  src->region, src_offset + src_depth_offset[i],
+				  0, 0, width, height, GL_COPY);
+      if (!success) {
+	 GLubyte *src_ptr, *dst_ptr;
+
+	 src_ptr = intel_region_map(intel, src->region);
+	 dst_ptr = intel_region_map(intel, dst->region);
+
+	 _mesa_copy_rect(dst_ptr + dst_offset + dst_depth_offset[i],
+			 dst->cpp,
+			 dst->pitch,
+			 0, 0, width, height,
+			 src_ptr + src_offset + src_depth_offset[i],
+			 src->pitch,
+			 0, 0);
+	 intel_region_unmap(intel, src->region);
+	 intel_region_unmap(intel, dst->region);
+      }
    }
-
 }
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
index 4060b9df78..c890b2a0d0 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
@@ -126,6 +126,7 @@ struct intel_mipmap_tree
 
 struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel,
                                                GLenum target,
+                                               GLenum base_format,
                                                GLenum internal_format,
                                                GLuint first_level,
                                                GLuint last_level,
@@ -148,6 +149,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
 
 int intel_miptree_pitch_align (struct intel_context *intel,
 			       struct intel_mipmap_tree *mt,
+			       uint32_t tiling,
 			       int pitch);
 
 void intel_miptree_reference(struct intel_mipmap_tree **dst,
@@ -194,6 +196,11 @@ void intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
                                   GLuint x, GLuint y,
                                   GLuint w, GLuint h, GLuint d);
 
+void intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt,
+                                       GLuint level,
+                                       GLuint img, GLuint x, GLuint y,
+                                       GLuint offset);
+
 void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
                                     GLuint level,
                                     GLuint img, GLuint x, GLuint y);
@@ -218,10 +225,13 @@ void intel_miptree_image_copy(struct intel_context *intel,
 /* i915_mipmap_tree.c:
  */
 GLboolean i915_miptree_layout(struct intel_context *intel,
-			      struct intel_mipmap_tree *mt);
+			      struct intel_mipmap_tree *mt,
+			      uint32_t tiling);
 GLboolean i945_miptree_layout(struct intel_context *intel,
-			      struct intel_mipmap_tree *mt);
+			      struct intel_mipmap_tree *mt,
+			      uint32_t tiling);
 GLboolean brw_miptree_layout(struct intel_context *intel,
-			     struct intel_mipmap_tree *mt);
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index fc0ac0b79c..a300141655 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -27,9 +27,12 @@
 
 #include "main/enums.h"
 #include "main/state.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/enable.h"
 #include "main/matrix.h"
+#include "main/texstate.h"
+#include "main/varray.h"
 #include "main/viewport.h"
 #include "swrast/swrast.h"
 #include "shader/arbprogram.h"
@@ -174,167 +177,6 @@ intel_check_blit_format(struct intel_region * region,
 }
 
 void
-intel_meta_set_passthrough_transform(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   intel->meta.saved_vp_x = ctx->Viewport.X;
-   intel->meta.saved_vp_y = ctx->Viewport.Y;
-   intel->meta.saved_vp_width = ctx->Viewport.Width;
-   intel->meta.saved_vp_height = ctx->Viewport.Height;
-   intel->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
-
-   intel->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-   intel->internal_viewport_call = GL_FALSE;
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-void
-intel_meta_restore_transform(struct intel_context *intel)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_MatrixMode(intel->meta.saved_matrix_mode);
-
-   intel->internal_viewport_call = GL_TRUE;
-   _mesa_Viewport(intel->meta.saved_vp_x, intel->meta.saved_vp_y,
-		  intel->meta.saved_vp_width, intel->meta.saved_vp_height);
-   intel->internal_viewport_call = GL_FALSE;
-}
-
-/**
- * Set up a vertex program to pass through the position and first texcoord
- * for pixel path.
- */
-void
-intel_meta_set_passthrough_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   static const char *vp =
-      "!!ARBvp1.0\n"
-      "TEMP vertexClip;\n"
-      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
-      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
-      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
-      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
-      "MOV result.position, vertexClip;\n"
-      "MOV result.texcoord[0], vertex.texcoord[0];\n"
-      "MOV result.color, vertex.color;\n"
-      "END\n";
-
-   assert(intel->meta.saved_vp == NULL);
-
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp,
-			    ctx->VertexProgram.Current);
-   if (intel->meta.passthrough_vp == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(vp), (const GLubyte *)vp);
-      _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp,
-			       ctx->VertexProgram.Current);
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.passthrough_vp);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &intel->meta.passthrough_vp->Base);
-
-   intel->meta.saved_vp_enable = ctx->VertexProgram.Enabled;
-   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous vertex program after
- * intel_meta_set_passthrough_vertex_program()
- */
-void
-intel_meta_restore_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.saved_vp);
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &ctx->VertexProgram.Current->Base);
-
-   if (!intel->meta.saved_vp_enable)
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
- * program object.
- */
-void
-intel_meta_set_fragment_program(struct intel_context *intel,
-				struct gl_fragment_program **prog,
-				const char *prog_string)
-{
-   GLcontext *ctx = &intel->ctx;
-   assert(intel->meta.saved_fp == NULL);
-
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp,
-			    ctx->FragmentProgram.Current);
-   if (*prog == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(prog_string), (const GLubyte *)prog_string);
-      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
-      /* Note that DeletePrograms unbinds the program on us */
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
-
-   intel->meta.saved_fp_enable = ctx->FragmentProgram.Enabled;
-   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous fragment program after
- * intel_meta_set_fragment_program()
- */
-void
-intel_meta_restore_fragment_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
-			    intel->meta.saved_fp);
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
-			   &ctx->FragmentProgram.Current->Base);
-
-   if (!intel->meta.saved_fp_enable)
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-void
 intelInitPixelFuncs(struct dd_function_table *functions)
 {
    functions->Accum = _swrast_Accum;
@@ -342,18 +184,7 @@ intelInitPixelFuncs(struct dd_function_table *functions)
       functions->Bitmap = intelBitmap;
       functions->CopyPixels = intelCopyPixels;
       functions->DrawPixels = intelDrawPixels;
-#ifdef I915
-      functions->ReadPixels = intelReadPixels;
-#endif
    }
-}
-
-void
-intel_free_pixel_state(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL);
-   _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL);
+   functions->ReadPixels = intelReadPixels;
 }
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h
index cb41fa182c..96a6dd17b2 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.h
+++ b/src/mesa/drivers/dri/intel/intel_pixel.h
@@ -31,16 +31,6 @@
 #include "main/mtypes.h"
 
 void intelInitPixelFuncs(struct dd_function_table *functions);
-void intel_meta_set_passthrough_transform(struct intel_context *intel);
-void intel_meta_restore_transform(struct intel_context *intel);
-void intel_meta_set_passthrough_vertex_program(struct intel_context *intel);
-void intel_meta_restore_vertex_program(struct intel_context *intel);
-void intel_meta_set_fragment_program(struct intel_context *intel,
-				     struct gl_fragment_program **prog,
-				     const char *prog_string);
-void intel_meta_restore_fragment_program(struct intel_context *intel);
-void intel_free_pixel_state(struct intel_context *intel);
-
 GLboolean intel_check_blit_fragment_ops(GLcontext * ctx,
 					GLboolean src_alpha_is_one);
 
@@ -76,6 +66,4 @@ void intelBitmap(GLcontext * ctx,
 		 const struct gl_pixelstore_attrib *unpack,
 		 const GLubyte * pixels);
 
-void intel_clear_tris(GLcontext *ctx, GLbitfield mask);
-
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index a2ccae1b7d..540e7620a9 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -42,6 +42,7 @@
 #include "main/varray.h"
 #include "main/attrib.h"
 #include "main/enable.h"
+#include "main/viewport.h"
 #include "shader/arbprogram.h"
 #include "glapi/dispatch.h"
 #include "swrast/swrast.h"
@@ -194,7 +195,7 @@ do_blit_bitmap( GLcontext *ctx,
    struct gl_framebuffer *fb = ctx->DrawBuffer;
    GLfloat tmpColor[4];
    GLubyte ubcolor[4];
-   GLuint color8888, color565;
+   GLuint color;
    unsigned int num_cliprects;
    drm_clip_rect_t *cliprects;
    int x_off, y_off;
@@ -232,8 +233,11 @@ do_blit_bitmap( GLcontext *ctx,
    UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[2], tmpColor[2]);
    UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]);
 
-   color8888 = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], ubcolor[2], ubcolor[3]);
-   color565 = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
+   if (dst->cpp == 2)
+      color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
+   else
+      color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1],
+				  ubcolor[2], ubcolor[3]);
 
    if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F))
       return GL_FALSE;
@@ -307,21 +311,21 @@ do_blit_bitmap( GLcontext *ctx,
 				   fb->Name == 0 ? GL_TRUE : GL_FALSE) == 0)
 		  continue;
 
-	       /* 
-		*/
-	       intelEmitImmediateColorExpandBlit( intel,
-						  dst->cpp,
-						  (GLubyte *)stipple, 
-						  sz,
-						  (dst->cpp == 2) ? color565 : color8888,
-						  dst->pitch,
-						  dst->buffer,
-						  0,
-						  dst->tiling,
-						  box_x + px,
-						  box_y + py,
-						  w, h,
-						  logic_op);
+	       if (!intelEmitImmediateColorExpandBlit(intel,
+						      dst->cpp,
+						      (GLubyte *)stipple,
+						      sz,
+						      color,
+						      dst->pitch,
+						      dst->buffer,
+						      0,
+						      dst->tiling,
+						      box_x + px,
+						      box_y + py,
+						      w, h,
+						      logic_op)) {
+		  return GL_FALSE;
+	       }
 	    } 
 	 } 
       }
@@ -360,7 +364,6 @@ intel_texture_bitmap(GLcontext * ctx,
       "END\n";
    GLuint texname;
    GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
    GLint old_active_texture;
    GLubyte *unpacked_bitmap;
    GLubyte *a8_bitmap;
@@ -409,6 +412,12 @@ intel_texture_bitmap(GLcontext * ctx,
       return GL_FALSE;
    }
 
+   if (ctx->Fog.Enabled) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glBitmap() fallback: fog\n");
+      return GL_FALSE;
+   }
+
    /* Check that we can load in a texture this big. */
    if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) ||
        height > (1 << (ctx->Const.MaxTextureLevels - 1))) {
@@ -467,15 +476,18 @@ intel_texture_bitmap(GLcontext * ctx,
 		    GL_ALPHA, GL_UNSIGNED_BYTE, a8_bitmap);
    _mesa_free(a8_bitmap);
 
-   intel_meta_set_fragment_program(intel, &intel->meta.bitmap_fp, fp);
+   meta_set_fragment_program(&intel->meta, &intel->meta.bitmap_fp, fp);
    _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0,
 				     ctx->Current.RasterColor);
-   intel_meta_set_passthrough_vertex_program(intel);
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_vertex_program(&intel->meta);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
    dst_z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
 
+   /* RasterPos[2] already takes into account the DepthRange mapping. */
+   _mesa_DepthRange(0.0, 1.0);
+
    vertices[0][0] = dst_x;
    vertices[0][1] = dst_y;
    vertices[0][2] = dst_z;
@@ -493,25 +505,15 @@ intel_texture_bitmap(GLcontext * ctx,
    vertices[3][2] = dst_z;
    vertices[3][3] = 1.0;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   meta_set_default_texrect(&intel->meta);
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_transform(intel);
-   intel_meta_restore_fragment_program(intel);
-   intel_meta_restore_vertex_program(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
 
    _mesa_PopClientAttrib();
    _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index d50dd68092..5d52335dee 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -272,6 +272,12 @@ do_blit_copypixels(GLcontext * ctx,
    drm_clip_rect_t *cliprects;
    int x_off, y_off;
 
+   if (type == GL_DEPTH || type == GL_STENCIL) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
+      return GL_FALSE;
+   }
+
    /* Update draw buffer bounds */
    _mesa_update_state(ctx);
 
@@ -362,14 +368,16 @@ do_blit_copypixels(GLcontext * ctx,
 				   &clip_x, &clip_y, &clip_w, &clip_h))
             continue;
 
-         intelEmitCopyBlit(intel, dst->cpp,
-			   src->pitch, src->buffer, 0, src->tiling,
-			   dst->pitch, dst->buffer, 0, dst->tiling,
-			   clip_x + delta_x, clip_y + delta_y, /* srcx, srcy */
-			   clip_x, clip_y, /* dstx, dsty */
-			   clip_w, clip_h,
-			   ctx->Color.ColorLogicOpEnabled ?
-			   ctx->Color.LogicOp : GL_COPY);
+	 if (!intel_region_copy(intel,
+				dst, 0, clip_x, clip_y,
+				src, 0, clip_x + delta_x, clip_y + delta_y,
+				clip_w, clip_h,
+				ctx->Color.ColorLogicOpEnabled ?
+				ctx->Color.LogicOp : GL_COPY)) {
+	    DBG("%s: blit failure\n", __FUNCTION__);
+	    UNLOCK_HARDWARE(intel);
+	    return GL_FALSE;
+	 }
       }
    }
 out:
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index d80069dd58..a6b6824164 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -45,6 +45,7 @@
 #include "main/depth.h"
 #include "main/hash.h"
 #include "main/blend.h"
+#include "main/viewport.h"
 #include "glapi/dispatch.h"
 #include "swrast/swrast.h"
 
@@ -70,7 +71,6 @@ intel_texture_drawpixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLuint texname;
    GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
    GLfloat z;
    GLint old_active_texture;
    GLenum internalFormat;
@@ -129,7 +129,7 @@ intel_texture_drawpixels(GLcontext * ctx,
    }
 
    _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
-		    GL_CURRENT_BIT);
+		    GL_CURRENT_BIT | GL_VIEWPORT_BIT);
    _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
 
    /* XXX: pixel store stuff */
@@ -150,11 +150,14 @@ intel_texture_drawpixels(GLcontext * ctx,
    _mesa_TexImage2D(GL_TEXTURE_2D, 0, internalFormat, width, height, 0, format,
 		    type, pixels);
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
    z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
 
+   /* RasterPos[2] already takes into account the DepthRange mapping. */
+   _mesa_DepthRange(0.0, 1.0);
+
    /* Create the vertex buffer based on the current raster pos.  The x and y
     * we're handed are ctx->Current.RasterPos[0,1] rounded to integers.
     * We also apply the depth.  However, the W component is already multiplied
@@ -177,23 +180,14 @@ intel_texture_drawpixels(GLcontext * ctx,
    vertices[3][2] = z;
    vertices[3][3] = 1.0;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   meta_set_default_texrect(&intel->meta);
+
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_transform(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_PopClientAttrib();
@@ -216,14 +210,14 @@ intel_stencil_drawpixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLuint texname, rb_name, fb_name, old_fb_name;
    GLfloat vertices[4][2];
-   GLfloat texcoords[4][2];
    struct intel_renderbuffer *irb;
    struct intel_renderbuffer *depth_irb;
    struct gl_renderbuffer *rb;
    struct gl_pixelstore_attrib old_unpack;
    GLstencil *stencil_pixels;
-   int row;
+   int row, y1, y2;
    GLint old_active_texture;
+   GLboolean rendering_to_fbo = ctx->DrawBuffer->Name != 0;
 
    if (format != GL_STENCIL_INDEX)
       return GL_FALSE;
@@ -358,34 +352,35 @@ intel_stencil_drawpixels(GLcontext * ctx,
    ctx->Unpack = old_unpack;
    _mesa_free(stencil_pixels);
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
+   /* Since we're rendering to the framebuffer as if it was an FBO,
+    * if it's the window system we have to flip the coordinates.
+    */
+   if (rendering_to_fbo) {
+      y1 = y;
+      y2 = y + height * ctx->Pixel.ZoomY;
+   } else {
+      y1 = irb->Base.Height - (y + height * ctx->Pixel.ZoomY);
+      y2 = irb->Base.Height - y;
+   }
    vertices[0][0] = x;
-   vertices[0][1] = y;
+   vertices[0][1] = y1;
    vertices[1][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[1][1] = y;
+   vertices[1][1] = y1;
    vertices[2][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[2][1] = y + height * ctx->Pixel.ZoomY;
+   vertices[2][1] = y2;
    vertices[3][0] = x;
-   vertices[3][1] = y + height * ctx->Pixel.ZoomY;
-
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
+   vertices[3][1] = y2;
 
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
-   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   meta_set_default_texrect(&intel->meta);
+
    _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_transform(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
 
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, old_fb_name);
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c
new file mode 100644
index 0000000000..8713463ace
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c
@@ -0,0 +1,324 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/state.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_blit.h"
+#include "intel_buffers.h"
+#include "intel_regions.h"
+#include "intel_pixel.h"
+#include "intel_buffer_objects.h"
+
+/* For many applications, the new ability to pull the source buffers
+ * back out of the GTT and then do the packing/conversion operations
+ * in software will be as much of an improvement as trying to get the
+ * blitter and/or texture engine to do the work.
+ *
+ * This step is gated on private backbuffers.
+ *
+ * Obviously the frontbuffer can't be pulled back, so that is either
+ * an argument for blit/texture readpixels, or for blitting to a
+ * temporary and then pulling that back.
+ *
+ * When the destination is a pbo, however, it's not clear if it is
+ * ever going to be pulled to main memory (though the access param
+ * will be a good hint).  So it sounds like we do want to be able to
+ * choose between blit/texture implementation on the gpu and pullback
+ * and cpu-based copying.
+ *
+ * Unless you can magically turn client memory into a PBO for the
+ * duration of this call, there will be a cpu-based copying step in
+ * any case.
+ */
+
+
+static GLboolean
+do_texture_readpixels(GLcontext * ctx,
+                      GLint x, GLint y, GLsizei width, GLsizei height,
+                      GLenum format, GLenum type,
+                      const struct gl_pixelstore_attrib *pack,
+                      struct intel_region *dest_region)
+{
+#if 0
+   struct intel_context *intel = intel_context(ctx);
+   intelScreenPrivate *screen = intel->intelScreen;
+   GLint pitch = pack->RowLength ? pack->RowLength : width;
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   int textureFormat;
+   GLenum glTextureFormat;
+   int destFormat, depthFormat, destPitch;
+   drm_clip_rect_t tmp;
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+
+   if (ctx->_ImageTransferState ||
+       pack->SwapBytes || pack->LsbFirst || !pack->Invert) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   intel->vtbl.meta_texrect_source(intel, intel_readbuf_region(intel));
+
+   if (!intel->vtbl.meta_render_dest(intel, dest_region, type, format)) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         fprintf(stderr, "%s: couldn't set dest %s/%s\n",
+                 __FUNCTION__,
+                 _mesa_lookup_enum_by_nr(type),
+                 _mesa_lookup_enum_by_nr(format));
+      return GL_FALSE;
+   }
+
+   LOCK_HARDWARE(intel);
+
+   if (intel->driDrawable->numClipRects) {
+      intel->vtbl.install_meta_state(intel);
+      intel->vtbl.meta_no_depth_write(intel);
+      intel->vtbl.meta_no_stencil_write(intel);
+
+      if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) {
+         UNLOCK_HARDWARE(intel);
+         SET_STATE(i830, state);
+         if (INTEL_DEBUG & DEBUG_PIXEL)
+            fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__);
+         return GL_TRUE;
+      }
+
+      y = dPriv->h - y - height;
+      x += dPriv->x;
+      y += dPriv->y;
+
+
+      /* Set the frontbuffer up as a large rectangular texture.
+       */
+      intel->vtbl.meta_tex_rect_source(intel, src_region, textureFormat);
+
+
+      intel->vtbl.meta_texture_blend_replace(i830, glTextureFormat);
+
+
+      /* Set the 3d engine to draw into the destination region:
+       */
+
+      intel->vtbl.meta_draw_region(intel, dest_region);
+      intel->vtbl.meta_draw_format(intel, destFormat, depthFormat);     /* ?? */
+
+
+      /* Draw a single quad, no cliprects:
+       */
+      intel->vtbl.meta_disable_cliprects(intel);
+
+      intel->vtbl.draw_quad(intel,
+                            0, width, 0, height,
+                            0x00ff00ff, x, x + width, y, y + height);
+
+      intel->vtbl.leave_meta_state(intel);
+   }
+   UNLOCK_HARDWARE(intel);
+
+   intel_region_wait_fence(ctx, dest_region);   /* required by GL */
+   return GL_TRUE;
+#endif
+
+   return GL_FALSE;
+}
+
+
+
+
+static GLboolean
+do_blit_readpixels(GLcontext * ctx,
+                   GLint x, GLint y, GLsizei width, GLsizei height,
+                   GLenum format, GLenum type,
+                   const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_region *src = intel_readbuf_region(intel);
+   struct intel_buffer_object *dst = intel_buffer_object(pack->BufferObj);
+   GLuint dst_offset;
+   GLuint rowLength;
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      _mesa_printf("%s\n", __FUNCTION__);
+
+   if (!src)
+      return GL_FALSE;
+
+   if (pack->BufferObj->Name) {
+      /* XXX This validation should be done by core mesa:
+       */
+      if (!_mesa_validate_pbo_access(2, pack, width, height, 1,
+                                     format, type, pixels)) {
+         _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels");
+         return GL_TRUE;
+      }
+   }
+   else {
+      /* PBO only for now:
+       */
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s - not PBO\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+
+   if (ctx->_ImageTransferState ||
+       !intel_check_blit_format(src, format, type)) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s - bad format for blit\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s: bad packing params\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if (pack->RowLength > 0)
+      rowLength = pack->RowLength;
+   else
+      rowLength = width;
+
+   if (pack->Invert) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+   else {
+      rowLength = -rowLength;
+   }
+
+   /* XXX 64-bit cast? */
+   dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height,
+                                             format, type, 0, 0, 0);
+
+
+   /* Although the blits go on the command buffer, need to do this and
+    * fire with lock held to guarentee cliprects are correct.
+    */
+   intelFlush(&intel->ctx);
+   LOCK_HARDWARE(intel);
+
+   if (intel->driDrawable->numClipRects) {
+      GLboolean all = (width * height * src->cpp == dst->Base.Size &&
+                       x == 0 && dst_offset == 0);
+
+      dri_bo *dst_buffer = intel_bufferobj_buffer(intel, dst,
+						  all ? INTEL_WRITE_FULL :
+						  INTEL_WRITE_PART);
+      __DRIdrawablePrivate *dPriv = intel->driDrawable;
+      int nbox = dPriv->numClipRects;
+      drm_clip_rect_t *box = dPriv->pClipRects;
+      drm_clip_rect_t rect;
+      drm_clip_rect_t src_rect;
+      int i;
+
+      src_rect.x1 = dPriv->x + x;
+      src_rect.y1 = dPriv->y + dPriv->h - (y + height);
+      src_rect.x2 = src_rect.x1 + width;
+      src_rect.y2 = src_rect.y1 + height;
+
+
+
+      for (i = 0; i < nbox; i++) {
+         if (!intel_intersect_cliprects(&rect, &src_rect, &box[i]))
+            continue;
+
+         if (!intelEmitCopyBlit(intel,
+				src->cpp,
+				src->pitch, src->buffer, 0, src->tiling,
+				rowLength, dst_buffer, dst_offset, GL_FALSE,
+				rect.x1,
+				rect.y1,
+				rect.x1 - src_rect.x1,
+				rect.y2 - src_rect.y2,
+				rect.x2 - rect.x1, rect.y2 - rect.y1,
+				GL_COPY)) {
+	    UNLOCK_HARDWARE(intel);
+	    return GL_FALSE;
+	 }
+      }
+   }
+   UNLOCK_HARDWARE(intel);
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      _mesa_printf("%s - DONE\n", __FUNCTION__);
+
+   return GL_TRUE;
+}
+
+void
+intelReadPixels(GLcontext * ctx,
+                GLint x, GLint y, GLsizei width, GLsizei height,
+                GLenum format, GLenum type,
+                const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
+{
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   intelFlush(ctx);
+
+#ifdef I915
+   if (do_blit_readpixels
+       (ctx, x, y, width, height, format, type, pack, pixels))
+      return;
+
+   if (do_texture_readpixels
+       (ctx, x, y, width, height, format, type, pack, pixels))
+      return;
+#else
+   (void)do_blit_readpixels;
+   (void)do_texture_readpixels;
+#endif
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
+
+   /* Update Mesa state before calling down into _swrast_ReadPixels, as
+    * the spans code requires the computed buffer states to be up to date,
+    * but _swrast_ReadPixels only updates Mesa state after setting up
+    * the spans code.
+    */
+
+   if (ctx->NewState)
+      _mesa_update_state(ctx);
+
+   _swrast_ReadPixels(ctx, x, y, width, height, format, type, pack, pixels);
+}
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index 57ac8f0cc1..d19f1bae34 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -189,6 +189,19 @@
 
 #define S7_DEPTH_OFFSET_CONST_MASK     ~0
 
+/* p143 */
+#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
+/* Dword 1 */
+#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
+#define BUF_3D_ID_DEPTH 	(0x7<<24)
+#define BUF_3D_USE_FENCE	(1<<23)
+#define BUF_3D_TILED_SURFACE	(1<<22)
+#define BUF_3D_TILE_WALK_X	0
+#define BUF_3D_TILE_WALK_Y	(1<<21)
+#define BUF_3D_PITCH(x)         (((x)/4)<<2)
+/* Dword 2 */
+#define BUF_3D_ADDR(x)		((x) & ~0x3)
+
 /* Primitive dispatch on 830-945 */
 #define _3DPRIMITIVE			(CMD_3D | (0x1f << 24))
 #define PRIM_INDIRECT            (1<<23)
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index 0aa5b8c02c..7525cd9c4d 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -52,17 +52,77 @@
 
 #define FILE_DEBUG_FLAG DEBUG_REGION
 
+/* This should be set to the maximum backtrace size desired.
+ * Set it to 0 to disable backtrace debugging.
+ */
+#define DEBUG_BACKTRACE_SIZE 0
+
+#if DEBUG_BACKTRACE_SIZE == 0
+/* Use the standard debug output */
+#define _DBG(...) DBG(__VA_ARGS__)
+#else
+/* Use backtracing debug output */
+#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);}
+
+/* Backtracing debug support */
+#include <execinfo.h>
+
+static void
+debug_backtrace(void)
+{
+   void *trace[DEBUG_BACKTRACE_SIZE];
+   char **strings = NULL;
+   int traceSize;
+   register int i;
+
+   traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE);
+   strings = backtrace_symbols(trace, traceSize);
+   if (strings == NULL) {
+      DBG("no backtrace:");
+      return;
+   }
+
+   /* Spit out all the strings with a colon separator.  Ignore
+    * the first, since we don't really care about the call
+    * to debug_backtrace() itself.  Skip until the final "/" in
+    * the trace to avoid really long lines.
+    */
+   for (i = 1; i < traceSize; i++) {
+      char *p = strings[i], *slash = strings[i];
+      while (*p) {
+         if (*p++ == '/') {
+            slash = p;
+         }
+      }
+
+      DBG("%s:", slash);
+   }
+
+   /* Free up the memory, and we're done */
+   free(strings);
+}
+
+#endif
+
+
+
 /* XXX: Thread safety?
  */
 GLubyte *
 intel_region_map(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   intelFlush(&intel->ctx);
+
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!region->map_refcount++) {
       if (region->pbo)
          intel_region_cow(intel, region);
 
-      dri_bo_map(region->buffer, GL_TRUE);
+      if (region->tiling != I915_TILING_NONE &&
+	  intel->intelScreen->kernel_exec_fencing)
+	 drm_intel_gem_bo_map_gtt(region->buffer);
+      else
+	 dri_bo_map(region->buffer, GL_TRUE);
       region->map = region->buffer->virtual;
    }
 
@@ -72,9 +132,13 @@ intel_region_map(struct intel_context *intel, struct intel_region *region)
 void
 intel_region_unmap(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!--region->map_refcount) {
-      dri_bo_unmap(region->buffer);
+      if (region->tiling != I915_TILING_NONE &&
+	  intel->intelScreen->kernel_exec_fencing)
+	 drm_intel_gem_bo_unmap_gtt(region->buffer);
+      else
+	 dri_bo_unmap(region->buffer);
       region->map = NULL;
    }
 }
@@ -87,10 +151,10 @@ intel_region_alloc_internal(struct intel_context *intel,
 {
    struct intel_region *region;
 
-   DBG("%s\n", __FUNCTION__);
-
-   if (buffer == NULL)
+   if (buffer == NULL) {
+      _DBG("%s <-- NULL\n", __FUNCTION__);
       return NULL;
+   }
 
    region = calloc(sizeof(*region), 1);
    region->cpp = cpp;
@@ -104,15 +168,18 @@ intel_region_alloc_internal(struct intel_context *intel,
    region->tiling = I915_TILING_NONE;
    region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE;
 
+   _DBG("%s <-- %p\n", __FUNCTION__, region);
    return region;
 }
 
 struct intel_region *
 intel_region_alloc(struct intel_context *intel,
+		   uint32_t tiling,
                    GLuint cpp, GLuint width, GLuint height, GLuint pitch,
 		   GLboolean expect_accelerated_upload)
 {
    dri_bo *buffer;
+   struct intel_region *region;
 
    if (expect_accelerated_upload) {
       buffer = drm_intel_bo_alloc_for_render(intel->bufmgr, "region",
@@ -122,7 +189,16 @@ intel_region_alloc(struct intel_context *intel,
 				  pitch * cpp * height, 64);
    }
 
-   return intel_region_alloc_internal(intel, cpp, width, height, pitch, buffer);
+   region = intel_region_alloc_internal(intel, cpp, width, height,
+					pitch, buffer);
+
+   if (tiling != I915_TILING_NONE) {
+      assert(((pitch * cpp) & 127) == 0);
+      drm_intel_bo_set_tiling(buffer, &tiling, pitch * cpp);
+      drm_intel_bo_get_tiling(buffer, &region->tiling, &region->bit_6_swizzle);
+   }
+
+   return region;
 }
 
 struct intel_region *
@@ -158,7 +234,7 @@ void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
    if (src)
-      DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
+      _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
 
    assert(*dst == NULL);
    if (src) {
@@ -172,10 +248,12 @@ intel_region_release(struct intel_region **region_handle)
 {
    struct intel_region *region = *region_handle;
 
-   if (region == NULL)
+   if (region == NULL) {
+      _DBG("%s NULL\n", __FUNCTION__);
       return;
+   }
 
-   DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
+   _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
 
    ASSERT(region->refcount > 0);
    region->refcount--;
@@ -249,9 +327,7 @@ intel_region_data(struct intel_context *intel,
                   const void *src, GLuint src_pitch,
                   GLuint srcx, GLuint srcy, GLuint width, GLuint height)
 {
-   GLboolean locked = GL_FALSE;
-
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -264,39 +340,33 @@ intel_region_data(struct intel_context *intel,
          intel_region_cow(intel, dst);
    }
 
-   if (!intel->locked) {
-      LOCK_HARDWARE(intel);
-      locked = GL_TRUE;
-   }
-
+   LOCK_HARDWARE(intel);
    _mesa_copy_rect(intel_region_map(intel, dst) + dst_offset,
                    dst->cpp,
                    dst->pitch,
                    dstx, dsty, width, height, src, src_pitch, srcx, srcy);
 
    intel_region_unmap(intel, dst);
-
-   if (locked)
-      UNLOCK_HARDWARE(intel);
-
+   UNLOCK_HARDWARE(intel);
 }
 
 /* Copy rectangular sub-regions. Need better logic about when to
  * push buffers into AGP - will currently do so whenever possible.
  */
-void
+GLboolean
 intel_region_copy(struct intel_context *intel,
                   struct intel_region *dst,
                   GLuint dst_offset,
                   GLuint dstx, GLuint dsty,
                   struct intel_region *src,
                   GLuint src_offset,
-                  GLuint srcx, GLuint srcy, GLuint width, GLuint height)
+                  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLenum logicop)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
-      return;
+      return GL_FALSE;
 
    if (dst->pbo) {
       if (dstx == 0 &&
@@ -308,41 +378,12 @@ intel_region_copy(struct intel_context *intel,
 
    assert(src->cpp == dst->cpp);
 
-   intelEmitCopyBlit(intel,
-                     dst->cpp,
-                     src->pitch, src->buffer, src_offset, src->tiling,
-                     dst->pitch, dst->buffer, dst_offset, dst->tiling,
-                     srcx, srcy, dstx, dsty, width, height,
-		     GL_COPY);
-}
-
-/* Fill a rectangular sub-region.  Need better logic about when to
- * push buffers into AGP - will currently do so whenever possible.
- */
-void
-intel_region_fill(struct intel_context *intel,
-                  struct intel_region *dst,
-                  GLuint dst_offset,
-                  GLuint dstx, GLuint dsty,
-                  GLuint width, GLuint height, GLuint color)
-{
-   DBG("%s\n", __FUNCTION__);
-
-   if (intel == NULL)
-      return;   
-
-   if (dst->pbo) {
-      if (dstx == 0 &&
-          dsty == 0 && width == dst->pitch && height == dst->height)
-         intel_region_release_pbo(intel, dst);
-      else
-         intel_region_cow(intel, dst);
-   }
-
-   intelEmitFillBlit(intel,
-                     dst->cpp,
-                     dst->pitch, dst->buffer, dst_offset, dst->tiling,
-                     dstx, dsty, width, height, color);
+   return intelEmitCopyBlit(intel,
+			    dst->cpp,
+			    src->pitch, src->buffer, src_offset, src->tiling,
+			    dst->pitch, dst->buffer, dst_offset, dst->tiling,
+			    srcx, srcy, dstx, dsty, width, height,
+			    logicop);
 }
 
 /* Attach to a pbo, discarding our data.  Effectively zero-copy upload
@@ -353,9 +394,13 @@ intel_region_attach_pbo(struct intel_context *intel,
                         struct intel_region *region,
                         struct intel_buffer_object *pbo)
 {
+   dri_bo *buffer;
+
    if (region->pbo == pbo)
       return;
 
+   _DBG("%s %p %p\n", __FUNCTION__, region, pbo);
+
    /* If there is already a pbo attached, break the cow tie now.
     * Don't call intel_region_release_pbo() as that would
     * unnecessarily allocate a new buffer we would have to immediately
@@ -371,10 +416,13 @@ intel_region_attach_pbo(struct intel_context *intel,
       region->buffer = NULL;
    }
 
+   /* make sure pbo has a buffer of its own */
+   buffer = intel_bufferobj_buffer(intel, pbo, INTEL_WRITE_FULL);
+
    region->pbo = pbo;
    region->pbo->region = region;
-   dri_bo_reference(pbo->buffer);
-   region->buffer = pbo->buffer;
+   dri_bo_reference(buffer);
+   region->buffer = buffer;
 }
 
 
@@ -385,6 +433,7 @@ void
 intel_region_release_pbo(struct intel_context *intel,
                          struct intel_region *region)
 {
+   _DBG("%s %p\n", __FUNCTION__, region);
    assert(region->buffer == region->pbo->buffer);
    region->pbo->region = NULL;
    region->pbo = NULL;
@@ -403,34 +452,25 @@ void
 intel_region_cow(struct intel_context *intel, struct intel_region *region)
 {
    struct intel_buffer_object *pbo = region->pbo;
-   GLboolean was_locked = intel->locked;
-
-   if (intel == NULL)
-      return;
 
    intel_region_release_pbo(intel, region);
 
    assert(region->cpp * region->pitch * region->height == pbo->Base.Size);
 
-   DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size);
+   _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size);
 
    /* Now blit from the texture buffer to the new buffer: 
     */
 
-   was_locked = intel->locked;
-   if (!was_locked)
-      LOCK_HARDWARE(intel);
-
-   intelEmitCopyBlit(intel,
-		     region->cpp,
-		     region->pitch, region->buffer, 0, region->tiling,
-		     region->pitch, pbo->buffer, 0, region->tiling,
-		     0, 0, 0, 0,
-		     region->pitch, region->height,
-		     GL_COPY);
-
-   if (!was_locked)
-      UNLOCK_HARDWARE(intel);
+   LOCK_HARDWARE(intel);
+   assert(intelEmitCopyBlit(intel,
+			    region->cpp,
+			    region->pitch, pbo->buffer, 0, region->tiling,
+			    region->pitch, region->buffer, 0, region->tiling,
+			    0, 0, 0, 0,
+			    region->pitch, region->height,
+			    GL_COPY));
+   UNLOCK_HARDWARE(intel);
 }
 
 dri_bo *
@@ -459,6 +499,10 @@ intel_recreate_static(struct intel_context *intel,
    if (region == NULL) {
       region = calloc(sizeof(*region), 1);
       region->refcount = 1;
+      _DBG("%s creating new region %p\n", __FUNCTION__, region);
+   }
+   else {
+      _DBG("%s %p\n", __FUNCTION__, region);
    }
 
    if (intel->ctx.Visual.rgbBits == 24)
diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h
index 45e2bf4e77..0d379bdc6e 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.h
+++ b/src/mesa/drivers/dri/intel/intel_regions.h
@@ -73,7 +73,8 @@ struct intel_region
  * copied by calling intel_reference_region().
  */
 struct intel_region *intel_region_alloc(struct intel_context *intel,
-                                        GLuint cpp, GLuint width,
+                                        uint32_t tiling,
+					GLuint cpp, GLuint width,
                                         GLuint height, GLuint pitch,
 					GLboolean expect_accelerated_upload);
 
@@ -109,21 +110,15 @@ void intel_region_data(struct intel_context *intel,
 
 /* Copy rectangular sub-regions
  */
-void intel_region_copy(struct intel_context *intel,
-                       struct intel_region *dest,
-                       GLuint dest_offset,
-                       GLuint destx, GLuint desty,
-                       struct intel_region *src,
-                       GLuint src_offset,
-                       GLuint srcx, GLuint srcy, GLuint width, GLuint height);
-
-/* Fill a rectangular sub-region
- */
-void intel_region_fill(struct intel_context *intel,
-                       struct intel_region *dest,
-                       GLuint dest_offset,
-                       GLuint destx, GLuint desty,
-                       GLuint width, GLuint height, GLuint color);
+GLboolean
+intel_region_copy(struct intel_context *intel,
+		  struct intel_region *dest,
+		  GLuint dest_offset,
+		  GLuint destx, GLuint desty,
+		  struct intel_region *src,
+		  GLuint src_offset,
+		  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLenum logicop);
 
 /* Helpers for zerocopy uploads, particularly texture image uploads:
  */
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index 0f278b3acd..6bbc995c1e 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -49,6 +49,10 @@
 #include "i915_drm.h"
 #include "i830_dri.h"
 
+#define DRI_CONF_TEXTURE_TILING(def) \
+	DRI_CONF_OPT_BEGIN(texture_tiling, bool, def)		\
+		DRI_CONF_DESC(en, "Enable texture tiling")	\
+	DRI_CONF_OPT_END					\
 
 PUBLIC const char __driConfigOptions[] =
    DRI_CONF_BEGIN
@@ -64,6 +68,13 @@ PUBLIC const char __driConfigOptions[] =
 	    DRI_CONF_ENUM(1, "Enable reuse of all sizes of buffer objects")
 	 DRI_CONF_DESC_END
       DRI_CONF_OPT_END
+
+      DRI_CONF_TEXTURE_TILING(false)
+
+      DRI_CONF_OPT_BEGIN(early_z, bool, false)
+	 DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).")
+      DRI_CONF_OPT_END
+
    DRI_CONF_SECTION_END
    DRI_CONF_SECTION_QUALITY
       DRI_CONF_FORCE_S3TC_ENABLE(false)
@@ -76,7 +87,7 @@ PUBLIC const char __driConfigOptions[] =
    DRI_CONF_SECTION_END
 DRI_CONF_END;
 
-const GLuint __driNConfigOptions = 8;
+const GLuint __driNConfigOptions = 10;
 
 #ifdef USE_NEW_INTERFACE
 static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
@@ -305,7 +316,7 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv)
 
    dri_bufmgr_destroy(intelScreen->bufmgr);
    intelUnmapScreenRegions(intelScreen);
-   driDestroyOptionCache(&intelScreen->optionCache);
+   driDestroyOptionInfo(&intelScreen->optionCache);
 
    FREE(intelScreen);
    sPriv->private = NULL;
@@ -587,7 +598,7 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen)
    GLboolean gem_supported;
    struct drm_i915_getparam gp;
    __DRIscreenPrivate *spriv = intelScreen->driScrnPriv;
-   int num_fences;
+   int num_fences = 0;
 
    intelScreen->no_hw = getenv("INTEL_NO_HW") != NULL;
 
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index ae0994b183..df63f29a42 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -158,81 +158,11 @@ timed_memcpy(void *dest, const void *src, size_t n)
 }
 #endif /* DO_DEBUG */
 
-/**
- * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
- * level).
- *
- * The texture object's miptree must be mapped.
- *
- * It would be really nice if this was just called by Mesa whenever mipmaps
- * needed to be regenerated, rather than us having to remember to do so in
- * each texture image modification path.
- *
- * This function should also include an accelerated path.
- */
-void
-intel_generate_mipmap(GLcontext *ctx, GLenum target,
-                      struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
-   int face, i;
-
-   _mesa_generate_mipmap(ctx, target, texObj);
-
-   /* Update the level information in our private data in the new images, since
-    * it didn't get set as part of a normal TexImage path.
-    */
-   for (face = 0; face < nr_faces; face++) {
-      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
-         struct intel_texture_image *intelImage;
-
-	 intelImage = intel_texture_image(texObj->Image[face][i]);
-	 if (intelImage == NULL)
-	    break;
-
-	 intelImage->level = i;
-	 intelImage->face = face;
-	 /* Unreference the miptree to signal that the new Data is a bare
-	  * pointer from mesa.
-	  */
-	 intel_miptree_release(intel, &intelImage->mt);
-      }
-   }
-}
-
-static void intelGenerateMipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-
-   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
-   intel_generate_mipmap(ctx, target, texObj);
-   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
-}
-
 void
 intelInitTextureFuncs(struct dd_function_table *functions)
 {
    functions->ChooseTextureFormat = intelChooseTextureFormat;
-   functions->TexImage1D = intelTexImage1D;
-   functions->TexImage2D = intelTexImage2D;
-   functions->TexImage3D = intelTexImage3D;
-   functions->TexSubImage1D = intelTexSubImage1D;
-   functions->TexSubImage2D = intelTexSubImage2D;
-   functions->TexSubImage3D = intelTexSubImage3D;
-   functions->CopyTexImage1D = intelCopyTexImage1D;
-   functions->CopyTexImage2D = intelCopyTexImage2D;
-   functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
-   functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
-   functions->GetTexImage = intelGetTexImage;
-   functions->GenerateMipmap = intelGenerateMipmap;
-
-   /* compressed texture functions */
-   functions->CompressedTexImage2D = intelCompressedTexImage2D;
-   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
-   functions->GetCompressedTexImage = intelGetCompressedTexImage;
+   functions->GenerateMipmap = intel_generate_mipmap;
 
    functions->NewTextureObject = intelNewTextureObject;
    functions->NewTextureImage = intelNewTextureImage;
diff --git a/src/mesa/drivers/dri/intel/intel_tex.h b/src/mesa/drivers/dri/intel/intel_tex.h
index f5372d82fb..471aa2a240 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.h
+++ b/src/mesa/drivers/dri/intel/intel_tex.h
@@ -35,116 +35,17 @@
 
 void intelInitTextureFuncs(struct dd_function_table *functions);
 
+void intelInitTextureImageFuncs(struct dd_function_table *functions);
+
+void intelInitTextureSubImageFuncs(struct dd_function_table *functions);
+
+void intelInitTextureCopyImageFuncs(struct dd_function_table *functions);
+
 const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx,
                                                          GLint internalFormat,
                                                          GLenum format,
                                                          GLenum type);
 
-
-void intelTexImage3D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint height, GLint depth,
-                     GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage3D(GLcontext * ctx,
-                        GLenum target,
-                        GLint level,
-                        GLint xoffset, GLint yoffset, GLint zoffset,
-                        GLsizei width, GLsizei height, GLsizei depth,
-                        GLenum format, GLenum type,
-                        const GLvoid * pixels,
-                        const struct gl_pixelstore_attrib *packing,
-                        struct gl_texture_object *texObj,
-                        struct gl_texture_image *texImage);
-
-void intelTexImage2D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint height, GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage2D(GLcontext * ctx,
-                        GLenum target,
-                        GLint level,
-                        GLint xoffset, GLint yoffset,
-                        GLsizei width, GLsizei height,
-                        GLenum format, GLenum type,
-                        const GLvoid * pixels,
-                        const struct gl_pixelstore_attrib *packing,
-                        struct gl_texture_object *texObj,
-                        struct gl_texture_image *texImage);
-
-void intelTexImage1D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage1D(GLcontext * ctx,
-                        GLenum target,
-                        GLint level,
-                        GLint xoffset,
-                        GLsizei width,
-                        GLenum format, GLenum type,
-                        const GLvoid * pixels,
-                        const struct gl_pixelstore_attrib *packing,
-                        struct gl_texture_object *texObj,
-                        struct gl_texture_image *texImage);
-
-void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
-                         GLenum internalFormat,
-                         GLint x, GLint y, GLsizei width, GLint border);
-
-void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
-                         GLenum internalFormat,
-                         GLint x, GLint y, GLsizei width, GLsizei height,
-                         GLint border);
-
-void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
-                            GLint xoffset, GLint x, GLint y, GLsizei width);
-
-void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
-                            GLint xoffset, GLint yoffset,
-                            GLint x, GLint y, GLsizei width, GLsizei height);
-
-void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level,
-                      GLenum format, GLenum type, GLvoid * pixels,
-                      struct gl_texture_object *texObj,
-                      struct gl_texture_image *texImage);
-
-void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-				GLint internalFormat,
-				GLint width, GLint height, GLint border,
-				GLsizei imageSize, const GLvoid *data,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage );
-
-void intelCompressedTexSubImage2D(GLcontext * ctx,
-				  GLenum target,
-				  GLint level,
-				  GLint xoffset, GLint yoffset,
-				  GLsizei width, GLsizei height,
-				  GLenum format, GLsizei imageSize,
-				  const GLvoid * pixels,
-				  struct gl_texture_object *texObj,
-				  struct gl_texture_image *texImage);
-
-void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
-				GLvoid *pixels,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage);
-
 void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 		       unsigned long long offset, GLint depth, GLuint pitch);
 void intelSetTexBuffer(__DRIcontext *pDRICtx,
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 90bbb8c6bb..028b49c14d 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -73,11 +73,8 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat)
       return NULL;
    case GL_RGBA:
    case GL_RGBA8:
-      return intel_readbuf_region(intel);
    case GL_RGB:
-      if (intel->ctx.Visual.rgbBits == 16)
-         return intel_readbuf_region(intel);
-      return NULL;
+      return intel_readbuf_region(intel);
    default:
       return NULL;
    }
@@ -99,14 +96,24 @@ do_copy_texsubimage(struct intel_context *intel,
 
    if (!intelImage->mt || !src) {
       if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s fail %p %p\n",
-		 __FUNCTION__, intelImage->mt, src);
+	 fprintf(stderr, "%s fail %p %p (0x%08x)\n",
+		 __FUNCTION__, intelImage->mt, src, internalFormat);
+      return GL_FALSE;
+   }
+
+   if (intelImage->mt->cpp != src->cpp) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "%s fail %d vs %d cpp\n",
+		 __FUNCTION__, intelImage->mt->cpp, src->cpp);
       return GL_FALSE;
    }
 
    intelFlush(ctx);
    LOCK_HARDWARE(intel);
    {
+      drm_intel_bo *dst_bo = intel_region_buffer(intel,
+						 intelImage->mt->region,
+						 INTEL_WRITE_PART);
       GLuint image_offset = intel_miptree_image_offset(intelImage->mt,
                                                        intelImage->face,
                                                        intelImage->level);
@@ -118,8 +125,12 @@ do_copy_texsubimage(struct intel_context *intel,
       dstx += x - orig_x;
       dsty += y - orig_y;
 
-      /* image_offset may be non-page-aligned, but that's illegal for tiling. */
-      assert(intelImage->mt->region->tiling == I915_TILING_NONE);
+      /* Can't blit to tiled buffers with non-tile-aligned offset. */
+      if (intelImage->mt->region->tiling != I915_TILING_NONE &&
+	  (image_offset & 4095) != 0) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
 
       if (ctx->ReadBuffer->Name == 0) {
 	 /* reading from a window, adjust x, y */
@@ -140,35 +151,35 @@ do_copy_texsubimage(struct intel_context *intel,
 	 src_pitch = src->pitch;
       }
 
-      intelEmitCopyBlit(intel,
-			intelImage->mt->cpp,
-			src_pitch,
-			src->buffer,
-			0,
-			src->tiling,
-			intelImage->mt->pitch,
-			intelImage->mt->region->buffer,
-			image_offset,
-			intelImage->mt->region->tiling,
-			x, y, dstx, dsty, width, height,
-			GL_COPY);
+      if (!intelEmitCopyBlit(intel,
+			     intelImage->mt->cpp,
+			     src_pitch,
+			     src->buffer,
+			     0,
+			     src->tiling,
+			     intelImage->mt->pitch,
+			     dst_bo,
+			     image_offset,
+			     intelImage->mt->region->tiling,
+			     x, y, dstx, dsty, width, height,
+			     GL_COPY)) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
    }
 
    UNLOCK_HARDWARE(intel);
 
    /* GL_SGIS_generate_mipmap */
    if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+      intel_generate_mipmap(ctx, target, texObj);
    }
 
    return GL_TRUE;
 }
 
 
-
-
-
-void
+static void
 intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                     GLenum internalFormat,
                     GLint x, GLint y, GLsizei width, GLint border)
@@ -214,7 +225,8 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                            width, border);
 }
 
-void
+
+static void
 intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
                     GLenum internalFormat,
                     GLint x, GLint y, GLsizei width, GLsizei height,
@@ -262,7 +274,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
 }
 
 
-void
+static void
 intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
                        GLint xoffset, GLint x, GLint y, GLsizei width)
 {
@@ -287,8 +299,7 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
 }
 
 
-
-void
+static void
 intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
                        GLint xoffset, GLint yoffset,
                        GLint x, GLint y, GLsizei width, GLsizei height)
@@ -301,7 +312,6 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
       _mesa_select_tex_image(ctx, texObj, target, level);
    GLenum internalFormat = texImage->InternalFormat;
 
-
    /* Need to check texture is compatible with source format. 
     */
 
@@ -316,3 +326,13 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
                                  xoffset, yoffset, x, y, width, height);
    }
 }
+
+
+void
+intelInitTextureCopyImageFuncs(struct dd_function_table *functions)
+{
+   functions->CopyTexImage1D = intelCopyTexImage1D;
+   functions->CopyTexImage2D = intelCopyTexImage2D;
+   functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
+   functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index 5e61e9e95e..c5f5220837 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -131,6 +131,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel,
       comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat);
    intelObj->mt = intel_miptree_create(intel,
                                        intelObj->base.Target,
+                                       intelImage->base._BaseFormat,
                                        intelImage->base.InternalFormat,
                                        firstLevel,
                                        lastLevel,
@@ -205,7 +206,7 @@ try_pbo_upload(struct intel_context *intel,
    GLuint src_offset, src_stride;
    GLuint dst_offset, dst_stride;
 
-   if (!pbo ||
+   if (unpack->BufferObj->Name == 0 ||
        intel->ctx._ImageTransferState ||
        unpack->SkipPixels || unpack->SkipRows) {
       DBG("%s: failure 1\n", __FUNCTION__);
@@ -235,12 +236,15 @@ try_pbo_upload(struct intel_context *intel,
 					       INTEL_WRITE_FULL);
 
 
-      intelEmitCopyBlit(intel,
-                        intelImage->mt->cpp,
-                        src_stride, src_buffer, src_offset, GL_FALSE,
-                        dst_stride, dst_buffer, dst_offset, GL_FALSE,
-                        0, 0, 0, 0, width, height,
-			GL_COPY);
+      if (!intelEmitCopyBlit(intel,
+			     intelImage->mt->cpp,
+			     src_stride, src_buffer, src_offset, GL_FALSE,
+			     dst_stride, dst_buffer, dst_offset, GL_FALSE,
+			     0, 0, 0, 0, width, height,
+			     GL_COPY)) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
    }
    UNLOCK_HARDWARE(intel);
 
@@ -248,7 +252,6 @@ try_pbo_upload(struct intel_context *intel,
 }
 
 
-
 static GLboolean
 try_pbo_zcopy(struct intel_context *intel,
               struct intel_texture_image *intelImage,
@@ -261,7 +264,7 @@ try_pbo_zcopy(struct intel_context *intel,
    GLuint src_offset, src_stride;
    GLuint dst_offset, dst_stride;
 
-   if (!pbo ||
+   if (unpack->BufferObj->Name == 0 ||
        intel->ctx._ImageTransferState ||
        unpack->SkipPixels || unpack->SkipRows) {
       DBG("%s: failure 1\n", __FUNCTION__);
@@ -293,10 +296,6 @@ try_pbo_zcopy(struct intel_context *intel,
 }
 
 
-
-
-
-
 static void
 intelTexImage(GLcontext * ctx,
               GLint dims,
@@ -307,7 +306,8 @@ intelTexImage(GLcontext * ctx,
               GLenum format, GLenum type, const void *pixels,
               const struct gl_pixelstore_attrib *unpack,
               struct gl_texture_object *texObj,
-              struct gl_texture_image *texImage, GLsizei imageSize, int compressed)
+              struct gl_texture_image *texImage, GLsizei imageSize,
+              GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_object *intelObj = intel_texture_object(texObj);
@@ -316,7 +316,6 @@ intelTexImage(GLcontext * ctx,
    GLint postConvHeight = height;
    GLint texelBytes, sizeInBytes;
    GLuint dstRowStride = 0, srcRowStride = texImage->RowStride;
-   GLboolean needs_map;
 
    DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
@@ -414,7 +413,9 @@ intelTexImage(GLcontext * ctx,
        * a miptree, so create one just for our level and store it in the image.
        * It'll get moved into the object miptree at validate time.
        */
-      intelImage->mt = intel_miptree_create(intel, target, internalFormat,
+      intelImage->mt = intel_miptree_create(intel, target,
+					    intelImage->base.TexFormat->BaseFormat,
+					    internalFormat,
 					    level, level,
 					    width, height, depth,
 					    intelImage->base.TexFormat->TexelBytes,
@@ -426,7 +427,7 @@ intelTexImage(GLcontext * ctx,
     */
    if (dims <= 2 &&
        intelImage->mt &&
-       intel_buffer_object(unpack->BufferObj) &&
+       unpack->BufferObj->Name != 0 &&
        check_pbo_format(internalFormat, format,
                         type, intelImage->base.TexFormat)) {
 
@@ -464,8 +465,6 @@ intelTexImage(GLcontext * ctx,
       DBG("pbo upload failed\n");
    }
 
-
-
    /* intelCopyTexImage calls this function with pixels == NULL, with
     * the expectation that the mipmap tree will be set up but nothing
     * more will be done.  This is where those calls return:
@@ -482,15 +481,8 @@ intelTexImage(GLcontext * ctx,
 
    LOCK_HARDWARE(intel);
 
-   /* Two cases where we need a mapping of the miptree: when the user supplied
-    * data is mapped as well (non-PBO, memcpy upload) or when we're going to do
-    * (software) mipmap generation.
-    */
-   needs_map = (pixels != NULL) || (level == texObj->BaseLevel &&
-				  texObj->GenerateMipmap);
-
    if (intelImage->mt) {
-      if (needs_map)
+      if (pixels != NULL)
          texImage->Data = intel_miptree_image_map(intel,
                                                   intelImage->mt,
                                                   intelImage->face,
@@ -547,25 +539,26 @@ intelTexImage(GLcontext * ctx,
 						   format, type, pixels, unpack)) {
 	   _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
        }
-
-       /* GL_SGIS_generate_mipmap */
-       if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-	  intel_generate_mipmap(ctx, target, texObj);
-       }
    }
 
    _mesa_unmap_teximage_pbo(ctx, unpack);
 
    if (intelImage->mt) {
-      if (needs_map)
+      if (pixels != NULL)
          intel_miptree_image_unmap(intel, intelImage->mt);
       texImage->Data = NULL;
    }
 
    UNLOCK_HARDWARE(intel);
+
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
 }
 
-void
+
+static void
 intelTexImage3D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -578,11 +571,11 @@ intelTexImage3D(GLcontext * ctx,
 {
    intelTexImage(ctx, 3, target, level,
                  internalFormat, width, height, depth, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
 
-void
+static void
 intelTexImage2D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -594,10 +587,11 @@ intelTexImage2D(GLcontext * ctx,
 {
    intelTexImage(ctx, 2, target, level,
                  internalFormat, width, height, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
-void
+
+static void
 intelTexImage1D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -609,21 +603,24 @@ intelTexImage1D(GLcontext * ctx,
 {
    intelTexImage(ctx, 1, target, level,
                  internalFormat, width, 1, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
-void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-				GLint internalFormat,
-				GLint width, GLint height, GLint border,
-				GLsizei imageSize, const GLvoid *data,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage )
+
+static void
+intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
+                           GLint internalFormat,
+                           GLint width, GLint height, GLint border,
+                           GLsizei imageSize, const GLvoid *data,
+                           struct gl_texture_object *texObj,
+                           struct gl_texture_image *texImage )
 {
    intelTexImage(ctx, 2, target, level,
 		 internalFormat, width, height, 1, border,
-		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1);
+		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, GL_TRUE);
 }
 
+
 /**
  * Need to map texture image into memory before copying image data,
  * then unmap it.
@@ -632,7 +629,7 @@ static void
 intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
 		    GLenum format, GLenum type, GLvoid * pixels,
 		    struct gl_texture_object *texObj,
-		    struct gl_texture_image *texImage, int compressed)
+		    struct gl_texture_image *texImage, GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_image *intelImage = intel_texture_image(texImage);
@@ -686,28 +683,29 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
    }
 }
 
-void
+
+static void
 intelGetTexImage(GLcontext * ctx, GLenum target, GLint level,
                  GLenum format, GLenum type, GLvoid * pixels,
                  struct gl_texture_object *texObj,
                  struct gl_texture_image *texImage)
 {
    intel_get_tex_image(ctx, target, level, format, type, pixels,
-		       texObj, texImage, 0);
-
-
+		       texObj, texImage, GL_FALSE);
 }
 
-void
+
+static void
 intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
 			   GLvoid *pixels,
 			   struct gl_texture_object *texObj,
 			   struct gl_texture_image *texImage)
 {
    intel_get_tex_image(ctx, target, level, 0, 0, pixels,
-		       texObj, texImage, 1);
+		       texObj, texImage, GL_TRUE);
 }
 
+
 void
 intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 		  unsigned long long offset, GLint depth, GLuint pitch)
@@ -816,3 +814,16 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
     */
    intelSetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
 }
+
+
+void
+intelInitTextureImageFuncs(struct dd_function_table *functions)
+{
+   functions->TexImage1D = intelTexImage1D;
+   functions->TexImage2D = intelTexImage2D;
+   functions->TexImage3D = intelTexImage3D;
+   functions->GetTexImage = intelGetTexImage;
+
+   functions->CompressedTexImage2D = intelCompressedTexImage2D;
+   functions->GetCompressedTexImage = intelGetCompressedTexImage;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c
index e6f9a41779..7d69ea4484 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c
@@ -35,26 +35,39 @@
 #include "intel_context.h"
 #include "main/macros.h"
 
-GLuint intel_compressed_alignment(GLenum internalFormat)
+void intel_get_texture_alignment_unit(GLenum internalFormat, GLuint *w, GLuint *h)
 {
-    GLuint alignment = 4;
-
     switch (internalFormat) {
     case GL_COMPRESSED_RGB_FXT1_3DFX:
     case GL_COMPRESSED_RGBA_FXT1_3DFX:
-        alignment = 8;
+        *w = 8;
+        *h = 4;
+        break;
+
+    case GL_RGB_S3TC:
+    case GL_RGB4_S3TC:
+    case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+    case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+    case GL_RGBA_S3TC:
+    case GL_RGBA4_S3TC:
+    case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+    case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+        *w = 4;
+        *h = 4;
         break;
 
     default:
+        *w = 4;
+        *h = 2;
         break;
     }
-
-    return alignment;
 }
 
-void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt )
+void i945_miptree_layout_2d( struct intel_context *intel,
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling )
 {
-   GLint align_h = 2, align_w = 4;
+   GLuint align_h = 2, align_w = 4;
    GLuint level;
    GLuint x = 0;
    GLuint y = 0;
@@ -62,9 +75,9 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr
    GLuint height = mt->height0;
 
    mt->pitch = mt->width0;
+   intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
 
    if (mt->compressed) {
-       align_w = intel_compressed_alignment(mt->internal_format);
        mt->pitch = ALIGN(mt->width0, align_w);
    }
 
@@ -92,7 +105,7 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr
    /* Pitch must be a whole number of dwords, even though we
     * express it in texels.
     */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->pitch);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->pitch);
    mt->total_height = 0;
 
    for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h
index dbc90e6f9b..c9de9b5678 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.h
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h
@@ -38,5 +38,7 @@ static GLuint minify( GLuint d )
    return MAX2(1, d>>1);
 }
 
-extern void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt );
-extern GLuint intel_compressed_alignment(GLenum);
+extern void i945_miptree_layout_2d(struct intel_context *intel,
+				   struct intel_mipmap_tree *mt,
+				   uint32_t tiling);
+extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *);
diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
index f86de56897..1f27131dac 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
@@ -101,11 +101,6 @@ intelTexSubimage(GLcontext * ctx,
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage");
    }
 
-   /* GL_SGIS_generate_mipmap */
-   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      intel_generate_mipmap(ctx, target, texObj);
-   }
-
    _mesa_unmap_teximage_pbo(ctx, packing);
 
    if (intelImage->mt) {
@@ -114,13 +109,15 @@ intelTexSubimage(GLcontext * ctx,
    }
 
    UNLOCK_HARDWARE(intel);
-}
-
-
 
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
+}
 
 
-void
+static void
 intelTexSubImage3D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -132,18 +129,15 @@ intelTexSubImage3D(GLcontext * ctx,
                    struct gl_texture_object *texObj,
                    struct gl_texture_image *texImage)
 {
-
    intelTexSubimage(ctx, 3,
                     target, level,
                     xoffset, yoffset, zoffset,
                     width, height, depth,
                     format, type, pixels, packing, texObj, texImage);
-
 }
 
 
-
-void
+static void
 intelTexSubImage2D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -155,17 +149,15 @@ intelTexSubImage2D(GLcontext * ctx,
                    struct gl_texture_object *texObj,
                    struct gl_texture_image *texImage)
 {
-
    intelTexSubimage(ctx, 2,
                     target, level,
                     xoffset, yoffset, 0,
                     width, height, 1,
                     format, type, pixels, packing, texObj, texImage);
-
 }
 
 
-void
+static void
 intelTexSubImage1D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -182,10 +174,9 @@ intelTexSubImage1D(GLcontext * ctx,
                     xoffset, 0, 0,
                     width, 1, 1,
                     format, type, pixels, packing, texObj, texImage);
-
 }
 
-void
+static void
 intelCompressedTexSubImage2D(GLcontext * ctx,
 			     GLenum target,
 			     GLint level,
@@ -199,3 +190,14 @@ intelCompressedTexSubImage2D(GLcontext * ctx,
    fprintf(stderr, "stubbed CompressedTexSubImage2D: %dx%d@%dx%d\n",
 	   width, height, xoffset, yoffset);
 }
+
+
+
+void
+intelInitTextureSubImageFuncs(struct dd_function_table *functions)
+{
+   functions->TexSubImage1D = intelTexSubImage1D;
+   functions->TexSubImage2D = intelTexSubImage2D;
+   functions->TexSubImage3D = intelTexSubImage3D;
+   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_validate.c b/src/mesa/drivers/dri/intel/intel_tex_validate.c
index 05a375e1f3..a284d5475f 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_validate.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_validate.c
@@ -199,6 +199,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
    if (!intelObj->mt) {
       intelObj->mt = intel_miptree_create(intel,
                                           intelObj->base.Target,
+                                          firstImage->base._BaseFormat,
                                           firstImage->base.InternalFormat,
                                           intelObj->firstLevel,
                                           intelObj->lastLevel,
@@ -241,7 +242,7 @@ intel_tex_map_level_images(struct intel_context *intel,
       struct intel_texture_image *intelImage =
 	 intel_texture_image(intelObj->base.Image[face][level]);
 
-      if (intelImage->mt) {
+      if (intelImage && intelImage->mt) {
 	 intelImage->base.Data =
 	    intel_miptree_image_map(intel,
 				    intelImage->mt,
@@ -268,7 +269,7 @@ intel_tex_unmap_level_images(struct intel_context *intel,
       struct intel_texture_image *intelImage =
 	 intel_texture_image(intelObj->base.Image[face][level]);
 
-      if (intelImage->mt) {
+      if (intelImage && intelImage->mt) {
 	 intel_miptree_image_unmap(intel, intelImage->mt);
 	 intelImage->base.Data = NULL;
       }
diff --git a/src/mesa/drivers/dri/r200/.gitignore b/src/mesa/drivers/dri/r200/.gitignore
index 3773d8ea73..2f9cd1a987 100644
--- a/src/mesa/drivers/dri/r200/.gitignore
+++ b/src/mesa/drivers/dri/r200/.gitignore
@@ -1,3 +1,15 @@
+radeon_bocs_wrapper.h
+radeon_bo_legacy.[ch]
 radeon_chipset.h
-radeon_screen.*
+radeon_cmdbuf.h
+radeon_common.[ch]
+radeon_common_context.[ch]
+radeon_cs_legacy.[ch]
+radeon_dma.[ch]
+radeon_fbo.c
+radeon_lock.[ch]
+radeon_mipmap_tree.[ch]
+radeon_screen.[ch]
+radeon_span.[ch]
+radeon_texture.[ch]
 server
diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile
index e9144ac75c..4686241957 100644
--- a/src/mesa/drivers/dri/r200/Makefile
+++ b/src/mesa/drivers/dri/r200/Makefile
@@ -3,6 +3,8 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
+CFLAGS += $(RADEON_CFLAGS)
+
 LIBNAME = r200_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c 
@@ -11,25 +13,41 @@ ifeq ($(USING_EGL), 1)
 EGL_SOURCES = server/radeon_egl.c
 endif
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
+RADEON_COMMON_SOURCES = \
+	radeon_texture.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_dma.c \
+	radeon_lock.c \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_mipmap_tree.c \
+	radeon_span.c \
+	radeon_fbo.c
+
+
 DRIVER_SOURCES = r200_context.c \
 		 r200_ioctl.c \
-		 r200_lock.c \
 		 r200_state.c \
 		 r200_state_init.c \
 		 r200_cmdbuf.c \
 		 r200_pixel.c \
 		 r200_tex.c \
-		 r200_texmem.c \
 		 r200_texstate.c \
 		 r200_tcl.c \
 		 r200_swtcl.c \
-		 r200_span.c \
 		 r200_maos.c \
 		 r200_sanity.c \
 		 r200_fragshader.c \
 		 r200_vertprog.c \
 		 radeon_screen.c \
-		 $(EGL_SOURCES)
+		 $(EGL_SOURCES) \
+		 $(RADEON_COMMON_SOURCES) \
+		 $(CS_SOURCES)
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
@@ -48,7 +66,31 @@ SYMLINKS = \
 COMMON_SYMLINKS = \
 	radeon_chipset.h \
 	radeon_screen.c \
-	radeon_screen.h
+	radeon_screen.h \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_bo_legacy.h \
+	radeon_cs_legacy.h \
+	radeon_bocs_wrapper.h \
+	radeon_span.h \
+	radeon_span.c \
+	radeon_lock.c \
+	radeon_lock.h \
+	radeon_common.c \
+	radeon_common_context.c \
+	radeon_common_context.h \
+	radeon_common.h \
+	radeon_cmdbuf.h \
+	radeon_mipmap_tree.c \
+	radeon_mipmap_tree.h \
+	radeon_texture.c \
+	radeon_texture.h \
+	radeon_dma.c \
+	radeon_dma.h \
+	radeon_fbo.c \
+	$(CS_SOURCES)
+
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
 ##### TARGETS #####
 
diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
index e1633772a1..d49f4fabe7 100644
--- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c
+++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
@@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "main/simple_list.h"
 
+#include "radeon_common.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -45,18 +46,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_sanity.h"
 #include "radeon_reg.h"
 
-static void print_state_atom( struct r200_state_atom *state )
-{
-   int i;
-
-   fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size);
-
-   if (0 & R200_DEBUG & DEBUG_VERBOSE) 
-      for (i = 0 ; i < state->cmd_size ; i++) 
-	 fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]);
-
-}
-
 /* The state atoms will be emitted in the order they appear in the atom list,
  * so this step is important.
  */
@@ -64,141 +53,85 @@ void r200SetUpAtomList( r200ContextPtr rmesa )
 {
    int i, mtu;
 
-   mtu = rmesa->glCtx->Const.MaxTextureUnits;
-
-   make_empty_list(&rmesa->hw.atomlist);
-   rmesa->hw.atomlist.name = "atom-list";
-
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ctx );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.set );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lin );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msk );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpt );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vtx );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vap );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vte );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msc );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cst );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.zbs );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcl );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msl );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcg );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.grd );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.fog );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tam );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tf );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.atf );
+   mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits;
+
+   make_empty_list(&rmesa->radeon.hw.atomlist);
+   rmesa->radeon.hw.atomlist.name = "atom-list";
+
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ctx );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.set );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.lin );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msk );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpt );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vtx );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vap );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vte );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msc );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.cst );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.zbs );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcl );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msl );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcg );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.grd );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.fog );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tam );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tf );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.atf );
    for (i = 0; i < mtu; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tex[i] );
+       insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i] );
    for (i = 0; i < mtu; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cube[i] );
+       insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i] );
    for (i = 0; i < 6; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pix[i] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[0] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[1] );
+       insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.pix[i] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[0] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[1] );
    for (i = 0; i < 8; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lit[i] );
+       insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i] );
    for (i = 0; i < 3 + mtu; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mat[i] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.eye );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.glt );
+       insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.eye );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.glt );
    for (i = 0; i < 2; ++i)
-      insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mtl[i] );
+      insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.mtl[i] );
    for (i = 0; i < 6; ++i)
-       insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ucp[i] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.spr );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ptp );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.prf );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pvs );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[0] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[1] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[0] );
-   insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[1] );
-}
-
-static void r200SaveHwState( r200ContextPtr rmesa )
-{
-   struct r200_state_atom *atom;
-   char * dest = rmesa->backup_store.cmd_buf;
-
-   if (R200_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   rmesa->backup_store.cmd_used = 0;
-
-   foreach( atom, &rmesa->hw.atomlist ) {
-      if ( atom->check( rmesa->glCtx, atom->idx ) ) {
-	 int size = atom->cmd_size * 4;
-	 memcpy( dest, atom->cmd, size);
-	 dest += size;
-	 rmesa->backup_store.cmd_used += size;
-	 if (R200_DEBUG & DEBUG_STATE)
-	    print_state_atom( atom );
-      }
-   }
-
-   assert( rmesa->backup_store.cmd_used <= R200_CMD_BUF_SZ );
-   if (R200_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Returning to r200EmitState\n");
+       insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.spr );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ptp );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.prf );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.pvs );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[0] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[1] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[0] );
+   insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[1] );
 }
 
-void r200EmitState( r200ContextPtr rmesa )
+void r200EmitScissor(r200ContextPtr rmesa)
 {
-   char *dest;
-   int mtu;
-   struct r200_state_atom *atom;
-
-   if (R200_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->save_on_next_emit) {
-      r200SaveHwState(rmesa);
-      rmesa->save_on_next_emit = GL_FALSE;
-   }
-
-   if (!rmesa->hw.is_dirty && !rmesa->hw.all_dirty)
-      return;
-
-   mtu = rmesa->glCtx->Const.MaxTextureUnits;
-
-   /* To avoid going across the entire set of states multiple times, just check
-    * for enough space for the case of emitting all state, and inline the
-    * r200AllocCmdBuf code here without all the checks.
-    */
-   r200EnsureCmdBufSpace( rmesa, rmesa->hw.max_state_size );
-
-   /* we need to calculate dest after EnsureCmdBufSpace
-      as we may flush the buffer - airlied */
-   dest = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-   if (R200_DEBUG & DEBUG_STATE) {
-      foreach( atom, &rmesa->hw.atomlist ) {
-	 if ( atom->dirty || rmesa->hw.all_dirty ) {
-	    if ( atom->check( rmesa->glCtx, atom->idx ) )
-	       print_state_atom( atom );
-	    else
-	       fprintf(stderr, "skip state %s\n", atom->name);
-	 }
-      }
-   }
-
-   foreach( atom, &rmesa->hw.atomlist ) {
-      if ( rmesa->hw.all_dirty )
-	 atom->dirty = GL_TRUE;
-      if ( atom->dirty ) {
-	 if ( atom->check( rmesa->glCtx, atom->idx ) ) {
-	    int size = atom->cmd_size * 4;
-	    memcpy( dest, atom->cmd, size);
-	    dest += size;
-	    rmesa->store.cmd_used += size;
-	    atom->dirty = GL_FALSE;
-	 }
-      }
-   }
-
-   assert( rmesa->store.cmd_used <= R200_CMD_BUF_SZ );
-
-   rmesa->hw.is_dirty = GL_FALSE;
-   rmesa->hw.all_dirty = GL_FALSE;
+    BATCH_LOCALS(&rmesa->radeon);
+    if (!rmesa->radeon.radeonScreen->kernel_mm) {
+       return;
+    }
+    if (rmesa->radeon.state.scissor.enabled) {
+        BEGIN_BATCH(8);
+        OUT_BATCH(CP_PACKET0(R200_RE_CNTL, 0));
+        OUT_BATCH(R200_SCISSOR_ENABLE | rmesa->hw.set.cmd[SET_RE_CNTL]);
+        OUT_BATCH(CP_PACKET0(R200_RE_AUX_SCISSOR_CNTL, 0));
+        OUT_BATCH(R200_SCISSOR_ENABLE_0);
+        OUT_BATCH(CP_PACKET0(R200_RE_SCISSOR_TL_0, 0));
+        OUT_BATCH((rmesa->radeon.state.scissor.rect.y1 << 16) |
+                  rmesa->radeon.state.scissor.rect.x1);
+        OUT_BATCH(CP_PACKET0(R200_RE_SCISSOR_BR_0, 0));
+        OUT_BATCH(((rmesa->radeon.state.scissor.rect.y2 - 1) << 16) |
+                  (rmesa->radeon.state.scissor.rect.x2 - 1));
+        END_BATCH();
+    } else {
+        BEGIN_BATCH(4);
+        OUT_BATCH(CP_PACKET0(R200_RE_CNTL, 0));
+        OUT_BATCH(rmesa->hw.set.cmd[SET_RE_CNTL] & ~R200_SCISSOR_ENABLE);
+        OUT_BATCH(CP_PACKET0(R200_RE_AUX_SCISSOR_CNTL, 0));
+        OUT_BATCH(0);
+        END_BATCH();
+    }
 }
 
 /* Fire a section of the retained (indexed_verts) buffer as a regular
@@ -208,51 +141,81 @@ void r200EmitVbufPrim( r200ContextPtr rmesa,
                        GLuint primitive,
                        GLuint vertex_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
+   BATCH_LOCALS(&rmesa->radeon);
 
    assert(!(primitive & R200_VF_PRIM_WALK_IND));
    
-   r200EmitState( rmesa );
+   radeonEmitState(&rmesa->radeon);
    
    if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
       fprintf(stderr, "%s cmd_used/4: %d prim %x nr %d\n", __FUNCTION__,
 	      rmesa->store.cmd_used/4, primitive, vertex_nr);
-   
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VBUF_BUFSZ,
-						  __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = R200_CP_CMD_3D_DRAW_VBUF_2;
-   cmd[2].i = (primitive | 
-	       R200_VF_PRIM_WALK_LIST |
-	       R200_VF_COLOR_ORDER_RGBA |
-	       (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
+   r200EmitScissor(rmesa);
+ 
+   BEGIN_BATCH(3);
+   OUT_BATCH_PACKET3_CLIP(R200_CP_CMD_3D_DRAW_VBUF_2, 0);
+   OUT_BATCH(primitive | R200_VF_PRIM_WALK_LIST | R200_VF_COLOR_ORDER_RGBA |
+	     (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
+   END_BATCH();
 }
 
+static void r200FireEB(r200ContextPtr rmesa, int vertex_count, int type)
+{
+	BATCH_LOCALS(&rmesa->radeon);
+
+	if (vertex_count > 0) {
+        r200EmitScissor(rmesa);
+		BEGIN_BATCH(8+2);
+		OUT_BATCH_PACKET3_CLIP(R200_CP_CMD_3D_DRAW_INDX_2, 0);
+		OUT_BATCH(R200_VF_PRIM_WALK_IND |
+			  R200_VF_COLOR_ORDER_RGBA | 
+			  ((vertex_count + 0) << 16) |
+			  type);
+		
+		if (!rmesa->radeon.radeonScreen->kernel_mm) {
+			OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2);
+			OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810);
+			OUT_BATCH_RELOC(rmesa->radeon.tcl.elt_dma_offset,
+					rmesa->radeon.tcl.elt_dma_bo,
+					rmesa->radeon.tcl.elt_dma_offset,
+					RADEON_GEM_DOMAIN_GTT, 0, 0);
+			OUT_BATCH((vertex_count + 1)/2);
+		} else {
+			OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2);
+			OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810);
+			OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset);
+			OUT_BATCH((vertex_count + 1)/2);
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.elt_dma_bo,
+					      RADEON_GEM_DOMAIN_GTT, 0, 0);
+		}
+		END_BATCH();
+	}
+}
 
-void r200FlushElts( r200ContextPtr rmesa )
+void r200FlushElts(GLcontext *ctx)
 {
-   int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start);
-   int dwords;
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 12)) / 2;
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   int nr, elt_used = rmesa->tcl.elt_used;
 
    if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
+     fprintf(stderr, "%s %x %d\n", __FUNCTION__, rmesa->tcl.hw_primitive, elt_used);
+
+   assert( rmesa->radeon.dma.flush == r200FlushElts );
+   rmesa->radeon.dma.flush = NULL;
 
-   assert( rmesa->dma.flush == r200FlushElts );
-   rmesa->dma.flush = NULL;
+   nr = elt_used / 2;
 
-   /* Cope with odd number of elts:
-    */
-   rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2;
-   dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4;
+   radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
 
-   cmd[1] |= (dwords - 3) << 16;
-   cmd[2] |= nr << R200_VF_VERTEX_NUMBER_SHIFT;
+   r200FireEB(rmesa, nr, rmesa->tcl.hw_primitive);
+
+   radeon_bo_unref(rmesa->radeon.tcl.elt_dma_bo);
+   rmesa->radeon.tcl.elt_dma_bo = NULL;
 
    if (R200_DEBUG & DEBUG_SYNC) {
       fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-      r200Finish( rmesa->glCtx );
+      radeonFinish( rmesa->radeon.glCtx );
    }
 }
 
@@ -261,38 +224,45 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 				    GLuint primitive,
 				    GLuint min_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
    GLushort *retval;
+   int ret;
 
    if (R200_DEBUG & DEBUG_IOCTL)
       fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
 
    assert((primitive & R200_VF_PRIM_WALK_IND));
    
-   r200EmitState( rmesa );
-   
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, ELTS_BUFSZ(min_nr),
-						__FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = R200_CP_CMD_3D_DRAW_INDX_2;
-   cmd[2].i = (primitive | 
-	       R200_VF_PRIM_WALK_IND |
-	       R200_VF_COLOR_ORDER_RGBA);
+   radeonEmitState(&rmesa->radeon);
+
+#ifdef RADEON_DEBUG_BO
+   rmesa->radeon.tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom,
+					  0, R200_ELT_BUF_SZ, 4,
+						 RADEON_GEM_DOMAIN_GTT, 0, "ELT");
+#else
+   rmesa->radeon.tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom,
+					  0, R200_ELT_BUF_SZ, 4,
+					  RADEON_GEM_DOMAIN_GTT, 0);
+#endif
+   rmesa->radeon.tcl.elt_dma_offset = 0;
+   rmesa->tcl.elt_used = min_nr * 2;
+
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.tcl.elt_dma_bo,
+				 RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret) {
+      fprintf(stderr,"failure to revalidate BOs - badness\n");
+   }
 
+   radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
+   retval = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
    
-   retval = (GLushort *)(cmd+3);
 
    if (R200_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x prim %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, primitive);
-
-   assert(!rmesa->dma.flush);
-   rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   rmesa->dma.flush = r200FlushElts;
+      fprintf(stderr, "%s: header prim %x \n",
+	      __FUNCTION__, primitive);
 
-   rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf;
+   assert(!rmesa->radeon.dma.flush);
+   rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
+   rmesa->radeon.dma.flush = r200FlushElts;
 
    return retval;
 }
@@ -300,129 +270,119 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 
 
 void r200EmitVertexAOS( r200ContextPtr rmesa,
-			  GLuint vertex_size,
-			  GLuint offset )
+			GLuint vertex_size,
+ 			struct radeon_bo *bo,
+			GLuint offset )
 {
-   drm_radeon_cmd_header_t *cmd;
+   BATCH_LOCALS(&rmesa->radeon);
 
    if (R200_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL))
       fprintf(stderr, "%s:  vertex_size 0x%x offset 0x%x \n",
 	      __FUNCTION__, vertex_size, offset);
 
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, VERT_AOS_BUFSZ,
-						  __FUNCTION__ );
 
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (2 << 16);
-   cmd[2].i = 1;
-   cmd[3].i = vertex_size | (vertex_size << 8);
-   cmd[4].i = offset;
+   BEGIN_BATCH(7);
+   OUT_BATCH_PACKET3(R200_CP_CMD_3D_LOAD_VBPNTR, 2);
+   OUT_BATCH(1);
+   OUT_BATCH(vertex_size | (vertex_size << 8));
+   OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   END_BATCH();
 }
-		       
 
-void r200EmitAOS( r200ContextPtr rmesa,
-		    struct r200_dma_region **component,
-		    GLuint nr,
-		    GLuint offset )
+void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset)
 {
-   drm_radeon_cmd_header_t *cmd;
-   int sz = AOS_BUFSZ(nr);
+   BATCH_LOCALS(&rmesa->radeon);
+   uint32_t voffset;
+   int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
    int i;
-   int *tmp;
-
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s nr arrays: %d\n", __FUNCTION__, nr);
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sz, __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (((sz / sizeof(int)) - 3) << 16);
-   cmd[2].i = nr;
-   tmp = &cmd[0].i;
-   cmd += 3;
-
-   for (i = 0 ; i < nr ; i++) {
-      if (i & 1) {
-	 cmd[0].i |= ((component[i]->aos_stride << 24) | 
-		      (component[i]->aos_size << 16));
-	 cmd[2].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-	 cmd += 3;
+   
+   if (RADEON_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr,
+	      offset);
+
+   BEGIN_BATCH(sz+2+ (nr*2));
+   OUT_BATCH_PACKET3(R200_CP_CMD_3D_LOAD_VBPNTR, sz - 1);
+   OUT_BATCH(nr);
+
+    
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+			
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i+1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
       }
-      else {
-	 cmd[0].i = ((component[i]->aos_stride << 8) | 
-		     (component[i]->aos_size << 0));
-	 cmd[1].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
+      
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[nr - 1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
+      }
+   } else {
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+	 
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH(voffset);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH(voffset);
+      }
+      
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH(voffset);
+      }
+      for (i = 0; i + 1 < nr; i += 2) {
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+0].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
+      }
+      if (nr & 1) {
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[nr-1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
       }
    }
-
-   if (R200_DEBUG & DEBUG_VERTS) {
-      fprintf(stderr, "%s:\n", __FUNCTION__);
-      for (i = 0 ; i < sz ; i++)
-	 fprintf(stderr, "   %d: %x\n", i, tmp[i]);
-   }
-}
-
-void r200EmitBlit( r200ContextPtr rmesa,
-		   GLuint color_fmt,
-		   GLuint src_pitch,
-		   GLuint src_offset,
-		   GLuint dst_pitch,
-		   GLuint dst_offset,
-		   GLint srcx, GLint srcy,
-		   GLint dstx, GLint dsty,
-		   GLuint w, GLuint h )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-	      __FUNCTION__, 
-	      src_pitch, src_offset, srcx, srcy,
-	      dst_pitch, dst_offset, dstx, dsty,
-	      w, h);
-
-   assert( (src_pitch & 63) == 0 );
-   assert( (dst_pitch & 63) == 0 );
-   assert( (src_offset & 1023) == 0 );
-   assert( (dst_offset & 1023) == 0 );
-   assert( w < (1<<16) );
-   assert( h < (1<<16) );
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 8 * sizeof(int),
-						  __FUNCTION__ );
-
-
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = R200_CP_CMD_BITBLT_MULTI | (5 << 16);
-   cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_BRUSH_NONE |
-	       (color_fmt << 8) |
-	       RADEON_GMC_SRC_DATATYPE_COLOR |
-	       RADEON_ROP3_S |
-	       RADEON_DP_SRC_SOURCE_MEMORY |
-	       RADEON_GMC_CLR_CMP_CNTL_DIS |
-	       RADEON_GMC_WR_MSK_DIS );
-
-   cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10);
-   cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10);
-   cmd[5].i = (srcx << 16) | srcy;
-   cmd[6].i = (dstx << 16) | dsty; /* dst */
-   cmd[7].i = (w << 16) | h;
-}
-
-
-void r200EmitWait( r200ContextPtr rmesa, GLuint flags )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) );
-
-   cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, 1 * sizeof(int),
-					   __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
-   cmd[0].wait.flags = flags;
+   END_BATCH();
 }
diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index 058296eab2..9a92a32079 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -54,7 +54,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
-#include "r200_span.h"
 #include "r200_pixel.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
@@ -62,6 +61,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_maos.h"
 #include "r200_vertprog.h"
 
+#include "radeon_span.h"
+
 #define need_GL_ARB_vertex_program
 #define need_GL_ATI_fragment_shader
 #define need_GL_EXT_blend_minmax
@@ -71,6 +72,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define need_GL_EXT_blend_func_separate
 #define need_GL_NV_vertex_program
 #define need_GL_ARB_point_parameters
+#define need_GL_EXT_framebuffer_object
 #include "extension_helper.h"
 
 #define DRIVER_DATE	"20060602"
@@ -78,9 +80,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h" /* for symbolic values of enum-type options */
-#ifndef R200_DEBUG
-int R200_DEBUG = (0);
-#endif
 
 /* Return various strings for glGetString().
  */
@@ -89,8 +88,8 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    static char buffer[128];
    unsigned   offset;
-   GLuint agp_mode = (rmesa->r200Screen->card_type == RADEON_CARD_PCI)? 0 :
-      rmesa->r200Screen->AGPMode;
+   GLuint agp_mode = (rmesa->radeon.radeonScreen->card_type == RADEON_CARD_PCI)? 0 :
+      rmesa->radeon.radeonScreen->AGPMode;
 
    switch ( name ) {
    case GL_VENDOR:
@@ -101,7 +100,7 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
 				     agp_mode );
 
       sprintf( & buffer[ offset ], " %sTCL",
-	       !(rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)
+	       !(rmesa->radeon.TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)
 	       ? "" : "NO-" );
 
       return (GLubyte *)buffer;
@@ -126,6 +125,7 @@ const struct dri_extension card_extensions[] =
     { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
     { "GL_EXT_blend_subtract",             NULL },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_packed_depth_stencil",	   NULL},
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -167,6 +167,11 @@ const struct dri_extension point_extensions[] = {
     { NULL,                                NULL }
 };
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 extern const struct tnl_pipeline_stage _r200_render_stage;
 extern const struct tnl_pipeline_stage _r200_tcl_stage;
 
@@ -234,6 +239,39 @@ static const struct dri_debug_control debug_control[] =
     { NULL,    0 }
 };
 
+static void r200_get_lock(radeonContextPtr radeon)
+{
+   r200ContextPtr rmesa = (r200ContextPtr)radeon;
+   drm_radeon_sarea_t *sarea = radeon->sarea;
+
+   R200_STATECHANGE( rmesa, ctx );
+   if (rmesa->radeon.sarea->tiling_enabled) {
+      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
+   }
+   else rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &= ~R200_COLOR_TILE_ENABLE;
+
+   if ( sarea->ctx_owner != rmesa->radeon.dri.hwContext ) {
+      sarea->ctx_owner = rmesa->radeon.dri.hwContext;
+      if (!radeon->radeonScreen->kernel_mm)
+         radeon_bo_legacy_texture_age(radeon->radeonScreen->bom);
+   }
+
+}
+
+static void r200_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+}
+
+
+static void r200_init_vtbl(radeonContextPtr radeon)
+{
+   radeon->vtbl.get_lock = r200_get_lock;
+   radeon->vtbl.update_viewport_offset = r200UpdateViewportOffset;
+   radeon->vtbl.emit_cs_header = r200_vtbl_emit_cs_header;
+   radeon->vtbl.swtcl_flush = r200_swtcl_flush;
+   radeon->vtbl.fallback = r200Fallback;
+}
+
 
 /* Create the device specific rendering context.
  */
@@ -245,9 +283,9 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    radeonScreenPtr screen = (radeonScreenPtr)(sPriv->private);
    struct dd_function_table functions;
    r200ContextPtr rmesa;
-   GLcontext *ctx, *shareCtx;
+   GLcontext *ctx;
    int i;
-   int tcl_mode, fthrottle_mode;
+   int tcl_mode;
 
    assert(glVisual);
    assert(driContextPriv);
@@ -257,7 +295,8 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    rmesa = (r200ContextPtr) CALLOC( sizeof(*rmesa) );
    if ( !rmesa )
       return GL_FALSE;
-      
+
+   r200_init_vtbl(&rmesa->radeon);
    /* init exp fog table data */
    r200InitStaticFogData();
 
@@ -265,12 +304,12 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
     * Do this here so that initialMaxAnisotropy is set before we create
     * the default textures.
     */
-   driParseConfigFiles (&rmesa->optionCache, &screen->optionCache,
+   driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache,
 			screen->driScreen->myNum, "r200");
-   rmesa->initialMaxAnisotropy = driQueryOptionf(&rmesa->optionCache,
-                                                 "def_max_anisotropy");
+   rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache,
+							"def_max_anisotropy");
 
-   if ( driQueryOptionb( &rmesa->optionCache, "hyperz" ) ) {
+   if ( driQueryOptionb( &rmesa->radeon.optionCache, "hyperz" ) ) {
       if ( sPriv->drm_version.minor < 13 )
 	 fprintf( stderr, "DRM version 1.%d too old to support HyperZ, "
 			  "disabling.\n", sPriv->drm_version.minor );
@@ -291,59 +330,15 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    r200InitTextureFuncs(&functions);
    r200InitShaderFuncs(&functions); 
 
-   /* Allocate and initialize the Mesa context */
-   if (sharedContextPrivate)
-      shareCtx = ((r200ContextPtr) sharedContextPrivate)->glCtx;
-   else
-      shareCtx = NULL;
-   rmesa->glCtx = _mesa_create_context(glVisual, shareCtx,
-                                       &functions, (void *) rmesa);
-   if (!rmesa->glCtx) {
-      FREE(rmesa);
-      return GL_FALSE;
-   }
-   driContextPriv->driverPrivate = rmesa;
-
-   /* Init r200 context data */
-   rmesa->dri.context = driContextPriv;
-   rmesa->dri.screen = sPriv;
-   rmesa->dri.drawable = NULL; /* Set by XMesaMakeCurrent */
-   rmesa->dri.hwContext = driContextPriv->hHWContext;
-   rmesa->dri.hwLock = &sPriv->pSAREA->lock;
-   rmesa->dri.fd = sPriv->fd;
-   rmesa->dri.drmMinor = sPriv->drm_version.minor;
-
-   rmesa->r200Screen = screen;
-   rmesa->sarea = (drm_radeon_sarea_t *)((GLubyte *)sPriv->pSAREA +
-				       screen->sarea_priv_offset);
-
-
-   rmesa->dma.buf0_address = rmesa->r200Screen->buffers->list[0].address;
-
-   (void) memset( rmesa->texture_heaps, 0, sizeof( rmesa->texture_heaps ) );
-   make_empty_list( & rmesa->swapped );
-
-   rmesa->nr_heaps = 1 /* screen->numTexHeaps */ ;
-   assert(rmesa->nr_heaps < RADEON_NR_TEX_HEAPS);
-   for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-      rmesa->texture_heaps[i] = driCreateTextureHeap( i, rmesa,
-	    screen->texSize[i],
-	    12,
-	    RADEON_NR_TEX_REGIONS,
-	    (drmTextureRegionPtr)rmesa->sarea->tex_list[i],
-	    & rmesa->sarea->tex_age[i],
-	    & rmesa->swapped,
-	    sizeof( r200TexObj ),
-	    (destroy_texture_object_t *) r200DestroyTexObj );
+   if (!radeonInitContext(&rmesa->radeon, &functions,
+			  glVisual, driContextPriv,
+			  sharedContextPrivate)) {
+     FREE(rmesa);
+     return GL_FALSE;
    }
-   rmesa->texture_depth = driQueryOptioni (&rmesa->optionCache,
-					   "texture_depth");
-   if (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
-      rmesa->texture_depth = ( screen->cpp == 4 ) ?
-	 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
 
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->hw.all_dirty = 1;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.hw.all_dirty = 1;
 
    /* Set the maximum texture size small enough that we can guarentee that
     * all texture units can bind a maximal texture and have all of them in
@@ -351,29 +346,20 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
     * setting allow larger textures.
     */
 
-   ctx = rmesa->glCtx;
-   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->optionCache,
+   ctx = rmesa->radeon.glCtx;
+   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache,
 						 "texture_units");
    ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
    ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
 
-   i = driQueryOptioni( &rmesa->optionCache, "allow_large_textures");
-
-   driCalculateMaxTextureLevels( rmesa->texture_heaps,
-				 rmesa->nr_heaps,
-				 & ctx->Const,
-				 4,
-				 11, /* max 2D texture size is 2048x2048 */
-#if ENABLE_HW_3D_TEXTURE
-				 8,  /* max 3D texture size is 256^3 */
-#else
-				 0,  /* 3D textures unsupported */
-#endif
-				 11, /* max cube texture size is 2048x2048 */
-				 11, /* max texture rectangle size is 2048x2048 */
-				 12,
-				 GL_FALSE,
-				 i );
+   i = driQueryOptioni( &rmesa->radeon.optionCache, "allow_large_textures");
+
+   /* FIXME: When no memory manager is available we should set this 
+    * to some reasonable value based on texture memory pool size */
+   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.Max3DTextureLevels = 9;
+   ctx->Const.MaxCubeTextureLevels = 12;
+   ctx->Const.MaxTextureRectSize = 2048;
 
    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 
@@ -383,7 +369,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    ctx->Const.MinPointSizeAA = 1.0;
    ctx->Const.MaxPointSizeAA = 1.0;
    ctx->Const.PointSizeGranularity = 0.0625;
-   if (rmesa->r200Screen->drmSupportsPointSprites)
+   if (rmesa->radeon.radeonScreen->drmSupportsPointSprites)
       ctx->Const.MaxPointSize = 2047.0;
    else
       ctx->Const.MaxPointSize = 1.0;
@@ -441,32 +427,35 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    _math_matrix_set_identity( &rmesa->tmpmat );
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
-   if (!(rmesa->r200Screen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) {
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     driInitExtensions(ctx, mm_extensions, GL_FALSE);
+   if (!(rmesa->radeon.radeonScreen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) {
      /* yuv textures don't work with some chips - R200 / rv280 okay so far
 	others get the bit ordering right but don't actually do YUV-RGB conversion */
       _mesa_enable_extension( ctx, "GL_MESA_ycbcr_texture" );
    }
-   if (rmesa->glCtx->Mesa_DXTn) {
+   if (rmesa->radeon.glCtx->Mesa_DXTn) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
       _mesa_enable_extension( ctx, "GL_S3_s3tc" );
    }
-   else if (driQueryOptionb (&rmesa->optionCache, "force_s3tc_enable")) {
+   else if (driQueryOptionb (&rmesa->radeon.optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
    }
 
-   if (rmesa->r200Screen->drmSupportsCubeMapsR200)
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR200)
       _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
-   if (rmesa->r200Screen->drmSupportsBlendColor) {
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
        driInitExtensions( ctx, blend_extensions, GL_FALSE );
    }
-   if(rmesa->r200Screen->drmSupportsVertexProgram)
+   if(rmesa->radeon.radeonScreen->drmSupportsVertexProgram)
       driInitSingleExtension( ctx, ARB_vp_extension );
-   if(driQueryOptionb(&rmesa->optionCache, "nv_vertex_program"))
+   if(driQueryOptionb(&rmesa->radeon.optionCache, "nv_vertex_program"))
       driInitSingleExtension( ctx, NV_vp_extension );
 
-   if ((ctx->Const.MaxTextureUnits == 6) && rmesa->r200Screen->drmSupportsFragShader)
+   if ((ctx->Const.MaxTextureUnits == 6) && rmesa->radeon.radeonScreen->drmSupportsFragShader)
       driInitSingleExtension( ctx, ATI_fs_extension );
-   if (rmesa->r200Screen->drmSupportsPointSprites)
+   if (rmesa->radeon.radeonScreen->drmSupportsPointSprites)
       driInitExtensions( ctx, point_extensions, GL_FALSE );
 #if 0
    r200InitDriverFuncs( ctx );
@@ -476,33 +465,16 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 #endif
    /* plug in a few more device driver functions */
    /* XXX these should really go right after _mesa_init_driver_functions() */
+   radeon_fbo_init(&rmesa->radeon);
+   radeonInitSpanFuncs( ctx );
    r200InitPixelFuncs( ctx );
-   r200InitSpanFuncs( ctx );
    r200InitTnlFuncs( ctx );
    r200InitState( rmesa );
    r200InitSwtcl( ctx );
 
-   fthrottle_mode = driQueryOptioni(&rmesa->optionCache, "fthrottle_mode");
-   rmesa->iw.irq_seq = -1;
-   rmesa->irqsEmitted = 0;
-   rmesa->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS &&
-		     rmesa->r200Screen->irq);
-
-   rmesa->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
-
-   if (!rmesa->do_irqs)
-      fprintf(stderr,
-	      "IRQ's not enabled, falling back to %s: %d %d\n",
-	      rmesa->do_usleeps ? "usleeps" : "busy waits",
-	      fthrottle_mode,
-	      rmesa->r200Screen->irq);
-
    rmesa->prefer_gart_client_texturing = 
       (getenv("R200_GART_CLIENT_TEXTURES") != 0);
 
-   (*sPriv->systemTime->getUST)( & rmesa->swap_ust );
-
-
 #if DO_DEBUG
    R200_DEBUG  = driParseDebugString( getenv( "R200_DEBUG" ),
 				      debug_control );
@@ -510,202 +482,21 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 				      debug_control );
 #endif
 
-   tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-   if (driQueryOptionb(&rmesa->optionCache, "no_rast")) {
+   tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode");
+   if (driQueryOptionb(&rmesa->radeon.optionCache, "no_rast")) {
       fprintf(stderr, "disabling 3D acceleration\n");
       FALLBACK(rmesa, R200_FALLBACK_DISABLE, 1);
    }
    else if (tcl_mode == DRI_CONF_TCL_SW || getenv("R200_NO_TCL") ||
-	    !(rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL)) {
-      if (rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL) {
-	 rmesa->r200Screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+	    !(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+      if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+	 rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
 	 fprintf(stderr, "Disabling HW TCL support\n");
       }
-      TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1);
+      TCL_FALLBACK(rmesa->radeon.glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1);
    }
 
    return GL_TRUE;
 }
 
 
-/* Destroy the device specific context.
- */
-/* Destroy the Mesa and driver specific context data.
- */
-void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate;
-   r200ContextPtr current = ctx ? R200_CONTEXT(ctx) : NULL;
-
-   /* check if we're deleting the currently bound context */
-   if (rmesa == current) {
-      R200_FIREVERTICES( rmesa );
-      _mesa_make_current(NULL, NULL, NULL);
-   }
-
-   /* Free r200 context resources */
-   assert(rmesa); /* should never be null */
-   if ( rmesa ) {
-      GLboolean   release_texture_heaps;
-
-
-      release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
-      _swsetup_DestroyContext( rmesa->glCtx );
-      _tnl_DestroyContext( rmesa->glCtx );
-      _vbo_DestroyContext( rmesa->glCtx );
-      _swrast_DestroyContext( rmesa->glCtx );
-
-      r200DestroySwtcl( rmesa->glCtx );
-      r200ReleaseArrays( rmesa->glCtx, ~0 );
-
-      if (rmesa->dma.current.buf) {
-	 r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-	 r200FlushCmdBuf( rmesa, __FUNCTION__ );
-      }
-
-      if (rmesa->state.scissor.pClipRects) {
-	 FREE(rmesa->state.scissor.pClipRects);
-	 rmesa->state.scissor.pClipRects = NULL;
-      }
-
-      if ( release_texture_heaps ) {
-         /* This share group is about to go away, free our private
-          * texture object data.
-          */
-         int i;
-
-         for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-	    driDestroyTextureHeap( rmesa->texture_heaps[ i ] );
-	    rmesa->texture_heaps[ i ] = NULL;
-         }
-
-	 assert( is_empty_list( & rmesa->swapped ) );
-      }
-
-      /* free the Mesa context */
-      rmesa->glCtx->DriverCtx = NULL;
-      _mesa_destroy_context( rmesa->glCtx );
-
-      /* free the option cache */
-      driDestroyOptionCache (&rmesa->optionCache);
-
-      FREE( rmesa );
-   }
-}
-
-
-
-
-void
-r200SwapBuffers( __DRIdrawablePrivate *dPriv )
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      r200ContextPtr rmesa;
-      GLcontext *ctx;
-      rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = rmesa->glCtx;
-      if (ctx->Visual.doubleBufferMode) {
-         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-         if ( rmesa->doPageFlip ) {
-            r200PageFlip( dPriv );
-         }
-         else {
-	     r200CopyBuffer( dPriv, NULL );
-         }
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__);
-   }
-}
-
-void
-r200CopySubBuffer( __DRIdrawablePrivate *dPriv,
-		   int x, int y, int w, int h )
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      r200ContextPtr rmesa;
-      GLcontext *ctx;
-      rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = rmesa->glCtx;
-      if (ctx->Visual.doubleBufferMode) {
-	 drm_clip_rect_t rect;
-	 rect.x1 = x + dPriv->x;
-	 rect.y1 = (dPriv->h - y - h) + dPriv->y;
-	 rect.x2 = rect.x1 + w;
-	 rect.y2 = rect.y1 + h;
-         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-	 r200CopyBuffer( dPriv, &rect );
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__);
-   }
-}
-
-/* Force the context `c' to be the current context and associate with it
- * buffer `b'.
- */
-GLboolean
-r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
-                   __DRIdrawablePrivate *driDrawPriv,
-                   __DRIdrawablePrivate *driReadPriv )
-{
-   if ( driContextPriv ) {
-      r200ContextPtr newCtx = 
-	 (r200ContextPtr) driContextPriv->driverPrivate;
-
-      if (R200_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)newCtx->glCtx);
-
-      newCtx->dri.readable = driReadPriv;
-
-      if ( newCtx->dri.drawable != driDrawPriv ||
-           newCtx->lastStamp != driDrawPriv->lastStamp ) {
-	 if (driDrawPriv->swap_interval == (unsigned)-1) {
-	    driDrawPriv->vblFlags = (newCtx->r200Screen->irq != 0)
-	       ? driGetDefaultVBlankFlags(&newCtx->optionCache)
-	       : VBLANK_FLAG_NO_IRQ;
-
-	    driDrawableInitVBlank( driDrawPriv );
-	 }
-
-	 newCtx->dri.drawable = driDrawPriv;
-
-	 r200SetCliprects(newCtx);
-	 r200UpdateViewportOffset( newCtx->glCtx );
-      }
-
-      _mesa_make_current( newCtx->glCtx,
-			  (GLframebuffer *) driDrawPriv->driverPrivate,
-			  (GLframebuffer *) driReadPriv->driverPrivate );
-
-      _mesa_update_state( newCtx->glCtx );
-      r200ValidateState( newCtx->glCtx );
-
-   } else {
-      if (R200_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
-      _mesa_make_current( NULL, NULL, NULL );
-   }
-
-   if (R200_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "End %s\n", __FUNCTION__);
-   return GL_TRUE;
-}
-
-/* Force the context `c' to be unbound from its buffer.
- */
-GLboolean
-r200UnbindContext( __DRIcontextPrivate *driContextPriv )
-{
-   r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate;
-
-   if (R200_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->glCtx);
-
-   return GL_TRUE;
-}
diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h
index 14a1dda46a..6267293817 100644
--- a/src/mesa/drivers/dri/r200/r200_context.h
+++ b/src/mesa/drivers/dri/r200/r200_context.h
@@ -53,51 +53,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #error This driver requires a newer libdrm to compile
 #endif
 
+#include "radeon_screen.h"
+#include "radeon_common.h"
+
+#include "radeon_lock.h"
+
 struct r200_context;
 typedef struct r200_context r200ContextRec;
 typedef struct r200_context *r200ContextPtr;
 
-/* This union is used to avoid warnings/miscompilation
-   with float to uint32_t casts due to strict-aliasing */
-typedef union { GLfloat f; uint32_t ui32; } float_ui32_type;
-
-#include "r200_lock.h"
-#include "radeon_screen.h"
 #include "main/mm.h"
 
-/* Flags for software fallback cases */
-/* See correponding strings in r200_swtcl.c */
-#define R200_FALLBACK_TEXTURE           0x01
-#define R200_FALLBACK_DRAW_BUFFER       0x02
-#define R200_FALLBACK_STENCIL           0x04
-#define R200_FALLBACK_RENDER_MODE       0x08
-#define R200_FALLBACK_DISABLE           0x10
-#define R200_FALLBACK_BORDER_MODE       0x20
-
-/* The blit width for texture uploads
- */
-#define BLIT_WIDTH_BYTES 1024
-
-/* Use the templated vertex format:
- */
-#define COLOR_IS_RGBA
-#define TAG(x) r200##x
-#include "tnl_dd/t_dd_vertex.h"
-#undef TAG
-
-typedef void (*r200_tri_func)( r200ContextPtr,
-				 r200Vertex *,
-				 r200Vertex *,
-				 r200Vertex * );
-
-typedef void (*r200_line_func)( r200ContextPtr,
-				  r200Vertex *,
-				  r200Vertex * );
-
-typedef void (*r200_point_func)( r200ContextPtr,
-				   r200Vertex * );
-
-
 struct r200_vertex_program {
         struct gl_vertex_program mesa_program; /* Must be first */
         int translated;
@@ -112,93 +78,11 @@ struct r200_vertex_program {
         int fogmode;
 };
 
-struct r200_colorbuffer_state {
-   GLuint clear;
-#if 000
-   GLint drawOffset, drawPitch;
-#endif
-   int roundEnable;
-};
-
-
-struct r200_depthbuffer_state {
-   GLuint clear;
-   GLfloat scale;
-};
-
-#if 000
-struct r200_pixel_state {
-   GLint readOffset, readPitch;
-};
-#endif
-
-struct r200_scissor_state {
-   drm_clip_rect_t rect;
-   GLboolean enabled;
-
-   GLuint numClipRects;			/* Cliprects active */
-   GLuint numAllocedClipRects;		/* Cliprects available */
-   drm_clip_rect_t *pClipRects;
-};
-
-struct r200_stencilbuffer_state {
-   GLboolean hwBuffer;
-   GLuint clear;			/* rb3d_stencilrefmask value */
-};
-
-struct r200_stipple_state {
-   GLuint mask[32];
-};
-
-
-
-#define TEX_0   0x1
-#define TEX_1   0x2
-#define TEX_2	0x4
-#define TEX_3	0x8
-#define TEX_4	0x10
-#define TEX_5	0x20
-#define TEX_ALL 0x3f
-
-typedef struct r200_tex_obj r200TexObj, *r200TexObjPtr;
-
-/* Texture object in locally shared texture space.
- */
-struct r200_tex_obj {
-   driTextureObject   base;
-
-   GLuint bufAddr;			/* Offset to start of locally
-					   shared texture block */
-
-   GLuint dirty_state;		        /* Flags (1 per texunit) for
-					   whether or not this texobj
-					   has dirty hardware state
-					   (pp_*) that needs to be
-					   brought into the
-					   texunit. */
-
-   drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS];
-					/* Six, for the cube faces */
-   GLboolean image_override;		/* Image overridden by GLX_EXT_tfp */
-
-   GLuint pp_txfilter;		        /* hardware register values */
-   GLuint pp_txformat;
-   GLuint pp_txformat_x;
-   GLuint pp_txoffset;		        /* Image location in texmem.
-					   All cube faces follow. */
-   GLuint pp_txsize;		        /* npot only */
-   GLuint pp_txpitch;		        /* npot only */
-   GLuint pp_border_color;
-   GLuint pp_cubic_faces;	        /* cube face 1,2,3,4 log2 sizes */
-
-   GLboolean  border_fallback;
-
-   GLuint tile_bits;			/* hw texture tile bits used on this texture */
-};
+#define R200_TEX_ALL 0x3f
 
 
 struct r200_texture_env_state {
-   r200TexObjPtr texobj;
+   radeonTexObjPtr texobj;
    GLuint outputreg;
    GLuint unitneeded;
 };
@@ -210,19 +94,6 @@ struct r200_texture_state {
 };
 
 
-struct r200_state_atom {
-   struct r200_state_atom *next, *prev;
-   const char *name;		         /* for debug */
-   int cmd_size;		         /* size in bytes */
-   GLuint idx;
-   int *cmd;			         /* one or more cmd's */
-   int *lastcmd;			 /* one or more cmd's */
-   GLboolean dirty;
-   GLboolean (*check)( GLcontext *, int );    /* is this state active? */
-};
-   
-
-
 /* Trying to keep these relatively short as the variables are becoming
  * extravagently long.  Drop the driver name prefix off the front of
  * everything - I think we know which driver we're in by now, and keep the
@@ -597,181 +468,79 @@ struct r200_state_atom {
 
 
 struct r200_hw_state {
-   /* Head of the linked list of state atoms. */
-   struct r200_state_atom atomlist;
-
    /* Hardware state, stored as cmdbuf commands:  
     *   -- Need to doublebuffer for
     *           - reviving state after loss of context
     *           - eliding noop statechange loops? (except line stipple count)
     */
-   struct r200_state_atom ctx;
-   struct r200_state_atom set;
-   struct r200_state_atom vte;
-   struct r200_state_atom lin;
-   struct r200_state_atom msk;
-   struct r200_state_atom vpt;
-   struct r200_state_atom vap;
-   struct r200_state_atom vtx;
-   struct r200_state_atom tcl;
-   struct r200_state_atom msl;
-   struct r200_state_atom tcg;
-   struct r200_state_atom msc;
-   struct r200_state_atom cst;
-   struct r200_state_atom tam;
-   struct r200_state_atom tf;
-   struct r200_state_atom tex[6];
-   struct r200_state_atom cube[6];
-   struct r200_state_atom zbs;
-   struct r200_state_atom mtl[2];
-   struct r200_state_atom mat[9];
-   struct r200_state_atom lit[8]; /* includes vec, scl commands */
-   struct r200_state_atom ucp[6];
-   struct r200_state_atom pix[6]; /* pixshader stages */
-   struct r200_state_atom eye; /* eye pos */
-   struct r200_state_atom grd; /* guard band clipping */
-   struct r200_state_atom fog;
-   struct r200_state_atom glt;
-   struct r200_state_atom prf;
-   struct r200_state_atom afs[2];
-   struct r200_state_atom pvs;
-   struct r200_state_atom vpi[2];
-   struct r200_state_atom vpp[2];
-   struct r200_state_atom atf;
-   struct r200_state_atom spr;
-   struct r200_state_atom ptp;
-
-   int max_state_size;	/* Number of bytes necessary for a full state emit. */
-   GLboolean is_dirty, all_dirty;
+   struct radeon_state_atom ctx;
+   struct radeon_state_atom set;
+   struct radeon_state_atom vte;
+   struct radeon_state_atom lin;
+   struct radeon_state_atom msk;
+   struct radeon_state_atom vpt;
+   struct radeon_state_atom vap;
+   struct radeon_state_atom vtx;
+   struct radeon_state_atom tcl;
+   struct radeon_state_atom msl;
+   struct radeon_state_atom tcg;
+   struct radeon_state_atom msc;
+   struct radeon_state_atom cst;
+   struct radeon_state_atom tam;
+   struct radeon_state_atom tf;
+   struct radeon_state_atom tex[6];
+   struct radeon_state_atom cube[6];
+   struct radeon_state_atom zbs;
+   struct radeon_state_atom mtl[2];
+   struct radeon_state_atom mat[9];
+   struct radeon_state_atom lit[8]; /* includes vec, scl commands */
+   struct radeon_state_atom ucp[6];
+   struct radeon_state_atom pix[6]; /* pixshader stages */
+   struct radeon_state_atom eye; /* eye pos */
+   struct radeon_state_atom grd; /* guard band clipping */
+   struct radeon_state_atom fog;
+   struct radeon_state_atom glt;
+   struct radeon_state_atom prf;
+   struct radeon_state_atom afs[2];
+   struct radeon_state_atom pvs;
+   struct radeon_state_atom vpi[2];
+   struct radeon_state_atom vpp[2];
+   struct radeon_state_atom atf;
+   struct radeon_state_atom spr;
+   struct radeon_state_atom ptp;
 };
 
 struct r200_state {
    /* Derived state for internal purposes:
     */
-   struct r200_colorbuffer_state color;
-   struct r200_depthbuffer_state depth;
-#if 00
-   struct r200_pixel_state pixel;
-#endif
-   struct r200_scissor_state scissor;
-   struct r200_stencilbuffer_state stencil;
-   struct r200_stipple_state stipple;
+   struct radeon_stipple_state stipple;
    struct r200_texture_state texture;
    GLuint envneeded;
 };
 
-/* Need refcounting on dma buffers:
- */
-struct r200_dma_buffer {
-   int refcount;		/* the number of retained regions in buf */
-   drmBufPtr buf;
-};
-
-#define GET_START(rvb) (rmesa->r200Screen->gart_buffer_offset +		\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-
-/* A retained region, eg vertices for indexed vertices.
- */
-struct r200_dma_region {
-   struct r200_dma_buffer *buf;
-   char *address;		/* == buf->address */
-   int start, end, ptr;		/* offsets from start of buf */
-   int aos_start;
-   int aos_stride;
-   int aos_size;
-};
-
-
-struct r200_dma {
-   /* Active dma region.  Allocations for vertices and retained
-    * regions come from here.  Also used for emitting random vertices,
-    * these may be flushed by calling flush_current();
-    */
-   struct r200_dma_region current;
-   
-   void (*flush)( r200ContextPtr );
-
-   char *buf0_address;		/* start of buf[0], for index calcs */
-   GLuint nr_released_bufs;	/* flush after so many buffers released */
-};
-
-struct r200_dri_mirror {
-   __DRIcontextPrivate	*context;	/* DRI context */
-   __DRIscreenPrivate	*screen;	/* DRI screen */
-   __DRIdrawablePrivate	*drawable;	/* DRI drawable bound to this ctx */
-   __DRIdrawablePrivate	*readable;	/* DRI readable bound to this ctx */
-
-   drm_context_t hwContext;
-   drm_hw_lock_t *hwLock;
-   int fd;
-   int drmMinor;
-};
-
-
 #define R200_CMD_BUF_SZ  (16*1024) 
 
-struct r200_store {
-   GLuint statenr;
-   GLuint primnr;
-   char cmd_buf[R200_CMD_BUF_SZ];
-   int cmd_used;   
-   int elts_start;
-};
-
-
+#define R200_ELT_BUF_SZ  (16*1024) 
 /* r200_tcl.c
  */
 struct r200_tcl_info {
    GLuint hw_primitive;
 
-/* hw can handle 12 components max */
-   struct r200_dma_region *aos_components[12];
-   GLuint nr_aos_components;
-
    GLuint *Elts;
 
-   struct r200_dma_region indexed_verts;
-   struct r200_dma_region vertex_data[15];
+   int elt_used;
+
 };
 
 
 /* r200_swtcl.c
  */
 struct r200_swtcl_info {
-   GLuint RenderIndex;
-   
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-   GLuint vertex_size;
 
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
 
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-   GLuint vertex_attr_count;
-
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-   GLubyte *verts;
-
-   /* Fallback rasterization functions
-    */
-   r200_point_func draw_point;
-   r200_line_func draw_line;
-   r200_tri_func draw_tri;
-
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
+   radeon_point_func draw_point;
+   radeon_line_func draw_line;
+   radeon_tri_func draw_tri;
 
    /**
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
@@ -787,27 +556,10 @@ struct r200_swtcl_info {
     * Should Mesa project vertex data or will the hardware do it?
     */
    GLboolean needproj;
-
-   struct r200_dma_region indexed_verts;
-};
-
-
-struct r200_ioctl {
-   GLuint vertex_offset;
-   GLuint vertex_size;
 };
 
 
 
-#define R200_MAX_PRIMS 64
-
-
-
-struct r200_prim {
-   GLuint start;
-   GLuint end;
-   GLuint prim;
-};
 
    /* A maximum total of 29 elements per vertex:  3 floats for position, 3
     * floats for normal, 4 floats for color, 4 bytes for secondary color,
@@ -822,9 +574,8 @@ struct r200_prim {
 
 #define R200_MAX_VERTEX_SIZE ((3*6)+11)
 
-
 struct r200_context {
-   GLcontext *glCtx;			/* Mesa context */
+   struct radeon_context radeon;
 
    /* Driver and hardware state management
     */
@@ -832,56 +583,15 @@ struct r200_context {
    struct r200_state state;
    struct r200_vertex_program *curr_vp_hw;
 
-   /* Texture object bookkeeping
-    */
-   unsigned              nr_heaps;
-   driTexHeap          * texture_heaps[ RADEON_NR_TEX_HEAPS ];
-   driTextureObject      swapped;
-   int                   texture_depth;
-   float                 initialMaxAnisotropy;
-
-   /* Rasterization and vertex state:
-    */
-   GLuint TclFallback;
-   GLuint Fallback;
-   GLuint NewGLState;
-   DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
-
    /* Vertex buffers
     */
-   struct r200_ioctl ioctl;
-   struct r200_dma dma;
-   struct r200_store store;
-   /* A full state emit as of the first state emit in the main store, in case
-    * the context is lost.
-    */
-   struct r200_store backup_store;
-
-   /* Page flipping
-    */
-   GLuint doPageFlip;
-
-   /* Busy waiting
-    */
-   GLuint do_usleeps;
-   GLuint do_irqs;
-   GLuint irqsEmitted;
-   drm_radeon_irq_wait_t iw;
+   struct radeon_ioctl ioctl;
+   struct radeon_store store;
 
    /* Clientdata textures;
     */
    GLuint prefer_gart_client_texturing;
 
-   /* Drawable, cliprect and scissor information
-    */
-   GLuint numClipRects;			/* Cliprects for the draw buffer */
-   drm_clip_rect_t *pClipRects;
-   unsigned int lastStamp;
-   GLboolean lost_context;
-   GLboolean save_on_next_emit;
-   radeonScreenPtr r200Screen;	/* Screen private DRI data */
-   drm_radeon_sarea_t *sarea;		/* Private SAREA data */
-
    /* TCL stuff
     */
    GLmatrix TexGenMatrix[R200_MAX_TEXTURE_UNITS];
@@ -893,15 +603,6 @@ struct r200_context {
    GLuint TexGenCompSel;
    GLmatrix tmpmat;
 
-   /* buffer swap
-    */
-   int64_t swap_ust;
-   int64_t swap_missed_ust;
-
-   GLuint swap_count;
-   GLuint swap_missed_count;
-
-
    /* r200_tcl.c
     */
    struct r200_tcl_info tcl;
@@ -910,14 +611,6 @@ struct r200_context {
     */
    struct r200_swtcl_info swtcl;
 
-   /* Mirrors of some DRI state
-    */
-   struct r200_dri_mirror dri;
-
-   /* Configuration cache
-    */
-   driOptionCache optionCache;
-
    GLboolean using_hyperz;
    GLboolean texmicrotile;
 
@@ -927,28 +620,10 @@ struct r200_context {
 #define R200_CONTEXT(ctx)		((r200ContextPtr)(ctx->DriverCtx))
 
 
-static INLINE GLuint r200PackColor( GLuint cpp,
-					GLubyte r, GLubyte g,
-					GLubyte b, GLubyte a )
-{
-   switch ( cpp ) {
-   case 2:
-      return PACK_COLOR_565( r, g, b );
-   case 4:
-      return PACK_COLOR_8888( a, r, g, b );
-   default:
-      return 0;
-   }
-}
-
-
 extern void r200DestroyContext( __DRIcontextPrivate *driContextPriv );
 extern GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 				    __DRIcontextPrivate *driContextPriv,
 				    void *sharedContextPrivate);
-extern void r200SwapBuffers( __DRIdrawablePrivate *dPriv );
-extern void r200CopySubBuffer( __DRIdrawablePrivate * dPriv,
-			       int x, int y, int w, int h );
 extern GLboolean r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
 				  __DRIdrawablePrivate *driDrawPriv,
 				  __DRIdrawablePrivate *driReadPriv );
@@ -957,28 +632,9 @@ extern GLboolean r200UnbindContext( __DRIcontextPrivate *driContextPriv );
 /* ================================================================
  * Debugging:
  */
-#define DO_DEBUG		1
 
-#if DO_DEBUG
-extern int R200_DEBUG;
-#else
-#define R200_DEBUG		0
-#endif
+#define R200_DEBUG RADEON_DEBUG
+
 
-#define DEBUG_TEXTURE	0x001
-#define DEBUG_STATE	0x002
-#define DEBUG_IOCTL	0x004
-#define DEBUG_PRIMS	0x008
-#define DEBUG_VERTS	0x010
-#define DEBUG_FALLBACKS	0x020
-#define DEBUG_VFMT	0x040
-#define DEBUG_CODEGEN	0x080
-#define DEBUG_VERBOSE	0x100
-#define DEBUG_DRI       0x200
-#define DEBUG_DMA       0x400
-#define DEBUG_SANITY    0x800
-#define DEBUG_SYNC      0x1000
-#define DEBUG_PIXEL     0x2000
-#define DEBUG_MEMORY    0x4000
 
 #endif /* __R200_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r200/r200_fragshader.c b/src/mesa/drivers/dri/r200/r200_fragshader.c
index d514b28219..85c1b7bdd1 100644
--- a/src/mesa/drivers/dri/r200/r200_fragshader.c
+++ b/src/mesa/drivers/dri/r200/r200_fragshader.c
@@ -522,7 +522,7 @@ static void r200UpdateFSConstants( GLcontext *ctx )
 	 CLAMPED_FLOAT_TO_UBYTE(con_byte[2], ctx->ATIFragmentShader.GlobalConstants[i][2]);
 	 CLAMPED_FLOAT_TO_UBYTE(con_byte[3], ctx->ATIFragmentShader.GlobalConstants[i][3]);
       }
-      rmesa->hw.atf.cmd[ATF_TFACTOR_0 + i] = r200PackColor (
+      rmesa->hw.atf.cmd[ATF_TFACTOR_0 + i] = radeonPackColor (
 	 4, con_byte[0], con_byte[1], con_byte[2], con_byte[3] );
    }
 }
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c
index 0741e57af7..4dbda39eb9 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.c
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.c
@@ -31,7 +31,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * Authors:
  *   Keith Whitwell <keith@tungstengraphics.com>
  */
- 
+
 #include <sched.h>
 #include <errno.h>
 
@@ -41,6 +41,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "swrast/swrast.h"
 
+
+
+#include "radeon_common.h"
+#include "radeon_lock.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -54,635 +58,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R200_TIMEOUT             512
 #define R200_IDLE_RETRY           16
 
-
-static void r200WaitForIdle( r200ContextPtr rmesa );
-
-
-/* At this point we were in FlushCmdBufLocked but we had lost our context, so
- * we need to unwire our current cmdbuf, hook the one with the saved state in
- * it, flush it, and then put the current one back.  This is so commands at the
- * start of a cmdbuf can rely on the state being kept from the previous one.
- */
-static void r200BackUpAndEmitLostStateLocked( r200ContextPtr rmesa )
-{
-   GLuint nr_released_bufs;
-   struct r200_store saved_store;
-
-   if (rmesa->backup_store.cmd_used == 0)
-      return;
-
-   if (R200_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Emitting backup state on lost context\n");
-
-   rmesa->lost_context = GL_FALSE;
-
-   nr_released_bufs = rmesa->dma.nr_released_bufs;
-   saved_store = rmesa->store;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->store = rmesa->backup_store;
-   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
-   rmesa->dma.nr_released_bufs = nr_released_bufs;
-   rmesa->store = saved_store;
-}
-
-int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller )
-{
-   int ret, i;
-   drm_radeon_cmd_buffer_t cmd;
-
-   if (rmesa->lost_context)
-      r200BackUpAndEmitLostStateLocked( rmesa );
-
-   if (R200_DEBUG & DEBUG_IOCTL) {
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-
-      if (0 & R200_DEBUG & DEBUG_VERBOSE) 
-	 for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 )
-	    fprintf(stderr, "%d: %x\n", i/4, 
-		    *(int *)(&rmesa->store.cmd_buf[i]));
-   }
-
-   if (R200_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__,
-	      rmesa->dma.nr_released_bufs);
-
-
-   if (R200_DEBUG & DEBUG_SANITY) {
-      if (rmesa->state.scissor.enabled) 
-	 ret = r200SanityCmdBuffer( rmesa, 
-				    rmesa->state.scissor.numClipRects,
-				    rmesa->state.scissor.pClipRects);
-      else
-	 ret = r200SanityCmdBuffer( rmesa, 
-				    rmesa->numClipRects,
-				    rmesa->pClipRects);
-      if (ret) {
-	 fprintf(stderr, "drmSanityCommandWrite: %d\n", ret);	 
-	 goto out;
-      }
-   }
-
-
-   if (R200_DEBUG & DEBUG_MEMORY) {
-      if (! driValidateTextureHeaps( rmesa->texture_heaps, rmesa->nr_heaps,
-				     & rmesa->swapped ) ) {
-	 fprintf( stderr, "%s: texture memory is inconsistent - expect "
-		  "mangled textures\n", __FUNCTION__ );
-      }
-   }
-
-
-   cmd.bufsz = rmesa->store.cmd_used;
-   cmd.buf = rmesa->store.cmd_buf;
-
-   if (rmesa->state.scissor.enabled) {
-      cmd.nbox = rmesa->state.scissor.numClipRects;
-      cmd.boxes = (drm_clip_rect_t *)rmesa->state.scissor.pClipRects;
-   } else {
-      cmd.nbox = rmesa->numClipRects;
-      cmd.boxes = (drm_clip_rect_t *)rmesa->pClipRects;
-   }
-
-   ret = drmCommandWrite( rmesa->dri.fd,
-			  DRM_RADEON_CMDBUF,
-			  &cmd, sizeof(cmd) );
-
-   if (ret)
-      fprintf(stderr, "drmCommandWrite: %d\n", ret);
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
-      r200WaitForIdleLocked( rmesa );
-   }
-
-
- out:
-   rmesa->store.primnr = 0;
-   rmesa->store.statenr = 0;
-   rmesa->store.cmd_used = 0;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->save_on_next_emit = 1;
-
-   return ret;
-}
-
-
-/* Note: does not emit any commands to avoid recursion on
- * r200AllocCmdBuf.
- */
-void r200FlushCmdBuf( r200ContextPtr rmesa, const char *caller )
-{
-   int ret;
-
-   LOCK_HARDWARE( rmesa );
-
-   ret = r200FlushCmdBufLocked( rmesa, caller );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if (ret) {
-      fprintf(stderr, "drmRadeonCmdBuffer: %d (exiting)\n", ret);
-      exit(ret);
-   }
-}
-
-
-/* =============================================================
- * Hardware vertex buffer handling
- */
-
-
-void r200RefillCurrentDmaRegion( r200ContextPtr rmesa )
-{
-   struct r200_dma_buffer *dmabuf;
-   int fd = rmesa->dri.fd;
-   int index = 0;
-   int size = 0;
-   drmDMAReq dma;
-   int ret;
-
-   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-      fprintf(stderr, "%s\n", __FUNCTION__);  
-
-   if (rmesa->dma.flush) {
-      rmesa->dma.flush( rmesa );
-   }
-
-   if (rmesa->dma.current.buf)
-      r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-
-   if (rmesa->dma.nr_released_bufs > 4)
-      r200FlushCmdBuf( rmesa, __FUNCTION__ );
-
-   dma.context = rmesa->dri.hwContext;
-   dma.send_count = 0;
-   dma.send_list = NULL;
-   dma.send_sizes = NULL;
-   dma.flags = 0;
-   dma.request_count = 1;
-   dma.request_size = RADEON_BUFFER_SIZE;
-   dma.request_list = &index;
-   dma.request_sizes = &size;
-   dma.granted_count = 0;
-
-   LOCK_HARDWARE(rmesa);	/* no need to validate */
-
-   while (1) {
-      ret = drmDMA( fd, &dma );
-      if (ret == 0)
-	 break;
-   
-      if (rmesa->dma.nr_released_bufs) {
-	 r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
-      }
-
-      if (rmesa->do_usleeps) {
-	 UNLOCK_HARDWARE( rmesa );
-	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa );
-      }
-   }
-
-   UNLOCK_HARDWARE(rmesa);
-
-   if (R200_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "Allocated buffer %d\n", index);
-
-   dmabuf = CALLOC_STRUCT( r200_dma_buffer );
-   dmabuf->buf = &rmesa->r200Screen->buffers->list[index];
-   dmabuf->refcount = 1;
-
-   rmesa->dma.current.buf = dmabuf;
-   rmesa->dma.current.address = dmabuf->buf->address;
-   rmesa->dma.current.end = dmabuf->buf->total;
-   rmesa->dma.current.start = 0;
-   rmesa->dma.current.ptr = 0;
-}
-
-void r200ReleaseDmaRegion( r200ContextPtr rmesa,
-			     struct r200_dma_region *region,
-			     const char *caller )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-   
-   if (!region->buf)
-      return;
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (--region->buf->refcount == 0) {
-      drm_radeon_cmd_header_t *cmd;
-
-      if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-	 fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__,
-		 region->buf->buf->idx);  
-      
-      cmd = (drm_radeon_cmd_header_t *)r200AllocCmdBuf( rmesa, sizeof(*cmd), 
-						     __FUNCTION__ );
-      cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD;
-      cmd->dma.buf_idx = region->buf->buf->idx;
-      FREE(region->buf);
-      rmesa->dma.nr_released_bufs++;
-   }
-
-   region->buf = NULL;
-   region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r200AllocDmaRegion( r200ContextPtr rmesa, 
-			   struct r200_dma_region *region,
-			   int bytes,
-			   int alignment )
-{
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (region->buf)
-      r200ReleaseDmaRegion( rmesa, region, __FUNCTION__ );
-
-   alignment--;
-   rmesa->dma.current.start = rmesa->dma.current.ptr = 
-      (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      r200RefillCurrentDmaRegion( rmesa );
-
-   region->start = rmesa->dma.current.start;
-   region->ptr = rmesa->dma.current.start;
-   region->end = rmesa->dma.current.start + bytes;
-   region->address = rmesa->dma.current.address;
-   region->buf = rmesa->dma.current.buf;
-   region->buf->refcount++;
-
-   rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */
-   rmesa->dma.current.start = 
-      rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;  
-
-   assert( rmesa->dma.current.ptr <= rmesa->dma.current.end );
-}
-
-/* ================================================================
- * SwapBuffers with client-side throttling
- */
-
-static uint32_t r200GetLastFrame(r200ContextPtr rmesa)
-{
-   drm_radeon_getparam_t gp;
-   int ret;
-   uint32_t frame;
-
-   gp.param = RADEON_PARAM_LAST_FRAME;
-   gp.value = (int *)&frame;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_GETPARAM,
-			      &gp, sizeof(gp) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-
-   return frame;
-}
-
-static void r200EmitIrqLocked( r200ContextPtr rmesa )
-{
-   drm_radeon_irq_emit_t ie;
-   int ret;
-
-   ie.irq_seq = &rmesa->iw.irq_seq;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_IRQ_EMIT, 
-			      &ie, sizeof(ie) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void r200WaitIrq( r200ContextPtr rmesa )
-{
-   int ret;
-
-   do {
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_IRQ_WAIT,
-			     &rmesa->iw, sizeof(rmesa->iw) );
-   } while (ret && (errno == EINTR || errno == EBUSY));
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void r200WaitForFrameCompletion( r200ContextPtr rmesa )
-{
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-
-   if (rmesa->do_irqs) {
-      if (r200GetLastFrame(rmesa) < sarea->last_frame) {
-	 if (!rmesa->irqsEmitted) {
-	    while (r200GetLastFrame (rmesa) < sarea->last_frame)
-	       ;
-	 }
-	 else {
-	    UNLOCK_HARDWARE( rmesa ); 
-	    r200WaitIrq( rmesa );	
-	    LOCK_HARDWARE( rmesa ); 
-	 }
-	 rmesa->irqsEmitted = 10;
-      }
-
-      if (rmesa->irqsEmitted) {
-	 r200EmitIrqLocked( rmesa );
-	 rmesa->irqsEmitted--;
-      }
-   } 
-   else {
-      while (r200GetLastFrame (rmesa) < sarea->last_frame) {
-	 UNLOCK_HARDWARE( rmesa ); 
-	 if (rmesa->do_usleeps) 
-	    DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa ); 
-      }
-   }
-}
-
-
-
-/* Copy the back color buffer to the front color buffer.
- */
-void r200CopyBuffer( __DRIdrawablePrivate *dPriv,
-		      const drm_clip_rect_t	 *rect)
-{
-   r200ContextPtr rmesa;
-   GLint nbox, i, ret;
-   GLboolean   missed_target;
-   int64_t ust;
-   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   if ( R200_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *)rmesa->glCtx );
-   }
-
-   R200_FIREVERTICES( rmesa );
-
-   LOCK_HARDWARE( rmesa );
-
-
-   /* Throttle the frame rate -- only allow one pending swap buffers
-    * request at a time.
-    */
-   r200WaitForFrameCompletion( rmesa );
-   if (!rect)
-   {
-       UNLOCK_HARDWARE( rmesa );
-       driWaitForVBlank( dPriv, & missed_target );
-       LOCK_HARDWARE( rmesa );
-   }
-
-   nbox = dPriv->numClipRects; /* must be in locked region */
-
-   for ( i = 0 ; i < nbox ; ) {
-      GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      GLint n = 0;
-
-      for ( ; i < nr ; i++ ) {
-
-	  *b = box[i];
-
-	  if (rect)
-	  {
-	     if (rect->x1 > b->x1)
-		 b->x1 = rect->x1;
-	     if (rect->y1 > b->y1)
-		 b->y1 = rect->y1;
-	     if (rect->x2 < b->x2)
-		 b->x2 = rect->x2;
-	     if (rect->y2 < b->y2)
-		 b->y2 = rect->y2;
-
-	     if (b->x1 >= b->x2 || b->y1 >= b->y2)
-		 continue;
-	  }
-
-	  b++;
-	  n++;
-      }
-      rmesa->sarea->nbox = n;
-
-      if (!n)
-	 continue;
-
-      ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
-
-      if ( ret ) {
-	 fprintf( stderr, "DRM_R200_SWAP_BUFFERS: return = %d\n", ret );
-	 UNLOCK_HARDWARE( rmesa );
-	 exit( 1 );
-      }
-   }
-
-   UNLOCK_HARDWARE( rmesa );
-   if (!rect)
-   {
-       rmesa->hw.all_dirty = GL_TRUE;
-
-       rmesa->swap_count++;
-       (*psp->systemTime->getUST)( & ust );
-       if ( missed_target ) {
-	   rmesa->swap_missed_count++;
-	   rmesa->swap_missed_ust = ust - rmesa->swap_ust;
-       }
-
-       rmesa->swap_ust = ust;
-
-       sched_yield();
-   }
-}
-
-void r200PageFlip( __DRIdrawablePrivate *dPriv )
-{
-   r200ContextPtr rmesa;
-   GLint ret;
-   GLboolean   missed_target;
-   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   if ( R200_DEBUG & DEBUG_IOCTL ) {
-      fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
-	      rmesa->sarea->pfCurrentPage);
-   }
-
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   if (!dPriv->numClipRects) {
-      UNLOCK_HARDWARE( rmesa );
-      usleep( 10000 );		/* throttle invisible client 10ms */
-      return;
-   }
-
-   /* Need to do this for the perf box placement:
-    */
-   {
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      b[0] = box[0];
-      rmesa->sarea->nbox = 1;
-   }
-
-   /* Throttle the frame rate -- only allow a few pending swap buffers
-    * request at a time.
-    */
-   r200WaitForFrameCompletion( rmesa );
-   UNLOCK_HARDWARE( rmesa );
-   driWaitForVBlank( dPriv, & missed_target );
-   if ( missed_target ) {
-      rmesa->swap_missed_count++;
-      (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust );
-   }
-   LOCK_HARDWARE( rmesa );
-
-   ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret );
-      exit( 1 );
-   }
-
-   rmesa->swap_count++;
-   (void) (*psp->systemTime->getUST)( & rmesa->swap_ust );
-
-#if 000
-   if ( rmesa->sarea->pfCurrentPage == 1 ) {
-	 rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
-	 rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
-   } else {
-	 rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
-	 rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
-   }
-
-   R200_STATECHANGE( rmesa, ctx );
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = rmesa->state.color.drawOffset
-					   + rmesa->r200Screen->fbLocation;
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH]  = rmesa->state.color.drawPitch;
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
-   }
-#else
-   /* Get ready for drawing next frame.  Update the renderbuffers'
-    * flippedOffset/Pitch fields so we draw into the right place.
-    */
-   driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-                        rmesa->sarea->pfCurrentPage);
-
-
-   r200UpdateDrawBuffer(rmesa->glCtx);
-#endif
-}
-
-
-/* ================================================================
- * Buffer clear
- */
-static void r200Clear( GLcontext *ctx, GLbitfield mask )
+static void r200KernelClear(GLcontext *ctx, GLuint flags)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   GLuint flags = 0;
-   GLuint color_mask = 0;
-   GLint ret, i;
-   GLint cx, cy, cw, ch;
-
-   if ( R200_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "r200Clear\n");
-   }
-
-   {
-      LOCK_HARDWARE( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      if ( dPriv->numClipRects == 0 ) 
-	 return;
-   }
-
-   r200Flush( ctx );
-
-   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
-      flags |= RADEON_FRONT;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_FRONT_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_BACK_LEFT ) {
-      flags |= RADEON_BACK;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_BACK_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_DEPTH ) {
-      flags |= RADEON_DEPTH;
-      mask &= ~BUFFER_BIT_DEPTH;
-   }
-
-   if ( (mask & BUFFER_BIT_STENCIL) && rmesa->state.stencil.hwBuffer ) {
-      flags |= RADEON_STENCIL;
-      mask &= ~BUFFER_BIT_STENCIL;
-   }
-
-   if ( mask ) {
-      if (R200_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
-      _swrast_Clear( ctx, mask );
-   }
-
-   if ( !flags ) 
-      return;
-
-   if (rmesa->using_hyperz) {
-      flags |= RADEON_USE_COMP_ZBUF;
-/*      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200)
-	 flags |= RADEON_USE_HIERZ; */
-      if (!(rmesa->state.stencil.hwBuffer) ||
-	 ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
-	    ((rmesa->state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) {
-	  flags |= RADEON_CLEAR_FASTZ;
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
-
-   /* compute region after locking: */
-   cx = ctx->DrawBuffer->_Xmin;
-   cy = ctx->DrawBuffer->_Ymin;
-   cw = ctx->DrawBuffer->_Xmax - cx;
-   ch = ctx->DrawBuffer->_Ymax - cy;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLint cx, cy, cw, ch, ret;
+   GLuint i;
 
-   /* Flip top to bottom */
-   cx += dPriv->x;
-   cy  = dPriv->y + dPriv->h - cy - ch;
+   LOCK_HARDWARE( &rmesa->radeon );
 
    /* Throttle the number of clear ioctls we do.
     */
@@ -693,7 +76,7 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 
       gp.param = RADEON_PARAM_LAST_CLEAR;
       gp.value = (int *)&clear;
-      ret = drmCommandWriteRead( rmesa->dri.fd,
+      ret = drmCommandWriteRead( rmesa->radeon.dri.fd,
 		      DRM_RADEON_GETPARAM, &gp, sizeof(gp) );
 
       if ( ret ) {
@@ -703,24 +86,34 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 
       /* Clear throttling needs more thought.
        */
-      if ( rmesa->sarea->last_clear - clear <= 25 ) {
+      if ( rmesa->radeon.sarea->last_clear - clear <= 25 ) {
 	 break;
       }
-      
-      if (rmesa->do_usleeps) {
-	 UNLOCK_HARDWARE( rmesa );
+
+      if (rmesa->radeon.do_usleeps) {
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa );
+	 LOCK_HARDWARE( &rmesa->radeon );
       }
    }
 
    /* Send current state to the hardware */
-   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+   rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
+
 
+  /* compute region after locking: */
+   cx = ctx->DrawBuffer->_Xmin;
+   cy = ctx->DrawBuffer->_Ymin;
+   cw = ctx->DrawBuffer->_Xmax - cx;
+   ch = ctx->DrawBuffer->_Ymax - cy;
+
+   /* Flip top to bottom */
+   cx += dPriv->x;
+   cy  = dPriv->y + dPriv->h - cy - ch;
    for ( i = 0 ; i < dPriv->numClipRects ; ) {
       GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, dPriv->numClipRects );
       drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
+      drm_clip_rect_t *b = rmesa->radeon.sarea->boxes;
       drm_radeon_clear_t clear;
       drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
       GLint n = 0;
@@ -755,17 +148,17 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 	 }
       }
 
-      rmesa->sarea->nbox = n;
+      rmesa->radeon.sarea->nbox = n;
 
       clear.flags       = flags;
-      clear.clear_color = rmesa->state.color.clear;
-      clear.clear_depth = rmesa->state.depth.clear;	/* needed for hyperz */
+      clear.clear_color = rmesa->radeon.state.color.clear;
+      clear.clear_depth = rmesa->radeon.state.depth.clear;	/* needed for hyperz */
       clear.color_mask  = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      clear.depth_mask  = rmesa->state.stencil.clear;
+      clear.depth_mask  = rmesa->radeon.state.stencil.clear;
       clear.depth_boxes = depth_boxes;
 
       n--;
-      b = rmesa->sarea->boxes;
+      b = rmesa->radeon.sarea->boxes;
       for ( ; n >= 0 ; n-- ) {
 	 depth_boxes[n].f[CLEAR_X1] = (float)b[n].x1;
 	 depth_boxes[n].f[CLEAR_Y1] = (float)b[n].y1;
@@ -774,84 +167,94 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
 	 depth_boxes[n].f[CLEAR_DEPTH] = ctx->Depth.Clear;
       }
 
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_CLEAR,
+      ret = drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_CLEAR,
 			     &clear, sizeof(clear));
 
 
       if ( ret ) {
-	 UNLOCK_HARDWARE( rmesa );
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 fprintf( stderr, "DRM_RADEON_CLEAR: return = %d\n", ret );
 	 exit( 1 );
       }
    }
-
-   UNLOCK_HARDWARE( rmesa );
-   rmesa->hw.all_dirty = GL_TRUE;
+   UNLOCK_HARDWARE( &rmesa->radeon );
 }
+/* ================================================================
+ * Buffer clear
+ */
+static void r200Clear( GLcontext *ctx, GLbitfield mask )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLuint flags = 0;
+   GLuint color_mask = 0;
+   GLuint orig_mask = mask;
 
+   if ( R200_DEBUG & DEBUG_IOCTL ) {
+	   if (rmesa->radeon.sarea)
+	       fprintf( stderr, "r200Clear %x %d\n", mask, rmesa->radeon.sarea->pfCurrentPage);
+	   else
+	       fprintf( stderr, "r200Clear %x radeon->sarea is NULL\n", mask);
+   }
 
-void r200WaitForIdleLocked( r200ContextPtr rmesa )
-{
-    int ret;
-    int i = 0;
-    
-    do {
-       ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_CP_IDLE);
-       if (ret) 
-	  DO_USLEEP( 1 );
-    } while (ret && ++i < 100);
-    
-    if ( ret < 0 ) {
-       UNLOCK_HARDWARE( rmesa );
-       fprintf( stderr, "Error: R200 timed out... exiting\n" );
-       exit( -1 );
-    }
-}
+   {
+      LOCK_HARDWARE( &rmesa->radeon );
+      UNLOCK_HARDWARE( &rmesa->radeon );
+      if ( dPriv->numClipRects == 0 )
+	 return;
+   }
 
+   radeonFlush( ctx );
 
-static void r200WaitForIdle( r200ContextPtr rmesa )
-{
-   LOCK_HARDWARE(rmesa);
-   r200WaitForIdleLocked( rmesa );
-   UNLOCK_HARDWARE(rmesa);
-}
+   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
+      flags |= RADEON_FRONT;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_FRONT_LEFT;
+   }
 
+   if ( mask & BUFFER_BIT_BACK_LEFT ) {
+      flags |= RADEON_BACK;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_BACK_LEFT;
+   }
 
-void r200Flush( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+   if ( mask & BUFFER_BIT_DEPTH ) {
+      flags |= RADEON_DEPTH;
+      mask &= ~BUFFER_BIT_DEPTH;
+   }
 
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   if ( (mask & BUFFER_BIT_STENCIL) ) {
+      flags |= RADEON_STENCIL;
+      mask &= ~BUFFER_BIT_STENCIL;
+   }
 
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
+   if ( mask ) {
+      if (R200_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
+      _swrast_Clear( ctx, mask );
+   }
 
-   r200EmitState( rmesa );
-   
-   if (rmesa->store.cmd_used)
-      r200FlushCmdBuf( rmesa, __FUNCTION__ );
-}
+   if ( !flags )
+      return;
 
-/* Make sure all commands have been sent to the hardware and have
- * completed processing.
- */
-void r200Finish( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200Flush( ctx );
+   if (rmesa->using_hyperz) {
+      flags |= RADEON_USE_COMP_ZBUF;
+/*      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200)
+	 flags |= RADEON_USE_HIERZ; */
+      if (!((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
+	    ((rmesa->radeon.state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) {
+	  flags |= RADEON_CLEAR_FASTZ;
+      }
+   }
 
-   if (rmesa->do_irqs) {
-      LOCK_HARDWARE( rmesa );
-      r200EmitIrqLocked( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      r200WaitIrq( rmesa );
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+      radeonUserClear(ctx, orig_mask);
+   else {
+      r200KernelClear(ctx, flags);
+      rmesa->radeon.hw.all_dirty = GL_TRUE;
    }
-   else 
-      r200WaitForIdle( rmesa );
 }
 
-
 /* This version of AllocateMemoryMESA allocates only GART memory, and
  * only does so after the point at which the driver has been
  * initialized.
@@ -862,7 +265,7 @@ void r200Finish( GLcontext *ctx )
  * device fd.
  */
 void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size,
-			     GLfloat readfreq, GLfloat writefreq, 
+			     GLfloat readfreq, GLfloat writefreq,
 			     GLfloat priority)
 {
    GET_CURRENT_CONTEXT(ctx);
@@ -872,10 +275,10 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size,
    int ret;
 
    if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq, 
+      fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq,
 	      writefreq, priority);
 
-   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->r200Screen->gartTextures.map)
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->radeon.radeonScreen->gartTextures.map)
       return NULL;
 
    if (getenv("R200_NO_ALLOC"))
@@ -886,17 +289,17 @@ void *r200AllocateMemoryMESA(__DRIscreen *screen, GLsizei size,
    alloc.size = size;
    alloc.region_offset = &region_offset;
 
-   ret = drmCommandWriteRead( rmesa->r200Screen->driScreen->fd,
+   ret = drmCommandWriteRead( rmesa->radeon.radeonScreen->driScreen->fd,
 			      DRM_RADEON_ALLOC,
 			      &alloc, sizeof(alloc));
-   
+
    if (ret) {
       fprintf(stderr, "%s: DRM_RADEON_ALLOC ret %d\n", __FUNCTION__, ret);
       return NULL;
    }
-   
+
    {
-      char *region_start = (char *)rmesa->r200Screen->gartTextures.map;
+      char *region_start = (char *)rmesa->radeon.radeonScreen->gartTextures.map;
       return (void *)(region_start + region_offset);
    }
 }
@@ -914,28 +317,28 @@ void r200FreeMemoryMESA(__DRIscreen *screen, GLvoid *pointer)
    if (R200_DEBUG & DEBUG_IOCTL)
       fprintf(stderr, "%s %p\n", __FUNCTION__, pointer);
 
-   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->r200Screen->gartTextures.map) {
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || !rmesa->radeon.radeonScreen->gartTextures.map) {
       fprintf(stderr, "%s: no context\n", __FUNCTION__);
       return;
    }
 
-   region_offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map;
+   region_offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map;
 
-   if (region_offset < 0 || 
-       region_offset > rmesa->r200Screen->gartTextures.size) {
+   if (region_offset < 0 ||
+       region_offset > rmesa->radeon.radeonScreen->gartTextures.size) {
       fprintf(stderr, "offset %d outside range 0..%d\n", region_offset,
-	      rmesa->r200Screen->gartTextures.size);
+	      rmesa->radeon.radeonScreen->gartTextures.size);
       return;
    }
 
    memfree.region = RADEON_MEM_REGION_GART;
    memfree.region_offset = region_offset;
-   
-   ret = drmCommandWrite( rmesa->r200Screen->driScreen->fd,
+
+   ret = drmCommandWrite( rmesa->radeon.radeonScreen->driScreen->fd,
 			  DRM_RADEON_FREE,
 			  &memfree, sizeof(memfree));
-   
-   if (ret) 
+
+   if (ret)
       fprintf(stderr, "%s: DRM_RADEON_FREE ret %d\n", __FUNCTION__, ret);
 }
 
@@ -956,32 +359,32 @@ GLuint r200GetMemoryOffsetMESA(__DRIscreen *screen, const GLvoid *pointer)
 
    card_offset = r200GartOffsetFromVirtual( rmesa, pointer );
 
-   return card_offset - rmesa->r200Screen->gart_base;
+   return card_offset - rmesa->radeon.radeonScreen->gart_base;
 }
 
 GLboolean r200IsGartMemory( r200ContextPtr rmesa, const GLvoid *pointer,
 			   GLint size )
 {
-   ptrdiff_t offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map;
+   ptrdiff_t offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map;
    int valid = (size >= 0 &&
 		offset >= 0 &&
-		offset + size < rmesa->r200Screen->gartTextures.size);
+		offset + size < rmesa->radeon.radeonScreen->gartTextures.size);
 
    if (R200_DEBUG & DEBUG_IOCTL)
       fprintf(stderr, "r200IsGartMemory( %p ) : %d\n", pointer, valid );
-   
+
    return valid;
 }
 
 
 GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer )
 {
-   ptrdiff_t offset = (char *)pointer - (char *)rmesa->r200Screen->gartTextures.map;
+   ptrdiff_t offset = (char *)pointer - (char *)rmesa->radeon.radeonScreen->gartTextures.map;
 
-   if (offset < 0 || offset > rmesa->r200Screen->gartTextures.size)
+   if (offset < 0 || offset > rmesa->radeon.radeonScreen->gartTextures.size)
       return ~0;
    else
-      return rmesa->r200Screen->gart_texture_offset + offset;
+      return rmesa->radeon.radeonScreen->gart_texture_offset + offset;
 }
 
 
@@ -989,7 +392,7 @@ GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer )
 void r200InitIoctlFuncs( struct dd_function_table *functions )
 {
     functions->Clear = r200Clear;
-    functions->Finish = r200Finish;
-    functions->Flush = r200Flush;
+    functions->Finish = radeonFinish;
+    functions->Flush = radeonFlush;
 }
 
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.h b/src/mesa/drivers/dri/r200/r200_ioctl.h
index f7458e4a0e..2a4b8a11f4 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.h
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.h
@@ -37,65 +37,30 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "main/simple_list.h"
 #include "radeon_dri.h"
-#include "r200_lock.h"
+
+#include "radeon_bocs_wrapper.h"
 
 #include "xf86drm.h"
 #include "drm.h"
 #include "radeon_drm.h"
 
-extern void r200EmitState( r200ContextPtr rmesa );
 extern void r200EmitVertexAOS( r200ContextPtr rmesa,
-				 GLuint vertex_size,
-				 GLuint offset );
+			       GLuint vertex_size,
+			       struct radeon_bo *bo,
+			       GLuint offset );
 
 extern void r200EmitVbufPrim( r200ContextPtr rmesa,
 				GLuint primitive,
 				GLuint vertex_nr );
 
-extern void r200FlushElts( r200ContextPtr rmesa );
+extern void r200FlushElts(GLcontext *ctx);
 
 extern GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
 					   GLuint primitive,
 					   GLuint min_nr );
 
-extern void r200EmitAOS( r200ContextPtr rmesa,
-			   struct r200_dma_region **regions,
-			   GLuint n,
-			   GLuint offset );
-
-extern void r200EmitBlit( r200ContextPtr rmesa,
-			  GLuint color_fmt,
-			  GLuint src_pitch,
-			  GLuint src_offset,
-			  GLuint dst_pitch,
-			  GLuint dst_offset,
-			  GLint srcx, GLint srcy,
-			  GLint dstx, GLint dsty,
-			  GLuint w, GLuint h );
-
-extern void r200EmitWait( r200ContextPtr rmesa, GLuint flags );
-
-extern void r200FlushCmdBuf( r200ContextPtr rmesa, const char * );
-extern int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller );
-
-extern void r200RefillCurrentDmaRegion( r200ContextPtr rmesa );
-
-extern void r200AllocDmaRegion( r200ContextPtr rmesa,
-				  struct r200_dma_region *region,
-				  int bytes, 
-				  int alignment );
-
-extern void r200ReleaseDmaRegion( r200ContextPtr rmesa,
-				    struct r200_dma_region *region,
-				    const char *caller );
-
-extern void r200CopyBuffer( __DRIdrawablePrivate *drawable,
-			    const drm_clip_rect_t      *rect);
-extern void r200PageFlip( __DRIdrawablePrivate *drawable );
-extern void r200Flush( GLcontext *ctx );
-extern void r200Finish( GLcontext *ctx );
-extern void r200WaitForIdleLocked( r200ContextPtr rmesa );
-extern void r200WaitForVBlank( r200ContextPtr rmesa );
+extern void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset);
+
 extern void r200InitIoctlFuncs( struct dd_function_table *functions );
 
 extern void *r200AllocateMemoryMESA( __DRIscreen *screen, GLsizei size, GLfloat readfreq,
@@ -119,8 +84,8 @@ void r200SetUpAtomList( r200ContextPtr rmesa );
  */
 #define R200_NEWPRIM( rmesa )			\
 do {						\
-   if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );	\
+   if ( rmesa->radeon.dma.flush )			\
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
 } while (0)
 
 /* Can accomodate several state changes and primitive changes without
@@ -130,7 +95,7 @@ do {						\
 do {								\
    R200_NEWPRIM( rmesa );					\
    rmesa->hw.ATOM.dirty = GL_TRUE;				\
-   rmesa->hw.is_dirty = GL_TRUE;				\
+   rmesa->radeon.hw.is_dirty = GL_TRUE;				\
 } while (0)
 
 #define R200_DB_STATE( ATOM )			        \
@@ -139,13 +104,13 @@ do {								\
 
 static INLINE int R200_DB_STATECHANGE( 
    r200ContextPtr rmesa,
-   struct r200_state_atom *atom )
+   struct radeon_state_atom *atom )
 {
    if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) {
-      int *tmp;
+      GLuint *tmp;
       R200_NEWPRIM( rmesa );
       atom->dirty = GL_TRUE;
-      rmesa->hw.is_dirty = GL_TRUE;
+      rmesa->radeon.hw.is_dirty = GL_TRUE;
       tmp = atom->cmd; 
       atom->cmd = atom->lastcmd;
       atom->lastcmd = tmp;
@@ -156,15 +121,6 @@ static INLINE int R200_DB_STATECHANGE(
 }
 
 
-/* Fire the buffered vertices no matter what.
- */
-#define R200_FIREVERTICES( rmesa )			\
-do {							\
-   if ( rmesa->store.cmd_used || rmesa->dma.flush ) {	\
-      r200Flush( rmesa->glCtx );			\
-   }							\
-} while (0)
-
 /* Command lengths.  Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ
  * are available, you will also be adding an rmesa->state.max_state_size because
  * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts.
@@ -174,36 +130,36 @@ do {							\
 #define ELTS_BUFSZ(nr)	(12 + nr * 2)
 #define VBUF_BUFSZ	(3 * sizeof(int))
 
-/* Ensure that a minimum amount of space is available in the command buffer.
- * This is used to ensure atomicity of state updates with the rendering requests
- * that rely on them.
- *
- * An alternative would be to implement a "soft lock" such that when the buffer
- * wraps at an inopportune time, we grab the lock, flush the current buffer,
- * and hang on to the lock until the critical section is finished and we flush
- * the buffer again and unlock.
- */
-static INLINE void r200EnsureCmdBufSpace( r200ContextPtr rmesa, int bytes )
+static inline uint32_t cmdpacket3(int cmd_type)
 {
-   if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
-      r200FlushCmdBuf( rmesa, __FUNCTION__ );
-   assert( bytes <= R200_CMD_BUF_SZ );
-}
+  drm_radeon_cmd_header_t cmd;
 
-/* Alloc space in the command buffer
- */
-static INLINE char *r200AllocCmdBuf( r200ContextPtr rmesa,
-					 int bytes, const char *where )
-{
-   char * head;
+  cmd.i = 0;
+  cmd.header.cmd_type = cmd_type;
 
-   if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
-      r200FlushCmdBuf( rmesa, where );
+  return (uint32_t)cmd.i;
 
-   head = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-   rmesa->store.cmd_used += bytes;
-   assert( rmesa->store.cmd_used <= R200_CMD_BUF_SZ );
-   return head;
 }
 
+#define OUT_BATCH_PACKET3(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3));				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+#define OUT_BATCH_PACKET3_CLIP(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3_CLIP));	      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+
 #endif /* __R200_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r200/r200_lock.c b/src/mesa/drivers/dri/r200/r200_lock.c
deleted file mode 100644
index 99661a4bfb..0000000000
--- a/src/mesa/drivers/dri/r200/r200_lock.c
+++ /dev/null
@@ -1,116 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
- 
-#include "r200_context.h"
-#include "r200_lock.h"
-#include "r200_tex.h"
-#include "r200_state.h"
-#include "r200_ioctl.h"
-
-#include "drirenderbuffer.h"
-
-
-#if DEBUG_LOCKING
-char *prevLockFile = NULL;
-int prevLockLine = 0;
-#endif
-
-/* Turn on/off page flipping according to the flags in the sarea:
- */
-static void
-r200UpdatePageFlipping( r200ContextPtr rmesa )
-{
-   rmesa->doPageFlip = rmesa->sarea->pfState;
-   if (rmesa->glCtx->WinSysDrawBuffer) {
-      driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-                           rmesa->sarea->pfCurrentPage);
-   }
-}
-
-
-
-/* Update the hardware state.  This is called if another main/context.has
- * grabbed the hardware lock, which includes the X server.  This
- * function also updates the driver's window state after the X server
- * moves, resizes or restacks a window -- the change will be reflected
- * in the drawable position and clip rects.  Since the X server grabs
- * the hardware lock when it changes the window state, this routine will
- * automatically be called after such a change.
- */
-void r200GetLock( r200ContextPtr rmesa, GLuint flags )
-{
-   __DRIdrawablePrivate *drawable = rmesa->dri.drawable;
-   __DRIdrawablePrivate *readable = rmesa->dri.readable;
-   __DRIscreenPrivate *sPriv = rmesa->dri.screen;
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-   int i;
-
-   drmGetLock( rmesa->dri.fd, rmesa->dri.hwContext, flags );
-
-   /* The window might have moved, so we might need to get new clip
-    * rects.
-    *
-    * NOTE: This releases and regrabs the hw lock to allow the X server
-    * to respond to the DRI protocol request for new drawable info.
-    * Since the hardware state depends on having the latest drawable
-    * clip rects, all state checking must be done _after_ this call.
-    */
-   DRI_VALIDATE_DRAWABLE_INFO( sPriv, drawable );
-   if (drawable != readable) {
-      DRI_VALIDATE_DRAWABLE_INFO( sPriv, readable );
-   }
-
-   if ( rmesa->lastStamp != drawable->lastStamp ) {
-      r200UpdatePageFlipping( rmesa );
-      r200SetCliprects( rmesa );
-      r200UpdateViewportOffset( rmesa->glCtx );
-      driUpdateFramebufferSize(rmesa->glCtx, drawable);
-   }
-
-   R200_STATECHANGE( rmesa, ctx );
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
-   }
-   else rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &= ~R200_COLOR_TILE_ENABLE;
-
-   if ( sarea->ctx_owner != rmesa->dri.hwContext ) {
-      sarea->ctx_owner = rmesa->dri.hwContext;
-   }
-
-   for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-      DRI_AGE_TEXTURES( rmesa->texture_heaps[ i ] );
-   }
-
-   rmesa->lost_context = GL_TRUE;
-}
diff --git a/src/mesa/drivers/dri/r200/r200_lock.h b/src/mesa/drivers/dri/r200/r200_lock.h
deleted file mode 100644
index 4ff98907fb..0000000000
--- a/src/mesa/drivers/dri/r200/r200_lock.h
+++ /dev/null
@@ -1,106 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_LOCK_H__
-#define __R200_LOCK_H__
-
-extern void r200GetLock( r200ContextPtr rmesa, GLuint flags );
-
-/* Turn DEBUG_LOCKING on to find locking conflicts.
- */
-#define DEBUG_LOCKING	0
-
-#if DEBUG_LOCKING
-extern char *prevLockFile;
-extern int prevLockLine;
-
-#define DEBUG_LOCK()							\
-   do {									\
-      prevLockFile = (__FILE__);					\
-      prevLockLine = (__LINE__);					\
-   } while (0)
-
-#define DEBUG_RESET()							\
-   do {									\
-      prevLockFile = 0;							\
-      prevLockLine = 0;							\
-   } while (0)
-
-#define DEBUG_CHECK_LOCK()						\
-   do {									\
-      if ( prevLockFile ) {						\
-	 fprintf( stderr,						\
-		  "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",	\
-		  prevLockFile, prevLockLine, __FILE__, __LINE__ );	\
-	 exit( 1 );							\
-      }									\
-   } while (0)
-
-#else
-
-#define DEBUG_LOCK()
-#define DEBUG_RESET()
-#define DEBUG_CHECK_LOCK()
-
-#endif
-
-/*
- * !!! We may want to separate locks from locks with validation.  This
- * could be used to improve performance for those things commands that
- * do not do any drawing !!!
- */
-
-
-/* Lock the hardware and validate our state.
- */
-#define LOCK_HARDWARE( rmesa )					\
-   do {								\
-      char __ret = 0;						\
-      DEBUG_CHECK_LOCK();					\
-      DRM_CAS( rmesa->dri.hwLock, rmesa->dri.hwContext,		\
-	       (DRM_LOCK_HELD | rmesa->dri.hwContext), __ret );	\
-      if ( __ret )						\
-	 r200GetLock( rmesa, 0 );				\
-      DEBUG_LOCK();						\
-   } while (0)
-
-#define UNLOCK_HARDWARE( rmesa )					\
-   do {									\
-      DRM_UNLOCK( rmesa->dri.fd,					\
-		  rmesa->dri.hwLock,					\
-		  rmesa->dri.hwContext );				\
-      DEBUG_RESET();							\
-   } while (0)
-
-#endif /* __R200_LOCK_H__ */
diff --git a/src/mesa/drivers/dri/r200/r200_maos.h b/src/mesa/drivers/dri/r200/r200_maos.h
index d3ed06d402..16a70475e1 100644
--- a/src/mesa/drivers/dri/r200/r200_maos.h
+++ b/src/mesa/drivers/dri/r200/r200_maos.h
@@ -38,6 +38,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 
 extern void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev );
-extern void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs );
 
 #endif
diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
index 8512b9af47..383a0c4b0d 100644
--- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c
+++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
@@ -50,110 +50,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_maos.h"
 #include "r200_tcl.h"
 
-
-#if 0
-/* Usage:
- *   - from r200_tcl_render
- *   - call r200EmitArrays to ensure uptodate arrays in dma
- *   - emit primitives (new type?) which reference the data
- *       -- need to use elts for lineloop, quads, quadstrip/flat
- *       -- other primitives are all well-formed (need tristrip-1,fake-poly)
- *
- */
-static void emit_ubyte_rgba3( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   r200_color_t *out = (r200_color_t *)(rvb->start + rvb->address);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p\n",
-	      __FUNCTION__, count, stride, (void *)out);
-
-   for (i = 0; i < count; i++) {
-      out->red   = *data;
-      out->green = *(data+1);
-      out->blue  = *(data+2);
-      out->alpha = 0xFF;
-      out++;
-      data += stride;
-   }
-}
-
-static void emit_ubyte_rgba4( GLcontext *ctx,
-			      struct r200_dma_region *rvb,
-			      char *data,
-			      int stride,
-			      int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 4) {
-      for (i = 0; i < count; i++)
-	 ((int *)out)[i] = LE32_TO_CPU(((int *)data)[i]);
-   } else {
-      for (i = 0; i < count; i++) {
-	 *(int *)out++ = LE32_TO_CPU(*(int *)data);
-	 data += stride;
-      }
-   }
-}
-
-
-static void emit_ubyte_rgba( GLcontext *ctx,
-			     struct r200_dma_region *rvb,
-			     char *data,
-			     int size,
-			     int stride,
-			     int count )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      r200AllocDmaRegion( rmesa, rvb, 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
-   }
-   else {
-      r200AllocDmaRegion( rmesa, rvb, 4 * count, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 3:
-      emit_ubyte_rgba3( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_ubyte_rgba4( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-}
-#endif
-
-
 #if defined(USE_X86_ASM)
 #define COPY_DWORDS( dst, src, nr )					\
 do {									\
@@ -174,204 +70,34 @@ do {						\
 } while (0)
 #endif
 
-
-static void emit_vecfog( GLcontext *ctx,
-			 struct r200_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
+static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos,
+			     GLvoid *data, int stride, int count)
 {
-   int i;
-   GLfloat *out;
-
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      r200AllocDmaRegion( rmesa, rvb, 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
-   }
-   else {
-      r200AllocDmaRegion( rmesa, rvb, count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
-
-   /* Emit the data
-    */
-   out = (GLfloat *)(rvb->address + rvb->start);
-   for (i = 0; i < count; i++) {
-      out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data );
-      out++;
-      data += stride;
-   }
-
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	uint32_t *out;
+	int i;
+	int size = 1;
+
+	if (stride == 0) {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		count = 1;
+		aos->stride = 0;
+	} else {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		aos->stride = size;
+	}
+
+	aos->components = size;
+	aos->count = count;
+
+	out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
+	for (i = 0; i < count; i++) {
+	  out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data );
+	  out++;
+	  data += stride;
+	}
 }
 
-
-static void emit_vec4( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 4)
-      COPY_DWORDS( out, data, count );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out++;
-	 data += stride;
-      }
-}
-
-
-static void emit_vec8( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 8)
-      COPY_DWORDS( out, data, count*2 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out += 2;
-	 data += stride;
-      }
-}
-
-static void emit_vec12( GLcontext *ctx,
-		       struct r200_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-	      __FUNCTION__, count, stride, (void *)out, (void *)data);
-
-   if (stride == 12)
-      COPY_DWORDS( out, data, count*3 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out += 3;
-	 data += stride;
-      }
-}
-
-static void emit_vec16( GLcontext *ctx,
-			struct r200_dma_region *rvb,
-			char *data,
-			int stride,
-			int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 16)
-      COPY_DWORDS( out, data, count*4 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out[3] = *(int *)(data+12);
-	 out += 4;
-	 data += stride;
-      }
-}
-
-
-static void emit_vector( GLcontext *ctx,
-			 struct r200_dma_region *rvb,
-			 char *data,
-			 int size,
-			 int stride,
-			 int count )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d size %d stride %d\n",
-	      __FUNCTION__, count, size, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      r200AllocDmaRegion( rmesa, rvb, size * 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = size;
-   }
-   else {
-      r200AllocDmaRegion( rmesa, rvb, size * count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = size;
-      rvb->aos_size = size;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 1:
-      emit_vec4( ctx, rvb, data, stride, count );
-      break;
-   case 2:
-      emit_vec8( ctx, rvb, data, stride, count );
-      break;
-   case 3:
-      emit_vec12( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_vec16( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-
-}
-
-
-
 /* Emit any changed arrays to new GART memory, re-emit a packet to
  * update the arrays.  
  */
@@ -379,12 +105,12 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 {
    r200ContextPtr rmesa = R200_CONTEXT( ctx );
    struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
-   struct r200_dma_region **component = rmesa->tcl.aos_components;
    GLuint nr = 0;
    GLuint vfmt0 = 0, vfmt1 = 0;
    GLuint count = VB->Count;
    GLuint i, emitsize;
 
+   //   fprintf(stderr,"emit arrays\n");
    for ( i = 0; i < 15; i++ ) {
       GLubyte attrib = vimap_rev[i];
       if (attrib != 255) {
@@ -416,20 +142,20 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 	 case 3:
 	    /* special handling to fix up fog. Will get us into trouble with vbos...*/
 	    assert(attrib == VERT_ATTRIB_FOG);
-	    if (!rmesa->tcl.vertex_data[i].buf) {
+	    if (!rmesa->radeon.tcl.aos[i].bo) {
 	       if (ctx->VertexProgram._Enabled)
-		  emit_vector( ctx,
-			 &(rmesa->tcl.vertex_data[i]),
-			 (char *)VB->AttribPtr[attrib]->data,
-			 1,
-			 VB->AttribPtr[attrib]->stride,
-			 count);
+		  rcommon_emit_vector( ctx,
+				       &(rmesa->radeon.tcl.aos[nr]),
+				       (char *)VB->AttribPtr[attrib]->data,
+				       1,
+				       VB->AttribPtr[attrib]->stride,
+				       count);
 	       else
-		  emit_vecfog( ctx,
-			 &(rmesa->tcl.vertex_data[i]),
-			 (char *)VB->AttribPtr[attrib]->data,
-			 VB->AttribPtr[attrib]->stride,
-			 count);
+		 r200_emit_vecfog( ctx,
+				   &(rmesa->radeon.tcl.aos[nr]),
+				   (char *)VB->AttribPtr[attrib]->data,
+				   VB->AttribPtr[attrib]->stride,
+				   count);
 	    }
 	    vfmt0 |= R200_VTX_DISCRETE_FOG;
 	    goto after_emit;
@@ -473,17 +199,17 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 	 default:
 	    assert(0);
 	 }
-	 if (!rmesa->tcl.vertex_data[i].buf) {
-	    emit_vector( ctx,
-			 &(rmesa->tcl.vertex_data[i]),
-			 (char *)VB->AttribPtr[attrib]->data,
-			 emitsize,
-			 VB->AttribPtr[attrib]->stride,
-			 count );
+	 if (!rmesa->radeon.tcl.aos[nr].bo) {
+	   rcommon_emit_vector( ctx,
+				&(rmesa->radeon.tcl.aos[nr]),
+				(char *)VB->AttribPtr[attrib]->data,
+				emitsize,
+				VB->AttribPtr[attrib]->stride,
+				count );
 	 }
 after_emit:
 	 assert(nr < 12);
-	 component[nr++] = &rmesa->tcl.vertex_data[i];
+	 nr++;
       }
    }
 
@@ -494,19 +220,6 @@ after_emit:
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1;
    }
 
-   rmesa->tcl.nr_aos_components = nr;
+   rmesa->radeon.tcl.aos_count = nr;
 }
 
-
-void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   /* only do it for changed inputs ? */
-   int i;
-   for (i = 0; i < 15; i++) {
-      if (newinputs & (1 << i))
-	 r200ReleaseDmaRegion( rmesa,
-	    &rmesa->tcl.vertex_data[i], __FUNCTION__ );
-   }
-}
diff --git a/src/mesa/drivers/dri/r200/r200_pixel.c b/src/mesa/drivers/dri/r200/r200_pixel.c
index 2797cbb3dc..654f2c6ae9 100644
--- a/src/mesa/drivers/dri/r200/r200_pixel.c
+++ b/src/mesa/drivers/dri/r200/r200_pixel.c
@@ -51,7 +51,7 @@ check_color( const GLcontext *ctx, GLenum type, GLenum format,
 	     const void *pixels, GLint sz, GLint pitch )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint cpp = rmesa->r200Screen->cpp;
+   GLuint cpp = rmesa->radeon.radeonScreen->cpp;
 
    if (R200_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
@@ -65,8 +65,8 @@ check_color( const GLcontext *ctx, GLenum type, GLenum format,
       return GL_FALSE;
    }
 
-   if ( type == GL_UNSIGNED_INT_8_8_8_8_REV && 
-	cpp == 4 && 
+   if ( type == GL_UNSIGNED_INT_8_8_8_8_REV &&
+	cpp == 4 &&
 	format == GL_BGRA ) {
       if (R200_DEBUG & DEBUG_PIXEL)
 	 fprintf(stderr, "%s: passed 2\n", __FUNCTION__);
@@ -83,7 +83,7 @@ static GLboolean
 check_color_per_fragment_ops( const GLcontext *ctx )
 {
    int result;
-   result = (!(     ctx->Color.AlphaEnabled || 
+   result = (!(     ctx->Color.AlphaEnabled ||
 		    ctx->Depth.Test ||
 		    ctx->Fog.Enabled ||
 		    ctx->Scissor.Enabled ||
@@ -96,7 +96,7 @@ check_color_per_fragment_ops( const GLcontext *ctx )
 		    ctx->Texture._EnabledUnits
            ) &&
 	   ctx->Current.RasterPosValid);
-   
+
    return result;
 }
 
@@ -137,8 +137,8 @@ clip_pixelrect( const GLcontext *ctx,
    if (*height <= 0)
       return GL_FALSE;
 
-   *size = ((*y + *height - 1) * rmesa->r200Screen->frontPitch +
-	    (*x + *width - 1) * rmesa->r200Screen->cpp);
+   *size = ((*y + *height - 1) * rmesa->radeon.radeonScreen->frontPitch +
+	    (*x + *width - 1) * rmesa->radeon.radeonScreen->cpp);
 
    return GL_TRUE;
 }
@@ -153,19 +153,20 @@ r200TryReadPixels( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLint pitch = pack->RowLength ? pack->RowLength : width;
    GLint blit_format;
-   GLuint cpp = rmesa->r200Screen->cpp;
+   GLuint cpp = rmesa->radeon.radeonScreen->cpp;
    GLint size = width * height * cpp;
 
+   return GL_FALSE;
+#if 0
    if (R200_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
    /* Only accelerate reading to GART buffers.
     */
-   if ( !r200IsGartMemory(rmesa, pixels, 
-			 pitch * height * rmesa->r200Screen->cpp ) ) {
+   if ( !r200IsGartMemory(rmesa, pixels,
+			 pitch * height * rmesa->radeon.radeonScreen->cpp ) ) {
       if (R200_DEBUG & DEBUG_PIXEL)
 	 fprintf(stderr, "%s: dest not GART\n", __FUNCTION__);
-      return GL_FALSE;
    }
 
    /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from
@@ -180,7 +181,7 @@ r200TryReadPixels( GLcontext *ctx,
    if (!check_color(ctx, type, format, pack, pixels, size, pitch))
       return GL_FALSE;
 
-   switch ( rmesa->r200Screen->cpp ) {
+   switch ( rmesa->radeon.radeonScreen->cpp ) {
    case 4:
       blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
       break;
@@ -197,14 +198,14 @@ r200TryReadPixels( GLcontext *ctx,
     * a full command buffer expects to be called unlocked.  As a
     * workaround, immediately flush the buffer on aquiring the lock.
     */
-   LOCK_HARDWARE( rmesa );
+   LOCK_HARDWARE( &rmesa->radeon );
 
    if (rmesa->store.cmd_used)
-      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
 
    if (!clip_pixelrect(ctx, ctx->ReadBuffer, &x, &y, &width, &height,
 		       &size)) {
-      UNLOCK_HARDWARE( rmesa );
+      UNLOCK_HARDWARE( &rmesa->radeon );
       if (R200_DEBUG & DEBUG_PIXEL)
 	 fprintf(stderr, "%s totally clipped -- nothing to do\n",
 		 __FUNCTION__);
@@ -212,18 +213,18 @@ r200TryReadPixels( GLcontext *ctx,
    }
 
    {
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+      __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable;
       driRenderbuffer *drb = (driRenderbuffer *) ctx->ReadBuffer->_ColorReadBuffer;
       int nbox = dPriv->numClipRects;
       int src_offset = drb->offset
-		     + rmesa->r200Screen->fbLocation;
+		     + rmesa->radeon.radeonScreen->fbLocation;
       int src_pitch = drb->pitch * drb->cpp;
       int dst_offset = r200GartOffsetFromVirtual( rmesa, pixels );
-      int dst_pitch = pitch * rmesa->r200Screen->cpp;
+      int dst_pitch = pitch * rmesa->radeon.radeonScreen->cpp;
       drm_clip_rect_t *box = dPriv->pClipRects;
       int i;
 
-      r200EmitWait( rmesa, RADEON_WAIT_3D ); 
+      r200EmitWait( rmesa, RADEON_WAIT_3D );
 
       y = dPriv->h - y - height;
       x += dPriv->x;
@@ -240,7 +241,7 @@ r200TryReadPixels( GLcontext *ctx,
 	 GLint by = box[i].y1;
 	 GLint bw = box[i].x2 - bx;
 	 GLint bh = box[i].y2 - by;
-	 
+
 	 if (bx < x) bw -= x - bx, bx = x;
 	 if (by < y) bh -= y - by, by = y;
 	 if (bx + bw > x + width) bw = x + width - bx;
@@ -257,12 +258,12 @@ r200TryReadPixels( GLcontext *ctx,
 		       bw, bh );
       }
 
-      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
    }
-   UNLOCK_HARDWARE( rmesa );
-
-   r200Finish( ctx ); /* required by GL */
+   UNLOCK_HARDWARE( &rmesa->radeon );
 
+   radeonFinish( ctx ); /* required by GL */
+#endif
    return GL_TRUE;
 }
 
@@ -276,9 +277,9 @@ r200ReadPixels( GLcontext *ctx,
    if (R200_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-   if (!r200TryReadPixels( ctx, x, y, width, height, format, type, pack, 
+   if (!r200TryReadPixels( ctx, x, y, width, height, format, type, pack,
 			   pixels))
-      _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack, 
+      _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack,
 			  pixels);
 }
 
@@ -292,7 +293,7 @@ static void do_draw_pix( GLcontext *ctx,
 			 GLuint planemask)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
    drm_clip_rect_t *box = dPriv->pClipRects;
    struct gl_renderbuffer *rb = ctx->ReadBuffer->_ColorDrawBuffers[0];
    driRenderbuffer *drb = (driRenderbuffer *) rb;
@@ -301,12 +302,12 @@ static void do_draw_pix( GLcontext *ctx,
    int blit_format;
    int size;
    int src_offset = r200GartOffsetFromVirtual( rmesa, pixels );
-   int src_pitch = pitch * rmesa->r200Screen->cpp;
+   int src_pitch = pitch * rmesa->radeon.radeonScreen->cpp;
 
    if (R200_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
-
-   switch ( rmesa->r200Screen->cpp ) {
+#if 0
+   switch ( rmesa->radeon.radeonScreen->cpp ) {
    case 2:
       blit_format = R200_CP_COLOR_FORMAT_RGB565;
       break;
@@ -318,17 +319,17 @@ static void do_draw_pix( GLcontext *ctx,
    }
 
 
-   LOCK_HARDWARE( rmesa );
+   LOCK_HARDWARE( &rmesa->radeon );
 
    if (rmesa->store.cmd_used)
-      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
 
    y -= height;			/* cope with pixel zoom */
-   
+
    if (!clip_pixelrect(ctx, ctx->DrawBuffer,
 		       &x, &y, &width, &height,
 		       &size)) {
-      UNLOCK_HARDWARE( rmesa );
+      UNLOCK_HARDWARE( &rmesa->radeon );
       return;
    }
 
@@ -357,15 +358,16 @@ static void do_draw_pix( GLcontext *ctx,
 		    blit_format,
 		    src_pitch, src_offset,
 		    drb->pitch * drb->cpp,
-		    drb->offset + rmesa->r200Screen->fbLocation,
+		    drb->offset + rmesa->radeon.radeonScreen->fbLocation,
 		    bx - x, by - y,
 		    bx, by,
 		    bw, bh );
    }
 
-   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
-   r200WaitForIdleLocked( rmesa ); /* required by GL */
-   UNLOCK_HARDWARE( rmesa );
+   rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
+   radeonWaitForIdleLocked( &rmesa->radeon ); /* required by GL */
+   UNLOCK_HARDWARE( &rmesa->radeon );
+#endif
 }
 
 
@@ -381,7 +383,7 @@ r200TryDrawPixels( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLint pitch = unpack->RowLength ? unpack->RowLength : width;
    GLuint planemask;
-   GLuint cpp = rmesa->r200Screen->cpp;
+   GLuint cpp = rmesa->radeon.radeonScreen->cpp;
    GLint size = height * pitch * cpp;
 
    if (R200_DEBUG & DEBUG_PIXEL)
@@ -395,7 +397,7 @@ r200TryDrawPixels( GLcontext *ctx,
    case GL_RGB:
    case GL_RGBA:
    case GL_BGRA:
-      planemask = r200PackColor(cpp,
+      planemask = radeonPackColor(cpp,
 				ctx->Color.ColorMask[RCOMP],
 				ctx->Color.ColorMask[GCOMP],
 				ctx->Color.ColorMask[BCOMP],
@@ -407,7 +409,7 @@ r200TryDrawPixels( GLcontext *ctx,
       if (planemask != ~0)
 	 return GL_FALSE;	/* fix me -- should be possible */
 
-      /* Can't do conversions on GART reads/draws. 
+      /* Can't do conversions on GART reads/draws.
        */
       if ( !r200IsGartMemory( rmesa, pixels, size ) ) {
 	 if (R200_DEBUG & DEBUG_PIXEL)
@@ -431,7 +433,7 @@ r200TryDrawPixels( GLcontext *ctx,
       return GL_FALSE;
    }
 
-   if ( r200IsGartMemory(rmesa, pixels, size) )
+   if (0)// r200IsGartMemory(rmesa, pixels, size) )
    {
       do_draw_pix( ctx, x, y, width, height, pitch, pixels, planemask );
       return GL_TRUE;
@@ -471,7 +473,7 @@ r200Bitmap( GLcontext *ctx, GLint px, GLint py,
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   if (rmesa->Fallback)
+   if (rmesa->radeon.Fallback)
       _swrast_Bitmap( ctx, px, py, width, height, unpack, bitmap );
    else
       r200PointsBitmap( ctx, px, py, width, height, unpack, bitmap );
@@ -482,9 +484,9 @@ r200Bitmap( GLcontext *ctx, GLint px, GLint py,
 void r200InitPixelFuncs( GLcontext *ctx )
 {
    if (!getenv("R200_NO_BLITS")) {
-      ctx->Driver.ReadPixels = r200ReadPixels;  
-      ctx->Driver.DrawPixels = r200DrawPixels; 
-      if (getenv("R200_HW_BITMAP")) 
+      ctx->Driver.ReadPixels = r200ReadPixels;
+      ctx->Driver.DrawPixels = r200DrawPixels;
+      if (getenv("R200_HW_BITMAP"))
 	 ctx->Driver.Bitmap = r200Bitmap;
    }
 }
diff --git a/src/mesa/drivers/dri/r200/r200_reg.h b/src/mesa/drivers/dri/r200/r200_reg.h
index 5ce287f7a5..526a624b69 100644
--- a/src/mesa/drivers/dri/r200/r200_reg.h
+++ b/src/mesa/drivers/dri/r200/r200_reg.h
@@ -463,8 +463,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define     R200_VSC_UPDATE_USER_COLOR_1_ENABLE    0x00020000
 /* gap */
 #define R200_SE_TCL_VECTOR_INDX_REG                0x2200
+#       define RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT  16
+#       define RADEON_VEC_INDX_DWORD_COUNT_SHIFT     28
 #define R200_SE_TCL_VECTOR_DATA_REG                0x2204
 #define R200_SE_TCL_SCALAR_INDX_REG                0x2208
+#       define RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT  16
 #define R200_SE_TCL_SCALAR_DATA_REG                0x220c
 /* gap */
 #define R200_SE_TCL_MATRIX_SEL_0                   0x2230
@@ -949,6 +952,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define     R200_LOD_BIAS_MASK                        (0xfff80000)
 #define     R200_LOD_BIAS_SHIFT                       19
 #define R200_PP_TXSIZE_0                  0x2c0c /* NPOT only */
+#define R200_PP_TX_WIDTHMASK_SHIFT 0
+#define R200_PP_TX_HEIGHTMASK_SHIFT 16
+
 #define R200_PP_TXPITCH_0                 0x2c10 /* NPOT only */
 #define R200_PP_BORDER_COLOR_0            0x2c14
 #define R200_PP_CUBIC_FACES_0             0x2c18
diff --git a/src/mesa/drivers/dri/r200/r200_span.c b/src/mesa/drivers/dri/r200/r200_span.c
deleted file mode 100644
index 9783678028..0000000000
--- a/src/mesa/drivers/dri/r200/r200_span.c
+++ /dev/null
@@ -1,307 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/colormac.h"
-#include "swrast/swrast.h"
-
-#include "r200_context.h"
-#include "r200_ioctl.h"
-#include "r200_state.h"
-#include "r200_span.h"
-#include "r200_tex.h"
-
-#define DBG 0
-
-/*
- * Note that all information needed to access pixels in a renderbuffer
- * should be obtained through the gl_renderbuffer parameter, not per-context
- * information.
- */
-#define LOCAL_VARS						\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;		\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
-   GLubyte *buf = (GLubyte *) drb->flippedData			\
-      + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;	\
-   GLuint p;							\
-   (void) p;
-
-#define LOCAL_DEPTH_VARS				\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;	\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;			\
-   GLuint xo = dPriv->x;				\
-   GLuint yo = dPriv->y;				\
-   GLubyte *buf = (GLubyte *) drb->Base.Data;
-
-#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
-
-#define Y_FLIP(Y) (bottom - (Y))
-
-#define HW_LOCK() 
-
-#define HW_UNLOCK()							
-
-
-
-/* ================================================================
- * Color buffer
- */
-
-/* 16 bit, RGB565 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    r200##x##_RGB565
-#define TAG2(x,y) r200##x##_RGB565##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
-#include "spantmp2.h"
-
-/* 32 bit, ARGB8888 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    r200##x##_ARGB8888
-#define TAG2(x,y) r200##x##_ARGB8888##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
-#include "spantmp2.h"
-
-
-/* ================================================================
- * Depth buffer
- */
-
-/* The Radeon family has depth tiling on all the time, so we have to convert
- * the x,y coordinates into the memory bus address (mba) in the same
- * manner as the engine.  In each case, the linear block address (ba)
- * is calculated, and then wired with x and y to produce the final
- * memory address.
- * The chip will do address translation on its own if the surface registers
- * are set up correctly. It is not quite enough to get it working with hyperz too...
- */
-
-/* extract bit 'b' of x, result is zero or one */
-#define BIT(x,b) ((x & (1<<b))>>b)
-
-static GLuint
-r200_mba_z32( driRenderbuffer *drb, GLint x, GLint y )
-{
-   GLuint pitch = drb->pitch;
-   if (drb->depthHasSurface) {
-      return 4 * (x + y * pitch);
-   }
-   else {
-      GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x7FF) >> 5);
-      GLuint a = 
-         (BIT(x,0) << 2) |
-         (BIT(y,0) << 3) |
-         (BIT(x,1) << 4) |
-         (BIT(y,1) << 5) |
-         (BIT(x,3) << 6) |
-         (BIT(x,4) << 7) |
-         (BIT(x,2) << 8) |
-         (BIT(y,2) << 9) |
-         (BIT(y,3) << 10) |
-         (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
-         ((b >> 1) << 12);
-      return a;
-   }
-}
-
-static GLuint
-r200_mba_z16( driRenderbuffer *drb, GLint x, GLint y )
-{
-   GLuint pitch = drb->pitch;
-   if (drb->depthHasSurface) {
-      return 2 * (x + y * pitch);
-   }
-   else {
-      GLuint b = ((y & 0x7FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x7FF) >> 6);
-      GLuint a = 
-         (BIT(x,0) << 1) |
-         (BIT(y,0) << 2) |
-         (BIT(x,1) << 3) |
-         (BIT(y,1) << 4) |
-         (BIT(x,2) << 5) |
-         (BIT(x,4) << 6) |
-         (BIT(x,5) << 7) |
-         (BIT(x,3) << 8) |
-         (BIT(y,2) << 9) |
-         (BIT(y,3) << 10) |
-         (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
-         ((b >> 1) << 12);
-      return a;
-   }
-}
-
-
-/* 16-bit depth buffer functions
- */
-#define VALUE_TYPE GLushort
-
-#define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo )) = d;
-
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)(buf + r200_mba_z16( drb, _x + xo, _y + yo ));
-
-#define TAG(x) r200##x##_z16
-#include "depthtmp.h"
-
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- */
-#define VALUE_TYPE GLuint
-
-#define WRITE_DEPTH( _x, _y, d )					\
-do {									\
-   GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xff000000;							\
-   tmp |= ((d) & 0x00ffffff);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint *)(buf + r200_mba_z32( drb, _x + xo,			\
-					 _y + yo )) & 0x00ffffff;
-
-#define TAG(x) r200##x##_z24_s8
-#include "depthtmp.h"
-
-
-/* ================================================================
- * Stencil buffer
- */
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- */
-#define WRITE_STENCIL( _x, _y, d )					\
-do {									\
-   GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0x00ffffff;							\
-   tmp |= (((d) & 0xff) << 24);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-
-#define READ_STENCIL( d, _x, _y )					\
-do {									\
-   GLuint offset = r200_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xff000000;							\
-   d = tmp >> 24;							\
-} while (0)
-
-#define TAG(x) r200##x##_z24_s8
-#include "stenciltmp.h"
-
-
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
-
-static void r200SpanRenderStart( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-   r200WaitForIdleLocked( rmesa );
-
-   /* Read & rewrite the first pixel in the frame buffer.  This should
-    * be a noop, right?  In fact without this conform fails as reading
-    * from the framebuffer sometimes produces old results -- the
-    * on-card read cache gets mixed up and doesn't notice that the
-    * framebuffer has been updated.
-    *
-    * In the worst case this is buggy too as p might get the wrong
-    * value first time, so really need a hidden pixel somewhere for this.
-    */
-   {
-      int p;
-      driRenderbuffer *drb =
-	 (driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0];
-      volatile int *buf =
-	 (volatile int *)(rmesa->dri.screen->pFB + drb->offset);
-      p = *buf;
-      *buf = p;
-   }
-}
-
-static void r200SpanRenderFinish( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   _swrast_flush( ctx );
-   UNLOCK_HARDWARE( rmesa );
-}
-
-void r200InitSpanFuncs( GLcontext *ctx )
-{
-   struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
-   swdd->SpanRenderStart          = r200SpanRenderStart;
-   swdd->SpanRenderFinish         = r200SpanRenderFinish; 
-}
-
-
-
-/**
- * Plug in the Get/Put routines for the given driRenderbuffer.
- */
-void
-radeonSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis)
-{
-   if (drb->Base.InternalFormat == GL_RGBA) {
-      if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
-         r200InitPointers_RGB565(&drb->Base);
-      }
-      else {
-         r200InitPointers_ARGB8888(&drb->Base);
-      }
-   }
-   else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
-      r200InitDepthPointers_z16(&drb->Base);
-   }
-   else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
-      r200InitDepthPointers_z24_s8(&drb->Base);
-   }
-   else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
-      r200InitStencilPointers_z24_s8(&drb->Base);
-   }
-}
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index 2fcc87c0f5..5a6fd20d8c 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -47,6 +47,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
 
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
@@ -77,7 +79,7 @@ static void r200AlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref )
 
    switch ( func ) {
    case GL_NEVER:
-      pp_misc |= R200_ALPHA_TEST_FAIL; 
+      pp_misc |= R200_ALPHA_TEST_FAIL;
       break;
    case GL_LESS:
       pp_misc |= R200_ALPHA_TEST_LESS;
@@ -114,8 +116,8 @@ static void r200BlendColor( GLcontext *ctx, const GLfloat cf[4] )
    CLAMPED_FLOAT_TO_UBYTE(color[1], cf[1]);
    CLAMPED_FLOAT_TO_UBYTE(color[2], cf[2]);
    CLAMPED_FLOAT_TO_UBYTE(color[3], cf[3]);
-   if (rmesa->r200Screen->drmSupportsBlendColor)
-      rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = r200PackColor( 4, color[0], color[1], color[2], color[3] );
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
+      rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = radeonPackColor( 4, color[0], color[1], color[2], color[3] );
 }
 
 /**
@@ -213,7 +215,7 @@ static void r200_set_blend_state( GLcontext * ctx )
 
    R200_STATECHANGE( rmesa, ctx );
 
-   if (rmesa->r200Screen->drmSupportsBlendColor) {
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
       if (ctx->Color.ColorLogicOpEnabled) {
          rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] =  cntl | R200_ROP_ENABLE;
          rmesa->hw.ctx.cmd[CTX_RB3D_ABLENDCNTL] = eqn | func;
@@ -278,7 +280,7 @@ static void r200_set_blend_state( GLcontext * ctx )
       return;
    }
 
-   if (!rmesa->r200Screen->drmSupportsBlendColor) {
+   if (!rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
       rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func;
       return;
    }
@@ -383,10 +385,10 @@ static void r200ClearDepth( GLcontext *ctx, GLclampd d )
 
    switch ( format ) {
    case R200_DEPTH_FORMAT_16BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x0000ffff;
+      rmesa->radeon.state.depth.clear = d * 0x0000ffff;
       break;
    case R200_DEPTH_FORMAT_24BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x00ffffff;
+      rmesa->radeon.state.depth.clear = d * 0x00ffffff;
       break;
    }
 }
@@ -477,10 +479,10 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 	 }
       }
       break;
-   case GL_FOG_COLOR: 
+   case GL_FOG_COLOR:
       R200_STATECHANGE( rmesa, ctx );
       UNCLAMPED_FLOAT_TO_RGB_CHAN( col, ctx->Fog.Color );
-      i = r200PackColor( 4, col[0], col[1], col[2], 0 );
+      i = radeonPackColor( 4, col[0], col[1], col[2], 0 );
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~R200_FOG_COLOR_MASK;
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= i;
       break;
@@ -505,7 +507,7 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 
       if (out_0 != rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0]) {
 	 R200_STATECHANGE( rmesa, vtx );
-	 rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] = out_0;	 
+	 rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] = out_0;
       }
 
       break;
@@ -521,102 +523,6 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
    }
 }
 
-
-/* =============================================================
- * Scissoring
- */
-
-
-static GLboolean intersect_rect( drm_clip_rect_t *out,
-				 drm_clip_rect_t *a,
-				 drm_clip_rect_t *b )
-{
-   *out = *a;
-   if ( b->x1 > out->x1 ) out->x1 = b->x1;
-   if ( b->y1 > out->y1 ) out->y1 = b->y1;
-   if ( b->x2 < out->x2 ) out->x2 = b->x2;
-   if ( b->y2 < out->y2 ) out->y2 = b->y2;
-   if ( out->x1 >= out->x2 ) return GL_FALSE;
-   if ( out->y1 >= out->y2 ) return GL_FALSE;
-   return GL_TRUE;
-}
-
-
-void r200RecalcScissorRects( r200ContextPtr rmesa )
-{
-   drm_clip_rect_t *out;
-   int i;
-
-   /* Grow cliprect store?
-    */
-   if (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-      while (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-	 rmesa->state.scissor.numAllocedClipRects += 1;	/* zero case */
-	 rmesa->state.scissor.numAllocedClipRects *= 2;
-      }
-
-      if (rmesa->state.scissor.pClipRects)
-	 FREE(rmesa->state.scissor.pClipRects);
-
-      rmesa->state.scissor.pClipRects = 
-	 MALLOC( rmesa->state.scissor.numAllocedClipRects * 
-		 sizeof(drm_clip_rect_t) );
-
-      if ( rmesa->state.scissor.pClipRects == NULL ) {
-	 rmesa->state.scissor.numAllocedClipRects = 0;
-	 return;
-      }
-   }
-   
-   out = rmesa->state.scissor.pClipRects;
-   rmesa->state.scissor.numClipRects = 0;
-
-   for ( i = 0 ; i < rmesa->numClipRects ;  i++ ) {
-      if ( intersect_rect( out, 
-			   &rmesa->pClipRects[i], 
-			   &rmesa->state.scissor.rect ) ) {
-	 rmesa->state.scissor.numClipRects++;
-	 out++;
-      }
-   }
-}
-
-
-static void r200UpdateScissor( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if ( rmesa->dri.drawable ) {
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-
-      int x = ctx->Scissor.X;
-      int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height;
-      int w = ctx->Scissor.X + ctx->Scissor.Width - 1;
-      int h = dPriv->h - ctx->Scissor.Y - 1;
-
-      rmesa->state.scissor.rect.x1 = x + dPriv->x;
-      rmesa->state.scissor.rect.y1 = y + dPriv->y;
-      rmesa->state.scissor.rect.x2 = w + dPriv->x + 1;
-      rmesa->state.scissor.rect.y2 = h + dPriv->y + 1;
-
-      r200RecalcScissorRects( rmesa );
-   }
-}
-
-
-static void r200Scissor( GLcontext *ctx,
-			   GLint x, GLint y, GLsizei w, GLsizei h )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if ( ctx->Scissor.Enabled ) {
-      R200_FIREVERTICES( rmesa );	/* don't pipeline cliprect changes */
-      r200UpdateScissor( ctx );
-   }
-
-}
-
-
 /* =============================================================
  * Culling
  */
@@ -668,6 +574,10 @@ static void r200FrontFace( GLcontext *ctx, GLenum mode )
    R200_STATECHANGE( rmesa, tcl );
    rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~R200_CULL_FRONT_IS_CCW;
 
+   /* Winding is inverted when rendering to FBO */
+   if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+      mode = (mode == GL_CW) ? GL_CCW : GL_CW;
+
    switch ( mode ) {
    case GL_CW:
       rmesa->hw.set.cmd[SET_SE_CNTL] |= R200_FFACE_CULL_CW;
@@ -790,7 +700,7 @@ static void r200LineStipple( GLcontext *ctx, GLint factor, GLushort pattern )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
    R200_STATECHANGE( rmesa, lin );
-   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] = 
+   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] =
       ((((GLuint)factor & 0xff) << 16) | ((GLuint)pattern));
 }
 
@@ -803,21 +713,27 @@ static void r200ColorMask( GLcontext *ctx,
 			   GLboolean b, GLboolean a )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint mask = r200PackColor( rmesa->r200Screen->cpp,
-				ctx->Color.ColorMask[RCOMP],
-				ctx->Color.ColorMask[GCOMP],
-				ctx->Color.ColorMask[BCOMP],
-				ctx->Color.ColorMask[ACOMP] );
-
+   GLuint mask;
+   struct radeon_renderbuffer *rrb;
    GLuint flag = rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] & ~R200_PLANE_MASK_ENABLE;
 
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
+   mask = radeonPackColor( rrb->cpp,
+			   ctx->Color.ColorMask[RCOMP],
+			   ctx->Color.ColorMask[GCOMP],
+			   ctx->Color.ColorMask[BCOMP],
+			   ctx->Color.ColorMask[ACOMP] );
+
+
    if (!(r && g && b && a))
       flag |= R200_PLANE_MASK_ENABLE;
 
-   if ( rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] != flag ) { 
-      R200_STATECHANGE( rmesa, ctx ); 
-      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = flag; 
-   } 
+   if ( rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] != flag ) {
+      R200_STATECHANGE( rmesa, ctx );
+      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = flag;
+   }
 
    if ( rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK] != mask ) {
       R200_STATECHANGE( rmesa, msk );
@@ -834,7 +750,8 @@ static void r200PolygonOffset( GLcontext *ctx,
 			       GLfloat factor, GLfloat units )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   float_ui32_type constant =  { units * rmesa->state.depth.scale };
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   float_ui32_type constant =  { units * depthScale };
    float_ui32_type factoru = { factor };
 
 /*    factor *= 2; */
@@ -861,15 +778,15 @@ static void r200PolygonStipple( GLcontext *ctx, const GLubyte *mask )
 
    /* TODO: push this into cmd mechanism
     */
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
+   radeon_firevertices(&rmesa->radeon);
+   LOCK_HARDWARE( &rmesa->radeon );
 
    /* FIXME: Use window x,y offsets into stipple RAM.
     */
    stipple.mask = rmesa->state.stipple.mask;
-   drmCommandWrite( rmesa->dri.fd, DRM_RADEON_STIPPLE, 
+   drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_STIPPLE,
                     &stipple, sizeof(stipple) );
-   UNLOCK_HARDWARE( rmesa );
+   UNLOCK_HARDWARE( &rmesa->radeon );
 }
 
 static void r200PolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
@@ -878,10 +795,10 @@ static void r200PolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
    GLboolean flag = (ctx->_TriangleCaps & DD_TRI_UNFILLED) != 0;
 
    /* Can't generally do unfilled via tcl, but some good special
-    * cases work. 
+    * cases work.
     */
    TCL_FALLBACK( ctx, R200_TCL_FALLBACK_UNFILLED, flag);
-   if (rmesa->TclFallback) {
+   if (rmesa->radeon.TclFallback) {
       r200ChooseRenderState( ctx );
       r200ChooseVertexState( ctx );
    }
@@ -920,34 +837,34 @@ static void r200UpdateSpecular( GLcontext *ctx )
 
    if (ctx->Light.Enabled &&
        ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
 	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT) |
-	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));	
+	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_0;
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_1;
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHTING_ENABLE;
       p |=  R200_SPECULAR_ENABLE;
-      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= 
+      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &=
 	 ~R200_DIFFUSE_SPECULAR_COMBINE;
    }
    else if (ctx->Light.Enabled) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
-	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));	
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
+	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_0;
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHTING_ENABLE;
    } else if (ctx->Fog.ColorSumEnabled ) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
 	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT) |
-	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));	
+	  (R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));
       p |=  R200_SPECULAR_ENABLE;
    } else {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
-	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));	
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
+	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT));
    }
 
    if (ctx->Fog.Enabled) {
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |= 
-	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));	
+      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] |=
+	 ((R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT));
       rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] |= R200_OUTPUT_COLOR_1;
    }
 
@@ -958,7 +875,7 @@ static void r200UpdateSpecular( GLcontext *ctx )
 
    /* Update vertex/render formats
     */
-   if (rmesa->TclFallback) { 
+   if (rmesa->radeon.TclFallback) {
       r200ChooseRenderState( ctx );
       r200ChooseVertexState( ctx );
    }
@@ -970,7 +887,7 @@ static void r200UpdateSpecular( GLcontext *ctx )
  */
 
 
-/* Update on colormaterial, material emmissive/ambient, 
+/* Update on colormaterial, material emmissive/ambient,
  * lightmodel.globalambient
  */
 static void update_global_ambient( GLcontext *ctx )
@@ -984,23 +901,23 @@ static void update_global_ambient( GLcontext *ctx )
     */
    if ((rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_1] &
        ((3 << R200_FRONT_EMISSIVE_SOURCE_SHIFT) |
-	(3 << R200_FRONT_AMBIENT_SOURCE_SHIFT))) == 0) 
+	(3 << R200_FRONT_AMBIENT_SOURCE_SHIFT))) == 0)
    {
-      COPY_3V( &fcmd[GLT_RED], 
+      COPY_3V( &fcmd[GLT_RED],
 	       ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_EMISSION]);
       ACC_SCALE_3V( &fcmd[GLT_RED],
 		   ctx->Light.Model.Ambient,
 		   ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT]);
-   } 
+   }
    else
    {
       COPY_3V( &fcmd[GLT_RED], ctx->Light.Model.Ambient );
    }
-   
+
    R200_DB_STATECHANGE(rmesa, &rmesa->hw.glt);
 }
 
-/* Update on change to 
+/* Update on change to
  *    - light[p].colors
  *    - light[p].enabled
  */
@@ -1014,10 +931,10 @@ static void update_light_colors( GLcontext *ctx, GLuint p )
       r200ContextPtr rmesa = R200_CONTEXT(ctx);
       float *fcmd = (float *)R200_DB_STATE( lit[p] );
 
-      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );	 
+      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );
       COPY_4V( &fcmd[LIT_DIFFUSE_RED], l->Diffuse );
       COPY_4V( &fcmd[LIT_SPECULAR_RED], l->Specular );
-      
+
       R200_DB_STATECHANGE( rmesa, &rmesa->hw.lit[p] );
    }
 }
@@ -1037,7 +954,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 
    if (ctx->Light.ColorMaterialEnabled) {
       GLuint mask = ctx->Light.ColorMaterialBitmask;
-   
+
       if (mask & MAT_BIT_FRONT_EMISSION) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_FRONT_EMISSIVE_SOURCE_SHIFT);
@@ -1053,7 +970,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
       else
          light_model_ctl1 |= (R200_LM1_SOURCE_MATERIAL_0 <<
 			     R200_FRONT_AMBIENT_SOURCE_SHIFT);
-	 
+
       if (mask & MAT_BIT_FRONT_DIFFUSE) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_FRONT_DIFFUSE_SOURCE_SHIFT);
@@ -1061,7 +978,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
       else
          light_model_ctl1 |= (R200_LM1_SOURCE_MATERIAL_0 <<
 			     R200_FRONT_DIFFUSE_SOURCE_SHIFT);
-   
+
       if (mask & MAT_BIT_FRONT_SPECULAR) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_FRONT_SPECULAR_SOURCE_SHIFT);
@@ -1070,7 +987,7 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
          light_model_ctl1 |= (R200_LM1_SOURCE_MATERIAL_0 <<
 			     R200_FRONT_SPECULAR_SOURCE_SHIFT);
       }
-   
+
       if (mask & MAT_BIT_BACK_EMISSION) {
 	 light_model_ctl1 |= (R200_LM1_SOURCE_VERTEX_COLOR_0 <<
 			     R200_BACK_EMISSIVE_SOURCE_SHIFT);
@@ -1120,8 +1037,8 @@ static void r200ColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
       R200_STATECHANGE( rmesa, tcl );
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_1] = light_model_ctl1;
    }
-   
-   
+
+
 }
 
 void r200UpdateMaterial( GLcontext *ctx )
@@ -1131,7 +1048,7 @@ void r200UpdateMaterial( GLcontext *ctx )
    GLfloat *fcmd = (GLfloat *)R200_DB_STATE( mtl[0] );
    GLfloat *fcmd2 = (GLfloat *)R200_DB_STATE( mtl[1] );
    GLuint mask = ~0;
-   
+
    /* Might be possible and faster to update everything unconditionally? */
    if (ctx->Light.ColorMaterialEnabled)
       mask &= ~ctx->Light.ColorMaterialBitmask;
@@ -1217,7 +1134,7 @@ void r200UpdateMaterial( GLcontext *ctx )
  *
  * which are calculated in light.c and are correct for the current
  * lighting space (model or eye), hence dependencies on _NEW_MODELVIEW
- * and _MESA_NEW_NEED_EYE_COORDS.  
+ * and _MESA_NEW_NEED_EYE_COORDS.
  */
 static void update_light( GLcontext *ctx )
 {
@@ -1234,8 +1151,8 @@ static void update_light( GLcontext *ctx )
 	 tmp &= ~R200_LIGHT_IN_MODELSPACE;
       else
 	 tmp |= R200_LIGHT_IN_MODELSPACE;
-      
-      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0]) 
+
+      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0])
       {
 	 R200_STATECHANGE( rmesa, tcl );
 	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] = tmp;
@@ -1259,10 +1176,10 @@ static void update_light( GLcontext *ctx )
 	 if (ctx->Light.Light[p].Enabled) {
 	    struct gl_light *l = &ctx->Light.Light[p];
 	    GLfloat *fcmd = (GLfloat *)R200_DB_STATE( lit[p] );
-	    
+
 	    if (l->EyePosition[3] == 0.0) {
-	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm ); 
-	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm ); 
+	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm );
+	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm );
 	       fcmd[LIT_POSITION_W] = 0;
 	       fcmd[LIT_DIRECTION_W] = 0;
 	    } else {
@@ -1286,21 +1203,21 @@ static void r200Lightfv( GLcontext *ctx, GLenum light,
    GLint p = light - GL_LIGHT0;
    struct gl_light *l = &ctx->Light.Light[p];
    GLfloat *fcmd = (GLfloat *)rmesa->hw.lit[p].cmd;
-   
+
 
    switch (pname) {
-   case GL_AMBIENT:		
+   case GL_AMBIENT:
    case GL_DIFFUSE:
    case GL_SPECULAR:
       update_light_colors( ctx, p );
       break;
 
-   case GL_SPOT_DIRECTION: 
-      /* picked up in update_light */	
+   case GL_SPOT_DIRECTION:
+      /* picked up in update_light */
       break;
 
    case GL_POSITION: {
-      /* positions picked up in update_light, but can do flag here */	
+      /* positions picked up in update_light, but can do flag here */
       GLuint flag = (p&1)? R200_LIGHT_1_IS_LOCAL : R200_LIGHT_0_IS_LOCAL;
       GLuint idx = TCL_PER_LIGHT_CTL_0 + p/2;
 
@@ -1416,7 +1333,7 @@ static void r200LightModelfv( GLcontext *ctx, GLenum pname,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
    switch (pname) {
-      case GL_LIGHT_MODEL_AMBIENT: 
+      case GL_LIGHT_MODEL_AMBIENT:
 	 update_global_ambient( ctx );
 	 break;
 
@@ -1430,7 +1347,7 @@ static void r200LightModelfv( GLcontext *ctx, GLenum pname,
 	    rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] |= R200_LIGHT_TWOSIDE;
 	 else
 	    rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= ~(R200_LIGHT_TWOSIDE);
-	 if (rmesa->TclFallback) {
+	 if (rmesa->radeon.TclFallback) {
 	    r200ChooseRenderState( ctx );
 	    r200ChooseVertexState( ctx );
 	 }
@@ -1675,7 +1592,7 @@ static void r200ClearStencil( GLcontext *ctx, GLint s )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   rmesa->state.stencil.clear = 
+   rmesa->radeon.state.stencil.clear =
       ((GLuint) (ctx->Stencil.Clear & 0xff) |
        (0xff << R200_STENCIL_MASK_SHIFT) |
        ((ctx->Stencil.WriteMask[0] & 0xff) << R200_STENCIL_WRITEMASK_SHIFT));
@@ -1700,19 +1617,29 @@ static void r200ClearStencil( GLcontext *ctx, GLint s )
 void r200UpdateWindow( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   GLfloat xoffset = (GLfloat)dPriv->x;
-   GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+   GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer ? (ctx->DrawBuffer->Name != 0) : 0);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat y_scale, y_bias;
+
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   } else {
+      y_scale = -1.0;
+      y_bias = yoffset;
+   }
 
    float_ui32_type sx = { v[MAT_SX] };
    float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X };
-   float_ui32_type sy = { - v[MAT_SY] };
-   float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y };
-   float_ui32_type sz = { v[MAT_SZ] * rmesa->state.depth.scale };
-   float_ui32_type tz = { v[MAT_TZ] * rmesa->state.depth.scale };
+   float_ui32_type sy = { v[MAT_SY] * y_scale };
+   float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y };
+   float_ui32_type sz = { v[MAT_SZ] * depthScale };
+   float_ui32_type tz = { v[MAT_TZ] * depthScale };
 
-   R200_FIREVERTICES( rmesa );
    R200_STATECHANGE( rmesa, vpt );
 
    rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE]  = sx.ui32;
@@ -1733,6 +1660,8 @@ static void r200Viewport( GLcontext *ctx, GLint x, GLint y,
     * values, or keep the originals hanging around.
     */
    r200UpdateWindow( ctx );
+
+   radeon_viewport(ctx, x, y, width, height);
 }
 
 static void r200DepthRange( GLcontext *ctx, GLclampd nearval,
@@ -1744,7 +1673,7 @@ static void r200DepthRange( GLcontext *ctx, GLclampd nearval,
 void r200UpdateViewportOffset( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
    GLfloat xoffset = (GLfloat)dPriv->x;
    GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
@@ -1774,8 +1703,8 @@ void r200UpdateViewportOffset( GLcontext *ctx )
                 R200_STIPPLE_Y_OFFSET_MASK);
 
          /* add magic offsets, then invert */
-         stx = 31 - ((rmesa->dri.drawable->x - 1) & R200_STIPPLE_COORD_MASK);
-         sty = 31 - ((rmesa->dri.drawable->y + rmesa->dri.drawable->h - 1)
+         stx = 31 - ((dPriv->x - 1) & R200_STIPPLE_COORD_MASK);
+         sty = 31 - ((dPriv->y + dPriv->h - 1)
                      & R200_STIPPLE_COORD_MASK);
 
          m |= ((stx << R200_STIPPLE_X_OFFSET_SHIFT) |
@@ -1788,7 +1717,7 @@ void r200UpdateViewportOffset( GLcontext *ctx )
       }
    }
 
-   r200UpdateScissor( ctx );
+   radeonUpdateScissor( ctx );
 }
 
 
@@ -1801,11 +1730,16 @@ static void r200ClearColor( GLcontext *ctx, const GLfloat c[4] )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLubyte color[4];
+   struct radeon_renderbuffer *rrb;
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
    CLAMPED_FLOAT_TO_UBYTE(color[0], c[0]);
    CLAMPED_FLOAT_TO_UBYTE(color[1], c[1]);
    CLAMPED_FLOAT_TO_UBYTE(color[2], c[2]);
    CLAMPED_FLOAT_TO_UBYTE(color[3], c[3]);
-   rmesa->state.color.clear = r200PackColor( rmesa->r200Screen->cpp,
+   rmesa->radeon.state.color.clear = radeonPackColor( rrb->cpp,
                                              color[0], color[1],
                                              color[2], color[3] );
 }
@@ -1848,96 +1782,6 @@ static void r200LogicOpCode( GLcontext *ctx, GLenum opcode )
    rmesa->hw.msk.cmd[MSK_RB3D_ROPCNTL] = r200_rop_tab[rop];
 }
 
-
-/*
- * Set up the cliprects for either front or back-buffer drawing.
- */
-void r200SetCliprects( r200ContextPtr rmesa )
-{
-   __DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-   __DRIdrawablePrivate *const readable = rmesa->dri.readable;
-   GLframebuffer *const draw_fb = (GLframebuffer*) drawable->driverPrivate;
-   GLframebuffer *const read_fb = (GLframebuffer*) readable->driverPrivate;
-
-   if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BIT_BACK_LEFT) {
-      /* Can't ignore 2d windows if we are page flipping.
-       */
-      if ( drawable->numBackClipRects == 0 || rmesa->doPageFlip ) {
-         rmesa->numClipRects = drawable->numClipRects;
-         rmesa->pClipRects = drawable->pClipRects;
-      }
-      else {
-         rmesa->numClipRects = drawable->numBackClipRects;
-         rmesa->pClipRects = drawable->pBackClipRects;
-      }
-   }
-   else {
-     /* front buffer (or none, or multiple buffers) */
-     rmesa->numClipRects = drawable->numClipRects;
-     rmesa->pClipRects = drawable->pClipRects;
-  }
-
-   if ((draw_fb->Width != drawable->w) || (draw_fb->Height != drawable->h)) {
-      _mesa_resize_framebuffer(rmesa->glCtx, draw_fb,
-			       drawable->w, drawable->h);
-      draw_fb->Initialized = GL_TRUE;
-   }
-
-   if (drawable != readable) {
-      if ((read_fb->Width != readable->w) ||
-	  (read_fb->Height != readable->h)) {
-	 _mesa_resize_framebuffer(rmesa->glCtx, read_fb,
-				  readable->w, readable->h);
-	 read_fb->Initialized = GL_TRUE;
-      }
-   }
-
-   if (rmesa->state.scissor.enabled)
-      r200RecalcScissorRects( rmesa );
-
-   rmesa->lastStamp = drawable->lastStamp;
-}
-
-
-static void r200DrawBuffer( GLcontext *ctx, GLenum mode )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s %s\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   R200_FIREVERTICES(rmesa);	/* don't pipeline cliprect changes */
-
-   if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) {
-      /* 0 (GL_NONE) buffers or multiple color drawing buffers */
-      FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   switch ( ctx->DrawBuffer->_ColorDrawBufferIndexes[0] ) {
-   case BUFFER_FRONT_LEFT:
-   case BUFFER_BACK_LEFT:
-      FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_FALSE );
-      break;
-   default:
-      FALLBACK( rmesa, R200_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   r200SetCliprects( rmesa );
-
-   /* We'll set the drawing engine's offset/pitch parameters later
-    * when we update other state.
-    */
-}
-
-
-static void r200ReadBuffer( GLcontext *ctx, GLenum mode )
-{
-   /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
-}
-
 /* =============================================================
  * State enable/disable
  */
@@ -1979,7 +1823,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_CLIP_PLANE2:
    case GL_CLIP_PLANE3:
    case GL_CLIP_PLANE4:
-   case GL_CLIP_PLANE5: 
+   case GL_CLIP_PLANE5:
       p = cap-GL_CLIP_PLANE0;
       R200_STATECHANGE( rmesa, tcl );
       if (state) {
@@ -2013,10 +1857,10 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
       R200_STATECHANGE(rmesa, ctx );
       if ( state ) {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->radeon.state.color.roundEnable;
       } else {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->radeon.state.color.roundEnable;
       }
       break;
 
@@ -2031,7 +1875,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
 	 rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~R200_TCL_FOG_MASK;
       }
       r200UpdateSpecular( ctx ); /* for PK_SPEC */
-      if (rmesa->TclFallback) 
+      if (rmesa->radeon.TclFallback)
 	 r200ChooseVertexState( ctx );
       _mesa_allow_light_in_model( ctx, !state );
       break;
@@ -2046,13 +1890,13 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_LIGHT7:
       R200_STATECHANGE(rmesa, tcl);
       p = cap - GL_LIGHT0;
-      if (p&1) 
+      if (p&1)
 	 flag = (R200_LIGHT_1_ENABLE |
-		 R200_LIGHT_1_ENABLE_AMBIENT | 
+		 R200_LIGHT_1_ENABLE_AMBIENT |
 		 R200_LIGHT_1_ENABLE_SPECULAR);
       else
 	 flag = (R200_LIGHT_0_ENABLE |
-		 R200_LIGHT_0_ENABLE_AMBIENT | 
+		 R200_LIGHT_0_ENABLE_AMBIENT |
 		 R200_LIGHT_0_ENABLE_SPECULAR);
 
       if (state)
@@ -2060,7 +1904,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
       else
 	 rmesa->hw.tcl.cmd[p/2 + TCL_PER_LIGHT_CTL_0] &= ~flag;
 
-      /* 
+      /*
        */
       update_light_colors( ctx, p );
       break;
@@ -2068,7 +1912,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_LIGHTING:
       r200UpdateSpecular(ctx);
       /* for reflection map fixup - might set recheck_texgen for all units too */
-      rmesa->NewGLState |= _NEW_TEXTURE;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE;
       break;
 
    case GL_LINE_SMOOTH:
@@ -2181,21 +2025,30 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    }
 
    case GL_SCISSOR_TEST:
-      R200_FIREVERTICES( rmesa );
-      rmesa->state.scissor.enabled = state;
-      r200UpdateScissor( ctx );
+      radeon_firevertices(&rmesa->radeon);
+      rmesa->radeon.state.scissor.enabled = state;
+      radeonUpdateScissor( ctx );
       break;
 
    case GL_STENCIL_TEST:
-      if ( rmesa->state.stencil.hwBuffer ) {
-	 R200_STATECHANGE( rmesa, ctx );
-	 if ( state ) {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_STENCIL_ENABLE;
+      {
+	 GLboolean hw_stencil = GL_FALSE;
+	 if (ctx->DrawBuffer) {
+	    struct radeon_renderbuffer *rrbStencil
+	       = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+	    hw_stencil = (rrbStencil && rrbStencil->bo);
+	 }
+
+	 if (hw_stencil) {
+	    R200_STATECHANGE( rmesa, ctx );
+	    if ( state ) {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_STENCIL_ENABLE;
+	    } else {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_STENCIL_ENABLE;
+	    }
 	 } else {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_STENCIL_ENABLE;
+	    FALLBACK( rmesa, R200_FALLBACK_STENCIL, state );
 	 }
-      } else {
-	 FALLBACK( rmesa, R200_FALLBACK_STENCIL, state );
       }
       break;
 
@@ -2205,7 +2058,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_TEXTURE_GEN_T:
       /* Picked up in r200UpdateTextureState.
        */
-      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE; 
+      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE;
       break;
 
    case GL_COLOR_SUM_EXT:
@@ -2322,7 +2175,7 @@ void r200LightingSpaceChange( GLcontext *ctx )
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLboolean tmp;
 
-   if (R200_DEBUG & DEBUG_STATE) 
+   if (R200_DEBUG & DEBUG_STATE)
       fprintf(stderr, "%s %d BEFORE %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0]);
 
@@ -2338,7 +2191,7 @@ void r200LightingSpaceChange( GLcontext *ctx )
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] &= ~R200_RESCALE_NORMALS;
    }
 
-   if (R200_DEBUG & DEBUG_STATE) 
+   if (R200_DEBUG & DEBUG_STATE)
       fprintf(stderr, "%s %d AFTER %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0]);
 }
@@ -2381,7 +2234,7 @@ static void update_texturematrix( GLcontext *ctx )
    GLuint compsel = rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL];
    int unit;
 
-   if (R200_DEBUG & DEBUG_STATE) 
+   if (R200_DEBUG & DEBUG_STATE)
       fprintf(stderr, "%s before COMPSEL: %x\n", __FUNCTION__,
 	      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL]);
 
@@ -2389,7 +2242,7 @@ static void update_texturematrix( GLcontext *ctx )
    rmesa->TexMatCompSel = 0;
 
    for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (!ctx->Texture.Unit[unit]._ReallyEnabled) 
+      if (!ctx->Texture.Unit[unit]._ReallyEnabled)
 	 continue;
 
       if (ctx->TextureMatrixStack[unit].Top->type != MATRIX_IDENTITY) {
@@ -2399,21 +2252,21 @@ static void update_texturematrix( GLcontext *ctx )
 	 rmesa->TexMatCompSel |= R200_OUTPUT_TEX_0 << unit;
 
 	 if (rmesa->TexGenEnabled & (R200_TEXMAT_0_ENABLE << unit)) {
-	    /* Need to preconcatenate any active texgen 
+	    /* Need to preconcatenate any active texgen
 	     * obj/eyeplane matrices:
 	     */
 	    _math_matrix_mul_matrix( &rmesa->tmpmat,
-				     ctx->TextureMatrixStack[unit].Top, 
+				     ctx->TextureMatrixStack[unit].Top,
 				     &rmesa->TexGenMatrix[unit] );
 	    upload_matrix( rmesa, rmesa->tmpmat.m, R200_MTX_TEX0+unit );
-	 } 
+	 }
 	 else {
-	    upload_matrix( rmesa, ctx->TextureMatrixStack[unit].Top->m, 
+	    upload_matrix( rmesa, ctx->TextureMatrixStack[unit].Top->m,
 			   R200_MTX_TEX0+unit );
 	 }
       }
       else if (rmesa->TexGenEnabled & (R200_TEXMAT_0_ENABLE << unit)) {
-	 upload_matrix( rmesa, rmesa->TexGenMatrix[unit].m, 
+	 upload_matrix( rmesa, rmesa->TexGenMatrix[unit].m,
 			R200_MTX_TEX0+unit );
       }
    }
@@ -2432,69 +2285,78 @@ static void update_texturematrix( GLcontext *ctx )
    }
 }
 
-
-
-/**
- * Tell the card where to render (offset, pitch).
- * Effected by glDrawBuffer, etc
- */
-void
-r200UpdateDrawBuffer(GLcontext *ctx)
+static GLboolean r200ValidateBuffers(GLcontext *ctx)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   driRenderbuffer *drb;
+   struct radeon_renderbuffer *rrb;
+   int i, ret;
 
-   if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-      /* draw to front */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
-   }
-   else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-      /* draw to back */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+   radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
-   else {
-      /* drawing to multiple buffers, or none */
-      return;
+
+   /* depth buffer */
+   rrb = radeon_get_depthbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
-   assert(drb);
-   assert(drb->flippedPitch);
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+      radeonTexObj *t;
 
-   R200_STATECHANGE( rmesa, ctx );
+      if (!ctx->Texture.Unit[i]._ReallyEnabled)
+	 continue;
 
-   /* Note: we used the (possibly) page-flipped values */
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET]
-     = ((drb->flippedOffset + rmesa->r200Screen->fbLocation)
-	& R200_COLOROFFSET_MASK);
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch;
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
+      t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
+      if (t->image_override && t->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+      else if (t->mt->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->mt->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
    }
-}
-
 
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret)
+       return GL_FALSE;
+   return GL_TRUE;
+}
 
-void r200ValidateState( GLcontext *ctx )
+GLboolean r200ValidateState( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint new_state = rmesa->NewGLState;
+   GLuint new_state = rmesa->radeon.NewGLState;
 
-   if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
-     r200UpdateDrawBuffer(ctx);
+   if (new_state & _NEW_BUFFERS) {
+      _mesa_update_framebuffer(ctx);
+      /* this updates the DrawBuffer's Width/Height if it's a FBO */
+      _mesa_update_draw_buffer_bounds(ctx);
+
+      R200_STATECHANGE(rmesa, ctx);
    }
 
-   if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM)) {
+   if (new_state & (_NEW_TEXTURE | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)) {
       r200UpdateTextureState( ctx );
-      new_state |= rmesa->NewGLState; /* may add TEXTURE_MATRIX */
+      new_state |= rmesa->radeon.NewGLState; /* may add TEXTURE_MATRIX */
       r200UpdateLocalViewer( ctx );
    }
 
+   /* we need to do a space check here */
+   if (!r200ValidateBuffers(ctx))
+     return GL_FALSE;
+
 /* FIXME: don't really need most of these when vertex progs are enabled */
 
    /* Need an event driven matrix update?
     */
-   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION)) 
+   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION))
       upload_matrix( rmesa, ctx->_ModelProjectMatrix.m, R200_MTX_MVP );
 
    /* Need these for lighting (shouldn't upload otherwise)
@@ -2518,11 +2380,12 @@ void r200ValidateState( GLcontext *ctx )
    /* emit all active clip planes if projection matrix changes.
     */
    if (new_state & (_NEW_PROJECTION)) {
-      if (ctx->Transform.ClipPlanesEnabled) 
+      if (ctx->Transform.ClipPlanesEnabled)
 	 r200UpdateClipPlanes( ctx );
    }
 
    if (new_state & (_NEW_PROGRAM|
+                    _NEW_PROGRAM_CONSTANTS |
    /* need to test for pretty much anything due to possible parameter bindings */
 	_NEW_MODELVIEW|_NEW_PROJECTION|_NEW_TRANSFORM|
 	_NEW_LIGHT|_NEW_TEXTURE|_NEW_TEXTURE_MATRIX|
@@ -2533,7 +2396,8 @@ void r200ValidateState( GLcontext *ctx )
       else TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, 0);
    }
 
-   rmesa->NewGLState = 0;
+   rmesa->radeon.NewGLState = 0;
+   return GL_TRUE;
 }
 
 
@@ -2544,7 +2408,7 @@ static void r200InvalidateState( GLcontext *ctx, GLuint new_state )
    _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
-   R200_CONTEXT(ctx)->NewGLState |= new_state;
+   R200_CONTEXT(ctx)->radeon.NewGLState |= new_state;
 }
 
 /* A hack.  The r200 can actually cope just fine with materials
@@ -2573,12 +2437,13 @@ static void r200WrapRunPipeline( GLcontext *ctx )
    GLboolean has_material;
 
    if (0)
-      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->NewGLState);
+      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->radeon.NewGLState);
 
    /* Validate state:
     */
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
+   if (rmesa->radeon.NewGLState)
+      if (!r200ValidateState( ctx ))
+	 FALLBACK(rmesa, RADEON_FALLBACK_TEXTURE, GL_TRUE);
 
    has_material = !ctx->VertexProgram._Enabled && ctx->Light.Enabled && check_material( ctx );
 
@@ -2587,7 +2452,7 @@ static void r200WrapRunPipeline( GLcontext *ctx )
    }
 
    /* Run the pipeline.
-    */ 
+    */
    _tnl_run_pipeline( ctx );
 
    if (has_material) {
@@ -2603,8 +2468,8 @@ void r200InitStateFuncs( struct dd_function_table *functions )
    functions->UpdateState		= r200InvalidateState;
    functions->LightingSpaceChange	= r200LightingSpaceChange;
 
-   functions->DrawBuffer		= r200DrawBuffer;
-   functions->ReadBuffer		= r200ReadBuffer;
+   functions->DrawBuffer		= radeonDrawBuffer;
+   functions->ReadBuffer		= radeonReadBuffer;
 
    functions->AlphaFunc			= r200AlphaFunc;
    functions->BlendColor		= r200BlendColor;
@@ -2636,7 +2501,7 @@ void r200InitStateFuncs( struct dd_function_table *functions )
    functions->PointParameterfv		= r200PointParameter;
    functions->PointSize			= r200PointSize;
    functions->RenderMode		= r200RenderMode;
-   functions->Scissor			= r200Scissor;
+   functions->Scissor			= radeonScissor;
    functions->ShadeModel		= r200ShadeModel;
    functions->StencilFuncSeparate	= r200StencilFuncSeparate;
    functions->StencilMaskSeparate	= r200StencilMaskSeparate;
diff --git a/src/mesa/drivers/dri/r200/r200_state.h b/src/mesa/drivers/dri/r200/r200_state.h
index a917163a00..23cf8aea66 100644
--- a/src/mesa/drivers/dri/r200/r200_state.h
+++ b/src/mesa/drivers/dri/r200/r200_state.h
@@ -43,23 +43,17 @@ extern void r200InitTnlFuncs( GLcontext *ctx );
 
 extern void r200UpdateMaterial( GLcontext *ctx );
 
-extern void r200SetCliprects( r200ContextPtr rmesa );
-extern void r200RecalcScissorRects( r200ContextPtr rmesa );
 extern void r200UpdateViewportOffset( GLcontext *ctx );
 extern void r200UpdateWindow( GLcontext *ctx );
 extern void r200UpdateDrawBuffer(GLcontext *ctx);
 
-extern void r200ValidateState( GLcontext *ctx );
-
-extern void r200PrintDirty( r200ContextPtr rmesa,
-			      const char *msg );
-
+extern GLboolean r200ValidateState( GLcontext *ctx );
 
 extern void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode );
 #define FALLBACK( rmesa, bit, mode ) do {				\
    if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n",		\
 		     __FUNCTION__, bit, mode );				\
-   r200Fallback( rmesa->glCtx, bit, mode );				\
+   r200Fallback( rmesa->radeon.glCtx, bit, mode );				\
 } while (0)
 
 extern void r200LightingSpaceChange( GLcontext *ctx );
diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index 9e4677eda4..bc871d9904 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -43,6 +43,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
 
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
@@ -52,31 +54,129 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "xmlpool.h"
 
+/* New (1.3) state mechanism.  3 commands (packet, scalar, vector) in
+ * 1.3 cmdbuffers allow all previous state to be updated as well as
+ * the tcl scalar and vector areas.
+ */
+static struct {
+	int start;
+	int len;
+	const char *name;
+} packet[RADEON_MAX_STATE_PACKETS] = {
+	{RADEON_PP_MISC, 7, "RADEON_PP_MISC"},
+	{RADEON_PP_CNTL, 3, "RADEON_PP_CNTL"},
+	{RADEON_RB3D_COLORPITCH, 1, "RADEON_RB3D_COLORPITCH"},
+	{RADEON_RE_LINE_PATTERN, 2, "RADEON_RE_LINE_PATTERN"},
+	{RADEON_SE_LINE_WIDTH, 1, "RADEON_SE_LINE_WIDTH"},
+	{RADEON_PP_LUM_MATRIX, 1, "RADEON_PP_LUM_MATRIX"},
+	{RADEON_PP_ROT_MATRIX_0, 2, "RADEON_PP_ROT_MATRIX_0"},
+	{RADEON_RB3D_STENCILREFMASK, 3, "RADEON_RB3D_STENCILREFMASK"},
+	{RADEON_SE_VPORT_XSCALE, 6, "RADEON_SE_VPORT_XSCALE"},
+	{RADEON_SE_CNTL, 2, "RADEON_SE_CNTL"},
+	{RADEON_SE_CNTL_STATUS, 1, "RADEON_SE_CNTL_STATUS"},
+	{RADEON_RE_MISC, 1, "RADEON_RE_MISC"},
+	{RADEON_PP_TXFILTER_0, 6, "RADEON_PP_TXFILTER_0"},
+	{RADEON_PP_BORDER_COLOR_0, 1, "RADEON_PP_BORDER_COLOR_0"},
+	{RADEON_PP_TXFILTER_1, 6, "RADEON_PP_TXFILTER_1"},
+	{RADEON_PP_BORDER_COLOR_1, 1, "RADEON_PP_BORDER_COLOR_1"},
+	{RADEON_PP_TXFILTER_2, 6, "RADEON_PP_TXFILTER_2"},
+	{RADEON_PP_BORDER_COLOR_2, 1, "RADEON_PP_BORDER_COLOR_2"},
+	{RADEON_SE_ZBIAS_FACTOR, 2, "RADEON_SE_ZBIAS_FACTOR"},
+	{RADEON_SE_TCL_OUTPUT_VTX_FMT, 11, "RADEON_SE_TCL_OUTPUT_VTX_FMT"},
+	{RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED, 17,
+		    "RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED"},
+	{R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0"},
+	{R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1"},
+	{R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2"},
+	{R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3"},
+	{R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4"},
+	{R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5"},
+	{R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6"},
+	{R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7"},
+	{R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0"},
+	{R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0"},
+	{R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0"},
+	{R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL"},
+	{R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0"},
+	{R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2"},
+	{R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL"},
+	{R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0"},
+	{R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1"},
+	{R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2"},
+	{R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3"},
+	{R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4"},
+	{R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5"},
+	{R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0"},
+	{R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1"},
+	{R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2"},
+	{R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3"},
+	{R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4"},
+	{R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5"},
+	{R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL"},
+	{R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1,
+	 "R200_SE_TCL_OUTPUT_VTX_COMP_SEL"},
+	{R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3"},
+	{R200_PP_CNTL_X, 1, "R200_PP_CNTL_X"},
+	{R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET"},
+	{R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL"},
+	{R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0"},
+	{R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1"},
+	{R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2"},
+	{R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS"},
+	{R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL"},
+	{R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE"},
+	{R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4,
+		    "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0"},
+	{R200_PP_CUBIC_FACES_0, 1, "R200_PP_CUBIC_FACES_0"},	/* 61 */
+	{R200_PP_CUBIC_OFFSET_F1_0, 5, "R200_PP_CUBIC_OFFSET_F1_0"}, /* 62 */
+	{R200_PP_CUBIC_FACES_1, 1, "R200_PP_CUBIC_FACES_1"},
+	{R200_PP_CUBIC_OFFSET_F1_1, 5, "R200_PP_CUBIC_OFFSET_F1_1"},
+	{R200_PP_CUBIC_FACES_2, 1, "R200_PP_CUBIC_FACES_2"},
+	{R200_PP_CUBIC_OFFSET_F1_2, 5, "R200_PP_CUBIC_OFFSET_F1_2"},
+	{R200_PP_CUBIC_FACES_3, 1, "R200_PP_CUBIC_FACES_3"},
+	{R200_PP_CUBIC_OFFSET_F1_3, 5, "R200_PP_CUBIC_OFFSET_F1_3"},
+	{R200_PP_CUBIC_FACES_4, 1, "R200_PP_CUBIC_FACES_4"},
+	{R200_PP_CUBIC_OFFSET_F1_4, 5, "R200_PP_CUBIC_OFFSET_F1_4"},
+	{R200_PP_CUBIC_FACES_5, 1, "R200_PP_CUBIC_FACES_5"},
+	{R200_PP_CUBIC_OFFSET_F1_5, 5, "R200_PP_CUBIC_OFFSET_F1_5"},
+	{RADEON_PP_TEX_SIZE_0, 2, "RADEON_PP_TEX_SIZE_0"},
+	{RADEON_PP_TEX_SIZE_1, 2, "RADEON_PP_TEX_SIZE_1"},
+	{RADEON_PP_TEX_SIZE_2, 2, "RADEON_PP_TEX_SIZE_2"},
+	{R200_RB3D_BLENDCOLOR, 3, "R200_RB3D_BLENDCOLOR"},
+	{R200_SE_TCL_POINT_SPRITE_CNTL, 1, "R200_SE_TCL_POINT_SPRITE_CNTL"},
+	{RADEON_PP_CUBIC_FACES_0, 1, "RADEON_PP_CUBIC_FACES_0"},
+	{RADEON_PP_CUBIC_OFFSET_T0_0, 5, "RADEON_PP_CUBIC_OFFSET_T0_0"},
+	{RADEON_PP_CUBIC_FACES_1, 1, "RADEON_PP_CUBIC_FACES_1"},
+	{RADEON_PP_CUBIC_OFFSET_T1_0, 5, "RADEON_PP_CUBIC_OFFSET_T1_0"},
+	{RADEON_PP_CUBIC_FACES_2, 1, "RADEON_PP_CUBIC_FACES_2"},
+	{RADEON_PP_CUBIC_OFFSET_T2_0, 5, "RADEON_PP_CUBIC_OFFSET_T2_0"},
+	{R200_PP_TRI_PERF, 2, "R200_PP_TRI_PERF"},
+	{R200_PP_TXCBLEND_8, 32, "R200_PP_AFS_0"},     /* 85 */
+	{R200_PP_TXCBLEND_0, 32, "R200_PP_AFS_1"},
+	{R200_PP_TFACTOR_0, 8, "R200_ATF_TFACTOR"},
+	{R200_PP_TXFILTER_0, 8, "R200_PP_TXCTLALL_0"},
+	{R200_PP_TXFILTER_1, 8, "R200_PP_TXCTLALL_1"},
+	{R200_PP_TXFILTER_2, 8, "R200_PP_TXCTLALL_2"},
+	{R200_PP_TXFILTER_3, 8, "R200_PP_TXCTLALL_3"},
+	{R200_PP_TXFILTER_4, 8, "R200_PP_TXCTLALL_4"},
+	{R200_PP_TXFILTER_5, 8, "R200_PP_TXCTLALL_5"},
+	{R200_VAP_PVS_CNTL_1, 2, "R200_VAP_PVS_CNTL"},
+};
+
 /* =============================================================
  * State initialization
  */
-
-void r200PrintDirty( r200ContextPtr rmesa, const char *msg )
+static int cmdpkt( r200ContextPtr rmesa, int id ) 
 {
-   struct r200_state_atom *l;
-
-   fprintf(stderr, msg);
-   fprintf(stderr, ": ");
+   drm_radeon_cmd_header_t h;
 
-   foreach(l, &rmesa->hw.atomlist) {
-      if (l->dirty || rmesa->hw.all_dirty)
-	 fprintf(stderr, "%s, ", l->name);
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     return CP_PACKET0(packet[id].start, packet[id].len - 1);
+   } else {
+     h.i = 0;
+     h.packet.cmd_type = RADEON_CMD_PACKET;
+     h.packet.packet_id = id;
    }
-
-   fprintf(stderr, "\n");
-}
-
-static int cmdpkt( int id ) 
-{
-   drm_radeon_cmd_header_t h;
-   h.i = 0;
-   h.packet.cmd_type = RADEON_CMD_PACKET;
-   h.packet.packet_id = id;
    return h.i;
 }
 
@@ -127,150 +227,502 @@ static int cmdscl2( int offset, int stride, int count )
 }
 
 #define CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom) \
 {							\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
    (void) rmesa;					\
-   return FLAG;						\
+   return (FLAG) ? atom->cmd_size : 0;			\
 }
 
 #define TCL_CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
-{							\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
-   return !rmesa->TclFallback && !ctx->VertexProgram._Enabled && (FLAG);	\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom) \
+{									\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);				\
+   return (!rmesa->radeon.TclFallback && !ctx->VertexProgram._Enabled && (FLAG)) ? atom->cmd_size : 0; \
 }
 
 #define TCL_OR_VP_CHECK( NM, FLAG )			\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \
 {							\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
-   return !rmesa->TclFallback && (FLAG);		\
+   return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size : 0;	\
 }
 
 #define VP_CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx, int idx )	\
-{							\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   (void) idx;						\
-   return !rmesa->TclFallback && ctx->VertexProgram._Enabled && (FLAG);		\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \
+{									\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);				\
+   (void) atom;								\
+   return (!rmesa->radeon.TclFallback && ctx->VertexProgram._Enabled && (FLAG)) ? atom->cmd_size : 0; \
 }
 
-
 CHECK( always, GL_TRUE )
 CHECK( never, GL_FALSE )
 CHECK( tex_any, ctx->Texture._EnabledUnits )
 CHECK( tf, (ctx->Texture._EnabledUnits && !ctx->ATIFragmentShader._Enabled) );
-CHECK( tex_pair, (rmesa->state.texture.unit[idx].unitneeded | rmesa->state.texture.unit[idx & ~1].unitneeded) )
-CHECK( tex, rmesa->state.texture.unit[idx].unitneeded )
+CHECK( tex_pair, (rmesa->state.texture.unit[atom->idx].unitneeded | rmesa->state.texture.unit[atom->idx & ~1].unitneeded) )
+CHECK( tex, rmesa->state.texture.unit[atom->idx].unitneeded )
 CHECK( pix_zero, !ctx->ATIFragmentShader._Enabled )
-CHECK( texenv, (rmesa->state.envneeded & (1 << idx) && !ctx->ATIFragmentShader._Enabled) )
+   CHECK( texenv, (rmesa->state.envneeded & (1 << (atom->idx)) && !ctx->ATIFragmentShader._Enabled) )
 CHECK( afs_pass1, (ctx->ATIFragmentShader._Enabled && (ctx->ATIFragmentShader.Current->NumPasses > 1)) )
 CHECK( afs, ctx->ATIFragmentShader._Enabled )
-CHECK( tex_cube, rmesa->state.texture.unit[idx].unitneeded & TEXTURE_CUBE_BIT )
+CHECK( tex_cube, rmesa->state.texture.unit[atom->idx].unitneeded & TEXTURE_CUBE_BIT )
 TCL_CHECK( tcl_fog, ctx->Fog.Enabled )
 TCL_CHECK( tcl, GL_TRUE )
-TCL_CHECK( tcl_tex, rmesa->state.texture.unit[idx].unitneeded )
+TCL_CHECK( tcl_tex, rmesa->state.texture.unit[atom->idx].unitneeded )
 TCL_CHECK( tcl_lighting, ctx->Light.Enabled )
-TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[idx].Enabled )
-TCL_OR_VP_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << idx)) )
+TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[atom->idx].Enabled )
+TCL_OR_VP_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << (atom->idx))) )
 TCL_OR_VP_CHECK( tcl_or_vp, GL_TRUE )
 VP_CHECK( tcl_vp, GL_TRUE )
 VP_CHECK( tcl_vp_size, ctx->VertexProgram.Current->Base.NumNativeInstructions > 64 )
 VP_CHECK( tcl_vpp_size, ctx->VertexProgram.Current->Base.NumNativeParameters > 96 )
 
+#define OUT_VEC(hdr, data) do {			\
+    drm_radeon_cmd_header_t h;					\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0));		\
+    OUT_BATCH(0);							\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0));		\
+    OUT_BATCH(h.vectors.offset | (h.vectors.stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, h.vectors.count - 1));	\
+    OUT_BATCH_TABLE((data), h.vectors.count);				\
+  } while(0)
+
+#define OUT_VECLINEAR(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    uint32_t _start, _sz;						\
+    h.i = hdr;								\
+    _start = h.veclinear.addr_lo | (h.veclinear.addr_hi << 8);		\
+    _sz = h.veclinear.count * 4;					\
+    if (r200->radeon.radeonScreen->kernel_mm && _sz) { \
+    BEGIN_BATCH_NO_AUTOSTATE(dwords); \
+    OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0));		\
+    OUT_BATCH(0);							\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0));		\
+    OUT_BATCH(_start | (1 << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT));	\
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, _sz - 1));	\
+    OUT_BATCH_TABLE((data), _sz);					\
+    END_BATCH(); \
+    } \
+  } while(0)
+
+#define OUT_SCL(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0));		\
+    OUT_BATCH((h.scalars.offset) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1));	\
+    OUT_BATCH_TABLE((data), h.scalars.count);				\
+  } while(0)
+
+#define OUT_SCL2(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0));		\
+    OUT_BATCH((h.scalars.offset + 0x100) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1));	\
+    OUT_BATCH_TABLE((data), h.scalars.count);				\
+  } while(0)
+
+static void mtl_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 6;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[MTL_CMD_0], (atom->cmd+1));
+   OUT_SCL2(atom->cmd[MTL_CMD_1], (atom->cmd + 18));
+   END_BATCH();
+}
 
-/* Initialize the context's hardware state.
- */
-void r200InitState( r200ContextPtr rmesa )
+static void lit_emit(GLcontext *ctx, struct radeon_state_atom *atom)
 {
-   GLcontext *ctx = rmesa->glCtx;
-   GLuint color_fmt, depth_fmt, i;
-   GLint drawPitch, drawOffset;
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 8;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[LIT_CMD_0], atom->cmd+1);
+   OUT_VEC(atom->cmd[LIT_CMD_1], atom->cmd+LIT_CMD_1+1);
+   END_BATCH();
+}
 
-   switch ( rmesa->r200Screen->cpp ) {
-   case 2:
-      color_fmt = R200_COLOR_FORMAT_RGB565;
-      break;
-   case 4:
-      color_fmt = R200_COLOR_FORMAT_ARGB8888;
-      break;
-   default:
-      fprintf( stderr, "Error: Unsupported pixel depth... exiting\n" );
-      exit( -1 );
+static void ptp_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 8;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[PTP_CMD_0], atom->cmd+1);
+   OUT_VEC(atom->cmd[PTP_CMD_1], atom->cmd+PTP_CMD_1+1);
+   END_BATCH();
+}
+
+static void veclinear_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 4;
+   OUT_VECLINEAR(atom->cmd[0], atom->cmd+1);
+}
+
+static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_SCL(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
+
+
+static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 4;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
+
+static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   struct radeon_renderbuffer *rrb;
+   uint32_t cbpitch;
+   uint32_t zbpitch, depth_fmt;
+   uint32_t dwords = atom->cmd_size;
+
+   /* output the first 7 bytes of context */
+   BEGIN_BATCH_NO_AUTOSTATE(dwords+2+2);
+   OUT_BATCH_TABLE(atom->cmd, 5);
+
+   rrb = radeon_get_depthbuffer(&r200->radeon);
+   if (!rrb) {
+     OUT_BATCH(0);
+     OUT_BATCH(0);
+   } else {
+     zbpitch = (rrb->pitch / rrb->cpp);
+     if (r200->using_hyperz)
+       zbpitch |= RADEON_DEPTH_HYPERZ;
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+     OUT_BATCH(zbpitch);
+     if (rrb->cpp == 4) 
+       depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z; 
+     else 
+       depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z; 
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK; 
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt; 
+   }
+     
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(atom->cmd[CTX_CMD_1]);
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+
+   rrb = radeon_get_colorbuffer(&r200->radeon);
+   if (!rrb || !rrb->bo) {
+     OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+     OUT_BATCH(atom->cmd[CTX_RB3D_COLOROFFSET]);
+   } else {
+     atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10); 
+     if (rrb->cpp == 4) 
+       atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888; 
+     else 
+       atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565; 
+ 
+     OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); 
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
    }
 
-   rmesa->state.color.clear = 0x00000000;
+   OUT_BATCH(atom->cmd[CTX_CMD_2]);
 
-   switch ( ctx->Visual.depthBits ) {
-   case 16:
-      rmesa->state.depth.clear = 0x0000ffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff;
-      depth_fmt = R200_DEPTH_FORMAT_16BIT_INT_Z;
-      rmesa->state.stencil.clear = 0x00000000;
-      break;
-   case 24:
-      rmesa->state.depth.clear = 0x00ffffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff;
-      depth_fmt = R200_DEPTH_FORMAT_24BIT_INT_Z;
-      rmesa->state.stencil.clear = 0xffff0000;
-      break;
-   default:
-      fprintf( stderr, "Error: Unsupported depth %d... exiting\n",
-	       ctx->Visual.depthBits );
-      exit( -1 );
+   if (!rrb || !rrb->bo) {
+     OUT_BATCH(atom->cmd[CTX_RB3D_COLORPITCH]);
+   } else {
+     cbpitch = (rrb->pitch / rrb->cpp);
+     if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= R200_COLOR_TILE_ENABLE;
+     OUT_BATCH(cbpitch);
    }
 
-   /* Only have hw stencil when depth buffer is 24 bits deep */
-   rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
-				     ctx->Visual.depthBits == 24 );
+   if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM)
+     OUT_BATCH_TABLE((atom->cmd + 14), 4);
+
+   END_BATCH();
+}
 
-   rmesa->Fallback = 0;
+static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   struct radeon_renderbuffer *rrb, *drb;
+   uint32_t cbpitch = 0;
+   uint32_t zbpitch = 0;
+   uint32_t dwords = atom->cmd_size;
+   uint32_t depth_fmt;
+
+   rrb = radeon_get_colorbuffer(&r200->radeon);
+   if (!rrb || !rrb->bo) {
+      return;
+   }
 
-   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
-      drawOffset = rmesa->r200Screen->backOffset;
-      drawPitch  = rmesa->r200Screen->backPitch;
-   } else {
-      drawOffset = rmesa->r200Screen->frontOffset;
-      drawPitch  = rmesa->r200Screen->frontPitch;
+   atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
+   if (rrb->cpp == 4)
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
+   else switch (rrb->base._ActualFormat) {
+   case GL_RGB5:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+	break;
+   case GL_RGBA4:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
+	break;
+   case GL_RGB5_A1:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
+	break;
+   }
+
+   cbpitch = (rrb->pitch / rrb->cpp);
+   if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= R200_COLOR_TILE_ENABLE;
+
+   drb = radeon_get_depthbuffer(&r200->radeon);
+   if (drb) {
+     zbpitch = (drb->pitch / drb->cpp);
+     if (drb->cpp == 4)
+        depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
+     else
+        depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
+   }
+
+   dwords = 10;
+   if (drb)
+     dwords += 6;
+   if (rrb)
+     dwords += 6;
+   if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM)
+     dwords += 4;
+
+   /* output the first 7 bytes of context */
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   /* In the CS case we need to split this up */
+   OUT_BATCH(CP_PACKET0(packet[0].start, 3));
+   OUT_BATCH_TABLE((atom->cmd + 1), 4);
+
+   if (drb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0));
+     OUT_BATCH_RELOC(0, drb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0));
+     OUT_BATCH(zbpitch);
    }
-#if 000
-   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
-      rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
-      rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+
+   OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0));
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1));
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+   OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+
+
+   if (rrb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0));
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0));
+     OUT_BATCH(cbpitch);
+   }
+
+   if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) {
+     OUT_BATCH_TABLE((atom->cmd + 14), 4);
+   }
+
+   END_BATCH();
+}
+
+static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+   int i = atom->idx;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   if (t && t->mt && !t->image_override)
+     dwords += 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   /* is this ok even with drm older than 1.18? */
+   OUT_BATCH_TABLE(atom->cmd, 10);
+
+   if (t && t->mt && !t->image_override) {
+     OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		  RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+   } else if (!t) {
+     /* workaround for old CS mechanism */
+     OUT_BATCH(r200->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]);
    } else {
-      rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
-      rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+     OUT_BATCH(t->override_offset);
    }
 
-   rmesa->state.pixel.readOffset = rmesa->state.color.drawOffset;
-   rmesa->state.pixel.readPitch  = rmesa->state.color.drawPitch;
-#endif
+   END_BATCH();
+}
+
+static void tex_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = atom->cmd_size;
+   int i = atom->idx;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+   int hastexture = 1;
+
+   if (!r200->state.texture.unit[i].unitneeded)
+        hastexture = 0;
+   if (!t)
+	hastexture = 0;
+   else {
+	if (!t->mt && !t->bo)
+		hastexture = 0;
+   }
+
+   if (hastexture)
+     dwords += 2;
+   else
+     dwords -= 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   OUT_BATCH(CP_PACKET0(R200_PP_TXFILTER_0 + (32 * i), 7));
+   OUT_BATCH_TABLE((atom->cmd + 1), 8);
+
+   if (hastexture) {
+     OUT_BATCH(CP_PACKET0(R200_PP_TXOFFSET_0 + (24 * i), 0));
+     if (t->mt && !t->image_override) {
+        OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		  RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+      } else {
+	if (t->bo)
+            OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
+                            RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+      }
+   }
+   END_BATCH();
+}
+
+
+static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = 3;
+   int i = atom->idx, j;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords + (3 * 5));
+   /* XXX that size won't really match with image_override... */
+   OUT_BATCH_TABLE(atom->cmd, 2);
+
+   if (t && !t->image_override) {
+     lvl = &t->mt->levels[0];
+     OUT_BATCH_TABLE((atom->cmd + 2), 1);
+     for (j = 1; j <= 5; j++) {
+       OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     }
+   }
+   END_BATCH();
+}
+
+static void cube_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r200ContextPtr r200 = R200_CONTEXT(ctx);
+   BATCH_LOCALS(&r200->radeon);
+   uint32_t dwords = 2;
+   int i = atom->idx, j;
+   radeonTexObj *t = r200->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords + (4 * 5));
+   OUT_BATCH_TABLE(atom->cmd, 2);
+
+   if (t && !t->image_override) {
+     lvl = &t->mt->levels[0];
+     for (j = 1; j <= 5; j++) {
+       OUT_BATCH(CP_PACKET0(R200_PP_CUBIC_OFFSET_F1_0 + (24*i) + (4 * (j-1)), 0));
+       OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     }
+   }
+   END_BATCH();
+}
+
+/* Initialize the context's hardware state.
+ */
+void r200InitState( r200ContextPtr rmesa )
+{
+   GLcontext *ctx = rmesa->radeon.glCtx;
+   GLuint i;
+
+   rmesa->radeon.state.color.clear = 0x00000000;
+
+   switch ( ctx->Visual.depthBits ) {
+   case 16:
+      rmesa->radeon.state.depth.clear = 0x0000ffff;
+      rmesa->radeon.state.stencil.clear = 0x00000000;
+      break;
+   case 24:
+   default:
+      rmesa->radeon.state.depth.clear = 0x00ffffff;
+      rmesa->radeon.state.stencil.clear = 0xffff0000;
+      break;
+   }
+
+   rmesa->radeon.Fallback = 0;
 
-   rmesa->hw.max_state_size = 0;
+   rmesa->radeon.hw.max_state_size = 0;
 
 #define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX )				\
    do {								\
       rmesa->hw.ATOM.cmd_size = SZ;				\
-      rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int));	\
-      rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.cmd = (GLuint *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.lastcmd = (GLuint *)CALLOC(SZ * sizeof(int));	\
       rmesa->hw.ATOM.name = NM;					\
       rmesa->hw.ATOM.idx = IDX;					\
       rmesa->hw.ATOM.check = check_##CHK;			\
       rmesa->hw.ATOM.dirty = GL_FALSE;				\
-      rmesa->hw.max_state_size += SZ * sizeof(int);		\
+      rmesa->radeon.hw.max_state_size += SZ * sizeof(int);		\
    } while (0)
 
 
    /* Allocate state buffers:
     */
-   if (rmesa->r200Screen->drmSupportsBlendColor)
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
       ALLOC_STATE( ctx, always, CTX_STATE_SIZE_NEWDRM, "CTX/context", 0 );
    else
       ALLOC_STATE( ctx, always, CTX_STATE_SIZE_OLDDRM, "CTX/context", 0 );
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     rmesa->hw.ctx.emit = ctx_emit_cs;
+   else
+     rmesa->hw.ctx.emit = ctx_emit;
    ALLOC_STATE( set, always, SET_STATE_SIZE, "SET/setup", 0 );
    ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 );
    ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 );
@@ -282,8 +734,8 @@ void r200InitState( r200ContextPtr rmesa )
    ALLOC_STATE( cst, always, CST_STATE_SIZE, "CST/constant", 0 );
    ALLOC_STATE( zbs, always, ZBS_STATE_SIZE, "ZBS/zbias", 0 );
    ALLOC_STATE( tf, tf, TF_STATE_SIZE, "TF/tfactor", 0 );
-   if (rmesa->r200Screen->drmSupportsFragShader) {
-      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) {
+   if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
+      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) {
       /* make sure texture units 0/1 are emitted pair-wise for r200 t0 hang workaround */
 	 ALLOC_STATE( tex[0], tex_pair, TEX_STATE_SIZE_NEWDRM, "TEX/tex-0", 0 );
 	 ALLOC_STATE( tex[1], tex_pair, TEX_STATE_SIZE_NEWDRM, "TEX/tex-1", 1 );
@@ -303,7 +755,7 @@ void r200InitState( r200ContextPtr rmesa )
       ALLOC_STATE( afs[1], afs, AFS_STATE_SIZE, "AFS/afsinst-1", 1 );
    }
    else {
-      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) {
+      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) {
 	 ALLOC_STATE( tex[0], tex_pair, TEX_STATE_SIZE_OLDDRM, "TEX/tex-0", 0 );
 	 ALLOC_STATE( tex[1], tex_pair, TEX_STATE_SIZE_OLDDRM, "TEX/tex-1", 1 );
 	 ALLOC_STATE( tam, tex_any, TAM_STATE_SIZE, "TAM/tam", 0 );
@@ -321,13 +773,24 @@ void r200InitState( r200ContextPtr rmesa )
       ALLOC_STATE( afs[0], never, AFS_STATE_SIZE, "AFS/afsinst-0", 0 );
       ALLOC_STATE( afs[1], never, AFS_STATE_SIZE, "AFS/afsinst-1", 1 );
    }
-   if (rmesa->r200Screen->drmSupportsCubeMapsR200) {
+
+   for (i = 0; i < 5; i++)
+      if (rmesa->radeon.radeonScreen->kernel_mm)
+          rmesa->hw.tex[i].emit = tex_emit_cs;
+      else
+          rmesa->hw.tex[i].emit = tex_emit;
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR200) {
       ALLOC_STATE( cube[0], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-0", 0 );
       ALLOC_STATE( cube[1], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-1", 1 );
       ALLOC_STATE( cube[2], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-2", 2 );
       ALLOC_STATE( cube[3], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-3", 3 );
       ALLOC_STATE( cube[4], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-4", 4 );
       ALLOC_STATE( cube[5], tex_cube, CUBE_STATE_SIZE, "CUBE/tex-5", 5 );
+      for (i = 0; i < 5; i++)
+          if (rmesa->radeon.radeonScreen->kernel_mm)
+              rmesa->hw.cube[i].emit = cube_emit_cs;
+          else
+              rmesa->hw.cube[i].emit = cube_emit;
    }
    else {
       ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/tex-0", 0 );
@@ -337,7 +800,8 @@ void r200InitState( r200ContextPtr rmesa )
       ALLOC_STATE( cube[4], never, CUBE_STATE_SIZE, "CUBE/tex-4", 4 );
       ALLOC_STATE( cube[5], never, CUBE_STATE_SIZE, "CUBE/tex-5", 5 );
    }
-   if (rmesa->r200Screen->drmSupportsVertexProgram) {
+
+   if (rmesa->radeon.radeonScreen->drmSupportsVertexProgram) {
       ALLOC_STATE( pvs, tcl_vp, PVS_STATE_SIZE, "PVS/pvscntl", 0 );
       ALLOC_STATE( vpi[0], tcl_vp, VPI_STATE_SIZE, "VP/vertexprog-0", 0 );
       ALLOC_STATE( vpi[1], tcl_vp_size, VPI_STATE_SIZE, "VP/vertexprog-1", 1 );
@@ -390,13 +854,13 @@ void r200InitState( r200ContextPtr rmesa )
    ALLOC_STATE( pix[3], texenv, PIX_STATE_SIZE, "PIX/pixstage-3", 3 );
    ALLOC_STATE( pix[4], texenv, PIX_STATE_SIZE, "PIX/pixstage-4", 4 );
    ALLOC_STATE( pix[5], texenv, PIX_STATE_SIZE, "PIX/pixstage-5", 5 );
-   if (rmesa->r200Screen->drmSupportsTriPerf) {
+   if (rmesa->radeon.radeonScreen->drmSupportsTriPerf) {
       ALLOC_STATE( prf, always, PRF_STATE_SIZE, "PRF/performance-tri", 0 );
    }
    else {
       ALLOC_STATE( prf, never, PRF_STATE_SIZE, "PRF/performance-tri", 0 );
    }
-   if (rmesa->r200Screen->drmSupportsPointSprites) {
+   if (rmesa->radeon.radeonScreen->drmSupportsPointSprites) {
       ALLOC_STATE( spr, always, SPR_STATE_SIZE, "SPR/pointsprite", 0 );
       ALLOC_STATE( ptp, tcl, PTP_STATE_SIZE, "PTP/pointparams", 0 );
    }
@@ -409,87 +873,115 @@ void r200InitState( r200ContextPtr rmesa )
 
    /* Fill in the packet headers:
     */
-   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC);
-   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL);
-   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH);
-   if (rmesa->r200Screen->drmSupportsBlendColor)
-      rmesa->hw.ctx.cmd[CTX_CMD_3] = cmdpkt(R200_EMIT_RB3D_BLENDCOLOR);
-   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN);
-   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH);
-   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK);
-   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE);
-   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL);
-   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC);
-   rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(R200_EMIT_PP_CNTL_X);
-   rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(R200_EMIT_RB3D_DEPTHXY_OFFSET);
-   rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(R200_EMIT_RE_AUX_SCISSOR_CNTL);
-   rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(R200_EMIT_RE_SCISSOR_TL_0);
-   rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(R200_EMIT_SE_VAP_CNTL_STATUS);
-   rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(R200_EMIT_RE_POINTSIZE);
-   rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0);
-   rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(R200_EMIT_PP_TAM_DEBUG3);
-   rmesa->hw.tf.cmd[TF_CMD_0] = cmdpkt(R200_EMIT_TFACTOR_0);
-   if (rmesa->r200Screen->drmSupportsFragShader) {
-      rmesa->hw.atf.cmd[ATF_CMD_0] = cmdpkt(R200_EMIT_ATF_TFACTOR);
-      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_0);
-      rmesa->hw.tex[0].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_0);
-      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_1);
-      rmesa->hw.tex[1].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_1);
-      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_2);
-      rmesa->hw.tex[2].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_2);
-      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_3);
-      rmesa->hw.tex[3].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_3);
-      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_4);
-      rmesa->hw.tex[4].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_4);
-      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCTLALL_5);
-      rmesa->hw.tex[5].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_5);
+   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_MISC);
+   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CNTL);
+   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(rmesa, RADEON_EMIT_RB3D_COLORPITCH);
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
+      rmesa->hw.ctx.cmd[CTX_CMD_3] = cmdpkt(rmesa, R200_EMIT_RB3D_BLENDCOLOR);
+   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_LINE_PATTERN);
+   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_LINE_WIDTH);
+   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RB3D_STENCILREFMASK);
+   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_VPORT_XSCALE);
+   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL);
+   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_MISC);
+   rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CNTL_X);
+   rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(rmesa, R200_EMIT_RB3D_DEPTHXY_OFFSET);
+   rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(rmesa, R200_EMIT_RE_AUX_SCISSOR_CNTL);
+   rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(rmesa, R200_EMIT_RE_SCISSOR_TL_0);
+   rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(rmesa, R200_EMIT_SE_VAP_CNTL_STATUS);
+   rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(rmesa, R200_EMIT_RE_POINTSIZE);
+   rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(rmesa, R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0);
+   rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TAM_DEBUG3);
+   rmesa->hw.tf.cmd[TF_CMD_0] = cmdpkt(rmesa, R200_EMIT_TFACTOR_0);
+   if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
+      rmesa->hw.atf.cmd[ATF_CMD_0] = cmdpkt(rmesa, R200_EMIT_ATF_TFACTOR);
+      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_0);
+      rmesa->hw.tex[0].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_0);
+      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_1);
+      rmesa->hw.tex[1].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_1);
+      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_2);
+      rmesa->hw.tex[2].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_2);
+      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_3);
+      rmesa->hw.tex[3].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_3);
+      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_4);
+      rmesa->hw.tex[4].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_4);
+      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCTLALL_5);
+      rmesa->hw.tex[5].cmd[TEX_CMD_1_NEWDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_5);
    } else {
-      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_0);
-      rmesa->hw.tex[0].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_0);
-      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_1);
-      rmesa->hw.tex[1].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_1);
-      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_2);
-      rmesa->hw.tex[2].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_2);
-      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_3);
-      rmesa->hw.tex[3].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_3);
-      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_4);
-      rmesa->hw.tex[4].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_4);
-      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_5);
-      rmesa->hw.tex[5].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(R200_EMIT_PP_TXOFFSET_5);
-   }
-   rmesa->hw.afs[0].cmd[AFS_CMD_0] = cmdpkt(R200_EMIT_PP_AFS_0);
-   rmesa->hw.afs[1].cmd[AFS_CMD_0] = cmdpkt(R200_EMIT_PP_AFS_1);
-   rmesa->hw.pvs.cmd[PVS_CMD_0] = cmdpkt(R200_EMIT_VAP_PVS_CNTL);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_0);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_0);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_1);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_1);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_2);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_2);
-   rmesa->hw.cube[3].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_3);
-   rmesa->hw.cube[3].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_3);
-   rmesa->hw.cube[4].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_4);
-   rmesa->hw.cube[4].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_4);
-   rmesa->hw.cube[5].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_5);
-   rmesa->hw.cube[5].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_5);
-   rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_0);
-   rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_1);
-   rmesa->hw.pix[2].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_2);
-   rmesa->hw.pix[3].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_3);
-   rmesa->hw.pix[4].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_4);
-   rmesa->hw.pix[5].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_5);
-   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR);
-   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(R200_EMIT_TCL_LIGHT_MODEL_CTL_0);
-   rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(R200_EMIT_TCL_UCP_VERT_BLEND_CTL);
-   rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(R200_EMIT_TEX_PROC_CTL_2);
-   rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(R200_EMIT_MATRIX_SELECT_0);
-   rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(R200_EMIT_VAP_CTL);
-   rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(R200_EMIT_VTX_FMT_0);
-   rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(R200_EMIT_OUTPUT_VTX_COMP_SEL);
-   rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(R200_EMIT_SE_VTX_STATE_CNTL);
-   rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(R200_EMIT_VTE_CNTL);
-   rmesa->hw.prf.cmd[PRF_CMD_0] = cmdpkt(R200_EMIT_PP_TRI_PERF_CNTL);
-   rmesa->hw.spr.cmd[SPR_CMD_0] = cmdpkt(R200_EMIT_TCL_POINT_SPRITE_CNTL);
+      rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_0);
+      rmesa->hw.tex[0].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_0);
+      rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_1);
+      rmesa->hw.tex[1].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_1);
+      rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_2);
+      rmesa->hw.tex[2].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_2);
+      rmesa->hw.tex[3].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_3);
+      rmesa->hw.tex[3].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_3);
+      rmesa->hw.tex[4].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_4);
+      rmesa->hw.tex[4].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_4);
+      rmesa->hw.tex[5].cmd[TEX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXFILTER_5);
+      rmesa->hw.tex[5].cmd[TEX_CMD_1_OLDDRM] = cmdpkt(rmesa, R200_EMIT_PP_TXOFFSET_5);
+   }
+   rmesa->hw.afs[0].cmd[AFS_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_AFS_0);
+   rmesa->hw.afs[1].cmd[AFS_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_AFS_1);
+   rmesa->hw.pvs.cmd[PVS_CMD_0] = cmdpkt(rmesa, R200_EMIT_VAP_PVS_CNTL);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_0);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_0);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_1);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_1);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_2);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_2);
+   rmesa->hw.cube[3].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_3);
+   rmesa->hw.cube[3].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_3);
+   rmesa->hw.cube[4].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_4);
+   rmesa->hw.cube[4].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_4);
+   rmesa->hw.cube[5].cmd[CUBE_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_FACES_5);
+   rmesa->hw.cube[5].cmd[CUBE_CMD_1] = cmdpkt(rmesa, R200_EMIT_PP_CUBIC_OFFSETS_5);
+   rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_0);
+   rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_1);
+   rmesa->hw.pix[2].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_2);
+   rmesa->hw.pix[3].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_3);
+   rmesa->hw.pix[4].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_4);
+   rmesa->hw.pix[5].cmd[PIX_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TXCBLEND_5);
+   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_ZBIAS_FACTOR);
+   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(rmesa, R200_EMIT_TCL_LIGHT_MODEL_CTL_0);
+   rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(rmesa, R200_EMIT_TCL_UCP_VERT_BLEND_CTL);
+   rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(rmesa, R200_EMIT_TEX_PROC_CTL_2);
+   rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(rmesa, R200_EMIT_MATRIX_SELECT_0);
+   rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(rmesa, R200_EMIT_VAP_CTL);
+   rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(rmesa, R200_EMIT_VTX_FMT_0);
+   rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(rmesa, R200_EMIT_OUTPUT_VTX_COMP_SEL);
+   rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(rmesa, R200_EMIT_SE_VTX_STATE_CNTL);
+   rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(rmesa, R200_EMIT_VTE_CNTL);
+   rmesa->hw.prf.cmd[PRF_CMD_0] = cmdpkt(rmesa, R200_EMIT_PP_TRI_PERF_CNTL);
+   rmesa->hw.spr.cmd[SPR_CMD_0] = cmdpkt(rmesa, R200_EMIT_TCL_POINT_SPRITE_CNTL);
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+	rmesa->hw.mtl[0].emit = mtl_emit;
+	rmesa->hw.mtl[1].emit = mtl_emit;
+
+	rmesa->hw.vpi[0].emit = veclinear_emit;
+	rmesa->hw.vpi[1].emit = veclinear_emit;
+	rmesa->hw.vpp[0].emit = veclinear_emit;
+	rmesa->hw.vpp[1].emit = veclinear_emit;
+
+	rmesa->hw.grd.emit = scl_emit;
+	rmesa->hw.fog.emit = vec_emit;
+	rmesa->hw.glt.emit = vec_emit;
+	rmesa->hw.eye.emit = vec_emit;
+
+	for (i = R200_MTX_MV; i <= R200_MTX_TEX5; i++)
+	  rmesa->hw.mat[i].emit = vec_emit;
+
+	for (i = 0; i < 8; i++)
+	  rmesa->hw.lit[i].emit = lit_emit;
+
+	for (i = 0; i < 6; i++)
+	  rmesa->hw.ucp[i].emit = vec_emit;
+
+	rmesa->hw.ptp.emit = ptp_emit;
+   }
+
+
+   
    rmesa->hw.mtl[0].cmd[MTL_CMD_0] = 
       cmdvec( R200_VS_MAT_0_EMISS, 1, 16 );
    rmesa->hw.mtl[0].cmd[MTL_CMD_1] = 
@@ -567,7 +1059,7 @@ void r200InitState( r200ContextPtr rmesa )
 				(R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
 				(R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT));
 
-   if (rmesa->r200Screen->drmSupportsBlendColor) {
+   if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
       rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = 0x00000000;
       rmesa->hw.ctx.cmd[CTX_RB3D_ABLENDCNTL] = (R200_COMB_FCN_ADD_CLAMP |
 				(R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
@@ -578,18 +1070,17 @@ void r200InitState( r200ContextPtr rmesa )
    }
 
    rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] =
-      rmesa->r200Screen->depthOffset + rmesa->r200Screen->fbLocation;
+      rmesa->radeon.radeonScreen->depthOffset + rmesa->radeon.radeonScreen->fbLocation;
 
    rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = 
-      ((rmesa->r200Screen->depthPitch &
+      ((rmesa->radeon.radeonScreen->depthPitch &
 	R200_DEPTHPITCH_MASK) |
        R200_DEPTH_ENDIAN_NO_SWAP);
    
    if (rmesa->using_hyperz)
       rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] |= R200_DEPTH_HYPERZ;
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (depth_fmt |
-					       R200_Z_TEST_LESS |
+   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (R200_Z_TEST_LESS |
 					       R200_STENCIL_TEST_ALWAYS |
 					       R200_STENCIL_FAIL_KEEP |
 					       R200_STENCIL_ZPASS_KEEP |
@@ -599,15 +1090,14 @@ void r200InitState( r200ContextPtr rmesa )
    if (rmesa->using_hyperz) {
       rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= R200_Z_COMPRESSION_ENABLE |
 						  R200_Z_DECOMPRESSION_ENABLE;
-/*      if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200)
+/*      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200)
 	 rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_HIERARCHY_ENABLE;*/
    }
 
    rmesa->hw.ctx.cmd[CTX_PP_CNTL] = (R200_ANTI_ALIAS_NONE 
  				     | R200_TEX_BLEND_0_ENABLE);
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = color_fmt;
-   switch ( driQueryOptioni( &rmesa->optionCache, "dither_mode" ) ) {
+   switch ( driQueryOptioni( &rmesa->radeon.optionCache, "dither_mode" ) ) {
    case DRI_CONF_DITHER_XERRORDIFFRESET:
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_INIT;
       break;
@@ -615,41 +1105,19 @@ void r200InitState( r200ContextPtr rmesa )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_SCALE_DITHER_ENABLE;
       break;
    }
-   if ( driQueryOptioni( &rmesa->optionCache, "round_mode" ) ==
+   if ( driQueryOptioni( &rmesa->radeon.optionCache, "round_mode" ) ==
 	DRI_CONF_ROUND_ROUND )
-      rmesa->state.color.roundEnable = R200_ROUND_ENABLE;
+      rmesa->radeon.state.color.roundEnable = R200_ROUND_ENABLE;
    else
-      rmesa->state.color.roundEnable = 0;
-   if ( driQueryOptioni (&rmesa->optionCache, "color_reduction" ) ==
+      rmesa->radeon.state.color.roundEnable = 0;
+   if ( driQueryOptioni (&rmesa->radeon.optionCache, "color_reduction" ) ==
 	DRI_CONF_COLOR_REDUCTION_DITHER )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= R200_DITHER_ENABLE;
    else
-      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable;
-
-#if 000
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((rmesa->state.color.drawOffset +
-					       rmesa->r200Screen->fbLocation)
-					      & R200_COLOROFFSET_MASK);
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((rmesa->state.color.drawPitch &
-					      R200_COLORPITCH_MASK) |
-					     R200_COLOR_ENDIAN_NO_SWAP);
-#else
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((drawOffset +
-					       rmesa->r200Screen->fbLocation)
-					      & R200_COLOROFFSET_MASK);
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((drawPitch &
-					      R200_COLORPITCH_MASK) |
-					     R200_COLOR_ENDIAN_NO_SWAP);
-#endif
-   /* (fixed size) sarea is initialized to zero afaics so can omit version check. Phew! */
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= R200_COLOR_TILE_ENABLE;
-   }
+      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable;
 
    rmesa->hw.prf.cmd[PRF_PP_TRI_PERF] = R200_TRI_CUTOFF_MASK - R200_TRI_CUTOFF_MASK * 
-			driQueryOptionf (&rmesa->optionCache,"texture_blend_quality");
+			driQueryOptionf (&rmesa->radeon.optionCache,"texture_blend_quality");
    rmesa->hw.prf.cmd[PRF_PP_PERF_CNTL] = 0;
 
    rmesa->hw.set.cmd[SET_SE_CNTL] = (R200_FFACE_CULL_CCW |
@@ -704,7 +1172,7 @@ void r200InitState( r200ContextPtr rmesa )
 						R200_VC_NO_SWAP;
 #endif
 
-   if (!(rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL)) {
+   if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
       /* Bypass TCL */
       rmesa->hw.cst.cmd[CST_SE_VAP_CNTL_STATUS] |= (1<<8);
    }
@@ -743,28 +1211,28 @@ void r200InitState( r200ContextPtr rmesa )
       rmesa->hw.tex[i].cmd[TEX_PP_TXFORMAT_X] =
          (/* R200_TEXCOORD_PROJ | */
           0x100000);	/* Small default bias */
-      if (rmesa->r200Screen->drmSupportsFragShader) {
+      if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
 	 rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET_NEWDRM] =
-	     rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	     rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
 	 rmesa->hw.tex[i].cmd[TEX_PP_CUBIC_FACES] = 0;
 	 rmesa->hw.tex[i].cmd[TEX_PP_TXMULTI_CTL] = 0;
       }
       else {
 	  rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET_OLDDRM] =
-	     rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	     rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
      }
 
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0;
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F1] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F2] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F3] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F4] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F5] =
-         rmesa->r200Screen->texOffset[RADEON_LOCAL_TEX_HEAP];
+         rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
 
       rmesa->hw.pix[i].cmd[PIX_PP_TXCBLEND] =
          (R200_TXC_ARG_A_ZERO |
@@ -967,5 +1435,7 @@ void r200InitState( r200ContextPtr rmesa )
 
    r200LightingSpaceChange( ctx );
 
-   rmesa->hw.all_dirty = GL_TRUE;
+   rmesa->radeon.hw.all_dirty = GL_TRUE;
+
+   rcommonInitCmdBuf(&rmesa->radeon);
 }
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c
index b25f028244..83e70b586d 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.c
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.c
@@ -55,27 +55,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 
 
-static void flush_last_swtcl_prim( r200ContextPtr rmesa  );
-
-
 /***********************************************************************
- *                         Initialization 
+ *                         Initialization
  ***********************************************************************/
 
 #define EMIT_ATTR( ATTR, STYLE, F0 )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
    fmt_0 |= F0;								\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 static void r200SetVertexFormat( GLcontext *ctx )
@@ -100,7 +97,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    }
 
    assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-   rmesa->swtcl.vertex_attr_count = 0;
+   rmesa->radeon.swtcl.vertex_attr_count = 0;
 
    /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
     * build up a hardware vertex.
@@ -121,7 +118,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    }
 
    rmesa->swtcl.coloroffset = offset;
-#if MESA_LITTLE_ENDIAN 
+#if MESA_LITTLE_ENDIAN
    EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA, (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT) );
 #else
    EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR, (R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT) );
@@ -132,7 +129,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
    if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) ||
        RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
 
-#if MESA_LITTLE_ENDIAN 
+#if MESA_LITTLE_ENDIAN
       if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
 	 rmesa->swtcl.specoffset = offset;
 	 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_RGB, (R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT) );
@@ -185,7 +182,7 @@ static void r200SetVertexFormat( GLcontext *ctx )
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= R200_FOG_USE_SPEC_ALPHA;
    }
 
-   if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) ||
+   if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) ||
 	(rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0) ||
 	(rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
       R200_NEWPRIM(rmesa);
@@ -193,26 +190,20 @@ static void r200SetVertexFormat( GLcontext *ctx )
       rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = fmt_0;
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = fmt_1;
 
-      rmesa->swtcl.vertex_size =
+      rmesa->radeon.swtcl.vertex_size =
 	  _tnl_install_attrs( ctx,
-			      rmesa->swtcl.vertex_attrs, 
-			      rmesa->swtcl.vertex_attr_count,
+			      rmesa->radeon.swtcl.vertex_attrs,
+			      rmesa->radeon.swtcl.vertex_attr_count,
 			      NULL, 0 );
-      rmesa->swtcl.vertex_size /= 4;
-      RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
+      rmesa->radeon.swtcl.vertex_size /= 4;
+      RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset );
    }
 }
 
 
 static void r200RenderStart( GLcontext *ctx )
 {
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
    r200SetVertexFormat( ctx );
-
-   if (rmesa->dma.flush != 0 && 
-       rmesa->dma.flush != flush_last_swtcl_prim)
-      rmesa->dma.flush( rmesa );
 }
 
 
@@ -232,7 +223,7 @@ void r200ChooseVertexState( GLcontext *ctx )
     * rasterization fallback.  As this function will be called again when we
     * leave a rasterization fallback, we can just skip it for now.
     */
-   if (rmesa->Fallback != 0)
+   if (rmesa->radeon.Fallback != 0)
       return;
 
    vte = rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL];
@@ -273,78 +264,27 @@ void r200ChooseVertexState( GLcontext *ctx )
    }
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( r200ContextPtr rmesa  )
+void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
 {
-   if (R200_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   rmesa->dma.flush = NULL;
-
-   if (rmesa->dma.current.buf) {
-      struct r200_dma_region *current = &rmesa->dma.current;
-      GLuint current_offset = (rmesa->r200Screen->gart_buffer_offset +
-			       current->buf->buf->idx * RADEON_BUFFER_SIZE + 
-			       current->start);
-
-      assert (!(rmesa->swtcl.hw_primitive & R200_VF_PRIM_WALK_IND));
-
-      assert (current->start + 
-	      rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	      current->ptr);
-
-      if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
-	 r200EnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ +
-			        rmesa->hw.max_state_size + VBUF_BUFSZ );
-	 r200EmitVertexAOS( rmesa,
-			      rmesa->swtcl.vertex_size,
-			      current_offset);
-
-	 r200EmitVbufPrim( rmesa,
-			   rmesa->swtcl.hw_primitive,
-			   rmesa->swtcl.numverts);
-      }
-
-      rmesa->swtcl.numverts = 0;
-      current->start = current->ptr;
-   }
-}
-
-
-/* Alloc space in the current dma region.
- */
-static INLINE void *
-r200AllocDmaLowVerts( r200ContextPtr rmesa, int nverts, int vsize )
-{
-   GLuint bytes = vsize * nverts;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      r200RefillCurrentDmaRegion( rmesa );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   rcommonEnsureCmdBufSpace(&rmesa->radeon,
+			    rmesa->radeon.hw.max_state_size + (12*sizeof(int)),
+			    __FUNCTION__);
 
-   if (!rmesa->dma.flush) {
-      rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-      rmesa->dma.flush = flush_last_swtcl_prim;
-   }
 
-   ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
-   ASSERT( rmesa->dma.flush == flush_last_swtcl_prim );
-   ASSERT( rmesa->dma.current.start + 
-	   rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	   rmesa->dma.current.ptr );
+   radeonEmitState(&rmesa->radeon);
+   r200EmitVertexAOS( rmesa,
+		      rmesa->radeon.swtcl.vertex_size,
+		      rmesa->radeon.dma.current,
+		      current_offset);
 
 
-   {
-      GLubyte *head = (GLubyte *) (rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->dma.current.ptr += bytes;
-      rmesa->swtcl.numverts += nverts;
-      return head;
-   }
+   r200EmitVbufPrim( rmesa,
+		     rmesa->radeon.swtcl.hw_primitive,
+		     rmesa->radeon.swtcl.numverts);
 
 }
 
-
 /**************************************************************************/
 
 
@@ -392,13 +332,13 @@ static void r200ResetLineStipple( GLcontext *ctx );
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
 #define CTX_ARG r200ContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) r200AllocDmaLowVerts( rmesa, n, size * 4 )
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 )
 #define LOCAL_VARS						\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   const char *r200verts = (char *)rmesa->swtcl.verts;
-#define VERT(x) (r200Vertex *)(r200verts + ((x) * vertsize * sizeof(int)))
-#define VERTEX r200Vertex 
+   const char *r200verts = (char *)rmesa->radeon.swtcl.verts;
+#define VERT(x) (radeonVertex *)(r200verts + ((x) * vertsize * sizeof(int)))
+#define VERTEX radeonVertex
 #define DO_DEBUG_VERTS (1 && (R200_DEBUG & DEBUG_VERTS))
 
 #undef TAG
@@ -456,11 +396,11 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int)))
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
 
 #define VERT_SET_RGBA( v, c )  					\
 do {								\
-   r200_color_t *color = (r200_color_t *)&((v)->ui[coloroffset]);	\
+   radeon_color_t *color = (radeon_color_t *)&((v)->ui[coloroffset]);	\
    UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]);		\
    UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]);		\
    UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]);		\
@@ -472,7 +412,7 @@ do {								\
 #define VERT_SET_SPEC( v, c )					\
 do {								\
    if (specoffset) {						\
-      r200_color_t *spec = (r200_color_t *)&((v)->ui[specoffset]);	\
+      radeon_color_t *spec = (radeon_color_t *)&((v)->ui[specoffset]);	\
       UNCLAMPED_FLOAT_TO_UBYTE(spec->red, (c)[0]);	\
       UNCLAMPED_FLOAT_TO_UBYTE(spec->green, (c)[1]);	\
       UNCLAMPED_FLOAT_TO_UBYTE(spec->blue, (c)[2]);	\
@@ -481,8 +421,8 @@ do {								\
 #define VERT_COPY_SPEC( v0, v1 )			\
 do {							\
    if (specoffset) {					\
-      r200_color_t *spec0 = (r200_color_t *)&((v0)->ui[specoffset]);	\
-      r200_color_t *spec1 = (r200_color_t *)&((v1)->ui[specoffset]);	\
+      radeon_color_t *spec0 = (radeon_color_t *)&((v0)->ui[specoffset]);	\
+      radeon_color_t *spec1 = (radeon_color_t *)&((v1)->ui[specoffset]);	\
       spec0->red   = spec1->red;	\
       spec0->green = spec1->green;	\
       spec0->blue  = spec1->blue; 	\
@@ -513,7 +453,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) r200RasterPrimitive( ctx, reduced_hw_prim(ctx, x) )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -569,8 +509,8 @@ static void init_rast_tab( void )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
    r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *r200verts = (char *)rmesa->swtcl.verts;		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *r200verts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -599,13 +539,13 @@ void r200ChooseRenderState( GLcontext *ctx )
    GLuint index = 0;
    GLuint flags = ctx->_TriangleCaps;
 
-   if (!rmesa->TclFallback || rmesa->Fallback) 
+   if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback)
       return;
 
    if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R200_TWOSIDE_BIT;
    if (flags & DD_TRI_UNFILLED)      index |= R200_UNFILLED_BIT;
 
-   if (index != rmesa->swtcl.RenderIndex) {
+   if (index != rmesa->radeon.swtcl.RenderIndex) {
       tnl->Driver.Render.Points = rast_tab[index].points;
       tnl->Driver.Render.Line = rast_tab[index].line;
       tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -622,7 +562,7 @@ void r200ChooseRenderState( GLcontext *ctx )
 	 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
       }
 
-      rmesa->swtcl.RenderIndex = index;
+      rmesa->radeon.swtcl.RenderIndex = index;
    }
 }
 
@@ -636,7 +576,7 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
-   if (rmesa->swtcl.hw_primitive != hwprim) {
+   if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
       /* need to disable perspective-correct texturing for point sprites */
       if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) {
 	 if (rmesa->hw.set.cmd[SET_RE_CNTL] & R200_PERSPECTIVE_ENABLE) {
@@ -649,15 +589,15 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 	 rmesa->hw.set.cmd[SET_RE_CNTL] |= R200_PERSPECTIVE_ENABLE;
       }
       R200_NEWPRIM( rmesa );
-      rmesa->swtcl.hw_primitive = hwprim;
+      rmesa->radeon.swtcl.hw_primitive = hwprim;
    }
 }
 
 static void r200RenderPrimitive( GLcontext *ctx, GLenum prim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   rmesa->swtcl.render_primitive = prim;
-   if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) 
+   rmesa->radeon.swtcl.render_primitive = prim;
+   if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED))
       r200RasterPrimitive( ctx, reduced_hw_prim(ctx, prim) );
 }
 
@@ -701,15 +641,15 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->Fallback;
+   GLuint oldfallback = rmesa->radeon.Fallback;
 
    if (mode) {
-      rmesa->Fallback |= bit;
+      rmesa->radeon.Fallback |= bit;
       if (oldfallback == 0) {
-	 R200_FIREVERTICES( rmesa );
+	 radeon_firevertices(&rmesa->radeon);
 	 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_TRUE );
 	 _swsetup_Wakeup( ctx );
-	 rmesa->swtcl.RenderIndex = ~0;
+	 rmesa->radeon.swtcl.RenderIndex = ~0;
          if (R200_DEBUG & DEBUG_FALLBACKS) {
             fprintf(stderr, "R200 begin rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
@@ -717,7 +657,7 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
       }
    }
    else {
-      rmesa->Fallback &= ~bit;
+      rmesa->radeon.Fallback &= ~bit;
       if (oldfallback == bit) {
 
 	 _swrast_flush( ctx );
@@ -731,14 +671,14 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 
 	 tnl->Driver.Render.ResetLineStipple = r200ResetLineStipple;
 	 TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_FALSE );
-	 if (rmesa->TclFallback) {
-	    /* These are already done if rmesa->TclFallback goes to
+	 if (rmesa->radeon.TclFallback) {
+	    /* These are already done if rmesa->radeon.TclFallback goes to
 	     * zero above. But not if it doesn't (R200_NO_TCL for
 	     * example?)
 	     */
 	    _tnl_invalidate_vertex_state( ctx, ~0 );
 	    _tnl_invalidate_vertices( ctx, ~0 );
-	    RENDERINPUTS_ZERO( rmesa->tnl_index_bitset );
+	    RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset );
 	    r200ChooseVertexState( ctx );
 	    r200ChooseRenderState( ctx );
 	 }
@@ -755,7 +695,7 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 
 /**
  * Cope with depth operations by drawing individual pixels as points.
- * 
+ *
  * \todo
  * The way the vertex state is set in this routine is hokey.  It seems to
  * work, but it's very hackish.  This whole routine is pretty hackish.  If
@@ -770,14 +710,14 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 		  const GLubyte *bitmap )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   const GLfloat *rc = ctx->Current.RasterColor; 
+   const GLfloat *rc = ctx->Current.RasterColor;
    GLint row, col;
-   r200Vertex vert;
+   radeonVertex vert;
    GLuint orig_vte;
    GLuint h;
 
 
-   /* Turn off tcl.  
+   /* Turn off tcl.
     */
    TCL_FALLBACK( ctx, R200_TCL_FALLBACK_BITMAP, 1 );
 
@@ -794,7 +734,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
       vte |= R200_VTX_W0_FMT;
       vap &= ~R200_VAP_FORCE_W_TO_ONE;
 
-      rmesa->swtcl.vertex_size = 5;
+      rmesa->radeon.swtcl.vertex_size = 5;
 
       if ( (rmesa->hw.vtx.cmd[VTX_VTXFMT_0] != fmt_0)
 	   || (rmesa->hw.vtx.cmd[VTX_VTXFMT_1] != fmt_1) ) {
@@ -828,7 +768,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 					   R200_VPORT_Z_SCALE_ENA |
 					   R200_VPORT_X_OFFSET_ENA |
 					   R200_VPORT_Y_OFFSET_ENA |
-					   R200_VPORT_Z_OFFSET_ENA); 
+					   R200_VPORT_Z_OFFSET_ENA);
 
    /* Turn off other stuff:  Stipple?, texture?, blending?, etc.
     */
@@ -871,16 +811,16 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 
    /* Update window height
     */
-   LOCK_HARDWARE( rmesa );
-   UNLOCK_HARDWARE( rmesa );
-   h = rmesa->dri.drawable->h + rmesa->dri.drawable->y;
-   px += rmesa->dri.drawable->x;
+   LOCK_HARDWARE( &rmesa->radeon );
+   UNLOCK_HARDWARE( &rmesa->radeon );
+   h = radeon_get_drawable(&rmesa->radeon)->h + radeon_get_drawable(&rmesa->radeon)->y;
+   px += radeon_get_drawable(&rmesa->radeon)->x;
 
    /* Clipping handled by existing mechansims in r200_ioctl.c?
     */
    for (row=0; row<height; row++) {
-      const GLubyte *src = (const GLubyte *) 
-	 _mesa_image_address2d(unpack, bitmap, width, height, 
+      const GLubyte *src = (const GLubyte *)
+	 _mesa_image_address2d(unpack, bitmap, width, height,
                                GL_COLOR_INDEX, GL_BITMAP, row, 0 );
 
       if (unpack->LsbFirst) {
@@ -929,7 +869,7 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 
    /* Need to restore vertexformat?
     */
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       r200ChooseVertexState( ctx );
 }
 
@@ -959,20 +899,12 @@ void r200InitSwtcl( GLcontext *ctx )
    tnl->Driver.Render.Interp = _tnl_interp;
 
    /* FIXME: what are these numbers? */
-   _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
+   _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12,
 		       36 * sizeof(GLfloat) );
-   
-   rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->swtcl.render_primitive = GL_TRIANGLES;
-   rmesa->swtcl.hw_primitive = 0;
-}
-
 
-void r200DestroySwtcl( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (rmesa->swtcl.indexed_verts.buf) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, __FUNCTION__ );
+   rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+   rmesa->radeon.swtcl.hw_primitive = 0;
 }
+
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h
index 8c29fd0c99..b0905879d7 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.h
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.h
@@ -39,7 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_context.h"
 
 extern void r200InitSwtcl( GLcontext *ctx );
-extern void r200DestroySwtcl( GLcontext *ctx );
 
 extern void r200ChooseRenderState( GLcontext *ctx );
 extern void r200ChooseVertexState( GLcontext *ctx );
@@ -52,15 +51,11 @@ extern void r200BuildVertices( GLcontext *ctx, GLuint start, GLuint count,
 extern void r200PrintSetupFlags(char *msg, GLuint flags );
 
 
-extern void r200_emit_indexed_verts( GLcontext *ctx,
-				       GLuint start,
-				       GLuint count );
-
 extern void r200_translate_vertex( GLcontext *ctx, 
-				     const r200Vertex *src, 
+				     const radeonVertex *src, 
 				     SWvertex *dst );
 
-extern void r200_print_vertex( GLcontext *ctx, const r200Vertex *v );
+extern void r200_print_vertex( GLcontext *ctx, const radeonVertex *v );
 
 extern void r200_import_float_colors( GLcontext *ctx );
 extern void r200_import_float_spec_colors( GLcontext *ctx );
@@ -70,5 +65,5 @@ extern void r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 			      const struct gl_pixelstore_attrib *unpack,
 			      const GLubyte *bitmap );
 
-
+void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
 #endif
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index 99aecfe1e9..580370933e 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -123,7 +123,7 @@ static GLboolean discrete_prim[0x10] = {
 
 #define RESET_STIPPLE() do {			\
    R200_STATECHANGE( rmesa, lin );		\
-   r200EmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);			\
 } while (0)
 
 #define AUTO_STIPPLE( mode )  do {		\
@@ -134,7 +134,7 @@ static GLboolean discrete_prim[0x10] = {
    else						\
       rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &=	\
 	 ~R200_LINE_PATTERN_AUTO_RESET;	\
-   r200EmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);			\
 } while (0)
 
 
@@ -142,26 +142,24 @@ static GLboolean discrete_prim[0x10] = {
 
 static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr ) 
 {
-   if (rmesa->dma.flush == r200FlushElts &&
-       rmesa->store.cmd_used + nr*2 < R200_CMD_BUF_SZ) {
+   if (rmesa->radeon.dma.flush == r200FlushElts &&
+       rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) {
 
-      GLushort *dest = (GLushort *)(rmesa->store.cmd_buf +
-				    rmesa->store.cmd_used);
+      GLushort *dest = (GLushort *)(rmesa->radeon.tcl.elt_dma_bo->ptr +
+				    rmesa->tcl.elt_used);
 
-      rmesa->store.cmd_used += nr*2;
+      rmesa->tcl.elt_used += nr*2;
 
       return dest;
    }
    else {
-      if (rmesa->dma.flush)
-	 rmesa->dma.flush( rmesa );
+      if (rmesa->radeon.dma.flush)
+	 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
 
-      r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			     rmesa->hw.max_state_size + ELTS_BUFSZ(nr) );
+      rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->radeon.tcl.aos_count), __FUNCTION__);
 
       r200EmitAOS( rmesa,
-		   rmesa->tcl.aos_components,
-		   rmesa->tcl.nr_aos_components, 0 );
+		   rmesa->radeon.tcl.aos_count, 0 );
 
       return r200AllocEltsOpenEnded( rmesa, rmesa->tcl.hw_primitive, nr );
    }
@@ -188,13 +186,14 @@ static void r200EmitPrim( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT( ctx );
    r200TclPrimitive( ctx, prim, hwprim );
    
-   r200EnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			  rmesa->hw.max_state_size + VBUF_BUFSZ );
+   //   fprintf(stderr,"Emit prim %d\n", rmesa->radeon.tcl.aos_count);
+   rcommonEnsureCmdBufSpace( &rmesa->radeon,
+			     AOS_BUFSZ(rmesa->radeon.tcl.aos_count) +
+			     rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ );
 
    r200EmitAOS( rmesa,
-		  rmesa->tcl.aos_components,
-		  rmesa->tcl.nr_aos_components,
-		  start );
+		rmesa->radeon.tcl.aos_count,
+		start );
    
    /* Why couldn't this packet have taken an offset param?
     */
@@ -394,7 +393,7 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
 
    /* TODO: separate this from the swtnl pipeline 
     */
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       return GL_TRUE;	/* fallback to software t&l */
 
    if (R200_DEBUG & DEBUG_PRIMS)
@@ -405,8 +404,9 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
 
    /* Validate state:
     */
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
+   if (rmesa->radeon.NewGLState)
+      if (!r200ValidateState( ctx ))
+         return GL_TRUE; /* fallback to sw t&l */
 
    if (!ctx->VertexProgram._Enabled) {
    /* NOTE: inputs != tnl->render_inputs - these are the untransformed
@@ -481,7 +481,7 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
 
    /* Do the actual work:
     */
-   r200ReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
+   radeonReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
    r200EmitArrays( ctx, vimap_rev );
 
    rmesa->tcl.Elts = VB->Elts;
@@ -545,7 +545,7 @@ static void transition_to_swtnl( GLcontext *ctx )
    tnl->Driver.NotifyMaterialChange = 
       _mesa_validate_all_lighting_tables;
 
-   r200ReleaseArrays( ctx, ~0 );
+   radeonReleaseArrays( ctx, ~0 );
 
    /* Still using the D3D based hardware-rasterizer from the radeon;
     * need to put the card into D3D mode to make it work:
@@ -565,15 +565,11 @@ static void transition_to_hwtnl( GLcontext *ctx )
 
    tnl->Driver.NotifyMaterialChange = r200UpdateMaterial;
 
-   if ( rmesa->dma.flush )			
-      rmesa->dma.flush( rmesa );	
+   if ( rmesa->radeon.dma.flush )			
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	
 
-   rmesa->dma.flush = NULL;
+   rmesa->radeon.dma.flush = NULL;
    
-   if (rmesa->swtcl.indexed_verts.buf) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
-			      __FUNCTION__ );
-
    R200_STATECHANGE( rmesa, vap );
    rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_TCL_ENABLE;
    rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_FORCE_W_TO_ONE;
@@ -631,10 +627,10 @@ static char *getFallbackString(GLuint bit)
 void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->TclFallback;
+   GLuint oldfallback = rmesa->radeon.TclFallback;
 
    if (mode) {
-      rmesa->TclFallback |= bit;
+      rmesa->radeon.TclFallback |= bit;
       if (oldfallback == 0) {
 	 if (R200_DEBUG & DEBUG_FALLBACKS) 
 	    fprintf(stderr, "R200 begin tcl fallback %s\n",
@@ -643,7 +639,7 @@ void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
       }
    }
    else {
-      rmesa->TclFallback &= ~bit;
+      rmesa->radeon.TclFallback &= ~bit;
       if (oldfallback == bit) {
 	 if (R200_DEBUG & DEBUG_FALLBACKS) 
 	    fprintf(stderr, "R200 end tcl fallback %s\n",
diff --git a/src/mesa/drivers/dri/r200/r200_tex.c b/src/mesa/drivers/dri/r200/r200_tex.c
index 259f35a34c..9f79157915 100644
--- a/src/mesa/drivers/dri/r200/r200_tex.c
+++ b/src/mesa/drivers/dri/r200/r200_tex.c
@@ -43,8 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/teximage.h"
 #include "main/texobj.h"
 
-#include "texmem.h"
-
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -63,10 +62,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \param twrap Wrap mode for the \a t texture coordinate
  */
 
-static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap )
+static void r200SetTexWrap( radeonTexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap )
 {
    GLboolean  is_clamp = GL_FALSE;
    GLboolean  is_clamp_to_border = GL_FALSE;
+   struct gl_texture_object *tObj = &t->base;
 
    t->pp_txfilter &= ~(R200_CLAMP_S_MASK | R200_CLAMP_T_MASK | R200_BORDER_MODE_D3D);
 
@@ -103,7 +103,7 @@ static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum
       _mesa_problem(NULL, "bad S wrap mode in %s", __FUNCTION__);
    }
 
-   if (t->base.tObj->Target != GL_TEXTURE_1D) {
+   if (tObj->Target != GL_TEXTURE_1D) {
       switch ( twrap ) {
       case GL_REPEAT:
          t->pp_txfilter |= R200_CLAMP_T_WRAP;
@@ -180,7 +180,7 @@ static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum
    t->border_fallback = (is_clamp && is_clamp_to_border);
 }
 
-static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max )
+static void r200SetTexMaxAnisotropy( radeonTexObjPtr t, GLfloat max )
 {
    t->pp_txfilter &= ~R200_MAX_ANISO_MASK;
 
@@ -205,10 +205,13 @@ static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max )
  * \param magf Texture magnification mode
  */
 
-static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf )
+static void r200SetTexFilter( radeonTexObjPtr t, GLenum minf, GLenum magf )
 {
    GLuint anisotropy = (t->pp_txfilter & R200_MAX_ANISO_MASK);
 
+   /* Force revalidation to account for switches from/to mipmapping. */
+   t->validated = GL_FALSE;
+
    t->pp_txfilter &= ~(R200_MIN_FILTER_MASK | R200_MAG_FILTER_MASK);
    t->pp_txformat_x &= ~R200_VOLUME_FILTER_MASK;
 
@@ -267,701 +270,16 @@ static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf )
    }
 }
 
-static void r200SetTexBorderColor( r200TexObjPtr t, const GLfloat color[4] )
+static void r200SetTexBorderColor( radeonTexObjPtr t, const GLfloat color[4] )
 {
    GLubyte c[4];
    CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
    CLAMPED_FLOAT_TO_UBYTE(c[1], color[1]);
    CLAMPED_FLOAT_TO_UBYTE(c[2], color[2]);
    CLAMPED_FLOAT_TO_UBYTE(c[3], color[3]);
-   t->pp_border_color = r200PackColor( 4, c[0], c[1], c[2], c[3] );
-}
-
-
-/**
- * Allocate space for and load the mesa images into the texture memory block.
- * This will happen before drawing with a new texture, or drawing with a
- * texture after it was swapped out or teximaged again.
- */
-
-static r200TexObjPtr r200AllocTexObj( struct gl_texture_object *texObj )
-{
-   r200TexObjPtr t;
-
-   t = CALLOC_STRUCT( r200_tex_obj );
-   texObj->DriverData = t;
-   if ( t != NULL ) {
-      if ( R200_DEBUG & DEBUG_TEXTURE ) {
-	 fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)texObj, 
-		  (void *)t );
-      }
-
-      /* Initialize non-image-dependent parts of the state:
-       */
-      t->base.tObj = texObj;
-      t->border_fallback = GL_FALSE;
-
-      make_empty_list( & t->base );
-
-      r200SetTexWrap( t, texObj->WrapS, texObj->WrapT, texObj->WrapR );
-      r200SetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
-      r200SetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
-      r200SetTexBorderColor( t, texObj->BorderColor );
-   }
-
-   return t;
-}
-
-/* try to find a format which will only need a memcopy */
-static const struct gl_texture_format *
-r200Choose8888TexFormat( GLenum srcFormat, GLenum srcType )
-{
-   const GLuint ui = 1;
-   const GLubyte littleEndian = *((const GLubyte *) &ui);
-
-   if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-       (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) {
-      return &_mesa_texformat_rgba8888;
-   }
-   else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-       (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-       (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) {
-      return &_mesa_texformat_rgba8888_rev;
-   }
-   else return _dri_texformat_argb8888;
-}
-
-static const struct gl_texture_format *
-r200ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
-                           GLenum format, GLenum type )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   const GLboolean do32bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32 );
-   const GLboolean force16bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16 );
-   (void) format;
-
-   switch ( internalFormat ) {
-   case 4:
-   case GL_RGBA:
-   case GL_COMPRESSED_RGBA:
-      switch ( type ) {
-      case GL_UNSIGNED_INT_10_10_10_2:
-      case GL_UNSIGNED_INT_2_10_10_10_REV:
-	 return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      default:
-         return do32bpt ?
-	    r200Choose8888TexFormat(format, type) : _dri_texformat_argb4444;
-      }
-
-   case 3:
-   case GL_RGB:
-   case GL_COMPRESSED_RGB:
-      switch ( type ) {
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_5_6_5:
-      case GL_UNSIGNED_SHORT_5_6_5_REV:
-	 return _dri_texformat_rgb565;
-      default:
-         return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-      }
-
-   case GL_RGBA8:
-   case GL_RGB10_A2:
-   case GL_RGBA12:
-   case GL_RGBA16:
-      return !force16bpt ?
-	  r200Choose8888TexFormat(format, type) : _dri_texformat_argb4444;
-
-   case GL_RGBA4:
-   case GL_RGBA2:
-      return _dri_texformat_argb4444;
-
-   case GL_RGB5_A1:
-      return _dri_texformat_argb1555;
-
-   case GL_RGB8:
-   case GL_RGB10:
-   case GL_RGB12:
-   case GL_RGB16:
-      return !force16bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-
-   case GL_RGB5:
-   case GL_RGB4:
-   case GL_R3_G3_B2:
-      return _dri_texformat_rgb565;
-
-   case GL_ALPHA:
-   case GL_ALPHA4:
-   case GL_ALPHA8:
-   case GL_ALPHA12:
-   case GL_ALPHA16:
-   case GL_COMPRESSED_ALPHA:
-   /* can't use a8 format since interpreting hw I8 as a8 would result
-      in wrong rgb values (same as alpha value instead of 0). */
-      return _dri_texformat_al88;
-
-   case 1:
-   case GL_LUMINANCE:
-   case GL_LUMINANCE4:
-   case GL_LUMINANCE8:
-   case GL_LUMINANCE12:
-   case GL_LUMINANCE16:
-   case GL_COMPRESSED_LUMINANCE:
-      return _dri_texformat_l8;
-
-   case 2:
-   case GL_LUMINANCE_ALPHA:
-   case GL_LUMINANCE4_ALPHA4:
-   case GL_LUMINANCE6_ALPHA2:
-   case GL_LUMINANCE8_ALPHA8:
-   case GL_LUMINANCE12_ALPHA4:
-   case GL_LUMINANCE12_ALPHA12:
-   case GL_LUMINANCE16_ALPHA16:
-   case GL_COMPRESSED_LUMINANCE_ALPHA:
-      return _dri_texformat_al88;
-
-   case GL_INTENSITY:
-   case GL_INTENSITY4:
-   case GL_INTENSITY8:
-   case GL_INTENSITY12:
-   case GL_INTENSITY16:
-   case GL_COMPRESSED_INTENSITY:
-       return _dri_texformat_i8;
-
-   case GL_YCBCR_MESA:
-      if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-          type == GL_UNSIGNED_BYTE)
-         return &_mesa_texformat_ycbcr;
-      else
-         return &_mesa_texformat_ycbcr_rev;
-
-   case GL_RGB_S3TC:
-   case GL_RGB4_S3TC:
-   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgb_dxt1;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgba_dxt1;
-
-   case GL_RGBA_S3TC:
-   case GL_RGBA4_S3TC:
-   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-      return &_mesa_texformat_rgba_dxt3;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-      return &_mesa_texformat_rgba_dxt5;
-
-   default:
-      _mesa_problem(ctx,
-         "unexpected internalFormat 0x%x in r200ChooseTextureFormat",
-         (int) internalFormat);
-      return NULL;
-   }
-
-   return NULL; /* never get here */
-}
-
-
-static GLboolean
-r200ValidateClientStorage( GLcontext *ctx, GLenum target,
-			   GLint internalFormat,
-			   GLint srcWidth, GLint srcHeight, 
-                           GLenum format, GLenum type,  const void *pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if ( R200_DEBUG & DEBUG_TEXTURE )
-      fprintf(stderr, "intformat %s format %s type %s\n",
-	      _mesa_lookup_enum_by_nr( internalFormat ),
-	      _mesa_lookup_enum_by_nr( format ),
-	      _mesa_lookup_enum_by_nr( type ));
-
-   if (!ctx->Unpack.ClientStorage)
-      return 0;
-
-   if (ctx->_ImageTransferState ||
-       texImage->IsCompressed ||
-       texObj->GenerateMipmap)
-      return 0;
-
-
-   /* This list is incomplete, may be different on ppc???
-    */
-   switch ( internalFormat ) {
-   case GL_RGBA:
-      if ( format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV ) {
-	 texImage->TexFormat = _dri_texformat_argb8888;
-      }
-      else
-	 return 0;
-      break;
-
-   case GL_RGB:
-      if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) {
-	 texImage->TexFormat = _dri_texformat_rgb565;
-      }
-      else
-	 return 0;
-      break;
-
-   case GL_YCBCR_MESA:
-      if ( format == GL_YCBCR_MESA && 
-	   type == GL_UNSIGNED_SHORT_8_8_REV_APPLE ) {
-	 texImage->TexFormat = &_mesa_texformat_ycbcr_rev;
-      }
-      else if ( format == GL_YCBCR_MESA && 
-		(type == GL_UNSIGNED_SHORT_8_8_APPLE || 
-		 type == GL_UNSIGNED_BYTE)) {
-	 texImage->TexFormat = &_mesa_texformat_ycbcr;
-      }
-      else
-	 return 0;
-      break;
-
-   default:
-      return 0;
-   }
-
-   /* Could deal with these packing issues, but currently don't:
-    */
-   if (packing->SkipPixels || 
-       packing->SkipRows || 
-       packing->SwapBytes ||
-       packing->LsbFirst) {
-      return 0;
-   }
-
-   {      
-      GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth,
-						  format, type);
-
-      
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf(stderr, "%s: srcRowStride %d/%x\n", 
-		 __FUNCTION__, srcRowStride, srcRowStride);
-
-      /* Could check this later in upload, pitch restrictions could be
-       * relaxed, but would need to store the image pitch somewhere,
-       * as packing details might change before image is uploaded:
-       */
-      if (!r200IsGartMemory( rmesa, pixels, srcHeight * srcRowStride ) ||
-	  (srcRowStride & 63))
-	 return 0;
-
-
-      /* Have validated that _mesa_transfer_teximage would be a straight
-       * memcpy at this point.  NOTE: future calls to TexSubImage will
-       * overwrite the client data.  This is explicitly mentioned in the
-       * extension spec.
-       */
-      texImage->Data = (void *)pixels;
-      texImage->IsClientData = GL_TRUE;
-      texImage->RowStride = srcRowStride / texImage->TexFormat->TexelBytes;
-
-      return 1;
-   }
-}
-
-
-static void r200TexImage1D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_teximage1d(ctx, target, level, internalFormat,
-                          width, border, format, type, pixels,
-                          &ctx->Unpack, texObj, texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void r200TexSubImage1D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset,
-                                 GLsizei width,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
-			     format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void r200TexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
-         return;
-      }
-   }
-
-   texImage->IsClientData = GL_FALSE;
-
-   if (r200ValidateClientStorage( ctx, target, 
-				  internalFormat, 
-				  width, height, 
-				  format, type, pixels, 
-				  packing, texObj, texImage)) {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); 
-   }
-   else {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); 
-
-      /* Normal path: copy (to cached memory) and eventually upload
-       * via another copy to GART memory and then a blit...  Could
-       * eliminate one copy by going straight to (permanent) GART.
-       *
-       * Note, this will call r200ChooseTextureFormat.
-       */
-      _mesa_store_teximage2d(ctx, target, level, internalFormat,
-			     width, height, border, format, type, pixels,
-			     &ctx->Unpack, texObj, texImage);
-      
-      t->dirty_images[face] |= (1 << level);
-   }
-}
-
-
-static void r200TexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-			     height, format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-static void r200CompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLsizei imageSize, const GLvoid *data,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexImage2D");
-         return;
-      }
-   }
-
-   texImage->IsClientData = GL_FALSE;
-/* can't call this, different parameters. Would never evaluate to true anyway currently
-   if (r200ValidateClientStorage( ctx, target, 
-				  internalFormat,
-				  width, height,
-				  format, type, pixels,
-				  packing, texObj, texImage)) {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__);
-   }
-   else */{
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__);
-
-      /* Normal path: copy (to cached memory) and eventually upload
-       * via another copy to GART memory and then a blit...  Could
-       * eliminate one copy by going straight to (permanent) GART.
-       *
-       * Note, this will call r200ChooseTextureFormat.
-       */
-      _mesa_store_compressed_teximage2d(ctx, target, level, internalFormat, width,
-                                 height, border, imageSize, data, texObj, texImage);
-
-      t->dirty_images[face] |= (1 << level);
-   }
+   t->pp_border_color = radeonPackColor( 4, c[0], c[1], c[2], c[3] );
 }
 
-
-static void r200CompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format,
-                                 GLsizei imageSize, const GLvoid *data,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-                            height, format, imageSize, data, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-#if ENABLE_HW_3D_TEXTURE
-static void r200TexImage3D( GLcontext *ctx, GLenum target, GLint level,
-                            GLint internalFormat,
-                            GLint width, GLint height, GLint depth,
-                            GLint border,
-                            GLenum format, GLenum type, const GLvoid *pixels,
-                            const struct gl_pixelstore_attrib *packing,
-                            struct gl_texture_object *texObj,
-                            struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage3D");
-         return;
-      }
-   }
-
-   texImage->IsClientData = GL_FALSE;
-
-#if 0
-   if (r200ValidateClientStorage( ctx, target, 
-				  internalFormat, 
-				  width, height, 
-				  format, type, pixels, 
-				  packing, texObj, texImage)) {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); 
-   }
-   else
-#endif
-   {
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); 
-
-      /* Normal path: copy (to cached memory) and eventually upload
-       * via another copy to GART memory and then a blit...  Could
-       * eliminate one copy by going straight to (permanent) GART.
-       *
-       * Note, this will call r200ChooseTextureFormat.
-       */
-      _mesa_store_teximage3d(ctx, target, level, internalFormat,
-			     width, height, depth, border,
-                             format, type, pixels,
-			     &ctx->Unpack, texObj, texImage);
-      
-      t->dirty_images[0] |= (1 << level);
-   }
-}
-#endif
-
-
-#if ENABLE_HW_3D_TEXTURE
-static void
-r200TexSubImage3D( GLcontext *ctx, GLenum target, GLint level,
-                   GLint xoffset, GLint yoffset, GLint zoffset,
-                   GLsizei width, GLsizei height, GLsizei depth,
-                   GLenum format, GLenum type,
-                   const GLvoid *pixels,
-                   const struct gl_pixelstore_attrib *packing,
-                   struct gl_texture_object *texObj,
-                   struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-/*     fprintf(stderr, "%s\n", __FUNCTION__); */
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) r200AllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage3D");
-         return;
-      }
-      texObj->DriverData = t;
-   }
-
-   _mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset,
-                             width, height, depth,
-                             format, type, pixels, packing, texObj, texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-#endif
-
-
-
 static void r200TexEnv( GLcontext *ctx, GLenum target,
 			  GLenum pname, const GLfloat *param )
 {
@@ -983,7 +301,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target,
       GLubyte c[4];
       GLuint envColor;
       UNCLAMPED_FLOAT_TO_RGBA_CHAN( c, texUnit->EnvColor );
-      envColor = r200PackColor( 4, c[0], c[1], c[2], c[3] );
+      envColor = radeonPackColor( 4, c[0], c[1], c[2], c[3] );
       if ( rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] != envColor ) {
 	 R200_STATECHANGE( rmesa, tf );
 	 rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] = envColor;
@@ -1002,7 +320,7 @@ static void r200TexEnv( GLcontext *ctx, GLenum target,
        * NOTE: Add a small bias to the bias for conform mipsel.c test.
        */
       bias = *param + .01;
-      min = driQueryOptionb (&rmesa->optionCache, "no_neg_lod_bias") ?
+      min = driQueryOptionb (&rmesa->radeon.optionCache, "no_neg_lod_bias") ?
 	  0.0 : -16.0;
       bias = CLAMP( bias, min, 16.0 );
       b = (int)(bias * fixed_one) & R200_LOD_BIAS_MASK;
@@ -1039,7 +357,7 @@ static void r200TexParameter( GLcontext *ctx, GLenum target,
 				struct gl_texture_object *texObj,
 				GLenum pname, const GLfloat *params )
 {
-   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+   radeonTexObj* t = radeon_tex_obj(texObj);
 
    if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
       fprintf( stderr, "%s( %s )\n", __FUNCTION__,
@@ -1073,59 +391,46 @@ static void r200TexParameter( GLcontext *ctx, GLenum target,
        * we just have to rely on loading the right subset of mipmap levels
        * to simulate a clamped LOD.
        */
-      driSwapOutTextureObject( (driTextureObject *) t );
+      if (t->mt) {
+         radeon_miptree_unreference(t->mt);
+	 t->mt = 0;
+	 t->validated = GL_FALSE;
+      }
       break;
 
    default:
       return;
    }
-
-   /* Mark this texobj as dirty (one bit per tex unit)
-    */
-   t->dirty_state = TEX_ALL;
 }
 
 
-
-static void r200BindTexture( GLcontext *ctx, GLenum target,
-			       struct gl_texture_object *texObj )
-{
-   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p ) unit=%d\n", __FUNCTION__, (void *)texObj,
-	       ctx->Texture.CurrentUnit );
-   }
-
-   if ( (target == GL_TEXTURE_1D)
-	|| (target == GL_TEXTURE_2D) 
-#if ENABLE_HW_3D_TEXTURE
-	|| (target == GL_TEXTURE_3D)
-#endif
-	|| (target == GL_TEXTURE_CUBE_MAP)
-	|| (target == GL_TEXTURE_RECTANGLE_NV) ) {
-      assert( texObj->DriverData != NULL );
-   }
-}
-
-
-static void r200DeleteTexture( GLcontext *ctx,
-				 struct gl_texture_object *texObj )
+static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj,
-	       _mesa_lookup_enum_by_nr( texObj->Target ) );
+   radeonTexObj* t = radeon_tex_obj(texObj);
+
+   if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+      fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+	      (void *)texObj,
+	      _mesa_lookup_enum_by_nr(texObj->Target));
+   }
+   
+   if (rmesa) {
+      int i;
+      radeon_firevertices(&rmesa->radeon);
+      for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) {
+	 if ( t == rmesa->state.texture.unit[i].texobj ) {
+	    rmesa->state.texture.unit[i].texobj = NULL;
+	    rmesa->hw.tex[i].dirty = GL_FALSE;
+	    rmesa->hw.cube[i].dirty = GL_FALSE;
+	 }
+      }      
    }
-
-   if ( t != NULL ) {
-      if ( rmesa ) {
-         R200_FIREVERTICES( rmesa );
-      }
-
-      driDestroyTextureObject( t );
+   
+   if (t->mt) {
+      radeon_miptree_unreference(t->mt);
+      t->mt = 0;
    }
-   /* Free mipmap images and the texture object itself */
    _mesa_delete_texture_object(ctx, texObj);
 }
 
@@ -1155,46 +460,59 @@ static void r200TexGen( GLcontext *ctx,
  * Called via ctx->Driver.NewTextureObject.
  * Note: this function will be called during context creation to
  * allocate the default texture objects.
- * Note: we could use containment here to 'derive' the driver-specific
- * texture object from the core mesa gl_texture_object.  Not done at this time.
  * Fixup MaxAnisotropy according to user preference.
  */
-static struct gl_texture_object *
-r200NewTextureObject( GLcontext *ctx, GLuint name, GLenum target )
+static struct gl_texture_object *r200NewTextureObject(GLcontext * ctx,
+						      GLuint name,
+						      GLenum target)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_object *obj;
-   obj = _mesa_new_texture_object(ctx, name, target);
-   if (!obj)
-      return NULL;
-   obj->MaxAnisotropy = rmesa->initialMaxAnisotropy;
-   r200AllocTexObj( obj );
-   return obj;
+   radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
+
+   if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+     fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+	     t, _mesa_lookup_enum_by_nr(target));
+   }
+
+   _mesa_initialize_texture_object(&t->base, name, target);
+   t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+   /* Initialize hardware state */
+   r200SetTexWrap( t, t->base.WrapS, t->base.WrapT, t->base.WrapR );
+   r200SetTexMaxAnisotropy( t, t->base.MaxAnisotropy );
+   r200SetTexFilter(t, t->base.MinFilter, t->base.MagFilter);
+   r200SetTexBorderColor(t, t->base.BorderColor);
+
+   return &t->base;
 }
 
 
+
 void r200InitTextureFuncs( struct dd_function_table *functions )
 {
    /* Note: we only plug in the functions we implement in the driver
     * since _mesa_init_driver_functions() was already called.
     */
-   functions->ChooseTextureFormat	= r200ChooseTextureFormat;
-   functions->TexImage1D		= r200TexImage1D;
-   functions->TexImage2D		= r200TexImage2D;
+   functions->ChooseTextureFormat	= radeonChooseTextureFormat_mesa;
+   functions->TexImage1D		= radeonTexImage1D;
+   functions->TexImage2D		= radeonTexImage2D;
 #if ENABLE_HW_3D_TEXTURE
-   functions->TexImage3D		= r200TexImage3D;
+   functions->TexImage3D		= radeonTexImage3D;
 #else
    functions->TexImage3D		= _mesa_store_teximage3d;
 #endif
-   functions->TexSubImage1D		= r200TexSubImage1D;
-   functions->TexSubImage2D		= r200TexSubImage2D;
+   functions->TexSubImage1D		= radeonTexSubImage1D;
+   functions->TexSubImage2D		= radeonTexSubImage2D;
 #if ENABLE_HW_3D_TEXTURE
-   functions->TexSubImage3D		= r200TexSubImage3D;
+   functions->TexSubImage3D		= radeonTexSubImage3D;
 #else
    functions->TexSubImage3D		= _mesa_store_texsubimage3d;
 #endif
+   functions->GetTexImage               = radeonGetTexImage;
+   functions->GetCompressedTexImage     = radeonGetCompressedTexImage;
    functions->NewTextureObject		= r200NewTextureObject;
-   functions->BindTexture		= r200BindTexture;
+   //   functions->BindTexture		= r200BindTexture;
    functions->DeleteTexture		= r200DeleteTexture;
    functions->IsTextureResident		= driIsTextureResident;
 
@@ -1202,22 +520,16 @@ void r200InitTextureFuncs( struct dd_function_table *functions )
    functions->TexParameter		= r200TexParameter;
    functions->TexGen			= r200TexGen;
 
-   functions->CompressedTexImage2D	= r200CompressedTexImage2D;
-   functions->CompressedTexSubImage2D	= r200CompressedTexSubImage2D;
+   functions->CompressedTexImage2D	= radeonCompressedTexImage2D;
+   functions->CompressedTexSubImage2D	= radeonCompressedTexSubImage2D;
 
-   driInitTextureFormats();
+   functions->GenerateMipmap = radeonGenerateMipmap;
 
-#if 000
-   /* moved or obsolete code */
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   driInitTextureObjects( ctx, & rmesa->swapped,
-			  DRI_TEXMGR_DO_TEXTURE_1D
-			  | DRI_TEXMGR_DO_TEXTURE_2D );
+   functions->NewTextureImage = radeonNewTextureImage;
+   functions->FreeTexImageData = radeonFreeTexImageData;
+   functions->MapTexture = radeonMapTexture;
+   functions->UnmapTexture = radeonUnmapTexture;
+
+   driInitTextureFormats();
 
-   /* Hack: r200NewTextureObject is not yet installed when the
-    * default textures are created. Therefore set MaxAnisotropy of the
-    * default 2D texture now. */
-   ctx->Shared->Default2D->MaxAnisotropy = driQueryOptionf (&rmesa->optionCache,
-							    "def_max_anisotropy");
-#endif
 }
diff --git a/src/mesa/drivers/dri/r200/r200_tex.h b/src/mesa/drivers/dri/r200/r200_tex.h
index 10ff8e8a66..e122de6e5e 100644
--- a/src/mesa/drivers/dri/r200/r200_tex.h
+++ b/src/mesa/drivers/dri/r200/r200_tex.h
@@ -35,15 +35,18 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __R200_TEX_H__
 #define __R200_TEX_H__
 
+extern void r200SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv);
+extern void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+			      __DRIdrawable *dPriv);
 extern void r200SetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 			     unsigned long long offset, GLint depth,
 			     GLuint pitch);
 
 extern void r200UpdateTextureState( GLcontext *ctx );
 
-extern int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face );
+extern int r200UploadTexImages( r200ContextPtr rmesa, radeonTexObjPtr t, GLuint face );
 
-extern void r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t );
+extern void r200DestroyTexObj( r200ContextPtr rmesa, radeonTexObjPtr t );
 
 extern void r200InitTextureFuncs( struct dd_function_table *functions );
 
diff --git a/src/mesa/drivers/dri/r200/r200_texmem.c b/src/mesa/drivers/dri/r200/r200_texmem.c
deleted file mode 100644
index 3b81ac0c80..0000000000
--- a/src/mesa/drivers/dri/r200/r200_texmem.c
+++ /dev/null
@@ -1,530 +0,0 @@
-/**************************************************************************
-
-Copyright (C) Tungsten Graphics 2002.  All Rights Reserved.  
-The Weather Channel, Inc. funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86
-license. This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation on the rights to use, copy, modify, merge, publish,
-distribute, sub license, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
-SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Kevin E. Martin <martin@valinux.com>
- *   Gareth Hughes <gareth@valinux.com>
- *
- */
- 
-#include <errno.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/colormac.h"
-#include "main/macros.h"
-#include "r200_context.h"
-#include "r200_ioctl.h"
-#include "r200_tex.h"
-#include "radeon_reg.h"
-
-#include <unistd.h>  /* for usleep() */
-
-
-/**
- * Destroy any device-dependent state associated with the texture.  This may
- * include NULLing out hardware state that points to the texture.
- */
-void
-r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t )
-{
-   if ( R200_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, 
-	       (void *)t, (void *)t->base.tObj );
-   }
-
-   if ( rmesa != NULL ) {
-      unsigned   i;
-
-
-      for ( i = 0 ; i < rmesa->glCtx->Const.MaxTextureUnits ; i++ ) {
-	 if ( t == rmesa->state.texture.unit[i].texobj ) {
-	    rmesa->state.texture.unit[i].texobj = NULL;
-	    rmesa->hw.tex[i].dirty = GL_FALSE;
-	    rmesa->hw.cube[i].dirty = GL_FALSE;
-	 }
-      }
-   }
-}
-
-
-/* ------------------------------------------------------------
- * Texture image conversions
- */
-
-
-static void r200UploadGARTClientSubImage( r200ContextPtr rmesa,
-					  r200TexObjPtr t, 
-					  struct gl_texture_image *texImage,
-					  GLint hwlevel,
-					  GLint x, GLint y, 
-					  GLint width, GLint height )
-{
-   const struct gl_texture_format *texFormat = texImage->TexFormat;
-   GLuint srcPitch, dstPitch;
-   int blit_format;
-   int srcOffset;
-
-   /*
-    * XXX it appears that we always upload the full image, not a subimage.
-    * I.e. x==0, y==0, width=texWidth, height=texWidth.  If this is ever
-    * changed, the src pitch will have to change.
-    */
-   switch ( texFormat->TexelBytes ) {
-   case 1:
-      blit_format = R200_CP_COLOR_FORMAT_CI8;
-      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-      break;
-   case 2:
-      blit_format = R200_CP_COLOR_FORMAT_RGB565;
-      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-      break;
-   case 4:
-      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
-      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-      break;
-   default:
-      return;
-   }
-
-   t->image[0][hwlevel].data = texImage->Data;
-   srcOffset = r200GartOffsetFromVirtual( rmesa, texImage->Data );
-
-   assert( srcOffset != ~0 );
-
-   /* Don't currently need to cope with small pitches?
-    */
-   width = texImage->Width;
-   height = texImage->Height;
-
-   r200EmitWait( rmesa, RADEON_WAIT_3D );
-
-   r200EmitBlit( rmesa, blit_format, 
-		 srcPitch,  
-		 srcOffset,   
-		 dstPitch,
-		 t->bufAddr,
-		 x, 
-		 y, 
-		 t->image[0][hwlevel].x + x,
-		 t->image[0][hwlevel].y + y, 
-		 width,
-		 height );
-
-   r200EmitWait( rmesa, RADEON_WAIT_2D );
-}
-
-static void r200UploadRectSubImage( r200ContextPtr rmesa,
-				    r200TexObjPtr t, 
-				    struct gl_texture_image *texImage,
-				    GLint x, GLint y, 
-				    GLint width, GLint height )
-{
-   const struct gl_texture_format *texFormat = texImage->TexFormat;
-   int blit_format, dstPitch, done;
-
-   switch ( texFormat->TexelBytes ) {
-   case 1:
-      blit_format = R200_CP_COLOR_FORMAT_CI8;
-      break;
-   case 2:
-      blit_format = R200_CP_COLOR_FORMAT_RGB565;
-      break;
-   case 4:
-      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
-      break;
-   default:
-      return;
-   }
-
-   t->image[0][0].data = texImage->Data;
-
-   /* Currently don't need to cope with small pitches.
-    */
-   width = texImage->Width;
-   height = texImage->Height;
-   dstPitch = t->pp_txpitch + 32;
-
-   if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) {
-      /* In this case, could also use GART texturing.  This is
-       * currently disabled, but has been tested & works.
-       */
-      if ( !t->image_override )
-         t->pp_txoffset = r200GartOffsetFromVirtual( rmesa, texImage->Data );
-      t->pp_txpitch = texImage->RowStride * texFormat->TexelBytes - 32;
-
-      if (R200_DEBUG & DEBUG_TEXTURE)
-	 fprintf(stderr, 
-		 "Using GART texturing for rectangular client texture\n");
-
-      /* Release FB memory allocated for this image:
-       */
-      /* FIXME This may not be correct as driSwapOutTextureObject sets
-       * FIXME dirty_images.  It may be fine, though.
-       */
-      if ( t->base.memBlock ) {
-	 driSwapOutTextureObject( (driTextureObject *) t );
-      }
-   }
-   else if (texImage->IsClientData) {
-      /* Data already in GART memory, with usable pitch.
-       */
-      GLuint srcPitch;
-      srcPitch = texImage->RowStride * texFormat->TexelBytes;
-      r200EmitBlit( rmesa, 
-		    blit_format, 
-		    srcPitch,
-		    r200GartOffsetFromVirtual( rmesa, texImage->Data ),   
-		    dstPitch, t->bufAddr,
-		    0, 0, 
-		    0, 0, 
-		    width, height );
-   }
-   else {
-      /* Data not in GART memory, or bad pitch.
-       */
-      for (done = 0; done < height ; ) {
-	 struct r200_dma_region region;
-	 int lines = MIN2( height - done, RADEON_BUFFER_SIZE / dstPitch );
-	 int src_pitch;
-	 char *tex;
-
-         src_pitch = texImage->RowStride * texFormat->TexelBytes;
-
-	 tex = (char *)texImage->Data + done * src_pitch;
-
-	 memset(&region, 0, sizeof(region));
-	 r200AllocDmaRegion( rmesa, &region, lines * dstPitch, 1024 );
-
-	 /* Copy texdata to dma:
-	  */
-	 if (0)
-	    fprintf(stderr, "%s: src_pitch %d dst_pitch %d\n",
-		    __FUNCTION__, src_pitch, dstPitch);
-
-	 if (src_pitch == dstPitch) {
-	    memcpy( region.address + region.start, tex, lines * src_pitch );
-	 } 
-	 else {
-	    char *buf = region.address + region.start;
-	    int i;
-	    for (i = 0 ; i < lines ; i++) {
-	       memcpy( buf, tex, src_pitch );
-	       buf += dstPitch;
-	       tex += src_pitch;
-	    }
-	 }
-
-	 r200EmitWait( rmesa, RADEON_WAIT_3D );
-
-	 /* Blit to framebuffer
-	  */
-	 r200EmitBlit( rmesa,
-		       blit_format,
-		       dstPitch, GET_START( &region ),
-		       dstPitch | (t->tile_bits >> 16),
-		       t->bufAddr,
-		       0, 0,
-		       0, done,
-		       width, lines );
-	 
-	 r200EmitWait( rmesa, RADEON_WAIT_2D );
-
-	 r200ReleaseDmaRegion( rmesa, &region, __FUNCTION__ );
-	 done += lines;
-      }
-   }
-}
-
-
-/**
- * Upload the texture image associated with texture \a t at the specified
- * level at the address relative to \a start.
- */
-static void uploadSubImage( r200ContextPtr rmesa, r200TexObjPtr t, 
-			    GLint hwlevel,
-			    GLint x, GLint y, GLint width, GLint height,
-			    GLuint face )
-{
-   struct gl_texture_image *texImage = NULL;
-   GLuint offset;
-   GLint imageWidth, imageHeight;
-   GLint ret;
-   drm_radeon_texture_t tex;
-   drm_radeon_tex_image_t tmp;
-   const int level = hwlevel + t->base.firstLevel;
-
-   if ( R200_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n", 
-	       __FUNCTION__, (void *)t, (void *)t->base.tObj,
-	       level, width, height, face );
-   }
-
-   ASSERT(face < 6);
-
-   /* Ensure we have a valid texture to upload */
-   if ( ( hwlevel < 0 ) || ( hwlevel >= RADEON_MAX_TEXTURE_LEVELS ) ) {
-      _mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
-      return;
-   }
-
-   texImage = t->base.tObj->Image[face][level];
-
-   if ( !texImage ) {
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: texImage %d is NULL!\n", __FUNCTION__, level );
-      return;
-   }
-   if ( !texImage->Data ) {
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is NULL!\n", __FUNCTION__ );
-      return;
-   }
-
-
-   if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-      assert(level == 0);
-      assert(hwlevel == 0);
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is rectangular\n", __FUNCTION__);
-      r200UploadRectSubImage( rmesa, t, texImage, x, y, width, height );
-      return;
-   }
-   else if (texImage->IsClientData) {
-      if ( R200_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is in GART client storage\n",
-		  __FUNCTION__);
-      r200UploadGARTClientSubImage( rmesa, t, texImage, hwlevel,
-				   x, y, width, height );
-      return;
-   }
-   else if ( R200_DEBUG & DEBUG_TEXTURE )
-      fprintf( stderr, "%s: image data is in normal memory\n",
-	       __FUNCTION__);
-      
-
-   imageWidth = texImage->Width;
-   imageHeight = texImage->Height;
-
-   offset = t->bufAddr + t->base.totalSize / 6 * face;
-
-   if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      GLint imageX = 0;
-      GLint imageY = 0;
-      GLint blitX = t->image[face][hwlevel].x;
-      GLint blitY = t->image[face][hwlevel].y;
-      GLint blitWidth = t->image[face][hwlevel].width;
-      GLint blitHeight = t->image[face][hwlevel].height;
-      fprintf( stderr, "   upload image: %d,%d at %d,%d\n",
-	       imageWidth, imageHeight, imageX, imageY );
-      fprintf( stderr, "   upload  blit: %d,%d at %d,%d\n",
-	       blitWidth, blitHeight, blitX, blitY );
-      fprintf( stderr, "       blit ofs: 0x%07x level: %d/%d\n",
-	       (GLuint)offset, hwlevel, level );
-   }
-
-   t->image[face][hwlevel].data = texImage->Data;
-
-   /* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
-    * NOTE: we're always use a 1KB-wide blit and I8 texture format.
-    * We used to use 1, 2 and 4-byte texels and used to use the texture
-    * width to dictate the blit width - but that won't work for compressed
-    * textures. (Brian)
-    * NOTE: can't do that with texture tiling. (sroland)
-    */
-   tex.offset = offset;
-   tex.image = &tmp;
-   /* copy (x,y,width,height,data) */
-   memcpy( &tmp, &t->image[face][hwlevel], sizeof(tmp) );
-   
-   if (texImage->TexFormat->TexelBytes) {
-      /* use multi-byte upload scheme */
-      tex.height = imageHeight;
-      tex.width = imageWidth;
-      tex.format = t->pp_txformat & R200_TXFORMAT_FORMAT_MASK;
-      if (tex.format == R200_TXFORMAT_ABGR8888) {
-	 /* drm will refuse abgr8888 textures. */
-	 tex.format = R200_TXFORMAT_ARGB8888;
-      }
-      tex.pitch = MAX2((texImage->Width * texImage->TexFormat->TexelBytes) / 64, 1);
-      tex.offset += tmp.x & ~1023;
-      tmp.x = tmp.x % 1024;
-      if (t->tile_bits & R200_TXO_MICRO_TILE) {
-	 /* need something like "tiled coordinates" ? */
-	 tmp.y = tmp.x / (tex.pitch * 128) * 2;
-	 tmp.x = tmp.x % (tex.pitch * 128) / 2 / texImage->TexFormat->TexelBytes;
-	 tex.pitch |= RADEON_DST_TILE_MICRO >> 22;
-      }
-      else {
-	 tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1);
-      }
-      if ((t->tile_bits & R200_TXO_MACRO_TILE) &&
-	 (texImage->Width * texImage->TexFormat->TexelBytes >= 256) &&
-	 ((!(t->tile_bits & R200_TXO_MICRO_TILE) && (texImage->Height >= 8)) ||
-	    (texImage->Height >= 16))) {
-	 /* weird: R200 disables macro tiling if mip width is smaller than 256 bytes,
-	    OR if height is smaller than 8 automatically, but if micro tiling is active
-	    the limit is height 16 instead ? */
-	 tex.pitch |= RADEON_DST_TILE_MACRO >> 22;
-      }
-   }
-   else {
-      /* In case of for instance 8x8 texture (2x2 dxt blocks), padding after the first two blocks is
-         needed (only with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */
-      /* set tex.height to 1/4 since 1 "macropixel" (dxt-block) has 4 real pixels. Needed
-         so the kernel module reads the right amount of data. */
-      tex.format = R200_TXFORMAT_I8; /* any 1-byte texel format */
-      tex.pitch = (BLIT_WIDTH_BYTES / 64);
-      tex.height = (imageHeight + 3) / 4;
-      tex.width = (imageWidth + 3) / 4;
-      switch (t->pp_txformat & R200_TXFORMAT_FORMAT_MASK) {
-      case R200_TXFORMAT_DXT1:
-           tex.width *= 8;
-           break;
-      case R200_TXFORMAT_DXT23:
-      case R200_TXFORMAT_DXT45:
-           tex.width *= 16;
-           break;
-      default:
-          fprintf(stderr, "unknown compressed tex format in uploadSubImage\n");
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
-   do {
-      ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_TEXTURE,
-                                 &tex, sizeof(drm_radeon_texture_t) );
-      if (ret) {
-	 if (R200_DEBUG & DEBUG_IOCTL)
-	    fprintf(stderr, "DRM_RADEON_TEXTURE:  again!\n");
-	 usleep(1);
-      }
-   } while ( ret == -EAGAIN );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret );
-      fprintf( stderr, "   offset=0x%08x\n",
-	       offset );
-      fprintf( stderr, "   image width=%d height=%d\n",
-	       imageWidth, imageHeight );
-      fprintf( stderr, "    blit width=%d height=%d data=%p\n",
-	       t->image[face][hwlevel].width, t->image[face][hwlevel].height,
-	       t->image[face][hwlevel].data );
-      exit( 1 );
-   }
-}
-
-
-/**
- * Upload the texture images associated with texture \a t.  This might
- * require the allocation of texture memory.
- * 
- * \param rmesa Context pointer
- * \param t Texture to be uploaded
- * \param face Cube map face to be uploaded.  Zero for non-cube maps.
- */
-
-int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face )
-{
-   const int numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      fprintf( stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
-	       (void *)rmesa->glCtx, (void *)t->base.tObj, t->base.totalSize,
-	       t->base.firstLevel, t->base.lastLevel );
-   }
-
-   if ( !t || t->base.totalSize == 0 || t->image_override )
-      return 0;
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      r200Finish( rmesa->glCtx );
-   }
-
-   LOCK_HARDWARE( rmesa );
-
-   if ( t->base.memBlock == NULL ) {
-      int heap;
-
-      heap = driAllocateTexture( rmesa->texture_heaps, rmesa->nr_heaps,
-				 (driTextureObject *) t );
-      if ( heap == -1 ) {
-	 UNLOCK_HARDWARE( rmesa );
-	 return -1;
-      }
-
-      /* Set the base offset of the texture image */
-      t->bufAddr = rmesa->r200Screen->texOffset[heap] 
-	   + t->base.memBlock->ofs;
-      t->pp_txoffset = t->bufAddr;
-       
-      if (!(t->base.tObj->Image[0][0]->IsClientData)) {
-	 /* hope it's safe to add that here... */
-	 t->pp_txoffset |= t->tile_bits;
-      }
-
-      /* Mark this texobj as dirty on all units:
-       */
-      t->dirty_state = TEX_ALL;
-   }
-
-   /* Let the world know we've used this memory recently.
-    */
-   driUpdateTextureLRU( (driTextureObject *) t );
-   UNLOCK_HARDWARE( rmesa );
-
-   /* Upload any images that are new */
-   if (t->base.dirty_images[face]) {
-      int i;
-      for ( i = 0 ; i < numLevels ; i++ ) {
-         if ( (t->base.dirty_images[face] & (1 << (i+t->base.firstLevel))) != 0 ) {
-            uploadSubImage( rmesa, t, i, 0, 0, t->image[face][i].width,
-			    t->image[face][i].height, face );
-         }
-      }
-      t->base.dirty_images[face] = 0;
-   }
-
-
-   if (R200_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      r200Finish( rmesa->glCtx );
-   }
-
-   return 0;
-}
diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c
index 0ad5651cd4..4e53672aee 100644
--- a/src/mesa/drivers/dri/r200/r200_texstate.c
+++ b/src/mesa/drivers/dri/r200/r200_texstate.c
@@ -37,9 +37,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "main/macros.h"
 #include "main/texformat.h"
+#include "main/teximage.h"
 #include "main/texobj.h"
 #include "main/enums.h"
 
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
 #include "r200_context.h"
 #include "r200_state.h"
 #include "r200_ioctl.h"
@@ -139,257 +142,6 @@ static const struct tx_table tx_table_le[] =
 #undef _ALPHA
 #undef _INVALID
 
-/**
- * This function computes the number of bytes of storage needed for
- * the given texture object (all mipmap levels, all cube faces).
- * The \c image[face][level].x/y/width/height parameters for upload/blitting
- * are computed here.  \c pp_txfilter, \c pp_txformat, etc. will be set here
- * too.
- * 
- * \param rmesa Context pointer
- * \param tObj GL texture object whose images are to be posted to
- *                 hardware state.
- */
-static void r200SetTexImages( r200ContextPtr rmesa,
-			      struct gl_texture_object *tObj )
-{
-   r200TexObjPtr t = (r200TexObjPtr)tObj->DriverData;
-   const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel];
-   GLint curOffset, blitWidth;
-   GLint i, texelBytes;
-   GLint numLevels;
-   GLint log2Width, log2Height, log2Depth;
-
-   /* Set the hardware texture format
-    */
-   if ( !t->image_override ) {
-      if ( VALID_FORMAT( baseImage->TexFormat->MesaFormat ) ) {
-	 const struct tx_table *table = _mesa_little_endian() ? tx_table_le :
-								tx_table_be;
-
-         t->pp_txformat &= ~(R200_TXFORMAT_FORMAT_MASK |
-                             R200_TXFORMAT_ALPHA_IN_MAP);
-         t->pp_txfilter &= ~R200_YUV_TO_RGB;
-
-	 t->pp_txformat |= table[ baseImage->TexFormat->MesaFormat ].format;
-	 t->pp_txfilter |= table[ baseImage->TexFormat->MesaFormat ].filter;
-      }
-      else {
-         _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__);
-         return;
-      }
-   }
-
-   texelBytes = baseImage->TexFormat->TexelBytes;
-
-   /* Compute which mipmap levels we really want to send to the hardware.
-    */
-
-   driCalculateTextureFirstLastLevel( (driTextureObject *) t );
-   log2Width  = tObj->Image[0][t->base.firstLevel]->WidthLog2;
-   log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2;
-   log2Depth  = tObj->Image[0][t->base.firstLevel]->DepthLog2;
-
-   numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   assert(numLevels <= RADEON_MAX_TEXTURE_LEVELS);
-
-   /* Calculate mipmap offsets and dimensions for blitting (uploading)
-    * The idea is that we lay out the mipmap levels within a block of
-    * memory organized as a rectangle of width BLIT_WIDTH_BYTES.
-    */
-   curOffset = 0;
-   blitWidth = BLIT_WIDTH_BYTES;
-   t->tile_bits = 0;
-
-   /* figure out if this texture is suitable for tiling. */
-   if (texelBytes) {
-      if (rmesa->texmicrotile  && (tObj->Target != GL_TEXTURE_RECTANGLE_NV) &&
-      /* texrect might be able to use micro tiling too in theory? */
-	 (baseImage->Height > 1)) {
-	 /* allow 32 (bytes) x 1 mip (which will use two times the space
-	 the non-tiled version would use) max if base texture is large enough */
-	 if ((numLevels == 1) ||
-	   (((baseImage->Width * texelBytes / baseImage->Height) <= 32) &&
-	       (baseImage->Width * texelBytes > 64)) ||
-	    ((baseImage->Width * texelBytes / baseImage->Height) <= 16)) {
-	    t->tile_bits |= R200_TXO_MICRO_TILE;
-	 }
-      }
-      if (tObj->Target != GL_TEXTURE_RECTANGLE_NV) {
-	 /* we can set macro tiling even for small textures, they will be untiled anyway */
-	 t->tile_bits |= R200_TXO_MACRO_TILE;
-      }
-   }
-
-   for (i = 0; i < numLevels; i++) {
-      const struct gl_texture_image *texImage;
-      GLuint size;
-
-      texImage = tObj->Image[0][i + t->base.firstLevel];
-      if ( !texImage )
-	 break;
-
-      /* find image size in bytes */
-      if (texImage->IsCompressed) {
-      /* need to calculate the size AFTER padding even though the texture is
-         submitted without padding.
-         Only handle pot textures currently - don't know if npot is even possible,
-         size calculation would certainly need (trivial) adjustments.
-         Align (and later pad) to 32byte, not sure what that 64byte blit width is
-         good for? */
-         if ((t->pp_txformat & R200_TXFORMAT_FORMAT_MASK) == R200_TXFORMAT_DXT1) {
-            /* RGB_DXT1/RGBA_DXT1, 8 bytes per block */
-            if ((texImage->Width + 3) < 8) /* width one block */
-               size = texImage->CompressedSize * 4;
-            else if ((texImage->Width + 3) < 16)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-         }
-         else /* DXT3/5, 16 bytes per block */
-            if ((texImage->Width + 3) < 8)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-      }
-      else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-	 size = ((texImage->Width * texelBytes + 63) & ~63) * texImage->Height;
-      }
-      else if (t->tile_bits & R200_TXO_MICRO_TILE) {
-	 /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
-	    though the actual offset may be different (if texture is less than
-	    32 bytes width) to the untiled case */
-	 int w = (texImage->Width * texelBytes * 2 + 31) & ~31;
-	 size = (w * ((texImage->Height + 1) / 2)) * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      else {
-	 int w = (texImage->Width * texelBytes + 31) & ~31;
-	 size = w * texImage->Height * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      assert(size > 0);
-
-      /* Align to 32-byte offset.  It is faster to do this unconditionally
-       * (no branch penalty).
-       */
-
-      curOffset = (curOffset + 0x1f) & ~0x1f;
-
-      if (texelBytes) {
-	 t->image[0][i].x = curOffset; /* fix x and y coords up later together with offset */
-	 t->image[0][i].y = 0;
-	 t->image[0][i].width = MIN2(size / texelBytes, blitWidth);
-	 t->image[0][i].height = (size / texelBytes) / t->image[0][i].width;
-      }
-      else {
-         t->image[0][i].x = curOffset % BLIT_WIDTH_BYTES;
-         t->image[0][i].y = curOffset / BLIT_WIDTH_BYTES;
-         t->image[0][i].width  = MIN2(size, BLIT_WIDTH_BYTES);
-         t->image[0][i].height = size / t->image[0][i].width;     
-      }
-
-#if 0
-      /* for debugging only and only  applicable to non-rectangle targets */
-      assert(size % t->image[0][i].width == 0);
-      assert(t->image[0][i].x == 0
-             || (size < BLIT_WIDTH_BYTES && t->image[0][i].height == 1));
-#endif
-
-      if (0)
-         fprintf(stderr,
-                 "level %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n",
-                 i, texImage->Width, texImage->Height,
-                 t->image[0][i].x, t->image[0][i].y,
-                 t->image[0][i].width, t->image[0][i].height, size, curOffset);
-
-      curOffset += size;
-
-   }
-
-   /* Align the total size of texture memory block.
-    */
-   t->base.totalSize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
-
-   /* Setup remaining cube face blits, if needed */
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      const GLuint faceSize = t->base.totalSize;
-      GLuint face;
-      /* reuse face 0 x/y/width/height - just update the offset when uploading */
-      for (face = 1; face < 6; face++) {
-         for (i = 0; i < numLevels; i++) {
-            t->image[face][i].x =  t->image[0][i].x;
-            t->image[face][i].y =  t->image[0][i].y;
-            t->image[face][i].width  = t->image[0][i].width;
-            t->image[face][i].height = t->image[0][i].height;
-         }
-      }
-      t->base.totalSize = 6 * faceSize; /* total texmem needed */
-   }
-
-
-   /* Hardware state:
-    */
-   t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK;
-   t->pp_txfilter |= (numLevels - 1) << R200_MAX_MIP_LEVEL_SHIFT;
-
-   t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK |
-		       R200_TXFORMAT_HEIGHT_MASK |
-                       R200_TXFORMAT_CUBIC_MAP_ENABLE |
-                       R200_TXFORMAT_F5_WIDTH_MASK |
-                       R200_TXFORMAT_F5_HEIGHT_MASK);
-   t->pp_txformat |= ((log2Width << R200_TXFORMAT_WIDTH_SHIFT) |
-		      (log2Height << R200_TXFORMAT_HEIGHT_SHIFT));
-
-   t->pp_txformat_x &= ~(R200_DEPTH_LOG2_MASK | R200_TEXCOORD_MASK);
-   if (tObj->Target == GL_TEXTURE_3D) {
-      t->pp_txformat_x |= (log2Depth << R200_DEPTH_LOG2_SHIFT);
-      t->pp_txformat_x |= R200_TEXCOORD_VOLUME;
-   }
-   else if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      ASSERT(log2Width == log2Height);
-      t->pp_txformat |= ((log2Width << R200_TXFORMAT_F5_WIDTH_SHIFT) |
-                         (log2Height << R200_TXFORMAT_F5_HEIGHT_SHIFT) |
-/* don't think we need this bit, if it exists at all - fglrx does not set it */
-                         (R200_TXFORMAT_CUBIC_MAP_ENABLE));
-      t->pp_txformat_x |= R200_TEXCOORD_CUBIC_ENV;
-      t->pp_cubic_faces = ((log2Width << R200_FACE_WIDTH_1_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_1_SHIFT) |
-                           (log2Width << R200_FACE_WIDTH_2_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_2_SHIFT) |
-                           (log2Width << R200_FACE_WIDTH_3_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_3_SHIFT) |
-                           (log2Width << R200_FACE_WIDTH_4_SHIFT) |
-                           (log2Height << R200_FACE_HEIGHT_4_SHIFT));
-   }
-   else {
-      /* If we don't in fact send enough texture coordinates, q will be 1,
-       * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
-       */
-      t->pp_txformat_x |= R200_TEXCOORD_PROJ;
-   }
-
-   t->pp_txsize = (((tObj->Image[0][t->base.firstLevel]->Width - 1) << 0) |
-                   ((tObj->Image[0][t->base.firstLevel]->Height - 1) << 16));
-
-   /* Only need to round to nearest 32 for textures, but the blitter
-    * requires 64-byte aligned pitches, and we may/may not need the
-    * blitter.   NPOT only!
-    */
-   if ( !t->image_override ) {
-      if (baseImage->IsCompressed)
-         t->pp_txpitch = (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63);
-      else
-         t->pp_txpitch = ((tObj->Image[0][t->base.firstLevel]->Width * texelBytes) + 63) & ~(63);
-      t->pp_txpitch -= 32;
-   }
-
-   t->dirty_state = TEX_ALL;
-
-   /* FYI: r200UploadTexImages( rmesa, t ) used to be called here */
-}
-
-
-
 /* ================================================================
  * Texture combine functions
  */
@@ -981,20 +733,19 @@ void r200SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 {
 	r200ContextPtr rmesa = pDRICtx->driverPrivate;
 	struct gl_texture_object *tObj =
-	    _mesa_lookup_texture(rmesa->glCtx, texname);
-	r200TexObjPtr t;
+	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
 
 	if (!tObj)
 		return;
 
-	t = (r200TexObjPtr) tObj->DriverData;
-
 	t->image_override = GL_TRUE;
 
 	if (!offset)
 		return;
 
-	t->pp_txoffset = offset;
+	t->bo = NULL;
+	t->override_offset = offset;
 	t->pp_txpitch = pitch - 32;
 
 	switch (depth) {
@@ -1014,6 +765,125 @@ void r200SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 	}
 }
 
+void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+		       __DRIdrawable *dPriv)
+{
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	r200ContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+	
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->tile_bits = 0;
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	t->pp_txpitch &= (1 << 13) -1;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = tx_table_le[MESA_FORMAT_RGB888].format;
+		else
+			t->pp_txformat = tx_table_le[MESA_FORMAT_ARGB8888].format;
+		t->pp_txfilter |= tx_table_le[MESA_FORMAT_ARGB8888].filter;
+		break;
+	case 3:
+	default:
+		t->pp_txformat = tx_table_le[MESA_FORMAT_RGB888].format;
+		t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB888].filter;
+		break;
+	case 2:
+		t->pp_txformat = tx_table_le[MESA_FORMAT_RGB565].format;
+		t->pp_txfilter |= tx_table_le[MESA_FORMAT_RGB565].filter;
+		break;
+	}
+        t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT);
+        t->pp_txformat |= R200_TXFORMAT_NON_POWER2;
+	t->pp_txpitch = pitch_val;
+        t->pp_txpitch -= 32;
+
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
+}
+
+
+void r200SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+        r200SetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
+
+
 #define REF_COLOR 1
 #define REF_ALPHA 2
 
@@ -1207,12 +1077,41 @@ static GLboolean r200UpdateAllTexEnv( GLcontext *ctx )
                                 R200_VOLUME_FILTER_MASK)
 
 
+static void disable_tex_obj_state( r200ContextPtr rmesa, 
+				   int unit )
+{
+   
+   R200_STATECHANGE( rmesa, vtx );
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
+
+   if (rmesa->radeon.TclFallback & (R200_TCL_FALLBACK_TEXGEN_0<<unit)) {
+      TCL_FALLBACK( rmesa->radeon.glCtx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
+   }
+
+   /* Actually want to keep all units less than max active texture
+    * enabled, right?  Fix this for >2 texunits.
+    */
+
+   {
+      GLuint tmp = rmesa->TexGenEnabled;
+
+      rmesa->TexGenEnabled &= ~(R200_TEXGEN_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenEnabled &= ~(R200_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenNeedNormals[unit] = GL_FALSE;
+      rmesa->TexGenCompSel &= ~(R200_OUTPUT_TEX_0 << unit);
+
+      if (tmp != rmesa->TexGenEnabled) {
+	 rmesa->recheck_texgen[unit] = GL_TRUE;
+	 rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
+      }
+   }
+}
 static void import_tex_obj_state( r200ContextPtr rmesa,
 				  int unit,
-				  r200TexObjPtr texobj )
+				  radeonTexObjPtr texobj )
 {
 /* do not use RADEON_DB_STATE to avoid stale texture caches */
-   int *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
+   GLuint *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
 
    R200_STATECHANGE( rmesa, tex[unit] );
 
@@ -1225,36 +1124,21 @@ static void import_tex_obj_state( r200ContextPtr rmesa,
    cmd[TEX_PP_TXSIZE] = texobj->pp_txsize; /* NPOT only! */
    cmd[TEX_PP_TXPITCH] = texobj->pp_txpitch; /* NPOT only! */
    cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color;
-   if (rmesa->r200Screen->drmSupportsFragShader) {
-      cmd[TEX_PP_TXOFFSET_NEWDRM] = texobj->pp_txoffset;
-   }
-   else {
-      cmd[TEX_PP_TXOFFSET_OLDDRM] = texobj->pp_txoffset;
-   }
 
-   if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      int *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
-      GLuint bytesPerFace = texobj->base.totalSize / 6;
-      ASSERT(texobj->base.totalSize % 6 == 0);
+   if (texobj->base.Target == GL_TEXTURE_CUBE_MAP) {
+      GLuint *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
 
       R200_STATECHANGE( rmesa, cube[unit] );
       cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
-      if (rmesa->r200Screen->drmSupportsFragShader) {
+      if (rmesa->radeon.radeonScreen->drmSupportsFragShader) {
 	 /* that value is submitted twice. could change cube atom
 	    to not include that command when new drm is used */
 	 cmd[TEX_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
       }
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F1] = texobj->pp_txoffset + 1 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F2] = texobj->pp_txoffset + 2 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F3] = texobj->pp_txoffset + 3 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F4] = texobj->pp_txoffset + 4 * bytesPerFace;
-      cube_cmd[CUBE_PP_CUBIC_OFFSET_F5] = texobj->pp_txoffset + 5 * bytesPerFace;
    }
 
-   texobj->dirty_state &= ~(1<<unit);
 }
 
-
 static void set_texgen_matrix( r200ContextPtr rmesa, 
 			       GLuint unit,
 			       const GLfloat *s_plane,
@@ -1377,7 +1261,6 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
    } else {
       tgcm |= R200_TEXGEN_COMP_T << (unit * 4);
    }
-
    if (texUnit->TexGenEnabled & R_BIT) {
       if (texUnit->GenR.Mode != mode)
 	 mixed_fallback = GL_TRUE;
@@ -1517,52 +1400,6 @@ static GLboolean r200_validate_texgen( GLcontext *ctx, GLuint unit )
    return GL_TRUE;
 }
 
-
-static void disable_tex( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE<<unit)) {
-      /* Texture unit disabled */
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
-
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= ~(1UL << unit);
-	 rmesa->state.texture.unit[unit].texobj = NULL;
-      }
-
-      R200_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= ~(R200_TEX_0_ENABLE << unit);
-	 
-      R200_STATECHANGE( rmesa, vtx );
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
-	 
-      if (rmesa->TclFallback & (R200_TCL_FALLBACK_TEXGEN_0<<unit)) {
-	 TCL_FALLBACK( ctx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
-      }
-
-      /* Actually want to keep all units less than max active texture
-       * enabled, right?  Fix this for >2 texunits.
-       */
-
-      {
-	 GLuint tmp = rmesa->TexGenEnabled;
-
-	 rmesa->TexGenEnabled &= ~(R200_TEXGEN_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenEnabled &= ~(R200_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenNeedNormals[unit] = GL_FALSE;
-	 rmesa->TexGenCompSel &= ~(R200_OUTPUT_TEX_0 << unit);
-
-	 if (tmp != rmesa->TexGenEnabled) {
-	    rmesa->recheck_texgen[unit] = GL_TRUE;
-	    rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
-	 }
-      }
-   }
-}
-
 void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
@@ -1579,237 +1416,169 @@ void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d )
    }
 }
 
-static GLboolean enable_tex_2d( GLcontext *ctx, int unit )
+/**
+ * Compute the cached hardware register values for the given texture object.
+ *
+ * \param rmesa Context pointer
+ * \param t the r300 texture object
+ */
+static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t)
 {
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
-   }
-
-   ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D);
-
-   if ( t->base.dirty_images[0] ) {
-      R200_FIREVERTICES( rmesa );
-      r200SetTexImages( rmesa, tObj );
-      r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock && !t->image_override ) 
-	 return GL_FALSE;
+   int firstlevel = t->mt ? t->mt->firstLevel : 0;
+   const struct gl_texture_image *firstImage = t->base.Image[0][firstlevel];
+   GLint log2Width, log2Height, log2Depth, texelBytes;
+   
+   if ( t->bo ) {
+       return;
    }
 
-   set_re_cntl_d3d( ctx, unit, GL_FALSE );
+   log2Width  = firstImage->WidthLog2;
+   log2Height = firstImage->HeightLog2;
+   log2Depth  = firstImage->DepthLog2;
+   texelBytes = firstImage->TexFormat->TexelBytes;
 
-   return GL_TRUE;
-}
 
-#if ENABLE_HW_3D_TEXTURE
-static GLboolean enable_tex_3d( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   /* Need to load the 3d images associated with this unit.
-    */
-   if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
+   if (!t->image_override) {
+      if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) {
+	 const struct tx_table *table = _mesa_little_endian() ? tx_table_le :
+	    tx_table_be;
+	 
+	 t->pp_txformat &= ~(R200_TXFORMAT_FORMAT_MASK |
+			     R200_TXFORMAT_ALPHA_IN_MAP);
+	 t->pp_txfilter &= ~R200_YUV_TO_RGB;
+	 
+	 t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format;
+	 t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter;
+      } else {
+	 _mesa_problem(NULL, "unexpected texture format in %s",
+		       __FUNCTION__);
+	 return;
+      }
    }
+   
+   t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK;
+   t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << R200_MAX_MIP_LEVEL_SHIFT;
+	
+   t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK |
+		       R200_TXFORMAT_HEIGHT_MASK |
+		       R200_TXFORMAT_CUBIC_MAP_ENABLE |
+		       R200_TXFORMAT_F5_WIDTH_MASK |
+		       R200_TXFORMAT_F5_HEIGHT_MASK);
+   t->pp_txformat |= ((log2Width << R200_TXFORMAT_WIDTH_SHIFT) |
+		      (log2Height << R200_TXFORMAT_HEIGHT_SHIFT));
+   
+   t->tile_bits = 0;
+   
+   t->pp_txformat_x &= ~(R200_DEPTH_LOG2_MASK | R200_TEXCOORD_MASK);
+   if (t->base.Target == GL_TEXTURE_3D) {
+      t->pp_txformat_x |= (log2Depth << R200_DEPTH_LOG2_SHIFT);
+      t->pp_txformat_x |= R200_TEXCOORD_VOLUME;
 
-   ASSERT(tObj->Target == GL_TEXTURE_3D);
-
-   /* R100 & R200 do not support mipmaps for 3D textures.
-    */
-   if ( (tObj->MinFilter != GL_NEAREST) && (tObj->MinFilter != GL_LINEAR) ) {
-      return GL_FALSE;
    }
-
-   if ( t->base.dirty_images[0] ) {
-      R200_FIREVERTICES( rmesa );
-      r200SetTexImages( rmesa, tObj );
-      r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock ) 
-	 return GL_FALSE;
+   else if (t->base.Target == GL_TEXTURE_CUBE_MAP) {
+      ASSERT(log2Width == log2Height);
+      t->pp_txformat |= ((log2Width << R200_TXFORMAT_F5_WIDTH_SHIFT) |
+			 (log2Height << R200_TXFORMAT_F5_HEIGHT_SHIFT) |
+			 /* don't think we need this bit, if it exists at all - fglrx does not set it */
+			 (R200_TXFORMAT_CUBIC_MAP_ENABLE));
+      t->pp_txformat_x |= R200_TEXCOORD_CUBIC_ENV;
+      t->pp_cubic_faces = ((log2Width << R200_FACE_WIDTH_1_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_1_SHIFT) |
+                           (log2Width << R200_FACE_WIDTH_2_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_2_SHIFT) |
+                           (log2Width << R200_FACE_WIDTH_3_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_3_SHIFT) |
+                           (log2Width << R200_FACE_WIDTH_4_SHIFT) |
+                           (log2Height << R200_FACE_HEIGHT_4_SHIFT));
    }
-
-   set_re_cntl_d3d( ctx, unit, GL_TRUE );
-
-   return GL_TRUE;
-}
-#endif
-
-static GLboolean enable_tex_cube( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-   GLuint face;
-
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & R200_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~R200_TXFORMAT_NON_POWER2;
-      for (face = 0; face < 6; face++)
-         t->base.dirty_images[face] = ~0;
+   else {
+      /* If we don't in fact send enough texture coordinates, q will be 1,
+       * making TEXCOORD_PROJ act like TEXCOORD_NONPROJ (Right?)
+       */
+      t->pp_txformat_x |= R200_TEXCOORD_PROJ;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
+   t->pp_txsize = (((firstImage->Width - 1) << R200_PP_TX_WIDTHMASK_SHIFT)
+		   | ((firstImage->Height - 1) << R200_PP_TX_HEIGHTMASK_SHIFT));
 
-   if ( t->base.dirty_images[0] || t->base.dirty_images[1] ||
-        t->base.dirty_images[2] || t->base.dirty_images[3] ||
-        t->base.dirty_images[4] || t->base.dirty_images[5] ) {
-      /* flush */
-      R200_FIREVERTICES( rmesa );
-      /* layout memory space, once for all faces */
-      r200SetTexImages( rmesa, tObj );
-   }
-
-   /* upload (per face) */
-   for (face = 0; face < 6; face++) {
-      if (t->base.dirty_images[face]) {
-         r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, face );
-      }
-   }
-      
-   if ( !t->base.memBlock ) {
-      /* texmem alloc failed, use s/w fallback */
-      return GL_FALSE;
+   if ( !t->image_override ) {
+      if (firstImage->IsCompressed)
+         t->pp_txpitch = (firstImage->Width + 63) & ~(63);
+      else
+         t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63);
+      t->pp_txpitch -= 32;
    }
 
-   set_re_cntl_d3d( ctx, unit, GL_TRUE );
-
-   return GL_TRUE;
-}
-
-static GLboolean enable_tex_rect( GLcontext *ctx, int unit )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   if (!(t->pp_txformat & R200_TXFORMAT_NON_POWER2)) {
+   if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) {
       t->pp_txformat |= R200_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
-   }
-
-   ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV);
-
-   if ( t->base.dirty_images[0] ) {
-      R200_FIREVERTICES( rmesa );
-      r200SetTexImages( rmesa, tObj );
-      r200UploadTexImages( rmesa, (r200TexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock &&
-           !t->image_override &&
-           !rmesa->prefer_gart_client_texturing ) 
-	 return GL_FALSE;
    }
 
-   set_re_cntl_d3d( ctx, unit, GL_FALSE );
-
-   return GL_TRUE;
 }
 
-
-static GLboolean update_tex_common( GLcontext *ctx, int unit )
+static GLboolean r200_validate_texture(GLcontext *ctx, struct gl_texture_object *texObj, int unit)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   r200TexObjPtr t = (r200TexObjPtr) tObj->DriverData;
-
-   /* Fallback if there's a texture border */
-   if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 )
-       return GL_FALSE;
-
-   /* Update state if this is a different texture object to last
-    * time.
-    */
-   if ( rmesa->state.texture.unit[unit].texobj != t ) {
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
+   radeonTexObj *t = radeon_tex_obj(texObj);
 
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= 
-	     ~(1UL << unit);
-      }
-
-      rmesa->state.texture.unit[unit].texobj = t;
-      t->base.bound |= (1UL << unit);
-      t->dirty_state |= 1<<unit;
-      driUpdateTextureLRU( (driTextureObject *) t ); /* XXX: should be locked! */
-   }
-
-
-   /* Newly enabled?
-    */
-   if ( 1|| !(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (R200_TEX_0_ENABLE<<unit))) {
-      R200_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_0_ENABLE << unit;
-
-      R200_STATECHANGE( rmesa, vtx );
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
-      rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] |= 4 << (unit * 3);
+   if (!radeon_validate_texture_miptree(ctx, texObj))
+      return GL_FALSE;
 
-      rmesa->recheck_texgen[unit] = GL_TRUE;
-   }
+   r200_validate_texgen(ctx, unit);
+   /* Configure the hardware registers (more precisely, the cached version
+    * of the hardware registers). */
+   setup_hardware_state(rmesa, t);
+
+   if (texObj->Target == GL_TEXTURE_RECTANGLE_NV ||
+       texObj->Target == GL_TEXTURE_2D ||
+       texObj->Target == GL_TEXTURE_1D)
+      set_re_cntl_d3d( ctx, unit, GL_FALSE );
+   else
+      set_re_cntl_d3d( ctx, unit, GL_TRUE );
+   R200_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_0_ENABLE << unit;
+   
+   R200_STATECHANGE( rmesa, vtx );
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] &= ~(7 << (unit * 3));
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] |= 4 << (unit * 3);
 
-   if (t->dirty_state & (1<<unit)) {
-      import_tex_obj_state( rmesa, unit, t );
-   }
+   rmesa->recheck_texgen[unit] = GL_TRUE;
+   import_tex_obj_state( rmesa, unit, t );
 
    if (rmesa->recheck_texgen[unit]) {
       GLboolean fallback = !r200_validate_texgen( ctx, unit );
       TCL_FALLBACK( ctx, (R200_TCL_FALLBACK_TEXGEN_0<<unit), fallback);
       rmesa->recheck_texgen[unit] = 0;
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 
-   FALLBACK( rmesa, R200_FALLBACK_BORDER_MODE, t->border_fallback );
-   return !t->border_fallback;
-}
+   t->validated = GL_TRUE;
 
+   FALLBACK( rmesa, RADEON_FALLBACK_BORDER_MODE, t->border_fallback );
 
+   return !t->border_fallback;
+}
 
-static GLboolean r200UpdateTextureUnit( GLcontext *ctx, int unit )
+static GLboolean r200UpdateTextureUnit(GLcontext *ctx, int unit)
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLuint unitneeded = rmesa->state.texture.unit[unit].unitneeded;
 
-   if ( unitneeded & (TEXTURE_RECT_BIT) ) {
-      return (enable_tex_rect( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( unitneeded & (TEXTURE_1D_BIT | TEXTURE_2D_BIT) ) {
-      return (enable_tex_2d( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-#if ENABLE_HW_3D_TEXTURE
-   else if ( unitneeded & (TEXTURE_3D_BIT) ) {
-      return (enable_tex_3d( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-#endif
-   else if ( unitneeded & (TEXTURE_CUBE_BIT) ) {
-      return (enable_tex_cube( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( unitneeded ) {
-      return GL_FALSE;
-   }
-   else {
-      disable_tex( ctx, unit );
-      return GL_TRUE;
+   if (!unitneeded) {
+      /* disable the unit */
+     disable_tex_obj_state(rmesa, unit);
+     return GL_TRUE;
    }
+
+   if (!r200_validate_texture(ctx, ctx->Texture.Unit[unit]._Current, unit)) {
+    _mesa_warning(ctx,
+		  "failed to validate texture for unit %d.\n",
+		  unit);
+    rmesa->state.texture.unit[unit].texobj = NULL;
+    return GL_FALSE;
+  }
+
+   rmesa->state.texture.unit[unit].texobj = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
+  return GL_TRUE;
 }
 
 
@@ -1850,11 +1619,11 @@ void r200UpdateTextureState( GLcontext *ctx )
 
    FALLBACK( rmesa, R200_FALLBACK_TEXTURE, !ok );
 
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       r200ChooseVertexState( ctx );
 
 
-   if (rmesa->r200Screen->chip_family == CHIP_FAMILY_R200) {
+   if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200) {
 
       /*
        * T0 hang workaround -------------
@@ -1867,7 +1636,7 @@ void r200UpdateTextureState( GLcontext *ctx )
 	 R200_STATECHANGE(rmesa, tex[1]);
 	 rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= R200_TEX_1_ENABLE;
 	 if (!(rmesa->hw.cst.cmd[CST_PP_CNTL_X] & R200_PPX_TEX_1_ENABLE))
-	    rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
+	   rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
 	 rmesa->hw.tex[1].cmd[TEX_PP_TXFORMAT] |= R200_TXFORMAT_LOOKUP_DISABLE;
       }
       else if (!ctx->ATIFragmentShader._Enabled) {
diff --git a/src/mesa/drivers/dri/r200/r200_vertprog.c b/src/mesa/drivers/dri/r200/r200_vertprog.c
index 4ce93b5145..620f29b5c6 100644
--- a/src/mesa/drivers/dri/r200/r200_vertprog.c
+++ b/src/mesa/drivers/dri/r200/r200_vertprog.c
@@ -1110,9 +1110,9 @@ void r200SetupVertexProg( GLcontext *ctx ) {
    }
    /* could optimize setting up vertex progs away for non-tcl hw */
    fallback = !(vp->native && r200VertexProgUpdateParams(ctx, vp) &&
-      rmesa->r200Screen->drmSupportsVertexProgram);
+      rmesa->radeon.radeonScreen->drmSupportsVertexProgram);
    TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, fallback);
-   if (rmesa->TclFallback) return;
+   if (rmesa->radeon.TclFallback) return;
 
    R200_STATECHANGE( rmesa, vap );
    /* FIXME: fglrx sets R200_VAP_SINGLE_BUF_STATE_ENABLE too. Do we need it?
diff --git a/src/mesa/drivers/dri/r300/.gitignore b/src/mesa/drivers/dri/r300/.gitignore
index 3689a6a78e..2f9cd1a987 100644
--- a/src/mesa/drivers/dri/r300/.gitignore
+++ b/src/mesa/drivers/dri/r300/.gitignore
@@ -1,4 +1,15 @@
+radeon_bocs_wrapper.h
+radeon_bo_legacy.[ch]
 radeon_chipset.h
+radeon_cmdbuf.h
+radeon_common.[ch]
+radeon_common_context.[ch]
+radeon_cs_legacy.[ch]
+radeon_dma.[ch]
+radeon_fbo.c
+radeon_lock.[ch]
+radeon_mipmap_tree.[ch]
 radeon_screen.[ch]
-radeon_span.h
+radeon_span.[ch]
+radeon_texture.[ch]
 server
diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 6ca934204f..a77209074a 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -3,6 +3,8 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
+CFLAGS += $(RADEON_CFLAGS)
+
 LIBNAME = r300_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c
@@ -11,6 +13,10 @@ ifeq ($(USING_EGL), 1)
 EGL_SOURCES = server/radeon_egl.c
 endif
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
 COMMON_SOURCES = \
 	../../common/driverfuncs.c \
 	../common/mm.c \
@@ -20,20 +26,26 @@ COMMON_SOURCES = \
 	../common/xmlconfig.c \
 	../common/dri_util.c
 
+RADEON_COMMON_SOURCES = \
+	radeon_texture.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_dma.c \
+	radeon_lock.c \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_mipmap_tree.c \
+	radeon_span.c \
+	radeon_fbo.c
+
 DRIVER_SOURCES = \
 		 radeon_screen.c \
-		 radeon_context.c \
-		 radeon_ioctl.c \
-		 radeon_lock.c \
-		 radeon_span.c \
-		 radeon_state.c \
-		 r300_mem.c \
 		 r300_context.c \
+		 r300_draw.c \
 		 r300_ioctl.c \
 		 r300_cmdbuf.c \
 		 r300_state.c \
 		 r300_render.c \
-		 r300_texmem.c \
 		 r300_tex.c \
 		 r300_texstate.c \
 		 radeon_program.c \
@@ -41,6 +53,7 @@ DRIVER_SOURCES = \
 		 radeon_program_pair.c \
 		 radeon_nqssadce.c \
 		 r300_vertprog.c \
+		 r300_fragprog_common.c \
 		 r300_fragprog.c \
 		 r300_fragprog_swizzle.c \
 		 r300_fragprog_emit.c \
@@ -49,12 +62,16 @@ DRIVER_SOURCES = \
 		 r300_shader.c \
 		 r300_emit.c \
 		 r300_swtcl.c \
-		 $(EGL_SOURCES)
+		 $(RADEON_COMMON_SOURCES) \
+		 $(EGL_SOURCES) \
+		 $(CS_SOURCES)
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
 DRIVER_DEFINES = -DCOMPILE_R300 -DR200_MERGED=0 \
-	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R300
+	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R300 \
+#	-DRADEON_BO_TRACK \
+	-Wall
 
 SYMLINKS = \
 	server/radeon_dri.c \
@@ -68,7 +85,30 @@ COMMON_SYMLINKS = \
 	radeon_chipset.h \
 	radeon_screen.c \
 	radeon_screen.h \
-	radeon_span.h
+	radeon_span.h \
+	radeon_span.c \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_bo_legacy.h \
+	radeon_cs_legacy.h \
+	radeon_bocs_wrapper.h \
+	radeon_lock.c \
+	radeon_lock.h \
+	radeon_common.c \
+	radeon_common.h \
+	radeon_common_context.c \
+	radeon_common_context.h \
+	radeon_cmdbuf.h \
+	radeon_dma.c \
+	radeon_dma.h \
+	radeon_mipmap_tree.c \
+	radeon_mipmap_tree.h \
+	radeon_texture.c \
+	radeon_texture.h \
+	radeon_fbo.c \
+	$(CS_SOURCES)
+
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
 ##### TARGETS #####
 
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index f4472756f1..af535037d0 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -44,245 +44,438 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drm.h"
 #include "radeon_drm.h"
 
-#include "radeon_ioctl.h"
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "radeon_reg.h"
 #include "r300_reg.h"
 #include "r300_cmdbuf.h"
 #include "r300_emit.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_mipmap_tree.h"
 #include "r300_state.h"
+#include "radeon_reg.h"
 
-// Set this to 1 for extremely verbose debugging of command buffers
-#define DEBUG_CMDBUF		0
-
-/**
- * Send the current command buffer via ioctl to the hardware.
+/** # of dwords reserved for additional instructions that may need to be written
+ * during flushing.
  */
-int r300FlushCmdBufLocked(r300ContextPtr r300, const char *caller)
+#define SPACE_FOR_FLUSHING	4
+
+static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt)
 {
-	int ret;
-	int i;
-	drm_radeon_cmd_buffer_t cmd;
-	int start;
-
-	if (r300->radeon.lost_context) {
-		start = 0;
-		r300->radeon.lost_context = GL_FALSE;
-	} else
-		start = r300->cmdbuf.count_reemit;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL) {
-		fprintf(stderr, "%s from %s - %i cliprects\n",
-			__FUNCTION__, caller, r300->radeon.numClipRects);
-
-		if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_VERBOSE)
-			for (i = start; i < r300->cmdbuf.count_used; ++i)
-				fprintf(stderr, "%d: %08x\n", i,
-					r300->cmdbuf.cmd_buf[i]);
-	}
+    if (r300->radeon.radeonScreen->kernel_mm) {
+        return ((((*pkt) >> 16) & 0x3FFF) + 1);
+    } else {
+        drm_r300_cmd_header_t *t = (drm_r300_cmd_header_t*)pkt;
+        return t->packet0.count;
+    }
+}
 
-	cmd.buf = (char *)(r300->cmdbuf.cmd_buf + start);
-	cmd.bufsz = (r300->cmdbuf.count_used - start) * 4;
+#define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count)
+#define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count)
 
-	if (r300->radeon.state.scissor.enabled) {
-		cmd.nbox = r300->radeon.state.scissor.numClipRects;
-		cmd.boxes =
-		    (drm_clip_rect_t *) r300->radeon.state.scissor.pClipRects;
-	} else {
-		cmd.nbox = r300->radeon.numClipRects;
-		cmd.boxes = (drm_clip_rect_t *) r300->radeon.pClipRects;
+void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	drm_r300_cmd_header_t cmd;
+	uint32_t addr, ndw, i;
+
+	if (!r300->radeon.radeonScreen->kernel_mm) {
+		uint32_t dwords;
+		dwords = (*atom->check) (ctx, atom);
+		BEGIN_BATCH_NO_AUTOSTATE(dwords);
+		OUT_BATCH_TABLE(atom->cmd, dwords);
+		END_BATCH();
+		return;
 	}
 
-	ret = drmCommandWrite(r300->radeon.dri.fd,
-			      DRM_RADEON_CMDBUF, &cmd, sizeof(cmd));
-
-	if (RADEON_DEBUG & DEBUG_SYNC) {
-		fprintf(stderr, "Syncing in %s (from %s)\n\n",
-			__FUNCTION__, caller);
-		radeonWaitForIdleLocked(&r300->radeon);
+	cmd.u = atom->cmd[0];
+	addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo;
+	ndw = cmd.vpu.count * 4;
+	if (ndw) {
+
+		if (r300->vap_flush_needed) {
+			BEGIN_BATCH_NO_AUTOSTATE(15 + ndw);
+
+			/* flush processing vertices */
+			OUT_BATCH_REGVAL(R300_SC_SCREENDOOR, 0);
+			OUT_BATCH_REGVAL(R300_RB3D_DSTCACHE_CTLSTAT, R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
+			OUT_BATCH_REGVAL(RADEON_WAIT_UNTIL, RADEON_WAIT_3D_IDLECLEAN);
+			OUT_BATCH_REGVAL(R300_SC_SCREENDOOR, 0xffffff);
+			OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0);
+			r300->vap_flush_needed = GL_FALSE;
+		} else {
+			BEGIN_BATCH_NO_AUTOSTATE(5 + ndw);
+		}
+		OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr);
+		OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR);
+		for (i = 0; i < ndw; i++) {
+			OUT_BATCH(atom->cmd[i+1]);
+		}
+		OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0);
+		END_BATCH();
 	}
-
-	r300->dma.nr_released_bufs = 0;
-	r300->cmdbuf.count_used = 0;
-	r300->cmdbuf.count_reemit = 0;
-
-	return ret;
 }
 
-int r300FlushCmdBuf(r300ContextPtr r300, const char *caller)
+void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom)
 {
-	int ret;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	drm_r300_cmd_header_t cmd;
+	uint32_t addr, ndw, i, sz;
+	int type, clamp, stride;
+
+	if (!r300->radeon.radeonScreen->kernel_mm) {
+		uint32_t dwords;
+		dwords = (*atom->check) (ctx, atom);
+		BEGIN_BATCH_NO_AUTOSTATE(dwords);
+		OUT_BATCH_TABLE(atom->cmd, dwords);
+		END_BATCH();
+		return;
+	}
 
-	LOCK_HARDWARE(&r300->radeon);
+	cmd.u = atom->cmd[0];
+	sz = cmd.r500fp.count;
+	addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo;
+	type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE);
+	clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP);
 
-	ret = r300FlushCmdBufLocked(r300, caller);
+	addr |= (type << 16);
+	addr |= (clamp << 17);
 
-	UNLOCK_HARDWARE(&r300->radeon);
+	stride = type ? 4 : 6;
 
-	if (ret) {
-		fprintf(stderr, "drmRadeonCmdBuffer: %d\n", ret);
-		_mesa_exit(ret);
-	}
+	ndw = sz * stride;
+	if (ndw) {
 
-	return ret;
+		BEGIN_BATCH_NO_AUTOSTATE(3 + ndw);
+		OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0));
+		OUT_BATCH(addr);
+		OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, ndw-1) | RADEON_ONE_REG_WR);
+		for (i = 0; i < ndw; i++) {
+			OUT_BATCH(atom->cmd[i+1]);
+		}
+		END_BATCH();
+	}
 }
 
-static void r300PrintStateAtom(r300ContextPtr r300, struct r300_state_atom *state)
+static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom)
 {
-	int i, j, reg;
-	int dwords = (*state->check) (r300, state);
-	drm_r300_cmd_header_t cmd;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	int numtmus = packet0_count(r300, r300->hw.tex.offset.cmd);
+	int notexture = 0;
 
-	fprintf(stderr, "  emit %s %d/%d\n", state->name, dwords,
-		state->cmd_size);
-
-	if (RADEON_DEBUG & DEBUG_VERBOSE) {
-		for (i = 0; i < dwords;) {
-			cmd = *((drm_r300_cmd_header_t *) &state->cmd[i]);
-			reg = (cmd.packet0.reghi << 8) | cmd.packet0.reglo;
-			fprintf(stderr, "      %s[%d]: cmdpacket0 (first reg=0x%04x, count=%d)\n",
-					state->name, i, reg, cmd.packet0.count);
-			++i;
-			for (j = 0; j < cmd.packet0.count; j++) {
-				fprintf(stderr, "      %s[%d]: 0x%04x = %08x\n",
-					state->name, i, reg, state->cmd[i]);
-				reg += 4;
-				++i;
-			}
+	if (numtmus) {
+		int i;
+
+		for(i = 0; i < numtmus; ++i) {
+		    radeonTexObj *t = r300->hw.textures[i];
+
+		    if (!t)
+			notexture = 1;
+		}
+
+		if (r300->radeon.radeonScreen->kernel_mm && notexture) {
+			return;
+		}
+		for(i = 0; i < numtmus; ++i) {
+		    radeonTexObj *t = r300->hw.textures[i];
+		    if (t && !t->image_override) {
+                BEGIN_BATCH_NO_AUTOSTATE(4);
+                OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+			    OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+					    RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+                END_BATCH();
+		    } else if (!t) {
+                /* Texture unit hasn't a texture bound nothings to do */
+		    } else { /* override cases */
+			    if (t->bo) {
+                    BEGIN_BATCH_NO_AUTOSTATE(4);
+                    OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+				    OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
+						    RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+                    END_BATCH();
+			    } else if (!r300->radeon.radeonScreen->kernel_mm) {
+                    BEGIN_BATCH_NO_AUTOSTATE(2);
+                    OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
+				    OUT_BATCH(t->override_offset);
+                    END_BATCH();
+			    } else {
+                    /* Texture unit hasn't a texture bound nothings to do */
+                }
+		    }
 		}
 	}
 }
 
-/**
- * Emit all atoms with a dirty field equal to dirty.
- *
- * The caller must have ensured that there is enough space in the command
- * buffer.
- */
-static INLINE void r300EmitAtoms(r300ContextPtr r300, GLboolean dirty)
+void r300_emit_scissor(GLcontext *ctx)
 {
-	struct r300_state_atom *atom;
-	uint32_t *dest;
-	int dwords;
-
-	dest = r300->cmdbuf.cmd_buf + r300->cmdbuf.count_used;
-
-	/* Emit WAIT */
-	*dest = cmdwait(R300_WAIT_3D | R300_WAIT_3D_CLEAN);
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	/* Emit cache flush */
-	*dest = cmdpacket0(R300_TX_INVALTAGS, 1);
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	*dest = R300_TX_FLUSH;
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	/* Emit END3D */
-	*dest = cmdpacify();
-	dest++;
-	r300->cmdbuf.count_used++;
-
-	/* Emit actual atoms */
-
-	foreach(atom, &r300->hw.atomlist) {
-		if ((atom->dirty || r300->hw.all_dirty) == dirty) {
-			dwords = (*atom->check) (r300, atom);
-			if (dwords) {
-				if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) {
-					r300PrintStateAtom(r300, atom);
-				}
-				memcpy(dest, atom->cmd, dwords * 4);
-				dest += dwords;
-				r300->cmdbuf.count_used += dwords;
-				atom->dirty = GL_FALSE;
-			} else {
-				if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) {
-					fprintf(stderr, "  skip state %s\n",
-						atom->name);
-				}
-			}
-		}
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    unsigned x1, y1, x2, y2;
+	struct radeon_renderbuffer *rrb;
+
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        return;
+    }
+	rrb = radeon_get_colorbuffer(&r300->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return;
 	}
+    if (r300->radeon.state.scissor.enabled) {
+        x1 = r300->radeon.state.scissor.rect.x1;
+        y1 = r300->radeon.state.scissor.rect.y1;
+        x2 = r300->radeon.state.scissor.rect.x2 - 1;
+        y2 = r300->radeon.state.scissor.rect.y2 - 1;
+    } else {
+        x1 = 0;
+        y1 = 0;
+        x2 = rrb->base.Width - 1;
+        y2 = rrb->base.Height - 1;
+    }
+    if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
+        x1 += R300_SCISSORS_OFFSET;
+        y1 += R300_SCISSORS_OFFSET;
+        x2 += R300_SCISSORS_OFFSET;
+        y2 += R300_SCISSORS_OFFSET;
+    }
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+    OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+    OUT_BATCH((x1 << R300_SCISSORS_X_SHIFT)|(y1 << R300_SCISSORS_Y_SHIFT));
+    OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT));
+    END_BATCH();
 }
 
-/**
- * Copy dirty hardware state atoms into the command buffer.
- *
- * We also copy out clean state if we're at the start of a buffer. That makes
- * it easy to recover from lost contexts.
- */
-void r300EmitState(r300ContextPtr r300)
+static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 {
-	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_PRIMS))
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (r300->cmdbuf.count_used && !r300->hw.is_dirty
-	    && !r300->hw.all_dirty)
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	struct radeon_renderbuffer *rrb;
+	uint32_t cbpitch;
+	uint32_t offset = r300->radeon.state.color.draw_offset;
+	uint32_t dw = 6;
+    int i;
+
+	rrb = radeon_get_colorbuffer(&r300->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
 		return;
+	}
+
+        if (RADEON_DEBUG & DEBUG_STATE)
+           fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height);
+	cbpitch = (rrb->pitch / rrb->cpp);
+	if (rrb->cpp == 4)
+		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
+	else switch (rrb->base._ActualFormat) {
+	case GL_RGB5:
+		cbpitch |= R300_COLOR_FORMAT_RGB565;
+		break;
+	case GL_RGBA4:
+		cbpitch |= R300_COLOR_FORMAT_ARGB4444;
+		break;
+	case GL_RGB5_A1:
+		cbpitch |= R300_COLOR_FORMAT_ARGB1555;
+		break;
+	}
 
-	/* To avoid going across the entire set of states multiple times, just check
-	 * for enough space for the case of emitting all state, and inline the
-	 * r300AllocCmdBuf code here without all the checks.
-	 */
-	r300EnsureCmdBufSpace(r300, r300->hw.max_state_size, __FUNCTION__);
+	if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+		cbpitch |= R300_COLOR_TILE_ENABLE;
+
+    	if (r300->radeon.radeonScreen->kernel_mm)
+		dw += 2;
+	BEGIN_BATCH_NO_AUTOSTATE(dw);
+	OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1);
+	OUT_BATCH_RELOC(offset, rrb->bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1);
+    	if (!r300->radeon.radeonScreen->kernel_mm)
+		OUT_BATCH(cbpitch);
+	else
+		OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	END_BATCH();
+    if (r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+            BEGIN_BATCH_NO_AUTOSTATE(3);
+            OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+            OUT_BATCH(0);
+            OUT_BATCH(((rrb->base.Width - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->base.Height - 1) << R300_SCISSORS_Y_SHIFT));
+            END_BATCH();
+            BEGIN_BATCH_NO_AUTOSTATE(16);
+            for (i = 0; i < 4; i++) {
+                OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
+                OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT));
+            }
+            OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
+            OUT_BATCH(0xAAAA);
+            OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1);
+            OUT_BATCH(0xffffff);
+            END_BATCH();
+        } else {
+            BEGIN_BATCH_NO_AUTOSTATE(3);
+            OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+            OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) |
+                    (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT));
+            OUT_BATCH(((rrb->base.Width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->base.Height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT));
+            END_BATCH();
+            BEGIN_BATCH_NO_AUTOSTATE(16);
+            for (i = 0; i < 4; i++) {
+                OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
+                OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) |
+                          ((R300_SCISSORS_OFFSET + rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT));
+            }
+            OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
+            OUT_BATCH(0xAAAA);
+            OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1);
+            OUT_BATCH(0xffffff);
+            END_BATCH();
+        }
+    }
+}
 
-	if (!r300->cmdbuf.count_used) {
-		if (RADEON_DEBUG & DEBUG_STATE)
-			fprintf(stderr, "Begin reemit state\n");
+static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	struct radeon_renderbuffer *rrb;
+	uint32_t zbpitch;
+	uint32_t dw;
+
+	rrb = radeon_get_depthbuffer(&r300->radeon);
+	if (!rrb)
+		return;
 
-		r300EmitAtoms(r300, GL_FALSE);
-		r300->cmdbuf.count_reemit = r300->cmdbuf.count_used;
+	zbpitch = (rrb->pitch / rrb->cpp);
+	if (!r300->radeon.radeonScreen->kernel_mm) {
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+	        zbpitch |= R300_DEPTHMACROTILE_ENABLE;
+	   }
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
+	        zbpitch |= R300_DEPTHMICROTILE_TILED;
+	    }
 	}
 
-	if (RADEON_DEBUG & DEBUG_STATE)
-		fprintf(stderr, "Begin dirty state\n");
+	dw = 6;
+    	if (r300->radeon.radeonScreen->kernel_mm)
+		dw += 2;
+	BEGIN_BATCH_NO_AUTOSTATE(dw);
+	OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1);
+	OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1);
+    	if (!r300->radeon.radeonScreen->kernel_mm)
+	    OUT_BATCH(zbpitch);
+	else
+	    OUT_BATCH_RELOC(cbpitch, rrb->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	END_BATCH();
+}
 
-	r300EmitAtoms(r300, GL_TRUE);
+static void emit_gb_misc(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        BEGIN_BATCH_NO_AUTOSTATE(4);
+        OUT_BATCH(atom->cmd[0]);
+        OUT_BATCH(atom->cmd[1]);
+        OUT_BATCH(atom->cmd[2]);
+        OUT_BATCH(atom->cmd[3]);
+        END_BATCH();
+    }
+}
 
-	assert(r300->cmdbuf.count_used < r300->cmdbuf.size);
+static void emit_threshold_misc(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+        BEGIN_BATCH_NO_AUTOSTATE(3);
+        OUT_BATCH(atom->cmd[0]);
+        OUT_BATCH(atom->cmd[1]);
+        OUT_BATCH(atom->cmd[2]);
+        END_BATCH();
+    }
+}
 
-	r300->hw.is_dirty = GL_FALSE;
-	r300->hw.all_dirty = GL_FALSE;
+static void emit_shade_misc(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        BEGIN_BATCH_NO_AUTOSTATE(2);
+        OUT_BATCH(atom->cmd[0]);
+        OUT_BATCH(atom->cmd[1]);
+        END_BATCH();
+    }
 }
 
-#define packet0_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->packet0.count)
-#define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count)
-#define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count)
+static void emit_zstencil_format(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+	struct radeon_renderbuffer *rrb;
+	uint32_t format = 0;
+
+	rrb = radeon_get_depthbuffer(&r300->radeon);
+	if (!rrb)
+	  format = 0;
+	else {
+	  if (rrb->cpp == 2)
+	    format = R300_DEPTHFORMAT_16BIT_INT_Z;
+	  else if (rrb->cpp == 4)
+	    format = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL;
+	}
+
+	OUT_BATCH(atom->cmd[0]);
+	atom->cmd[1] &= ~0xf;
+	atom->cmd[1] |= format;
+	OUT_BATCH(atom->cmd[1]);
+	OUT_BATCH(atom->cmd[2]);
+	OUT_BATCH(atom->cmd[3]);
+	OUT_BATCH(atom->cmd[4]);
+}
 
-static int check_always(r300ContextPtr r300, struct r300_state_atom *atom)
+static int check_always(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	return atom->cmd_size;
 }
 
-static int check_variable(r300ContextPtr r300, struct r300_state_atom *atom)
+static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int cnt;
-	cnt = packet0_count(atom->cmd);
+	if (atom->cmd[0] == CP_PACKET2) {
+		return 0;
+	}
+	cnt = packet0_count(r300, atom->cmd);
 	return cnt ? cnt + 1 : 0;
 }
 
-static int check_vpu(r300ContextPtr r300, struct r300_state_atom *atom)
+int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	int cnt;
+
 	cnt = vpu_count(atom->cmd);
 	return cnt ? (cnt * 4) + 1 : 0;
 }
 
-static int check_r500fp(r300ContextPtr r300, struct r300_state_atom *atom)
+int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	int cnt;
+
 	cnt = r500fp_count(atom->cmd);
 	return cnt ? (cnt * 6) + 1 : 0;
 }
 
-static int check_r500fp_const(r300ContextPtr r300, struct r300_state_atom *atom)
+int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	int cnt;
+
 	cnt = r500fp_count(atom->cmd);
 	return cnt ? (cnt * 4) + 1 : 0;
 }
@@ -295,8 +488,8 @@ static int check_r500fp_const(r300ContextPtr r300, struct r300_state_atom *atom)
       r300->hw.ATOM.idx = (IDX);					\
       r300->hw.ATOM.check = check_##CHK;				\
       r300->hw.ATOM.dirty = GL_FALSE;					\
-      r300->hw.max_state_size += (SZ);					\
-      insert_at_tail(&r300->hw.atomlist, &r300->hw.ATOM);		\
+      r300->radeon.hw.max_state_size += (SZ);					\
+      insert_at_tail(&r300->radeon.hw.atomlist, &r300->hw.ATOM);		\
    } while (0)
 /**
  * Allocate memory for the command buffer and initialize the state atom
@@ -304,18 +497,16 @@ static int check_r500fp_const(r300ContextPtr r300, struct r300_state_atom *atom)
  */
 void r300InitCmdBuf(r300ContextPtr r300)
 {
-	int size, mtu;
-	int has_tcl = 1;
+	int mtu;
+	int has_tcl;
 	int is_r500 = 0;
-	int i;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
 		is_r500 = 1;
 
-	r300->hw.max_state_size = 2 + 2;	/* reserve extra space for WAIT_IDLE and tex cache flush */
+	r300->radeon.hw.max_state_size = 2 + 2;	/* reserve extra space for WAIT_IDLE and tex cache flush */
 
 	mtu = r300->radeon.glCtx->Const.MaxTextureUnits;
 	if (RADEON_DEBUG & DEBUG_TEXTURE) {
@@ -323,232 +514,248 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	}
 
 	/* Setup the atom linked list */
-	make_empty_list(&r300->hw.atomlist);
-	r300->hw.atomlist.name = "atom-list";
+	make_empty_list(&r300->radeon.hw.atomlist);
+	r300->radeon.hw.atomlist.name = "atom-list";
 
 	/* Initialize state atoms */
 	ALLOC_STATE(vpt, always, R300_VPT_CMDSIZE, 0);
-	r300->hw.vpt.cmd[R300_VPT_CMD_0] = cmdpacket0(R300_SE_VPORT_XSCALE, 6);
+	r300->hw.vpt.cmd[R300_VPT_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_SE_VPORT_XSCALE, 6);
 	ALLOC_STATE(vap_cntl, always, R300_VAP_CNTL_SIZE, 0);
-	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH] = cmdpacket0(R300_VAP_PVS_STATE_FLUSH_REG, 1);
+	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PVS_STATE_FLUSH_REG, 1);
 	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_FLUSH_1] = 0;
-	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_CMD] = cmdpacket0(R300_VAP_CNTL, 1);
-	if (is_r500) {
+	r300->hw.vap_cntl.cmd[R300_VAP_CNTL_CMD] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CNTL, 1);
+	if (is_r500 && !r300->radeon.radeonScreen->kernel_mm) {
 	    ALLOC_STATE(vap_index_offset, always, 2, 0);
-	    r300->hw.vap_index_offset.cmd[0] = cmdpacket0(R500_VAP_INDEX_OFFSET, 1);
+	    r300->hw.vap_index_offset.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_VAP_INDEX_OFFSET, 1);
 	    r300->hw.vap_index_offset.cmd[1] = 0;
 	}
 	ALLOC_STATE(vte, always, 3, 0);
-	r300->hw.vte.cmd[0] = cmdpacket0(R300_SE_VTE_CNTL, 2);
+	r300->hw.vte.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SE_VTE_CNTL, 2);
 	ALLOC_STATE(vap_vf_max_vtx_indx, always, 3, 0);
-	r300->hw.vap_vf_max_vtx_indx.cmd[0] = cmdpacket0(R300_VAP_VF_MAX_VTX_INDX, 2);
+	r300->hw.vap_vf_max_vtx_indx.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_VF_MAX_VTX_INDX, 2);
 	ALLOC_STATE(vap_cntl_status, always, 2, 0);
-	r300->hw.vap_cntl_status.cmd[0] = cmdpacket0(R300_VAP_CNTL_STATUS, 1);
+	r300->hw.vap_cntl_status.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CNTL_STATUS, 1);
 	ALLOC_STATE(vir[0], variable, R300_VIR_CMDSIZE, 0);
 	r300->hw.vir[0].cmd[R300_VIR_CMD_0] =
-	    cmdpacket0(R300_VAP_PROG_STREAM_CNTL_0, 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PROG_STREAM_CNTL_0, 1);
 	ALLOC_STATE(vir[1], variable, R300_VIR_CMDSIZE, 1);
 	r300->hw.vir[1].cmd[R300_VIR_CMD_0] =
-	    cmdpacket0(R300_VAP_PROG_STREAM_CNTL_EXT_0, 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PROG_STREAM_CNTL_EXT_0, 1);
 	ALLOC_STATE(vic, always, R300_VIC_CMDSIZE, 0);
-	r300->hw.vic.cmd[R300_VIC_CMD_0] = cmdpacket0(R300_VAP_VTX_STATE_CNTL, 2);
+	r300->hw.vic.cmd[R300_VIC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_VTX_STATE_CNTL, 2);
 	ALLOC_STATE(vap_psc_sgn_norm_cntl, always, 2, 0);
-	r300->hw.vap_psc_sgn_norm_cntl.cmd[0] = cmdpacket0(R300_VAP_PSC_SGN_NORM_CNTL, SGN_NORM_ZERO_CLAMP_MINUS_ONE);
+	r300->hw.vap_psc_sgn_norm_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PSC_SGN_NORM_CNTL, SGN_NORM_ZERO_CLAMP_MINUS_ONE);
 
 	if (has_tcl) {
 		ALLOC_STATE(vap_clip_cntl, always, 2, 0);
-		r300->hw.vap_clip_cntl.cmd[0] = cmdpacket0(R300_VAP_CLIP_CNTL, 1);
+		r300->hw.vap_clip_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_CLIP_CNTL, 1);
 		ALLOC_STATE(vap_clip, always, 5, 0);
-		r300->hw.vap_clip.cmd[0] = cmdpacket0(R300_VAP_GB_VERT_CLIP_ADJ, 4);
+		r300->hw.vap_clip.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_VAP_GB_VERT_CLIP_ADJ, 4);
 		ALLOC_STATE(vap_pvs_vtx_timeout_reg, always, 2, 0);
-		r300->hw.vap_pvs_vtx_timeout_reg.cmd[0] = cmdpacket0(VAP_PVS_VTX_TIMEOUT_REG, 1);
+		r300->hw.vap_pvs_vtx_timeout_reg.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, VAP_PVS_VTX_TIMEOUT_REG, 1);
 	}
 
 	ALLOC_STATE(vof, always, R300_VOF_CMDSIZE, 0);
 	r300->hw.vof.cmd[R300_VOF_CMD_0] =
-	    cmdpacket0(R300_VAP_OUTPUT_VTX_FMT_0, 2);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_OUTPUT_VTX_FMT_0, 2);
 
 	if (has_tcl) {
 		ALLOC_STATE(pvs, always, R300_PVS_CMDSIZE, 0);
 		r300->hw.pvs.cmd[R300_PVS_CMD_0] =
-		    cmdpacket0(R300_VAP_PVS_CODE_CNTL_0, 3);
+		    cmdpacket0(r300->radeon.radeonScreen, R300_VAP_PVS_CODE_CNTL_0, 3);
 	}
 
 	ALLOC_STATE(gb_enable, always, 2, 0);
-	r300->hw.gb_enable.cmd[0] = cmdpacket0(R300_GB_ENABLE, 1);
+	r300->hw.gb_enable.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_ENABLE, 1);
 	ALLOC_STATE(gb_misc, always, R300_GB_MISC_CMDSIZE, 0);
-	r300->hw.gb_misc.cmd[0] = cmdpacket0(R300_GB_MSPOS0, 5);
+	r300->hw.gb_misc.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_MSPOS0, 3);
+	r300->hw.gb_misc.emit = emit_gb_misc;
+	ALLOC_STATE(gb_misc2, always, R300_GB_MISC2_CMDSIZE, 0);
+    r300->hw.gb_misc2.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x401C, 2);
 	ALLOC_STATE(txe, always, R300_TXE_CMDSIZE, 0);
-	r300->hw.txe.cmd[R300_TXE_CMD_0] = cmdpacket0(R300_TX_ENABLE, 1);
+	r300->hw.txe.cmd[R300_TXE_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_ENABLE, 1);
 	ALLOC_STATE(ga_point_s0, always, 5, 0);
-	r300->hw.ga_point_s0.cmd[0] = cmdpacket0(R300_GA_POINT_S0, 4);
+	r300->hw.ga_point_s0.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POINT_S0, 4);
 	ALLOC_STATE(ga_triangle_stipple, always, 2, 0);
-	r300->hw.ga_triangle_stipple.cmd[0] = cmdpacket0(R300_GA_TRIANGLE_STIPPLE, 1);
+	r300->hw.ga_triangle_stipple.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_TRIANGLE_STIPPLE, 1);
 	ALLOC_STATE(ps, always, R300_PS_CMDSIZE, 0);
-	r300->hw.ps.cmd[0] = cmdpacket0(R300_GA_POINT_SIZE, 1);
+	r300->hw.ps.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POINT_SIZE, 1);
 	ALLOC_STATE(ga_point_minmax, always, 4, 0);
-	r300->hw.ga_point_minmax.cmd[0] = cmdpacket0(R300_GA_POINT_MINMAX, 3);
+	r300->hw.ga_point_minmax.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POINT_MINMAX, 3);
 	ALLOC_STATE(lcntl, always, 2, 0);
-	r300->hw.lcntl.cmd[0] = cmdpacket0(R300_GA_LINE_CNTL, 1);
+	r300->hw.lcntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_LINE_CNTL, 1);
 	ALLOC_STATE(ga_line_stipple, always, 4, 0);
-	r300->hw.ga_line_stipple.cmd[0] = cmdpacket0(R300_GA_LINE_STIPPLE_VALUE, 3);
-	ALLOC_STATE(shade, always, 5, 0);
-	r300->hw.shade.cmd[0] = cmdpacket0(R300_GA_ENHANCE, 4);
+	r300->hw.ga_line_stipple.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_LINE_STIPPLE_VALUE, 3);
+	ALLOC_STATE(shade, always, 2, 0);
+	r300->hw.shade.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_ENHANCE, 1);
+	r300->hw.shade.emit = emit_shade_misc;
+	ALLOC_STATE(shade2, always, 4, 0);
+	r300->hw.shade2.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x4278, 3);
 	ALLOC_STATE(polygon_mode, always, 4, 0);
-	r300->hw.polygon_mode.cmd[0] = cmdpacket0(R300_GA_POLY_MODE, 3);
+	r300->hw.polygon_mode.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POLY_MODE, 3);
 	ALLOC_STATE(fogp, always, 3, 0);
-	r300->hw.fogp.cmd[0] = cmdpacket0(R300_GA_FOG_SCALE, 2);
+	r300->hw.fogp.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_FOG_SCALE, 2);
 	ALLOC_STATE(zbias_cntl, always, 2, 0);
-	r300->hw.zbias_cntl.cmd[0] = cmdpacket0(R300_SU_TEX_WRAP, 1);
+	r300->hw.zbias_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_TEX_WRAP, 1);
 	ALLOC_STATE(zbs, always, R300_ZBS_CMDSIZE, 0);
 	r300->hw.zbs.cmd[R300_ZBS_CMD_0] =
-	    cmdpacket0(R300_SU_POLY_OFFSET_FRONT_SCALE, 4);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_SU_POLY_OFFSET_FRONT_SCALE, 4);
 	ALLOC_STATE(occlusion_cntl, always, 2, 0);
-	r300->hw.occlusion_cntl.cmd[0] = cmdpacket0(R300_SU_POLY_OFFSET_ENABLE, 1);
+	r300->hw.occlusion_cntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_POLY_OFFSET_ENABLE, 1);
 	ALLOC_STATE(cul, always, R300_CUL_CMDSIZE, 0);
-	r300->hw.cul.cmd[R300_CUL_CMD_0] = cmdpacket0(R300_SU_CULL_MODE, 1);
+	r300->hw.cul.cmd[R300_CUL_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_CULL_MODE, 1);
 	ALLOC_STATE(su_depth_scale, always, 3, 0);
-	r300->hw.su_depth_scale.cmd[0] = cmdpacket0(R300_SU_DEPTH_SCALE, 2);
+	r300->hw.su_depth_scale.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SU_DEPTH_SCALE, 2);
 	ALLOC_STATE(rc, always, R300_RC_CMDSIZE, 0);
-	r300->hw.rc.cmd[R300_RC_CMD_0] = cmdpacket0(R300_RS_COUNT, 2);
+	r300->hw.rc.cmd[R300_RC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_COUNT, 2);
 	if (is_r500) {
-		ALLOC_STATE(ri, always, R500_RI_CMDSIZE, 0);
-		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(R500_RS_IP_0, 16);
-		for (i = 0; i < 8; i++) {
-			r300->hw.ri.cmd[R300_RI_CMD_0 + i +1] =
-			  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
-                          (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
-                          (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
-                          (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT);
-		}
+		ALLOC_STATE(ri, variable, R500_RI_CMDSIZE, 0);
+		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_IP_0, 16);
 		ALLOC_STATE(rr, variable, R300_RR_CMDSIZE, 0);
-		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R500_RS_INST_0, 1);
+		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, 1);
 	} else {
-		ALLOC_STATE(ri, always, R300_RI_CMDSIZE, 0);
-		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(R300_RS_IP_0, 8);
+		ALLOC_STATE(ri, variable, R300_RI_CMDSIZE, 0);
+		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_IP_0, 8);
 		ALLOC_STATE(rr, variable, R300_RR_CMDSIZE, 0);
-		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R300_RS_INST_0, 1);
+		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, 1);
 	}
 	ALLOC_STATE(sc_hyperz, always, 3, 0);
-	r300->hw.sc_hyperz.cmd[0] = cmdpacket0(R300_SC_HYPERZ, 2);
+	r300->hw.sc_hyperz.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SC_HYPERZ, 2);
 	ALLOC_STATE(sc_screendoor, always, 2, 0);
-	r300->hw.sc_screendoor.cmd[0] = cmdpacket0(R300_SC_SCREENDOOR, 1);
+	r300->hw.sc_screendoor.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_SC_SCREENDOOR, 1);
 	ALLOC_STATE(us_out_fmt, always, 6, 0);
-	r300->hw.us_out_fmt.cmd[0] = cmdpacket0(R300_US_OUT_FMT, 5);
+	r300->hw.us_out_fmt.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_OUT_FMT, 5);
 
 	if (is_r500) {
 		ALLOC_STATE(fp, always, R500_FP_CMDSIZE, 0);
-		r300->hw.fp.cmd[R500_FP_CMD_0] = cmdpacket0(R500_US_CONFIG, 2);
+		r300->hw.fp.cmd[R500_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_US_CONFIG, 2);
 		r300->hw.fp.cmd[R500_FP_CNTL] = R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO;
-		r300->hw.fp.cmd[R500_FP_CMD_1] = cmdpacket0(R500_US_CODE_ADDR, 3);
-		r300->hw.fp.cmd[R500_FP_CMD_2] = cmdpacket0(R500_US_FC_CTRL, 1);
+		r300->hw.fp.cmd[R500_FP_CMD_1] = cmdpacket0(r300->radeon.radeonScreen, R500_US_CODE_ADDR, 3);
+		r300->hw.fp.cmd[R500_FP_CMD_2] = cmdpacket0(r300->radeon.radeonScreen, R500_US_FC_CTRL, 1);
 		r300->hw.fp.cmd[R500_FP_FC_CNTL] = 0; /* FIXME when we add flow control */
 
 		ALLOC_STATE(r500fp, r500fp, R500_FPI_CMDSIZE, 0);
-		r300->hw.r500fp.cmd[R300_FPI_CMD_0] = cmdr500fp(0, 0, 0, 0);
+		r300->hw.r500fp.cmd[R300_FPI_CMD_0] =
+			cmdr500fp(r300->radeon.radeonScreen, 0, 0, 0, 0);
+		r300->hw.r500fp.emit = emit_r500fp;
 		ALLOC_STATE(r500fp_const, r500fp_const, R500_FPP_CMDSIZE, 0);
-		r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] = cmdr500fp(0, 0, 1, 0);
+		r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] =
+			cmdr500fp(r300->radeon.radeonScreen, 0, 0, 1, 0);
+		r300->hw.r500fp_const.emit = emit_r500fp;
 	} else {
 		ALLOC_STATE(fp, always, R300_FP_CMDSIZE, 0);
-		r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(R300_US_CONFIG, 3);
-		r300->hw.fp.cmd[R300_FP_CMD_1] = cmdpacket0(R300_US_CODE_ADDR_0, 4);
+		r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CONFIG, 3);
+		r300->hw.fp.cmd[R300_FP_CMD_1] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CODE_ADDR_0, 4);
+
 		ALLOC_STATE(fpt, variable, R300_FPT_CMDSIZE, 0);
-		r300->hw.fpt.cmd[R300_FPT_CMD_0] = cmdpacket0(R300_US_TEX_INST_0, 0);
+		r300->hw.fpt.cmd[R300_FPT_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_TEX_INST_0, 0);
 
 		ALLOC_STATE(fpi[0], variable, R300_FPI_CMDSIZE, 0);
-		r300->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_INST_0, 1);
+		r300->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_RGB_INST_0, 1);
 		ALLOC_STATE(fpi[1], variable, R300_FPI_CMDSIZE, 1);
-		r300->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_ADDR_0, 1);
+		r300->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_RGB_ADDR_0, 1);
 		ALLOC_STATE(fpi[2], variable, R300_FPI_CMDSIZE, 2);
-		r300->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_INST_0, 1);
+		r300->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_ALPHA_INST_0, 1);
 		ALLOC_STATE(fpi[3], variable, R300_FPI_CMDSIZE, 3);
-		r300->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_ADDR_0, 1);
+		r300->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_ALU_ALPHA_ADDR_0, 1);
 		ALLOC_STATE(fpp, variable, R300_FPP_CMDSIZE, 0);
-		r300->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(R300_PFS_PARAM_0_X, 0);
+		r300->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_PFS_PARAM_0_X, 0);
 	}
 	ALLOC_STATE(fogs, always, R300_FOGS_CMDSIZE, 0);
-	r300->hw.fogs.cmd[R300_FOGS_CMD_0] = cmdpacket0(R300_FG_FOG_BLEND, 1);
+	r300->hw.fogs.cmd[R300_FOGS_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_FOG_BLEND, 1);
 	ALLOC_STATE(fogc, always, R300_FOGC_CMDSIZE, 0);
-	r300->hw.fogc.cmd[R300_FOGC_CMD_0] = cmdpacket0(R300_FG_FOG_COLOR_R, 3);
+	r300->hw.fogc.cmd[R300_FOGC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_FOG_COLOR_R, 3);
 	ALLOC_STATE(at, always, R300_AT_CMDSIZE, 0);
-	r300->hw.at.cmd[R300_AT_CMD_0] = cmdpacket0(R300_FG_ALPHA_FUNC, 2);
+	r300->hw.at.cmd[R300_AT_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_ALPHA_FUNC, 2);
 	ALLOC_STATE(fg_depth_src, always, 2, 0);
-	r300->hw.fg_depth_src.cmd[0] = cmdpacket0(R300_FG_DEPTH_SRC, 1);
+	r300->hw.fg_depth_src.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_FG_DEPTH_SRC, 1);
 	ALLOC_STATE(rb3d_cctl, always, 2, 0);
-	r300->hw.rb3d_cctl.cmd[0] = cmdpacket0(R300_RB3D_CCTL, 1);
+	r300->hw.rb3d_cctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_CCTL, 1);
 	ALLOC_STATE(bld, always, R300_BLD_CMDSIZE, 0);
-	r300->hw.bld.cmd[R300_BLD_CMD_0] = cmdpacket0(R300_RB3D_CBLEND, 2);
+	r300->hw.bld.cmd[R300_BLD_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_CBLEND, 2);
 	ALLOC_STATE(cmk, always, R300_CMK_CMDSIZE, 0);
-	r300->hw.cmk.cmd[R300_CMK_CMD_0] = cmdpacket0(RB3D_COLOR_CHANNEL_MASK, 1);
+	r300->hw.cmk.cmd[R300_CMK_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, RB3D_COLOR_CHANNEL_MASK, 1);
 	if (is_r500) {
 		ALLOC_STATE(blend_color, always, 3, 0);
-		r300->hw.blend_color.cmd[0] = cmdpacket0(R500_RB3D_CONSTANT_COLOR_AR, 2);
+		r300->hw.blend_color.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_RB3D_CONSTANT_COLOR_AR, 2);
 	} else {
 		ALLOC_STATE(blend_color, always, 2, 0);
-		r300->hw.blend_color.cmd[0] = cmdpacket0(R300_RB3D_BLEND_COLOR, 1);
+		r300->hw.blend_color.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_BLEND_COLOR, 1);
 	}
 	ALLOC_STATE(rop, always, 2, 0);
-	r300->hw.rop.cmd[0] = cmdpacket0(R300_RB3D_ROPCNTL, 1);
+	r300->hw.rop.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_ROPCNTL, 1);
 	ALLOC_STATE(cb, always, R300_CB_CMDSIZE, 0);
-	r300->hw.cb.cmd[R300_CB_CMD_0] = cmdpacket0(R300_RB3D_COLOROFFSET0, 1);
-	r300->hw.cb.cmd[R300_CB_CMD_1] = cmdpacket0(R300_RB3D_COLORPITCH0, 1);
+	r300->hw.cb.emit = &emit_cb_offset;
 	ALLOC_STATE(rb3d_dither_ctl, always, 10, 0);
-	r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(R300_RB3D_DITHER_CTL, 9);
+	r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DITHER_CTL, 9);
 	ALLOC_STATE(rb3d_aaresolve_ctl, always, 2, 0);
-	r300->hw.rb3d_aaresolve_ctl.cmd[0] = cmdpacket0(R300_RB3D_AARESOLVE_CTL, 1);
-	ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, always, 3, 0);
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[0] = cmdpacket0(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 2);
+	r300->hw.rb3d_aaresolve_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_AARESOLVE_CTL, 1);
+    ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, always, 3, 0);
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 2);
+	r300->hw.rb3d_discard_src_pixel_lte_threshold.emit = emit_threshold_misc;
 	ALLOC_STATE(zs, always, R300_ZS_CMDSIZE, 0);
 	r300->hw.zs.cmd[R300_ZS_CMD_0] =
-	    cmdpacket0(R300_ZB_CNTL, 3);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_ZB_CNTL, 3);
+
 	ALLOC_STATE(zstencil_format, always, 5, 0);
 	r300->hw.zstencil_format.cmd[0] =
-	    cmdpacket0(R300_ZB_FORMAT, 4);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_ZB_FORMAT, 4);
+	r300->hw.zstencil_format.emit = emit_zstencil_format;
+
 	ALLOC_STATE(zb, always, R300_ZB_CMDSIZE, 0);
-	r300->hw.zb.cmd[R300_ZB_CMD_0] = cmdpacket0(R300_ZB_DEPTHOFFSET, 2);
+	r300->hw.zb.emit = emit_zb_offset;
 	ALLOC_STATE(zb_depthclearvalue, always, 2, 0);
-	r300->hw.zb_depthclearvalue.cmd[0] = cmdpacket0(R300_ZB_DEPTHCLEARVALUE, 1);
-	ALLOC_STATE(unk4F30, always, 3, 0);
-	r300->hw.unk4F30.cmd[0] = cmdpacket0(0x4F30, 2);
+	r300->hw.zb_depthclearvalue.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_DEPTHCLEARVALUE, 1);
+	ALLOC_STATE(zb_zmask, always, 3, 0);
+	r300->hw.zb_zmask.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_ZMASK_OFFSET, 2);
 	ALLOC_STATE(zb_hiz_offset, always, 2, 0);
-	r300->hw.zb_hiz_offset.cmd[0] = cmdpacket0(R300_ZB_HIZ_OFFSET, 1);
+	r300->hw.zb_hiz_offset.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_HIZ_OFFSET, 1);
 	ALLOC_STATE(zb_hiz_pitch, always, 2, 0);
-	r300->hw.zb_hiz_pitch.cmd[0] = cmdpacket0(R300_ZB_HIZ_PITCH, 1);
+	r300->hw.zb_hiz_pitch.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_HIZ_PITCH, 1);
 
 	/* VPU only on TCL */
 	if (has_tcl) {
    	        int i;
 		ALLOC_STATE(vpi, vpu, R300_VPI_CMDSIZE, 0);
-		r300->hw.vpi.cmd[R300_VPI_CMD_0] =
-		    cmdvpu(R300_PVS_CODE_START, 0);
+		r300->hw.vpi.cmd[0] =
+		    cmdvpu(r300->radeon.radeonScreen, R300_PVS_CODE_START, 0);
+		r300->hw.vpi.emit = emit_vpu;
 
 		if (is_r500) {
 		    ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
-		    r300->hw.vpp.cmd[R300_VPP_CMD_0] =
-			cmdvpu(R500_PVS_CONST_START, 0);
+		    r300->hw.vpp.cmd[0] =
+			cmdvpu(r300->radeon.radeonScreen, R500_PVS_CONST_START, 0);
+		    r300->hw.vpp.emit = emit_vpu;
 
 		    ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
-		    r300->hw.vps.cmd[R300_VPS_CMD_0] =
-			cmdvpu(R500_POINT_VPORT_SCALE_OFFSET, 1);
+		    r300->hw.vps.cmd[0] =
+			cmdvpu(r300->radeon.radeonScreen, R500_POINT_VPORT_SCALE_OFFSET, 1);
+		    r300->hw.vps.emit = emit_vpu;
 
 			for (i = 0; i < 6; i++) {
-				ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0);
-				r300->hw.vpucp[i].cmd[R300_VPUCP_CMD_0] =
-					cmdvpu(R500_PVS_UCP_START + i, 1);
+			  ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0);
+			  r300->hw.vpucp[i].cmd[0] =
+				  cmdvpu(r300->radeon.radeonScreen,
+                           R500_PVS_UCP_START + i, 1);
+				r300->hw.vpucp[i].emit = emit_vpu;
 			}
 		} else {
 		    ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
-		    r300->hw.vpp.cmd[R300_VPP_CMD_0] =
-			cmdvpu(R300_PVS_CONST_START, 0);
+		    r300->hw.vpp.cmd[0] =
+			cmdvpu(r300->radeon.radeonScreen, R300_PVS_CONST_START, 0);
+		    r300->hw.vpp.emit = emit_vpu;
 
 		    ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
-		    r300->hw.vps.cmd[R300_VPS_CMD_0] =
-			cmdvpu(R300_POINT_VPORT_SCALE_OFFSET, 1);
+		    r300->hw.vps.cmd[0] =
+			cmdvpu(r300->radeon.radeonScreen, R300_POINT_VPORT_SCALE_OFFSET, 1);
+		    r300->hw.vps.emit = emit_vpu;
 
 			for (i = 0; i < 6; i++) {
 				ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0);
-				r300->hw.vpucp[i].cmd[R300_VPUCP_CMD_0] =
-					cmdvpu(R300_PVS_UCP_START + i, 1);
+				r300->hw.vpucp[i].cmd[0] =
+					cmdvpu(r300->radeon.radeonScreen,
+					       R300_PVS_UCP_START + i, 1);
+				r300->hw.vpucp[i].emit = emit_vpu;
 			}
 		}
 	}
@@ -556,130 +763,37 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	/* Textures */
 	ALLOC_STATE(tex.filter, variable, mtu + 1, 0);
 	r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER0_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, 0);
 
 	ALLOC_STATE(tex.filter_1, variable, mtu + 1, 0);
 	r300->hw.tex.filter_1.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER1_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER1_0, 0);
 
 	ALLOC_STATE(tex.size, variable, mtu + 1, 0);
-	r300->hw.tex.size.cmd[R300_TEX_CMD_0] = cmdpacket0(R300_TX_SIZE_0, 0);
+	r300->hw.tex.size.cmd[R300_TEX_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_SIZE_0, 0);
 
 	ALLOC_STATE(tex.format, variable, mtu + 1, 0);
 	r300->hw.tex.format.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FORMAT_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT_0, 0);
 
 	ALLOC_STATE(tex.pitch, variable, mtu + 1, 0);
-	r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] = cmdpacket0(R300_TX_FORMAT2_0, 0);
+	r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT2_0, 0);
 
-	ALLOC_STATE(tex.offset, variable, mtu + 1, 0);
+	ALLOC_STATE(tex.offset, variable, 1, 0);
 	r300->hw.tex.offset.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_OFFSET_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_OFFSET_0, 0);
+	r300->hw.tex.offset.emit = &emit_tex_offsets;
 
 	ALLOC_STATE(tex.chroma_key, variable, mtu + 1, 0);
 	r300->hw.tex.chroma_key.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_CHROMA_KEY_0, 0);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_CHROMA_KEY_0, 0);
 
 	ALLOC_STATE(tex.border_color, variable, mtu + 1, 0);
 	r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_BORDER_COLOR_0, 0);
-
-	r300->hw.is_dirty = GL_TRUE;
-	r300->hw.all_dirty = GL_TRUE;
-
-	/* Initialize command buffer */
-	size =
-	    256 * driQueryOptioni(&r300->radeon.optionCache,
-				  "command_buffer_size");
-	if (size < 2 * r300->hw.max_state_size) {
-		size = 2 * r300->hw.max_state_size + 65535;
-	}
-	if (size > 64 * 256)
-		size = 64 * 256;
-
-	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) {
-		fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n",
-			sizeof(drm_r300_cmd_header_t));
-		fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%zd\n",
-			sizeof(drm_radeon_cmd_buffer_t));
-		fprintf(stderr,
-			"Allocating %d bytes command buffer (max state is %d bytes)\n",
-			size * 4, r300->hw.max_state_size * 4);
-	}
-
-	r300->cmdbuf.size = size;
-	r300->cmdbuf.cmd_buf = (uint32_t *) CALLOC(size * 4);
-	r300->cmdbuf.count_used = 0;
-	r300->cmdbuf.count_reemit = 0;
-}
-
-/**
- * Destroy the command buffer and state atoms.
- */
-void r300DestroyCmdBuf(r300ContextPtr r300)
-{
-	struct r300_state_atom *atom;
-
-	FREE(r300->cmdbuf.cmd_buf);
-
-	foreach(atom, &r300->hw.atomlist) {
-		FREE(atom->cmd);
-	}
-}
-
-void r300EmitBlit(r300ContextPtr rmesa,
-		  GLuint color_fmt,
-		  GLuint src_pitch,
-		  GLuint src_offset,
-		  GLuint dst_pitch,
-		  GLuint dst_offset,
-		  GLint srcx, GLint srcy,
-		  GLint dstx, GLint dsty, GLuint w, GLuint h)
-{
-	drm_r300_cmd_header_t *cmd;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr,
-			"%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-			__FUNCTION__, src_pitch, src_offset, srcx, srcy,
-			dst_pitch, dst_offset, dstx, dsty, w, h);
-
-	assert((src_pitch & 63) == 0);
-	assert((dst_pitch & 63) == 0);
-	assert((src_offset & 1023) == 0);
-	assert((dst_offset & 1023) == 0);
-	assert(w < (1 << 16));
-	assert(h < (1 << 16));
-
-	cmd = (drm_r300_cmd_header_t *) r300AllocCmdBuf(rmesa, 8, __FUNCTION__);
-
-	cmd[0].header.cmd_type = R300_CMD_PACKET3;
-	cmd[0].header.pad0 = R300_CMD_PACKET3_RAW;
-	cmd[1].u = R300_CP_CMD_BITBLT_MULTI | (5 << 16);
-	cmd[2].u = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-		    RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-		    RADEON_GMC_BRUSH_NONE |
-		    (color_fmt << 8) |
-		    RADEON_GMC_SRC_DATATYPE_COLOR |
-		    RADEON_ROP3_S |
-		    RADEON_DP_SRC_SOURCE_MEMORY |
-		    RADEON_GMC_CLR_CMP_CNTL_DIS | RADEON_GMC_WR_MSK_DIS);
-
-	cmd[3].u = ((src_pitch / 64) << 22) | (src_offset >> 10);
-	cmd[4].u = ((dst_pitch / 64) << 22) | (dst_offset >> 10);
-	cmd[5].u = (srcx << 16) | srcy;
-	cmd[6].u = (dstx << 16) | dsty;	/* dst */
-	cmd[7].u = (w << 16) | h;
-}
-
-void r300EmitWait(r300ContextPtr rmesa, GLuint flags)
-{
-	drm_r300_cmd_header_t *cmd;
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, 0);
 
-	assert(!(flags & ~(R300_WAIT_2D | R300_WAIT_3D)));
+	r300->radeon.hw.is_dirty = GL_TRUE;
+	r300->radeon.hw.all_dirty = GL_TRUE;
 
-	cmd = (drm_r300_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].u = 0;
-	cmd[0].wait.cmd_type = R300_CMD_WAIT;
-	cmd[0].wait.flags = flags;
+	rcommonInitCmdBuf(&r300->radeon);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h
index a8eaa580bd..53bcc0eeb4 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h
@@ -38,79 +38,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "r300_context.h"
 
-extern int r300FlushCmdBufLocked(r300ContextPtr r300, const char *caller);
-extern int r300FlushCmdBuf(r300ContextPtr r300, const char *caller);
-
-extern void r300EmitState(r300ContextPtr r300);
-
 extern void r300InitCmdBuf(r300ContextPtr r300);
-extern void r300DestroyCmdBuf(r300ContextPtr r300);
-
-/**
- * Make sure that enough space is available in the command buffer
- * by flushing if necessary.
- *
- * \param dwords The number of dwords we need to be free on the command buffer
- */
-static INLINE void r300EnsureCmdBufSpace(r300ContextPtr r300,
-					     int dwords, const char *caller)
-{
-	assert(dwords < r300->cmdbuf.size);
-
-	if (r300->cmdbuf.count_used + dwords > r300->cmdbuf.size)
-		r300FlushCmdBuf(r300, caller);
-}
-
-/**
- * Allocate the given number of dwords in the command buffer and return
- * a pointer to the allocated area.
- * When necessary, these functions cause a flush. r300AllocCmdBuf() also
- * causes state reemission after a flush. This is necessary to ensure
- * correct hardware state after an unlock.
- */
-static INLINE uint32_t *r300RawAllocCmdBuf(r300ContextPtr r300,
-					       int dwords, const char *caller)
-{
-	uint32_t *ptr;
-
-	r300EnsureCmdBufSpace(r300, dwords, caller);
-
-	ptr = &r300->cmdbuf.cmd_buf[r300->cmdbuf.count_used];
-	r300->cmdbuf.count_used += dwords;
-	return ptr;
-}
-
-static INLINE uint32_t *r300AllocCmdBuf(r300ContextPtr r300,
-					    int dwords, const char *caller)
-{
-	uint32_t *ptr;
-
-	r300EnsureCmdBufSpace(r300, dwords, caller);
-
-	if (!r300->cmdbuf.count_used) {
-		if (RADEON_DEBUG & DEBUG_IOCTL)
-			fprintf(stderr,
-				"Reemit state after flush (from %s)\n", caller);
-		r300EmitState(r300);
-	}
-
-	ptr = &r300->cmdbuf.cmd_buf[r300->cmdbuf.count_used];
-	r300->cmdbuf.count_used += dwords;
-	return ptr;
-}
+void r300_emit_scissor(GLcontext *ctx);
 
-extern void r300EmitBlit(r300ContextPtr rmesa,
-			 GLuint color_fmt,
-			 GLuint src_pitch,
-			 GLuint src_offset,
-			 GLuint dst_pitch,
-			 GLuint dst_offset,
-			 GLint srcx, GLint srcy,
-			 GLint dstx, GLint dsty, GLuint w, GLuint h);
+void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom);
+int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom);
 
-extern void r300EmitWait(r300ContextPtr rmesa, GLuint flags);
-extern void r300EmitLOAD_VBPNTR(r300ContextPtr rmesa, int start);
-extern void r300EmitVertexShader(r300ContextPtr rmesa);
-extern void r300EmitPixelShader(r300ContextPtr rmesa);
+void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom);
+int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom);
+int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom);
 
 #endif				/* __R300_CMDBUF_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 7d6705058f..6f3aab986d 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -43,8 +43,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/matrix.h"
 #include "main/extensions.h"
 #include "main/state.h"
-#include "main/texobj.h"
 #include "main/bufferobj.h"
+#include "main/texobj.h"
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
@@ -56,42 +56,40 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "drivers/common/driverfuncs.h"
 
-#include "radeon_ioctl.h"
-#include "radeon_span.h"
 #include "r300_context.h"
+#include "radeon_context.h"
+#include "radeon_span.h"
 #include "r300_cmdbuf.h"
 #include "r300_state.h"
 #include "r300_ioctl.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
+#include "r300_render.h"
 #include "r300_swtcl.h"
+#include "radeon_bocs_wrapper.h"
 
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
 
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h"		/* for symbolic values of enum-type options */
 
-/* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
-int future_hw_tcl_on = 1;
-int hw_tcl_on = 1;
-
 #define need_GL_VERSION_2_0
 #define need_GL_ARB_point_parameters
 #define need_GL_ARB_vertex_program
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
 #define need_GL_EXT_blend_minmax
+#define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_stencil_two_side
 #define need_GL_ATI_separate_stencil
 #define need_GL_NV_vertex_program
+
 #include "extension_helper.h"
 
+
 const struct dri_extension card_extensions[] = {
   /* *INDENT-OFF* */
   {"GL_ARB_depth_texture",		NULL},
@@ -112,6 +110,7 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_blend_func_separate",	GL_EXT_blend_func_separate_functions},
   {"GL_EXT_blend_minmax",		GL_EXT_blend_minmax_functions},
   {"GL_EXT_blend_subtract",		NULL},
+  {"GL_EXT_packed_depth_stencil",	NULL},
   {"GL_EXT_fog_coord",			GL_EXT_fog_coord_functions },
   {"GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions},
   {"GL_EXT_secondary_color", 		GL_EXT_secondary_color_functions},
@@ -125,6 +124,8 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_texture_lod_bias",		NULL},
   {"GL_EXT_texture_mirror_clamp",	NULL},
   {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_EXT_texture_sRGB",		NULL},
+  {"GL_EXT_vertex_array_bgra",		NULL},
   {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
   {"GL_ATI_texture_env_combine3",	NULL},
   {"GL_ATI_texture_mirror_once",	NULL},
@@ -139,6 +140,11 @@ const struct dri_extension card_extensions[] = {
 };
 
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 /**
  * The GL 2.0 functions are needed to make display lists work with
  * functions added by GL_ATI_separate_stencil.
@@ -147,16 +153,8 @@ const struct dri_extension gl_20_extension[] = {
   {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
 };
 
-
-extern struct tnl_pipeline_stage _r300_render_stage;
-extern const struct tnl_pipeline_stage _r300_tcl_stage;
-
 static const struct tnl_pipeline_stage *r300_pipeline[] = {
 
-	/* Try and go straight to t&l
-	 */
-	&_r300_tcl_stage,
-
 	/* Catch any t&l fallbacks
 	 */
 	&_tnl_vertex_transform_stage,
@@ -165,6 +163,7 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	&_tnl_fog_coordinate_stage,
 	&_tnl_texgen_stage,
 	&_tnl_texture_transform_stage,
+	&_tnl_point_attenuation_stage,
 	&_tnl_vertex_program_stage,
 
 	/* Try again to go to tcl?
@@ -184,6 +183,192 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	0,
 };
 
+static void r300_get_lock(radeonContextPtr rmesa)
+{
+	drm_radeon_sarea_t *sarea = rmesa->sarea;
+
+	if (sarea->ctx_owner != rmesa->dri.hwContext) {
+		sarea->ctx_owner = rmesa->dri.hwContext;
+		if (!rmesa->radeonScreen->kernel_mm)
+			radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
+	}
+}
+
+static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+    /* please flush pipe do all pending work */
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_SC_SCREENDOOR, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_SC_SCREENDOOR, 1));
+    radeon_cs_write_dword(cs, 0x00FFFFFF);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_SC_HYPERZ, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_US_CONFIG, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_ZB_CNTL, 1));
+    radeon_cs_write_dword(cs, 0x0);
+    radeon_cs_write_dword(cs, cmdwait(rmesa->radeonScreen, R300_WAIT_3D));
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_RB3D_DSTCACHE_CTLSTAT, 1));
+    radeon_cs_write_dword(cs, R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
+    radeon_cs_write_dword(cs, cmdpacket0(rmesa->radeonScreen,
+                                  R300_ZB_ZCACHE_CTLSTAT, 1));
+    radeon_cs_write_dword(cs, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE);
+    radeon_cs_write_dword(cs, cmdwait(rmesa->radeonScreen,
+                               R300_WAIT_3D | R300_WAIT_3D_CLEAN));
+}
+
+static void r300_vtbl_pre_emit_atoms(radeonContextPtr radeon)
+{
+	r300ContextPtr r300 = (r300ContextPtr)radeon;
+	BATCH_LOCALS(radeon);
+
+	r300->vap_flush_needed = GL_TRUE;
+
+	cp_wait(radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+	BEGIN_BATCH_NO_AUTOSTATE(2);
+	OUT_BATCH_REGVAL(R300_TX_INVALTAGS, R300_TX_FLUSH);
+	END_BATCH();
+	end_3d(radeon);
+}
+
+static void r300_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	if (mode)
+		r300->radeon.Fallback |= bit;
+	else
+		r300->radeon.Fallback &= ~bit;
+}
+
+static void r300_init_vtbl(radeonContextPtr radeon)
+{
+	radeon->vtbl.get_lock = r300_get_lock;
+	radeon->vtbl.update_viewport_offset = r300UpdateViewportOffset;
+	radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header;
+	radeon->vtbl.swtcl_flush = r300_swtcl_flush;
+	radeon->vtbl.pre_emit_atoms = r300_vtbl_pre_emit_atoms;
+	radeon->vtbl.fallback = r300_fallback;
+}
+
+static void r300InitConstValues(GLcontext *ctx, radeonScreenPtr screen)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+	ctx->Const.MaxTextureImageUnits =
+	    driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
+	ctx->Const.MaxTextureCoordUnits =
+	    driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
+	ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureImageUnits,
+		 ctx->Const.MaxTextureCoordUnits);
+
+	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
+	ctx->Const.MaxTextureLodBias = 16.0;
+
+	if (screen->chip_family >= CHIP_FAMILY_RV515) {
+		ctx->Const.MaxTextureLevels = 13;
+		ctx->Const.MaxCubeTextureLevels = 13;
+		ctx->Const.MaxTextureRectSize = 4096;
+	}
+	else {
+		ctx->Const.MaxTextureLevels = 12;
+		ctx->Const.MaxCubeTextureLevels = 12;
+		ctx->Const.MaxTextureRectSize = 2048;
+	}
+
+	ctx->Const.MinPointSize = 1.0;
+	ctx->Const.MinPointSizeAA = 1.0;
+	ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
+	ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
+
+	ctx->Const.MinLineWidth = 1.0;
+	ctx->Const.MinLineWidthAA = 1.0;
+	ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
+	ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
+
+	ctx->Const.MaxDrawBuffers = 1;
+
+	/* currently bogus data */
+	if (r300->options.hw_tcl_enabled) {
+		ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
+		ctx->Const.VertexProgram.MaxNativeInstructions =
+		  VSF_MAX_FRAGMENT_LENGTH / 4;
+		ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
+		ctx->Const.VertexProgram.MaxTemps = 32;
+		ctx->Const.VertexProgram.MaxNativeTemps =
+		  /*VSF_MAX_FRAGMENT_TEMPS */ 32;
+		ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
+		ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
+	}
+
+	if (screen->chip_family >= CHIP_FAMILY_RV515) {
+		ctx->Const.FragmentProgram.MaxNativeTemps = R500_PFS_NUM_TEMP_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+		ctx->Const.FragmentProgram.MaxNativeParameters = R500_PFS_NUM_CONST_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAluInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexIndirections = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+	} else {
+		ctx->Const.FragmentProgram.MaxNativeTemps = R300_PFS_NUM_TEMP_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+		ctx->Const.FragmentProgram.MaxNativeParameters = R300_PFS_NUM_CONST_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAluInstructions = R300_PFS_MAX_ALU_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexInstructions = R300_PFS_MAX_TEX_INST;
+		ctx->Const.FragmentProgram.MaxNativeInstructions = R300_PFS_MAX_ALU_INST + R300_PFS_MAX_TEX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexIndirections = R300_PFS_MAX_TEX_INDIRECT;
+		ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+	}
+}
+
+static void r300ParseOptions(r300ContextPtr r300, radeonScreenPtr screen)
+{
+	struct r300_options options = { 0 };
+
+	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
+			    screen->driScreen->myNum, "r300");
+
+	r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache, "def_max_anisotropy");
+
+	options.stencil_two_side_disabled = driQueryOptionb(&r300->radeon.optionCache, "disable_stencil_two_side");
+	options.s3tc_force_enabled = driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable");
+	options.s3tc_force_disabled = driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc");
+
+	if (!(screen->chip_flags & RADEON_CHIPSET_TCL) || driQueryOptioni(&r300->radeon.optionCache, "tcl_mode") == DRI_CONF_TCL_SW)
+		options.hw_tcl_enabled = 0;
+	else
+		options.hw_tcl_enabled = 1;
+
+	options.conformance_mode = !driQueryOptionb(&r300->radeon.optionCache, "disable_lowimpact_fallback");
+
+	r300->options = options;
+}
+
+static void r300InitGLExtensions(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+	driInitExtensions(ctx, card_extensions, GL_TRUE);
+	if (r300->radeon.radeonScreen->kernel_mm)
+		driInitExtensions(ctx, mm_extensions, GL_FALSE);
+
+	if (r300->options.stencil_two_side_disabled)
+		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
+
+	if (ctx->Mesa_DXTn && !r300->options.s3tc_force_enabled) {
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+		_mesa_enable_extension(ctx, "GL_S3_s3tc");
+	} else if (r300->options.s3tc_force_disabled) {
+		_mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+	}
+}
+
 /* Create the device specific rendering context.
  */
 GLboolean r300CreateContext(const __GLcontextModes * glVisual,
@@ -195,42 +380,25 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	struct dd_function_table functions;
 	r300ContextPtr r300;
 	GLcontext *ctx;
-	int tcl_mode, i;
 
 	assert(glVisual);
 	assert(driContextPriv);
 	assert(screen);
 
-	/* Allocate the R300 context */
 	r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
 	if (!r300)
 		return GL_FALSE;
 
-	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
-		hw_tcl_on = future_hw_tcl_on = 0;
+	r300ParseOptions(r300, screen);
 
-	/* Parse configuration files.
-	 * Do this here so that initialMaxAnisotropy is set before we create
-	 * the default textures.
-	 */
-	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
-			    screen->driScreen->myNum, "r300");
-	r300->initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
-						     "def_max_anisotropy");
+	r300_init_vtbl(&r300->radeon);
 
-	/* Init default driver functions then plug in our R300-specific functions
-	 * (the texture functions are especially important)
-	 */
 	_mesa_init_driver_functions(&functions);
 	r300InitIoctlFuncs(&functions);
 	r300InitStateFuncs(&functions);
 	r300InitTextureFuncs(&functions);
 	r300InitShaderFuncs(&functions);
 
-#ifdef USER_BUFFERS
-	r300_mem_init(r300);
-#endif
-
 	if (!radeonInitContext(&r300->radeon, &functions,
 			       glVisual, driContextPriv,
 			       sharedContextPrivate)) {
@@ -238,94 +406,15 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 		return GL_FALSE;
 	}
 
-	/* Init r300 context data */
-	r300->dma.buf0_address =
-	    r300->radeon.radeonScreen->buffers->list[0].address;
-
-	(void)memset(r300->texture_heaps, 0, sizeof(r300->texture_heaps));
-	make_empty_list(&r300->swapped);
-
-	r300->nr_heaps = 1 /* screen->numTexHeaps */ ;
-	assert(r300->nr_heaps < RADEON_NR_TEX_HEAPS);
-	for (i = 0; i < r300->nr_heaps; i++) {
-		/* *INDENT-OFF* */
-		r300->texture_heaps[i] = driCreateTextureHeap(i, r300,
-							       screen->
-							       texSize[i], 12,
-							       RADEON_NR_TEX_REGIONS,
-							       (drmTextureRegionPtr)
-							       r300->radeon.sarea->
-							       tex_list[i],
-							       &r300->radeon.sarea->
-							       tex_age[i],
-							       &r300->swapped,
-							       sizeof
-							       (r300TexObj),
-							       (destroy_texture_object_t
-								*)
-							       r300DestroyTexObj);
-		/* *INDENT-ON* */
-	}
-	r300->texture_depth = driQueryOptioni(&r300->radeon.optionCache,
-					      "texture_depth");
-	if (r300->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
-		r300->texture_depth = (screen->cpp == 4) ?
-		    DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
-
-	/* Set the maximum texture size small enough that we can guarentee that
-	 * all texture units can bind a maximal texture and have them both in
-	 * texturable memory at once.
-	 */
-
 	ctx = r300->radeon.glCtx;
 
-	ctx->Const.MaxTextureImageUnits =
-	    driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
-	ctx->Const.MaxTextureCoordUnits =
-	    driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
-	ctx->Const.MaxTextureUnits =
-	    MIN2(ctx->Const.MaxTextureImageUnits,
-		 ctx->Const.MaxTextureCoordUnits);
-	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
-	ctx->Const.MaxTextureLodBias = 16.0;
+	r300->fallback = 0;
+	if (r300->options.hw_tcl_enabled)
+		ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
 
-	if (screen->chip_family >= CHIP_FAMILY_RV515)
-	    ctx->Const.MaxTextureLevels = 13;
-	else
-	    ctx->Const.MaxTextureLevels = 12;
-
-        driCalculateMaxTextureLevels( r300->texture_heaps,
-                                      r300->nr_heaps,
-                                      & ctx->Const,
-                                      4,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      MIN2(ctx->Const.MaxTextureLevels,
-                                           MAX_3D_TEXTURE_LEVELS) - 1,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      ctx->Const.MaxTextureLevels - 1,
-                                      GL_FALSE,
-                                      2 );
-
-	ctx->Const.MinPointSize = 1.0;
-	ctx->Const.MinPointSizeAA = 1.0;
-	ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
-	ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
-
-	ctx->Const.MinLineWidth = 1.0;
-	ctx->Const.MinLineWidthAA = 1.0;
-	ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
-	ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
-
-#ifdef USER_BUFFERS
-	/* Needs further modifications */
-#if 0
-	ctx->Const.MaxArrayLockSize =
-	    ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
-#endif
-#endif
+	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 
-	ctx->Const.MaxDrawBuffers = 1;
+	r300InitConstValues(ctx, screen);
 
 	_mesa_set_mvp_with_dp4( ctx, GL_TRUE );
 
@@ -336,16 +425,12 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	_tnl_CreateContext(ctx);
 	_swsetup_CreateContext(ctx);
 	_swsetup_Wakeup(ctx);
-	_ae_create_context(ctx);
 
 	/* Install the customized pipeline:
 	 */
 	_tnl_destroy_pipeline(ctx);
 	_tnl_install_pipeline(ctx, r300_pipeline);
-
-	/* Try and keep materials and vertices separate:
-	 */
-/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
+	TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
 
 	/* Configure swrast and TNL to match hardware characteristics:
 	 */
@@ -354,226 +439,25 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	_tnl_allow_pixel_fog(ctx, GL_FALSE);
 	_tnl_allow_vertex_fog(ctx, GL_TRUE);
 
-	/* currently bogus data */
-	if (screen->chip_flags & RADEON_CHIPSET_TCL) {
-	        ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
-		ctx->Const.VertexProgram.MaxNativeInstructions =
-		  VSF_MAX_FRAGMENT_LENGTH / 4;
-		ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
-		ctx->Const.VertexProgram.MaxTemps = 32;
-		ctx->Const.VertexProgram.MaxNativeTemps =
-		  /*VSF_MAX_FRAGMENT_TEMPS */ 32;
-		ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
-		ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
+	if (r300->options.hw_tcl_enabled) {
+		r300InitDraw(ctx);
+	} else {
+		r300InitSwtcl(ctx);
 	}
 
-	ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
-	ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
-	ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
-	ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
-	ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
-	ctx->Const.FragmentProgram.MaxNativeInstructions =
-	    PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
-	ctx->Const.FragmentProgram.MaxNativeTexIndirections =
-	    PFS_MAX_TEX_INDIRECT;
-	ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;	/* and these are?? */
-	ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
-	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
-
-	driInitExtensions(ctx, card_extensions, GL_TRUE);
-
-	if (driQueryOptionb
-	    (&r300->radeon.optionCache, "disable_stencil_two_side"))
-		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
-
-	if (r300->radeon.glCtx->Mesa_DXTn
-	    && !driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc")) {
-		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
-		_mesa_enable_extension(ctx, "GL_S3_s3tc");
-	} else
-	    if (driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable"))
-	{
-		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
-	}
-
-	r300->disable_lowimpact_fallback =
-	    driQueryOptionb(&r300->radeon.optionCache,
-			    "disable_lowimpact_fallback");
-
-	radeonInitSpanFuncs(ctx);
+	radeon_fbo_init(&r300->radeon);
+	radeonInitSpanFuncs( ctx );
 	r300InitCmdBuf(r300);
 	r300InitState(r300);
-	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
-	        r300InitSwtcl(ctx);
-
-	TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
+	r300InitShaderFunctions(r300);
 
-	tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
-	if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
-		fprintf(stderr, "disabling 3D acceleration\n");
-#if R200_MERGED
-		FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
-#endif
-	}
-	if (tcl_mode == DRI_CONF_TCL_SW ||
-	    !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
-		if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-			r300->radeon.radeonScreen->chip_flags &=
-			    ~RADEON_CHIPSET_TCL;
-			fprintf(stderr, "Disabling HW TCL support\n");
-		}
-		TCL_FALLBACK(r300->radeon.glCtx,
-			     RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
+	if (screen->chip_family == CHIP_FAMILY_RS600 ||	screen->chip_family == CHIP_FAMILY_RS690 ||
+		screen->chip_family == CHIP_FAMILY_RS740) {
+		r300->radeon.texture_row_align = 64;
 	}
 
-	return GL_TRUE;
-}
-
-static void r300FreeGartAllocations(r300ContextPtr r300)
-{
-	int i, ret, tries = 0, done_age, in_use = 0;
-	drm_radeon_mem_free_t memfree;
+	r300InitGLExtensions(ctx);
 
-	memfree.region = RADEON_MEM_REGION_GART;
-
-#ifdef USER_BUFFERS
-	for (i = r300->rmm->u_last; i > 0; i--) {
-		if (r300->rmm->u_list[i].ptr == NULL) {
-			continue;
-		}
-
-		/* check whether this buffer is still in use */
-		if (r300->rmm->u_list[i].pending) {
-			in_use++;
-		}
-	}
-	/* Cannot flush/lock if no context exists. */
-	if (in_use)
-		r300FlushCmdBuf(r300, __FUNCTION__);
-
-	done_age = radeonGetAge((radeonContextPtr) r300);
-
-	for (i = r300->rmm->u_last; i > 0; i--) {
-		if (r300->rmm->u_list[i].ptr == NULL) {
-			continue;
-		}
-
-		/* check whether this buffer is still in use */
-		if (!r300->rmm->u_list[i].pending) {
-			continue;
-		}
-
-		assert(r300->rmm->u_list[i].h_pending == 0);
-
-		tries = 0;
-		while (r300->rmm->u_list[i].age > done_age && tries++ < 1000) {
-			usleep(10);
-			done_age = radeonGetAge((radeonContextPtr) r300);
-		}
-		if (tries >= 1000) {
-			WARN_ONCE("Failed to idle region!");
-		}
-
-		memfree.region_offset = (char *)r300->rmm->u_list[i].ptr -
-		    (char *)r300->radeon.radeonScreen->gartTextures.map;
-
-		ret = drmCommandWrite(r300->radeon.radeonScreen->driScreen->fd,
-				      DRM_RADEON_FREE, &memfree,
-				      sizeof(memfree));
-		if (ret) {
-			fprintf(stderr, "Failed to free at %p\nret = %s\n",
-				r300->rmm->u_list[i].ptr, strerror(-ret));
-		} else {
-			if (i == r300->rmm->u_last)
-				r300->rmm->u_last--;
-
-			r300->rmm->u_list[i].pending = 0;
-			r300->rmm->u_list[i].ptr = NULL;
-		}
-	}
-	r300->rmm->u_head = i;
-#endif				/* USER_BUFFERS */
+	return GL_TRUE;
 }
 
-/* Destroy the device specific context.
- */
-void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
-{
-	GET_CURRENT_CONTEXT(ctx);
-	r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate;
-	radeonContextPtr radeon = (radeonContextPtr) r300;
-	radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
-	int i;
-
-	if (RADEON_DEBUG & DEBUG_DRI) {
-		fprintf(stderr, "Destroying context !\n");
-	}
-
-	/* check if we're deleting the currently bound context */
-	if (&r300->radeon == current) {
-		radeonFlush(r300->radeon.glCtx);
-		_mesa_make_current(NULL, NULL, NULL);
-	}
-
-	/* Free r300 context resources */
-	assert(r300);		/* should never be null */
-
-	if (r300) {
-		GLboolean release_texture_heaps;
-
-		release_texture_heaps =
-		    (r300->radeon.glCtx->Shared->RefCount == 1);
-		_swsetup_DestroyContext(r300->radeon.glCtx);
-		_tnl_DestroyContext(r300->radeon.glCtx);
-		_vbo_DestroyContext(r300->radeon.glCtx);
-		_swrast_DestroyContext(r300->radeon.glCtx);
-
-		if (r300->dma.current.buf) {
-			r300ReleaseDmaRegion(r300, &r300->dma.current,
-					     __FUNCTION__);
-#ifndef USER_BUFFERS
-			r300FlushCmdBuf(r300, __FUNCTION__);
-#endif
-		}
-		r300FreeGartAllocations(r300);
-		r300DestroyCmdBuf(r300);
-
-		if (radeon->state.scissor.pClipRects) {
-			FREE(radeon->state.scissor.pClipRects);
-			radeon->state.scissor.pClipRects = NULL;
-		}
-
-		if (release_texture_heaps) {
-			/* This share group is about to go away, free our private
-			 * texture object data.
-			 */
-			int i;
-
-			for (i = 0; i < r300->nr_heaps; i++) {
-				driDestroyTextureHeap(r300->texture_heaps[i]);
-				r300->texture_heaps[i] = NULL;
-			}
-
-			assert(is_empty_list(&r300->swapped));
-		}
-
-                /* Drop texture object references from current hardware state */
-		for (i = 0; i < 8; i++) {
-			_mesa_reference_texobj(&r300->state.texture.unit[i].texobj, NULL);
-		}
-
-		radeonCleanupContext(&r300->radeon);
-
-#ifdef USER_BUFFERS
-		/* the memory manager might be accessed when Mesa frees the shared
-		 * state, so don't destroy it earlier
-		 */
-		r300_mem_destroy(r300);
-#endif
-
-		/* free the option cache */
-		driDestroyOptionCache(&r300->radeon.optionCache);
-
-		FREE(r300);
-	}
-}
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 96a3205f1a..f7af7d4e57 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -37,26 +37,20 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __R300_CONTEXT_H__
 #define __R300_CONTEXT_H__
 
-#include "tnl/t_vertex.h"
 #include "drm.h"
 #include "radeon_drm.h"
 #include "dri_util.h"
-#include "texmem.h"
+#include "radeon_common.h"
 
-#include "main/macros.h"
 #include "main/mtypes.h"
-#include "main/colormac.h"
-
-#define USER_BUFFERS
+#include "shader/prog_instruction.h"
 
 struct r300_context;
 typedef struct r300_context r300ContextRec;
 typedef struct r300_context *r300ContextPtr;
 
-#include "radeon_lock.h"
-#include "main/mm.h"
 
-/* From http://gcc.gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html .
+/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html .
    I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble
    with other compilers ... GLUE!
 */
@@ -73,180 +67,14 @@ typedef struct r300_context *r300ContextPtr;
 	}
 
 #include "r300_vertprog.h"
-#include "r500_fragprog.h"
-
-/**
- * This function takes a float and packs it into a uint32_t
- */
-static INLINE uint32_t r300PackFloat32(float fl)
-{
-	union {
-		float fl;
-		uint32_t u;
-	} u;
-
-	u.fl = fl;
-	return u.u;
-}
-
-/* This is probably wrong for some values, I need to test this
- * some more.  Range checking would be a good idea also..
- *
- * But it works for most things.  I'll fix it later if someone
- * else with a better clue doesn't
- */
-static INLINE uint32_t r300PackFloat24(float f)
-{
-	float mantissa;
-	int exponent;
-	uint32_t float24 = 0;
-
-	if (f == 0.0)
-		return 0;
-
-	mantissa = frexpf(f, &exponent);
-
-	/* Handle -ve */
-	if (mantissa < 0) {
-		float24 |= (1 << 23);
-		mantissa = mantissa * -1.0;
-	}
-	/* Handle exponent, bias of 63 */
-	exponent += 62;
-	float24 |= (exponent << 16);
-	/* Kill 7 LSB of mantissa */
-	float24 |= (r300PackFloat32(mantissa) & 0x7FFFFF) >> 7;
-
-	return float24;
-}
-
-/************ DMA BUFFERS **************/
-
-/* Need refcounting on dma buffers:
- */
-struct r300_dma_buffer {
-	int refcount;		/**< the number of retained regions in buf */
-	drmBufPtr buf;
-	int id;
-};
-#undef GET_START
-#ifdef USER_BUFFERS
-#define GET_START(rvb) (r300GartOffsetFromVirtual(rmesa, (rvb)->address+(rvb)->start))
-#else
-#define GET_START(rvb) (rmesa->radeon.radeonScreen->gart_buffer_offset +		\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-#endif
-/* A retained region, eg vertices for indexed vertices.
- */
-struct r300_dma_region {
-	struct r300_dma_buffer *buf;
-	char *address;		/* == buf->address */
-	int start, end, ptr;	/* offsets from start of buf */
-
-	int aos_offset;		/* address in GART memory */
-	int aos_stride;		/* distance between elements, in dwords */
-	int aos_size;		/* number of components (1-4) */
-};
-
-struct r300_dma {
-	/* Active dma region.  Allocations for vertices and retained
-	 * regions come from here.  Also used for emitting random vertices,
-	 * these may be flushed by calling flush_current();
-	 */
-	struct r300_dma_region current;
 
-	void (*flush) (r300ContextPtr);
-
-	char *buf0_address;	/* start of buf[0], for index calcs */
-
-	/* Number of "in-flight" DMA buffers, i.e. the number of buffers
-	 * for which a DISCARD command is currently queued in the command buffer.
-	 */
-	GLuint nr_released_bufs;
-};
-
-       /* Texture related */
-
-typedef struct r300_tex_obj r300TexObj, *r300TexObjPtr;
-
-/* Maximum number of mipmap levels supported by any supported GPU
- */
-#define R300_MAX_TEXTURE_LEVELS 13
-
-/* Texture object in locally shared texture space.
- */
-struct r300_tex_obj {
-	driTextureObject base;
-
-	GLuint bufAddr;		/* Offset to start of locally
-				   shared texture block */
-
-	drm_radeon_tex_image_t image[6][R300_MAX_TEXTURE_LEVELS];
-	/* Six, for the cube faces */
-
-	GLboolean image_override;	/* Image overridden by GLX_EXT_tfp */
-
-	GLuint pitch;		/* this isn't sent to hardware just used in calculations */
-	/* hardware register values */
-	/* Note that R200 has 8 registers per texture and R300 only 7 */
-	GLuint filter;
-	GLuint filter_1;
-	GLuint pitch_reg;
-	GLuint size;		/* npot only */
-	GLuint format;
-	GLuint offset;		/* Image location in the card's address space.
-				   All cube faces follow. */
-	GLuint unknown4;
-	GLuint unknown5;
-	/* end hardware registers */
-
-	/* registers computed by r200 code - keep them here to
-	   compare against what is actually written.
-
-	   to be removed later.. */
-	GLuint pp_border_color;
-	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
-	GLuint format_x;
-
-	GLboolean border_fallback;
-
-	GLuint tile_bits;	/* hw texture tile bits used on this texture */
-};
-
-struct r300_texture_env_state {
-	struct gl_texture_object *texobj;
-	GLenum format;
-	GLenum envMode;
-};
 
 /* The blit width for texture uploads
  */
 #define R300_BLIT_WIDTH_BYTES 1024
 #define R300_MAX_TEXTURE_UNITS 8
 
-struct r300_texture_state {
-	struct r300_texture_env_state unit[R300_MAX_TEXTURE_UNITS];
-	int tc_count;		/* number of incoming texture coordinates from VAP */
-};
 
-/**
- * A block of hardware state.
- *
- * When check returns non-zero, the returned number of dwords must be
- * copied verbatim into the command buffer in order to update a state atom
- * when it is dirty.
- */
-struct r300_state_atom {
-	struct r300_state_atom *next, *prev;
-	const char *name;	/* for debug */
-	int cmd_size;		/* maximum size in dwords */
-	GLuint idx;		/* index in an array (e.g. textures) */
-	uint32_t *cmd;
-	GLboolean dirty;
-
-	int (*check) (r300ContextPtr, struct r300_state_atom * atom);
-};
 
 #define R300_VPT_CMD_0		0
 #define R300_VPT_XSCALE		1
@@ -288,9 +116,11 @@ struct r300_state_atom {
 #define R300_GB_MISC_MSPOS_0		1
 #define R300_GB_MISC_MSPOS_1		2
 #define R300_GB_MISC_TILE_CONFIG	3
-#define R300_GB_MISC_SELECT		4
-#define R300_GB_MISC_AA_CONFIG		5
-#define R300_GB_MISC_CMDSIZE		6
+#define R300_GB_MISC_CMDSIZE		4
+#define R300_GB_MISC2_CMD_0		    0
+#define R300_GB_MISC2_SELECT		1
+#define R300_GB_MISC2_AA_CONFIG		2
+#define R300_GB_MISC2_CMDSIZE		3
 
 #define R300_TXE_CMD_0		0
 #define R300_TXE_ENABLE		1
@@ -463,124 +293,100 @@ struct r300_state_atom {
  * Cache for hardware register state.
  */
 struct r300_hw_state {
-	struct r300_state_atom atomlist;
-
-	GLboolean is_dirty;
-	GLboolean all_dirty;
-	int max_state_size;	/* in dwords */
-
-	struct r300_state_atom vpt;	/* viewport (1D98) */
-	struct r300_state_atom vap_cntl;
-        struct r300_state_atom vap_index_offset; /* 0x208c r5xx only */
-	struct r300_state_atom vof;	/* VAP output format register 0x2090 */
-	struct r300_state_atom vte;	/* (20B0) */
-	struct r300_state_atom vap_vf_max_vtx_indx;	/* Maximum Vertex Indx Clamp (2134) */
-	struct r300_state_atom vap_cntl_status;
-	struct r300_state_atom vir[2];	/* vap input route (2150/21E0) */
-	struct r300_state_atom vic;	/* vap input control (2180) */
-	struct r300_state_atom vap_psc_sgn_norm_cntl; /* Programmable Stream Control Signed Normalize Control (21DC) */
-	struct r300_state_atom vap_clip_cntl;
-	struct r300_state_atom vap_clip;
-	struct r300_state_atom vap_pvs_vtx_timeout_reg;	/* Vertex timeout register (2288) */
-	struct r300_state_atom pvs;	/* pvs_cntl (22D0) */
-	struct r300_state_atom gb_enable;	/* (4008) */
-	struct r300_state_atom gb_misc;	/* Multisampling position shifts ? (4010) */
-	struct r300_state_atom ga_point_s0;	/* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) (4200) */
-	struct r300_state_atom ga_triangle_stipple;	/* (4214) */
-	struct r300_state_atom ps;	/* pointsize (421C) */
-	struct r300_state_atom ga_point_minmax;	/* (4230) */
-	struct r300_state_atom lcntl;	/* line control */
-	struct r300_state_atom ga_line_stipple;	/* (4260) */
-	struct r300_state_atom shade;
-	struct r300_state_atom polygon_mode;
-	struct r300_state_atom fogp;	/* fog parameters (4294) */
-	struct r300_state_atom ga_soft_reset;	/* (429C) */
-	struct r300_state_atom zbias_cntl;
-	struct r300_state_atom zbs;	/* zbias (42A4) */
-	struct r300_state_atom occlusion_cntl;
-	struct r300_state_atom cul;	/* cull cntl (42B8) */
-	struct r300_state_atom su_depth_scale;	/* (42C0) */
-	struct r300_state_atom rc;	/* rs control (4300) */
-	struct r300_state_atom ri;	/* rs interpolators (4310) */
-	struct r300_state_atom rr;	/* rs route (4330) */
-	struct r300_state_atom sc_hyperz;	/* (43A4) */
-	struct r300_state_atom sc_screendoor;	/* (43E8) */
-	struct r300_state_atom fp;	/* fragment program cntl + nodes (4600) */
-	struct r300_state_atom fpt;	/* texi - (4620) */
-	struct r300_state_atom us_out_fmt;	/* (46A4) */
-	struct r300_state_atom r500fp;	/* r500 fp instructions */
-	struct r300_state_atom r500fp_const;	/* r500 fp constants */
-	struct r300_state_atom fpi[4];	/* fp instructions (46C0/47C0/48C0/49C0) */
-	struct r300_state_atom fogs;	/* fog state (4BC0) */
-	struct r300_state_atom fogc;	/* fog color (4BC8) */
-	struct r300_state_atom at;	/* alpha test (4BD4) */
-	struct r300_state_atom fg_depth_src;	/* (4BD8) */
-	struct r300_state_atom fpp;	/* 0x4C00 and following */
-	struct r300_state_atom rb3d_cctl;	/* (4E00) */
-	struct r300_state_atom bld;	/* blending (4E04) */
-	struct r300_state_atom cmk;	/* colormask (4E0C) */
-	struct r300_state_atom blend_color;	/* constant blend color */
-	struct r300_state_atom rop;	/* ropcntl */
-	struct r300_state_atom cb;	/* colorbuffer (4E28) */
-	struct r300_state_atom rb3d_dither_ctl;	/* (4E50) */
-	struct r300_state_atom rb3d_aaresolve_ctl;	/* (4E88) */
-	struct r300_state_atom rb3d_discard_src_pixel_lte_threshold;	/* (4E88) I saw it only written on RV350 hardware..  */
-	struct r300_state_atom zs;	/* zstencil control (4F00) */
-	struct r300_state_atom zstencil_format;
-	struct r300_state_atom zb;	/* z buffer (4F20) */
-	struct r300_state_atom zb_depthclearvalue;	/* (4F28) */
-	struct r300_state_atom unk4F30;	/* (4F30) */
-	struct r300_state_atom zb_hiz_offset;	/* (4F44) */
-	struct r300_state_atom zb_hiz_pitch;	/* (4F54) */
-
-	struct r300_state_atom vpi;	/* vp instructions */
-	struct r300_state_atom vpp;	/* vp parameters */
-	struct r300_state_atom vps;	/* vertex point size (?) */
-	struct r300_state_atom vpucp[6];	/* vp user clip plane - 6 */
+	struct radeon_state_atom vpt;	/* viewport (1D98) */
+	struct radeon_state_atom vap_cntl;
+	struct radeon_state_atom vap_index_offset; /* 0x208c r5xx only */
+	struct radeon_state_atom vof;	/* VAP output format register 0x2090 */
+	struct radeon_state_atom vte;	/* (20B0) */
+	struct radeon_state_atom vap_vf_max_vtx_indx;	/* Maximum Vertex Indx Clamp (2134) */
+	struct radeon_state_atom vap_cntl_status;
+	struct radeon_state_atom vir[2];	/* vap input route (2150/21E0) */
+	struct radeon_state_atom vic;	/* vap input control (2180) */
+	struct radeon_state_atom vap_psc_sgn_norm_cntl; /* Programmable Stream Control Signed Normalize Control (21DC) */
+	struct radeon_state_atom vap_clip_cntl;
+	struct radeon_state_atom vap_clip;
+	struct radeon_state_atom vap_pvs_vtx_timeout_reg;	/* Vertex timeout register (2288) */
+	struct radeon_state_atom pvs;	/* pvs_cntl (22D0) */
+	struct radeon_state_atom gb_enable;	/* (4008) */
+	struct radeon_state_atom gb_misc;	/* Multisampling position shifts ? (4010) */
+	struct radeon_state_atom gb_misc2;	/* Multisampling position shifts ? (4010) */
+	struct radeon_state_atom ga_point_s0;	/* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) (4200) */
+	struct radeon_state_atom ga_triangle_stipple;	/* (4214) */
+	struct radeon_state_atom ps;	/* pointsize (421C) */
+	struct radeon_state_atom ga_point_minmax;	/* (4230) */
+	struct radeon_state_atom lcntl;	/* line control */
+	struct radeon_state_atom ga_line_stipple;	/* (4260) */
+	struct radeon_state_atom shade;
+	struct radeon_state_atom shade2;
+	struct radeon_state_atom polygon_mode;
+	struct radeon_state_atom fogp;	/* fog parameters (4294) */
+	struct radeon_state_atom ga_soft_reset;	/* (429C) */
+	struct radeon_state_atom zbias_cntl;
+	struct radeon_state_atom zbs;	/* zbias (42A4) */
+	struct radeon_state_atom occlusion_cntl;
+	struct radeon_state_atom cul;	/* cull cntl (42B8) */
+	struct radeon_state_atom su_depth_scale;	/* (42C0) */
+	struct radeon_state_atom rc;	/* rs control (4300) */
+	struct radeon_state_atom ri;	/* rs interpolators (4310) */
+	struct radeon_state_atom rr;	/* rs route (4330) */
+	struct radeon_state_atom sc_hyperz;	/* (43A4) */
+	struct radeon_state_atom sc_screendoor;	/* (43E8) */
+	struct radeon_state_atom fp;	/* fragment program cntl + nodes (4600) */
+	struct radeon_state_atom fpt;	/* texi - (4620) */
+	struct radeon_state_atom us_out_fmt;	/* (46A4) */
+	struct radeon_state_atom r500fp;	/* r500 fp instructions */
+	struct radeon_state_atom r500fp_const;	/* r500 fp constants */
+	struct radeon_state_atom fpi[4];	/* fp instructions (46C0/47C0/48C0/49C0) */
+	struct radeon_state_atom fogs;	/* fog state (4BC0) */
+	struct radeon_state_atom fogc;	/* fog color (4BC8) */
+	struct radeon_state_atom at;	/* alpha test (4BD4) */
+	struct radeon_state_atom fg_depth_src;	/* (4BD8) */
+	struct radeon_state_atom fpp;	/* 0x4C00 and following */
+	struct radeon_state_atom rb3d_cctl;	/* (4E00) */
+	struct radeon_state_atom bld;	/* blending (4E04) */
+	struct radeon_state_atom cmk;	/* colormask (4E0C) */
+	struct radeon_state_atom blend_color;	/* constant blend color */
+	struct radeon_state_atom rop;	/* ropcntl */
+	struct radeon_state_atom cb;	/* colorbuffer (4E28) */
+	struct radeon_state_atom rb3d_dither_ctl;	/* (4E50) */
+	struct radeon_state_atom rb3d_aaresolve_ctl;	/* (4E88) */
+	struct radeon_state_atom rb3d_discard_src_pixel_lte_threshold;	/* (4E88) I saw it only written on RV350 hardware..  */
+	struct radeon_state_atom zs;	/* zstencil control (4F00) */
+	struct radeon_state_atom zstencil_format;
+	struct radeon_state_atom zb;	/* z buffer (4F20) */
+	struct radeon_state_atom zb_depthclearvalue;	/* (4F28) */
+	struct radeon_state_atom zb_zmask;	/* (4F30) */
+	struct radeon_state_atom zb_hiz_offset;	/* (4F44) */
+	struct radeon_state_atom zb_hiz_pitch;	/* (4F54) */
+
+	struct radeon_state_atom vpi;	/* vp instructions */
+	struct radeon_state_atom vpp;	/* vp parameters */
+	struct radeon_state_atom vps;	/* vertex point size (?) */
+	struct radeon_state_atom vpucp[6];	/* vp user clip plane - 6 */
 	/* 8 texture units */
 	/* the state is grouped by function and not by
 	   texture unit. This makes single unit updates
 	   really awkward - we are much better off
 	   updating the whole thing at once */
 	struct {
-		struct r300_state_atom filter;
-		struct r300_state_atom filter_1;
-		struct r300_state_atom size;
-		struct r300_state_atom format;
-		struct r300_state_atom pitch;
-		struct r300_state_atom offset;
-		struct r300_state_atom chroma_key;
-		struct r300_state_atom border_color;
+		struct radeon_state_atom filter;
+		struct radeon_state_atom filter_1;
+		struct radeon_state_atom size;
+		struct radeon_state_atom format;
+		struct radeon_state_atom pitch;
+		struct radeon_state_atom offset;
+		struct radeon_state_atom chroma_key;
+		struct radeon_state_atom border_color;
 	} tex;
-	struct r300_state_atom txe;	/* tex enable (4104) */
-};
+	struct radeon_state_atom txe;	/* tex enable (4104) */
 
-/**
- * This structure holds the command buffer while it is being constructed.
- *
- * The first batch of commands in the buffer is always the state that needs
- * to be re-emitted when the context is lost. This batch can be skipped
- * otherwise.
- */
-struct r300_cmdbuf {
-	int size;		/* DWORDs allocated for buffer */
-	uint32_t *cmd_buf;
-	int count_used;		/* DWORDs filled so far */
-	int count_reemit;	/* size of re-emission batch */
+	radeonTexObj *textures[R300_MAX_TEXTURE_UNITS];
 };
 
 /**
  * State cache
  */
 
-struct r300_depthbuffer_state {
-	GLfloat scale;
-};
-
-struct r300_stencilbuffer_state {
-	GLboolean hw_stencil;
-};
-
 /* Vertex shader state */
 
 /* Perhaps more if we store programs in vmem? */
@@ -593,73 +399,60 @@ struct r300_stencilbuffer_state {
 #define STATE_R300_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
 #define STATE_R300_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1)
 
-struct r300_vertex_shader_fragment {
-	int length;
-	union {
-		GLuint d[VSF_MAX_FRAGMENT_LENGTH];
-		float f[VSF_MAX_FRAGMENT_LENGTH];
-		GLuint i[VSF_MAX_FRAGMENT_LENGTH];
-	} body;
-};
-
-struct r300_vertex_shader_state {
-	struct r300_vertex_shader_fragment program;
-};
-
-extern int hw_tcl_on;
-
 #define COLOR_IS_RGBA
 #define TAG(x) r300##x
 #include "tnl_dd/t_dd_vertex.h"
 #undef TAG
 
-//#define CURRENT_VERTEX_SHADER(ctx) (ctx->VertexProgram._Current)
-#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->selected_vp)
-
-/* Should but doesnt work */
-//#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->curr_vp)
-
-/* r300_vertex_shader_state and r300_vertex_program should probably be merged together someday.
- * Keeping them them seperate for now should ensure fixed pipeline keeps functioning properly.
- */
-
-struct r300_vertex_program_key {
-	GLuint InputsRead;
-	GLuint OutputsWritten;
-	GLuint OutputsAdded;
-};
-
 struct r300_vertex_program {
+	struct gl_vertex_program *Base;
 	struct r300_vertex_program *next;
-	struct r300_vertex_program_key key;
-	int translated;
 
-	struct r300_vertex_shader_fragment program;
+	struct r300_vertex_program_key {
+		GLuint FpReads;
+		GLuint FogAttr;
+		GLuint WPosAttr;
+	} key;
+	
+	struct r300_vertex_shader_hw_code {
+		int length;
+		union {
+			GLuint d[VSF_MAX_FRAGMENT_LENGTH];
+			float f[VSF_MAX_FRAGMENT_LENGTH];
+		} body;
+	} hw_code;
+
+	GLboolean translated;
+	GLboolean error;
 
 	int pos_end;
 	int num_temporaries;	/* Number of temp vars used by program */
-	int wpos_idx;
 	int inputs[VERT_ATTRIB_MAX];
 	int outputs[VERT_RESULT_MAX];
-	int native;
-	int ref_count;
-	int use_ref_count;
 };
 
 struct r300_vertex_program_cont {
-	struct gl_vertex_program mesa_program;	/* Must be first */
-	struct r300_vertex_shader_fragment params;
+	/* This is the unmodified vertex program mesa provided us with.
+	 * We need to keep it unchanged because we may need to create another
+	 * hw specific vertex program based on this.
+	 */
+	struct gl_vertex_program mesa_program;
+	/* This is the list of hw specific vertex programs derived from mesa_program */
 	struct r300_vertex_program *progs;
 };
 
-#define PFS_MAX_ALU_INST	64
-#define PFS_MAX_TEX_INST	64
-#define PFS_MAX_TEX_INDIRECT 4
-#define PFS_NUM_TEMP_REGS	32
-#define PFS_NUM_CONST_REGS	16
+#define R300_PFS_MAX_ALU_INST	64
+#define R300_PFS_MAX_TEX_INST	32
+#define R300_PFS_MAX_TEX_INDIRECT 4
+#define R300_PFS_NUM_TEMP_REGS	32
+#define R300_PFS_NUM_CONST_REGS	32
 
-struct r300_pfs_compile_state;
+#define R500_PFS_MAX_INST 512
+#define R500_PFS_NUM_TEMP_REGS 128
+#define R500_PFS_NUM_CONST_REGS 256
 
+struct r300_pfs_compile_state;
+struct r500_pfs_compile_state;
 
 /**
  * Stores state that influences the compilation of a fragment program.
@@ -702,7 +495,7 @@ struct r300_fragment_program_node {
 struct r300_fragment_program_code {
 	struct {
 		int length; /**< total # of texture instructions used */
-		GLuint inst[PFS_MAX_TEX_INST];
+		GLuint inst[R300_PFS_MAX_TEX_INST];
 	} tex;
 
 	struct {
@@ -712,7 +505,7 @@ struct r300_fragment_program_code {
 			GLuint inst1;
 			GLuint inst2;
 			GLuint inst3;
-		} inst[PFS_MAX_ALU_INST];
+		} inst[R300_PFS_MAX_ALU_INST];
 	} alu;
 
 	struct r300_fragment_program_node node[4];
@@ -723,53 +516,12 @@ struct r300_fragment_program_code {
 	 * Remember which program register a given hardware constant
 	 * belongs to.
 	 */
-	struct prog_src_register constant[PFS_NUM_CONST_REGS];
+	struct prog_src_register constant[R300_PFS_NUM_CONST_REGS];
 	int const_nr;
 
 	int max_temp_idx;
 };
 
-/**
- * Store everything about a fragment program that is needed
- * to render with that program.
- */
-struct r300_fragment_program {
-	struct gl_fragment_program mesa_program;
-
-	GLboolean translated;
-	GLboolean error;
-
-	struct r300_fragment_program_external_state state;
-	struct r300_fragment_program_code code;
-
-	GLboolean WritesDepth;
-	GLuint optimization;
-};
-
-struct r500_pfs_compile_state;
-
-struct r500_fragment_program_external_state {
-	struct {
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is:
-		 *  0 - GL_LUMINANCE
-		 *  1 - GL_INTENSITY
-		 *  2 - GL_ALPHA
-		 * depending on the depth texture mode.
-		 */
-		GLuint depth_texture_mode : 2;
-
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is (texture_compare_func - GL_NEVER).
-		 * [e.g. if compare function is GL_LEQUAL, this field is 3]
-		 *
-		 * Otherwise, this field is 0.
-		 */
-		GLuint texture_compare_func : 3;
-	} unit[16];
-};
 
 struct r500_fragment_program_code {
 	struct {
@@ -779,7 +531,7 @@ struct r500_fragment_program_code {
 		GLuint inst3;
 		GLuint inst4;
 		GLuint inst5;
-	} inst[512];
+	} inst[R500_PFS_MAX_INST];
 
 	int inst_offset;
 	int inst_end;
@@ -788,94 +540,63 @@ struct r500_fragment_program_code {
 	 * Remember which program register a given hardware constant
 	 * belongs to.
 	 */
-	struct prog_src_register constant[PFS_NUM_CONST_REGS];
+	struct prog_src_register constant[R500_PFS_NUM_CONST_REGS];
 	int const_nr;
 
 	int max_temp_idx;
 };
 
-struct r500_fragment_program {
-	struct gl_fragment_program mesa_program;
+/**
+* Store everything about a fragment program that is needed
+* to render with that program.
+*/
+struct r300_fragment_program {
+	struct gl_program *Base;
 
-	GLcontext *ctx;
 	GLboolean translated;
 	GLboolean error;
 
-	struct r500_fragment_program_external_state state;
-	struct r500_fragment_program_code code;
+	struct r300_fragment_program_external_state state;
+	union rX00_fragment_program_code {
+		struct r300_fragment_program_code r300;
+		struct r500_fragment_program_code r500;
+	} code;
 
 	GLboolean writes_depth;
-
 	GLuint optimization;
-};
-
-#define R300_MAX_AOS_ARRAYS		16
-
-#define REG_COORDS	0
-#define REG_COLOR0	1
-#define REG_TEX0	2
-
-struct r300_state {
-	struct r300_depthbuffer_state depth;
-	struct r300_texture_state texture;
-	int sw_tcl_inputs[VERT_ATTRIB_MAX];
-	struct r300_vertex_shader_state vertex_shader;
-	struct r300_dma_region aos[R300_MAX_AOS_ARRAYS];
-	int aos_count;
 
-	GLuint *Elts;
-	struct r300_dma_region elt_dma;
+	struct r300_fragment_program *next;
 
-	struct r300_dma_region swtcl_dma;
-	DECLARE_RENDERINPUTS(render_inputs_bitset);	/* actual render inputs that R300 was configured for.
-							   They are the same as tnl->render_inputs for fixed pipeline */
+	/* attribute that we are sending the WPOS in */
+	gl_frag_attrib wpos_attr;
+	/* attribute that we are sending the fog coordinate in */
+	gl_frag_attrib fog_attr;
+};
 
-	struct r300_stencilbuffer_state stencil;
+struct r300_fragment_program_cont {
+	/* This is the unmodified fragment program mesa provided us with.
+	 * We need to keep it unchanged because we may need to create another
+	 * hw specific fragment program based on this.
+	 */
+	struct gl_fragment_program Base;
+	/* This is the list of hw specific fragment programs derived from Base */
+	struct r300_fragment_program *progs;
+};
 
+struct r300_fragment_program_compiler {
+	r300ContextPtr r300;
+	struct r300_fragment_program *fp;
+	union rX00_fragment_program_code *code;
+	struct gl_program *program;
 };
 
-#define R300_FALLBACK_NONE 0
-#define R300_FALLBACK_TCL 1
-#define R300_FALLBACK_RAST 2
+#define R300_MAX_AOS_ARRAYS		16
+
 
 /* r300_swtcl.c
  */
 struct r300_swtcl_info {
-   GLuint RenderIndex;
-
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-   GLuint vertex_size;
-
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-   GLuint vertex_attr_count;
-
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-   GLubyte *verts;
-
-   /* Fallback rasterization functions
-    */
-  //   r200_point_func draw_point;
-  //   r200_line_func draw_line;
-  //   r200_tri_func draw_tri;
-
-   GLuint hw_primitive;
-   GLenum render_primitive;
-   GLuint numverts;
-
-   /**
+  /*
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
     */
    GLuint coloroffset;
@@ -884,15 +605,44 @@ struct r300_swtcl_info {
     * Offset of the 3UB specular color data within a hardware (swtcl) vertex.
     */
    GLuint specoffset;
+};
 
-   /**
-    * Should Mesa project vertex data or will the hardware do it?
-    */
-   GLboolean needproj;
+struct r300_vtable {
+	void (* SetupRSUnit)(GLcontext *ctx);
+	void (* SetupFragmentShaderTextures)(GLcontext *ctx, int *tmu_mappings);
+	GLboolean (* BuildFragmentProgramHwCode)(struct r300_fragment_program_compiler *compiler);
+	void (* FragmentProgramDump)(union rX00_fragment_program_code *code);
+	void (* SetupPixelShader)(GLcontext *ctx);
+};
 
-   struct r300_dma_region indexed_verts;
+struct r300_vertex_buffer {
+	struct vertex_attribute {
+		/* generic */
+		GLubyte element;
+		GLvoid *data;
+		GLboolean free_needed;
+		GLuint stride;
+		GLuint dwords;
+		GLubyte size; /* number of components */
+
+		/* hw specific */
+		uint32_t data_type:4;
+		uint32_t dst_loc:5;
+		uint32_t _signed:1;
+		uint32_t normalize:1;
+		uint32_t swizzle:12;
+		uint32_t write_mask:4;
+	} attribs[VERT_ATTRIB_MAX];
+
+	GLubyte num_attribs;
 };
 
+struct r300_index_buffer {
+	GLvoid *ptr;
+	GLboolean is_32bit;
+	GLboolean free_needed;
+	GLuint count;
+};
 
 /**
  * \brief R300 context structure.
@@ -900,46 +650,34 @@ struct r300_swtcl_info {
 struct r300_context {
 	struct radeon_context radeon;	/* parent class, must be first */
 
+	struct r300_vtable vtbl;
+
 	struct r300_hw_state hw;
-	struct r300_cmdbuf cmdbuf;
-	struct r300_state state;
-	struct gl_vertex_program *curr_vp;
+
 	struct r300_vertex_program *selected_vp;
+	struct r300_fragment_program *selected_fp;
 
 	/* Vertex buffers
 	 */
-	struct r300_dma dma;
-	GLboolean save_on_next_unlock;
-	GLuint NewGLState;
-
-	/* Texture object bookkeeping
-	 */
-	unsigned nr_heaps;
-	driTexHeap *texture_heaps[RADEON_NR_TEX_HEAPS];
-	driTextureObject swapped;
-	int texture_depth;
-	float initialMaxAnisotropy;
-
-	/* Clientdata textures;
-	 */
-	GLuint prefer_gart_client_texturing;
-
-#ifdef USER_BUFFERS
-	struct r300_memory_manager *rmm;
-#endif
-
 	GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
 	GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
 
-	GLboolean disable_lowimpact_fallback;
-
-	DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
+	struct r300_options {
+		uint32_t conformance_mode:1;
+		uint32_t hw_tcl_enabled:1;
+		uint32_t s3tc_force_enabled:1;
+		uint32_t s3tc_force_disabled:1;
+		uint32_t stencil_two_side_disabled:1;
+	} options;
+	
 	struct r300_swtcl_info swtcl;
-};
+	struct r300_vertex_buffer vbuf;
+	struct r300_index_buffer ind_buf;
+	GLboolean vap_flush_needed;
 
-struct r300_buffer_object {
-	struct gl_buffer_object mesa_obj;
-	int id;
+	uint32_t fallback;
+
+	DECLARE_RENDERINPUTS(render_inputs_bitset);
 };
 
 #define R300_CONTEXT(ctx)		((r300ContextPtr)(ctx->DriverCtx))
@@ -949,15 +687,13 @@ extern GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 				   __DRIcontextPrivate * driContextPriv,
 				   void *sharedContextPrivate);
 
-extern void r300SelectVertexShader(r300ContextPtr r300);
 extern void r300InitShaderFuncs(struct dd_function_table *functions);
-extern int r300VertexProgUpdateParams(GLcontext * ctx,
-				      struct r300_vertex_program_cont *vp,
-				      float *dst);
-
-#define RADEON_D_CAPTURE 0
-#define RADEON_D_PLAYBACK 1
-#define RADEON_D_PLAYBACK_RAW 2
-#define RADEON_D_T 3
+
+extern void r300InitShaderFunctions(r300ContextPtr r300);
+
+extern void r300InitDraw(GLcontext *ctx);
+
+#define r300PackFloat32 radeonPackFloat32
+#define r300PackFloat24 radeonPackFloat24
 
 #endif				/* __R300_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
new file mode 100644
index 0000000000..9769ff5399
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -0,0 +1,501 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Maciej Cencora
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHOR(S) AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <stdlib.h>
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/state.h"
+#include "main/api_validate.h"
+#include "main/enums.h"
+
+#include "r300_reg.h"
+#include "r300_context.h"
+#include "r300_emit.h"
+#include "r300_render.h"
+#include "r300_state.h"
+#include "r300_tex.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
+#include "vbo/vbo_context.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+
+static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_index_buffer *ind_buf = &r300->ind_buf;
+	GLvoid *src_ptr;
+
+	if (!mesa_ind_buf) {
+		ind_buf->ptr = NULL;
+		return;
+	}
+
+	ind_buf->count = mesa_ind_buf->count;
+	if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
+		bo[*nr_bo] = mesa_ind_buf->obj;
+		(*nr_bo)++;
+		ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+		assert(mesa_ind_buf->obj->Pointer != NULL);
+	}
+	src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
+
+	if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) {
+		GLubyte *in = (GLubyte *)src_ptr;
+		GLuint *out = _mesa_malloc(sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1));
+		int i;
+
+		ind_buf->ptr = out;
+
+		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
+			*out++ = in[i] | in[i + 1] << 16;
+		}
+
+		if (i < mesa_ind_buf->count) {
+			*out++ = in[i];
+		}
+
+		ind_buf->free_needed = GL_TRUE;
+		ind_buf->is_32bit = GL_FALSE;
+	} else if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) {
+#if MESA_BIG_ENDIAN
+		GLushort *in = (GLushort *)src_ptr;
+		GLuint *out = _mesa_malloc(sizeof(GLushort) *
+					   ((mesa_ind_buf->count + 1) & ~1));
+		int i;
+
+		ind_buf->ptr = out;
+
+		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
+			*out++ = in[i] | in[i + 1] << 16;
+		}
+
+		if (i < mesa_ind_buf->count) {
+			*out++ = in[i];
+		}
+
+		ind_buf->free_needed = GL_TRUE;
+#else
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+#endif
+		ind_buf->is_32bit = GL_FALSE;
+	} else {
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+		ind_buf->is_32bit = GL_TRUE;
+	}
+}
+
+static int getTypeSize(GLenum type)
+{
+	switch (type) {
+		case GL_DOUBLE:
+			return sizeof(GLdouble);
+		case GL_FLOAT:
+			return sizeof(GLfloat);
+		case GL_INT:
+			return sizeof(GLint);
+		case GL_UNSIGNED_INT:
+			return sizeof(GLuint);
+		case GL_SHORT:
+			return sizeof(GLshort);
+		case GL_UNSIGNED_SHORT:
+			return sizeof(GLushort);
+		case GL_BYTE:
+			return sizeof(GLbyte);
+		case GL_UNSIGNED_BYTE:
+			return sizeof(GLubyte);
+		default:
+			assert(0);
+			return 0;
+	}
+}
+
+#define CONVERT( TYPE, MACRO ) do {		\
+	GLuint i, j, sz;				\
+	sz = input->Size;				\
+	if (input->Normalized) {			\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = MACRO(*in);		\
+				in++;				\
+			}					\
+			src_ptr += stride;			\
+		}						\
+	} else {					\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = (GLfloat)(*in);		\
+				in++;				\
+			}					\
+			src_ptr += stride;			\
+		}						\
+	}						\
+} while (0)
+
+static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	struct vertex_attribute r300_attr;
+	const void *src_ptr;
+	GLenum type;
+	GLuint stride;
+
+	if (input->BufferObj->Name) {
+		if (!input->BufferObj->Pointer) {
+			bo[*nr_bo] = input->BufferObj;
+			(*nr_bo)++;
+			ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
+			assert(input->BufferObj->Pointer != NULL);
+		}
+
+		src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
+	} else
+		src_ptr = input->Ptr;
+
+	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
+
+	if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT ||
+#if MESA_BIG_ENDIAN
+	    getTypeSize(input->Type) != 4 ||
+#endif
+	    stride < 4) {
+		if (RADEON_DEBUG & DEBUG_FALLBACKS) {
+			fprintf(stderr, "%s: Converting vertex attributes, attribute data format %x,", __FUNCTION__, input->Type);
+			fprintf(stderr, "stride %d, components %d\n", stride, input->Size);
+		}
+
+		GLfloat *dst_ptr, *tmp;
+
+		/* Convert value for first element only */
+		if (input->StrideB == 0)
+			count = 1;
+
+		tmp = dst_ptr = _mesa_malloc(sizeof(GLfloat) * input->Size * count);
+
+		switch (input->Type) {
+			case GL_DOUBLE:
+				CONVERT(GLdouble, (GLfloat));
+				break;
+			case GL_UNSIGNED_INT:
+				CONVERT(GLuint, UINT_TO_FLOAT);
+				break;
+			case GL_INT:
+				CONVERT(GLint, INT_TO_FLOAT);
+				break;
+			case GL_UNSIGNED_SHORT:
+				CONVERT(GLushort, USHORT_TO_FLOAT);
+				break;
+			case GL_SHORT:
+				CONVERT(GLshort, SHORT_TO_FLOAT);
+				break;
+			case GL_UNSIGNED_BYTE:
+				assert(input->Format != GL_BGRA);
+				CONVERT(GLubyte, UBYTE_TO_FLOAT);
+				break;
+			case GL_BYTE:
+				CONVERT(GLbyte, BYTE_TO_FLOAT);
+				break;
+			default:
+				assert(0);
+				break;
+		}
+
+		type = GL_FLOAT;
+		r300_attr.free_needed = GL_TRUE;
+		r300_attr.data = tmp;
+		if (input->StrideB == 0) {
+			r300_attr.stride = 0;
+		} else {
+			r300_attr.stride = sizeof(GLfloat) * input->Size;
+		}
+		r300_attr.dwords = input->Size;
+	} else {
+		type = input->Type;
+		r300_attr.free_needed = GL_FALSE;
+		r300_attr.data = (GLvoid *)src_ptr;
+		r300_attr.stride = input->StrideB;
+		r300_attr.dwords = (getTypeSize(type) * input->Size  + 3)/ 4;
+	}
+
+	r300_attr.size = input->Size;
+	r300_attr.element = attr;
+	r300_attr.dst_loc = vbuf->num_attribs;
+
+	switch (type) {
+		case GL_FLOAT:
+			switch (input->Size) {
+				case 1: r300_attr.data_type = R300_DATA_TYPE_FLOAT_1; break;
+				case 2: r300_attr.data_type = R300_DATA_TYPE_FLOAT_2; break;
+				case 3: r300_attr.data_type = R300_DATA_TYPE_FLOAT_3; break;
+				case 4: r300_attr.data_type = R300_DATA_TYPE_FLOAT_4; break;
+			}
+			r300_attr._signed = 0;
+			r300_attr.normalize = 0;
+			break;
+		case GL_SHORT:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			switch (input->Size) {
+				case 1:
+				case 2:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+					break;
+				case 3:
+				case 4:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+					break;
+			}
+			break;
+		case GL_BYTE:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+		case GL_UNSIGNED_SHORT:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			switch (input->Size) {
+				case 1:
+				case 2:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+					break;
+				case 3:
+				case 4:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+					break;
+			}
+			break;
+		case GL_UNSIGNED_BYTE:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			if (input->Format == GL_BGRA)
+				r300_attr.data_type = R300_DATA_TYPE_D3DCOLOR;
+			else
+				r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+
+		default:
+		case GL_DOUBLE:
+		case GL_INT:
+		case GL_UNSIGNED_INT:
+			assert(0);
+			break;
+	}
+
+	switch (input->Size) {
+		case 4:
+			r300_attr.swizzle = SWIZZLE_XYZW;
+			break;
+		case 3:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			break;
+		case 2:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+		case 1:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+	}
+
+	r300_attr.write_mask = MASK_XYZW;
+
+	vbuf->attribs[vbuf->num_attribs] = r300_attr;
+	++vbuf->num_attribs;
+}
+
+static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+
+	{
+		int i, tmp;
+
+		tmp = r300->selected_vp->Base->Base.InputsRead;
+		i = 0;
+		vbuf->num_attribs = 0;
+		while (tmp) {
+			/* find first enabled bit */
+			while (!(tmp & 1)) {
+				tmp >>= 1;
+				++i;
+			}
+
+			r300TranslateAttrib(ctx, i, count, arrays[i], bo, nr_bo);
+
+			tmp >>= 1;
+			++i;
+		}
+	}
+
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback)
+		return;
+
+	{
+		int i;
+
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i],
+						vbuf->attribs[i].data, vbuf->attribs[i].dwords,
+						vbuf->attribs[i].stride, count);
+		}
+
+		r300->radeon.tcl.aos_count = vbuf->num_attribs;
+	}
+}
+
+static void r300FreeData(GLcontext *ctx, struct gl_buffer_object **bo, GLuint nr_bo)
+{
+	{
+		struct r300_vertex_buffer *vbuf = &R300_CONTEXT(ctx)->vbuf;
+		int i;
+
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			if (vbuf->attribs[i].free_needed)
+				_mesa_free(vbuf->attribs[i].data);
+		}
+	}
+
+	{
+		struct r300_index_buffer *ind_buf = &R300_CONTEXT(ctx)->ind_buf;
+		if (ind_buf->free_needed)
+			_mesa_free(ind_buf->ptr);
+	}
+
+	{
+		int i;
+
+		for (i = 0; i < nr_bo; ++i) {
+			ctx->Driver.UnmapBuffer(ctx, 0, bo[i]);
+		}
+	}
+}
+
+static GLboolean r300TryDrawPrims(GLcontext *ctx,
+					 const struct gl_client_array *arrays[],
+					 const struct _mesa_prim *prim,
+					 GLuint nr_prims,
+					 const struct _mesa_index_buffer *ib,
+					 GLuint min_index,
+					 GLuint max_index )
+{
+	struct r300_context *r300 = R300_CONTEXT(ctx);
+	struct gl_buffer_object *bo[VERT_ATTRIB_MAX+1];
+	GLuint i, nr_bo = 0;
+
+	if (ctx->NewState)
+		_mesa_update_state( ctx );
+
+	if (r300->options.hw_tcl_enabled)
+		_tnl_UpdateFixedFunctionProgram(ctx);
+
+	r300UpdateShaders(r300);
+
+	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
+
+	r300FixupIndexBuffer(ctx, ib, bo, &nr_bo);
+
+	/* ensure we have the cmd buf space in advance to cover
+ 	 * the state + DMA AOS pointers */
+	rcommonEnsureCmdBufSpace(&r300->radeon,
+                           r300->radeon.hw.max_state_size + (50*sizeof(int)),
+                           __FUNCTION__);
+
+	r300SetVertexFormat(ctx, arrays, max_index + 1, bo, &nr_bo);
+
+	if (r300->fallback)
+		return GL_FALSE;
+
+	r300SetupVAP(ctx, r300->selected_vp->Base->Base.InputsRead, r300->selected_vp->Base->Base.OutputsWritten);
+
+	r300UpdateShaderStates(r300);
+
+	r300EmitCacheFlush(r300);
+	radeonEmitState(&r300->radeon);
+
+	for (i = 0; i < nr_prims; ++i) {
+		r300RunRenderPrimitive(ctx, prim[i].start, prim[i].start + prim[i].count, prim[i].mode);
+	}
+
+	r300EmitCacheFlush(r300);
+
+	radeonReleaseArrays(ctx, ~0);
+
+	r300FreeData(ctx, bo, nr_bo);
+
+	return GL_TRUE;
+}
+
+static void r300DrawPrims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+	struct split_limits limits;
+	GLboolean retval;
+
+	if (ib)
+		limits.max_verts = 0xffffffff;
+	else
+		limits.max_verts = 65535;
+
+	limits.max_indices = 65535;
+	limits.max_vb_size = 1024*1024;
+
+	if (min_index) {
+		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
+		return;
+	}
+	if ((ib && ib->count > 65535)) {
+		vbo_split_prims (ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims, &limits);
+		return;
+	}
+
+	/* Make an attempt at drawing */
+	retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+
+	/* If failed run tnl pipeline - it should take care of fallbacks */
+	if (!retval)
+		_tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
+
+void r300InitDraw(GLcontext *ctx)
+{
+	struct vbo_context *vbo = vbo_context(ctx);
+
+	vbo->draw_prims = r300DrawPrims;
+}
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index 28c3157427..feb3370f37 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -31,6 +31,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \file
  *
  * \author Keith Whitwell <keith@tungstengraphics.com>
+ * \author Maciej Cencora <m.cencora@gmail.com>
  */
 
 #include "main/glheader.h"
@@ -46,222 +47,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_context.h"
 
 #include "r300_context.h"
-#include "radeon_ioctl.h"
 #include "r300_state.h"
 #include "r300_emit.h"
 #include "r300_ioctl.h"
-
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
-
-#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
-    SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
-    SWIZZLE_Z != R300_INPUT_ROUTE_SELECT_Z || \
-    SWIZZLE_W != R300_INPUT_ROUTE_SELECT_W || \
-    SWIZZLE_ZERO != R300_INPUT_ROUTE_SELECT_ZERO || \
-    SWIZZLE_ONE != R300_INPUT_ROUTE_SELECT_ONE
-#error Cannot change these!
-#endif
-
-#define DEBUG_ALL DEBUG_VERTS
-
-#if defined(USE_X86_ASM)
-#define COPY_DWORDS( dst, src, nr )					\
-do {									\
-	int __tmp;							\
-	__asm__ __volatile__( "rep ; movsl"				\
-			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
-			      : "0" (nr),				\
-			        "D" ((long)dst),			\
-			        "S" ((long)src) );			\
-} while (0)
-#else
-#define COPY_DWORDS( dst, src, nr )		\
-do {						\
-   int j;					\
-   for ( j = 0 ; j < nr ; j++ )			\
-      dst[j] = ((int *)src)[j];			\
-   dst += nr;					\
-} while (0)
-#endif
-
-static void r300EmitVec4(GLcontext * ctx, struct r300_dma_region *rvb,
-			 GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 4)
-		COPY_DWORDS(out, data, count);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out++;
-			data += stride;
-		}
-}
-
-static void r300EmitVec8(GLcontext * ctx, struct r300_dma_region *rvb,
-			 GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 8)
-		COPY_DWORDS(out, data, count * 2);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out[1] = *(int *)(data + 4);
-			out += 2;
-			data += stride;
-		}
-}
-
-static void r300EmitVec12(GLcontext * ctx, struct r300_dma_region *rvb,
-			  GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 12)
-		COPY_DWORDS(out, data, count * 3);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out[1] = *(int *)(data + 4);
-			out[2] = *(int *)(data + 8);
-			out += 3;
-			data += stride;
-		}
-}
-
-static void r300EmitVec16(GLcontext * ctx, struct r300_dma_region *rvb,
-			  GLvoid * data, int stride, int count)
-{
-	int i;
-	int *out = (int *)(rvb->address + rvb->start);
-
-	if (RADEON_DEBUG & DEBUG_VERTS)
-		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-			__FUNCTION__, count, stride, (void *)out, (void *)data);
-
-	if (stride == 16)
-		COPY_DWORDS(out, data, count * 4);
-	else
-		for (i = 0; i < count; i++) {
-			out[0] = *(int *)data;
-			out[1] = *(int *)(data + 4);
-			out[2] = *(int *)(data + 8);
-			out[3] = *(int *)(data + 12);
-			out += 4;
-			data += stride;
-		}
-}
-
-static void r300EmitVec(GLcontext * ctx, struct r300_dma_region *rvb,
-			GLvoid * data, int size, int stride, int count)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (stride == 0) {
-		r300AllocDmaRegion(rmesa, rvb, size * 4, 4);
-		count = 1;
-		rvb->aos_offset = GET_START(rvb);
-		rvb->aos_stride = 0;
-	} else {
-		r300AllocDmaRegion(rmesa, rvb, size * count * 4, 4);
-		rvb->aos_offset = GET_START(rvb);
-		rvb->aos_stride = size;
-	}
-
-	switch (size) {
-	case 1:
-		r300EmitVec4(ctx, rvb, data, stride, count);
-		break;
-	case 2:
-		r300EmitVec8(ctx, rvb, data, stride, count);
-		break;
-	case 3:
-		r300EmitVec12(ctx, rvb, data, stride, count);
-		break;
-	case 4:
-		r300EmitVec16(ctx, rvb, data, stride, count);
-		break;
-	default:
-		assert(0);
-		break;
-	}
-}
-
-#define DW_SIZE(x) ((inputs[tab[(x)]] << R300_DST_VEC_LOC_SHIFT) |	\
-		    (attribptr[tab[(x)]]->size - 1) << R300_DATA_TYPE_0_SHIFT)
-
-GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
-				 int *inputs, GLint * tab, GLuint nr)
-{
-	GLuint i, dw;
-
-	/* type, inputs, stop bit, size */
-	for (i = 0; i < nr; i += 2) {
-		/* make sure input is valid, would lockup the gpu */
-		assert(inputs[tab[i]] != -1);
-		dw = (R300_SIGNED | DW_SIZE(i));
-		if (i + 1 == nr) {
-			dw |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
-		} else {
-			assert(inputs[tab[i + 1]] != -1);
-			dw |= (R300_SIGNED |
-			       DW_SIZE(i + 1)) << R300_DATA_TYPE_1_SHIFT;
-			if (i + 2 == nr) {
-				dw |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
-			}
-		}
-		dst[i >> 1] = dw;
-	}
-
-	return (nr + 1) >> 1;
-}
-
-static GLuint r300VAPInputRoute1Swizzle(int swizzle[4])
-{
-	return (swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) |
-	    (swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) |
-	    (swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) |
-	    (swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT);
-}
-
-GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr)
-{
-	GLuint i, dw;
-
-	for (i = 0; i < nr; i += 2) {
-		dw = (r300VAPInputRoute1Swizzle(swizzle[i]) |
-		      ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
-			R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE0_SHIFT;
-		if (i + 1 < nr) {
-			dw |= (r300VAPInputRoute1Swizzle(swizzle[i + 1]) |
-			       ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
-				 R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
-		}
-		dst[i >> 1] = dw;
-	}
-
-	return (nr + 1) >> 1;
-}
+#include "r300_render.h"
+#include "r300_swtcl.h"
 
 GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
 {
@@ -272,7 +62,6 @@ GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
 
 GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	GLuint i, vic_1 = 0;
 
 	if (InputsRead & (1 << VERT_ATTRIB_POS))
@@ -284,281 +73,103 @@ GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 	if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
 		vic_1 |= R300_INPUT_CNTL_COLOR;
 
-	rmesa->state.texture.tc_count = 0;
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
 		if (InputsRead & (1 << (VERT_ATTRIB_TEX0 + i))) {
-			rmesa->state.texture.tc_count++;
 			vic_1 |= R300_INPUT_CNTL_TC0 << i;
 		}
 
 	return vic_1;
 }
 
-GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes)
 {
 	GLuint ret = 0;
 
-	if (OutputsWritten & (1 << VERT_RESULT_HPOS))
+	if (vp_writes & (1 << VERT_RESULT_HPOS))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_COL0))
+	if (vp_writes & (1 << VERT_RESULT_COL0))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_COL1))
+	if (vp_writes & (1 << VERT_RESULT_COL1))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_BFC0)
-	    || OutputsWritten & (1 << VERT_RESULT_BFC1))
-		ret |=
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT |
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
+	/* Two sided lighting works only if all 4 colors are written */
+	if (vp_writes & (1 << VERT_RESULT_BFC0) || vp_writes & (1 << VERT_RESULT_BFC1))
+		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
+			   R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_PSIZ))
+	if (vp_writes & (1 << VERT_RESULT_PSIZ))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
 
 	return ret;
 }
 
-GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes)
 {
 	GLuint i, ret = 0, first_free_texcoord = 0;
 
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (OutputsWritten & (1 << (VERT_RESULT_TEX0 + i))) {
-			ret |= (4 << (3 * i));
+		if (vp_writes & (1 << (VERT_RESULT_TEX0 + i))) {
+			ret |= (4 << (3 * first_free_texcoord));
 			++first_free_texcoord;
 		}
 	}
 
-	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
-		if (first_free_texcoord > 8) {
-			fprintf(stderr, "\tout of free texcoords to write fog coord\n");
-			_mesa_exit(-1);
-		}
-		ret |= 4 << (3 * first_free_texcoord);
+	if (first_free_texcoord > 8) {
+		fprintf(stderr, "\tout of free texcoords\n");
+		_mesa_exit(-1);
 	}
 
 	return ret;
 }
 
-/* Emit vertex data to GART memory
- * Route inputs to the vertex processor
- * This function should never return R300_FALLBACK_TCL when using software tcl.
- */
-int r300EmitArrays(GLcontext * ctx)
+GLboolean r300EmitArrays(GLcontext * ctx)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
-	GLuint nr;
-	GLuint count = vb->Count;
-	GLuint i;
-	GLuint InputsRead = 0, OutputsWritten = 0;
-	int *inputs = NULL;
-	int vir_inputs[VERT_ATTRIB_MAX];
-	GLint tab[VERT_ATTRIB_MAX];
-	int swizzle[VERT_ATTRIB_MAX][4];
-	struct r300_vertex_program *prog =
-	    (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-
-	if (hw_tcl_on) {
-		inputs = prog->inputs;
-		InputsRead = prog->key.InputsRead;
-		OutputsWritten = prog->key.OutputsWritten;
-	} else {
-		inputs = rmesa->state.sw_tcl_inputs;
-
-		DECLARE_RENDERINPUTS(render_inputs_bitset);
-		RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset);
-
-		vb->AttribPtr[VERT_ATTRIB_POS] = vb->ClipPtr;
-
-		assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS));
-		assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_NORMAL) == 0);
-		//assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR0));
-
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS)) {
-			InputsRead |= 1 << VERT_ATTRIB_POS;
-			OutputsWritten |= 1 << VERT_RESULT_HPOS;
-		}
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	GLuint InputsRead, OutputsWritten;
 
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR0)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR0;
-			OutputsWritten |= 1 << VERT_RESULT_COL0;
-		}
+	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
 
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR1)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR1;
-			OutputsWritten |= 1 << VERT_RESULT_COL1;
-		}
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback & R300_RASTER_FALLBACK_MASK)
+		return GL_FALSE;
 
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-			if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_TEX(i))) {
-				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
-				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-			}
-		}
+	{
+		struct vertex_buffer *mesa_vb = &TNL_CONTEXT(ctx)->vb;
+		GLuint attr, i;
 
-		for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-			if (InputsRead & (1 << i)) {
-				inputs[i] = nr++;
-			} else {
-				inputs[i] = -1;
-			}
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			attr = vbuf->attribs[i].element;
+			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i], mesa_vb->AttribPtr[attr]->data,
+					mesa_vb->AttribPtr[attr]->size, mesa_vb->AttribPtr[attr]->stride, mesa_vb->Count);
 		}
 
-		/* Fixed, apply to vir0 only */
-		memcpy(vir_inputs, inputs, VERT_ATTRIB_MAX * sizeof(int));
-		inputs = vir_inputs;
-		if (InputsRead & VERT_ATTRIB_POS)
-			inputs[VERT_ATTRIB_POS] = 0;
-		if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
-			inputs[VERT_ATTRIB_COLOR0] = 2;
-		if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
-			inputs[VERT_ATTRIB_COLOR1] = 3;
-		for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
-			if (InputsRead & (1 << i))
-				inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
-
-		RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
-	}
+		r300->radeon.tcl.aos_count = vbuf->num_attribs;
 
-	assert(InputsRead);
-	assert(OutputsWritten);
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			tab[nr++] = i;
-		}
+		/* Fill index buffer info */
+		r300->ind_buf.ptr = mesa_vb->Elts;
+		r300->ind_buf.is_32bit = GL_TRUE;
+		r300->ind_buf.free_needed = GL_FALSE;
 	}
 
-	if (nr > R300_MAX_AOS_ARRAYS) {
-		return R300_FALLBACK_TCL;
-	}
-
-	for (i = 0; i < nr; i++) {
-		int ci, fix, found = 0;
-
-		swizzle[i][0] = SWIZZLE_ZERO;
-		swizzle[i][1] = SWIZZLE_ZERO;
-		swizzle[i][2] = SWIZZLE_ZERO;
-		swizzle[i][3] = SWIZZLE_ONE;
-
-		for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
-			swizzle[i][ci] = ci;
-		}
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
-		if (r300IsGartMemory(rmesa, vb->AttribPtr[tab[i]]->data, 4)) {
-			if (vb->AttribPtr[tab[i]]->stride % 4) {
-				return R300_FALLBACK_TCL;
-			}
-			rmesa->state.aos[i].address = (void *)(vb->AttribPtr[tab[i]]->data);
-			rmesa->state.aos[i].start = 0;
-			rmesa->state.aos[i].aos_offset = r300GartOffsetFromVirtual(rmesa, vb->AttribPtr[tab[i]]->data);
-			rmesa->state.aos[i].aos_stride = vb->AttribPtr[tab[i]]->stride / 4;
-			rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size;
-		} else {
-			r300EmitVec(ctx, &rmesa->state.aos[i],
-				    vb->AttribPtr[tab[i]]->data,
-				    vb->AttribPtr[tab[i]]->size,
-				    vb->AttribPtr[tab[i]]->stride, count);
-		}
-
-		rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size;
-
-		for (fix = 0; fix <= 4 - vb->AttribPtr[tab[i]]->size; fix++) {
-			if ((rmesa->state.aos[i].aos_offset - _mesa_sizeof_type(GL_FLOAT) * fix) % 4) {
-				continue;
-			}
-			found = 1;
-			break;
-		}
-
-		if (found) {
-			if (fix > 0) {
-				WARN_ONCE("Feeling lucky?\n");
-			}
-			rmesa->state.aos[i].aos_offset -= _mesa_sizeof_type(GL_FLOAT) * fix;
-			for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
-				swizzle[i][ci] += fix;
-			}
-		} else {
-			WARN_ONCE
-			    ("Cannot handle offset %x with stride %d, comp %d\n",
-			     rmesa->state.aos[i].aos_offset,
-			     rmesa->state.aos[i].aos_stride,
-			     vb->AttribPtr[tab[i]]->size);
-			return R300_FALLBACK_TCL;
-		}
-	}
-
-	/* Setup INPUT_ROUTE. */
-	R300_STATECHANGE(rmesa, vir[0]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
-	    r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
-			       vb->AttribPtr, inputs, tab, nr);
-	R300_STATECHANGE(rmesa, vir[1]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
-	    r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
-			       nr);
-
-	/* Setup INPUT_CNTL. */
-	R300_STATECHANGE(rmesa, vic);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-
-	/* Setup OUTPUT_VTX_FMT. */
-	R300_STATECHANGE(rmesa, vof);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] =
-	    r300VAPOutputCntl0(ctx, OutputsWritten);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] =
-	    r300VAPOutputCntl1(ctx, OutputsWritten);
-
-	rmesa->state.aos_count = nr;
-
-	return R300_FALLBACK_NONE;
-}
-
-#ifdef USER_BUFFERS
-void r300UseArrays(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int i;
-
-	if (rmesa->state.elt_dma.buf)
-		r300_mem_use(rmesa, rmesa->state.elt_dma.buf->id);
-
-	for (i = 0; i < rmesa->state.aos_count; i++) {
-		if (rmesa->state.aos[i].buf)
-			r300_mem_use(rmesa, rmesa->state.aos[i].buf->id);
-	}
-}
-#endif
-
-void r300ReleaseArrays(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int i;
-
-	r300ReleaseDmaRegion(rmesa, &rmesa->state.elt_dma, __FUNCTION__);
-	for (i = 0; i < rmesa->state.aos_count; i++) {
-		r300ReleaseDmaRegion(rmesa, &rmesa->state.aos[i], __FUNCTION__);
-	}
+	return GL_TRUE;
 }
 
 void r300EmitCacheFlush(r300ContextPtr rmesa)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-
-	drm_radeon_cmd_header_t *cmd = NULL;
-
-	reg_start(R300_RB3D_DSTCACHE_CTLSTAT, 0);
-	e32(R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
-	    R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
-
-	reg_start(R300_ZB_ZCACHE_CTLSTAT, 0);
-	e32(R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
-	    R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+	BATCH_LOCALS(&rmesa->radeon);
+
+	BEGIN_BATCH_NO_AUTOSTATE(4);
+	OUT_BATCH_REGVAL(R300_RB3D_DSTCACHE_CTLSTAT,
+		R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
+		R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
+	OUT_BATCH_REGVAL(R300_ZB_ZCACHE_CTLSTAT,
+		R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+		R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+	END_BATCH();
+	COMMIT_BATCH();
 }
diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h
index 89d738339f..3f8c60ffae 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.h
+++ b/src/mesa/drivers/dri/r300/r300_emit.h
@@ -44,28 +44,31 @@
 #include "r300_cmdbuf.h"
 #include "radeon_reg.h"
 
-/* TODO: move these defines (and the ones from DRM) into r300_reg.h and sync up
- * with DRM */
-#define CP_PACKET0(reg, n)	(RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2))
-#define CP_PACKET3( pkt, n )						\
-	(RADEON_CP_PACKET3 | (pkt) | ((n) << 16))
-
-static INLINE uint32_t cmdpacket0(int reg, int count)
+static INLINE uint32_t cmdpacket0(struct radeon_screen *rscrn,
+                                  int reg, int count)
 {
-	drm_r300_cmd_header_t cmd;
-
-	cmd.packet0.cmd_type = R300_CMD_PACKET0;
-	cmd.packet0.count = count;
-	cmd.packet0.reghi = ((unsigned int)reg & 0xFF00) >> 8;
-	cmd.packet0.reglo = ((unsigned int)reg & 0x00FF);
-
-	return cmd.u;
+    if (!rscrn->kernel_mm) {
+	    drm_r300_cmd_header_t cmd;
+
+	cmd.u = 0;
+    	cmd.packet0.cmd_type = R300_CMD_PACKET0;
+	    cmd.packet0.count = count;
+    	cmd.packet0.reghi = ((unsigned int)reg & 0xFF00) >> 8;
+	    cmd.packet0.reglo = ((unsigned int)reg & 0x00FF);
+
+    	return cmd.u;
+    }
+    if (count) {
+        return CP_PACKET0(reg, count - 1);
+    }
+    return CP_PACKET2;
 }
 
-static INLINE uint32_t cmdvpu(int addr, int count)
+static INLINE uint32_t cmdvpu(struct radeon_screen *rscrn, int addr, int count)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.vpu.cmd_type = R300_CMD_VPU;
 	cmd.vpu.count = count;
 	cmd.vpu.adrhi = ((unsigned int)addr & 0xFF00) >> 8;
@@ -74,10 +77,12 @@ static INLINE uint32_t cmdvpu(int addr, int count)
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdr500fp(int addr, int count, int type, int clamp)
+static INLINE uint32_t cmdr500fp(struct radeon_screen *rscrn,
+                                 int addr, int count, int type, int clamp)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.r500fp.cmd_type = R300_CMD_R500FP;
 	cmd.r500fp.count = count;
 	cmd.r500fp.adrhi_flags = ((unsigned int)addr & 0x100) >> 8;
@@ -88,181 +93,139 @@ static INLINE uint32_t cmdr500fp(int addr, int count, int type, int clamp)
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdpacket3(int packet)
+static INLINE uint32_t cmdpacket3(struct radeon_screen *rscrn, int packet)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.packet3.cmd_type = R300_CMD_PACKET3;
 	cmd.packet3.packet = packet;
 
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdcpdelay(unsigned short count)
+static INLINE uint32_t cmdcpdelay(struct radeon_screen *rscrn,  
+                                  unsigned short count)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
+
 	cmd.delay.cmd_type = R300_CMD_CP_DELAY;
 	cmd.delay.count = count;
 
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdwait(unsigned char flags)
+static INLINE uint32_t cmdwait(struct radeon_screen *rscrn,
+                               unsigned char flags)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.wait.cmd_type = R300_CMD_WAIT;
 	cmd.wait.flags = flags;
 
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdpacify(void)
+static INLINE uint32_t cmdpacify(struct radeon_screen *rscrn)
 {
 	drm_r300_cmd_header_t cmd;
 
+	cmd.u = 0;
 	cmd.header.cmd_type = R300_CMD_END3D;
 
 	return cmd.u;
 }
 
 /**
- * Prepare to write a register value to register at address reg.
- * If num_extra > 0 then the following extra values are written
- * to registers with address +4, +8 and so on..
- */
-#define reg_start(reg, num_extra)					\
-	do {								\
-		int _n;							\
-		_n=(num_extra);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+2),				\
-					__FUNCTION__);			\
-		cmd_reserved=_n+2;					\
-		cmd_written=1;						\
-		cmd[0].i=cmdpacket0((reg), _n+1);			\
-	} while (0);
-
-/**
- * Emit GLuint freestyle
+ * Write the header of a packet3 to the command buffer.
+ * Outputs 2 dwords and expects (num_extra+1) additional dwords afterwards.
  */
-#define e32(dword)							\
-	do {								\
-		if(cmd_written<cmd_reserved) {				\
-			cmd[cmd_written].i=(dword);			\
-			cmd_written++;					\
-		} else {						\
-			fprintf(stderr,					\
-				"e32 but no previous packet "		\
-				"declaration.\n"			\
-				"Aborting! in %s::%s at line %d, "	\
-				"cmd_written=%d cmd_reserved=%d\n",	\
-				__FILE__, __FUNCTION__, __LINE__,	\
-				cmd_written, cmd_reserved);		\
-			_mesa_exit(-1);					\
-		}							\
+#define OUT_BATCH_PACKET3(packet, num_extra) do {\
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		\
+    	OUT_BATCH(cmdpacket3(b_l_rmesa->radeonScreen,\
+                  R300_CMD_PACKET3_RAW)); \
+    } else b_l_rmesa->cmdbuf.cs->section_cdw++;\
+	OUT_BATCH(CP_PACKET3((packet), (num_extra))); \
 	} while(0)
 
-#define	efloat(f) e32(r300PackFloat32(f))
-
-#define vsf_start_fragment(dest, length)				\
-	do {								\
-		int _n;							\
-		_n = (length);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+1),				\
-					__FUNCTION__);			\
-		cmd_reserved = _n+2;					\
-		cmd_written =1;						\
-		cmd[0].i = cmdvpu((dest), _n/4);			\
-	} while (0);
-
-#define r500fp_start_fragment(dest, length)				\
-	do {								\
-		int _n;							\
-		_n = (length);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+1),				\
-					__FUNCTION__);			\
-		cmd_reserved = _n+1;					\
-		cmd_written =1;						\
-		cmd[0].i = cmdr500fp((dest), _n/6, 0, 0);		\
-	} while (0);
-
-#define start_packet3(packet, count)					\
-	{								\
-		int _n;							\
-		GLuint _p;						\
-		_n = (count);						\
-		_p = (packet);						\
-		cmd = (drm_radeon_cmd_header_t*)			\
-			r300AllocCmdBuf(rmesa,				\
-					(_n+3),				\
-					__FUNCTION__);			\
-		cmd_reserved = _n+3;					\
-		cmd_written = 2;					\
-		if(_n > 0x3fff) {					\
-			fprintf(stderr,"Too big packet3 %08x: cannot "	\
-				"store %d dwords\n",			\
-				_p, _n);				\
-			_mesa_exit(-1);					\
-		}							\
-		cmd[0].i = cmdpacket3(R300_CMD_PACKET3_RAW);		\
-		cmd[1].i = _p | ((_n & 0x3fff)<<16);			\
-	}
-
 /**
  * Must be sent to switch to 2d commands
  */
-void static INLINE end_3d(r300ContextPtr rmesa)
+void static INLINE end_3d(radeonContextPtr radeon)
 {
-	drm_radeon_cmd_header_t *cmd = NULL;
+	BATCH_LOCALS(radeon);
 
-	cmd =
-	    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].header.cmd_type = R300_CMD_END3D;
+	if (!radeon->radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(1);
+		OUT_BATCH(cmdpacify(radeon->radeonScreen));
+		END_BATCH();
+	}
 }
 
 void static INLINE cp_delay(r300ContextPtr rmesa, unsigned short count)
 {
-	drm_radeon_cmd_header_t *cmd = NULL;
+	BATCH_LOCALS(&rmesa->radeon);
 
-	cmd =
-	    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].i = cmdcpdelay(count);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(1);
+		OUT_BATCH(cmdcpdelay(rmesa->radeon.radeonScreen, count));
+		END_BATCH();
+	}
 }
 
-void static INLINE cp_wait(r300ContextPtr rmesa, unsigned char flags)
+void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags)
 {
-	drm_radeon_cmd_header_t *cmd = NULL;
-
-	cmd =
-	    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa, 1, __FUNCTION__);
-	cmd[0].i = cmdwait(flags);
+	BATCH_LOCALS(radeon);
+	uint32_t wait_until;
+
+	if (!radeon->radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(1);
+		OUT_BATCH(cmdwait(radeon->radeonScreen, flags));
+		END_BATCH();
+	} else {
+		switch(flags) {
+		case R300_WAIT_2D:
+			wait_until = (1 << 14);
+			break;
+		case R300_WAIT_3D:
+			wait_until = (1 << 15);
+			break;
+		case R300_NEW_WAIT_2D_3D:
+			wait_until = (1 << 14) | (1 << 15);
+			break;
+		case R300_NEW_WAIT_2D_2D_CLEAN:
+			wait_until = (1 << 14) | (1 << 16) | (1 << 18);
+			break;
+		case R300_NEW_WAIT_3D_3D_CLEAN:
+			wait_until = (1 << 15) | (1 << 17) | (1 << 18);
+			break;
+		case R300_NEW_WAIT_2D_2D_CLEAN_3D_3D_CLEAN:
+			wait_until  = (1 << 14) | (1 << 16) | (1 << 18);
+			wait_until |= (1 << 15) | (1 << 17) | (1 << 18);
+			break;
+		default:
+			return;
+		}
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		OUT_BATCH(CP_PACKET0(RADEON_WAIT_UNTIL, 0));
+		OUT_BATCH(wait_until);
+		END_BATCH();
+	}
 }
 
-extern int r300EmitArrays(GLcontext * ctx);
-
-#ifdef USER_BUFFERS
-void r300UseArrays(GLcontext * ctx);
-#endif
+extern GLboolean r300EmitArrays(GLcontext * ctx);
 
-extern void r300ReleaseArrays(GLcontext * ctx);
 extern int r300PrimitiveType(r300ContextPtr rmesa, int prim);
 extern int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim);
 
 extern void r300EmitCacheFlush(r300ContextPtr rmesa);
 
-extern GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
-				 int *inputs, GLint * tab, GLuint nr);
-extern GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr);
 extern GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead);
 extern GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead);
-extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten);
-extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten);
+extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes);
+extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c
index 873cde4414..55c1cfe631 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.c
@@ -25,32 +25,12 @@
  *
  */
 
-/**
- * \file
- *
- * Fragment program compiler. Perform transformations on the intermediate
- * representation until the program is in a form where we can translate
- * it more or less directly into machine-readable form.
- *
- * \author Ben Skeggs <darktama@iinet.net.au>
- * \author Jerome Glisse <j.glisse@gmail.com>
- */
+#include "r300_fragprog.h"
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "shader/prog_instruction.h"
 #include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
 
 #include "r300_context.h"
-#include "r300_fragprog.h"
 #include "r300_fragprog_swizzle.h"
-#include "r300_state.h"
-
-#include "radeon_nqssadce.h"
-#include "radeon_program_alu.h"
-
 
 static void reset_srcreg(struct prog_src_register* reg)
 {
@@ -81,7 +61,7 @@ static struct prog_src_register shadow_ambient(struct gl_program *program, int t
  * \todo If/when r5xx uses the radeon_program architecture, this can probably
  * be reused.
  */
-static GLboolean transform_TEX(
+GLboolean r300_transform_TEX(
 	struct radeon_transform_context *t,
 	struct prog_instruction* orig_inst, void* data)
 {
@@ -160,6 +140,8 @@ static GLboolean transform_TEX(
 			inst.DstReg.Index = tempreg;
 			inst.DstReg.WriteMask = WRITEMASK_XYZW;
 			destredirect = GL_TRUE;
+		} else if (inst.SaturateMode) {
+			destredirect = GL_TRUE;
 		}
 	}
 
@@ -175,7 +157,7 @@ static GLboolean transform_TEX(
 		inst.SrcReg[0].File = PROGRAM_TEMPORARY;
 		inst.SrcReg[0].Index = tmpreg;
 	}
-	
+
 	tgt = radeonAppendInstructions(t->Program, 1);
 	_mesa_copy_instructions(tgt, &inst, 1);
 
@@ -239,6 +221,7 @@ static GLboolean transform_TEX(
 
 		tgt->Opcode = OPCODE_MOV;
 		tgt->DstReg = orig_inst->DstReg;
+		tgt->SaturateMode = inst.SaturateMode;
 		tgt->SrcReg[0].File = PROGRAM_TEMPORARY;
 		tgt->SrcReg[0].Index = inst.DstReg.Index;
 	}
@@ -246,241 +229,10 @@ static GLboolean transform_TEX(
 	return GL_TRUE;
 }
 
-
-static void update_params(r300ContextPtr r300, struct r300_fragment_program *fp)
-{
-	struct gl_fragment_program *mp = &fp->mesa_program;
-
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (mp->Base.Parameters)
-		_mesa_load_state_parameters(r300->radeon.glCtx, mp->Base.Parameters);
-}
-
-
-/**
- * Transform the program to support fragment.position.
- *
- * Introduce a small fragment at the start of the program that will be
- * the only code that directly reads the FRAG_ATTRIB_WPOS input.
- * All other code pieces that reference that input will be rewritten
- * to read from a newly allocated temporary.
- *
- * \todo if/when r5xx supports the radeon_program architecture, this is a
- * likely candidate for code sharing.
- */
-static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
-{
-	GLuint InputsRead = compiler->fp->mesa_program.Base.InputsRead;
-
-	if (!(InputsRead & FRAG_BIT_WPOS))
-		return;
-
-	static gl_state_index tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
-	};
-	struct prog_instruction *fpi;
-	GLuint window_index;
-	int i = 0;
-	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
-
-	_mesa_insert_instructions(compiler->program, 0, 3);
-	fpi = compiler->program->Instructions;
-
-	/* perspective divide */
-	fpi[i].Opcode = OPCODE_RCP;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_W;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	fpi[i].Opcode = OPCODE_MUL;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[1].Index = tempregi;
-	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	/* viewport transformation */
-	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
-
-	fpi[i].Opcode = OPCODE_MAD;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[0].Index = tempregi;
-	fpi[i].SrcReg[0].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[1].Index = window_index;
-	fpi[i].SrcReg[1].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[2].Index = window_index;
-	fpi[i].SrcReg[2].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-	i++;
-
-	for (; i < compiler->program->NumInstructions; ++i) {
-		int reg;
-		for (reg = 0; reg < 3; reg++) {
-			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
-			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
-				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
-				fpi[i].SrcReg[reg].Index = tempregi;
-			}
-		}
-	}
-}
-
-
-static void nqssadce_init(struct nqssadce_state* s)
-{
-	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
-	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
-}
-
-
-static GLuint build_dtm(GLuint depthmode)
-{
-	switch(depthmode) {
-	default:
-	case GL_LUMINANCE: return 0;
-	case GL_INTENSITY: return 1;
-	case GL_ALPHA: return 2;
-	}
-}
-
-static GLuint build_func(GLuint comparefunc)
-{
-	return comparefunc - GL_NEVER;
-}
-
-
-/**
- * Collect all external state that is relevant for compiling the given
- * fragment program.
- */
-static void build_state(
-	r300ContextPtr r300,
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_external_state *state)
-{
-	int unit;
-
-	_mesa_bzero(state, sizeof(*state));
-
-	for(unit = 0; unit < 16; ++unit) {
-		if (fp->mesa_program.Base.ShadowSamplers & (1 << unit)) {
-			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
-
-			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
-			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
-		}
-	}
-}
-
-
-void r300TranslateFragmentShader(r300ContextPtr r300,
-				 struct r300_fragment_program *fp)
-{
-	struct r300_fragment_program_external_state state;
-
-	build_state(r300, fp, &state);
-	if (_mesa_memcmp(&fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		fp->translated = GL_FALSE;
-		_mesa_memcpy(&fp->state, &state, sizeof(state));
-	}
-
-	if (!fp->translated) {
-		struct r300_fragment_program_compiler compiler;
-
-		compiler.r300 = r300;
-		compiler.fp = fp;
-		compiler.code = &fp->code;
-		compiler.program = _mesa_clone_program(r300->radeon.glCtx, &fp->mesa_program.Base);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: Initial program:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		insert_WPOS_trailer(&compiler);
-
-		struct radeon_program_transformation transformations[] = {
-			{ &transform_TEX, &compiler },
-			{ &radeonTransformALU, 0 },
-			{ &radeonTransformTrigSimple, 0 }
-		};
-		radeonLocalTransform(
-			r300->radeon.glCtx,
-			compiler.program,
-			3, transformations);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: After native rewrite:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		struct radeon_nqssadce_descr nqssadce = {
-			.Init = &nqssadce_init,
-			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
-			.BuildSwizzle = &r300FPBuildSwizzle,
-			.RewriteDepthOut = GL_TRUE
-		};
-		radeonNqssaDce(r300->radeon.glCtx, compiler.program, &nqssadce);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		if (!r300FragmentProgramEmit(&compiler))
-			fp->error = GL_TRUE;
-
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->mesa_program.Base.Parameters);
-		fp->mesa_program.Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
-
-		_mesa_reference_program(r300->radeon.glCtx, &compiler.program, NULL);
-
-		if (!fp->error)
-			fp->translated = GL_TRUE;
-		if (fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
-			r300FragmentProgramDump(fp, &fp->code);
-		r300UpdateStateParameters(r300->radeon.glCtx, _NEW_PROGRAM);
-	}
-
-	update_params(r300, fp);
-}
-
 /* just some random things... */
-void r300FragmentProgramDump(
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_code *code)
+void r300FragmentProgramDump(union rX00_fragment_program_code *c)
 {
+	struct r300_fragment_program_code *code = &c->r300;
 	int n, i, j;
 	static int pc = 0;
 
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.h b/src/mesa/drivers/dri/r300/r300_fragprog.h
index 94fb554fb3..5ce6f33cee 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.h
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.h
@@ -33,9 +33,6 @@
 #ifndef __R300_FRAGPROG_H_
 #define __R300_FRAGPROG_H_
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
 #include "shader/program.h"
 #include "shader/prog_instruction.h"
 
@@ -105,28 +102,10 @@
 
 #endif
 
-struct r300_fragment_program;
+extern GLboolean r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler);
 
-extern void r300TranslateFragmentShader(r300ContextPtr r300,
-					struct r300_fragment_program *fp);
+extern void r300FragmentProgramDump(union rX00_fragment_program_code *c);
 
-
-/**
- * Used internally by the r300 fragment program code to store compile-time
- * only data.
- */
-struct r300_fragment_program_compiler {
-	r300ContextPtr r300;
-	struct r300_fragment_program *fp;
-	struct r300_fragment_program_code *code;
-	struct gl_program *program;
-};
-
-extern GLboolean r300FragmentProgramEmit(struct r300_fragment_program_compiler *compiler);
-
-
-extern void r300FragmentProgramDump(
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_code *code);
+extern GLboolean r300_transform_TEX(struct radeon_transform_context *t, struct prog_instruction* orig_inst, void* data);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
new file mode 100644
index 0000000000..f5c4c0f4a0
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -0,0 +1,399 @@
+/*
+ * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+/**
+ * \file
+ *
+ * Fragment program compiler. Perform transformations on the intermediate
+ * representation until the program is in a form where we can translate
+ * it more or less directly into machine-readable form.
+ *
+ * \author Ben Skeggs <darktama@iinet.net.au>
+ * \author Jerome Glisse <j.glisse@gmail.com>
+ */
+
+#include "r300_fragprog_common.h"
+
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "r300_state.h"
+#include "r300_fragprog.h"
+#include "r300_fragprog_swizzle.h"
+#include "r500_fragprog.h"
+
+#include "radeon_program.h"
+#include "radeon_program_alu.h"
+
+static void nqssadce_init(struct nqssadce_state* s)
+{
+	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
+	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
+}
+
+/**
+ * Transform the program to support fragment.position.
+ *
+ * Introduce a small fragment at the start of the program that will be
+ * the only code that directly reads the FRAG_ATTRIB_WPOS input.
+ * All other code pieces that reference that input will be rewritten
+ * to read from a newly allocated temporary.
+ *
+ */
+static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
+{
+	GLuint InputsRead = compiler->fp->Base->InputsRead;
+
+	if (!(InputsRead & FRAG_BIT_WPOS)) {
+		compiler->fp->wpos_attr = FRAG_ATTRIB_MAX;
+		return;
+	}
+
+	static gl_state_index tokens[STATE_LENGTH] = {
+		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
+	};
+	struct prog_instruction *fpi;
+	GLuint window_index;
+	int i = 0;
+
+	for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
+	{
+		if (!(InputsRead & (1 << i))) {
+			InputsRead &= ~(1 << FRAG_ATTRIB_WPOS);
+			InputsRead |= 1 << i;
+			compiler->fp->Base->InputsRead = InputsRead;
+			compiler->fp->wpos_attr = i;
+			break;
+		}
+	}
+
+	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
+
+	_mesa_insert_instructions(compiler->program, 0, 3);
+	fpi = compiler->program->Instructions;
+	i = 0;
+
+	/* perspective divide */
+	fpi[i].Opcode = OPCODE_RCP;
+
+	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
+	fpi[i].DstReg.Index = tempregi;
+	fpi[i].DstReg.WriteMask = WRITEMASK_W;
+	fpi[i].DstReg.CondMask = COND_TR;
+
+	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
+	fpi[i].SrcReg[0].Index = compiler->fp->wpos_attr;
+	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
+	i++;
+
+	fpi[i].Opcode = OPCODE_MUL;
+
+	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
+	fpi[i].DstReg.Index = tempregi;
+	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
+	fpi[i].DstReg.CondMask = COND_TR;
+
+	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
+	fpi[i].SrcReg[0].Index = compiler->fp->wpos_attr;
+	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
+	fpi[i].SrcReg[1].Index = tempregi;
+	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
+	i++;
+
+	/* viewport transformation */
+	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
+
+	fpi[i].Opcode = OPCODE_MAD;
+
+	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
+	fpi[i].DstReg.Index = tempregi;
+	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
+	fpi[i].DstReg.CondMask = COND_TR;
+
+	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
+	fpi[i].SrcReg[0].Index = tempregi;
+	fpi[i].SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
+	fpi[i].SrcReg[1].Index = window_index;
+	fpi[i].SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
+	fpi[i].SrcReg[2].Index = window_index;
+	fpi[i].SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+	i++;
+
+	for (; i < compiler->program->NumInstructions; ++i) {
+		int reg;
+		for (reg = 0; reg < 3; reg++) {
+			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
+			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
+				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
+				fpi[i].SrcReg[reg].Index = tempregi;
+			}
+		}
+	}
+}
+
+
+/**
+ * Rewrite fragment.fogcoord to use a texture coordinate slot.
+ * Note that fogcoord is forced into an X001 pattern, and this enforcement
+ * is done here.
+ *
+ * See also the counterpart rewriting for vertex programs.
+ */
+static void rewriteFog(struct r300_fragment_program_compiler *compiler)
+{
+	struct r300_fragment_program *fp = compiler->fp;
+	GLuint InputsRead;
+	int i;
+
+	InputsRead = fp->Base->InputsRead;
+
+	if (!(InputsRead & FRAG_BIT_FOGC)) {
+		fp->fog_attr = FRAG_ATTRIB_MAX;
+		return;
+	}
+
+	for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
+	{
+		if (!(InputsRead & (1 << i))) {
+			InputsRead &= ~(1 << FRAG_ATTRIB_FOGC);
+			InputsRead |= 1 << i;
+			fp->Base->InputsRead = InputsRead;
+			fp->fog_attr = i;
+			break;
+		}
+	}
+
+	{
+		struct prog_instruction *inst;
+
+		inst = compiler->program->Instructions;
+		while (inst->Opcode != OPCODE_END) {
+			const int src_regs = _mesa_num_inst_src_regs(inst->Opcode);
+			for (i = 0; i < src_regs; ++i) {
+				if (inst->SrcReg[i].File == PROGRAM_INPUT && inst->SrcReg[i].Index == FRAG_ATTRIB_FOGC) {
+					inst->SrcReg[i].Index = fp->fog_attr;
+					inst->SrcReg[i].Swizzle = combine_swizzles(
+						MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE),
+						inst->SrcReg[i].Swizzle);
+				}
+			}
+			++inst;
+		}
+	}
+}
+
+static GLuint build_dtm(GLuint depthmode)
+{
+	switch(depthmode) {
+	default:
+	case GL_LUMINANCE: return 0;
+	case GL_INTENSITY: return 1;
+	case GL_ALPHA: return 2;
+	}
+}
+
+static GLuint build_func(GLuint comparefunc)
+{
+	return comparefunc - GL_NEVER;
+}
+
+/**
+ * Collect all external state that is relevant for compiling the given
+ * fragment program.
+ */
+static void build_state(
+	r300ContextPtr r300,
+	struct gl_fragment_program *fp,
+	struct r300_fragment_program_external_state *state)
+{
+	int unit;
+
+	_mesa_bzero(state, sizeof(*state));
+
+	for(unit = 0; unit < 16; ++unit) {
+		if (fp->Base.ShadowSamplers & (1 << unit)) {
+			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
+
+			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
+			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
+		}
+	}
+}
+
+static void rewrite_depth_out(struct gl_program *prog)
+{
+	struct prog_instruction *inst;
+
+	for (inst = prog->Instructions; inst->Opcode != OPCODE_END; ++inst) {
+		if (inst->DstReg.File != PROGRAM_OUTPUT || inst->DstReg.Index != FRAG_RESULT_DEPTH)
+			continue;
+
+		if (inst->DstReg.WriteMask & WRITEMASK_Z) {
+			inst->DstReg.WriteMask = WRITEMASK_W;
+		} else {
+			inst->DstReg.WriteMask = 0;
+			continue;
+		}
+
+		switch (inst->Opcode) {
+			case OPCODE_FRC:
+			case OPCODE_MOV:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				break;
+			case OPCODE_ADD:
+			case OPCODE_MAX:
+			case OPCODE_MIN:
+			case OPCODE_MUL:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
+				break;
+			case OPCODE_CMP:
+			case OPCODE_MAD:
+				inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
+				inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
+				inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]);
+				break;
+			default:
+				// Scalar instructions needn't be reswizzled
+				break;
+		}
+	}
+}
+
+void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *fp)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program_compiler compiler;
+
+	compiler.r300 = r300;
+	compiler.fp = fp;
+	compiler.code = &fp->code;
+	compiler.program = fp->Base;
+
+	if (RADEON_DEBUG & DEBUG_PIXEL) {
+		fflush(stdout);
+		_mesa_printf("Fragment Program: Initial program:\n");
+		_mesa_print_program(compiler.program);
+		fflush(stdout);
+	}
+
+	insert_WPOS_trailer(&compiler);
+
+	rewriteFog(&compiler);
+
+	rewrite_depth_out(compiler.program);
+
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		struct radeon_program_transformation transformations[] = {
+			{ &r500_transform_TEX, &compiler },
+			{ &radeonTransformALU, 0 },
+			{ &radeonTransformDeriv, 0 },
+			{ &radeonTransformTrigScale, 0 }
+		};
+		radeonLocalTransform(ctx, compiler.program, 4, transformations);
+	} else {
+		struct radeon_program_transformation transformations[] = {
+			{ &r300_transform_TEX, &compiler },
+			{ &radeonTransformALU, 0 },
+			{ &radeonTransformTrigSimple, 0 }
+		};
+		radeonLocalTransform(ctx, compiler.program, 3, transformations);
+	}
+
+	if (RADEON_DEBUG & DEBUG_PIXEL) {
+		_mesa_printf("Fragment Program: After native rewrite:\n");
+		_mesa_print_program(compiler.program);
+		fflush(stdout);
+	}
+
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadce_init,
+			.IsNativeSwizzle = &r500FPIsNativeSwizzle,
+			.BuildSwizzle = &r500FPBuildSwizzle
+		};
+		radeonNqssaDce(ctx, compiler.program, &nqssadce);
+	} else {
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadce_init,
+			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
+			.BuildSwizzle = &r300FPBuildSwizzle
+		};
+		radeonNqssaDce(ctx, compiler.program, &nqssadce);
+	}
+
+	if (RADEON_DEBUG & DEBUG_PIXEL) {
+		_mesa_printf("Compiler: after NqSSA-DCE:\n");
+		_mesa_print_program(compiler.program);
+		fflush(stdout);
+	}
+
+	if (!r300->vtbl.BuildFragmentProgramHwCode(&compiler))
+		fp->error = GL_TRUE;
+
+	fp->translated = GL_TRUE;
+
+	if (fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
+		r300->vtbl.FragmentProgramDump(&fp->code);
+}
+
+struct r300_fragment_program *r300SelectFragmentShader(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program_cont *fp_list;
+	struct r300_fragment_program *fp;
+	struct r300_fragment_program_external_state state;
+
+	fp_list = (struct r300_fragment_program_cont *)ctx->FragmentProgram._Current;
+	build_state(r300, ctx->FragmentProgram._Current, &state);
+
+	fp = fp_list->progs;
+	while (fp) {
+		if (_mesa_memcmp(&fp->state, &state, sizeof(state)) == 0) {
+			return r300->selected_fp = fp;
+		}
+		fp = fp->next;
+	}
+
+	fp = _mesa_calloc(sizeof(struct r300_fragment_program));
+
+	fp->state = state;
+	fp->translated = GL_FALSE;
+	fp->Base = _mesa_clone_program(ctx, &ctx->FragmentProgram._Current->Base);
+
+	fp->next = fp_list->progs;
+	fp_list->progs = fp;
+
+	return r300->selected_fp = fp;
+}
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.h b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
new file mode 100644
index 0000000000..5e103ee408
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __R300_FRAGPROG_COMMON_H_
+#define __R300_FRAGPROG_COMMON_H_
+
+#include "main/mtypes.h"
+
+#include "r300_context.h"
+
+extern void r300TranslateFragmentShader(GLcontext *ctx, struct r300_fragment_program *fp);
+
+struct r300_fragment_program *r300SelectFragmentShader(GLcontext *ctx);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_emit.c b/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
index 9f0b7e3534..b75656e7ee 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
@@ -47,7 +47,7 @@
 
 #define PROG_CODE \
 	struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \
-	struct r300_fragment_program_code *code = c->code
+	struct r300_fragment_program_code *code = &c->code->r300
 
 #define error(fmt, args...) do {			\
 		fprintf(stderr, "%s::%s(): " fmt "\n",	\
@@ -66,7 +66,7 @@ static GLboolean emit_const(void* data, GLuint file, GLuint index, GLuint *hwind
 	}
 
 	if (*hwindex >= code->const_nr) {
-		if (*hwindex >= PFS_NUM_CONST_REGS) {
+		if (*hwindex >= R300_PFS_NUM_CONST_REGS) {
 			error("Out of hw constants!\n");
 			return GL_FALSE;
 		}
@@ -138,7 +138,7 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
 {
 	PROG_CODE;
 
-	if (code->alu.length >= PFS_MAX_ALU_INST) {
+	if (code->alu.length >= R300_PFS_MAX_ALU_INST) {
 		error("Too many ALU instructions");
 		return GL_FALSE;
 	}
@@ -201,7 +201,7 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
 	if (inst->Alpha.DepthWriteMask) {
 		code->alu.inst[ip].inst3 |= R300_ALU_DSTA_DEPTH;
 		code->node[code->cur_node].flags |= R300_W_OUT;
-		c->fp->WritesDepth = GL_TRUE;
+		c->fp->writes_depth = GL_TRUE;
 	}
 
 	return GL_TRUE;
@@ -213,7 +213,7 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
  */
 static GLboolean finish_node(struct r300_fragment_program_compiler *c)
 {
-	struct r300_fragment_program_code *code = c->code;
+	struct r300_fragment_program_code *code = &c->code->r300;
 	struct r300_fragment_program_node *node = &code->node[code->cur_node];
 
 	if (node->alu_end < 0) {
@@ -275,7 +275,7 @@ static GLboolean emit_tex(void* data, struct prog_instruction* inst)
 {
 	PROG_CODE;
 
-	if (code->tex.length >= PFS_MAX_TEX_INST) {
+	if (code->tex.length >= R300_PFS_MAX_TEX_INST) {
 		error("Too many TEX instructions");
 		return GL_FALSE;
 	}
@@ -318,16 +318,16 @@ static const struct radeon_pair_handler pair_handler = {
 	.EmitPaired = &emit_alu,
 	.EmitTex = &emit_tex,
 	.BeginTexBlock = &begin_tex,
-	.MaxHwTemps = PFS_NUM_TEMP_REGS
+	.MaxHwTemps = R300_PFS_NUM_TEMP_REGS
 };
 
 /**
  * Final compilation step: Turn the intermediate radeon_program into
  * machine-readable instructions.
  */
-GLboolean r300FragmentProgramEmit(struct r300_fragment_program_compiler *compiler)
+GLboolean r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler)
 {
-	struct r300_fragment_program_code *code = compiler->code;
+	struct r300_fragment_program_code *code = &compiler->code->r300;
 
 	_mesa_bzero(code, sizeof(struct r300_fragment_program_code));
 	code->node[0].alu_end = -1;
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index ee85e229f0..ddabd53992 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -46,8 +46,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "swrast/swrast.h"
 
+#include "radeon_common.h"
+#include "radeon_lock.h"
 #include "r300_context.h"
-#include "radeon_ioctl.h"
 #include "r300_ioctl.h"
 #include "r300_cmdbuf.h"
 #include "r300_state.h"
@@ -55,71 +56,85 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_reg.h"
 #include "r300_emit.h"
 #include "r300_fragprog.h"
+#include "r300_context.h"
 
 #include "vblank.h"
 
+#define R200_3D_DRAW_IMMD_2      0xC0003500
+
 #define CLEARBUFFER_COLOR	0x1
 #define CLEARBUFFER_DEPTH	0x2
 #define CLEARBUFFER_STENCIL	0x4
 
-static void r300ClearBuffer(r300ContextPtr r300, int flags, int buffer)
+static void r300EmitClearState(GLcontext * ctx);
+
+static void r300ClearBuffer(r300ContextPtr r300, int flags,
+			    struct radeon_renderbuffer *rrb,
+			    struct radeon_renderbuffer *rrbd)
 {
+	BATCH_LOCALS(&r300->radeon);
 	GLcontext *ctx = r300->radeon.glCtx;
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
-	GLuint cboffset, cbpitch;
-	drm_r300_cmd_header_t *cmd2;
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
+	GLuint cbpitch = 0;
 	r300ContextPtr rmesa = r300;
 
 	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s: %s buffer (%i,%i %ix%i)\n",
-			__FUNCTION__, buffer ? "back" : "front",
-			dPriv->x, dPriv->y, dPriv->w, dPriv->h);
-
-	if (buffer) {
-		cboffset = r300->radeon.radeonScreen->backOffset;
-		cbpitch = r300->radeon.radeonScreen->backPitch;
-	} else {
-		cboffset = r300->radeon.radeonScreen->frontOffset;
-		cbpitch = r300->radeon.radeonScreen->frontPitch;
+		fprintf(stderr, "%s: buffer %p (%i,%i %ix%i)\n",
+			__FUNCTION__, rrb, dPriv->x, dPriv->y,
+			dPriv->w, dPriv->h);
+
+	if (rrb) {
+		cbpitch = (rrb->pitch / rrb->cpp);
+		if (rrb->cpp == 4)
+			cbpitch |= R300_COLOR_FORMAT_ARGB8888;
+		else
+			cbpitch |= R300_COLOR_FORMAT_RGB565;
+
+		if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){
+			cbpitch |= R300_COLOR_TILE_ENABLE;
+        }
 	}
 
-	cboffset += r300->radeon.radeonScreen->fbLocation;
-
-	cp_wait(r300, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
-	end_3d(rmesa);
-
-	R300_STATECHANGE(r300, cb);
-	reg_start(R300_RB3D_COLOROFFSET0, 0);
-	e32(cboffset);
-
-	if (r300->radeon.radeonScreen->cpp == 4)
-		cbpitch |= R300_COLOR_FORMAT_ARGB8888;
-	else
-		cbpitch |= R300_COLOR_FORMAT_RGB565;
-
-	if (r300->radeon.sarea->tiling_enabled)
-		cbpitch |= R300_COLOR_TILE_ENABLE;
-
-	reg_start(R300_RB3D_COLORPITCH0, 0);
-	e32(cbpitch);
-
-	R300_STATECHANGE(r300, cmk);
-	reg_start(RB3D_COLOR_CHANNEL_MASK, 0);
+	/* TODO in bufmgr */
+	cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+	end_3d(&rmesa->radeon);
 
 	if (flags & CLEARBUFFER_COLOR) {
-		e32((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) |
-		    (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) |
-		    (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) |
-		    (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0));
+		assert(rrb != 0);
+		BEGIN_BATCH_NO_AUTOSTATE(6);
+		OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1);
+		OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+		OUT_BATCH_REGVAL(R300_RB3D_COLORPITCH0, cbpitch);
+		END_BATCH();
+	}
+#if 1
+	if (flags & (CLEARBUFFER_DEPTH | CLEARBUFFER_STENCIL)) {
+		assert(rrbd != 0);
+		cbpitch = (rrbd->pitch / rrbd->cpp);
+		if (rrbd->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){
+			cbpitch |= R300_DEPTHMACROTILE_ENABLE;
+        }
+		if (rrbd->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
+            cbpitch |= R300_DEPTHMICROTILE_TILED;
+        }
+		BEGIN_BATCH_NO_AUTOSTATE(6);
+		OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1);
+		OUT_BATCH_RELOC(0, rrbd->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+		OUT_BATCH_REGVAL(R300_ZB_DEPTHPITCH, cbpitch);
+		END_BATCH();
+	}
+#endif
+	BEGIN_BATCH_NO_AUTOSTATE(6);
+	OUT_BATCH_REGSEQ(RB3D_COLOR_CHANNEL_MASK, 1);
+	if (flags & CLEARBUFFER_COLOR) {
+		OUT_BATCH((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) |
+			  (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) |
+			  (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) |
+			  (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0));
 	} else {
-		e32(0x0);
+		OUT_BATCH(0);
 	}
 
-	R300_STATECHANGE(r300, zs);
-	reg_start(R300_ZB_CNTL, 2);
 
 	{
 		uint32_t t1, t2;
@@ -146,73 +161,92 @@ static void r300ClearBuffer(r300ContextPtr r300, int flags, int buffer)
 			     R300_S_FRONT_ZFAIL_OP_SHIFT);
 		}
 
-		e32(t1);
-		e32(t2);
-		e32(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) << R300_STENCILWRITEMASK_SHIFT) |
-		    (ctx->Stencil.Clear & R300_STENCILREF_MASK));
+		OUT_BATCH_REGSEQ(R300_ZB_CNTL, 3);
+		OUT_BATCH(t1);
+		OUT_BATCH(t2);
+		OUT_BATCH(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) <<
+                   R300_STENCILWRITEMASK_SHIFT) |
+			  (ctx->Stencil.Clear & R300_STENCILREF_MASK));
+		END_BATCH();
 	}
 
-	cmd2 = (drm_r300_cmd_header_t *) r300AllocCmdBuf(r300, 9, __FUNCTION__);
-	cmd2[0].packet3.cmd_type = R300_CMD_PACKET3;
-	cmd2[0].packet3.packet = R300_CMD_PACKET3_CLEAR;
-	cmd2[1].u = r300PackFloat32(dPriv->w / 2.0);
-	cmd2[2].u = r300PackFloat32(dPriv->h / 2.0);
-	cmd2[3].u = r300PackFloat32(ctx->Depth.Clear);
-	cmd2[4].u = r300PackFloat32(1.0);
-	cmd2[5].u = r300PackFloat32(ctx->Color.ClearColor[0]);
-	cmd2[6].u = r300PackFloat32(ctx->Color.ClearColor[1]);
-	cmd2[7].u = r300PackFloat32(ctx->Color.ClearColor[2]);
-	cmd2[8].u = r300PackFloat32(ctx->Color.ClearColor[3]);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		BEGIN_BATCH_NO_AUTOSTATE(9);
+		OUT_BATCH(cmdpacket3(r300->radeon.radeonScreen, R300_CMD_PACKET3_CLEAR));
+		OUT_BATCH_FLOAT32(dPriv->w / 2.0);
+		OUT_BATCH_FLOAT32(dPriv->h / 2.0);
+		OUT_BATCH_FLOAT32(ctx->Depth.Clear);
+		OUT_BATCH_FLOAT32(1.0);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]);
+		END_BATCH();
+	} else {
+		OUT_BATCH(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8));
+		OUT_BATCH(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
+			  (1 << R300_PRIM_NUM_VERTICES_SHIFT));
+		OUT_BATCH_FLOAT32(dPriv->w / 2.0);
+		OUT_BATCH_FLOAT32(dPriv->h / 2.0);
+		OUT_BATCH_FLOAT32(ctx->Depth.Clear);
+		OUT_BATCH_FLOAT32(1.0);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]);
+		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]);
+	}
 
 	r300EmitCacheFlush(rmesa);
-	cp_wait(rmesa, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+	cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
+
+	R300_STATECHANGE(r300, cb);
+	R300_STATECHANGE(r300, cmk);
+	R300_STATECHANGE(r300, zs);
 }
 
 static void r300EmitClearState(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	r300ContextPtr rmesa = r300;
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	BATCH_LOCALS(&r300->radeon);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	int i;
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
-	int has_tcl = 1;
+	int has_tcl;
 	int is_r500 = 0;
 	GLuint vap_cntl;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
-
-        if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-                is_r500 = 1;
+	has_tcl = r300->options.hw_tcl_enabled;
 
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
+		is_r500 = 1;
 
-	/* FIXME: the values written to R300_VAP_INPUT_ROUTE_0_0 and
-	 * R300_VAP_INPUT_ROUTE_0_1 are in fact known, however, the values are
-	 * quite complex; see the functions in r300_emit.c.
+	/* State atom dirty tracking is a little subtle here.
 	 *
-	 * I believe it would be a good idea to extend the functions in
-	 * r300_emit.c so that they can be used to setup the default values for
-	 * these registers, as well as the actual values used for rendering.
+	 * On the one hand, we need to make sure base state is emitted
+	 * here if we start with an empty batch buffer, otherwise clear
+	 * works incorrectly with multiple processes. Therefore, the first
+	 * BEGIN_BATCH cannot be a BEGIN_BATCH_NO_AUTOSTATE.
+	 *
+	 * On the other hand, implicit state emission clears the state atom
+	 * dirty bits, so we have to call R300_STATECHANGE later than the
+	 * first BEGIN_BATCH.
+	 *
+	 * The final trickiness is that, because we change state, we need
+	 * to ensure that any stored swtcl primitives are flushed properly
+	 * before we start changing state. See the R300_NEWPRIM in r300Clear
+	 * for this.
 	 */
-	R300_STATECHANGE(r300, vir[0]);
-	reg_start(R300_VAP_PROG_STREAM_CNTL_0, 0);
+	BEGIN_BATCH(31);
+	OUT_BATCH_REGSEQ(R300_VAP_PROG_STREAM_CNTL_0, 1);
 	if (!has_tcl)
-	    e32(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
+		OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
 		 ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)));
 	else
-	    e32(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
+		OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) |
 		 ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT)));
 
-	/* disable fog */
-	R300_STATECHANGE(r300, fogs);
-	reg_start(R300_FG_FOG_BLEND, 0);
-	e32(0x0);
-
-	R300_STATECHANGE(r300, vir[1]);
-	reg_start(R300_VAP_PROG_STREAM_CNTL_EXT_0, 0);
-	e32(((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) |
+	OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0);
+	OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0,
+	   ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) |
 	       (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) |
 	       (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) |
 	       (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) |
@@ -226,619 +260,442 @@ static void r300EmitClearState(GLcontext * ctx)
 	      << R300_SWIZZLE1_SHIFT)));
 
 	/* R300_VAP_INPUT_CNTL_0, R300_VAP_INPUT_CNTL_1 */
-	R300_STATECHANGE(r300, vic);
-	reg_start(R300_VAP_VTX_STATE_CNTL, 1);
-	e32((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT));
-	e32(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0);
+	OUT_BATCH_REGSEQ(R300_VAP_VTX_STATE_CNTL, 2);
+	OUT_BATCH((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT));
+	OUT_BATCH(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0);
 
-	R300_STATECHANGE(r300, vte);
 	/* comes from fglrx startup of clear */
-	reg_start(R300_SE_VTE_CNTL, 1);
-	e32(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA |
-	    R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
-	    R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
-	    R300_VPORT_Z_OFFSET_ENA);
-	e32(0x8);
+	OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2);
+	OUT_BATCH(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA |
+		  R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
+		  R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
+		  R300_VPORT_Z_OFFSET_ENA);
+	OUT_BATCH(0x8);
 
-	reg_start(R300_VAP_PSC_SGN_NORM_CNTL, 0);
-	e32(0xaaaaaaaa);
+	OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa);
 
-	R300_STATECHANGE(r300, vof);
-	reg_start(R300_VAP_OUTPUT_VTX_FMT_0, 1);
-	e32(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT |
-	    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT);
-	e32(0x0);		/* no textures */
+	OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
+	OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT |
+		  R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT);
+	OUT_BATCH(0); /* no textures */
 
-	R300_STATECHANGE(r300, txe);
-	reg_start(R300_TX_ENABLE, 0);
-	e32(0x0);
+	OUT_BATCH_REGVAL(R300_TX_ENABLE, 0);
 
-	R300_STATECHANGE(r300, vpt);
-	reg_start(R300_SE_VPORT_XSCALE, 5);
-	efloat(1.0);
-	efloat(dPriv->x);
-	efloat(1.0);
-	efloat(dPriv->y);
-	efloat(1.0);
-	efloat(0.0);
+	OUT_BATCH_REGSEQ(R300_SE_VPORT_XSCALE, 6);
+	OUT_BATCH_FLOAT32(1.0);
+	OUT_BATCH_FLOAT32(dPriv->x);
+	OUT_BATCH_FLOAT32(1.0);
+	OUT_BATCH_FLOAT32(dPriv->y);
+	OUT_BATCH_FLOAT32(1.0);
+	OUT_BATCH_FLOAT32(0.0);
 
-	R300_STATECHANGE(r300, at);
-	reg_start(R300_FG_ALPHA_FUNC, 0);
-	e32(0x0);
+	OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0);
 
+	OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2);
+	OUT_BATCH(0x0);
+	OUT_BATCH(0x0);
+	END_BATCH();
+
+	R300_STATECHANGE(r300, vir[0]);
+	R300_STATECHANGE(r300, fogs);
+	R300_STATECHANGE(r300, vir[1]);
+	R300_STATECHANGE(r300, vic);
+	R300_STATECHANGE(r300, vte);
+	R300_STATECHANGE(r300, vof);
+	R300_STATECHANGE(r300, txe);
+	R300_STATECHANGE(r300, vpt);
+	R300_STATECHANGE(r300, at);
 	R300_STATECHANGE(r300, bld);
-	reg_start(R300_RB3D_CBLEND, 1);
-	e32(0x0);
-	e32(0x0);
+	R300_STATECHANGE(r300, ps);
 
 	if (has_tcl) {
-	    R300_STATECHANGE(r300, vap_clip_cntl);
-	    reg_start(R300_VAP_CLIP_CNTL, 0);
-	    e32(R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE);
+		R300_STATECHANGE(r300, vap_clip_cntl);
+
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE);
+		END_BATCH();
         }
 
-	R300_STATECHANGE(r300, ps);
-	reg_start(R300_GA_POINT_SIZE, 0);
-	e32(((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) |
-	    ((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT));
+	BEGIN_BATCH_NO_AUTOSTATE(2);
+	OUT_BATCH_REGVAL(R300_GA_POINT_SIZE,
+		((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) |
+		((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT));
+	END_BATCH();
 
 	if (!is_r500) {
 		R300_STATECHANGE(r300, ri);
-		reg_start(R300_RS_IP_0, 7);
-		for (i = 0; i < 8; ++i) {
-			e32(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3));
-		}
-
 		R300_STATECHANGE(r300, rc);
-		/* The second constant is needed to get glxgears display anything .. */
-		reg_start(R300_RS_COUNT, 1);
-		e32((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
-		e32(0x0);
-
 		R300_STATECHANGE(r300, rr);
-		reg_start(R300_RS_INST_0, 0);
-		e32(R300_RS_INST_COL_CN_WRITE);
+
+		BEGIN_BATCH(14);
+		OUT_BATCH_REGSEQ(R300_RS_IP_0, 8);
+		for (i = 0; i < 8; ++i)
+			OUT_BATCH(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3));
+
+		OUT_BATCH_REGSEQ(R300_RS_COUNT, 2);
+		OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+		OUT_BATCH(0x0);
+
+		OUT_BATCH_REGVAL(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE);
+		END_BATCH();
 	} else {
 		R300_STATECHANGE(r300, ri);
-		reg_start(R500_RS_IP_0, 7);
+		R300_STATECHANGE(r300, rc);
+		R300_STATECHANGE(r300, rr);
+
+		BEGIN_BATCH(14);
+		OUT_BATCH_REGSEQ(R500_RS_IP_0, 8);
 		for (i = 0; i < 8; ++i) {
-			e32((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
-			    (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
-			    (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
-			    (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT));
+			OUT_BATCH((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
+				  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
+				  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
+				  (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT));
 		}
 
-		R300_STATECHANGE(r300, rc);
-		/* The second constant is needed to get glxgears display anything .. */
-		reg_start(R300_RS_COUNT, 1);
-		e32((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
-		e32(0x0);
-
-		R300_STATECHANGE(r300, rr);
-		reg_start(R500_RS_INST_0, 0);
-		e32(R500_RS_INST_COL_CN_WRITE);
+		OUT_BATCH_REGSEQ(R300_RS_COUNT, 2);
+		OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+		OUT_BATCH(0x0);
 
+		OUT_BATCH_REGVAL(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE);
+		END_BATCH();
 	}
 
 	if (!is_r500) {
 		R300_STATECHANGE(r300, fp);
-		reg_start(R300_US_CONFIG, 2);
-		e32(0x0);
-		e32(0x0);
-		e32(0x0);
-		reg_start(R300_US_CODE_ADDR_0, 3);
-		e32(0x0);
-		e32(0x0);
-		e32(0x0);
-		e32(R300_RGBA_OUT);
-
 		R300_STATECHANGE(r300, fpi[0]);
 		R300_STATECHANGE(r300, fpi[1]);
 		R300_STATECHANGE(r300, fpi[2]);
 		R300_STATECHANGE(r300, fpi[3]);
 
-		reg_start(R300_US_ALU_RGB_INST_0, 0);
-		e32(FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO)));
-
-		reg_start(R300_US_ALU_RGB_ADDR_0, 0);
-		e32(FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0));
-
-		reg_start(R300_US_ALU_ALPHA_INST_0, 0);
-		e32(FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO)));
-
-		reg_start(R300_US_ALU_ALPHA_ADDR_0, 0);
-		e32(FP_SELA(0, NO, W, FP_TMP(0), 0, 0));
+		BEGIN_BATCH(17);
+		OUT_BATCH_REGSEQ(R300_US_CONFIG, 3);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH(0x0);
+		OUT_BATCH(R300_RGBA_OUT);
+
+		OUT_BATCH_REGVAL(R300_US_ALU_RGB_INST_0,
+			FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO)));
+		OUT_BATCH_REGVAL(R300_US_ALU_RGB_ADDR_0,
+			FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0));
+		OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_INST_0,
+			FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO)));
+		OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_ADDR_0,
+			FP_SELA(0, NO, W, FP_TMP(0), 0, 0));
+		END_BATCH();
 	} else {
- 		R300_STATECHANGE(r300, fp);
- 		reg_start(R500_US_CONFIG, 1);
- 		e32(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
- 		e32(0x0);
- 		reg_start(R500_US_CODE_ADDR, 2);
- 		e32(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1));
- 		e32(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1));
- 		e32(R500_US_CODE_OFFSET_ADDR(0));
+		struct radeon_state_atom r500fp;
+		uint32_t _cmd[10];
 
+		R300_STATECHANGE(r300, fp);
 		R300_STATECHANGE(r300, r500fp);
-		r500fp_start_fragment(0, 6);
-
-		e32(R500_INST_TYPE_OUT |
-		    R500_INST_TEX_SEM_WAIT |
-		    R500_INST_LAST |
-		    R500_INST_RGB_OMASK_R |
-		    R500_INST_RGB_OMASK_G |
-		    R500_INST_RGB_OMASK_B |
-		    R500_INST_ALPHA_OMASK |
-		    R500_INST_RGB_CLAMP |
-		    R500_INST_ALPHA_CLAMP);
-
-		e32(R500_RGB_ADDR0(0) |
-		    R500_RGB_ADDR1(0) |
-		    R500_RGB_ADDR1_CONST |
-		    R500_RGB_ADDR2(0) |
-		    R500_RGB_ADDR2_CONST);
-
-		e32(R500_ALPHA_ADDR0(0) |
-		    R500_ALPHA_ADDR1(0) |
-		    R500_ALPHA_ADDR1_CONST |
-		    R500_ALPHA_ADDR2(0) |
-		    R500_ALPHA_ADDR2_CONST);
-
-		e32(R500_ALU_RGB_SEL_A_SRC0 |
-		    R500_ALU_RGB_R_SWIZ_A_R |
-		    R500_ALU_RGB_G_SWIZ_A_G |
-		    R500_ALU_RGB_B_SWIZ_A_B |
-		    R500_ALU_RGB_SEL_B_SRC0 |
-		    R500_ALU_RGB_R_SWIZ_B_R |
-		    R500_ALU_RGB_B_SWIZ_B_G |
-		    R500_ALU_RGB_G_SWIZ_B_B);
-
-		e32(R500_ALPHA_OP_CMP |
-		    R500_ALPHA_SWIZ_A_A |
-		    R500_ALPHA_SWIZ_B_A);
-
-		e32(R500_ALU_RGBA_OP_CMP |
-		    R500_ALU_RGBA_R_SWIZ_0 |
-		    R500_ALU_RGBA_G_SWIZ_0 |
-		    R500_ALU_RGBA_B_SWIZ_0 |
-		    R500_ALU_RGBA_A_SWIZ_0);
+
+		BEGIN_BATCH(7);
+		OUT_BATCH_REGSEQ(R500_US_CONFIG, 2);
+		OUT_BATCH(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
+		OUT_BATCH(0x0);
+		OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3);
+		OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1));
+		OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1));
+		OUT_BATCH(R500_US_CODE_OFFSET_ADDR(0));
+		END_BATCH();
+
+		r500fp.check = check_r500fp;
+		r500fp.cmd = _cmd;
+		r500fp.cmd[0] = cmdr500fp(r300->radeon.radeonScreen, 0, 1, 0, 0);
+		r500fp.cmd[1] = R500_INST_TYPE_OUT |
+			R500_INST_TEX_SEM_WAIT |
+			R500_INST_LAST |
+			R500_INST_RGB_OMASK_R |
+			R500_INST_RGB_OMASK_G |
+			R500_INST_RGB_OMASK_B |
+			R500_INST_ALPHA_OMASK |
+			R500_INST_RGB_CLAMP |
+			R500_INST_ALPHA_CLAMP;
+		r500fp.cmd[2] = R500_RGB_ADDR0(0) |
+			R500_RGB_ADDR1(0) |
+			R500_RGB_ADDR1_CONST |
+			R500_RGB_ADDR2(0) |
+			R500_RGB_ADDR2_CONST;
+		r500fp.cmd[3] = R500_ALPHA_ADDR0(0) |
+			R500_ALPHA_ADDR1(0) |
+			R500_ALPHA_ADDR1_CONST |
+			R500_ALPHA_ADDR2(0) |
+			R500_ALPHA_ADDR2_CONST;
+		r500fp.cmd[4] = R500_ALU_RGB_SEL_A_SRC0 |
+			R500_ALU_RGB_R_SWIZ_A_R |
+			R500_ALU_RGB_G_SWIZ_A_G |
+			R500_ALU_RGB_B_SWIZ_A_B |
+			R500_ALU_RGB_SEL_B_SRC0 |
+			R500_ALU_RGB_R_SWIZ_B_R |
+			R500_ALU_RGB_B_SWIZ_B_G |
+			R500_ALU_RGB_G_SWIZ_B_B;
+		r500fp.cmd[5] = R500_ALPHA_OP_CMP |
+			R500_ALPHA_SWIZ_A_A |
+			R500_ALPHA_SWIZ_B_A;
+		r500fp.cmd[6] = R500_ALU_RGBA_OP_CMP |
+			R500_ALU_RGBA_R_SWIZ_0 |
+			R500_ALU_RGBA_G_SWIZ_0 |
+			R500_ALU_RGBA_B_SWIZ_0 |
+			R500_ALU_RGBA_A_SWIZ_0;
+
+		r500fp.cmd[7] = 0;
+		emit_r500fp(ctx, &r500fp);
 	}
 
-	reg_start(R300_VAP_PVS_STATE_FLUSH_REG, 0);
-	e32(0x00000000);
+	BEGIN_BATCH(2);
+	OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0);
+	END_BATCH();
+
 	if (has_tcl) {
-	    vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
+		vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
 			(5 << R300_PVS_NUM_CNTLRS_SHIFT) |
 			(12 << R300_VF_MAX_VTX_NUM_SHIFT));
-	    if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		vap_cntl |= R500_TCL_STATE_OPTIMIZATION;
-	} else
-	    vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
+		if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
+			vap_cntl |= R500_TCL_STATE_OPTIMIZATION;
+	} else {
+		vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) |
 			(5 << R300_PVS_NUM_CNTLRS_SHIFT) |
 			(5 << R300_VF_MAX_VTX_NUM_SHIFT));
+	}
 
 	if (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV515)
-	    vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT);
 	else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV530) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV560) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV570))
-	    vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT);
 	else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV410) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R420))
-	    vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT);
 	else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R520) ||
 		 (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R580))
-	    vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT);
 	else
-	    vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT);
+		vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT);
 
-	R300_STATECHANGE(rmesa, vap_cntl);
-	reg_start(R300_VAP_CNTL, 0);
-	e32(vap_cntl);
+	R300_STATECHANGE(r300, vap_cntl);
+
+	BEGIN_BATCH(2);
+	OUT_BATCH_REGVAL(R300_VAP_CNTL, vap_cntl);
+	END_BATCH();
 
 	if (has_tcl) {
+        struct radeon_state_atom vpu;
+        uint32_t _cmd[10];
 		R300_STATECHANGE(r300, pvs);
-		reg_start(R300_VAP_PVS_CODE_CNTL_0, 2);
-
-		e32((0 << R300_PVS_FIRST_INST_SHIFT) |
-		    (0 << R300_PVS_XYZW_VALID_INST_SHIFT) |
-		    (1 << R300_PVS_LAST_INST_SHIFT));
-		e32((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-		    (0 << R300_PVS_MAX_CONST_ADDR_SHIFT));
-		e32(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-
 		R300_STATECHANGE(r300, vpi);
-		vsf_start_fragment(0x0, 8);
-
-		e32(PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 0, 0xf, PVS_DST_REG_OUT));
-		e32(PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(0x0);
 
-		e32(PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf, PVS_DST_REG_OUT));
-		e32(PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_SELECT_FORCE_0, PVS_SRC_REG_INPUT, VSF_FLAG_NONE));
-		e32(0x0);
+		BEGIN_BATCH(4);
+		OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3);
+		OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) |
+			  (0 << R300_PVS_XYZW_VALID_INST_SHIFT) |
+			  (1 << R300_PVS_LAST_INST_SHIFT));
+		OUT_BATCH((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
+			  (0 << R300_PVS_MAX_CONST_ADDR_SHIFT));
+		OUT_BATCH(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
+		END_BATCH();
+
+		vpu.check = check_vpu;
+		vpu.cmd = _cmd;
+		vpu.cmd[0] = cmdvpu(r300->radeon.radeonScreen, 0, 2);
+
+		vpu.cmd[1] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE,
+                                         0, 0xf, PVS_DST_REG_OUT);
+		vpu.cmd[2] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y,
+                                      PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W,
+                                      PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
+		vpu.cmd[3] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
+		vpu.cmd[4] = 0x0;
+
+		vpu.cmd[5] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf,
+                                         PVS_DST_REG_OUT);
+		vpu.cmd[6] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X,
+                                      PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z,
+                                      PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT,
+
+                                      VSF_FLAG_NONE);
+		vpu.cmd[7] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_SELECT_FORCE_0,
+                                      PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
+		vpu.cmd[8] = 0x0;
+
+		r300->vap_flush_needed = GL_TRUE;
+		emit_vpu(ctx, &vpu);
 	}
 }
 
-/**
- * Buffer clear
- */
-static void r300Clear(GLcontext * ctx, GLbitfield mask)
+static int r300KernelClear(GLcontext *ctx, GLuint flags)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
-	int flags = 0;
-	int bits = 0;
-	int swapped;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
+	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+	struct radeon_renderbuffer *rrb;
+	struct radeon_renderbuffer *rrbd;
+	int bits = 0, ret;
 
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "r300Clear\n");
+	/* Make sure it fits there. */
+	radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs);
+	
+	if (flags & BUFFER_BIT_COLOR0) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
+	}
 
-	{
-		LOCK_HARDWARE(&r300->radeon);
-		UNLOCK_HARDWARE(&r300->radeon);
-		if (dPriv->numClipRects == 0)
-			return;
+	if (flags & BUFFER_BIT_FRONT_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (mask & BUFFER_BIT_FRONT_LEFT) {
-		flags |= BUFFER_BIT_FRONT_LEFT;
-		mask &= ~BUFFER_BIT_FRONT_LEFT;
+	if (flags & BUFFER_BIT_BACK_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (mask & BUFFER_BIT_BACK_LEFT) {
-		flags |= BUFFER_BIT_BACK_LEFT;
-		mask &= ~BUFFER_BIT_BACK_LEFT;
+	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rrbd) {
+		radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+						  rrbd->bo, 0, RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (mask & BUFFER_BIT_DEPTH) {
+	ret = radeon_cs_space_check(r300->radeon.cmdbuf.cs);
+	if (ret)
+	  return -1;
+
+	rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__);
+	if (flags || bits)
+		r300EmitClearState(ctx);
+
+	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rrbd && (flags & BUFFER_BIT_DEPTH))
 		bits |= CLEARBUFFER_DEPTH;
-		mask &= ~BUFFER_BIT_DEPTH;
-	}
 
-	if ((mask & BUFFER_BIT_STENCIL) && r300->state.stencil.hw_stencil) {
+	if (rrbd && (flags & BUFFER_BIT_STENCIL))
 		bits |= CLEARBUFFER_STENCIL;
-		mask &= ~BUFFER_BIT_STENCIL;
-	}
 
-	if (mask) {
-		if (RADEON_DEBUG & DEBUG_FALLBACKS)
-			fprintf(stderr, "%s: swrast clear, mask: %x\n",
-				__FUNCTION__, mask);
-		_swrast_Clear(ctx, mask);
+	if (flags & BUFFER_BIT_COLOR0) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0);
+		r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL);
+		bits = 0;
 	}
 
-	swapped = r300->radeon.sarea->pfCurrentPage == 1;
-
-	/* Make sure it fits there. */
-	r300EnsureCmdBufSpace(r300, 421 * 3, __FUNCTION__);
-	if (flags || bits)
-		r300EmitClearState(ctx);
-
 	if (flags & BUFFER_BIT_FRONT_LEFT) {
-		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, swapped);
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
+		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
 		bits = 0;
 	}
 
 	if (flags & BUFFER_BIT_BACK_LEFT) {
-		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, swapped ^ 1);
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
+		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
 		bits = 0;
 	}
 
 	if (bits)
-		r300ClearBuffer(r300, bits, 0);
+		r300ClearBuffer(r300, bits, NULL, rrbd);
 
+	COMMIT_BATCH();
+	return 0;
 }
 
-void r300Flush(GLcontext * ctx)
+/**
+ * Buffer clear
+ */
+static void r300Clear(GLcontext * ctx, GLbitfield mask)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
+	const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+	GLbitfield swrast_mask = 0, tri_mask = 0;
+	int i, ret;
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
 
 	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush( rmesa );
-
-	if (rmesa->cmdbuf.count_used > rmesa->cmdbuf.count_reemit)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-}
-
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-
-void r300RefillCurrentDmaRegion(r300ContextPtr rmesa, int size)
-{
-	struct r300_dma_buffer *dmabuf;
-	size = MAX2(size, RADEON_BUFFER_SIZE * 16);
-
-	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->dma.flush) {
-		rmesa->dma.flush(rmesa);
-	}
+		fprintf(stderr, "r300Clear\n");
 
-	if (rmesa->dma.current.buf) {
-#ifdef USER_BUFFERS
-		r300_mem_use(rmesa, rmesa->dma.current.buf->id);
-#endif
-		r300ReleaseDmaRegion(rmesa, &rmesa->dma.current, __FUNCTION__);
+	if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+		LOCK_HARDWARE(&r300->radeon);
+		UNLOCK_HARDWARE(&r300->radeon);
+		if (dPriv->numClipRects == 0)
+			return;
 	}
-	if (rmesa->dma.nr_released_bufs > 4)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-	dmabuf = CALLOC_STRUCT(r300_dma_buffer);
-	dmabuf->buf = (void *)1;	/* hack */
-	dmabuf->refcount = 1;
 
-	dmabuf->id = r300_mem_alloc(rmesa, 4, size);
-	if (dmabuf->id == 0) {
-		LOCK_HARDWARE(&rmesa->radeon);	/* no need to validate */
-
-		r300FlushCmdBufLocked(rmesa, __FUNCTION__);
-		radeonWaitForIdleLocked(&rmesa->radeon);
+	/* Flush swtcl vertices if necessary, because we will change hardware
+	 * state during clear. See also the state-related comment in
+	 * r300EmitClearState.
+	 */
+	R300_NEWPRIM(r300);
 
-		dmabuf->id = r300_mem_alloc(rmesa, 4, size);
+	if (colorMask == ~0)
+	  tri_mask |= (mask & BUFFER_BITS_COLOR);
+	else
+	  tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
 
-		UNLOCK_HARDWARE(&rmesa->radeon);
 
-		if (dmabuf->id == 0) {
-			fprintf(stderr,
-				"Error: Could not get dma buffer... exiting\n");
-			_mesa_exit(-1);
-		}
+	/* HW stencil */
+	if (mask & BUFFER_BIT_STENCIL) {
+		tri_mask |= BUFFER_BIT_STENCIL;
 	}
 
-	rmesa->dma.current.buf = dmabuf;
-	rmesa->dma.current.address = r300_mem_ptr(rmesa, dmabuf->id);
-	rmesa->dma.current.end = size;
-	rmesa->dma.current.start = 0;
-	rmesa->dma.current.ptr = 0;
-}
-
-void r300ReleaseDmaRegion(r300ContextPtr rmesa,
-			  struct r300_dma_region *region, const char *caller)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-	if (!region->buf)
-		return;
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (--region->buf->refcount == 0) {
-		r300_mem_free(rmesa, region->buf->id);
-		FREE(region->buf);
-		rmesa->dma.nr_released_bufs++;
+	/* HW depth */
+	if (mask & BUFFER_BIT_DEPTH) {
+    	        tri_mask |= BUFFER_BIT_DEPTH;
 	}
 
-	region->buf = 0;
-	region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r300AllocDmaRegion(r300ContextPtr rmesa,
-			struct r300_dma_region *region,
-			int bytes, int alignment)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (region->buf)
-		r300ReleaseDmaRegion(rmesa, region, __FUNCTION__);
-
-	alignment--;
-	rmesa->dma.current.start = rmesa->dma.current.ptr =
-	    (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-	if (rmesa->dma.current.ptr + bytes > rmesa->dma.current.end)
-		r300RefillCurrentDmaRegion(rmesa, (bytes + 0x7) & ~0x7);
-
-	region->start = rmesa->dma.current.start;
-	region->ptr = rmesa->dma.current.start;
-	region->end = rmesa->dma.current.start + bytes;
-	region->address = rmesa->dma.current.address;
-	region->buf = rmesa->dma.current.buf;
-	region->buf->refcount++;
-
-	rmesa->dma.current.ptr += bytes;	/* bug - if alignment > 7 */
-	rmesa->dma.current.start =
-	    rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;
-
-	assert(rmesa->dma.current.ptr <= rmesa->dma.current.end);
-}
+	/* If we're doing a tri pass for depth/stencil, include a likely color
+	 * buffer with it.
+	 */
 
-#else
-static void r300RefillCurrentDmaRegion(r300ContextPtr rmesa)
-{
-	struct r300_dma_buffer *dmabuf;
-	int fd = rmesa->radeon.dri.fd;
-	int index = 0;
-	int size = 0;
-	drmDMAReq dma;
-	int ret;
-
-	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->dma.flush) {
-		rmesa->dma.flush(rmesa);
+	for (i = 0; i < BUFFER_COUNT; i++) {
+	  GLuint bufBit = 1 << i;
+	  if ((tri_mask) & bufBit) {
+	    if (!fb->Attachment[i].Renderbuffer->ClassID) {
+	      tri_mask &= ~bufBit;
+	      swrast_mask |= bufBit;
+	    }
+	  }
 	}
 
-	if (rmesa->dma.current.buf)
-		r300ReleaseDmaRegion(rmesa, &rmesa->dma.current, __FUNCTION__);
-
-	if (rmesa->dma.nr_released_bufs > 4)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-	dma.context = rmesa->radeon.dri.hwContext;
-	dma.send_count = 0;
-	dma.send_list = NULL;
-	dma.send_sizes = NULL;
-	dma.flags = 0;
-	dma.request_count = 1;
-	dma.request_size = RADEON_BUFFER_SIZE;
-	dma.request_list = &index;
-	dma.request_sizes = &size;
-	dma.granted_count = 0;
-
-	LOCK_HARDWARE(&rmesa->radeon);	/* no need to validate */
-
-	ret = drmDMA(fd, &dma);
-
-	if (ret != 0) {
-		/* Try to release some buffers and wait until we can't get any more */
-		if (rmesa->dma.nr_released_bufs) {
-			r300FlushCmdBufLocked(rmesa, __FUNCTION__);
-		}
-
-		if (RADEON_DEBUG & DEBUG_DMA)
-			fprintf(stderr, "Waiting for buffers\n");
-
-		radeonWaitForIdleLocked(&rmesa->radeon);
-		ret = drmDMA(fd, &dma);
-
-		if (ret != 0) {
-			UNLOCK_HARDWARE(&rmesa->radeon);
-			fprintf(stderr,
-				"Error: Could not get dma buffer... exiting\n");
-			_mesa_exit(-1);
+	/* SW fallback clearing */
+	swrast_mask = mask & ~tri_mask;
+
+	ret = 0;
+	if (tri_mask) {
+		if (r300->radeon.radeonScreen->kernel_mm)
+			radeonUserClear(ctx, tri_mask);
+		else {
+			/* if kernel clear fails due to size restraints fallback */
+			ret = r300KernelClear(ctx, tri_mask);
+			if (ret < 0)
+				swrast_mask |= tri_mask;
 		}
 	}
 
-	UNLOCK_HARDWARE(&rmesa->radeon);
-
-	if (RADEON_DEBUG & DEBUG_DMA)
-		fprintf(stderr, "Allocated buffer %d\n", index);
-
-	dmabuf = CALLOC_STRUCT(r300_dma_buffer);
-	dmabuf->buf = &rmesa->radeon.radeonScreen->buffers->list[index];
-	dmabuf->refcount = 1;
-
-	rmesa->dma.current.buf = dmabuf;
-	rmesa->dma.current.address = dmabuf->buf->address;
-	rmesa->dma.current.end = dmabuf->buf->total;
-	rmesa->dma.current.start = 0;
-	rmesa->dma.current.ptr = 0;
-}
-
-void r300ReleaseDmaRegion(r300ContextPtr rmesa,
-			  struct r300_dma_region *region, const char *caller)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-	if (!region->buf)
-		return;
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (--region->buf->refcount == 0) {
-		drm_radeon_cmd_header_t *cmd;
-
-		if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
-			fprintf(stderr, "%s -- DISCARD BUF %d\n",
-				__FUNCTION__, region->buf->buf->idx);
-		cmd =
-		    (drm_radeon_cmd_header_t *) r300AllocCmdBuf(rmesa,
-								sizeof
-								(*cmd) / 4,
-								__FUNCTION__);
-		cmd->dma.cmd_type = R300_CMD_DMA_DISCARD;
-		cmd->dma.buf_idx = region->buf->buf->idx;
-
-		FREE(region->buf);
-		rmesa->dma.nr_released_bufs++;
+	if (swrast_mask) {
+		if (RADEON_DEBUG & DEBUG_FALLBACKS)
+			fprintf(stderr, "%s: swrast clear, mask: %x\n",
+				__FUNCTION__, swrast_mask);
+		_swrast_Clear(ctx, swrast_mask);
 	}
-
-	region->buf = 0;
-	region->start = 0;
 }
 
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void r300AllocDmaRegion(r300ContextPtr rmesa,
-			struct r300_dma_region *region,
-			int bytes, int alignment)
-{
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-	if (rmesa->dma.flush)
-		rmesa->dma.flush(rmesa);
-
-	if (region->buf)
-		r300ReleaseDmaRegion(rmesa, region, __FUNCTION__);
-
-	alignment--;
-	rmesa->dma.current.start = rmesa->dma.current.ptr =
-	    (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-	if (rmesa->dma.current.ptr + bytes > rmesa->dma.current.end)
-		r300RefillCurrentDmaRegion(rmesa);
-
-	region->start = rmesa->dma.current.start;
-	region->ptr = rmesa->dma.current.start;
-	region->end = rmesa->dma.current.start + bytes;
-	region->address = rmesa->dma.current.address;
-	region->buf = rmesa->dma.current.buf;
-	region->buf->refcount++;
-
-	rmesa->dma.current.ptr += bytes;	/* bug - if alignment > 7 */
-	rmesa->dma.current.start =
-	    rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;
-
-	assert(rmesa->dma.current.ptr <= rmesa->dma.current.end);
-}
-
-#endif
-
-GLboolean r300IsGartMemory(r300ContextPtr rmesa, const GLvoid * pointer,
-			   GLint size)
-{
-	int offset =
-	    (char *)pointer -
-	    (char *)rmesa->radeon.radeonScreen->gartTextures.map;
-	int valid = (size >= 0 && offset >= 0
-		     && offset + size <
-		     rmesa->radeon.radeonScreen->gartTextures.size);
-
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "r300IsGartMemory( %p ) : %d\n", pointer,
-			valid);
-
-	return valid;
-}
-
-GLuint r300GartOffsetFromVirtual(r300ContextPtr rmesa, const GLvoid * pointer)
-{
-	int offset =
-	    (char *)pointer -
-	    (char *)rmesa->radeon.radeonScreen->gartTextures.map;
-
-	//fprintf(stderr, "offset=%08x\n", offset);
-
-	if (offset < 0
-	    || offset > rmesa->radeon.radeonScreen->gartTextures.size)
-		return ~0;
-	else
-		return rmesa->radeon.radeonScreen->gart_texture_offset + offset;
-}
 
 void r300InitIoctlFuncs(struct dd_function_table *functions)
 {
 	functions->Clear = r300Clear;
 	functions->Finish = radeonFinish;
-	functions->Flush = r300Flush;
+	functions->Flush = radeonFlush;
 }
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.h b/src/mesa/drivers/dri/r300/r300_ioctl.h
index e1143fb6c3..3abfa71a6e 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.h
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.h
@@ -39,22 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_context.h"
 #include "radeon_drm.h"
 
-extern GLboolean r300IsGartMemory(r300ContextPtr rmesa,
-				  const GLvoid * pointer, GLint size);
-
-extern GLuint r300GartOffsetFromVirtual(r300ContextPtr rmesa,
-					const GLvoid * pointer);
-
-extern void r300Flush(GLcontext * ctx);
-
-extern void r300ReleaseDmaRegion(r300ContextPtr rmesa,
-				 struct r300_dma_region *region,
-				 const char *caller);
-extern void r300AllocDmaRegion(r300ContextPtr rmesa,
-			       struct r300_dma_region *region, int bytes,
-			       int alignment);
-
 extern void r300InitIoctlFuncs(struct dd_function_table *functions);
 
-extern void r300RefillCurrentDmaRegion(r300ContextPtr rmesa, int size);
 #endif				/* __R300_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_mem.c b/src/mesa/drivers/dri/r300/r300_mem.c
deleted file mode 100644
index f8f9d4fcdf..0000000000
--- a/src/mesa/drivers/dri/r300/r300_mem.c
+++ /dev/null
@@ -1,385 +0,0 @@
-/*
- * Copyright (C) 2005 Aapo Tahkola.
- *
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining
- * a copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sublicense, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial
- * portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-/**
- * \file
- *
- * \author Aapo Tahkola <aet@rasterburn.org>
- */
-
-#include <unistd.h>
-
-#include "r300_context.h"
-#include "r300_cmdbuf.h"
-#include "r300_ioctl.h"
-#include "r300_mem.h"
-#include "radeon_ioctl.h"
-
-#ifdef USER_BUFFERS
-
-static void resize_u_list(r300ContextPtr rmesa)
-{
-	void *temp;
-	int nsize;
-
-	temp = rmesa->rmm->u_list;
-	nsize = rmesa->rmm->u_size * 2;
-
-	rmesa->rmm->u_list = _mesa_malloc(nsize * sizeof(*rmesa->rmm->u_list));
-	_mesa_memset(rmesa->rmm->u_list, 0,
-		     nsize * sizeof(*rmesa->rmm->u_list));
-
-	if (temp) {
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-		_mesa_memcpy(rmesa->rmm->u_list, temp,
-			     rmesa->rmm->u_size * sizeof(*rmesa->rmm->u_list));
-		_mesa_free(temp);
-	}
-
-	rmesa->rmm->u_size = nsize;
-}
-
-void r300_mem_init(r300ContextPtr rmesa)
-{
-	rmesa->rmm = malloc(sizeof(struct r300_memory_manager));
-	memset(rmesa->rmm, 0, sizeof(struct r300_memory_manager));
-
-	rmesa->rmm->u_size = 128;
-	resize_u_list(rmesa);
-}
-
-void r300_mem_destroy(r300ContextPtr rmesa)
-{
-	_mesa_free(rmesa->rmm->u_list);
-	rmesa->rmm->u_list = NULL;
-
-	_mesa_free(rmesa->rmm);
-	rmesa->rmm = NULL;
-}
-
-void *r300_mem_ptr(r300ContextPtr rmesa, int id)
-{
-	assert(id <= rmesa->rmm->u_last);
-	return rmesa->rmm->u_list[id].ptr;
-}
-
-int r300_mem_find(r300ContextPtr rmesa, void *ptr)
-{
-	int i;
-
-	for (i = 1; i < rmesa->rmm->u_size + 1; i++)
-		if (rmesa->rmm->u_list[i].ptr &&
-		    ptr >= rmesa->rmm->u_list[i].ptr &&
-		    ptr <
-		    rmesa->rmm->u_list[i].ptr + rmesa->rmm->u_list[i].size)
-			break;
-
-	if (i < rmesa->rmm->u_size + 1)
-		return i;
-
-	fprintf(stderr, "%p failed\n", ptr);
-	return 0;
-}
-
-//#define MM_DEBUG
-int r300_mem_alloc(r300ContextPtr rmesa, int alignment, int size)
-{
-	drm_radeon_mem_alloc_t alloc;
-	int offset = 0, ret;
-	int i, free = -1;
-	int done_age;
-	drm_radeon_mem_free_t memfree;
-	int tries = 0;
-	static int bytes_wasted = 0, allocated = 0;
-
-	if (size < 4096)
-		bytes_wasted += 4096 - size;
-
-	allocated += size;
-
-#if 0
-	static int t = 0;
-	if (t != time(NULL)) {
-		t = time(NULL);
-		fprintf(stderr, "slots used %d, wasted %d kb, allocated %d\n",
-			rmesa->rmm->u_last, bytes_wasted / 1024,
-			allocated / 1024);
-	}
-#endif
-
-	memfree.region = RADEON_MEM_REGION_GART;
-
-      again:
-
-	done_age = radeonGetAge((radeonContextPtr) rmesa);
-
-	if (rmesa->rmm->u_last + 1 >= rmesa->rmm->u_size)
-		resize_u_list(rmesa);
-
-	for (i = rmesa->rmm->u_last + 1; i > 0; i--) {
-		if (rmesa->rmm->u_list[i].ptr == NULL) {
-			free = i;
-			continue;
-		}
-
-		if (rmesa->rmm->u_list[i].h_pending == 0 &&
-		    rmesa->rmm->u_list[i].pending
-		    && rmesa->rmm->u_list[i].age <= done_age) {
-			memfree.region_offset =
-			    (char *)rmesa->rmm->u_list[i].ptr -
-			    (char *)rmesa->radeon.radeonScreen->gartTextures.
-			    map;
-
-			ret =
-			    drmCommandWrite(rmesa->radeon.radeonScreen->
-					    driScreen->fd, DRM_RADEON_FREE,
-					    &memfree, sizeof(memfree));
-
-			if (ret) {
-				fprintf(stderr, "Failed to free at %p\n",
-					rmesa->rmm->u_list[i].ptr);
-				fprintf(stderr, "ret = %s\n", strerror(-ret));
-				exit(1);
-			} else {
-#ifdef MM_DEBUG
-				fprintf(stderr, "really freed %d at age %x\n",
-					i,
-					radeonGetAge((radeonContextPtr) rmesa));
-#endif
-				if (i == rmesa->rmm->u_last)
-					rmesa->rmm->u_last--;
-
-				if (rmesa->rmm->u_list[i].size < 4096)
-					bytes_wasted -=
-					    4096 - rmesa->rmm->u_list[i].size;
-
-				allocated -= rmesa->rmm->u_list[i].size;
-				rmesa->rmm->u_list[i].pending = 0;
-				rmesa->rmm->u_list[i].ptr = NULL;
-				free = i;
-			}
-		}
-	}
-	rmesa->rmm->u_head = i;
-
-	if (free == -1) {
-		WARN_ONCE("Ran out of slots!\n");
-		//usleep(100);
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-		tries++;
-		if (tries > 100) {
-			WARN_ONCE("Ran out of slots!\n");
-			exit(1);
-		}
-		goto again;
-	}
-
-	alloc.region = RADEON_MEM_REGION_GART;
-	alloc.alignment = alignment;
-	alloc.size = size;
-	alloc.region_offset = &offset;
-
-	ret =
-	    drmCommandWriteRead(rmesa->radeon.dri.fd, DRM_RADEON_ALLOC, &alloc,
-				sizeof(alloc));
-	if (ret) {
-#if 0
-		WARN_ONCE("Ran out of mem!\n");
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-		//usleep(100);
-		tries2++;
-		tries = 0;
-		if (tries2 > 100) {
-			WARN_ONCE("Ran out of GART memory!\n");
-			exit(1);
-		}
-		goto again;
-#else
-		WARN_ONCE
-		    ("Ran out of GART memory (for %d)!\nPlease consider adjusting GARTSize option.\n",
-		     size);
-		return 0;
-#endif
-	}
-
-	i = free;
-
-	if (i > rmesa->rmm->u_last)
-		rmesa->rmm->u_last = i;
-
-	rmesa->rmm->u_list[i].ptr =
-	    ((GLubyte *) rmesa->radeon.radeonScreen->gartTextures.map) + offset;
-	rmesa->rmm->u_list[i].size = size;
-	rmesa->rmm->u_list[i].age = 0;
-	//fprintf(stderr, "alloc %p at id %d\n", rmesa->rmm->u_list[i].ptr, i);
-
-#ifdef MM_DEBUG
-	fprintf(stderr, "allocated %d at age %x\n", i,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-
-	return i;
-}
-
-void r300_mem_use(r300ContextPtr rmesa, int id)
-{
-	uint64_t ull;
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-	drm_r300_cmd_header_t *cmd;
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (id == 0)
-		return;
-
-	cmd =
-	    (drm_r300_cmd_header_t *) r300AllocCmdBuf(rmesa,
-						      2 + sizeof(ull) / 4,
-						      __FUNCTION__);
-	cmd[0].scratch.cmd_type = R300_CMD_SCRATCH;
-	cmd[0].scratch.reg = R300_MEM_SCRATCH;
-	cmd[0].scratch.n_bufs = 1;
-	cmd[0].scratch.flags = 0;
-	cmd++;
-
-	ull = (uint64_t) (intptr_t) & rmesa->rmm->u_list[id].age;
-	_mesa_memcpy(cmd, &ull, sizeof(ull));
-	cmd += sizeof(ull) / 4;
-
-	cmd[0].u = /*id */ 0;
-
-	LOCK_HARDWARE(&rmesa->radeon);	/* Protect from DRM. */
-	rmesa->rmm->u_list[id].h_pending++;
-	UNLOCK_HARDWARE(&rmesa->radeon);
-}
-
-unsigned long r300_mem_offset(r300ContextPtr rmesa, int id)
-{
-	unsigned long offset;
-
-	assert(id <= rmesa->rmm->u_last);
-
-	offset = (char *)rmesa->rmm->u_list[id].ptr -
-	    (char *)rmesa->radeon.radeonScreen->gartTextures.map;
-	offset += rmesa->radeon.radeonScreen->gart_texture_offset;
-
-	return offset;
-}
-
-void *r300_mem_map(r300ContextPtr rmesa, int id, int access)
-{
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-	void *ptr;
-	int tries = 0;
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (access == R300_MEM_R) {
-
-		if (rmesa->rmm->u_list[id].mapped == 1)
-			WARN_ONCE("buffer %d already mapped\n", id);
-
-		rmesa->rmm->u_list[id].mapped = 1;
-		ptr = r300_mem_ptr(rmesa, id);
-
-		return ptr;
-	}
-
-	if (rmesa->rmm->u_list[id].h_pending)
-		r300FlushCmdBuf(rmesa, __FUNCTION__);
-
-	if (rmesa->rmm->u_list[id].h_pending) {
-		return NULL;
-	}
-
-	while (rmesa->rmm->u_list[id].age >
-	       radeonGetAge((radeonContextPtr) rmesa) && tries++ < 1000)
-		usleep(10);
-
-	if (tries >= 1000) {
-		fprintf(stderr, "Idling failed (%x vs %x)\n",
-			rmesa->rmm->u_list[id].age,
-			radeonGetAge((radeonContextPtr) rmesa));
-		return NULL;
-	}
-
-	if (rmesa->rmm->u_list[id].mapped == 1)
-		WARN_ONCE("buffer %d already mapped\n", id);
-
-	rmesa->rmm->u_list[id].mapped = 1;
-	ptr = r300_mem_ptr(rmesa, id);
-
-	return ptr;
-}
-
-void r300_mem_unmap(r300ContextPtr rmesa, int id)
-{
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (rmesa->rmm->u_list[id].mapped == 0)
-		WARN_ONCE("buffer %d not mapped\n", id);
-
-	rmesa->rmm->u_list[id].mapped = 0;
-}
-
-void r300_mem_free(r300ContextPtr rmesa, int id)
-{
-#ifdef MM_DEBUG
-	fprintf(stderr, "%s: %d at age %x\n", __FUNCTION__, id,
-		radeonGetAge((radeonContextPtr) rmesa));
-#endif
-
-	assert(id <= rmesa->rmm->u_last);
-
-	if (id == 0)
-		return;
-
-	if (rmesa->rmm->u_list[id].ptr == NULL) {
-		WARN_ONCE("Not allocated!\n");
-		return;
-	}
-
-	if (rmesa->rmm->u_list[id].pending) {
-		WARN_ONCE("%p already pended!\n", rmesa->rmm->u_list[id].ptr);
-		return;
-	}
-
-	rmesa->rmm->u_list[id].pending = 1;
-}
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_mem.h b/src/mesa/drivers/dri/r300/r300_mem.h
deleted file mode 100644
index 625a7f6d8d..0000000000
--- a/src/mesa/drivers/dri/r300/r300_mem.h
+++ /dev/null
@@ -1,37 +0,0 @@
-#ifndef __R300_MEM_H__
-#define __R300_MEM_H__
-
-//#define R300_MEM_PDL 0
-#define R300_MEM_UL 1
-
-#define R300_MEM_R 1
-#define R300_MEM_W 2
-#define R300_MEM_RW (R300_MEM_R | R300_MEM_W)
-
-#define R300_MEM_SCRATCH 2
-
-struct r300_memory_manager {
-	struct {
-		void *ptr;
-		uint32_t size;
-		uint32_t age;
-		uint32_t h_pending;
-		int pending;
-		int mapped;
-	} *u_list;
-	int u_head, u_size, u_last;
-
-};
-
-extern void r300_mem_init(r300ContextPtr rmesa);
-extern void r300_mem_destroy(r300ContextPtr rmesa);
-extern void *r300_mem_ptr(r300ContextPtr rmesa, int id);
-extern int r300_mem_find(r300ContextPtr rmesa, void *ptr);
-extern int r300_mem_alloc(r300ContextPtr rmesa, int alignment, int size);
-extern void r300_mem_use(r300ContextPtr rmesa, int id);
-extern unsigned long r300_mem_offset(r300ContextPtr rmesa, int id);
-extern void *r300_mem_map(r300ContextPtr rmesa, int id, int access);
-extern void r300_mem_unmap(r300ContextPtr rmesa, int id);
-extern void r300_mem_free(r300ContextPtr rmesa, int id);
-
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index 8f1a6630d5..357c600af9 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -1467,6 +1467,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_TX_FORMAT_3D		   (1 << 25)
 #	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
 
+#	define R300_TX_FORMAT_GAMMA			(1 << 21)
+
 	/* gap */
 	/* Floating point formats */
 	/* Note - hardware supports both 16 and 32 bit floating point */
@@ -1531,6 +1533,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R500_SEL_FILTER4_TC3		 (3 << 18)
 
 #define R300_TX_OFFSET_0                    0x4540
+#define R300_TX_OFFSET_1                    0x4544
+#define R300_TX_OFFSET_2                    0x4548
+#define R300_TX_OFFSET_3                    0x454C
+#define R300_TX_OFFSET_4                    0x4550
+#define R300_TX_OFFSET_5                    0x4554
+#define R300_TX_OFFSET_6                    0x4558
+#define R300_TX_OFFSET_7                    0x455C
 	/* BEGIN: Guess from R200 */
 #       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
 #       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
@@ -2425,6 +2434,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* Z Buffer Clear Value */
 #define R300_ZB_DEPTHCLEARVALUE                  0x4f28
 
+#define R300_ZB_ZMASK_OFFSET                     0x4f30
+#define R300_ZB_ZMASK_PITCH                      0x4f34
+#define R300_ZB_ZMASK_WRINDEX                    0x4f38
+#define R300_ZB_ZMASK_DWORD                      0x4f3c
+#define R300_ZB_ZMASK_RDINDEX                    0x4f40
+
 /* Hierarchical Z Memory Offset */
 #define R300_ZB_HIZ_OFFSET                       0x4f44
 
@@ -3165,6 +3180,9 @@ enum {
 #   define R300_W_SRC_RAS				(1 << 2)
 
 
+/* Packet0 field ordering to write all values to the same reg */
+#define RADEON_ONE_REG_WR        (1 << 15)
+
 /* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
  * Two parameter dwords:
  * 0. VAP_VTX_FMT: The first parameter is not written to hardware
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 16ce4a1199..08d67b73ed 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -50,6 +50,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * no bugs...
  */
 
+#include "r300_render.h"
+
 #include "main/glheader.h"
 #include "main/state.h"
 #include "main/imports.h"
@@ -66,16 +68,14 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_vp_build.h"
 #include "radeon_reg.h"
 #include "radeon_macros.h"
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "r300_state.h"
 #include "r300_reg.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
-extern int future_hw_tcl_on;
+#include "r300_fragprog_common.h"
+#include "r300_swtcl.h"
 
 /**
  * \brief Convert a OpenGL primitive type into a R300 primitive type.
@@ -172,96 +172,186 @@ int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim)
 	return num_verts - verts_off;
 }
 
-static void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts)
+static void r300EmitElts(GLcontext * ctx, unsigned long n_elts)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_dma_region *rvb = &rmesa->state.elt_dma;
 	void *out;
+	GLuint size;
 
-	if (r300IsGartMemory(rmesa, elts, n_elts * 4)) {
-		rvb->address = rmesa->radeon.radeonScreen->gartTextures.map;
-		rvb->start = ((char *)elts) - rvb->address;
-		rvb->aos_offset =
-		    rmesa->radeon.radeonScreen->gart_texture_offset +
-		    rvb->start;
-		return;
-	} else if (r300IsGartMemory(rmesa, elts, 1)) {
-		WARN_ONCE("Pointer not within GART memory!\n");
-		_mesa_exit(-1);
-	}
-
-	r300AllocDmaRegion(rmesa, rvb, n_elts * 4, 4);
-	rvb->aos_offset = GET_START(rvb);
+	size = ((rmesa->ind_buf.is_32bit ? 4 : 2) * n_elts + 3) & ~3;
 
-	out = rvb->address + rvb->start;
-	memcpy(out, elts, n_elts * 4);
+	radeonAllocDmaRegion(&rmesa->radeon, &rmesa->radeon.tcl.elt_dma_bo,
+			     &rmesa->radeon.tcl.elt_dma_offset, size, 4);
+	radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
+	out = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
+	memcpy(out, rmesa->ind_buf.ptr, size);
+	radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
 }
 
-static void r300FireEB(r300ContextPtr rmesa, unsigned long addr,
-		       int vertex_count, int type)
+static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
-
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0), 0);
-	e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count << 16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
-
-	start_packet3(CP_PACKET3(R300_PACKET3_INDX_BUFFER, 2), 2);
-	e32(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
-	    (R300_VAP_PORT_IDX0 >> 2));
-	e32(addr);
-	e32(vertex_count);
+	BATCH_LOCALS(&rmesa->radeon);
+
+    r300_emit_scissor(rmesa->radeon.glCtx);
+	if (vertex_count > 0) {
+		int size;
+
+		BEGIN_BATCH(10);
+		OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0);
+		if (rmesa->ind_buf.is_32bit) {
+			size = vertex_count;
+			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+			  ((vertex_count + 0) << 16) | type |
+			  R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
+		} else {
+			size = (vertex_count + 1) >> 1;
+			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+			   ((vertex_count + 0) << 16) | type);
+		}
+
+		if (!rmesa->radeon.radeonScreen->kernel_mm) {
+			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
+			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+	    			 (R300_VAP_PORT_IDX0 >> 2));
+			OUT_BATCH_RELOC(rmesa->radeon.tcl.elt_dma_offset,
+					rmesa->radeon.tcl.elt_dma_bo,
+					rmesa->radeon.tcl.elt_dma_offset,
+					RADEON_GEM_DOMAIN_GTT, 0, 0);
+			OUT_BATCH(size);
+		} else {
+			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
+			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+	    			 (R300_VAP_PORT_IDX0 >> 2));
+			OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset);
+			OUT_BATCH(size);
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.elt_dma_bo,
+					      RADEON_GEM_DOMAIN_GTT, 0, 0);
+		}
+		END_BATCH();
+	}
 }
 
 static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 {
+	BATCH_LOCALS(&rmesa->radeon);
+	uint32_t voffset;
 	int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
 	int i;
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
 
 	if (RADEON_DEBUG & DEBUG_VERTS)
 		fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr,
 			offset);
 
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1), sz - 1);
-	e32(nr);
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		BEGIN_BATCH(sz+2+(nr * 2));
+		OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1);
+		OUT_BATCH(nr);
+
+		for (i = 0; i + 1 < nr; i += 2) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+				  (rmesa->radeon.tcl.aos[i].stride << 8) |
+				  (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+				  (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+			OUT_BATCH_RELOC(voffset,
+					rmesa->radeon.tcl.aos[i].bo,
+					voffset,
+					RADEON_GEM_DOMAIN_GTT,
+					0, 0);
+			voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+			  offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+			OUT_BATCH_RELOC(voffset,
+					rmesa->radeon.tcl.aos[i+1].bo,
+					voffset,
+					RADEON_GEM_DOMAIN_GTT,
+					0, 0);
+		}
 
-	for (i = 0; i + 1 < nr; i += 2) {
-		e32((rmesa->state.aos[i].aos_size << 0) |
-		    (rmesa->state.aos[i].aos_stride << 8) |
-		    (rmesa->state.aos[i + 1].aos_size << 16) |
-		    (rmesa->state.aos[i + 1].aos_stride << 24));
+		if (nr & 1) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+				  (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+			voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+			OUT_BATCH_RELOC(voffset,
+					rmesa->radeon.tcl.aos[nr - 1].bo,
+					voffset,
+					RADEON_GEM_DOMAIN_GTT,
+					0, 0);
+		}
+		END_BATCH();
+	} else {
 
-		e32(rmesa->state.aos[i].aos_offset + offset * 4 * rmesa->state.aos[i].aos_stride);
-		e32(rmesa->state.aos[i + 1].aos_offset + offset * 4 * rmesa->state.aos[i + 1].aos_stride);
-	}
+		BEGIN_BATCH(sz+2+(nr * 2));
+		OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1);
+		OUT_BATCH(nr);
+
+		for (i = 0; i + 1 < nr; i += 2) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+				  (rmesa->radeon.tcl.aos[i].stride << 8) |
+				  (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+				  (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+			OUT_BATCH(voffset);
+			voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+			OUT_BATCH(voffset);
+		}
 
-	if (nr & 1) {
-		e32((rmesa->state.aos[nr - 1].aos_size << 0) |
-		    (rmesa->state.aos[nr - 1].aos_stride << 8));
-		e32(rmesa->state.aos[nr - 1].aos_offset + offset * 4 * rmesa->state.aos[nr - 1].aos_stride);
+		if (nr & 1) {
+			OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+			  (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+			voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+			OUT_BATCH(voffset);
+		}
+		for (i = 0; i + 1 < nr; i += 2) {
+			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.aos[i+0].bo,
+					      RADEON_GEM_DOMAIN_GTT,
+					      0, 0);
+			voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.aos[i+1].bo,
+					      RADEON_GEM_DOMAIN_GTT,
+					      0, 0);
+		}
+		if (nr & 1) {
+			voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+				offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+					      rmesa->radeon.tcl.aos[nr-1].bo,
+					      RADEON_GEM_DOMAIN_GTT,
+					      0, 0);
+		}
+		END_BATCH();
 	}
+
 }
 
 static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int type)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
-	drm_radeon_cmd_header_t *cmd = NULL;
+	BATCH_LOCALS(&rmesa->radeon);
 
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0), 0);
-	e32(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (vertex_count << 16) | type);
+    r300_emit_scissor(rmesa->radeon.glCtx);
+	BEGIN_BATCH(3);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
+	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (vertex_count << 16) | type);
+	END_BATCH();
 }
 
-static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
-				   int start, int end, int prim)
+void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 {
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&rmesa->radeon);
 	int type, num_verts;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
 
 	type = r300PrimitiveType(rmesa, prim);
 	num_verts = r300NumVerts(rmesa, end - start, prim);
@@ -269,131 +359,150 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	if (type < 0 || num_verts <= 0)
 		return;
 
-	if (vb->Elts) {
-		if (num_verts > 65535) {
-			/* not implemented yet */
-			WARN_ONCE("Too many elts\n");
-			return;
+	/* Make space for at least 128 dwords.
+	 * This is supposed to ensure that we can get all rendering
+	 * commands into a single command buffer.
+	 */
+	rcommonEnsureCmdBufSpace(&rmesa->radeon, 128, __FUNCTION__);
+
+	if (rmesa->ind_buf.ptr) {
+		r300EmitElts(ctx, num_verts);
+		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, 0);
+		if (rmesa->radeon.radeonScreen->kernel_mm) {
+			BEGIN_BATCH_NO_AUTOSTATE(2);
+			OUT_BATCH_REGSEQ(R300_VAP_VF_MAX_VTX_INDX, 1);
+			OUT_BATCH(rmesa->radeon.tcl.aos[0].count);
+			END_BATCH();
 		}
-		/* Note: The following is incorrect, but it's the best I can do
-		 * without a major refactoring of how DMA memory is handled.
-		 * The problem: Ensuring that both vertex arrays *and* index
-		 * arrays are at the right position, and then ensuring that
-		 * the LOAD_VBPNTR, DRAW_INDX and INDX_BUFFER packets are emitted
-		 * at once.
-		 *
-		 * So why is the following incorrect? Well, it seems like
-		 * allocating the index array might actually evict the vertex
-		 * arrays. *sigh*
-		 */
-		r300EmitElts(ctx, vb->Elts, num_verts);
-		r300EmitAOS(rmesa, rmesa->state.aos_count, start);
-		r300FireEB(rmesa, rmesa->state.elt_dma.aos_offset, num_verts, type);
+		r300FireEB(rmesa, num_verts, type);
 	} else {
-		r300EmitAOS(rmesa, rmesa->state.aos_count, start);
+		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start);
 		r300FireAOS(rmesa, num_verts, type);
 	}
+	COMMIT_BATCH();
 }
 
-static GLboolean r300RunRender(GLcontext * ctx,
-			       struct tnl_pipeline_stage *stage)
+static void r300RunRender(GLcontext * ctx, struct tnl_pipeline_stage *stage)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	int i;
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	struct vertex_buffer *vb = &tnl->vb;
 
-
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
 	r300UpdateShaders(rmesa);
-	if (r300EmitArrays(ctx))
-		return GL_TRUE;
+	r300EmitArrays(ctx);
 
 	r300UpdateShaderStates(rmesa);
 
 	r300EmitCacheFlush(rmesa);
-	r300EmitState(rmesa);
+	radeonEmitState(&rmesa->radeon);
 
 	for (i = 0; i < vb->PrimitiveCount; i++) {
 		GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
 		GLuint start = vb->Primitive[i].start;
 		GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
-		r300RunRenderPrimitive(rmesa, ctx, start, end, prim);
+		r300RunRenderPrimitive(ctx, start, end, prim);
 	}
 
 	r300EmitCacheFlush(rmesa);
 
-#ifdef USER_BUFFERS
-	r300UseArrays(ctx);
-#endif
+	radeonReleaseArrays(ctx, ~0);
+}
 
-	r300ReleaseArrays(ctx);
 
-	return GL_FALSE;
+static const char *getFallbackString(uint32_t bit)
+{
+	switch (bit) {
+		case R300_FALLBACK_VERTEX_PROGRAM :
+			return "vertex program";
+		case R300_FALLBACK_LINE_SMOOTH:
+			return "smooth lines";
+		case R300_FALLBACK_POINT_SMOOTH:
+			return "smooth points";
+		case R300_FALLBACK_POLYGON_SMOOTH:
+			return "smooth polygons";
+		case R300_FALLBACK_LINE_STIPPLE:
+			return "line stipple";
+		case R300_FALLBACK_POLYGON_STIPPLE:
+			return "polygon stipple";
+		case R300_FALLBACK_STENCIL_TWOSIDE:
+			return "two-sided stencil";
+		case R300_FALLBACK_RENDER_MODE:
+			return "render mode != GL_RENDER";
+		case R300_FALLBACK_FRAGMENT_PROGRAM:
+			return "fragment program";
+		case R300_FALLBACK_AOS_LIMIT:
+			return "aos limit";
+		case R300_FALLBACK_INVALID_BUFFERS:
+			return "invalid buffers";
+		default:
+			return "unknown";
+	}
 }
 
-#define FALLBACK_IF(expr)						\
-	do {								\
-		if (expr) {						\
-			if (1 || RADEON_DEBUG & DEBUG_FALLBACKS)	\
-				WARN_ONCE("Software fallback:%s\n",	\
-					  #expr);			\
-			return R300_FALLBACK_RAST;			\
-		}							\
-	} while(0)
-
-static int r300Fallback(GLcontext * ctx)
+void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 {
-	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	const unsigned back = ctx->Stencil._BackFace;
-
-	/* Do we need to use new-style shaders?
-	 * Also is there a better way to do this? */
-	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-		struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-		if (fp) {
-			if (!fp->translated) {
-				r500TranslateFragmentShader(r300, fp);
-				FALLBACK_IF(!fp->translated);
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	uint32_t old_fallback = rmesa->fallback;
+	static uint32_t fallback_warn = 0;
+	
+	if (mode) {
+		if ((fallback_warn & bit) == 0) {
+			if (RADEON_DEBUG & DEBUG_FALLBACKS)
+				_mesa_fprintf(stderr, "WARNING! Falling back to software for %s\n", getFallbackString(bit));
+			fallback_warn |= bit;
+		}
+		rmesa->fallback |= bit;
+
+		/* update only if we change from no tcl fallbacks to some tcl fallbacks */
+		if (rmesa->options.hw_tcl_enabled) {
+			if (((old_fallback & R300_TCL_FALLBACK_MASK) == 0) &&
+				((bit & R300_TCL_FALLBACK_MASK) > 0)) {
+				R300_STATECHANGE(rmesa, vap_cntl_status);
+				rmesa->hw.vap_cntl_status.cmd[1] |= R300_VAP_TCL_BYPASS;
 			}
 		}
+
+		/* update only if we change from no raster fallbacks to some raster fallbacks */
+		if (((old_fallback & R300_RASTER_FALLBACK_MASK) == 0) &&
+			((bit & R300_RASTER_FALLBACK_MASK) > 0)) {
+			
+			radeon_firevertices(&rmesa->radeon);
+			rmesa->radeon.swtcl.RenderIndex = ~0;
+			_swsetup_Wakeup( ctx );
+		}
 	} else {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-		if (fp) {
-			if (!fp->translated) {
-				r300TranslateFragmentShader(r300, fp);
-				FALLBACK_IF(!fp->translated);
+		rmesa->fallback &= ~bit;
+
+		/* update only if we have disabled all tcl fallbacks */
+		if (rmesa->options.hw_tcl_enabled) {
+			if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
+				R300_STATECHANGE(rmesa, vap_cntl_status);
+				rmesa->hw.vap_cntl_status.cmd[1] &= ~R300_VAP_TCL_BYPASS;
 			}
 		}
-	}
-
-	FALLBACK_IF(ctx->RenderMode != GL_RENDER);
 
-	/* If GL_EXT_stencil_two_side is disabled, this fallback check can
-	 * be removed.
-	 */
-	FALLBACK_IF(ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
-		    || ctx->Stencil.ValueMask[0] !=
-		    ctx->Stencil.ValueMask[back]
-		    || ctx->Stencil.WriteMask[0] !=
-		    ctx->Stencil.WriteMask[back]);
-
-	if (ctx->Extensions.NV_point_sprite || ctx->Extensions.ARB_point_sprite)
-		FALLBACK_IF(ctx->Point.PointSprite);
-
-	if (!r300->disable_lowimpact_fallback) {
-		FALLBACK_IF(ctx->Polygon.StippleFlag);
-		FALLBACK_IF(ctx->Multisample._Enabled);
-		FALLBACK_IF(ctx->Line.StippleFlag);
-		FALLBACK_IF(ctx->Line.SmoothFlag);
-		FALLBACK_IF(ctx->Point.SmoothFlag);
+		/* update only if we have disabled all raster fallbacks */
+		if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
+			_swrast_flush( ctx );
+			
+			tnl->Driver.Render.Start = r300RenderStart;
+			tnl->Driver.Render.Finish = r300RenderFinish;
+			tnl->Driver.Render.PrimitiveNotify = r300RenderPrimitive;
+			tnl->Driver.Render.ResetLineStipple = r300ResetLineStipple;
+			tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+			tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+			tnl->Driver.Render.Interp = _tnl_interp;
+			
+			_tnl_invalidate_vertex_state( ctx, ~0 );
+			_tnl_invalidate_vertices( ctx, ~0 );
+		}
 	}
-
-	return R300_FALLBACK_NONE;
+	
 }
 
 static GLboolean r300RunNonTCLRender(GLcontext * ctx,
@@ -404,43 +513,15 @@ static GLboolean r300RunNonTCLRender(GLcontext * ctx,
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
-	if (r300Fallback(ctx) >= R300_FALLBACK_RAST)
+	if (rmesa->fallback & R300_RASTER_FALLBACK_MASK)
 		return GL_TRUE;
 
-	if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
- 	        return GL_TRUE;
-
-	return r300RunRender(ctx, stage);
-}
-
-static GLboolean r300RunTCLRender(GLcontext * ctx,
-				  struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_vertex_program *vp;
-
-	hw_tcl_on = future_hw_tcl_on;
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (hw_tcl_on == GL_FALSE)
-		return GL_TRUE;
-
-	if (r300Fallback(ctx) >= R300_FALLBACK_TCL) {
-		hw_tcl_on = GL_FALSE;
+	if (rmesa->options.hw_tcl_enabled == GL_FALSE)
 		return GL_TRUE;
-	}
 
-	r300UpdateShaders(rmesa);
-
-	vp = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-	if (vp->native == GL_FALSE) {
-		hw_tcl_on = GL_FALSE;
-		return GL_TRUE;
-	}
+	r300RunRender(ctx, stage);
 
-	return r300RunRender(ctx, stage);
+	return GL_FALSE;
 }
 
 const struct tnl_pipeline_stage _r300_render_stage = {
@@ -451,12 +532,3 @@ const struct tnl_pipeline_stage _r300_render_stage = {
 	NULL,
 	r300RunNonTCLRender
 };
-
-const struct tnl_pipeline_stage _r300_tcl_stage = {
-	"r300 Hardware Transform, Clipping and Lighting",
-	NULL,
-	NULL,
-	NULL,
-	NULL,
-	r300RunTCLRender
-};
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
new file mode 100644
index 0000000000..ec785474a6
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -0,0 +1,69 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __R300_RENDER_H__
+#define __R300_RENDER_H__
+
+#include "main/mtypes.h"
+
+#define R300_FALLBACK_VERTEX_PROGRAM    (1 << 0)
+#define R300_TCL_FALLBACK_MASK           0x0000ffff
+
+#define R300_FALLBACK_LINE_SMOOTH       (1 << 16)
+#define R300_FALLBACK_POINT_SMOOTH      (1 << 17)
+#define R300_FALLBACK_POLYGON_SMOOTH    (1 << 18)
+#define R300_FALLBACK_LINE_STIPPLE      (1 << 19)
+#define R300_FALLBACK_POLYGON_STIPPLE   (1 << 20)
+#define R300_FALLBACK_STENCIL_TWOSIDE   (1 << 21)
+#define R300_FALLBACK_RENDER_MODE       (1 << 22)
+#define R300_FALLBACK_FRAGMENT_PROGRAM  (1 << 23)
+#define R300_FALLBACK_AOS_LIMIT         (1 << 30)
+#define R300_FALLBACK_INVALID_BUFFERS   (1 << 31)
+#define R300_RASTER_FALLBACK_MASK        0xffff0000
+
+#define MASK_XYZW (R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W)
+#define MASK_X R300_WRITE_ENA_X
+#define MASK_Y R300_WRITE_ENA_Y
+#define MASK_Z R300_WRITE_ENA_Z
+#define MASK_W R300_WRITE_ENA_W
+
+#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
+    SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
+    SWIZZLE_Z != R300_INPUT_ROUTE_SELECT_Z || \
+    SWIZZLE_W != R300_INPUT_ROUTE_SELECT_W || \
+    SWIZZLE_ZERO != R300_INPUT_ROUTE_SELECT_ZERO || \
+    SWIZZLE_ONE != R300_INPUT_ROUTE_SELECT_ONE
+#error Cannot change these!
+#endif
+
+extern const struct tnl_pipeline_stage _r300_render_stage;
+
+extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
+
+extern void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index f30fd986e0..62228a3786 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -1,47 +1,78 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
 
 #include "main/glheader.h"
 
 #include "shader/program.h"
 #include "tnl/tnl.h"
 #include "r300_context.h"
-#include "r300_fragprog.h"
+#include "r300_fragprog_common.h"
+
+static void freeFragProgCache(GLcontext *ctx, struct r300_fragment_program_cont *cache)
+{
+	struct r300_fragment_program *tmp, *fp = cache->progs;
+
+	while (fp) {
+		tmp = fp->next;
+		_mesa_reference_program(ctx, &fp->Base, NULL);
+		_mesa_free(fp);
+		fp = tmp;
+	}
+}
+
+static void freeVertProgCache(GLcontext *ctx, struct r300_vertex_program_cont *cache)
+{
+	struct r300_vertex_program *tmp, *vp = cache->progs;
+
+	while (vp) {
+		tmp = vp->next;
+		_mesa_reference_vertprog(ctx, &vp->Base, NULL);
+		_mesa_free(vp);
+		vp = tmp;
+	}
+}
 
 static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 					 GLuint id)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program_cont *vp;
-	struct r300_fragment_program *r300_fp;
-	struct r500_fragment_program *r500_fp;
+	struct r300_fragment_program_cont *fp;
 
 	switch (target) {
 	case GL_VERTEX_STATE_PROGRAM_NV:
 	case GL_VERTEX_PROGRAM_ARB:
 		vp = CALLOC_STRUCT(r300_vertex_program_cont);
-		return _mesa_init_vertex_program(ctx, &vp->mesa_program,
-						 target, id);
-	case GL_FRAGMENT_PROGRAM_ARB:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			r500_fp = CALLOC_STRUCT(r500_fragment_program);
-			r500_fp->ctx = ctx;
-			return _mesa_init_fragment_program(ctx, &r500_fp->mesa_program,
-							   target, id);
-		} else {
-			r300_fp = CALLOC_STRUCT(r300_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r300_fp->mesa_program,
-							   target, id);
-		}
+		return _mesa_init_vertex_program(ctx, &vp->mesa_program, target, id);
 
 	case GL_FRAGMENT_PROGRAM_NV:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			r500_fp = CALLOC_STRUCT(r500_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r500_fp->mesa_program,
-							   target, id);
-		} else {
-			r300_fp = CALLOC_STRUCT(r300_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r300_fp->mesa_program,
-							   target, id);
-		}
+	case GL_FRAGMENT_PROGRAM_ARB:
+		fp = CALLOC_STRUCT(r300_fragment_program_cont);
+		return _mesa_init_fragment_program(ctx, &fp->Base, target, id);
+
 	default:
 		_mesa_problem(ctx, "Bad target in r300NewProgram");
 	}
@@ -51,26 +82,35 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 
 static void r300DeleteProgram(GLcontext * ctx, struct gl_program *prog)
 {
+	struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog;
+	struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog;
+
+	switch (prog->Target) {
+		case GL_VERTEX_PROGRAM_ARB:
+			freeVertProgCache(ctx, vp);
+			break;
+		case GL_FRAGMENT_PROGRAM_ARB:
+			freeFragProgCache(ctx, fp);
+			break;
+	}
+
 	_mesa_delete_program(ctx, prog);
 }
 
 static void
 r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_vertex_program_cont *vp = (void *)prog;
-	struct r300_fragment_program *r300_fp = (struct r300_fragment_program *)prog;
-	struct r500_fragment_program *r500_fp = (struct r500_fragment_program *)prog;
+	struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog;
+	struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog;
 
 	switch (target) {
 	case GL_VERTEX_PROGRAM_ARB:
+		freeVertProgCache(ctx, vp);
 		vp->progs = NULL;
 		break;
 	case GL_FRAGMENT_PROGRAM_ARB:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-			r500_fp->translated = GL_FALSE;
-		else
-			r300_fp->translated = GL_FALSE;
+		freeFragProgCache(ctx, fp);
+		fp->progs = NULL;
 		break;
 	}
 
@@ -81,7 +121,19 @@ r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 static GLboolean
 r300IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	return GL_TRUE;
+	if (target == GL_FRAGMENT_PROGRAM_ARB) {
+		struct r300_fragment_program *fp = r300SelectFragmentShader(ctx);
+		if (!fp->translated)
+			r300TranslateFragmentShader(ctx, fp);
+
+		return !fp->error;
+	} else {
+		struct r300_vertex_program *vp = r300SelectVertexShader(ctx);
+		if (!vp->translated)
+			r300TranslateVertexShader(vp);
+
+		return !vp->error;
+	}
 }
 
 void r300InitShaderFuncs(struct dd_function_table *functions)
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 79f0b3625c..12fbf281d9 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/macros.h"
 #include "main/context.h"
 #include "main/dd.h"
+#include "main/framebuffer.h"
 #include "main/simple_list.h"
 #include "main/api_arrayelt.h"
 #include "main/texformat.h"
@@ -52,22 +53,22 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "shader/prog_statevars.h"
 #include "vbo/vbo.h"
 #include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
 
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "r300_state.h"
 #include "r300_reg.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
 #include "r300_tex.h"
+#include "r300_fragprog_common.h"
+#include "r300_fragprog.h"
+#include "r500_fragprog.h"
+#include "r300_render.h"
+#include "r300_vertprog.h"
 
 #include "drirenderbuffer.h"
 
-extern int future_hw_tcl_on;
-extern void _tnl_UpdateFixedFunctionProgram(GLcontext * ctx);
-
 static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4])
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
@@ -366,7 +367,7 @@ static void r300ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 	GLint *ip;
 
 	/* no VAP UCP on non-TCL chipsets */
-	if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (!rmesa->options.hw_tcl_enabled)
 			return;
 
 	p = (GLint) plane - (GLint) GL_CLIP_PLANE0;
@@ -385,7 +386,7 @@ static void r300SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state)
 	GLuint p;
 
 	/* no VAP UCP on non-TCL chipsets */
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (!r300->options.hw_tcl_enabled)
 		return;
 
 	p = cap - GL_CLIP_PLANE0;
@@ -433,6 +434,10 @@ static void r300UpdateCulling(GLcontext * ctx)
 		break;
 	}
 
+	/* Winding is inverted when rendering to FBO */
+	if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+		val ^= R300_FRONT_FACE_CW;
+
 	R300_STATECHANGE(r300, cul);
 	r300->hw.cul.cmd[R300_CUL_CULL] = val;
 }
@@ -453,22 +458,14 @@ static GLboolean current_fragment_program_writes_depth(GLcontext* ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
-	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)
-			(char *)ctx->FragmentProgram._Current;
-		return (fp && fp->WritesDepth);
-	} else {
-		struct r500_fragment_program* fp =
-			(struct r500_fragment_program*)(char*)
-			ctx->FragmentProgram._Current;
-		return (fp && fp->writes_depth);
-	}
+	return ctx->FragmentProgram._Current && r300->selected_fp->writes_depth;
 }
 
 static void r300SetEarlyZState(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	GLuint topZ = R300_ZTOP_ENABLE;
+	GLuint w_fmt, fgdepthsrc;
 
 	if (ctx->Color.AlphaEnabled && ctx->Color.AlphaFunc != GL_ALWAYS)
 		topZ = R300_ZTOP_DISABLE;
@@ -485,6 +482,26 @@ static void r300SetEarlyZState(GLcontext * ctx)
 		R300_STATECHANGE(r300, zstencil_format);
 		r300->hw.zstencil_format.cmd[2] = topZ;
 	}
+
+	/* w_fmt value is set to get best performance
+	* see p.130 R5xx 3D acceleration guide v1.3 */
+	if (current_fragment_program_writes_depth(ctx)) {
+		fgdepthsrc = R300_FG_DEPTH_SRC_SHADER;
+		w_fmt = R300_W_FMT_W24 | R300_W_SRC_US;
+	} else {
+		fgdepthsrc = R300_FG_DEPTH_SRC_SCAN;
+		w_fmt = R300_W_FMT_W0 | R300_W_SRC_US;
+	}
+
+	if (w_fmt != r300->hw.us_out_fmt.cmd[5]) {
+		R300_STATECHANGE(r300, us_out_fmt);
+		r300->hw.us_out_fmt.cmd[5] = w_fmt;
+	}
+
+	if (fgdepthsrc != r300->hw.fg_depth_src.cmd[1]) {
+		R300_STATECHANGE(r300, fg_depth_src);
+		r300->hw.fg_depth_src.cmd[1] = fgdepthsrc;
+	}
 }
 
 static void r300SetAlphaState(GLcontext * ctx)
@@ -535,8 +552,6 @@ static void r300SetAlphaState(GLcontext * ctx)
 	R300_STATECHANGE(r300, at);
 	r300->hw.at.cmd[R300_AT_ALPHA_TEST] = pp_misc;
 	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
-
-	r300SetEarlyZState(ctx);
 }
 
 static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref)
@@ -584,15 +599,35 @@ static void r300SetDepthState(GLcontext * ctx)
 		r300->hw.zs.cmd[R300_ZS_CNTL_1] |=
 		    translate_func(ctx->Depth.Func) << R300_Z_FUNC_SHIFT;
 	}
+}
 
-	r300SetEarlyZState(ctx);
+static void r300CatchStencilFallback(GLcontext *ctx)
+{
+	const unsigned back = ctx->Stencil._BackFace;
+
+	if (ctx->Stencil._Enabled && (ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
+		|| ctx->Stencil.ValueMask[0] != ctx->Stencil.ValueMask[back]
+		|| ctx->Stencil.WriteMask[0] != ctx->Stencil.WriteMask[back])) {
+		r300SwitchFallback(ctx, R300_FALLBACK_STENCIL_TWOSIDE, GL_TRUE);
+	} else {
+		r300SwitchFallback(ctx, R300_FALLBACK_STENCIL_TWOSIDE, GL_FALSE);
+	}
 }
 
 static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	GLboolean hw_stencil = GL_FALSE;
 
-	if (r300->state.stencil.hw_stencil) {
+	r300CatchStencilFallback(ctx);
+
+	if (ctx->DrawBuffer) {
+		struct radeon_renderbuffer *rrbStencil
+			= radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+		hw_stencil = (rrbStencil && rrbStencil->bo);
+	}
+
+	if (hw_stencil) {
 		R300_STATECHANGE(r300, zs);
 		if (state) {
 			r300->hw.zs.cmd[R300_ZS_CNTL_0] |=
@@ -601,10 +636,6 @@ static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 			r300->hw.zs.cmd[R300_ZS_CNTL_0] &=
 			    ~R300_STENCIL_ENABLE;
 		}
-	} else {
-#if R200_MERGED
-		FALLBACK(&r300->radeon, RADEON_FALLBACK_STENCIL, state);
-#endif
 	}
 }
 
@@ -737,7 +768,12 @@ static void r300ColorMask(GLcontext * ctx,
 static void r300PointSize(GLcontext * ctx, GLfloat size)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        /* same size limits for AA, non-AA points */
+
+	/* We need to clamp to user defined range here, because
+	 * the HW clamping happens only for per vertex point size. */
+	size = CLAMP(size, ctx->Point.MinSize, ctx->Point.MaxSize);
+
+	/* same size limits for AA, non-AA points */
 	size = CLAMP(size, ctx->Const.MinPointSize, ctx->Const.MaxPointSize);
 
 	R300_STATECHANGE(r300, ps);
@@ -830,29 +866,33 @@ static void r300ShadeModel(GLcontext * ctx, GLenum mode)
 
 	R300_STATECHANGE(rmesa, shade);
 	rmesa->hw.shade.cmd[1] = 0x00000002;
+	R300_STATECHANGE(rmesa, shade2);
 	switch (mode) {
 	case GL_FLAT:
-		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_FLAT;
+		rmesa->hw.shade2.cmd[1] = R300_RE_SHADE_MODEL_FLAT;
 		break;
 	case GL_SMOOTH:
-		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_SMOOTH;
+		rmesa->hw.shade2.cmd[1] = R300_RE_SHADE_MODEL_SMOOTH;
 		break;
 	default:
 		return;
 	}
-	rmesa->hw.shade.cmd[3] = 0x00000000;
-	rmesa->hw.shade.cmd[4] = 0x00000000;
+	rmesa->hw.shade2.cmd[2] = 0x00000000;
+	rmesa->hw.shade2.cmd[3] = 0x00000000;
 }
 
 static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
 				    GLenum func, GLint ref, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint refmask =
-	    ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
-	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
-	const unsigned back = ctx->Stencil._BackFace;
+	GLuint refmask;
 	GLuint flag;
+	const unsigned back = ctx->Stencil._BackFace;
+
+	r300CatchStencilFallback(ctx);
+
+	refmask = ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
+	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
 
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_0] |= R300_STENCIL_FRONT_BACK;
@@ -880,6 +920,8 @@ static void r300StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
+	r300CatchStencilFallback(ctx);
+
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_2] &=
 	    ~(R300_STENCILREF_MASK <<
@@ -896,6 +938,8 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	const unsigned back = ctx->Stencil._BackFace;
 
+	r300CatchStencilFallback(ctx);
+
 	R300_STATECHANGE(rmesa, zs);
 	/* It is easier to mask what's left.. */
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &=
@@ -924,28 +968,32 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
  * Window position and viewport transformation
  */
 
-/*
- * To correctly position primitives:
- */
-#define SUBPIXEL_X 0.125
-#define SUBPIXEL_Y 0.125
-
 static void r300UpdateWindow(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
 	GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
 	GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
+	const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+	const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
+	GLfloat y_scale, y_bias;
+
+	if (render_to_fbo) {
+		y_scale = 1.0;
+		y_bias = 0;
+	} else {
+		y_scale = -1.0;
+		y_bias = yoffset;
+	}
 
 	GLfloat sx = v[MAT_SX];
-	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
-	GLfloat sy = -v[MAT_SY];
-	GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
-	GLfloat sz = v[MAT_SZ] * rmesa->state.depth.scale;
-	GLfloat tz = v[MAT_TZ] * rmesa->state.depth.scale;
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat sy = v[MAT_SY] * y_scale;
+	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias;
+	GLfloat sz = v[MAT_SZ] * depthScale;
+	GLfloat tz = v[MAT_TZ] * depthScale;
 
-	R300_FIREVERTICES(rmesa);
 	R300_STATECHANGE(rmesa, vpt);
 
 	rmesa->hw.vpt.cmd[R300_VPT_XSCALE] = r300PackFloat32(sx);
@@ -964,6 +1012,8 @@ static void r300Viewport(GLcontext * ctx, GLint x, GLint y,
 	 * values, or keep the originals hanging around.
 	 */
 	r300UpdateWindow(ctx);
+
+	radeon_viewport(ctx, x, y, width, height);
 }
 
 static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
@@ -974,13 +1024,13 @@ static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
 void r300UpdateViewportOffset(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = ((radeonContextPtr) rmesa)->dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
 	GLfloat xoffset = (GLfloat) dPriv->x;
 	GLfloat yoffset = (GLfloat) dPriv->y + dPriv->h;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
 
-	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
-	GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat ty = (-v[MAT_TY]) + yoffset;
 
 	if (rmesa->hw.vpt.cmd[R300_VPT_XOFFSET] != r300PackFloat32(tx) ||
 	    rmesa->hw.vpt.cmd[R300_VPT_YOFFSET] != r300PackFloat32(ty)) {
@@ -996,64 +1046,6 @@ void r300UpdateViewportOffset(GLcontext * ctx)
 	radeonUpdateScissor(ctx);
 }
 
-/**
- * Tell the card where to render (offset, pitch).
- * Effected by glDrawBuffer, etc
- */
-void r300UpdateDrawBuffer(GLcontext * ctx)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	r300ContextPtr r300 = rmesa;
-	struct gl_framebuffer *fb = ctx->DrawBuffer;
-	driRenderbuffer *drb;
-
-	if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-		/* draw to front */
-		drb =
-		    (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].
-		    Renderbuffer;
-	} else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-		/* draw to back */
-		drb =
-		    (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].
-		    Renderbuffer;
-	} else {
-		/* drawing to multiple buffers, or none */
-		return;
-	}
-
-	assert(drb);
-	assert(drb->flippedPitch);
-
-	R300_STATECHANGE(rmesa, cb);
-
-	r300->hw.cb.cmd[R300_CB_OFFSET] = drb->flippedOffset +	//r300->radeon.state.color.drawOffset +
-	    r300->radeon.radeonScreen->fbLocation;
-	r300->hw.cb.cmd[R300_CB_PITCH] = drb->flippedPitch;	//r300->radeon.state.color.drawPitch;
-
-	if (r300->radeon.radeonScreen->cpp == 4)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888;
-	else
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565;
-
-	if (r300->radeon.sarea->tiling_enabled)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE;
-#if 0
-	R200_STATECHANGE(rmesa, ctx);
-
-	/* Note: we used the (possibly) page-flipped values */
-	rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET]
-	    = ((drb->flippedOffset + rmesa->r200Screen->fbLocation)
-	       & R200_COLOROFFSET_MASK);
-	rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch;
-
-	if (rmesa->sarea->tiling_enabled) {
-		rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |=
-		    R200_COLOR_TILE_ENABLE;
-	}
-#endif
-}
-
 static void
 r300FetchStateParameter(GLcontext * ctx,
 			const gl_state_index state[STATE_LENGTH],
@@ -1064,12 +1056,14 @@ r300FetchStateParameter(GLcontext * ctx,
 	switch (state[0]) {
 	case STATE_INTERNAL:
 		switch (state[1]) {
-		case STATE_R300_WINDOW_DIMENSION:
-			value[0] = r300->radeon.dri.drawable->w * 0.5f;	/* width*0.5 */
-			value[1] = r300->radeon.dri.drawable->h * 0.5f;	/* height*0.5 */
-			value[2] = 0.5F;	/* for moving range [-1 1] -> [0 1] */
-			value[3] = 1.0F;	/* not used */
-			break;
+		case STATE_R300_WINDOW_DIMENSION: {
+				__DRIdrawablePrivate * drawable = radeon_get_drawable(&r300->radeon);
+				value[0] = drawable->w * 0.5f;	/* width*0.5 */
+				value[1] = drawable->h * 0.5f;	/* height*0.5 */
+				value[2] = 0.5F;	/* for moving range [-1 1] -> [0 1] */
+				value[3] = 1.0F;	/* not used */
+				break;
+			}
 
 		case STATE_R300_TEXRECT_FACTOR:{
 				struct gl_texture_object *t =
@@ -1103,24 +1097,25 @@ r300FetchStateParameter(GLcontext * ctx,
  * Update R300's own internal state parameters.
  * For now just STATE_R300_WINDOW_DIMENSION
  */
-void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
+static void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 {
-	struct r300_fragment_program *fp;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct gl_program_parameter_list *paramList;
 	GLuint i;
 
-	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM)))
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
 		return;
 
-	fp = (struct r300_fragment_program *)ctx->FragmentProgram._Current;
-	if (!fp)
+	if (!ctx->FragmentProgram._Current || !rmesa->selected_fp)
 		return;
 
-	paramList = fp->mesa_program.Base.Parameters;
+	paramList = rmesa->selected_fp->Base->Parameters;
 
 	if (!paramList)
 		return;
 
+	_mesa_load_state_parameters(ctx, paramList);
+
 	for (i = 0; i < paramList->NumParameters; i++) {
 		if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) {
 			r300FetchStateParameter(ctx,
@@ -1235,9 +1230,7 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-	struct r300_fragment_program_code *code = &fp->code;
+	struct r300_fragment_program_code *code = &r300->selected_fp->code.r300;
 
 	R300_STATECHANGE(r300, fpt);
 
@@ -1271,15 +1264,15 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 	}
 
 	r300->hw.fpt.cmd[R300_FPT_CMD_0] =
-		cmdpacket0(R300_US_TEX_INST_0, code->tex.length);
+		cmdpacket0(r300->radeon.radeonScreen,
+                   R300_US_TEX_INST_0, code->tex.length);
 }
 
 static void r500SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-	struct r500_fragment_program_code *code = &fp->code;
+	struct r500_fragment_program_code *code = &r300->selected_fp->code.r500;
 
 	/* find all the texture instructions and relocate the texture units */
 	for (i = 0; i < code->inst_end + 1; i++) {
@@ -1322,13 +1315,11 @@ static GLuint translate_lod_bias(GLfloat bias)
 static void r300SetupTextures(GLcontext * ctx)
 {
 	int i, mtu;
-	struct r300_tex_obj *t;
+	struct radeon_tex_obj *t;
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int hw_tmu = 0;
 	int last_hw_tmu = -1;	/* -1 translates into no setup costs for fields */
 	int tmu_mappings[R300_MAX_TEXTURE_UNITS] = { -1, };
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
 
 	R300_STATECHANGE(r300, txe);
 	R300_STATECHANGE(r300, tex.filter);
@@ -1356,21 +1347,16 @@ static void r300SetupTextures(GLcontext * ctx)
 	/* We cannot let disabled tmu offsets pass DRM */
 	for (i = 0; i < mtu; i++) {
 		if (ctx->Texture.Unit[i]._ReallyEnabled) {
-
-#if 0				/* Enables old behaviour */
-			hw_tmu = i;
-#endif
 			tmu_mappings[i] = hw_tmu;
 
-			t = (r300TexObjPtr) r300->state.texture.unit[i].texobj->DriverData;
-			/* XXX questionable fix for bug 9170: */
+			t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
 			if (!t)
 				continue;
 
-			if ((t->format & 0xffffff00) == 0xffffff00) {
+			if ((t->pp_txformat & 0xffffff00) == 0xffffff00) {
 				WARN_ONCE
 				    ("unknown texture format (entry %x) encountered. Help me !\n",
-				     t->format & 0xff);
+				     t->pp_txformat & 0xff);
 			}
 
 			if (RADEON_DEBUG & DEBUG_STATE)
@@ -1381,29 +1367,28 @@ static void r300SetupTextures(GLcontext * ctx)
 
 			r300->hw.tex.filter.cmd[R300_TEX_VALUE_0 +
 						hw_tmu] =
-			    gen_fixed_filter(t->filter) | (hw_tmu << 28);
+			    gen_fixed_filter(t->pp_txfilter) | (hw_tmu << 28);
 			/* Note: There is a LOD bias per texture unit and a LOD bias
 			 * per texture object. We add them here to get the correct behaviour.
 			 * (The per-texture object LOD bias was introduced in OpenGL 1.4
 			 * and is not present in the EXT_texture_object extension).
 			 */
 			r300->hw.tex.filter_1.cmd[R300_TEX_VALUE_0 + hw_tmu] =
-				t->filter_1 |
-				translate_lod_bias(ctx->Texture.Unit[i].LodBias + t->base.tObj->LodBias);
+				t->pp_txfilter_1 |
+				translate_lod_bias(ctx->Texture.Unit[i].LodBias + t->base.LodBias);
 			r300->hw.tex.size.cmd[R300_TEX_VALUE_0 + hw_tmu] =
-			    t->size;
+			    t->pp_txsize;
 			r300->hw.tex.format.cmd[R300_TEX_VALUE_0 +
-						hw_tmu] = t->format;
+						hw_tmu] = t->pp_txformat;
 			r300->hw.tex.pitch.cmd[R300_TEX_VALUE_0 + hw_tmu] =
-			    t->pitch_reg;
-			r300->hw.tex.offset.cmd[R300_TEX_VALUE_0 +
-						hw_tmu] = t->offset;
+			  t->pp_txpitch;
+			r300->hw.textures[hw_tmu] = t;
 
-			if (t->offset & R300_TXO_MACRO_TILE) {
+			if (t->tile_bits & R300_TXO_MACRO_TILE) {
 				WARN_ONCE("macro tiling enabled!\n");
 			}
 
-			if (t->offset & R300_TXO_MICRO_TILE) {
+			if (t->tile_bits & R300_TXO_MICRO_TILE) {
 				WARN_ONCE("micro tiling enabled!\n");
 			}
 
@@ -1420,37 +1405,33 @@ static void r300SetupTextures(GLcontext * ctx)
 	}
 
 	r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER0_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, last_hw_tmu + 1);
 	r300->hw.tex.filter_1.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FILTER1_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER1_0, last_hw_tmu + 1);
 	r300->hw.tex.size.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_SIZE_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_SIZE_0, last_hw_tmu + 1);
 	r300->hw.tex.format.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FORMAT_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT_0, last_hw_tmu + 1);
 	r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_FORMAT2_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT2_0, last_hw_tmu + 1);
 	r300->hw.tex.offset.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_OFFSET_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_OFFSET_0, last_hw_tmu + 1);
 	r300->hw.tex.chroma_key.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_CHROMA_KEY_0, last_hw_tmu + 1);
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_CHROMA_KEY_0, last_hw_tmu + 1);
 	r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] =
-	    cmdpacket0(R300_TX_BORDER_COLOR_0, last_hw_tmu + 1);
-
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
+	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, last_hw_tmu + 1);
 
 	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		if (fp->mesa_program.UsesKill && last_hw_tmu < 0) {
+		if (ctx->FragmentProgram._Current->UsesKill && last_hw_tmu < 0) {
 			// The KILL operation requires the first texture unit
 			// to be enabled.
 			r300->hw.txe.cmd[R300_TXE_ENABLE] |= 1;
 			r300->hw.tex.filter.cmd[R300_TEX_VALUE_0] = 0;
 			r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
-				cmdpacket0(R300_TX_FILTER0_0, 1);
+				cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, 1);
 		}
-		r300SetupFragmentShaderTextures(ctx, tmu_mappings);
-	} else
-		r500SetupFragmentShaderTextures(ctx, tmu_mappings);
+	}
+	r300->vtbl.SetupFragmentShaderTextures(ctx, tmu_mappings);
 
 	if (RADEON_DEBUG & DEBUG_STATE)
 		fprintf(stderr, "TX_ENABLE: %08x  last_hw_tmu=%d\n",
@@ -1469,26 +1450,21 @@ union r300_outputs_written {
 static void r300SetupRSUnit(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
 	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int col_ip, tex_ip;
 	int rs_tex_count = 0;
-	int i, count, col_fmt;
+	int i, col_fmt, hw_tcl_on;
+
+	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->Base->Base.OutputsWritten;
 	else
-		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset);
+		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	if (ctx->FragmentProgram._Current)
-		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-	else {
-		fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
-		return;		/* This should only ever happen once.. */
-	}
+	InputsRead = r300->selected_fp->Base->InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1507,15 +1483,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R300_RS_INST_COL_ID(col_ip) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL0;
 			++col_ip;
@@ -1527,15 +1495,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL1) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R300_RS_INST_COL_ID(col_ip) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL1;
 			++col_ip;
@@ -1545,6 +1505,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
+	/* We always route 4 texcoord components */
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (! ( InputsRead & FRAG_BIT_TEX(i) ) )
 		    continue;
@@ -1554,64 +1515,27 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		    continue;
 		}
 
-		int swiz;
-
-		/* with TCL we always seem to route 4 components */
-		if (hw_tcl_on)
-		  count = 4;
-		else
-		  count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size;
-
-		switch(count) {
-		case 4: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3); break;
-		case 3: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(R300_RS_SEL_K1); break;
-		default:
-		case 1:
-		case 2: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1); break;
-		};
-
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz | R300_RS_TEX_PTR(rs_tex_count);
+		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count);
 		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
 		InputsRead &= ~(FRAG_BIT_TEX0 << i);
-		rs_tex_count += count;
-		++tex_ip;
-		++fp_reg;
-	}
-
-	if (InputsRead & FRAG_BIT_FOGC) {
-		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |=  R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) |  R300_RS_TEX_PTR(rs_tex_count);
-			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
-			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 4;
-			++tex_ip;
-			++fp_reg;
-		} else {
-			WARN_ONCE("fragprog wants fogc, vp doesn't provide it\n");
-		}
-	}
-
-	if (InputsRead & FRAG_BIT_WPOS) {
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |=  R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) |  R300_RS_TEX_PTR(rs_tex_count);
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
 		rs_tex_count += 4;
 		++tex_ip;
 		++fp_reg;
 	}
-	InputsRead &= ~FRAG_BIT_WPOS;
 
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
-		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001);
+		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_ADDR(0);
+		r300->hw.ri.cmd[R300_RI_INTERP_0] = R300_RS_COL_PTR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001);
 		++col_ip;
 	}
 
 	high_rr = (col_ip > tex_ip) ? col_ip : tex_ip;
-	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT)  | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
+	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
 	r300->hw.rc.cmd[2] |= high_rr - 1;
 
-	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R300_RS_INST_0, high_rr);
+	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, high_rr);
+	r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_IP_0, high_rr);
 
 	if (InputsRead)
 		WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead);
@@ -1620,26 +1544,21 @@ static void r300SetupRSUnit(GLcontext * ctx)
 static void r500SetupRSUnit(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
 	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int col_ip, tex_ip;
 	int rs_tex_count = 0;
-	int i, count, col_fmt;
+	int i, col_fmt, hw_tcl_on;
+
+	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->Base->Base.OutputsWritten;
 	else
-		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset);
+		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	if (ctx->FragmentProgram._Current)
-		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-	else {
-		fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
-		return;		/* This should only ever happen once.. */
-	}
+	InputsRead = r300->selected_fp->Base->InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1658,15 +1577,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R500_RS_INST_COL_ID(col_ip) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL0;
 			++col_ip;
@@ -1678,15 +1589,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL1) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R500_RS_INST_COL_ID(col_ip) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL1;
 			++col_ip;
@@ -1696,7 +1599,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
-
+	/* We always route 4 texcoord components */
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (! ( InputsRead & FRAG_BIT_TEX(i) ) )
 		    continue;
@@ -1706,74 +1609,13 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		    continue;
 		}
 
-		int swiz = 0;
-
-		/* with TCL we always seem to route 4 components */
-		if (hw_tcl_on)
-		  count = 4;
-		else
-		  count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size;
-
-		if (count == 4) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= (rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 3) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 2) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 1) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else {
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		}
-
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz;
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~(FRAG_BIT_TEX0 << i);
-		rs_tex_count += count;
-		++tex_ip;
-		++fp_reg;
-	}
-
-	if (InputsRead & FRAG_BIT_FOGC) {
-		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
-
-			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 4;
-			++tex_ip;
-			++fp_reg;
-		} else {
-			WARN_ONCE("fragprog wants fogc, vp doesn't provide it\n");
-		}
-	}
-
-	if (InputsRead & FRAG_BIT_WPOS) {
 		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
+			((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
+			((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
+			((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
 
 		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
+		InputsRead &= ~(FRAG_BIT_TEX0 << i);
 		rs_tex_count += 4;
 		++tex_ip;
 		++fp_reg;
@@ -1781,72 +1623,25 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
-		r300->hw.rr.cmd[R300_RR_INST_0] |= R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001);
+		r300->hw.rr.cmd[R300_RR_INST_0] = R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_ADDR(0);
+		r300->hw.ri.cmd[R300_RI_INTERP_0] = R500_RS_COL_PTR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001);
 		++col_ip;
 	}
 
 	high_rr = (col_ip > tex_ip) ? col_ip : tex_ip;
-	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT)  | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
-	r300->hw.rc.cmd[2] |= 0xC0 | (high_rr - 1);
+	r300->hw.rc.cmd[1] = (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
+	r300->hw.rc.cmd[2] = 0xC0 | (high_rr - 1);
 
-	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R500_RS_INST_0, high_rr);
+	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, high_rr);
+	r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_IP_0, high_rr);
 
 	if (InputsRead)
 		WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead);
 }
 
-
-
-
-#define bump_vpu_count(ptr, new_count)   do{\
-	drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr));\
-	int _nc=(new_count)/4; \
-	assert(_nc < 256); \
-	if(_nc>_p->vpu.count)_p->vpu.count=_nc;\
-	}while(0)
-
-static INLINE void r300SetupVertexProgramFragment(r300ContextPtr r300, int dest, struct r300_vertex_shader_fragment *vsf)
-{
-	int i;
-
-	if (vsf->length == 0)
-		return;
-
-	if (vsf->length & 0x3) {
-		fprintf(stderr, "VERTEX_SHADER_FRAGMENT must have length divisible by 4\n");
-		_mesa_exit(-1);
-	}
-
-	switch ((dest >> 8) & 0xf) {
-	case 0:
-		R300_STATECHANGE(r300, vpi);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vpi.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-
-	case 2:
-		R300_STATECHANGE(r300, vpp);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vpp.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-	case 4:
-		R300_STATECHANGE(r300, vps);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vps.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-	default:
-		fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
-		_mesa_exit(-1);
-	}
-}
-
 #define MIN3(a, b, c)	((a) < (b) ? MIN2(a, c) : MIN2(b, c))
 
-
-static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
+void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
 			GLuint output_count, GLuint temp_count)
 {
     int vtx_mem_size;
@@ -1870,7 +1665,7 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
     pvs_num_cntrls = MIN2(6, vtx_mem_size/temp_count);
 
     R300_STATECHANGE(rmesa, vap_cntl);
-    if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+    if (rmesa->options.hw_tcl_enabled) {
 	rmesa->hw.vap_cntl.cmd[R300_VAP_CNTL_INSTR] =
 	    (pvs_num_slots << R300_PVS_NUM_SLOTS_SHIFT) |
 	    (pvs_num_cntrls << R300_PVS_NUM_CNTLRS_SHIFT) |
@@ -1900,114 +1695,6 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
 
 }
 
-static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa)
-{
-	struct r300_vertex_shader_state *prog = &(rmesa->state.vertex_shader);
-	GLuint o_reg = 0;
-	GLuint i_reg = 0;
-	int i;
-	int inst_count = 0;
-	int param_count = 0;
-	int program_end = 0;
-
-	for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) {
-		if (rmesa->state.sw_tcl_inputs[i] != -1) {
-			prog->program.body.i[program_end + 0] = PVS_OP_DST_OPERAND(VE_MULTIPLY, GL_FALSE, GL_FALSE, o_reg++, VSF_FLAG_ALL, PVS_DST_REG_OUT);
-			prog->program.body.i[program_end + 1] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			prog->program.body.i[program_end + 2] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			prog->program.body.i[program_end + 3] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			program_end += 4;
-			i_reg++;
-		}
-	}
-
-	prog->program.length = program_end;
-
-	r300SetupVertexProgramFragment(rmesa, R300_PVS_CODE_START,
-				       &(prog->program));
-	inst_count = (prog->program.length / 4) - 1;
-
-	r300VapCntl(rmesa, i_reg, o_reg, 0);
-
-	R300_STATECHANGE(rmesa, pvs);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] =
-	    (0 << R300_PVS_FIRST_INST_SHIFT) |
-	    (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-	    (inst_count << R300_PVS_LAST_INST_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] =
-	    (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-	    (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] =
-	    (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-}
-
-static int bit_count (int x)
-{
-    x = ((x & 0xaaaaaaaaU) >> 1) + (x & 0x55555555U);
-    x = ((x & 0xccccccccU) >> 2) + (x & 0x33333333U);
-    x = (x >> 16) + (x & 0xffff);
-    x = ((x & 0xf0f0) >> 4) + (x & 0x0f0f);
-    return (x >> 8) + (x & 0x00ff);
-}
-
-static void r300SetupRealVertexProgram(r300ContextPtr rmesa)
-{
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_vertex_program *prog = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-	int inst_count = 0;
-	int param_count = 0;
-
-	/* FIXME: r300SetupVertexProgramFragment */
-	R300_STATECHANGE(rmesa, vpp);
-	param_count =
-	    r300VertexProgUpdateParams(ctx,
-				       (struct r300_vertex_program_cont *)
-				       ctx->VertexProgram._Current,
-				       (float *)&rmesa->hw.vpp.
-				       cmd[R300_VPP_PARAM_0]);
-	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
-	param_count /= 4;
-
-	r300SetupVertexProgramFragment(rmesa, R300_PVS_CODE_START, &(prog->program));
-	inst_count = (prog->program.length / 4) - 1;
-
-	r300VapCntl(rmesa, bit_count(prog->key.InputsRead),
-		    bit_count(prog->key.OutputsWritten), prog->num_temporaries);
-
-	R300_STATECHANGE(rmesa, pvs);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] =
-	  (0 << R300_PVS_FIRST_INST_SHIFT) |
-	  (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-	  (inst_count << R300_PVS_LAST_INST_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] =
-	  (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-	  (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] =
-	  (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-}
-
-static void r300SetupVertexProgram(r300ContextPtr rmesa)
-{
-	GLcontext *ctx = rmesa->radeon.glCtx;
-
-	/* Reset state, in case we don't use something */
-	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
-	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
-	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
-
-	/* Not sure why this doesnt work...
-	   0x400 area might have something to do with pixel shaders as it appears right after pfs programming.
-	   0x406 is set to { 0.0, 0.0, 1.0, 0.0 } most of the time but should change with smooth points and in other rare cases. */
-	//setup_vertex_shader_fragment(rmesa, 0x406, &unk4);
-	if (hw_tcl_on && ((struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx))->translated) {
-		r300SetupRealVertexProgram(rmesa);
-	} else {
-		/* FIXME: This needs to be replaced by vertex shader generation code. */
-		r300SetupDefaultVertexProgram(rmesa);
-	}
-
-}
-
 /**
  * Enable/Disable states.
  *
@@ -2015,20 +1702,13 @@ static void r300SetupVertexProgram(r300ContextPtr rmesa)
  */
 static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 {
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	if (RADEON_DEBUG & DEBUG_STATE)
 		fprintf(stderr, "%s( %s = %s )\n", __FUNCTION__,
 			_mesa_lookup_enum_by_nr(cap),
 			state ? "GL_TRUE" : "GL_FALSE");
 
 	switch (cap) {
-	case GL_TEXTURE_1D:
-	case GL_TEXTURE_2D:
-	case GL_TEXTURE_3D:
-		/* empty */
-		break;
-	case GL_FOG:
-		/* empty */
-		break;
 	case GL_ALPHA_TEST:
 		r300SetAlphaState(ctx);
 		break;
@@ -2046,22 +1726,46 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 	case GL_CLIP_PLANE5:
 		r300SetClipPlaneState(ctx, cap, state);
 		break;
+	case GL_CULL_FACE:
+		r300UpdateCulling(ctx);
+		break;
 	case GL_DEPTH_TEST:
 		r300SetDepthState(ctx);
 		break;
-	case GL_STENCIL_TEST:
-		r300SetStencilState(ctx, state);
+	case GL_LINE_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_LINE_SMOOTH, ctx->Line.SmoothFlag);
 		break;
-	case GL_CULL_FACE:
-		r300UpdateCulling(ctx);
+	case GL_LINE_STIPPLE:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_LINE_STIPPLE, ctx->Line.StippleFlag);
+		break;
+	case GL_POINT_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POINT_SMOOTH, ctx->Point.SmoothFlag);
+		break;
+	case GL_POLYGON_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POLYGON_SMOOTH, ctx->Polygon.SmoothFlag);
+		break;
+	case GL_POLYGON_STIPPLE:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POLYGON_STIPPLE, ctx->Polygon.StippleFlag);
 		break;
 	case GL_POLYGON_OFFSET_POINT:
 	case GL_POLYGON_OFFSET_LINE:
 	case GL_POLYGON_OFFSET_FILL:
 		r300SetPolygonOffsetState(ctx, state);
 		break;
+	case GL_SCISSOR_TEST:
+		radeon_firevertices(&rmesa->radeon);
+		rmesa->radeon.state.scissor.enabled = state;
+		radeonUpdateScissor( ctx );
+		break;
+	case GL_STENCIL_TEST:
+		r300SetStencilState(ctx, state);
+		break;
 	default:
-		radeonEnable(ctx, cap, state);
 		break;
 	}
 }
@@ -2072,15 +1776,14 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 static void r300ResetHwState(r300ContextPtr r300)
 {
 	GLcontext *ctx = r300->radeon.glCtx;
-	int has_tcl = 1;
+	int has_tcl;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
 	if (RADEON_DEBUG & DEBUG_STATE)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
-	r300UpdateWindow(ctx);
+	radeon_firevertices(&r300->radeon);
 
 	r300ColorMask(ctx,
 		      ctx->Color.ColorMask[RCOMP],
@@ -2102,8 +1805,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300UpdateCulling(ctx);
 
-	r300UpdateTextureState(ctx);
-
 	r300SetBlendState(ctx);
 	r300SetLogicOpState(ctx);
 
@@ -2182,8 +1883,8 @@ static void r300ResetHwState(r300ContextPtr r300)
 	}
 
 	/* XXX: Enable anti-aliasing? */
-	r300->hw.gb_misc.cmd[R300_GB_MISC_AA_CONFIG] = GB_AA_CONFIG_AA_DISABLE;
-	r300->hw.gb_misc.cmd[R300_GB_MISC_SELECT] = 0;
+	r300->hw.gb_misc2.cmd[R300_GB_MISC2_AA_CONFIG] = GB_AA_CONFIG_AA_DISABLE;
+	r300->hw.gb_misc2.cmd[R300_GB_MISC2_SELECT] = 0;
 
 	r300->hw.ga_point_s0.cmd[1] = r300PackFloat32(0.0);
 	r300->hw.ga_point_s0.cmd[2] = r300PackFloat32(0.0);
@@ -2242,20 +1943,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300BlendColor(ctx, ctx->Color.BlendColor);
 
-	/* Again, r300ClearBuffer uses this */
-	r300->hw.cb.cmd[R300_CB_OFFSET] =
-	    r300->radeon.state.color.drawOffset +
-	    r300->radeon.radeonScreen->fbLocation;
-	r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.state.color.drawPitch;
-
-	if (r300->radeon.radeonScreen->cpp == 4)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888;
-	else
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565;
-
-	if (r300->radeon.sarea->tiling_enabled)
-		r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE;
-
 	r300->hw.rb3d_dither_ctl.cmd[1] = 0;
 	r300->hw.rb3d_dither_ctl.cmd[2] = 0;
 	r300->hw.rb3d_dither_ctl.cmd[3] = 0;
@@ -2268,44 +1955,18 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300->hw.rb3d_aaresolve_ctl.cmd[1] = 0;
 
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
-
-	r300->hw.zb.cmd[R300_ZB_OFFSET] =
-	    r300->radeon.radeonScreen->depthOffset +
-	    r300->radeon.radeonScreen->fbLocation;
-	r300->hw.zb.cmd[R300_ZB_PITCH] = r300->radeon.radeonScreen->depthPitch;
-
-	if (r300->radeon.sarea->tiling_enabled) {
-		/* XXX: Turn off when clearing buffers ? */
-		r300->hw.zb.cmd[R300_ZB_PITCH] |= R300_DEPTHMACROTILE_ENABLE;
-
-		if (ctx->Visual.depthBits == 24)
-			r300->hw.zb.cmd[R300_ZB_PITCH] |=
-			    R300_DEPTHMICROTILE_TILED;
-	}
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
 
 	r300->hw.zb_depthclearvalue.cmd[1] = 0;
 
-	switch (ctx->Visual.depthBits) {
-	case 16:
-		r300->hw.zstencil_format.cmd[1] = R300_DEPTHFORMAT_16BIT_INT_Z;
-		break;
-	case 24:
-		r300->hw.zstencil_format.cmd[1] = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL;
-		break;
-	default:
-		fprintf(stderr, "Error: Unsupported depth %d... exiting\n", ctx->Visual.depthBits);
-		_mesa_exit(-1);
-	}
-
 	r300->hw.zstencil_format.cmd[2] = R300_ZTOP_DISABLE;
 	r300->hw.zstencil_format.cmd[3] = 0x00000003;
 	r300->hw.zstencil_format.cmd[4] = 0x00000000;
 	r300SetEarlyZState(ctx);
 
-	r300->hw.unk4F30.cmd[1] = 0;
-	r300->hw.unk4F30.cmd[2] = 0;
+	r300->hw.zb_zmask.cmd[1] = 0;
+	r300->hw.zb_zmask.cmd[2] = 0;
 
 	r300->hw.zb_hiz_offset.cmd[1] = 0;
 
@@ -2319,49 +1980,59 @@ static void r300ResetHwState(r300ContextPtr r300)
 		r300->hw.vps.cmd[R300_VPS_ZERO_3] = 0;
 	}
 
-	r300->hw.all_dirty = GL_TRUE;
+	r300->radeon.hw.all_dirty = GL_TRUE;
 }
 
 void r300UpdateShaders(r300ContextPtr rmesa)
 {
-	GLcontext *ctx;
-	struct r300_vertex_program *vp;
-	int i;
+	GLcontext *ctx = rmesa->radeon.glCtx;
 
-	ctx = rmesa->radeon.glCtx;
+	/* should only happenen once, just after context is created */
+	/* TODO: shouldn't we fallback to sw here? */
+	if (!ctx->FragmentProgram._Current) {
+		_mesa_fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
+		return;
+	}
 
-	if (rmesa->NewGLState && hw_tcl_on) {
-		rmesa->NewGLState = 0;
+	{
+		struct r300_fragment_program *fp;
 
-		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-			rmesa->temp_attrib[i] =
-			    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
-			TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
-			    &rmesa->dummy_attrib[i];
-		}
+		fp = r300SelectFragmentShader(ctx);
+		if (!fp->translated)
+			r300TranslateFragmentShader(ctx, fp);
 
-		_tnl_UpdateFixedFunctionProgram(ctx);
+		r300SwitchFallback(ctx, R300_FALLBACK_FRAGMENT_PROGRAM, fp->error);
+	}
 
-		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-			TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
-			    rmesa->temp_attrib[i];
-		}
+	if (rmesa->options.hw_tcl_enabled) {
+		struct r300_vertex_program *vp;
 
-		r300SelectVertexShader(rmesa);
-		vp = (struct r300_vertex_program *)
-		    CURRENT_VERTEX_SHADER(ctx);
-		/*if (vp->translated == GL_FALSE)
-		   r300TranslateVertexShader(vp); */
-		if (vp->translated == GL_FALSE) {
-			fprintf(stderr, "Failing back to sw-tcl\n");
-			hw_tcl_on = future_hw_tcl_on = 0;
-			r300ResetHwState(rmesa);
-
-			r300UpdateStateParameters(ctx, _NEW_PROGRAM);
-			return;
+		if (rmesa->radeon.NewGLState) {
+			int i;
+			for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+				rmesa->temp_attrib[i] =
+				    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
+				TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
+				    &rmesa->dummy_attrib[i];
+			}
+
+			_tnl_UpdateFixedFunctionProgram(ctx);
+
+			for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+				TNL_CONTEXT(ctx)->vb.AttribPtr[i] =
+				    rmesa->temp_attrib[i];
+			}
 		}
+
+		vp = r300SelectVertexShader(ctx);
+		if (!vp->translated)
+			r300TranslateVertexShader(vp);
+
+		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, vp->error);
 	}
-	r300UpdateStateParameters(ctx, _NEW_PROGRAM);
+
+	r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+	rmesa->radeon.NewGLState = 0;
 }
 
 static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
@@ -2385,35 +2056,23 @@ static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
 }
 
 
-static void r300SetupPixelShader(r300ContextPtr rmesa)
+static void r300SetupPixelShader(GLcontext *ctx)
 {
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = rmesa->selected_fp;
 	struct r300_fragment_program_code *code;
 	int i, k;
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
-	r300TranslateFragmentShader(rmesa, fp);
-	if (!fp->translated) {
-		fprintf(stderr, "%s: No valid fragment shader, exiting\n",
-			__FUNCTION__);
-		return;
-	}
-	code = &fp->code;
-
-	r300SetupTextures(ctx);
+	code = &fp->code.r300;
 
 	R300_STATECHANGE(rmesa, fpi[0]);
 	R300_STATECHANGE(rmesa, fpi[1]);
 	R300_STATECHANGE(rmesa, fpi[2]);
 	R300_STATECHANGE(rmesa, fpi[3]);
-	rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_INST_0, code->alu.length);
-	rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_ADDR_0, code->alu.length);
-	rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_INST_0, code->alu.length);
-	rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
+	rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_RGB_INST_0, code->alu.length);
+	rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_RGB_ADDR_0, code->alu.length);
+	rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_ALPHA_INST_0, code->alu.length);
+	rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
 	for (i = 0; i < code->alu.length; i++) {
 		rmesa->hw.fpi[0].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst0;
 		rmesa->hw.fpi[1].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst1;
@@ -2444,10 +2103,9 @@ static void r300SetupPixelShader(r300ContextPtr rmesa)
 	}
 
 	R300_STATECHANGE(rmesa, fpp);
-	rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(R300_PFS_PARAM_0_X, code->const_nr * 4);
+	rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_PFS_PARAM_0_X, code->const_nr * 4);
 	for (i = 0; i < code->const_nr; i++) {
-		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->mesa_program.Base, code->constant[i]);
+		const GLfloat *constant = get_fragmentprogram_constant(ctx, fp->Base, code->constant[i]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat24(constant[0]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat24(constant[1]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat24(constant[2]);
@@ -2469,29 +2127,17 @@ static void r300SetupPixelShader(r300ContextPtr rmesa)
 	if(_nc>_p->r500fp.count)_p->r500fp.count=_nc;\
 } while(0)
 
-static void r500SetupPixelShader(r300ContextPtr rmesa)
+static void r500SetupPixelShader(GLcontext *ctx)
 {
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = rmesa->selected_fp;
 	int i;
 	struct r500_fragment_program_code *code;
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
 	((drm_r300_cmd_header_t *) rmesa->hw.r500fp.cmd)->r500fp.count = 0;
 	((drm_r300_cmd_header_t *) rmesa->hw.r500fp_const.cmd)->r500fp.count = 0;
 
-	r500TranslateFragmentShader(rmesa, fp);
-	if (!fp->translated) {
-		fprintf(stderr, "%s: No valid fragment shader, exiting\n",
-			__FUNCTION__);
-		return;
-	}
-	code = &fp->code;
-
-	r300SetupTextures(ctx);
+	code = &fp->code.r500;
 
 	R300_STATECHANGE(rmesa, fp);
 	rmesa->hw.fp.cmd[R500_FP_PIXSIZE] = code->max_temp_idx;
@@ -2520,59 +2166,93 @@ static void r500SetupPixelShader(r300ContextPtr rmesa)
 
 	R300_STATECHANGE(rmesa, r500fp_const);
 	for (i = 0; i < code->const_nr; i++) {
-		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->mesa_program.Base, code->constant[i]);
+		const GLfloat *constant = get_fragmentprogram_constant(ctx, fp->Base, code->constant[i]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat32(constant[0]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat32(constant[1]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat32(constant[2]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 3] = r300PackFloat32(constant[3]);
 	}
 	bump_r500fp_const_count(rmesa->hw.r500fp_const.cmd, code->const_nr * 4);
-
 }
 
-void r300UpdateShaderStates(r300ContextPtr rmesa)
+void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten)
 {
-	GLcontext *ctx;
-	ctx = rmesa->radeon.glCtx;
+	r300ContextPtr rmesa = R300_CONTEXT( ctx );
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
+	int i, j, reg_count;
+	uint32_t *vir0 = &rmesa->hw.vir[0].cmd[1];
+	uint32_t *vir1 = &rmesa->hw.vir[1].cmd[1];
 
-	r300UpdateTextureState(ctx);
-	r300SetEarlyZState(ctx);
+	for (i = 0; i < R300_VIR_CMDSIZE-1; ++i)
+		vir0[i] = vir1[i] = 0;
 
-	/* w_fmt value is set to get best performance
-	 * see p.130 R5xx 3D acceleration guide v1.3 */
-	GLuint w_fmt, fgdepthsrc;
-	if (current_fragment_program_writes_depth(ctx)) {
-		fgdepthsrc = R300_FG_DEPTH_SRC_SHADER;
-		w_fmt = R300_W_FMT_W24 | R300_W_SRC_US;
-	} else {
-		fgdepthsrc = R300_FG_DEPTH_SRC_SCAN;
-		w_fmt = R300_W_FMT_W0 | R300_W_SRC_US;
+	for (i = 0, j = 0; i < rmesa->vbuf.num_attribs; ++i) {
+		int tmp;
+
+		tmp = attrs[i].data_type | (attrs[i].dst_loc << R300_DST_VEC_LOC_SHIFT);
+		if (attrs[i]._signed)
+			tmp |= R300_SIGNED;
+		if (attrs[i].normalize)
+			tmp |= R300_NORMALIZE;
+
+		if (i % 2 == 0) {
+			vir0[j] = tmp << R300_DATA_TYPE_0_SHIFT;
+			vir1[j] = attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT);
+		} else {
+			vir0[j] |= tmp << R300_DATA_TYPE_1_SHIFT;
+			vir1[j] |= (attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
+			++j;
+		}
 	}
 
-	if (w_fmt != rmesa->hw.us_out_fmt.cmd[5]) {
-		R300_STATECHANGE(rmesa, us_out_fmt);
-		rmesa->hw.us_out_fmt.cmd[5] = w_fmt;
+	reg_count = (rmesa->vbuf.num_attribs + 1) >> 1;
+	if (rmesa->vbuf.num_attribs % 2 != 0) {
+		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
+	} else {
+		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
 	}
 
-	if (fgdepthsrc != rmesa->hw.fg_depth_src.cmd[1]) {
-		R300_STATECHANGE(rmesa, fg_depth_src);
-		rmesa->hw.fg_depth_src.cmd[1] = fgdepthsrc;
+	R300_STATECHANGE(rmesa, vir[0]);
+	R300_STATECHANGE(rmesa, vir[1]);
+	R300_STATECHANGE(rmesa, vof);
+	R300_STATECHANGE(rmesa, vic);
+
+	if (rmesa->radeon.radeonScreen->kernel_mm) {
+		rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
+		rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
+		rmesa->hw.vir[0].cmd[0] |= (reg_count & 0x3FFF) << 16;
+		rmesa->hw.vir[1].cmd[0] |= (reg_count & 0x3FFF) << 16;
+	} else {
+		((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count = reg_count;
+		((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count = reg_count;
 	}
 
-	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		r500SetupPixelShader(rmesa);
-	else
-		r300SetupPixelShader(rmesa);
+	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
+	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten);
+}
 
-	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		r500SetupRSUnit(ctx);
-	else
-		r300SetupRSUnit(ctx);
+void r300UpdateShaderStates(r300ContextPtr rmesa)
+{
+	GLcontext *ctx;
+	ctx = rmesa->radeon.glCtx;
 
-	if ((rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		r300SetupVertexProgram(rmesa);
+	/* should only happenen once, just after context is created */
+	if (!ctx->FragmentProgram._Current)
+		return;
+
+	r300SetEarlyZState(ctx);
 
+	r300SetupTextures(ctx);
+
+	rmesa->vtbl.SetupPixelShader(ctx);
+
+	rmesa->vtbl.SetupRSUnit(ctx);
+
+	if (rmesa->options.hw_tcl_enabled) {
+		r300SetupVertexProgram(rmesa);
+	}
 }
 
 /**
@@ -2586,15 +2266,16 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
 	_swsetup_InvalidateState(ctx, new_state);
 	_vbo_InvalidateState(ctx, new_state);
 	_tnl_InvalidateState(ctx, new_state);
-	_ae_invalidate_state(ctx, new_state);
 
-	if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
-		r300UpdateDrawBuffer(ctx);
-	}
+	if (new_state & _NEW_BUFFERS) {
+		_mesa_update_framebuffer(ctx);
+		/* this updates the DrawBuffer's Width/Height if it's a FBO */
+		_mesa_update_draw_buffer_bounds(ctx);
 
-	r300UpdateStateParameters(ctx, new_state);
+		R300_STATECHANGE(r300, cb);
+	}
 
-	r300->NewGLState |= new_state;
+	r300->radeon.NewGLState |= new_state;
 }
 
 /**
@@ -2604,58 +2285,12 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
  */
 void r300InitState(r300ContextPtr r300)
 {
-	GLcontext *ctx = r300->radeon.glCtx;
-	GLuint depth_fmt;
-
-	radeonInitState(&r300->radeon);
-
-	switch (ctx->Visual.depthBits) {
-	case 16:
-		r300->state.depth.scale = 1.0 / (GLfloat) 0xffff;
-		depth_fmt = R300_DEPTHFORMAT_16BIT_INT_Z;
-		break;
-	case 24:
-		r300->state.depth.scale = 1.0 / (GLfloat) 0xffffff;
-		depth_fmt = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL;
-		break;
-	default:
-		fprintf(stderr, "Error: Unsupported depth %d... exiting\n",
-			ctx->Visual.depthBits);
-		_mesa_exit(-1);
-	}
-
-	/* Only have hw stencil when depth buffer is 24 bits deep */
-	r300->state.stencil.hw_stencil = (ctx->Visual.stencilBits > 0 &&
-					  ctx->Visual.depthBits == 24);
-
-	memset(&(r300->state.texture), 0, sizeof(r300->state.texture));
-
 	r300ResetHwState(r300);
 }
 
 static void r300RenderMode(GLcontext * ctx, GLenum mode)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	(void)rmesa;
-	(void)mode;
-}
-
-void r300UpdateClipPlanes( GLcontext *ctx )
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint p;
-
-	for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
-		if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
-			GLint *ip = (GLint *)ctx->Transform._ClipUserPlane[p];
-
-			R300_STATECHANGE( rmesa, vpucp[p] );
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_X] = ip[0];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_Y] = ip[1];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_Z] = ip[2];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_W] = ip[3];
-		}
-	}
+	r300SwitchFallback(ctx, R300_FALLBACK_RENDER_MODE, ctx->RenderMode != GL_RENDER);
 }
 
 /**
@@ -2663,7 +2298,6 @@ void r300UpdateClipPlanes( GLcontext *ctx )
  */
 void r300InitStateFuncs(struct dd_function_table *functions)
 {
-	radeonInitStateFuncs(functions);
 
 	functions->UpdateState = r300InvalidateState;
 	functions->AlphaFunc = r300AlphaFunc;
@@ -2699,4 +2333,25 @@ void r300InitStateFuncs(struct dd_function_table *functions)
 	functions->RenderMode = r300RenderMode;
 
 	functions->ClipPlane = r300ClipPlane;
+	functions->Scissor = radeonScissor;
+
+	functions->DrawBuffer		= radeonDrawBuffer;
+	functions->ReadBuffer		= radeonReadBuffer;
+}
+
+void r300InitShaderFunctions(r300ContextPtr r300)
+{
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		r300->vtbl.SetupRSUnit = r500SetupRSUnit;
+		r300->vtbl.SetupPixelShader = r500SetupPixelShader;
+		r300->vtbl.SetupFragmentShaderTextures = r500SetupFragmentShaderTextures;
+		r300->vtbl.BuildFragmentProgramHwCode = r500BuildFragmentProgramHwCode;
+		r300->vtbl.FragmentProgramDump = r500FragmentProgramDump;
+	} else {
+		r300->vtbl.SetupRSUnit = r300SetupRSUnit;
+		r300->vtbl.SetupPixelShader = r300SetupPixelShader;
+		r300->vtbl.SetupFragmentShaderTextures = r300SetupFragmentShaderTextures;
+		r300->vtbl.BuildFragmentProgramHwCode = r300BuildFragmentProgramHwCode;
+		r300->vtbl.FragmentProgramDump = r300FragmentProgramDump;
+	}
 }
diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h
index 0589ab7cad..d46bf9f179 100644
--- a/src/mesa/drivers/dri/r300/r300_state.h
+++ b/src/mesa/drivers/dri/r300/r300_state.h
@@ -39,42 +39,24 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define R300_NEWPRIM( rmesa )			\
   do {						\
-    if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );		\
+  if ( rmesa->radeon.dma.flush )			\
+    rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
   } while (0)
 
 #define R300_STATECHANGE(r300, atom) \
 	do {						\
 	  R300_NEWPRIM(r300);				\
 		r300->hw.atom.dirty = GL_TRUE;		\
-		r300->hw.is_dirty = GL_TRUE;		\
+		r300->radeon.hw.is_dirty = GL_TRUE;		\
 	} while(0)
 
-#define R300_PRINT_STATE(r300, atom) \
-		r300PrintStateAtom(r300, &r300->hw.atom)
-
-/* Fire the buffered vertices no matter what.
-   TODO: This has not been implemented yet
- */
-#define R300_FIREVERTICES( r300 )			\
-do {							\
-    \
-   if ( (r300)->cmdbuf.count_used || (r300)->dma.flush ) {	\
-      r300Flush( (r300)->radeon.glCtx );		\
-   }							\
-    \
-} while (0)
-
-// r300_state.c
-extern int future_hw_tcl_on;
-void _tnl_UpdateFixedFunctionProgram (GLcontext * ctx);
 void r300UpdateViewportOffset (GLcontext * ctx);
 void r300UpdateDrawBuffer (GLcontext * ctx);
-void r300UpdateStateParameters (GLcontext * ctx, GLuint new_state);
 void r300UpdateShaders (r300ContextPtr rmesa);
 void r300UpdateShaderStates (r300ContextPtr rmesa);
 void r300InitState (r300ContextPtr r300);
-void r300UpdateClipPlanes (GLcontext * ctx);
 void r300InitStateFuncs (struct dd_function_table *functions);
+void r300VapCntl(r300ContextPtr rmesa, GLuint input_count, GLuint output_count, GLuint temp_count);
+void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten);
 
 #endif				/* __R300_STATE_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index ba3621b16b..56ed519cf4 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -28,362 +28,231 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Dave Airlie <airlied@linux.ie>
+ *   Maciej Cencora <m.cencora@gmail.com>
  */
 
-/* derived from r200 swtcl path */
-
-
-
-#include "main/glheader.h"
-#include "main/mtypes.h"
-#include "main/colormac.h"
-#include "main/enums.h"
-#include "main/image.h"
-#include "main/imports.h"
-#include "main/light.h"
-#include "main/macros.h"
-
-#include "swrast/s_context.h"
-#include "swrast/s_fog.h"
-#include "swrast_setup/swrast_setup.h"
-#include "math/m_translate.h"
 #include "tnl/tnl.h"
-#include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
 
-#include "r300_context.h"
-#include "r300_swtcl.h"
 #include "r300_state.h"
-#include "r300_ioctl.h"
+#include "r300_swtcl.h"
 #include "r300_emit.h"
-#include "r300_mem.h"
-
-static void flush_last_swtcl_prim( r300ContextPtr rmesa  );
+#include "r300_tex.h"
+#include "r300_render.h"
 
-
-void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, GLuint offset);
-void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr);
 #define EMIT_ATTR( ATTR, STYLE )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
+} while (0)
+
+#define ADD_ATTR(_attr, _format, _dst_loc, _swizzle, _write_mask, _normalize) \
+do { \
+	attrs[num_attrs].element = (_attr); \
+	attrs[num_attrs].data_type = (_format); \
+	attrs[num_attrs].dst_loc = (_dst_loc); \
+	attrs[num_attrs].swizzle = (_swizzle); \
+	attrs[num_attrs].write_mask = (_write_mask); \
+	attrs[num_attrs]._signed = 0; \
+	attrs[num_attrs].normalize = (_normalize); \
+	++num_attrs; \
 } while (0)
 
-static void r300SetVertexFormat( GLcontext *ctx )
+void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_OutputsWritten)
 {
 	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	struct vertex_buffer *VB = &tnl->vb;
-	DECLARE_RENDERINPUTS(index_bitset);
-	GLuint InputsRead = 0, OutputsWritten = 0;
-	int vap_fmt_1 = 0;
-	int offset = 0;
-	int vte = 0;
-	int fog_id;
-	GLint inputs[VERT_ATTRIB_MAX];
-	GLint tab[VERT_ATTRIB_MAX];
-	int swizzle[VERT_ATTRIB_MAX][4];
-	GLuint i, nr;
-	GLuint sz;
-
-	DECLARE_RENDERINPUTS(render_inputs_bitset);
-	RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset);
-	RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );
-	RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
-
-	vte = rmesa->hw.vte.cmd[1];
-	vte &= ~(R300_VTX_XY_FMT | R300_VTX_Z_FMT | R300_VTX_W0_FMT);
-	/* Important:
-	 */
-	if ( VB->NdcPtr != NULL ) {
-		VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
-		vte |= R300_VTX_XY_FMT | R300_VTX_Z_FMT;
-	}
-	else {
-		VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
-		vte |= R300_VTX_W0_FMT;
-	}
-
-	assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-	rmesa->swtcl.vertex_attr_count = 0;
-
-	/* EMIT_ATTR's must be in order as they tell t_vertex.c how to
-	 * build up a hardware vertex.
-	 */
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POS)) {
-		sz = VB->AttribPtr[VERT_ATTRIB_POS]->size;
-		InputsRead |= 1 << VERT_ATTRIB_POS;
-		OutputsWritten |= 1 << VERT_RESULT_HPOS;
-		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_1F + sz - 1 );
-		offset = sz;
-	} else {
-		offset = 4;
-		EMIT_PAD(4 * sizeof(float));
-	}
-/*
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) {
-		EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F );
-		offset += 1;
-	}
-*/
-	if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR0)) {
-		sz = VB->AttribPtr[VERT_ATTRIB_COLOR0]->size;
-	        rmesa->swtcl.coloroffset = offset;
+	int first_free_tex = 0;
+	GLuint InputsRead = 0;
+	GLuint OutputsWritten = 0;
+	int num_attrs = 0;
+	GLuint fp_reads = rmesa->selected_fp->Base->InputsRead;
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
+
+	rmesa->swtcl.coloroffset = rmesa->swtcl.specoffset = 0;
+	rmesa->radeon.swtcl.vertex_attr_count = 0;
+
+	/* We always want non Ndc coords format */
+	VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
+
+	/* Always write position vector */
+	InputsRead |= 1 << VERT_ATTRIB_POS;
+	OutputsWritten |= 1 << VERT_RESULT_HPOS;
+	EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F );
+	ADD_ATTR(VERT_ATTRIB_POS, R300_DATA_TYPE_FLOAT_4, SWTCL_OVM_POS, SWIZZLE_XYZW, MASK_XYZW, 0);
+	rmesa->swtcl.coloroffset = 4;
+
+	if (fp_reads & FRAG_BIT_COL0) {
 		InputsRead |= 1 << VERT_ATTRIB_COLOR0;
 		OutputsWritten |= 1 << VERT_RESULT_COL0;
-		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_1F + sz - 1 );
-		offset += sz;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_COLOR0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_COLOR0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW, 1);
+#endif
 	}
 
-	rmesa->swtcl.specoffset = 0;
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
-		sz = VB->AttribPtr[VERT_ATTRIB_COLOR1]->size;
-		rmesa->swtcl.specoffset = offset;
-		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_1F + sz - 1 );
+	if (fp_reads & FRAG_BIT_COL1) {
+		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
 		InputsRead |= 1 << VERT_ATTRIB_COLOR1;
 		OutputsWritten |= 1 << VERT_RESULT_COL1;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_COLOR1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR1, swiz, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_COLOR1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR1, swiz, MASK_XYZW, 1);
+#endif
+		rmesa->swtcl.specoffset = rmesa->swtcl.coloroffset + 1;
 	}
 
-	fog_id = -1;
-	if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_FOG)) {
-		/* find first free tex coord slot */
-		if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-			int i;
-			for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-				if (!RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
-					fog_id = i;
-					break;
-				}
-			}
-		} else {
-			fog_id = 0;
+	if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
+		VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1];
+		OutputsWritten |= 1 << VERT_RESULT_BFC0;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
+#endif
+		if (fp_reads & FRAG_BIT_COL1) {
+			VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1];
+			GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			OutputsWritten |= 1 << VERT_RESULT_BFC1;
+#if MESA_LITTLE_ENDIAN
+			EMIT_ATTR( _TNL_ATTRIB_GENERIC1, EMIT_4UB_4F_RGBA );
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
+#else
+			EMIT_ATTR( _TNL_ATTRIB_GENERIC1, EMIT_4UB_4F_ABGR );
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
+#endif
 		}
+	}
 
-		if (fog_id == -1) {
-			fprintf(stderr, "\tout of free texcoords to do fog\n");
-			_mesa_exit(-1);
-		}
+	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_POINTSIZE )) {
+		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO);
+		InputsRead |= 1 << VERT_ATTRIB_POINT_SIZE;
+		OutputsWritten |= 1 << VERT_RESULT_PSIZ;
+		EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F );
+		ADD_ATTR(VERT_ATTRIB_POINT_SIZE, R300_DATA_TYPE_FLOAT_1, SWTCL_OVM_POINT_SIZE, swiz, MASK_X, 0);
+	}
+
+	if (rmesa->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0;
 
-		sz = VB->AttribPtr[VERT_ATTRIB_FOG]->size;
-		EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1F + sz - 1);
-		InputsRead |= 1 << VERT_ATTRIB_FOG;
-		OutputsWritten |= 1 << VERT_RESULT_FOGC;
-		vap_fmt_1 |= sz << (3 * fog_id);
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
 	}
 
-	if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-		int i;
+	if (rmesa->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
+		int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0;
 
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-			if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
-				sz = VB->TexCoordPtr[i]->size;
-				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
-				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-				EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1 );
-				vap_fmt_1 |= sz << (3 * i);
-			}
-		}
+		VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG];
+		RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id);
 	}
 
-	/* RS can't put fragment position on the pixel stack, so stuff it in texcoord if needed */
-	if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POS) && (ctx->FragmentProgram._Current->Base.InputsRead & FRAG_BIT_WPOS)) {
-		int first_free_tex = -1;
-		if (fog_id >= 0) {
-			first_free_tex = fog_id+1;
-		} else {
-			if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-				int i;
-				for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-					if (!RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
-						first_free_tex = i;
+	/**
+	 *  Sending only one texcoord component may lead to lock up,
+	 *  so for all textures always output 4 texcoord components to RS.
+	 */
+	{
+		int i;
+		GLuint swiz, format, hw_format;
+		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+			if (fp_reads & FRAG_BIT_TEX(i)) {
+				switch (VB->TexCoordPtr[i]->size) {
+					case 1:
+						format = EMIT_1F;
+						hw_format = R300_DATA_TYPE_FLOAT_1;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+						break;
+					case 2:
+						format = EMIT_2F;
+						hw_format = R300_DATA_TYPE_FLOAT_2;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
+						break;
+					case 3:
+						format = EMIT_3F;
+						hw_format = R300_DATA_TYPE_FLOAT_3;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+						break;
+					case 4:
+						format = EMIT_4F;
+						hw_format = R300_DATA_TYPE_FLOAT_4;
+						swiz = SWIZZLE_XYZW;
 						break;
-					}
+					default:
+						continue;
 				}
-			} else {
-				first_free_tex = 0;
+				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
+				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
+				EMIT_ATTR(_TNL_ATTRIB_TEX(i), format);
+				ADD_ATTR(VERT_ATTRIB_TEX0 + i, hw_format, SWTCL_OVM_TEX(first_free_tex), swiz, MASK_XYZW, 0);
+				++first_free_tex;
 			}
 		}
-
-		if (first_free_tex == -1) {
-			fprintf(stderr, "\tout of free texcoords to write w pos\n");
-			_mesa_exit(-1);
-		}
-
-		sz = VB->AttribPtr[VERT_ATTRIB_POS]->size;
-		InputsRead |= 1 << (VERT_ATTRIB_TEX0 + first_free_tex);
-		OutputsWritten |= 1 << (VERT_RESULT_TEX0 + first_free_tex);
-		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_1F + sz - 1 );
-		vap_fmt_1 |= sz << (3 * first_free_tex);
-	}
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			inputs[i] = nr++;
-		} else {
-			inputs[i] = -1;
-		}
 	}
 
-	/* Fixed, apply to vir0 only */
-	if (InputsRead & (1 << VERT_ATTRIB_POS))
-		inputs[VERT_ATTRIB_POS] = 0;
-	if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
-		inputs[VERT_ATTRIB_COLOR0] = 2;
-	if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
-		inputs[VERT_ATTRIB_COLOR1] = 3;
-	if (InputsRead & (1 << VERT_ATTRIB_FOG))
-		inputs[VERT_ATTRIB_FOG] = 6 + fog_id;
-	for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
-		if (InputsRead & (1 << i))
-			inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			tab[nr++] = i;
-		}
-	}
-
-	for (i = 0; i < nr; i++) {
-		int ci;
-
-		swizzle[i][0] = SWIZZLE_ZERO;
-		swizzle[i][1] = SWIZZLE_ZERO;
-		swizzle[i][2] = SWIZZLE_ZERO;
-		swizzle[i][3] = SWIZZLE_ONE;
-
-		for (ci = 0; ci < VB->AttribPtr[tab[i]]->size; ci++) {
-			swizzle[i][ci] = ci;
-		}
+	if (first_free_tex >= ctx->Const.MaxTextureUnits) {
+		fprintf(stderr, "\tout of free texcoords to write fog coordinate\n");
+		_mesa_exit(-1);
 	}
 
 	R300_NEWPRIM(rmesa);
-	R300_STATECHANGE(rmesa, vir[0]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
-		r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
-				   VB->AttribPtr, inputs, tab, nr);
-	R300_STATECHANGE(rmesa, vir[1]);
-	((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
-		r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
-				   nr);
-
-	R300_STATECHANGE(rmesa, vic);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-
-	R300_STATECHANGE(rmesa, vof);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = vap_fmt_1;
-
-	rmesa->swtcl.vertex_size =
-		_tnl_install_attrs( ctx,
-				    rmesa->swtcl.vertex_attrs,
-				    rmesa->swtcl.vertex_attr_count,
-				    NULL, 0 );
-
-	rmesa->swtcl.vertex_size /= 4;
+	rmesa->vbuf.num_attribs = num_attrs;
+	*_InputsRead = InputsRead;
+	*_OutputsWritten = OutputsWritten;
 
-	RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
-
-
-	R300_STATECHANGE(rmesa, vte);
-	rmesa->hw.vte.cmd[1] = vte;
-	rmesa->hw.vte.cmd[2] = rmesa->swtcl.vertex_size;
+	RENDERINPUTS_COPY(rmesa->render_inputs_bitset, tnl->render_inputs_bitset);
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( r300ContextPtr rmesa  )
+static void r300PrepareVertices(GLcontext *ctx)
 {
-	if (RADEON_DEBUG & DEBUG_IOCTL)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	rmesa->dma.flush = NULL;
-
-	if (rmesa->dma.current.buf) {
-		struct r300_dma_region *current = &rmesa->dma.current;
-		GLuint current_offset = GET_START(current);
-
-		assert (current->start +
-			rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-			current->ptr);
-
-		if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
-
-			r300EnsureCmdBufSpace( rmesa, rmesa->hw.max_state_size + (12*sizeof(int)), __FUNCTION__);
-
-			r300EmitState(rmesa);
-
-			r300EmitVertexAOS( rmesa,
-					   rmesa->swtcl.vertex_size,
-					   current_offset);
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	GLuint InputsRead, OutputsWritten;
 
-			r300EmitVbufPrim( rmesa,
-					  rmesa->swtcl.hw_primitive,
-					  rmesa->swtcl.numverts);
+	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
-			r300EmitCacheFlush(rmesa);
-		}
+	rmesa->radeon.swtcl.vertex_size =
+		_tnl_install_attrs( ctx,
+				    rmesa->radeon.swtcl.vertex_attrs,
+				    rmesa->radeon.swtcl.vertex_attr_count,
+				    NULL, 0 );
 
-		rmesa->swtcl.numverts = 0;
-		current->start = current->ptr;
-	}
+	rmesa->radeon.swtcl.vertex_size /= 4;
 }
 
-/* Alloc space in the current dma region.
- */
-static void *
-r300AllocDmaLowVerts( r300ContextPtr rmesa, int nverts, int vsize )
-{
-	GLuint bytes = vsize * nverts;
-
-	if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end )
-		r300RefillCurrentDmaRegion( rmesa, bytes);
-
-	if (!rmesa->dma.flush) {
-		rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-		rmesa->dma.flush = flush_last_swtcl_prim;
-	}
-
-	ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
-	ASSERT( rmesa->dma.flush == flush_last_swtcl_prim );
-	ASSERT( rmesa->dma.current.start +
-		rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-		rmesa->dma.current.ptr );
-
-	{
-		GLubyte *head = (GLubyte *) (rmesa->dma.current.address + rmesa->dma.current.ptr);
-		rmesa->dma.current.ptr += bytes;
-		rmesa->swtcl.numverts += nverts;
-		return head;
-	}
-}
 
 static GLuint reduced_prim[] = {
-  GL_POINTS,
-  GL_LINES,
-  GL_LINES,
-  GL_LINES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
-  GL_TRIANGLES,
+	GL_POINTS,
+	GL_LINES,
+	GL_LINES,
+	GL_LINES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
+	GL_TRIANGLES,
 };
 
 static void r300RasterPrimitive( GLcontext *ctx, GLuint prim );
-static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
-//static void r300ResetLineStipple( GLcontext *ctx );
 
 /***********************************************************************
  *                    Emit primitives as inline vertices               *
@@ -405,15 +274,13 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
 #define CTX_ARG r300ContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) r300AllocDmaLowVerts( rmesa, n, size * 4 )
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 )
 #define LOCAL_VARS						\
    r300ContextPtr rmesa = R300_CONTEXT(ctx);		\
-   const char *r300verts = (char *)rmesa->swtcl.verts;
+   const char *r300verts = (char *)rmesa->radeon.swtcl.verts;
 #define VERT(x) (r300Vertex *)(r300verts + ((x) * vertsize * sizeof(int)))
 #define VERTEX r300Vertex
-#define DO_DEBUG_VERTS (1 && (RADEON_DEBUG & DEBUG_VERTS))
-#define PRINT_VERTEX(x)
 #undef TAG
 #define TAG(x) r300_##x
 #include "tnl_dd/t_dd_triemit.h"
@@ -433,9 +300,8 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
  *              Build render functions from dd templates               *
  ***********************************************************************/
 
-#define R300_TWOSIDE_BIT	0x01
-#define R300_UNFILLED_BIT	0x02
-#define R300_MAX_TRIFUNC	0x04
+#define R300_UNFILLED_BIT	0x01
+#define R300_MAX_TRIFUNC	0x02
 
 static struct {
    tnl_points_func	        points;
@@ -446,9 +312,9 @@ static struct {
 
 #define DO_FALLBACK  0
 #define DO_UNFILLED (IND & R300_UNFILLED_BIT)
-#define DO_TWOSIDE  (IND & R300_TWOSIDE_BIT)
+#define DO_TWOSIDE   0
 #define DO_FLAT      0
-#define DO_OFFSET     0
+#define DO_OFFSET    0
 #define DO_TRI       1
 #define DO_QUAD      1
 #define DO_LINE      1
@@ -468,33 +334,39 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + (e*rmesa->swtcl.vertex_size*sizeof(int)))
-
-/* Only used to pull back colors into vertices (ie, we know color is
- * floating point).
- */
-#define R300_COLOR( dst, src )				\
-do {							\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]);	\
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + (e*rmesa->radeon.swtcl.vertex_size*sizeof(int)))
+
+#define VERT_SET_RGBA( v, c ) \
+do { \
+   r300_color_t *color = (r300_color_t *)&((v)->ui[coloroffset]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(color->alpha, (c)[3]); \
 } while (0)
 
-#define VERT_SET_RGBA( v, c )    if (coloroffset) R300_COLOR( v->ub4[coloroffset], c )
-#define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
-#define VERT_SAVE_RGBA( idx )    if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
-#define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
+#define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
 
-#define R300_SPEC( dst, src )				\
-do {							\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);	\
-   UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);	\
+#define VERT_SET_SPEC( v0, c ) \
+do { \
+   if (specoffset) { \
+   UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.red, (c)[0]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.green, (c)[1]); \
+   UNCLAMPED_FLOAT_TO_UBYTE(v0->v.specular.blue, (c)[2]); \
+   } \
 } while (0)
 
-#define VERT_SET_SPEC( v, c )    if (specoffset) R300_SPEC( v->ub4[specoffset], c )
-#define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
+#define VERT_COPY_SPEC( v0, v1 ) \
+do { \
+   if (specoffset) { \
+       v0->v.specular.red = v1->v.specular.red; \
+       v0->v.specular.green = v1->v.specular.green; \
+       v0->v.specular.blue = v1->v.specular.blue; \
+   } \
+} while (0)
+
+#define VERT_SAVE_RGBA( idx )    color[idx] = v[idx]->ui[coloroffset]
+#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
 #define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 
@@ -514,7 +386,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) r300RasterPrimitive( ctx, reduced_prim[x] )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -530,26 +402,15 @@ do {							\
 #define TAG(x) x
 #include "tnl_dd/t_dd_tritmp.h"
 
-#define IND (R300_TWOSIDE_BIT)
-#define TAG(x) x##_twoside
-#include "tnl_dd/t_dd_tritmp.h"
-
 #define IND (R300_UNFILLED_BIT)
 #define TAG(x) x##_unfilled
 #include "tnl_dd/t_dd_tritmp.h"
 
-#define IND (R300_TWOSIDE_BIT|R300_UNFILLED_BIT)
-#define TAG(x) x##_twoside_unfilled
-#include "tnl_dd/t_dd_tritmp.h"
-
-
 
 static void init_rast_tab( void )
 {
    init();
-   init_twoside();
    init_unfilled();
-   init_twoside_unfilled();
 }
 
 /**********************************************************************/
@@ -571,8 +432,8 @@ static void init_rast_tab( void )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
    r300ContextPtr rmesa = R300_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *r300verts = (char *)rmesa->swtcl.verts;		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *r300verts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -601,10 +462,9 @@ static void r300ChooseRenderState( GLcontext *ctx )
 	GLuint index = 0;
 	GLuint flags = ctx->_TriangleCaps;
 
-	if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R300_TWOSIDE_BIT;
 	if (flags & DD_TRI_UNFILLED)      index |= R300_UNFILLED_BIT;
 
-	if (index != rmesa->swtcl.RenderIndex) {
+	if (index != rmesa->radeon.swtcl.RenderIndex) {
 		tnl->Driver.Render.Points = rast_tab[index].points;
 		tnl->Driver.Render.Line = rast_tab[index].line;
 		tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -621,30 +481,34 @@ static void r300ChooseRenderState( GLcontext *ctx )
 			tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
 		}
 
-		rmesa->swtcl.RenderIndex = index;
+		rmesa->radeon.swtcl.RenderIndex = index;
 	}
 }
 
 
-static void r300RenderStart(GLcontext *ctx)
+void r300RenderStart(GLcontext *ctx)
 {
-        r300ContextPtr rmesa = R300_CONTEXT( ctx );
+	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 
 	r300ChooseRenderState(ctx);
-	r300SetVertexFormat(ctx);
 
 	r300UpdateShaders(rmesa);
+
+	r300PrepareVertices(ctx);
+
+	r300ValidateBuffers(ctx);
+
 	r300UpdateShaderStates(rmesa);
 
 	r300EmitCacheFlush(rmesa);
 
-	if (rmesa->dma.flush != 0 &&
-	    rmesa->dma.flush != flush_last_swtcl_prim)
-		rmesa->dma.flush( rmesa );
-
+	/* investigate if we can put back flush optimisation if needed */
+	if (rmesa->radeon.dma.flush != NULL) {
+		rmesa->radeon.dma.flush(ctx);
+	}
 }
 
-static void r300RenderFinish(GLcontext *ctx)
+void r300RenderFinish(GLcontext *ctx)
 {
 }
 
@@ -652,28 +516,26 @@ static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
-	if (rmesa->swtcl.hw_primitive != hwprim) {
-	        R300_NEWPRIM( rmesa );
-		rmesa->swtcl.hw_primitive = hwprim;
+	if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
+		R300_NEWPRIM( rmesa );
+		rmesa->radeon.swtcl.hw_primitive = hwprim;
 	}
 }
 
-static void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
+void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
 {
 
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	rmesa->swtcl.render_primitive = prim;
+	rmesa->radeon.swtcl.render_primitive = prim;
 
 	if ((prim == GL_TRIANGLES) && (ctx->_TriangleCaps & DD_TRI_UNFILLED))
-	  return;
+		return;
 
 	r300RasterPrimitive( ctx, reduced_prim[prim] );
 }
 
-static void r300ResetLineStipple(GLcontext *ctx)
+void r300ResetLineStipple(GLcontext *ctx)
 {
-
-
 }
 
 void r300InitSwtcl(GLcontext *ctx)
@@ -699,50 +561,68 @@ void r300InitSwtcl(GLcontext *ctx)
 	_tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12,
 			    48 * sizeof(GLfloat) );
 
-	rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-	rmesa->swtcl.RenderIndex = ~0;
-	rmesa->swtcl.render_primitive = GL_TRIANGLES;
-	rmesa->swtcl.hw_primitive = 0;
+	rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+	rmesa->radeon.swtcl.RenderIndex = ~0;
+	rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+	rmesa->radeon.swtcl.hw_primitive = 0;
 
 	_tnl_invalidate_vertex_state( ctx, ~0 );
 	_tnl_invalidate_vertices( ctx, ~0 );
-	RENDERINPUTS_ZERO( rmesa->tnl_index_bitset );
 
 	_tnl_need_projected_coords( ctx, GL_FALSE );
-	r300ChooseRenderState(ctx);
 }
 
 void r300DestroySwtcl(GLcontext *ctx)
 {
 }
 
-void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, GLuint offset)
+static void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, GLuint offset)
 {
-	int cmd_reserved = 0;
-	int cmd_written = 0;
+	BATCH_LOCALS(&rmesa->radeon);
 
-	drm_radeon_cmd_header_t *cmd = NULL;
 	if (RADEON_DEBUG & DEBUG_VERTS)
-	  fprintf(stderr, "%s:  vertex_size %d, offset 0x%x \n",
-		  __FUNCTION__, vertex_size, offset);
-
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 2), 2);
-	e32(1);
-	e32(vertex_size | (vertex_size << 8));
-	e32(offset);
+		fprintf(stderr, "%s:  vertex_size %d, offset 0x%x \n",
+			__FUNCTION__, vertex_size, offset);
+
+	BEGIN_BATCH(7);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 2);
+	OUT_BATCH(1);
+	OUT_BATCH(vertex_size | (vertex_size << 8));
+	OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+	END_BATCH();
 }
 
-void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr)
+static void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr)
 {
-
-	int cmd_reserved = 0;
-	int cmd_written = 0;
+	BATCH_LOCALS(&rmesa->radeon);
 	int type, num_verts;
-	drm_radeon_cmd_header_t *cmd = NULL;
 
 	type = r300PrimitiveType(rmesa, primitive);
 	num_verts = r300NumVerts(rmesa, vertex_nr, primitive);
 
-	start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0), 0);
-	e32(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type);
+	BEGIN_BATCH(3);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
+	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type);
+	END_BATCH();
+}
+
+void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
+{
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+
+	rcommonEnsureCmdBufSpace(&rmesa->radeon,
+			   rmesa->radeon.hw.max_state_size + (12*sizeof(int)),
+			   __FUNCTION__);
+	radeonEmitState(&rmesa->radeon);
+    r300_emit_scissor(ctx);
+	r300EmitVertexAOS(rmesa,
+			rmesa->radeon.swtcl.vertex_size,
+			rmesa->radeon.dma.current,
+			current_offset);
+
+	r300EmitVbufPrim(rmesa,
+		   rmesa->radeon.swtcl.hw_primitive,
+		   rmesa->radeon.swtcl.numverts);
+	r300EmitCacheFlush(rmesa);
+	COMMIT_BATCH();
 }
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.h b/src/mesa/drivers/dri/r300/r300_swtcl.h
index 55df53c1ad..c271d26546 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.h
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.h
@@ -39,7 +39,27 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "r300_context.h"
 
+/*
+ * Here are definitions of OVM locations of vertex attributes for non TCL hw
+ */
+#define SWTCL_OVM_POS 0
+#define SWTCL_OVM_COLOR0 2
+#define SWTCL_OVM_COLOR1 3
+#define SWTCL_OVM_COLOR2 4
+#define SWTCL_OVM_COLOR3 5
+#define SWTCL_OVM_TEX(n) ((n) + 6)
+#define SWTCL_OVM_POINT_SIZE 15
+
+extern void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *InputsRead,  GLuint *OutputsWritten);
+
 extern void r300InitSwtcl( GLcontext *ctx );
 extern void r300DestroySwtcl( GLcontext *ctx );
 
+extern void r300RenderStart(GLcontext *ctx);
+extern void r300RenderFinish(GLcontext *ctx);
+extern void r300RenderPrimitive(GLcontext *ctx, GLenum prim);
+extern void r300ResetLineStipple(GLcontext *ctx);
+
+extern void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
+
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c
index 7c699ec572..0af5bb4f46 100644
--- a/src/mesa/drivers/dri/r300/r300_tex.c
+++ b/src/mesa/drivers/dri/r300/r300_tex.c
@@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "main/enums.h"
 #include "main/image.h"
+#include "main/mipmap.h"
 #include "main/simple_list.h"
 #include "main/texformat.h"
 #include "main/texstore.h"
@@ -49,6 +50,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_context.h"
 #include "r300_state.h"
 #include "r300_ioctl.h"
+#include "radeon_mipmap_tree.h"
 #include "r300_tex.h"
 
 #include "xmlpool.h"
@@ -77,20 +79,20 @@ static unsigned int translate_wrap_mode(GLenum wrapmode)
  *
  * \param t Texture object whose wrap modes are to be set
  */
-static void r300UpdateTexWrap(r300TexObjPtr t)
+static void r300UpdateTexWrap(radeonTexObjPtr t)
 {
-	struct gl_texture_object *tObj = t->base.tObj;
+	struct gl_texture_object *tObj = &t->base;
 
-	t->filter &=
+	t->pp_txfilter &=
 	    ~(R300_TX_WRAP_S_MASK | R300_TX_WRAP_T_MASK | R300_TX_WRAP_R_MASK);
 
-	t->filter |= translate_wrap_mode(tObj->WrapS) << R300_TX_WRAP_S_SHIFT;
+	t->pp_txfilter |= translate_wrap_mode(tObj->WrapS) << R300_TX_WRAP_S_SHIFT;
 
 	if (tObj->Target != GL_TEXTURE_1D) {
-		t->filter |= translate_wrap_mode(tObj->WrapT) << R300_TX_WRAP_T_SHIFT;
+		t->pp_txfilter |= translate_wrap_mode(tObj->WrapT) << R300_TX_WRAP_T_SHIFT;
 
 		if (tObj->Target == GL_TEXTURE_3D)
-			t->filter |= translate_wrap_mode(tObj->WrapR) << R300_TX_WRAP_R_SHIFT;
+			t->pp_txfilter |= translate_wrap_mode(tObj->WrapR) << R300_TX_WRAP_R_SHIFT;
 	}
 }
 
@@ -117,10 +119,13 @@ static GLuint aniso_filter(GLfloat anisotropy)
  * \param magf Texture magnification mode
  * \param anisotropy Maximum anisotropy level
  */
-static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat anisotropy)
+static void r300SetTexFilter(radeonTexObjPtr t, GLenum minf, GLenum magf, GLfloat anisotropy)
 {
-	t->filter &= ~(R300_TX_MIN_FILTER_MASK | R300_TX_MIN_FILTER_MIP_MASK | R300_TX_MAG_FILTER_MASK | R300_TX_MAX_ANISO_MASK);
-	t->filter_1 &= ~R300_EDGE_ANISO_EDGE_ONLY;
+	/* Force revalidation to account for switches from/to mipmapping. */
+	t->validated = GL_FALSE;
+
+	t->pp_txfilter &= ~(R300_TX_MIN_FILTER_MASK | R300_TX_MIN_FILTER_MIP_MASK | R300_TX_MAG_FILTER_MASK | R300_TX_MAX_ANISO_MASK);
+	t->pp_txfilter_1 &= ~R300_EDGE_ANISO_EDGE_ONLY;
 
 	/* Note that EXT_texture_filter_anisotropic is extremely vague about
 	 * how anisotropic filtering interacts with the "normal" filter modes.
@@ -128,7 +133,7 @@ static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat
 	 * filter settings completely. This includes driconf's settings.
 	 */
 	if (anisotropy >= 2.0 && (minf != GL_NEAREST) && (magf != GL_NEAREST)) {
-		t->filter |= R300_TX_MAG_FILTER_ANISO
+		t->pp_txfilter |= R300_TX_MAG_FILTER_ANISO
 			| R300_TX_MIN_FILTER_ANISO
 			| R300_TX_MIN_FILTER_MIP_LINEAR
 			| aniso_filter(anisotropy);
@@ -139,22 +144,22 @@ static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat
 
 	switch (minf) {
 	case GL_NEAREST:
-		t->filter |= R300_TX_MIN_FILTER_NEAREST;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_NEAREST;
 		break;
 	case GL_LINEAR:
-		t->filter |= R300_TX_MIN_FILTER_LINEAR;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_LINEAR;
 		break;
 	case GL_NEAREST_MIPMAP_NEAREST:
-		t->filter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_NEAREST;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_NEAREST;
 		break;
 	case GL_NEAREST_MIPMAP_LINEAR:
-		t->filter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_LINEAR;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_NEAREST|R300_TX_MIN_FILTER_MIP_LINEAR;
 		break;
 	case GL_LINEAR_MIPMAP_NEAREST:
-		t->filter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_NEAREST;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_NEAREST;
 		break;
 	case GL_LINEAR_MIPMAP_LINEAR:
-		t->filter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_LINEAR;
+		t->pp_txfilter |= R300_TX_MIN_FILTER_LINEAR|R300_TX_MIN_FILTER_MIP_LINEAR;
 		break;
 	}
 
@@ -163,15 +168,15 @@ static void r300SetTexFilter(r300TexObjPtr t, GLenum minf, GLenum magf, GLfloat
 	 */
 	switch (magf) {
 	case GL_NEAREST:
-		t->filter |= R300_TX_MAG_FILTER_NEAREST;
+		t->pp_txfilter |= R300_TX_MAG_FILTER_NEAREST;
 		break;
 	case GL_LINEAR:
-		t->filter |= R300_TX_MAG_FILTER_LINEAR;
+		t->pp_txfilter |= R300_TX_MAG_FILTER_LINEAR;
 		break;
 	}
 }
 
-static void r300SetTexBorderColor(r300TexObjPtr t, const GLfloat color[4])
+static void r300SetTexBorderColor(radeonTexObjPtr t, const GLfloat color[4])
 {
 	GLubyte c[4];
 	CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
@@ -182,729 +187,6 @@ static void r300SetTexBorderColor(r300TexObjPtr t, const GLfloat color[4])
 }
 
 /**
- * Allocate space for and load the mesa images into the texture memory block.
- * This will happen before drawing with a new texture, or drawing with a
- * texture after it was swapped out or teximaged again.
- */
-
-static r300TexObjPtr r300AllocTexObj(struct gl_texture_object *texObj)
-{
-	r300TexObjPtr t;
-
-	t = CALLOC_STRUCT(r300_tex_obj);
-	texObj->DriverData = t;
-	if (t != NULL) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE) {
-			fprintf(stderr, "%s( %p, %p )\n", __FUNCTION__,
-				(void *)texObj, (void *)t);
-		}
-
-		/* Initialize non-image-dependent parts of the state:
-		 */
-		t->base.tObj = texObj;
-		t->border_fallback = GL_FALSE;
-
-		make_empty_list(&t->base);
-
-		r300UpdateTexWrap(t);
-		r300SetTexFilter(t, texObj->MinFilter, texObj->MagFilter, texObj->MaxAnisotropy);
-		r300SetTexBorderColor(t, texObj->BorderColor);
-	}
-
-	return t;
-}
-
-/* try to find a format which will only need a memcopy */
-static const struct gl_texture_format *r300Choose8888TexFormat(GLenum srcFormat,
-							       GLenum srcType)
-{
-	const GLuint ui = 1;
-	const GLubyte littleEndian = *((const GLubyte *)&ui);
-
-	if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-	    (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
-	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) {
-		return &_mesa_texformat_rgba8888;
-	} else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
-		   (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) ||
-		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
-		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) {
-		return &_mesa_texformat_rgba8888_rev;
-	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
-					    srcType == GL_UNSIGNED_INT_8_8_8_8)) {
-		return &_mesa_texformat_argb8888_rev;
-	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) ||
-					    srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) {
-		return &_mesa_texformat_argb8888;
-	} else
-		return _dri_texformat_argb8888;
-}
-
-static const struct gl_texture_format *r300ChooseTextureFormat(GLcontext * ctx,
-							       GLint
-							       internalFormat,
-							       GLenum format,
-							       GLenum type)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	const GLboolean do32bpt =
-	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32);
-	const GLboolean force16bpt =
-	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16);
-	(void)format;
-
-#if 0
-	fprintf(stderr, "InternalFormat=%s(%d) type=%s format=%s\n",
-		_mesa_lookup_enum_by_nr(internalFormat), internalFormat,
-		_mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
-	fprintf(stderr, "do32bpt=%d force16bpt=%d\n", do32bpt, force16bpt);
-#endif
-
-	switch (internalFormat) {
-	case 4:
-	case GL_RGBA:
-	case GL_COMPRESSED_RGBA:
-		switch (type) {
-		case GL_UNSIGNED_INT_10_10_10_2:
-		case GL_UNSIGNED_INT_2_10_10_10_REV:
-			return do32bpt ? _dri_texformat_argb8888 :
-			    _dri_texformat_argb1555;
-		case GL_UNSIGNED_SHORT_4_4_4_4:
-		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-			return _dri_texformat_argb4444;
-		case GL_UNSIGNED_SHORT_5_5_5_1:
-		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-			return _dri_texformat_argb1555;
-		default:
-			return do32bpt ? r300Choose8888TexFormat(format, type) :
-			    _dri_texformat_argb4444;
-		}
-
-	case 3:
-	case GL_RGB:
-	case GL_COMPRESSED_RGB:
-		switch (type) {
-		case GL_UNSIGNED_SHORT_4_4_4_4:
-		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-			return _dri_texformat_argb4444;
-		case GL_UNSIGNED_SHORT_5_5_5_1:
-		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-			return _dri_texformat_argb1555;
-		case GL_UNSIGNED_SHORT_5_6_5:
-		case GL_UNSIGNED_SHORT_5_6_5_REV:
-			return _dri_texformat_rgb565;
-		default:
-			return do32bpt ? _dri_texformat_argb8888 :
-			    _dri_texformat_rgb565;
-		}
-
-	case GL_RGBA8:
-	case GL_RGB10_A2:
-	case GL_RGBA12:
-	case GL_RGBA16:
-		return !force16bpt ?
-		    r300Choose8888TexFormat(format,
-					    type) : _dri_texformat_argb4444;
-
-	case GL_RGBA4:
-	case GL_RGBA2:
-		return _dri_texformat_argb4444;
-
-	case GL_RGB5_A1:
-		return _dri_texformat_argb1555;
-
-	case GL_RGB8:
-	case GL_RGB10:
-	case GL_RGB12:
-	case GL_RGB16:
-		return !force16bpt ? _dri_texformat_argb8888 :
-		    _dri_texformat_rgb565;
-
-	case GL_RGB5:
-	case GL_RGB4:
-	case GL_R3_G3_B2:
-		return _dri_texformat_rgb565;
-
-	case GL_ALPHA:
-	case GL_ALPHA4:
-	case GL_ALPHA8:
-	case GL_ALPHA12:
-	case GL_ALPHA16:
-	case GL_COMPRESSED_ALPHA:
-		return _dri_texformat_a8;
-
-	case 1:
-	case GL_LUMINANCE:
-	case GL_LUMINANCE4:
-	case GL_LUMINANCE8:
-	case GL_LUMINANCE12:
-	case GL_LUMINANCE16:
-	case GL_COMPRESSED_LUMINANCE:
-		return _dri_texformat_l8;
-
-	case 2:
-	case GL_LUMINANCE_ALPHA:
-	case GL_LUMINANCE4_ALPHA4:
-	case GL_LUMINANCE6_ALPHA2:
-	case GL_LUMINANCE8_ALPHA8:
-	case GL_LUMINANCE12_ALPHA4:
-	case GL_LUMINANCE12_ALPHA12:
-	case GL_LUMINANCE16_ALPHA16:
-	case GL_COMPRESSED_LUMINANCE_ALPHA:
-		return _dri_texformat_al88;
-
-	case GL_INTENSITY:
-	case GL_INTENSITY4:
-	case GL_INTENSITY8:
-	case GL_INTENSITY12:
-	case GL_INTENSITY16:
-	case GL_COMPRESSED_INTENSITY:
-		return _dri_texformat_i8;
-
-	case GL_YCBCR_MESA:
-		if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-		    type == GL_UNSIGNED_BYTE)
-			return &_mesa_texformat_ycbcr;
-		else
-			return &_mesa_texformat_ycbcr_rev;
-
-	case GL_RGB_S3TC:
-	case GL_RGB4_S3TC:
-	case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-		return &_mesa_texformat_rgb_dxt1;
-
-	case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-		return &_mesa_texformat_rgba_dxt1;
-
-	case GL_RGBA_S3TC:
-	case GL_RGBA4_S3TC:
-	case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-		return &_mesa_texformat_rgba_dxt3;
-
-	case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-		return &_mesa_texformat_rgba_dxt5;
-
-	case GL_ALPHA16F_ARB:
-		return &_mesa_texformat_alpha_float16;
-	case GL_ALPHA32F_ARB:
-		return &_mesa_texformat_alpha_float32;
-	case GL_LUMINANCE16F_ARB:
-		return &_mesa_texformat_luminance_float16;
-	case GL_LUMINANCE32F_ARB:
-		return &_mesa_texformat_luminance_float32;
-	case GL_LUMINANCE_ALPHA16F_ARB:
-		return &_mesa_texformat_luminance_alpha_float16;
-	case GL_LUMINANCE_ALPHA32F_ARB:
-		return &_mesa_texformat_luminance_alpha_float32;
-	case GL_INTENSITY16F_ARB:
-		return &_mesa_texformat_intensity_float16;
-	case GL_INTENSITY32F_ARB:
-		return &_mesa_texformat_intensity_float32;
-	case GL_RGB16F_ARB:
-		return &_mesa_texformat_rgba_float16;
-	case GL_RGB32F_ARB:
-		return &_mesa_texformat_rgba_float32;
-	case GL_RGBA16F_ARB:
-		return &_mesa_texformat_rgba_float16;
-	case GL_RGBA32F_ARB:
-		return &_mesa_texformat_rgba_float32;
-
-	case GL_DEPTH_COMPONENT:
-	case GL_DEPTH_COMPONENT16:
-	case GL_DEPTH_COMPONENT24:
-	case GL_DEPTH_COMPONENT32:
-#if 0
-		switch (type) {
-		case GL_UNSIGNED_BYTE:
-		case GL_UNSIGNED_SHORT:
-			return &_mesa_texformat_z16;
-		case GL_UNSIGNED_INT:
-			return &_mesa_texformat_z32;
-		case GL_UNSIGNED_INT_24_8_EXT:
-		default:
-			return &_mesa_texformat_z24_s8;
-		}
-#else
-		return &_mesa_texformat_z16;
-#endif
-
-	default:
-		_mesa_problem(ctx,
-			      "unexpected internalFormat 0x%x in r300ChooseTextureFormat",
-			      (int)internalFormat);
-		return NULL;
-	}
-
-	return NULL;		/* never get here */
-}
-
-static GLboolean
-r300ValidateClientStorage(GLcontext * ctx, GLenum target,
-			  GLint internalFormat,
-			  GLint srcWidth, GLint srcHeight,
-			  GLenum format, GLenum type, const void *pixels,
-			  const struct gl_pixelstore_attrib *packing,
-			  struct gl_texture_object *texObj,
-			  struct gl_texture_image *texImage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "intformat %s format %s type %s\n",
-			_mesa_lookup_enum_by_nr(internalFormat),
-			_mesa_lookup_enum_by_nr(format),
-			_mesa_lookup_enum_by_nr(type));
-
-	if (!ctx->Unpack.ClientStorage)
-		return 0;
-
-	if (ctx->_ImageTransferState ||
-	    texImage->IsCompressed || texObj->GenerateMipmap)
-		return 0;
-
-	/* This list is incomplete, may be different on ppc???
-	 */
-	switch (internalFormat) {
-	case GL_RGBA:
-		if (format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV) {
-			texImage->TexFormat = _dri_texformat_argb8888;
-		} else
-			return 0;
-		break;
-
-	case GL_RGB:
-		if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) {
-			texImage->TexFormat = _dri_texformat_rgb565;
-		} else
-			return 0;
-		break;
-
-	case GL_YCBCR_MESA:
-		if (format == GL_YCBCR_MESA &&
-		    type == GL_UNSIGNED_SHORT_8_8_REV_APPLE) {
-			texImage->TexFormat = &_mesa_texformat_ycbcr_rev;
-		} else if (format == GL_YCBCR_MESA &&
-			   (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-			    type == GL_UNSIGNED_BYTE)) {
-			texImage->TexFormat = &_mesa_texformat_ycbcr;
-		} else
-			return 0;
-		break;
-
-	default:
-		return 0;
-	}
-
-	/* Could deal with these packing issues, but currently don't:
-	 */
-	if (packing->SkipPixels ||
-	    packing->SkipRows || packing->SwapBytes || packing->LsbFirst) {
-		return 0;
-	}
-
-	GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth,
-						    format, type);
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "%s: srcRowStride %d/%x\n",
-			__FUNCTION__, srcRowStride, srcRowStride);
-
-	/* Could check this later in upload, pitch restrictions could be
-	 * relaxed, but would need to store the image pitch somewhere,
-	 * as packing details might change before image is uploaded:
-	 */
-	if (!r300IsGartMemory(rmesa, pixels, srcHeight * srcRowStride)
-	    || (srcRowStride & 63))
-		return 0;
-
-	/* Have validated that _mesa_transfer_teximage would be a straight
-	 * memcpy at this point.  NOTE: future calls to TexSubImage will
-	 * overwrite the client data.  This is explicitly mentioned in the
-	 * extension spec.
-	 */
-	texImage->Data = (void *)pixels;
-	texImage->IsClientData = GL_TRUE;
-	texImage->RowStride = srcRowStride / texImage->TexFormat->TexelBytes;
-
-	return 1;
-}
-
-static void r300TexImage1D(GLcontext * ctx, GLenum target, GLint level,
-			   GLint internalFormat,
-			   GLint width, GLint border,
-			   GLenum format, GLenum type, const GLvoid * pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
-			return;
-		}
-	}
-
-	/* Note, this will call ChooseTextureFormat */
-	_mesa_store_teximage1d(ctx, target, level, internalFormat,
-			       width, border, format, type, pixels,
-			       &ctx->Unpack, texObj, texImage);
-
-	t->dirty_images[0] |= (1 << level);
-}
-
-static void r300TexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
-			      GLint xoffset,
-			      GLsizei width,
-			      GLenum format, GLenum type,
-			      const GLvoid * pixels,
-			      const struct gl_pixelstore_attrib *packing,
-			      struct gl_texture_object *texObj,
-			      struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
-			return;
-		}
-	}
-
-	_mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
-				  format, type, pixels, packing, texObj,
-				  texImage);
-
-	t->dirty_images[0] |= (1 << level);
-}
-
-static void r300TexImage2D(GLcontext * ctx, GLenum target, GLint level,
-			   GLint internalFormat,
-			   GLint width, GLint height, GLint border,
-			   GLenum format, GLenum type, const GLvoid * pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	if (t != NULL) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
-			return;
-		}
-	}
-
-	texImage->IsClientData = GL_FALSE;
-
-	if (r300ValidateClientStorage(ctx, target,
-				      internalFormat,
-				      width, height,
-				      format, type, pixels,
-				      packing, texObj, texImage)) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using client storage\n",
-				__FUNCTION__);
-	} else {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using normal storage\n",
-				__FUNCTION__);
-
-		/* Normal path: copy (to cached memory) and eventually upload
-		 * via another copy to GART memory and then a blit...  Could
-		 * eliminate one copy by going straight to (permanent) GART.
-		 *
-		 * Note, this will call r300ChooseTextureFormat.
-		 */
-		_mesa_store_teximage2d(ctx, target, level, internalFormat,
-				       width, height, border, format, type,
-				       pixels, &ctx->Unpack, texObj, texImage);
-
-		t->dirty_images[face] |= (1 << level);
-	}
-}
-
-static void r300TexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
-			      GLint xoffset, GLint yoffset,
-			      GLsizei width, GLsizei height,
-			      GLenum format, GLenum type,
-			      const GLvoid * pixels,
-			      const struct gl_pixelstore_attrib *packing,
-			      struct gl_texture_object *texObj,
-			      struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
-			return;
-		}
-	}
-
-	_mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-				  height, format, type, pixels, packing, texObj,
-				  texImage);
-
-	t->dirty_images[face] |= (1 << level);
-}
-
-static void r300CompressedTexImage2D(GLcontext * ctx, GLenum target,
-				     GLint level, GLint internalFormat,
-				     GLint width, GLint height, GLint border,
-				     GLsizei imageSize, const GLvoid * data,
-				     struct gl_texture_object *texObj,
-				     struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	if (t != NULL) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY,
-				    "glCompressedTexImage2D");
-			return;
-		}
-	}
-
-	texImage->IsClientData = GL_FALSE;
-
-	/* can't call this, different parameters. Would never evaluate to true anyway currently */
-#if 0
-	if (r300ValidateClientStorage(ctx, target,
-				      internalFormat,
-				      width, height,
-				      format, type, pixels,
-				      packing, texObj, texImage)) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using client storage\n",
-				__FUNCTION__);
-	} else
-#endif
-	{
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using normal storage\n",
-				__FUNCTION__);
-
-		/* Normal path: copy (to cached memory) and eventually upload
-		 * via another copy to GART memory and then a blit...  Could
-		 * eliminate one copy by going straight to (permanent) GART.
-		 *
-		 * Note, this will call r300ChooseTextureFormat.
-		 */
-		_mesa_store_compressed_teximage2d(ctx, target, level,
-						  internalFormat, width, height,
-						  border, imageSize, data,
-						  texObj, texImage);
-
-		t->dirty_images[face] |= (1 << level);
-	}
-}
-
-static void r300CompressedTexSubImage2D(GLcontext * ctx, GLenum target,
-					GLint level, GLint xoffset,
-					GLint yoffset, GLsizei width,
-					GLsizei height, GLenum format,
-					GLsizei imageSize, const GLvoid * data,
-					struct gl_texture_object *texObj,
-					struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-	GLuint face;
-
-	/* which cube face or ordinary 2D image */
-	switch (target) {
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-		face =
-		    (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-		ASSERT(face < 6);
-		break;
-	default:
-		face = 0;
-	}
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY,
-				    "glCompressedTexSubImage3D");
-			return;
-		}
-	}
-
-	_mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset,
-					     yoffset, width, height, format,
-					     imageSize, data, texObj, texImage);
-
-	t->dirty_images[face] |= (1 << level);
-}
-
-static void r300TexImage3D(GLcontext * ctx, GLenum target, GLint level,
-			   GLint internalFormat,
-			   GLint width, GLint height, GLint depth,
-			   GLint border,
-			   GLenum format, GLenum type, const GLvoid * pixels,
-			   const struct gl_pixelstore_attrib *packing,
-			   struct gl_texture_object *texObj,
-			   struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage3D");
-			return;
-		}
-	}
-
-	texImage->IsClientData = GL_FALSE;
-
-#if 0
-	if (r300ValidateClientStorage(ctx, target,
-				      internalFormat,
-				      width, height,
-				      format, type, pixels,
-				      packing, texObj, texImage)) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using client storage\n",
-				__FUNCTION__);
-	} else
-#endif
-	{
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: Using normal storage\n",
-				__FUNCTION__);
-
-		/* Normal path: copy (to cached memory) and eventually upload
-		 * via another copy to GART memory and then a blit...  Could
-		 * eliminate one copy by going straight to (permanent) GART.
-		 *
-		 * Note, this will call r300ChooseTextureFormat.
-		 */
-		_mesa_store_teximage3d(ctx, target, level, internalFormat,
-				       width, height, depth, border,
-				       format, type, pixels,
-				       &ctx->Unpack, texObj, texImage);
-
-		t->dirty_images[0] |= (1 << level);
-	}
-}
-
-static void
-r300TexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
-		  GLint xoffset, GLint yoffset, GLint zoffset,
-		  GLsizei width, GLsizei height, GLsizei depth,
-		  GLenum format, GLenum type,
-		  const GLvoid * pixels,
-		  const struct gl_pixelstore_attrib *packing,
-		  struct gl_texture_object *texObj,
-		  struct gl_texture_image *texImage)
-{
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
-
-/*     fprintf(stderr, "%s\n", __FUNCTION__); */
-
-	assert(t);		/* this _should_ be true */
-	if (t) {
-		driSwapOutTextureObject(t);
-	} else {
-		t = (driTextureObject *) r300AllocTexObj(texObj);
-		if (!t) {
-			_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage3D");
-			return;
-		}
-		texObj->DriverData = t;
-	}
-
-	_mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset,
-				  width, height, depth,
-				  format, type, pixels, packing, texObj,
-				  texImage);
-
-	t->dirty_images[0] |= (1 << level);
-}
-
-/**
  * Changes variables and flags for a state update, which will happen at the
  * next UpdateTextureState
  */
@@ -913,7 +195,7 @@ static void r300TexParameter(GLcontext * ctx, GLenum target,
 			     struct gl_texture_object *texObj,
 			     GLenum pname, const GLfloat * params)
 {
-	r300TexObjPtr t = (r300TexObjPtr) texObj->DriverData;
+	radeonTexObj* t = radeon_tex_obj(texObj);
 
 	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
 		fprintf(stderr, "%s( %s )\n", __FUNCTION__,
@@ -946,7 +228,11 @@ static void r300TexParameter(GLcontext * ctx, GLenum target,
 		 * we just have to rely on loading the right subset of mipmap levels
 		 * to simulate a clamped LOD.
 		 */
-		driSwapOutTextureObject((driTextureObject *) t);
+		if (t->mt) {
+			radeon_miptree_unreference(t->mt);
+			t->mt = 0;
+			t->validated = GL_FALSE;
+		}
 		break;
 
 	case GL_DEPTH_TEXTURE_MODE:
@@ -969,27 +255,10 @@ static void r300TexParameter(GLcontext * ctx, GLenum target,
 	}
 }
 
-static void r300BindTexture(GLcontext * ctx, GLenum target,
-			    struct gl_texture_object *texObj)
-{
-	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
-		fprintf(stderr, "%s( %p ) unit=%d\n", __FUNCTION__,
-			(void *)texObj, ctx->Texture.CurrentUnit);
-	}
-
-	if ((target == GL_TEXTURE_1D)
-	    || (target == GL_TEXTURE_2D)
-	    || (target == GL_TEXTURE_3D)
-	    || (target == GL_TEXTURE_CUBE_MAP)
-	    || (target == GL_TEXTURE_RECTANGLE_NV)) {
-		assert(texObj->DriverData != NULL);
-	}
-}
-
 static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	driTextureObject *t = (driTextureObject *) texObj->DriverData;
+	radeonTexObj* t = radeon_tex_obj(texObj);
 
 	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
 		fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
@@ -997,14 +266,24 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
 			_mesa_lookup_enum_by_nr(texObj->Target));
 	}
 
-	if (t != NULL) {
-		if (rmesa) {
-			R300_FIREVERTICES(rmesa);
-		}
+	if (rmesa) {
+		int i;
+		radeon_firevertices(&rmesa->radeon);
+
+		for(i = 0; i < R300_MAX_TEXTURE_UNITS; ++i)
+			if (rmesa->hw.textures[i] == t)
+				rmesa->hw.textures[i] = 0;
+	}
 
-		driDestroyTextureObject(t);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = 0;
 	}
-	/* Free mipmap images and the texture object itself */
 	_mesa_delete_texture_object(ctx, texObj);
 }
 
@@ -1013,8 +292,6 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
  * Called via ctx->Driver.NewTextureObject.
  * Note: this function will be called during context creation to
  * allocate the default texture objects.
- * Note: we could use containment here to 'derive' the driver-specific
- * texture object from the core mesa gl_texture_object.  Not done at this time.
  * Fixup MaxAnisotropy according to user preference.
  */
 static struct gl_texture_object *r300NewTextureObject(GLcontext * ctx,
@@ -1022,14 +299,23 @@ static struct gl_texture_object *r300NewTextureObject(GLcontext * ctx,
 						      GLenum target)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_object *obj;
-	obj = _mesa_new_texture_object(ctx, name, target);
-	if (!obj)
-		return NULL;
-	obj->MaxAnisotropy = rmesa->initialMaxAnisotropy;
+	radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
 
-	r300AllocTexObj(obj);
-	return obj;
+	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+		fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+			t, _mesa_lookup_enum_by_nr(target));
+	}
+
+	_mesa_initialize_texture_object(&t->base, name, target);
+	t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+	/* Initialize hardware state */
+	r300UpdateTexWrap(t);
+	r300SetTexFilter(t, t->base.MinFilter, t->base.MagFilter, t->base.MaxAnisotropy);
+	r300SetTexBorderColor(t, t->base.BorderColor);
+
+	return &t->base;
 }
 
 void r300InitTextureFuncs(struct dd_function_table *functions)
@@ -1037,22 +323,30 @@ void r300InitTextureFuncs(struct dd_function_table *functions)
 	/* Note: we only plug in the functions we implement in the driver
 	 * since _mesa_init_driver_functions() was already called.
 	 */
-	functions->ChooseTextureFormat = r300ChooseTextureFormat;
-	functions->TexImage1D = r300TexImage1D;
-	functions->TexImage2D = r300TexImage2D;
-	functions->TexImage3D = r300TexImage3D;
-	functions->TexSubImage1D = r300TexSubImage1D;
-	functions->TexSubImage2D = r300TexSubImage2D;
-	functions->TexSubImage3D = r300TexSubImage3D;
+	functions->NewTextureImage = radeonNewTextureImage;
+	functions->FreeTexImageData = radeonFreeTexImageData;
+	functions->MapTexture = radeonMapTexture;
+	functions->UnmapTexture = radeonUnmapTexture;
+
+	functions->ChooseTextureFormat = radeonChooseTextureFormat_mesa;
+	functions->TexImage1D = radeonTexImage1D;
+	functions->TexImage2D = radeonTexImage2D;
+	functions->TexImage3D = radeonTexImage3D;
+	functions->TexSubImage1D = radeonTexSubImage1D;
+	functions->TexSubImage2D = radeonTexSubImage2D;
+	functions->TexSubImage3D = radeonTexSubImage3D;
+	functions->GetTexImage = radeonGetTexImage;
+	functions->GetCompressedTexImage = radeonGetCompressedTexImage;
 	functions->NewTextureObject = r300NewTextureObject;
-	functions->BindTexture = r300BindTexture;
 	functions->DeleteTexture = r300DeleteTexture;
 	functions->IsTextureResident = driIsTextureResident;
 
 	functions->TexParameter = r300TexParameter;
 
-	functions->CompressedTexImage2D = r300CompressedTexImage2D;
-	functions->CompressedTexSubImage2D = r300CompressedTexSubImage2D;
+	functions->CompressedTexImage2D = radeonCompressedTexImage2D;
+	functions->CompressedTexSubImage2D = radeonCompressedTexSubImage2D;
+
+	functions->GenerateMipmap = radeonGenerateMipmap;
 
 	driInitTextureFormats();
 }
diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h
index b86d45bfe0..8a653ea2d1 100644
--- a/src/mesa/drivers/dri/r300/r300_tex.h
+++ b/src/mesa/drivers/dri/r300/r300_tex.h
@@ -37,16 +37,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 extern void r300SetDepthTexMode(struct gl_texture_object *tObj);
 
+extern void r300SetTexBuffer(__DRIcontext *pDRICtx, GLint target,
+			     __DRIdrawable *dPriv);
+
+extern void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
+			      GLint format, __DRIdrawable *dPriv);
+
 extern void r300SetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 			     unsigned long long offset, GLint depth,
 			     GLuint pitch);
 
-extern void r300UpdateTextureState(GLcontext * ctx);
-
-extern int r300UploadTexImages(r300ContextPtr rmesa, r300TexObjPtr t,
-			       GLuint face);
-
-extern void r300DestroyTexObj(r300ContextPtr rmesa, r300TexObjPtr t);
+extern GLboolean r300ValidateBuffers(GLcontext * ctx);
 
 extern void r300InitTextureFuncs(struct dd_function_table *functions);
 
diff --git a/src/mesa/drivers/dri/r300/r300_texmem.c b/src/mesa/drivers/dri/r300/r300_texmem.c
deleted file mode 100644
index a89ab83d94..0000000000
--- a/src/mesa/drivers/dri/r300/r300_texmem.c
+++ /dev/null
@@ -1,568 +0,0 @@
-/**************************************************************************
-
-Copyright (C) Tungsten Graphics 2002.  All Rights Reserved.
-The Weather Channel, Inc. funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86
-license. This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation on the rights to use, copy, modify, merge, publish,
-distribute, sub license, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
-SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
-**************************************************************************/
-
-/**
- * \file
- *
- * \author Gareth Hughes <gareth@valinux.com>
- *
- * \author Kevin E. Martin <martin@valinux.com>
- */
-
-#include <errno.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/colormac.h"
-#include "main/macros.h"
-#include "main/simple_list.h"
-#include "main/texobj.h"
-#include "radeon_reg.h"		/* gets definition for usleep */
-#include "r300_context.h"
-#include "r300_state.h"
-#include "r300_cmdbuf.h"
-#include "radeon_ioctl.h"
-#include "r300_tex.h"
-#include "r300_ioctl.h"
-#include <unistd.h>		/* for usleep() */
-
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
-
-/**
- * Destroy any device-dependent state associated with the texture.  This may
- * include NULLing out hardware state that points to the texture.
- */
-void r300DestroyTexObj(r300ContextPtr rmesa, r300TexObjPtr t)
-{
-	int i;
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE) {
-		fprintf(stderr, "%s( %p, %p )\n", __FUNCTION__,
-			(void *)t, (void *)t->base.tObj);
-	}
-
-	for (i = 0; i < rmesa->radeon.glCtx->Const.MaxTextureUnits; i++) {
-		if (rmesa->state.texture.unit[i].texobj == t->base.tObj) {
-			_mesa_reference_texobj(&rmesa->state.texture.unit[i].texobj, NULL);
-		}
-	}
-}
-
-/* ------------------------------------------------------------
- * Texture image conversions
- */
-
-static void r300UploadGARTClientSubImage(r300ContextPtr rmesa,
-					 r300TexObjPtr t,
-					 struct gl_texture_image *texImage,
-					 GLint hwlevel,
-					 GLint x, GLint y,
-					 GLint width, GLint height)
-{
-	const struct gl_texture_format *texFormat = texImage->TexFormat;
-	GLuint srcPitch, dstPitch;
-	int blit_format;
-	int srcOffset;
-
-	/*
-	 * XXX it appears that we always upload the full image, not a subimage.
-	 * I.e. x==0, y==0, width=texWidth, height=texWidth.  If this is ever
-	 * changed, the src pitch will have to change.
-	 */
-	switch (texFormat->TexelBytes) {
-	case 1:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	case 2:
-		blit_format = R300_CP_COLOR_FORMAT_RGB565;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	case 4:
-		blit_format = R300_CP_COLOR_FORMAT_ARGB8888;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	case 8:
-	case 16:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
-		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
-		break;
-	default:
-		return;
-	}
-
-	t->image[0][hwlevel].data = texImage->Data;
-	srcOffset = r300GartOffsetFromVirtual(rmesa, texImage->Data);
-
-	assert(srcOffset != ~0);
-
-	/* Don't currently need to cope with small pitches?
-	 */
-	width = texImage->Width;
-	height = texImage->Height;
-
-	if (texFormat->TexelBytes > 4) {
-		width *= texFormat->TexelBytes;
-	}
-
-	r300EmitWait(rmesa, R300_WAIT_3D);
-
-	r300EmitBlit(rmesa, blit_format,
-		     srcPitch,
-		     srcOffset,
-		     dstPitch,
-		     t->bufAddr,
-		     x,
-		     y,
-		     t->image[0][hwlevel].x + x,
-		     t->image[0][hwlevel].y + y, width, height);
-
-	r300EmitWait(rmesa, R300_WAIT_2D);
-}
-
-static void r300UploadRectSubImage(r300ContextPtr rmesa,
-				   r300TexObjPtr t,
-				   struct gl_texture_image *texImage,
-				   GLint x, GLint y, GLint width, GLint height)
-{
-	const struct gl_texture_format *texFormat = texImage->TexFormat;
-	int blit_format, dstPitch, done;
-
-	switch (texFormat->TexelBytes) {
-	case 1:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		break;
-	case 2:
-		blit_format = R300_CP_COLOR_FORMAT_RGB565;
-		break;
-	case 4:
-		blit_format = R300_CP_COLOR_FORMAT_ARGB8888;
-		break;
-	case 8:
-	case 16:
-		blit_format = R300_CP_COLOR_FORMAT_CI8;
-		break;
-	default:
-		return;
-	}
-
-	t->image[0][0].data = texImage->Data;
-
-	/* Currently don't need to cope with small pitches.
-	 */
-	width = texImage->Width;
-	height = texImage->Height;
-	dstPitch = t->pitch;
-
-	if (texFormat->TexelBytes > 4) {
-		width *= texFormat->TexelBytes;
-	}
-
-	if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) {
-		/* In this case, could also use GART texturing.  This is
-		 * currently disabled, but has been tested & works.
-		 */
-		t->offset = r300GartOffsetFromVirtual(rmesa, texImage->Data);
-		t->pitch = texImage->RowStride * texFormat->TexelBytes - 32;
-
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr,
-				"Using GART texturing for rectangular client texture\n");
-
-		/* Release FB memory allocated for this image:
-		 */
-		/* FIXME This may not be correct as driSwapOutTextureObject sets
-		 * FIXME dirty_images.  It may be fine, though.
-		 */
-		if (t->base.memBlock) {
-			driSwapOutTextureObject((driTextureObject *) t);
-		}
-	} else if (texImage->IsClientData) {
-		/* Data already in GART memory, with usable pitch.
-		 */
-		GLuint srcPitch;
-		srcPitch = texImage->RowStride * texFormat->TexelBytes;
-		r300EmitBlit(rmesa,
-			     blit_format,
-			     srcPitch,
-			     r300GartOffsetFromVirtual(rmesa, texImage->Data),
-			     dstPitch, t->bufAddr, 0, 0, 0, 0, width, height);
-	} else {
-		/* Data not in GART memory, or bad pitch.
-		 */
-		for (done = 0; done < height;) {
-			struct r300_dma_region region;
-			int lines =
-			    MIN2(height - done, RADEON_BUFFER_SIZE / dstPitch);
-			int src_pitch;
-			char *tex;
-
-			src_pitch = texImage->RowStride * texFormat->TexelBytes;
-
-			tex = (char *)texImage->Data + done * src_pitch;
-
-			memset(&region, 0, sizeof(region));
-			r300AllocDmaRegion(rmesa, &region, lines * dstPitch,
-					   1024);
-
-			/* Copy texdata to dma:
-			 */
-			if (RADEON_DEBUG & DEBUG_TEXTURE)
-				fprintf(stderr,
-					"%s: src_pitch %d dst_pitch %d\n",
-					__FUNCTION__, src_pitch, dstPitch);
-
-			if (src_pitch == dstPitch) {
-				memcpy(region.address + region.start, tex,
-				       lines * src_pitch);
-			} else {
-				char *buf = region.address + region.start;
-				int i;
-				for (i = 0; i < lines; i++) {
-					memcpy(buf, tex, src_pitch);
-					buf += dstPitch;
-					tex += src_pitch;
-				}
-			}
-
-			r300EmitWait(rmesa, R300_WAIT_3D);
-
-			/* Blit to framebuffer
-			 */
-			r300EmitBlit(rmesa,
-				     blit_format,
-				     dstPitch, GET_START(&region),
-				     dstPitch | (t->tile_bits >> 16),
-				     t->bufAddr, 0, 0, 0, done, width, lines);
-
-			r300EmitWait(rmesa, R300_WAIT_2D);
-#ifdef USER_BUFFERS
-			r300_mem_use(rmesa, region.buf->id);
-#endif
-
-			r300ReleaseDmaRegion(rmesa, &region, __FUNCTION__);
-			done += lines;
-		}
-	}
-}
-
-/**
- * Upload the texture image associated with texture \a t at the specified
- * level at the address relative to \a start.
- */
-static void r300UploadSubImage(r300ContextPtr rmesa, r300TexObjPtr t,
-			       GLint hwlevel,
-			       GLint x, GLint y, GLint width, GLint height,
-			       GLuint face)
-{
-	struct gl_texture_image *texImage = NULL;
-	GLuint offset;
-	GLint imageWidth, imageHeight;
-	GLint ret;
-	drm_radeon_texture_t tex;
-	drm_radeon_tex_image_t tmp;
-	const int level = hwlevel + t->base.firstLevel;
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE) {
-		fprintf(stderr,
-			"%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n",
-			__FUNCTION__, (void *)t, (void *)t->base.tObj, level,
-			width, height, face);
-	}
-
-	ASSERT(face < 6);
-
-	/* Ensure we have a valid texture to upload */
-	if ((hwlevel < 0) || (hwlevel >= R300_MAX_TEXTURE_LEVELS)) {
-		_mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
-		return;
-	}
-
-	texImage = t->base.tObj->Image[face][level];
-
-	if (!texImage) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: texImage %d is NULL!\n",
-				__FUNCTION__, level);
-		return;
-	}
-	if (!texImage->Data) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: image data is NULL!\n",
-				__FUNCTION__);
-		return;
-	}
-
-	if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-		assert(level == 0);
-		assert(hwlevel == 0);
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr, "%s: image data is rectangular\n",
-				__FUNCTION__);
-		r300UploadRectSubImage(rmesa, t, texImage, x, y, width, height);
-		return;
-	} else if (texImage->IsClientData) {
-		if (RADEON_DEBUG & DEBUG_TEXTURE)
-			fprintf(stderr,
-				"%s: image data is in GART client storage\n",
-				__FUNCTION__);
-		r300UploadGARTClientSubImage(rmesa, t, texImage, hwlevel, x, y,
-					     width, height);
-		return;
-	} else if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "%s: image data is in normal memory\n",
-			__FUNCTION__);
-
-	imageWidth = texImage->Width;
-	imageHeight = texImage->Height;
-
-	offset = t->bufAddr;
-
-	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
-		GLint imageX = 0;
-		GLint imageY = 0;
-		GLint blitX = t->image[face][hwlevel].x;
-		GLint blitY = t->image[face][hwlevel].y;
-		GLint blitWidth = t->image[face][hwlevel].width;
-		GLint blitHeight = t->image[face][hwlevel].height;
-		fprintf(stderr, "   upload image: %d,%d at %d,%d\n",
-			imageWidth, imageHeight, imageX, imageY);
-		fprintf(stderr, "   upload  blit: %d,%d at %d,%d\n",
-			blitWidth, blitHeight, blitX, blitY);
-		fprintf(stderr, "       blit ofs: 0x%07x level: %d/%d\n",
-			(GLuint) offset, hwlevel, level);
-	}
-
-	t->image[face][hwlevel].data = texImage->Data;
-
-	/* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
-	 * NOTE: we're always use a 1KB-wide blit and I8 texture format.
-	 * We used to use 1, 2 and 4-byte texels and used to use the texture
-	 * width to dictate the blit width - but that won't work for compressed
-	 * textures. (Brian)
-	 * NOTE: can't do that with texture tiling. (sroland)
-	 */
-	tex.offset = offset;
-	tex.image = &tmp;
-	/* copy (x,y,width,height,data) */
-	memcpy(&tmp, &t->image[face][hwlevel], sizeof(tmp));
-
-	if (texImage->TexFormat->TexelBytes > 4) {
-		const int log2TexelBytes =
-		    (3 + (texImage->TexFormat->TexelBytes >> 4));
-		tex.format = RADEON_TXFORMAT_I8;	/* any 1-byte texel format */
-		tex.pitch =
-		    MAX2((texImage->Width * texImage->TexFormat->TexelBytes) /
-			 64, 1);
-		tex.height = imageHeight;
-		tex.width = imageWidth << log2TexelBytes;
-		tex.offset += (tmp.x << log2TexelBytes) & ~1023;
-		tmp.x = tmp.x % (1024 >> log2TexelBytes);
-		tmp.width = tmp.width << log2TexelBytes;
-	} else if (texImage->TexFormat->TexelBytes) {
-		/* use multi-byte upload scheme */
-		tex.height = imageHeight;
-		tex.width = imageWidth;
-		switch (texImage->TexFormat->TexelBytes) {
-		case 1:
-			tex.format = RADEON_TXFORMAT_I8;
-			break;
-		case 2:
-			tex.format = RADEON_TXFORMAT_AI88;
-			break;
-		case 4:
-			tex.format = RADEON_TXFORMAT_ARGB8888;
-			break;
-		}
-		tex.pitch =
-		    MAX2((texImage->Width * texImage->TexFormat->TexelBytes) /
-			 64, 1);
-		tex.offset += tmp.x & ~1023;
-		tmp.x = tmp.x % 1024;
-
-		if (t->tile_bits & R300_TXO_MICRO_TILE) {
-			/* need something like "tiled coordinates" ? */
-			tmp.y = tmp.x / (tex.pitch * 128) * 2;
-			tmp.x =
-			    tmp.x % (tex.pitch * 128) / 2 /
-			    texImage->TexFormat->TexelBytes;
-			tex.pitch |= RADEON_DST_TILE_MICRO >> 22;
-		} else {
-			tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1);
-		}
-#if 1
-		if ((t->tile_bits & R300_TXO_MACRO_TILE) &&
-		    (texImage->Width * texImage->TexFormat->TexelBytes >= 256)
-		    && ((!(t->tile_bits & R300_TXO_MICRO_TILE)
-			 && (texImage->Height >= 8))
-			|| (texImage->Height >= 16))) {
-			/* weird: R200 disables macro tiling if mip width is smaller than 256 bytes,
-			   OR if height is smaller than 8 automatically, but if micro tiling is active
-			   the limit is height 16 instead ? */
-			tex.pitch |= RADEON_DST_TILE_MACRO >> 22;
-		}
-#endif
-	} else {
-		/* In case of for instance 8x8 texture (2x2 dxt blocks),
-		   padding after the first two blocks is needed (only
-		   with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */
-		/* set tex.height to 1/4 since 1 "macropixel" (dxt-block)
-		   has 4 real pixels. Needed so the kernel module reads
-		   the right amount of data. */
-		tex.format = RADEON_TXFORMAT_I8;	/* any 1-byte texel format */
-		tex.pitch = (R300_BLIT_WIDTH_BYTES / 64);
-		tex.height = (imageHeight + 3) / 4;
-		tex.width = (imageWidth + 3) / 4;
-		if ((t->format & R300_TX_FORMAT_DXT1) == R300_TX_FORMAT_DXT1) {
-			tex.width *= 8;
-		} else {
-			tex.width *= 16;
-		}
-	}
-
-	LOCK_HARDWARE(&rmesa->radeon);
-	do {
-		ret =
-		    drmCommandWriteRead(rmesa->radeon.dri.fd,
-					DRM_RADEON_TEXTURE, &tex,
-					sizeof(drm_radeon_texture_t));
-		if (ret) {
-			if (RADEON_DEBUG & DEBUG_IOCTL)
-				fprintf(stderr,
-					"DRM_RADEON_TEXTURE:  again!\n");
-			usleep(1);
-		}
-	} while (ret == -EAGAIN);
-
-	UNLOCK_HARDWARE(&rmesa->radeon);
-
-	if (ret) {
-		fprintf(stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret);
-		fprintf(stderr, "   offset=0x%08x\n", offset);
-		fprintf(stderr, "   image width=%d height=%d\n",
-			imageWidth, imageHeight);
-		fprintf(stderr, "    blit width=%d height=%d data=%p\n",
-			t->image[face][hwlevel].width,
-			t->image[face][hwlevel].height,
-			t->image[face][hwlevel].data);
-		_mesa_exit(-1);
-	}
-}
-
-/**
- * Upload the texture images associated with texture \a t.  This might
- * require the allocation of texture memory.
- *
- * \param rmesa Context pointer
- * \param t Texture to be uploaded
- * \param face Cube map face to be uploaded.  Zero for non-cube maps.
- */
-
-int r300UploadTexImages(r300ContextPtr rmesa, r300TexObjPtr t, GLuint face)
-{
-	const int numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-	if (t->image_override)
-		return 0;
-
-	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
-		fprintf(stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
-			(void *)rmesa->radeon.glCtx, (void *)t->base.tObj,
-			t->base.totalSize, t->base.firstLevel,
-			t->base.lastLevel);
-	}
-
-	if (t->base.totalSize == 0)
-		return 0;
-
-	if (RADEON_DEBUG & DEBUG_SYNC) {
-		fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-		radeonFinish(rmesa->radeon.glCtx);
-	}
-
-	LOCK_HARDWARE(&rmesa->radeon);
-
-	if (t->base.memBlock == NULL) {
-		int heap;
-
-		heap = driAllocateTexture(rmesa->texture_heaps, rmesa->nr_heaps,
-					  (driTextureObject *) t);
-		if (heap == -1) {
-			UNLOCK_HARDWARE(&rmesa->radeon);
-			return -1;
-		}
-
-		/* Set the base offset of the texture image */
-		t->bufAddr = rmesa->radeon.radeonScreen->texOffset[heap]
-		    + t->base.memBlock->ofs;
-		t->offset = t->bufAddr;
-
-		if (!(t->base.tObj->Image[0][0]->IsClientData)) {
-			/* hope it's safe to add that here... */
-			t->offset |= t->tile_bits;
-		}
-	}
-
-	/* Let the world know we've used this memory recently.
-	 */
-	driUpdateTextureLRU((driTextureObject *) t);
-	UNLOCK_HARDWARE(&rmesa->radeon);
-
-	/* Upload any images that are new */
-	if (t->base.dirty_images[face]) {
-		int i;
-		for (i = 0; i < numLevels; i++) {
-			if ((t->base.
-			     dirty_images[face] & (1 <<
-						   (i + t->base.firstLevel))) !=
-			    0) {
-				r300UploadSubImage(rmesa, t, i, 0, 0,
-						   t->image[face][i].width,
-						   t->image[face][i].height,
-						   face);
-			}
-		}
-		t->base.dirty_images[face] = 0;
-	}
-
-	if (RADEON_DEBUG & DEBUG_SYNC) {
-		fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-		radeonFinish(rmesa->radeon.glCtx);
-	}
-
-	return 0;
-}
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index f6ae4b675b..6f489ace7b 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -47,7 +47,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_context.h"
 #include "r300_state.h"
 #include "r300_ioctl.h"
-#include "radeon_ioctl.h"
+#include "radeon_mipmap_tree.h"
 #include "r300_tex.h"
 #include "r300_reg.h"
 
@@ -117,7 +117,12 @@ static const struct tx_table {
 	_ASSIGN(INTENSITY_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, X, FL_I16)),
 	_ASSIGN(Z16, R300_EASY_TX_FORMAT(X, X, X, X, X16)),
 	_ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)),
+	_ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)),
 	_ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)),
+	/* EXT_texture_sRGB */
+	_ASSIGN(SRGBA8, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA),
+	_ASSIGN(SLA8, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA),
+	_ASSIGN(SL8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA),
 	/* *INDENT-ON* */
 };
 
@@ -143,13 +148,12 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj)
 		},
 	};
 	const GLuint *format;
-	r300TexObjPtr t;
+	radeonTexObjPtr t;
 
 	if (!tObj)
 		return;
 
-	t = (r300TexObjPtr) tObj->DriverData;
-
+	t = radeon_tex_obj(tObj);
 
 	switch (tObj->Image[0][tObj->BaseLevel]->TexFormat->MesaFormat) {
 	case MESA_FORMAT_Z16:
@@ -171,13 +175,13 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj)
 
 	switch (tObj->DepthMode) {
 	case GL_LUMINANCE:
-		t->format = format[0];
+		t->pp_txformat = format[0];
 		break;
 	case GL_INTENSITY:
-		t->format = format[1];
+		t->pp_txformat = format[1];
 		break;
 	case GL_ALPHA:
-		t->format = format[2];
+		t->pp_txformat = format[2];
 		break;
 	default:
 		/* Error...which should have already been caught by higher
@@ -190,486 +194,307 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj)
 
 
 /**
- * Compute sizes and fill in offset and blit information for the given
- * image (determined by \p face and \p level).
- *
- * \param curOffset points to the offset at which the image is to be stored
- * and is updated by this function according to the size of the image.
- */
-static void compute_tex_image_offset(
-	struct gl_texture_object *tObj,
-	GLuint face,
-	GLint level,
-	GLint* curOffset)
-{
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-	const struct gl_texture_image* texImage;
-	GLuint blitWidth = R300_BLIT_WIDTH_BYTES;
-	GLuint texelBytes;
-	GLuint size;
-
-	texImage = tObj->Image[0][level + t->base.firstLevel];
-	if (!texImage)
-		return;
-
-	texelBytes = texImage->TexFormat->TexelBytes;
-
-	/* find image size in bytes */
-	if (texImage->IsCompressed) {
-		if ((t->format & R300_TX_FORMAT_DXT1) ==
-			R300_TX_FORMAT_DXT1) {
-			// fprintf(stderr,"DXT 1 %d %08X\n", texImage->Width, t->format);
-			if ((texImage->Width + 3) < 8)	/* width one block */
-				size = texImage->CompressedSize * 4;
-			else if ((texImage->Width + 3) < 16)
-				size = texImage->CompressedSize * 2;
-			else
-				size = texImage->CompressedSize;
-		} else {
-			/* DXT3/5, 16 bytes per block */
-			WARN_ONCE
-				("DXT 3/5 suffers from multitexturing problems!\n");
-			// fprintf(stderr,"DXT 3/5 %d\n", texImage->Width);
-			if ((texImage->Width + 3) < 8)
-				size = texImage->CompressedSize * 2;
-			else
-				size = texImage->CompressedSize;
-		}
-	} else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-		size =
-			((texImage->Width * texelBytes +
-			63) & ~63) * texImage->Height;
-		blitWidth = 64 / texelBytes;
-	} else if (t->tile_bits & R300_TXO_MICRO_TILE) {
-		/* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
-			though the actual offset may be different (if texture is less than
-			32 bytes width) to the untiled case */
-		int w = (texImage->Width * texelBytes * 2 + 31) & ~31;
-		size =
-			(w * ((texImage->Height + 1) / 2)) *
-			texImage->Depth;
-		blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-	} else {
-		int w = (texImage->Width * texelBytes + 31) & ~31;
-		size = w * texImage->Height * texImage->Depth;
-		blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-	}
-	assert(size > 0);
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr, "w=%d h=%d d=%d tb=%d intFormat=%d\n",
-			texImage->Width, texImage->Height,
-			texImage->Depth,
-			texImage->TexFormat->TexelBytes,
-			texImage->InternalFormat);
-
-	/* All images are aligned to a 32-byte offset */
-	*curOffset = (*curOffset + 0x1f) & ~0x1f;
-
-	if (texelBytes) {
-		/* fix x and y coords up later together with offset */
-		t->image[face][level].x = *curOffset;
-		t->image[face][level].y = 0;
-		t->image[face][level].width =
-			MIN2(size / texelBytes, blitWidth);
-		t->image[face][level].height =
-			(size / texelBytes) / t->image[face][level].width;
-	} else {
-		t->image[face][level].x = *curOffset % R300_BLIT_WIDTH_BYTES;
-		t->image[face][level].y = *curOffset / R300_BLIT_WIDTH_BYTES;
-		t->image[face][level].width =
-			MIN2(size, R300_BLIT_WIDTH_BYTES);
-		t->image[face][level].height = size / t->image[face][level].width;
-	}
-
-	if (RADEON_DEBUG & DEBUG_TEXTURE)
-		fprintf(stderr,
-			"level %d, face %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n",
-			level, face, texImage->Width, texImage->Height,
-			t->image[face][level].x, t->image[face][level].y,
-			t->image[face][level].width, t->image[face][level].height,
-			size, *curOffset);
-
-	*curOffset += size;
-}
-
-
-
-/**
- * This function computes the number of bytes of storage needed for
- * the given texture object (all mipmap levels, all cube faces).
- * The \c image[face][level].x/y/width/height parameters for upload/blitting
- * are computed here.  \c filter, \c format, etc. will be set here
- * too.
+ * Compute the cached hardware register values for the given texture object.
  *
  * \param rmesa Context pointer
- * \param tObj GL texture object whose images are to be posted to
- *                 hardware state.
+ * \param t the r300 texture object
  */
-static void r300SetTexImages(r300ContextPtr rmesa,
-			     struct gl_texture_object *tObj)
+static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t)
 {
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-	const struct gl_texture_image *baseImage =
-	    tObj->Image[0][tObj->BaseLevel];
-	GLint curOffset;
-	GLint i, texelBytes;
-	GLint numLevels;
-	GLint log2Width, log2Height, log2Depth;
-
-	/* Set the hardware texture format
-	 */
+	const struct gl_texture_image *firstImage;
+	int firstlevel = t->mt ? t->mt->firstLevel : 0;
+	    
+	firstImage = t->base.Image[0][firstlevel];
+
 	if (!t->image_override
-	    && VALID_FORMAT(baseImage->TexFormat->MesaFormat)) {
-		if (baseImage->TexFormat->BaseFormat == GL_DEPTH_COMPONENT) {
-			r300SetDepthTexMode(tObj);
+	    && VALID_FORMAT(firstImage->TexFormat->MesaFormat)) {
+		if (firstImage->TexFormat->BaseFormat == GL_DEPTH_COMPONENT) {
+			r300SetDepthTexMode(&t->base);
 		} else {
-			t->format = tx_table[baseImage->TexFormat->MesaFormat].format;
+			t->pp_txformat = tx_table[firstImage->TexFormat->MesaFormat].format;
 		}
 
-		t->filter |= tx_table[baseImage->TexFormat->MesaFormat].filter;
+		t->pp_txfilter |= tx_table[firstImage->TexFormat->MesaFormat].filter;
 	} else if (!t->image_override) {
 		_mesa_problem(NULL, "unexpected texture format in %s",
 			      __FUNCTION__);
 		return;
 	}
 
-	texelBytes = baseImage->TexFormat->TexelBytes;
-
-	/* Compute which mipmap levels we really want to send to the hardware.
-	 */
-	driCalculateTextureFirstLastLevel((driTextureObject *) t);
-	log2Width = tObj->Image[0][t->base.firstLevel]->WidthLog2;
-	log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2;
-	log2Depth = tObj->Image[0][t->base.firstLevel]->DepthLog2;
-
-	numLevels = t->base.lastLevel - t->base.firstLevel + 1;
+	if (t->image_override && t->bo)
+		return;
 
-	assert(numLevels <= R300_MAX_TEXTURE_LEVELS);
+	t->pp_txsize = (((firstImage->Width - 1) << R300_TX_WIDTHMASK_SHIFT)
+			| ((firstImage->Height - 1) << R300_TX_HEIGHTMASK_SHIFT)
+			| ((firstImage->DepthLog2) << R300_TX_DEPTHMASK_SHIFT)
+			| ((t->mt->lastLevel - t->mt->firstLevel) << R300_TX_MAX_MIP_LEVEL_SHIFT));
 
-	/* Calculate mipmap offsets and dimensions for blitting (uploading)
-	 * The idea is that we lay out the mipmap levels within a block of
-	 * memory organized as a rectangle of width BLIT_WIDTH_BYTES.
-	 */
 	t->tile_bits = 0;
 
-	/* figure out if this texture is suitable for tiling. */
-#if 0				/* Disabled for now */
-	if (texelBytes) {
-		if ((tObj->Target != GL_TEXTURE_RECTANGLE_NV) &&
-		    /* texrect might be able to use micro tiling too in theory? */
-		    (baseImage->Height > 1)) {
-
-			/* allow 32 (bytes) x 1 mip (which will use two times the space
-			   the non-tiled version would use) max if base texture is large enough */
-			if ((numLevels == 1) ||
-			    (((baseImage->Width * texelBytes /
-			       baseImage->Height) <= 32)
-			     && (baseImage->Width * texelBytes > 64))
-			    ||
-			    ((baseImage->Width * texelBytes /
-			      baseImage->Height) <= 16)) {
-				t->tile_bits |= R300_TXO_MICRO_TILE;
-			}
-		}
-
-		if (tObj->Target != GL_TEXTURE_RECTANGLE_NV) {
-			/* we can set macro tiling even for small textures, they will be untiled anyway */
-			t->tile_bits |= R300_TXO_MACRO_TILE;
-		}
-	}
-#endif
-
-	curOffset = 0;
+	if (t->base.Target == GL_TEXTURE_CUBE_MAP)
+		t->pp_txformat |= R300_TX_FORMAT_CUBIC_MAP;
+	if (t->base.Target == GL_TEXTURE_3D)
+		t->pp_txformat |= R300_TX_FORMAT_3D;
 
-	if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-		ASSERT(log2Width == log2Height);
-		t->format |= R300_TX_FORMAT_CUBIC_MAP;
 
-		for(i = 0; i < numLevels; i++) {
-			GLuint face;
-			for(face = 0; face < 6; face++)
-				compute_tex_image_offset(tObj, face, i, &curOffset);
-		}
-	} else {
-		if (tObj->Target == GL_TEXTURE_3D)
-                	t->format |= R300_TX_FORMAT_3D;
-
-		for (i = 0; i < numLevels; i++)
-			compute_tex_image_offset(tObj, 0, i, &curOffset);
-	}
-
-	/* Align the total size of texture memory block.
-	 */
-	t->base.totalSize =
-	    (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
-
-	t->size =
-	    (((tObj->Image[0][t->base.firstLevel]->Width -
-	       1) << R300_TX_WIDTHMASK_SHIFT)
-	     | ((tObj->Image[0][t->base.firstLevel]->Height - 1) <<
-		R300_TX_HEIGHTMASK_SHIFT)
-	     | ((tObj->Image[0][t->base.firstLevel]->DepthLog2) <<
-		R300_TX_DEPTHMASK_SHIFT))
-	    | ((numLevels - 1) << R300_TX_MAX_MIP_LEVEL_SHIFT);
-
-	t->pitch = 0;
-
-	/* Only need to round to nearest 32 for textures, but the blitter
-	 * requires 64-byte aligned pitches, and we may/may not need the
-	 * blitter.   NPOT only!
-	 */
-	if (baseImage->IsCompressed) {
-		t->pitch |=
-		    (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63);
-	} else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-		unsigned int align = (64 / texelBytes) - 1;
-		t->pitch |= ((tObj->Image[0][t->base.firstLevel]->Width *
-			     texelBytes) + 63) & ~(63);
-		t->size |= R300_TX_SIZE_TXPITCH_EN;
+	if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) {
+		unsigned int align = (64 / t->mt->bpp) - 1;
+		t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN;
 		if (!t->image_override)
-			t->pitch_reg =
-			    (((tObj->Image[0][t->base.firstLevel]->Width) +
-			      align) & ~align) - 1;
-	} else {
-		t->pitch |=
-		    ((tObj->Image[0][t->base.firstLevel]->Width *
-		      texelBytes) + 63) & ~(63);
+			t->pp_txpitch = ((firstImage->Width + align) & ~align) - 1;
 	}
 
 	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-	    if (tObj->Image[0][t->base.firstLevel]->Width > 2048)
-		t->pitch_reg |= R500_TXWIDTH_BIT11;
-	    if (tObj->Image[0][t->base.firstLevel]->Height > 2048)
-		t->pitch_reg |= R500_TXHEIGHT_BIT11;
+	    if (firstImage->Width > 2048)
+		t->pp_txpitch |= R500_TXWIDTH_BIT11;
+	    if (firstImage->Height > 2048)
+		t->pp_txpitch |= R500_TXHEIGHT_BIT11;
 	}
 }
 
-/* ================================================================
- * Texture unit state management
+/**
+ * Ensure the given texture is ready for rendering.
+ *
+ * Mostly this means populating the texture object's mipmap tree.
  */
-
-static GLboolean r300EnableTexture2D(GLcontext * ctx, int unit)
+static GLboolean r300_validate_texture(GLcontext * ctx, struct gl_texture_object *texObj)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
+	radeonTexObj *t = radeon_tex_obj(texObj);
 
-	ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D);
-
-	if (t->base.dirty_images[0]) {
-		R300_FIREVERTICES(rmesa);
+	if (!radeon_validate_texture_miptree(ctx, texObj))
+		return GL_FALSE;
 
-		r300SetTexImages(rmesa, tObj);
-		r300UploadTexImages(rmesa, (r300TexObjPtr) tObj->DriverData, 0);
-		if (!t->base.memBlock && !t->image_override)
-			return GL_FALSE;
-	}
+	/* Configure the hardware registers (more precisely, the cached version
+	 * of the hardware registers). */
+	setup_hardware_state(rmesa, t);
 
+	t->validated = GL_TRUE;
 	return GL_TRUE;
 }
 
-static GLboolean r300EnableTexture3D(GLcontext * ctx, int unit)
+/**
+ * Ensure all enabled and complete textures are uploaded along with any buffers being used.
+ */
+GLboolean r300ValidateBuffers(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
+	struct radeon_renderbuffer *rrb;
+	int i;
+	int ret;
 
-	ASSERT(tObj->Target == GL_TEXTURE_3D);
+	radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
 
-	/* r300 does not support mipmaps for 3D textures. */
-	if ((tObj->MinFilter != GL_NEAREST) && (tObj->MinFilter != GL_LINEAR)) {
-		return GL_FALSE;
+	rrb = radeon_get_colorbuffer(&rmesa->radeon);
+	/* color buffer */
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
 	}
 
-	if (t->base.dirty_images[0]) {
-		R300_FIREVERTICES(rmesa);
-		r300SetTexImages(rmesa, tObj);
-		r300UploadTexImages(rmesa, (r300TexObjPtr) tObj->DriverData, 0);
-		if (!t->base.memBlock)
-			return GL_FALSE;
+	/* depth buffer */
+	rrb = radeon_get_depthbuffer(&rmesa->radeon);
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
 	}
+	
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+		radeonTexObj *t;
 
-	return GL_TRUE;
-}
-
-static GLboolean r300EnableTextureCube(GLcontext * ctx, int unit)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-	GLuint face;
-
-	ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
-
-	if (t->base.dirty_images[0] || t->base.dirty_images[1] ||
-	    t->base.dirty_images[2] || t->base.dirty_images[3] ||
-	    t->base.dirty_images[4] || t->base.dirty_images[5]) {
-		/* flush */
-		R300_FIREVERTICES(rmesa);
-		/* layout memory space, once for all faces */
-		r300SetTexImages(rmesa, tObj);
-	}
+		if (!ctx->Texture.Unit[i]._ReallyEnabled)
+			continue;
 
-	/* upload (per face) */
-	for (face = 0; face < 6; face++) {
-		if (t->base.dirty_images[face]) {
-			r300UploadTexImages(rmesa,
-					    (r300TexObjPtr) tObj->DriverData,
-					    face);
+		if (!r300_validate_texture(ctx, ctx->Texture.Unit[i]._Current)) {
+			_mesa_warning(ctx,
+				      "failed to validate texture for unit %d.\n",
+				      i);
 		}
+		t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
+		if (t->image_override && t->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+		else if (t->mt->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->mt->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
 	}
 
-	if (!t->base.memBlock) {
-		/* texmem alloc failed, use s/w fallback */
+	ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+	if (ret)
 		return GL_FALSE;
-	}
-
 	return GL_TRUE;
 }
 
-static GLboolean r300EnableTextureRect(GLcontext * ctx, int unit)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_Current;
-	r300TexObjPtr t = (r300TexObjPtr) tObj->DriverData;
-
-	ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV);
-
-	if (t->base.dirty_images[0]) {
-		R300_FIREVERTICES(rmesa);
-
-		r300SetTexImages(rmesa, tObj);
-		r300UploadTexImages(rmesa, (r300TexObjPtr) tObj->DriverData, 0);
-		if (!t->base.memBlock && !t->image_override &&
-		    !rmesa->prefer_gart_client_texturing)
-			return GL_FALSE;
-	}
-
-	return GL_TRUE;
-}
-
-static GLboolean r300UpdateTexture(GLcontext * ctx, int unit)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-	struct gl_texture_object *tObj = texUnit->_ReallyEnabled ?
-		texUnit->_Current : NULL;
-	r300TexObjPtr t = tObj ? (r300TexObjPtr) tObj->DriverData : NULL;
-
-	/* Fallback if there's a texture border */
-	if (tObj && tObj->Image[0][tObj->BaseLevel]->Border > 0) {
-		tObj = NULL;
-		t = NULL;
-	}
-
-	/* Update state if this is a different texture object to last
-	 * time.
-	 */
-	if (rmesa->state.texture.unit[unit].texobj != tObj) {
-		if (rmesa->state.texture.unit[unit].texobj != NULL) {
-			r300TexObjPtr t_old = (r300TexObjPtr) rmesa->state.texture.unit[unit].texobj->DriverData;
-
-			/* The old texture is no longer bound to this texture unit.
-			 * Mark it as such.
-			 */
-
-			t_old->base.bound &= ~(1 << unit);
-		}
-
-		_mesa_reference_texobj(&rmesa->state.texture.unit[unit].texobj, tObj);
-
-		if (t) {
-			t->base.bound |= (1 << unit);
-			driUpdateTextureLRU(&t->base);	/* XXX: should be locked! */
-		}
-	}
-
-	return !t || !t->border_fallback;
-}
-
 void r300SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 		      unsigned long long offset, GLint depth, GLuint pitch)
 {
 	r300ContextPtr rmesa = pDRICtx->driverPrivate;
 	struct gl_texture_object *tObj =
 	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
-	r300TexObjPtr t;
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
 	uint32_t pitch_val;
 
 	if (!tObj)
 		return;
 
-	t = (r300TexObjPtr) tObj->DriverData;
-
 	t->image_override = GL_TRUE;
 
 	if (!offset)
 		return;
 
-	t->offset = offset;
-	t->pitch_reg &= (1 << 13) -1;
+	t->bo = NULL;
+	t->override_offset = offset;
+	t->pp_txpitch &= (1 << 13) -1;
 	pitch_val = pitch;
 
 	switch (depth) {
 	case 32:
-		t->format = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
-		t->filter |= tx_table[2].filter;
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[2].filter;
 		pitch_val /= 4;
 		break;
 	case 24:
 	default:
-		t->format = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
-		t->filter |= tx_table[4].filter;
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[4].filter;
 		pitch_val /= 4;
 		break;
 	case 16:
-		t->format = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5);
-		t->filter |= tx_table[5].filter;
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5);
+		t->pp_txfilter |= tx_table[5].filter;
 		pitch_val /= 2;
 		break;
 	}
 	pitch_val--;
 
-	t->pitch_reg |= pitch_val;
+	t->pp_txpitch |= pitch_val;
 }
 
-static GLboolean r300UpdateTextureUnit(GLcontext * ctx, int unit)
+void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format, __DRIdrawable *dPriv)
 {
-	struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-
-	if (texUnit->_ReallyEnabled & (TEXTURE_RECT_BIT)) {
-		return (r300EnableTextureRect(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled & (TEXTURE_1D_BIT | TEXTURE_2D_BIT)) {
-		return (r300EnableTexture2D(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled & (TEXTURE_3D_BIT)) {
-		return (r300EnableTexture3D(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled & (TEXTURE_CUBE_BIT)) {
-		return (r300EnableTextureCube(ctx, unit) &&
-			r300UpdateTexture(ctx, unit));
-	} else if (texUnit->_ReallyEnabled) {
-		return GL_FALSE;
-	} else {
-		return r300UpdateTexture(ctx, unit);
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	r300ContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
 	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+	
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->tile_bits = 0;
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	t->pp_txpitch &= (1 << 13) -1;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		else
+			t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[2].filter;
+		pitch_val /= 4;
+		break;
+	case 3:
+	default:
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		t->pp_txfilter |= tx_table[4].filter;
+		pitch_val /= 4;
+		break;
+	case 2:
+		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5);
+		t->pp_txfilter |= tx_table[5].filter;
+		pitch_val /= 2;
+		break;
+	}
+	pitch_val--;
+	t->pp_txsize = ((rb->base.Width - 1) << R300_TX_WIDTHMASK_SHIFT) |
+              ((rb->base.Height - 1) << R300_TX_HEIGHTMASK_SHIFT);
+	t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN;
+	t->pp_txpitch |= pitch_val;
+
+	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+	    if (rb->base.Width > 2048)
+		t->pp_txpitch |= R500_TXWIDTH_BIT11;
+	    if (rb->base.Height > 2048)
+		t->pp_txpitch |= R500_TXHEIGHT_BIT11;
+	}
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
 }
 
-void r300UpdateTextureState(GLcontext * ctx)
+void r300SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
 {
-	int i;
-
-	for (i = 0; i < 8; i++) {
-		if (!r300UpdateTextureUnit(ctx, i)) {
-			_mesa_warning(ctx,
-				      "failed to update texture state for unit %d.\n",
-				      i);
-		}
-	}
+        r300SetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 146daa367c..de32013032 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -32,12 +32,17 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/macros.h"
 #include "main/enums.h"
 #include "shader/program.h"
+#include "shader/programopt.h"
 #include "shader/prog_instruction.h"
+#include "shader/prog_optimize.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "tnl/tnl.h"
 
+#include "radeon_nqssadce.h"
 #include "r300_context.h"
+#include "r300_state.h"
 
 /* TODO: Get rid of t_src_class call */
 #define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \
@@ -64,20 +69,18 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 		int u_temp_used = (VSF_MAX_FRAGMENT_TEMPS - 1) - u_temp_i; \
 		if((vp->num_temporaries + u_temp_used) > VSF_MAX_FRAGMENT_TEMPS) { \
 			WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_used); \
-			vp->native = GL_FALSE; \
+			vp->error = GL_TRUE; \
 		} \
 		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
 	} while (0)
 
-int r300VertexProgUpdateParams(GLcontext * ctx,
-			       struct r300_vertex_program_cont *vp, float *dst)
+static int r300VertexProgUpdateParams(GLcontext * ctx, struct gl_vertex_program *vp, float *dst)
 {
 	int pi;
-	struct gl_vertex_program *mesa_vp = &vp->mesa_program;
 	float *dst_o = dst;
 	struct gl_program_parameter_list *paramList;
 
-	if (mesa_vp->IsNVProgram) {
+	if (vp->IsNVProgram) {
 		_mesa_load_tracked_matrices(ctx);
 
 		for (pi = 0; pi < MAX_NV_VERTEX_PROGRAM_PARAMS; pi++) {
@@ -89,16 +92,18 @@ int r300VertexProgUpdateParams(GLcontext * ctx,
 		return dst - dst_o;
 	}
 
-	assert(mesa_vp->Base.Parameters);
-	_mesa_load_state_parameters(ctx, mesa_vp->Base.Parameters);
+	if (!vp->Base.Parameters)
+		return 0;
 
-	if (mesa_vp->Base.Parameters->NumParameters * 4 >
+	_mesa_load_state_parameters(ctx, vp->Base.Parameters);
+
+	if (vp->Base.Parameters->NumParameters * 4 >
 	    VSF_MAX_FRAGMENT_LENGTH) {
 		fprintf(stderr, "%s:Params exhausted\n", __FUNCTION__);
 		_mesa_exit(-1);
 	}
 
-	paramList = mesa_vp->Base.Parameters;
+	paramList = vp->Base.Parameters;
 	for (pi = 0; pi < paramList->NumParameters; pi++) {
 		switch (paramList->Parameters[pi].Type) {
 		case PROGRAM_STATE_VAR:
@@ -214,21 +219,8 @@ static void vp_dump_inputs(struct r300_vertex_program *vp, char *caller)
 static unsigned long t_src_index(struct r300_vertex_program *vp,
 				 struct prog_src_register *src)
 {
-	int i;
-	int max_reg = -1;
-
 	if (src->File == PROGRAM_INPUT) {
-		if (vp->inputs[src->Index] != -1)
-			return vp->inputs[src->Index];
-
-		for (i = 0; i < VERT_ATTRIB_MAX; i++)
-			if (vp->inputs[i] > max_reg)
-				max_reg = vp->inputs[i];
-
-		vp->inputs[src->Index] = max_reg + 1;
-
-		//vp_dump_inputs(vp, __FUNCTION__);
-
+		assert(vp->inputs[src->Index] != -1);
 		return vp->inputs[src->Index];
 	} else {
 		if (src->Index < 0) {
@@ -943,60 +935,80 @@ static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp,
 static void t_inputs_outputs(struct r300_vertex_program *vp)
 {
 	int i;
-	int cur_reg = 0;
+	int cur_reg;
+	GLuint OutputsWritten, InputsRead;
 
-	for (i = 0; i < VERT_ATTRIB_MAX; i++)
-		vp->inputs[i] = -1;
+	OutputsWritten = vp->Base->Base.OutputsWritten;
+	InputsRead = vp->Base->Base.InputsRead;
+
+	cur_reg = -1;
+	for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+		if (InputsRead & (1 << i))
+			vp->inputs[i] = ++cur_reg;
+		else
+			vp->inputs[i] = -1;
+	}
 
+	cur_reg = 0;
 	for (i = 0; i < VERT_RESULT_MAX; i++)
 		vp->outputs[i] = -1;
 
-	assert(vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS));
+	assert(OutputsWritten & (1 << VERT_RESULT_HPOS));
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)) {
+	if (OutputsWritten & (1 << VERT_RESULT_HPOS)) {
 		vp->outputs[VERT_RESULT_HPOS] = cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
+	if (OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
 		vp->outputs[VERT_RESULT_PSIZ] = cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) {
+	/* If we're writing back facing colors we need to send
+	 * four colors to make front/back face colors selection work.
+	 * If the vertex program doesn't write all 4 colors, lets
+	 * pretend it does by skipping output index reg so the colors
+	 * get written into appropriate output vectors.
+	 */
+	if (OutputsWritten & (1 << VERT_RESULT_COL0)) {
 		vp->outputs[VERT_RESULT_COL0] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) {
-		vp->outputs[VERT_RESULT_COL1] =
-		    vp->outputs[VERT_RESULT_COL0] + 1;
-		cur_reg = vp->outputs[VERT_RESULT_COL1] + 1;
+	if (OutputsWritten & (1 << VERT_RESULT_COL1)) {
+		vp->outputs[VERT_RESULT_COL1] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
-		vp->outputs[VERT_RESULT_BFC0] =
-		    vp->outputs[VERT_RESULT_COL0] + 2;
-		cur_reg = vp->outputs[VERT_RESULT_BFC0] + 2;
+	if (OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+		vp->outputs[VERT_RESULT_BFC0] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
-		vp->outputs[VERT_RESULT_BFC1] =
-		    vp->outputs[VERT_RESULT_COL0] + 3;
-		cur_reg = vp->outputs[VERT_RESULT_BFC1] + 1;
+	if (OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		vp->outputs[VERT_RESULT_BFC1] = cur_reg++;
+	} else if (OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+		cur_reg++;
 	}
 
 	for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) {
-		if (vp->key.OutputsWritten & (1 << i)) {
+		if (OutputsWritten & (1 << i)) {
 			vp->outputs[i] = cur_reg++;
 		}
 	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC)) {
+	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
 		vp->outputs[VERT_RESULT_FOGC] = cur_reg++;
 	}
 }
 
-static void r300TranslateVertexShader(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi)
+void r300TranslateVertexShader(struct r300_vertex_program *vp)
 {
+	struct prog_instruction *vpi = vp->Base->Base.Instructions;
 	int i;
 	GLuint *inst;
 	unsigned long num_operands;
@@ -1007,14 +1019,13 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 	struct prog_src_register src[3];
 
 	vp->pos_end = 0;	/* Not supported yet */
-	vp->program.length = 0;
-	/*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */
+	vp->hw_code.length = 0;
 	vp->translated = GL_TRUE;
-	vp->native = GL_TRUE;
+	vp->error = GL_FALSE;
 
 	t_inputs_outputs(vp);
 
-	for (inst = vp->program.body.i; vpi->Opcode != OPCODE_END;
+	for (inst = vp->hw_code.body.d; vpi->Opcode != OPCODE_END;
 	     vpi++, inst += 4) {
 
 		FREE_TEMPS();
@@ -1176,293 +1187,541 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 						      &u_temp_i);
 			break;
 		default:
-			assert(0);
+			vp->error = GL_TRUE;
 			break;
 		}
 	}
 
-	/* Some outputs may be artificially added, to match the inputs
-	   of the fragment program. Blank the outputs here. */
-	for (i = 0; i < VERT_RESULT_MAX; i++) {
-		if (vp->key.OutputsAdded & (1 << i)) {
-			inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-						     GL_FALSE,
-						     GL_FALSE,
-						     vp->outputs[i],
-						     VSF_FLAG_ALL,
-						     PVS_DST_REG_OUT);
-			inst[1] = __CONST(0, SWIZZLE_ZERO);
-			inst[2] = __CONST(0, SWIZZLE_ZERO);
-			inst[3] = __CONST(0, SWIZZLE_ZERO);
-			inst += 4;
-		}
-	}
-
-	vp->program.length = (inst - vp->program.body.i);
-	if (vp->program.length >= VSF_MAX_FRAGMENT_LENGTH) {
-		vp->program.length = 0;
-		vp->native = GL_FALSE;
+	vp->hw_code.length = (inst - vp->hw_code.body.d);
+	if (vp->hw_code.length >= VSF_MAX_FRAGMENT_LENGTH) {
+		vp->error = GL_TRUE;
 	}
-#if 0
-	fprintf(stderr, "hw program:\n");
-	for (i = 0; i < vp->program.length; i++)
-		fprintf(stderr, "%08x\n", vp->program.body.d[i]);
-#endif
 }
 
-/* DP4 version seems to trigger some hw peculiarity */
-//#define PREFER_DP4
+static void insert_wpos(struct gl_program *prog, GLuint temp_index, int tex_id)
+{
+	struct prog_instruction *vpi;
+
+	_mesa_insert_instructions(prog, prog->NumInstructions - 1, 2);
+
+	vpi = &prog->Instructions[prog->NumInstructions - 3];
+
+	vpi->Opcode = OPCODE_MOV;
+
+	vpi->DstReg.File = PROGRAM_OUTPUT;
+	vpi->DstReg.Index = VERT_RESULT_HPOS;
+	vpi->DstReg.WriteMask = WRITEMASK_XYZW;
+	vpi->DstReg.CondMask = COND_TR;
 
-static void position_invariant(struct gl_program *prog)
+	vpi->SrcReg[0].File = PROGRAM_TEMPORARY;
+	vpi->SrcReg[0].Index = temp_index;
+	vpi->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	++vpi;
+
+	vpi->Opcode = OPCODE_MOV;
+
+	vpi->DstReg.File = PROGRAM_OUTPUT;
+	vpi->DstReg.Index = VERT_RESULT_TEX0 + tex_id;
+	vpi->DstReg.WriteMask = WRITEMASK_XYZW;
+	vpi->DstReg.CondMask = COND_TR;
+
+	vpi->SrcReg[0].File = PROGRAM_TEMPORARY;
+	vpi->SrcReg[0].Index = temp_index;
+	vpi->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	++vpi;
+
+	vpi->Opcode = OPCODE_END;
+}
+
+static void pos_as_texcoord(struct gl_program *prog, int tex_id)
 {
 	struct prog_instruction *vpi;
-	struct gl_program_parameter_list *paramList;
-	int i;
+	GLuint tempregi = prog->NumTemporaries;
 
-	gl_state_index tokens[STATE_LENGTH] = { STATE_MVP_MATRIX, 0, 0, 0, 0 };
+	prog->NumTemporaries++;
 
-	/* tokens[4] = matrix modifier */
-#ifdef PREFER_DP4
-	tokens[4] = 0;		/* not transposed or inverted */
-#else
-	tokens[4] = STATE_MATRIX_TRANSPOSE;
-#endif
-	paramList = prog->Parameters;
-
-	vpi = _mesa_alloc_instructions(prog->NumInstructions + 4);
-	_mesa_init_instructions(vpi, prog->NumInstructions + 4);
-
-	for (i = 0; i < 4; i++) {
-		GLint idx;
-		tokens[2] = tokens[3] = i;	/* matrix row[i]..row[i] */
-		idx = _mesa_add_state_reference(paramList, tokens);
-#ifdef PREFER_DP4
-		vpi[i].Opcode = OPCODE_DP4;
-		vpi[i].StringPos = 0;
-		vpi[i].Data = 0;
-
-		vpi[i].DstReg.File = PROGRAM_OUTPUT;
-		vpi[i].DstReg.Index = VERT_RESULT_HPOS;
-		vpi[i].DstReg.WriteMask = 1 << i;
-		vpi[i].DstReg.CondMask = COND_TR;
-
-		vpi[i].SrcReg[0].File = PROGRAM_STATE_VAR;
-		vpi[i].SrcReg[0].Index = idx;
-		vpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-		vpi[i].SrcReg[1].File = PROGRAM_INPUT;
-		vpi[i].SrcReg[1].Index = VERT_ATTRIB_POS;
-		vpi[i].SrcReg[1].Swizzle = SWIZZLE_XYZW;
-#else
-		if (i == 0)
-			vpi[i].Opcode = OPCODE_MUL;
-		else
-			vpi[i].Opcode = OPCODE_MAD;
+	for (vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++) {
+		if (vpi->DstReg.File == PROGRAM_OUTPUT && vpi->DstReg.Index == VERT_RESULT_HPOS) {
+			vpi->DstReg.File = PROGRAM_TEMPORARY;
+			vpi->DstReg.Index = tempregi;
+		}
+	}
 
-		vpi[i].Data = 0;
+	insert_wpos(prog, tempregi, tex_id);
 
-		if (i == 3)
-			vpi[i].DstReg.File = PROGRAM_OUTPUT;
-		else
-			vpi[i].DstReg.File = PROGRAM_TEMPORARY;
-		vpi[i].DstReg.Index = 0;
-		vpi[i].DstReg.WriteMask = 0xf;
-		vpi[i].DstReg.CondMask = COND_TR;
-
-		vpi[i].SrcReg[0].File = PROGRAM_STATE_VAR;
-		vpi[i].SrcReg[0].Index = idx;
-		vpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-		vpi[i].SrcReg[1].File = PROGRAM_INPUT;
-		vpi[i].SrcReg[1].Index = VERT_ATTRIB_POS;
-		vpi[i].SrcReg[1].Swizzle = MAKE_SWIZZLE4(i, i, i, i);
-
-		if (i > 0) {
-			vpi[i].SrcReg[2].File = PROGRAM_TEMPORARY;
-			vpi[i].SrcReg[2].Index = 0;
-			vpi[i].SrcReg[2].Swizzle = SWIZZLE_XYZW;
+	prog->OutputsWritten |= 1 << (VERT_RESULT_TEX0 + tex_id);
+}
+
+/**
+ * The fogcoord attribute is special in that only the first component
+ * is relevant, and the remaining components are always fixed (when read
+ * from by the fragment program) to yield an X001 pattern.
+ *
+ * We need to enforce this either in the vertex program or in the fragment
+ * program, and this code chooses not to enforce it in the vertex program.
+ * This is slightly cheaper, as long as the fragment program does not use
+ * weird swizzles.
+ *
+ * And it seems that usually, weird swizzles are not used, so...
+ *
+ * See also the counterpart rewriting for fragment programs.
+ */
+static void fog_as_texcoord(struct gl_program *prog, int tex_id)
+{
+	struct prog_instruction *vpi;
+
+	vpi = prog->Instructions;
+	while (vpi->Opcode != OPCODE_END) {
+		if (vpi->DstReg.File == PROGRAM_OUTPUT && vpi->DstReg.Index == VERT_RESULT_FOGC) {
+			vpi->DstReg.Index = VERT_RESULT_TEX0 + tex_id;
+			vpi->DstReg.WriteMask = WRITEMASK_X;
 		}
-#endif
+
+		++vpi;
 	}
 
-	_mesa_copy_instructions(&vpi[i], prog->Instructions,
-				prog->NumInstructions);
+	prog->OutputsWritten &= ~(1 << VERT_RESULT_FOGC);
+	prog->OutputsWritten |= 1 << (VERT_RESULT_TEX0 + tex_id);
+}
 
-	free(prog->Instructions);
+static int translateABS(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
 
-	prog->Instructions = vpi;
+	inst = &prog->Instructions[pos];
 
-	prog->NumInstructions += 4;
-	vpi = &prog->Instructions[prog->NumInstructions - 1];
+	inst->Opcode = OPCODE_MAX;
+	inst->SrcReg[1] = inst->SrcReg[0];
+	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
 
-	assert(vpi->Opcode == OPCODE_END);
+	return 0;
 }
 
-static void insert_wpos(struct r300_vertex_program *vp, struct gl_program *prog,
-			GLuint temp_index)
+static int translateDP3(struct gl_program *prog, int pos)
 {
-	struct prog_instruction *vpi;
-	struct prog_instruction *vpi_insert;
-	int i = 0;
+	struct prog_instruction *inst;
 
-	vpi = _mesa_alloc_instructions(prog->NumInstructions + 2);
-	_mesa_init_instructions(vpi, prog->NumInstructions + 2);
-	/* all but END */
-	_mesa_copy_instructions(vpi, prog->Instructions,
-				prog->NumInstructions - 1);
-	/* END */
-	_mesa_copy_instructions(&vpi[prog->NumInstructions + 1],
-				&prog->Instructions[prog->NumInstructions - 1],
-				1);
-	vpi_insert = &vpi[prog->NumInstructions - 1];
+	inst = &prog->Instructions[pos];
 
-	vpi_insert[i].Opcode = OPCODE_MOV;
+	inst->Opcode = OPCODE_DP4;
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
 
-	vpi_insert[i].DstReg.File = PROGRAM_OUTPUT;
-	vpi_insert[i].DstReg.Index = VERT_RESULT_HPOS;
-	vpi_insert[i].DstReg.WriteMask = WRITEMASK_XYZW;
-	vpi_insert[i].DstReg.CondMask = COND_TR;
+	return 0;
+}
 
-	vpi_insert[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	vpi_insert[i].SrcReg[0].Index = temp_index;
-	vpi_insert[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-	i++;
+static int translateDPH(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
 
-	vpi_insert[i].Opcode = OPCODE_MOV;
+	inst = &prog->Instructions[pos];
 
-	vpi_insert[i].DstReg.File = PROGRAM_OUTPUT;
-	vpi_insert[i].DstReg.Index = VERT_RESULT_TEX0 + vp->wpos_idx;
-	vpi_insert[i].DstReg.WriteMask = WRITEMASK_XYZW;
-	vpi_insert[i].DstReg.CondMask = COND_TR;
+	inst->Opcode = OPCODE_DP4;
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
 
-	vpi_insert[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	vpi_insert[i].SrcReg[0].Index = temp_index;
-	vpi_insert[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-	i++;
+	return 0;
+}
+
+static int translateFLR(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+	struct prog_dst_register dst;
+	int tmp_idx;
 
-	free(prog->Instructions);
+	tmp_idx = prog->NumTemporaries++;
 
-	prog->Instructions = vpi;
+	_mesa_insert_instructions(prog, pos + 1, 1);
 
-	prog->NumInstructions += i;
-	vpi = &prog->Instructions[prog->NumInstructions - 1];
+	inst = &prog->Instructions[pos];
+	dst = inst->DstReg;
 
-	assert(vpi->Opcode == OPCODE_END);
+	inst->Opcode = OPCODE_FRC;
+	inst->DstReg.File = PROGRAM_TEMPORARY;
+	inst->DstReg.Index = tmp_idx;
+	++inst;
+
+	inst->Opcode = OPCODE_ADD;
+	inst->DstReg = dst;
+	inst->SrcReg[0] = (inst-1)->SrcReg[0];
+	inst->SrcReg[1].File = PROGRAM_TEMPORARY;
+	inst->SrcReg[1].Index = tmp_idx;
+	inst->SrcReg[1].Negate = NEGATE_XYZW;
+
+	return 1;
 }
 
-static void pos_as_texcoord(struct r300_vertex_program *vp,
-			    struct gl_program *prog)
+static int translateSUB(struct gl_program *prog, int pos)
 {
-	struct prog_instruction *vpi;
-	GLuint tempregi = prog->NumTemporaries;
-	/* should do something else if no temps left... */
-	prog->NumTemporaries++;
+	struct prog_instruction *inst;
 
-	for (vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++) {
-		if (vpi->DstReg.File == PROGRAM_OUTPUT
-		    && vpi->DstReg.Index == VERT_RESULT_HPOS) {
-			vpi->DstReg.File = PROGRAM_TEMPORARY;
-			vpi->DstReg.Index = tempregi;
+	inst = &prog->Instructions[pos];
+
+	inst->Opcode = OPCODE_ADD;
+	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
+
+	return 0;
+}
+
+static int translateSWZ(struct gl_program *prog, int pos)
+{
+	prog->Instructions[pos].Opcode = OPCODE_MOV;
+
+	return 0;
+}
+
+static int translateXPD(struct gl_program *prog, int pos)
+{
+	struct prog_instruction *inst;
+	int tmp_idx;
+
+	tmp_idx = prog->NumTemporaries++;
+
+	_mesa_insert_instructions(prog, pos + 1, 1);
+
+	inst = &prog->Instructions[pos];
+
+	*(inst+1) = *inst;
+
+	inst->Opcode = OPCODE_MUL;
+	inst->DstReg.File = PROGRAM_TEMPORARY;
+	inst->DstReg.Index = tmp_idx;
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
+	inst->SrcReg[1].Swizzle = combine_swizzles4(inst->SrcReg[1].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
+	++inst;
+
+	inst->Opcode = OPCODE_MAD;
+	inst->SrcReg[0].Swizzle = combine_swizzles4(inst->SrcReg[0].Swizzle, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W);
+	inst->SrcReg[1].Swizzle = combine_swizzles4(inst->SrcReg[1].Swizzle, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W);
+	inst->SrcReg[1].Negate ^= NEGATE_XYZW;
+	inst->SrcReg[2].File = PROGRAM_TEMPORARY;
+	inst->SrcReg[2].Index = tmp_idx;
+
+	return 1;
+}
+
+static void translateInsts(struct gl_program *prog)
+{
+	struct prog_instruction *inst;
+	int i;
+
+	for (i = 0; i < prog->NumInstructions; ++i) {
+		inst = &prog->Instructions[i];
+
+		switch (inst->Opcode) {
+			case OPCODE_ABS:
+				i += translateABS(prog, i);
+				break;
+			case OPCODE_DP3:
+				i += translateDP3(prog, i);
+				break;
+			case OPCODE_DPH:
+				i += translateDPH(prog, i);
+				break;
+			case OPCODE_FLR:
+				i += translateFLR(prog, i);
+				break;
+			case OPCODE_SUB:
+				i += translateSUB(prog, i);
+				break;
+			case OPCODE_SWZ:
+				i += translateSWZ(prog, i);
+				break;
+			case OPCODE_XPD:
+				i += translateXPD(prog, i);
+				break;
+			default:
+				break;
 		}
 	}
-	insert_wpos(vp, prog, tempregi);
 }
 
-static struct r300_vertex_program *build_program(struct r300_vertex_program_key
-						 *wanted_key, struct gl_vertex_program
-						 *mesa_vp, GLint wpos_idx)
+#define ADD_OUTPUT(fp_attr, vp_result) \
+	do { \
+		if ((FpReads & (1 << (fp_attr))) && !(prog->OutputsWritten & (1 << (vp_result)))) { \
+			OutputsAdded |= 1 << (vp_result); \
+			count++; \
+		} \
+	} while (0)
+
+static void addArtificialOutputs(GLcontext *ctx, struct gl_program *prog)
 {
-	struct r300_vertex_program *vp;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	GLuint OutputsAdded, FpReads;
+	int i, count;
 
-	vp = _mesa_calloc(sizeof(*vp));
-	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
-	vp->wpos_idx = wpos_idx;
+	OutputsAdded = 0;
+	count = 0;
+	FpReads = r300->selected_fp->Base->InputsRead;
+
+	ADD_OUTPUT(FRAG_ATTRIB_COL0, VERT_RESULT_COL0);
+	ADD_OUTPUT(FRAG_ATTRIB_COL1, VERT_RESULT_COL1);
+
+	for (i = 0; i < 7; ++i) {
+		ADD_OUTPUT(FRAG_ATTRIB_TEX0 + i, VERT_RESULT_TEX0 + i);
+	}
+
+	/* Some outputs may be artificially added, to match the inputs of the fragment program.
+	 * Issue 16 of vertex program spec says that all vertex attributes that are unwritten by
+	 * vertex program are undefined, so just use MOV [vertex_result], CONST[0]
+	 */
+	if (count > 0) {
+		struct prog_instruction *inst;
+
+		_mesa_insert_instructions(prog, prog->NumInstructions - 1, count);
+		inst = &prog->Instructions[prog->NumInstructions - 1 - count];
+
+		for (i = 0; i < VERT_RESULT_MAX; ++i) {
+			if (OutputsAdded & (1 << i)) {
+				inst->Opcode = OPCODE_MOV;
+
+				inst->DstReg.File = PROGRAM_OUTPUT;
+				inst->DstReg.Index = i;
+				inst->DstReg.WriteMask = WRITEMASK_XYZW;
+				inst->DstReg.CondMask = COND_TR;
+
+				inst->SrcReg[0].File = PROGRAM_CONSTANT;
+				inst->SrcReg[0].Index = 0;
+				inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
 
-	if (mesa_vp->IsPositionInvariant) {
-		position_invariant(&mesa_vp->Base);
+				++inst;
+			}
+		}
+
+		prog->OutputsWritten |= OutputsAdded;
 	}
+}
+
+#undef ADD_OUTPUT
+
+static void nqssadceInit(struct nqssadce_state* s)
+{
+	r300ContextPtr r300 = R300_CONTEXT(s->Ctx);
+	GLuint fp_reads;
+
+	fp_reads = r300->selected_fp->Base->InputsRead;
+	{
+		if (fp_reads & FRAG_BIT_COL0) {
+				s->Outputs[VERT_RESULT_COL0].Sourced = WRITEMASK_XYZW;
+				s->Outputs[VERT_RESULT_BFC0].Sourced = WRITEMASK_XYZW;
+		}
 
-	if (wpos_idx > -1) {
-		pos_as_texcoord(vp, &mesa_vp->Base);
+		if (fp_reads & FRAG_BIT_COL1) {
+				s->Outputs[VERT_RESULT_COL1].Sourced = WRITEMASK_XYZW;
+				s->Outputs[VERT_RESULT_BFC1].Sourced = WRITEMASK_XYZW;
+		}
 	}
 
-	assert(mesa_vp->Base.NumInstructions);
-	vp->num_temporaries = mesa_vp->Base.NumTemporaries;
-	r300TranslateVertexShader(vp, mesa_vp->Base.Instructions);
+	{
+		int i;
+		for (i = 0; i < 8; ++i) {
+			if (fp_reads & FRAG_BIT_TEX(i)) {
+				s->Outputs[VERT_RESULT_TEX0 + i].Sourced = WRITEMASK_XYZW;
+			}
+		}
+	}
 
-	return vp;
+	s->Outputs[VERT_RESULT_HPOS].Sourced = WRITEMASK_XYZW;
+	if (s->Program->OutputsWritten & (1 << VERT_RESULT_PSIZ))
+		s->Outputs[VERT_RESULT_PSIZ].Sourced = WRITEMASK_X;
 }
 
-static void add_outputs(struct r300_vertex_program_key *key, GLint vert)
+static GLboolean swizzleIsNative(GLuint opcode, struct prog_src_register reg)
 {
-	if (key->OutputsWritten & (1 << vert))
-		return;
+	(void) opcode;
+	(void) reg;
 
-	key->OutputsWritten |= 1 << vert;
-	key->OutputsAdded |= 1 << vert;
+	return GL_TRUE;
 }
 
-void r300SelectVertexShader(r300ContextPtr r300)
+static struct r300_vertex_program *build_program(GLcontext *ctx,
+						 struct r300_vertex_program_key *wanted_key,
+						 const struct gl_vertex_program *mesa_vp)
 {
-	GLcontext *ctx = ctx = r300->radeon.glCtx;
-	GLuint InputsRead;
-	struct r300_vertex_program_key wanted_key = { 0 };
-	GLint i;
-	struct r300_vertex_program_cont *vpc;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	struct r300_vertex_program *vp;
-	GLint wpos_idx;
+	struct gl_program *prog;
 
-	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
-	wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead;
-	wanted_key.OutputsWritten = vpc->mesa_program.Base.OutputsWritten;
-	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-
-	wpos_idx = -1;
-	if (InputsRead & FRAG_BIT_WPOS) {
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-			if (!(InputsRead & (FRAG_BIT_TEX0 << i)))
-				break;
+	vp = _mesa_calloc(sizeof(*vp));
+	vp->Base = (struct gl_vertex_program *) _mesa_clone_program(ctx, &mesa_vp->Base);
+	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
 
-		if (i == ctx->Const.MaxTextureUnits) {
-			fprintf(stderr, "\tno free texcoord found\n");
-			_mesa_exit(-1);
-		}
+	prog = &vp->Base->Base;
 
-		wanted_key.OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-		wpos_idx = i;
+	if (RADEON_DEBUG & DEBUG_VERTS) {
+		fprintf(stderr, "Initial vertex program:\n");
+		_mesa_print_program(prog);
+		fflush(stdout);
 	}
 
-	add_outputs(&wanted_key, VERT_RESULT_HPOS);
+	if (vp->Base->IsPositionInvariant) {
+		_mesa_insert_mvp_code(ctx, vp->Base);
+	}
+
+	if (r300->selected_fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		pos_as_texcoord(&vp->Base->Base, r300->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0);
+	}
 
-	if (InputsRead & FRAG_BIT_COL0) {
-		add_outputs(&wanted_key, VERT_RESULT_COL0);
+	if (r300->selected_fp->fog_attr != FRAG_ATTRIB_MAX) {
+		fog_as_texcoord(&vp->Base->Base, r300->selected_fp->fog_attr - FRAG_ATTRIB_TEX0);
 	}
 
-	if (InputsRead & FRAG_BIT_COL1) {
-		add_outputs(&wanted_key, VERT_RESULT_COL1);
+	addArtificialOutputs(ctx, prog);
+
+	translateInsts(prog);
+
+	if (RADEON_DEBUG & DEBUG_VERTS) {
+		fprintf(stderr, "Vertex program after native rewrite:\n");
+		_mesa_print_program(prog);
+		fflush(stdout);
 	}
 
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (InputsRead & (FRAG_BIT_TEX0 << i)) {
-			add_outputs(&wanted_key, VERT_RESULT_TEX0 + i);
+	{
+		struct radeon_nqssadce_descr nqssadce = {
+			.Init = &nqssadceInit,
+			.IsNativeSwizzle = &swizzleIsNative,
+			.BuildSwizzle = NULL
+		};
+		radeonNqssaDce(ctx, prog, &nqssadce);
+
+		/* We need this step for reusing temporary registers */
+		_mesa_optimize_program(ctx, prog);
+
+		if (RADEON_DEBUG & DEBUG_VERTS) {
+			fprintf(stderr, "Vertex program after NQSSADCE:\n");
+			_mesa_print_program(prog);
+			fflush(stdout);
 		}
 	}
 
-	if (vpc->mesa_program.IsPositionInvariant) {
-		/* we wan't position don't we ? */
-		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
+	assert(prog->NumInstructions);
+	{
+		struct prog_instruction *inst;
+		int max, i, tmp;
+
+		inst = prog->Instructions;
+		max = -1;
+		while (inst->Opcode != OPCODE_END) {
+			tmp = _mesa_num_inst_src_regs(inst->Opcode);
+			for (i = 0; i < tmp; ++i) {
+				if (inst->SrcReg[i].File == PROGRAM_TEMPORARY) {
+					if ((int) inst->SrcReg[i].Index > max) {
+						max = inst->SrcReg[i].Index;
+					}
+				}
+			}
+
+			if (_mesa_num_inst_dst_regs(inst->Opcode)) {
+				if (inst->DstReg.File == PROGRAM_TEMPORARY) {
+					if ((int) inst->DstReg.Index > max) {
+						max = inst->DstReg.Index;
+					}
+				}
+			}
+			++inst;
+		}
+
+		/* We actually want highest index of used temporary register,
+		 * not the number of temporaries used.
+		 * These values aren't always the same.
+		 */
+		vp->num_temporaries = max + 1;
 	}
 
-	for (vp = vpc->progs; vp; vp = vp->next)
+	return vp;
+}
+
+struct r300_vertex_program * r300SelectVertexShader(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_program_key wanted_key = { 0 };
+	struct r300_vertex_program_cont *vpc;
+	struct r300_vertex_program *vp;
+
+	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
+	wanted_key.FpReads = r300->selected_fp->Base->InputsRead;
+	wanted_key.FogAttr = r300->selected_fp->fog_attr;
+	wanted_key.WPosAttr = r300->selected_fp->wpos_attr;
+
+	for (vp = vpc->progs; vp; vp = vp->next) {
 		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
 		    == 0) {
-			r300->selected_vp = vp;
-			return;
+			return r300->selected_vp = vp;
 		}
-	//_mesa_print_program(&vpc->mesa_program.Base);
+	}
 
-	vp = build_program(&wanted_key, &vpc->mesa_program, wpos_idx);
+	vp = build_program(ctx, &wanted_key, &vpc->mesa_program);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
-	r300->selected_vp = vp;
+
+	return r300->selected_vp = vp;
+}
+
+#define bump_vpu_count(ptr, new_count)   do { \
+		drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr)); \
+		int _nc=(new_count)/4; \
+		assert(_nc < 256); \
+		if(_nc>_p->vpu.count)_p->vpu.count=_nc; \
+	} while(0)
+
+static void r300EmitVertexProgram(r300ContextPtr r300, int dest, struct r300_vertex_shader_hw_code *code)
+{
+	int i;
+
+	assert((code->length > 0) && (code->length % 4 == 0));
+
+	switch ((dest >> 8) & 0xf) {
+		case 0:
+			R300_STATECHANGE(r300, vpi);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vpi.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		case 2:
+			R300_STATECHANGE(r300, vpp);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vpp.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		case 4:
+			R300_STATECHANGE(r300, vps);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vps.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		default:
+			fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
+			_mesa_exit(-1);
+	}
+}
+
+void r300SetupVertexProgram(r300ContextPtr rmesa)
+{
+	GLcontext *ctx = rmesa->radeon.glCtx;
+	struct r300_vertex_program *prog = rmesa->selected_vp;
+	int inst_count = 0;
+	int param_count = 0;
+
+	/* Reset state, in case we don't use something */
+	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
+	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
+	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
+
+	R300_STATECHANGE(rmesa, vpp);
+	param_count = r300VertexProgUpdateParams(ctx, prog->Base, (float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
+	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
+	param_count /= 4;
+
+	r300EmitVertexProgram(rmesa, R300_PVS_CODE_START, &(prog->hw_code));
+	inst_count = (prog->hw_code.length / 4) - 1;
+
+	r300VapCntl(rmesa, _mesa_bitcount(prog->Base->Base.InputsRead),
+				 _mesa_bitcount(prog->Base->Base.OutputsWritten), prog->num_temporaries);
+
+	R300_STATECHANGE(rmesa, pvs);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = (0 << R300_PVS_FIRST_INST_SHIFT) | (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
+				(inst_count << R300_PVS_LAST_INST_SHIFT);
+
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
 }
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.h b/src/mesa/drivers/dri/r300/r300_vertprog.h
index 2f35f02bc8..2dab11c337 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.h
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.h
@@ -32,4 +32,10 @@
 
 #endif
 
+void r300SetupVertexProgram(r300ContextPtr rmesa);
+
+struct r300_vertex_program * r300SelectVertexShader(GLcontext *ctx);
+
+void r300TranslateVertexShader(struct r300_vertex_program *vp);
+
 #endif
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.c b/src/mesa/drivers/dri/r300/r500_fragprog.c
index 292573de89..4d58cf2162 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r500_fragprog.c
@@ -27,10 +27,6 @@
 
 #include "r500_fragprog.h"
 
-#include "radeon_nqssadce.h"
-#include "radeon_program_alu.h"
-
-
 static void reset_srcreg(struct prog_src_register* reg)
 {
 	_mesa_bzero(reg, sizeof(*reg));
@@ -58,12 +54,12 @@ static struct prog_src_register shadow_ambient(struct gl_program *program, int t
  *  - introduce a temporary register when write masks are needed
  *
  */
-static GLboolean transform_TEX(
+GLboolean r500_transform_TEX(
 	struct radeon_transform_context *t,
 	struct prog_instruction* orig_inst, void* data)
 {
-	struct r500_fragment_program_compiler *compiler =
-		(struct r500_fragment_program_compiler*)data;
+	struct r300_fragment_program_compiler *compiler =
+		(struct r300_fragment_program_compiler*)data;
 	struct prog_instruction inst = *orig_inst;
 	struct prog_instruction* tgt;
 	GLboolean destredirect = GL_FALSE;
@@ -188,121 +184,7 @@ static GLboolean transform_TEX(
 	return GL_TRUE;
 }
 
-
-static void update_params(r300ContextPtr r300, struct r500_fragment_program *fp)
-{
-	struct gl_fragment_program *mp = &fp->mesa_program;
-
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (mp->Base.Parameters)
-		_mesa_load_state_parameters(r300->radeon.glCtx, mp->Base.Parameters);
-}
-
-
-/**
- * Transform the program to support fragment.position.
- *
- * Introduce a small fragment at the start of the program that will be
- * the only code that directly reads the FRAG_ATTRIB_WPOS input.
- * All other code pieces that reference that input will be rewritten
- * to read from a newly allocated temporary.
- *
- * \todo if/when r5xx supports the radeon_program architecture, this is a
- * likely candidate for code sharing.
- */
-static void insert_WPOS_trailer(struct r500_fragment_program_compiler *compiler)
-{
-	GLuint InputsRead = compiler->fp->mesa_program.Base.InputsRead;
-
-	if (!(InputsRead & FRAG_BIT_WPOS))
-		return;
-
-	static gl_state_index tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
-	};
-	struct prog_instruction *fpi;
-	GLuint window_index;
-	int i = 0;
-	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
-
-	_mesa_insert_instructions(compiler->program, 0, 3);
-	fpi = compiler->program->Instructions;
-
-	/* perspective divide */
-	fpi[i].Opcode = OPCODE_RCP;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_W;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	fpi[i].Opcode = OPCODE_MUL;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[1].Index = tempregi;
-	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	/* viewport transformation */
-	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
-
-	fpi[i].Opcode = OPCODE_MAD;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[0].Index = tempregi;
-	fpi[i].SrcReg[0].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[1].Index = window_index;
-	fpi[i].SrcReg[1].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[2].Index = window_index;
-	fpi[i].SrcReg[2].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-	i++;
-
-	for (; i < compiler->program->NumInstructions; ++i) {
-		int reg;
-		for (reg = 0; reg < 3; reg++) {
-			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
-			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
-				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
-				fpi[i].SrcReg[reg].Index = tempregi;
-			}
-		}
-	}
-}
-
-
-static void nqssadce_init(struct nqssadce_state* s)
-{
-	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
-	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
-}
-
-static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
+GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
 {
 	GLuint relevant;
 	int i;
@@ -314,22 +196,20 @@ static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
 		if (reg.Abs)
 			return GL_FALSE;
 
+		if (opcode == OPCODE_KIL && (reg.Swizzle != SWIZZLE_NOOP || reg.Negate != NEGATE_NONE))
+			return GL_FALSE;
+
 		if (reg.Negate)
 			reg.Negate ^= NEGATE_XYZW;
 
-		if (opcode == OPCODE_KIL) {
-			if (reg.Swizzle != SWIZZLE_NOOP)
-				return GL_FALSE;
-		} else {
-			for(i = 0; i < 4; ++i) {
-				GLuint swz = GET_SWZ(reg.Swizzle, i);
-				if (swz == SWIZZLE_NIL) {
-					reg.Negate &= ~(1 << i);
-					continue;
-				}
-				if (swz >= 4)
-					return GL_FALSE;
+		for(i = 0; i < 4; ++i) {
+			GLuint swz = GET_SWZ(reg.Swizzle, i);
+			if (swz == SWIZZLE_NIL) {
+				reg.Negate &= ~(1 << i);
+				continue;
 			}
+			if (swz >= 4)
+				return GL_FALSE;
 		}
 
 		if (reg.Negate)
@@ -367,8 +247,7 @@ static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
  * The only thing we *cannot* do in an ALU instruction is per-component
  * negation. Therefore, we split the MOV into two instructions when necessary.
  */
-static void nqssadce_build_swizzle(struct nqssadce_state *s,
-	struct prog_dst_register dst, struct prog_src_register src)
+void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src)
 {
 	struct prog_instruction *inst;
 	GLuint negatebase[2] = { 0, 0 };
@@ -392,129 +271,12 @@ static void nqssadce_build_swizzle(struct nqssadce_state *s,
 		inst->DstReg = dst;
 		inst->DstReg.WriteMask = negatebase[i];
 		inst->SrcReg[0] = src;
+		inst->SrcReg[0].Negate = (i == 0) ? NEGATE_NONE : NEGATE_XYZW;
 		inst++;
 		s->IP++;
 	}
 }
 
-static GLuint build_dtm(GLuint depthmode)
-{
-	switch(depthmode) {
-	default:
-	case GL_LUMINANCE: return 0;
-	case GL_INTENSITY: return 1;
-	case GL_ALPHA: return 2;
-	}
-}
-
-static GLuint build_func(GLuint comparefunc)
-{
-	return comparefunc - GL_NEVER;
-}
-
-
-/**
- * Collect all external state that is relevant for compiling the given
- * fragment program.
- */
-static void build_state(
-	r300ContextPtr r300,
-	struct r500_fragment_program *fp,
-	struct r500_fragment_program_external_state *state)
-{
-	int unit;
-
-	_mesa_bzero(state, sizeof(*state));
-
-	for(unit = 0; unit < 16; ++unit) {
-		if (fp->mesa_program.Base.ShadowSamplers & (1 << unit)) {
-			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
-
-			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
-			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
-		}
-	}
-}
-
-static void dump_program(struct r500_fragment_program_code *code);
-
-void r500TranslateFragmentShader(r300ContextPtr r300,
-				 struct r500_fragment_program *fp)
-{
-	struct r500_fragment_program_external_state state;
-
-	build_state(r300, fp, &state);
-	if (_mesa_memcmp(&fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		fp->translated = GL_FALSE;
-		_mesa_memcpy(&fp->state, &state, sizeof(state));
-	}
-
-	if (!fp->translated) {
-		struct r500_fragment_program_compiler compiler;
-
-		compiler.r300 = r300;
-		compiler.fp = fp;
-		compiler.code = &fp->code;
-		compiler.program = _mesa_clone_program(r300->radeon.glCtx, &fp->mesa_program.Base);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: Initial program:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		insert_WPOS_trailer(&compiler);
-
-		struct radeon_program_transformation transformations[] = {
-			{ &transform_TEX, &compiler },
-			{ &radeonTransformALU, 0 },
-			{ &radeonTransformDeriv, 0 },
-			{ &radeonTransformTrigScale, 0 }
-		};
-		radeonLocalTransform(r300->radeon.glCtx, compiler.program,
-			4, transformations);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after native rewrite:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		struct radeon_nqssadce_descr nqssadce = {
-			.Init = &nqssadce_init,
-			.IsNativeSwizzle = &is_native_swizzle,
-			.BuildSwizzle = &nqssadce_build_swizzle,
-			.RewriteDepthOut = GL_TRUE
-		};
-		radeonNqssaDce(r300->radeon.glCtx, compiler.program, &nqssadce);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		fp->translated = r500FragmentProgramEmit(&compiler);
-
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->mesa_program.Base.Parameters);
-		fp->mesa_program.Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
-
-		_mesa_reference_program(r300->radeon.glCtx, &compiler.program, 0);
-
-		r300UpdateStateParameters(r300->radeon.glCtx, _NEW_PROGRAM);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			if (fp->translated) {
-				_mesa_printf("Machine-readable code:\n");
-				dump_program(&fp->code);
-			}
-		}
-
-	}
-
-	update_params(r300, fp);
-
-}
 
 static char *toswiz(int swiz_val) {
   switch(swiz_val) {
@@ -613,9 +375,9 @@ static char *to_texop(int val)
   return NULL;
 }
 
-static void dump_program(struct r500_fragment_program_code *code)
+void r500FragmentProgramDump(union rX00_fragment_program_code *c)
 {
-
+  struct r500_fragment_program_code *code = &c->r500;
   fprintf(stderr, "R500 Fragment Program:\n--------\n");
 
   int n;
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.h b/src/mesa/drivers/dri/r300/r500_fragprog.h
index 1e45538f80..1179bf6607 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog.h
+++ b/src/mesa/drivers/dri/r300/r500_fragprog.h
@@ -33,30 +33,20 @@
 #ifndef __R500_FRAGPROG_H_
 #define __R500_FRAGPROG_H_
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
 #include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
-#include "shader/program.h"
 #include "shader/prog_instruction.h"
 
 #include "r300_context.h"
-#include "r300_state.h"
-#include "radeon_program.h"
+#include "radeon_nqssadce.h"
 
-struct r500_fragment_program;
+extern GLboolean r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler);
 
-extern void r500TranslateFragmentShader(r300ContextPtr r300,
-					struct r500_fragment_program *fp);
+extern void r500FragmentProgramDump(union rX00_fragment_program_code *c);
 
-struct r500_fragment_program_compiler {
-	r300ContextPtr r300;
-	struct r500_fragment_program *fp;
-	struct r500_fragment_program_code *code;
-	struct gl_program *program;
-};
+extern GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg);
 
-extern GLboolean r500FragmentProgramEmit(struct r500_fragment_program_compiler *compiler);
+extern void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src);
+
+extern GLboolean r500_transform_TEX(struct radeon_transform_context *t, struct prog_instruction* orig_inst, void* data);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
index 4631235f0d..30f4514897 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
@@ -49,8 +49,8 @@
 
 
 #define PROG_CODE \
-	struct r500_fragment_program_compiler *c = (struct r500_fragment_program_compiler*)data; \
-	struct r500_fragment_program_code *code = c->code
+	struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \
+	struct r500_fragment_program_code *code = &c->code->r500
 
 #define error(fmt, args...) do {			\
 		fprintf(stderr, "%s::%s(): " fmt "\n",	\
@@ -72,7 +72,7 @@ static GLboolean emit_const(void *data, GLuint file, GLuint idx, GLuint *hwindex
 	}
 
 	if (*hwindex >= code->const_nr) {
-		if (*hwindex >= PFS_NUM_CONST_REGS) {
+		if (*hwindex >= R500_PFS_NUM_CONST_REGS) {
 			error("Out of hw constants!\n");
 			return GL_FALSE;
 		}
@@ -299,9 +299,9 @@ static const struct radeon_pair_handler pair_handler = {
 	.MaxHwTemps = 128
 };
 
-GLboolean r500FragmentProgramEmit(struct r500_fragment_program_compiler *compiler)
+GLboolean r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler)
 {
-	struct r500_fragment_program_code *code = compiler->code;
+	struct r500_fragment_program_code *code = &compiler->code->r500;
 
 	_mesa_bzero(code, sizeof(*code));
 	code->max_temp_idx = 1;
diff --git a/src/mesa/drivers/dri/r300/radeon_context.c b/src/mesa/drivers/dri/r300/radeon_context.c
deleted file mode 100644
index 5267fe9a77..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_context.c
+++ /dev/null
@@ -1,330 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/**
- * \file radeon_context.c
- * Common context initialization.
- *
- * \author Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include <dlfcn.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/state.h"
-#include "main/matrix.h"
-#include "main/framebuffer.h"
-
-#include "drivers/common/driverfuncs.h"
-#include "swrast/swrast.h"
-
-#include "radeon_screen.h"
-#include "radeon_ioctl.h"
-#include "radeon_macros.h"
-#include "radeon_reg.h"
-
-#include "radeon_state.h"
-#include "r300_state.h"
-
-#include "utils.h"
-#include "vblank.h"
-#include "xmlpool.h"		/* for symbolic values of enum-type options */
-
-#define DRIVER_DATE "20060815"
-
-
-/* Return various strings for glGetString().
- */
-static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-	static char buffer[128];
-
-	switch (name) {
-	case GL_VENDOR:
-		if (IS_R300_CLASS(radeon->radeonScreen))
-			return (GLubyte *) "DRI R300 Project";
-		else
-			return (GLubyte *) "Tungsten Graphics, Inc.";
-
-	case GL_RENDERER:
-	{
-		unsigned offset;
-		GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
-			radeon->radeonScreen->AGPMode;
-		const char* chipname;
-
-		if (IS_R300_CLASS(radeon->radeonScreen))
-			chipname = "R300";
-		else
-			chipname = "R200";
-
-		offset = driGetRendererString(buffer, chipname, DRIVER_DATE,
-					      agp_mode);
-
-		if (IS_R300_CLASS(radeon->radeonScreen)) {
-		sprintf(&buffer[offset], " %sTCL",
-			(radeon->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)
-			? "" : "NO-");
-		} else {
-			sprintf(&buffer[offset], " %sTCL",
-			!(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)
-			? "" : "NO-");
-		}
-
-		return (GLubyte *) buffer;
-	}
-
-	default:
-		return NULL;
-	}
-}
-
-/* Initialize the driver's misc functions.
- */
-static void radeonInitDriverFuncs(struct dd_function_table *functions)
-{
-	functions->GetString = radeonGetString;
-}
-
-
-/**
- * Create and initialize all common fields of the context,
- * including the Mesa context itself.
- */
-GLboolean radeonInitContext(radeonContextPtr radeon,
-			    struct dd_function_table* functions,
-			    const __GLcontextModes * glVisual,
-			    __DRIcontextPrivate * driContextPriv,
-			    void *sharedContextPrivate)
-{
-	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
-	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
-	GLcontext* ctx;
-	GLcontext* shareCtx;
-	int fthrottle_mode;
-
-	/* Fill in additional standard functions. */
-	radeonInitDriverFuncs(functions);
-
-	radeon->radeonScreen = screen;
-	/* Allocate and initialize the Mesa context */
-	if (sharedContextPrivate)
-		shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx;
-	else
-		shareCtx = NULL;
-	radeon->glCtx = _mesa_create_context(glVisual, shareCtx,
-					    functions, (void *)radeon);
-	if (!radeon->glCtx)
-		return GL_FALSE;
-
-	ctx = radeon->glCtx;
-	driContextPriv->driverPrivate = radeon;
-
-	/* DRI fields */
-	radeon->dri.context = driContextPriv;
-	radeon->dri.screen = sPriv;
-	radeon->dri.drawable = NULL;
-	radeon->dri.readable = NULL;
-	radeon->dri.hwContext = driContextPriv->hHWContext;
-	radeon->dri.hwLock = &sPriv->pSAREA->lock;
-	radeon->dri.fd = sPriv->fd;
-	radeon->dri.drmMinor = sPriv->drm_version.minor;
-
-	radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA +
-					       screen->sarea_priv_offset);
-
-	/* Setup IRQs */
-	fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode");
-	radeon->iw.irq_seq = -1;
-	radeon->irqsEmitted = 0;
-	radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS &&
-			  radeon->radeonScreen->irq);
-
-	radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
-
-	if (!radeon->do_irqs)
-		fprintf(stderr,
-			"IRQ's not enabled, falling back to %s: %d %d\n",
-			radeon->do_usleeps ? "usleeps" : "busy waits",
-			fthrottle_mode, radeon->radeonScreen->irq);
-
-	(*sPriv->systemTime->getUST) (&radeon->swap_ust);
-
-	return GL_TRUE;
-}
-
-
-/**
- * Cleanup common context fields.
- * Called by r200DestroyContext/r300DestroyContext
- */
-void radeonCleanupContext(radeonContextPtr radeon)
-{
-	/* _mesa_destroy_context() might result in calls to functions that
-	 * depend on the DriverCtx, so don't set it to NULL before.
-	 *
-	 * radeon->glCtx->DriverCtx = NULL;
-	 */
-
-	/* free the Mesa context */
-	_mesa_destroy_context(radeon->glCtx);
-
-	if (radeon->state.scissor.pClipRects) {
-		FREE(radeon->state.scissor.pClipRects);
-		radeon->state.scissor.pClipRects = 0;
-	}
-}
-
-
-/**
- * Swap front and back buffer.
- */
-void radeonSwapBuffers(__DRIdrawablePrivate * dPriv)
-{
-	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-		radeonContextPtr radeon;
-		GLcontext *ctx;
-
-		radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-		ctx = radeon->glCtx;
-
-		if (ctx->Visual.doubleBufferMode) {
-			_mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
-			if (radeon->doPageFlip) {
-				radeonPageFlip(dPriv);
-			} else {
-			    radeonCopyBuffer(dPriv, NULL);
-			}
-		}
-	} else {
-		/* XXX this shouldn't be an error but we can't handle it for now */
-		_mesa_problem(NULL, "%s: drawable has no context!",
-			      __FUNCTION__);
-	}
-}
-
-void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-			 int x, int y, int w, int h )
-{
-    if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-	radeonContextPtr radeon;
-	GLcontext *ctx;
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-	ctx = radeon->glCtx;
-
-	if (ctx->Visual.doubleBufferMode) {
-	    drm_clip_rect_t rect;
-	    rect.x1 = x + dPriv->x;
-	    rect.y1 = (dPriv->h - y - h) + dPriv->y;
-	    rect.x2 = rect.x1 + w;
-	    rect.y2 = rect.y1 + h;
-	    _mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
-	    radeonCopyBuffer(dPriv, &rect);
-	}
-    } else {
-	/* XXX this shouldn't be an error but we can't handle it for now */
-	_mesa_problem(NULL, "%s: drawable has no context!",
-		      __FUNCTION__);
-    }
-}
-
-/* Force the context `c' to be the current context and associate with it
- * buffer `b'.
- */
-GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
-			    __DRIdrawablePrivate * driDrawPriv,
-			    __DRIdrawablePrivate * driReadPriv)
-{
-	if (driContextPriv) {
-		radeonContextPtr radeon =
-			(radeonContextPtr) driContextPriv->driverPrivate;
-
-		if (RADEON_DEBUG & DEBUG_DRI)
-			fprintf(stderr, "%s ctx %p\n", __FUNCTION__,
-				radeon->glCtx);
-
-		if (radeon->dri.drawable != driDrawPriv) {
-			if (driDrawPriv->swap_interval == (unsigned)-1) {
-				driDrawPriv->vblFlags =
-					(radeon->radeonScreen->irq != 0)
-					? driGetDefaultVBlankFlags(&radeon->
-								   optionCache)
-					: VBLANK_FLAG_NO_IRQ;
-
-				driDrawableInitVBlank(driDrawPriv);
-			}
-		}
-
-		radeon->dri.readable = driReadPriv;
-
-		if (radeon->dri.drawable != driDrawPriv ||
-		    radeon->lastStamp != driDrawPriv->lastStamp) {
-			radeon->dri.drawable = driDrawPriv;
-
-			radeonSetCliprects(radeon);
-			r300UpdateViewportOffset(radeon->glCtx);
-		}
-
-		_mesa_make_current(radeon->glCtx,
-				    (GLframebuffer *) driDrawPriv->
-				    driverPrivate,
-				    (GLframebuffer *) driReadPriv->
-				    driverPrivate);
-
-		_mesa_update_state(radeon->glCtx);		
-
-		radeonUpdatePageFlipping(radeon);
-	} else {
-		if (RADEON_DEBUG & DEBUG_DRI)
-			fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
-		_mesa_make_current(0, 0, 0);
-	}
-
-	if (RADEON_DEBUG & DEBUG_DRI)
-		fprintf(stderr, "End %s\n", __FUNCTION__);
-	return GL_TRUE;
-}
-
-/* Force the context `c' to be unbound from its buffer.
- */
-GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv)
-{
-	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
-
-	if (RADEON_DEBUG & DEBUG_DRI)
-		fprintf(stderr, "%s ctx %p\n", __FUNCTION__,
-			radeon->glCtx);
-
-	return GL_TRUE;
-}
-
diff --git a/src/mesa/drivers/dri/r300/radeon_context.h b/src/mesa/drivers/dri/r300/radeon_context.h
index 47cbc22a72..250570f6b8 100644
--- a/src/mesa/drivers/dri/r300/radeon_context.h
+++ b/src/mesa/drivers/dri/r300/radeon_context.h
@@ -49,20 +49,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drm.h"
 #include "dri_util.h"
 
-struct radeon_context;
-typedef struct radeon_context radeonContextRec;
-typedef struct radeon_context *radeonContextPtr;
-
-/* Rasterizing fallbacks */
-/* See correponding strings in r200_swtcl.c */
-#define RADEON_FALLBACK_TEXTURE		0x0001
-#define RADEON_FALLBACK_DRAW_BUFFER	0x0002
-#define RADEON_FALLBACK_STENCIL		0x0004
-#define RADEON_FALLBACK_RENDER_MODE	0x0008
-#define RADEON_FALLBACK_BLEND_EQ	0x0010
-#define RADEON_FALLBACK_BLEND_FUNC	0x0020
-#define RADEON_FALLBACK_DISABLE		0x0040
-#define RADEON_FALLBACK_BORDER_MODE	0x0080
+#include "radeon_screen.h"
 
 #if R200_MERGED
 extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
@@ -79,155 +66,11 @@ extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
 /* TCL fallbacks */
 extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
 
-#define RADEON_TCL_FALLBACK_RASTER		0x0001	/* rasterization */
-#define RADEON_TCL_FALLBACK_UNFILLED		0x0002	/* unfilled tris */
-#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE	0x0004	/* twoside tris */
-#define RADEON_TCL_FALLBACK_MATERIAL		0x0008	/* material in vb */
-#define RADEON_TCL_FALLBACK_TEXGEN_0		0x0010	/* texgen, unit 0 */
-#define RADEON_TCL_FALLBACK_TEXGEN_1		0x0020	/* texgen, unit 1 */
-#define RADEON_TCL_FALLBACK_TEXGEN_2		0x0040	/* texgen, unit 2 */
-#define RADEON_TCL_FALLBACK_TEXGEN_3		0x0080	/* texgen, unit 3 */
-#define RADEON_TCL_FALLBACK_TEXGEN_4		0x0100	/* texgen, unit 4 */
-#define RADEON_TCL_FALLBACK_TEXGEN_5		0x0200	/* texgen, unit 5 */
-#define RADEON_TCL_FALLBACK_TCL_DISABLE		0x0400	/* user disable */
-#define RADEON_TCL_FALLBACK_BITMAP		0x0800	/* draw bitmap with points */
-#define RADEON_TCL_FALLBACK_VERTEX_PROGRAM	0x1000	/* vertex program active */
-
 #if R200_MERGED
 #define TCL_FALLBACK( ctx, bit, mode )	radeonTclFallback( ctx, bit, mode )
 #else
 #define TCL_FALLBACK( ctx, bit, mode )	;
 #endif
 
-struct radeon_dri_mirror {
-	__DRIcontextPrivate *context;	/* DRI context */
-	__DRIscreenPrivate *screen;	/* DRI screen */
-	/**
-	 * DRI drawable bound to this context for drawing.
-	 */
-	__DRIdrawablePrivate *drawable;
-
-	/**
-	 * DRI drawable bound to this context for reading.
-	 */
-	__DRIdrawablePrivate *readable;
-
-	drm_context_t hwContext;
-	drm_hw_lock_t *hwLock;
-	int fd;
-	int drmMinor;
-};
-
-/**
- * Derived state for internal purposes.
- */
-struct radeon_scissor_state {
-	drm_clip_rect_t rect;
-	GLboolean enabled;
-
-	GLuint numClipRects;	/* Cliprects active */
-	GLuint numAllocedClipRects;	/* Cliprects available */
-	drm_clip_rect_t *pClipRects;
-};
-
-struct radeon_colorbuffer_state {
-	GLuint clear;
-	GLint drawOffset, drawPitch;
-};
-
-struct radeon_state {
-	struct radeon_colorbuffer_state color;
-	struct radeon_scissor_state scissor;
-};
-
-/**
- * Common per-context variables shared by R200 and R300.
- * R200- and R300-specific code "derive" their own context from this
- * structure.
- */
-struct radeon_context {
-	GLcontext *glCtx;	/* Mesa context */
-	radeonScreenPtr radeonScreen;	/* Screen private DRI data */
-
-	/* Fallback state */
-	GLuint Fallback;
-	GLuint TclFallback;
-
-	/* Page flipping */
-	GLuint doPageFlip;
-
-	/* Drawable, cliprect and scissor information */
-	GLuint numClipRects;	/* Cliprects for the draw buffer */
-	drm_clip_rect_t *pClipRects;
-	unsigned int lastStamp;
-	GLboolean lost_context;
-	drm_radeon_sarea_t *sarea;	/* Private SAREA data */
-
-	/* Mirrors of some DRI state */
-	struct radeon_dri_mirror dri;
-
-	/* Busy waiting */
-	GLuint do_usleeps;
-	GLuint do_irqs;
-	GLuint irqsEmitted;
-	drm_radeon_irq_wait_t iw;
-
-	/* buffer swap */
-	int64_t swap_ust;
-	int64_t swap_missed_ust;
-
-	GLuint swap_count;
-	GLuint swap_missed_count;
-
-	/* Derived state */
-	struct radeon_state state;
-
-	/* Configuration cache
-	 */
-	driOptionCache optionCache;
-};
-
-#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx))
-
-extern void radeonSwapBuffers(__DRIdrawablePrivate * dPriv);
-extern void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-				int x, int y, int w, int h);
-extern GLboolean radeonInitContext(radeonContextPtr radeon,
-				   struct dd_function_table *functions,
-				   const __GLcontextModes * glVisual,
-				   __DRIcontextPrivate * driContextPriv,
-				   void *sharedContextPrivate);
-extern void radeonCleanupContext(radeonContextPtr radeon);
-extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
-				   __DRIdrawablePrivate * driDrawPriv,
-				   __DRIdrawablePrivate * driReadPriv);
-extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
-
-/* ================================================================
- * Debugging:
- */
-#define DO_DEBUG		1
-
-#if DO_DEBUG
-extern int RADEON_DEBUG;
-#else
-#define RADEON_DEBUG		0
-#endif
-
-#define DEBUG_TEXTURE	0x0001
-#define DEBUG_STATE	0x0002
-#define DEBUG_IOCTL	0x0004
-#define DEBUG_PRIMS	0x0008
-#define DEBUG_VERTS	0x0010
-#define DEBUG_FALLBACKS	0x0020
-#define DEBUG_VFMT	0x0040
-#define DEBUG_CODEGEN	0x0080
-#define DEBUG_VERBOSE	0x0100
-#define DEBUG_DRI       0x0200
-#define DEBUG_DMA       0x0400
-#define DEBUG_SANITY    0x0800
-#define DEBUG_SYNC      0x1000
-#define DEBUG_PIXEL     0x2000
-#define DEBUG_MEMORY    0x4000
 
 #endif				/* __RADEON_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.c b/src/mesa/drivers/dri/r300/radeon_ioctl.c
deleted file mode 100644
index f042a7b943..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_ioctl.c
+++ /dev/null
@@ -1,396 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include <sched.h>
-#include <errno.h>
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/macros.h"
-#include "main/context.h"
-#include "swrast/swrast.h"
-#include "r300_context.h"
-#include "radeon_ioctl.h"
-#include "r300_ioctl.h"
-#include "r300_state.h"
-#include "radeon_reg.h"
-
-#include "drirenderbuffer.h"
-#include "vblank.h"
-
-static void radeonWaitForIdle(radeonContextPtr radeon);
-
-/* ================================================================
- * SwapBuffers with client-side throttling
- */
-
-static uint32_t radeonGetLastFrame(radeonContextPtr radeon)
-{
-	drm_radeon_getparam_t gp;
-	int ret;
-	uint32_t frame = 0;
-
-	gp.param = RADEON_PARAM_LAST_FRAME;
-	gp.value = (int *)&frame;
-	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
-				  &gp, sizeof(gp));
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-
-	return frame;
-}
-
-uint32_t radeonGetAge(radeonContextPtr radeon)
-{
-	drm_radeon_getparam_t gp;
-	int ret;
-	uint32_t age = 0;
-
-	gp.param = RADEON_PARAM_LAST_CLEAR;
-	gp.value = (int *)&age;
-	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
-				  &gp, sizeof(gp));
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-
-	return age;
-}
-
-static void radeonEmitIrqLocked(radeonContextPtr radeon)
-{
-	drm_radeon_irq_emit_t ie;
-	int ret;
-
-	ie.irq_seq = &radeon->iw.irq_seq;
-	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT,
-				  &ie, sizeof(ie));
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-}
-
-static void radeonWaitIrq(radeonContextPtr radeon)
-{
-	int ret;
-
-	do {
-		ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT,
-				      &radeon->iw, sizeof(radeon->iw));
-	} while (ret && (errno == EINTR || errno == EBUSY));
-
-	if (ret) {
-		fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__,
-			ret);
-		exit(1);
-	}
-}
-
-static void radeonWaitForFrameCompletion(radeonContextPtr radeon)
-{
-	drm_radeon_sarea_t *sarea = radeon->sarea;
-
-	if (radeon->do_irqs) {
-		if (radeonGetLastFrame(radeon) < sarea->last_frame) {
-			if (!radeon->irqsEmitted) {
-				while (radeonGetLastFrame(radeon) <
-				       sarea->last_frame) ;
-			} else {
-				UNLOCK_HARDWARE(radeon);
-				radeonWaitIrq(radeon);
-				LOCK_HARDWARE(radeon);
-			}
-			radeon->irqsEmitted = 10;
-		}
-
-		if (radeon->irqsEmitted) {
-			radeonEmitIrqLocked(radeon);
-			radeon->irqsEmitted--;
-		}
-	} else {
-		while (radeonGetLastFrame(radeon) < sarea->last_frame) {
-			UNLOCK_HARDWARE(radeon);
-			if (radeon->do_usleeps)
-				DO_USLEEP(1);
-			LOCK_HARDWARE(radeon);
-		}
-	}
-}
-
-/* Copy the back color buffer to the front color buffer.
- */
-void radeonCopyBuffer(__DRIdrawablePrivate * dPriv,
-		      const drm_clip_rect_t	 * rect)
-{
-	radeonContextPtr radeon;
-	GLint nbox, i, ret;
-	GLboolean missed_target;
-	int64_t ust;
-	__DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-	assert(dPriv);
-	assert(dPriv->driContextPriv);
-	assert(dPriv->driContextPriv->driverPrivate);
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL) {
-		fprintf(stderr, "\n%s( %p )\n\n", __FUNCTION__,
-			(void *)radeon->glCtx);
-	}
-
-	r300Flush(radeon->glCtx);
-
-	LOCK_HARDWARE(radeon);
-
-	/* Throttle the frame rate -- only allow one pending swap buffers
-	 * request at a time.
-	 */
-	radeonWaitForFrameCompletion(radeon);
-	if (!rect)
-	{
-	    UNLOCK_HARDWARE(radeon);
-	    driWaitForVBlank(dPriv, &missed_target);
-	    LOCK_HARDWARE(radeon);
-	}
-
-	nbox = dPriv->numClipRects;	/* must be in locked region */
-
-	for (i = 0; i < nbox;) {
-		GLint nr = MIN2(i + RADEON_NR_SAREA_CLIPRECTS, nbox);
-		drm_clip_rect_t *box = dPriv->pClipRects;
-		drm_clip_rect_t *b = radeon->sarea->boxes;
-		GLint n = 0;
-
-		for ( ; i < nr ; i++ ) {
-
-		    *b = box[i];
-
-		    if (rect)
-		    {
-			if (rect->x1 > b->x1)
-			    b->x1 = rect->x1;
-			if (rect->y1 > b->y1)
-			    b->y1 = rect->y1;
-			if (rect->x2 < b->x2)
-			    b->x2 = rect->x2;
-			if (rect->y2 < b->y2)
-			    b->y2 = rect->y2;
-
-			if (b->x1 >= b->x2 || b->y1 >= b->y2)
-			    continue;
-		    }
-
-		    b++;
-		    n++;
-		}
-		radeon->sarea->nbox = n;
-
-		if (!n)
-		   continue;
-
-		ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_SWAP);
-
-		if (ret) {
-			fprintf(stderr, "DRM_RADEON_SWAP: return = %d\n",
-				ret);
-			UNLOCK_HARDWARE(radeon);
-			exit(1);
-		}
-	}
-
-	UNLOCK_HARDWARE(radeon);
-	if (!rect)
-	{
-	    ((r300ContextPtr)radeon)->hw.all_dirty = GL_TRUE;
-
-	    radeon->swap_count++;
-	    (*psp->systemTime->getUST) (&ust);
-	    if (missed_target) {
-		radeon->swap_missed_count++;
-		radeon->swap_missed_ust = ust - radeon->swap_ust;
-	    }
-
-	    radeon->swap_ust = ust;
-
-	    sched_yield();
-	}
-}
-
-void radeonPageFlip(__DRIdrawablePrivate * dPriv)
-{
-	radeonContextPtr radeon;
-	GLint ret;
-	GLboolean missed_target;
-	__DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-	assert(dPriv);
-	assert(dPriv->driContextPriv);
-	assert(dPriv->driContextPriv->driverPrivate);
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-
-	if (RADEON_DEBUG & DEBUG_IOCTL) {
-		fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
-			radeon->sarea->pfCurrentPage);
-	}
-
-	r300Flush(radeon->glCtx);
-	LOCK_HARDWARE(radeon);
-
-	if (!dPriv->numClipRects) {
-		UNLOCK_HARDWARE(radeon);
-		usleep(10000);	/* throttle invisible client 10ms */
-		return;
-	}
-
-	/* Need to do this for the perf box placement:
-	 */
-	{
-		drm_clip_rect_t *box = dPriv->pClipRects;
-		drm_clip_rect_t *b = radeon->sarea->boxes;
-		b[0] = box[0];
-		radeon->sarea->nbox = 1;
-	}
-
-	/* Throttle the frame rate -- only allow a few pending swap buffers
-	 * request at a time.
-	 */
-	radeonWaitForFrameCompletion(radeon);
-	UNLOCK_HARDWARE(radeon);
-	driWaitForVBlank(dPriv, &missed_target);
-	if (missed_target) {
-		radeon->swap_missed_count++;
-		(void)(*psp->systemTime->getUST) (&radeon->swap_missed_ust);
-	}
-	LOCK_HARDWARE(radeon);
-
-	ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_FLIP);
-
-	UNLOCK_HARDWARE(radeon);
-
-	if (ret) {
-		fprintf(stderr, "DRM_RADEON_FLIP: return = %d\n", ret);
-		exit(1);
-	}
-
-	radeon->swap_count++;
-	(void)(*psp->systemTime->getUST) (&radeon->swap_ust);
-
-        driFlipRenderbuffers(radeon->glCtx->WinSysDrawBuffer, 
-                             radeon->sarea->pfCurrentPage);
-
-	if (radeon->sarea->pfCurrentPage == 1) {
-		radeon->state.color.drawOffset = radeon->radeonScreen->frontOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->frontPitch;
-	} else {
-		radeon->state.color.drawOffset = radeon->radeonScreen->backOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->backPitch;
-	}
-
-	if (IS_R300_CLASS(radeon->radeonScreen)) {
-		r300ContextPtr r300 = (r300ContextPtr)radeon;
-		R300_STATECHANGE(r300, cb);
-		r300->hw.cb.cmd[R300_CB_OFFSET] = r300->radeon.state.color.drawOffset + 
-						r300->radeon.radeonScreen->fbLocation;
-		r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.state.color.drawPitch;
-		
-		if (r300->radeon.radeonScreen->cpp == 4)
-			r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888;
-		else
-			r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565;
-	
-		if (r300->radeon.sarea->tiling_enabled)
-			r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE;
-	}
-}
-
-void radeonWaitForIdleLocked(radeonContextPtr radeon)
-{
-	int ret;
-	int i = 0;
-
-	do {
-		ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE);
-		if (ret)
-			DO_USLEEP(1);
-	} while (ret && ++i < 100);
-
-	if (ret < 0) {
-		UNLOCK_HARDWARE(radeon);
-		fprintf(stderr, "Error: R300 timed out... exiting\n");
-		exit(-1);
-	}
-}
-
-static void radeonWaitForIdle(radeonContextPtr radeon)
-{
-	LOCK_HARDWARE(radeon);
-	radeonWaitForIdleLocked(radeon);
-	UNLOCK_HARDWARE(radeon);
-}
-
-void radeonFlush(GLcontext * ctx)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	if (IS_R300_CLASS(radeon->radeonScreen))
-		r300Flush(ctx);
-}
-
-
-/* Make sure all commands have been sent to the hardware and have
- * completed processing.
- */
-void radeonFinish(GLcontext * ctx)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	radeonFlush(ctx);
-
-	if (radeon->do_irqs) {
-		LOCK_HARDWARE(radeon);
-		radeonEmitIrqLocked(radeon);
-		UNLOCK_HARDWARE(radeon);
-		radeonWaitIrq(radeon);
-	} else
-		radeonWaitForIdle(radeon);
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_lock.c b/src/mesa/drivers/dri/r300/radeon_lock.c
deleted file mode 100644
index 4f47afd5dc..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_lock.c
+++ /dev/null
@@ -1,137 +0,0 @@
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Gareth Hughes <gareth@valinux.com>
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Kevin E. Martin <martin@valinux.com>
- */
-
-#include "radeon_lock.h"
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
-#include "r300_context.h"
-#include "r300_state.h"
-
-#include "main/framebuffer.h"
-
-#include "drirenderbuffer.h"
-
-#if DEBUG_LOCKING
-char *prevLockFile = NULL;
-int prevLockLine = 0;
-#endif
-
-/* Turn on/off page flipping according to the flags in the sarea:
- */
-void radeonUpdatePageFlipping(radeonContextPtr rmesa)
-{
-	int use_back;
-
-	rmesa->doPageFlip = rmesa->sarea->pfState;
-	if (rmesa->glCtx->WinSysDrawBuffer) {
-		driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-				     rmesa->sarea->pfCurrentPage);
-		r300UpdateDrawBuffer(rmesa->glCtx);
-	}
-
-	use_back = rmesa->glCtx->DrawBuffer ?
-	    (rmesa->glCtx->DrawBuffer->_ColorDrawBufferIndexes[0] ==
-	     BUFFER_BACK_LEFT) : 1;
-	use_back ^= (rmesa->sarea->pfCurrentPage == 1);
-
-	if (use_back) {
-		rmesa->state.color.drawOffset =
-		    rmesa->radeonScreen->backOffset;
-		rmesa->state.color.drawPitch = rmesa->radeonScreen->backPitch;
-	} else {
-		rmesa->state.color.drawOffset =
-		    rmesa->radeonScreen->frontOffset;
-		rmesa->state.color.drawPitch =
-		    rmesa->radeonScreen->frontPitch;
-	}
-}
-
-/* Update the hardware state.  This is called if another context has
- * grabbed the hardware lock, which includes the X server.  This
- * function also updates the driver's window state after the X server
- * moves, resizes or restacks a window -- the change will be reflected
- * in the drawable position and clip rects.  Since the X server grabs
- * the hardware lock when it changes the window state, this routine will
- * automatically be called after such a change.
- */
-void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
-{
-	__DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-	__DRIdrawablePrivate *const readable = rmesa->dri.readable;
-	__DRIscreenPrivate *sPriv = rmesa->dri.screen;
-	drm_radeon_sarea_t *sarea = rmesa->sarea;
-	r300ContextPtr r300 = (r300ContextPtr) rmesa;
-
-	assert(drawable != NULL);
-
-	drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags);
-
-	/* The window might have moved, so we might need to get new clip
-	 * rects.
-	 *
-	 * NOTE: This releases and regrabs the hw lock to allow the X server
-	 * to respond to the DRI protocol request for new drawable info.
-	 * Since the hardware state depends on having the latest drawable
-	 * clip rects, all state checking must be done _after_ this call.
-	 */
-	DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable);
-	if (drawable != readable) {
-		DRI_VALIDATE_DRAWABLE_INFO(sPriv, readable);
-	}
-
-	if (rmesa->lastStamp != drawable->lastStamp) {
-		radeonUpdatePageFlipping(rmesa);
-		radeonSetCliprects(rmesa);
-		r300UpdateViewportOffset(rmesa->glCtx);
-		driUpdateFramebufferSize(rmesa->glCtx, drawable);
-	}
-
-	if (sarea->ctx_owner != rmesa->dri.hwContext) {
-		int i;
-
-		sarea->ctx_owner = rmesa->dri.hwContext;
-		for (i = 0; i < r300->nr_heaps; i++) {
-			DRI_AGE_TEXTURES(r300->texture_heaps[i]);
-		}
-	}
-
-	rmesa->lost_context = GL_TRUE;
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index 4a2e1cba40..840c9733b1 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -46,6 +46,7 @@ static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint fil
 	switch(file) {
 	case PROGRAM_TEMPORARY: return &s->Temps[index];
 	case PROGRAM_OUTPUT: return &s->Outputs[index];
+	case PROGRAM_ADDRESS: return &s->Address;
 	default: return 0;
 	}
 }
@@ -56,7 +57,7 @@ static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint fil
  *
  * @note Works correctly only for X, Y, Z, W swizzles, not for constant swizzles.
  */
-static struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg)
+struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg)
 {
 	struct prog_src_register tmp = srcreg;
 	int i;
@@ -114,47 +115,19 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
 		deswz_source = sourced;
 	}
 
-	struct register_state *regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index);
+	struct register_state *regstate;
+
+	if (inst->SrcReg[src].RelAddr)
+		regstate = get_reg_state(s, PROGRAM_ADDRESS, 0);
+	else
+		regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index);
+
 	if (regstate)
 		regstate->Sourced |= deswz_source & 0xf;
 
 	return inst;
 }
 
-
-static void rewrite_depth_out(struct prog_instruction *inst)
-{
-	if (inst->DstReg.WriteMask & WRITEMASK_Z) {
-		inst->DstReg.WriteMask = WRITEMASK_W;
-	} else {
-		inst->DstReg.WriteMask = 0;
-		return;
-	}
-
-	switch (inst->Opcode) {
-	case OPCODE_FRC:
-	case OPCODE_MOV:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		break;
-	case OPCODE_ADD:
-	case OPCODE_MAX:
-	case OPCODE_MIN:
-	case OPCODE_MUL:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
-		break;
-	case OPCODE_CMP:
-	case OPCODE_MAD:
-		inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]);
-		inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]);
-		inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]);
-		break;
-	default:
-		// Scalar instructions needn't be reswizzled
-		break;
-	}
-}
-
 static void unalias_srcregs(struct prog_instruction *inst, GLuint oldindex, GLuint newindex)
 {
 	int nsrc = _mesa_num_inst_src_regs(inst->Opcode);
@@ -189,11 +162,6 @@ static void process_instruction(struct nqssadce_state* s)
 		return;
 
 	if (inst->Opcode != OPCODE_KIL) {
-		if (s->Descr->RewriteDepthOut) {
-			if (inst->DstReg.File == PROGRAM_OUTPUT && inst->DstReg.Index == FRAG_RESULT_DEPTH)
-				rewrite_depth_out(inst);
-		}
-
 		struct register_state *regstate = get_reg_state(s, inst->DstReg.File, inst->DstReg.Index);
 		if (!regstate) {
 			_mesa_problem(s->Ctx, "NqssaDce: bad destination register (%i[%i])\n",
@@ -217,6 +185,7 @@ static void process_instruction(struct nqssadce_state* s)
 	 * might change the instruction stream under us, so we have
 	 * to be careful with the inst pointer. */
 	switch (inst->Opcode) {
+	case OPCODE_ARL:
 	case OPCODE_DDX:
 	case OPCODE_DDY:
 	case OPCODE_FRC:
@@ -227,6 +196,8 @@ static void process_instruction(struct nqssadce_state* s)
 	case OPCODE_MAX:
 	case OPCODE_MIN:
 	case OPCODE_MUL:
+	case OPCODE_SGE:
+	case OPCODE_SLT:
 		inst = track_used_srcreg(s, inst, 0, inst->DstReg.WriteMask);
 		inst = track_used_srcreg(s, inst, 1, inst->DstReg.WriteMask);
 		break;
@@ -258,12 +229,51 @@ static void process_instruction(struct nqssadce_state* s)
 	case OPCODE_TXP:
 		inst = track_used_srcreg(s, inst, 0, 0xf);
 		break;
+	case OPCODE_DST:
+		inst = track_used_srcreg(s, inst, 0, 0x6);
+		inst = track_used_srcreg(s, inst, 1, 0xa);
+		break;
+	case OPCODE_EXP:
+	case OPCODE_LOG:
+	case OPCODE_POW:
+		inst = track_used_srcreg(s, inst, 0, 0x3);
+		break;
+	case OPCODE_LIT:
+		inst = track_used_srcreg(s, inst, 0, 0xb);
+		break;
 	default:
 		_mesa_problem(s->Ctx, "NqssaDce: Unknown opcode %d\n", inst->Opcode);
 		return;
 	}
 }
 
+static void calculateInputsOutputs(struct gl_program *p)
+{
+	struct prog_instruction *inst;
+	GLuint InputsRead, OutputsWritten;
+
+	inst = p->Instructions;
+	InputsRead = 0;
+	OutputsWritten = 0;
+	while (inst->Opcode != OPCODE_END)
+	{
+		int i, num_src_regs;
+
+		num_src_regs = _mesa_num_inst_src_regs(inst->Opcode);
+		for (i = 0; i < num_src_regs; ++i) {
+			if (inst->SrcReg[i].File == PROGRAM_INPUT)
+				InputsRead |= 1 << inst->SrcReg[i].Index;
+		}
+
+		if (inst->DstReg.File == PROGRAM_OUTPUT)
+			OutputsWritten |= 1 << inst->DstReg.Index;
+
+		++inst;
+	}
+
+	p->InputsRead = InputsRead;
+	p->OutputsWritten = OutputsWritten;
+}
 
 void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce_descr* descr)
 {
@@ -280,4 +290,6 @@ void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce
 		s.IP--;
 		process_instruction(&s);
 	}
+
+	calculateInputsOutputs(p);
 }
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.h b/src/mesa/drivers/dri/r300/radeon_nqssadce.h
index a4f94abcb6..8626f21c25 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.h
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.h
@@ -58,6 +58,7 @@ struct nqssadce_state {
 	 */
 	struct register_state Temps[MAX_PROGRAM_TEMPS];
 	struct register_state Outputs[VERT_RESULT_MAX];
+	struct register_state Address;
 };
 
 
@@ -83,14 +84,10 @@ struct radeon_nqssadce_descr {
 	 */
 	void (*BuildSwizzle)(struct nqssadce_state*, struct prog_dst_register dst, struct prog_src_register src);
 
-	/**
-	 * Rewrite instructions that write to DEPR.z to write to DEPR.w
-	 * instead (rewriting is done *before* the WriteMask test).
-	 */
-	GLboolean RewriteDepthOut;
 	void *Data;
 };
 
 void radeonNqssaDce(GLcontext *ctx, struct gl_program *p, struct radeon_nqssadce_descr* descr);
+struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg);
 
 #endif /* __RADEON_PROGRAM_NQSSADCE_H_ */
diff --git a/src/mesa/drivers/dri/r300/radeon_program.h b/src/mesa/drivers/dri/r300/radeon_program.h
index b411f69bc8..88474d43a2 100644
--- a/src/mesa/drivers/dri/r300/radeon_program.h
+++ b/src/mesa/drivers/dri/r300/radeon_program.h
@@ -52,6 +52,38 @@ enum {
 #define SWIZZLE_0000 MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO)
 #define SWIZZLE_1111 MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE)
 
+static inline GLuint get_swz(GLuint swz, GLuint idx)
+{
+	if (idx & 0x4)
+		return idx;
+	return GET_SWZ(swz, idx);
+}
+
+static inline GLuint combine_swizzles4(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz_z, GLuint swz_w)
+{
+	GLuint ret = 0;
+
+	ret |= get_swz(src, swz_x);
+	ret |= get_swz(src, swz_y) << 3;
+	ret |= get_swz(src, swz_z) << 6;
+	ret |= get_swz(src, swz_w) << 9;
+
+	return ret;
+}
+
+static inline GLuint combine_swizzles(GLuint src, GLuint swz)
+{
+	GLuint ret = 0;
+
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_X));
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Y)) << 3;
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Z)) << 6;
+	ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_W)) << 9;
+
+	return ret;
+}
+
+
 /**
  * Transformation context that is passed to local transformations.
  *
diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.c b/src/mesa/drivers/dri/r300/radeon_program_pair.c
index 2e21f7bf66..d6fb474cf2 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_pair.c
+++ b/src/mesa/drivers/dri/r300/radeon_program_pair.c
@@ -35,7 +35,7 @@
 
 #include "radeon_program_pair.h"
 
-#include "radeon_context.h"
+#include "radeon_common.h"
 
 #include "shader/prog_print.h"
 
@@ -47,7 +47,6 @@
 
 struct pair_state_instruction {
 	GLuint IsTex:1; /**< Is a texture instruction */
-	GLuint IsOutput:1; /**< Is output instruction */
 	GLuint NeedRGB:1; /**< Needs the RGB ALU */
 	GLuint NeedAlpha:1; /**< Needs the Alpha ALU */
 	GLuint IsTranscendent:1; /**< Is a special transcendent instruction */
@@ -124,7 +123,6 @@ struct pair_state {
 	GLboolean Debug;
 	GLboolean Verbose;
 	void *UserData;
-	GLubyte NumKillInsts;
 
 	/**
 	 * Translate Mesa registers to hardware registers
@@ -151,11 +149,6 @@ struct pair_state {
 	struct pair_state_instruction *ReadyTEX;
 
 	/**
-	 * Linked list of deferred instructions
-	 */
-	struct pair_state_instruction *DeferredInsts;
-
-	/**
 	 * Pool of @ref reg_value structures for fast allocation.
 	 */
 	struct reg_value *ValuePool;
@@ -238,9 +231,7 @@ static void instruction_ready(struct pair_state *s, int ip)
 	if (s->Verbose)
 		_mesa_printf("instruction_ready(%i)\n", ip);
 
-	if (s->NumKillInsts > 0 && pairinst->IsOutput)
-		add_pairinst_to_list(&s->DeferredInsts, pairinst);
-	else if (pairinst->IsTex)
+	if (pairinst->IsTex)
 		add_pairinst_to_list(&s->ReadyTEX, pairinst);
 	else if (!pairinst->NeedAlpha)
 		add_pairinst_to_list(&s->ReadyRGB, pairinst);
@@ -348,8 +339,6 @@ static void classify_instruction(struct pair_state *s,
 		error("Unknown opcode %d\n", inst->Opcode);
 		break;
 	}
-
-	pairinst->IsOutput = (inst->DstReg.File == PROGRAM_OUTPUT);
 }
 
 
@@ -613,16 +602,14 @@ static void emit_all_tex(struct pair_state *s)
 		struct prog_instruction *inst = s->Program->Instructions + ip;
 		commit_instruction(s, ip);
 
-		if (inst->Opcode == OPCODE_KIL)
-			--s->NumKillInsts;
-		else
+		if (inst->Opcode != OPCODE_KIL)
 			inst->DstReg.Index = get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index);
-
 		inst->SrcReg[0].Index = get_hw_reg(s, inst->SrcReg[0].File, inst->SrcReg[0].Index);
 
 		if (s->Debug) {
 			_mesa_printf("   ");
 			_mesa_print_instruction(inst);
+			fflush(stdout);
 		}
 		s->Error = s->Error || !s->Handler->EmitTex(s->UserData, inst);
 	}
@@ -875,17 +862,6 @@ static void emit_alu(struct pair_state *s)
 	s->Error = s->Error || !s->Handler->EmitPaired(s->UserData, &pair);
 }
 
-static GLubyte countKillInsts(struct gl_program *prog)
-{
-	GLubyte i, count = 0;
-
-	for (i = 0; i < prog->NumInstructions; ++i) {
-		if (prog->Instructions[i].Opcode == OPCODE_KIL)
-			++count;
-	}
-
-	return count;
-}
 
 GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
 	const struct radeon_pair_handler* handler, void *userdata)
@@ -894,12 +870,11 @@ GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
 
 	_mesa_bzero(&s, sizeof(s));
 	s.Ctx = ctx;
-	s.Program = program;
+	s.Program = _mesa_clone_program(ctx, program);
 	s.Handler = handler;
 	s.UserData = userdata;
 	s.Debug = (RADEON_DEBUG & DEBUG_PIXEL) ? GL_TRUE : GL_FALSE;
 	s.Verbose = GL_FALSE && s.Debug;
-	s.NumKillInsts = countKillInsts(program);
 
 	s.Instructions = (struct pair_state_instruction*)_mesa_calloc(
 		sizeof(struct pair_state_instruction)*s.Program->NumInstructions);
@@ -918,21 +893,6 @@ GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
 		if (s.ReadyTEX)
 			emit_all_tex(&s);
 
-		if (!s.NumKillInsts) {
-			struct pair_state_instruction *pairinst = s.DeferredInsts;
-			while (pairinst) {
-				if (!pairinst->NeedAlpha)
-					add_pairinst_to_list(&s.ReadyRGB, pairinst);
-				else if (!pairinst->NeedRGB)
-					add_pairinst_to_list(&s.ReadyAlpha, pairinst);
-				else
-					add_pairinst_to_list(&s.ReadyFullALU, pairinst);
-
-				pairinst = pairinst->NextReady;
-			}
-			s.DeferredInsts = NULL;
-		}
-
 		while(s.ReadyFullALU || s.ReadyRGB || s.ReadyAlpha)
 			emit_alu(&s);
 	}
@@ -944,6 +904,8 @@ GLboolean radeonPairProgram(GLcontext *ctx, struct gl_program *program,
 	_mesa_free(s.ValuePool);
 	_mesa_free(s.ReaderPool);
 
+	_mesa_reference_program(ctx, &s.Program, NULL);
+
 	return !s.Error;
 }
 
diff --git a/src/mesa/drivers/dri/r300/radeon_span.c b/src/mesa/drivers/dri/r300/radeon_span.c
deleted file mode 100644
index 16f9fb99e6..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_span.c
+++ /dev/null
@@ -1,349 +0,0 @@
-/**************************************************************************
-
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Kevin E. Martin <martin@valinux.com>
- *   Gareth Hughes <gareth@valinux.com>
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#include "main/glheader.h"
-#include "swrast/swrast.h"
-
-#include "r300_state.h"
-#include "radeon_ioctl.h"
-#include "r300_ioctl.h"
-#include "radeon_span.h"
-
-#include "drirenderbuffer.h"
-
-#define DBG 0
-
-/*
- * Note that all information needed to access pixels in a renderbuffer
- * should be obtained through the gl_renderbuffer parameter, not per-context
- * information.
- */
-#define LOCAL_VARS						\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;		\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
-   GLubyte *buf = (GLubyte *) drb->flippedData			\
-      + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;	\
-   GLuint p;							\
-   (void) p;
-
-#define LOCAL_DEPTH_VARS				\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;	\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;			\
-   GLuint xo = dPriv->x;				\
-   GLuint yo = dPriv->y;				\
-   GLubyte *buf = (GLubyte *) drb->Base.Data;
-
-#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
-
-#define Y_FLIP(Y) (bottom - (Y))
-
-#define HW_LOCK()
-
-#define HW_UNLOCK()
-
-/* ================================================================
- * Color buffer
- */
-
-/* 16 bit, RGB565 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    radeon##x##_RGB565
-#define TAG2(x,y) radeon##x##_RGB565##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
-#include "spantmp2.h"
-
-/* 32 bit, ARGB8888 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    radeon##x##_ARGB8888
-#define TAG2(x,y) radeon##x##_ARGB8888##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
-#include "spantmp2.h"
-
-/* ================================================================
- * Depth buffer
- */
-
-/* The Radeon family has depth tiling on all the time, so we have to convert
- * the x,y coordinates into the memory bus address (mba) in the same
- * manner as the engine.  In each case, the linear block address (ba)
- * is calculated, and then wired with x and y to produce the final
- * memory address.
- * The chip will do address translation on its own if the surface registers
- * are set up correctly. It is not quite enough to get it working with hyperz
- * too...
- */
-
-static GLuint radeon_mba_z32(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 4 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0..1] = 0           */
-
-#ifdef COMPILE_R300
-		ba = (y / 8) * (pitch / 8) + (x / 8);
-#else
-		ba = (y / 16) * (pitch / 16) + (x / 16);
-#endif
-
-		address |= (x & 0x7) << 2;	/* a[2..4] = x[0..2]     */
-		address |= (y & 0x3) << 5;	/* a[5..6] = y[0..1]     */
-		address |= (((x & 0x10) >> 2) ^ (y & 0x4)) << 5;	/* a[7]    = x[4] ^ y[2] */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= (((x & 0x8) << 1) ^ (y & 0x10)) << 7;	/* a[11]   = x[3] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
-static INLINE GLuint
-radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 2 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0]    = 0           */
-
-		ba = (y / 16) * (pitch / 32) + (x / 32);
-
-		address |= (x & 0x7) << 1;	/* a[1..3] = x[0..2]     */
-		address |= (y & 0x7) << 4;	/* a[4..6] = y[0..2]     */
-		address |= (x & 0x8) << 4;	/* a[7]    = x[3]        */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= ((x & 0x10) ^ (y & 0x10)) << 7;	/* a[11]   = x[4] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
-/* 16-bit depth buffer functions
- */
-#define VALUE_TYPE GLushort
-
-#define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )) = d;
-
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo ));
-
-#define TAG(x) radeon##x##_z16
-#include "depthtmp.h"
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- *
- * Careful: It looks like the R300 uses ZZZS byte order while the R200
- * uses SZZZ for 24 bit depth, 8 bit stencil mode.
- */
-#define VALUE_TYPE GLuint
-
-#ifdef COMPILE_R300
-#define WRITE_DEPTH( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0x000000ff;							\
-   tmp |= ((d << 8) & 0xffffff00);					\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#else
-#define WRITE_DEPTH( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xff000000;							\
-   tmp |= ((d) & 0x00ffffff);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#endif
-
-#ifdef COMPILE_R300
-#define READ_DEPTH( d, _x, _y )						\
-  do { \
-    d = (*(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,		\
-					 _y + yo )) & 0xffffff00) >> 8; \
-  }while(0)
-#else
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,			\
-					 _y + yo )) & 0x00ffffff;
-#endif
-
-#define TAG(x) radeon##x##_z24_s8
-#include "depthtmp.h"
-
-/* ================================================================
- * Stencil buffer
- */
-
-/* 24 bit depth, 8 bit stencil depthbuffer functions
- */
-#ifdef COMPILE_R300
-#define WRITE_STENCIL( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0xffffff00;							\
-   tmp |= (d) & 0xff;							\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#else
-#define WRITE_STENCIL( _x, _y, d )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   tmp &= 0x00ffffff;							\
-   tmp |= (((d) & 0xff) << 24);						\
-   *(GLuint *)(buf + offset) = tmp;					\
-} while (0)
-#endif
-
-#ifdef COMPILE_R300
-#define READ_STENCIL( d, _x, _y )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   d = tmp & 0x000000ff;						\
-} while (0)
-#else
-#define READ_STENCIL( d, _x, _y )					\
-do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
-   d = (tmp & 0xff000000) >> 24;					\
-} while (0)
-#endif
-
-#define TAG(x) radeon##x##_z24_s8
-#include "stenciltmp.h"
-
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
-
-static void radeonSpanRenderStart(GLcontext * ctx)
-{
-	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-#ifdef COMPILE_R300
-	r300ContextPtr r300 = (r300ContextPtr) rmesa;
-	R300_FIREVERTICES(r300);
-#else
-	RADEON_FIREVERTICES(rmesa);
-#endif
-	LOCK_HARDWARE(rmesa);
-	radeonWaitForIdleLocked(rmesa);
-
-	/* Read the first pixel in the frame buffer.  This should
-	 * be a noop, right?  In fact without this conform fails as reading
-	 * from the framebuffer sometimes produces old results -- the
-	 * on-card read cache gets mixed up and doesn't notice that the
-	 * framebuffer has been updated.
-	 *
-	 * Note that we should probably be reading some otherwise unused
-	 * region of VRAM, otherwise we might get incorrect results when
-	 * reading pixels from the top left of the screen.
-	 *
-	 * I found this problem on an R420 with glean's texCube test.
-	 * Note that the R200 span code also *writes* the first pixel in the
-	 * framebuffer, but I've found this to be unnecessary.
-	 *  -- Nicolai Hähnle, June 2008
-	 */
-	{
-		int p;
-		driRenderbuffer *drb =
-			(driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0];
-		volatile int *buf =
-			(volatile int *)(rmesa->dri.screen->pFB + drb->offset);
-		p = *buf;
-	}
-}
-
-static void radeonSpanRenderFinish(GLcontext * ctx)
-{
-	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-	_swrast_flush(ctx);
-	UNLOCK_HARDWARE(rmesa);
-}
-
-void radeonInitSpanFuncs(GLcontext * ctx)
-{
-	struct swrast_device_driver *swdd =
-	    _swrast_GetDeviceDriverReference(ctx);
-	swdd->SpanRenderStart = radeonSpanRenderStart;
-	swdd->SpanRenderFinish = radeonSpanRenderFinish;
-}
-
-/**
- * Plug in the Get/Put routines for the given driRenderbuffer.
- */
-void radeonSetSpanFunctions(driRenderbuffer * drb, const GLvisual * vis)
-{
-	if (drb->Base.InternalFormat == GL_RGBA) {
-		if (vis->redBits == 5 && vis->greenBits == 6
-		    && vis->blueBits == 5) {
-			radeonInitPointers_RGB565(&drb->Base);
-		} else {
-			radeonInitPointers_ARGB8888(&drb->Base);
-		}
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
-		radeonInitDepthPointers_z16(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
-		radeonInitDepthPointers_z24_s8(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
-		radeonInitStencilPointers_z24_s8(&drb->Base);
-	}
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_state.c b/src/mesa/drivers/dri/r300/radeon_state.c
deleted file mode 100644
index c401da6c54..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_state.c
+++ /dev/null
@@ -1,244 +0,0 @@
-/**************************************************************************
-
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/api_arrayelt.h"
-#include "main/enums.h"
-#include "main/framebuffer.h"
-#include "main/colormac.h"
-#include "main/light.h"
-
-#include "swrast/swrast.h"
-#include "vbo/vbo.h"
-#include "tnl/tnl.h"
-#include "tnl/t_pipeline.h"
-#include "swrast_setup/swrast_setup.h"
-
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
-#include "r300_ioctl.h"
-
-
-/* =============================================================
- * Scissoring
- */
-
-static GLboolean intersect_rect(drm_clip_rect_t * out,
-				drm_clip_rect_t * a, drm_clip_rect_t * b)
-{
-	*out = *a;
-	if (b->x1 > out->x1)
-		out->x1 = b->x1;
-	if (b->y1 > out->y1)
-		out->y1 = b->y1;
-	if (b->x2 < out->x2)
-		out->x2 = b->x2;
-	if (b->y2 < out->y2)
-		out->y2 = b->y2;
-	if (out->x1 >= out->x2)
-		return GL_FALSE;
-	if (out->y1 >= out->y2)
-		return GL_FALSE;
-	return GL_TRUE;
-}
-
-void radeonRecalcScissorRects(radeonContextPtr radeon)
-{
-	drm_clip_rect_t *out;
-	int i;
-
-	/* Grow cliprect store?
-	 */
-	if (radeon->state.scissor.numAllocedClipRects < radeon->numClipRects) {
-		while (radeon->state.scissor.numAllocedClipRects <
-		       radeon->numClipRects) {
-			radeon->state.scissor.numAllocedClipRects += 1;	/* zero case */
-			radeon->state.scissor.numAllocedClipRects *= 2;
-		}
-
-		if (radeon->state.scissor.pClipRects)
-			FREE(radeon->state.scissor.pClipRects);
-
-		radeon->state.scissor.pClipRects =
-		    MALLOC(radeon->state.scissor.numAllocedClipRects *
-			   sizeof(drm_clip_rect_t));
-
-		if (radeon->state.scissor.pClipRects == NULL) {
-			radeon->state.scissor.numAllocedClipRects = 0;
-			return;
-		}
-	}
-
-	out = radeon->state.scissor.pClipRects;
-	radeon->state.scissor.numClipRects = 0;
-
-	for (i = 0; i < radeon->numClipRects; i++) {
-		if (intersect_rect(out,
-				   &radeon->pClipRects[i],
-				   &radeon->state.scissor.rect)) {
-			radeon->state.scissor.numClipRects++;
-			out++;
-		}
-	}
-}
-
-void radeonUpdateScissor(GLcontext* ctx)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	if (radeon->dri.drawable) {
-		__DRIdrawablePrivate *dPriv = radeon->dri.drawable;
-		int x1 = dPriv->x + ctx->Scissor.X;
-		int y1 = dPriv->y + dPriv->h - (ctx->Scissor.Y + ctx->Scissor.Height);
-
-		radeon->state.scissor.rect.x1 = x1;
-		radeon->state.scissor.rect.y1 = y1;
-		radeon->state.scissor.rect.x2 = x1 + ctx->Scissor.Width;
-		radeon->state.scissor.rect.y2 = y1 + ctx->Scissor.Height;
-
-		radeonRecalcScissorRects(radeon);
-	}
-}
-
-static void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h)
-{
-	if (ctx->Scissor.Enabled) {
-		/* We don't pipeline cliprect changes */
-		r300Flush(ctx);
-		radeonUpdateScissor(ctx);
-	}
-}
-
-
-/**
- * Update cliprects and scissors.
- */
-void radeonSetCliprects(radeonContextPtr radeon)
-{
-	__DRIdrawablePrivate *const drawable = radeon->dri.drawable;
-	__DRIdrawablePrivate *const readable = radeon->dri.readable;
-	GLframebuffer *const draw_fb = (GLframebuffer*)drawable->driverPrivate;
-	GLframebuffer *const read_fb = (GLframebuffer*)readable->driverPrivate;
-
-	if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-		/* Can't ignore 2d windows if we are page flipping. */
-		if (drawable->numBackClipRects == 0 || radeon->doPageFlip ||
-		    radeon->sarea->pfCurrentPage == 1) {
-			radeon->numClipRects = drawable->numClipRects;
-			radeon->pClipRects = drawable->pClipRects;
-		} else {
-			radeon->numClipRects = drawable->numBackClipRects;
-			radeon->pClipRects = drawable->pBackClipRects;
-		}
-	} else {
-		/* front buffer (or none, or multiple buffers */
-		radeon->numClipRects = drawable->numClipRects;
-		radeon->pClipRects = drawable->pClipRects;
-	}
-
-	if ((draw_fb->Width != drawable->w) ||
-	    (draw_fb->Height != drawable->h)) {
-		_mesa_resize_framebuffer(radeon->glCtx, draw_fb,
-					 drawable->w, drawable->h);
-		draw_fb->Initialized = GL_TRUE;
-	}
-
-	if (drawable != readable) {
-		if ((read_fb->Width != readable->w) ||
-		    (read_fb->Height != readable->h)) {
-			_mesa_resize_framebuffer(radeon->glCtx, read_fb,
-						 readable->w, readable->h);
-			read_fb->Initialized = GL_TRUE;
-		}
-	}
-
-	if (radeon->state.scissor.enabled)
-		radeonRecalcScissorRects(radeon);
-
-	radeon->lastStamp = drawable->lastStamp;
-}
-
-
-/**
- * Handle common enable bits.
- * Called as a fallback by r200Enable/r300Enable.
- */
-void radeonEnable(GLcontext* ctx, GLenum cap, GLboolean state)
-{
-	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
-
-	switch(cap) {
-	case GL_SCISSOR_TEST:
-		/* We don't pipeline cliprect & scissor changes */
-		r300Flush(ctx);
-
-		radeon->state.scissor.enabled = state;
-		radeonUpdateScissor(ctx);
-		break;
-
-	default:
-		return;
-	}
-}
-
-
-/**
- * Initialize default state.
- * This function is called once at context init time from
- * r200InitState/r300InitState
- */
-void radeonInitState(radeonContextPtr radeon)
-{
-	radeon->Fallback = 0;
-
-	if (radeon->glCtx->Visual.doubleBufferMode && radeon->sarea->pfCurrentPage == 0) {
-		radeon->state.color.drawOffset = radeon->radeonScreen->backOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->backPitch;
-	} else {
-		radeon->state.color.drawOffset = radeon->radeonScreen->frontOffset;
-		radeon->state.color.drawPitch = radeon->radeonScreen->frontPitch;
-	}
-}
-
-
-/**
- * Initialize common state functions.
- * Called by r200InitStateFuncs/r300InitStateFuncs
- */
-void radeonInitStateFuncs(struct dd_function_table *functions)
-{
-	functions->Scissor = radeonScissor;
-}
diff --git a/src/mesa/drivers/dri/r300/radeon_state.h b/src/mesa/drivers/dri/r300/radeon_state.h
deleted file mode 100644
index 821cb40c7e..0000000000
--- a/src/mesa/drivers/dri/r300/radeon_state.h
+++ /dev/null
@@ -1,43 +0,0 @@
-/*
-Copyright (C) 2004 Nicolai Haehnle.  All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Nicolai Haehnle <prefect_@gmx.net>
- */
-
-#ifndef __RADEON_STATE_H__
-#define __RADEON_STATE_H__
-
-extern void radeonRecalcScissorRects(radeonContextPtr radeon);
-extern void radeonSetCliprects(radeonContextPtr radeon);
-extern void radeonUpdateScissor(GLcontext* ctx);
-
-extern void radeonEnable(GLcontext* ctx, GLenum cap, GLboolean state);
-
-extern void radeonInitState(radeonContextPtr radeon);
-extern void radeonInitStateFuncs(struct dd_function_table* functions);
-
-#endif
diff --git a/src/mesa/drivers/dri/r600/Lindent b/src/mesa/drivers/dri/r600/Lindent
new file mode 100755
index 0000000000..7d8d8896e3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/Lindent
@@ -0,0 +1,2 @@
+#!/bin/sh
+indent -npro -kr -i8 -ts8 -sob -l80 -ss -ncs "$@"
diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile
new file mode 100644
index 0000000000..2f225f7cb5
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/Makefile
@@ -0,0 +1,121 @@
+# src/mesa/drivers/dri/r300/Makefile
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+CFLAGS += $(RADEON_CFLAGS)
+
+LIBNAME = r600_dri.so
+
+MINIGLX_SOURCES = server/radeon_dri.c
+
+ifeq ($(USING_EGL), 1)
+EGL_SOURCES = server/radeon_egl.c
+endif
+
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
+COMMON_SOURCES = \
+	../../common/driverfuncs.c \
+	../common/mm.c \
+	../common/utils.c \
+	../common/texmem.c \
+	../common/vblank.c \
+	../common/xmlconfig.c \
+	../common/dri_util.c
+
+RADEON_COMMON_SOURCES = \
+	radeon_texture.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_dma.c \
+	radeon_lock.c \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_mipmap_tree.c \
+	radeon_span.c \
+	radeon_fbo.c
+
+DRIVER_SOURCES = \
+		 radeon_screen.c \
+		 r600_context.c \
+		 r600_cmdbuf.c \
+		 r600_emit.c       \
+		 r700_assembler.c  \
+		 r700_fragprog.c \
+		 r700_vertprog.c \
+		 r700_shader.c \
+		 r700_shaderinst.c \
+		 r700_ioctl.c \
+		 r700_oglprog.c \
+		 r700_chip.c     \
+		 r700_state.c    \
+		 r700_clear.c    \
+		 r700_render.c   \
+		 r600_tex.c      \
+		 r600_texstate.c      \
+		 r700_debug.c    \
+		 $(RADEON_COMMON_SOURCES) \
+		 $(EGL_SOURCES) \
+		 $(CS_SOURCES)
+
+C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
+
+DRIVER_DEFINES = -DCOMPILE_R600 -DR200_MERGED=0 \
+	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R600 \
+#	-DRADEON_BO_TRACK \
+	-Wall
+
+SYMLINKS = \
+	server/radeon_dri.c \
+	server/radeon_dri.h \
+	server/radeon.h \
+	server/radeon_macros.h \
+	server/radeon_reg.h \
+	server/radeon_egl.c
+
+COMMON_SYMLINKS = \
+	radeon_chipset.h \
+	radeon_screen.c \
+	radeon_screen.h \
+	radeon_span.h \
+	radeon_span.c \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_bo_legacy.h \
+	radeon_cs_legacy.h \
+	radeon_bocs_wrapper.h \
+	radeon_lock.c \
+	radeon_lock.h \
+	radeon_common.c \
+	radeon_common.h \
+	radeon_common_context.c \
+	radeon_common_context.h \
+	radeon_cmdbuf.h \
+	radeon_dma.c \
+	radeon_dma.h \
+	radeon_mipmap_tree.c \
+	radeon_mipmap_tree.h \
+	radeon_texture.c \
+	radeon_texture.h \
+	radeon_fbo.c \
+	$(CS_SOURCES)
+
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
+
+##### TARGETS #####
+
+include ../Makefile.template
+
+server:
+	mkdir -p server
+
+$(SYMLINKS): server
+	@[ -e $@ ] || ln -sf ../../radeon/$@ server/
+
+$(COMMON_SYMLINKS):
+	@[ -e $@ ] || ln -sf ../radeon/$@ ./
+
+symlinks: $(SYMLINKS) $(COMMON_SYMLINKS)
diff --git a/src/mesa/drivers/dri/r600/defaultendian.h b/src/mesa/drivers/dri/r600/defaultendian.h
new file mode 100644
index 0000000000..32caf32cd2
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/defaultendian.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef _DEFINEENDIAN_H_
+#define _DEFINEENDIAN_H_
+
+//We have to choose a reg bits orientation if there is no compile flag for it.
+#if defined(LITTLEENDIAN_CPU)
+#elif defined(BIGENDIAN_CPU)
+#else
+#define LITTLEENDIAN_CPU
+#endif
+
+#endif //_DEFINEENDIAN_H_
diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c
new file mode 100644
index 0000000000..74fec02584
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c
@@ -0,0 +1,555 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * Mostly coppied from \radeon\radeon_cs_legacy.c
+ */
+
+#include <errno.h>
+
+#include "main/glheader.h"
+#include "main/state.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "main/simple_list.h"
+#include "swrast/swrast.h"
+
+#include "drm.h"
+#include "radeon_drm.h"
+
+#include "r600_context.h"
+#include "radeon_reg.h"
+#include "r600_cmdbuf.h"
+#include "r600_emit.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_mipmap_tree.h"
+#include "radeon_reg.h"
+
+
+
+static struct radeon_cs * r600_cs_create(struct radeon_cs_manager *csm,
+                                   uint32_t ndw)
+{
+    struct radeon_cs *cs;
+
+    cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs));
+    if (cs == NULL) {
+        return NULL;
+    }
+    cs->csm = csm;
+    cs->ndw = (ndw + 0x3FF) & (~0x3FF);
+    cs->packets = (uint32_t*)malloc(4*cs->ndw);
+    if (cs->packets == NULL) {
+        free(cs);
+        return NULL;
+    }
+    cs->relocs_total_size = 0;
+    return cs;
+}
+
+int r600_cs_write_reloc(struct radeon_cs *cs,
+                        struct radeon_bo *bo,
+                        uint32_t read_domain,
+                        uint32_t write_domain,
+                        uint32_t flags,
+                        offset_modifiers* poffset_mod)
+{
+    struct r600_cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct r600_cs_reloc_legacy *)cs->relocs;
+    /* check domains */
+    if ((read_domain && write_domain) || (!read_domain && !write_domain)) {
+        /* in one CS a bo can only be in read or write domain but not
+         * in read & write domain at the same sime
+         */
+        return -EINVAL;
+    }
+    if (read_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    if (write_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    /* check if bo is already referenced */
+    for(i = 0; i < cs->crelocs; i++) {
+        uint32_t *indices;
+        uint32_t *reloc_indices;
+
+        if (relocs[i].base.bo->handle == bo->handle) {
+            /* Check domains must be in read or write. As we check already
+             * checked that in argument one of the read or write domain was
+             * set we only need to check that if previous reloc as the read
+             * domain set then the read_domain should also be set for this
+             * new relocation.
+             */
+            if (relocs[i].base.read_domain && !read_domain) {
+                return -EINVAL;
+            }
+            if (relocs[i].base.write_domain && !write_domain) {
+                return -EINVAL;
+            }
+            relocs[i].base.read_domain |= read_domain;
+            relocs[i].base.write_domain |= write_domain;
+            /* save indice */
+            relocs[i].cindices++;
+            indices = (uint32_t*)realloc(relocs[i].indices,
+                                         relocs[i].cindices * 4);
+            reloc_indices = (uint32_t*)realloc(relocs[i].reloc_indices,
+                                               relocs[i].cindices * 4);
+            if ( (indices == NULL) || (reloc_indices == NULL) ) {
+                relocs[i].cindices -= 1;
+                return -ENOMEM;
+            }
+            relocs[i].indices = indices;
+            relocs[i].reloc_indices = reloc_indices;
+            relocs[i].indices[relocs[i].cindices - 1] = cs->cdw - 1;
+            relocs[i].reloc_indices[relocs[i].cindices - 1] = cs->section_cdw;
+            cs->section_ndw += 2;
+            cs->section_cdw += 2;
+
+            relocs[i].offset_mod.shift     = poffset_mod->shift;
+            relocs[i].offset_mod.shiftbits = poffset_mod->shiftbits;
+            relocs[i].offset_mod.mask      = poffset_mod->mask;
+
+            return 0;
+        }
+    }
+    /* add bo to reloc */
+    relocs = (struct r600_cs_reloc_legacy*)
+             realloc(cs->relocs,
+                     sizeof(struct r600_cs_reloc_legacy) * (cs->crelocs + 1));
+    if (relocs == NULL) {
+        return -ENOMEM;
+    }
+    cs->relocs = relocs;
+    relocs[cs->crelocs].base.bo = bo;
+    relocs[cs->crelocs].base.read_domain = read_domain;
+    relocs[cs->crelocs].base.write_domain = write_domain;
+    relocs[cs->crelocs].base.flags = flags;
+    relocs[cs->crelocs].indices = (uint32_t*)malloc(4);
+    relocs[cs->crelocs].reloc_indices = (uint32_t*)malloc(4);
+    if ( (relocs[cs->crelocs].indices == NULL) || (relocs[cs->crelocs].reloc_indices == NULL) )
+    {
+        return -ENOMEM;
+    }
+    relocs[cs->crelocs].offset_mod.shift     = poffset_mod->shift;
+    relocs[cs->crelocs].offset_mod.shiftbits = poffset_mod->shiftbits;
+    relocs[cs->crelocs].offset_mod.mask      = poffset_mod->mask;
+
+    relocs[cs->crelocs].indices[0] = cs->cdw - 1;
+    relocs[cs->crelocs].reloc_indices[0] = cs->section_cdw;
+    cs->section_ndw += 2;
+    cs->section_cdw += 2;
+    relocs[cs->crelocs].cindices = 1;
+    cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo);
+    cs->crelocs++;
+
+    radeon_bo_ref(bo);
+
+    return 0;
+}
+
+static int r600_cs_begin(struct radeon_cs *cs,
+                    uint32_t ndw,
+                    const char *file,
+                    const char *func,
+                    int line)
+{
+    if (cs->section) {
+        fprintf(stderr, "CS already in a section(%s,%s,%d)\n",
+                cs->section_file, cs->section_func, cs->section_line);
+        fprintf(stderr, "CS can't start section(%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+
+    if (cs->cdw + ndw + 32 > cs->ndw) { /* Left 32 DWORD (8 offset+pitch) spare room for reloc indices */
+        uint32_t tmp, *ptr;
+	int num = (ndw > 0x3FF) ? ndw : 0x3FF;
+
+        tmp = (cs->cdw + 1 + num) & (~num);
+        ptr = (uint32_t*)realloc(cs->packets, 4 * tmp);
+        if (ptr == NULL) {
+            return -ENOMEM;
+        }
+        cs->packets = ptr;
+        cs->ndw = tmp;
+    }
+
+    cs->section = 1;
+    cs->section_ndw = 0; 
+    cs->section_cdw = cs->cdw + ndw; /* start of reloc indices. */
+    cs->section_file = file;
+    cs->section_func = func;
+    cs->section_line = line;
+
+    return 0;
+}
+
+static int r600_cs_end(struct radeon_cs *cs,
+                  const char *file,
+                  const char *func,
+                  int line)
+
+{
+    if (!cs->section) {
+        fprintf(stderr, "CS no section to end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    cs->section = 0;
+
+    if ( (cs->section_ndw + cs->cdw) != cs->section_cdw ) 
+    {
+        fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n",
+                cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw);
+        fprintf(stderr, "cs->section_ndw = %d, cs->cdw = %d, cs->section_cdw = %d \n",
+                cs->section_ndw, cs->cdw, cs->section_cdw);
+        fprintf(stderr, "CS section end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+
+    cs->cdw = cs->section_cdw;
+    return 0;
+}
+
+static int r600_cs_process_relocs(struct radeon_cs *cs, 
+                                  uint32_t * reloc_chunk,
+                                  uint32_t * length_dw_reloc_chunk) 
+{
+    struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm;
+    struct r600_cs_reloc_legacy *relocs;
+    int i, j, r;
+
+    uint32_t offset_dw = 0;
+
+    csm = (struct r600_cs_manager_legacy*)cs->csm;
+    relocs = (struct r600_cs_reloc_legacy *)cs->relocs;
+restart:
+    for (i = 0; i < cs->crelocs; i++) 
+    {
+        for (j = 0; j < relocs[i].cindices; j++) 
+        {
+            uint32_t soffset, eoffset, asicoffset;
+
+            r = radeon_bo_legacy_validate(relocs[i].base.bo,
+                                           &soffset, &eoffset);
+	        if (r == -EAGAIN)
+            {
+	             goto restart;
+            }
+            if (r) 
+            {
+                fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+                        relocs[i].base.bo, soffset, eoffset);
+                return r;
+            }
+            asicoffset = soffset;
+            if (asicoffset >= eoffset) 
+            {
+	      /*                radeon_bo_debug(relocs[i].base.bo, 12); */
+                fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+                        relocs[i].base.bo, soffset, eoffset);
+                fprintf(stderr, "above end: %p 0x%08X 0x%08X\n",
+                        relocs[i].base.bo,
+                        cs->packets[relocs[i].indices[j]],
+                        eoffset);
+                exit(0);
+                return -EINVAL;
+            }
+            /* apply offset operator */
+            switch (relocs[i].offset_mod.shift)
+            {
+            case NO_SHIFT:
+                asicoffset = asicoffset & relocs[i].offset_mod.mask;
+                break;
+            case LEFT_SHIFT:
+                asicoffset = (asicoffset << relocs[i].offset_mod.shiftbits) & relocs[i].offset_mod.mask;
+                break;
+            case RIGHT_SHIFT:
+                asicoffset = (asicoffset >> relocs[i].offset_mod.shiftbits) & relocs[i].offset_mod.mask;
+                break;
+            default:
+                break;
+            };              
+
+            /* pkt3 nop header in ib chunk */
+            cs->packets[relocs[i].reloc_indices[j]] = 0xC0001000;
+
+            /* reloc index in ib chunk */
+            cs->packets[relocs[i].reloc_indices[j] + 1] = offset_dw;
+            
+            /* asic offset in reloc chunk */ /* see alex drm r600_nomm_relocate */
+            reloc_chunk[offset_dw] = asicoffset;
+            reloc_chunk[offset_dw + 3] = 0;
+
+            offset_dw += 4;
+        }
+    }
+
+    *length_dw_reloc_chunk = offset_dw;
+
+    return 0;
+}
+
+static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */
+{
+    struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm;
+    struct r600_cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct r600_cs_reloc_legacy *)cs->relocs;
+    for (i = 0; i < cs->crelocs; i++) {
+        radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age);
+        radeon_bo_unref(relocs[i].base.bo);
+    }
+    return 0;
+}
+
+static void dump_cmdbuf(struct radeon_cs *cs)
+{
+	int i;
+	fprintf(stderr,"--start--\n");
+	for (i = 0; i < cs->cdw; i++){
+		fprintf(stderr,"0x%08x\n", cs->packets[i]);
+	}
+	fprintf(stderr,"--end--\n");
+
+}
+
+static int r600_cs_emit(struct radeon_cs *cs)
+{
+    struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm;
+    struct drm_radeon_cs       cs_cmd;
+    struct drm_radeon_cs_chunk cs_chunk[2];
+    drm_radeon_cmd_buffer_t cmd; 
+    /* drm_r300_cmd_header_t age; */
+    uint32_t length_dw_reloc_chunk;
+    uint64_t ull;
+    uint64_t chunk_ptrs[2];
+    uint32_t reloc_chunk[128]; 
+    int r;
+    int retry = 0;
+
+    /* TODO : put chip level things here if need. */
+    /* csm->ctx->vtbl.emit_cs_header(cs, csm->ctx); */
+
+    BATCH_LOCALS(csm->ctx);
+    drm_radeon_getparam_t gp;
+    uint32_t              current_scratchx_age;
+
+    gp.param = RADEON_PARAM_LAST_CLEAR;
+    gp.value = (int *)&current_scratchx_age;
+    r = drmCommandWriteRead(cs->csm->fd, 
+                            DRM_RADEON_GETPARAM,
+                            &gp, 
+                            sizeof(gp));
+    if (r) 
+    {
+        fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, r);
+        exit(1);
+    }
+
+    csm->pending_age = 0;
+    csm->pending_count = 1;
+
+    current_scratchx_age++;
+    csm->pending_age = current_scratchx_age;
+
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+    R600_OUT_BATCH((SCRATCH_REG2 - R600_SET_CONFIG_REG_OFFSET) >> 2);
+    R600_OUT_BATCH(current_scratchx_age);
+    END_BATCH();
+    COMMIT_BATCH();
+
+    //TODO ioctl to get back cs id assigned in drm
+    //csm->pending_age = cs_id_back;
+    
+    r = r600_cs_process_relocs(cs, &(reloc_chunk[0]), &length_dw_reloc_chunk);
+    if (r) {
+        return 0;
+    }
+      
+    /* raw ib chunk */
+    cs_chunk[0].chunk_id   = RADEON_CHUNK_ID_IB;
+    cs_chunk[0].length_dw  = cs->cdw;
+    cs_chunk[0].chunk_data = (unsigned long)(cs->packets);
+
+    /* reloc chaunk */
+    cs_chunk[1].chunk_id   = RADEON_CHUNK_ID_RELOCS;
+    cs_chunk[1].length_dw  = length_dw_reloc_chunk;
+    cs_chunk[1].chunk_data = (unsigned long)&(reloc_chunk[0]);
+
+    chunk_ptrs[0] = (uint64_t)(unsigned long)&(cs_chunk[0]);
+    chunk_ptrs[1] = (uint64_t)(unsigned long)&(cs_chunk[1]);
+
+    cs_cmd.num_chunks = 2;
+    /* cs_cmd.cs_id      = 0; */
+    cs_cmd.chunks     = (uint64_t)(unsigned long)chunk_ptrs;
+
+    //dump_cmdbuf(cs);
+
+    do 
+    {
+        r = drmCommandWriteRead(cs->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd));
+        retry++;
+    } while (r == -EAGAIN && retry < 1000);
+
+    if (r) {
+        return r;
+    }
+
+    r600_cs_set_age(cs);
+
+    cs->csm->read_used = 0;
+    cs->csm->vram_write_used = 0;
+    cs->csm->gart_write_used = 0;
+
+    return 0;
+}
+
+static void inline r600_cs_free_reloc(void *relocs_p, int crelocs)
+{
+    struct r600_cs_reloc_legacy *relocs = relocs_p;
+    int i;
+    if (!relocs_p)
+      return;
+    for (i = 0; i < crelocs; i++)
+    {
+        free(relocs[i].indices);
+        free(relocs[i].reloc_indices);
+    }
+}
+
+static int r600_cs_destroy(struct radeon_cs *cs)
+{
+    r600_cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    free(cs->packets);
+    free(cs);
+    return 0;
+}
+
+static int r600_cs_erase(struct radeon_cs *cs)
+{
+    r600_cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    cs->relocs_total_size = 0;
+    cs->relocs = NULL;
+    cs->crelocs = 0;
+    cs->cdw = 0;
+    cs->section = 0;
+    return 0;
+}
+
+static int r600_cs_need_flush(struct radeon_cs *cs)
+{
+    /* this function used to flush when the BO usage got to
+     * a certain size, now the higher levels handle this better */
+    return 0;
+}
+
+static void r600_cs_print(struct radeon_cs *cs, FILE *file)
+{
+}
+
+static struct radeon_cs_funcs  r600_cs_funcs = {
+    r600_cs_create,
+    r600_cs_write_reloc,
+    r600_cs_begin,
+    r600_cs_end,
+    r600_cs_emit,
+    r600_cs_destroy,
+    r600_cs_erase,
+    r600_cs_need_flush,
+    r600_cs_print
+};
+
+struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx)
+{
+    struct r600_cs_manager_legacy *csm;
+
+    csm = (struct r600_cs_manager_legacy*)
+          calloc(1, sizeof(struct r600_cs_manager_legacy));
+    if (csm == NULL) {
+        return NULL;
+    }
+    csm->base.funcs = &r600_cs_funcs;
+    csm->base.fd = ctx->dri.fd;
+    csm->ctx = ctx;
+    csm->pending_age = 1;
+    return (struct radeon_cs_manager*)csm;
+}
+
+void r600InitCmdBuf(context_t *r600) /* from rcommonInitCmdBuf */
+{
+    radeonContextPtr rmesa = &r600->radeon;
+	
+    GLuint size;
+	/* Initialize command buffer */
+	size = 256 * driQueryOptioni(&rmesa->optionCache,
+				     "command_buffer_size");
+	if (size < 2 * rmesa->hw.max_state_size) {
+		size = 2 * rmesa->hw.max_state_size + 65535;
+	}
+	if (size > 64 * 256)
+		size = 64 * 256;
+
+	if (rmesa->radeonScreen->kernel_mm) {
+		int fd = rmesa->radeonScreen->driScreen->fd;
+		rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd);
+	} else {
+		rmesa->cmdbuf.csm = r600_radeon_cs_manager_legacy_ctor(rmesa);
+	}
+	if (rmesa->cmdbuf.csm == NULL) {
+		/* FIXME: fatal error */
+		return;
+	}
+	rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size);
+	assert(rmesa->cmdbuf.cs != NULL);
+	rmesa->cmdbuf.size = size;
+	
+	if (!rmesa->radeonScreen->kernel_mm) {
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]);
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size);
+	} else {
+		struct drm_radeon_gem_info mminfo;
+
+		if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo)))
+		{
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_visible);
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size);
+		}
+	}
+}
+
diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h
new file mode 100644
index 0000000000..bd1ed7fdff
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h
@@ -0,0 +1,229 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+#ifndef __R600_CMDBUF_H__
+#define __R600_CMDBUF_H__
+
+#include "r600_context.h"
+#include "r600_emit.h"
+
+#define RADEON_CP_PACKET3_NOP                       0xC0001000
+#define RADEON_CP_PACKET3_NEXT_CHAR                 0xC0001900
+#define RADEON_CP_PACKET3_PLY_NEXTSCAN              0xC0001D00
+#define RADEON_CP_PACKET3_SET_SCISSORS              0xC0001E00
+#define RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM     0xC0002300
+#define RADEON_CP_PACKET3_LOAD_MICROCODE            0xC0002400
+#define RADEON_CP_PACKET3_WAIT_FOR_IDLE             0xC0002600
+#define RADEON_CP_PACKET3_3D_DRAW_VBUF              0xC0002800
+#define RADEON_CP_PACKET3_3D_DRAW_IMMD              0xC0002900
+#define RADEON_CP_PACKET3_3D_DRAW_INDX              0xC0002A00
+#define RADEON_CP_PACKET3_LOAD_PALETTE              0xC0002C00
+#define RADEON_CP_PACKET3_3D_LOAD_VBPNTR            0xC0002F00
+#define RADEON_CP_PACKET3_CNTL_PAINT                0xC0009100
+#define RADEON_CP_PACKET3_CNTL_BITBLT               0xC0009200
+#define RADEON_CP_PACKET3_CNTL_SMALLTEXT            0xC0009300
+#define RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT         0xC0009400
+#define RADEON_CP_PACKET3_CNTL_POLYLINE             0xC0009500
+#define RADEON_CP_PACKET3_CNTL_POLYSCANLINES        0xC0009800
+#define RADEON_CP_PACKET3_CNTL_PAINT_MULTI          0xC0009A00
+#define RADEON_CP_PACKET3_CNTL_BITBLT_MULTI         0xC0009B00
+#define RADEON_CP_PACKET3_CNTL_TRANS_BITBLT         0xC0009C00
+
+/* r6xx/r7xx packet 3 type offsets */
+#define R600_SET_CONFIG_REG_OFFSET                  0x00008000
+#define R600_SET_CONFIG_REG_END                     0x0000ac00
+#define R600_SET_CONTEXT_REG_OFFSET                 0x00028000
+#define R600_SET_CONTEXT_REG_END                    0x00029000
+#define R600_SET_ALU_CONST_OFFSET                   0x00030000
+#define R600_SET_ALU_CONST_END                      0x00032000
+#define R600_SET_RESOURCE_OFFSET                    0x00038000
+#define R600_SET_RESOURCE_END                       0x0003c000
+#define R600_SET_SAMPLER_OFFSET                     0x0003c000
+#define R600_SET_SAMPLER_END                        0x0003cff0
+#define R600_SET_CTL_CONST_OFFSET                   0x0003cff0
+#define R600_SET_CTL_CONST_END                      0x0003e200
+#define R600_SET_LOOP_CONST_OFFSET                  0x0003e200
+#define R600_SET_LOOP_CONST_END                     0x0003e380
+#define R600_SET_BOOL_CONST_OFFSET                  0x0003e380
+#define R600_SET_BOOL_CONST_END                     0x00040000
+
+/* r6xx/r7xx packet 3 types */
+#define R600_IT_INDIRECT_BUFFER_END               0x00001700
+#define R600_IT_SET_PREDICATION                   0x00002000
+#define R600_IT_REG_RMW                           0x00002100
+#define R600_IT_COND_EXEC                         0x00002200
+#define R600_IT_PRED_EXEC                         0x00002300
+#define R600_IT_START_3D_CMDBUF                   0x00002400
+#define R600_IT_DRAW_INDEX_2                      0x00002700
+#define R600_IT_CONTEXT_CONTROL                   0x00002800
+#define R600_IT_DRAW_INDEX_IMMD_BE                0x00002900
+#define R600_IT_INDEX_TYPE                        0x00002A00
+#define R600_IT_DRAW_INDEX                        0x00002B00
+#define R600_IT_DRAW_INDEX_AUTO                   0x00002D00
+#define R600_IT_DRAW_INDEX_IMMD                   0x00002E00
+#define R600_IT_NUM_INSTANCES                     0x00002F00
+#define R600_IT_STRMOUT_BUFFER_UPDATE             0x00003400
+#define R600_IT_INDIRECT_BUFFER_MP                0x00003800
+#define R600_IT_MEM_SEMAPHORE                     0x00003900
+#define R600_IT_MPEG_INDEX                        0x00003A00
+#define R600_IT_WAIT_REG_MEM                      0x00003C00
+#define R600_IT_MEM_WRITE                         0x00003D00
+#define R600_IT_INDIRECT_BUFFER                   0x00003200
+#define R600_IT_CP_INTERRUPT                      0x00004000
+#define R600_IT_SURFACE_SYNC                      0x00004300
+#define R600_IT_ME_INITIALIZE                     0x00004400
+#define R600_IT_COND_WRITE                        0x00004500
+#define R600_IT_EVENT_WRITE                       0x00004600
+#define R600_IT_EVENT_WRITE_EOP                   0x00004700
+#define R600_IT_ONE_REG_WRITE                     0x00005700
+#define R600_IT_SET_CONFIG_REG                    0x00006800
+#define R600_IT_SET_CONTEXT_REG                   0x00006900
+#define R600_IT_SET_ALU_CONST                     0x00006A00
+#define R600_IT_SET_BOOL_CONST                    0x00006B00
+#define R600_IT_SET_LOOP_CONST                    0x00006C00
+#define R600_IT_SET_RESOURCE                      0x00006D00
+#define R600_IT_SET_SAMPLER                       0x00006E00
+#define R600_IT_SET_CTL_CONST                     0x00006F00
+#define R600_IT_SURFACE_BASE_UPDATE               0x00007300
+
+struct r600_cs_manager_legacy
+{
+    struct radeon_cs_manager    base;
+    struct radeon_context       *ctx;
+    /* hack for scratch stuff */
+    uint32_t                    pending_age;
+    uint32_t                    pending_count;
+};
+
+struct r600_cs_reloc_legacy {
+    struct radeon_cs_reloc  base;
+    uint32_t                cindices;
+    uint32_t                *indices;
+    uint32_t                *reloc_indices;
+    struct offset_modifiers offset_mod;
+};
+
+extern int r600_cs_write_reloc(struct radeon_cs *cs,
+                        struct radeon_bo *bo,
+                        uint32_t read_domain,
+                        uint32_t write_domain,
+                        uint32_t flags,
+                        offset_modifiers* poffset_mod);
+
+static inline void r600_cs_write_dword(struct radeon_cs *cs, uint32_t dword)
+{
+    cs->packets[cs->cdw++] = dword;
+}
+
+struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx);
+
+/**
+ * Write one dword to the command buffer.
+ */
+#define R600_OUT_BATCH(data) \
+	do { \
+        r600_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);\
+	} while(0)
+
+/**
+ * Write n dwords from ptr to the command buffer.
+ */
+#define R600_OUT_BATCH_TABLE(ptr,n) \
+	do { \
+		int _i; \
+        for (_i=0; _i < n; _i++) {\
+            r600_cs_write_dword(b_l_rmesa->cmdbuf.cs, ptr[_i]);\
+        }\
+	} while(0)
+
+/**
+ * Write a relocated dword to the command buffer.
+ */
+#define R600_OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags, offset_mod) 	\
+	do { 							\
+        if (0 && offset) {					\
+            fprintf(stderr, "(%s:%s:%d) offset : %d\n",		\
+            __FILE__, __FUNCTION__, __LINE__, offset);		\
+        }							\
+        r600_cs_write_dword(b_l_rmesa->cmdbuf.cs, offset);	\
+        r600_cs_write_reloc(b_l_rmesa->cmdbuf.cs, 		\
+                              bo, rd, wd, flags, offset_mod);		\
+	} while(0)
+
+/* R600/R700 */
+#define R600_OUT_BATCH_REGS(reg, num)					\
+do {								\
+	if ((reg) >= R600_SET_CONFIG_REG_OFFSET && (reg) < R600_SET_CONFIG_REG_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_CONFIG_REG_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_CONTEXT_REG_OFFSET && (reg) < R600_SET_CONTEXT_REG_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONTEXT_REG, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_CONTEXT_REG_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_ALU_CONST_OFFSET && (reg) < R600_SET_ALU_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_ALU_CONST_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_RESOURCE_OFFSET && (reg) < R600_SET_RESOURCE_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_RESOURCE_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_SAMPLER_OFFSET && (reg) < R600_SET_SAMPLER_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_SAMPLER_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_CTL_CONST_OFFSET && (reg) < R600_SET_CTL_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_CTL_CONST_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_LOOP_CONST_OFFSET && (reg) < R600_SET_LOOP_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_LOOP_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_LOOP_CONST_OFFSET) >> 2);	\
+	} else if ((reg) >= R600_SET_BOOL_CONST_OFFSET && (reg) < R600_SET_BOOL_CONST_END) { \
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_BOOL_CONST, (num)));	\
+		R600_OUT_BATCH(((reg) - R600_SET_BOOL_CONST_OFFSET) >> 2);	\
+	} else {							\
+		R600_OUT_BATCH(CP_PACKET0((reg), (num))); \
+	}								\
+} while (0)
+
+/** Single register write to command buffer; requires 3 dwords for most things. */
+#define R600_OUT_BATCH_REGVAL(reg, val)		\
+	R600_OUT_BATCH_REGS((reg), 1);		\
+	R600_OUT_BATCH((val))
+
+/** Continuous register range write to command buffer; requires 1 dword,
+ * expects count dwords afterwards for register contents. */
+#define R600_OUT_BATCH_REGSEQ(reg, count)	\
+	R600_OUT_BATCH_REGS((reg), (count))
+
+extern void r600InitCmdBuf(context_t *r600);
+
+#endif				/* __R600_CMDBUF_H__ */
diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c
new file mode 100644
index 0000000000..78bad8726b
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_context.c
@@ -0,0 +1,413 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ *
+ * \author Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+#include "main/glheader.h"
+#include "main/api_arrayelt.h"
+#include "main/context.h"
+#include "main/simple_list.h"
+#include "main/imports.h"
+#include "main/matrix.h"
+#include "main/extensions.h"
+#include "main/state.h"
+#include "main/bufferobj.h"
+#include "main/texobj.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "vbo/vbo.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "tnl/t_vp_build.h"
+
+#include "drivers/common/driverfuncs.h"
+
+#include "r600_context.h"
+#include "radeon_context.h"
+#include "radeon_span.h"
+#include "r600_cmdbuf.h"
+#include "r600_emit.h"
+#include "radeon_bocs_wrapper.h"
+
+#include "r700_state.h"
+#include "r700_ioctl.h"
+
+
+#include "vblank.h"
+#include "utils.h"
+#include "xmlpool.h"		/* for symbolic values of enum-type options */
+
+/* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
+int future_hw_tcl_on = 1;
+int hw_tcl_on = 1;
+
+#define need_GL_VERSION_2_0
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_vertex_program
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_gpu_program_parameters
+#define need_GL_EXT_secondary_color
+#define need_GL_EXT_stencil_two_side
+#define need_GL_ATI_separate_stencil
+#define need_GL_NV_vertex_program
+
+#include "extension_helper.h"
+
+extern const struct tnl_pipeline_stage *r700_pipeline[];
+
+const struct dri_extension card_extensions[] = {
+  /* *INDENT-OFF* */
+  {"GL_ARB_depth_texture",		NULL},
+  {"GL_ARB_fragment_program",		NULL},
+  {"GL_ARB_multitexture",		NULL},
+  {"GL_ARB_point_parameters",		GL_ARB_point_parameters_functions},
+  {"GL_ARB_shadow",			NULL},
+  {"GL_ARB_shadow_ambient",		NULL},
+  {"GL_ARB_texture_border_clamp",	NULL},
+  {"GL_ARB_texture_cube_map",		NULL},
+  {"GL_ARB_texture_env_add",		NULL},
+  {"GL_ARB_texture_env_combine",	NULL},
+  {"GL_ARB_texture_env_crossbar",	NULL},
+  {"GL_ARB_texture_env_dot3",		NULL},
+  {"GL_ARB_texture_mirrored_repeat",	NULL},
+  {"GL_ARB_vertex_program",		GL_ARB_vertex_program_functions},
+  {"GL_EXT_blend_equation_separate",	GL_EXT_blend_equation_separate_functions},
+  {"GL_EXT_blend_func_separate",	GL_EXT_blend_func_separate_functions},
+  {"GL_EXT_blend_minmax",		GL_EXT_blend_minmax_functions},
+  {"GL_EXT_blend_subtract",		NULL},
+  {"GL_EXT_packed_depth_stencil",	NULL},
+  {"GL_EXT_fog_coord",			GL_EXT_fog_coord_functions },
+  {"GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions},
+  {"GL_EXT_secondary_color", 		GL_EXT_secondary_color_functions},
+  {"GL_EXT_shadow_funcs",		NULL},
+  {"GL_EXT_stencil_two_side",		GL_EXT_stencil_two_side_functions},
+  {"GL_EXT_stencil_wrap",		NULL},
+  {"GL_EXT_texture_edge_clamp",		NULL},
+  {"GL_EXT_texture_env_combine", 	NULL},
+  {"GL_EXT_texture_env_dot3", 		NULL},
+  {"GL_EXT_texture_filter_anisotropic",	NULL},
+  {"GL_EXT_texture_lod_bias",		NULL},
+  {"GL_EXT_texture_mirror_clamp",	NULL},
+  {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
+  {"GL_ATI_texture_env_combine3",	NULL},
+  {"GL_ATI_texture_mirror_once",	NULL},
+  {"GL_MESA_pack_invert",		NULL},
+  {"GL_MESA_ycbcr_texture",		NULL},
+  {"GL_MESAX_texture_float",		NULL},
+  {"GL_NV_blend_square",		NULL},
+  {"GL_NV_vertex_program",		GL_NV_vertex_program_functions},
+  {"GL_SGIS_generate_mipmap",		NULL},
+  {NULL,				NULL}
+  /* *INDENT-ON* */
+};
+
+
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
+/**
+ * The GL 2.0 functions are needed to make display lists work with
+ * functions added by GL_ATI_separate_stencil.
+ */
+const struct dri_extension gl_20_extension[] = {
+  {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
+};
+
+
+static void r600RunPipeline(GLcontext * ctx)
+{
+    _mesa_lock_context_textures(ctx);
+
+    if (ctx->NewState)
+        _mesa_update_state_locked(ctx);
+    
+    _tnl_run_pipeline(ctx);
+    _mesa_unlock_context_textures(ctx);
+}
+
+static void r600_get_lock(radeonContextPtr rmesa)
+{
+	drm_radeon_sarea_t *sarea = rmesa->sarea;
+
+	if (sarea->ctx_owner != rmesa->dri.hwContext) {
+		sarea->ctx_owner = rmesa->dri.hwContext;
+		if (!rmesa->radeonScreen->kernel_mm)
+			radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
+	}
+}		  
+
+static void r600_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+    /* please flush pipe do all pending work */
+    /* to be enabled */
+}
+
+static void r600_vtbl_pre_emit_atoms(radeonContextPtr radeon)
+{
+	/* to be enabled */
+}
+
+static void r600_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	if (mode)
+		context->radeon.Fallback |= bit;
+	else
+		context->radeon.Fallback &= ~bit;
+}
+
+static void r600_init_vtbl(radeonContextPtr radeon)
+{
+	radeon->vtbl.get_lock = r600_get_lock;
+	radeon->vtbl.update_viewport_offset = r700UpdateViewportOffset;
+	radeon->vtbl.emit_cs_header = r600_vtbl_emit_cs_header;
+	radeon->vtbl.swtcl_flush = NULL;
+	radeon->vtbl.pre_emit_atoms = r600_vtbl_pre_emit_atoms;
+	radeon->vtbl.fallback = r600_fallback;
+}
+
+/* Create the device specific rendering context.
+ */
+GLboolean r600CreateContext(const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate)
+{
+	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+	struct dd_function_table functions;
+	context_t *r600;
+	GLcontext *ctx;
+	int tcl_mode;
+
+	assert(glVisual);
+	assert(driContextPriv);
+	assert(screen);
+
+	/* Allocate the R600 context */
+	r600 = (context_t*) CALLOC(sizeof(*r600));
+	if (!r600)
+		return GL_FALSE;
+
+	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
+		hw_tcl_on = future_hw_tcl_on = 0;
+
+	r600_init_vtbl(&r600->radeon);
+	/* Parse configuration files.
+	 * Do this here so that initialMaxAnisotropy is set before we create
+	 * the default textures.
+	 */
+	driParseConfigFiles(&r600->radeon.optionCache, &screen->optionCache,
+			    screen->driScreen->myNum, "r600");
+
+	r600->radeon.initialMaxAnisotropy = driQueryOptionf(&r600->radeon.optionCache,
+						     "def_max_anisotropy");
+
+	/* Init default driver functions then plug in our R600-specific functions
+	 * (the texture functions are especially important)
+	 */
+	_mesa_init_driver_functions(&functions);
+
+	r700InitChipObject(r600);  /* let the eag... */
+
+	r700InitStateFuncs(&functions);
+	r600InitTextureFuncs(&functions);
+	r700InitShaderFuncs(&functions);
+	r700InitIoctlFuncs(&functions);
+
+	if (!radeonInitContext(&r600->radeon, &functions,
+			       glVisual, driContextPriv,
+			       sharedContextPrivate)) {
+		FREE(r600);
+		return GL_FALSE;
+	}
+
+	/* Init r600 context data */
+	/* Set the maximum texture size small enough that we can guarentee that
+	 * all texture units can bind a maximal texture and have them both in
+	 * texturable memory at once.
+	 */
+
+	ctx = r600->radeon.glCtx;
+
+	ctx->Const.MaxTextureImageUnits =
+	    driQueryOptioni(&r600->radeon.optionCache, "texture_image_units");
+	ctx->Const.MaxTextureCoordUnits =
+	    driQueryOptioni(&r600->radeon.optionCache, "texture_coord_units");
+	ctx->Const.MaxTextureUnits =
+	    MIN2(ctx->Const.MaxTextureImageUnits,
+		 ctx->Const.MaxTextureCoordUnits);
+	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
+	ctx->Const.MaxTextureLodBias = 16.0;
+
+	ctx->Const.MaxTextureLevels = 13;
+	ctx->Const.MaxTextureRectSize = 4096;
+
+	ctx->Const.MinPointSize   = 0x0001 / 8.0;
+	ctx->Const.MinPointSizeAA = 0x0001 / 8.0;
+	ctx->Const.MaxPointSize   = 0xffff / 8.0;
+	ctx->Const.MaxPointSizeAA = 0xffff / 8.0;
+
+	ctx->Const.MinLineWidth   = 0x0001 / 8.0;
+	ctx->Const.MinLineWidthAA = 0x0001 / 8.0;
+	ctx->Const.MaxLineWidth   = 0xffff / 8.0;
+	ctx->Const.MaxLineWidthAA = 0xffff / 8.0;
+
+	/* Needs further modifications */
+#if 0
+	ctx->Const.MaxArrayLockSize =
+	    ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
+#endif
+
+	ctx->Const.MaxDrawBuffers = 1;
+
+	/* Initialize the software rasterizer and helper modules.
+	 */
+	_swrast_CreateContext(ctx);
+	_vbo_CreateContext(ctx);
+	_tnl_CreateContext(ctx);
+	_swsetup_CreateContext(ctx);
+	_swsetup_Wakeup(ctx);
+	_ae_create_context(ctx);
+
+	/* Install the customized pipeline:
+	 */
+	_tnl_destroy_pipeline(ctx);
+	_tnl_install_pipeline(ctx, r700_pipeline);
+
+	/* Try and keep materials and vertices separate:
+	 */
+/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
+
+	/* Configure swrast and TNL to match hardware characteristics:
+	 */
+	_swrast_allow_pixel_fog(ctx, GL_FALSE);
+	_swrast_allow_vertex_fog(ctx, GL_TRUE);
+	_tnl_allow_pixel_fog(ctx, GL_FALSE);
+	_tnl_allow_vertex_fog(ctx, GL_TRUE);
+
+	/* currently bogus data */
+	ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
+	ctx->Const.VertexProgram.MaxNativeInstructions =
+		VSF_MAX_FRAGMENT_LENGTH / 4;
+	ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
+	ctx->Const.VertexProgram.MaxTemps = 32;
+	ctx->Const.VertexProgram.MaxNativeTemps =
+		/*VSF_MAX_FRAGMENT_TEMPS */ 32;
+	ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
+	ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
+
+	ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
+	ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+	ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
+	ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
+	ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
+	ctx->Const.FragmentProgram.MaxNativeInstructions =
+	    PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
+	ctx->Const.FragmentProgram.MaxNativeTexIndirections =
+	    PFS_MAX_TEX_INDIRECT;
+	ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;	/* and these are?? */
+	ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
+	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
+
+	driInitExtensions(ctx, card_extensions, GL_TRUE);
+	if (r600->radeon.radeonScreen->kernel_mm)
+	  driInitExtensions(ctx, mm_extensions, GL_FALSE);
+
+	if (driQueryOptionb
+	    (&r600->radeon.optionCache, "disable_stencil_two_side"))
+		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
+
+	if (r600->radeon.glCtx->Mesa_DXTn
+	    && !driQueryOptionb(&r600->radeon.optionCache, "disable_s3tc")) {
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+		_mesa_enable_extension(ctx, "GL_S3_s3tc");
+	} else
+	    if (driQueryOptionb(&r600->radeon.optionCache, "force_s3tc_enable"))
+	{
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+	}
+
+	radeon_fbo_init(&r600->radeon);
+   	radeonInitSpanFuncs( ctx );
+
+	r600InitCmdBuf(r600);
+
+	r700InitState(r600->radeon.glCtx);
+
+	TNL_CONTEXT(ctx)->Driver.RunPipeline = r600RunPipeline;
+
+	tcl_mode = driQueryOptioni(&r600->radeon.optionCache, "tcl_mode");
+	if (driQueryOptionb(&r600->radeon.optionCache, "no_rast")) {
+		fprintf(stderr, "disabling 3D acceleration\n");
+#if R200_MERGED
+		FALLBACK(&r600->radeon, RADEON_FALLBACK_DISABLE, 1);
+#endif
+	}
+	if (tcl_mode == DRI_CONF_TCL_SW ||
+	    !(r600->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+		if (r600->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+			r600->radeon.radeonScreen->chip_flags &=
+			    ~RADEON_CHIPSET_TCL;
+			fprintf(stderr, "Disabling HW TCL support\n");
+		}
+		TCL_FALLBACK(r600->radeon.glCtx,
+			     RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
+	}
+
+	return GL_TRUE;
+}
+
+/* Clean our own things only, radeonDestroyContext will do every thing else. */
+void
+r600DestroyContext (__DRIcontextPrivate * driContextPriv)
+{
+    GET_CURRENT_CONTEXT (ctx);
+    context_t *context = ctx ? R700_CONTEXT(ctx) : NULL;
+
+    if (context)
+	    FREE(context->hw.pStateList);
+}
+
+
+
diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h
new file mode 100644
index 0000000000..bcb33e1386
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_context.h
@@ -0,0 +1,185 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ * \author Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+#ifndef __R600_CONTEXT_H__
+#define __R600_CONTEXT_H__
+
+#include "tnl/t_vertex.h"
+#include "drm.h"
+#include "radeon_drm.h"
+#include "dri_util.h"
+#include "texmem.h"
+#include "radeon_common.h"
+
+#include "main/macros.h"
+#include "main/mtypes.h"
+#include "main/colormac.h"
+
+#include "r700_chip.h"
+#include "r600_tex.h"
+#include "r700_oglprog.h"
+
+struct r600_context;
+typedef struct r600_context context_t;
+
+GLboolean r700SendPSState(context_t *context);
+GLboolean r700SendVSState(context_t *context);
+GLboolean r700SendSQConfig(context_t *context);
+
+#include "main/mm.h"
+
+/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html .
+   I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble
+   with other compilers ... GLUE!
+*/
+#define WARN_ONCE(a, ...)	{ \
+	static int warn##__LINE__=1; \
+	if(warn##__LINE__){ \
+		fprintf(stderr, "*********************************WARN_ONCE*********************************\n"); \
+		fprintf(stderr, "File %s function %s line %d\n", \
+			__FILE__, __FUNCTION__, __LINE__); \
+		fprintf(stderr,  a, ## __VA_ARGS__);\
+		fprintf(stderr, "***************************************************************************\n"); \
+		warn##__LINE__=0;\
+		} \
+	}
+
+/************ DMA BUFFERS **************/
+
+/* The blit width for texture uploads
+ */
+#define R600_BLIT_WIDTH_BYTES 1024
+#define R600_MAX_TEXTURE_UNITS 8
+
+struct r600_texture_state {
+	int tc_count;		/* number of incoming texture coordinates from VAP */
+};
+
+/* Perhaps more if we store programs in vmem? */
+/* drm_r600_cmd_header_t->vpu->count is unsigned char */
+#define VSF_MAX_FRAGMENT_LENGTH (255*4)
+
+/* Can be tested with colormat currently. */
+#define VSF_MAX_FRAGMENT_TEMPS (14)
+
+#define STATE_R600_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
+#define STATE_R600_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1)
+
+extern int hw_tcl_on;
+
+#define COLOR_IS_RGBA
+#define TAG(x) r600##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+#define PFS_MAX_ALU_INST	64
+#define PFS_MAX_TEX_INST	64
+#define PFS_MAX_TEX_INDIRECT 4
+#define PFS_NUM_TEMP_REGS	32
+#define PFS_NUM_CONST_REGS	16
+
+#define R600_MAX_AOS_ARRAYS		16
+
+#define REG_COORDS	0
+#define REG_COLOR0	1
+#define REG_TEX0	2
+
+#define R600_FALLBACK_NONE 0
+#define R600_FALLBACK_TCL 1
+#define R600_FALLBACK_RAST 2
+
+enum 
+{
+    NO_SHIFT    = 0,
+    LEFT_SHIFT  = 1,
+    RIGHT_SHIFT = 2,
+};
+
+typedef struct offset_modifiers
+{
+    GLuint shift;
+    GLuint shiftbits;
+    GLuint mask;
+} offset_modifiers;
+
+/**
+ * \brief R600 context structure.
+ */
+struct r600_context {
+	struct radeon_context radeon;	/* parent class, must be first */
+
+	/* ------ */
+	R700_CHIP_CONTEXT hw;
+
+	/* Vertex buffers
+	 */
+	GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
+	GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
+
+};
+
+#define R700_CONTEXT(ctx)		((context_t *)(ctx->DriverCtx))
+#define GL_CONTEXT(context)     ((GLcontext *)(context->radeon.glCtx))
+
+extern void r600DestroyContext(__DRIcontextPrivate * driContextPriv);
+extern GLboolean r600CreateContext(const __GLcontextModes * glVisual,
+				   __DRIcontextPrivate * driContextPriv,
+				   void *sharedContextPrivate);
+
+#define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(&context->hw))
+
+extern GLboolean r700InitChipObject(context_t *context);
+extern GLboolean r700SendContextStates(context_t *context);
+extern GLboolean r700SendViewportState(context_t *context, int id);
+extern GLboolean r700SendRenderTargetState(context_t *context, int id);
+
+extern int       r700SetupStreams(GLcontext * ctx);
+extern void      r700SetupVTXConstants(GLcontext  * ctx, 
+				       unsigned int nStreamID,
+				       void *       pAos,
+				       unsigned int size,      /* number of elements in vector */
+				       unsigned int stride,
+				       unsigned int Count);    /* number of vectors in stream */
+
+#define RADEON_D_CAPTURE 0
+#define RADEON_D_PLAYBACK 1
+#define RADEON_D_PLAYBACK_RAW 2
+#define RADEON_D_T 3
+
+#define r600PackFloat32 radeonPackFloat32
+#define r600PackFloat24 radeonPackFloat24
+
+#endif				/* __R600_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c
new file mode 100644
index 0000000000..685f7fe473
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_emit.c
@@ -0,0 +1,122 @@
+/**************************************************************************
+
+Copyright 2008, 2009 Advanced Micro Devices Inc. (AMD)
+
+Copyright (C) Advanced Micro Devices Inc. (AMD)  2009.  All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/colormac.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/image.h"
+
+#include "swrast_setup/swrast_setup.h"
+#include "math/m_translate.h"
+#include "tnl/tnl.h"
+#include "tnl/t_context.h"
+
+#include "r600_context.h"
+#include "r600_emit.h"
+
+void r600EmitCacheFlush(context_t *rmesa)
+{
+	BATCH_LOCALS(&rmesa->radeon);
+}
+
+GLboolean r600EmitShader(GLcontext * ctx, 
+                         void ** shaderbo,
+			             GLvoid * data, 
+                         int sizeinDWORD,
+                         char * szShaderUsage) 
+{
+    radeonContextPtr radeonctx = RADEON_CONTEXT(ctx);
+
+    struct radeon_bo * pbo;
+    uint32_t *out;
+
+shader_again_alloc:	
+#ifdef RADEON_DEBUG_BO
+	pbo = radeon_bo_open(radeonctx->radeonScreen->bom,
+					     0, 
+                         sizeinDWORD * 4, 
+                         256, 
+                         RADEON_GEM_DOMAIN_GTT,
+					     0,
+                         szShaderUsage);
+#else
+    pbo = radeon_bo_open(radeonctx->radeonScreen->bom,
+					     0, 
+                         sizeinDWORD * 4, 
+                         256, 
+                         RADEON_GEM_DOMAIN_GTT,
+					     0);
+#endif /* RADEON_DEBUG_BO */
+
+	if (!pbo) 
+    {
+		rcommonFlushCmdBuf(radeonctx, __FUNCTION__);
+		goto shader_again_alloc;
+	}
+
+        if (radeon_cs_space_check_with_bo(radeonctx->cmdbuf.cs,
+                                          pbo,
+                                          RADEON_GEM_DOMAIN_GTT, 0))
+                fprintf(stderr,"failure to revalidate BOs - badness\n");
+
+
+	radeon_bo_map(pbo, 1);
+
+    radeon_bo_ref(pbo);
+
+    out = (uint32_t*)(pbo->ptr);
+
+    memcpy(out, data, sizeinDWORD * 4);
+
+    radeon_bo_unmap(pbo); 
+
+    *shaderbo = (void*)pbo;
+
+    return GL_TRUE;
+}
+
+GLboolean r600DeleteShader(GLcontext * ctx, 
+                           void * shaderbo) 
+{
+    struct radeon_bo * pbo = (struct radeon_bo *)shaderbo;
+
+    if (pbo) {
+	    radeon_bo_unmap(pbo);
+	    radeon_bo_unref(pbo); /* when bo->cref <= 0, bo will be bo_free */
+    }
+
+    return GL_TRUE;
+}
diff --git a/src/mesa/drivers/dri/r200/r200_span.h b/src/mesa/drivers/dri/r600/r600_emit.h
index bae5644309..661774d11e 100644
--- a/src/mesa/drivers/dri/r200/r200_span.h
+++ b/src/mesa/drivers/dri/r600/r600_emit.h
@@ -1,9 +1,8 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+/**************************************************************************
+
+Copyright 2008, 2009 Advanced Micro Devices Inc. (AMD)
 
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
+Copyright (C) Advanced Micro Devices Inc. (AMD)  2009.  All Rights Reserved.
 
 Permission is hereby granted, free of charge, to any person obtaining
 a copy of this software and associated documentation files (the
@@ -29,17 +28,28 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 /*
  * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
  */
 
-#ifndef __R200_SPAN_H__
-#define __R200_SPAN_H__
 
-#include "drirenderbuffer.h"
+#ifndef __R600_EMIT_H__
+#define __R600_EMIT_H__
+
+#include "main/glheader.h"
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+#include "radeon_reg.h"
+
+void r600EmitCacheFlush(context_t *rmesa);
 
-extern void r200InitSpanFuncs( GLcontext *ctx );
+extern GLboolean r600EmitShader(GLcontext * ctx, 
+                                void ** shaderbo,
+			                    GLvoid * data, 
+                                int sizeinDWORD,
+                                char * szShaderUsage); 
 
-extern void
-radeonSetSpanFunctions(driRenderbuffer *rb, const GLvisual *vis);
+extern GLboolean r600DeleteShader(GLcontext * ctx, 
+                                 void * shaderbo);
 
 #endif
diff --git a/src/mesa/drivers/dri/r600/r600_reg.h b/src/mesa/drivers/dri/r600/r600_reg.h
new file mode 100644
index 0000000000..ffe5ee4f74
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg.h
@@ -0,0 +1,121 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _R600_REG_H_
+#define _R600_REG_H_
+
+/*
+ * Register definitions
+ */
+
+#include "r600_reg_auto_r6xx.h"
+#include "r600_reg_r6xx.h"
+#include "r600_reg_r7xx.h"
+
+
+/* SET_*_REG offsets + ends */
+enum 
+{
+    SET_CONFIG_REG_offset                = 0x00008000,
+    SET_CONFIG_REG_end                   = 0x0000ac00,
+    SET_CONTEXT_REG_offset               = 0x00028000,
+    SET_CONTEXT_REG_end                  = 0x00029000,
+    SET_ALU_CONST_offset                 = 0x00030000,
+    SET_ALU_CONST_end                    = 0x00032000,
+    SET_RESOURCE_offset                  = 0x00038000,
+    SET_RESOURCE_end                     = 0x0003c000,
+    SET_SAMPLER_offset                   = 0x0003c000,
+    SET_SAMPLER_end                      = 0x0003cff0,
+    SET_CTL_CONST_offset                 = 0x0003cff0,
+    SET_CTL_CONST_end                    = 0x0003e200,
+    SET_LOOP_CONST_offset                = 0x0003e200,
+    SET_LOOP_CONST_end                   = 0x0003e380,
+    SET_BOOL_CONST_offset                = 0x0003e380,
+    SET_BOOL_CONST_end                   = 0x00040000,
+};
+
+/* packet3 IT_SURFACE_BASE_UPDATE bits */
+enum 
+{
+    DEPTH_BASE                           = (1 << 0),
+    COLOR0_BASE                          = (1 << 1),
+    COLOR1_BASE                          = (1 << 2),
+    COLOR2_BASE                          = (1 << 3),
+    COLOR3_BASE                          = (1 << 4),
+    COLOR4_BASE                          = (1 << 5),
+    COLOR5_BASE                          = (1 << 6),
+    COLOR6_BASE                          = (1 << 7),
+    COLOR7_BASE                          = (1 << 8),
+    STRMOUT_BASE0                        = (1 << 9),
+    STRMOUT_BASE1                        = (1 << 10),
+    STRMOUT_BASE2                        = (1 << 11),
+    STRMOUT_BASE3                        = (1 << 12),
+    COHER_BASE0                          = (1 << 13),
+    COHER_BASE1                          = (1 << 14),
+};
+
+/* Packet3 commands */
+enum 
+{
+    IT_NOP                               = 0x10,
+    IT_INDIRECT_BUFFER_END               = 0x17,
+    IT_SET_PREDICATION                   = 0x20,
+    IT_REG_RMW                           = 0x21,
+    IT_COND_EXEC                         = 0x22,
+    IT_PRED_EXEC                         = 0x23,
+    IT_START_3D_CMDBUF                   = 0x24,
+    IT_DRAW_INDEX_2                      = 0x27,
+    IT_CONTEXT_CONTROL                   = 0x28,
+    IT_DRAW_INDEX_IMMD_BE                = 0x29,
+    IT_INDEX_TYPE                        = 0x2A,
+    IT_DRAW_INDEX                        = 0x2B,
+    IT_DRAW_INDEX_AUTO                   = 0x2D,
+    IT_DRAW_INDEX_IMMD                   = 0x2E,
+    IT_NUM_INSTANCES                     = 0x2F,
+    IT_STRMOUT_BUFFER_UPDATE             = 0x34,
+    IT_INDIRECT_BUFFER_MP                = 0x38,
+    IT_MEM_SEMAPHORE                     = 0x39,
+    IT_MPEG_INDEX                        = 0x3A,
+    IT_WAIT_REG_MEM                      = 0x3C,
+    IT_MEM_WRITE                         = 0x3D,
+    IT_INDIRECT_BUFFER                   = 0x32,
+    IT_CP_INTERRUPT                      = 0x40,
+    IT_SURFACE_SYNC                      = 0x43,
+    IT_ME_INITIALIZE                     = 0x44,
+    IT_COND_WRITE                        = 0x45,
+    IT_EVENT_WRITE                       = 0x46,
+    IT_EVENT_WRITE_EOP                   = 0x47,
+    IT_ONE_REG_WRITE                     = 0x57,
+    IT_SET_CONFIG_REG                    = 0x68,
+    IT_SET_CONTEXT_REG                   = 0x69,
+    IT_SET_ALU_CONST                     = 0x6A,
+    IT_SET_BOOL_CONST                    = 0x6B,
+    IT_SET_LOOP_CONST                    = 0x6C,
+    IT_SET_RESOURCE                      = 0x6D,
+    IT_SET_SAMPLER                       = 0x6E,
+    IT_SET_CTL_CONST                     = 0x6F,
+    IT_SURFACE_BASE_UPDATE               = 0x73,
+};
+
+#endif
diff --git a/src/mesa/drivers/dri/r600/r600_reg_auto_r6xx.h b/src/mesa/drivers/dri/r600/r600_reg_auto_r6xx.h
new file mode 100644
index 0000000000..9d5aa3c7e4
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg_auto_r6xx.h
@@ -0,0 +1,3087 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _AUTOREGS
+#define _AUTOREGS
+
+enum {
+
+    VGT_VTX_VECT_EJECT_REG                                = 0x000088b0,
+	PRIM_COUNT_mask                                   = 0x3ff << 0,
+	PRIM_COUNT_shift                                  = 0,
+    VGT_LAST_COPY_STATE                                   = 0x000088c0,
+	SRC_STATE_ID_mask                                 = 0x07 << 0,
+	SRC_STATE_ID_shift                                = 0,
+	DST_STATE_ID_mask                                 = 0x07 << 16,
+	DST_STATE_ID_shift                                = 16,
+    VGT_CACHE_INVALIDATION                                = 0x000088c4,
+	CACHE_INVALIDATION_mask                           = 0x03 << 0,
+	CACHE_INVALIDATION_shift                          = 0,
+	    VC_ONLY                                       = 0x00,
+	    TC_ONLY                                       = 0x01,
+	    VC_AND_TC                                     = 0x02,
+	VS_NO_EXTRA_BUFFER_bit                            = 1 << 5,
+    VGT_GS_PER_ES                                         = 0x000088c8,
+    VGT_ES_PER_GS                                         = 0x000088cc,
+    VGT_GS_VERTEX_REUSE                                   = 0x000088d4,
+	VERT_REUSE_mask                                   = 0x1f << 0,
+	VERT_REUSE_shift                                  = 0,
+    VGT_MC_LAT_CNTL                                       = 0x000088d8,
+	MC_TIME_STAMP_RES_mask                            = 0x03 << 0,
+	MC_TIME_STAMP_RES_shift                           = 0,
+	    X_0_992_MAX_LATENCY                           = 0x00,
+	    X_0_496_MAX_LATENCY                           = 0x01,
+	    X_0_248_MAX_LATENCY                           = 0x02,
+	    X_0_124_MAX_LATENCY                           = 0x03,
+    VGT_GS_PER_VS                                         = 0x000088e8,
+	GS_PER_VS_mask                                    = 0x0f << 0,
+	GS_PER_VS_shift                                   = 0,
+    VGT_CNTL_STATUS                                       = 0x000088f0,
+	VGT_OUT_INDX_BUSY_bit                             = 1 << 0,
+	VGT_OUT_BUSY_bit                                  = 1 << 1,
+	VGT_PT_BUSY_bit                                   = 1 << 2,
+	VGT_TE_BUSY_bit                                   = 1 << 3,
+	VGT_VR_BUSY_bit                                   = 1 << 4,
+	VGT_GRP_BUSY_bit                                  = 1 << 5,
+	VGT_DMA_REQ_BUSY_bit                              = 1 << 6,
+	VGT_DMA_BUSY_bit                                  = 1 << 7,
+	VGT_GS_BUSY_bit                                   = 1 << 8,
+	VGT_BUSY_bit                                      = 1 << 9,
+    VGT_PRIMITIVE_TYPE                                    = 0x00008958,
+	VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask                = 0x3f << 0,
+	VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift               = 0,
+	    DI_PT_NONE                                    = 0x00,
+	    DI_PT_POINTLIST                               = 0x01,
+	    DI_PT_LINELIST                                = 0x02,
+	    DI_PT_LINESTRIP                               = 0x03,
+	    DI_PT_TRILIST                                 = 0x04,
+	    DI_PT_TRIFAN                                  = 0x05,
+	    DI_PT_TRISTRIP                                = 0x06,
+	    DI_PT_UNUSED_0                                = 0x07,
+	    DI_PT_UNUSED_1                                = 0x08,
+	    DI_PT_UNUSED_2                                = 0x09,
+	    DI_PT_LINELIST_ADJ                            = 0x0a,
+	    DI_PT_LINESTRIP_ADJ                           = 0x0b,
+	    DI_PT_TRILIST_ADJ                             = 0x0c,
+	    DI_PT_TRISTRIP_ADJ                            = 0x0d,
+	    DI_PT_UNUSED_3                                = 0x0e,
+	    DI_PT_UNUSED_4                                = 0x0f,
+	    DI_PT_TRI_WITH_WFLAGS                         = 0x10,
+	    DI_PT_RECTLIST                                = 0x11,
+	    DI_PT_LINELOOP                                = 0x12,
+	    DI_PT_QUADLIST                                = 0x13,
+	    DI_PT_QUADSTRIP                               = 0x14,
+	    DI_PT_POLYGON                                 = 0x15,
+	    DI_PT_2D_COPY_RECT_LIST_V0                    = 0x16,
+	    DI_PT_2D_COPY_RECT_LIST_V1                    = 0x17,
+	    DI_PT_2D_COPY_RECT_LIST_V2                    = 0x18,
+	    DI_PT_2D_COPY_RECT_LIST_V3                    = 0x19,
+	    DI_PT_2D_FILL_RECT_LIST                       = 0x1a,
+	    DI_PT_2D_LINE_STRIP                           = 0x1b,
+	    DI_PT_2D_TRI_STRIP                            = 0x1c,
+    VGT_INDEX_TYPE                                        = 0x0000895c,
+	INDEX_TYPE_mask                                   = 0x03 << 0,
+	INDEX_TYPE_shift                                  = 0,
+	    DI_INDEX_SIZE_16_BIT                          = 0x00,
+	    DI_INDEX_SIZE_32_BIT                          = 0x01,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_0                      = 0x00008960,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_1                      = 0x00008964,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_2                      = 0x00008968,
+    VGT_STRMOUT_BUFFER_FILLED_SIZE_3                      = 0x0000896c,
+    VGT_NUM_INDICES                                       = 0x00008970,
+    VGT_NUM_INSTANCES                                     = 0x00008974,
+    PA_CL_CNTL_STATUS                                     = 0x00008a10,
+	CL_BUSY_bit                                       = 1 << 31,
+    PA_CL_ENHANCE                                         = 0x00008a14,
+	CLIP_VTX_REORDER_ENA_bit                          = 1 << 0,
+	NUM_CLIP_SEQ_mask                                 = 0x03 << 1,
+	NUM_CLIP_SEQ_shift                                = 1,
+	CLIPPED_PRIM_SEQ_STALL_bit                        = 1 << 3,
+	VE_NAN_PROC_DISABLE_bit                           = 1 << 4,
+    PA_SU_CNTL_STATUS                                     = 0x00008a50,
+	SU_BUSY_bit                                       = 1 << 31,
+    PA_SC_LINE_STIPPLE_STATE                              = 0x00008b10,
+	CURRENT_PTR_mask                                  = 0x0f << 0,
+	CURRENT_PTR_shift                                 = 0,
+	CURRENT_COUNT_mask                                = 0xff << 8,
+	CURRENT_COUNT_shift                               = 8,
+    PA_SC_MULTI_CHIP_CNTL                                 = 0x00008b20,
+	LOG2_NUM_CHIPS_mask                               = 0x07 << 0,
+	LOG2_NUM_CHIPS_shift                              = 0,
+	MULTI_CHIP_TILE_SIZE_mask                         = 0x03 << 3,
+	MULTI_CHIP_TILE_SIZE_shift                        = 3,
+	    X_16_X_16_PIXEL_TILE_PER_CHIP                 = 0x00,
+	    X_32_X_32_PIXEL_TILE_PER_CHIP                 = 0x01,
+	    X_64_X_64_PIXEL_TILE_PER_CHIP                 = 0x02,
+	    X_128X128_PIXEL_TILE_PER_CHIP                 = 0x03,
+	CHIP_TILE_X_LOC_mask                              = 0x07 << 5,
+	CHIP_TILE_X_LOC_shift                             = 5,
+	CHIP_TILE_Y_LOC_mask                              = 0x07 << 8,
+	CHIP_TILE_Y_LOC_shift                             = 8,
+	CHIP_SUPER_TILE_B_bit                             = 1 << 11,
+    PA_SC_AA_SAMPLE_LOCS_2S                               = 0x00008b40,
+	S0_X_mask                                         = 0x0f << 0,
+	S0_X_shift                                        = 0,
+	S0_Y_mask                                         = 0x0f << 4,
+	S0_Y_shift                                        = 4,
+	S1_X_mask                                         = 0x0f << 8,
+	S1_X_shift                                        = 8,
+	S1_Y_mask                                         = 0x0f << 12,
+	S1_Y_shift                                        = 12,
+    PA_SC_AA_SAMPLE_LOCS_4S                               = 0x00008b44,
+/* 	S0_X_mask                                         = 0x0f << 0, */
+/* 	S0_X_shift                                        = 0, */
+/* 	S0_Y_mask                                         = 0x0f << 4, */
+/* 	S0_Y_shift                                        = 4, */
+/* 	S1_X_mask                                         = 0x0f << 8, */
+/* 	S1_X_shift                                        = 8, */
+/* 	S1_Y_mask                                         = 0x0f << 12, */
+/* 	S1_Y_shift                                        = 12, */
+	S2_X_mask                                         = 0x0f << 16,
+	S2_X_shift                                        = 16,
+	S2_Y_mask                                         = 0x0f << 20,
+	S2_Y_shift                                        = 20,
+	S3_X_mask                                         = 0x0f << 24,
+	S3_X_shift                                        = 24,
+	S3_Y_mask                                         = 0x0f << 28,
+	S3_Y_shift                                        = 28,
+    PA_SC_AA_SAMPLE_LOCS_8S_WD0                           = 0x00008b48,
+/* 	S0_X_mask                                         = 0x0f << 0, */
+/* 	S0_X_shift                                        = 0, */
+/* 	S0_Y_mask                                         = 0x0f << 4, */
+/* 	S0_Y_shift                                        = 4, */
+/* 	S1_X_mask                                         = 0x0f << 8, */
+/* 	S1_X_shift                                        = 8, */
+/* 	S1_Y_mask                                         = 0x0f << 12, */
+/* 	S1_Y_shift                                        = 12, */
+/* 	S2_X_mask                                         = 0x0f << 16, */
+/* 	S2_X_shift                                        = 16, */
+/* 	S2_Y_mask                                         = 0x0f << 20, */
+/* 	S2_Y_shift                                        = 20, */
+/* 	S3_X_mask                                         = 0x0f << 24, */
+/* 	S3_X_shift                                        = 24, */
+/* 	S3_Y_mask                                         = 0x0f << 28, */
+/* 	S3_Y_shift                                        = 28, */
+    PA_SC_AA_SAMPLE_LOCS_8S_WD1                           = 0x00008b4c,
+	S4_X_mask                                         = 0x0f << 0,
+	S4_X_shift                                        = 0,
+	S4_Y_mask                                         = 0x0f << 4,
+	S4_Y_shift                                        = 4,
+	S5_X_mask                                         = 0x0f << 8,
+	S5_X_shift                                        = 8,
+	S5_Y_mask                                         = 0x0f << 12,
+	S5_Y_shift                                        = 12,
+	S6_X_mask                                         = 0x0f << 16,
+	S6_X_shift                                        = 16,
+	S6_Y_mask                                         = 0x0f << 20,
+	S6_Y_shift                                        = 20,
+	S7_X_mask                                         = 0x0f << 24,
+	S7_X_shift                                        = 24,
+	S7_Y_mask                                         = 0x0f << 28,
+	S7_Y_shift                                        = 28,
+    PA_SC_CNTL_STATUS                                     = 0x00008be0,
+	MPASS_OVERFLOW_bit                                = 1 << 30,
+    PA_SC_ENHANCE                                         = 0x00008bf0,
+	FORCE_EOV_MAX_CLK_CNT_mask                        = 0xfff << 0,
+	FORCE_EOV_MAX_CLK_CNT_shift                       = 0,
+	FORCE_EOV_MAX_TILE_CNT_mask                       = 0xfff << 12,
+	FORCE_EOV_MAX_TILE_CNT_shift                      = 12,
+    SQ_CONFIG                                             = 0x00008c00,
+	VC_ENABLE_bit                                     = 1 << 0,
+	EXPORT_SRC_C_bit                                  = 1 << 1,
+	DX9_CONSTS_bit                                    = 1 << 2,
+	ALU_INST_PREFER_VECTOR_bit                        = 1 << 3,
+	SQ_CONFIG__DX10_CLAMP_bit                         = 1 << 4,
+	ALU_PREFER_ONE_WATERFALL_bit                      = 1 << 5,
+	ALU_MAX_ONE_WATERFALL_bit                         = 1 << 6,
+	CLAUSE_SEQ_PRIO_mask                              = 0x03 << 8,
+	CLAUSE_SEQ_PRIO_shift                             = 8,
+	    SQ_CL_PRIO_RND_ROBIN                          = 0x00,
+	    SQ_CL_PRIO_MACRO_SEQ                          = 0x01,
+	    SQ_CL_PRIO_NONE                               = 0x02,
+	PS_PRIO_mask                                      = 0x03 << 24,
+	PS_PRIO_shift                                     = 24,
+	VS_PRIO_mask                                      = 0x03 << 26,
+	VS_PRIO_shift                                     = 26,
+	GS_PRIO_mask                                      = 0x03 << 28,
+	GS_PRIO_shift                                     = 28,
+	ES_PRIO_mask                                      = 0x03 << 30,
+	ES_PRIO_shift                                     = 30,
+    SQ_GPR_RESOURCE_MGMT_1                                = 0x00008c04,
+	NUM_PS_GPRS_mask                                  = 0xff << 0,
+	NUM_PS_GPRS_shift                                 = 0,
+	NUM_VS_GPRS_mask                                  = 0xff << 16,
+	NUM_VS_GPRS_shift                                 = 16,
+	NUM_CLAUSE_TEMP_GPRS_mask                         = 0x0f << 28,
+	NUM_CLAUSE_TEMP_GPRS_shift                        = 28,
+    SQ_GPR_RESOURCE_MGMT_2                                = 0x00008c08,
+	NUM_GS_GPRS_mask                                  = 0xff << 0,
+	NUM_GS_GPRS_shift                                 = 0,
+	NUM_ES_GPRS_mask                                  = 0xff << 16,
+	NUM_ES_GPRS_shift                                 = 16,
+    SQ_THREAD_RESOURCE_MGMT                               = 0x00008c0c,
+	NUM_PS_THREADS_mask                               = 0xff << 0,
+	NUM_PS_THREADS_shift                              = 0,
+	NUM_VS_THREADS_mask                               = 0xff << 8,
+	NUM_VS_THREADS_shift                              = 8,
+	NUM_GS_THREADS_mask                               = 0xff << 16,
+	NUM_GS_THREADS_shift                              = 16,
+	NUM_ES_THREADS_mask                               = 0xff << 24,
+	NUM_ES_THREADS_shift                              = 24,
+    SQ_STACK_RESOURCE_MGMT_1                              = 0x00008c10,
+	NUM_PS_STACK_ENTRIES_mask                         = 0xfff << 0,
+	NUM_PS_STACK_ENTRIES_shift                        = 0,
+	NUM_VS_STACK_ENTRIES_mask                         = 0xfff << 16,
+	NUM_VS_STACK_ENTRIES_shift                        = 16,
+    SQ_STACK_RESOURCE_MGMT_2                              = 0x00008c14,
+	NUM_GS_STACK_ENTRIES_mask                         = 0xfff << 0,
+	NUM_GS_STACK_ENTRIES_shift                        = 0,
+	NUM_ES_STACK_ENTRIES_mask                         = 0xfff << 16,
+	NUM_ES_STACK_ENTRIES_shift                        = 16,
+    SQ_ESGS_RING_BASE                                     = 0x00008c40,
+    SQ_ESGS_RING_SIZE                                     = 0x00008c44,
+    SQ_GSVS_RING_BASE                                     = 0x00008c48,
+    SQ_GSVS_RING_SIZE                                     = 0x00008c4c,
+    SQ_ESTMP_RING_BASE                                    = 0x00008c50,
+    SQ_ESTMP_RING_SIZE                                    = 0x00008c54,
+    SQ_GSTMP_RING_BASE                                    = 0x00008c58,
+    SQ_GSTMP_RING_SIZE                                    = 0x00008c5c,
+    SQ_VSTMP_RING_BASE                                    = 0x00008c60,
+    SQ_VSTMP_RING_SIZE                                    = 0x00008c64,
+    SQ_PSTMP_RING_BASE                                    = 0x00008c68,
+    SQ_PSTMP_RING_SIZE                                    = 0x00008c6c,
+    SQ_FBUF_RING_BASE                                     = 0x00008c70,
+    SQ_FBUF_RING_SIZE                                     = 0x00008c74,
+    SQ_REDUC_RING_BASE                                    = 0x00008c78,
+    SQ_REDUC_RING_SIZE                                    = 0x00008c7c,
+    SQ_ALU_WORD1_OP3                                      = 0x00008dfc,
+	SRC2_SEL_mask                                     = 0x1ff << 0,
+	SRC2_SEL_shift                                    = 0,
+	    SQ_ALU_SRC_0                                  = 0xf8,
+	    SQ_ALU_SRC_1                                  = 0xf9,
+	    SQ_ALU_SRC_1_INT                              = 0xfa,
+	    SQ_ALU_SRC_M_1_INT                            = 0xfb,
+	    SQ_ALU_SRC_0_5                                = 0xfc,
+	    SQ_ALU_SRC_LITERAL                            = 0xfd,
+	    SQ_ALU_SRC_PV                                 = 0xfe,
+	    SQ_ALU_SRC_PS                                 = 0xff,
+	SRC2_REL_bit                                      = 1 << 9,
+	SRC2_CHAN_mask                                    = 0x03 << 10,
+	SRC2_CHAN_shift                                   = 10,
+	    SQ_CHAN_X                                     = 0x00,
+	    SQ_CHAN_Y                                     = 0x01,
+	    SQ_CHAN_Z                                     = 0x02,
+	    SQ_CHAN_W                                     = 0x03,
+	SRC2_NEG_bit                                      = 1 << 12,
+	SQ_ALU_WORD1_OP3__ALU_INST_mask                   = 0x1f << 13,
+	SQ_ALU_WORD1_OP3__ALU_INST_shift                  = 13,
+	    SQ_OP3_INST_MUL_LIT                           = 0x0c,
+	    SQ_OP3_INST_MUL_LIT_M2                        = 0x0d,
+	    SQ_OP3_INST_MUL_LIT_M4                        = 0x0e,
+	    SQ_OP3_INST_MUL_LIT_D2                        = 0x0f,
+	    SQ_OP3_INST_MULADD                            = 0x10,
+	    SQ_OP3_INST_MULADD_M2                         = 0x11,
+	    SQ_OP3_INST_MULADD_M4                         = 0x12,
+	    SQ_OP3_INST_MULADD_D2                         = 0x13,
+	    SQ_OP3_INST_MULADD_IEEE                       = 0x14,
+	    SQ_OP3_INST_MULADD_IEEE_M2                    = 0x15,
+	    SQ_OP3_INST_MULADD_IEEE_M4                    = 0x16,
+	    SQ_OP3_INST_MULADD_IEEE_D2                    = 0x17,
+	    SQ_OP3_INST_CNDE                              = 0x18,
+	    SQ_OP3_INST_CNDGT                             = 0x19,
+	    SQ_OP3_INST_CNDGE                             = 0x1a,
+	    SQ_OP3_INST_CNDE_INT                          = 0x1c,
+	    SQ_OP3_INST_CNDGT_INT                         = 0x1d,
+	    SQ_OP3_INST_CNDGE_INT                         = 0x1e,
+    SQ_TEX_WORD2                                          = 0x00008dfc,
+	OFFSET_X_mask                                     = 0x1f << 0,
+	OFFSET_X_shift                                    = 0,
+	OFFSET_Y_mask                                     = 0x1f << 5,
+	OFFSET_Y_shift                                    = 5,
+	OFFSET_Z_mask                                     = 0x1f << 10,
+	OFFSET_Z_shift                                    = 10,
+	SAMPLER_ID_mask                                   = 0x1f << 15,
+	SAMPLER_ID_shift                                  = 15,
+	SQ_TEX_WORD2__SRC_SEL_X_mask                      = 0x07 << 20,
+	SQ_TEX_WORD2__SRC_SEL_X_shift                     = 20,
+	    SQ_SEL_X                                      = 0x00,
+	    SQ_SEL_Y                                      = 0x01,
+	    SQ_SEL_Z                                      = 0x02,
+	    SQ_SEL_W                                      = 0x03,
+	    SQ_SEL_0                                      = 0x04,
+	    SQ_SEL_1                                      = 0x05,
+	SRC_SEL_Y_mask                                    = 0x07 << 23,
+	SRC_SEL_Y_shift                                   = 23,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SRC_SEL_Z_mask                                    = 0x07 << 26,
+	SRC_SEL_Z_shift                                   = 26,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SRC_SEL_W_mask                                    = 0x07 << 29,
+	SRC_SEL_W_shift                                   = 29,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+    SQ_CF_ALLOC_EXPORT_WORD1                              = 0x00008dfc,
+	BURST_COUNT_mask                                  = 0x0f << 17,
+	BURST_COUNT_shift                                 = 17,
+	END_OF_PROGRAM_bit                                = 1 << 21,
+	VALID_PIXEL_MODE_bit                              = 1 << 22,
+	SQ_CF_ALLOC_EXPORT_WORD1__CF_INST_mask            = 0x7f << 23,
+	SQ_CF_ALLOC_EXPORT_WORD1__CF_INST_shift           = 23,
+	    SQ_CF_INST_MEM_STREAM0                        = 0x20,
+	    SQ_CF_INST_MEM_STREAM1                        = 0x21,
+	    SQ_CF_INST_MEM_STREAM2                        = 0x22,
+	    SQ_CF_INST_MEM_STREAM3                        = 0x23,
+	    SQ_CF_INST_MEM_SCRATCH                        = 0x24,
+	    SQ_CF_INST_MEM_REDUCTION                      = 0x25,
+	    SQ_CF_INST_MEM_RING                           = 0x26,
+	    SQ_CF_INST_EXPORT                             = 0x27,
+	    SQ_CF_INST_EXPORT_DONE                        = 0x28,
+	WHOLE_QUAD_MODE_bit                               = 1 << 30,
+	BARRIER_bit                                       = 1 << 31,
+    SQ_CF_ALU_WORD1                                       = 0x00008dfc,
+	KCACHE_MODE1_mask                                 = 0x03 << 0,
+	KCACHE_MODE1_shift                                = 0,
+	    SQ_CF_KCACHE_NOP                              = 0x00,
+	    SQ_CF_KCACHE_LOCK_1                           = 0x01,
+	    SQ_CF_KCACHE_LOCK_2                           = 0x02,
+	    SQ_CF_KCACHE_LOCK_LOOP_INDEX                  = 0x03,
+	KCACHE_ADDR0_mask                                 = 0xff << 2,
+	KCACHE_ADDR0_shift                                = 2,
+	KCACHE_ADDR1_mask                                 = 0xff << 10,
+	KCACHE_ADDR1_shift                                = 10,
+	SQ_CF_ALU_WORD1__COUNT_mask                       = 0x7f << 18,
+	SQ_CF_ALU_WORD1__COUNT_shift                      = 18,
+	SQ_CF_ALU_WORD1__ALT_CONST_bit                    = 1 << 25,
+	SQ_CF_ALU_WORD1__CF_INST_mask                     = 0x0f << 26,
+	SQ_CF_ALU_WORD1__CF_INST_shift                    = 26,
+	    SQ_CF_INST_ALU                                = 0x08,
+	    SQ_CF_INST_ALU_PUSH_BEFORE                    = 0x09,
+	    SQ_CF_INST_ALU_POP_AFTER                      = 0x0a,
+	    SQ_CF_INST_ALU_POP2_AFTER                     = 0x0b,
+	    SQ_CF_INST_ALU_CONTINUE                       = 0x0d,
+	    SQ_CF_INST_ALU_BREAK                          = 0x0e,
+	    SQ_CF_INST_ALU_ELSE_AFTER                     = 0x0f,
+/* 	WHOLE_QUAD_MODE_bit                               = 1 << 30, */
+/* 	BARRIER_bit                                       = 1 << 31, */
+    SQ_TEX_WORD1                                          = 0x00008dfc,
+	SQ_TEX_WORD1__DST_GPR_mask                        = 0x7f << 0,
+	SQ_TEX_WORD1__DST_GPR_shift                       = 0,
+	SQ_TEX_WORD1__DST_REL_bit                         = 1 << 7,
+	SQ_TEX_WORD1__DST_SEL_X_mask                      = 0x07 << 9,
+	SQ_TEX_WORD1__DST_SEL_X_shift                     = 9,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	    SQ_SEL_MASK                                   = 0x07,
+	SQ_TEX_WORD1__DST_SEL_Y_mask                      = 0x07 << 12,
+	SQ_TEX_WORD1__DST_SEL_Y_shift                     = 12,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_TEX_WORD1__DST_SEL_Z_mask                      = 0x07 << 15,
+	SQ_TEX_WORD1__DST_SEL_Z_shift                     = 15,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_TEX_WORD1__DST_SEL_W_mask                      = 0x07 << 18,
+	SQ_TEX_WORD1__DST_SEL_W_shift                     = 18,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_TEX_WORD1__LOD_BIAS_mask                       = 0x7f << 21,
+	SQ_TEX_WORD1__LOD_BIAS_shift                      = 21,
+	COORD_TYPE_X_bit                                  = 1 << 28,
+	COORD_TYPE_Y_bit                                  = 1 << 29,
+	COORD_TYPE_Z_bit                                  = 1 << 30,
+	COORD_TYPE_W_bit                                  = 1 << 31,
+    SQ_VTX_WORD0                                          = 0x00008dfc,
+	VTX_INST_mask                                     = 0x1f << 0,
+	VTX_INST_shift                                    = 0,
+	    SQ_VTX_INST_FETCH                             = 0x00,
+	    SQ_VTX_INST_SEMANTIC                          = 0x01,
+	FETCH_TYPE_mask                                   = 0x03 << 5,
+	FETCH_TYPE_shift                                  = 5,
+	    SQ_VTX_FETCH_VERTEX_DATA                      = 0x00,
+	    SQ_VTX_FETCH_INSTANCE_DATA                    = 0x01,
+	    SQ_VTX_FETCH_NO_INDEX_OFFSET                  = 0x02,
+	FETCH_WHOLE_QUAD_bit                              = 1 << 7,
+	BUFFER_ID_mask                                    = 0xff << 8,
+	BUFFER_ID_shift                                   = 8,
+	SRC_GPR_mask                                      = 0x7f << 16,
+	SRC_GPR_shift                                     = 16,
+	SRC_REL_bit                                       = 1 << 23,
+	SQ_VTX_WORD0__SRC_SEL_X_mask                      = 0x03 << 24,
+	SQ_VTX_WORD0__SRC_SEL_X_shift                     = 24,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+	MEGA_FETCH_COUNT_mask                             = 0x3f << 26,
+	MEGA_FETCH_COUNT_shift                            = 26,
+    SQ_CF_ALLOC_EXPORT_WORD1_SWIZ                         = 0x00008dfc,
+	SEL_X_mask                                        = 0x07 << 0,
+	SEL_X_shift                                       = 0,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SEL_Y_mask                                        = 0x07 << 3,
+	SEL_Y_shift                                       = 3,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SEL_Z_mask                                        = 0x07 << 6,
+	SEL_Z_shift                                       = 6,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SEL_W_mask                                        = 0x07 << 9,
+	SEL_W_shift                                       = 9,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+    SQ_ALU_WORD1                                          = 0x00008dfc,
+	ENCODING_mask                                     = 0x07 << 15,
+	ENCODING_shift                                    = 15,
+	BANK_SWIZZLE_mask                                 = 0x07 << 18,
+	BANK_SWIZZLE_shift                                = 18,
+	    SQ_ALU_VEC_012                                = 0x00,
+	    SQ_ALU_VEC_021                                = 0x01,
+	    SQ_ALU_VEC_120                                = 0x02,
+	    SQ_ALU_VEC_102                                = 0x03,
+	    SQ_ALU_VEC_201                                = 0x04,
+	    SQ_ALU_VEC_210                                = 0x05,
+	SQ_ALU_WORD1__DST_GPR_mask                        = 0x7f << 21,
+	SQ_ALU_WORD1__DST_GPR_shift                       = 21,
+	SQ_ALU_WORD1__DST_REL_bit                         = 1 << 28,
+	DST_CHAN_mask                                     = 0x03 << 29,
+	DST_CHAN_shift                                    = 29,
+	    CHAN_X                                        = 0x00,
+	    CHAN_Y                                        = 0x01,
+	    CHAN_Z                                        = 0x02,
+	    CHAN_W                                        = 0x03,
+	SQ_ALU_WORD1__CLAMP_bit                           = 1 << 31,
+    SQ_CF_ALU_WORD0                                       = 0x00008dfc,
+	SQ_CF_ALU_WORD0__ADDR_mask                        = 0x3fffff << 0,
+	SQ_CF_ALU_WORD0__ADDR_shift                       = 0,
+	KCACHE_BANK0_mask                                 = 0x0f << 22,
+	KCACHE_BANK0_shift                                = 22,
+	KCACHE_BANK1_mask                                 = 0x0f << 26,
+	KCACHE_BANK1_shift                                = 26,
+	KCACHE_MODE0_mask                                 = 0x03 << 30,
+	KCACHE_MODE0_shift                                = 30,
+/* 	    SQ_CF_KCACHE_NOP                              = 0x00, */
+/* 	    SQ_CF_KCACHE_LOCK_1                           = 0x01, */
+/* 	    SQ_CF_KCACHE_LOCK_2                           = 0x02, */
+/* 	    SQ_CF_KCACHE_LOCK_LOOP_INDEX                  = 0x03, */
+    SQ_VTX_WORD2                                          = 0x00008dfc,
+	SQ_VTX_WORD2__OFFSET_mask                         = 0xffff << 0,
+	SQ_VTX_WORD2__OFFSET_shift                        = 0,
+	SQ_VTX_WORD2__ENDIAN_SWAP_mask                    = 0x03 << 16,
+	SQ_VTX_WORD2__ENDIAN_SWAP_shift                   = 16,
+	    SQ_ENDIAN_NONE                                = 0x00,
+	    SQ_ENDIAN_8IN16                               = 0x01,
+	    SQ_ENDIAN_8IN32                               = 0x02,
+	CONST_BUF_NO_STRIDE_bit                           = 1 << 18,
+	MEGA_FETCH_bit                                    = 1 << 19,
+	SQ_VTX_WORD2__ALT_CONST_bit                       = 1 << 20,
+    SQ_ALU_WORD1_OP2_V2                                   = 0x00008dfc,
+	SRC0_ABS_bit                                      = 1 << 0,
+	SRC1_ABS_bit                                      = 1 << 1,
+	UPDATE_EXECUTE_MASK_bit                           = 1 << 2,
+	UPDATE_PRED_bit                                   = 1 << 3,
+	WRITE_MASK_bit                                    = 1 << 4,
+	SQ_ALU_WORD1_OP2_V2__OMOD_mask                    = 0x03 << 5,
+	SQ_ALU_WORD1_OP2_V2__OMOD_shift                   = 5,
+	    SQ_ALU_OMOD_OFF                               = 0x00,
+	    SQ_ALU_OMOD_M2                                = 0x01,
+	    SQ_ALU_OMOD_M4                                = 0x02,
+	    SQ_ALU_OMOD_D2                                = 0x03,
+	SQ_ALU_WORD1_OP2_V2__ALU_INST_mask                = 0x7ff << 7,
+	SQ_ALU_WORD1_OP2_V2__ALU_INST_shift               = 7,
+	    SQ_OP2_INST_ADD                               = 0x00,
+	    SQ_OP2_INST_MUL                               = 0x01,
+	    SQ_OP2_INST_MUL_IEEE                          = 0x02,
+	    SQ_OP2_INST_MAX                               = 0x03,
+	    SQ_OP2_INST_MIN                               = 0x04,
+	    SQ_OP2_INST_MAX_DX10                          = 0x05,
+	    SQ_OP2_INST_MIN_DX10                          = 0x06,
+	    SQ_OP2_INST_SETE                              = 0x08,
+	    SQ_OP2_INST_SETGT                             = 0x09,
+	    SQ_OP2_INST_SETGE                             = 0x0a,
+	    SQ_OP2_INST_SETNE                             = 0x0b,
+	    SQ_OP2_INST_SETE_DX10                         = 0x0c,
+	    SQ_OP2_INST_SETGT_DX10                        = 0x0d,
+	    SQ_OP2_INST_SETGE_DX10                        = 0x0e,
+	    SQ_OP2_INST_SETNE_DX10                        = 0x0f,
+	    SQ_OP2_INST_FRACT                             = 0x10,
+	    SQ_OP2_INST_TRUNC                             = 0x11,
+	    SQ_OP2_INST_CEIL                              = 0x12,
+	    SQ_OP2_INST_RNDNE                             = 0x13,
+	    SQ_OP2_INST_FLOOR                             = 0x14,
+	    SQ_OP2_INST_MOVA                              = 0x15,
+	    SQ_OP2_INST_MOVA_FLOOR                        = 0x16,
+	    SQ_OP2_INST_MOVA_INT                          = 0x18,
+	    SQ_OP2_INST_MOV                               = 0x19,
+	    SQ_OP2_INST_NOP                               = 0x1a,
+	    SQ_OP2_INST_PRED_SETGT_UINT                   = 0x1e,
+	    SQ_OP2_INST_PRED_SETGE_UINT                   = 0x1f,
+	    SQ_OP2_INST_PRED_SETE                         = 0x20,
+	    SQ_OP2_INST_PRED_SETGT                        = 0x21,
+	    SQ_OP2_INST_PRED_SETGE                        = 0x22,
+	    SQ_OP2_INST_PRED_SETNE                        = 0x23,
+	    SQ_OP2_INST_PRED_SET_INV                      = 0x24,
+	    SQ_OP2_INST_PRED_SET_POP                      = 0x25,
+	    SQ_OP2_INST_PRED_SET_CLR                      = 0x26,
+	    SQ_OP2_INST_PRED_SET_RESTORE                  = 0x27,
+	    SQ_OP2_INST_PRED_SETE_PUSH                    = 0x28,
+	    SQ_OP2_INST_PRED_SETGT_PUSH                   = 0x29,
+	    SQ_OP2_INST_PRED_SETGE_PUSH                   = 0x2a,
+	    SQ_OP2_INST_PRED_SETNE_PUSH                   = 0x2b,
+	    SQ_OP2_INST_KILLE                             = 0x2c,
+	    SQ_OP2_INST_KILLGT                            = 0x2d,
+	    SQ_OP2_INST_KILLGE                            = 0x2e,
+	    SQ_OP2_INST_KILLNE                            = 0x2f,
+	    SQ_OP2_INST_AND_INT                           = 0x30,
+	    SQ_OP2_INST_OR_INT                            = 0x31,
+	    SQ_OP2_INST_XOR_INT                           = 0x32,
+	    SQ_OP2_INST_NOT_INT                           = 0x33,
+	    SQ_OP2_INST_ADD_INT                           = 0x34,
+	    SQ_OP2_INST_SUB_INT                           = 0x35,
+	    SQ_OP2_INST_MAX_INT                           = 0x36,
+	    SQ_OP2_INST_MIN_INT                           = 0x37,
+	    SQ_OP2_INST_MAX_UINT                          = 0x38,
+	    SQ_OP2_INST_MIN_UINT                          = 0x39,
+	    SQ_OP2_INST_SETE_INT                          = 0x3a,
+	    SQ_OP2_INST_SETGT_INT                         = 0x3b,
+	    SQ_OP2_INST_SETGE_INT                         = 0x3c,
+	    SQ_OP2_INST_SETNE_INT                         = 0x3d,
+	    SQ_OP2_INST_SETGT_UINT                        = 0x3e,
+	    SQ_OP2_INST_SETGE_UINT                        = 0x3f,
+	    SQ_OP2_INST_KILLGT_UINT                       = 0x40,
+	    SQ_OP2_INST_KILLGE_UINT                       = 0x41,
+	    SQ_OP2_INST_PRED_SETE_INT                     = 0x42,
+	    SQ_OP2_INST_PRED_SETGT_INT                    = 0x43,
+	    SQ_OP2_INST_PRED_SETGE_INT                    = 0x44,
+	    SQ_OP2_INST_PRED_SETNE_INT                    = 0x45,
+	    SQ_OP2_INST_KILLE_INT                         = 0x46,
+	    SQ_OP2_INST_KILLGT_INT                        = 0x47,
+	    SQ_OP2_INST_KILLGE_INT                        = 0x48,
+	    SQ_OP2_INST_KILLNE_INT                        = 0x49,
+	    SQ_OP2_INST_PRED_SETE_PUSH_INT                = 0x4a,
+	    SQ_OP2_INST_PRED_SETGT_PUSH_INT               = 0x4b,
+	    SQ_OP2_INST_PRED_SETGE_PUSH_INT               = 0x4c,
+	    SQ_OP2_INST_PRED_SETNE_PUSH_INT               = 0x4d,
+	    SQ_OP2_INST_PRED_SETLT_PUSH_INT               = 0x4e,
+	    SQ_OP2_INST_PRED_SETLE_PUSH_INT               = 0x4f,
+	    SQ_OP2_INST_DOT4                              = 0x50,
+	    SQ_OP2_INST_DOT4_IEEE                         = 0x51,
+	    SQ_OP2_INST_CUBE                              = 0x52,
+	    SQ_OP2_INST_MAX4                              = 0x53,
+	    SQ_OP2_INST_MOVA_GPR_INT                      = 0x60,
+	    SQ_OP2_INST_EXP_IEEE                          = 0x61,
+	    SQ_OP2_INST_LOG_CLAMPED                       = 0x62,
+	    SQ_OP2_INST_LOG_IEEE                          = 0x63,
+	    SQ_OP2_INST_RECIP_CLAMPED                     = 0x64,
+	    SQ_OP2_INST_RECIP_FF                          = 0x65,
+	    SQ_OP2_INST_RECIP_IEEE                        = 0x66,
+	    SQ_OP2_INST_RECIPSQRT_CLAMPED                 = 0x67,
+	    SQ_OP2_INST_RECIPSQRT_FF                      = 0x68,
+	    SQ_OP2_INST_RECIPSQRT_IEEE                    = 0x69,
+	    SQ_OP2_INST_SQRT_IEEE                         = 0x6a,
+	    SQ_OP2_INST_FLT_TO_INT                        = 0x6b,
+	    SQ_OP2_INST_INT_TO_FLT                        = 0x6c,
+	    SQ_OP2_INST_UINT_TO_FLT                       = 0x6d,
+	    SQ_OP2_INST_SIN                               = 0x6e,
+	    SQ_OP2_INST_COS                               = 0x6f,
+	    SQ_OP2_INST_ASHR_INT                          = 0x70,
+	    SQ_OP2_INST_LSHR_INT                          = 0x71,
+	    SQ_OP2_INST_LSHL_INT                          = 0x72,
+	    SQ_OP2_INST_MULLO_INT                         = 0x73,
+	    SQ_OP2_INST_MULHI_INT                         = 0x74,
+	    SQ_OP2_INST_MULLO_UINT                        = 0x75,
+	    SQ_OP2_INST_MULHI_UINT                        = 0x76,
+	    SQ_OP2_INST_RECIP_INT                         = 0x77,
+	    SQ_OP2_INST_RECIP_UINT                        = 0x78,
+	    SQ_OP2_INST_FLT_TO_UINT                       = 0x79,
+    SQ_CF_ALLOC_EXPORT_WORD1_BUF                          = 0x00008dfc,
+	ARRAY_SIZE_mask                                   = 0xfff << 0,
+	ARRAY_SIZE_shift                                  = 0,
+	COMP_MASK_mask                                    = 0x0f << 12,
+	COMP_MASK_shift                                   = 12,
+    SQ_CF_WORD0                                           = 0x00008dfc,
+    SQ_CF_ALLOC_EXPORT_WORD0                              = 0x00008dfc,
+	ARRAY_BASE_mask                                   = 0x1fff << 0,
+	ARRAY_BASE_shift                                  = 0,
+	SQ_CF_ALLOC_EXPORT_WORD0__TYPE_mask               = 0x03 << 13,
+	SQ_CF_ALLOC_EXPORT_WORD0__TYPE_shift              = 13,
+	    SQ_EXPORT_PIXEL                               = 0x00,
+	    SQ_EXPORT_POS                                 = 0x01,
+	    SQ_EXPORT_PARAM                               = 0x02,
+	    X_UNUSED_FOR_SX_EXPORTS                       = 0x03,
+	RW_GPR_mask                                       = 0x7f << 15,
+	RW_GPR_shift                                      = 15,
+	RW_REL_bit                                        = 1 << 22,
+	INDEX_GPR_mask                                    = 0x7f << 23,
+	INDEX_GPR_shift                                   = 23,
+	ELEM_SIZE_mask                                    = 0x03 << 30,
+	ELEM_SIZE_shift                                   = 30,
+    SQ_VTX_WORD1                                          = 0x00008dfc,
+	SQ_VTX_WORD1__DST_SEL_X_mask                      = 0x07 << 9,
+	SQ_VTX_WORD1__DST_SEL_X_shift                     = 9,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_VTX_WORD1__DST_SEL_Y_mask                      = 0x07 << 12,
+	SQ_VTX_WORD1__DST_SEL_Y_shift                     = 12,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_VTX_WORD1__DST_SEL_Z_mask                      = 0x07 << 15,
+	SQ_VTX_WORD1__DST_SEL_Z_shift                     = 15,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	SQ_VTX_WORD1__DST_SEL_W_mask                      = 0x07 << 18,
+	SQ_VTX_WORD1__DST_SEL_W_shift                     = 18,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+/* 	    SQ_SEL_MASK                                   = 0x07, */
+	USE_CONST_FIELDS_bit                              = 1 << 21,
+	SQ_VTX_WORD1__DATA_FORMAT_mask                    = 0x3f << 22,
+	SQ_VTX_WORD1__DATA_FORMAT_shift                   = 22,
+	SQ_VTX_WORD1__NUM_FORMAT_ALL_mask                 = 0x03 << 28,
+	SQ_VTX_WORD1__NUM_FORMAT_ALL_shift                = 28,
+	    SQ_NUM_FORMAT_NORM                            = 0x00,
+	    SQ_NUM_FORMAT_INT                             = 0x01,
+	    SQ_NUM_FORMAT_SCALED                          = 0x02,
+	SQ_VTX_WORD1__FORMAT_COMP_ALL_bit                 = 1 << 30,
+	SQ_VTX_WORD1__SRF_MODE_ALL_bit                    = 1 << 31,
+    SQ_ALU_WORD1_OP2                                      = 0x00008dfc,
+/* 	SRC0_ABS_bit                                      = 1 << 0, */
+/* 	SRC1_ABS_bit                                      = 1 << 1, */
+/* 	UPDATE_EXECUTE_MASK_bit                           = 1 << 2, */
+/* 	UPDATE_PRED_bit                                   = 1 << 3, */
+/* 	WRITE_MASK_bit                                    = 1 << 4, */
+	FOG_MERGE_bit                                     = 1 << 5,
+	SQ_ALU_WORD1_OP2__OMOD_mask                       = 0x03 << 6,
+	SQ_ALU_WORD1_OP2__OMOD_shift                      = 6,
+/* 	    SQ_ALU_OMOD_OFF                               = 0x00, */
+/* 	    SQ_ALU_OMOD_M2                                = 0x01, */
+/* 	    SQ_ALU_OMOD_M4                                = 0x02, */
+/* 	    SQ_ALU_OMOD_D2                                = 0x03, */
+	SQ_ALU_WORD1_OP2__ALU_INST_mask                   = 0x3ff << 8,
+	SQ_ALU_WORD1_OP2__ALU_INST_shift                  = 8,
+/* 	    SQ_OP2_INST_ADD                               = 0x00, */
+/* 	    SQ_OP2_INST_MUL                               = 0x01, */
+/* 	    SQ_OP2_INST_MUL_IEEE                          = 0x02, */
+/* 	    SQ_OP2_INST_MAX                               = 0x03, */
+/* 	    SQ_OP2_INST_MIN                               = 0x04, */
+/* 	    SQ_OP2_INST_MAX_DX10                          = 0x05, */
+/* 	    SQ_OP2_INST_MIN_DX10                          = 0x06, */
+/* 	    SQ_OP2_INST_SETE                              = 0x08, */
+/* 	    SQ_OP2_INST_SETGT                             = 0x09, */
+/* 	    SQ_OP2_INST_SETGE                             = 0x0a, */
+/* 	    SQ_OP2_INST_SETNE                             = 0x0b, */
+/* 	    SQ_OP2_INST_SETE_DX10                         = 0x0c, */
+/* 	    SQ_OP2_INST_SETGT_DX10                        = 0x0d, */
+/* 	    SQ_OP2_INST_SETGE_DX10                        = 0x0e, */
+/* 	    SQ_OP2_INST_SETNE_DX10                        = 0x0f, */
+/* 	    SQ_OP2_INST_FRACT                             = 0x10, */
+/* 	    SQ_OP2_INST_TRUNC                             = 0x11, */
+/* 	    SQ_OP2_INST_CEIL                              = 0x12, */
+/* 	    SQ_OP2_INST_RNDNE                             = 0x13, */
+/* 	    SQ_OP2_INST_FLOOR                             = 0x14, */
+/* 	    SQ_OP2_INST_MOVA                              = 0x15, */
+/* 	    SQ_OP2_INST_MOVA_FLOOR                        = 0x16, */
+/* 	    SQ_OP2_INST_MOVA_INT                          = 0x18, */
+/* 	    SQ_OP2_INST_MOV                               = 0x19, */
+/* 	    SQ_OP2_INST_NOP                               = 0x1a, */
+/* 	    SQ_OP2_INST_PRED_SETGT_UINT                   = 0x1e, */
+/* 	    SQ_OP2_INST_PRED_SETGE_UINT                   = 0x1f, */
+/* 	    SQ_OP2_INST_PRED_SETE                         = 0x20, */
+/* 	    SQ_OP2_INST_PRED_SETGT                        = 0x21, */
+/* 	    SQ_OP2_INST_PRED_SETGE                        = 0x22, */
+/* 	    SQ_OP2_INST_PRED_SETNE                        = 0x23, */
+/* 	    SQ_OP2_INST_PRED_SET_INV                      = 0x24, */
+/* 	    SQ_OP2_INST_PRED_SET_POP                      = 0x25, */
+/* 	    SQ_OP2_INST_PRED_SET_CLR                      = 0x26, */
+/* 	    SQ_OP2_INST_PRED_SET_RESTORE                  = 0x27, */
+/* 	    SQ_OP2_INST_PRED_SETE_PUSH                    = 0x28, */
+/* 	    SQ_OP2_INST_PRED_SETGT_PUSH                   = 0x29, */
+/* 	    SQ_OP2_INST_PRED_SETGE_PUSH                   = 0x2a, */
+/* 	    SQ_OP2_INST_PRED_SETNE_PUSH                   = 0x2b, */
+/* 	    SQ_OP2_INST_KILLE                             = 0x2c, */
+/* 	    SQ_OP2_INST_KILLGT                            = 0x2d, */
+/* 	    SQ_OP2_INST_KILLGE                            = 0x2e, */
+/* 	    SQ_OP2_INST_KILLNE                            = 0x2f, */
+/* 	    SQ_OP2_INST_AND_INT                           = 0x30, */
+/* 	    SQ_OP2_INST_OR_INT                            = 0x31, */
+/* 	    SQ_OP2_INST_XOR_INT                           = 0x32, */
+/* 	    SQ_OP2_INST_NOT_INT                           = 0x33, */
+/* 	    SQ_OP2_INST_ADD_INT                           = 0x34, */
+/* 	    SQ_OP2_INST_SUB_INT                           = 0x35, */
+/* 	    SQ_OP2_INST_MAX_INT                           = 0x36, */
+/* 	    SQ_OP2_INST_MIN_INT                           = 0x37, */
+/* 	    SQ_OP2_INST_MAX_UINT                          = 0x38, */
+/* 	    SQ_OP2_INST_MIN_UINT                          = 0x39, */
+/* 	    SQ_OP2_INST_SETE_INT                          = 0x3a, */
+/* 	    SQ_OP2_INST_SETGT_INT                         = 0x3b, */
+/* 	    SQ_OP2_INST_SETGE_INT                         = 0x3c, */
+/* 	    SQ_OP2_INST_SETNE_INT                         = 0x3d, */
+/* 	    SQ_OP2_INST_SETGT_UINT                        = 0x3e, */
+/* 	    SQ_OP2_INST_SETGE_UINT                        = 0x3f, */
+/* 	    SQ_OP2_INST_KILLGT_UINT                       = 0x40, */
+/* 	    SQ_OP2_INST_KILLGE_UINT                       = 0x41, */
+/* 	    SQ_OP2_INST_PRED_SETE_INT                     = 0x42, */
+/* 	    SQ_OP2_INST_PRED_SETGT_INT                    = 0x43, */
+/* 	    SQ_OP2_INST_PRED_SETGE_INT                    = 0x44, */
+/* 	    SQ_OP2_INST_PRED_SETNE_INT                    = 0x45, */
+/* 	    SQ_OP2_INST_KILLE_INT                         = 0x46, */
+/* 	    SQ_OP2_INST_KILLGT_INT                        = 0x47, */
+/* 	    SQ_OP2_INST_KILLGE_INT                        = 0x48, */
+/* 	    SQ_OP2_INST_KILLNE_INT                        = 0x49, */
+/* 	    SQ_OP2_INST_PRED_SETE_PUSH_INT                = 0x4a, */
+/* 	    SQ_OP2_INST_PRED_SETGT_PUSH_INT               = 0x4b, */
+/* 	    SQ_OP2_INST_PRED_SETGE_PUSH_INT               = 0x4c, */
+/* 	    SQ_OP2_INST_PRED_SETNE_PUSH_INT               = 0x4d, */
+/* 	    SQ_OP2_INST_PRED_SETLT_PUSH_INT               = 0x4e, */
+/* 	    SQ_OP2_INST_PRED_SETLE_PUSH_INT               = 0x4f, */
+/* 	    SQ_OP2_INST_DOT4                              = 0x50, */
+/* 	    SQ_OP2_INST_DOT4_IEEE                         = 0x51, */
+/* 	    SQ_OP2_INST_CUBE                              = 0x52, */
+/* 	    SQ_OP2_INST_MAX4                              = 0x53, */
+/* 	    SQ_OP2_INST_MOVA_GPR_INT                      = 0x60, */
+/* 	    SQ_OP2_INST_EXP_IEEE                          = 0x61, */
+/* 	    SQ_OP2_INST_LOG_CLAMPED                       = 0x62, */
+/* 	    SQ_OP2_INST_LOG_IEEE                          = 0x63, */
+/* 	    SQ_OP2_INST_RECIP_CLAMPED                     = 0x64, */
+/* 	    SQ_OP2_INST_RECIP_FF                          = 0x65, */
+/* 	    SQ_OP2_INST_RECIP_IEEE                        = 0x66, */
+/* 	    SQ_OP2_INST_RECIPSQRT_CLAMPED                 = 0x67, */
+/* 	    SQ_OP2_INST_RECIPSQRT_FF                      = 0x68, */
+/* 	    SQ_OP2_INST_RECIPSQRT_IEEE                    = 0x69, */
+/* 	    SQ_OP2_INST_SQRT_IEEE                         = 0x6a, */
+/* 	    SQ_OP2_INST_FLT_TO_INT                        = 0x6b, */
+/* 	    SQ_OP2_INST_INT_TO_FLT                        = 0x6c, */
+/* 	    SQ_OP2_INST_UINT_TO_FLT                       = 0x6d, */
+/* 	    SQ_OP2_INST_SIN                               = 0x6e, */
+/* 	    SQ_OP2_INST_COS                               = 0x6f, */
+/* 	    SQ_OP2_INST_ASHR_INT                          = 0x70, */
+/* 	    SQ_OP2_INST_LSHR_INT                          = 0x71, */
+/* 	    SQ_OP2_INST_LSHL_INT                          = 0x72, */
+/* 	    SQ_OP2_INST_MULLO_INT                         = 0x73, */
+/* 	    SQ_OP2_INST_MULHI_INT                         = 0x74, */
+/* 	    SQ_OP2_INST_MULLO_UINT                        = 0x75, */
+/* 	    SQ_OP2_INST_MULHI_UINT                        = 0x76, */
+/* 	    SQ_OP2_INST_RECIP_INT                         = 0x77, */
+/* 	    SQ_OP2_INST_RECIP_UINT                        = 0x78, */
+/* 	    SQ_OP2_INST_FLT_TO_UINT                       = 0x79, */
+    SQ_CF_WORD1                                           = 0x00008dfc,
+	POP_COUNT_mask                                    = 0x07 << 0,
+	POP_COUNT_shift                                   = 0,
+	CF_CONST_mask                                     = 0x1f << 3,
+	CF_CONST_shift                                    = 3,
+	COND_mask                                         = 0x03 << 8,
+	COND_shift                                        = 8,
+	    SQ_CF_COND_ACTIVE                             = 0x00,
+	    SQ_CF_COND_FALSE                              = 0x01,
+	    SQ_CF_COND_BOOL                               = 0x02,
+	    SQ_CF_COND_NOT_BOOL                           = 0x03,
+	SQ_CF_WORD1__COUNT_mask                           = 0x07 << 10,
+	SQ_CF_WORD1__COUNT_shift                          = 10,
+	CALL_COUNT_mask                                   = 0x3f << 13,
+	CALL_COUNT_shift                                  = 13,
+	COUNT_3_bit                                       = 1 << 19,
+/* 	END_OF_PROGRAM_bit                                = 1 << 21, */
+/* 	VALID_PIXEL_MODE_bit                              = 1 << 22, */
+	SQ_CF_WORD1__CF_INST_mask                         = 0x7f << 23,
+	SQ_CF_WORD1__CF_INST_shift                        = 23,
+	    SQ_CF_INST_NOP                                = 0x00,
+	    SQ_CF_INST_TEX                                = 0x01,
+	    SQ_CF_INST_VTX                                = 0x02,
+	    SQ_CF_INST_VTX_TC                             = 0x03,
+	    SQ_CF_INST_LOOP_START                         = 0x04,
+	    SQ_CF_INST_LOOP_END                           = 0x05,
+	    SQ_CF_INST_LOOP_START_DX10                    = 0x06,
+	    SQ_CF_INST_LOOP_START_NO_AL                   = 0x07,
+	    SQ_CF_INST_LOOP_CONTINUE                      = 0x08,
+	    SQ_CF_INST_LOOP_BREAK                         = 0x09,
+	    SQ_CF_INST_JUMP                               = 0x0a,
+	    SQ_CF_INST_PUSH                               = 0x0b,
+	    SQ_CF_INST_PUSH_ELSE                          = 0x0c,
+	    SQ_CF_INST_ELSE                               = 0x0d,
+	    SQ_CF_INST_POP                                = 0x0e,
+	    SQ_CF_INST_POP_JUMP                           = 0x0f,
+	    SQ_CF_INST_POP_PUSH                           = 0x10,
+	    SQ_CF_INST_POP_PUSH_ELSE                      = 0x11,
+	    SQ_CF_INST_CALL                               = 0x12,
+	    SQ_CF_INST_CALL_FS                            = 0x13,
+	    SQ_CF_INST_RETURN                             = 0x14,
+	    SQ_CF_INST_EMIT_VERTEX                        = 0x15,
+	    SQ_CF_INST_EMIT_CUT_VERTEX                    = 0x16,
+	    SQ_CF_INST_CUT_VERTEX                         = 0x17,
+	    SQ_CF_INST_KILL                               = 0x18,
+/* 	WHOLE_QUAD_MODE_bit                               = 1 << 30, */
+/* 	BARRIER_bit                                       = 1 << 31, */
+    SQ_VTX_WORD1_SEM                                      = 0x00008dfc,
+	SEMANTIC_ID_mask                                  = 0xff << 0,
+	SEMANTIC_ID_shift                                 = 0,
+    SQ_TEX_WORD0                                          = 0x00008dfc,
+	TEX_INST_mask                                     = 0x1f << 0,
+	TEX_INST_shift                                    = 0,
+	    SQ_TEX_INST_VTX_FETCH                         = 0x00,
+	    SQ_TEX_INST_VTX_SEMANTIC                      = 0x01,
+	    SQ_TEX_INST_LD                                = 0x03,
+	    SQ_TEX_INST_GET_TEXTURE_RESINFO               = 0x04,
+	    SQ_TEX_INST_GET_NUMBER_OF_SAMPLES             = 0x05,
+	    SQ_TEX_INST_GET_LOD                           = 0x06,
+	    SQ_TEX_INST_GET_GRADIENTS_H                   = 0x07,
+	    SQ_TEX_INST_GET_GRADIENTS_V                   = 0x08,
+	    SQ_TEX_INST_GET_LERP                          = 0x09,
+	    SQ_TEX_INST_RESERVED_10                       = 0x0a,
+	    SQ_TEX_INST_SET_GRADIENTS_H                   = 0x0b,
+	    SQ_TEX_INST_SET_GRADIENTS_V                   = 0x0c,
+	    SQ_TEX_INST_PASS                              = 0x0d,
+	    X_Z_SET_INDEX_FOR_ARRAY_OF_CUBEMAPS           = 0x0e,
+	    SQ_TEX_INST_SAMPLE                            = 0x10,
+	    SQ_TEX_INST_SAMPLE_L                          = 0x11,
+	    SQ_TEX_INST_SAMPLE_LB                         = 0x12,
+	    SQ_TEX_INST_SAMPLE_LZ                         = 0x13,
+	    SQ_TEX_INST_SAMPLE_G                          = 0x14,
+	    SQ_TEX_INST_SAMPLE_G_L                        = 0x15,
+	    SQ_TEX_INST_SAMPLE_G_LB                       = 0x16,
+	    SQ_TEX_INST_SAMPLE_G_LZ                       = 0x17,
+	    SQ_TEX_INST_SAMPLE_C                          = 0x18,
+	    SQ_TEX_INST_SAMPLE_C_L                        = 0x19,
+	    SQ_TEX_INST_SAMPLE_C_LB                       = 0x1a,
+	    SQ_TEX_INST_SAMPLE_C_LZ                       = 0x1b,
+	    SQ_TEX_INST_SAMPLE_C_G                        = 0x1c,
+	    SQ_TEX_INST_SAMPLE_C_G_L                      = 0x1d,
+	    SQ_TEX_INST_SAMPLE_C_G_LB                     = 0x1e,
+	    SQ_TEX_INST_SAMPLE_C_G_LZ                     = 0x1f,
+	BC_FRAC_MODE_bit                                  = 1 << 5,
+/* 	FETCH_WHOLE_QUAD_bit                              = 1 << 7, */
+	RESOURCE_ID_mask                                  = 0xff << 8,
+	RESOURCE_ID_shift                                 = 8,
+/* 	SRC_GPR_mask                                      = 0x7f << 16, */
+/* 	SRC_GPR_shift                                     = 16, */
+/* 	SRC_REL_bit                                       = 1 << 23, */
+	SQ_TEX_WORD0__ALT_CONST_bit                       = 1 << 24,
+    SQ_VTX_WORD1_GPR                                      = 0x00008dfc,
+	SQ_VTX_WORD1_GPR__DST_GPR_mask                    = 0x7f << 0,
+	SQ_VTX_WORD1_GPR__DST_GPR_shift                   = 0,
+	SQ_VTX_WORD1_GPR__DST_REL_bit                     = 1 << 7,
+    SQ_ALU_WORD0                                          = 0x00008dfc,
+	SRC0_SEL_mask                                     = 0x1ff << 0,
+	SRC0_SEL_shift                                    = 0,
+/* 	    SQ_ALU_SRC_0                                  = 0xf8, */
+/* 	    SQ_ALU_SRC_1                                  = 0xf9, */
+/* 	    SQ_ALU_SRC_1_INT                              = 0xfa, */
+/* 	    SQ_ALU_SRC_M_1_INT                            = 0xfb, */
+/* 	    SQ_ALU_SRC_0_5                                = 0xfc, */
+/* 	    SQ_ALU_SRC_LITERAL                            = 0xfd, */
+/* 	    SQ_ALU_SRC_PV                                 = 0xfe, */
+/* 	    SQ_ALU_SRC_PS                                 = 0xff, */
+	SRC0_REL_bit                                      = 1 << 9,
+	SRC0_CHAN_mask                                    = 0x03 << 10,
+	SRC0_CHAN_shift                                   = 10,
+/* 	    SQ_CHAN_X                                     = 0x00, */
+/* 	    SQ_CHAN_Y                                     = 0x01, */
+/* 	    SQ_CHAN_Z                                     = 0x02, */
+/* 	    SQ_CHAN_W                                     = 0x03, */
+	SRC0_NEG_bit                                      = 1 << 12,
+	SRC1_SEL_mask                                     = 0x1ff << 13,
+	SRC1_SEL_shift                                    = 13,
+/* 	    SQ_ALU_SRC_0                                  = 0xf8, */
+/* 	    SQ_ALU_SRC_1                                  = 0xf9, */
+/* 	    SQ_ALU_SRC_1_INT                              = 0xfa, */
+/* 	    SQ_ALU_SRC_M_1_INT                            = 0xfb, */
+/* 	    SQ_ALU_SRC_0_5                                = 0xfc, */
+/* 	    SQ_ALU_SRC_LITERAL                            = 0xfd, */
+/* 	    SQ_ALU_SRC_PV                                 = 0xfe, */
+/* 	    SQ_ALU_SRC_PS                                 = 0xff, */
+	SRC1_REL_bit                                      = 1 << 22,
+	SRC1_CHAN_mask                                    = 0x03 << 23,
+	SRC1_CHAN_shift                                   = 23,
+/* 	    SQ_CHAN_X                                     = 0x00, */
+/* 	    SQ_CHAN_Y                                     = 0x01, */
+/* 	    SQ_CHAN_Z                                     = 0x02, */
+/* 	    SQ_CHAN_W                                     = 0x03, */
+	SRC1_NEG_bit                                      = 1 << 25,
+	INDEX_MODE_mask                                   = 0x07 << 26,
+	INDEX_MODE_shift                                  = 26,
+	    SQ_INDEX_AR_X                                 = 0x00,
+	    SQ_INDEX_AR_Y                                 = 0x01,
+	    SQ_INDEX_AR_Z                                 = 0x02,
+	    SQ_INDEX_AR_W                                 = 0x03,
+	    SQ_INDEX_LOOP                                 = 0x04,
+	PRED_SEL_mask                                     = 0x03 << 29,
+	PRED_SEL_shift                                    = 29,
+	    SQ_PRED_SEL_OFF                               = 0x00,
+	    SQ_PRED_SEL_ZERO                              = 0x02,
+	    SQ_PRED_SEL_ONE                               = 0x03,
+	LAST_bit                                          = 1 << 31,
+    SX_EXPORT_BUFFER_SIZES                                = 0x0000900c,
+	COLOR_BUFFER_SIZE_mask                            = 0xff << 0,
+	COLOR_BUFFER_SIZE_shift                           = 0,
+	POSITION_BUFFER_SIZE_mask                         = 0xff << 8,
+	POSITION_BUFFER_SIZE_shift                        = 8,
+	SMX_BUFFER_SIZE_mask                              = 0xff << 16,
+	SMX_BUFFER_SIZE_shift                             = 16,
+    SX_MEMORY_EXPORT_BASE                                 = 0x00009010,
+    SX_MEMORY_EXPORT_SIZE                                 = 0x00009014,
+    SPI_CONFIG_CNTL                                       = 0x00009100,
+	GPR_WRITE_PRIORITY_mask                           = 0x1f << 0,
+	GPR_WRITE_PRIORITY_shift                          = 0,
+	    X_PRIORITY_ORDER                              = 0x00,
+	    X_PRIORITY_ORDER_VS                           = 0x01,
+	DISABLE_INTERP_1_bit                              = 1 << 5,
+	DEBUG_THREAD_TYPE_SEL_mask                        = 0x03 << 6,
+	DEBUG_THREAD_TYPE_SEL_shift                       = 6,
+	DEBUG_GROUP_SEL_mask                              = 0x1f << 8,
+	DEBUG_GROUP_SEL_shift                             = 8,
+	DEBUG_GRBM_OVERRIDE_bit                           = 1 << 13,
+    SPI_CONFIG_CNTL_1                                     = 0x0000913c,
+	VTX_DONE_DELAY_mask                               = 0x0f << 0,
+	VTX_DONE_DELAY_shift                              = 0,
+	    X_DELAY_10_CLKS                               = 0x00,
+	    X_DELAY_11_CLKS                               = 0x01,
+	    X_DELAY_12_CLKS                               = 0x02,
+	    X_DELAY_13_CLKS                               = 0x03,
+	    X_DELAY_14_CLKS                               = 0x04,
+	    X_DELAY_15_CLKS                               = 0x05,
+	    X_DELAY_16_CLKS                               = 0x06,
+	    X_DELAY_17_CLKS                               = 0x07,
+	    X_DELAY_2_CLKS                                = 0x08,
+	    X_DELAY_3_CLKS                                = 0x09,
+	    X_DELAY_4_CLKS                                = 0x0a,
+	    X_DELAY_5_CLKS                                = 0x0b,
+	    X_DELAY_6_CLKS                                = 0x0c,
+	    X_DELAY_7_CLKS                                = 0x0d,
+	    X_DELAY_8_CLKS                                = 0x0e,
+	    X_DELAY_9_CLKS                                = 0x0f,
+	INTERP_ONE_PRIM_PER_ROW_bit                       = 1 << 4,
+    TD_FILTER4                                            = 0x00009400,
+	WEIGHT_1_mask                                     = 0x7ff << 0,
+	WEIGHT_1_shift                                    = 0,
+	WEIGHT_0_mask                                     = 0x7ff << 11,
+	WEIGHT_0_shift                                    = 11,
+	WEIGHT_PAIR_bit                                   = 1 << 22,
+	PHASE_mask                                        = 0x0f << 23,
+	PHASE_shift                                       = 23,
+	DIRECTION_bit                                     = 1 << 27,
+    TD_FILTER4_1                                          = 0x00009404,
+	TD_FILTER4_1_num                                  = 35,
+/* 	WEIGHT_1_mask                                     = 0x7ff << 0, */
+/* 	WEIGHT_1_shift                                    = 0, */
+/* 	WEIGHT_0_mask                                     = 0x7ff << 11, */
+/* 	WEIGHT_0_shift                                    = 11, */
+    TD_CNTL                                               = 0x00009490,
+	SYNC_PHASE_SH_mask                                = 0x03 << 0,
+	SYNC_PHASE_SH_shift                               = 0,
+	SYNC_PHASE_VC_SMX_mask                            = 0x03 << 4,
+	SYNC_PHASE_VC_SMX_shift                           = 4,
+    TD0_CNTL                                              = 0x00009494,
+	TD0_CNTL_num                                      = 4,
+	ID_OVERRIDE_mask                                  = 0x03 << 28,
+	ID_OVERRIDE_shift                                 = 28,
+    TD0_STATUS                                            = 0x000094a4,
+	TD0_STATUS_num                                    = 4,
+	BUSY_bit                                          = 1 << 31,
+    TA_CNTL                                               = 0x00009504,
+	GRADIENT_CREDIT_mask                              = 0x1f << 0,
+	GRADIENT_CREDIT_shift                             = 0,
+	WALKER_CREDIT_mask                                = 0x1f << 8,
+	WALKER_CREDIT_shift                               = 8,
+	ALIGNER_CREDIT_mask                               = 0x1f << 16,
+	ALIGNER_CREDIT_shift                              = 16,
+	TD_FIFO_CREDIT_mask                               = 0x3ff << 22,
+	TD_FIFO_CREDIT_shift                              = 22,
+    TA_CNTL_AUX                                           = 0x00009508,
+	DISABLE_CUBE_WRAP_bit                             = 1 << 0,
+	SYNC_GRADIENT_bit                                 = 1 << 24,
+	SYNC_WALKER_bit                                   = 1 << 25,
+	SYNC_ALIGNER_bit                                  = 1 << 26,
+	BILINEAR_PRECISION_bit                            = 1 << 31,
+    TA0_CNTL                                              = 0x00009510,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA1_CNTL                                              = 0x00009514,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA2_CNTL                                              = 0x00009518,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA3_CNTL                                              = 0x0000951c,
+/* 	ID_OVERRIDE_mask                                  = 0x03 << 28, */
+/* 	ID_OVERRIDE_shift                                 = 28, */
+    TA0_STATUS                                            = 0x00009520,
+	FG_PFIFO_EMPTYB_bit                               = 1 << 12,
+	FG_LFIFO_EMPTYB_bit                               = 1 << 13,
+	FG_SFIFO_EMPTYB_bit                               = 1 << 14,
+	FL_PFIFO_EMPTYB_bit                               = 1 << 16,
+	FL_LFIFO_EMPTYB_bit                               = 1 << 17,
+	FL_SFIFO_EMPTYB_bit                               = 1 << 18,
+	FA_PFIFO_EMPTYB_bit                               = 1 << 20,
+	FA_LFIFO_EMPTYB_bit                               = 1 << 21,
+	FA_SFIFO_EMPTYB_bit                               = 1 << 22,
+	IN_BUSY_bit                                       = 1 << 24,
+	FG_BUSY_bit                                       = 1 << 25,
+	FL_BUSY_bit                                       = 1 << 27,
+	TA_BUSY_bit                                       = 1 << 28,
+	FA_BUSY_bit                                       = 1 << 29,
+	AL_BUSY_bit                                       = 1 << 30,
+/* 	BUSY_bit                                          = 1 << 31, */
+    TA1_STATUS                                            = 0x00009524,
+/* 	FG_PFIFO_EMPTYB_bit                               = 1 << 12, */
+/* 	FG_LFIFO_EMPTYB_bit                               = 1 << 13, */
+/* 	FG_SFIFO_EMPTYB_bit                               = 1 << 14, */
+/* 	FL_PFIFO_EMPTYB_bit                               = 1 << 16, */
+/* 	FL_LFIFO_EMPTYB_bit                               = 1 << 17, */
+/* 	FL_SFIFO_EMPTYB_bit                               = 1 << 18, */
+/* 	FA_PFIFO_EMPTYB_bit                               = 1 << 20, */
+/* 	FA_LFIFO_EMPTYB_bit                               = 1 << 21, */
+/* 	FA_SFIFO_EMPTYB_bit                               = 1 << 22, */
+/* 	IN_BUSY_bit                                       = 1 << 24, */
+/* 	FG_BUSY_bit                                       = 1 << 25, */
+/* 	FL_BUSY_bit                                       = 1 << 27, */
+/* 	TA_BUSY_bit                                       = 1 << 28, */
+/* 	FA_BUSY_bit                                       = 1 << 29, */
+/* 	AL_BUSY_bit                                       = 1 << 30, */
+/* 	BUSY_bit                                          = 1 << 31, */
+    TA2_STATUS                                            = 0x00009528,
+/* 	FG_PFIFO_EMPTYB_bit                               = 1 << 12, */
+/* 	FG_LFIFO_EMPTYB_bit                               = 1 << 13, */
+/* 	FG_SFIFO_EMPTYB_bit                               = 1 << 14, */
+/* 	FL_PFIFO_EMPTYB_bit                               = 1 << 16, */
+/* 	FL_LFIFO_EMPTYB_bit                               = 1 << 17, */
+/* 	FL_SFIFO_EMPTYB_bit                               = 1 << 18, */
+/* 	FA_PFIFO_EMPTYB_bit                               = 1 << 20, */
+/* 	FA_LFIFO_EMPTYB_bit                               = 1 << 21, */
+/* 	FA_SFIFO_EMPTYB_bit                               = 1 << 22, */
+/* 	IN_BUSY_bit                                       = 1 << 24, */
+/* 	FG_BUSY_bit                                       = 1 << 25, */
+/* 	FL_BUSY_bit                                       = 1 << 27, */
+/* 	TA_BUSY_bit                                       = 1 << 28, */
+/* 	FA_BUSY_bit                                       = 1 << 29, */
+/* 	AL_BUSY_bit                                       = 1 << 30, */
+/* 	BUSY_bit                                          = 1 << 31, */
+    TA3_STATUS                                            = 0x0000952c,
+/* 	FG_PFIFO_EMPTYB_bit                               = 1 << 12, */
+/* 	FG_LFIFO_EMPTYB_bit                               = 1 << 13, */
+/* 	FG_SFIFO_EMPTYB_bit                               = 1 << 14, */
+/* 	FL_PFIFO_EMPTYB_bit                               = 1 << 16, */
+/* 	FL_LFIFO_EMPTYB_bit                               = 1 << 17, */
+/* 	FL_SFIFO_EMPTYB_bit                               = 1 << 18, */
+/* 	FA_PFIFO_EMPTYB_bit                               = 1 << 20, */
+/* 	FA_LFIFO_EMPTYB_bit                               = 1 << 21, */
+/* 	FA_SFIFO_EMPTYB_bit                               = 1 << 22, */
+/* 	IN_BUSY_bit                                       = 1 << 24, */
+/* 	FG_BUSY_bit                                       = 1 << 25, */
+/* 	FL_BUSY_bit                                       = 1 << 27, */
+/* 	TA_BUSY_bit                                       = 1 << 28, */
+/* 	FA_BUSY_bit                                       = 1 << 29, */
+/* 	AL_BUSY_bit                                       = 1 << 30, */
+/* 	BUSY_bit                                          = 1 << 31, */
+    TC_STATUS                                             = 0x00009600,
+	TC_BUSY_bit                                       = 1 << 0,
+    TC_INVALIDATE                                         = 0x00009604,
+	START_bit                                         = 1 << 0,
+    TC_CNTL                                               = 0x00009608,
+	FORCE_HIT_bit                                     = 1 << 0,
+	FORCE_MISS_bit                                    = 1 << 1,
+	L2_SIZE_mask                                      = 0x0f << 5,
+	L2_SIZE_shift                                     = 5,
+	    _256K                                         = 0x00,
+	    _224K                                         = 0x01,
+	    _192K                                         = 0x02,
+	    _160K                                         = 0x03,
+	    _128K                                         = 0x04,
+	    _96K                                          = 0x05,
+	    _64K                                          = 0x06,
+	    _32K                                          = 0x07,
+	L2_DISABLE_LATE_HIT_bit                           = 1 << 9,
+	DISABLE_VERT_PERF_bit                             = 1 << 10,
+	DISABLE_INVAL_BUSY_bit                            = 1 << 11,
+	DISABLE_INVAL_SAME_SURFACE_bit                    = 1 << 12,
+	PARTITION_MODE_mask                               = 0x03 << 13,
+	PARTITION_MODE_shift                              = 13,
+	    X_VERTEX                                      = 0x00,
+	MISS_ARB_MODE_bit                                 = 1 << 15,
+	HIT_ARB_MODE_bit                                  = 1 << 16,
+	DISABLE_WRITE_DELAY_bit                           = 1 << 17,
+	HIT_FIFO_DEPTH_bit                                = 1 << 18,
+    VC_CNTL                                               = 0x00009700,
+	L2_INVALIDATE_bit                                 = 1 << 0,
+	RESERVED_bit                                      = 1 << 1,
+	CC_FORCE_MISS_bit                                 = 1 << 2,
+	MI_CHAN_SEL_mask                                  = 0x03 << 3,
+	MI_CHAN_SEL_shift                                 = 3,
+	    X_MC0_USES_CH_0_1                             = 0x00,
+	    X_MC0_USES_CH_0_3                             = 0x01,
+	    X_VC_MC0_IS_ACTIVE                            = 0x02,
+	    X_VC_MC1_IS_DISABLED                          = 0x03,
+	MI_STEER_DISABLE_bit                              = 1 << 5,
+	MI_CREDIT_CTR_mask                                = 0x0f << 6,
+	MI_CREDIT_CTR_shift                               = 6,
+	MI_CREDIT_WE_bit                                  = 1 << 10,
+	MI_REQ_STALL_THLD_mask                            = 0x07 << 11,
+	MI_REQ_STALL_THLD_shift                           = 11,
+	    X_LATENCY_EXCEEDS_399_CLOCKS                  = 0x00,
+	    X_LATENCY_EXCEEDS_415_CLOCKS                  = 0x01,
+	    X_LATENCY_EXCEEDS_431_CLOCKS                  = 0x02,
+	    X_LATENCY_EXCEEDS_447_CLOCKS                  = 0x03,
+	    X_LATENCY_EXCEEDS_463_CLOCKS                  = 0x04,
+	    X_LATENCY_EXCEEDS_479_CLOCKS                  = 0x05,
+	    X_LATENCY_EXCEEDS_495_CLOCKS                  = 0x06,
+	    X_LATENCY_EXCEEDS_511_CLOCKS                  = 0x07,
+	VC_CNTL__MI_TIMESTAMP_RES_mask                    = 0x1f << 14,
+	VC_CNTL__MI_TIMESTAMP_RES_shift                   = 14,
+	    X_1X_SYSTEM_CLOCK                             = 0x00,
+	    X_2X_SYSTEM_CLOCK                             = 0x01,
+	    X_4X_SYSTEM_CLOCK                             = 0x02,
+	    X_8X_SYSTEM_CLOCK                             = 0x03,
+	    X_16X_SYSTEM_CLOCK                            = 0x04,
+	    X_32X_SYSTEM_CLOCK                            = 0x05,
+	    X_64X_SYSTEM_CLOCK                            = 0x06,
+	    X_128X_SYSTEM_CLOCK                           = 0x07,
+	    X_256X_SYSTEM_CLOCK                           = 0x08,
+	    X_512X_SYSTEM_CLOCK                           = 0x09,
+	    X_1024X_SYSTEM_CLOCK                          = 0x0a,
+	    X_2048X_SYSTEM_CLOCK                          = 0x0b,
+	    X_4092X_SYSTEM_CLOCK                          = 0x0c,
+	    X_8192X_SYSTEM_CLOCK                          = 0x0d,
+	    X_16384X_SYSTEM_CLOCK                         = 0x0e,
+	    X_32768X_SYSTEM_CLOCK                         = 0x0f,
+    VC_CNTL_STATUS                                        = 0x00009704,
+	RP_BUSY_bit                                       = 1 << 0,
+	RG_BUSY_bit                                       = 1 << 1,
+	VC_BUSY_bit                                       = 1 << 2,
+	CLAMP_DETECT_bit                                  = 1 << 3,
+    VC_CONFIG                                             = 0x00009718,
+	WRITE_DIS_bit                                     = 1 << 0,
+	GPR_DATA_PHASE_ADJ_mask                           = 0x07 << 1,
+	GPR_DATA_PHASE_ADJ_shift                          = 1,
+	    X_LATENCY_BASE_0_CYCLES                       = 0x00,
+	    X_LATENCY_BASE_1_CYCLES                       = 0x01,
+	    X_LATENCY_BASE_2_CYCLES                       = 0x02,
+	    X_LATENCY_BASE_3_CYCLES                       = 0x03,
+	TD_SIMD_SYNC_ADJ_mask                             = 0x07 << 4,
+	TD_SIMD_SYNC_ADJ_shift                            = 4,
+	    X_0_CYCLES_DELAY                              = 0x00,
+	    X_1_CYCLES_DELAY                              = 0x01,
+	    X_2_CYCLES_DELAY                              = 0x02,
+	    X_3_CYCLES_DELAY                              = 0x03,
+	    X_4_CYCLES_DELAY                              = 0x04,
+	    X_5_CYCLES_DELAY                              = 0x05,
+	    X_6_CYCLES_DELAY                              = 0x06,
+	    X_7_CYCLES_DELAY                              = 0x07,
+    SMX_DC_CTL0                                           = 0x0000a020,
+	WR_GATHER_STREAM0_bit                             = 1 << 0,
+	WR_GATHER_STREAM1_bit                             = 1 << 1,
+	WR_GATHER_STREAM2_bit                             = 1 << 2,
+	WR_GATHER_STREAM3_bit                             = 1 << 3,
+	WR_GATHER_SCRATCH_bit                             = 1 << 4,
+	WR_GATHER_REDUC_BUF_bit                           = 1 << 5,
+	WR_GATHER_RING_BUF_bit                            = 1 << 6,
+	WR_GATHER_F_BUF_bit                               = 1 << 7,
+	DISABLE_CACHES_bit                                = 1 << 8,
+	AUTO_FLUSH_INVAL_EN_bit                           = 1 << 10,
+	AUTO_FLUSH_EN_bit                                 = 1 << 11,
+	AUTO_FLUSH_CNT_mask                               = 0xffff << 12,
+	AUTO_FLUSH_CNT_shift                              = 12,
+	MC_RD_STALL_FACTOR_mask                           = 0x03 << 28,
+	MC_RD_STALL_FACTOR_shift                          = 28,
+	MC_WR_STALL_FACTOR_mask                           = 0x03 << 30,
+	MC_WR_STALL_FACTOR_shift                          = 30,
+    SMX_DC_CTL1                                           = 0x0000a024,
+	OP_FIFO_SKID_mask                                 = 0x7f << 0,
+	OP_FIFO_SKID_shift                                = 0,
+	CACHE_LINE_SIZE_bit                               = 1 << 8,
+	MULTI_FLUSH_MODE_bit                              = 1 << 9,
+	MULTI_FLUSH_REQ_ABORT_IDX_FIFO_SKID_mask          = 0x0f << 10,
+	MULTI_FLUSH_REQ_ABORT_IDX_FIFO_SKID_shift         = 10,
+	DISABLE_WR_GATHER_RD_HIT_FORCE_EVICT_bit          = 1 << 16,
+	DISABLE_WR_GATHER_RD_HIT_COMP_VLDS_CHECK_bit      = 1 << 17,
+	DISABLE_FLUSH_ES_ALSO_INVALS_bit                  = 1 << 18,
+	DISABLE_FLUSH_GS_ALSO_INVALS_bit                  = 1 << 19,
+    SMX_DC_CTL2                                           = 0x0000a028,
+	INVALIDATE_CACHES_bit                             = 1 << 0,
+	CACHES_INVALID_bit                                = 1 << 1,
+	CACHES_DIRTY_bit                                  = 1 << 2,
+	FLUSH_ALL_bit                                     = 1 << 4,
+	FLUSH_GS_THREADS_bit                              = 1 << 8,
+	FLUSH_ES_THREADS_bit                              = 1 << 9,
+    SMX_DC_MC_INTF_CTL                                    = 0x0000a02c,
+	MC_RD_REQ_CRED_mask                               = 0xff << 0,
+	MC_RD_REQ_CRED_shift                              = 0,
+	MC_WR_REQ_CRED_mask                               = 0xff << 16,
+	MC_WR_REQ_CRED_shift                              = 16,
+    TD_PS_SAMPLER0_BORDER_RED                             = 0x0000a400,
+	TD_PS_SAMPLER0_BORDER_RED_num                     = 18,
+	TD_PS_SAMPLER0_BORDER_RED_offset                  = 16,
+    TD_PS_SAMPLER0_BORDER_GREEN                           = 0x0000a404,
+	TD_PS_SAMPLER0_BORDER_GREEN_num                   = 18,
+	TD_PS_SAMPLER0_BORDER_GREEN_offset                = 16,
+    TD_PS_SAMPLER0_BORDER_BLUE                            = 0x0000a408,
+	TD_PS_SAMPLER0_BORDER_BLUE_num                    = 18,
+	TD_PS_SAMPLER0_BORDER_BLUE_offset                 = 16,
+    TD_PS_SAMPLER0_BORDER_ALPHA                           = 0x0000a40c,
+	TD_PS_SAMPLER0_BORDER_ALPHA_num                   = 18,
+	TD_PS_SAMPLER0_BORDER_ALPHA_offset                = 16,
+    TD_VS_SAMPLER0_BORDER_RED                             = 0x0000a600,
+	TD_VS_SAMPLER0_BORDER_RED_num                     = 18,
+	TD_VS_SAMPLER0_BORDER_RED_offset                  = 16,
+    TD_VS_SAMPLER0_BORDER_GREEN                           = 0x0000a604,
+	TD_VS_SAMPLER0_BORDER_GREEN_num                   = 18,
+	TD_VS_SAMPLER0_BORDER_GREEN_offset                = 16,
+    TD_VS_SAMPLER0_BORDER_BLUE                            = 0x0000a608,
+	TD_VS_SAMPLER0_BORDER_BLUE_num                    = 18,
+	TD_VS_SAMPLER0_BORDER_BLUE_offset                 = 16,
+    TD_VS_SAMPLER0_BORDER_ALPHA                           = 0x0000a60c,
+	TD_VS_SAMPLER0_BORDER_ALPHA_num                   = 18,
+	TD_VS_SAMPLER0_BORDER_ALPHA_offset                = 16,
+    TD_GS_SAMPLER0_BORDER_RED                             = 0x0000a800,
+	TD_GS_SAMPLER0_BORDER_RED_num                     = 18,
+	TD_GS_SAMPLER0_BORDER_RED_offset                  = 16,
+    TD_GS_SAMPLER0_BORDER_GREEN                           = 0x0000a804,
+	TD_GS_SAMPLER0_BORDER_GREEN_num                   = 18,
+	TD_GS_SAMPLER0_BORDER_GREEN_offset                = 16,
+    TD_GS_SAMPLER0_BORDER_BLUE                            = 0x0000a808,
+	TD_GS_SAMPLER0_BORDER_BLUE_num                    = 18,
+	TD_GS_SAMPLER0_BORDER_BLUE_offset                 = 16,
+    TD_GS_SAMPLER0_BORDER_ALPHA                           = 0x0000a80c,
+	TD_GS_SAMPLER0_BORDER_ALPHA_num                   = 18,
+	TD_GS_SAMPLER0_BORDER_ALPHA_offset                = 16,
+    TD_PS_SAMPLER0_CLEARTYPE_KERNEL                       = 0x0000aa00,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL_num               = 18,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__WIDTH_mask       = 0x07 << 0,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__WIDTH_shift      = 0,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__HEIGHT_mask      = 0x07 << 3,
+	TD_PS_SAMPLER0_CLEARTYPE_KERNEL__HEIGHT_shift     = 3,
+    DB_DEPTH_SIZE                                         = 0x00028000,
+	PITCH_TILE_MAX_mask                               = 0x3ff << 0,
+	PITCH_TILE_MAX_shift                              = 0,
+	SLICE_TILE_MAX_mask                               = 0xfffff << 10,
+	SLICE_TILE_MAX_shift                              = 10,
+    DB_DEPTH_VIEW                                         = 0x00028004,
+	SLICE_START_mask                                  = 0x7ff << 0,
+	SLICE_START_shift                                 = 0,
+	SLICE_MAX_mask                                    = 0x7ff << 13,
+	SLICE_MAX_shift                                   = 13,
+    DB_DEPTH_BASE                                         = 0x0002800c,
+    DB_DEPTH_INFO                                         = 0x00028010,
+	DB_DEPTH_INFO__FORMAT_mask                        = 0x07 << 0,
+	DB_DEPTH_INFO__FORMAT_shift                       = 0,
+	    DEPTH_INVALID                                 = 0x00,
+	    DEPTH_16                                      = 0x01,
+	    DEPTH_X8_24                                   = 0x02,
+	    DEPTH_8_24                                    = 0x03,
+	    DEPTH_X8_24_FLOAT                             = 0x04,
+	    DEPTH_8_24_FLOAT                              = 0x05,
+	    DEPTH_32_FLOAT                                = 0x06,
+	    DEPTH_X24_8_32_FLOAT                          = 0x07,
+	DB_DEPTH_INFO__READ_SIZE_bit                      = 1 << 3,
+	DB_DEPTH_INFO__ARRAY_MODE_mask                    = 0x0f << 15,
+	DB_DEPTH_INFO__ARRAY_MODE_shift                   = 15,
+	    ARRAY_2D_TILED_THIN1                          = 0x04,
+	TILE_SURFACE_ENABLE_bit                           = 1 << 25,
+	TILE_COMPACT_bit                                  = 1 << 26,
+	ZRANGE_PRECISION_bit                              = 1 << 31,
+    DB_HTILE_DATA_BASE                                    = 0x00028014,
+    DB_STENCIL_CLEAR                                      = 0x00028028,
+	DB_STENCIL_CLEAR__CLEAR_mask                      = 0xff << 0,
+	DB_STENCIL_CLEAR__CLEAR_shift                     = 0,
+	MIN_mask                                          = 0xff << 16,
+	MIN_shift                                         = 16,
+    DB_DEPTH_CLEAR                                        = 0x0002802c,
+    PA_SC_SCREEN_SCISSOR_TL                               = 0x00028030,
+	PA_SC_SCREEN_SCISSOR_TL__TL_X_mask                = 0x7fff << 0,
+	PA_SC_SCREEN_SCISSOR_TL__TL_X_shift               = 0,
+	PA_SC_SCREEN_SCISSOR_TL__TL_Y_mask                = 0x7fff << 16,
+	PA_SC_SCREEN_SCISSOR_TL__TL_Y_shift               = 16,
+    PA_SC_SCREEN_SCISSOR_BR                               = 0x00028034,
+	PA_SC_SCREEN_SCISSOR_BR__BR_X_mask                = 0x7fff << 0,
+	PA_SC_SCREEN_SCISSOR_BR__BR_X_shift               = 0,
+	PA_SC_SCREEN_SCISSOR_BR__BR_Y_mask                = 0x7fff << 16,
+	PA_SC_SCREEN_SCISSOR_BR__BR_Y_shift               = 16,
+    CB_COLOR0_BASE                                        = 0x00028040,
+	CB_COLOR0_BASE_num                                = 8,
+    CB_COLOR0_SIZE                                        = 0x00028060,
+	CB_COLOR0_SIZE_num                                = 8,
+/* 	PITCH_TILE_MAX_mask                               = 0x3ff << 0, */
+/* 	PITCH_TILE_MAX_shift                              = 0, */
+/* 	SLICE_TILE_MAX_mask                               = 0xfffff << 10, */
+/* 	SLICE_TILE_MAX_shift                              = 10, */
+    CB_COLOR0_VIEW                                        = 0x00028080,
+	CB_COLOR0_VIEW_num                                = 8,
+/* 	SLICE_START_mask                                  = 0x7ff << 0, */
+/* 	SLICE_START_shift                                 = 0, */
+/* 	SLICE_MAX_mask                                    = 0x7ff << 13, */
+/* 	SLICE_MAX_shift                                   = 13, */
+    CB_COLOR0_INFO                                        = 0x000280a0,
+	CB_COLOR0_INFO_num                                = 8,
+	ENDIAN_mask                                       = 0x03 << 0,
+	ENDIAN_shift                                      = 0,
+	    ENDIAN_NONE                                   = 0x00,
+	    ENDIAN_8IN16                                  = 0x01,
+	    ENDIAN_8IN32                                  = 0x02,
+	    ENDIAN_8IN64                                  = 0x03,
+	CB_COLOR0_INFO__FORMAT_mask                       = 0x3f << 2,
+	CB_COLOR0_INFO__FORMAT_shift                      = 2,
+	    COLOR_INVALID                                 = 0x00,
+	    COLOR_8                                       = 0x01,
+	    COLOR_4_4                                     = 0x02,
+	    COLOR_3_3_2                                   = 0x03,
+	    COLOR_16                                      = 0x05,
+	    COLOR_16_FLOAT                                = 0x06,
+	    COLOR_8_8                                     = 0x07,
+	    COLOR_5_6_5                                   = 0x08,
+	    COLOR_6_5_5                                   = 0x09,
+	    COLOR_1_5_5_5                                 = 0x0a,
+	    COLOR_4_4_4_4                                 = 0x0b,
+	    COLOR_5_5_5_1                                 = 0x0c,
+	    COLOR_32                                      = 0x0d,
+	    COLOR_32_FLOAT                                = 0x0e,
+	    COLOR_16_16                                   = 0x0f,
+	    COLOR_16_16_FLOAT                             = 0x10,
+	    COLOR_8_24                                    = 0x11,
+	    COLOR_8_24_FLOAT                              = 0x12,
+	    COLOR_24_8                                    = 0x13,
+	    COLOR_24_8_FLOAT                              = 0x14,
+	    COLOR_10_11_11                                = 0x15,
+	    COLOR_10_11_11_FLOAT                          = 0x16,
+	    COLOR_11_11_10                                = 0x17,
+	    COLOR_11_11_10_FLOAT                          = 0x18,
+	    COLOR_2_10_10_10                              = 0x19,
+	    COLOR_8_8_8_8                                 = 0x1a,
+	    COLOR_10_10_10_2                              = 0x1b,
+	    COLOR_X24_8_32_FLOAT                          = 0x1c,
+	    COLOR_32_32                                   = 0x1d,
+	    COLOR_32_32_FLOAT                             = 0x1e,
+	    COLOR_16_16_16_16                             = 0x1f,
+	    COLOR_16_16_16_16_FLOAT                       = 0x20,
+	    COLOR_32_32_32_32                             = 0x22,
+	    COLOR_32_32_32_32_FLOAT                       = 0x23,
+	CB_COLOR0_INFO__ARRAY_MODE_mask                   = 0x0f << 8,
+	CB_COLOR0_INFO__ARRAY_MODE_shift                  = 8,
+	    ARRAY_LINEAR_GENERAL                          = 0x00,
+	    ARRAY_LINEAR_ALIGNED                          = 0x01,
+/* 	    ARRAY_2D_TILED_THIN1                          = 0x04, */
+	NUMBER_TYPE_mask                                  = 0x07 << 12,
+	NUMBER_TYPE_shift                                 = 12,
+	    NUMBER_UNORM                                  = 0x00,
+	    NUMBER_SNORM                                  = 0x01,
+	    NUMBER_USCALED                                = 0x02,
+	    NUMBER_SSCALED                                = 0x03,
+	    NUMBER_UINT                                   = 0x04,
+	    NUMBER_SINT                                   = 0x05,
+	    NUMBER_SRGB                                   = 0x06,
+	    NUMBER_FLOAT                                  = 0x07,
+	CB_COLOR0_INFO__READ_SIZE_bit                     = 1 << 15,
+	COMP_SWAP_mask                                    = 0x03 << 16,
+	COMP_SWAP_shift                                   = 16,
+	    SWAP_STD                                      = 0x00,
+	    SWAP_ALT                                      = 0x01,
+	    SWAP_STD_REV                                  = 0x02,
+	    SWAP_ALT_REV                                  = 0x03,
+	CB_COLOR0_INFO__TILE_MODE_mask                    = 0x03 << 18,
+	CB_COLOR0_INFO__TILE_MODE_shift                   = 18,
+	    TILE_DISABLE                                  = 0x00,
+	    TILE_CLEAR_ENABLE                             = 0x01,
+	    TILE_FRAG_ENABLE                              = 0x02,
+	BLEND_CLAMP_bit                                   = 1 << 20,
+	CLEAR_COLOR_bit                                   = 1 << 21,
+	BLEND_BYPASS_bit                                  = 1 << 22,
+	BLEND_FLOAT32_bit                                 = 1 << 23,
+	SIMPLE_FLOAT_bit                                  = 1 << 24,
+	CB_COLOR0_INFO__ROUND_MODE_bit                    = 1 << 25,
+/* 	TILE_COMPACT_bit                                  = 1 << 26, */
+	SOURCE_FORMAT_bit                                 = 1 << 27,
+    CB_COLOR0_TILE                                        = 0x000280c0,
+	CB_COLOR0_TILE_num                                = 8,
+    CB_COLOR0_FRAG                                        = 0x000280e0,
+	CB_COLOR0_FRAG_num                                = 8,
+    CB_COLOR0_MASK                                        = 0x00028100,
+	CB_COLOR0_MASK_num                                = 8,
+	CMASK_BLOCK_MAX_mask                              = 0xfff << 0,
+	CMASK_BLOCK_MAX_shift                             = 0,
+	FMASK_TILE_MAX_mask                               = 0xfffff << 12,
+	FMASK_TILE_MAX_shift                              = 12,
+    CB_CLEAR_RED                                          = 0x00028120,
+    CB_CLEAR_GREEN                                        = 0x00028124,
+    CB_CLEAR_BLUE                                         = 0x00028128,
+    CB_CLEAR_ALPHA                                        = 0x0002812c,
+    SQ_ALU_CONST_BUFFER_SIZE_PS_0                         = 0x00028140,
+	SQ_ALU_CONST_BUFFER_SIZE_PS_0_num                 = 16,
+	SQ_ALU_CONST_BUFFER_SIZE_PS_0__DATA_mask          = 0x1ff << 0,
+	SQ_ALU_CONST_BUFFER_SIZE_PS_0__DATA_shift         = 0,
+    SQ_ALU_CONST_BUFFER_SIZE_VS_0                         = 0x00028180,
+	SQ_ALU_CONST_BUFFER_SIZE_VS_0_num                 = 16,
+	SQ_ALU_CONST_BUFFER_SIZE_VS_0__DATA_mask          = 0x1ff << 0,
+	SQ_ALU_CONST_BUFFER_SIZE_VS_0__DATA_shift         = 0,
+    SQ_ALU_CONST_BUFFER_SIZE_GS_0                         = 0x000281c0,
+	SQ_ALU_CONST_BUFFER_SIZE_GS_0_num                 = 16,
+	SQ_ALU_CONST_BUFFER_SIZE_GS_0__DATA_mask          = 0x1ff << 0,
+	SQ_ALU_CONST_BUFFER_SIZE_GS_0__DATA_shift         = 0,
+    PA_SC_WINDOW_OFFSET                                   = 0x00028200,
+	WINDOW_X_OFFSET_mask                              = 0x7fff << 0,
+	WINDOW_X_OFFSET_shift                             = 0,
+	WINDOW_Y_OFFSET_mask                              = 0x7fff << 16,
+	WINDOW_Y_OFFSET_shift                             = 16,
+    PA_SC_WINDOW_SCISSOR_TL                               = 0x00028204,
+	PA_SC_WINDOW_SCISSOR_TL__TL_X_mask                = 0x3fff << 0,
+	PA_SC_WINDOW_SCISSOR_TL__TL_X_shift               = 0,
+	PA_SC_WINDOW_SCISSOR_TL__TL_Y_mask                = 0x3fff << 16,
+	PA_SC_WINDOW_SCISSOR_TL__TL_Y_shift               = 16,
+	WINDOW_OFFSET_DISABLE_bit                         = 1 << 31,
+    PA_SC_WINDOW_SCISSOR_BR                               = 0x00028208,
+	PA_SC_WINDOW_SCISSOR_BR__BR_X_mask                = 0x3fff << 0,
+	PA_SC_WINDOW_SCISSOR_BR__BR_X_shift               = 0,
+	PA_SC_WINDOW_SCISSOR_BR__BR_Y_mask                = 0x3fff << 16,
+	PA_SC_WINDOW_SCISSOR_BR__BR_Y_shift               = 16,
+    PA_SC_CLIPRECT_RULE                                   = 0x0002820c,
+	CLIP_RULE_mask                                    = 0xffff << 0,
+	CLIP_RULE_shift                                   = 0,
+    PA_SC_CLIPRECT_0_TL                                   = 0x00028210,
+	PA_SC_CLIPRECT_0_TL_num                           = 4,
+	PA_SC_CLIPRECT_0_TL_offset                        = 8,
+	PA_SC_CLIPRECT_0_TL__TL_X_mask                    = 0x3fff << 0,
+	PA_SC_CLIPRECT_0_TL__TL_X_shift                   = 0,
+	PA_SC_CLIPRECT_0_TL__TL_Y_mask                    = 0x3fff << 16,
+	PA_SC_CLIPRECT_0_TL__TL_Y_shift                   = 16,
+    PA_SC_CLIPRECT_0_BR                                   = 0x00028214,
+	PA_SC_CLIPRECT_0_BR_num                           = 4,
+	PA_SC_CLIPRECT_0_BR_offset                        = 8,
+	PA_SC_CLIPRECT_0_BR__BR_X_mask                    = 0x3fff << 0,
+	PA_SC_CLIPRECT_0_BR__BR_X_shift                   = 0,
+	PA_SC_CLIPRECT_0_BR__BR_Y_mask                    = 0x3fff << 16,
+	PA_SC_CLIPRECT_0_BR__BR_Y_shift                   = 16,
+    CB_TARGET_MASK                                        = 0x00028238,
+	TARGET0_ENABLE_mask                               = 0x0f << 0,
+	TARGET0_ENABLE_shift                              = 0,
+	TARGET1_ENABLE_mask                               = 0x0f << 4,
+	TARGET1_ENABLE_shift                              = 4,
+	TARGET2_ENABLE_mask                               = 0x0f << 8,
+	TARGET2_ENABLE_shift                              = 8,
+	TARGET3_ENABLE_mask                               = 0x0f << 12,
+	TARGET3_ENABLE_shift                              = 12,
+	TARGET4_ENABLE_mask                               = 0x0f << 16,
+	TARGET4_ENABLE_shift                              = 16,
+	TARGET5_ENABLE_mask                               = 0x0f << 20,
+	TARGET5_ENABLE_shift                              = 20,
+	TARGET6_ENABLE_mask                               = 0x0f << 24,
+	TARGET6_ENABLE_shift                              = 24,
+	TARGET7_ENABLE_mask                               = 0x0f << 28,
+	TARGET7_ENABLE_shift                              = 28,
+    CB_SHADER_MASK                                        = 0x0002823c,
+	OUTPUT0_ENABLE_mask                               = 0x0f << 0,
+	OUTPUT0_ENABLE_shift                              = 0,
+	OUTPUT1_ENABLE_mask                               = 0x0f << 4,
+	OUTPUT1_ENABLE_shift                              = 4,
+	OUTPUT2_ENABLE_mask                               = 0x0f << 8,
+	OUTPUT2_ENABLE_shift                              = 8,
+	OUTPUT3_ENABLE_mask                               = 0x0f << 12,
+	OUTPUT3_ENABLE_shift                              = 12,
+	OUTPUT4_ENABLE_mask                               = 0x0f << 16,
+	OUTPUT4_ENABLE_shift                              = 16,
+	OUTPUT5_ENABLE_mask                               = 0x0f << 20,
+	OUTPUT5_ENABLE_shift                              = 20,
+	OUTPUT6_ENABLE_mask                               = 0x0f << 24,
+	OUTPUT6_ENABLE_shift                              = 24,
+	OUTPUT7_ENABLE_mask                               = 0x0f << 28,
+	OUTPUT7_ENABLE_shift                              = 28,
+    PA_SC_GENERIC_SCISSOR_TL                              = 0x00028240,
+	PA_SC_GENERIC_SCISSOR_TL__TL_X_mask               = 0x3fff << 0,
+	PA_SC_GENERIC_SCISSOR_TL__TL_X_shift              = 0,
+	PA_SC_GENERIC_SCISSOR_TL__TL_Y_mask               = 0x3fff << 16,
+	PA_SC_GENERIC_SCISSOR_TL__TL_Y_shift              = 16,
+/* 	WINDOW_OFFSET_DISABLE_bit                         = 1 << 31, */
+    PA_SC_GENERIC_SCISSOR_BR                              = 0x00028244,
+	PA_SC_GENERIC_SCISSOR_BR__BR_X_mask               = 0x3fff << 0,
+	PA_SC_GENERIC_SCISSOR_BR__BR_X_shift              = 0,
+	PA_SC_GENERIC_SCISSOR_BR__BR_Y_mask               = 0x3fff << 16,
+	PA_SC_GENERIC_SCISSOR_BR__BR_Y_shift              = 16,
+    PA_SC_VPORT_SCISSOR_0_TL                              = 0x00028250,
+	PA_SC_VPORT_SCISSOR_0_TL_num                      = 16,
+	PA_SC_VPORT_SCISSOR_0_TL_offset                   = 8,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_X_mask               = 0x3fff << 0,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_X_shift              = 0,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_Y_mask               = 0x3fff << 16,
+	PA_SC_VPORT_SCISSOR_0_TL__TL_Y_shift              = 16,
+/* 	WINDOW_OFFSET_DISABLE_bit                         = 1 << 31, */
+    PA_SC_VPORT_SCISSOR_0_BR                              = 0x00028254,
+	PA_SC_VPORT_SCISSOR_0_BR_num                      = 16,
+	PA_SC_VPORT_SCISSOR_0_BR_offset                   = 8,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_X_mask               = 0x3fff << 0,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_X_shift              = 0,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask               = 0x3fff << 16,
+	PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift              = 16,
+    PA_SC_VPORT_ZMIN_0                                    = 0x000282d0,
+	PA_SC_VPORT_ZMIN_0_num                            = 16,
+	PA_SC_VPORT_ZMIN_0_offset                         = 8,
+    PA_SC_VPORT_ZMAX_0                                    = 0x000282d4,
+	PA_SC_VPORT_ZMAX_0_num                            = 16,
+	PA_SC_VPORT_ZMAX_0_offset                         = 8,
+    SX_MISC                                               = 0x00028350,
+	MULTIPASS_bit                                     = 1 << 0,
+    SQ_VTX_SEMANTIC_0                                     = 0x00028380,
+	SQ_VTX_SEMANTIC_0_num                             = 32,
+/* 	SEMANTIC_ID_mask                                  = 0xff << 0, */
+/* 	SEMANTIC_ID_shift                                 = 0, */
+    VGT_MAX_VTX_INDX                                      = 0x00028400,
+    VGT_MIN_VTX_INDX                                      = 0x00028404,
+    VGT_INDX_OFFSET                                       = 0x00028408,
+    VGT_MULTI_PRIM_IB_RESET_INDX                          = 0x0002840c,
+    SX_ALPHA_TEST_CONTROL                                 = 0x00028410,
+	ALPHA_FUNC_mask                                   = 0x07 << 0,
+	ALPHA_FUNC_shift                                  = 0,
+	    REF_NEVER                                     = 0x00,
+	    REF_LESS                                      = 0x01,
+	    REF_EQUAL                                     = 0x02,
+	    REF_LEQUAL                                    = 0x03,
+	    REF_GREATER                                   = 0x04,
+	    REF_NOTEQUAL                                  = 0x05,
+	    REF_GEQUAL                                    = 0x06,
+	    REF_ALWAYS                                    = 0x07,
+	ALPHA_TEST_ENABLE_bit                             = 1 << 3,
+	ALPHA_TEST_BYPASS_bit                             = 1 << 8,
+    CB_BLEND_RED                                          = 0x00028414,
+    CB_BLEND_GREEN                                        = 0x00028418,
+    CB_BLEND_BLUE                                         = 0x0002841c,
+    CB_BLEND_ALPHA                                        = 0x00028420,
+    CB_FOG_RED                                            = 0x00028424,
+    CB_FOG_GREEN                                          = 0x00028428,
+    CB_FOG_BLUE                                           = 0x0002842c,
+    DB_STENCILREFMASK                                     = 0x00028430,
+	STENCILREF_mask                                   = 0xff << 0,
+	STENCILREF_shift                                  = 0,
+	STENCILMASK_mask                                  = 0xff << 8,
+	STENCILMASK_shift                                 = 8,
+	STENCILWRITEMASK_mask                             = 0xff << 16,
+	STENCILWRITEMASK_shift                            = 16,
+    DB_STENCILREFMASK_BF                                  = 0x00028434,
+	STENCILREF_BF_mask                                = 0xff << 0,
+	STENCILREF_BF_shift                               = 0,
+	STENCILMASK_BF_mask                               = 0xff << 8,
+	STENCILMASK_BF_shift                              = 8,
+	STENCILWRITEMASK_BF_mask                          = 0xff << 16,
+	STENCILWRITEMASK_BF_shift                         = 16,
+    SX_ALPHA_REF                                          = 0x00028438,
+    PA_CL_VPORT_XSCALE_0                                  = 0x0002843c,
+	PA_CL_VPORT_XSCALE_0_num                          = 16,
+	PA_CL_VPORT_XSCALE_0_offset                       = 24,
+    PA_CL_VPORT_XOFFSET_0                                 = 0x00028440,
+	PA_CL_VPORT_XOFFSET_0_num                         = 16,
+	PA_CL_VPORT_XOFFSET_0_offset                      = 24,
+    PA_CL_VPORT_YSCALE_0                                  = 0x00028444,
+	PA_CL_VPORT_YSCALE_0_num                          = 16,
+	PA_CL_VPORT_YSCALE_0_offset                       = 24,
+    PA_CL_VPORT_YOFFSET_0                                 = 0x00028448,
+	PA_CL_VPORT_YOFFSET_0_num                         = 16,
+	PA_CL_VPORT_YOFFSET_0_offset                      = 24,
+    PA_CL_VPORT_ZSCALE_0                                  = 0x0002844c,
+	PA_CL_VPORT_ZSCALE_0_num                          = 16,
+	PA_CL_VPORT_ZSCALE_0_offset                       = 24,
+    PA_CL_VPORT_ZOFFSET_0                                 = 0x00028450,
+	PA_CL_VPORT_ZOFFSET_0_num                         = 16,
+	PA_CL_VPORT_ZOFFSET_0_offset                      = 24,
+    SPI_VS_OUT_ID_0                                       = 0x00028614,
+	SPI_VS_OUT_ID_0_num                               = 10,
+	SEMANTIC_0_mask                                   = 0xff << 0,
+	SEMANTIC_0_shift                                  = 0,
+	SEMANTIC_1_mask                                   = 0xff << 8,
+	SEMANTIC_1_shift                                  = 8,
+	SEMANTIC_2_mask                                   = 0xff << 16,
+	SEMANTIC_2_shift                                  = 16,
+	SEMANTIC_3_mask                                   = 0xff << 24,
+	SEMANTIC_3_shift                                  = 24,
+    SPI_PS_INPUT_CNTL_0                                   = 0x00028644,
+	SPI_PS_INPUT_CNTL_0_num                           = 32,
+	SEMANTIC_mask                                     = 0xff << 0,
+	SEMANTIC_shift                                    = 0,
+	DEFAULT_VAL_mask                                  = 0x03 << 8,
+	DEFAULT_VAL_shift                                 = 8,
+	    X_0_0F                                        = 0x00,
+	FLAT_SHADE_bit                                    = 1 << 10,
+	SEL_CENTROID_bit                                  = 1 << 11,
+	SEL_LINEAR_bit                                    = 1 << 12,
+	CYL_WRAP_mask                                     = 0x0f << 13,
+	CYL_WRAP_shift                                    = 13,
+	PT_SPRITE_TEX_bit                                 = 1 << 17,
+	SEL_SAMPLE_bit                                    = 1 << 18,
+    SPI_VS_OUT_CONFIG                                     = 0x000286c4,
+	VS_PER_COMPONENT_bit                              = 1 << 0,
+	VS_EXPORT_COUNT_mask                              = 0x1f << 1,
+	VS_EXPORT_COUNT_shift                             = 1,
+	VS_EXPORTS_FOG_bit                                = 1 << 8,
+	VS_OUT_FOG_VEC_ADDR_mask                          = 0x1f << 9,
+	VS_OUT_FOG_VEC_ADDR_shift                         = 9,
+    SPI_PS_IN_CONTROL_0                                   = 0x000286cc,
+	NUM_INTERP_mask                                   = 0x3f << 0,
+	NUM_INTERP_shift                                  = 0,
+	POSITION_ENA_bit                                  = 1 << 8,
+	POSITION_CENTROID_bit                             = 1 << 9,
+	POSITION_ADDR_mask                                = 0x1f << 10,
+	POSITION_ADDR_shift                               = 10,
+	PARAM_GEN_mask                                    = 0x0f << 15,
+	PARAM_GEN_shift                                   = 15,
+	PARAM_GEN_ADDR_mask                               = 0x7f << 19,
+	PARAM_GEN_ADDR_shift                              = 19,
+	BARYC_SAMPLE_CNTL_mask                            = 0x03 << 26,
+	BARYC_SAMPLE_CNTL_shift                           = 26,
+	    CENTROIDS_ONLY                                = 0x00,
+	    CENTERS_ONLY                                  = 0x01,
+	    CENTROIDS_AND_CENTERS                         = 0x02,
+	    UNDEF                                         = 0x03,
+	PERSP_GRADIENT_ENA_bit                            = 1 << 28,
+	LINEAR_GRADIENT_ENA_bit                           = 1 << 29,
+	POSITION_SAMPLE_bit                               = 1 << 30,
+	BARYC_AT_SAMPLE_ENA_bit                           = 1 << 31,
+    SPI_PS_IN_CONTROL_1                                   = 0x000286d0,
+	GEN_INDEX_PIX_bit                                 = 1 << 0,
+	GEN_INDEX_PIX_ADDR_mask                           = 0x7f << 1,
+	GEN_INDEX_PIX_ADDR_shift                          = 1,
+	FRONT_FACE_ENA_bit                                = 1 << 8,
+	FRONT_FACE_CHAN_mask                              = 0x03 << 9,
+	FRONT_FACE_CHAN_shift                             = 9,
+	FRONT_FACE_ALL_BITS_bit                           = 1 << 11,
+	FRONT_FACE_ADDR_mask                              = 0x1f << 12,
+	FRONT_FACE_ADDR_shift                             = 12,
+	FOG_ADDR_mask                                     = 0x7f << 17,
+	FOG_ADDR_shift                                    = 17,
+	FIXED_PT_POSITION_ENA_bit                         = 1 << 24,
+	FIXED_PT_POSITION_ADDR_mask                       = 0x1f << 25,
+	FIXED_PT_POSITION_ADDR_shift                      = 25,
+    SPI_INTERP_CONTROL_0                                  = 0x000286d4,
+	FLAT_SHADE_ENA_bit                                = 1 << 0,
+	PNT_SPRITE_ENA_bit                                = 1 << 1,
+	PNT_SPRITE_OVRD_X_mask                            = 0x07 << 2,
+	PNT_SPRITE_OVRD_X_shift                           = 2,
+	    SPI_PNT_SPRITE_SEL_0                          = 0x00,
+	    SPI_PNT_SPRITE_SEL_1                          = 0x01,
+	    SPI_PNT_SPRITE_SEL_S                          = 0x02,
+	    SPI_PNT_SPRITE_SEL_T                          = 0x03,
+	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04,
+	PNT_SPRITE_OVRD_Y_mask                            = 0x07 << 5,
+	PNT_SPRITE_OVRD_Y_shift                           = 5,
+/* 	    SPI_PNT_SPRITE_SEL_0                          = 0x00, */
+/* 	    SPI_PNT_SPRITE_SEL_1                          = 0x01, */
+/* 	    SPI_PNT_SPRITE_SEL_S                          = 0x02, */
+/* 	    SPI_PNT_SPRITE_SEL_T                          = 0x03, */
+/* 	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04, */
+	PNT_SPRITE_OVRD_Z_mask                            = 0x07 << 8,
+	PNT_SPRITE_OVRD_Z_shift                           = 8,
+/* 	    SPI_PNT_SPRITE_SEL_0                          = 0x00, */
+/* 	    SPI_PNT_SPRITE_SEL_1                          = 0x01, */
+/* 	    SPI_PNT_SPRITE_SEL_S                          = 0x02, */
+/* 	    SPI_PNT_SPRITE_SEL_T                          = 0x03, */
+/* 	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04, */
+	PNT_SPRITE_OVRD_W_mask                            = 0x07 << 11,
+	PNT_SPRITE_OVRD_W_shift                           = 11,
+/* 	    SPI_PNT_SPRITE_SEL_0                          = 0x00, */
+/* 	    SPI_PNT_SPRITE_SEL_1                          = 0x01, */
+/* 	    SPI_PNT_SPRITE_SEL_S                          = 0x02, */
+/* 	    SPI_PNT_SPRITE_SEL_T                          = 0x03, */
+/* 	    SPI_PNT_SPRITE_SEL_NONE                       = 0x04, */
+	PNT_SPRITE_TOP_1_bit                              = 1 << 14,
+    SPI_INPUT_Z                                           = 0x000286d8,
+	PROVIDE_Z_TO_SPI_bit                              = 1 << 0,
+    SPI_FOG_CNTL                                          = 0x000286dc,
+	PASS_FOG_THROUGH_PS_bit                           = 1 << 0,
+	PIXEL_FOG_FUNC_mask                               = 0x03 << 1,
+	PIXEL_FOG_FUNC_shift                              = 1,
+	    SPI_FOG_NONE                                  = 0x00,
+	    SPI_FOG_EXP                                   = 0x01,
+	    SPI_FOG_EXP2                                  = 0x02,
+	    SPI_FOG_LINEAR                                = 0x03,
+	PIXEL_FOG_SRC_SEL_bit                             = 1 << 3,
+	VS_FOG_CLAMP_DISABLE_bit                          = 1 << 4,
+    SPI_FOG_FUNC_SCALE                                    = 0x000286e0,
+    SPI_FOG_FUNC_BIAS                                     = 0x000286e4,
+    CB_BLEND0_CONTROL                                     = 0x00028780,
+	CB_BLEND0_CONTROL_num                             = 8,
+	COLOR_SRCBLEND_mask                               = 0x1f << 0,
+	COLOR_SRCBLEND_shift                              = 0,
+	COLOR_COMB_FCN_mask                               = 0x07 << 5,
+	COLOR_COMB_FCN_shift                              = 5,
+	COLOR_DESTBLEND_mask                              = 0x1f << 8,
+	COLOR_DESTBLEND_shift                             = 8,
+	OPACITY_WEIGHT_bit                                = 1 << 13,
+	ALPHA_SRCBLEND_mask                               = 0x1f << 16,
+	ALPHA_SRCBLEND_shift                              = 16,
+	ALPHA_COMB_FCN_mask                               = 0x07 << 21,
+	ALPHA_COMB_FCN_shift                              = 21,
+	ALPHA_DESTBLEND_mask                              = 0x1f << 24,
+	ALPHA_DESTBLEND_shift                             = 24,
+	SEPARATE_ALPHA_BLEND_bit                          = 1 << 29,
+    VGT_DMA_BASE_HI                                       = 0x000287e4,
+	VGT_DMA_BASE_HI__BASE_ADDR_mask                   = 0xff << 0,
+	VGT_DMA_BASE_HI__BASE_ADDR_shift                  = 0,
+    VGT_DMA_BASE                                          = 0x000287e8,
+    VGT_DRAW_INITIATOR                                    = 0x000287f0,
+	SOURCE_SELECT_mask                                = 0x03 << 0,
+	SOURCE_SELECT_shift                               = 0,
+	    DI_SRC_SEL_DMA                                = 0x00,
+	    DI_SRC_SEL_IMMEDIATE                          = 0x01,
+	    DI_SRC_SEL_AUTO_INDEX                         = 0x02,
+	    DI_SRC_SEL_RESERVED                           = 0x03,
+	MAJOR_MODE_mask                                   = 0x03 << 2,
+	MAJOR_MODE_shift                                  = 2,
+	    DI_MAJOR_MODE_0                               = 0x00,
+	    DI_MAJOR_MODE_1                               = 0x01,
+	SPRITE_EN_bit                                     = 1 << 4,
+	NOT_EOP_bit                                       = 1 << 5,
+	USE_OPAQUE_bit                                    = 1 << 6,
+    VGT_IMMED_DATA                                        = 0x000287f4,
+    VGT_EVENT_ADDRESS_REG                                 = 0x000287f8,
+	ADDRESS_LOW_mask                                  = 0xfffffff << 0,
+	ADDRESS_LOW_shift                                 = 0,
+    DB_DEPTH_CONTROL                                      = 0x00028800,
+	STENCIL_ENABLE_bit                                = 1 << 0,
+	Z_ENABLE_bit                                      = 1 << 1,
+	Z_WRITE_ENABLE_bit                                = 1 << 2,
+	ZFUNC_mask                                        = 0x07 << 4,
+	ZFUNC_shift                                       = 4,
+	    FRAG_NEVER                                    = 0x00,
+	    FRAG_LESS                                     = 0x01,
+	    FRAG_EQUAL                                    = 0x02,
+	    FRAG_LEQUAL                                   = 0x03,
+	    FRAG_GREATER                                  = 0x04,
+	    FRAG_NOTEQUAL                                 = 0x05,
+	    FRAG_GEQUAL                                   = 0x06,
+	    FRAG_ALWAYS                                   = 0x07,
+	BACKFACE_ENABLE_bit                               = 1 << 7,
+	STENCILFUNC_mask                                  = 0x07 << 8,
+	STENCILFUNC_shift                                 = 8,
+/* 	    REF_NEVER                                     = 0x00, */
+/* 	    REF_LESS                                      = 0x01, */
+/* 	    REF_EQUAL                                     = 0x02, */
+/* 	    REF_LEQUAL                                    = 0x03, */
+/* 	    REF_GREATER                                   = 0x04, */
+/* 	    REF_NOTEQUAL                                  = 0x05, */
+/* 	    REF_GEQUAL                                    = 0x06, */
+/* 	    REF_ALWAYS                                    = 0x07, */
+	STENCILFAIL_mask                                  = 0x07 << 11,
+	STENCILFAIL_shift                                 = 11,
+	    STENCIL_KEEP                                  = 0x00,
+	    STENCIL_ZERO                                  = 0x01,
+	    STENCIL_REPLACE                               = 0x02,
+	    STENCIL_INCR_CLAMP                            = 0x03,
+	    STENCIL_DECR_CLAMP                            = 0x04,
+	    STENCIL_INVERT                                = 0x05,
+	    STENCIL_INCR_WRAP                             = 0x06,
+	    STENCIL_DECR_WRAP                             = 0x07,
+	STENCILZPASS_mask                                 = 0x07 << 14,
+	STENCILZPASS_shift                                = 14,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILZFAIL_mask                                 = 0x07 << 17,
+	STENCILZFAIL_shift                                = 17,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILFUNC_BF_mask                               = 0x07 << 20,
+	STENCILFUNC_BF_shift                              = 20,
+/* 	    REF_NEVER                                     = 0x00, */
+/* 	    REF_LESS                                      = 0x01, */
+/* 	    REF_EQUAL                                     = 0x02, */
+/* 	    REF_LEQUAL                                    = 0x03, */
+/* 	    REF_GREATER                                   = 0x04, */
+/* 	    REF_NOTEQUAL                                  = 0x05, */
+/* 	    REF_GEQUAL                                    = 0x06, */
+/* 	    REF_ALWAYS                                    = 0x07, */
+	STENCILFAIL_BF_mask                               = 0x07 << 23,
+	STENCILFAIL_BF_shift                              = 23,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILZPASS_BF_mask                              = 0x07 << 26,
+	STENCILZPASS_BF_shift                             = 26,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+	STENCILZFAIL_BF_mask                              = 0x07 << 29,
+	STENCILZFAIL_BF_shift                             = 29,
+/* 	    STENCIL_KEEP                                  = 0x00, */
+/* 	    STENCIL_ZERO                                  = 0x01, */
+/* 	    STENCIL_REPLACE                               = 0x02, */
+/* 	    STENCIL_INCR_CLAMP                            = 0x03, */
+/* 	    STENCIL_DECR_CLAMP                            = 0x04, */
+/* 	    STENCIL_INVERT                                = 0x05, */
+/* 	    STENCIL_INCR_WRAP                             = 0x06, */
+/* 	    STENCIL_DECR_WRAP                             = 0x07, */
+    CB_BLEND_CONTROL                                      = 0x00028804,
+/* 	COLOR_SRCBLEND_mask                               = 0x1f << 0, */
+/* 	COLOR_SRCBLEND_shift                              = 0, */
+	    BLEND_ZERO                                    = 0x00,
+	    BLEND_ONE                                     = 0x01,
+	    BLEND_SRC_COLOR                               = 0x02,
+	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03,
+	    BLEND_SRC_ALPHA                               = 0x04,
+	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05,
+	    BLEND_DST_ALPHA                               = 0x06,
+	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07,
+	    BLEND_DST_COLOR                               = 0x08,
+	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09,
+	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a,
+	    BLEND_BOTH_SRC_ALPHA                          = 0x0b,
+	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c,
+	    BLEND_CONSTANT_COLOR                          = 0x0d,
+	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e,
+	    BLEND_SRC1_COLOR                              = 0x0f,
+	    BLEND_INV_SRC1_COLOR                          = 0x10,
+	    BLEND_SRC1_ALPHA                              = 0x11,
+	    BLEND_INV_SRC1_ALPHA                          = 0x12,
+	    BLEND_CONSTANT_ALPHA                          = 0x13,
+	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14,
+/* 	COLOR_COMB_FCN_mask                               = 0x07 << 5, */
+/* 	COLOR_COMB_FCN_shift                              = 5, */
+	    COMB_DST_PLUS_SRC                             = 0x00,
+	    COMB_SRC_MINUS_DST                            = 0x01,
+	    COMB_MIN_DST_SRC                              = 0x02,
+	    COMB_MAX_DST_SRC                              = 0x03,
+	    COMB_DST_MINUS_SRC                            = 0x04,
+/* 	COLOR_DESTBLEND_mask                              = 0x1f << 8, */
+/* 	COLOR_DESTBLEND_shift                             = 8, */
+/* 	    BLEND_ZERO                                    = 0x00, */
+/* 	    BLEND_ONE                                     = 0x01, */
+/* 	    BLEND_SRC_COLOR                               = 0x02, */
+/* 	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03, */
+/* 	    BLEND_SRC_ALPHA                               = 0x04, */
+/* 	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05, */
+/* 	    BLEND_DST_ALPHA                               = 0x06, */
+/* 	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07, */
+/* 	    BLEND_DST_COLOR                               = 0x08, */
+/* 	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09, */
+/* 	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a, */
+/* 	    BLEND_BOTH_SRC_ALPHA                          = 0x0b, */
+/* 	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c, */
+/* 	    BLEND_CONSTANT_COLOR                          = 0x0d, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e, */
+/* 	    BLEND_SRC1_COLOR                              = 0x0f, */
+/* 	    BLEND_INV_SRC1_COLOR                          = 0x10, */
+/* 	    BLEND_SRC1_ALPHA                              = 0x11, */
+/* 	    BLEND_INV_SRC1_ALPHA                          = 0x12, */
+/* 	    BLEND_CONSTANT_ALPHA                          = 0x13, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14, */
+/* 	OPACITY_WEIGHT_bit                                = 1 << 13, */
+/* 	ALPHA_SRCBLEND_mask                               = 0x1f << 16, */
+/* 	ALPHA_SRCBLEND_shift                              = 16, */
+/* 	    BLEND_ZERO                                    = 0x00, */
+/* 	    BLEND_ONE                                     = 0x01, */
+/* 	    BLEND_SRC_COLOR                               = 0x02, */
+/* 	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03, */
+/* 	    BLEND_SRC_ALPHA                               = 0x04, */
+/* 	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05, */
+/* 	    BLEND_DST_ALPHA                               = 0x06, */
+/* 	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07, */
+/* 	    BLEND_DST_COLOR                               = 0x08, */
+/* 	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09, */
+/* 	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a, */
+/* 	    BLEND_BOTH_SRC_ALPHA                          = 0x0b, */
+/* 	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c, */
+/* 	    BLEND_CONSTANT_COLOR                          = 0x0d, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e, */
+/* 	    BLEND_SRC1_COLOR                              = 0x0f, */
+/* 	    BLEND_INV_SRC1_COLOR                          = 0x10, */
+/* 	    BLEND_SRC1_ALPHA                              = 0x11, */
+/* 	    BLEND_INV_SRC1_ALPHA                          = 0x12, */
+/* 	    BLEND_CONSTANT_ALPHA                          = 0x13, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14, */
+/* 	ALPHA_COMB_FCN_mask                               = 0x07 << 21, */
+/* 	ALPHA_COMB_FCN_shift                              = 21, */
+/* 	    COMB_DST_PLUS_SRC                             = 0x00, */
+/* 	    COMB_SRC_MINUS_DST                            = 0x01, */
+/* 	    COMB_MIN_DST_SRC                              = 0x02, */
+/* 	    COMB_MAX_DST_SRC                              = 0x03, */
+/* 	    COMB_DST_MINUS_SRC                            = 0x04, */
+/* 	ALPHA_DESTBLEND_mask                              = 0x1f << 24, */
+/* 	ALPHA_DESTBLEND_shift                             = 24, */
+/* 	    BLEND_ZERO                                    = 0x00, */
+/* 	    BLEND_ONE                                     = 0x01, */
+/* 	    BLEND_SRC_COLOR                               = 0x02, */
+/* 	    BLEND_ONE_MINUS_SRC_COLOR                     = 0x03, */
+/* 	    BLEND_SRC_ALPHA                               = 0x04, */
+/* 	    BLEND_ONE_MINUS_SRC_ALPHA                     = 0x05, */
+/* 	    BLEND_DST_ALPHA                               = 0x06, */
+/* 	    BLEND_ONE_MINUS_DST_ALPHA                     = 0x07, */
+/* 	    BLEND_DST_COLOR                               = 0x08, */
+/* 	    BLEND_ONE_MINUS_DST_COLOR                     = 0x09, */
+/* 	    BLEND_SRC_ALPHA_SATURATE                      = 0x0a, */
+/* 	    BLEND_BOTH_SRC_ALPHA                          = 0x0b, */
+/* 	    BLEND_BOTH_INV_SRC_ALPHA                      = 0x0c, */
+/* 	    BLEND_CONSTANT_COLOR                          = 0x0d, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_COLOR                = 0x0e, */
+/* 	    BLEND_SRC1_COLOR                              = 0x0f, */
+/* 	    BLEND_INV_SRC1_COLOR                          = 0x10, */
+/* 	    BLEND_SRC1_ALPHA                              = 0x11, */
+/* 	    BLEND_INV_SRC1_ALPHA                          = 0x12, */
+/* 	    BLEND_CONSTANT_ALPHA                          = 0x13, */
+/* 	    BLEND_ONE_MINUS_CONSTANT_ALPHA                = 0x14, */
+/* 	SEPARATE_ALPHA_BLEND_bit                          = 1 << 29, */
+    CB_COLOR_CONTROL                                      = 0x00028808,
+	FOG_ENABLE_bit                                    = 1 << 0,
+	MULTIWRITE_ENABLE_bit                             = 1 << 1,
+	DITHER_ENABLE_bit                                 = 1 << 2,
+	DEGAMMA_ENABLE_bit                                = 1 << 3,
+	SPECIAL_OP_mask                                   = 0x07 << 4,
+	SPECIAL_OP_shift                                  = 4,
+	    SPECIAL_NORMAL                                = 0x00,
+	    SPECIAL_DISABLE                               = 0x01,
+	    SPECIAL_FAST_CLEAR                            = 0x02,
+	    SPECIAL_FORCE_CLEAR                           = 0x03,
+	    SPECIAL_EXPAND_COLOR                          = 0x04,
+	    SPECIAL_EXPAND_TEXTURE                        = 0x05,
+	    SPECIAL_EXPAND_SAMPLES                        = 0x06,
+	    SPECIAL_RESOLVE_BOX                           = 0x07,
+	PER_MRT_BLEND_bit                                 = 1 << 7,
+	TARGET_BLEND_ENABLE_mask                          = 0xff << 8,
+	TARGET_BLEND_ENABLE_shift                         = 8,
+	ROP3_mask                                         = 0xff << 16,
+	ROP3_shift                                        = 16,
+    DB_SHADER_CONTROL                                     = 0x0002880c,
+	Z_EXPORT_ENABLE_bit                               = 1 << 0,
+	STENCIL_REF_EXPORT_ENABLE_bit                     = 1 << 1,
+	Z_ORDER_mask                                      = 0x03 << 4,
+	Z_ORDER_shift                                     = 4,
+	    LATE_Z                                        = 0x00,
+	    EARLY_Z_THEN_LATE_Z                           = 0x01,
+	    RE_Z                                          = 0x02,
+	    EARLY_Z_THEN_RE_Z                             = 0x03,
+	KILL_ENABLE_bit                                   = 1 << 6,
+	COVERAGE_TO_MASK_ENABLE_bit                       = 1 << 7,
+	MASK_EXPORT_ENABLE_bit                            = 1 << 8,
+	DUAL_EXPORT_ENABLE_bit                            = 1 << 9,
+	EXEC_ON_HIER_FAIL_bit                             = 1 << 10,
+	EXEC_ON_NOOP_bit                                  = 1 << 11,
+    PA_CL_CLIP_CNTL                                       = 0x00028810,
+	UCP_ENA_0_bit                                     = 1 << 0,
+	UCP_ENA_1_bit                                     = 1 << 1,
+	UCP_ENA_2_bit                                     = 1 << 2,
+	UCP_ENA_3_bit                                     = 1 << 3,
+	UCP_ENA_4_bit                                     = 1 << 4,
+	UCP_ENA_5_bit                                     = 1 << 5,
+	PS_UCP_Y_SCALE_NEG_bit                            = 1 << 13,
+	PS_UCP_MODE_mask                                  = 0x03 << 14,
+	PS_UCP_MODE_shift                                 = 14,
+	CLIP_DISABLE_bit                                  = 1 << 16,
+	UCP_CULL_ONLY_ENA_bit                             = 1 << 17,
+	BOUNDARY_EDGE_FLAG_ENA_bit                        = 1 << 18,
+	DX_CLIP_SPACE_DEF_bit                             = 1 << 19,
+	DIS_CLIP_ERR_DETECT_bit                           = 1 << 20,
+	VTX_KILL_OR_bit                                   = 1 << 21,
+	DX_LINEAR_ATTR_CLIP_ENA_bit                       = 1 << 24,
+	VTE_VPORT_PROVOKE_DISABLE_bit                     = 1 << 25,
+	ZCLIP_NEAR_DISABLE_bit                            = 1 << 26,
+	ZCLIP_FAR_DISABLE_bit                             = 1 << 27,
+    PA_SU_SC_MODE_CNTL                                    = 0x00028814,
+	CULL_FRONT_bit                                    = 1 << 0,
+	CULL_BACK_bit                                     = 1 << 1,
+	FACE_bit                                          = 1 << 2,
+	POLY_MODE_mask                                    = 0x03 << 3,
+	POLY_MODE_shift                                   = 3,
+	    X_DISABLE_POLY_MODE                           = 0x00,
+	    X_DUAL_MODE                                   = 0x01,
+	POLYMODE_FRONT_PTYPE_mask                         = 0x07 << 5,
+	POLYMODE_FRONT_PTYPE_shift                        = 5,
+	    X_DRAW_POINTS                                 = 0x00,
+	    X_DRAW_LINES                                  = 0x01,
+	    X_DRAW_TRIANGLES                              = 0x02,
+	POLYMODE_BACK_PTYPE_mask                          = 0x07 << 8,
+	POLYMODE_BACK_PTYPE_shift                         = 8,
+/* 	    X_DRAW_POINTS                                 = 0x00, */
+/* 	    X_DRAW_LINES                                  = 0x01, */
+/* 	    X_DRAW_TRIANGLES                              = 0x02, */
+	POLY_OFFSET_FRONT_ENABLE_bit                      = 1 << 11,
+	POLY_OFFSET_BACK_ENABLE_bit                       = 1 << 12,
+	POLY_OFFSET_PARA_ENABLE_bit                       = 1 << 13,
+	VTX_WINDOW_OFFSET_ENABLE_bit                      = 1 << 16,
+	PROVOKING_VTX_LAST_bit                            = 1 << 19,
+	PERSP_CORR_DIS_bit                                = 1 << 20,
+	MULTI_PRIM_IB_ENA_bit                             = 1 << 21,
+    PA_CL_VTE_CNTL                                        = 0x00028818,
+	VPORT_X_SCALE_ENA_bit                             = 1 << 0,
+	VPORT_X_OFFSET_ENA_bit                            = 1 << 1,
+	VPORT_Y_SCALE_ENA_bit                             = 1 << 2,
+	VPORT_Y_OFFSET_ENA_bit                            = 1 << 3,
+	VPORT_Z_SCALE_ENA_bit                             = 1 << 4,
+	VPORT_Z_OFFSET_ENA_bit                            = 1 << 5,
+	VTX_XY_FMT_bit                                    = 1 << 8,
+	VTX_Z_FMT_bit                                     = 1 << 9,
+	VTX_W0_FMT_bit                                    = 1 << 10,
+	PERFCOUNTER_REF_bit                               = 1 << 11,
+    PA_CL_VS_OUT_CNTL                                     = 0x0002881c,
+	CLIP_DIST_ENA_0_bit                               = 1 << 0,
+	CLIP_DIST_ENA_1_bit                               = 1 << 1,
+	CLIP_DIST_ENA_2_bit                               = 1 << 2,
+	CLIP_DIST_ENA_3_bit                               = 1 << 3,
+	CLIP_DIST_ENA_4_bit                               = 1 << 4,
+	CLIP_DIST_ENA_5_bit                               = 1 << 5,
+	CLIP_DIST_ENA_6_bit                               = 1 << 6,
+	CLIP_DIST_ENA_7_bit                               = 1 << 7,
+	CULL_DIST_ENA_0_bit                               = 1 << 8,
+	CULL_DIST_ENA_1_bit                               = 1 << 9,
+	CULL_DIST_ENA_2_bit                               = 1 << 10,
+	CULL_DIST_ENA_3_bit                               = 1 << 11,
+	CULL_DIST_ENA_4_bit                               = 1 << 12,
+	CULL_DIST_ENA_5_bit                               = 1 << 13,
+	CULL_DIST_ENA_6_bit                               = 1 << 14,
+	CULL_DIST_ENA_7_bit                               = 1 << 15,
+	USE_VTX_POINT_SIZE_bit                            = 1 << 16,
+	USE_VTX_EDGE_FLAG_bit                             = 1 << 17,
+	USE_VTX_RENDER_TARGET_INDX_bit                    = 1 << 18,
+	USE_VTX_VIEWPORT_INDX_bit                         = 1 << 19,
+	USE_VTX_KILL_FLAG_bit                             = 1 << 20,
+	VS_OUT_MISC_VEC_ENA_bit                           = 1 << 21,
+	VS_OUT_CCDIST0_VEC_ENA_bit                        = 1 << 22,
+	VS_OUT_CCDIST1_VEC_ENA_bit                        = 1 << 23,
+    PA_CL_NANINF_CNTL                                     = 0x00028820,
+	VTE_XY_INF_DISCARD_bit                            = 1 << 0,
+	VTE_Z_INF_DISCARD_bit                             = 1 << 1,
+	VTE_W_INF_DISCARD_bit                             = 1 << 2,
+	VTE_0XNANINF_IS_0_bit                             = 1 << 3,
+	VTE_XY_NAN_RETAIN_bit                             = 1 << 4,
+	VTE_Z_NAN_RETAIN_bit                              = 1 << 5,
+	VTE_W_NAN_RETAIN_bit                              = 1 << 6,
+	VTE_W_RECIP_NAN_IS_0_bit                          = 1 << 7,
+	VS_XY_NAN_TO_INF_bit                              = 1 << 8,
+	VS_XY_INF_RETAIN_bit                              = 1 << 9,
+	VS_Z_NAN_TO_INF_bit                               = 1 << 10,
+	VS_Z_INF_RETAIN_bit                               = 1 << 11,
+	VS_W_NAN_TO_INF_bit                               = 1 << 12,
+	VS_W_INF_RETAIN_bit                               = 1 << 13,
+	VS_CLIP_DIST_INF_DISCARD_bit                      = 1 << 14,
+	VTE_NO_OUTPUT_NEG_0_bit                           = 1 << 20,
+    SQ_PGM_START_PS                                       = 0x00028840,
+    SQ_PGM_RESOURCES_PS                                   = 0x00028850,
+	NUM_GPRS_mask                                     = 0xff << 0,
+	NUM_GPRS_shift                                    = 0,
+	STACK_SIZE_mask                                   = 0xff << 8,
+	STACK_SIZE_shift                                  = 8,
+	SQ_PGM_RESOURCES_PS__DX10_CLAMP_bit               = 1 << 21,
+	FETCH_CACHE_LINES_mask                            = 0x07 << 24,
+	FETCH_CACHE_LINES_shift                           = 24,
+	UNCACHED_FIRST_INST_bit                           = 1 << 28,
+	CLAMP_CONSTS_bit                                  = 1 << 31,
+    SQ_PGM_EXPORTS_PS                                     = 0x00028854,
+	EXPORT_MODE_mask                                  = 0x1f << 0,
+	EXPORT_MODE_shift                                 = 0,
+    SQ_PGM_START_VS                                       = 0x00028858,
+    SQ_PGM_RESOURCES_VS                                   = 0x00028868,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_VS__DX10_CLAMP_bit               = 1 << 21,
+/* 	FETCH_CACHE_LINES_mask                            = 0x07 << 24, */
+/* 	FETCH_CACHE_LINES_shift                           = 24, */
+/* 	UNCACHED_FIRST_INST_bit                           = 1 << 28, */
+    SQ_PGM_START_GS                                       = 0x0002886c,
+    SQ_PGM_RESOURCES_GS                                   = 0x0002887c,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_GS__DX10_CLAMP_bit               = 1 << 21,
+/* 	FETCH_CACHE_LINES_mask                            = 0x07 << 24, */
+/* 	FETCH_CACHE_LINES_shift                           = 24, */
+/* 	UNCACHED_FIRST_INST_bit                           = 1 << 28, */
+    SQ_PGM_START_ES                                       = 0x00028880,
+    SQ_PGM_RESOURCES_ES                                   = 0x00028890,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_ES__DX10_CLAMP_bit               = 1 << 21,
+/* 	FETCH_CACHE_LINES_mask                            = 0x07 << 24, */
+/* 	FETCH_CACHE_LINES_shift                           = 24, */
+/* 	UNCACHED_FIRST_INST_bit                           = 1 << 28, */
+    SQ_PGM_START_FS                                       = 0x00028894,
+    SQ_PGM_RESOURCES_FS                                   = 0x000288a4,
+/* 	NUM_GPRS_mask                                     = 0xff << 0, */
+/* 	NUM_GPRS_shift                                    = 0, */
+/* 	STACK_SIZE_mask                                   = 0xff << 8, */
+/* 	STACK_SIZE_shift                                  = 8, */
+	SQ_PGM_RESOURCES_FS__DX10_CLAMP_bit               = 1 << 21,
+    SQ_ESGS_RING_ITEMSIZE                                 = 0x000288a8,
+	ITEMSIZE_mask                                     = 0x7fff << 0,
+	ITEMSIZE_shift                                    = 0,
+    SQ_GSVS_RING_ITEMSIZE                                 = 0x000288ac,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_ESTMP_RING_ITEMSIZE                                = 0x000288b0,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_GSTMP_RING_ITEMSIZE                                = 0x000288b4,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_VSTMP_RING_ITEMSIZE                                = 0x000288b8,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_PSTMP_RING_ITEMSIZE                                = 0x000288bc,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_FBUF_RING_ITEMSIZE                                 = 0x000288c0,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_REDUC_RING_ITEMSIZE                                = 0x000288c4,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_GS_VERT_ITEMSIZE                                   = 0x000288c8,
+/* 	ITEMSIZE_mask                                     = 0x7fff << 0, */
+/* 	ITEMSIZE_shift                                    = 0, */
+    SQ_PGM_CF_OFFSET_PS                                   = 0x000288cc,
+	PGM_CF_OFFSET_mask                                = 0xfffff << 0,
+	PGM_CF_OFFSET_shift                               = 0,
+    SQ_PGM_CF_OFFSET_VS                                   = 0x000288d0,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_PGM_CF_OFFSET_GS                                   = 0x000288d4,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_PGM_CF_OFFSET_ES                                   = 0x000288d8,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_PGM_CF_OFFSET_FS                                   = 0x000288dc,
+/* 	PGM_CF_OFFSET_mask                                = 0xfffff << 0, */
+/* 	PGM_CF_OFFSET_shift                               = 0, */
+    SQ_VTX_SEMANTIC_CLEAR                                 = 0x000288e0,
+    SQ_ALU_CONST_CACHE_PS_0                               = 0x00028940,
+	SQ_ALU_CONST_CACHE_PS_0_num                       = 16,
+    SQ_ALU_CONST_CACHE_VS_0                               = 0x00028980,
+	SQ_ALU_CONST_CACHE_VS_0_num                       = 16,
+    SQ_ALU_CONST_CACHE_GS_0                               = 0x000289c0,
+	SQ_ALU_CONST_CACHE_GS_0_num                       = 16,
+    PA_SU_POINT_SIZE                                      = 0x00028a00,
+	PA_SU_POINT_SIZE__HEIGHT_mask                     = 0xffff << 0,
+	PA_SU_POINT_SIZE__HEIGHT_shift                    = 0,
+	PA_SU_POINT_SIZE__WIDTH_mask                      = 0xffff << 16,
+	PA_SU_POINT_SIZE__WIDTH_shift                     = 16,
+    PA_SU_POINT_MINMAX                                    = 0x00028a04,
+	MIN_SIZE_mask                                     = 0xffff << 0,
+	MIN_SIZE_shift                                    = 0,
+	MAX_SIZE_mask                                     = 0xffff << 16,
+	MAX_SIZE_shift                                    = 16,
+    PA_SU_LINE_CNTL                                       = 0x00028a08,
+	PA_SU_LINE_CNTL__WIDTH_mask                       = 0xffff << 0,
+	PA_SU_LINE_CNTL__WIDTH_shift                      = 0,
+    PA_SC_LINE_STIPPLE                                    = 0x00028a0c,
+	LINE_PATTERN_mask                                 = 0xffff << 0,
+	LINE_PATTERN_shift                                = 0,
+	REPEAT_COUNT_mask                                 = 0xff << 16,
+	REPEAT_COUNT_shift                                = 16,
+	PATTERN_BIT_ORDER_bit                             = 1 << 28,
+	AUTO_RESET_CNTL_mask                              = 0x03 << 29,
+	AUTO_RESET_CNTL_shift                             = 29,
+    VGT_OUTPUT_PATH_CNTL                                  = 0x00028a10,
+	PATH_SELECT_mask                                  = 0x03 << 0,
+	PATH_SELECT_shift                                 = 0,
+	    VGT_OUTPATH_VTX_REUSE                         = 0x00,
+	    VGT_OUTPATH_TESS_EN                           = 0x01,
+	    VGT_OUTPATH_PASSTHRU                          = 0x02,
+	    VGT_OUTPATH_GS_BLOCK                          = 0x03,
+    VGT_HOS_CNTL                                          = 0x00028a14,
+	TESS_MODE_mask                                    = 0x03 << 0,
+	TESS_MODE_shift                                   = 0,
+    VGT_HOS_MAX_TESS_LEVEL                                = 0x00028a18,
+    VGT_HOS_MIN_TESS_LEVEL                                = 0x00028a1c,
+    VGT_HOS_REUSE_DEPTH                                   = 0x00028a20,
+	REUSE_DEPTH_mask                                  = 0xff << 0,
+	REUSE_DEPTH_shift                                 = 0,
+    VGT_GROUP_PRIM_TYPE                                   = 0x00028a24,
+	VGT_GROUP_PRIM_TYPE__PRIM_TYPE_mask               = 0x1f << 0,
+	VGT_GROUP_PRIM_TYPE__PRIM_TYPE_shift              = 0,
+	    VGT_GRP_3D_POINT                              = 0x00,
+	    VGT_GRP_3D_LINE                               = 0x01,
+	    VGT_GRP_3D_TRI                                = 0x02,
+	    VGT_GRP_3D_RECT                               = 0x03,
+	    VGT_GRP_3D_QUAD                               = 0x04,
+	    VGT_GRP_2D_COPY_RECT_V0                       = 0x05,
+	    VGT_GRP_2D_COPY_RECT_V1                       = 0x06,
+	    VGT_GRP_2D_COPY_RECT_V2                       = 0x07,
+	    VGT_GRP_2D_COPY_RECT_V3                       = 0x08,
+	    VGT_GRP_2D_FILL_RECT                          = 0x09,
+	    VGT_GRP_2D_LINE                               = 0x0a,
+	    VGT_GRP_2D_TRI                                = 0x0b,
+	    VGT_GRP_PRIM_INDEX_LINE                       = 0x0c,
+	    VGT_GRP_PRIM_INDEX_TRI                        = 0x0d,
+	    VGT_GRP_PRIM_INDEX_QUAD                       = 0x0e,
+	    VGT_GRP_3D_LINE_ADJ                           = 0x0f,
+	    VGT_GRP_3D_TRI_ADJ                            = 0x10,
+	RETAIN_ORDER_bit                                  = 1 << 14,
+	RETAIN_QUADS_bit                                  = 1 << 15,
+	PRIM_ORDER_mask                                   = 0x07 << 16,
+	PRIM_ORDER_shift                                  = 16,
+	    VGT_GRP_LIST                                  = 0x00,
+	    VGT_GRP_STRIP                                 = 0x01,
+	    VGT_GRP_FAN                                   = 0x02,
+	    VGT_GRP_LOOP                                  = 0x03,
+	    VGT_GRP_POLYGON                               = 0x04,
+    VGT_GROUP_FIRST_DECR                                  = 0x00028a28,
+	FIRST_DECR_mask                                   = 0x0f << 0,
+	FIRST_DECR_shift                                  = 0,
+    VGT_GROUP_DECR                                        = 0x00028a2c,
+	DECR_mask                                         = 0x0f << 0,
+	DECR_shift                                        = 0,
+    VGT_GROUP_VECT_0_CNTL                                 = 0x00028a30,
+	COMP_X_EN_bit                                     = 1 << 0,
+	COMP_Y_EN_bit                                     = 1 << 1,
+	COMP_Z_EN_bit                                     = 1 << 2,
+	COMP_W_EN_bit                                     = 1 << 3,
+	VGT_GROUP_VECT_0_CNTL__STRIDE_mask                = 0xff << 8,
+	VGT_GROUP_VECT_0_CNTL__STRIDE_shift               = 8,
+	SHIFT_mask                                        = 0xff << 16,
+	SHIFT_shift                                       = 16,
+    VGT_GROUP_VECT_1_CNTL                                 = 0x00028a34,
+/* 	COMP_X_EN_bit                                     = 1 << 0, */
+/* 	COMP_Y_EN_bit                                     = 1 << 1, */
+/* 	COMP_Z_EN_bit                                     = 1 << 2, */
+/* 	COMP_W_EN_bit                                     = 1 << 3, */
+	VGT_GROUP_VECT_1_CNTL__STRIDE_mask                = 0xff << 8,
+	VGT_GROUP_VECT_1_CNTL__STRIDE_shift               = 8,
+/* 	SHIFT_mask                                        = 0xff << 16, */
+/* 	SHIFT_shift                                       = 16, */
+    VGT_GROUP_VECT_0_FMT_CNTL                             = 0x00028a38,
+	X_CONV_mask                                       = 0x0f << 0,
+	X_CONV_shift                                      = 0,
+	    VGT_GRP_INDEX_16                              = 0x00,
+	    VGT_GRP_INDEX_32                              = 0x01,
+	    VGT_GRP_UINT_16                               = 0x02,
+	    VGT_GRP_UINT_32                               = 0x03,
+	    VGT_GRP_SINT_16                               = 0x04,
+	    VGT_GRP_SINT_32                               = 0x05,
+	    VGT_GRP_FLOAT_32                              = 0x06,
+	    VGT_GRP_AUTO_PRIM                             = 0x07,
+	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08,
+	X_OFFSET_mask                                     = 0x0f << 4,
+	X_OFFSET_shift                                    = 4,
+	Y_CONV_mask                                       = 0x0f << 8,
+	Y_CONV_shift                                      = 8,
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+	Y_OFFSET_mask                                     = 0x0f << 12,
+	Y_OFFSET_shift                                    = 12,
+	Z_CONV_mask                                       = 0x0f << 16,
+	Z_CONV_shift                                      = 16,
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+	Z_OFFSET_mask                                     = 0x0f << 20,
+	Z_OFFSET_shift                                    = 20,
+	W_CONV_mask                                       = 0x0f << 24,
+	W_CONV_shift                                      = 24,
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+	W_OFFSET_mask                                     = 0x0f << 28,
+	W_OFFSET_shift                                    = 28,
+    VGT_GROUP_VECT_1_FMT_CNTL                             = 0x00028a3c,
+/* 	X_CONV_mask                                       = 0x0f << 0, */
+/* 	X_CONV_shift                                      = 0, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	X_OFFSET_mask                                     = 0x0f << 4, */
+/* 	X_OFFSET_shift                                    = 4, */
+/* 	Y_CONV_mask                                       = 0x0f << 8, */
+/* 	Y_CONV_shift                                      = 8, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	Y_OFFSET_mask                                     = 0x0f << 12, */
+/* 	Y_OFFSET_shift                                    = 12, */
+/* 	Z_CONV_mask                                       = 0x0f << 16, */
+/* 	Z_CONV_shift                                      = 16, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	Z_OFFSET_mask                                     = 0x0f << 20, */
+/* 	Z_OFFSET_shift                                    = 20, */
+/* 	W_CONV_mask                                       = 0x0f << 24, */
+/* 	W_CONV_shift                                      = 24, */
+/* 	    VGT_GRP_INDEX_16                              = 0x00, */
+/* 	    VGT_GRP_INDEX_32                              = 0x01, */
+/* 	    VGT_GRP_UINT_16                               = 0x02, */
+/* 	    VGT_GRP_UINT_32                               = 0x03, */
+/* 	    VGT_GRP_SINT_16                               = 0x04, */
+/* 	    VGT_GRP_SINT_32                               = 0x05, */
+/* 	    VGT_GRP_FLOAT_32                              = 0x06, */
+/* 	    VGT_GRP_AUTO_PRIM                             = 0x07, */
+/* 	    VGT_GRP_FIX_1_23_TO_FLOAT                     = 0x08, */
+/* 	W_OFFSET_mask                                     = 0x0f << 28, */
+/* 	W_OFFSET_shift                                    = 28, */
+    VGT_GS_MODE                                           = 0x00028a40,
+	MODE_mask                                         = 0x03 << 0,
+	MODE_shift                                        = 0,
+	    GS_OFF                                        = 0x00,
+	    GS_SCENARIO_A                                 = 0x01,
+	    GS_SCENARIO_B                                 = 0x02,
+	    GS_SCENARIO_G                                 = 0x03,
+	ES_PASSTHRU_bit                                   = 1 << 2,
+	CUT_MODE_mask                                     = 0x03 << 3,
+	CUT_MODE_shift                                    = 3,
+	    GS_CUT_1024                                   = 0x00,
+	    GS_CUT_512                                    = 0x01,
+	    GS_CUT_256                                    = 0x02,
+	    GS_CUT_128                                    = 0x03,
+    PA_SC_MPASS_PS_CNTL                                   = 0x00028a48,
+	MPASS_PIX_VEC_PER_PASS_mask                       = 0xfffff << 0,
+	MPASS_PIX_VEC_PER_PASS_shift                      = 0,
+	MPASS_PS_ENA_bit                                  = 1 << 31,
+    PA_SC_MODE_CNTL                                       = 0x00028a4c,
+	MSAA_ENABLE_bit                                   = 1 << 0,
+	CLIPRECT_ENABLE_bit                               = 1 << 1,
+	LINE_STIPPLE_ENABLE_bit                           = 1 << 2,
+	MULTI_CHIP_PRIM_DISCARD_ENAB_bit                  = 1 << 3,
+	WALK_ORDER_ENABLE_bit                             = 1 << 4,
+	HALVE_DETAIL_SAMPLE_PERF_bit                      = 1 << 5,
+	WALK_SIZE_bit                                     = 1 << 6,
+	WALK_ALIGNMENT_bit                                = 1 << 7,
+	WALK_ALIGN8_PRIM_FITS_ST_bit                      = 1 << 8,
+	TILE_COVER_NO_SCISSOR_bit                         = 1 << 9,
+	KILL_PIX_POST_HI_Z_bit                            = 1 << 10,
+	KILL_PIX_POST_DETAIL_MASK_bit                     = 1 << 11,
+	MULTI_CHIP_SUPERTILE_ENABLE_bit                   = 1 << 12,
+	TILE_COVER_DISABLE_bit                            = 1 << 13,
+	FORCE_EOV_CNTDWN_ENABLE_bit                       = 1 << 14,
+	FORCE_EOV_TILE_ENABLE_bit                         = 1 << 15,
+	FORCE_EOV_REZ_ENABLE_bit                          = 1 << 16,
+	PS_ITER_SAMPLE_bit                                = 1 << 17,
+    VGT_ENHANCE                                           = 0x00028a50,
+	VGT_ENHANCE__MI_TIMESTAMP_RES_mask                = 0x03 << 0,
+	VGT_ENHANCE__MI_TIMESTAMP_RES_shift               = 0,
+	    X_0_992_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_32   = 0x00,
+	    X_0_496_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_16   = 0x01,
+	    X_0_248_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_8    = 0x02,
+	    X_0_124_CLOCKS_LATENCY_RANGE_IN_STEPS_OF_4    = 0x03,
+	MISC_mask                                         = 0x3fffffff << 2,
+	MISC_shift                                        = 2,
+    VGT_GS_OUT_PRIM_TYPE                                  = 0x00028a6c,
+	OUTPRIM_TYPE_mask                                 = 0x3f << 0,
+	OUTPRIM_TYPE_shift                                = 0,
+	    POINTLIST                                     = 0x00,
+	    LINESTRIP                                     = 0x01,
+	    TRISTRIP                                      = 0x02,
+    VGT_DMA_SIZE                                          = 0x00028a74,
+    VGT_DMA_INDEX_TYPE                                    = 0x00028a7c,
+/* 	INDEX_TYPE_mask                                   = 0x03 << 0, */
+/* 	INDEX_TYPE_shift                                  = 0, */
+	    VGT_INDEX_16                                  = 0x00,
+	    VGT_INDEX_32                                  = 0x01,
+	SWAP_MODE_mask                                    = 0x03 << 2,
+	SWAP_MODE_shift                                   = 2,
+	    VGT_DMA_SWAP_NONE                             = 0x00,
+	    VGT_DMA_SWAP_16_BIT                           = 0x01,
+	    VGT_DMA_SWAP_32_BIT                           = 0x02,
+	    VGT_DMA_SWAP_WORD                             = 0x03,
+    VGT_PRIMITIVEID_EN                                    = 0x00028a84,
+	PRIMITIVEID_EN_bit                                = 1 << 0,
+    VGT_DMA_NUM_INSTANCES                                 = 0x00028a88,
+    VGT_EVENT_INITIATOR                                   = 0x00028a90,
+	EVENT_TYPE_mask                                   = 0x3f << 0,
+	EVENT_TYPE_shift                                  = 0,
+	    CACHE_FLUSH_TS                                = 0x04,
+	    CONTEXT_DONE                                  = 0x05,
+	    CACHE_FLUSH                                   = 0x06,
+	    VIZQUERY_START                                = 0x07,
+	    VIZQUERY_END                                  = 0x08,
+	    SC_WAIT_WC                                    = 0x09,
+	    MPASS_PS_CP_REFETCH                           = 0x0a,
+	    MPASS_PS_RST_START                            = 0x0b,
+	    MPASS_PS_INCR_START                           = 0x0c,
+	    RST_PIX_CNT                                   = 0x0d,
+	    RST_VTX_CNT                                   = 0x0e,
+	    VS_PARTIAL_FLUSH                              = 0x0f,
+	    PS_PARTIAL_FLUSH                              = 0x10,
+	    CACHE_FLUSH_AND_INV_TS_EVENT                  = 0x14,
+	    ZPASS_DONE                                    = 0x15,
+	    CACHE_FLUSH_AND_INV_EVENT                     = 0x16,
+	    PERFCOUNTER_START                             = 0x17,
+	    PERFCOUNTER_STOP                              = 0x18,
+	    PIPELINESTAT_START                            = 0x19,
+	    PIPELINESTAT_STOP                             = 0x1a,
+	    PERFCOUNTER_SAMPLE                            = 0x1b,
+	    FLUSH_ES_OUTPUT                               = 0x1c,
+	    FLUSH_GS_OUTPUT                               = 0x1d,
+	    SAMPLE_PIPELINESTAT                           = 0x1e,
+	    SO_VGTSTREAMOUT_FLUSH                         = 0x1f,
+	    SAMPLE_STREAMOUTSTATS                         = 0x20,
+	    RESET_VTX_CNT                                 = 0x21,
+	    BLOCK_CONTEXT_DONE                            = 0x22,
+	    CR_CONTEXT_DONE                               = 0x23,
+	    VGT_FLUSH                                     = 0x24,
+	    CR_DONE_TS                                    = 0x25,
+	    SQ_NON_EVENT                                  = 0x26,
+	    SC_SEND_DB_VPZ                                = 0x27,
+	    BOTTOM_OF_PIPE_TS                             = 0x28,
+	    DB_CACHE_FLUSH_AND_INV                        = 0x2a,
+	ADDRESS_HI_mask                                   = 0xff << 19,
+	ADDRESS_HI_shift                                  = 19,
+	EXTENDED_EVENT_bit                                = 1 << 27,
+    VGT_MULTI_PRIM_IB_RESET_EN                            = 0x00028a94,
+	RESET_EN_bit                                      = 1 << 0,
+    VGT_INSTANCE_STEP_RATE_0                              = 0x00028aa0,
+    VGT_INSTANCE_STEP_RATE_1                              = 0x00028aa4,
+    VGT_STRMOUT_EN                                        = 0x00028ab0,
+	STREAMOUT_bit                                     = 1 << 0,
+    VGT_REUSE_OFF                                         = 0x00028ab4,
+	REUSE_OFF_bit                                     = 1 << 0,
+    VGT_VTX_CNT_EN                                        = 0x00028ab8,
+	VTX_CNT_EN_bit                                    = 1 << 0,
+    VGT_STRMOUT_BUFFER_SIZE_0                             = 0x00028ad0,
+    VGT_STRMOUT_VTX_STRIDE_0                              = 0x00028ad4,
+	VGT_STRMOUT_VTX_STRIDE_0__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_0__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_0                             = 0x00028ad8,
+    VGT_STRMOUT_BUFFER_OFFSET_0                           = 0x00028adc,
+    VGT_STRMOUT_BUFFER_SIZE_1                             = 0x00028ae0,
+    VGT_STRMOUT_VTX_STRIDE_1                              = 0x00028ae4,
+	VGT_STRMOUT_VTX_STRIDE_1__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_1__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_1                             = 0x00028ae8,
+    VGT_STRMOUT_BUFFER_OFFSET_1                           = 0x00028aec,
+    VGT_STRMOUT_BUFFER_SIZE_2                             = 0x00028af0,
+    VGT_STRMOUT_VTX_STRIDE_2                              = 0x00028af4,
+	VGT_STRMOUT_VTX_STRIDE_2__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_2__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_2                             = 0x00028af8,
+    VGT_STRMOUT_BUFFER_OFFSET_2                           = 0x00028afc,
+    VGT_STRMOUT_BUFFER_SIZE_3                             = 0x00028b00,
+    VGT_STRMOUT_VTX_STRIDE_3                              = 0x00028b04,
+	VGT_STRMOUT_VTX_STRIDE_3__STRIDE_mask             = 0x3ff << 0,
+	VGT_STRMOUT_VTX_STRIDE_3__STRIDE_shift            = 0,
+    VGT_STRMOUT_BUFFER_BASE_3                             = 0x00028b08,
+    VGT_STRMOUT_BUFFER_OFFSET_3                           = 0x00028b0c,
+    VGT_STRMOUT_BASE_OFFSET_0                             = 0x00028b10,
+    VGT_STRMOUT_BASE_OFFSET_1                             = 0x00028b14,
+    VGT_STRMOUT_BASE_OFFSET_2                             = 0x00028b18,
+    VGT_STRMOUT_BASE_OFFSET_3                             = 0x00028b1c,
+    VGT_STRMOUT_BUFFER_EN                                 = 0x00028b20,
+	BUFFER_0_EN_bit                                   = 1 << 0,
+	BUFFER_1_EN_bit                                   = 1 << 1,
+	BUFFER_2_EN_bit                                   = 1 << 2,
+	BUFFER_3_EN_bit                                   = 1 << 3,
+    VGT_STRMOUT_DRAW_OPAQUE_OFFSET                        = 0x00028b28,
+    VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE            = 0x00028b2c,
+    VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE                 = 0x00028b30,
+    VGT_STRMOUT_BASE_OFFSET_HI_0                          = 0x00028b44,
+	VGT_STRMOUT_BASE_OFFSET_HI_0__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_0__BASE_OFFSET_shift   = 0,
+    VGT_STRMOUT_BASE_OFFSET_HI_1                          = 0x00028b48,
+	VGT_STRMOUT_BASE_OFFSET_HI_1__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_1__BASE_OFFSET_shift   = 0,
+    VGT_STRMOUT_BASE_OFFSET_HI_2                          = 0x00028b4c,
+	VGT_STRMOUT_BASE_OFFSET_HI_2__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_2__BASE_OFFSET_shift   = 0,
+    VGT_STRMOUT_BASE_OFFSET_HI_3                          = 0x00028b50,
+	VGT_STRMOUT_BASE_OFFSET_HI_3__BASE_OFFSET_mask    = 0x3f << 0,
+	VGT_STRMOUT_BASE_OFFSET_HI_3__BASE_OFFSET_shift   = 0,
+    PA_SC_LINE_CNTL                                       = 0x00028c00,
+	BRES_CNTL_mask                                    = 0xff << 0,
+	BRES_CNTL_shift                                   = 0,
+	USE_BRES_CNTL_bit                                 = 1 << 8,
+	EXPAND_LINE_WIDTH_bit                             = 1 << 9,
+	LAST_PIXEL_bit                                    = 1 << 10,
+    PA_SC_AA_CONFIG                                       = 0x00028c04,
+	MSAA_NUM_SAMPLES_mask                             = 0x03 << 0,
+	MSAA_NUM_SAMPLES_shift                            = 0,
+	AA_MASK_CENTROID_DTMN_bit                         = 1 << 4,
+	MAX_SAMPLE_DIST_mask                              = 0x0f << 13,
+	MAX_SAMPLE_DIST_shift                             = 13,
+    PA_SU_VTX_CNTL                                        = 0x00028c08,
+	PIX_CENTER_bit                                    = 1 << 0,
+	PA_SU_VTX_CNTL__ROUND_MODE_mask                   = 0x03 << 1,
+	PA_SU_VTX_CNTL__ROUND_MODE_shift                  = 1,
+	    X_TRUNCATE                                    = 0x00,
+	    X_ROUND                                       = 0x01,
+	    X_ROUND_TO_EVEN                               = 0x02,
+	    X_ROUND_TO_ODD                                = 0x03,
+	QUANT_MODE_mask                                   = 0x07 << 3,
+	QUANT_MODE_shift                                  = 3,
+	    X_1_16TH                                      = 0x00,
+	    X_1_8TH                                       = 0x01,
+	    X_1_4TH                                       = 0x02,
+	    X_1_2                                         = 0x03,
+	    X_1                                           = 0x04,
+	    X_1_256TH                                     = 0x05,
+    PA_CL_GB_VERT_CLIP_ADJ                                = 0x00028c0c,
+    PA_CL_GB_VERT_DISC_ADJ                                = 0x00028c10,
+    PA_CL_GB_HORZ_CLIP_ADJ                                = 0x00028c14,
+    PA_CL_GB_HORZ_DISC_ADJ                                = 0x00028c18,
+    PA_SC_AA_SAMPLE_LOCS_MCTX                             = 0x00028c1c,
+/* 	S0_X_mask                                         = 0x0f << 0, */
+/* 	S0_X_shift                                        = 0, */
+/* 	S0_Y_mask                                         = 0x0f << 4, */
+/* 	S0_Y_shift                                        = 4, */
+/* 	S1_X_mask                                         = 0x0f << 8, */
+/* 	S1_X_shift                                        = 8, */
+/* 	S1_Y_mask                                         = 0x0f << 12, */
+/* 	S1_Y_shift                                        = 12, */
+/* 	S2_X_mask                                         = 0x0f << 16, */
+/* 	S2_X_shift                                        = 16, */
+/* 	S2_Y_mask                                         = 0x0f << 20, */
+/* 	S2_Y_shift                                        = 20, */
+/* 	S3_X_mask                                         = 0x0f << 24, */
+/* 	S3_X_shift                                        = 24, */
+/* 	S3_Y_mask                                         = 0x0f << 28, */
+/* 	S3_Y_shift                                        = 28, */
+    PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX                      = 0x00028c20,
+/* 	S4_X_mask                                         = 0x0f << 0, */
+/* 	S4_X_shift                                        = 0, */
+/* 	S4_Y_mask                                         = 0x0f << 4, */
+/* 	S4_Y_shift                                        = 4, */
+/* 	S5_X_mask                                         = 0x0f << 8, */
+/* 	S5_X_shift                                        = 8, */
+/* 	S5_Y_mask                                         = 0x0f << 12, */
+/* 	S5_Y_shift                                        = 12, */
+/* 	S6_X_mask                                         = 0x0f << 16, */
+/* 	S6_X_shift                                        = 16, */
+/* 	S6_Y_mask                                         = 0x0f << 20, */
+/* 	S6_Y_shift                                        = 20, */
+/* 	S7_X_mask                                         = 0x0f << 24, */
+/* 	S7_X_shift                                        = 24, */
+/* 	S7_Y_mask                                         = 0x0f << 28, */
+/* 	S7_Y_shift                                        = 28, */
+    CB_CLRCMP_CONTROL                                     = 0x00028c30,
+	CLRCMP_FCN_SRC_mask                               = 0x07 << 0,
+	CLRCMP_FCN_SRC_shift                              = 0,
+	    CLRCMP_DRAW_ALWAYS                            = 0x00,
+	    CLRCMP_DRAW_NEVER                             = 0x01,
+	    CLRCMP_DRAW_ON_NEQ                            = 0x04,
+	    CLRCMP_DRAW_ON_EQ                             = 0x05,
+	CLRCMP_FCN_DST_mask                               = 0x07 << 8,
+	CLRCMP_FCN_DST_shift                              = 8,
+/* 	    CLRCMP_DRAW_ALWAYS                            = 0x00, */
+/* 	    CLRCMP_DRAW_NEVER                             = 0x01, */
+/* 	    CLRCMP_DRAW_ON_NEQ                            = 0x04, */
+/* 	    CLRCMP_DRAW_ON_EQ                             = 0x05, */
+	CLRCMP_FCN_SEL_mask                               = 0x03 << 24,
+	CLRCMP_FCN_SEL_shift                              = 24,
+	    CLRCMP_SEL_DST                                = 0x00,
+	    CLRCMP_SEL_SRC                                = 0x01,
+	    CLRCMP_SEL_AND                                = 0x02,
+    CB_CLRCMP_SRC                                         = 0x00028c34,
+    CB_CLRCMP_DST                                         = 0x00028c38,
+    CB_CLRCMP_MSK                                         = 0x00028c3c,
+    PA_SC_AA_MASK                                         = 0x00028c48,
+    VGT_VERTEX_REUSE_BLOCK_CNTL                           = 0x00028c58,
+	VTX_REUSE_DEPTH_mask                              = 0xff << 0,
+	VTX_REUSE_DEPTH_shift                             = 0,
+    VGT_OUT_DEALLOC_CNTL                                  = 0x00028c5c,
+	DEALLOC_DIST_mask                                 = 0x7f << 0,
+	DEALLOC_DIST_shift                                = 0,
+    DB_RENDER_CONTROL                                     = 0x00028d0c,
+	DEPTH_CLEAR_ENABLE_bit                            = 1 << 0,
+	STENCIL_CLEAR_ENABLE_bit                          = 1 << 1,
+	DEPTH_COPY_bit                                    = 1 << 2,
+	STENCIL_COPY_bit                                  = 1 << 3,
+	RESUMMARIZE_ENABLE_bit                            = 1 << 4,
+	STENCIL_COMPRESS_DISABLE_bit                      = 1 << 5,
+	DEPTH_COMPRESS_DISABLE_bit                        = 1 << 6,
+	COPY_CENTROID_bit                                 = 1 << 7,
+	COPY_SAMPLE_mask                                  = 0x07 << 8,
+	COPY_SAMPLE_shift                                 = 8,
+	ZPASS_INCREMENT_DISABLE_bit                       = 1 << 11,
+    DB_RENDER_OVERRIDE                                    = 0x00028d10,
+	FORCE_HIZ_ENABLE_mask                             = 0x03 << 0,
+	FORCE_HIZ_ENABLE_shift                            = 0,
+	    FORCE_OFF                                     = 0x00,
+	    FORCE_ENABLE                                  = 0x01,
+	    FORCE_DISABLE                                 = 0x02,
+	    FORCE_RESERVED                                = 0x03,
+	FORCE_HIS_ENABLE0_mask                            = 0x03 << 2,
+	FORCE_HIS_ENABLE0_shift                           = 2,
+/* 	    FORCE_OFF                                     = 0x00, */
+/* 	    FORCE_ENABLE                                  = 0x01, */
+/* 	    FORCE_DISABLE                                 = 0x02, */
+/* 	    FORCE_RESERVED                                = 0x03, */
+	FORCE_HIS_ENABLE1_mask                            = 0x03 << 4,
+	FORCE_HIS_ENABLE1_shift                           = 4,
+/* 	    FORCE_OFF                                     = 0x00, */
+/* 	    FORCE_ENABLE                                  = 0x01, */
+/* 	    FORCE_DISABLE                                 = 0x02, */
+/* 	    FORCE_RESERVED                                = 0x03, */
+	FORCE_SHADER_Z_ORDER_bit                          = 1 << 6,
+	FAST_Z_DISABLE_bit                                = 1 << 7,
+	FAST_STENCIL_DISABLE_bit                          = 1 << 8,
+	NOOP_CULL_DISABLE_bit                             = 1 << 9,
+	FORCE_COLOR_KILL_bit                              = 1 << 10,
+	FORCE_Z_READ_bit                                  = 1 << 11,
+	FORCE_STENCIL_READ_bit                            = 1 << 12,
+	FORCE_FULL_Z_RANGE_mask                           = 0x03 << 13,
+	FORCE_FULL_Z_RANGE_shift                          = 13,
+/* 	    FORCE_OFF                                     = 0x00, */
+/* 	    FORCE_ENABLE                                  = 0x01, */
+/* 	    FORCE_DISABLE                                 = 0x02, */
+/* 	    FORCE_RESERVED                                = 0x03, */
+	FORCE_QC_SMASK_CONFLICT_bit                       = 1 << 15,
+	DISABLE_VIEWPORT_CLAMP_bit                        = 1 << 16,
+	IGNORE_SC_ZRANGE_bit                              = 1 << 17,
+    DB_HTILE_SURFACE                                      = 0x00028d24,
+	HTILE_WIDTH_bit                                   = 1 << 0,
+	HTILE_HEIGHT_bit                                  = 1 << 1,
+	LINEAR_bit                                        = 1 << 2,
+	FULL_CACHE_bit                                    = 1 << 3,
+	HTILE_USES_PRELOAD_WIN_bit                        = 1 << 4,
+	PRELOAD_bit                                       = 1 << 5,
+	PREFETCH_WIDTH_mask                               = 0x3f << 6,
+	PREFETCH_WIDTH_shift                              = 6,
+	PREFETCH_HEIGHT_mask                              = 0x3f << 12,
+	PREFETCH_HEIGHT_shift                             = 12,
+    DB_SRESULTS_COMPARE_STATE1                            = 0x00028d2c,
+	COMPAREFUNC1_mask                                 = 0x07 << 0,
+	COMPAREFUNC1_shift                                = 0,
+/* 	    REF_NEVER                                     = 0x00, */
+/* 	    REF_LESS                                      = 0x01, */
+/* 	    REF_EQUAL                                     = 0x02, */
+/* 	    REF_LEQUAL                                    = 0x03, */
+/* 	    REF_GREATER                                   = 0x04, */
+/* 	    REF_NOTEQUAL                                  = 0x05, */
+/* 	    REF_GEQUAL                                    = 0x06, */
+/* 	    REF_ALWAYS                                    = 0x07, */
+	COMPAREVALUE1_mask                                = 0xff << 4,
+	COMPAREVALUE1_shift                               = 4,
+	COMPAREMASK1_mask                                 = 0xff << 12,
+	COMPAREMASK1_shift                                = 12,
+	ENABLE1_bit                                       = 1 << 24,
+    DB_PRELOAD_CONTROL                                    = 0x00028d30,
+	START_X_mask                                      = 0xff << 0,
+	START_X_shift                                     = 0,
+	START_Y_mask                                      = 0xff << 8,
+	START_Y_shift                                     = 8,
+	MAX_X_mask                                        = 0xff << 16,
+	MAX_X_shift                                       = 16,
+	MAX_Y_mask                                        = 0xff << 24,
+	MAX_Y_shift                                       = 24,
+    DB_PREFETCH_LIMIT                                     = 0x00028d34,
+	DEPTH_HEIGHT_TILE_MAX_mask                        = 0x3ff << 0,
+	DEPTH_HEIGHT_TILE_MAX_shift                       = 0,
+    PA_SU_POLY_OFFSET_DB_FMT_CNTL                         = 0x00028df8,
+	POLY_OFFSET_NEG_NUM_DB_BITS_mask                  = 0xff << 0,
+	POLY_OFFSET_NEG_NUM_DB_BITS_shift                 = 0,
+	POLY_OFFSET_DB_IS_FLOAT_FMT_bit                   = 1 << 8,
+    PA_SU_POLY_OFFSET_CLAMP                               = 0x00028dfc,
+    PA_SU_POLY_OFFSET_FRONT_SCALE                         = 0x00028e00,
+    PA_SU_POLY_OFFSET_FRONT_OFFSET                        = 0x00028e04,
+    PA_SU_POLY_OFFSET_BACK_SCALE                          = 0x00028e08,
+    PA_SU_POLY_OFFSET_BACK_OFFSET                         = 0x00028e0c,
+    PA_CL_POINT_X_RAD                                     = 0x00028e10,
+    PA_CL_POINT_Y_RAD                                     = 0x00028e14,
+    PA_CL_POINT_SIZE                                      = 0x00028e18,
+    PA_CL_POINT_CULL_RAD                                  = 0x00028e1c,
+    PA_CL_UCP_0_X                                         = 0x00028e20,
+	PA_CL_UCP_0_X_num                                 = 6,
+	PA_CL_UCP_0_X_offset                              = 16,
+    PA_CL_UCP_0_Y                                         = 0x00028e24,
+	PA_CL_UCP_0_Y_num                                 = 6,
+	PA_CL_UCP_0_Y_offset                              = 16,
+    PA_CL_UCP_0_Z                                         = 0x00028e28,
+	PA_CL_UCP_0_Z_num                                 = 6,
+	PA_CL_UCP_0_Z_offset                              = 16,
+    SQ_ALU_CONSTANT0_0                                    = 0x00030000,
+    SQ_ALU_CONSTANT1_0                                    = 0x00030004,
+    SQ_ALU_CONSTANT2_0                                    = 0x00030008,
+    SQ_ALU_CONSTANT3_0                                    = 0x0003000c,
+    SQ_VTX_CONSTANT_WORD0_0                               = 0x00038000,
+    SQ_TEX_RESOURCE_WORD0_0                               = 0x00038000,
+	DIM_mask                                          = 0x07 << 0,
+	DIM_shift                                         = 0,
+	    SQ_TEX_DIM_1D                                 = 0x00,
+	    SQ_TEX_DIM_2D                                 = 0x01,
+	    SQ_TEX_DIM_3D                                 = 0x02,
+	    SQ_TEX_DIM_CUBEMAP                            = 0x03,
+	    SQ_TEX_DIM_1D_ARRAY                           = 0x04,
+	    SQ_TEX_DIM_2D_ARRAY                           = 0x05,
+	    SQ_TEX_DIM_2D_MSAA                            = 0x06,
+	    SQ_TEX_DIM_2D_ARRAY_MSAA                      = 0x07,
+	SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_mask           = 0x0f << 3,
+	SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_shift          = 3,
+	TILE_TYPE_bit                                     = 1 << 7,
+	PITCH_mask                                        = 0x7ff << 8,
+	PITCH_shift                                       = 8,
+	TEX_WIDTH_mask                                    = 0x1fff << 19,
+	TEX_WIDTH_shift                                   = 19,
+    SQ_VTX_CONSTANT_WORD1_0                               = 0x00038004,
+    SQ_TEX_RESOURCE_WORD1_0                               = 0x00038004,
+	TEX_HEIGHT_mask                                   = 0x1fff << 0,
+	TEX_HEIGHT_shift                                  = 0,
+	TEX_DEPTH_mask                                    = 0x1fff << 13,
+	TEX_DEPTH_shift                                   = 13,
+	SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask         = 0x3f << 26,
+	SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift        = 26,
+    SQ_VTX_CONSTANT_WORD2_0                               = 0x00038008,
+	BASE_ADDRESS_HI_mask                              = 0xff << 0,
+	BASE_ADDRESS_HI_shift                             = 0,
+	SQ_VTX_CONSTANT_WORD2_0__STRIDE_mask              = 0x7ff << 8,
+	SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift             = 8,
+	SQ_VTX_CONSTANT_WORD2_0__CLAMP_X_bit              = 1 << 19,
+	SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask         = 0x3f << 20,
+	SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift        = 20,
+	SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask      = 0x03 << 26,
+	SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift     = 26,
+/* 	    SQ_NUM_FORMAT_NORM                            = 0x00, */
+/* 	    SQ_NUM_FORMAT_INT                             = 0x01, */
+/* 	    SQ_NUM_FORMAT_SCALED                          = 0x02, */
+	SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit      = 1 << 28,
+	SQ_VTX_CONSTANT_WORD2_0__SRF_MODE_ALL_bit         = 1 << 29,
+	SQ_VTX_CONSTANT_WORD2_0__ENDIAN_SWAP_mask         = 0x03 << 30,
+	SQ_VTX_CONSTANT_WORD2_0__ENDIAN_SWAP_shift        = 30,
+/* 	    SQ_ENDIAN_NONE                                = 0x00, */
+/* 	    SQ_ENDIAN_8IN16                               = 0x01, */
+/* 	    SQ_ENDIAN_8IN32                               = 0x02, */
+    SQ_TEX_RESOURCE_WORD2_0                               = 0x00038008,
+    SQ_VTX_CONSTANT_WORD3_0                               = 0x0003800c,
+	MEM_REQUEST_SIZE_mask                             = 0x03 << 0,
+	MEM_REQUEST_SIZE_shift                            = 0,
+    SQ_TEX_RESOURCE_WORD3_0                               = 0x0003800c,
+    SQ_TEX_RESOURCE_WORD4_0                               = 0x00038010,
+	FORMAT_COMP_X_mask                                = 0x03 << 0,
+	FORMAT_COMP_X_shift                               = 0,
+	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00,
+	    SQ_FORMAT_COMP_SIGNED                         = 0x01,
+	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02,
+	FORMAT_COMP_Y_mask                                = 0x03 << 2,
+	FORMAT_COMP_Y_shift                               = 2,
+/* 	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00, */
+/* 	    SQ_FORMAT_COMP_SIGNED                         = 0x01, */
+/* 	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02, */
+	FORMAT_COMP_Z_mask                                = 0x03 << 4,
+	FORMAT_COMP_Z_shift                               = 4,
+/* 	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00, */
+/* 	    SQ_FORMAT_COMP_SIGNED                         = 0x01, */
+/* 	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02, */
+	FORMAT_COMP_W_mask                                = 0x03 << 6,
+	FORMAT_COMP_W_shift                               = 6,
+/* 	    SQ_FORMAT_COMP_UNSIGNED                       = 0x00, */
+/* 	    SQ_FORMAT_COMP_SIGNED                         = 0x01, */
+/* 	    SQ_FORMAT_COMP_UNSIGNED_BIASED                = 0x02, */
+	SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_mask      = 0x03 << 8,
+	SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_shift     = 8,
+/* 	    SQ_NUM_FORMAT_NORM                            = 0x00, */
+/* 	    SQ_NUM_FORMAT_INT                             = 0x01, */
+/* 	    SQ_NUM_FORMAT_SCALED                          = 0x02, */
+	SQ_TEX_RESOURCE_WORD4_0__SRF_MODE_ALL_bit         = 1 << 10,
+	SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit        = 1 << 11,
+	SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_mask         = 0x03 << 12,
+	SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_shift        = 12,
+/* 	    SQ_ENDIAN_NONE                                = 0x00, */
+/* 	    SQ_ENDIAN_8IN16                               = 0x01, */
+/* 	    SQ_ENDIAN_8IN32                               = 0x02, */
+	REQUEST_SIZE_mask                                 = 0x03 << 14,
+	REQUEST_SIZE_shift                                = 14,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask           = 0x07 << 16,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift          = 16,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask           = 0x07 << 19,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift          = 19,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask           = 0x07 << 22,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift          = 22,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask           = 0x07 << 25,
+	SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift          = 25,
+/* 	    SQ_SEL_X                                      = 0x00, */
+/* 	    SQ_SEL_Y                                      = 0x01, */
+/* 	    SQ_SEL_Z                                      = 0x02, */
+/* 	    SQ_SEL_W                                      = 0x03, */
+/* 	    SQ_SEL_0                                      = 0x04, */
+/* 	    SQ_SEL_1                                      = 0x05, */
+	BASE_LEVEL_mask                                   = 0x0f << 28,
+	BASE_LEVEL_shift                                  = 28,
+    SQ_TEX_RESOURCE_WORD5_0                               = 0x00038014,
+	LAST_LEVEL_mask                                   = 0x0f << 0,
+	LAST_LEVEL_shift                                  = 0,
+	BASE_ARRAY_mask                                   = 0x1fff << 4,
+	BASE_ARRAY_shift                                  = 4,
+	LAST_ARRAY_mask                                   = 0x1fff << 17,
+	LAST_ARRAY_shift                                  = 17,
+    SQ_TEX_RESOURCE_WORD6_0                               = 0x00038018,
+	MPEG_CLAMP_mask                                   = 0x03 << 0,
+	MPEG_CLAMP_shift                                  = 0,
+	    SQ_TEX_MPEG_CLAMP_OFF                         = 0x00,
+	    SQ_TEX_MPEG_9                                 = 0x01,
+	    SQ_TEX_MPEG_10                                = 0x02,
+	PERF_MODULATION_mask                              = 0x07 << 5,
+	PERF_MODULATION_shift                             = 5,
+	INTERLACED_bit                                    = 1 << 8,
+	SQ_TEX_RESOURCE_WORD6_0__TYPE_mask                = 0x03 << 30,
+	SQ_TEX_RESOURCE_WORD6_0__TYPE_shift               = 30,
+	    SQ_TEX_VTX_INVALID_TEXTURE                    = 0x00,
+	    SQ_TEX_VTX_INVALID_BUFFER                     = 0x01,
+	    SQ_TEX_VTX_VALID_TEXTURE                      = 0x02,
+	    SQ_TEX_VTX_VALID_BUFFER                       = 0x03,
+    SQ_VTX_CONSTANT_WORD6_0                               = 0x00038018,
+	SQ_VTX_CONSTANT_WORD6_0__TYPE_mask                = 0x03 << 30,
+	SQ_VTX_CONSTANT_WORD6_0__TYPE_shift               = 30,
+/* 	    SQ_TEX_VTX_INVALID_TEXTURE                    = 0x00, */
+/* 	    SQ_TEX_VTX_INVALID_BUFFER                     = 0x01, */
+/* 	    SQ_TEX_VTX_VALID_TEXTURE                      = 0x02, */
+/* 	    SQ_TEX_VTX_VALID_BUFFER                       = 0x03, */
+    SQ_TEX_SAMPLER_WORD0_0                                = 0x0003c000,
+	SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_mask              = 0x07 << 0,
+	SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_shift             = 0,
+	    SQ_TEX_WRAP                                   = 0x00,
+	    SQ_TEX_MIRROR                                 = 0x01,
+	    SQ_TEX_CLAMP_LAST_TEXEL                       = 0x02,
+	    SQ_TEX_MIRROR_ONCE_LAST_TEXEL                 = 0x03,
+	    SQ_TEX_CLAMP_HALF_BORDER                      = 0x04,
+	    SQ_TEX_MIRROR_ONCE_HALF_BORDER                = 0x05,
+	    SQ_TEX_CLAMP_BORDER                           = 0x06,
+	    SQ_TEX_MIRROR_ONCE_BORDER                     = 0x07,
+	CLAMP_Y_mask                                      = 0x07 << 3,
+	CLAMP_Y_shift                                     = 3,
+/* 	    SQ_TEX_WRAP                                   = 0x00, */
+/* 	    SQ_TEX_MIRROR                                 = 0x01, */
+/* 	    SQ_TEX_CLAMP_LAST_TEXEL                       = 0x02, */
+/* 	    SQ_TEX_MIRROR_ONCE_LAST_TEXEL                 = 0x03, */
+/* 	    SQ_TEX_CLAMP_HALF_BORDER                      = 0x04, */
+/* 	    SQ_TEX_MIRROR_ONCE_HALF_BORDER                = 0x05, */
+/* 	    SQ_TEX_CLAMP_BORDER                           = 0x06, */
+/* 	    SQ_TEX_MIRROR_ONCE_BORDER                     = 0x07, */
+	CLAMP_Z_mask                                      = 0x07 << 6,
+	CLAMP_Z_shift                                     = 6,
+/* 	    SQ_TEX_WRAP                                   = 0x00, */
+/* 	    SQ_TEX_MIRROR                                 = 0x01, */
+/* 	    SQ_TEX_CLAMP_LAST_TEXEL                       = 0x02, */
+/* 	    SQ_TEX_MIRROR_ONCE_LAST_TEXEL                 = 0x03, */
+/* 	    SQ_TEX_CLAMP_HALF_BORDER                      = 0x04, */
+/* 	    SQ_TEX_MIRROR_ONCE_HALF_BORDER                = 0x05, */
+/* 	    SQ_TEX_CLAMP_BORDER                           = 0x06, */
+/* 	    SQ_TEX_MIRROR_ONCE_BORDER                     = 0x07, */
+	XY_MAG_FILTER_mask                                = 0x07 << 9,
+	XY_MAG_FILTER_shift                               = 9,
+	    SQ_TEX_XY_FILTER_POINT                        = 0x00,
+	    SQ_TEX_XY_FILTER_BILINEAR                     = 0x01,
+	    SQ_TEX_XY_FILTER_BICUBIC                      = 0x02,
+	XY_MIN_FILTER_mask                                = 0x07 << 12,
+	XY_MIN_FILTER_shift                               = 12,
+/* 	    SQ_TEX_XY_FILTER_POINT                        = 0x00, */
+/* 	    SQ_TEX_XY_FILTER_BILINEAR                     = 0x01, */
+/* 	    SQ_TEX_XY_FILTER_BICUBIC                      = 0x02, */
+	Z_FILTER_mask                                     = 0x03 << 15,
+	Z_FILTER_shift                                    = 15,
+	    SQ_TEX_Z_FILTER_NONE                          = 0x00,
+	    SQ_TEX_Z_FILTER_POINT                         = 0x01,
+	    SQ_TEX_Z_FILTER_LINEAR                        = 0x02,
+	MIP_FILTER_mask                                   = 0x03 << 17,
+	MIP_FILTER_shift                                  = 17,
+/* 	    SQ_TEX_Z_FILTER_NONE                          = 0x00, */
+/* 	    SQ_TEX_Z_FILTER_POINT                         = 0x01, */
+/* 	    SQ_TEX_Z_FILTER_LINEAR                        = 0x02, */
+	BORDER_COLOR_TYPE_mask                            = 0x03 << 22,
+	BORDER_COLOR_TYPE_shift                           = 22,
+	    SQ_TEX_BORDER_COLOR_TRANS_BLACK               = 0x00,
+	    SQ_TEX_BORDER_COLOR_OPAQUE_BLACK              = 0x01,
+	    SQ_TEX_BORDER_COLOR_OPAQUE_WHITE              = 0x02,
+	    SQ_TEX_BORDER_COLOR_REGISTER                  = 0x03,
+	POINT_SAMPLING_CLAMP_bit                          = 1 << 24,
+	TEX_ARRAY_OVERRIDE_bit                            = 1 << 25,
+	DEPTH_COMPARE_FUNCTION_mask                       = 0x07 << 26,
+	DEPTH_COMPARE_FUNCTION_shift                      = 26,
+	    SQ_TEX_DEPTH_COMPARE_NEVER                    = 0x00,
+	    SQ_TEX_DEPTH_COMPARE_LESS                     = 0x01,
+	    SQ_TEX_DEPTH_COMPARE_EQUAL                    = 0x02,
+	    SQ_TEX_DEPTH_COMPARE_LESSEQUAL                = 0x03,
+	    SQ_TEX_DEPTH_COMPARE_GREATER                  = 0x04,
+	    SQ_TEX_DEPTH_COMPARE_NOTEQUAL                 = 0x05,
+	    SQ_TEX_DEPTH_COMPARE_GREATEREQUAL             = 0x06,
+	    SQ_TEX_DEPTH_COMPARE_ALWAYS                   = 0x07,
+	CHROMA_KEY_mask                                   = 0x03 << 29,
+	CHROMA_KEY_shift                                  = 29,
+	    SQ_TEX_CHROMA_KEY_DISABLED                    = 0x00,
+	    SQ_TEX_CHROMA_KEY_KILL                        = 0x01,
+	    SQ_TEX_CHROMA_KEY_BLEND                       = 0x02,
+	LOD_USES_MINOR_AXIS_bit                           = 1 << 31,
+    SQ_TEX_SAMPLER_WORD1_0                                = 0x0003c004,
+	MIN_LOD_mask                                      = 0x3ff << 0,
+	MIN_LOD_shift                                     = 0,
+	MAX_LOD_mask                                      = 0x3ff << 10,
+	MAX_LOD_shift                                     = 10,
+	SQ_TEX_SAMPLER_WORD1_0__LOD_BIAS_mask             = 0xfff << 20,
+	SQ_TEX_SAMPLER_WORD1_0__LOD_BIAS_shift            = 20,
+    SQ_TEX_SAMPLER_WORD2_0                                = 0x0003c008,
+	LOD_BIAS_SEC_mask                                 = 0xfff << 0,
+	LOD_BIAS_SEC_shift                                = 0,
+	MC_COORD_TRUNCATE_bit                             = 1 << 12,
+	SQ_TEX_SAMPLER_WORD2_0__FORCE_DEGAMMA_bit         = 1 << 13,
+	HIGH_PRECISION_FILTER_bit                         = 1 << 14,
+	PERF_MIP_mask                                     = 0x07 << 15,
+	PERF_MIP_shift                                    = 15,
+	PERF_Z_mask                                       = 0x03 << 18,
+	PERF_Z_shift                                      = 18,
+	FETCH_4_bit                                       = 1 << 26,
+	SAMPLE_IS_PCF_bit                                 = 1 << 27,
+	SQ_TEX_SAMPLER_WORD2_0__TYPE_bit                  = 1 << 31,
+    SQ_VTX_BASE_VTX_LOC                                   = 0x0003cff0,
+    SQ_VTX_START_INST_LOC                                 = 0x0003cff4,
+    SQ_LOOP_CONST_DX10_0                                  = 0x0003e200,
+    SQ_LOOP_CONST_0                                       = 0x0003e200,
+	SQ_LOOP_CONST_0__COUNT_mask                       = 0xfff << 0,
+	SQ_LOOP_CONST_0__COUNT_shift                      = 0,
+	INIT_mask                                         = 0xfff << 12,
+	INIT_shift                                        = 12,
+	INC_mask                                          = 0xff << 24,
+	INC_shift                                         = 24,
+    SQ_BOOL_CONST_0                                       = 0x0003e380,
+	SQ_BOOL_CONST_0_num                               = 3,
+
+} ;
+
+#endif /* _AUTOREGS */
+
diff --git a/src/mesa/drivers/dri/r600/r600_reg_r6xx.h b/src/mesa/drivers/dri/r600/r600_reg_r6xx.h
new file mode 100644
index 0000000000..f7702c46de
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg_r6xx.h
@@ -0,0 +1,492 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _R600_REG_R6xx_H_
+#define _R600_REG_R6xx_H_
+
+/*
+ * Registers for R6xx chips that are not documented yet
+ */
+
+enum {
+
+    MM_INDEX                                              = 0x0000,
+    MM_DATA                                               = 0x0004,
+
+    SRBM_STATUS                                           = 0x0e50,
+	RLC_RQ_PENDING_bit                                = 1 << 3,
+	RCU_RQ_PENDING_bit                                = 1 << 4,
+	GRBM_RQ_PENDING_bit                               = 1 << 5,
+	HI_RQ_PENDING_bit                                 = 1 << 6,
+	IO_EXTERN_SIGNAL_bit                              = 1 << 7,
+	VMC_BUSY_bit                                      = 1 << 8,
+	MCB_BUSY_bit                                      = 1 << 9,
+	MCDZ_BUSY_bit                                     = 1 << 10,
+	MCDY_BUSY_bit                                     = 1 << 11,
+	MCDX_BUSY_bit                                     = 1 << 12,
+	MCDW_BUSY_bit                                     = 1 << 13,
+	SEM_BUSY_bit                                      = 1 << 14,
+	SRBM_STATUS__RLC_BUSY_bit                         = 1 << 15,
+	PDMA_BUSY_bit                                     = 1 << 16,
+	IH_BUSY_bit                                       = 1 << 17,
+	CSC_BUSY_bit                                      = 1 << 20,
+	CMC7_BUSY_bit                                     = 1 << 21,
+	CMC6_BUSY_bit                                     = 1 << 22,
+	CMC5_BUSY_bit                                     = 1 << 23,
+	CMC4_BUSY_bit                                     = 1 << 24,
+	CMC3_BUSY_bit                                     = 1 << 25,
+	CMC2_BUSY_bit                                     = 1 << 26,
+	CMC1_BUSY_bit                                     = 1 << 27,
+	CMC0_BUSY_bit                                     = 1 << 28,
+	BIF_BUSY_bit                                      = 1 << 29,
+	IDCT_BUSY_bit                                     = 1 << 30,
+
+    SRBM_READ_ERROR                                       = 0x0e98,
+	READ_ADDRESS_mask                                 = 0xffff << 2,
+	READ_ADDRESS_shift                                = 2,
+	READ_REQUESTER_HI_bit                             = 1 << 24,
+	READ_REQUESTER_GRBM_bit                           = 1 << 25,
+	READ_REQUESTER_RCU_bit                            = 1 << 26,
+	READ_REQUESTER_RLC_bit                            = 1 << 27,
+	READ_ERROR_bit                                    = 1 << 31,
+
+    SRBM_INT_STATUS                                       = 0x0ea4,
+	RDERR_INT_STAT_bit                                = 1 << 0,
+	GFX_CNTX_SWITCH_INT_STAT_bit                      = 1 << 1,
+    SRBM_INT_ACK                                          = 0x0ea8,
+	RDERR_INT_ACK_bit                                 = 1 << 0,
+	GFX_CNTX_SWITCH_INT_ACK_bit                       = 1 << 1,
+
+    R6XX_MC_VM_FB_LOCATION                                = 0x2180,
+
+    VENDOR_DEVICE_ID                                      = 0x4000,
+
+    D1GRPH_PRIMARY_SURFACE_ADDRESS                        = 0x6110,
+    D1GRPH_PITCH                                          = 0x6120,
+    D1GRPH_Y_END                                          = 0x6138,
+
+    GRBM_STATUS                                           = 0x8010,
+	CMDFIFO_AVAIL_mask                                = 0x1f << 0,
+	CMDFIFO_AVAIL_shift                               = 0,
+	SRBM_RQ_PENDING_bit                               = 1 << 5,
+	CP_RQ_PENDING_bit                                 = 1 << 6,
+	CF_RQ_PENDING_bit                                 = 1 << 7,
+	PF_RQ_PENDING_bit                                 = 1 << 8,
+	GRBM_EE_BUSY_bit                                  = 1 << 10,
+	GRBM_STATUS__VC_BUSY_bit                          = 1 << 11,
+	DB03_CLEAN_bit                                    = 1 << 12,
+	CB03_CLEAN_bit                                    = 1 << 13,
+	VGT_BUSY_NO_DMA_bit                               = 1 << 16,
+	GRBM_STATUS__VGT_BUSY_bit                         = 1 << 17,
+	TA03_BUSY_bit                                     = 1 << 18,
+	GRBM_STATUS__TC_BUSY_bit                          = 1 << 19,
+	SX_BUSY_bit                                       = 1 << 20,
+	SH_BUSY_bit                                       = 1 << 21,
+	SPI03_BUSY_bit                                    = 1 << 22,
+	SMX_BUSY_bit                                      = 1 << 23,
+	SC_BUSY_bit                                       = 1 << 24,
+	PA_BUSY_bit                                       = 1 << 25,
+	DB03_BUSY_bit                                     = 1 << 26,
+	CR_BUSY_bit                                       = 1 << 27,
+	CP_COHERENCY_BUSY_bit                             = 1 << 28,
+	GRBM_STATUS__CP_BUSY_bit                          = 1 << 29,
+	CB03_BUSY_bit                                     = 1 << 30,
+	GUI_ACTIVE_bit                                    = 1 << 31,
+    GRBM_STATUS2                                          = 0x8014,
+	CR_CLEAN_bit                                      = 1 << 0,
+	SMX_CLEAN_bit                                     = 1 << 1,
+	SPI0_BUSY_bit                                     = 1 << 8,
+	SPI1_BUSY_bit                                     = 1 << 9,
+	SPI2_BUSY_bit                                     = 1 << 10,
+	SPI3_BUSY_bit                                     = 1 << 11,
+	TA0_BUSY_bit                                      = 1 << 12,
+	TA1_BUSY_bit                                      = 1 << 13,
+	TA2_BUSY_bit                                      = 1 << 14,
+	TA3_BUSY_bit                                      = 1 << 15,
+	DB0_BUSY_bit                                      = 1 << 16,
+	DB1_BUSY_bit                                      = 1 << 17,
+	DB2_BUSY_bit                                      = 1 << 18,
+	DB3_BUSY_bit                                      = 1 << 19,
+	CB0_BUSY_bit                                      = 1 << 20,
+	CB1_BUSY_bit                                      = 1 << 21,
+	CB2_BUSY_bit                                      = 1 << 22,
+	CB3_BUSY_bit                                      = 1 << 23,
+    GRBM_SOFT_RESET                                       = 0x8020,
+	SOFT_RESET_CP_bit                                 = 1 << 0,
+	SOFT_RESET_CB_bit                                 = 1 << 1,
+	SOFT_RESET_CR_bit                                 = 1 << 2,
+	SOFT_RESET_DB_bit                                 = 1 << 3,
+	SOFT_RESET_PA_bit                                 = 1 << 5,
+	SOFT_RESET_SC_bit                                 = 1 << 6,
+	SOFT_RESET_SMX_bit                                = 1 << 7,
+	SOFT_RESET_SPI_bit                                = 1 << 8,
+	SOFT_RESET_SH_bit                                 = 1 << 9,
+	SOFT_RESET_SX_bit                                 = 1 << 10,
+	SOFT_RESET_TC_bit                                 = 1 << 11,
+	SOFT_RESET_TA_bit                                 = 1 << 12,
+	SOFT_RESET_VC_bit                                 = 1 << 13,
+	SOFT_RESET_VGT_bit                                = 1 << 14,
+	SOFT_RESET_GRBM_GCA_bit                           = 1 << 15,
+
+    WAIT_UNTIL                                            = 0x8040,
+	WAIT_CP_DMA_IDLE_bit                              = 1 << 8,
+	WAIT_CMDFIFO_bit                                  = 1 << 10,
+	WAIT_2D_IDLE_bit                                  = 1 << 14,
+	WAIT_3D_IDLE_bit                                  = 1 << 15,
+	WAIT_2D_IDLECLEAN_bit                             = 1 << 16,
+	WAIT_3D_IDLECLEAN_bit                             = 1 << 17,
+	WAIT_EXTERN_SIG_bit                               = 1 << 19,
+	CMDFIFO_ENTRIES_mask                              = 0x1f << 20,
+	CMDFIFO_ENTRIES_shift                             = 20,
+
+    GRBM_READ_ERROR                                       = 0x8058,
+/* 	READ_ADDRESS_mask                                 = 0xffff << 2, */
+/* 	READ_ADDRESS_shift                                = 2, */
+	READ_REQUESTER_SRBM_bit                           = 1 << 28,
+	READ_REQUESTER_CP_bit                             = 1 << 29,
+	READ_REQUESTER_WU_POLL_bit                        = 1 << 30,
+/* 	READ_ERROR_bit                                    = 1 << 31, */
+
+    SCRATCH_REG0		                          = 0x8500,
+    SCRATCH_REG1		                          = 0x8504,
+    SCRATCH_REG2		                          = 0x8508,
+    SCRATCH_REG3		                          = 0x850c,
+    SCRATCH_REG4		                          = 0x8510,
+    SCRATCH_REG5		                          = 0x8514,
+    SCRATCH_REG6		                          = 0x8518,
+    SCRATCH_REG7		                          = 0x851c,
+    SCRATCH_UMSK		                          = 0x8540,
+    SCRATCH_ADDR		                          = 0x8544,
+
+    CP_COHER_CNTL                                         = 0x85f0,
+	DEST_BASE_0_ENA_bit                               = 1 << 0,
+	DEST_BASE_1_ENA_bit                               = 1 << 1,
+	SO0_DEST_BASE_ENA_bit                             = 1 << 2,
+	SO1_DEST_BASE_ENA_bit                             = 1 << 3,
+	SO2_DEST_BASE_ENA_bit                             = 1 << 4,
+	SO3_DEST_BASE_ENA_bit                             = 1 << 5,
+	CB0_DEST_BASE_ENA_bit                             = 1 << 6,
+	CB1_DEST_BASE_ENA_bit                             = 1 << 7,
+	CB2_DEST_BASE_ENA_bit                             = 1 << 8,
+	CB3_DEST_BASE_ENA_bit                             = 1 << 9,
+	CB4_DEST_BASE_ENA_bit                             = 1 << 10,
+	CB5_DEST_BASE_ENA_bit                             = 1 << 11,
+	CB6_DEST_BASE_ENA_bit                             = 1 << 12,
+	CB7_DEST_BASE_ENA_bit                             = 1 << 13,
+	DB_DEST_BASE_ENA_bit                              = 1 << 14,
+	CR_DEST_BASE_ENA_bit                              = 1 << 15,
+	TC_ACTION_ENA_bit                                 = 1 << 23,
+	VC_ACTION_ENA_bit                                 = 1 << 24,
+	CB_ACTION_ENA_bit                                 = 1 << 25,
+	DB_ACTION_ENA_bit                                 = 1 << 26,
+	SH_ACTION_ENA_bit                                 = 1 << 27,
+	SMX_ACTION_ENA_bit                                = 1 << 28,
+	CR0_ACTION_ENA_bit                                = 1 << 29,
+	CR1_ACTION_ENA_bit                                = 1 << 30,
+	CR2_ACTION_ENA_bit                                = 1 << 31,
+    CP_COHER_SIZE                                         = 0x85f4,
+    CP_COHER_BASE                                         = 0x85f8,
+    CP_COHER_STATUS                                       = 0x85fc,
+	MATCHING_GFX_CNTX_mask                            = 0xff << 0,
+	MATCHING_GFX_CNTX_shift                           = 0,
+	MATCHING_CR_CNTX_mask                             = 0xffff << 8,
+	MATCHING_CR_CNTX_shift                            = 8,
+	STATUS_bit                                        = 1 << 31,
+
+    CP_STALLED_STAT1                                      = 0x8674,
+	RBIU_TO_DMA_NOT_RDY_TO_RCV_bit                    = 1 << 0,
+	RBIU_TO_IBS_NOT_RDY_TO_RCV_bit                    = 1 << 1,
+	RBIU_TO_SEM_NOT_RDY_TO_RCV_bit                    = 1 << 2,
+	RBIU_TO_2DREGS_NOT_RDY_TO_RCV_bit                 = 1 << 3,
+	RBIU_TO_MEMWR_NOT_RDY_TO_RCV_bit                  = 1 << 4,
+	RBIU_TO_MEMRD_NOT_RDY_TO_RCV_bit                  = 1 << 5,
+	RBIU_TO_EOPD_NOT_RDY_TO_RCV_bit                   = 1 << 6,
+	RBIU_TO_RECT_NOT_RDY_TO_RCV_bit                   = 1 << 7,
+	RBIU_TO_STRMO_NOT_RDY_TO_RCV_bit                  = 1 << 8,
+	RBIU_TO_PSTAT_NOT_RDY_TO_RCV_bit                  = 1 << 9,
+	MIU_WAITING_ON_RDREQ_FREE_bit                     = 1 << 16,
+	MIU_WAITING_ON_WRREQ_FREE_bit                     = 1 << 17,
+	MIU_NEEDS_AVAIL_WRREQ_PHASE_bit                   = 1 << 18,
+	RCIU_WAITING_ON_GRBM_FREE_bit                     = 1 << 24,
+	RCIU_WAITING_ON_VGT_FREE_bit                      = 1 << 25,
+	RCIU_STALLED_ON_ME_READ_bit                       = 1 << 26,
+	RCIU_STALLED_ON_DMA_READ_bit                      = 1 << 27,
+	RCIU_HALTED_BY_REG_VIOLATION_bit                  = 1 << 28,
+    CP_STALLED_STAT2                                      = 0x8678,
+	PFP_TO_CSF_NOT_RDY_TO_RCV_bit                     = 1 << 0,
+	PFP_TO_MEQ_NOT_RDY_TO_RCV_bit                     = 1 << 1,
+	PFP_TO_VGT_NOT_RDY_TO_RCV_bit                     = 1 << 2,
+	PFP_HALTED_BY_INSTR_VIOLATION_bit                 = 1 << 3,
+	MULTIPASS_IB_PENDING_IN_PFP_bit                   = 1 << 4,
+	ME_BRUSH_WC_NOT_RDY_TO_RCV_bit                    = 1 << 8,
+	ME_STALLED_ON_BRUSH_LOGIC_bit                     = 1 << 9,
+	CR_CNTX_NOT_AVAIL_TO_ME_bit                       = 1 << 10,
+	GFX_CNTX_NOT_AVAIL_TO_ME_bit                      = 1 << 11,
+	ME_RCIU_NOT_RDY_TO_RCV_bit                        = 1 << 12,
+	ME_TO_CONST_NOT_RDY_TO_RCV_bit                    = 1 << 13,
+	ME_WAITING_DATA_FROM_PFP_bit                      = 1 << 14,
+	ME_WAITING_ON_PARTIAL_FLUSH_bit                   = 1 << 15,
+	RECT_FIFO_NEEDS_CR_RECT_DONE_bit                  = 1 << 16,
+	RECT_FIFO_NEEDS_WR_CONFIRM_bit                    = 1 << 17,
+	EOPD_FIFO_NEEDS_SC_EOP_DONE_bit                   = 1 << 18,
+	EOPD_FIFO_NEEDS_SMX_EOP_DONE_bit                  = 1 << 19,
+	EOPD_FIFO_NEEDS_WR_CONFIRM_bit                    = 1 << 20,
+	EOPD_FIFO_NEEDS_SIGNAL_SEM_bit                    = 1 << 21,
+	SO_NUMPRIM_FIFO_NEEDS_SOADDR_bit                  = 1 << 22,
+	SO_NUMPRIM_FIFO_NEEDS_NUMPRIM_bit                 = 1 << 23,
+	PIPE_STATS_FIFO_NEEDS_SAMPLE_bit                  = 1 << 24,
+	SURF_SYNC_NEEDS_IDLE_CNTXS_bit                    = 1 << 30,
+	SURF_SYNC_NEEDS_ALL_CLEAN_bit                     = 1 << 31,
+    CP_BUSY_STAT                                          = 0x867c,
+	REG_BUS_FIFO_BUSY_bit                             = 1 << 0,
+	RING_FETCHING_DATA_bit                            = 1 << 1,
+	INDR1_FETCHING_DATA_bit                           = 1 << 2,
+	INDR2_FETCHING_DATA_bit                           = 1 << 3,
+	STATE_FETCHING_DATA_bit                           = 1 << 4,
+	PRED_FETCHING_DATA_bit                            = 1 << 5,
+	COHER_CNTR_NEQ_ZERO_bit                           = 1 << 6,
+	PFP_PARSING_PACKETS_bit                           = 1 << 7,
+	ME_PARSING_PACKETS_bit                            = 1 << 8,
+	RCIU_PFP_BUSY_bit                                 = 1 << 9,
+	RCIU_ME_BUSY_bit                                  = 1 << 10,
+	OUTSTANDING_READ_TAGS_bit                         = 1 << 11,
+	SEM_CMDFIFO_NOT_EMPTY_bit                         = 1 << 12,
+	SEM_FAILED_AND_HOLDING_bit                        = 1 << 13,
+	SEM_POLLING_FOR_PASS_bit                          = 1 << 14,
+	_3D_BUSY_bit                                      = 1 << 15,
+	_2D_BUSY_bit                                      = 1 << 16,
+    CP_STAT                                               = 0x8680,
+	CSF_RING_BUSY_bit                                 = 1 << 0,
+	CSF_WPTR_POLL_BUSY_bit                            = 1 << 1,
+	CSF_INDIRECT1_BUSY_bit                            = 1 << 2,
+	CSF_INDIRECT2_BUSY_bit                            = 1 << 3,
+	CSF_STATE_BUSY_bit                                = 1 << 4,
+	CSF_PREDICATE_BUSY_bit                            = 1 << 5,
+	CSF_BUSY_bit                                      = 1 << 6,
+	MIU_RDREQ_BUSY_bit                                = 1 << 7,
+	MIU_WRREQ_BUSY_bit                                = 1 << 8,
+	ROQ_RING_BUSY_bit                                 = 1 << 9,
+	ROQ_INDIRECT1_BUSY_bit                            = 1 << 10,
+	ROQ_INDIRECT2_BUSY_bit                            = 1 << 11,
+	ROQ_STATE_BUSY_bit                                = 1 << 12,
+	ROQ_PREDICATE_BUSY_bit                            = 1 << 13,
+	ROQ_ALIGN_BUSY_bit                                = 1 << 14,
+	PFP_BUSY_bit                                      = 1 << 15,
+	MEQ_BUSY_bit                                      = 1 << 16,
+	ME_BUSY_bit                                       = 1 << 17,
+	QUERY_BUSY_bit                                    = 1 << 18,
+	SEMAPHORE_BUSY_bit                                = 1 << 19,
+	INTERRUPT_BUSY_bit                                = 1 << 20,
+	SURFACE_SYNC_BUSY_bit                             = 1 << 21,
+	DMA_BUSY_bit                                      = 1 << 22,
+	RCIU_BUSY_bit                                     = 1 << 23,
+	CP_STAT__CP_BUSY_bit                              = 1 << 31,
+
+    CP_ME_CNTL                                            = 0x86d8,
+	ME_STATMUX_mask                                   = 0xff << 0,
+	ME_STATMUX_shift                                  = 0,
+	ME_HALT_bit                                       = 1 << 28,
+    CP_ME_STATUS                                          = 0x86dc,
+
+    CP_RB_RPTR                                            = 0x8700,
+	RB_RPTR_mask                                      = 0xfffff << 0,
+	RB_RPTR_shift                                     = 0,
+    CP_RB_WPTR_DELAY                                      = 0x8704,
+	PRE_WRITE_TIMER_mask                              = 0xfffffff << 0,
+	PRE_WRITE_TIMER_shift                             = 0,
+	PRE_WRITE_LIMIT_mask                              = 0x0f << 28,
+	PRE_WRITE_LIMIT_shift                             = 28,
+
+    CP_ROQ_RB_STAT                                        = 0x8780,
+	ROQ_RPTR_PRIMARY_mask                             = 0x3ff << 0,
+	ROQ_RPTR_PRIMARY_shift                            = 0,
+	ROQ_WPTR_PRIMARY_mask                             = 0x3ff << 16,
+	ROQ_WPTR_PRIMARY_shift                            = 16,
+    CP_ROQ_IB1_STAT                                       = 0x8784,
+	ROQ_RPTR_INDIRECT1_mask                           = 0x3ff << 0,
+	ROQ_RPTR_INDIRECT1_shift                          = 0,
+	ROQ_WPTR_INDIRECT1_mask                           = 0x3ff << 16,
+	ROQ_WPTR_INDIRECT1_shift                          = 16,
+    CP_ROQ_IB2_STAT                                       = 0x8788,
+	ROQ_RPTR_INDIRECT2_mask                           = 0x3ff << 0,
+	ROQ_RPTR_INDIRECT2_shift                          = 0,
+	ROQ_WPTR_INDIRECT2_mask                           = 0x3ff << 16,
+	ROQ_WPTR_INDIRECT2_shift                          = 16,
+
+    CP_MEQ_STAT                                           = 0x8794,
+	MEQ_RPTR_mask                                     = 0x3ff << 0,
+	MEQ_RPTR_shift                                    = 0,
+	MEQ_WPTR_mask                                     = 0x3ff << 16,
+	MEQ_WPTR_shift                                    = 16,
+
+    CC_GC_SHADER_PIPE_CONFIG                              = 0x8950,
+	INACTIVE_QD_PIPES_mask                            = 0xff << 8,
+	INACTIVE_QD_PIPES_shift                           = 8,
+	    R6XX_MAX_QD_PIPES                             = 8,
+	INACTIVE_SIMDS_mask                               = 0xff << 16,
+	INACTIVE_SIMDS_shift                              = 16,
+	    R6XX_MAX_SIMDS                                = 8,
+    GC_USER_SHADER_PIPE_CONFIG                            = 0x8954,
+
+    VC_ENHANCE                                            = 0x9714,
+    DB_DEBUG                                              = 0x9830,
+        PREZ_MUST_WAIT_FOR_POSTZ_DONE                     = 1 << 31,
+
+    DB_WATERMARKS                                         = 0x00009838,
+	DEPTH_FREE_mask                                   = 0x1f << 0,
+	DEPTH_FREE_shift                                  = 0,
+	DEPTH_FLUSH_mask                                  = 0x3f << 5,
+	DEPTH_FLUSH_shift                                 = 5,
+	FORCE_SUMMARIZE_mask                              = 0x0f << 11,
+	FORCE_SUMMARIZE_shift                             = 11,
+	DEPTH_PENDING_FREE_mask                           = 0x1f << 15,
+	DEPTH_PENDING_FREE_shift                          = 15,
+	DEPTH_CACHELINE_FREE_mask                         = 0x1f << 20,
+	DEPTH_CACHELINE_FREE_shift                        = 20,
+	EARLY_Z_PANIC_DISABLE_bit                         = 1 << 25,
+	LATE_Z_PANIC_DISABLE_bit                          = 1 << 26,
+	RE_Z_PANIC_DISABLE_bit                            = 1 << 27,
+	DB_EXTRA_DEBUG_mask                               = 0x0f << 28,
+	DB_EXTRA_DEBUG_shift                              = 28,
+
+    CP_RB_BASE                                            = 0xc100,
+    CP_RB_CNTL                                            = 0xc104,
+        RB_BUFSZ_mask                                     = 0x3f << 0,
+    CP_RB_WPTR                                            = 0xc114,
+	RB_WPTR_mask                                      = 0xfffff << 0,
+	RB_WPTR_shift                                     = 0,
+    CP_RB_RPTR_WR                                         = 0xc108,
+	RB_RPTR_WR_mask                                   = 0xfffff << 0,
+	RB_RPTR_WR_shift                                  = 0,
+
+    CP_INT_STATUS                                         = 0xc128,
+	DISABLE_CNTX_SWITCH_INT_STAT_bit                  = 1 << 0,
+	ENABLE_CNTX_SWITCH_INT_STAT_bit                   = 1 << 1,
+	SEM_SIGNAL_INT_STAT_bit                           = 1 << 18,
+	CNTX_BUSY_INT_STAT_bit                            = 1 << 19,
+	CNTX_EMPTY_INT_STAT_bit                           = 1 << 20,
+	WAITMEM_SEM_INT_STAT_bit                          = 1 << 21,
+	PRIV_INSTR_INT_STAT_bit                           = 1 << 22,
+	PRIV_REG_INT_STAT_bit                             = 1 << 23,
+	OPCODE_ERROR_INT_STAT_bit                         = 1 << 24,
+	SCRATCH_INT_STAT_bit                              = 1 << 25,
+	TIME_STAMP_INT_STAT_bit                           = 1 << 26,
+	RESERVED_BIT_ERROR_INT_STAT_bit                   = 1 << 27,
+	DMA_INT_STAT_bit                                  = 1 << 28,
+	IB2_INT_STAT_bit                                  = 1 << 29,
+	IB1_INT_STAT_bit                                  = 1 << 30,
+	RB_INT_STAT_bit                                   = 1 << 31,
+
+//  SX_ALPHA_TEST_CONTROL                                 = 0x00028410,
+	ALPHA_FUNC__REF_NEVER                             = 0,
+	ALPHA_FUNC__REF_ALWAYS                            = 7,
+//  DB_SHADER_CONTROL                                     = 0x0002880c,
+	Z_ORDER__EARLY_Z_THEN_LATE_Z                      = 2,
+//  PA_SU_SC_MODE_CNTL                                    = 0x00028814,
+//	POLY_MODE_mask                                    = 0x03 << 3,
+	POLY_MODE__TRIANGLES = 0, POLY_MODE__DUAL_MODE,
+//	POLYMODE_FRONT_PTYPE_mask                         = 0x07 << 5,
+	POLYMODE_PTYPE__POINTS = 0, POLYMODE_PTYPE__LINES, POLYMODE_PTYPE__TRIANGLES,
+    PA_SC_AA_SAMPLE_LOCS_8S_WD1_M                         = 0x00028c20,
+    DB_SRESULTS_COMPARE_STATE0                            = 0x00028d28,	/* See autoregs: DB_SRESULTS_COMPARE_STATE1 */
+//  DB_SRESULTS_COMPARE_STATE1                            = 0x00028d2c,
+    DB_ALPHA_TO_MASK                                      = 0x00028d44,
+	ALPHA_TO_MASK_ENABLE                              = 1 << 0,
+	ALPHA_TO_MASK_OFFSET0_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET0_shift                       = 8,
+	ALPHA_TO_MASK_OFFSET1_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET1_shift                       = 10,
+	ALPHA_TO_MASK_OFFSET2_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET2_shift                       = 12,
+	ALPHA_TO_MASK_OFFSET3_mask                        = 0x03 << 8,
+	ALPHA_TO_MASK_OFFSET3_shift                       = 14,
+
+//  SQ_VTX_CONSTANT_WORD2_0                               = 0x00038008,
+//    	SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask         = 0x3f << 20,
+	FMT_INVALID=0,      FMT_8,          FMT_4_4,            FMT_3_3_2,
+	                    FMT_16=5,       FMT_16_FLOAT,       FMT_8_8,
+	FMT_5_6_5,          FMT_6_5_5,      FMT_1_5_5_5,        FMT_4_4_4_4,
+	FMT_5_5_5_1,        FMT_32,         FMT_32_FLOAT,       FMT_16_16,
+	FMT_16_16_FLOAT=16, FMT_8_24,       FMT_8_24_FLOAT,     FMT_24_8,
+	FMT_24_8_FLOAT,     FMT_10_11_11,   FMT_10_11_11_FLOAT, FMT_11_11_10,
+	FMT_11_11_10_FLOAT, FMT_2_10_10_10, FMT_8_8_8_8,        FMT_10_10_10_2,
+	FMT_X24_8_32_FLOAT, FMT_32_32,      FMT_32_32_FLOAT,    FMT_16_16_16_16,
+	FMT_16_16_16_16_FLOAT=32,           FMT_32_32_32_32=34, FMT_32_32_32_32_FLOAT,
+	                    FMT_1 = 37,                         FMT_GB_GR=39,
+	FMT_BG_RG,          FMT_32_AS_8,    FMT_32_AS_8_8,      FMT_5_9_9_9_SHAREDEXP,
+	FMT_8_8_8,          FMT_16_16_16,   FMT_16_16_16_FLOAT, FMT_32_32_32,
+	FMT_32_32_32_FLOAT=48,
+
+//  High level register file lengths
+    SQ_ALU_CONSTANT                                       = SQ_ALU_CONSTANT0_0,	/* 256 PS, 256 VS */
+    SQ_ALU_CONSTANT_ps_num                                = 256,
+    SQ_ALU_CONSTANT_vs_num                                = 256,
+    SQ_ALU_CONSTANT_all_num                               = 512,
+    SQ_ALU_CONSTANT_offset                                = 16,
+    SQ_ALU_CONSTANT_ps                                    = 0,
+    SQ_ALU_CONSTANT_vs                                    = SQ_ALU_CONSTANT_ps + SQ_ALU_CONSTANT_ps_num,
+    SQ_TEX_RESOURCE                                       = SQ_TEX_RESOURCE_WORD0_0,	/* 160 PS, 160 VS, 16 FS, 160 GS */
+    SQ_TEX_RESOURCE_ps_num                                = 160,
+    SQ_TEX_RESOURCE_vs_num                                = 160,
+    SQ_TEX_RESOURCE_fs_num                                = 16,
+    SQ_TEX_RESOURCE_gs_num                                = 160,
+    SQ_TEX_RESOURCE_all_num                               = 496,
+    SQ_TEX_RESOURCE_offset                                = 28,
+    SQ_TEX_RESOURCE_ps                                    = 0,
+    SQ_TEX_RESOURCE_vs                                    = SQ_TEX_RESOURCE_ps + SQ_TEX_RESOURCE_ps_num,
+    SQ_TEX_RESOURCE_fs                                    = SQ_TEX_RESOURCE_vs + SQ_TEX_RESOURCE_vs_num,
+    SQ_TEX_RESOURCE_gs                                    = SQ_TEX_RESOURCE_fs + SQ_TEX_RESOURCE_fs_num,
+    SQ_VTX_RESOURCE                                       = SQ_VTX_CONSTANT_WORD0_0,	/* 160 PS, 160 VS, 16 FS, 160 GS */
+    SQ_VTX_RESOURCE_ps_num                                = 160,
+    SQ_VTX_RESOURCE_vs_num                                = 160,
+    SQ_VTX_RESOURCE_fs_num                                = 16,
+    SQ_VTX_RESOURCE_gs_num                                = 160,
+    SQ_VTX_RESOURCE_all_num                               = 496,
+    SQ_VTX_RESOURCE_offset                                = 28,
+    SQ_VTX_RESOURCE_ps                                    = 0,
+    SQ_VTX_RESOURCE_vs                                    = SQ_VTX_RESOURCE_ps + SQ_VTX_RESOURCE_ps_num,
+    SQ_VTX_RESOURCE_fs                                    = SQ_VTX_RESOURCE_vs + SQ_VTX_RESOURCE_vs_num,
+    SQ_VTX_RESOURCE_gs                                    = SQ_VTX_RESOURCE_fs + SQ_VTX_RESOURCE_fs_num,
+    SQ_TEX_SAMPLER_WORD                                   = SQ_TEX_SAMPLER_WORD0_0,	/* 18 per PS, VS, GS */
+    SQ_TEX_SAMPLER_WORD_ps_num                            = 18,
+    SQ_TEX_SAMPLER_WORD_vs_num                            = 18,
+    SQ_TEX_SAMPLER_WORD_gs_num                            = 18,
+    SQ_TEX_SAMPLER_WORD_all_num                           = 54,
+    SQ_TEX_SAMPLER_WORD_offset                            = 12,
+    SQ_TEX_SAMPLER_WORD_ps                                = 0,
+    SQ_TEX_SAMPLER_WORD_vs                                = SQ_TEX_SAMPLER_WORD_ps + SQ_TEX_SAMPLER_WORD_ps_num,
+    SQ_TEX_SAMPLER_WORD_gs                                = SQ_TEX_SAMPLER_WORD_vs + SQ_TEX_SAMPLER_WORD_vs_num,
+    SQ_LOOP_CONST                                         = SQ_LOOP_CONST_0,		/* 32 per PS, VS, GS */
+    SQ_LOOP_CONST_ps_num                                  = 32,
+    SQ_LOOP_CONST_vs_num                                  = 32,
+    SQ_LOOP_CONST_gs_num                                  = 32,
+    SQ_LOOP_CONST_all_num                                 = 96,
+    SQ_LOOP_CONST_offset                                  = 4,
+    SQ_LOOP_CONST_ps                                      = 0,
+    SQ_LOOP_CONST_vs                                      = SQ_LOOP_CONST_ps + SQ_LOOP_CONST_ps_num,
+    SQ_LOOP_CONST_gs                                      = SQ_LOOP_CONST_vs + SQ_LOOP_CONST_vs_num,
+} ;
+
+
+#endif
diff --git a/src/mesa/drivers/dri/r600/r600_reg_r7xx.h b/src/mesa/drivers/dri/r600/r600_reg_r7xx.h
new file mode 100644
index 0000000000..e5c01c861a
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_reg_r7xx.h
@@ -0,0 +1,149 @@
+/*
+ * RadeonHD R6xx, R7xx Register documentation
+ *
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ * Copyright (C) 2008-2009  Matthias Hopf
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _R600_REG_R7xx_H_
+#define _R600_REG_R7xx_H_
+
+/*
+ * Register update for R7xx chips
+ */
+
+enum {
+
+    R7XX_MC_VM_FB_LOCATION                                = 0x00002024,
+
+//  GRBM_STATUS                                           = 0x00008010,
+	R7XX_TA_BUSY_bit                                  = 1 << 14,
+
+    R7xx_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ                     = 0x00008d8c,
+	RING0_OFFSET_mask                                 = 0xff << 0,
+	RING0_OFFSET_shift                                = 0,
+	ISOLATE_ES_ENABLE_bit                             = 1 << 12,
+	ISOLATE_GS_ENABLE_bit                             = 1 << 13,
+	VS_PC_LIMIT_ENABLE_bit                            = 1 << 14,
+
+//  SQ_ALU_WORD0                                          = 0x00008dfc,
+//	SRC0_SEL_mask                                     = 0x1ff << 0,
+// 	SRC1_SEL_mask                                     = 0x1ff << 13,
+	    R7xx_SQ_ALU_SRC_1_DBL_L                       = 0xf4,
+	    R7xx_SQ_ALU_SRC_1_DBL_M                       = 0xf5,
+	    R7xx_SQ_ALU_SRC_0_5_DBL_L                     = 0xf6,
+	    R7xx_SQ_ALU_SRC_0_5_DBL_M                     = 0xf7,
+// 	INDEX_MODE_mask                                   = 0x07 << 26,
+	    R7xx_SQ_INDEX_GLOBAL                          = 0x05,
+	    R7xx_SQ_INDEX_GLOBAL_AR_X                     = 0x06,
+    R6xx_SQ_ALU_WORD1_OP2                                 = 0x00008dfc,
+    R7xx_SQ_ALU_WORD1_OP2_V2                              = 0x00008dfc,
+	R6xx_FOG_MERGE_bit                                = 1 << 5,
+	R6xx_OMOD_mask                                    = 0x03 << 6,
+	R7xx_OMOD_mask                                    = 0x03 << 5,
+	R6xx_OMOD_shift                                   = 6,
+	R7xx_OMOD_shift                                   = 5,
+	R6xx_SQ_ALU_WORD1_OP2__ALU_INST_mask              = 0x3ff << 8,
+	R7xx_SQ_ALU_WORD1_OP2_V2__ALU_INST_mask           = 0x7ff << 7,
+	R6xx_SQ_ALU_WORD1_OP2__ALU_INST_shift             = 8,
+	R7xx_SQ_ALU_WORD1_OP2_V2__ALU_INST_shift          = 7,
+	    R7xx_SQ_OP2_INST_FREXP_64                     = 0x07,
+	    R7xx_SQ_OP2_INST_ADD_64                       = 0x17,
+	    R7xx_SQ_OP2_INST_MUL_64                       = 0x1b,
+	    R7xx_SQ_OP2_INST_FLT64_TO_FLT32               = 0x1c,
+	    R7xx_SQ_OP2_INST_FLT32_TO_FLT64               = 0x1d,
+	    R7xx_SQ_OP2_INST_LDEXP_64                     = 0x7a,
+	    R7xx_SQ_OP2_INST_FRACT_64                     = 0x7b,
+	    R7xx_SQ_OP2_INST_PRED_SETGT_64                = 0x7c,
+	    R7xx_SQ_OP2_INST_PRED_SETE_64                 = 0x7d,
+	    R7xx_SQ_OP2_INST_PRED_SETGE_64                = 0x7e,
+//  SQ_ALU_WORD1_OP3                                      = 0x00008dfc,
+//	SRC2_SEL_mask                                     = 0x1ff << 0,
+//	    R7xx_SQ_ALU_SRC_1_DBL_L                       = 0xf4,
+//	    R7xx_SQ_ALU_SRC_1_DBL_M                       = 0xf5,
+//	    R7xx_SQ_ALU_SRC_0_5_DBL_L                     = 0xf6,
+//	    R7xx_SQ_ALU_SRC_0_5_DBL_M                     = 0xf7,
+// 	SQ_ALU_WORD1_OP3__ALU_INST_mask                   = 0x1f << 13,
+	    R7xx_SQ_OP3_INST_MULADD_64                    = 0x08,
+	    R7xx_SQ_OP3_INST_MULADD_64_M2                 = 0x09,
+	    R7xx_SQ_OP3_INST_MULADD_64_M4                 = 0x0a,
+	    R7xx_SQ_OP3_INST_MULADD_64_D2                 = 0x0b,
+//  SQ_CF_ALU_WORD1                                       = 0x00008dfc,
+	R6xx_USES_WATERFALL_bit                           = 1 << 25,
+	R7xx_SQ_CF_ALU_WORD1__ALT_CONST_bit               = 1 << 25,
+//  SQ_CF_ALLOC_EXPORT_WORD0                              = 0x00008dfc,
+//	ARRAY_BASE_mask                                   = 0x1fff << 0,
+//	TYPE_mask                                         = 0x03 << 13,
+//	    SQ_EXPORT_PARAM                               = 0x02,
+//	    X_UNUSED_FOR_SX_EXPORTS                       = 0x03,
+//	ELEM_SIZE_mask                                    = 0x03 << 30,
+//  SQ_CF_ALLOC_EXPORT_WORD1                              = 0x00008dfc,
+//	SQ_CF_ALLOC_EXPORT_WORD1__CF_INST_mask            = 0x7f << 23,
+	    R7xx_SQ_CF_INST_MEM_EXPORT                    = 0x3a,
+//  SQ_CF_WORD1                                           = 0x00008dfc,
+//	SQ_CF_WORD1__COUNT_mask                           = 0x07 << 10,
+	R7xx_COUNT_3_bit                                  = 1 << 19,
+//	SQ_CF_WORD1__CF_INST_mask                         = 0x7f << 23,
+	    R7xx_SQ_CF_INST_END_PROGRAM                   = 0x19,
+	    R7xx_SQ_CF_INST_WAIT_ACK                      = 0x1a,
+	    R7xx_SQ_CF_INST_TEX_ACK                       = 0x1b,
+	    R7xx_SQ_CF_INST_VTX_ACK                       = 0x1c,
+	    R7xx_SQ_CF_INST_VTX_TC_ACK                    = 0x1d,
+//  SQ_VTX_WORD0                                          = 0x00008dfc,
+//	VTX_INST_mask                                     = 0x1f << 0,
+	    R7xx_SQ_VTX_INST_MEM                          = 0x02,
+//  SQ_VTX_WORD2                                          = 0x00008dfc,
+	R7xx_SQ_VTX_WORD2__ALT_CONST_bit                  = 1 << 20,
+
+//  SQ_TEX_WORD0                                          = 0x00008dfc,
+//	TEX_INST_mask                                     = 0x1f << 0,
+	    R7xx_X_MEMORY_READ                            = 0x02,
+	    R7xx_SQ_TEX_INST_KEEP_GRADIENTS               = 0x0a,
+	    R7xx_X_FETCH4_LOAD4_INSTRUCTION_FOR_DX10_1    = 0x0f,
+	R7xx_SQ_TEX_WORD0__ALT_CONST_bit                  = 1 << 24,
+
+    R7xx_PA_SC_EDGERULE                                   = 0x00028230,
+    R7xx_SPI_THREAD_GROUPING                              = 0x000286c8,
+	PS_GROUPING_mask                                  = 0x1f << 0,
+	PS_GROUPING_shift                                 = 0,
+	VS_GROUPING_mask                                  = 0x1f << 8,
+	VS_GROUPING_shift                                 = 8,
+	GS_GROUPING_mask                                  = 0x1f << 16,
+	GS_GROUPING_shift                                 = 16,
+	ES_GROUPING_mask                                  = 0x1f << 24,
+	ES_GROUPING_shift                                 = 24,
+    R7xx_CB_SHADER_CONTROL                                = 0x000287a0,
+	RT0_ENABLE_bit                                    = 1 << 0,
+	RT1_ENABLE_bit                                    = 1 << 1,
+	RT2_ENABLE_bit                                    = 1 << 2,
+	RT3_ENABLE_bit                                    = 1 << 3,
+	RT4_ENABLE_bit                                    = 1 << 4,
+	RT5_ENABLE_bit                                    = 1 << 5,
+	RT6_ENABLE_bit                                    = 1 << 6,
+	RT7_ENABLE_bit                                    = 1 << 7,
+//  DB_ALPHA_TO_MASK                                      = 0x00028d44,
+	R7xx_OFFSET_ROUND_bit                             = 1 << 16,
+//  SQ_TEX_SAMPLER_MISC_0                                 = 0x0003d03c,
+	R7xx_TRUNCATE_COORD_bit                           = 1 << 9,
+	R7xx_DISABLE_CUBE_WRAP_bit                        = 1 << 10,
+
+} ;
+
+#endif /* _R600_REG_R7xx_H_ */
diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c
new file mode 100644
index 0000000000..43d9f641af
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_tex.c
@@ -0,0 +1,437 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+*/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mipmap.h"
+#include "main/simple_list.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+
+#include "texmem.h"
+
+#include "r600_context.h"
+#include "r700_state.h"
+#include "radeon_mipmap_tree.h"
+#include "r600_tex.h"
+
+#include "xmlpool.h"
+
+
+static unsigned int translate_wrap_mode(GLenum wrapmode)
+{
+	switch(wrapmode) {
+	case GL_REPEAT: return SQ_TEX_WRAP;
+	case GL_CLAMP: return SQ_TEX_CLAMP_HALF_BORDER;
+	case GL_CLAMP_TO_EDGE: return SQ_TEX_CLAMP_LAST_TEXEL;
+	case GL_CLAMP_TO_BORDER: return SQ_TEX_CLAMP_BORDER;
+	case GL_MIRRORED_REPEAT: return SQ_TEX_MIRROR_ONCE_HALF_BORDER;
+	case GL_MIRROR_CLAMP_EXT: return SQ_TEX_MIRROR;
+	case GL_MIRROR_CLAMP_TO_EDGE_EXT: return SQ_TEX_MIRROR_ONCE_BORDER;
+	case GL_MIRROR_CLAMP_TO_BORDER_EXT: return SQ_TEX_MIRROR_ONCE_LAST_TEXEL;
+	default:
+		_mesa_problem(NULL, "bad wrap mode in %s", __FUNCTION__);
+		return 0;
+	}
+}
+
+
+/**
+ * Update the cached hardware registers based on the current texture wrap modes.
+ *
+ * \param t Texture object whose wrap modes are to be set
+ */
+static void r600UpdateTexWrap(radeonTexObjPtr t)
+{
+	struct gl_texture_object *tObj = &t->base;
+
+        SETfield(t->SQ_TEX_SAMPLER0, translate_wrap_mode(tObj->WrapS),
+                 SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_shift, SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_mask);
+
+	if (tObj->Target != GL_TEXTURE_1D) {
+		SETfield(t->SQ_TEX_SAMPLER0, translate_wrap_mode(tObj->WrapT),
+			 CLAMP_Y_shift, CLAMP_Y_mask);
+
+		if (tObj->Target == GL_TEXTURE_3D)
+			SETfield(t->SQ_TEX_SAMPLER0, translate_wrap_mode(tObj->WrapR),
+				 CLAMP_Z_shift, CLAMP_Z_mask);
+	}
+}
+
+static void r600SetTexDefaultState(radeonTexObjPtr t)
+{
+        /* Init text object to default states. */
+        t->SQ_TEX_RESOURCE0              = 0;
+        SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_2D, DIM_shift, DIM_mask);
+        SETfield(t->SQ_TEX_RESOURCE0, ARRAY_LINEAR_GENERAL,
+                 SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_shift, SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_mask);
+        CLEARbit(t->SQ_TEX_RESOURCE0, TILE_TYPE_bit);
+
+        t->SQ_TEX_RESOURCE1                = 0;
+        SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+                 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+        t->SQ_TEX_RESOURCE2                = 0;
+        t->SQ_TEX_RESOURCE3                = 0;
+
+        t->SQ_TEX_RESOURCE4                   = 0;
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_X_shift, FORMAT_COMP_X_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_Y_shift, FORMAT_COMP_Y_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_Z_shift, FORMAT_COMP_Z_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
+                 FORMAT_COMP_W_shift, FORMAT_COMP_W_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_NUM_FORMAT_NORM,
+                 SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_shift, SQ_TEX_RESOURCE_WORD4_0__NUM_FORMAT_ALL_mask);
+        CLEARbit(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__SRF_MODE_ALL_bit);
+        CLEARbit(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit);
+        SETfield(t->SQ_TEX_RESOURCE4, SQ_ENDIAN_NONE,
+                 SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_shift, SQ_TEX_RESOURCE_WORD4_0__ENDIAN_SWAP_mask);
+        SETfield(t->SQ_TEX_RESOURCE4, 1, REQUEST_SIZE_shift, REQUEST_SIZE_mask);
+        t->SQ_TEX_RESOURCE4 |= SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift
+		              |SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift
+		              |SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift
+		              |SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift;
+        SETfield(t->SQ_TEX_RESOURCE4, 0, BASE_LEVEL_shift, BASE_LEVEL_mask); /* mip-maps */
+
+        t->SQ_TEX_RESOURCE5 = 0;
+        t->SQ_TEX_RESOURCE6 = 0;
+
+        SETfield(t->SQ_TEX_RESOURCE6, SQ_TEX_VTX_VALID_TEXTURE,
+                 SQ_TEX_RESOURCE_WORD6_0__TYPE_shift, SQ_TEX_RESOURCE_WORD6_0__TYPE_mask);
+
+        /* Initialize sampler registers */
+        t->SQ_TEX_SAMPLER0                           = 0;
+        t->SQ_TEX_SAMPLER0 |=
+                         SQ_TEX_WRAP << SQ_TEX_SAMPLER_WORD0_0__CLAMP_X_shift
+                        |SQ_TEX_WRAP << CLAMP_Y_shift
+                        |SQ_TEX_WRAP << CLAMP_Z_shift
+                        |SQ_TEX_XY_FILTER_POINT << XY_MAG_FILTER_shift
+                        |SQ_TEX_XY_FILTER_POINT << XY_MIN_FILTER_shift
+                        |SQ_TEX_Z_FILTER_NONE << Z_FILTER_shift
+                        |SQ_TEX_Z_FILTER_NONE << MIP_FILTER_shift
+                        |SQ_TEX_BORDER_COLOR_TRANS_BLACK << BORDER_COLOR_TYPE_shift;
+
+        t->SQ_TEX_SAMPLER1 = 0x7FF << MAX_LOD_shift;
+
+        t->SQ_TEX_SAMPLER2                          = 0;
+        SETbit(t->SQ_TEX_SAMPLER2, SQ_TEX_SAMPLER_WORD2_0__TYPE_bit);
+}
+
+
+static GLuint aniso_filter(GLfloat anisotropy)
+{
+#if 0
+	if (anisotropy >= 16.0) {
+		return R300_TX_MAX_ANISO_16_TO_1;
+	} else if (anisotropy >= 8.0) {
+		return R300_TX_MAX_ANISO_8_TO_1;
+	} else if (anisotropy >= 4.0) {
+		return R300_TX_MAX_ANISO_4_TO_1;
+	} else if (anisotropy >= 2.0) {
+		return R300_TX_MAX_ANISO_2_TO_1;
+	} else {
+		return R300_TX_MAX_ANISO_1_TO_1;
+	}
+#endif
+	return 0;
+}
+
+/**
+ * Set the texture magnification and minification modes.
+ *
+ * \param t Texture whose filter modes are to be set
+ * \param minf Texture minification mode
+ * \param magf Texture magnification mode
+ * \param anisotropy Maximum anisotropy level
+ */
+static void r600SetTexFilter(radeonTexObjPtr t, GLenum minf, GLenum magf, GLfloat anisotropy)
+{
+	/* Force revalidation to account for switches from/to mipmapping. */
+	t->validated = GL_FALSE;
+
+	/* Note that EXT_texture_filter_anisotropic is extremely vague about
+	 * how anisotropic filtering interacts with the "normal" filter modes.
+	 * When anisotropic filtering is enabled, we override min and mag
+	 * filter settings completely. This includes driconf's settings.
+	 */
+	if (anisotropy >= 2.0 && (minf != GL_NEAREST) && (magf != GL_NEAREST)) {
+		/*t->pp_txfilter |= R300_TX_MAG_FILTER_ANISO
+			| R300_TX_MIN_FILTER_ANISO
+			| R300_TX_MIN_FILTER_MIP_LINEAR
+			| aniso_filter(anisotropy);*/
+		if (RADEON_DEBUG & DEBUG_TEXTURE)
+			fprintf(stderr, "Using maximum anisotropy of %f\n", anisotropy);
+		return;
+	}
+
+	switch (minf) {
+	case GL_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_None,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_None,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_NEAREST_MIPMAP_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Point,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_NEAREST_MIPMAP_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Linear,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_LINEAR_MIPMAP_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Point,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	case GL_LINEAR_MIPMAP_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MIN_FILTER_shift, XY_MIN_FILTER_mask);
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_MipFilter_Linear,
+			 MIP_FILTER_shift, MIP_FILTER_mask);
+		break;
+	}
+
+	/* Note we don't have 3D mipmaps so only use the mag filter setting
+	 * to set the 3D texture filter mode.
+	 */
+	switch (magf) {
+	case GL_NEAREST:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Point,
+			 XY_MAG_FILTER_shift, XY_MAG_FILTER_mask);
+		break;
+	case GL_LINEAR:
+		SETfield(t->SQ_TEX_SAMPLER0, TEX_XYFilter_Linear,
+			 XY_MAG_FILTER_shift, XY_MAG_FILTER_mask);
+		break;
+	}
+}
+
+static void r600SetTexBorderColor(radeonTexObjPtr t, const GLfloat color[4])
+{
+#if 0
+	GLubyte c[4];
+	CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
+	CLAMPED_FLOAT_TO_UBYTE(c[1], color[1]);
+	CLAMPED_FLOAT_TO_UBYTE(c[2], color[2]);
+	CLAMPED_FLOAT_TO_UBYTE(c[3], color[3]);
+	t->pp_border_color = PACK_COLOR_8888(c[3], c[0], c[1], c[2]);
+#endif
+}
+
+/**
+ * Changes variables and flags for a state update, which will happen at the
+ * next UpdateTextureState
+ */
+
+static void r600TexParameter(GLcontext * ctx, GLenum target,
+			     struct gl_texture_object *texObj,
+			     GLenum pname, const GLfloat * params)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+
+	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+		fprintf(stderr, "%s( %s )\n", __FUNCTION__,
+			_mesa_lookup_enum_by_nr(pname));
+	}
+
+	switch (pname) {
+	case GL_TEXTURE_MIN_FILTER:
+	case GL_TEXTURE_MAG_FILTER:
+	case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+		r600SetTexFilter(t, texObj->MinFilter, texObj->MagFilter, texObj->MaxAnisotropy);
+		break;
+
+	case GL_TEXTURE_WRAP_S:
+	case GL_TEXTURE_WRAP_T:
+	case GL_TEXTURE_WRAP_R:
+		r600UpdateTexWrap(t);
+		break;
+
+	case GL_TEXTURE_BORDER_COLOR:
+		r600SetTexBorderColor(t, texObj->BorderColor);
+		break;
+
+	case GL_TEXTURE_BASE_LEVEL:
+	case GL_TEXTURE_MAX_LEVEL:
+	case GL_TEXTURE_MIN_LOD:
+	case GL_TEXTURE_MAX_LOD:
+		/* This isn't the most efficient solution but there doesn't appear to
+		 * be a nice alternative.  Since there's no LOD clamping,
+		 * we just have to rely on loading the right subset of mipmap levels
+		 * to simulate a clamped LOD.
+		 */
+		if (t->mt) {
+			radeon_miptree_unreference(t->mt);
+			t->mt = 0;
+			t->validated = GL_FALSE;
+		}
+		break;
+
+	case GL_DEPTH_TEXTURE_MODE:
+		if (!texObj->Image[0][texObj->BaseLevel])
+			return;
+		if (texObj->Image[0][texObj->BaseLevel]->TexFormat->BaseFormat
+		    == GL_DEPTH_COMPONENT) {
+			r600SetDepthTexMode(texObj);
+			break;
+		} else {
+			/* If the texture isn't a depth texture, changing this
+			 * state won't cause any changes to the hardware.
+			 * Don't force a flush of texture state.
+			 */
+			return;
+		}
+
+	default:
+		return;
+	}
+}
+
+static void r600DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj)
+{
+	context_t* rmesa = R700_CONTEXT(ctx);
+	radeonTexObj* t = radeon_tex_obj(texObj);
+
+	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+		fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+			(void *)texObj,
+			_mesa_lookup_enum_by_nr(texObj->Target));
+	}
+
+	if (rmesa) {
+		int i;
+		radeon_firevertices(&rmesa->radeon);
+
+		for(i = 0; i < R700_MAX_TEXTURE_UNITS; ++i)
+			if (rmesa->hw.textures[i] == t)
+				rmesa->hw.textures[i] = 0;
+	}
+
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = 0;
+	}
+	_mesa_delete_texture_object(ctx, texObj);
+}
+
+/**
+ * Allocate a new texture object.
+ * Called via ctx->Driver.NewTextureObject.
+ * Note: this function will be called during context creation to
+ * allocate the default texture objects.
+ * Fixup MaxAnisotropy according to user preference.
+ */
+static struct gl_texture_object *r600NewTextureObject(GLcontext * ctx,
+						      GLuint name,
+						      GLenum target)
+{
+	context_t* rmesa = R700_CONTEXT(ctx);
+	radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
+
+	if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) {
+		fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__,
+			t, _mesa_lookup_enum_by_nr(target));
+	}
+
+	_mesa_initialize_texture_object(&t->base, name, target);
+	t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+	/* Initialize hardware state */
+	r600SetTexDefaultState(t);
+	r600UpdateTexWrap(t);
+	r600SetTexFilter(t, t->base.MinFilter, t->base.MagFilter, t->base.MaxAnisotropy);
+	r600SetTexBorderColor(t, t->base.BorderColor);
+
+	return &t->base;
+}
+
+void r600InitTextureFuncs(struct dd_function_table *functions)
+{
+	/* Note: we only plug in the functions we implement in the driver
+	 * since _mesa_init_driver_functions() was already called.
+	 */
+	functions->NewTextureImage = radeonNewTextureImage;
+	functions->FreeTexImageData = radeonFreeTexImageData;
+	functions->MapTexture = radeonMapTexture;
+	functions->UnmapTexture = radeonUnmapTexture;
+
+	functions->ChooseTextureFormat = radeonChooseTextureFormat_mesa;
+	functions->TexImage1D = radeonTexImage1D;
+	functions->TexImage2D = radeonTexImage2D;
+	functions->TexImage3D = radeonTexImage3D;
+	functions->TexSubImage1D = radeonTexSubImage1D;
+	functions->TexSubImage2D = radeonTexSubImage2D;
+	functions->TexSubImage3D = radeonTexSubImage3D;
+	functions->GetTexImage = radeonGetTexImage;
+	functions->GetCompressedTexImage = radeonGetCompressedTexImage;
+	functions->NewTextureObject = r600NewTextureObject;
+	functions->DeleteTexture = r600DeleteTexture;
+	functions->IsTextureResident = driIsTextureResident;
+
+	functions->TexParameter = r600TexParameter;
+
+	functions->CompressedTexImage2D = radeonCompressedTexImage2D;
+	functions->CompressedTexSubImage2D = radeonCompressedTexSubImage2D;
+
+	functions->GenerateMipmap = radeonGenerateMipmap;
+
+	driInitTextureFormats();
+}
diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.h b/src/mesa/drivers/dri/r600/r600_tex.h
index 3add775b82..fb0e1a023e 100644
--- a/src/mesa/drivers/dri/r300/radeon_ioctl.h
+++ b/src/mesa/drivers/dri/r600/r600_tex.h
@@ -32,26 +32,32 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Keith Whitwell <keith@tungstengraphics.com>
  */
 
-#ifndef __RADEON_IOCTL_H__
-#define __RADEON_IOCTL_H__
-
-#include "main/simple_list.h"
-#include "radeon_dri.h"
-#include "radeon_lock.h"
-
-#include "xf86drm.h"
-#include "drm.h"
-#if 0
-#include "r200context.h"
-#endif
-#include "radeon_drm.h"
-
-extern void radeonCopyBuffer(__DRIdrawablePrivate * drawable,
-			     const drm_clip_rect_t	* rect);
-extern void radeonPageFlip(__DRIdrawablePrivate * drawable);
-extern void radeonFlush(GLcontext * ctx);
-extern void radeonFinish(GLcontext * ctx);
-extern void radeonWaitForIdleLocked(radeonContextPtr radeon);
-extern uint32_t radeonGetAge(radeonContextPtr radeon);
-
-#endif				/* __RADEON_IOCTL_H__ */
+#ifndef __r600_TEX_H__
+#define __r600_TEX_H__
+
+/* TODO : review this after texture load code. */
+#define R700_BLIT_WIDTH_BYTES 1024
+/* The BASE_ADDRESS and MIP_ADDRESS fields are 256-byte-aligned */
+#define R700_TEXTURE_ALIGNMENT_MASK     0x255
+/* Texel pitch is 8 alignment. */
+#define R700_TEXEL_PITCH_ALIGNMENT_MASK 0x7
+
+#define R700_MAX_TEXTURE_UNITS 8 /* TODO : should be 16, lets make it work, review later */
+
+extern void r600SetDepthTexMode(struct gl_texture_object *tObj);
+
+extern void r600SetTexBuffer(__DRIcontext *pDRICtx, GLint target,
+			     __DRIdrawable *dPriv);
+
+extern void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
+			      GLint format, __DRIdrawable *dPriv);
+
+extern void r600SetTexOffset(__DRIcontext *pDRICtx, GLint texname,
+			     unsigned long long offset, GLint depth,
+			     GLuint pitch);
+
+extern GLboolean r600ValidateBuffers(GLcontext * ctx);
+
+extern void r600InitTextureFuncs(struct dd_function_table *functions);
+
+#endif				/* __r600_TEX_H__ */
diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c
new file mode 100644
index 0000000000..2466aa9595
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r600_texstate.c
@@ -0,0 +1,802 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file
+ *
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ *
+ * \todo Enable R300 texture tiling code?
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+#include "main/enums.h"
+
+#include "r600_context.h"
+#include "r700_state.h"
+#include "radeon_mipmap_tree.h"
+#include "r600_tex.h"
+
+void r600UpdateTextureState(GLcontext * ctx);
+
+void r600UpdateTextureState(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	struct gl_texture_unit *texUnit;
+	struct radeon_tex_obj *t;
+	GLuint    unit;
+
+	for (unit = 0; unit < R700_MAX_TEXTURE_UNITS; unit++) {
+		texUnit = &ctx->Texture.Unit[unit];
+		t = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
+
+		if (texUnit->_ReallyEnabled) {
+			if (!t)
+				continue;
+			r700->textures[unit] = t;
+		}
+	}
+}
+
+static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_format)
+{
+	radeonTexObj *t = radeon_tex_obj(tObj);
+
+	t->SQ_TEX_RESOURCE4 &= ~( SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask
+				  |SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask
+				  |SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask
+				  |SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask );
+
+	switch (mesa_format) /* This is mesa format. */
+	{
+	case MESA_FORMAT_RGBA8888:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGBA8888_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ARGB8888:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ARGB8888_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGB888:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGB565:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGB565_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ARGB4444:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_4_4_4_4,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ARGB4444_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_4_4_4_4,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ARGB1555:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_1_5_5_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ARGB1555_REV:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_1_5_5_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_AL88:
+	case MESA_FORMAT_AL88_REV: /* TODO : Check this. */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGB332:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_3_3_2,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_A8: /* ZERO, ZERO, ZERO, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_L8: /* X, X, X, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_I8: /* X, X, X, X */
+	case MESA_FORMAT_CI8:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+		/* YUV422 TODO conversion */  /* X, Y, Z, ONE, G8R8_G8B8 */
+		/*
+		  case MESA_FORMAT_YCBCR:
+		  t->SQ_TEX_RESOURCE1.bitfields.DATA_FORMAT = ;
+		  break;
+		*/
+		/* VUY422 TODO conversion */  /* X, Y, Z, ONE, G8R8_G8B8 */
+		/*
+		  case MESA_FORMAT_YCBCR_REV:
+		  t->SQ_TEX_RESOURCE1.bitfields.DATA_FORMAT = ;
+		  break;
+		*/
+	case MESA_FORMAT_RGB_DXT1: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_DXT1: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_DXT3: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_DXT5: /* not supported yet */
+
+		break;
+	case MESA_FORMAT_RGBA_FLOAT32:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_32_32_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGBA_FLOAT16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_16_16_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGB_FLOAT32: /* X, Y, Z, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_32_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_RGB_FLOAT16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_16_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ALPHA_FLOAT32: /* ZERO, ZERO, ZERO, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_ALPHA_FLOAT16: /* ZERO, ZERO, ZERO, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_LUMINANCE_FLOAT32: /* X, X, X, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_LUMINANCE_FLOAT16: /* X, X, X, ONE */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_INTENSITY_FLOAT32: /* X, X, X, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_32_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_INTENSITY_FLOAT16: /* X, X, X, X */
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_16_FLOAT,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		break;
+	case MESA_FORMAT_Z16:
+	case MESA_FORMAT_Z24_S8:
+	case MESA_FORMAT_Z32:
+		switch (mesa_format) {
+		case MESA_FORMAT_Z16:
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_16,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+			break;
+		case MESA_FORMAT_Z24_S8:
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_24_8,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+			break;
+		case MESA_FORMAT_Z32:
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_32,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+			break;
+		};
+		switch (tObj->DepthMode) {
+		case GL_LUMINANCE:  /* X, X, X, ONE */
+			t->SQ_TEX_RESOURCE4 |=
+				(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+				|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+			break;
+		case GL_INTENSITY:  /* X, X, X, X */
+			t->SQ_TEX_RESOURCE4 |=
+				(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+			break;
+		case GL_ALPHA:     /* ZERO, ZERO, ZERO, X */
+			t->SQ_TEX_RESOURCE4 |=
+				(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+				|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+				|(SQ_SEL_0 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+			break;
+		default:
+			return GL_FALSE;
+		}
+		break;
+	default:
+		/* Not supported format */
+		return GL_FALSE;
+	};
+
+	return GL_TRUE;
+}
+
+void r600SetDepthTexMode(struct gl_texture_object *tObj)
+{
+	const GLuint *format;
+	radeonTexObjPtr t;
+
+	if (!tObj)
+		return;
+
+	t = radeon_tex_obj(tObj);
+
+	r600GetTexFormat(tObj, tObj->Image[0][tObj->BaseLevel]->TexFormat->MesaFormat);
+
+}
+
+/**
+ * Compute the cached hardware register values for the given texture object.
+ *
+ * \param rmesa Context pointer
+ * \param t the r300 texture object
+ */
+static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *texObj)
+{
+	radeonTexObj *t = radeon_tex_obj(texObj);
+	const struct gl_texture_image *firstImage;
+	int firstlevel = t->mt ? t->mt->firstLevel : 0;
+	GLuint uTexelPitch;
+
+	firstImage = t->base.Image[0][firstlevel];
+
+	if (!t->image_override) {
+		if (!r600GetTexFormat(texObj, firstImage->TexFormat->MesaFormat)) {
+			_mesa_problem(NULL, "unexpected texture format in %s",
+				      __FUNCTION__);
+			return;
+		}
+	}
+
+	if (t->image_override && t->bo)
+		return;
+
+	switch (texObj->Target) {
+        case GL_TEXTURE_1D:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_1D, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        case GL_TEXTURE_2D:
+        case GL_TEXTURE_RECTANGLE_NV:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_2D, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        case GL_TEXTURE_3D:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_3D, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, firstImage->Depth - 1, // ???
+			 TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        case GL_TEXTURE_CUBE_MAP:
+		SETfield(t->SQ_TEX_RESOURCE0, SQ_TEX_DIM_CUBEMAP, DIM_shift, DIM_mask);
+		SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask);
+		break;
+        default:
+		_mesa_problem(NULL, "unexpected texture target type in %s", __FUNCTION__);
+		return;
+	}
+
+	uTexelPitch = (firstImage->Width + R700_TEXEL_PITCH_ALIGNMENT_MASK)
+		& ~R700_TEXEL_PITCH_ALIGNMENT_MASK;
+
+	SETfield(t->SQ_TEX_RESOURCE0, (uTexelPitch/8)-1, PITCH_shift, PITCH_mask);
+	SETfield(t->SQ_TEX_RESOURCE0, firstImage->Width - 1,
+		 TEX_WIDTH_shift, TEX_WIDTH_mask);
+	SETfield(t->SQ_TEX_RESOURCE1, firstImage->Height - 1,
+		 TEX_HEIGHT_shift, TEX_HEIGHT_mask);
+
+}
+
+/**
+ * Ensure the given texture is ready for rendering.
+ *
+ * Mostly this means populating the texture object's mipmap tree.
+ */
+static GLboolean r600_validate_texture(GLcontext * ctx, struct gl_texture_object *texObj)
+{
+	context_t *rmesa = R700_CONTEXT(ctx);
+	radeonTexObj *t = radeon_tex_obj(texObj);
+
+	if (!radeon_validate_texture_miptree(ctx, texObj))
+		return GL_FALSE;
+
+	/* Configure the hardware registers (more precisely, the cached version
+	 * of the hardware registers). */
+	setup_hardware_state(rmesa, texObj);
+
+	t->validated = GL_TRUE;
+	return GL_TRUE;
+}
+
+/**
+ * Ensure all enabled and complete textures are uploaded along with any buffers being used.
+ */
+GLboolean r600ValidateBuffers(GLcontext * ctx)
+{
+	context_t *rmesa = R700_CONTEXT(ctx);
+	struct radeon_renderbuffer *rrb;
+	int i;
+	int ret;
+
+	radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
+
+	rrb = radeon_get_colorbuffer(&rmesa->radeon);
+	/* color buffer */
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
+	}
+
+	/* depth buffer */
+	rrb = radeon_get_depthbuffer(&rmesa->radeon);
+	if (rrb && rrb->bo) {
+		radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+						  rrb->bo, 0,
+						  RADEON_GEM_DOMAIN_VRAM);
+	}
+	
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+		radeonTexObj *t;
+
+		if (!ctx->Texture.Unit[i]._ReallyEnabled)
+			continue;
+
+		if (!r600_validate_texture(ctx, ctx->Texture.Unit[i]._Current)) {
+			_mesa_warning(ctx,
+				      "failed to validate texture for unit %d.\n",
+				      i);
+		}
+		t = radeon_tex_obj(ctx->Texture.Unit[i]._Current);
+		if (t->image_override && t->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+		else if (t->mt->bo)
+			radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs,
+							  t->mt->bo,
+							  RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+	}
+
+	ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+	if (ret)
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+void r600SetTexOffset(__DRIcontext * pDRICtx, GLint texname,
+		      unsigned long long offset, GLint depth, GLuint pitch)
+{
+	context_t *rmesa = pDRICtx->driverPrivate;
+	struct gl_texture_object *tObj =
+	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
+	uint32_t pitch_val;
+
+	if (!tObj)
+		return;
+
+	t->image_override = GL_TRUE;
+
+	if (!offset)
+		return;
+
+	t->bo = NULL;
+	t->override_offset = offset;
+	pitch_val = pitch;
+	switch (depth) {
+	case 32:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		pitch_val /= 4;
+		break;
+	case 24:
+	default:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		pitch_val /= 4;
+		break;
+	case 16:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		pitch_val /= 2;
+		break;
+	}
+
+	pitch_val = (pitch_val + R700_TEXEL_PITCH_ALIGNMENT_MASK)
+		& ~R700_TEXEL_PITCH_ALIGNMENT_MASK;
+	SETfield(t->SQ_TEX_RESOURCE0, (pitch_val/8)-1, PITCH_shift, PITCH_mask);
+}
+
+void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format, __DRIdrawable *dPriv)
+{
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	context_t *rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) {
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+			t->SQ_TEX_RESOURCE4 |=
+				(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+				|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+				|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+				|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		} else {
+			SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+				 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+			t->SQ_TEX_RESOURCE4 |=
+				(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+				|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+				|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+				|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		}
+		pitch_val /= 4;
+		break;
+	case 3:
+	default:
+		// FMT_8_8_8 ???
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_W << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		pitch_val /= 4;
+		break;
+	case 2:
+		SETfield(t->SQ_TEX_RESOURCE1, FMT_5_6_5,
+			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+		t->SQ_TEX_RESOURCE4 |=
+			(SQ_SEL_Z << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift)
+			|(SQ_SEL_Y << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift)
+			|(SQ_SEL_X << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift)
+			|(SQ_SEL_1 << SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift);
+		pitch_val /= 2;
+		break;
+	}
+
+	pitch_val = (pitch_val + R700_TEXEL_PITCH_ALIGNMENT_MASK)
+		& ~R700_TEXEL_PITCH_ALIGNMENT_MASK;
+
+	SETfield(t->SQ_TEX_RESOURCE0, (pitch_val/8)-1, PITCH_shift, PITCH_mask);
+	SETfield(t->SQ_TEX_RESOURCE0, rb->base.Width - 1,
+		 TEX_WIDTH_shift, TEX_WIDTH_mask);
+	SETfield(t->SQ_TEX_RESOURCE1, rb->base.Height - 1,
+		 TEX_HEIGHT_shift, TEX_HEIGHT_mask);
+
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
+}
+
+void r600SetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+        r600SetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c
new file mode 100644
index 0000000000..1d41c5cf78
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_assembler.c
@@ -0,0 +1,4071 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/mtypes.h"
+#include "main/imports.h"
+
+#include "r600_context.h"
+#include "r700_debug.h"
+
+#include "r700_assembler.h"
+
+BITS addrmode_PVSDST(PVSDST * pPVSDST)
+{
+	return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1);
+}
+
+void setaddrmode_PVSDST(PVSDST * pPVSDST, BITS addrmode) 
+{
+	pPVSDST->addrmode0 = addrmode & 1;
+	pPVSDST->addrmode1 = (addrmode >> 1) & 1;
+}
+
+void nomask_PVSDST(PVSDST * pPVSDST) 
+{
+	pPVSDST->writex = pPVSDST->writey = pPVSDST->writez = pPVSDST->writew = 1;
+}
+
+BITS addrmode_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	return pPVSSRC->addrmode0 | ((BITS)pPVSSRC->addrmode1 << 1);
+}
+
+void setaddrmode_PVSSRC(PVSSRC* pPVSSRC, BITS addrmode) 
+{
+	pPVSSRC->addrmode0 = addrmode & 1;
+	pPVSSRC->addrmode1 = (addrmode >> 1) & 1;
+}
+
+
+void setswizzle_PVSSRC(PVSSRC* pPVSSRC, BITS swz) 
+{
+	pPVSSRC->swizzlex = 
+	pPVSSRC->swizzley = 
+	pPVSSRC->swizzlez = 
+	pPVSSRC->swizzlew = swz;
+}
+
+void noswizzle_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->swizzlex = SQ_SEL_X;
+	pPVSSRC->swizzley = SQ_SEL_Y;
+	pPVSSRC->swizzlez = SQ_SEL_Z;
+	pPVSSRC->swizzlew = SQ_SEL_W;
+}
+
+void
+swizzleagain_PVSSRC(PVSSRC * pPVSSRC, BITS x, BITS y, BITS z, BITS w)
+{
+    switch (x) 
+    {
+        case SQ_SEL_X: x = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: x = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: x = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: x = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    switch (y) 
+    {
+        case SQ_SEL_X: y = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: y = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: y = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: y = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    switch (z) 
+    {
+        case SQ_SEL_X: z = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: z = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: z = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: z = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    switch (w) 
+    {
+        case SQ_SEL_X: w = pPVSSRC->swizzlex; 
+            break;
+        case SQ_SEL_Y: w = pPVSSRC->swizzley; 
+            break;
+        case SQ_SEL_Z: w = pPVSSRC->swizzlez; 
+            break;
+        case SQ_SEL_W: w = pPVSSRC->swizzlew; 
+            break;
+        default:;
+    }
+
+    pPVSSRC->swizzlex = x;
+    pPVSSRC->swizzley = y;
+    pPVSSRC->swizzlez = z;
+    pPVSSRC->swizzlew = w;
+}
+
+void neg_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->negx = 1;
+	pPVSSRC->negy = 1;
+	pPVSSRC->negz = 1;
+	pPVSSRC->negw = 1;
+}
+
+void noneg_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->negx = 0;
+	pPVSSRC->negy = 0;
+	pPVSSRC->negz = 0;
+	pPVSSRC->negw = 0;
+}
+
+// negate argument (for SUB instead of ADD and alike)
+void flipneg_PVSSRC(PVSSRC* pPVSSRC) 
+{
+	pPVSSRC->negx = !pPVSSRC->negx;
+	pPVSSRC->negy = !pPVSSRC->negy;
+	pPVSSRC->negz = !pPVSSRC->negz;
+	pPVSSRC->negw = !pPVSSRC->negw;
+}
+
+void zerocomp_PVSSRC(PVSSRC* pPVSSRC, int c) 
+{
+	switch (c) 
+	{
+		case 0: pPVSSRC->swizzlex = SQ_SEL_0; pPVSSRC->negx = 0; break;
+		case 1: pPVSSRC->swizzley = SQ_SEL_0; pPVSSRC->negy = 0; break;
+		case 2: pPVSSRC->swizzlez = SQ_SEL_0; pPVSSRC->negz = 0; break;
+		case 3: pPVSSRC->swizzlew = SQ_SEL_0; pPVSSRC->negw = 0; break;
+		default:;
+	} 
+}
+
+void onecomp_PVSSRC(PVSSRC* pPVSSRC, int c) 
+{
+	switch (c) 
+	{
+		case 0: pPVSSRC->swizzlex = SQ_SEL_1; pPVSSRC->negx = 0; break;
+		case 1: pPVSSRC->swizzley = SQ_SEL_1; pPVSSRC->negy = 0; break;
+		case 2: pPVSSRC->swizzlez = SQ_SEL_1; pPVSSRC->negz = 0; break;
+		case 3: pPVSSRC->swizzlew = SQ_SEL_1; pPVSSRC->negw = 0; break;
+		default:;
+	} 
+}
+
+BITS is_misc_component_exported(VAP_OUT_VTX_FMT_0* pOutVTXFmt0)  
+{
+	  return (pOutVTXFmt0->point_size            |
+			  pOutVTXFmt0->edge_flag             |
+			  pOutVTXFmt0->rta_index             |
+			  pOutVTXFmt0->kill_flag             |
+			  pOutVTXFmt0->viewport_index);
+}
+
+BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) 
+{
+	  return (pFPOutFmt->depth            | 
+			  pFPOutFmt->stencil_ref      | 
+			  pFPOutFmt->mask             | 
+			  pFPOutFmt->coverage_to_mask);
+}
+
+GLboolean is_reduction_opcode(PVSDWORD* dest)
+{
+    if (dest->dst.op3 == 0) 
+    {
+        if ( (dest->dst.opcode == SQ_OP2_INST_DOT4 || dest->dst.opcode == SQ_OP2_INST_DOT4_IEEE) ) 
+        {
+            return GL_TRUE;
+        }
+    }
+    return GL_FALSE;
+}
+
+GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size)
+{
+    GLuint format = FMT_INVALID;
+	GLuint uiElemSize = 0;
+
+    switch (eType)
+    {
+        case GL_BYTE:
+        case GL_UNSIGNED_BYTE:
+			uiElemSize = 1;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_8; break;
+                case 2:
+                    format = FMT_8_8; break;
+                case 3:
+                    format = FMT_8_8_8; break;
+                case 4:
+                    format = FMT_8_8_8_8; break;
+                default:
+                    break;
+            }
+            break;
+
+        case GL_UNSIGNED_SHORT:
+        case GL_SHORT:
+			uiElemSize = 2;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_16; break;
+                case 2:
+                    format = FMT_16_16; break;
+                case 3:
+                    format = FMT_16_16_16; break;
+                case 4:
+                    format = FMT_16_16_16_16; break;
+                default:
+                    break;
+            }
+            break;
+
+        case GL_UNSIGNED_INT:
+        case GL_INT:
+			uiElemSize = 4;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_32; break;
+                case 2:
+                    format = FMT_32_32; break;
+                case 3:
+                    format = FMT_32_32_32; break;
+                case 4:
+                    format = FMT_32_32_32_32; break;
+                default:
+                    break;
+            }
+            break;
+
+        case GL_FLOAT:
+			uiElemSize = 4;
+			switch(nChannels)
+            {
+                case 1:
+                    format = FMT_32_FLOAT; break;
+                case 2:
+                    format = FMT_32_32_FLOAT; break;
+                case 3:
+                    format = FMT_32_32_32_FLOAT; break;
+                case 4:
+                    format = FMT_32_32_32_32_FLOAT; break;
+                default:
+                    break;
+            }
+			break;
+        case GL_DOUBLE:
+			uiElemSize = 8;
+            switch(nChannels)
+            {
+                case 1:
+                    format = FMT_32_FLOAT; break;
+                case 2:
+                    format = FMT_32_32_FLOAT; break;
+                case 3:
+                    format = FMT_32_32_32_FLOAT; break;
+                case 4:
+                    format = FMT_32_32_32_32_FLOAT; break;
+                default:
+                    break;
+            }
+            break;
+        default:
+			;
+            //GL_ASSERT_NO_CASE();
+    }
+
+    if(NULL != pClient_size)
+    {
+	    *pClient_size = uiElemSize * nChannels;
+    }
+
+    return(format);
+}
+
+unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm)
+{
+    if(pAsm->D.dst.op3)
+    {
+        return 3;
+    }
+
+    switch (pAsm->D.dst.opcode)
+    {
+    case SQ_OP2_INST_ADD:                          
+    case SQ_OP2_INST_MUL: 
+    case SQ_OP2_INST_MAX:
+    case SQ_OP2_INST_MIN:
+    //case SQ_OP2_INST_MAX_DX10:
+    //case SQ_OP2_INST_MIN_DX10:
+    case SQ_OP2_INST_SETGT:
+    case SQ_OP2_INST_SETGE:
+    case SQ_OP2_INST_PRED_SETE:
+    case SQ_OP2_INST_PRED_SETGT:
+    case SQ_OP2_INST_PRED_SETGE:
+    case SQ_OP2_INST_PRED_SETNE:
+    case SQ_OP2_INST_DOT4:
+    case SQ_OP2_INST_DOT4_IEEE:
+        return 2;  
+
+    case SQ_OP2_INST_MOV: 
+    case SQ_OP2_INST_FRACT:
+    case SQ_OP2_INST_FLOOR:
+    case SQ_OP2_INST_KILLGT:
+    case SQ_OP2_INST_EXP_IEEE:
+    case SQ_OP2_INST_LOG_CLAMPED:
+    case SQ_OP2_INST_LOG_IEEE:
+    case SQ_OP2_INST_RECIP_IEEE:
+    case SQ_OP2_INST_RECIPSQRT_IEEE:
+    case SQ_OP2_INST_FLT_TO_INT:
+    case SQ_OP2_INST_SIN:
+    case SQ_OP2_INST_COS:
+        return 1;
+        
+    default: r700_error(TODO_ASM_NEEDIMPINST, 
+                        "Need instruction operand number. \n");;
+    };
+
+    return 3;
+}
+
+int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader)
+{
+    GLuint i;
+
+    Init_R700_Shader(pShader);
+    pAsm->pR700Shader = pShader;
+    pAsm->currentShaderType = spt;
+
+	pAsm->cf_last_export_ptr   = NULL;
+
+	pAsm->cf_current_export_clause_ptr = NULL;
+	pAsm->cf_current_alu_clause_ptr    = NULL;
+	pAsm->cf_current_tex_clause_ptr    = NULL;
+	pAsm->cf_current_vtx_clause_ptr    = NULL;
+	pAsm->cf_current_cf_clause_ptr     = NULL;
+
+	// No clause has been created yet
+	pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE;
+
+	pAsm->number_of_colorandz_exports = 0;
+	pAsm->number_of_exports           = 0;
+	pAsm->number_of_export_opcodes    = 0;
+
+
+	pAsm->D.bits = 0;
+	pAsm->S[0].bits = 0;
+	pAsm->S[1].bits = 0;
+	pAsm->S[2].bits = 0;
+
+	pAsm->uLastPosUpdate = 0; 
+	
+	*(BITS *) &pAsm->fp_stOutFmt0 = 0;
+
+	pAsm->uIIns = 0;
+	pAsm->uOIns = 0;
+	pAsm->number_used_registers = 0;
+	pAsm->uUsedConsts = 256; 
+
+
+	// Fragment programs
+	pAsm->uBoolConsts = 0;
+	pAsm->uIntConsts = 0;
+	pAsm->uInsts = 0;
+	pAsm->uConsts = 0;
+
+	pAsm->FCSP = 0;
+	pAsm->fc_stack[0].type = FC_NONE;
+
+	pAsm->branch_depth     = 0;
+	pAsm->max_branch_depth = 0;
+
+	pAsm->aArgSubst[0] =
+	pAsm->aArgSubst[1] =
+	pAsm->aArgSubst[2] =
+	pAsm->aArgSubst[3] = (-1);
+
+	pAsm->uOutputs = 0;
+
+	for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) 
+	{
+		pAsm->color_export_register_number[i] = (-1);
+	}
+
+
+	pAsm->depth_export_register_number = (-1);
+	pAsm->stencil_export_register_number = (-1);
+	pAsm->coverage_to_mask_export_register_number = (-1);
+	pAsm->mask_export_register_number = (-1);
+
+	pAsm->starting_export_register_number = 0;
+	pAsm->starting_vfetch_register_number = 0;
+	pAsm->starting_temp_register_number   = 0;
+	pAsm->uFirstHelpReg = 0;
+
+
+	pAsm->input_position_is_used = GL_FALSE;
+	pAsm->input_normal_is_used   = GL_FALSE;
+
+
+	for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) 
+	{
+		pAsm->input_color_is_used[ i ] = GL_FALSE;
+	}
+
+	for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) 
+	{
+		pAsm->input_texture_unit_is_used[ i ] = GL_FALSE;
+	}
+
+	for (i=0; i<VERT_ATTRIB_MAX; i++) 
+	{
+		pAsm->vfetch_instruction_ptr_array[ i ] = NULL;
+	}
+
+	pAsm->number_of_inputs = 0;
+
+	return 0;
+}
+
+GLboolean IsTex(gl_inst_opcode Opcode)
+{
+    if( (OPCODE_TEX==Opcode) || (OPCODE_TXP==Opcode) || (OPCODE_TXB==Opcode) )
+    {
+        return GL_TRUE;
+    }
+    return GL_FALSE;
+}
+
+GLboolean IsAlu(gl_inst_opcode Opcode)
+{
+    //TODO : more for fc and ex for higher spec.
+    if( IsTex(Opcode) )
+    {
+        return GL_FALSE;
+    }
+    return GL_TRUE;
+}
+
+int check_current_clause(r700_AssemblerBase* pAsm,
+					     CF_CLAUSE_TYPE      new_clause_type)
+{
+	if (pAsm->cf_current_clause_type != new_clause_type) 
+	{	//Close last open clause
+		switch (pAsm->cf_current_clause_type) 
+		{
+		case CF_ALU_CLAUSE:
+			if ( pAsm->cf_current_alu_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_alu_clause_ptr = NULL;
+            }
+			break;
+		case CF_VTX_CLAUSE:
+			if ( pAsm->cf_current_vtx_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_vtx_clause_ptr = NULL;
+            }
+			break;
+		case CF_TEX_CLAUSE:
+			if ( pAsm->cf_current_tex_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_tex_clause_ptr = NULL;
+            }
+			break;
+		case CF_EXPORT_CLAUSE:
+			if ( pAsm->cf_current_export_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_export_clause_ptr = NULL;
+            }
+			break;
+		case CF_OTHER_CLAUSE:
+			if ( pAsm->cf_current_cf_clause_ptr != NULL) 
+            {
+                pAsm->cf_current_cf_clause_ptr = NULL;
+            }
+			break;
+		case CF_EMPTY_CLAUSE:
+			break;
+		default:
+            r700_error(ERROR_ASM_VTX_CLAUSE,
+                       "Unknown CF_CLAUSE_TYPE (%d) in check_current_clause. \n", (int) new_clause_type);
+			return GL_FALSE;
+		}
+
+        pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE;
+
+		// Create new clause
+        switch (new_clause_type) 
+	    {
+        case CF_ALU_CLAUSE:
+            pAsm->cf_current_clause_type = CF_ALU_CLAUSE;
+            break;
+        case CF_VTX_CLAUSE:
+            pAsm->cf_current_clause_type = CF_VTX_CLAUSE;
+            break;
+        case CF_TEX_CLAUSE:        
+            pAsm->cf_current_clause_type = CF_TEX_CLAUSE;
+            break;
+        case CF_EXPORT_CLAUSE:
+            {
+                R700ControlFlowSXClause* pR700ControlFlowSXClause 
+                            = (R700ControlFlowSXClause*) CALLOC_STRUCT(R700ControlFlowSXClause); 
+            
+                // Add new export instruction to control flow program        
+                if (pR700ControlFlowSXClause != 0) 
+                {
+                    pAsm->cf_current_export_clause_ptr = pR700ControlFlowSXClause;
+                    Init_R700ControlFlowSXClause(pR700ControlFlowSXClause);
+                    AddCFInstruction( pAsm->pR700Shader, 
+                                      (R700ControlFlowInstruction *)pR700ControlFlowSXClause );
+                }
+                else 
+                {
+                    r700_error(ERROR_ASM_ALLOCEXPORTCF,
+                               "Error allocating new EXPORT CF instruction in check_current_clause. \n");
+                    return GL_FALSE;
+                }
+                pAsm->cf_current_clause_type = CF_EXPORT_CLAUSE;
+            }
+            break;
+        case CF_EMPTY_CLAUSE:
+            break;
+        case CF_OTHER_CLAUSE:
+            pAsm->cf_current_clause_type = CF_OTHER_CLAUSE;
+            break;
+        default:
+            r700_error(ERROR_ASM_UNKOWNCLAUSE,
+                       "Unknown CF_CLAUSE_TYPE (%d) in check_current_clause. \n", (int) new_clause_type);
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean add_vfetch_instruction(r700_AssemblerBase*     pAsm,
+								 R700VertexInstruction*  vertex_instruction_ptr)
+{
+	if( GL_FALSE == check_current_clause(pAsm,  CF_VTX_CLAUSE) )
+	{
+		return GL_FALSE;
+	}
+
+    if( pAsm->cf_current_vtx_clause_ptr == NULL ||
+        ( (pAsm->cf_current_vtx_clause_ptr != NULL) && 
+         (pAsm->cf_current_vtx_clause_ptr->m_Word1.f.count >= GetCFMaxInstructions(pAsm->cf_current_vtx_clause_ptr->m_ShaderInstType)-1) 
+        ) ) 
+    { 
+		// Create new Vfetch control flow instruction for this new clause
+		pAsm->cf_current_vtx_clause_ptr = (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause);
+
+		if (pAsm->cf_current_vtx_clause_ptr != NULL) 
+		{
+			Init_R700ControlFlowGenericClause(pAsm->cf_current_vtx_clause_ptr);
+			AddCFInstruction( pAsm->pR700Shader, 
+                              (R700ControlFlowInstruction *)pAsm->cf_current_vtx_clause_ptr );
+		}
+		else 
+		{
+            r700_error(ERROR_ASM_ALLOCVTXCF, "Could not allocate a new VFetch CF instruction.");
+			return GL_FALSE;
+		}
+
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.pop_count        = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.cf_const         = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.cond             = SQ_CF_COND_ACTIVE;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.count            = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.end_of_program   = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.cf_inst          = SQ_CF_INST_VTX;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.whole_quad_mode  = 0x0;
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.barrier          = 0x1;
+
+		LinkVertexInstruction(pAsm->cf_current_vtx_clause_ptr, vertex_instruction_ptr );
+	}
+	else
+	{
+		pAsm->cf_current_vtx_clause_ptr->m_Word1.f.count++;
+	}
+
+	AddVTXInstruction(pAsm->pR700Shader, vertex_instruction_ptr);
+
+	return GL_TRUE;
+}
+
+GLboolean add_tex_instruction(r700_AssemblerBase*     pAsm,
+                              R700TextureInstruction* tex_instruction_ptr)
+{ 
+    if ( GL_FALSE == check_current_clause(pAsm, CF_TEX_CLAUSE) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( pAsm->cf_current_tex_clause_ptr == NULL ||
+         ( (pAsm->cf_current_tex_clause_ptr != NULL) && 
+           (pAsm->cf_current_tex_clause_ptr->m_Word1.f.count >= GetCFMaxInstructions(pAsm->cf_current_tex_clause_ptr->m_ShaderInstType)-1) 
+         ) ) 
+    {
+        // new tex cf instruction for this new clause  
+        pAsm->cf_current_tex_clause_ptr = (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause);
+
+		if (pAsm->cf_current_tex_clause_ptr != NULL) 
+		{
+			Init_R700ControlFlowGenericClause(pAsm->cf_current_tex_clause_ptr);
+			AddCFInstruction( pAsm->pR700Shader, 
+                              (R700ControlFlowInstruction *)pAsm->cf_current_tex_clause_ptr );
+		}
+		else 
+		{
+            r700_error(ERROR_ASM_ALLOCTEXCF, "Could not allocate a new TEX CF instruction.");
+			return GL_FALSE;
+		}
+        
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.pop_count        = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.cf_const         = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.cond             = SQ_CF_COND_ACTIVE;
+
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.end_of_program   = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.cf_inst          = SQ_CF_INST_TEX;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.whole_quad_mode  = 0x0;
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.barrier          = 0x0;   //0x1;
+    }
+    else 
+    {        
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.count++;
+    }
+
+    // If this clause constains any TEX instruction that is dependent on a previous instruction, 
+    // set the barrier bit
+    if( pAsm->pInstDeps[pAsm->uiCurInst].nDstDep > (-1) )
+    {
+        pAsm->cf_current_tex_clause_ptr->m_Word1.f.barrier = 0x1;  
+    }
+
+    if(NULL == pAsm->cf_current_tex_clause_ptr->m_pLinkedTEXInstruction)
+    {
+        pAsm->cf_current_tex_clause_ptr->m_pLinkedTEXInstruction = tex_instruction_ptr;
+        tex_instruction_ptr->m_pLinkedGenericClause = pAsm->cf_current_tex_clause_ptr;
+    }
+
+    AddTEXInstruction(pAsm->pR700Shader, tex_instruction_ptr);
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_vfetch_instruction(r700_AssemblerBase* pAsm,
+								GLuint gl_client_id,
+                                GLuint destination_register,
+								GLuint number_of_elements,
+                                GLenum dataElementType,
+								VTX_FETCH_METHOD* pFetchMethod)
+{
+    GLuint client_size_inbyte;
+	GLuint data_format;
+    GLuint mega_fetch_count;
+	GLuint is_mega_fetch_flag;
+
+	R700VertexGenericFetch*   vfetch_instruction_ptr;
+	R700VertexGenericFetch*   assembled_vfetch_instruction_ptr = pAsm->vfetch_instruction_ptr_array[ gl_client_id ];
+
+	if (assembled_vfetch_instruction_ptr == NULL) 
+	{
+		vfetch_instruction_ptr = (R700VertexGenericFetch*) CALLOC_STRUCT(R700VertexGenericFetch);
+		if (vfetch_instruction_ptr == NULL) 
+		{
+			return GL_FALSE;
+		}
+        Init_R700VertexGenericFetch(vfetch_instruction_ptr);
+    }
+	else 
+	{
+		vfetch_instruction_ptr = assembled_vfetch_instruction_ptr;
+	}
+
+	data_format = GetSurfaceFormat(dataElementType, number_of_elements, &client_size_inbyte);
+
+	if(GL_TRUE == pFetchMethod->bEnableMini) //More conditions here
+	{
+		//TODO : mini fetch
+	}
+	else
+	{
+		mega_fetch_count = MEGA_FETCH_BYTES - 1;
+		is_mega_fetch_flag       = 0x1;
+		pFetchMethod->mega_fetch_remainder = MEGA_FETCH_BYTES - client_size_inbyte;
+	}
+
+	vfetch_instruction_ptr->m_Word0.f.vtx_inst         = SQ_VTX_INST_FETCH;
+	vfetch_instruction_ptr->m_Word0.f.fetch_type       = SQ_VTX_FETCH_VERTEX_DATA;
+	vfetch_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0;
+
+	vfetch_instruction_ptr->m_Word0.f.buffer_id        = gl_client_id;
+	vfetch_instruction_ptr->m_Word0.f.src_gpr          = 0x0; 
+	vfetch_instruction_ptr->m_Word0.f.src_rel          = SQ_ABSOLUTE;
+	vfetch_instruction_ptr->m_Word0.f.src_sel_x        = SQ_SEL_X;
+	vfetch_instruction_ptr->m_Word0.f.mega_fetch_count = mega_fetch_count;
+
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_x        = (number_of_elements < 1) ? SQ_SEL_0 : SQ_SEL_X;
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_y        = (number_of_elements < 2) ? SQ_SEL_0 : SQ_SEL_Y;
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_z        = (number_of_elements < 3) ? SQ_SEL_0 : SQ_SEL_Z;
+	vfetch_instruction_ptr->m_Word1.f.dst_sel_w        = (number_of_elements < 4) ? SQ_SEL_1 : SQ_SEL_W;
+
+	vfetch_instruction_ptr->m_Word1.f.use_const_fields = 1;
+
+	// Destination register
+	vfetch_instruction_ptr->m_Word1_GPR.f.dst_gpr = destination_register; 
+	vfetch_instruction_ptr->m_Word1_GPR.f.dst_rel = SQ_ABSOLUTE;
+
+	vfetch_instruction_ptr->m_Word2.f.offset              = 0;
+	vfetch_instruction_ptr->m_Word2.f.const_buf_no_stride = 0x0;
+
+	vfetch_instruction_ptr->m_Word2.f.mega_fetch          = is_mega_fetch_flag;
+
+	if (assembled_vfetch_instruction_ptr == NULL) 
+	{
+		if ( GL_FALSE == add_vfetch_instruction(pAsm, (R700VertexInstruction *)vfetch_instruction_ptr) ) 
+        {   
+			return GL_FALSE;
+		}
+
+		if (pAsm->vfetch_instruction_ptr_array[ gl_client_id ] != NULL) 
+		{
+			return GL_FALSE;
+		}
+		else 
+		{
+			pAsm->vfetch_instruction_ptr_array[ gl_client_id ] = vfetch_instruction_ptr;
+		}
+	}
+
+	return GL_TRUE;
+}
+
+GLuint gethelpr(r700_AssemblerBase* pAsm) 
+{
+    GLuint r = pAsm->uHelpReg;
+    pAsm->uHelpReg++;
+    if (pAsm->uHelpReg > pAsm->number_used_registers)
+    {
+        pAsm->number_used_registers = pAsm->uHelpReg;
+	}
+    return r;
+}
+void resethelpr(r700_AssemblerBase* pAsm) 
+{
+    pAsm->uHelpReg = pAsm->uFirstHelpReg;
+}
+
+void checkop_init(r700_AssemblerBase* pAsm)
+{
+    resethelpr(pAsm);
+    pAsm->aArgSubst[0] =
+    pAsm->aArgSubst[1] =
+    pAsm->aArgSubst[2] =
+    pAsm->aArgSubst[3] = -1;
+}
+
+GLboolean mov_temp(r700_AssemblerBase* pAsm, int src)
+{
+    GLuint tmp = gethelpr(pAsm);
+
+    //mov src to temp helper gpr.
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+  
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    if( GL_FALSE == assemble_src(pAsm, src, 0) )
+    {
+        return GL_FALSE;
+    }
+
+    noswizzle_PVSSRC(&(pAsm->S[0].src));
+    noneg_PVSSRC(&(pAsm->S[0].src));
+   
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->aArgSubst[1 + src] = tmp;
+
+    return GL_TRUE;
+}
+
+GLboolean checkop1(r700_AssemblerBase* pAsm)
+{
+    checkop_init(pAsm);
+    return GL_TRUE;
+}
+
+GLboolean checkop2(r700_AssemblerBase* pAsm)
+{
+    GLboolean bSrcConst[2];
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    checkop_init(pAsm);
+
+    if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[0] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[0] = GL_FALSE;
+    }
+    if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[1] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[1] = GL_FALSE;
+    }
+
+    if( (bSrcConst[0] == GL_TRUE) && (bSrcConst[1] == GL_TRUE) )
+    {
+        if(pILInst->SrcReg[0].Index != pILInst->SrcReg[1].Index)
+        {
+            if( GL_FALSE == mov_temp(pAsm, 1) )
+            {
+                return GL_FALSE;
+            }
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean checkop3(r700_AssemblerBase* pAsm)
+{
+    GLboolean bSrcConst[3];
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    checkop_init(pAsm);
+
+    if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[0] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[0] = GL_FALSE;
+    }
+    if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[1] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[1] = GL_FALSE;
+    }
+    if( (pILInst->SrcReg[2].File == PROGRAM_CONSTANT)    ||
+        (pILInst->SrcReg[2].File == PROGRAM_LOCAL_PARAM) ||
+        (pILInst->SrcReg[2].File == PROGRAM_ENV_PARAM)   ||
+        (pILInst->SrcReg[2].File == PROGRAM_STATE_VAR) )
+    {
+        bSrcConst[2] = GL_TRUE;
+    }
+    else
+    {
+        bSrcConst[2] = GL_FALSE;
+    }
+
+    if( (GL_TRUE == bSrcConst[0]) && 
+        (GL_TRUE == bSrcConst[1]) && 
+        (GL_TRUE == bSrcConst[2]) ) 
+    {
+        if( GL_FALSE == mov_temp(pAsm, 1) )
+        {
+            return GL_FALSE;
+        }
+        if( GL_FALSE == mov_temp(pAsm, 2) )
+        {
+            return GL_FALSE;
+        }
+
+        return GL_TRUE;
+    }
+    else if( (GL_TRUE == bSrcConst[0]) && 
+             (GL_TRUE == bSrcConst[1]) ) 
+    {
+        if(pILInst->SrcReg[0].Index != pILInst->SrcReg[1].Index)    
+	    {
+            if( GL_FALSE == mov_temp(pAsm, 1) )
+            {
+                return 1;
+            }
+        }
+
+        return GL_TRUE;
+    }
+    else if ( (GL_TRUE == bSrcConst[0]) && 
+              (GL_TRUE == bSrcConst[2]) )  
+    {
+        if(pILInst->SrcReg[0].Index != pILInst->SrcReg[2].Index)     
+	    {
+            if( GL_FALSE == mov_temp(pAsm, 2) )
+            {
+                return GL_FALSE;
+            }
+        }
+
+        return GL_TRUE;
+    }
+    else if( (GL_TRUE == bSrcConst[1]) && 
+             (GL_TRUE == bSrcConst[2]) ) 
+    {
+        if(pILInst->SrcReg[1].Index != pILInst->SrcReg[2].Index)
+	    {
+            if( GL_FALSE == mov_temp(pAsm, 2) )
+            {
+                return GL_FALSE;
+            }
+        }
+
+        return GL_TRUE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_src(r700_AssemblerBase *pAsm,
+                       int src, 
+                       int fld)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    if (fld == -1)
+    {
+        fld = src;
+    }
+
+    if(pAsm->aArgSubst[1+src] >= 0) 
+    {
+        setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);
+        pAsm->S[fld].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[fld].src.reg   = pAsm->aArgSubst[1+src];
+    }
+    else 
+    {
+        switch (pILInst->SrcReg[src].File)
+        {
+        case PROGRAM_TEMPORARY:
+            setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);
+            pAsm->S[fld].src.rtype = SRC_REG_TEMPORARY;
+            pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index + pAsm->starting_temp_register_number;
+            break;
+        case PROGRAM_CONSTANT:
+        case PROGRAM_LOCAL_PARAM:
+        case PROGRAM_ENV_PARAM:
+        case PROGRAM_STATE_VAR:
+            if (1 == pILInst->SrcReg[src].RelAddr)
+            {
+                setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_RELATIVE_A0);
+            }
+            else
+            {
+                setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);              
+            }
+
+            pAsm->S[fld].src.rtype = SRC_REG_CONSTANT;
+            pAsm->S[fld].src.reg   = pILInst->SrcReg[src].Index;
+            break;      
+        case PROGRAM_INPUT:
+            setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE);
+            pAsm->S[fld].src.rtype = SRC_REG_INPUT;
+            switch (pAsm->currentShaderType)
+            {
+            case SPT_FP:
+                pAsm->S[fld].src.reg = pAsm->uiFP_AttributeMap[pILInst->SrcReg[src].Index];
+                break;
+            case SPT_VP:
+                pAsm->S[fld].src.reg = pAsm->ucVP_AttributeMap[pILInst->SrcReg[src].Index];
+                break;
+            }
+            break;      
+        default:
+            r700_error(ERROR_ASM_SRCARGUMENT, "Invalid source argument type");          
+            return GL_FALSE;
+        }
+    } 
+
+    pAsm->S[fld].src.swizzlex = pILInst->SrcReg[src].Swizzle & 0x7;
+    pAsm->S[fld].src.swizzley = (pILInst->SrcReg[src].Swizzle >> 3) & 0x7;
+    pAsm->S[fld].src.swizzlez = (pILInst->SrcReg[src].Swizzle >> 6) & 0x7;
+    pAsm->S[fld].src.swizzlew = (pILInst->SrcReg[src].Swizzle >> 9) & 0x7;
+
+    pAsm->S[fld].src.negx = pILInst->SrcReg[src].Negate & 0x1;
+    pAsm->S[fld].src.negy = (pILInst->SrcReg[src].Negate >> 1) & 0x1;
+    pAsm->S[fld].src.negz = (pILInst->SrcReg[src].Negate >> 2) & 0x1;
+    pAsm->S[fld].src.negw = (pILInst->SrcReg[src].Negate >> 3) & 0x1;
+     
+    return GL_TRUE;
+}
+
+GLboolean assemble_dst(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+    switch (pILInst->DstReg.File) 
+    {
+    case PROGRAM_TEMPORARY:
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg = pILInst->DstReg.Index + pAsm->starting_temp_register_number;
+        break;
+    case PROGRAM_ADDRESS:
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_A0;
+        pAsm->D.dst.reg = 0;
+        break;
+    case PROGRAM_OUTPUT:
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_OUT;
+        switch (pAsm->currentShaderType)
+        {
+        case SPT_FP:
+            pAsm->D.dst.reg = pAsm->uiFP_OutputMap[pILInst->DstReg.Index];
+            break;
+        case SPT_VP:
+            pAsm->D.dst.reg = pAsm->ucVP_OutputMap[pILInst->DstReg.Index];
+            break;
+        }
+        break;   
+    default:
+        r700_error(ERROR_ASM_DSTARGUMENT, "Invalid destination output argument type");
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.writex = pILInst->DstReg.WriteMask & 0x1;
+    pAsm->D.dst.writey = (pILInst->DstReg.WriteMask >> 1) & 0x1;
+    pAsm->D.dst.writez = (pILInst->DstReg.WriteMask >> 2) & 0x1;
+    pAsm->D.dst.writew = (pILInst->DstReg.WriteMask >> 3) & 0x1;
+  
+    return GL_TRUE;
+}
+
+GLboolean tex_dst(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    if(PROGRAM_TEMPORARY == pILInst->DstReg.File)
+    {
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = pAsm->pILInst[pAsm->uiCurInst].DstReg.Index + pAsm->starting_temp_register_number;
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    }
+    else if(PROGRAM_OUTPUT == pILInst->DstReg.File)
+    {
+        pAsm->D.dst.rtype = DST_REG_OUT;
+        switch (pAsm->currentShaderType)
+        {
+        case SPT_FP:
+            pAsm->D.dst.reg = pAsm->uiFP_OutputMap[pILInst->DstReg.Index];
+            break;
+        case SPT_VP:
+            pAsm->D.dst.reg = pAsm->ucVP_OutputMap[pILInst->DstReg.Index];
+            break;
+        }
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    }
+    else 
+    {
+        r700_error(ERROR_ASM_DSTARGUMENT, "Invalid destination output argument type");
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.writex = pILInst->DstReg.WriteMask & 0x1;
+    pAsm->D.dst.writey = (pILInst->DstReg.WriteMask >> 1) & 0x1;
+    pAsm->D.dst.writez = (pILInst->DstReg.WriteMask >> 2) & 0x1;
+    pAsm->D.dst.writew = (pILInst->DstReg.WriteMask >> 3) & 0x1;
+  
+    return GL_TRUE;
+}
+
+GLboolean tex_src(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+ 
+    GLboolean bValidTexCoord = GL_FALSE;
+
+    switch (pILInst->SrcReg[0].File)
+    {
+    case PROGRAM_TEMPORARY:
+        bValidTexCoord = GL_TRUE;
+
+        pAsm->S[0].src.reg   = pILInst->SrcReg[0].Index + pAsm->starting_temp_register_number;
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+
+        break;
+    case PROGRAM_INPUT:
+        switch (pILInst->SrcReg[0].Index)
+        {
+        case FRAG_ATTRIB_COL0:
+        case FRAG_ATTRIB_COL1:
+        case FRAG_ATTRIB_TEX0:
+        case FRAG_ATTRIB_TEX1:
+        case FRAG_ATTRIB_TEX2:
+        case FRAG_ATTRIB_TEX3:
+        case FRAG_ATTRIB_TEX4:
+        case FRAG_ATTRIB_TEX5:
+        case FRAG_ATTRIB_TEX6:
+        case FRAG_ATTRIB_TEX7:
+            bValidTexCoord = GL_TRUE;
+
+            pAsm->S[0].src.reg   = pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index];
+            pAsm->S[0].src.rtype = SRC_REG_INPUT;
+        }
+        break;
+    }
+
+    if(GL_TRUE == bValidTexCoord)
+    { 
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    }
+    else
+    {
+        r700_error(ERROR_ASM_BADTEXSRC, "Invalid source texcoord for TEX instruction");
+        return GL_FALSE;
+    }
+
+    pAsm->S[0].src.swizzlex = pILInst->SrcReg[0].Swizzle & 0x7;
+    pAsm->S[0].src.swizzley = (pILInst->SrcReg[0].Swizzle >> 3) & 0x7;
+    pAsm->S[0].src.swizzlez = (pILInst->SrcReg[0].Swizzle >> 6) & 0x7;
+    pAsm->S[0].src.swizzlew = (pILInst->SrcReg[0].Swizzle >> 9) & 0x7;
+
+    pAsm->S[0].src.negx = pILInst->SrcReg[0].Negate & 0x1;
+    pAsm->S[0].src.negy = (pILInst->SrcReg[0].Negate >> 1) & 0x1;
+    pAsm->S[0].src.negz = (pILInst->SrcReg[0].Negate >> 2) & 0x1;
+    pAsm->S[0].src.negw = (pILInst->SrcReg[0].Negate >> 3) & 0x1;
+     
+    return GL_TRUE;
+}
+
+GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm)
+{
+    PVSSRC *   texture_coordinate_source;
+    PVSSRC *   texture_unit_source;
+    
+    R700TextureInstruction* tex_instruction_ptr = (R700TextureInstruction*) CALLOC_STRUCT(R700TextureInstruction);
+	if (tex_instruction_ptr == NULL) 
+	{
+		return GL_FALSE;
+	}
+    Init_R700TextureInstruction(tex_instruction_ptr);
+
+    texture_coordinate_source = &(pAsm->S[0].src);
+    texture_unit_source       = &(pAsm->S[1].src);
+
+    tex_instruction_ptr->m_Word0.f.tex_inst         = pAsm->D.dst.opcode;
+    tex_instruction_ptr->m_Word0.f.bc_frac_mode     = 0x0;
+    tex_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0;
+
+    tex_instruction_ptr->m_Word0.f.resource_id      = texture_unit_source->reg;
+
+    tex_instruction_ptr->m_Word1.f.lod_bias     = 0x0;
+    tex_instruction_ptr->m_Word1.f.coord_type_x = SQ_TEX_NORMALIZED;
+    tex_instruction_ptr->m_Word1.f.coord_type_y = SQ_TEX_NORMALIZED;
+    tex_instruction_ptr->m_Word1.f.coord_type_z = SQ_TEX_NORMALIZED;
+    tex_instruction_ptr->m_Word1.f.coord_type_w = SQ_TEX_NORMALIZED;
+
+    tex_instruction_ptr->m_Word2.f.offset_x   = 0x0;
+    tex_instruction_ptr->m_Word2.f.offset_y   = 0x0;
+    tex_instruction_ptr->m_Word2.f.offset_z   = 0x0;
+
+    tex_instruction_ptr->m_Word2.f.sampler_id = texture_unit_source->reg;
+
+    // dst
+    if ( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || 
+         (pAsm->D.dst.rtype == DST_REG_OUT) ) 
+    {
+        tex_instruction_ptr->m_Word0.f.src_gpr    = texture_coordinate_source->reg;
+        tex_instruction_ptr->m_Word0.f.src_rel    = SQ_ABSOLUTE;
+
+        tex_instruction_ptr->m_Word1.f.dst_gpr    = pAsm->D.dst.reg;
+        tex_instruction_ptr->m_Word1.f.dst_rel    = SQ_ABSOLUTE;
+
+        tex_instruction_ptr->m_Word1.f.dst_sel_x  = (pAsm->D.dst.writex ? texture_unit_source->swizzlex : SQ_SEL_MASK);
+        tex_instruction_ptr->m_Word1.f.dst_sel_y  = (pAsm->D.dst.writey ? texture_unit_source->swizzley : SQ_SEL_MASK);
+        tex_instruction_ptr->m_Word1.f.dst_sel_z  = (pAsm->D.dst.writez ? texture_unit_source->swizzlez : SQ_SEL_MASK);
+        tex_instruction_ptr->m_Word1.f.dst_sel_w  = (pAsm->D.dst.writew ? texture_unit_source->swizzlew : SQ_SEL_MASK);
+
+
+        tex_instruction_ptr->m_Word2.f.src_sel_x  = texture_coordinate_source->swizzlex;
+        tex_instruction_ptr->m_Word2.f.src_sel_y  = texture_coordinate_source->swizzley;
+        tex_instruction_ptr->m_Word2.f.src_sel_z  = texture_coordinate_source->swizzlez;
+        tex_instruction_ptr->m_Word2.f.src_sel_w  = texture_coordinate_source->swizzlew;
+    }
+    else 
+    {
+        r700_error(ERROR_ASM_TEXDSTBADTYPE, "Only temp destination registers supported for TEX dest regs.");
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == add_tex_instruction(pAsm, tex_instruction_ptr) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+void initialize(r700_AssemblerBase *pAsm)
+{
+    GLuint cycle, component;
+
+    for (cycle=0; cycle<NUMBER_OF_CYCLES; cycle++) 
+    {
+        for (component=0; component<NUMBER_OF_COMPONENTS; component++) 
+        {
+            pAsm->hw_gpr[cycle][component] = (-1);
+        }
+    }
+    for (component=0; component<NUMBER_OF_COMPONENTS; component++) 
+    {
+        pAsm->hw_cfile_addr[component] = (-1);
+        pAsm->hw_cfile_chan[component] = (-1);
+    }
+}
+
+GLboolean assemble_alu_src(R700ALUInstruction*  alu_instruction_ptr,
+                           int                  source_index,
+                           PVSSRC*              pSource,
+                           BITS                 scalar_channel_index)
+{
+    BITS src_sel;
+    BITS src_rel;
+    BITS src_chan;
+    BITS src_neg;
+
+    //--------------------------------------------------------------------------
+    // Source for operands src0, src1. 
+    // Values [0,127] correspond to GPR[0..127]. 
+    // Values [256,511] correspond to cfile constants c[0..255]. 
+
+    //--------------------------------------------------------------------------
+    // Other special values are shown in the list below.
+
+    // 248	SQ_ALU_SRC_0: special constant 0.0.
+    // 249	SQ_ALU_SRC_1: special constant 1.0 float.
+
+    // 250	SQ_ALU_SRC_1_INT: special constant 1 integer.
+    // 251	SQ_ALU_SRC_M_1_INT: special constant -1 integer.
+
+    // 252	SQ_ALU_SRC_0_5: special constant 0.5 float.
+    // 253	SQ_ALU_SRC_LITERAL: literal constant.
+
+    // 254	SQ_ALU_SRC_PV: previous vector result.
+    // 255	SQ_ALU_SRC_PS: previous scalar result.
+    //--------------------------------------------------------------------------
+
+    BITS channel_swizzle;
+    switch (scalar_channel_index) 
+    {
+        case 0: channel_swizzle = pSource->swizzlex; break;
+        case 1: channel_swizzle = pSource->swizzley; break;
+        case 2: channel_swizzle = pSource->swizzlez; break;
+        case 3: channel_swizzle = pSource->swizzlew; break;
+        default: channel_swizzle = SQ_SEL_MASK; break;
+    }
+
+    if(channel_swizzle == SQ_SEL_0) 
+    {
+        src_sel = SQ_ALU_SRC_0; 
+    }
+    else if (channel_swizzle == SQ_SEL_1) 
+    {
+        src_sel = SQ_ALU_SRC_1; 
+    }
+    else 
+    {
+        if ( (pSource->rtype == SRC_REG_TEMPORARY) || 
+             (pSource->rtype == SRC_REG_INPUT)
+        ) 
+        {
+            src_sel = pSource->reg;
+        }
+        else if (pSource->rtype == SRC_REG_CONSTANT)
+        {
+            src_sel = pSource->reg + CFILE_REGISTER_OFFSET;            
+        }
+        else
+        {
+            r700_error(ERROR_ASM_ALUSRCBADTYPE, "Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.", 
+                     source_index, pSource->rtype);
+            return GL_FALSE;
+        }
+    }
+
+    if( ADDR_ABSOLUTE == addrmode_PVSSRC(pSource) ) 
+    {
+        src_rel = SQ_ABSOLUTE;
+    }
+    else 
+    {
+        src_rel = SQ_RELATIVE;
+    }
+
+    switch (channel_swizzle) 
+    {
+        case SQ_SEL_X: 
+            src_chan = SQ_CHAN_X; 
+            break;
+        case SQ_SEL_Y: 
+            src_chan = SQ_CHAN_Y; 
+            break;
+        case SQ_SEL_Z: 
+            src_chan = SQ_CHAN_Z; 
+            break;
+        case SQ_SEL_W: 
+            src_chan = SQ_CHAN_W; 
+            break;
+        case SQ_SEL_0:
+        case SQ_SEL_1:
+            // Does not matter since src_sel controls
+            src_chan = SQ_CHAN_X; 
+            break;
+        default:
+            r700_error(ERROR_ASM_ALUSRCSELECT, "Unknown source select value (%d) in assemble_alu_src().");
+            return GL_FALSE;
+            break;
+    }
+
+    switch (scalar_channel_index) 
+    {
+        case 0: src_neg = pSource->negx; break;
+        case 1: src_neg = pSource->negy; break;
+        case 2: src_neg = pSource->negz; break;
+        case 3: src_neg = pSource->negw; break;
+        default: src_neg = 0; break;
+    }
+
+    switch (source_index) 
+    {
+        case 0:
+            alu_instruction_ptr->m_Word0.f.src0_sel  = src_sel;
+            alu_instruction_ptr->m_Word0.f.src0_rel  = src_rel;
+            alu_instruction_ptr->m_Word0.f.src0_chan = src_chan;
+            alu_instruction_ptr->m_Word0.f.src0_neg  = src_neg;
+            break;
+        case 1:
+            alu_instruction_ptr->m_Word0.f.src1_sel  = src_sel;
+            alu_instruction_ptr->m_Word0.f.src1_rel  = src_rel;
+            alu_instruction_ptr->m_Word0.f.src1_chan = src_chan;
+            alu_instruction_ptr->m_Word0.f.src1_neg  = src_neg;
+            break;
+        case 2:
+            alu_instruction_ptr->m_Word1_OP3.f.src2_sel  = src_sel;
+            alu_instruction_ptr->m_Word1_OP3.f.src2_rel  = src_rel;
+            alu_instruction_ptr->m_Word1_OP3.f.src2_chan = src_chan;
+            alu_instruction_ptr->m_Word1_OP3.f.src2_neg  = src_neg;
+            break;
+        default:
+            r700_error(ERROR_ASM_ALUSRCNUMBER, "Only three sources allowed in ALU opcodes.");
+          return GL_FALSE;
+          break;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean add_alu_instruction(r700_AssemblerBase* pAsm,
+                              R700ALUInstruction* alu_instruction_ptr,
+                              GLuint              contiguous_slots_needed)
+{
+    if( GL_FALSE == check_current_clause(pAsm, CF_ALU_CLAUSE) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( pAsm->cf_current_alu_clause_ptr == NULL ||
+         ( (pAsm->cf_current_alu_clause_ptr != NULL) && 
+           (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-contiguous_slots_needed-1) )
+         ) ) 
+    {
+
+        //new cf inst for this clause
+        pAsm->cf_current_alu_clause_ptr = (R700ControlFlowALUClause*) CALLOC_STRUCT(R700ControlFlowALUClause);
+            
+        // link the new cf to cf segment    
+        if(NULL != pAsm->cf_current_alu_clause_ptr) 
+        {
+            Init_R700ControlFlowALUClause(pAsm->cf_current_alu_clause_ptr);
+			AddCFInstruction( pAsm->pR700Shader, 
+                              (R700ControlFlowInstruction *)pAsm->cf_current_alu_clause_ptr );            
+        }
+        else 
+        {
+            r700_error(ERROR_ASM_ALLOCALUCF, "Could not allocate a new ALU CF instruction.");
+            return GL_FALSE;
+        }
+
+        pAsm->cf_current_alu_clause_ptr->m_Word0.f.kcache_bank0 = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word0.f.kcache_bank1 = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word0.f.kcache_mode0 = SQ_CF_KCACHE_NOP;
+
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_mode1 = SQ_CF_KCACHE_NOP;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr0 = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr1 = 0x0;
+
+        //cf_current_alu_clause_ptr->m_Word1.f.count           = number_of_scalar_operations - 1;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.count           = 0x0;
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst         = SQ_CF_INST_ALU;
+
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x0;
+
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.barrier         = 0x1;
+    }
+    else 
+    {
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.count++;
+    }
+
+    // If this clause constains any instruction that is forward dependent on a TEX instruction, 
+    // set the whole_quad_mode for this clause
+    if ( pAsm->pInstDeps[pAsm->uiCurInst].nDstDep > (-1) ) 
+    {
+        pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x1;   
+    }
+
+    if (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-1) ) 
+    {
+        alu_instruction_ptr->m_Word0.f.last = 1;
+    }
+
+    if(NULL == pAsm->cf_current_alu_clause_ptr->m_pLinkedALUInstruction)
+    {
+        pAsm->cf_current_alu_clause_ptr->m_pLinkedALUInstruction = alu_instruction_ptr;
+        alu_instruction_ptr->m_pLinkedALUClause = pAsm->cf_current_alu_clause_ptr;
+    }
+    
+    AddALUInstruction(pAsm->pR700Shader, alu_instruction_ptr);
+
+    return GL_TRUE;
+}
+
+void get_src_properties(R700ALUInstruction*  alu_instruction_ptr,
+                        int                  source_index,
+                        BITS*                psrc_sel,
+                        BITS*                psrc_rel,
+                        BITS*                psrc_chan,
+                        BITS*                psrc_neg)
+{
+    switch (source_index) 
+    {
+        case 0:
+            *psrc_sel  = alu_instruction_ptr->m_Word0.f.src0_sel ;
+            *psrc_rel  = alu_instruction_ptr->m_Word0.f.src0_rel ;
+            *psrc_chan = alu_instruction_ptr->m_Word0.f.src0_chan;
+            *psrc_neg  = alu_instruction_ptr->m_Word0.f.src0_neg ;
+            break;
+
+        case 1:
+            *psrc_sel  = alu_instruction_ptr->m_Word0.f.src1_sel ;
+            *psrc_rel  = alu_instruction_ptr->m_Word0.f.src1_rel ;
+            *psrc_chan = alu_instruction_ptr->m_Word0.f.src1_chan;
+            *psrc_neg  = alu_instruction_ptr->m_Word0.f.src1_neg ;
+            break;
+
+        case 2:
+            *psrc_sel  = alu_instruction_ptr->m_Word1_OP3.f.src2_sel;
+            *psrc_rel  = alu_instruction_ptr->m_Word1_OP3.f.src2_rel;
+            *psrc_chan = alu_instruction_ptr->m_Word1_OP3.f.src2_chan;
+            *psrc_neg  = alu_instruction_ptr->m_Word1_OP3.f.src2_neg;
+            break;
+    }
+}
+
+int is_cfile(BITS sel) 
+{
+    if (sel > 255 && sel < 512) 
+    {
+        return 1;
+    }
+    return 0;
+}
+
+int is_const(BITS sel) 
+{
+    if (is_cfile(sel)) 
+    {
+        return 1;
+    }
+    else if(sel >= SQ_ALU_SRC_0 && sel <= SQ_ALU_SRC_LITERAL) 
+    {
+        return 1;
+    }
+    return 0;
+}
+
+int is_gpr(BITS sel) 
+{
+    if (sel >= 0 && sel < 128) 
+    {
+        return 1;
+    }
+    return 0;
+}
+
+const GLuint BANK_SWIZZLE_VEC[8] = {SQ_ALU_VEC_210,  //000
+                                    SQ_ALU_VEC_120,  //001
+                                    SQ_ALU_VEC_102,  //010
+
+                                    SQ_ALU_VEC_201,  //011
+                                    SQ_ALU_VEC_012,  //100
+                                    SQ_ALU_VEC_021,  //101
+
+                                    SQ_ALU_VEC_012,  //110
+                                    SQ_ALU_VEC_012}; //111
+
+const GLuint BANK_SWIZZLE_SCL[8] = {SQ_ALU_SCL_210,  //000
+                                    SQ_ALU_SCL_122,  //001 
+                                    SQ_ALU_SCL_122,  //010
+
+                                    SQ_ALU_SCL_221,  //011
+                                    SQ_ALU_SCL_212,  //100
+                                    SQ_ALU_SCL_122,  //101
+
+                                    SQ_ALU_SCL_122,  //110
+                                    SQ_ALU_SCL_122}; //111
+
+GLboolean reserve_cfile(r700_AssemblerBase* pAsm, 
+                        GLuint sel, 
+                        GLuint chan)
+{
+    int res_match = (-1);
+    int res_empty = (-1);
+
+    GLint res;
+
+    for (res=3; res>=0; res--) 
+    {
+        if(pAsm->hw_cfile_addr[ res] < 0)  
+        {
+            res_empty = res;
+        }
+        else if( (pAsm->hw_cfile_addr[res] == (int)sel)
+                 &&
+                 (pAsm->hw_cfile_chan[ res ] == (int) chan) ) 
+        {
+            res_match = res;
+        }
+    }
+
+    if(res_match >= 0) 
+    {
+        // Read for this scalar component already reserved, nothing to do here.
+        ;
+    }
+    else if(res_empty >= 0) 
+    {
+        pAsm->hw_cfile_addr[ res_empty ] = sel;
+        pAsm->hw_cfile_chan[ res_empty ] = chan;
+    }
+    else 
+    {
+        r700_error(ERROR_ASM_CONSTCHANNEL, "All cfile read ports are used, cannot reference C$sel, channel $chan.");
+        return GL_FALSE;
+    }
+    return GL_TRUE;
+}
+
+GLboolean reserve_gpr(r700_AssemblerBase* pAsm, GLuint sel, GLuint chan, GLuint cycle)
+{
+    if(pAsm->hw_gpr[cycle][chan] < 0) 
+    {
+        pAsm->hw_gpr[cycle][chan] = sel;
+    }
+    else if(pAsm->hw_gpr[cycle][chan] != (int)sel) 
+    {
+        r700_error(ERROR_ASM_BADGPRRESERVE, "Another scalar operation has already used GPR read port for given channel");
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean cycle_for_scalar_bank_swizzle(const int swiz, const int sel, GLuint* pCycle)
+{
+    switch (swiz) 
+    {
+        case SQ_ALU_SCL_210:
+            {
+                int table[3] = {2,	1,	0};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        case SQ_ALU_SCL_122:
+            {
+                int table[3] = {1,	2,	2};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        case SQ_ALU_SCL_212:
+            {	
+                int table[3] = {2,	1,	2};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        case SQ_ALU_SCL_221:
+            {
+                int table[3] = {2, 2, 1};
+                *pCycle = table[sel];
+                return GL_TRUE;
+            }
+            break;
+        default:
+            r700_error(ERROR_ASM_BADSCALARBZ, "Bad Scalar bank swizzle value");
+            break;
+    }
+
+    return GL_FALSE;
+}
+
+GLboolean cycle_for_vector_bank_swizzle(const int swiz, const int sel, GLuint* pCycle)
+{
+    switch (swiz) 
+    {
+        case SQ_ALU_VEC_012:
+            {
+                int table[3] = {0, 1, 2};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_021:
+            {
+                int table[3] = {0, 2,	1};
+                *pCycle = table[sel];
+            }
+            break;        
+        case SQ_ALU_VEC_120:
+            {
+                int table[3] = {1, 2,	0};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_102:
+            {
+                int table[3] = {1, 0,	2};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_201:
+            {
+                int table[3] = {2, 0,	1};
+                *pCycle = table[sel];
+            }
+            break;
+        case SQ_ALU_VEC_210:
+            {
+                int table[3] = {2, 1,	0};
+                *pCycle = table[sel];
+            }
+            break;
+        default:
+            r700_error(ERROR_ASM_BADVECTORBZ, "Bad Vec bank swizzle value");
+            return GL_FALSE;
+            break;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean check_scalar(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr)
+{
+    GLuint cycle;
+    GLuint bank_swizzle;
+    GLuint const_count = 0;
+
+    BITS sel;
+    BITS chan;
+    BITS rel;
+    BITS neg;
+
+    GLuint src;
+
+    BITS src_sel [3] = {0,0,0};
+    BITS src_chan[3] = {0,0,0};
+    BITS src_rel [3] = {0,0,0};
+    BITS src_neg [3] = {0,0,0};
+
+    GLuint swizzle_key;
+
+    GLuint number_of_operands = r700GetNumOperands(pAsm);
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        get_src_properties(alu_instruction_ptr,
+                           src,
+                           &(src_sel[src]), 
+                           &(src_rel[src]), 
+                           &(src_chan[src]), 
+                           &(src_neg[src]) );
+    }
+
+
+    swizzle_key = ( (is_const( src_sel[0] ) ? 4 : 0) + 
+                    (is_const( src_sel[1] ) ? 2 : 0) + 
+                    (is_const( src_sel[2] ) ? 1 : 0) );
+  
+    alu_instruction_ptr->m_Word1.f.bank_swizzle = BANK_SWIZZLE_SCL[ swizzle_key ];
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        sel  = src_sel [src];
+        chan = src_chan[src];
+        rel  = src_rel [src];
+        neg  = src_neg [src];
+
+        if (is_const( sel )) 
+        {
+            // Any constant, including literal and inline constants
+            const_count++;
+
+            if (is_cfile( sel )) 
+            {
+                reserve_cfile(pAsm, sel, chan);
+            }
+
+        }
+    }
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        sel  = src_sel [src];
+        chan = src_chan[src];
+        rel  = src_rel [src];
+        neg  = src_neg [src];
+
+        if( is_gpr(sel) ) 
+        {
+            bank_swizzle = alu_instruction_ptr->m_Word1.f.bank_swizzle;
+
+            if( GL_FALSE == cycle_for_scalar_bank_swizzle(bank_swizzle, src, &cycle) )
+            {
+                return GL_FALSE;
+            }
+
+            if(cycle < const_count) 
+            {
+                if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) )
+                {
+                    return GL_FALSE;
+                }
+            }
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean check_vector(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr)
+{
+    GLuint cycle;
+    GLuint bank_swizzle;
+    GLuint const_count = 0;
+
+    GLuint src;
+
+    BITS sel;
+    BITS chan;
+    BITS rel;
+    BITS neg;
+
+    BITS src_sel [3] = {0,0,0};
+    BITS src_chan[3] = {0,0,0};
+    BITS src_rel [3] = {0,0,0};
+    BITS src_neg [3] = {0,0,0};
+
+    GLuint swizzle_key;
+
+    GLuint number_of_operands = r700GetNumOperands(pAsm);
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        get_src_properties(alu_instruction_ptr,
+                           src,
+                           &(src_sel[src]), 
+                           &(src_rel[src]), 
+                           &(src_chan[src]), 
+                           &(src_neg[src]) );
+    }
+
+
+    swizzle_key = ( (is_const( src_sel[0] ) ? 4 : 0) + 
+                           (is_const( src_sel[1] ) ? 2 : 0) + 
+                           (is_const( src_sel[2] ) ? 1 : 0) 
+                         );
+
+    alu_instruction_ptr->m_Word1.f.bank_swizzle = BANK_SWIZZLE_VEC[swizzle_key];
+
+    for (src=0; src<number_of_operands; src++) 
+    {
+        sel  = src_sel [src];
+        chan = src_chan[src];
+        rel  = src_rel [src];
+        neg  = src_neg [src];
+
+
+        bank_swizzle = alu_instruction_ptr->m_Word1.f.bank_swizzle;
+
+        if( is_gpr(sel) ) 
+        {
+            if( GL_FALSE == cycle_for_vector_bank_swizzle(bank_swizzle, src, &cycle) )
+            {
+                return GL_FALSE;
+            }
+
+            if ( (src  == 1)          && 
+                 (sel  == src_sel[0]) &&
+                 (chan == src_chan[0]) ) 
+            {        
+            }
+            else 
+            {
+                if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) )
+                {
+                    return GL_FALSE;
+                }
+            }
+        }
+        else if( is_const(sel) ) 
+        {                  
+            const_count++;
+
+            if( is_cfile(sel) ) 
+            {        
+                if( GL_FALSE == reserve_cfile(pAsm, sel, chan) )
+                {
+                    return GL_FALSE;
+                }
+            }
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm)
+{
+    GLuint    number_of_scalar_operations;
+    GLboolean is_single_scalar_operation;
+    GLuint    scalar_channel_index;
+
+    PVSSRC * pcurrent_source;
+    int    current_source_index;
+    GLuint contiguous_slots_needed;
+
+    GLuint    uNumSrc = r700GetNumOperands(pAsm);
+    GLuint    channel_swizzle, j;
+    GLuint    chan_counter[4] = {0, 0, 0, 0};
+    PVSSRC *  pSource[3];
+    GLboolean bSplitInst = GL_FALSE;
+
+    if (1 == pAsm->D.dst.math) 
+    {
+        is_single_scalar_operation = GL_TRUE;
+        number_of_scalar_operations = 1;
+    }
+    else 
+    {
+        is_single_scalar_operation = GL_FALSE;
+        number_of_scalar_operations = 4;
+        
+        /* check read port, only very preliminary algorithm, not count in 
+           src0/1 same comp case and prev slot repeat case; also not count relative
+           addressing. TODO: improve performance. */
+        for(j=0; j<uNumSrc; j++)
+        {
+            pSource[j] = &(pAsm->S[j].src);
+        }
+        for(scalar_channel_index=0; scalar_channel_index<4; scalar_channel_index++) 
+        {
+            for(j=0; j<uNumSrc; j++) 
+            {
+                switch (scalar_channel_index) 
+                {
+                    case 0: channel_swizzle = pSource[j]->swizzlex; break;
+                    case 1: channel_swizzle = pSource[j]->swizzley; break;
+                    case 2: channel_swizzle = pSource[j]->swizzlez; break;
+                    case 3: channel_swizzle = pSource[j]->swizzlew; break;
+                    default: channel_swizzle = SQ_SEL_MASK; break;
+                }
+                if ( ((pSource[j]->rtype == SRC_REG_TEMPORARY) || 
+                     (pSource[j]->rtype == SRC_REG_INPUT))
+                     && (channel_swizzle <= SQ_SEL_W) )
+                {                    
+                    chan_counter[channel_swizzle]++;                        
+                }
+            }
+        }
+        if(   (chan_counter[SQ_SEL_X] > 3)
+           || (chan_counter[SQ_SEL_Y] > 3)
+           || (chan_counter[SQ_SEL_Z] > 3)
+           || (chan_counter[SQ_SEL_W] > 3) ) /* each chan bank has only 3 ports. */
+        {
+            bSplitInst = GL_TRUE;
+        }
+    }
+
+    contiguous_slots_needed = 0;
+
+    if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) 
+    {
+        contiguous_slots_needed = 4;
+    }
+
+    initialize(pAsm);    
+
+    for (scalar_channel_index=0;
+            scalar_channel_index < number_of_scalar_operations; 
+                scalar_channel_index++) 
+    {
+        R700ALUInstruction* alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction);
+        if (alu_instruction_ptr == NULL) 
+		{
+			return GL_FALSE;
+		}
+        Init_R700ALUInstruction(alu_instruction_ptr);
+        
+        //src 0
+        current_source_index = 0;
+        pcurrent_source = &(pAsm->S[0].src);
+
+        if (GL_FALSE == assemble_alu_src(alu_instruction_ptr,
+                                         current_source_index,
+                                         pcurrent_source, 
+                                         scalar_channel_index) )     
+        {
+            return GL_FALSE;
+        }
+   
+        if (pAsm->D.dst.math == 0) 
+        {            
+            // Process source 1            
+            current_source_index = 1;
+            pcurrent_source = &(pAsm->S[current_source_index].src);
+
+            if (GL_FALSE == assemble_alu_src(alu_instruction_ptr,
+                                             current_source_index,
+                                             pcurrent_source, 
+                                             scalar_channel_index) ) 
+            {
+                return GL_FALSE;
+            }
+        }
+
+        //other bits
+        alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP;
+
+        if(   (is_single_scalar_operation == GL_TRUE) 
+           || (GL_TRUE == bSplitInst) )
+        {
+            alu_instruction_ptr->m_Word0.f.last = 1;
+        }
+        else 
+        {
+            alu_instruction_ptr->m_Word0.f.last = (scalar_channel_index == 3) ?  1 : 0;
+        }
+
+        alu_instruction_ptr->m_Word0.f.pred_sel                = 0x0;
+        alu_instruction_ptr->m_Word1_OP2.f.update_pred         = 0x0;  
+        alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0;
+
+        // dst
+        if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || 
+            (pAsm->D.dst.rtype == DST_REG_OUT) ) 
+        {
+            alu_instruction_ptr->m_Word1.f.dst_gpr  = pAsm->D.dst.reg;
+        }
+        else 
+        {
+            r700_error(ERROR_ASM_ALUDSTBADTYPE, "Only temp destination registers supported for ALU dest regs.");
+            return GL_FALSE;
+        }
+
+        alu_instruction_ptr->m_Word1.f.dst_rel  = SQ_ABSOLUTE;  //D.rtype
+
+        if ( is_single_scalar_operation == GL_TRUE ) 
+        {
+            // Override scalar_channel_index since only one scalar value will be written
+            if(pAsm->D.dst.writex) 
+            {
+                scalar_channel_index = 0;
+            }
+            else if(pAsm->D.dst.writey) 
+            {
+                scalar_channel_index = 1;
+            }
+            else if(pAsm->D.dst.writez) 
+            {
+                scalar_channel_index = 2;
+            }
+            else if(pAsm->D.dst.writew) 
+            {
+                scalar_channel_index = 3;
+            }
+        }
+
+        alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index;
+
+        alu_instruction_ptr->m_Word1.f.clamp    = pAsm->pILInst[pAsm->uiCurInst].SaturateMode;
+
+        if (pAsm->D.dst.op3) 
+        {            
+            //op3
+
+            alu_instruction_ptr->m_Word1_OP3.f.alu_inst = pAsm->D.dst.opcode;
+
+            //There's 3rd src for op3
+            current_source_index = 2;
+            pcurrent_source = &(pAsm->S[current_source_index].src);
+
+            if ( GL_FALSE == assemble_alu_src(alu_instruction_ptr,
+                                              current_source_index,
+                                              pcurrent_source, 
+                                              scalar_channel_index) ) 
+            {
+                return GL_FALSE;
+            }
+        }
+        else 
+        {
+            //op2
+            if (pAsm->bR6xx)
+            {
+                alu_instruction_ptr->m_Word1_OP2.f6.alu_inst           = pAsm->D.dst.opcode;
+
+                alu_instruction_ptr->m_Word1_OP2.f6.src0_abs           = 0x0;
+                alu_instruction_ptr->m_Word1_OP2.f6.src1_abs           = 0x0;
+
+                //alu_instruction_ptr->m_Word1_OP2.f6.update_execute_mask = 0x0;
+                //alu_instruction_ptr->m_Word1_OP2.f6.update_pred         = 0x0;
+                switch (scalar_channel_index) 
+                {
+                    case 0: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writex; 
+                        break;
+                    case 1: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writey; 
+                        break;
+                    case 2: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writez; 
+                        break;
+                    case 3: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = pAsm->D.dst.writew; 
+                        break;
+                    default: 
+                        alu_instruction_ptr->m_Word1_OP2.f6.write_mask = SQ_SEL_MASK; 
+                        break;
+                }            
+                alu_instruction_ptr->m_Word1_OP2.f6.omod               = SQ_ALU_OMOD_OFF;
+            }
+            else
+            {
+                alu_instruction_ptr->m_Word1_OP2.f.alu_inst           = pAsm->D.dst.opcode;
+
+                alu_instruction_ptr->m_Word1_OP2.f.src0_abs           = 0x0;
+                alu_instruction_ptr->m_Word1_OP2.f.src1_abs           = 0x0;
+
+                //alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0;
+                //alu_instruction_ptr->m_Word1_OP2.f.update_pred         = 0x0;
+                switch (scalar_channel_index) 
+                {
+                    case 0: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writex; 
+                        break;
+                    case 1: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writey; 
+                        break;
+                    case 2: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writez; 
+                        break;
+                    case 3: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = pAsm->D.dst.writew; 
+                        break;
+                    default: 
+                        alu_instruction_ptr->m_Word1_OP2.f.write_mask = SQ_SEL_MASK; 
+                        break;
+                }            
+                alu_instruction_ptr->m_Word1_OP2.f.omod               = SQ_ALU_OMOD_OFF;
+            }
+        }
+
+        if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) )
+        {
+            return GL_FALSE;
+        }
+
+        /*
+         * Judge the type of current instruction, is it vector or scalar 
+         * instruction.
+         */        
+        if (is_single_scalar_operation) 
+        {
+            if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) )
+            {
+                return GL_FALSE;
+            }
+        }
+        else 
+        {
+            if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) )
+            {
+                return 1;
+            }
+        }
+
+        contiguous_slots_needed = 0;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean next_ins(r700_AssemblerBase *pAsm)
+{
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+    if( GL_TRUE == IsTex(pILInst->Opcode) )
+    {
+        if( GL_FALSE == assemble_tex_instruction(pAsm) ) 
+        {
+            r700_error(ERROR_ASM_TEXINSTRUCTION, "Error assembling TEX instruction");
+            return GL_FALSE;
+        }
+    }
+    else 
+    {   //ALU      
+        if( GL_FALSE == assemble_alu_instruction(pAsm) ) 
+        {
+            r700_error(ERROR_ASM_TEXINSTRUCTION, "Error assembling ALU instruction");
+            return GL_FALSE;
+        }
+    } 
+      
+    if(pAsm->D.dst.rtype == DST_REG_OUT) 
+    {
+        if(pAsm->D.dst.op3) 
+        {        
+            // There is no mask for OP3 instructions, so all channels are written        
+            pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] = 0xF;
+        }
+        else 
+        {
+            pAsm->pucOutMask[pAsm->D.dst.reg - pAsm->starting_export_register_number] 
+               |= (unsigned char)pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask;
+        }
+    }
+    
+    //reset for next inst.
+    pAsm->D.bits    = 0;
+    pAsm->S[0].bits = 0;
+    pAsm->S[1].bits = 0;
+    pAsm->S[2].bits = 0;
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode)
+{
+    BITS tmp;
+
+    checkop1(pAsm);
+
+    tmp = gethelpr(pAsm);
+
+    // opcode  tmp.x,    a.x
+    // MOV     dst,      tmp.x
+
+    pAsm->D.dst.opcode = opcode;
+    pAsm->D.dst.math = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg    = tmp;
+    pAsm->D.dst.writex = 1;
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // Now replicate result to all necessary channels in destination
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_ABS(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MAX;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+ 
+    pAsm->S[1].bits = pAsm->S[0].bits;
+    flipneg_PVSSRC(&(pAsm->S[1].src));
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_ADD(r700_AssemblerBase *pAsm)
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_ADD;
+ 
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+ 
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_SUB)
+    {
+        flipneg_PVSSRC(&(pAsm->S[1].src));
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_BAD(char *opcode_str) 
+{
+    r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction (%s)", opcode_str);
+    return GL_FALSE;
+}
+
+GLboolean assemble_CMP(r700_AssemblerBase *pAsm)
+{
+    int tmp;
+
+    if( GL_FALSE == checkop3(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_CNDGE;
+    pAsm->D.dst.op3     = 1;  
+
+    tmp = (-1);
+
+    if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask)
+    {
+        //OP3 has no support for write mask
+        tmp = gethelpr(pAsm);
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = tmp;
+
+        nomask_PVSDST(&(pAsm->D.dst));
+    }
+    else 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+              
+    if( GL_FALSE == assemble_src(pAsm, 2, 1) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, 2) ) 
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if (0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask) 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+
+        pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+        //tmp for source
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[0].src.reg   = tmp;
+
+        noneg_PVSSRC(&(pAsm->S[0].src));
+        noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+        if( GL_FALSE == next_ins(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_COS(r700_AssemblerBase *pAsm)
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_COS);
+}
+ 
+GLboolean assemble_DOT(r700_AssemblerBase *pAsm)
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+        return GL_FALSE;
+    }
+ 
+    pAsm->D.dst.opcode = SQ_OP2_INST_DOT4;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if(OPCODE_DP3 == pAsm->pILInst[pAsm->uiCurInst].Opcode)
+    {
+        zerocomp_PVSSRC(&(pAsm->S[0].src), 3);
+        zerocomp_PVSSRC(&(pAsm->S[1].src), 3);
+    }
+    else if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_DPH) 
+    {
+        onecomp_PVSSRC(&(pAsm->S[1].src), 3);
+    } 
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_DST(r700_AssemblerBase *pAsm)
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    onecomp_PVSSRC(&(pAsm->S[0].src), 0);
+    onecomp_PVSSRC(&(pAsm->S[0].src), 3);
+
+    onecomp_PVSSRC(&(pAsm->S[1].src), 0);
+    onecomp_PVSSRC(&(pAsm->S[1].src), 2);
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_EX2(r700_AssemblerBase *pAsm)
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_EXP_IEEE);
+}
+ 
+GLboolean assemble_FLR(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_FLOOR;  
+
+    if ( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm)
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_FLT_TO_INT);
+}
+
+GLboolean assemble_FRC(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_FRACT; 
+
+    if ( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_KIL(r700_AssemblerBase *pAsm)
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_KILLGT;  
+  
+    if ( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.writex = 0;
+    pAsm->D.dst.writey = 0;
+    pAsm->D.dst.writez = 0;
+    pAsm->D.dst.writew = 0;
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = 0;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_0);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+
+    if(PROGRAM_TEMPORARY == pAsm->pILInst[pAsm->uiCurInst].DstReg.File)
+    {
+        pAsm->S[1].src.reg = pAsm->pILInst[pAsm->uiCurInst].DstReg.Index + pAsm->starting_temp_register_number;
+    }
+    else
+    {   //PROGRAM_OUTPUT
+        pAsm->S[1].src.reg = pAsm->uiFP_OutputMap[pAsm->pILInst[pAsm->uiCurInst].DstReg.Index];
+    }
+  
+    setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+    noswizzle_PVSSRC(&(pAsm->S[1].src));
+  
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->pR700Shader->killIsUsed = GL_TRUE;
+    
+    return GL_TRUE;
+}
+
+GLboolean assemble_LG2(r700_AssemblerBase *pAsm) 
+{ 
+    return assemble_math_function(pAsm, SQ_OP2_INST_LOG_IEEE);
+}
+
+GLboolean assemble_LRP(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+    if( GL_FALSE == checkop3(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    tmp = gethelpr(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_ADD;
+
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    nomask_PVSDST(&(pAsm->D.dst));
+
+          
+    if( GL_FALSE == assemble_src(pAsm, 1, 0) ) 
+    {
+	    return GL_FALSE;
+    }
+
+    if ( GL_FALSE == assemble_src(pAsm, 2, 1) )   
+    {
+	    return GL_FALSE;
+    }
+
+    neg_PVSSRC(&(pAsm->S[1].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+	    return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+
+    if( GL_FALSE == assemble_src(pAsm, 0, 1) ) 
+    {
+        return GL_FALSE;
+    }
+    if( GL_FALSE == assemble_src(pAsm, 2, -1) ) 
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_MAD(struct r700_AssemblerBase *pAsm) 
+{
+    int tmp, ii;
+    GLboolean bReplaceDst = GL_FALSE;
+    struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]);
+
+	if( GL_FALSE == checkop3(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+	pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;  
+	pAsm->D.dst.op3     = 1; 
+
+	tmp = (-1);
+
+    if(PROGRAM_TEMPORARY == pILInst->DstReg.File)
+    {   /* TODO : more investigation on MAD src and dst using same register */
+        for(ii=0; ii<3; ii++)
+        {
+            if(   (PROGRAM_TEMPORARY == pILInst->SrcReg[ii].File)
+               && (pILInst->DstReg.Index == pILInst->SrcReg[ii].Index) )
+            {
+                bReplaceDst = GL_TRUE;
+                break;
+            }
+        }
+    }
+    if(0xF != pILInst->DstReg.WriteMask)
+    {   /* OP3 has no support for write mask */
+        bReplaceDst = GL_TRUE;
+    }
+
+	if(GL_TRUE == bReplaceDst)
+    {
+        tmp = gethelpr(pAsm);
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = tmp;
+
+        nomask_PVSDST(&(pAsm->D.dst));
+    }
+    else 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+              
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 2, -1) ) 
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+	if (GL_TRUE == bReplaceDst) 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+
+        pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+        //tmp for source
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[0].src.reg   = tmp;
+
+        noneg_PVSSRC(&(pAsm->S[0].src));
+        noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+        if( GL_FALSE == next_ins(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+/* LIT dst, src */
+GLboolean assemble_LIT(r700_AssemblerBase *pAsm)
+{
+    unsigned int dstReg;
+    unsigned int dstType;
+    unsigned int srcReg;
+    unsigned int srcType;
+    checkop1(pAsm);
+    int tmp = gethelpr(pAsm);
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+    dstReg  = pAsm->D.dst.reg;
+    dstType = pAsm->D.dst.rtype;
+    srcReg  = pAsm->S[0].src.reg;
+    srcType = pAsm->S[0].src.rtype;
+
+    /* dst.xw, <- 1.0  */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_MOV;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 1;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 1;
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_1;
+    pAsm->S[0].src.swizzley = SQ_SEL_1;
+    pAsm->S[0].src.swizzlez = SQ_SEL_1;
+    pAsm->S[0].src.swizzlew = SQ_SEL_1;
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* dst.y = max(src.x, 0.0) */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_MAX;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 0;
+    pAsm->D.dst.writey   = 1;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 0;
+    pAsm->S[0].src.rtype = srcType;
+    pAsm->S[0].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_X;
+    pAsm->S[0].src.swizzley = SQ_SEL_X;
+    pAsm->S[0].src.swizzlez = SQ_SEL_X;
+    pAsm->S[0].src.swizzlew = SQ_SEL_X;
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[1].src.reg   = tmp;
+    setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[1].src));
+    pAsm->S[1].src.swizzlex = SQ_SEL_0;
+    pAsm->S[1].src.swizzley = SQ_SEL_0;
+    pAsm->S[1].src.swizzlez = SQ_SEL_0;
+    pAsm->S[1].src.swizzlew = SQ_SEL_0;
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* before: dst.w = log(src.y)
+     * after : dst.x = log(src.y)
+     * why change dest register is that dst.w has been initialized as 1 before
+     */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_LOG_CLAMPED;
+    pAsm->D.dst.math     = 1;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 1;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 0;
+    pAsm->S[0].src.rtype = srcType;
+    pAsm->S[0].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_Y;
+    pAsm->S[0].src.swizzley = SQ_SEL_Y;
+    pAsm->S[0].src.swizzlez = SQ_SEL_Y;
+    pAsm->S[0].src.swizzlew = SQ_SEL_Y;
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* before: tmp.x = amd MUL_LIT(src.w, dst.w, src.x ) */
+    /* after : tmp.x = amd MUL_LIT(src.w, dst.x, src.x ) */
+    pAsm->D.dst.opcode   = SQ_OP3_INST_MUL_LIT;
+    pAsm->D.dst.op3      = 1;
+    pAsm->D.dst.rtype    = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg      = tmp;
+    pAsm->D.dst.writex   = 1;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 0;
+    pAsm->D.dst.writew   = 0;
+
+    pAsm->S[0].src.rtype = srcType;
+    pAsm->S[0].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_W;
+    pAsm->S[0].src.swizzley = SQ_SEL_W;
+    pAsm->S[0].src.swizzlez = SQ_SEL_W;
+    pAsm->S[0].src.swizzlew = SQ_SEL_W;
+
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[1].src.reg   = dstReg;
+    setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[1].src));
+    pAsm->S[1].src.swizzlex = SQ_SEL_X;
+    pAsm->S[1].src.swizzley = SQ_SEL_X;
+    pAsm->S[1].src.swizzlez = SQ_SEL_X;
+    pAsm->S[1].src.swizzlew = SQ_SEL_X;
+
+    pAsm->S[2].src.rtype = srcType;
+    pAsm->S[2].src.reg   = srcReg;
+    setaddrmode_PVSSRC(&(pAsm->S[2].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[2].src));
+    pAsm->S[2].src.swizzlex = SQ_SEL_X;
+    pAsm->S[2].src.swizzley = SQ_SEL_X;
+    pAsm->S[2].src.swizzlez = SQ_SEL_X;
+    pAsm->S[2].src.swizzlew = SQ_SEL_X;
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    /* dst.z = exp(tmp.x) */
+    pAsm->D.dst.opcode   = SQ_OP2_INST_EXP_IEEE;
+    pAsm->D.dst.math     = 1;
+    pAsm->D.dst.rtype    = dstType;
+    pAsm->D.dst.reg      = dstReg;
+    pAsm->D.dst.writex   = 0;
+    pAsm->D.dst.writey   = 0;
+    pAsm->D.dst.writez   = 1;
+    pAsm->D.dst.writew   = 0;
+
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+    pAsm->S[0].src.swizzlex = SQ_SEL_X;
+    pAsm->S[0].src.swizzley = SQ_SEL_X;
+    pAsm->S[0].src.swizzlez = SQ_SEL_X;
+    pAsm->S[0].src.swizzlew = SQ_SEL_X;
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MAX(r700_AssemblerBase *pAsm) 
+{
+	if( GL_FALSE == checkop2(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	pAsm->D.dst.opcode = SQ_OP2_INST_MAX; 
+	
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MIN(r700_AssemblerBase *pAsm) 
+{
+	if( GL_FALSE == checkop2(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	pAsm->D.dst.opcode = SQ_OP2_INST_MIN;  
+
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+	{
+		return GL_FALSE;
+	}
+ 
+	if( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MOV(r700_AssemblerBase *pAsm) 
+{
+    checkop1(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if (GL_FALSE == assemble_dst(pAsm))
+    {
+        return GL_FALSE;
+    }
+
+    if (GL_FALSE == assemble_src(pAsm, 0, -1))
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_MUL(r700_AssemblerBase *pAsm) 
+{
+	if( GL_FALSE == checkop2(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == next_ins(pAsm) ) 
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_POW(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+    checkop1(pAsm);
+
+    tmp = gethelpr(pAsm);
+
+    // LG2 tmp.x,     a.swizzle
+    pAsm->D.dst.opcode = SQ_OP2_INST_LOG_IEEE;  
+    pAsm->D.dst.math = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // MUL tmp.x,     tmp.x, b.swizzle
+    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // EX2 dst.mask,          tmp.x
+    // EX2 tmp.x,             tmp.x
+    pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE;
+    pAsm->D.dst.math = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    // Now replicate result to all necessary channels in destination
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_RCP(r700_AssemblerBase *pAsm) 
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_RECIP_IEEE);
+}
+ 
+GLboolean assemble_RSQ(r700_AssemblerBase *pAsm) 
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_RECIPSQRT_IEEE);
+}
+ 
+GLboolean assemble_SIN(r700_AssemblerBase *pAsm) 
+{
+    return assemble_math_function(pAsm, SQ_OP2_INST_SIN);
+}
+ 
+GLboolean assemble_SCS(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+	checkop1(pAsm);
+
+	tmp = gethelpr(pAsm);
+
+	// COS tmp.x,    a.x
+	pAsm->D.dst.opcode = SQ_OP2_INST_COS;
+	pAsm->D.dst.math = 1;
+
+	setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+	pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+	pAsm->D.dst.reg = tmp;
+	pAsm->D.dst.writex = 1;
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if ( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	// SIN tmp.y,    a.x
+	pAsm->D.dst.opcode = SQ_OP2_INST_SIN;
+	pAsm->D.dst.math = 1;
+
+	setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+	pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+	pAsm->D.dst.reg = tmp;
+	pAsm->D.dst.writey = 1;
+
+	if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+	{
+		return GL_FALSE;
+	}
+
+	if( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	// MOV dst.mask,     tmp
+	pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+	if( GL_FALSE == assemble_dst(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+	setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+	pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+	pAsm->S[0].src.reg = tmp;
+
+	noswizzle_PVSSRC(&(pAsm->S[0].src));
+	pAsm->S[0].src.swizzlez = SQ_SEL_0;
+	pAsm->S[0].src.swizzlew = SQ_SEL_0;
+
+	if ( GL_FALSE == next_ins(pAsm) )
+	{
+		return GL_FALSE;
+	}
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_SGE(r700_AssemblerBase *pAsm) 
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_SETGE;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+	    return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+	    return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+	    return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_SLT(r700_AssemblerBase *pAsm) 
+{
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_SETGT;  
+
+    if( GL_FALSE == assemble_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+                
+    if( GL_FALSE == assemble_src(pAsm, 0, 1) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, 0) )  
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_STP(r700_AssemblerBase *pAsm) 
+{
+    return GL_TRUE;
+}
+ 
+GLboolean assemble_TEX(r700_AssemblerBase *pAsm) 
+{
+    GLboolean src_const;
+
+    switch (pAsm->pILInst[pAsm->uiCurInst].SrcReg[0].File)
+    {
+    case PROGRAM_CONSTANT:
+    case PROGRAM_LOCAL_PARAM:
+    case PROGRAM_ENV_PARAM:
+    case PROGRAM_STATE_VAR:
+        src_const = GL_TRUE;
+    case PROGRAM_TEMPORARY:
+    case PROGRAM_INPUT:
+        src_const = GL_FALSE;
+    }
+
+    if (GL_TRUE == src_const) 
+    {
+        r700_error(TODO_ASM_CONSTTEXADDR, "TODO: Texture coordinates from a constant register not supported.");
+        return GL_FALSE;
+    }
+
+    switch (pAsm->pILInst[pAsm->uiCurInst].Opcode) 
+    {
+        case OPCODE_TEX:
+            pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;            
+            break;
+        case OPCODE_TXB:            
+            r700_error(TODO_ASM_TXB, "do not support TXB yet");
+            return GL_FALSE;
+            break;
+        case OPCODE_TXP:            
+            /* TODO : tex proj version : divid first 3 components by 4th */ 
+            pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;
+            break;
+        default:
+            r700_error(ERROR_ASM_BADTEXINST, "Internal error: bad texture op (not TEX)");
+            return GL_FALSE;
+            break;
+    }
+
+    // Set src1 to tex unit id
+    pAsm->S[1].src.reg   = pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit;
+    pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+
+    //No sw info from mesa compiler, so hard code here.
+    pAsm->S[1].src.swizzlex = SQ_SEL_X;
+    pAsm->S[1].src.swizzley = SQ_SEL_Y;
+    pAsm->S[1].src.swizzlez = SQ_SEL_Z;
+    pAsm->S[1].src.swizzlew = SQ_SEL_W;
+
+    if( GL_FALSE == tex_dst(pAsm) )
+    {
+        return GL_FALSE;
+    }
+ 
+    if( GL_FALSE == tex_src(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    if ( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_XPD(r700_AssemblerBase *pAsm) 
+{
+    BITS tmp;
+
+    if( GL_FALSE == checkop2(pAsm) )
+    {
+	    return GL_FALSE;
+    }
+
+    tmp = gethelpr(pAsm);
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = tmp;
+    nomask_PVSDST(&(pAsm->D.dst));
+  
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+ 
+    swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Z, SQ_SEL_X, SQ_SEL_Y, SQ_SEL_0);
+    swizzleagain_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Y, SQ_SEL_Z, SQ_SEL_X, SQ_SEL_0);
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
+    if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask)
+    {
+        tmp = gethelpr(pAsm);
+
+        setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+        pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+        pAsm->D.dst.reg   = tmp;
+
+        nomask_PVSDST(&(pAsm->D.dst));
+    }
+    else 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+    {
+        return GL_FALSE;
+    }
+
+    if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+    {
+        return GL_FALSE;
+    }
+ 
+    swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Y, SQ_SEL_Z, SQ_SEL_X, SQ_SEL_0);
+    swizzleagain_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Z, SQ_SEL_X, SQ_SEL_Y, SQ_SEL_0);
+
+    // result1 + (neg) result0
+    setaddrmode_PVSSRC(&(pAsm->S[2].src),ADDR_ABSOLUTE);
+    pAsm->S[2].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[2].src.reg   = tmp;
+
+    neg_PVSSRC(&(pAsm->S[2].src));
+    noswizzle_PVSSRC(&(pAsm->S[2].src));
+
+    if( GL_FALSE == next_ins(pAsm) ) 
+    {
+        return GL_FALSE;
+    }
+
+
+    if(0xF != pAsm->pILInst[pAsm->uiCurInst].DstReg.WriteMask) 
+    {
+        if( GL_FALSE == assemble_dst(pAsm) )
+        {
+            return GL_FALSE;
+        }
+
+        pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+        // Use tmp as source
+        setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+        pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+        pAsm->S[0].src.reg   = tmp;
+
+        noneg_PVSSRC(&(pAsm->S[0].src));
+        noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+        if( GL_FALSE == next_ins(pAsm) )
+        {
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm)
+{
+    return GL_TRUE;
+}
+
+GLboolean assemble_IF(r700_AssemblerBase *pAsm)
+{
+    return GL_TRUE;
+}
+
+GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm)
+{
+    return GL_TRUE;
+}
+
+GLboolean AssembleInstr(GLuint uiNumberInsts,
+                        struct prog_instruction *pILInst, 
+						r700_AssemblerBase *pR700AsmCode)
+{
+    GLuint i;
+
+    pR700AsmCode->pILInst = pILInst;
+	for(i=0; i<uiNumberInsts; i++)
+    {
+        pR700AsmCode->uiCurInst = i;
+
+        switch (pILInst[i].Opcode)
+        {
+        case OPCODE_ABS: 
+            if ( GL_FALSE == assemble_ABS(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_ADD: 
+        case OPCODE_SUB: 
+            if ( GL_FALSE == assemble_ADD(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_ARL: 
+            r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_ARL ");
+            //if ( GL_FALSE == assemble_BAD("ARL") ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_ARR: 
+            r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_ARR ");
+            //if ( GL_FALSE == assemble_BAD("ARR") ) 
+                return GL_FALSE;
+            break;
+
+        case OPCODE_CMP: 
+            if ( GL_FALSE == assemble_CMP(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_COS: 
+            if ( GL_FALSE == assemble_COS(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_DP3: 
+        case OPCODE_DP4: 
+        case OPCODE_DPH: 
+            if ( GL_FALSE == assemble_DOT(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_DST: 
+            if ( GL_FALSE == assemble_DST(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_EX2: 
+            if ( GL_FALSE == assemble_EX2(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_EXP: 
+            r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_EXP ");
+            //if ( GL_FALSE == assemble_BAD("EXP") ) 
+                return GL_FALSE;
+            break; // approx of EX2
+
+        case OPCODE_FLR:     
+            if ( GL_FALSE == assemble_FLR(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        //case OP_FLR_INT: 
+        //    if ( GL_FALSE == assemble_FLR_INT() ) 
+        //        return GL_FALSE;
+        //    break;  
+
+        case OPCODE_FRC: 
+            if ( GL_FALSE == assemble_FRC(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_KIL: 
+            if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_LG2: 
+            if ( GL_FALSE == assemble_LG2(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_LIT:
+            if ( GL_FALSE == assemble_LIT(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_LRP: 
+            if ( GL_FALSE == assemble_LRP(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_LOG: 
+            r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_LOG ");
+            //if ( GL_FALSE == assemble_BAD("LOG") ) 
+                return GL_FALSE;
+            break; // approx of LG2
+
+        case OPCODE_MAD: 
+            if ( GL_FALSE == assemble_MAD(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_MAX: 
+            if ( GL_FALSE == assemble_MAX(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_MIN: 
+            if ( GL_FALSE == assemble_MIN(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_MOV: 
+            if ( GL_FALSE == assemble_MOV(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_MUL: 
+            if ( GL_FALSE == assemble_MUL(pR700AsmCode) ) 
+                return GL_FALSE;
+            break; 
+
+        case OPCODE_POW: 
+            if ( GL_FALSE == assemble_POW(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_RCP: 
+            if ( GL_FALSE == assemble_RCP(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_RSQ: 
+            if ( GL_FALSE == assemble_RSQ(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_SIN: 
+            if ( GL_FALSE == assemble_SIN(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+        case OPCODE_SCS: 
+            if ( GL_FALSE == assemble_SCS(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_SGE: 
+            if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) 
+                return GL_FALSE;
+            break; 
+        case OPCODE_SLT: 
+            if ( GL_FALSE == assemble_SLT(pR700AsmCode) ) 
+                return GL_FALSE;
+            break; 
+
+        //case OP_STP: 
+        //    if ( GL_FALSE == assemble_STP(pR700AsmCode) ) 
+        //        return GL_FALSE;
+        //    break;
+
+        case OPCODE_SWZ: 
+            if ( GL_FALSE == assemble_MOV(pR700AsmCode) ) 
+            {
+                return GL_FALSE; 
+            }
+            else
+            {
+                if( (i+1)<uiNumberInsts )
+                {
+                    if(OPCODE_END != pILInst[i+1].Opcode)
+                    {
+                        if( GL_TRUE == IsTex(pILInst[i+1].Opcode) )
+                        {
+                            pR700AsmCode->pInstDeps[i+1].nDstDep = i+1; //=1?
+                        }
+                    }
+                }
+            }
+            break;
+
+        case OPCODE_TEX: 
+        case OPCODE_TXB:  
+        case OPCODE_TXP: 
+            if ( GL_FALSE == assemble_TEX(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+
+        case OPCODE_XPD: 
+            if ( GL_FALSE == assemble_XPD(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;  
+
+        case OPCODE_IF   : 
+            if ( GL_FALSE == assemble_IF(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_ELSE : 
+            r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_ELSE ");
+            //if ( GL_FALSE == assemble_BAD("ELSE") ) 
+                return GL_FALSE;
+            break;
+        case OPCODE_ENDIF: 
+            if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) 
+                return GL_FALSE;
+            break;
+
+        //case OPCODE_EXPORT: 
+        //    if ( GL_FALSE == assemble_EXPORT() ) 
+        //        return GL_FALSE;
+        //    break;
+
+        case OPCODE_END: 
+			//pR700AsmCode->uiCurInst = i;
+			//This is to remaind that if in later exoort there is depth/stencil
+			//export, we need a mov to re-arrange DST channel, where using a
+			//psuedo inst, we will use this end inst to do it.
+            return GL_TRUE;
+
+        default:
+            r700_error(ERROR_ASM_UNKNOWNILINST, "internal: unknown instruction");
+            return GL_FALSE;
+        }
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean Process_Export(r700_AssemblerBase* pAsm,
+                         GLuint type,
+                         GLuint export_starting_index,
+                         GLuint export_count, 
+                         GLuint starting_register_number,
+                         GLboolean is_depth_export)
+{
+    unsigned char ucWriteMask;
+
+    check_current_clause(pAsm, CF_EMPTY_CLAUSE);
+    check_current_clause(pAsm, CF_EXPORT_CLAUSE); //alloc the cf_current_export_clause_ptr
+
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.type = type;
+
+    switch (type) 
+    {
+        case SQ_EXPORT_PIXEL:
+            if(GL_TRUE == is_depth_export) 
+            {
+                pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = SQ_CF_PIXEL_Z;
+            }
+            else 
+            {
+                pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = SQ_CF_PIXEL_MRT0 + export_starting_index;
+            }
+            break;
+
+        case SQ_EXPORT_POS:
+            pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = SQ_CF_POS_0 + export_starting_index; 
+            break;
+
+        case SQ_EXPORT_PARAM:
+            pAsm->cf_current_export_clause_ptr->m_Word0.f.array_base  = 0x0 + export_starting_index; 
+            break;
+
+        default:
+            r700_error(ERROR_ASM_BADEXPORTTYPE, "Unknown export type: %d", type);
+            return GL_FALSE;
+            break;
+    }
+
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.rw_gpr      = starting_register_number;
+
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.rw_rel      = SQ_ABSOLUTE;
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.index_gpr   = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word0.f.elem_size   = 0x3; 
+
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.burst_count      = (export_count - 1);
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.end_of_program   = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.cf_inst          = SQ_CF_INST_EXPORT;  // _DONE
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.whole_quad_mode  = 0x0;
+    pAsm->cf_current_export_clause_ptr->m_Word1.f.barrier          = 0x1;
+
+    if (export_count == 1) 
+    {
+        ucWriteMask = pAsm->pucOutMask[starting_register_number - pAsm->starting_export_register_number];
+
+        if( (ucWriteMask & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_X;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_MASK;
+        }
+        if( ((ucWriteMask>>1) & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_Y;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_MASK;
+        }
+        if( ((ucWriteMask>>2) & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_Z;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_MASK;
+        }
+        if( ((ucWriteMask>>3) & 0x1) != 0)
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_W;
+        }
+        else
+        {
+            pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_MASK;
+        }
+    }
+    else 
+    {
+        // This should only be used if all components for all registers have been written
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_X;
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_Y;
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_Z;
+        pAsm->cf_current_export_clause_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_W;
+    }
+
+    pAsm->cf_last_export_ptr = pAsm->cf_current_export_clause_ptr;
+
+    return GL_TRUE;
+}
+
+GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, BITS depth_channel_select)
+{
+	gl_inst_opcode Opcode_save = pAsm->pILInst[pAsm->uiCurInst].Opcode; //Should be OPCODE_END
+    pAsm->pILInst[pAsm->uiCurInst].Opcode = OPCODE_MOV;
+
+    // MOV depth_export_register.hw_depth_channel, depth_export_register.depth_channel_select
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg   = pAsm->depth_export_register_number;
+
+    pAsm->D.dst.writex = 1;   // depth          goes in R channel for HW                       
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = pAsm->depth_export_register_number;
+
+    setswizzle_PVSSRC(&(pAsm->S[0].src), depth_channel_select);
+
+    noneg_PVSSRC(&(pAsm->S[0].src));
+
+    if( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->pILInst[pAsm->uiCurInst].Opcode = Opcode_save;
+
+    return GL_TRUE;
+}
+ 
+GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode,
+                                   GLbitfield          OutputsWritten)  
+{ 
+    unsigned int unBit;
+
+    if(pR700AsmCode->depth_export_register_number >= 0) 
+    {
+        if( GL_FALSE == Move_Depth_Exports_To_Correct_Channels(pR700AsmCode, SQ_SEL_Z) )  // depth
+		{
+			return GL_FALSE;
+		}
+    }
+
+    unBit = 1 << FRAG_RESULT_COLOR;
+	if(OutputsWritten & unBit)
+	{
+		if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PIXEL, 
+                                       0, 
+                                       1, 
+                                       pR700AsmCode->uiFP_OutputMap[FRAG_RESULT_COLOR], 
+                                       GL_FALSE) ) 
+        {
+            return GL_FALSE;
+        }
+	}
+	unBit = 1 << FRAG_RESULT_DEPTH;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PIXEL, 
+                                       0, 
+                                       1, 
+                                       pR700AsmCode->uiFP_OutputMap[FRAG_RESULT_DEPTH], 
+                                       GL_TRUE)) 
+        {
+            return GL_FALSE;
+        }
+	}
+
+    if(pR700AsmCode->cf_last_export_ptr != NULL) 
+    {
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst        = SQ_CF_INST_EXPORT_DONE;
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.end_of_program = 0x1;
+    }
+
+    return GL_TRUE;
+}
+
+GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode,
+                                 GLbitfield          OutputsWritten)  
+{
+    unsigned int unBit;
+    unsigned int i;
+
+    GLuint export_starting_index  = 0;
+    GLuint export_count           = pR700AsmCode->number_of_exports;
+
+    unBit = 1 << VERT_RESULT_HPOS;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode, 
+                                       SQ_EXPORT_POS, 
+                                       export_starting_index, 
+                                       1, 
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_HPOS],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_count--;
+
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE;
+	}
+
+    pR700AsmCode->number_of_exports = export_count;
+
+	unBit = 1 << VERT_RESULT_COL0;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode, 
+                                       SQ_EXPORT_PARAM, 
+                                       export_starting_index, 
+                                       1, 
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_COL0],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_starting_index++;
+	}
+
+	unBit = 1 << VERT_RESULT_COL1;
+	if(OutputsWritten & unBit)
+	{
+        if( GL_FALSE == Process_Export(pR700AsmCode, 
+                                       SQ_EXPORT_PARAM, 
+                                       export_starting_index, 
+                                       1, 
+                                       pR700AsmCode->ucVP_OutputMap[VERT_RESULT_COL1],
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+
+        export_starting_index++;
+	}
+
+	for(i=0; i<8; i++)
+	{
+		unBit = 1 << (VERT_RESULT_TEX0 + i);
+		if(OutputsWritten & unBit)
+		{
+            if( GL_FALSE == Process_Export(pR700AsmCode,
+                                          SQ_EXPORT_PARAM, 
+                                          export_starting_index, 
+                                          1, 
+                                          pR700AsmCode->ucVP_OutputMap[VERT_RESULT_TEX0 + i],
+                                          GL_FALSE) )
+            {
+                return GL_FALSE;
+            }
+
+            export_starting_index++;
+		}
+	}
+
+    // At least one param should be exported
+    if (export_count) 
+    {
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE;    
+    }
+    else
+    {
+        if( GL_FALSE == Process_Export(pR700AsmCode,
+                                       SQ_EXPORT_PARAM, 
+                                       0, 
+                                       1, 
+                                       pR700AsmCode->starting_export_register_number,
+                                       GL_FALSE) )
+        {
+            return GL_FALSE;
+        }
+      
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_x = SQ_SEL_0;
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_y = SQ_SEL_0;
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_z = SQ_SEL_0;
+        pR700AsmCode->cf_last_export_ptr->m_Word1_SWIZ.f.sel_w = SQ_SEL_1;
+        pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE;
+    }
+
+    pR700AsmCode->cf_last_export_ptr->m_Word1.f.end_of_program = 0x1;
+
+    return GL_TRUE;
+}
+
+GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode)
+{
+    FREE(pR700AsmCode->pucOutMask);
+    FREE(pR700AsmCode->pInstDeps);
+    return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h
new file mode 100644
index 0000000000..e9b21b802e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_assembler.h
@@ -0,0 +1,512 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_ASSEMBLER_H_
+#define _R700_ASSEMBLER_H_
+
+#include "main/mtypes.h"
+#include "shader/prog_instruction.h"
+
+#include "r700_chip.h"
+#include "r700_shaderinst.h"
+#include "r700_shader.h"
+
+typedef enum SHADER_PIPE_TYPE 
+{
+    SPT_VP = 0,
+    SPT_FP = 1
+} SHADER_PIPE_TYPE;
+
+typedef enum ConstantCycles 
+{
+    NUMBER_OF_CYCLES     = 3,
+    NUMBER_OF_COMPONENTS = 4
+} ConstantCycles;
+
+typedef enum  HARDWARE_LIMIT_VALUES  
+{
+   TEMPORARY_REGISTER_OFFSET = SQ_ALU_SRC_GPR_BASE,
+   MAX_TEMPORARY_REGISTERS   = SQ_ALU_SRC_GPR_SIZE,
+   MAX_CONSTANT_REGISTERS    = SQ_ALU_SRC_CFILE_SIZE,
+   CFILE_REGISTER_OFFSET     = SQ_ALU_SRC_CFILE_BASE,
+   NUMBER_OF_INPUT_COLORS    = 2,
+   NUMBER_OF_OUTPUT_COLORS   = 8,
+   NUMBER_OF_TEXTURE_UNITS   = 16,
+   MEGA_FETCH_BYTES          = 32
+} HARDWARE_LIMIT_VALUES;
+
+typedef enum AddressMode 
+{
+    ADDR_ABSOLUTE          = 0,
+    ADDR_RELATIVE_A0       = 1,
+    ADDR_RELATIVE_FLI_0    = 2,
+    NUMBER_OF_ADDR_MOD     = 3
+} AddressMode;
+
+typedef enum SrcRegisterType 
+{
+    SRC_REG_TEMPORARY      = 0,
+    SRC_REG_INPUT          = 1,
+    SRC_REG_CONSTANT       = 2,
+    SRC_REG_ALT_TEMPORARY  = 3,
+    NUMBER_OF_SRC_REG_TYPE = 4
+} SrcRegisterType;
+
+typedef enum DstRegisterType 
+{
+    DST_REG_TEMPORARY      = 0,
+    DST_REG_A0             = 1,
+    DST_REG_OUT            = 2,
+    DST_REG_OUT_X_REPL     = 3,
+    DST_REG_ALT_TEMPORARY  = 4,
+    DST_REG_INPUT          = 5,
+    NUMBER_OF_DST_REG_TYPE = 6
+} DstRegisterType;
+
+typedef unsigned int BITS;
+
+typedef struct PVSDSTtag 
+{
+	BITS opcode:8;     //(:6)  //@@@ really should be 10 bits for OP2
+	BITS math:1;
+	BITS predicated:1; //10   //8
+	BITS pred_inv  :1; //11   //8
+
+	BITS rtype:3;
+	BITS reg:10;       //24   //20
+
+	BITS writex:1;
+	BITS writey:1;
+	BITS writez:1;
+	BITS writew:1;     //28
+
+	BITS op3:1;       // 29  Represents *_OP3_* ALU opcode
+
+	BITS dualop:1;    // 30  //26
+
+	BITS addrmode0:1; //31   //29
+	BITS addrmode1:1; //32
+} PVSDST;
+
+typedef struct PVSSRCtag 
+{
+	BITS rtype:4;            
+	BITS addrmode0:1;        
+	BITS reg:10;      //15     (8)
+	BITS swizzlex:3;
+	BITS swizzley:3;
+	BITS swizzlez:3;
+	BITS swizzlew:3;  //27        
+
+	BITS negx:1;
+	BITS negy:1;
+	BITS negz:1;
+	BITS negw:1;      //31
+	//BITS addrsel:2;
+	BITS addrmode1:1; //32
+} PVSSRC;
+
+typedef struct PVSMATHtag 
+{
+	BITS rtype:4;
+	BITS spare:1;
+	BITS reg:8;
+	BITS swizzlex:3;
+	BITS swizzley:3;
+	BITS dstoff:2; // 2 bits of dest offset into alt ram
+	BITS opcode:4;
+	BITS negx:1;
+	BITS negy:1;
+	BITS dstcomp:2; // select dest component
+	BITS spare2:3;
+} PVSMATH;
+
+typedef union PVSDWORDtag 
+{
+	BITS    bits;
+	PVSDST  dst;
+	PVSSRC  src;
+	PVSMATH math;
+	float   f;
+} PVSDWORD;
+
+typedef struct VAP_OUT_VTX_FMT_0tag 
+{
+	BITS pos:1;      // 0
+	BITS misc:1;
+	BITS clip_dist0:1;
+	BITS clip_dist1:1;
+	BITS pos_param:1; // 4
+
+	BITS color0:1;    // 5
+	BITS color1:1;
+	BITS color2:1;
+	BITS color3:1;
+	BITS color4:1;
+	BITS color5:1;
+	BITS color6:1;
+	BITS color7:1;
+
+	BITS normal:1;    
+
+	BITS depth:1;          // 14
+
+	BITS point_size:1;     // 15   
+	BITS edge_flag:1;      
+	BITS rta_index:1;      //     shares same channel as kill_flag
+	BITS kill_flag:1;
+	BITS viewport_index:1; // 19   
+
+	BITS resvd1:12;        // 20
+} VAP_OUT_VTX_FMT_0;
+
+typedef struct VAP_OUT_VTX_FMT_1tag 
+{
+	BITS tex0comp:3;
+	BITS tex1comp:3;
+	BITS tex2comp:3;
+	BITS tex3comp:3;
+	BITS tex4comp:3;
+	BITS tex5comp:3;
+	BITS tex6comp:3;
+	BITS tex7comp:3;
+
+	BITS resvd:8;
+} VAP_OUT_VTX_FMT_1;
+
+typedef struct VAP_OUT_VTX_FMT_2tag 
+{
+	BITS tex8comp :3;
+	BITS tex9comp :3;
+	BITS tex10comp:3;
+	BITS tex11comp:3;
+	BITS tex12comp:3;
+	BITS tex13comp:3;
+	BITS tex14comp:3;
+	BITS tex15comp:3;
+
+	BITS resvd:8;
+} VAP_OUT_VTX_FMT_2;
+
+typedef struct OUT_FRAGMENT_FMT_0tag 
+{
+	BITS color0:1;
+	BITS color1:1;
+	BITS color2:1;
+	BITS color3:1;
+	BITS color4:1;
+	BITS color5:1;
+	BITS color6:1;
+	BITS color7:1;
+
+	BITS depth:1;
+	BITS stencil_ref:1;
+	BITS coverage_to_mask:1;
+	BITS mask:1;
+
+	BITS resvd1:20;
+} OUT_FRAGMENT_FMT_0;
+
+typedef enum  CF_CLAUSE_TYPE 
+{
+   CF_EXPORT_CLAUSE,
+   CF_ALU_CLAUSE,
+   CF_TEX_CLAUSE,
+   CF_VTX_CLAUSE,
+   CF_OTHER_CLAUSE,
+   CF_EMPTY_CLAUSE,
+   NUMBER_CF_CLAUSE_TYPES
+} CF_CLAUSE_TYPE;
+
+enum 
+{
+    MAX_BOOL_CONSTANTS   = 32,
+    MAX_INT_CONSTANTS    = 32,
+    MAX_FLOAT_CONSTANTS  = 256,
+
+    FC_NONE = 0,
+    FC_IF = 1,
+    FC_LOOP = 2,
+    FC_REP = 3,
+
+    COND_NONE = 0,
+    COND_BOOL = 1,
+    COND_PRED = 2,
+    COND_ALU = 3,
+
+    SAFEDIST_TEX = 6, ///< safe distance for using result of texture lookup in alu or another tex lookup
+    SAFEDIST_ALU = 6 ///< the same for alu->fc
+};
+
+typedef struct FC_LEVEL 
+{
+	unsigned int           first; ///< first fc instruction on level (if, rep, loop)
+	unsigned int*          mid; ///< middle instructions - else or all breaks on this level
+	unsigned int           midLen;
+	unsigned int           type;
+	unsigned int           cond;
+	unsigned int           inv;
+	unsigned int           bpush; ///< 1 if first instruction does branch stack push
+			 int           id; ///< id of bool or int variable
+} FC_LEVEL;
+
+typedef struct VTX_FETCH_METHOD 
+{
+	GLboolean bEnableMini;
+	GLuint mega_fetch_remainder;
+} VTX_FETCH_METHOD;
+
+typedef struct r700_AssemblerBase 
+{
+	R700ControlFlowSXClause*      cf_last_export_ptr;
+	R700ControlFlowSXClause*      cf_current_export_clause_ptr;
+	R700ControlFlowALUClause*     cf_current_alu_clause_ptr;
+	R700ControlFlowGenericClause* cf_current_tex_clause_ptr;
+	R700ControlFlowGenericClause* cf_current_vtx_clause_ptr;
+	R700ControlFlowGenericClause* cf_current_cf_clause_ptr;
+
+    //Result shader
+    R700_Shader * pR700Shader;
+
+	// No clause has been created yet
+	CF_CLAUSE_TYPE cf_current_clause_type;
+
+	GLuint number_of_exports;
+	GLuint number_of_colorandz_exports;
+	GLuint number_of_export_opcodes;
+
+	PVSDWORD D;
+	PVSDWORD S[3];
+
+	unsigned int uLastPosUpdate;
+
+	OUT_FRAGMENT_FMT_0     fp_stOutFmt0;
+
+	unsigned int uIIns;
+	unsigned int uOIns;
+	unsigned int number_used_registers;
+	unsigned int uUsedConsts; 
+
+	// Fragment programs
+	unsigned int uiFP_AttributeMap[FRAG_ATTRIB_MAX];
+	unsigned int uiFP_OutputMap[FRAG_RESULT_MAX];
+	unsigned int uBoolConsts;
+	unsigned int uIntConsts;
+	unsigned int uInsts;
+	unsigned int uConsts;
+
+	// Vertex programs
+	unsigned char ucVP_AttributeMap[VERT_ATTRIB_MAX];
+	unsigned char ucVP_OutputMap[VERT_RESULT_MAX];
+
+    unsigned char * pucOutMask;
+
+	//-----------------------------------------------------------------------------------
+	// flow control members
+	//-----------------------------------------------------------------------------------
+	unsigned int FCSP;
+	FC_LEVEL fc_stack[32];
+
+	unsigned int branch_depth;
+	unsigned int max_branch_depth;
+
+	//-----------------------------------------------------------------------------------
+	// ArgSubst used in Assemble_Source() function
+	//-----------------------------------------------------------------------------------
+	int aArgSubst[4];
+
+    GLint hw_gpr[ NUMBER_OF_CYCLES ][ NUMBER_OF_COMPONENTS ];
+    GLint hw_cfile_addr[ NUMBER_OF_COMPONENTS ];
+    GLint hw_cfile_chan[ NUMBER_OF_COMPONENTS ];
+
+    GLuint uOutputs;
+  
+    GLint color_export_register_number[NUMBER_OF_OUTPUT_COLORS];
+	GLint depth_export_register_number;
+
+	GLint stencil_export_register_number;
+	GLint coverage_to_mask_export_register_number;
+	GLint mask_export_register_number;
+
+	GLuint starting_export_register_number;
+	GLuint starting_vfetch_register_number;
+	GLuint starting_temp_register_number;
+	GLuint uHelpReg;
+	GLuint uFirstHelpReg;
+
+	GLboolean input_position_is_used;
+	GLboolean input_normal_is_used;
+
+	GLboolean input_color_is_used[NUMBER_OF_INPUT_COLORS];
+  
+	GLboolean input_texture_unit_is_used[NUMBER_OF_TEXTURE_UNITS];
+  
+    R700VertexGenericFetch* vfetch_instruction_ptr_array[VERT_ATTRIB_MAX];
+  
+	GLuint number_of_inputs;
+
+    InstDeps *pInstDeps;
+
+    SHADER_PIPE_TYPE currentShaderType;
+    struct prog_instruction * pILInst;
+    GLuint             uiCurInst;
+    GLboolean   bR6xx;
+} r700_AssemblerBase;
+
+//Internal use
+BITS addrmode_PVSDST(PVSDST * pPVSDST);
+void setaddrmode_PVSDST(PVSDST * pPVSDST, BITS addrmode);
+void nomask_PVSDST(PVSDST * pPVSDST);
+BITS addrmode_PVSSRC(PVSSRC* pPVSSRC);
+void setaddrmode_PVSSRC(PVSSRC* pPVSSRC, BITS addrmode);
+void setswizzle_PVSSRC(PVSSRC* pPVSSRC, BITS swz);
+void noswizzle_PVSSRC(PVSSRC* pPVSSRC);
+void swizzleagain_PVSSRC(PVSSRC * pPVSSRC, BITS x, BITS y, BITS z, BITS w);
+void neg_PVSSRC(PVSSRC* pPVSSRC);
+void noneg_PVSSRC(PVSSRC* pPVSSRC);
+void flipneg_PVSSRC(PVSSRC* pPVSSRC);
+void zerocomp_PVSSRC(PVSSRC* pPVSSRC, int c);
+void onecomp_PVSSRC(PVSSRC* pPVSSRC, int c);
+BITS is_misc_component_exported(VAP_OUT_VTX_FMT_0* pOutVTXFmt0);
+BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) ;
+GLboolean is_reduction_opcode(PVSDWORD * dest);
+GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size);
+
+unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm);
+
+GLboolean IsTex(gl_inst_opcode Opcode);
+GLboolean IsAlu(gl_inst_opcode Opcode);
+int check_current_clause(r700_AssemblerBase* pAsm,
+					     CF_CLAUSE_TYPE      new_clause_type);
+GLboolean add_vfetch_instruction(r700_AssemblerBase*     pAsm,
+								 R700VertexInstruction*  vertex_instruction_ptr);
+GLboolean add_tex_instruction(r700_AssemblerBase*     pAsm,
+                              R700TextureInstruction* tex_instruction_ptr);
+GLboolean assemble_vfetch_instruction(r700_AssemblerBase* pAsm,
+								GLuint gl_client_id,
+                                GLuint destination_register,
+								GLuint number_of_elements,
+                                GLenum dataElementType,
+								VTX_FETCH_METHOD* pFetchMethod);
+GLuint gethelpr(r700_AssemblerBase* pAsm);
+void resethelpr(r700_AssemblerBase* pAsm);
+void checkop_init(r700_AssemblerBase* pAsm);
+GLboolean mov_temp(r700_AssemblerBase* pAsm, int src);
+GLboolean checkop1(r700_AssemblerBase* pAsm);
+GLboolean checkop2(r700_AssemblerBase* pAsm);
+GLboolean checkop3(r700_AssemblerBase* pAsm);
+GLboolean assemble_src(r700_AssemblerBase *pAsm,
+                       int src, 
+                       int fld);
+GLboolean assemble_dst(r700_AssemblerBase *pAsm);
+GLboolean tex_dst(r700_AssemblerBase *pAsm);
+GLboolean tex_src(r700_AssemblerBase *pAsm);
+GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm);
+void initialize(r700_AssemblerBase *pAsm);
+GLboolean assemble_alu_src(R700ALUInstruction*  alu_instruction_ptr,
+                           int                  source_index,
+                           PVSSRC*              pSource,
+                           BITS                 scalar_channel_index);
+GLboolean add_alu_instruction(r700_AssemblerBase* pAsm,
+                              R700ALUInstruction* alu_instruction_ptr,
+                              GLuint              contiguous_slots_needed);
+void get_src_properties(R700ALUInstruction*  alu_instruction_ptr,
+                        int                  source_index,
+                        BITS*                psrc_sel,
+                        BITS*                psrc_rel,
+                        BITS*                psrc_chan,
+                        BITS*                psrc_neg);
+int is_cfile(BITS sel);
+int is_const(BITS sel);
+int is_gpr(BITS sel);
+GLboolean reserve_cfile(r700_AssemblerBase* pAsm, 
+                        GLuint sel, 
+                        GLuint chan);
+GLboolean reserve_gpr(r700_AssemblerBase* pAsm, GLuint sel, GLuint chan, GLuint cycle);
+GLboolean cycle_for_scalar_bank_swizzle(const int swiz, const int sel, GLuint* pCycle);
+GLboolean cycle_for_vector_bank_swizzle(const int swiz, const int sel, GLuint* pCycle);
+GLboolean check_scalar(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr);
+GLboolean check_vector(r700_AssemblerBase* pAsm,
+                       R700ALUInstruction* alu_instruction_ptr);
+GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm);
+GLboolean next_ins(r700_AssemblerBase *pAsm);
+GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode);
+GLboolean assemble_ABS(r700_AssemblerBase *pAsm);
+GLboolean assemble_ADD(r700_AssemblerBase *pAsm);
+GLboolean assemble_BAD(char *opcode_str);
+GLboolean assemble_CMP(r700_AssemblerBase *pAsm);
+GLboolean assemble_COS(r700_AssemblerBase *pAsm);
+GLboolean assemble_DOT(r700_AssemblerBase *pAsm);
+GLboolean assemble_DST(r700_AssemblerBase *pAsm);
+GLboolean assemble_EX2(r700_AssemblerBase *pAsm);
+GLboolean assemble_FLR(r700_AssemblerBase *pAsm);
+GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm);
+GLboolean assemble_FRC(r700_AssemblerBase *pAsm);
+GLboolean assemble_KIL(r700_AssemblerBase *pAsm);
+GLboolean assemble_LG2(r700_AssemblerBase *pAsm);
+GLboolean assemble_LRP(r700_AssemblerBase *pAsm);
+GLboolean assemble_MAD(r700_AssemblerBase *pAsm);
+GLboolean assemble_LIT(r700_AssemblerBase *pAsm);
+GLboolean assemble_MAX(r700_AssemblerBase *pAsm);
+GLboolean assemble_MIN(r700_AssemblerBase *pAsm);
+GLboolean assemble_MOV(r700_AssemblerBase *pAsm);
+GLboolean assemble_MUL(r700_AssemblerBase *pAsm);
+GLboolean assemble_POW(r700_AssemblerBase *pAsm);
+GLboolean assemble_RCP(r700_AssemblerBase *pAsm);
+GLboolean assemble_RSQ(r700_AssemblerBase *pAsm);
+GLboolean assemble_SIN(r700_AssemblerBase *pAsm);
+GLboolean assemble_SCS(r700_AssemblerBase *pAsm);
+GLboolean assemble_SGE(r700_AssemblerBase *pAsm);
+GLboolean assemble_SLT(r700_AssemblerBase *pAsm);
+GLboolean assemble_STP(r700_AssemblerBase *pAsm);
+GLboolean assemble_TEX(r700_AssemblerBase *pAsm);
+GLboolean assemble_XPD(r700_AssemblerBase *pAsm);
+GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm);
+GLboolean assemble_IF(r700_AssemblerBase *pAsm);
+GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm);
+
+GLboolean Process_Export(r700_AssemblerBase* pAsm,
+                         GLuint type, 
+                         GLuint export_starting_index,
+                         GLuint export_count, 
+                         GLuint starting_register_number,
+                         GLboolean is_depth_export);
+GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, 
+                                                 BITS depth_channel_select);
+
+
+//Interface
+GLboolean AssembleInstr(GLuint uiNumberInsts,
+                        struct prog_instruction *pILInst, 
+						r700_AssemblerBase *pR700AsmCode);
+GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten);  
+GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten);
+
+int       Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader);
+GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode);
+
+#endif //_R700_ASSEMBLER_H_
diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c
new file mode 100644
index 0000000000..087d17312e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_chip.c
@@ -0,0 +1,724 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "main/imports.h"
+#include "main/glheader.h"
+
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r700_state.h"
+#include "r600_tex.h"
+#include "r700_oglprog.h"
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+#include "r700_ioctl.h"
+
+#define LINK_STATES(reg)                                            \
+do                                                                  \
+{                                                                   \
+    pStateListWork->puiValue = (unsigned int*)&(r700->reg);         \
+    pStateListWork->unOffset = mm##reg - ASIC_CONTEXT_BASE_INDEX; \
+    pStateListWork->pNext    = pStateListWork + 1;                  \
+    pStateListWork++;                                               \
+}while(0)
+
+GLboolean r700InitChipObject(context_t *context)
+{
+    ContextState * pStateListWork;
+
+    R700_CHIP_CONTEXT *r700 = &context->hw;
+
+    /* init state list */
+    r700->pStateList = (ContextState*) MALLOC (sizeof(ContextState)*sizeof(R700_CHIP_CONTEXT)/sizeof(unsigned int));
+    pStateListWork = r700->pStateList;
+
+    // misc
+    LINK_STATES(TA_CNTL_AUX);
+    LINK_STATES(VC_ENHANCE);
+    LINK_STATES(SQ_DYN_GPR_CNTL_PS_FLUSH_REQ);
+    LINK_STATES(DB_DEBUG);
+    LINK_STATES(DB_WATERMARKS);
+
+    // SC
+    LINK_STATES(PA_SC_SCREEN_SCISSOR_TL);
+    LINK_STATES(PA_SC_SCREEN_SCISSOR_BR);
+    LINK_STATES(PA_SC_WINDOW_OFFSET);
+    LINK_STATES(PA_SC_WINDOW_SCISSOR_TL);
+    LINK_STATES(PA_SC_WINDOW_SCISSOR_BR);
+    LINK_STATES(PA_SC_CLIPRECT_RULE);
+    LINK_STATES(PA_SC_CLIPRECT_0_TL);
+    LINK_STATES(PA_SC_CLIPRECT_0_BR);
+    LINK_STATES(PA_SC_CLIPRECT_1_TL);
+    LINK_STATES(PA_SC_CLIPRECT_1_BR);
+    LINK_STATES(PA_SC_CLIPRECT_2_TL);
+    LINK_STATES(PA_SC_CLIPRECT_2_BR);
+    LINK_STATES(PA_SC_CLIPRECT_3_TL);
+    LINK_STATES(PA_SC_CLIPRECT_3_BR);
+    LINK_STATES(PA_SC_EDGERULE);
+    LINK_STATES(PA_SC_GENERIC_SCISSOR_TL);
+    LINK_STATES(PA_SC_GENERIC_SCISSOR_BR);
+    LINK_STATES(PA_SC_LINE_STIPPLE);
+    LINK_STATES(PA_SC_MPASS_PS_CNTL);
+    LINK_STATES(PA_SC_MODE_CNTL);
+    LINK_STATES(PA_SC_LINE_CNTL);
+    LINK_STATES(PA_SC_AA_CONFIG);
+    LINK_STATES(PA_SC_AA_SAMPLE_LOCS_MCTX);
+    LINK_STATES(PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX);
+    LINK_STATES(PA_SC_AA_MASK);
+
+    // SU
+    LINK_STATES(PA_SU_POINT_SIZE);
+    LINK_STATES(PA_SU_POINT_MINMAX);
+    LINK_STATES(PA_SU_LINE_CNTL);
+    LINK_STATES(PA_SU_SC_MODE_CNTL);
+    LINK_STATES(PA_SU_VTX_CNTL);
+    LINK_STATES(PA_SU_POLY_OFFSET_DB_FMT_CNTL);
+    LINK_STATES(PA_SU_POLY_OFFSET_CLAMP);
+    LINK_STATES(PA_SU_POLY_OFFSET_FRONT_SCALE);
+    LINK_STATES(PA_SU_POLY_OFFSET_FRONT_OFFSET);
+    LINK_STATES(PA_SU_POLY_OFFSET_BACK_SCALE);
+    LINK_STATES(PA_SU_POLY_OFFSET_BACK_OFFSET);
+
+    // CL
+    LINK_STATES(PA_CL_CLIP_CNTL);
+    LINK_STATES(PA_CL_VTE_CNTL);
+    LINK_STATES(PA_CL_VS_OUT_CNTL);
+    LINK_STATES(PA_CL_NANINF_CNTL);
+    LINK_STATES(PA_CL_GB_VERT_CLIP_ADJ);
+    LINK_STATES(PA_CL_GB_VERT_DISC_ADJ);
+    LINK_STATES(PA_CL_GB_HORZ_CLIP_ADJ);
+    LINK_STATES(PA_CL_GB_HORZ_DISC_ADJ);
+
+    // CB
+    LINK_STATES(CB_CLEAR_RED_R6XX);
+    LINK_STATES(CB_CLEAR_GREEN_R6XX);
+    LINK_STATES(CB_CLEAR_BLUE_R6XX);
+    LINK_STATES(CB_CLEAR_ALPHA_R6XX);
+    LINK_STATES(CB_TARGET_MASK);
+    LINK_STATES(CB_SHADER_MASK);
+    LINK_STATES(CB_BLEND_RED);
+    LINK_STATES(CB_BLEND_GREEN);
+    LINK_STATES(CB_BLEND_BLUE);
+    LINK_STATES(CB_BLEND_ALPHA);
+    LINK_STATES(CB_FOG_RED_R6XX);
+    LINK_STATES(CB_FOG_GREEN_R6XX);
+    LINK_STATES(CB_FOG_BLUE_R6XX);
+    LINK_STATES(CB_SHADER_CONTROL);
+    LINK_STATES(CB_COLOR_CONTROL);
+    LINK_STATES(CB_CLRCMP_CONTROL);
+    LINK_STATES(CB_CLRCMP_SRC);
+    LINK_STATES(CB_CLRCMP_DST);
+    LINK_STATES(CB_CLRCMP_MSK);
+    LINK_STATES(CB_BLEND_CONTROL);
+
+    // SX
+    LINK_STATES(SX_MISC);
+    LINK_STATES(SX_ALPHA_TEST_CONTROL);
+
+    // VGT
+    LINK_STATES(VGT_MAX_VTX_INDX);
+    LINK_STATES(VGT_MIN_VTX_INDX);
+    LINK_STATES(VGT_INDX_OFFSET);
+    LINK_STATES(VGT_MULTI_PRIM_IB_RESET_INDX);
+    LINK_STATES(VGT_OUTPUT_PATH_CNTL);
+    LINK_STATES(VGT_HOS_CNTL);
+    LINK_STATES(VGT_HOS_MAX_TESS_LEVEL);
+    LINK_STATES(VGT_HOS_MIN_TESS_LEVEL);
+    LINK_STATES(VGT_HOS_REUSE_DEPTH);
+    LINK_STATES(VGT_GROUP_PRIM_TYPE);
+    LINK_STATES(VGT_GROUP_FIRST_DECR);
+    LINK_STATES(VGT_GROUP_DECR);
+    LINK_STATES(VGT_GROUP_VECT_0_CNTL);
+    LINK_STATES(VGT_GROUP_VECT_1_CNTL);
+    LINK_STATES(VGT_GROUP_VECT_0_FMT_CNTL);
+    LINK_STATES(VGT_GROUP_VECT_1_FMT_CNTL);
+    LINK_STATES(VGT_GS_MODE);
+    LINK_STATES(VGT_PRIMITIVEID_EN);
+    LINK_STATES(VGT_DMA_NUM_INSTANCES);
+    LINK_STATES(VGT_MULTI_PRIM_IB_RESET_EN);
+    LINK_STATES(VGT_INSTANCE_STEP_RATE_0);
+    LINK_STATES(VGT_INSTANCE_STEP_RATE_1);
+    LINK_STATES(VGT_STRMOUT_EN);
+    LINK_STATES(VGT_REUSE_OFF);
+    LINK_STATES(VGT_VTX_CNT_EN);
+    LINK_STATES(VGT_STRMOUT_BUFFER_EN);
+
+    LINK_STATES(SQ_VTX_SEMANTIC_0);
+    LINK_STATES(SQ_VTX_SEMANTIC_1);
+    LINK_STATES(SQ_VTX_SEMANTIC_2);
+    LINK_STATES(SQ_VTX_SEMANTIC_3);
+    LINK_STATES(SQ_VTX_SEMANTIC_4);
+    LINK_STATES(SQ_VTX_SEMANTIC_5);
+    LINK_STATES(SQ_VTX_SEMANTIC_6);
+    LINK_STATES(SQ_VTX_SEMANTIC_7);
+    LINK_STATES(SQ_VTX_SEMANTIC_8);
+    LINK_STATES(SQ_VTX_SEMANTIC_9);
+    LINK_STATES(SQ_VTX_SEMANTIC_10);
+    LINK_STATES(SQ_VTX_SEMANTIC_11);
+    LINK_STATES(SQ_VTX_SEMANTIC_12);
+    LINK_STATES(SQ_VTX_SEMANTIC_13);
+    LINK_STATES(SQ_VTX_SEMANTIC_14);
+    LINK_STATES(SQ_VTX_SEMANTIC_15);
+    LINK_STATES(SQ_VTX_SEMANTIC_16);
+    LINK_STATES(SQ_VTX_SEMANTIC_17);
+    LINK_STATES(SQ_VTX_SEMANTIC_18);
+    LINK_STATES(SQ_VTX_SEMANTIC_19);
+    LINK_STATES(SQ_VTX_SEMANTIC_20);
+    LINK_STATES(SQ_VTX_SEMANTIC_21);
+    LINK_STATES(SQ_VTX_SEMANTIC_22);
+    LINK_STATES(SQ_VTX_SEMANTIC_23);
+    LINK_STATES(SQ_VTX_SEMANTIC_24);
+    LINK_STATES(SQ_VTX_SEMANTIC_25);
+    LINK_STATES(SQ_VTX_SEMANTIC_26);
+    LINK_STATES(SQ_VTX_SEMANTIC_27);
+    LINK_STATES(SQ_VTX_SEMANTIC_28);
+    LINK_STATES(SQ_VTX_SEMANTIC_29);
+    LINK_STATES(SQ_VTX_SEMANTIC_30);
+    LINK_STATES(SQ_VTX_SEMANTIC_31);
+
+    // SPI
+    LINK_STATES(SPI_VS_OUT_ID_0);
+    LINK_STATES(SPI_VS_OUT_ID_1);
+    LINK_STATES(SPI_VS_OUT_ID_2);
+    LINK_STATES(SPI_VS_OUT_ID_3);
+    LINK_STATES(SPI_VS_OUT_ID_4);
+    LINK_STATES(SPI_VS_OUT_ID_5);
+    LINK_STATES(SPI_VS_OUT_ID_6);
+    LINK_STATES(SPI_VS_OUT_ID_7);
+    LINK_STATES(SPI_VS_OUT_ID_8);
+    LINK_STATES(SPI_VS_OUT_ID_9);
+
+    LINK_STATES(SPI_PS_INPUT_CNTL_0);
+    LINK_STATES(SPI_PS_INPUT_CNTL_1);
+    LINK_STATES(SPI_PS_INPUT_CNTL_2);
+    LINK_STATES(SPI_PS_INPUT_CNTL_3);
+    LINK_STATES(SPI_PS_INPUT_CNTL_4);
+    LINK_STATES(SPI_PS_INPUT_CNTL_5);
+    LINK_STATES(SPI_PS_INPUT_CNTL_6);
+    LINK_STATES(SPI_PS_INPUT_CNTL_7);
+    LINK_STATES(SPI_PS_INPUT_CNTL_8);
+    LINK_STATES(SPI_PS_INPUT_CNTL_9);
+    LINK_STATES(SPI_PS_INPUT_CNTL_10);
+    LINK_STATES(SPI_PS_INPUT_CNTL_11);
+    LINK_STATES(SPI_PS_INPUT_CNTL_12);
+    LINK_STATES(SPI_PS_INPUT_CNTL_13);
+    LINK_STATES(SPI_PS_INPUT_CNTL_14);
+    LINK_STATES(SPI_PS_INPUT_CNTL_15);
+    LINK_STATES(SPI_PS_INPUT_CNTL_16);
+    LINK_STATES(SPI_PS_INPUT_CNTL_17);
+    LINK_STATES(SPI_PS_INPUT_CNTL_18);
+    LINK_STATES(SPI_PS_INPUT_CNTL_19);
+    LINK_STATES(SPI_PS_INPUT_CNTL_20);
+    LINK_STATES(SPI_PS_INPUT_CNTL_21);
+    LINK_STATES(SPI_PS_INPUT_CNTL_22);
+    LINK_STATES(SPI_PS_INPUT_CNTL_23);
+    LINK_STATES(SPI_PS_INPUT_CNTL_24);
+    LINK_STATES(SPI_PS_INPUT_CNTL_25);
+    LINK_STATES(SPI_PS_INPUT_CNTL_26);
+    LINK_STATES(SPI_PS_INPUT_CNTL_27);
+    LINK_STATES(SPI_PS_INPUT_CNTL_28);
+    LINK_STATES(SPI_PS_INPUT_CNTL_29);
+    LINK_STATES(SPI_PS_INPUT_CNTL_30);
+    LINK_STATES(SPI_PS_INPUT_CNTL_31);
+
+    LINK_STATES(SPI_VS_OUT_CONFIG);
+    LINK_STATES(SPI_THREAD_GROUPING);
+    LINK_STATES(SPI_PS_IN_CONTROL_0);
+    LINK_STATES(SPI_PS_IN_CONTROL_1);
+    LINK_STATES(SPI_INTERP_CONTROL_0);
+    LINK_STATES(SPI_INPUT_Z);
+    LINK_STATES(SPI_FOG_CNTL);
+    LINK_STATES(SPI_FOG_FUNC_SCALE);
+    LINK_STATES(SPI_FOG_FUNC_BIAS);
+
+    // SQ
+    LINK_STATES(SQ_ESGS_RING_ITEMSIZE);
+    LINK_STATES(SQ_GSVS_RING_ITEMSIZE);
+    LINK_STATES(SQ_ESTMP_RING_ITEMSIZE);
+    LINK_STATES(SQ_GSTMP_RING_ITEMSIZE);
+    LINK_STATES(SQ_VSTMP_RING_ITEMSIZE);
+    LINK_STATES(SQ_PSTMP_RING_ITEMSIZE);
+    LINK_STATES(SQ_FBUF_RING_ITEMSIZE);
+    LINK_STATES(SQ_REDUC_RING_ITEMSIZE);
+    //LINK_STATES(SQ_GS_VERT_ITEMSIZE);
+
+    pStateListWork->puiValue = (unsigned int*)&(r700->SQ_GS_VERT_ITEMSIZE);
+    pStateListWork->unOffset = mmSQ_GS_VERT_ITEMSIZE - ASIC_CONTEXT_BASE_INDEX;
+    pStateListWork->pNext    = NULL;  /* END OF STATE LIST */
+
+    return GL_TRUE;
+}
+
+void r700SetupVTXConstants(GLcontext  * ctx,
+			   unsigned int nStreamID,
+			   void *       pAos,
+			   unsigned int size,      /* number of elements in vector */
+			   unsigned int stride,
+			   unsigned int count)     /* number of vectors in stream */
+{
+    context_t *context = R700_CONTEXT(ctx);
+    uint32_t *dest;
+    struct radeon_aos * paos = (struct radeon_aos *)pAos;
+    offset_modifiers offset_mod = {NO_SHIFT, 0, 0xFFFFFFFF};
+
+    BATCH_LOCALS(&context->radeon);
+
+    unsigned int uSQ_VTX_CONSTANT_WORD0_0;
+    unsigned int uSQ_VTX_CONSTANT_WORD1_0;
+    unsigned int uSQ_VTX_CONSTANT_WORD2_0 = 0;
+    unsigned int uSQ_VTX_CONSTANT_WORD3_0 = 0;
+    unsigned int uSQ_VTX_CONSTANT_WORD6_0 = 0;
+
+    if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710))
+	    r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, TC_ACTION_ENA_bit);
+    else
+	    r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, VC_ACTION_ENA_bit);
+
+    uSQ_VTX_CONSTANT_WORD0_0 = paos->offset;
+    uSQ_VTX_CONSTANT_WORD1_0 = count * (size * 4) - 1;
+
+    uSQ_VTX_CONSTANT_WORD2_0 |= 0 << BASE_ADDRESS_HI_shift /* TODO */
+	    |stride << SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift
+	    |GetSurfaceFormat(GL_FLOAT, size, NULL) << SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift /* TODO : trace back api for initial data type, not only GL_FLOAT */
+	    |SQ_NUM_FORMAT_SCALED << SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift
+	    |SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit;
+
+    uSQ_VTX_CONSTANT_WORD3_0 |= 1 << MEM_REQUEST_SIZE_shift;
+
+    uSQ_VTX_CONSTANT_WORD6_0 |= SQ_TEX_VTX_VALID_BUFFER << SQ_TEX_RESOURCE_WORD6_0__TYPE_shift;
+
+    BEGIN_BATCH_NO_AUTOSTATE(9);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
+    R600_OUT_BATCH((nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE);
+
+    R600_OUT_BATCH_RELOC(uSQ_VTX_CONSTANT_WORD0_0,
+                         paos->bo,
+                         uSQ_VTX_CONSTANT_WORD0_0,
+                         RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD3_0);
+    R600_OUT_BATCH(0);
+    R600_OUT_BATCH(0);
+    R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD6_0);
+
+    END_BATCH();
+    COMMIT_BATCH();
+
+}
+
+int r700SetupStreams(GLcontext * ctx)
+{
+    context_t         *context = R700_CONTEXT(ctx);
+
+    BATCH_LOCALS(&context->radeon);
+
+    struct r700_vertex_program *vpc
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+	struct vertex_buffer *vb = &tnl->vb;
+
+    unsigned int unBit;
+    unsigned int i, j = 0;
+
+    BEGIN_BATCH_NO_AUTOSTATE(6);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
+	R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
+    R600_OUT_BATCH(0);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
+    R600_OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX);
+    R600_OUT_BATCH(0);
+    END_BATCH();
+    COMMIT_BATCH();
+
+    context->radeon.tcl.aos_count = 0;
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(vpc->mesa_program.Base.InputsRead & unBit)
+		{
+			rcommon_emit_vector(ctx,
+					    &context->radeon.tcl.aos[j],
+					    vb->AttribPtr[i]->data,
+					    vb->AttribPtr[i]->size,
+					    vb->AttribPtr[i]->stride,
+					    vb->Count);
+
+			/* currently aos are packed */
+			r700SetupVTXConstants(ctx,
+					      i,
+					      (void*)(&context->radeon.tcl.aos[j]),
+					      (unsigned int)context->radeon.tcl.aos[j].components,
+					      (unsigned int)context->radeon.tcl.aos[j].stride * 4,
+					      (unsigned int)context->radeon.tcl.aos[j].count);
+			j++;
+			context->radeon.tcl.aos_count++;
+		}
+	}
+
+    return R600_FALLBACK_NONE;
+}
+
+GLboolean r700SendContextStates(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+
+    R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+
+    ContextState * pState = r700->pStateList;
+    ContextState * pInit;
+    unsigned int   toSend;
+    unsigned int   ui;
+
+    while(NULL != pState)
+    {
+        toSend = 1;
+
+        pInit = pState;
+
+	while(NULL != pState->pNext)
+	{
+                if ((pState->pNext->unOffset - pState->unOffset) > 1)
+                {
+			break;
+                }
+                else
+                {
+			pState = pState->pNext;
+			toSend++;
+                }
+	}
+
+        pState = pState->pNext;
+
+        BEGIN_BATCH_NO_AUTOSTATE(toSend + 2);
+        R600_OUT_BATCH_REGSEQ(((pInit->unOffset + ASIC_CONTEXT_BASE_INDEX)<<2), toSend);
+        for(ui=0; ui<toSend; ui++)
+        {
+                R600_OUT_BATCH(*(pInit->puiValue));
+		pInit = pInit->pNext;
+        };
+        END_BATCH();
+    };
+    COMMIT_BATCH();
+
+    return GL_TRUE;
+}
+
+
+GLboolean r700SendDepthTargetState(context_t *context, int id)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	offset_modifiers offset_mod;
+	BATCH_LOCALS(&context->radeon);
+
+	rrb = radeon_get_depthbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return GL_FALSE;
+	}
+
+	offset_mod.shift     = NO_SHIFT;
+	offset_mod.shiftbits = 0;
+	offset_mod.mask      = 0xFFFFFFFF;
+
+        BEGIN_BATCH_NO_AUTOSTATE(9);
+	R600_OUT_BATCH_REGSEQ(DB_DEPTH_SIZE, 2);
+	R600_OUT_BATCH(r700->DB_DEPTH_SIZE.u32All);
+	R600_OUT_BATCH(r700->DB_DEPTH_VIEW.u32All);
+	R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 3);
+	R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All,
+			     rrb->bo,
+			     r700->DB_DEPTH_BASE.u32All,
+			     0, RADEON_GEM_DOMAIN_VRAM, 0, &offset_mod);
+	R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All);
+	R600_OUT_BATCH(r700->DB_HTILE_DATA_BASE.u32All);
+        END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(24);
+	R600_OUT_BATCH_REGSEQ(DB_STENCIL_CLEAR, 2);
+	R600_OUT_BATCH(r700->DB_STENCIL_CLEAR.u32All);
+	R600_OUT_BATCH(r700->DB_DEPTH_CLEAR.u32All);
+
+	R600_OUT_BATCH_REGSEQ(DB_STENCILREFMASK, 2);
+	R600_OUT_BATCH(r700->DB_STENCILREFMASK.u32All);
+	R600_OUT_BATCH(r700->DB_STENCILREFMASK_BF.u32All);
+
+	R600_OUT_BATCH_REGVAL(DB_DEPTH_CONTROL, r700->DB_DEPTH_CONTROL.u32All);
+	R600_OUT_BATCH_REGVAL(DB_SHADER_CONTROL, r700->DB_SHADER_CONTROL.u32All);
+
+	R600_OUT_BATCH_REGSEQ(DB_RENDER_CONTROL, 2);
+	R600_OUT_BATCH(r700->DB_RENDER_CONTROL.u32All);
+	R600_OUT_BATCH(r700->DB_RENDER_OVERRIDE.u32All);
+
+	R600_OUT_BATCH_REGVAL(DB_HTILE_SURFACE, r700->DB_HTILE_SURFACE.u32All);
+	R600_OUT_BATCH_REGVAL(DB_ALPHA_TO_MASK, r700->DB_ALPHA_TO_MASK.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+	r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM,
+		     DB_ACTION_ENA_bit | DB_DEST_BASE_ENA_bit);
+
+	return GL_TRUE;
+}
+
+GLboolean r700SendRenderTargetState(context_t *context, int id)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	offset_modifiers offset_mod;
+	BATCH_LOCALS(&context->radeon);
+
+	rrb = radeon_get_colorbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return GL_FALSE;
+	}
+
+	if (id > R700_MAX_RENDER_TARGETS)
+		return GL_FALSE;
+
+	if (!r700->render_target[id].enabled)
+		return GL_FALSE;
+
+	offset_mod.shift     = NO_SHIFT;
+	offset_mod.shiftbits = 0;
+	offset_mod.mask      = 0xFFFFFFFF;
+
+        BEGIN_BATCH_NO_AUTOSTATE(3);
+	R600_OUT_BATCH_REGSEQ(CB_COLOR0_BASE + (4 * id), 1);
+	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
+			     rrb->bo,
+			     r700->render_target[id].CB_COLOR0_BASE.u32All,
+			     0, RADEON_GEM_DOMAIN_VRAM, 0, &offset_mod);
+        END_BATCH();
+
+	if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) &&
+	    (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) {
+		BEGIN_BATCH_NO_AUTOSTATE(2);
+		R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_BASE_UPDATE, 0));
+		R600_OUT_BATCH((2 << id));
+		END_BATCH();
+	}
+
+        BEGIN_BATCH_NO_AUTOSTATE(18);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), r700->render_target[id].CB_COLOR0_SIZE.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), r700->render_target[id].CB_COLOR0_VIEW.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_TILE + (4 * id), r700->render_target[id].CB_COLOR0_TILE.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_FRAG + (4 * id), r700->render_target[id].CB_COLOR0_FRAG.u32All);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), r700->render_target[id].CB_COLOR0_MASK.u32All);
+        END_BATCH();
+
+	if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) {
+		BEGIN_BATCH_NO_AUTOSTATE(3);
+		R600_OUT_BATCH_REGVAL(CB_BLEND0_CONTROL + (4 * id), r700->render_target[id].CB_BLEND0_CONTROL.u32All);
+		END_BATCH();
+	}
+
+	COMMIT_BATCH();
+
+	r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM,
+		     CB_ACTION_ENA_bit | (1 << (id + 6)));
+
+	return GL_TRUE;
+}
+
+GLboolean r700SendPSState(context_t *context)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	struct radeon_bo * pbo;
+	offset_modifiers offset_mod;
+	BATCH_LOCALS(&context->radeon);
+
+	pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context));
+
+	offset_mod.shift     = NO_SHIFT;
+	offset_mod.shiftbits = 0;
+	offset_mod.mask      = 0xFFFFFFFF;
+
+	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
+
+        BEGIN_BATCH_NO_AUTOSTATE(3);
+	R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1);
+	R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All,
+			     pbo,
+			     r700->ps.SQ_PGM_START_PS.u32All,
+			     RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod);
+	END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(9);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+	return GL_TRUE;
+}
+
+GLboolean r700SendVSState(context_t *context)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	struct radeon_bo * pbo;
+	offset_modifiers offset_mod;
+	BATCH_LOCALS(&context->radeon);
+
+	pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context));
+
+	offset_mod.shift     = NO_SHIFT;
+	offset_mod.shiftbits = 0;
+	offset_mod.mask      = 0xFFFFFFFF;
+
+	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
+
+        BEGIN_BATCH_NO_AUTOSTATE(3);
+	R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1);
+	R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All,
+			     pbo,
+			     r700->vs.SQ_PGM_START_VS.u32All,
+			     RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod);
+	END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+	return GL_TRUE;
+}
+
+GLboolean r700SendFSState(context_t *context)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	struct radeon_bo * pbo;
+	offset_modifiers offset_mod;
+	BATCH_LOCALS(&context->radeon);
+
+	/* XXX fixme
+	 * R6xx chips require a FS be emitted, even if it's not used.
+	 * since we aren't using FS yet, just send the VS address to make
+	 * the kernel command checker happy
+	 */
+	pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context));
+	r700->fs.SQ_PGM_START_FS.u32All = r700->vs.SQ_PGM_START_VS.u32All;
+	r700->fs.SQ_PGM_RESOURCES_FS.u32All = 0;
+	r700->fs.SQ_PGM_CF_OFFSET_FS.u32All = 0;
+	/* XXX */
+
+	offset_mod.shift     = NO_SHIFT;
+	offset_mod.shiftbits = 0;
+	offset_mod.mask      = 0xFFFFFFFF;
+
+	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
+
+        BEGIN_BATCH_NO_AUTOSTATE(3);
+	R600_OUT_BATCH_REGSEQ(SQ_PGM_START_FS, 1);
+	R600_OUT_BATCH_RELOC(r700->fs.SQ_PGM_START_FS.u32All,
+			     pbo,
+			     r700->fs.SQ_PGM_START_FS.u32All,
+			     RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod);
+	END_BATCH();
+
+        BEGIN_BATCH_NO_AUTOSTATE(6);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_FS, r700->fs.SQ_PGM_RESOURCES_FS.u32All);
+	R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_FS, r700->fs.SQ_PGM_CF_OFFSET_FS.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+	return GL_TRUE;
+}
+
+GLboolean r700SendViewportState(context_t *context, int id)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	struct radeon_renderbuffer *rrb;
+	offset_modifiers offset_mod;
+	BATCH_LOCALS(&context->radeon);
+
+	if (id > R700_MAX_VIEWPORTS)
+		return GL_FALSE;
+
+	if (!r700->viewport[id].enabled)
+		return GL_FALSE;
+
+        BEGIN_BATCH_NO_AUTOSTATE(16);
+	R600_OUT_BATCH_REGSEQ(PA_SC_VPORT_SCISSOR_0_TL + (8 * id), 2);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All);
+	R600_OUT_BATCH_REGSEQ(PA_SC_VPORT_ZMIN_0 + (8 * id), 2);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_ZMIN_0.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_SC_VPORT_ZMAX_0.u32All);
+	R600_OUT_BATCH_REGSEQ(PA_CL_VPORT_XSCALE_0 + (24 * id), 6);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_XSCALE.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_XOFFSET.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_YSCALE.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_YOFFSET.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_ZSCALE.u32All);
+	R600_OUT_BATCH(r700->viewport[id].PA_CL_VPORT_ZOFFSET.u32All);
+        END_BATCH();
+
+	COMMIT_BATCH();
+
+	return GL_TRUE;
+}
+
+GLboolean r700SendSQConfig(context_t *context)
+{
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	BATCH_LOCALS(&context->radeon);
+
+        BEGIN_BATCH_NO_AUTOSTATE(8);
+	R600_OUT_BATCH_REGSEQ(SQ_CONFIG, 6);
+	R600_OUT_BATCH(r700->sq_config.SQ_CONFIG.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All);
+	R600_OUT_BATCH(r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All);
+        END_BATCH();
+	COMMIT_BATCH();
+
+	return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h
new file mode 100644
index 0000000000..04af4bc3b9
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_chip.h
@@ -0,0 +1,521 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_CHIP_H_
+#define _R700_CHIP_H_
+
+#include "r600_context.h"
+
+#include "r600_reg.h"
+#include "r600_reg_auto_r6xx.h"
+#include "r600_reg_r6xx.h"
+#include "r600_reg_r7xx.h"
+
+#include "r700_chipoffset.h"
+
+#define SETfield(x, val, shift, mask)  ( (x) = ((x) & ~(mask)) | ((val) << (shift)) ) /* u32All */
+#define CLEARfield(x, mask)            ( (x) &= ~(mask) )
+#define SETbit(x, bit)                 ( (x) |= (bit) )
+#define CLEARbit(x, bit)               ( (x) &= ~(bit) )
+
+#define R700_TEXTURE_NUMBERUNITS 16
+#define R700_MAX_RENDER_TARGETS  8
+#define R700_MAX_VIEWPORTS       16
+#define R700_MAX_SHADER_EXPORTS  32
+#define R700_MAX_UCP             6
+
+/* Enum not show in r600_*.h */
+
+#define FETCH_RESOURCE_STRIDE 7
+
+#define ASIC_CONFIG_BASE_INDEX    0x2000
+#define ASIC_CONTEXT_BASE_INDEX   0xA000
+#define ASIC_CTL_CONST_BASE_INDEX 0xF3FC
+
+
+enum 
+{
+    SQ_ABSOLUTE                              = 0x00000000,
+    SQ_RELATIVE                              = 0x00000001,
+};
+
+enum 
+{
+    SQ_ALU_SCL_210                           = 0x00000000,
+    SQ_ALU_SCL_122                           = 0x00000001,
+    SQ_ALU_SCL_212                           = 0x00000002,
+    SQ_ALU_SCL_221                           = 0x00000003,
+};
+
+enum 
+{
+    SQ_TEX_UNNORMALIZED                      = 0x00000000,
+    SQ_TEX_NORMALIZED                        = 0x00000001,
+};
+
+enum 
+{
+    SQ_CF_PIXEL_MRT0                         = 0x00000000,
+    SQ_CF_PIXEL_MRT1                         = 0x00000001,
+    SQ_CF_PIXEL_MRT2                         = 0x00000002,
+    SQ_CF_PIXEL_MRT3                         = 0x00000003,
+    SQ_CF_PIXEL_MRT4                         = 0x00000004,
+    SQ_CF_PIXEL_MRT5                         = 0x00000005,
+    SQ_CF_PIXEL_MRT6                         = 0x00000006,
+    SQ_CF_PIXEL_MRT7                         = 0x00000007,
+    SQ_CF_PIXEL_Z                            = 0x0000003d,
+};
+
+typedef enum ENUM_SQ_CF_ARRAY_BASE_POS {
+SQ_CF_POS_0                              = 0x0000003c,
+SQ_CF_POS_1                              = 0x0000003d,
+SQ_CF_POS_2                              = 0x0000003e,
+SQ_CF_POS_3                              = 0x0000003f,
+} ENUM_SQ_CF_ARRAY_BASE_POS;
+
+enum
+{
+    PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit = 23,
+};
+
+enum 
+{
+    TEX_XYFilter_Point                       = 0x00000000,
+    TEX_XYFilter_Linear                      = 0x00000001,
+    TEX_XYFilter_Cubic                       = 0x00000002,
+    TEX_XYFilter_Cleartype                   = 0x00000003,
+
+    TEX_MipFilter_None                       = 0x00000000,
+    TEX_MipFilter_Point                      = 0x00000001,
+    TEX_MipFilter_Linear                     = 0x00000002,
+};
+
+enum 
+{
+    SQ_EXPORT_WRITE                          = 0x00000000,
+    SQ_EXPORT_WRITE_IND                      = 0x00000001,
+    SQ_EXPORT_WRITE_ACK                      = 0x00000002,
+    SQ_EXPORT_WRITE_IND_ACK                  = 0x00000003,
+};
+
+/* --------------------------------- */
+
+enum
+{
+    R700_PM4_PACKET0_NOP = 0x00000000,
+    R700_PM4_PACKET1_NOP = 0x40000000,
+    R700_PM4_PACKET2_NOP = 0x80000000,
+    R700_PM4_PACKET3_NOP = 0xC0000000,
+};
+
+#define  PM4_OPCODE_SET_INDEX_TYPE      (R700_PM4_PACKET3_NOP | (IT_INDEX_TYPE << 8))
+
+#define  PM4_OPCODE_DRAW_INDEX_AUTO     (R700_PM4_PACKET3_NOP | (IT_DRAW_INDEX_AUTO << 8))
+#define  PM4_OPCODE_DRAW_INDEX_IMMD     (R700_PM4_PACKET3_NOP | (IT_DRAW_INDEX_IMMD << 8))
+#define  PM4_OPCODE_WAIT_REG_MEM        (R700_PM4_PACKET3_NOP | (IT_WAIT_REG_MEM << 8))
+#define  PM4_OPCODE_SET_CONTEXT_REG     (R700_PM4_PACKET3_NOP | (IT_SET_CONTEXT_REG << 8))
+#define  PM4_OPCODE_SET_CONFIG_REG      (R700_PM4_PACKET3_NOP | (IT_SET_CONFIG_REG << 8))
+#define  PM4_OPCODE_SET_ALU_CONST       (R700_PM4_PACKET3_NOP | (IT_SET_ALU_CONST << 8))
+#define  PM4_OPCODE_SET_RESOURCE        (R700_PM4_PACKET3_NOP | (IT_SET_RESOURCE << 8))
+#define  PM4_OPCODE_SET_SAMPLER         (R700_PM4_PACKET3_NOP | (IT_SET_SAMPLER << 8))
+#define  PM4_OPCODE_CONTEXT_CONTROL     (R700_PM4_PACKET3_NOP | (IT_CONTEXT_CONTROL << 8))
+
+union UINT_FLOAT 
+{
+    unsigned int u32All;
+    float	f32All;
+};
+
+#if 0
+typedef struct _TEXTURE_STATE_STRUCT
+{
+    union UINT_FLOAT     SQ_TEX_RESOURCE0;
+    union UINT_FLOAT     SQ_TEX_RESOURCE1;
+    union UINT_FLOAT     SQ_TEX_RESOURCE2;
+    union UINT_FLOAT     SQ_TEX_RESOURCE3;
+    union UINT_FLOAT     SQ_TEX_RESOURCE4;
+    union UINT_FLOAT     SQ_TEX_RESOURCE5;
+    union UINT_FLOAT     SQ_TEX_RESOURCE6;
+    GLboolean                         enabled;
+} TEXTURE_STATE_STRUCT;
+
+typedef struct _SAMPLER_STATE_STRUCT
+{
+    union UINT_FLOAT      SQ_TEX_SAMPLER0;
+    union UINT_FLOAT      SQ_TEX_SAMPLER1;
+    union UINT_FLOAT      SQ_TEX_SAMPLER2;
+    GLboolean                         enabled;
+} SAMPLER_STATE_STRUCT;
+
+typedef struct _R700_TEXTURE_STATES
+{
+    TEXTURE_STATE_STRUCT *textures[R700_TEXTURE_NUMBERUNITS];
+    SAMPLER_STATE_STRUCT *samplers[R700_TEXTURE_NUMBERUNITS];
+} R700_TEXTURE_STATES;
+#endif
+
+typedef struct _RENDER_TARGET_STATE_STRUCT
+{
+	union UINT_FLOAT            	CB_COLOR0_BASE;  /* 0xA010 */
+	union UINT_FLOAT            	CB_COLOR0_SIZE;  /* 0xA018 */
+	union UINT_FLOAT            	CB_COLOR0_VIEW;  /* 0xA020 */
+	union UINT_FLOAT            	CB_COLOR0_INFO;  /* 0xA028 */
+	union UINT_FLOAT            	CB_COLOR0_TILE;  /* 0xA030 */
+	union UINT_FLOAT            	CB_COLOR0_FRAG;  /* 0xA038 */
+	union UINT_FLOAT            	CB_COLOR0_MASK;  /* 0xA040 */
+	union UINT_FLOAT         	CB_BLEND0_CONTROL;  /* 0xA1E0 */
+	GLboolean                         enabled;
+} RENDER_TARGET_STATE_STRUCT;
+
+typedef struct _VIEWPORT_STATE_STRUCT
+{
+	union UINT_FLOAT  	PA_SC_VPORT_SCISSOR_0_TL;  /* 0xA094 */
+	union UINT_FLOAT  	PA_SC_VPORT_SCISSOR_0_BR;  /* 0xA095 */
+	union UINT_FLOAT        PA_SC_VPORT_ZMIN_0;        /* 0xA0B4 */
+	union UINT_FLOAT        PA_SC_VPORT_ZMAX_0;        /* 0xA0B5 */
+	union UINT_FLOAT        PA_CL_VPORT_XSCALE;        /* 0xA10F */
+	union UINT_FLOAT       	PA_CL_VPORT_XOFFSET;       /* 0xA110 */
+	union UINT_FLOAT        PA_CL_VPORT_YSCALE;        /* 0xA111 */
+	union UINT_FLOAT       	PA_CL_VPORT_YOFFSET;       /* 0xA112 */
+	union UINT_FLOAT        PA_CL_VPORT_ZSCALE;        /* 0xA113 */
+	union UINT_FLOAT       	PA_CL_VPORT_ZOFFSET;       /* 0xA114 */
+	GLboolean                         enabled;
+} VIEWPORT_STATE_STRUCT;
+
+typedef struct _UCP_STATE_STRUCT
+{
+	union UINT_FLOAT        PA_CL_UCP_0_X;
+	union UINT_FLOAT        PA_CL_UCP_0_Y;
+	union UINT_FLOAT        PA_CL_UCP_0_Z;
+	union UINT_FLOAT        PA_CL_UCP_0_W;
+	GLboolean                         enabled;
+} UCP_STATE_STRUCT;
+
+typedef struct _PS_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_PS           ;  /* 0xA210 */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_PS       ;  /* 0xA214 */
+	union UINT_FLOAT         	SQ_PGM_EXPORTS_PS         ;  /* 0xA215 */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_PS       ;  /* 0xA233 */
+} PS_STATE_STRUCT;
+
+typedef struct _VS_STATE_STRUCT
+{
+ 	union UINT_FLOAT           	SQ_PGM_START_VS           ;  /* 0xA216 */
+	union UINT_FLOAT  		SQ_PGM_RESOURCES_VS       ;  /* 0xA21A */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_VS       ;  /* 0xA234 */
+} VS_STATE_STRUCT;
+
+typedef struct _GS_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_GS           ;  /* 0xA21B */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_GS       ;  /* 0xA21F */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_GS       ;  /* 0xA235 */
+} GS_STATE_STRUCT;
+
+typedef struct _ES_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_ES           ;  /* 0xA220 */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_ES       ;  /* 0xA224 */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_ES       ;  /* 0xA236 */
+} ES_STATE_STRUCT;
+
+typedef struct _FS_STATE_STRUCT
+{
+	union UINT_FLOAT           	SQ_PGM_START_FS           ;  /* 0xA225 */
+	union UINT_FLOAT       	        SQ_PGM_RESOURCES_FS       ;  /* 0xA229 */
+	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_FS       ;  /* 0xA237 */
+} FS_STATE_STRUCT;
+
+typedef struct _SQ_CONFIG_STRUCT
+{
+	union UINT_FLOAT     	        SQ_CONFIG                 ;  /* 0x2300 */
+	union UINT_FLOAT     	        SQ_GPR_RESOURCE_MGMT_1    ;  /* 0x2301 */
+	union UINT_FLOAT     	        SQ_GPR_RESOURCE_MGMT_2    ;  /* 0x2302 */
+	union UINT_FLOAT     	        SQ_THREAD_RESOURCE_MGMT   ;  /* 0x2303 */
+	union UINT_FLOAT     	        SQ_STACK_RESOURCE_MGMT_1  ;  /* 0x2304 */
+	union UINT_FLOAT     	        SQ_STACK_RESOURCE_MGMT_2  ;  /* 0x2305 */
+} SQ_CONFIG_STRUCT;
+
+typedef struct ContextState
+{
+    unsigned int * puiValue;
+    unsigned int   unOffset;
+    struct ContextState * pNext;
+} ContextState;
+
+typedef struct _R700_CHIP_CONTEXT
+{
+	// misc
+	union UINT_FLOAT             	TA_CNTL_AUX               ;  /* 0x2542 */
+	union UINT_FLOAT             	VC_ENHANCE                ;  /* 0x25C5 */
+	union UINT_FLOAT             	SQ_DYN_GPR_CNTL_PS_FLUSH_REQ;  /* 0x2363 */
+	union UINT_FLOAT             	DB_DEBUG                  ;  /* 0x260C */
+	union UINT_FLOAT             	DB_WATERMARKS             ;  /* 0x260E */
+
+	// DB
+	union UINT_FLOAT             	DB_DEPTH_SIZE             ;  /* 0xA000 */
+	union UINT_FLOAT             	DB_DEPTH_VIEW             ;  /* 0xA001 */
+	union UINT_FLOAT             	DB_DEPTH_BASE             ;  /* 0xA003 */
+	union UINT_FLOAT             	DB_DEPTH_INFO             ;  /* 0xA004 */
+	union UINT_FLOAT                DB_HTILE_DATA_BASE        ;  /* 0xA005 */
+	union UINT_FLOAT          	DB_STENCIL_CLEAR          ;  /* 0xA00A */
+	union UINT_FLOAT            	DB_DEPTH_CLEAR            ;  /* 0xA00B */
+	union UINT_FLOAT            	DB_STENCILREFMASK         ;  /* 0xA10C */
+	union UINT_FLOAT            	DB_STENCILREFMASK_BF      ;  /* 0xA10D */
+	union UINT_FLOAT         	DB_RENDER_CONTROL         ;  /* 0xA343 */
+	union UINT_FLOAT        	DB_RENDER_OVERRIDE        ;  /* 0xA344 */
+	union UINT_FLOAT          	DB_HTILE_SURFACE          ;  /* 0xA349 */
+	union UINT_FLOAT          	DB_ALPHA_TO_MASK          ;  /* 0xA351 */
+	union UINT_FLOAT          	DB_DEPTH_CONTROL          ;  /* 0xA200 */
+	union UINT_FLOAT         	DB_SHADER_CONTROL         ;  /* 0xA203 */
+
+	// SC
+	union UINT_FLOAT   	        PA_SC_SCREEN_SCISSOR_TL   ;  /* 0xA00C */
+	union UINT_FLOAT   	        PA_SC_SCREEN_SCISSOR_BR   ;  /* 0xA00D */
+	union UINT_FLOAT       	        PA_SC_WINDOW_OFFSET       ;  /* 0xA080 */
+	union UINT_FLOAT   	        PA_SC_WINDOW_SCISSOR_TL   ;  /* 0xA081 */
+	union UINT_FLOAT   	        PA_SC_WINDOW_SCISSOR_BR   ;  /* 0xA082 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_RULE       ;  /* 0xA083 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_0_TL       ;  /* 0xA084 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_0_BR       ;  /* 0xA085 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_1_TL       ;  /* 0xA086 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_1_BR       ;  /* 0xA087 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_2_TL       ;  /* 0xA088 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_2_BR       ;  /* 0xA089 */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_3_TL       ;  /* 0xA08A */
+	union UINT_FLOAT       	        PA_SC_CLIPRECT_3_BR       ;  /* 0xA08B */
+	union UINT_FLOAT            	PA_SC_EDGERULE            ;  /* 0xA08C */
+	union UINT_FLOAT  	        PA_SC_GENERIC_SCISSOR_TL  ;  /* 0xA090 */
+	union UINT_FLOAT  	        PA_SC_GENERIC_SCISSOR_BR  ;  /* 0xA091 */
+	union UINT_FLOAT        	PA_SC_LINE_STIPPLE        ;  /* 0xA283 */
+	union UINT_FLOAT           	PA_SC_LINE_CNTL           ;  /* 0xA300 */
+	union UINT_FLOAT           	PA_SC_AA_CONFIG           ;  /* 0xA301 */
+	union UINT_FLOAT       	        PA_SC_MPASS_PS_CNTL       ;  /* 0xA292 */
+	union UINT_FLOAT           	PA_SC_MODE_CNTL           ;  /* 0xA293 */
+	union UINT_FLOAT 	        PA_SC_AA_SAMPLE_LOCS_MCTX ;  /* 0xA307 */
+	union UINT_FLOAT                PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX; /* 0xA308 */
+	union UINT_FLOAT             	PA_SC_AA_MASK             ;  /* 0xA312 */
+
+	// CL
+	union UINT_FLOAT           	PA_CL_CLIP_CNTL           ;  /* 0xA204 */
+	union UINT_FLOAT            	PA_CL_VTE_CNTL            ;  /* 0xA206 */
+	union UINT_FLOAT         	PA_CL_VS_OUT_CNTL         ;  /* 0xA207 */
+	union UINT_FLOAT         	PA_CL_NANINF_CNTL         ;  /* 0xA208 */
+	union UINT_FLOAT    	        PA_CL_GB_VERT_CLIP_ADJ    ;  /* 0xA303 */
+	union UINT_FLOAT    	        PA_CL_GB_VERT_DISC_ADJ    ;  /* 0xA304 */
+	union UINT_FLOAT    	        PA_CL_GB_HORZ_CLIP_ADJ    ;  /* 0xA305 */
+	union UINT_FLOAT    	        PA_CL_GB_HORZ_DISC_ADJ    ;  /* 0xA306 */
+
+	// SU
+	union UINT_FLOAT        	PA_SU_SC_MODE_CNTL        ;  /* 0xA205 */
+	union UINT_FLOAT          	PA_SU_POINT_SIZE          ;  /* 0xA280 */
+	union UINT_FLOAT        	PA_SU_POINT_MINMAX        ;  /* 0xA281 */
+	union UINT_FLOAT           	PA_SU_LINE_CNTL           ;  /* 0xA282 */
+	union UINT_FLOAT            	PA_SU_VTX_CNTL            ;  /* 0xA302 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_DB_FMT_CNTL;   /* 0xA37E */
+	union UINT_FLOAT   	        PA_SU_POLY_OFFSET_CLAMP   ;      /* 0xA37F */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_FRONT_SCALE;   /* 0xA380 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_FRONT_OFFSET; /* 0xA381 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_BACK_SCALE;    /* 0xA382 */
+	union UINT_FLOAT                PA_SU_POLY_OFFSET_BACK_OFFSET;   /* 0xA383 */
+
+	VIEWPORT_STATE_STRUCT           viewport[R700_MAX_VIEWPORTS];
+	UCP_STATE_STRUCT                ucp[R700_MAX_UCP];
+
+	// CB
+	union UINT_FLOAT              	CB_CLEAR_RED_R6XX         ;  /* 0xA048 */
+	union UINT_FLOAT            	CB_CLEAR_GREEN_R6XX       ;  /* 0xA049 */
+	union UINT_FLOAT             	CB_CLEAR_BLUE_R6XX        ;  /* 0xA04A */
+	union UINT_FLOAT            	CB_CLEAR_ALPHA_R6XX       ;  /* 0xA04B */
+	union UINT_FLOAT            	CB_TARGET_MASK            ;  /* 0xA08E */
+	union UINT_FLOAT            	CB_SHADER_MASK            ;  /* 0xA08F */
+	union UINT_FLOAT              	CB_BLEND_RED              ;  /* 0xA105 */
+	union UINT_FLOAT            	CB_BLEND_GREEN            ;  /* 0xA106 */
+	union UINT_FLOAT             	CB_BLEND_BLUE             ;  /* 0xA107 */
+	union UINT_FLOAT            	CB_BLEND_ALPHA            ;  /* 0xA108 */
+	union UINT_FLOAT              	CB_FOG_RED_R6XX           ;  /* 0xA109 */
+	union UINT_FLOAT            	CB_FOG_GREEN_R6XX         ;  /* 0xA10A */
+	union UINT_FLOAT             	CB_FOG_BLUE_R6XX          ;  /* 0xA10B */
+	union UINT_FLOAT         	CB_SHADER_CONTROL         ;  /* 0xA1E8 */
+	union UINT_FLOAT          	CB_COLOR_CONTROL          ;  /* 0xA202 */
+	union UINT_FLOAT         	CB_CLRCMP_CONTROL         ;  /* 0xA30C */
+	union UINT_FLOAT             	CB_CLRCMP_SRC             ;  /* 0xA30D */
+	union UINT_FLOAT             	CB_CLRCMP_DST             ;  /* 0xA30E */
+	union UINT_FLOAT             	CB_CLRCMP_MSK             ;  /* 0xA30F */
+	union UINT_FLOAT             	CB_BLEND_CONTROL          ;  /* 0xABD0 */
+	RENDER_TARGET_STATE_STRUCT      render_target[R700_MAX_RENDER_TARGETS];
+
+	// SX
+	union UINT_FLOAT                SX_MISC                   ;  /* 0xA0D4 */
+	union UINT_FLOAT     	        SX_ALPHA_TEST_CONTROL     ;  /* 0xA104 */
+
+	// VGT
+	union UINT_FLOAT          	VGT_MAX_VTX_INDX          ;  /* 0xA100 */
+	union UINT_FLOAT          	VGT_MIN_VTX_INDX          ;  /* 0xA101 */
+	union UINT_FLOAT           	VGT_INDX_OFFSET           ;  /* 0xA102 */
+	union UINT_FLOAT                VGT_MULTI_PRIM_IB_RESET_INDX;  /* 0xA103 */
+	union UINT_FLOAT      	        VGT_OUTPUT_PATH_CNTL      ;  /* 0xA284 */
+	union UINT_FLOAT      	        VGT_HOS_CNTL              ;  /* 0xA285 */
+	union UINT_FLOAT      	        VGT_HOS_MAX_TESS_LEVEL    ;  /* 0xA286 */
+	union UINT_FLOAT      	        VGT_HOS_MIN_TESS_LEVEL    ;  /* 0xA287 */
+	union UINT_FLOAT      	        VGT_HOS_REUSE_DEPTH       ;  /* 0xA288 */
+	union UINT_FLOAT      	        VGT_GROUP_PRIM_TYPE       ;  /* 0xA289 */
+	union UINT_FLOAT      	        VGT_GROUP_FIRST_DECR      ;  /* 0xA28A */
+	union UINT_FLOAT      	        VGT_GROUP_DECR            ;  /* 0xA28B */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_0_CNTL     ;  /* 0xA28C */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_1_CNTL     ;  /* 0xA28D */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_0_FMT_CNTL ;  /* 0xA28E */
+	union UINT_FLOAT      	        VGT_GROUP_VECT_1_FMT_CNTL ;  /* 0xA28F */
+	union UINT_FLOAT               	VGT_GS_MODE               ;  /* 0xA290 */
+	union UINT_FLOAT        	VGT_PRIMITIVEID_EN        ;  /* 0xA2A1 */
+	union UINT_FLOAT     	        VGT_DMA_NUM_INSTANCES     ;  /* 0xA2A2 */
+	union UINT_FLOAT	        VGT_MULTI_PRIM_IB_RESET_EN;  /* 0xA2A5 */
+	union UINT_FLOAT  	        VGT_INSTANCE_STEP_RATE_0  ;  /* 0xA2A8 */
+	union UINT_FLOAT  	        VGT_INSTANCE_STEP_RATE_1  ;  /* 0xA2A9 */
+	union UINT_FLOAT            	VGT_STRMOUT_EN            ;  /* 0xA2AC */
+	union UINT_FLOAT             	VGT_REUSE_OFF             ;  /* 0xA2AD */
+	union UINT_FLOAT             	VGT_VTX_CNT_EN            ;  /* 0xA2AE */
+	union UINT_FLOAT            	VGT_STRMOUT_BUFFER_EN     ;  /* 0xA2C8 */
+
+	// SPI
+	union UINT_FLOAT           	SPI_VS_OUT_ID_0           ;  /* 0xA185 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_1           ;  /* 0xA186 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_2           ;  /* 0xA187 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_3           ;  /* 0xA188 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_4           ;  /* 0xA189 */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_5           ;  /* 0xA18A */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_6           ;  /* 0xA18B */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_7           ;  /* 0xA18C */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_8           ;  /* 0xA18D */
+	union UINT_FLOAT           	SPI_VS_OUT_ID_9           ;  /* 0xA18E */
+	union UINT_FLOAT                SPI_VS_OUT_CONFIG         ;  /* 0xA1B1 */
+	union UINT_FLOAT       	        SPI_THREAD_GROUPING       ;  /* 0xA1B2 */
+	union UINT_FLOAT       	        SPI_PS_IN_CONTROL_0       ;  /* 0xA1B3 */
+	union UINT_FLOAT       	        SPI_PS_IN_CONTROL_1       ;  /* 0xA1B4 */
+	union UINT_FLOAT       	        SPI_INTERP_CONTROL_0      ;  /* 0xA1B5 */
+ 	union UINT_FLOAT               	SPI_INPUT_Z               ;  /* 0xA1B6 */
+	union UINT_FLOAT              	SPI_FOG_CNTL              ;  /* 0xA1B7 */
+	union UINT_FLOAT              	SPI_FOG_FUNC_SCALE        ;  /* 0xA1B8 */
+	union UINT_FLOAT              	SPI_FOG_FUNC_BIAS         ;  /* 0xA1B9 */
+
+    union UINT_FLOAT         	SQ_VTX_SEMANTIC_0         ;  /* 0xA0E0 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_1         ;  /* 0xA0E1 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_2         ;  /* 0xA0E2 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_3         ;  /* 0xA0E3 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_4         ;  /* 0xA0E4 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_5         ;  /* 0xA0E5 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_6         ;  /* 0xA0E6 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_7         ;  /* 0xA0E7 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_8         ;  /* 0xA0E8 */
+	union UINT_FLOAT         	SQ_VTX_SEMANTIC_9         ;  /* 0xA0E9 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_10        ;  /* 0xA0EA */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_11        ;  /* 0xA0EB */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_12        ;  /* 0xA0EC */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_13        ;  /* 0xA0ED */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_14        ;  /* 0xA0EE */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_15        ;  /* 0xA0EF */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_16        ;  /* 0xA0F0 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_17        ;  /* 0xA0F1 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_18        ;  /* 0xA0F2 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_19        ;  /* 0xA0F3 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_20        ;  /* 0xA0F4 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_21        ;  /* 0xA0F5 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_22        ;  /* 0xA0F6 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_23        ;  /* 0xA0F7 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_24        ;  /* 0xA0F8 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_25        ;  /* 0xA0F9 */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_26        ;  /* 0xA0FA */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_27        ;  /* 0xA0FB */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_28        ;  /* 0xA0FC */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_29        ;  /* 0xA0FD */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_30        ;  /* 0xA0FE */
+	union UINT_FLOAT        	SQ_VTX_SEMANTIC_31        ;  /* 0xA0FF */
+
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_0       ;  /* 0xA191 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_1       ;  /* 0xA192 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_2       ;  /* 0xA193 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_3       ;  /* 0xA194 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_4       ;  /* 0xA195 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_5       ;  /* 0xA196 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_6       ;  /* 0xA197 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_7       ;  /* 0xA198 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_8       ;  /* 0xA199 */
+	union UINT_FLOAT       	SPI_PS_INPUT_CNTL_9       ;  /* 0xA19A */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_10      ;  /* 0xA19B */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_11      ;  /* 0xA19C */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_12      ;  /* 0xA19D */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_13      ;  /* 0xA19E */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_14      ;  /* 0xA19F */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_15      ;  /* 0xA1A0 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_16      ;  /* 0xA1A1 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_17      ;  /* 0xA1A2 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_18      ;  /* 0xA1A3 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_19      ;  /* 0xA1A4 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_20      ;  /* 0xA1A5 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_21      ;  /* 0xA1A6 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_22      ;  /* 0xA1A7 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_23      ;  /* 0xA1A8 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_24      ;  /* 0xA1A9 */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_25      ;  /* 0xA1AA */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_26      ;  /* 0xA1AB */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_27      ;  /* 0xA1AC */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_28      ;  /* 0xA1AD */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_29      ;  /* 0xA1AE */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_30      ;  /* 0xA1AF */
+	union UINT_FLOAT      	SPI_PS_INPUT_CNTL_31      ;  /* 0xA1B0 */
+
+	// shaders
+	PS_STATE_STRUCT                 ps;
+	VS_STATE_STRUCT                 vs;
+	GS_STATE_STRUCT                 gs;
+	ES_STATE_STRUCT                 es;
+	FS_STATE_STRUCT                 fs;
+
+	// SQ CONFIG
+	SQ_CONFIG_STRUCT                sq_config;
+
+	// SQ
+	union UINT_FLOAT     	        SQ_ESGS_RING_ITEMSIZE     ;  /* 0xA22A */
+	union UINT_FLOAT     	        SQ_GSVS_RING_ITEMSIZE     ;  /* 0xA22B */
+	union UINT_FLOAT    	        SQ_ESTMP_RING_ITEMSIZE    ;  /* 0xA22C */
+	union UINT_FLOAT    	        SQ_GSTMP_RING_ITEMSIZE    ;  /* 0xA22D */
+	union UINT_FLOAT    	        SQ_VSTMP_RING_ITEMSIZE    ;  /* 0xA22E */
+	union UINT_FLOAT    	        SQ_PSTMP_RING_ITEMSIZE    ;  /* 0xA22F */
+	union UINT_FLOAT     	        SQ_FBUF_RING_ITEMSIZE     ;  /* 0xA230 */
+	union UINT_FLOAT    	        SQ_REDUC_RING_ITEMSIZE    ;  /* 0xA231 */
+	union UINT_FLOAT       	        SQ_GS_VERT_ITEMSIZE       ;  /* 0xA232 */
+
+	ContextState*                   pStateList;
+
+	radeonTexObj*                   textures[R700_TEXTURE_NUMBERUNITS];
+
+	GLboolean                       bEnablePerspective;
+
+} R700_CHIP_CONTEXT;
+
+#endif /* _R700_CHIP_H_ */
+
diff --git a/src/mesa/drivers/dri/r600/r700_chipoffset.h b/src/mesa/drivers/dri/r600/r700_chipoffset.h
new file mode 100644
index 0000000000..4d73fb99a7
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_chipoffset.h
@@ -0,0 +1,693 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef _R700_CHIPOFFSET_H_
+#define _R700_CHIPOFFSET_H_
+
+#define mmWAIT_UNTIL                                    0x2010
+#define mmSCRATCH_REG0                                  0x2140
+#define mmGUI_SCRATCH_REG0                              0x2140
+#define mmSCRATCH_REG1                                  0x2141
+#define mmGUI_SCRATCH_REG1                              0x2141
+#define mmSCRATCH_REG2                                  0x2142
+#define mmGUI_SCRATCH_REG2                              0x2142
+#define mmSCRATCH_REG3                                  0x2143
+#define mmGUI_SCRATCH_REG3                              0x2143
+#define mmSCRATCH_REG4                                  0x2144
+#define mmGUI_SCRATCH_REG4                              0x2144
+#define mmSCRATCH_REG5                                  0x2145
+#define mmGUI_SCRATCH_REG5                              0x2145
+#define mmSCRATCH_REG6                                  0x2146
+#define mmGUI_SCRATCH_REG6                              0x2146
+#define mmSCRATCH_REG7                                  0x2147
+#define mmGUI_SCRATCH_REG7                              0x2147
+
+#define mmCP_COHER_CNTL                                 0x217C
+#define mmCP_COHER_SIZE                                 0x217D
+#define mmCP_COHER_BASE                                 0x217E
+#define mmCP_COHER_STATUS                               0x217F
+
+#define mmTA_CNTL_AUX                                   0x2542
+#define mmVC_ENHANCE                                    0x25C5
+#define mmSQ_DYN_GPR_CNTL_PS_FLUSH_REQ                  0x2363
+#define mmDB_DEBUG                                      0x260C
+#define mmDB_WATERMARKS                                 0x260E
+
+#define mmPA_CL_VPORT_XSCALE                            0xA10F
+#define mmPA_CL_VPORT_XOFFSET                           0xA110
+#define mmPA_CL_VPORT_YSCALE                            0xA111
+#define mmPA_CL_VPORT_YOFFSET                           0xA112
+#define mmPA_CL_VPORT_ZSCALE                            0xA113
+#define mmPA_CL_VPORT_ZOFFSET                           0xA114
+#define mmPA_CL_VPORT_XSCALE_1                          0xA115
+#define mmPA_CL_VPORT_XSCALE_2                          0xA11B
+#define mmPA_CL_VPORT_XSCALE_3                          0xA121
+#define mmPA_CL_VPORT_XSCALE_4                          0xA127
+#define mmPA_CL_VPORT_XSCALE_5                          0xA12D
+#define mmPA_CL_VPORT_XSCALE_6                          0xA133
+#define mmPA_CL_VPORT_XSCALE_7                          0xA139
+#define mmPA_CL_VPORT_XSCALE_8                          0xA13F
+#define mmPA_CL_VPORT_XSCALE_9                          0xA145
+#define mmPA_CL_VPORT_XSCALE_10                         0xA14B
+#define mmPA_CL_VPORT_XSCALE_11                         0xA151
+#define mmPA_CL_VPORT_XSCALE_12                         0xA157
+#define mmPA_CL_VPORT_XSCALE_13                         0xA15D
+#define mmPA_CL_VPORT_XSCALE_14                         0xA163
+#define mmPA_CL_VPORT_XSCALE_15                         0xA169
+#define mmPA_CL_VPORT_XOFFSET_1                         0xA116
+#define mmPA_CL_VPORT_XOFFSET_2                         0xA11C
+#define mmPA_CL_VPORT_XOFFSET_3                         0xA122
+#define mmPA_CL_VPORT_XOFFSET_4                         0xA128
+#define mmPA_CL_VPORT_XOFFSET_5                         0xA12E
+#define mmPA_CL_VPORT_XOFFSET_6                         0xA134
+#define mmPA_CL_VPORT_XOFFSET_7                         0xA13A
+#define mmPA_CL_VPORT_XOFFSET_8                         0xA140
+#define mmPA_CL_VPORT_XOFFSET_9                         0xA146
+#define mmPA_CL_VPORT_XOFFSET_10                        0xA14C
+#define mmPA_CL_VPORT_XOFFSET_11                        0xA152
+#define mmPA_CL_VPORT_XOFFSET_12                        0xA158
+#define mmPA_CL_VPORT_XOFFSET_13                        0xA15E
+#define mmPA_CL_VPORT_XOFFSET_14                        0xA164
+#define mmPA_CL_VPORT_XOFFSET_15                        0xA16A
+#define mmPA_CL_VPORT_YSCALE_1                          0xA117
+#define mmPA_CL_VPORT_YSCALE_2                          0xA11D
+#define mmPA_CL_VPORT_YSCALE_3                          0xA123
+#define mmPA_CL_VPORT_YSCALE_4                          0xA129
+#define mmPA_CL_VPORT_YSCALE_5                          0xA12F
+#define mmPA_CL_VPORT_YSCALE_6                          0xA135
+#define mmPA_CL_VPORT_YSCALE_7                          0xA13B
+#define mmPA_CL_VPORT_YSCALE_8                          0xA141
+#define mmPA_CL_VPORT_YSCALE_9                          0xA147
+#define mmPA_CL_VPORT_YSCALE_10                         0xA14D
+#define mmPA_CL_VPORT_YSCALE_11                         0xA153
+#define mmPA_CL_VPORT_YSCALE_12                         0xA159
+#define mmPA_CL_VPORT_YSCALE_13                         0xA15F
+#define mmPA_CL_VPORT_YSCALE_14                         0xA165
+#define mmPA_CL_VPORT_YSCALE_15                         0xA16B
+#define mmPA_CL_VPORT_YOFFSET_1                         0xA118
+#define mmPA_CL_VPORT_YOFFSET_2                         0xA11E
+#define mmPA_CL_VPORT_YOFFSET_3                         0xA124
+#define mmPA_CL_VPORT_YOFFSET_4                         0xA12A
+#define mmPA_CL_VPORT_YOFFSET_5                         0xA130
+#define mmPA_CL_VPORT_YOFFSET_6                         0xA136
+#define mmPA_CL_VPORT_YOFFSET_7                         0xA13C
+#define mmPA_CL_VPORT_YOFFSET_8                         0xA142
+#define mmPA_CL_VPORT_YOFFSET_9                         0xA148
+#define mmPA_CL_VPORT_YOFFSET_10                        0xA14E
+#define mmPA_CL_VPORT_YOFFSET_11                        0xA154
+#define mmPA_CL_VPORT_YOFFSET_12                        0xA15A
+#define mmPA_CL_VPORT_YOFFSET_13                        0xA160
+#define mmPA_CL_VPORT_YOFFSET_14                        0xA166
+#define mmPA_CL_VPORT_YOFFSET_15                        0xA16C
+#define mmPA_CL_VPORT_ZSCALE_1                          0xA119
+#define mmPA_CL_VPORT_ZSCALE_2                          0xA11F
+#define mmPA_CL_VPORT_ZSCALE_3                          0xA125
+#define mmPA_CL_VPORT_ZSCALE_4                          0xA12B
+#define mmPA_CL_VPORT_ZSCALE_5                          0xA131
+#define mmPA_CL_VPORT_ZSCALE_6                          0xA137
+#define mmPA_CL_VPORT_ZSCALE_7                          0xA13D
+#define mmPA_CL_VPORT_ZSCALE_8                          0xA143
+#define mmPA_CL_VPORT_ZSCALE_9                          0xA149
+#define mmPA_CL_VPORT_ZSCALE_10                         0xA14F
+#define mmPA_CL_VPORT_ZSCALE_11                         0xA155
+#define mmPA_CL_VPORT_ZSCALE_12                         0xA15B
+#define mmPA_CL_VPORT_ZSCALE_13                         0xA161
+#define mmPA_CL_VPORT_ZSCALE_14                         0xA167
+#define mmPA_CL_VPORT_ZSCALE_15                         0xA16D
+#define mmPA_CL_VPORT_ZOFFSET_1                         0xA11A
+#define mmPA_CL_VPORT_ZOFFSET_2                         0xA120
+#define mmPA_CL_VPORT_ZOFFSET_3                         0xA126
+#define mmPA_CL_VPORT_ZOFFSET_4                         0xA12C
+#define mmPA_CL_VPORT_ZOFFSET_5                         0xA132
+#define mmPA_CL_VPORT_ZOFFSET_6                         0xA138
+#define mmPA_CL_VPORT_ZOFFSET_7                         0xA13E
+#define mmPA_CL_VPORT_ZOFFSET_8                         0xA144
+#define mmPA_CL_VPORT_ZOFFSET_9                         0xA14A
+#define mmPA_CL_VPORT_ZOFFSET_10                        0xA150
+#define mmPA_CL_VPORT_ZOFFSET_11                        0xA156
+#define mmPA_CL_VPORT_ZOFFSET_12                        0xA15C
+#define mmPA_CL_VPORT_ZOFFSET_13                        0xA162
+#define mmPA_CL_VPORT_ZOFFSET_14                        0xA168
+#define mmPA_CL_VPORT_ZOFFSET_15                        0xA16E
+#define mmPA_CL_VTE_CNTL                                0xA206
+#define mmPA_CL_VS_OUT_CNTL                             0xA207
+#define mmPA_CL_NANINF_CNTL                             0xA208
+#define mmPA_CL_CLIP_CNTL                               0xA204
+#define mmPA_CL_GB_VERT_CLIP_ADJ                        0xA303
+#define mmPA_CL_GB_VERT_DISC_ADJ                        0xA304
+#define mmPA_CL_GB_HORZ_CLIP_ADJ                        0xA305
+#define mmPA_CL_GB_HORZ_DISC_ADJ                        0xA306
+#define mmPA_CL_UCP_0_X                                 0xA388
+#define mmPA_CL_UCP_0_Y                                 0xA389
+#define mmPA_CL_UCP_0_Z                                 0xA38A
+#define mmPA_CL_UCP_0_W                                 0xA38B
+#define mmPA_CL_UCP_1_X                                 0xA38C
+#define mmPA_CL_UCP_1_Y                                 0xA38D
+#define mmPA_CL_UCP_1_Z                                 0xA38E
+#define mmPA_CL_UCP_1_W                                 0xA38F
+#define mmPA_CL_UCP_2_X                                 0xA390
+#define mmPA_CL_UCP_2_Y                                 0xA391
+#define mmPA_CL_UCP_2_Z                                 0xA392
+#define mmPA_CL_UCP_2_W                                 0xA393
+#define mmPA_CL_UCP_3_X                                 0xA394
+#define mmPA_CL_UCP_3_Y                                 0xA395
+#define mmPA_CL_UCP_3_Z                                 0xA396
+#define mmPA_CL_UCP_3_W                                 0xA397
+#define mmPA_CL_UCP_4_X                                 0xA398
+#define mmPA_CL_UCP_4_Y                                 0xA399
+#define mmPA_CL_UCP_4_Z                                 0xA39A
+#define mmPA_CL_UCP_4_W                                 0xA39B
+#define mmPA_CL_UCP_5_X                                 0xA39C
+#define mmPA_CL_UCP_5_Y                                 0xA39D
+#define mmPA_CL_UCP_5_Z                                 0xA39E
+#define mmPA_CL_UCP_5_W                                 0xA39F
+#define mmPA_CL_POINT_X_RAD                             0xA384
+#define mmPA_CL_POINT_Y_RAD                             0xA385
+#define mmPA_CL_POINT_SIZE                              0xA386
+#define mmPA_CL_POINT_CULL_RAD                          0xA387
+
+#define mmPA_SU_VTX_CNTL                                0xA302
+#define mmPA_SU_POINT_SIZE                              0xA280
+#define mmPA_SU_POINT_MINMAX                            0xA281
+#define mmPA_SU_LINE_CNTL                               0xA282
+#define mmPA_SU_SC_MODE_CNTL                            0xA205
+#define mmPA_SU_POLY_OFFSET_DB_FMT_CNTL                 0xA37E
+#define mmPA_SU_POLY_OFFSET_CLAMP                       0xA37F
+#define mmPA_SU_POLY_OFFSET_FRONT_SCALE                 0xA380
+#define mmPA_SU_POLY_OFFSET_FRONT_OFFSET                0xA381
+#define mmPA_SU_POLY_OFFSET_BACK_SCALE                  0xA382
+#define mmPA_SU_POLY_OFFSET_BACK_OFFSET                 0xA383
+
+#define mmPA_SC_WINDOW_OFFSET                           0xA080
+#define mmPA_SC_AA_CONFIG                               0xA301
+#define mmPA_SC_AA_MASK                                 0xA312
+#define mmPA_SC_AA_SAMPLE_LOCS_MCTX                     0xA307
+#define mmPA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX              0xA308
+#define mmPA_SC_LINE_STIPPLE                            0xA283
+#define mmPA_SC_LINE_CNTL                               0xA300
+#define mmPA_SC_SCREEN_SCISSOR_TL                       0xA00C
+#define mmPA_SC_SCREEN_SCISSOR_BR                       0xA00D
+#define mmPA_SC_WINDOW_SCISSOR_TL                       0xA081
+#define mmPA_SC_WINDOW_SCISSOR_BR                       0xA082
+#define mmPA_SC_CLIPRECT_RULE                           0xA083
+#define mmPA_SC_CLIPRECT_0_TL                           0xA084
+#define mmPA_SC_CLIPRECT_0_BR                           0xA085
+#define mmPA_SC_CLIPRECT_1_TL                           0xA086
+#define mmPA_SC_CLIPRECT_1_BR                           0xA087
+#define mmPA_SC_CLIPRECT_2_TL                           0xA088
+#define mmPA_SC_CLIPRECT_2_BR                           0xA089
+#define mmPA_SC_CLIPRECT_3_TL                           0xA08A
+#define mmPA_SC_CLIPRECT_3_BR                           0xA08B
+#define mmPA_SC_EDGERULE                                0xA08C
+#define mmPA_SC_GENERIC_SCISSOR_TL                      0xA090
+#define mmPA_SC_GENERIC_SCISSOR_BR                      0xA091
+#define mmPA_SC_VPORT_SCISSOR_0_TL                      0xA094
+#define mmPA_SC_VPORT_SCISSOR_1_TL                      0xA096
+#define mmPA_SC_VPORT_SCISSOR_2_TL                      0xA098
+#define mmPA_SC_VPORT_SCISSOR_3_TL                      0xA09A
+#define mmPA_SC_VPORT_SCISSOR_4_TL                      0xA09C
+#define mmPA_SC_VPORT_SCISSOR_5_TL                      0xA09E
+#define mmPA_SC_VPORT_SCISSOR_6_TL                      0xA0A0
+#define mmPA_SC_VPORT_SCISSOR_7_TL                      0xA0A2
+#define mmPA_SC_VPORT_SCISSOR_8_TL                      0xA0A4
+#define mmPA_SC_VPORT_SCISSOR_9_TL                      0xA0A6
+#define mmPA_SC_VPORT_SCISSOR_10_TL                     0xA0A8
+#define mmPA_SC_VPORT_SCISSOR_11_TL                     0xA0AA
+#define mmPA_SC_VPORT_SCISSOR_12_TL                     0xA0AC
+#define mmPA_SC_VPORT_SCISSOR_13_TL                     0xA0AE
+#define mmPA_SC_VPORT_SCISSOR_14_TL                     0xA0B0
+#define mmPA_SC_VPORT_SCISSOR_15_TL                     0xA0B2
+#define mmPA_SC_VPORT_SCISSOR_0_BR                      0xA095
+#define mmPA_SC_VPORT_SCISSOR_1_BR                      0xA097
+#define mmPA_SC_VPORT_SCISSOR_2_BR                      0xA099
+#define mmPA_SC_VPORT_SCISSOR_3_BR                      0xA09B
+#define mmPA_SC_VPORT_SCISSOR_4_BR                      0xA09D
+#define mmPA_SC_VPORT_SCISSOR_5_BR                      0xA09F
+#define mmPA_SC_VPORT_SCISSOR_6_BR                      0xA0A1
+#define mmPA_SC_VPORT_SCISSOR_7_BR                      0xA0A3
+#define mmPA_SC_VPORT_SCISSOR_8_BR                      0xA0A5
+#define mmPA_SC_VPORT_SCISSOR_9_BR                      0xA0A7
+#define mmPA_SC_VPORT_SCISSOR_10_BR                     0xA0A9
+#define mmPA_SC_VPORT_SCISSOR_11_BR                     0xA0AB
+#define mmPA_SC_VPORT_SCISSOR_12_BR                     0xA0AD
+#define mmPA_SC_VPORT_SCISSOR_13_BR                     0xA0AF
+#define mmPA_SC_VPORT_SCISSOR_14_BR                     0xA0B1
+#define mmPA_SC_VPORT_SCISSOR_15_BR                     0xA0B3
+#define mmPA_SC_VPORT_ZMIN_0                            0xA0B4
+#define mmPA_SC_VPORT_ZMIN_1                            0xA0B6
+#define mmPA_SC_VPORT_ZMIN_2                            0xA0B8
+#define mmPA_SC_VPORT_ZMIN_3                            0xA0BA
+#define mmPA_SC_VPORT_ZMIN_4                            0xA0BC
+#define mmPA_SC_VPORT_ZMIN_5                            0xA0BE
+#define mmPA_SC_VPORT_ZMIN_6                            0xA0C0
+#define mmPA_SC_VPORT_ZMIN_7                            0xA0C2
+#define mmPA_SC_VPORT_ZMIN_8                            0xA0C4
+#define mmPA_SC_VPORT_ZMIN_9                            0xA0C6
+#define mmPA_SC_VPORT_ZMIN_10                           0xA0C8
+#define mmPA_SC_VPORT_ZMIN_11                           0xA0CA
+#define mmPA_SC_VPORT_ZMIN_12                           0xA0CC
+#define mmPA_SC_VPORT_ZMIN_13                           0xA0CE
+#define mmPA_SC_VPORT_ZMIN_14                           0xA0D0
+#define mmPA_SC_VPORT_ZMIN_15                           0xA0D2
+#define mmPA_SC_VPORT_ZMAX_0                            0xA0B5
+#define mmPA_SC_VPORT_ZMAX_1                            0xA0B7
+#define mmPA_SC_VPORT_ZMAX_2                            0xA0B9
+#define mmPA_SC_VPORT_ZMAX_3                            0xA0BB
+#define mmPA_SC_VPORT_ZMAX_4                            0xA0BD
+#define mmPA_SC_VPORT_ZMAX_5                            0xA0BF
+#define mmPA_SC_VPORT_ZMAX_6                            0xA0C1
+#define mmPA_SC_VPORT_ZMAX_7                            0xA0C3
+#define mmPA_SC_VPORT_ZMAX_8                            0xA0C5
+#define mmPA_SC_VPORT_ZMAX_9                            0xA0C7
+#define mmPA_SC_VPORT_ZMAX_10                           0xA0C9
+#define mmPA_SC_VPORT_ZMAX_11                           0xA0CB
+#define mmPA_SC_VPORT_ZMAX_12                           0xA0CD
+#define mmPA_SC_VPORT_ZMAX_13                           0xA0CF
+#define mmPA_SC_VPORT_ZMAX_14                           0xA0D1
+#define mmPA_SC_VPORT_ZMAX_15                           0xA0D3
+#define mmPA_SC_MODE_CNTL                               0xA293
+#define mmPA_SC_MPASS_PS_CNTL                           0xA292
+
+#define mmVGT_DRAW_INITIATOR                            0xA1FC
+#define mmVGT_EVENT_INITIATOR                           0xA2A4
+#define mmVGT_EVENT_ADDRESS_REG                         0xA1FE
+#define mmVGT_DMA_BASE_HI                               0xA1F9
+#define mmVGT_DMA_BASE                                  0xA1FA
+#define mmVGT_DMA_INDEX_TYPE                            0xA29F
+#define mmVGT_DMA_NUM_INSTANCES                         0xA2A2
+#define mmVGT_DMA_SIZE                                  0xA29D
+
+#define mmVGT_IMMED_DATA                                0xA1FD
+#define mmVGT_INDEX_TYPE                                0x2257
+#define mmVGT_NUM_INDICES                               0x225C
+#define mmVGT_NUM_INSTANCES                             0x225D
+#define mmVGT_PRIMITIVE_TYPE                            0x2256
+#define mmVGT_PRIMITIVEID_EN                            0xA2A1
+#define mmVGT_VTX_CNT_EN                                0xA2AE
+#define mmVGT_REUSE_OFF                                 0xA2AD
+#define mmVGT_INSTANCE_STEP_RATE_0                      0xA2A8
+#define mmVGT_INSTANCE_STEP_RATE_1                      0xA2A9
+#define mmVGT_MAX_VTX_INDX                              0xA100
+#define mmVGT_MIN_VTX_INDX                              0xA101
+#define mmVGT_INDX_OFFSET                               0xA102
+#define mmVGT_VERTEX_REUSE_BLOCK_CNTL                   0xA316
+#define mmVGT_OUT_DEALLOC_CNTL                          0xA317
+#define mmVGT_MULTI_PRIM_IB_RESET_INDX                  0xA103
+#define mmVGT_MULTI_PRIM_IB_RESET_EN                    0xA2A5
+#define mmVGT_ENHANCE                                   0xA294
+#define mmVGT_OUTPUT_PATH_CNTL                          0xA284
+#define mmVGT_HOS_CNTL                                  0xA285
+#define mmVGT_HOS_MAX_TESS_LEVEL                        0xA286
+#define mmVGT_HOS_MIN_TESS_LEVEL                        0xA287
+#define mmVGT_HOS_REUSE_DEPTH                           0xA288
+#define mmVGT_GROUP_PRIM_TYPE                           0xA289
+#define mmVGT_GROUP_FIRST_DECR                          0xA28A
+#define mmVGT_GROUP_DECR                                0xA28B
+#define mmVGT_GROUP_VECT_0_CNTL                         0xA28C
+#define mmVGT_GROUP_VECT_1_CNTL                         0xA28D
+#define mmVGT_GROUP_VECT_0_FMT_CNTL                     0xA28E
+#define mmVGT_GROUP_VECT_1_FMT_CNTL                     0xA28F
+#define mmVGT_GS_MODE                                   0xA290
+#define mmVGT_GS_OUT_PRIM_TYPE                          0xA29B
+
+#define mmVGT_STRMOUT_EN                                0xA2AC
+#define mmVGT_STRMOUT_BUFFER_SIZE_0                     0xA2B4
+#define mmVGT_STRMOUT_BUFFER_SIZE_1                     0xA2B8
+#define mmVGT_STRMOUT_BUFFER_SIZE_2                     0xA2BC
+#define mmVGT_STRMOUT_BUFFER_SIZE_3                     0xA2C0
+#define mmVGT_STRMOUT_BUFFER_OFFSET_0                   0xA2B7
+#define mmVGT_STRMOUT_BUFFER_OFFSET_1                   0xA2BB
+#define mmVGT_STRMOUT_BUFFER_OFFSET_2                   0xA2BF
+#define mmVGT_STRMOUT_BUFFER_OFFSET_3                   0xA2C3
+#define mmVGT_STRMOUT_VTX_STRIDE_0                      0xA2B5
+#define mmVGT_STRMOUT_VTX_STRIDE_1                      0xA2B9
+#define mmVGT_STRMOUT_VTX_STRIDE_2                      0xA2BD
+#define mmVGT_STRMOUT_VTX_STRIDE_3                      0xA2C1
+#define mmVGT_STRMOUT_BUFFER_BASE_0                     0xA2B6
+#define mmVGT_STRMOUT_BUFFER_BASE_1                     0xA2BA
+#define mmVGT_STRMOUT_BUFFER_BASE_2                     0xA2BE
+#define mmVGT_STRMOUT_BUFFER_BASE_3                     0xA2C2
+#define mmVGT_STRMOUT_BUFFER_EN                         0xA2C8
+#define mmVGT_STRMOUT_BASE_OFFSET_0                     0xA2C4
+#define mmVGT_STRMOUT_BASE_OFFSET_1                     0xA2C5
+#define mmVGT_STRMOUT_BASE_OFFSET_2                     0xA2C6
+#define mmVGT_STRMOUT_BASE_OFFSET_3                     0xA2C7
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_0                  0xA2D1
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_1                  0xA2D2
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_2                  0xA2D3
+#define mmVGT_STRMOUT_BASE_OFFSET_HI_3                  0xA2D4
+#define mmVGT_STRMOUT_DRAW_OPAQUE_OFFSET                0xA2CA
+#define mmVGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE    0xA2CB
+#define mmVGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE         0xA2CC
+
+#define mmSQ_PGM_START_PS                               0xA210
+#define mmSQ_PGM_CF_OFFSET_PS                           0xA233
+#define mmSQ_PGM_RESOURCES_PS                           0xA214
+#define mmSQ_PGM_EXPORTS_PS                             0xA215
+#define mmSQ_PGM_START_VS                               0xA216
+#define mmSQ_PGM_CF_OFFSET_VS                           0xA234
+#define mmSQ_PGM_RESOURCES_VS                           0xA21A
+#define mmSQ_PGM_START_GS                               0xA21B
+#define mmSQ_PGM_CF_OFFSET_GS                           0xA235
+#define mmSQ_PGM_RESOURCES_GS                           0xA21F
+#define mmSQ_PGM_START_ES                               0xA220
+#define mmSQ_PGM_CF_OFFSET_ES                           0xA236
+#define mmSQ_PGM_RESOURCES_ES                           0xA224
+#define mmSQ_PGM_START_FS                               0xA225
+#define mmSQ_PGM_CF_OFFSET_FS                           0xA237
+#define mmSQ_PGM_RESOURCES_FS                           0xA229
+#define mmSQ_ESGS_RING_ITEMSIZE                         0xA22A
+#define mmSQ_GSVS_RING_ITEMSIZE                         0xA22B
+#define mmSQ_ESTMP_RING_ITEMSIZE                        0xA22C
+#define mmSQ_GSTMP_RING_ITEMSIZE                        0xA22D
+#define mmSQ_VSTMP_RING_ITEMSIZE                        0xA22E
+#define mmSQ_PSTMP_RING_ITEMSIZE                        0xA22F
+#define mmSQ_FBUF_RING_ITEMSIZE                         0xA230
+#define mmSQ_REDUC_RING_ITEMSIZE                        0xA231
+#define mmSQ_GS_VERT_ITEMSIZE                           0xA232
+#define mmSQ_VTX_SEMANTIC_CLEAR                         0xA238
+
+#define mmSQ_VTX_SEMANTIC_0                             0xA0E0
+#define mmSQ_VTX_SEMANTIC_1                             0xA0E1
+#define mmSQ_VTX_SEMANTIC_2                             0xA0E2
+#define mmSQ_VTX_SEMANTIC_3                             0xA0E3
+#define mmSQ_VTX_SEMANTIC_4                             0xA0E4
+#define mmSQ_VTX_SEMANTIC_5                             0xA0E5
+#define mmSQ_VTX_SEMANTIC_6                             0xA0E6
+#define mmSQ_VTX_SEMANTIC_7                             0xA0E7
+#define mmSQ_VTX_SEMANTIC_8                             0xA0E8
+#define mmSQ_VTX_SEMANTIC_9                             0xA0E9
+#define mmSQ_VTX_SEMANTIC_10                            0xA0EA
+#define mmSQ_VTX_SEMANTIC_11                            0xA0EB
+#define mmSQ_VTX_SEMANTIC_12                            0xA0EC
+#define mmSQ_VTX_SEMANTIC_13                            0xA0ED
+#define mmSQ_VTX_SEMANTIC_14                            0xA0EE
+#define mmSQ_VTX_SEMANTIC_15                            0xA0EF
+#define mmSQ_VTX_SEMANTIC_16                            0xA0F0
+#define mmSQ_VTX_SEMANTIC_17                            0xA0F1
+#define mmSQ_VTX_SEMANTIC_18                            0xA0F2
+#define mmSQ_VTX_SEMANTIC_19                            0xA0F3
+#define mmSQ_VTX_SEMANTIC_20                            0xA0F4
+#define mmSQ_VTX_SEMANTIC_21                            0xA0F5
+#define mmSQ_VTX_SEMANTIC_22                            0xA0F6
+#define mmSQ_VTX_SEMANTIC_23                            0xA0F7
+#define mmSQ_VTX_SEMANTIC_24                            0xA0F8
+#define mmSQ_VTX_SEMANTIC_25                            0xA0F9
+#define mmSQ_VTX_SEMANTIC_26                            0xA0FA
+#define mmSQ_VTX_SEMANTIC_27                            0xA0FB
+#define mmSQ_VTX_SEMANTIC_28                            0xA0FC
+#define mmSQ_VTX_SEMANTIC_29                            0xA0FD
+#define mmSQ_VTX_SEMANTIC_30                            0xA0FE
+#define mmSQ_VTX_SEMANTIC_31                            0xA0FF
+
+#define mmSQ_ALU_CONST_CACHE_PS_0                       0xA250
+#define mmSQ_ALU_CONST_CACHE_PS_1                       0xA251
+#define mmSQ_ALU_CONST_CACHE_PS_2                       0xA252
+#define mmSQ_ALU_CONST_CACHE_PS_3                       0xA253
+#define mmSQ_ALU_CONST_CACHE_PS_4                       0xA254
+#define mmSQ_ALU_CONST_CACHE_PS_5                       0xA255
+#define mmSQ_ALU_CONST_CACHE_PS_6                       0xA256
+#define mmSQ_ALU_CONST_CACHE_PS_7                       0xA257
+#define mmSQ_ALU_CONST_CACHE_PS_8                       0xA258
+#define mmSQ_ALU_CONST_CACHE_PS_9                       0xA259
+#define mmSQ_ALU_CONST_CACHE_PS_10                      0xA25A
+#define mmSQ_ALU_CONST_CACHE_PS_11                      0xA25B
+#define mmSQ_ALU_CONST_CACHE_PS_12                      0xA25C
+#define mmSQ_ALU_CONST_CACHE_PS_13                      0xA25D
+#define mmSQ_ALU_CONST_CACHE_PS_14                      0xA25E
+#define mmSQ_ALU_CONST_CACHE_PS_15                      0xA25F
+#define mmSQ_ALU_CONST_CACHE_VS_0                       0xA260
+#define mmSQ_ALU_CONST_CACHE_VS_1                       0xA261
+#define mmSQ_ALU_CONST_CACHE_VS_2                       0xA262
+#define mmSQ_ALU_CONST_CACHE_VS_3                       0xA263
+#define mmSQ_ALU_CONST_CACHE_VS_4                       0xA264
+#define mmSQ_ALU_CONST_CACHE_VS_5                       0xA265
+#define mmSQ_ALU_CONST_CACHE_VS_6                       0xA266
+#define mmSQ_ALU_CONST_CACHE_VS_7                       0xA267
+#define mmSQ_ALU_CONST_CACHE_VS_8                       0xA268
+#define mmSQ_ALU_CONST_CACHE_VS_9                       0xA269
+#define mmSQ_ALU_CONST_CACHE_VS_10                      0xA26A
+#define mmSQ_ALU_CONST_CACHE_VS_11                      0xA26B
+#define mmSQ_ALU_CONST_CACHE_VS_12                      0xA26C
+#define mmSQ_ALU_CONST_CACHE_VS_13                      0xA26D
+#define mmSQ_ALU_CONST_CACHE_VS_14                      0xA26E
+#define mmSQ_ALU_CONST_CACHE_VS_15                      0xA26F
+#define mmSQ_ALU_CONST_CACHE_GS_0                       0xA270
+#define mmSQ_ALU_CONST_CACHE_GS_1                       0xA271
+#define mmSQ_ALU_CONST_CACHE_GS_2                       0xA272
+#define mmSQ_ALU_CONST_CACHE_GS_3                       0xA273
+#define mmSQ_ALU_CONST_CACHE_GS_4                       0xA274
+#define mmSQ_ALU_CONST_CACHE_GS_5                       0xA275
+#define mmSQ_ALU_CONST_CACHE_GS_6                       0xA276
+#define mmSQ_ALU_CONST_CACHE_GS_7                       0xA277
+#define mmSQ_ALU_CONST_CACHE_GS_8                       0xA278
+#define mmSQ_ALU_CONST_CACHE_GS_9                       0xA279
+#define mmSQ_ALU_CONST_CACHE_GS_10                      0xA27A
+#define mmSQ_ALU_CONST_CACHE_GS_11                      0xA27B
+#define mmSQ_ALU_CONST_CACHE_GS_12                      0xA27C
+#define mmSQ_ALU_CONST_CACHE_GS_13                      0xA27D
+#define mmSQ_ALU_CONST_CACHE_GS_14                      0xA27E
+#define mmSQ_ALU_CONST_CACHE_GS_15                      0xA27F
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_0                 0xA050
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_1                 0xA051
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_2                 0xA052
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_3                 0xA053
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_4                 0xA054
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_5                 0xA055
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_6                 0xA056
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_7                 0xA057
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_8                 0xA058
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_9                 0xA059
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_10                0xA05A
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_11                0xA05B
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_12                0xA05C
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_13                0xA05D
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_14                0xA05E
+#define mmSQ_ALU_CONST_BUFFER_SIZE_PS_15                0xA05F
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_0                 0xA060
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_1                 0xA061
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_2                 0xA062
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_3                 0xA063
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_4                 0xA064
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_5                 0xA065
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_6                 0xA066
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_7                 0xA067
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_8                 0xA068
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_9                 0xA069
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_10                0xA06A
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_11                0xA06B
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_12                0xA06C
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_13                0xA06D
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_14                0xA06E
+#define mmSQ_ALU_CONST_BUFFER_SIZE_VS_15                0xA06F
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_0                 0xA070
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_1                 0xA071
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_2                 0xA072
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_3                 0xA073
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_4                 0xA074
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_5                 0xA075
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_6                 0xA076
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_7                 0xA077
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_8                 0xA078
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_9                 0xA079
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_10                0xA07A
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_11                0xA07B
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_12                0xA07C
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_13                0xA07D
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_14                0xA07E
+#define mmSQ_ALU_CONST_BUFFER_SIZE_GS_15                0xA07F
+
+#define mmSPI_VS_OUT_ID_0                               0xA185
+#define mmSPI_VS_OUT_ID_1                               0xA186
+#define mmSPI_VS_OUT_ID_2                               0xA187
+#define mmSPI_VS_OUT_ID_3                               0xA188
+#define mmSPI_VS_OUT_ID_4                               0xA189
+#define mmSPI_VS_OUT_ID_5                               0xA18A
+#define mmSPI_VS_OUT_ID_6                               0xA18B
+#define mmSPI_VS_OUT_ID_7                               0xA18C
+#define mmSPI_VS_OUT_ID_8                               0xA18D
+#define mmSPI_VS_OUT_ID_9                               0xA18E
+#define mmSPI_PS_INPUT_CNTL_0                           0xA191
+#define mmSPI_PS_INPUT_CNTL_1                           0xA192
+#define mmSPI_PS_INPUT_CNTL_2                           0xA193
+#define mmSPI_PS_INPUT_CNTL_3                           0xA194
+#define mmSPI_PS_INPUT_CNTL_4                           0xA195
+#define mmSPI_PS_INPUT_CNTL_5                           0xA196
+#define mmSPI_PS_INPUT_CNTL_6                           0xA197
+#define mmSPI_PS_INPUT_CNTL_7                           0xA198
+#define mmSPI_PS_INPUT_CNTL_8                           0xA199
+#define mmSPI_PS_INPUT_CNTL_9                           0xA19A
+#define mmSPI_PS_INPUT_CNTL_10                          0xA19B
+#define mmSPI_PS_INPUT_CNTL_11                          0xA19C
+#define mmSPI_PS_INPUT_CNTL_12                          0xA19D
+#define mmSPI_PS_INPUT_CNTL_13                          0xA19E
+#define mmSPI_PS_INPUT_CNTL_14                          0xA19F
+#define mmSPI_PS_INPUT_CNTL_15                          0xA1A0
+#define mmSPI_PS_INPUT_CNTL_16                          0xA1A1
+#define mmSPI_PS_INPUT_CNTL_17                          0xA1A2
+#define mmSPI_PS_INPUT_CNTL_18                          0xA1A3
+#define mmSPI_PS_INPUT_CNTL_19                          0xA1A4
+#define mmSPI_PS_INPUT_CNTL_20                          0xA1A5
+#define mmSPI_PS_INPUT_CNTL_21                          0xA1A6
+#define mmSPI_PS_INPUT_CNTL_22                          0xA1A7
+#define mmSPI_PS_INPUT_CNTL_23                          0xA1A8
+#define mmSPI_PS_INPUT_CNTL_24                          0xA1A9
+#define mmSPI_PS_INPUT_CNTL_25                          0xA1AA
+#define mmSPI_PS_INPUT_CNTL_26                          0xA1AB
+#define mmSPI_PS_INPUT_CNTL_27                          0xA1AC
+#define mmSPI_PS_INPUT_CNTL_28                          0xA1AD
+#define mmSPI_PS_INPUT_CNTL_29                          0xA1AE
+#define mmSPI_PS_INPUT_CNTL_30                          0xA1AF
+#define mmSPI_PS_INPUT_CNTL_31                          0xA1B0
+#define mmSPI_VS_OUT_CONFIG                             0xA1B1
+#define mmSPI_THREAD_GROUPING                           0xA1B2
+#define mmSPI_PS_IN_CONTROL_0                           0xA1B3
+#define mmSPI_PS_IN_CONTROL_1                           0xA1B4
+#define mmSPI_INTERP_CONTROL_0                          0xA1B5
+#define mmSPI_INPUT_Z                                   0xA1B6
+#define mmSPI_FOG_CNTL                                  0xA1B7
+#define mmSPI_FOG_FUNC_SCALE                            0xA1B8
+#define mmSPI_FOG_FUNC_BIAS                             0xA1B9
+
+#define mmSX_MISC                                       0xA0D4
+#define mmSX_ALPHA_TEST_CONTROL                         0xA104
+#define mmSX_ALPHA_REF                                  0xA10E
+
+#define mmDB_DEPTH_BASE                                 0xA003
+#define mmDB_DEPTH_INFO                                 0xA004
+#define mmDB_HTILE_DATA_BASE                            0xA005
+#define mmDB_DEPTH_SIZE                                 0xA000
+#define mmDB_DEPTH_VIEW                                 0xA001
+#define mmDB_RENDER_CONTROL                             0xA343
+#define mmDB_RENDER_OVERRIDE                            0xA344
+#define mmDB_SHADER_CONTROL                             0xA203
+#define mmDB_STENCIL_CLEAR                              0xA00A
+#define mmDB_DEPTH_CLEAR                                0xA00B
+#define mmDB_HTILE_SURFACE                              0xA349
+#define mmDB_PRELOAD_CONTROL                            0xA34C
+#define mmDB_PREFETCH_LIMIT                             0xA34D
+#define mmDB_STENCILREFMASK                             0xA10C
+#define mmDB_STENCILREFMASK_BF                          0xA10D
+#define mmDB_SRESULTS_COMPARE_STATE0                    0xA34A
+#define mmDB_SRESULTS_COMPARE_STATE1                    0xA34B
+#define mmDB_DEPTH_CONTROL                              0xA200
+#define mmDB_ALPHA_TO_MASK                              0xA351
+
+#define mmCB_CLEAR_RED_R6XX                             0xA048
+#define mmCB_CLEAR_GREEN_R6XX                           0xA049
+#define mmCB_CLEAR_BLUE_R6XX                            0xA04A
+#define mmCB_CLEAR_ALPHA_R6XX                           0xA04B
+#define mmCB_BLEND_RED                                  0xA105
+#define mmCB_BLEND_GREEN                                0xA106
+#define mmCB_BLEND_BLUE                                 0xA107
+#define mmCB_BLEND_ALPHA                                0xA108
+#define mmCB_FOG_RED_R6XX                               0xA109
+#define mmCB_FOG_GREEN_R6XX                             0xA10A
+#define mmCB_FOG_BLUE_R6XX                              0xA10B
+#define mmCB_BLEND_CONTROL                              0xA201
+#define mmCB_COLOR_CONTROL                              0xA202
+#define mmCB_BLEND0_CONTROL                             0xA1E0
+#define mmCB_BLEND1_CONTROL                             0xA1E1
+#define mmCB_BLEND2_CONTROL                             0xA1E2
+#define mmCB_BLEND3_CONTROL                             0xA1E3
+#define mmCB_BLEND4_CONTROL                             0xA1E4
+#define mmCB_BLEND5_CONTROL                             0xA1E5
+#define mmCB_BLEND6_CONTROL                             0xA1E6
+#define mmCB_BLEND7_CONTROL                             0xA1E7
+#define mmCB_CLRCMP_CONTROL                             0xA30C
+#define mmCB_CLRCMP_SRC                                 0xA30D
+#define mmCB_CLRCMP_DST                                 0xA30E
+#define mmCB_CLRCMP_MSK                                 0xA30F
+#define mmCB_COLOR0_BASE                                0xA010
+#define mmCB_COLOR1_BASE                                0xA011
+#define mmCB_COLOR2_BASE                                0xA012
+#define mmCB_COLOR3_BASE                                0xA013
+#define mmCB_COLOR4_BASE                                0xA014
+#define mmCB_COLOR5_BASE                                0xA015
+#define mmCB_COLOR6_BASE                                0xA016
+#define mmCB_COLOR7_BASE                                0xA017
+#define mmCB_COLOR0_SIZE                                0xA018
+#define mmCB_COLOR1_SIZE                                0xA019
+#define mmCB_COLOR2_SIZE                                0xA01A
+#define mmCB_COLOR3_SIZE                                0xA01B
+#define mmCB_COLOR4_SIZE                                0xA01C
+#define mmCB_COLOR5_SIZE                                0xA01D
+#define mmCB_COLOR6_SIZE                                0xA01E
+#define mmCB_COLOR7_SIZE                                0xA01F
+#define mmCB_COLOR0_VIEW                                0xA020
+#define mmCB_COLOR1_VIEW                                0xA021
+#define mmCB_COLOR2_VIEW                                0xA022
+#define mmCB_COLOR3_VIEW                                0xA023
+#define mmCB_COLOR4_VIEW                                0xA024
+#define mmCB_COLOR5_VIEW                                0xA025
+#define mmCB_COLOR6_VIEW                                0xA026
+#define mmCB_COLOR7_VIEW                                0xA027
+#define mmCB_COLOR0_INFO                                0xA028
+#define mmCB_COLOR1_INFO                                0xA029
+#define mmCB_COLOR2_INFO                                0xA02A
+#define mmCB_COLOR3_INFO                                0xA02B
+#define mmCB_COLOR4_INFO                                0xA02C
+#define mmCB_COLOR5_INFO                                0xA02D
+#define mmCB_COLOR6_INFO                                0xA02E
+#define mmCB_COLOR7_INFO                                0xA02F
+#define mmCB_COLOR0_TILE                                0xA030
+#define mmCB_COLOR1_TILE                                0xA031
+#define mmCB_COLOR2_TILE                                0xA032
+#define mmCB_COLOR3_TILE                                0xA033
+#define mmCB_COLOR4_TILE                                0xA034
+#define mmCB_COLOR5_TILE                                0xA035
+#define mmCB_COLOR6_TILE                                0xA036
+#define mmCB_COLOR7_TILE                                0xA037
+#define mmCB_COLOR0_FRAG                                0xA038
+#define mmCB_COLOR1_FRAG                                0xA039
+#define mmCB_COLOR2_FRAG                                0xA03A
+#define mmCB_COLOR3_FRAG                                0xA03B
+#define mmCB_COLOR4_FRAG                                0xA03C
+#define mmCB_COLOR5_FRAG                                0xA03D
+#define mmCB_COLOR6_FRAG                                0xA03E
+#define mmCB_COLOR7_FRAG                                0xA03F
+#define mmCB_COLOR0_MASK                                0xA040
+#define mmCB_COLOR1_MASK                                0xA041
+#define mmCB_COLOR2_MASK                                0xA042
+#define mmCB_COLOR3_MASK                                0xA043
+#define mmCB_COLOR4_MASK                                0xA044
+#define mmCB_COLOR5_MASK                                0xA045
+#define mmCB_COLOR6_MASK                                0xA046
+#define mmCB_COLOR7_MASK                                0xA047
+#define mmCB_CLEAR_RED_R6XX                             0xA048
+#define mmCB_CLEAR_GREEN_R6XX                           0xA049
+#define mmCB_CLEAR_BLUE_R6XX                            0xA04A
+#define mmCB_CLEAR_ALPHA_R6XX                           0xA04B
+#define mmCB_TARGET_MASK                                0xA08E
+#define mmCB_SHADER_MASK                                0xA08F
+#define mmCB_SHADER_CONTROL                             0xA1E8
+
+#define mmSQ_VTX_BASE_VTX_LOC                           0xF3FC
+#define mmSQ_VTX_START_INST_LOC                         0xF3FD
+
+#endif /* _R700_CHIPOFFSET_H_ */
+
diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c
new file mode 100644
index 0000000000..e84be38622
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_clear.c
@@ -0,0 +1,122 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+ 
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/enums.h"
+
+#include "radeon_lock.h"
+#include "r600_context.h"
+
+#include "r700_shaderinst.h"
+#include "r600_emit.h"
+#include "r700_clear.h"
+
+static GLboolean r700ClearFast(context_t *context, GLbitfield mask)
+{
+    /* TODO, fast clear need implementation */
+    return GL_FALSE;
+}
+
+#define R600_NEWPRIM( rmesa )			\
+  do {						\
+  if ( rmesa->radeon.dma.flush )			\
+    rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
+  } while (0)
+
+void r700Clear(GLcontext * ctx, GLbitfield mask)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    __DRIdrawablePrivate *dPriv = radeon_get_drawable(&context->radeon);
+    const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+    GLbitfield swrast_mask = 0, tri_mask = 0;
+    int i;
+    struct gl_framebuffer *fb = ctx->DrawBuffer;
+
+    if( GL_TRUE == r700ClearFast(context, mask) )
+    {
+        return;
+    }
+	if (!context->radeon.radeonScreen->driScreen->dri2.enabled) {
+		LOCK_HARDWARE(&context->radeon);
+		UNLOCK_HARDWARE(&context->radeon);
+		if (dPriv->numClipRects == 0)
+			return;
+	}
+
+	R600_NEWPRIM(context);
+
+	if (colorMask == ~0)
+	  tri_mask |= (mask & BUFFER_BITS_COLOR);
+	else
+	  tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
+
+
+	/* HW stencil */
+	if (mask & BUFFER_BIT_STENCIL) {
+		tri_mask |= BUFFER_BIT_STENCIL;
+	}
+
+	/* HW depth */
+	if (mask & BUFFER_BIT_DEPTH) {
+    	        tri_mask |= BUFFER_BIT_DEPTH;
+	}
+
+	/* If we're doing a tri pass for depth/stencil, include a likely color
+	 * buffer with it.
+	 */
+
+	for (i = 0; i < BUFFER_COUNT; i++) {
+	  GLuint bufBit = 1 << i;
+	  if ((tri_mask) & bufBit) {
+	    if (!fb->Attachment[i].Renderbuffer->ClassID) {
+	      tri_mask &= ~bufBit;
+	      swrast_mask |= bufBit;
+	    }
+	  }
+	}
+
+	/* SW fallback clearing */
+	swrast_mask = mask & ~tri_mask;
+
+	if (tri_mask) {
+		radeonUserClear(ctx, tri_mask);
+	}
+
+	if (swrast_mask) {
+		if (RADEON_DEBUG & DEBUG_FALLBACKS)
+			fprintf(stderr, "%s: swrast clear, mask: %x\n",
+				__FUNCTION__, swrast_mask);
+		_swrast_Clear(ctx, swrast_mask);
+	}
+
+}
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_clear.h b/src/mesa/drivers/dri/r600/r700_clear.h
new file mode 100644
index 0000000000..bed1d3a90e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_clear.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef __r700_CLEAR_H__
+#define __r700_CLEAR_H__
+
+extern void r700Clear(GLcontext * ctx, GLbitfield mask);
+
+#endif /* __r700_CLEAR_H__ */
diff --git a/src/mesa/drivers/dri/r600/r700_debug.c b/src/mesa/drivers/dri/r600/r700_debug.c
new file mode 100644
index 0000000000..ecdb75ad48
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_debug.c
@@ -0,0 +1,83 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "main/glheader.h"
+
+#include "r700_debug.h"
+#include "r600_context.h"
+
+void NormalizeLogErrorCode(int nError)
+{
+    //TODO
+}
+
+void r700_error(int nLocalError, char* fmt, ...)
+{
+    va_list args;
+
+    NormalizeLogErrorCode(nLocalError);
+
+	va_start(args, fmt);
+    fprintf(stderr, fmt, args);
+    va_end(args);
+}
+
+void DumpHwBinary(int type, void *addr, int size)
+{
+    int i;
+    unsigned int *pHw = (unsigned int *)addr;
+
+    return;
+
+    switch (type)
+    {
+        case DUMP_PIXEL_SHADER:
+            DEBUGF("Pixel Shader\n");
+        break;
+        case DUMP_VERTEX_SHADER:
+            DEBUGF("Vertex Shader\n");
+        break;
+        case DUMP_FETCH_SHADER:
+            DEBUGF("Fetch Shader\n");
+        break;
+    }
+
+    for (i = 0; i < size; i++)
+    {
+        DEBUGP("0x%08x,\t", *pHw);
+        if (i%4 == 3)
+            DEBUGP("\n", *pHw);
+        pHw++;
+
+    }
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_debug.h b/src/mesa/drivers/dri/r600/r700_debug.h
new file mode 100644
index 0000000000..e810e6da08
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_debug.h
@@ -0,0 +1,93 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef _R700_DEBUG_H_
+#define _R700_DEBUG_H_
+
+enum R700_ERROR
+{
+    ERROR_ASM_VTX_CLAUSE     = 0x1000,
+    ERROR_ASM_UNKOWNCLAUSE   = 0x1001,
+    ERROR_ASM_ALLOCEXPORTCF  = 0x1002,
+    ERROR_ASM_ALLOCVTXCF     = 0x1003,
+    ERROR_ASM_ALLOCTEXCF     = 0x1004,
+    ERROR_ASM_ALLOCALUCF     = 0x1005,
+    ERROR_ASM_UNKNOWNILINST  = 0x1006,
+    ERROR_ASM_SRCARGUMENT    = 0x1007,
+    ERROR_ASM_DSTARGUMENT    = 0x1008,
+    ERROR_ASM_TEXINSTRUCTION = 0x1009,
+    ERROR_ASM_ALUINSTRUCTION = 0x100A,
+    ERROR_ASM_INSTDSTTRACK   = 0x100B,
+    ERROR_ASM_TEXDSTBADTYPE  = 0x100C,
+    ERROR_ASM_ALUSRCBADTYPE  = 0x100D,
+    ERROR_ASM_ALUSRCSELECT   = 0x100E,
+    ERROR_ASM_ALUSRCNUMBER   = 0x100F,
+    ERROR_ASM_ALUDSTBADTYPE  = 0x1010,
+    ERROR_ASM_CONSTCHANNEL   = 0x1011,
+    ERROR_ASM_BADSCALARBZ    = 0x1012,
+    ERROR_ASM_BADGPRRESERVE  = 0x1013,
+    ERROR_ASM_BADVECTORBZ    = 0x1014,
+    ERROR_ASM_BADTEXINST     = 0x1015,
+    ERROR_ASM_BADTEXSRC      = 0x1016,
+    ERROR_ASM_BADEXPORTTYPE  = 0x1017,
+
+
+    TODO_ASM_CONSTTEXADDR   = 0x8000,
+    TODO_ASM_NEEDIMPINST    = 0x8001,
+    TODO_ASM_TXB            = 0x8002,
+    TODO_ASM_TXP            = 0x8003
+};
+
+enum R700_DUMP_TYPE
+{
+    DUMP_VERTEX_SHADER      = 0x1,
+    DUMP_PIXEL_SHADER       = 0x2,
+    DUMP_FETCH_SHADER       = 0x4,
+};
+
+#define DEBUGF printf
+#define DEBUGP printf
+
+void NormalizeLogErrorCode(int nError);
+/*NormalizeLogErrorCode(nLocalError); */
+void r700_error(int nLocalError, char *fmt, ...);      
+extern void DumpHwBinary(int, void *, int);
+
+#ifdef STANDALONE_COMPILER
+#ifdef __cplusplus
+extern "C"
+{
+#endif //__cplusplus
+
+void LogString(char* szStr);
+
+#ifdef __cplusplus
+}
+#endif //__cplusplus
+#endif /*STANDALONE_COMPILER*/
+
+#endif /*_R700_DEBUG_H_*/
diff --git a/src/mesa/drivers/dri/r600/r700_driconf.h b/src/mesa/drivers/dri/r600/r700_driconf.h
new file mode 100644
index 0000000000..a9e2152344
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_driconf.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_DRICONF_H_
+#define _R700_DRICONF_H_
+
+#define DRI_CONF_FP_OPTIMIZATION_SPEED   0
+#define DRI_CONF_FP_OPTIMIZATION_QUALITY 1
+
+#endif /* _R700_DRICONF_H_ */
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c
new file mode 100644
index 0000000000..44de2aebee
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.c
@@ -0,0 +1,369 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/imports.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r700_fragprog.h"
+
+#include "r700_debug.h"
+
+//TODO : Validate FP input with VP output.
+void Map_Fragment_Program(r700_AssemblerBase         *pAsm,
+						  struct gl_fragment_program *mesa_fp)
+{
+	unsigned int unBit;
+    unsigned int i;
+    GLuint       ui;
+
+	pAsm->number_used_registers = 0;
+
+//Input mapping : mesa_fp->Base.InputsRead set the flag, set in 
+	//The flags parsed in parse_attrib_binding. FRAG_ATTRIB_COLx, FRAG_ATTRIB_TEXx, ...
+	//MUST match order in Map_Vertex_Output
+	unBit = 1 << FRAG_ATTRIB_COL0;
+	if(mesa_fp->Base.InputsRead & unBit)
+	{
+		pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0] = pAsm->number_used_registers++;
+	}
+
+	unBit = 1 << FRAG_ATTRIB_COL1;
+	if(mesa_fp->Base.InputsRead & unBit)
+	{
+		pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++;
+	}
+
+	for(i=0; i<8; i++)
+	{
+		unBit = 1 << (FRAG_ATTRIB_TEX0 + i);
+		if(mesa_fp->Base.InputsRead & unBit)
+		{
+			pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i] = pAsm->number_used_registers++;
+		}
+	}
+
+/* Map temporary registers (GPRs) */
+    pAsm->starting_temp_register_number = pAsm->number_used_registers;
+
+    if(mesa_fp->Base.NumNativeTemporaries >= mesa_fp->Base.NumTemporaries)
+    {
+	    pAsm->number_used_registers += mesa_fp->Base.NumNativeTemporaries;
+    }
+    else
+    {
+        pAsm->number_used_registers += mesa_fp->Base.NumTemporaries;
+    }
+
+/* Output mapping */
+	pAsm->number_of_exports = 0;
+	pAsm->number_of_colorandz_exports = 0; /* don't include stencil and mask out. */
+	pAsm->starting_export_register_number = pAsm->number_used_registers;
+	unBit = 1 << FRAG_RESULT_COLOR;
+	if(mesa_fp->Base.OutputsWritten & unBit)
+	{
+		pAsm->uiFP_OutputMap[FRAG_RESULT_COLOR] = pAsm->number_used_registers++;
+		pAsm->number_of_exports++;
+		pAsm->number_of_colorandz_exports++;
+	}
+	unBit = 1 << FRAG_RESULT_DEPTH;
+	if(mesa_fp->Base.OutputsWritten & unBit)
+	{
+        pAsm->depth_export_register_number = pAsm->number_used_registers;
+		pAsm->uiFP_OutputMap[FRAG_RESULT_DEPTH] = pAsm->number_used_registers++;
+		pAsm->number_of_exports++;
+		pAsm->number_of_colorandz_exports++;
+	}
+
+    pAsm->pucOutMask = (unsigned char*) MALLOC(pAsm->number_of_exports);    
+    for(ui=0; ui<pAsm->number_of_exports; ui++)
+    {
+        pAsm->pucOutMask[ui] = 0x0;
+    }
+	
+	pAsm->uFirstHelpReg = pAsm->number_used_registers;
+}
+
+GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp,
+					                	struct gl_fragment_program   *mesa_fp)
+{
+    GLuint i, j;
+    GLint * puiTEMPwrites;
+    struct prog_instruction * pILInst;
+    InstDeps         *pInstDeps;
+    struct prog_instruction * texcoord_DepInst;
+    GLint              nDepInstID;
+
+    puiTEMPwrites = (GLint*) MALLOC(sizeof(GLuint)*mesa_fp->Base.NumTemporaries);
+    for(i=0; i<mesa_fp->Base.NumTemporaries; i++)
+    {
+        puiTEMPwrites[i] = -1;
+    }
+
+    pInstDeps = (InstDeps*)MALLOC(sizeof(InstDeps)*mesa_fp->Base.NumInstructions);
+
+    for(i=0; i<mesa_fp->Base.NumInstructions; i++)
+    {
+        pInstDeps[i].nDstDep = -1;
+        pILInst = &(mesa_fp->Base.Instructions[i]);
+
+        //Dst
+        if(pILInst->DstReg.File == PROGRAM_TEMPORARY)
+        {
+            //Set lastwrite for the temp
+            puiTEMPwrites[pILInst->DstReg.Index] = i;
+        }
+
+        //Src
+        for(j=0; j<3; j++)
+        {
+            if(pILInst->SrcReg[j].File == PROGRAM_TEMPORARY)
+            {
+                //Set dep.
+                pInstDeps[i].nSrcDeps[j] = puiTEMPwrites[pILInst->SrcReg[j].Index];
+            }
+            else
+            {
+                pInstDeps[i].nSrcDeps[j] = -1;
+            }
+        }
+    }
+
+    fp->r700AsmCode.pInstDeps = pInstDeps;
+
+    FREE(puiTEMPwrites);
+
+    //Find dep for tex inst    
+    for(i=0; i<mesa_fp->Base.NumInstructions; i++)
+    {
+        pILInst = &(mesa_fp->Base.Instructions[i]);
+
+        if(GL_TRUE == IsTex(pILInst->Opcode))
+        {   //src0 is the tex coord register, src1 is texunit, src2 is textype
+            nDepInstID = pInstDeps[i].nSrcDeps[0];
+            if(nDepInstID >= 0)
+            {
+                texcoord_DepInst = &(mesa_fp->Base.Instructions[nDepInstID]);
+                if(GL_TRUE == IsAlu(texcoord_DepInst->Opcode) )
+                {
+                    pInstDeps[nDepInstID].nDstDep = i;
+                    pInstDeps[i].nDstDep = i;
+                }
+                else if(GL_TRUE == IsTex(texcoord_DepInst->Opcode) )
+                {
+                    pInstDeps[i].nDstDep = i;
+                }
+                else
+                {   //... other deps?
+                }
+            }
+        }
+	}
+
+    return GL_TRUE;
+}
+
+GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
+							     struct gl_fragment_program   *mesa_fp)
+{
+	GLuint    number_of_colors_exported;
+	GLboolean z_enabled = GL_FALSE;
+	GLuint    unBit;
+
+    //Init_Program
+	Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) );
+	Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp);
+
+    if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) )
+	{
+		return GL_FALSE;
+    }
+	
+	if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions,
+                                  &(mesa_fp->Base.Instructions[0]), 
+                                  &(fp->r700AsmCode)) )
+	{
+		return GL_FALSE;
+	}
+
+    if(GL_FALSE == Process_Fragment_Exports(&(fp->r700AsmCode), mesa_fp->Base.OutputsWritten) )
+    {
+        return GL_FALSE;
+    }
+
+    fp->r700Shader.nRegs = (fp->r700AsmCode.number_used_registers == 0) ? 0 
+                         : (fp->r700AsmCode.number_used_registers - 1);
+
+	fp->r700Shader.nParamExports = fp->r700AsmCode.number_of_exports;
+
+	number_of_colors_exported = fp->r700AsmCode.number_of_colorandz_exports;
+
+	unBit = 1 << FRAG_RESULT_DEPTH;
+	if(mesa_fp->Base.OutputsWritten & unBit)
+	{
+		z_enabled = GL_TRUE;
+		number_of_colors_exported--;
+	}
+
+	fp->r700Shader.exportMode = number_of_colors_exported << 1 | z_enabled;
+
+    fp->translated = GL_TRUE;
+
+	return GL_TRUE;
+}
+
+void * r700GetActiveFpShaderBo(GLcontext * ctx)
+{
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	                                   (ctx->FragmentProgram._Current);
+
+    return fp->shaderbo;
+}
+
+GLboolean r700SetupFragmentProgram(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);   
+    BATCH_LOCALS(&context->radeon);
+    
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	                                   (ctx->FragmentProgram._Current);
+
+    struct gl_program_parameter_list *paramList;
+    unsigned int unNumParamData;
+    unsigned int ui;
+
+    unsigned int unNumOfReg;
+    
+    if(GL_FALSE == fp->loaded)
+    {
+        if(fp->r700Shader.bNeedsAssembly == GL_TRUE)
+	    {
+		    Assemble( &(fp->r700Shader) );
+	    }
+
+        /* Load fp to gpu */
+        r600EmitShader(ctx, 
+                       &(fp->shaderbo), 
+                       (GLvoid *)(fp->r700Shader.pProgram),
+                       fp->r700Shader.uShaderBinaryDWORDSize,
+                       "FS");    			                
+
+        fp->loaded = GL_TRUE;
+    }
+
+    DumpHwBinary(DUMP_PIXEL_SHADER, (GLvoid *)(fp->r700Shader.pProgram),
+                 fp->r700Shader.uShaderBinaryDWORDSize);
+
+    /* TODO : enable this after MemUse fixed *=
+    (context->chipobj.MemUse)(context, fp->shadercode.buf->id);
+    */
+
+    r700->ps.SQ_PGM_START_PS.u32All = 0; /* set from buffer obj */
+
+    unNumOfReg = fp->r700Shader.nRegs + 1;
+
+    ui = (r700->SPI_PS_IN_CONTROL_0.u32All & NUM_INTERP_mask) / (1 << NUM_INTERP_shift);
+
+    ui = ui ? ui : unNumOfReg;
+
+    SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask); 
+    
+    CLEARbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, UNCACHED_FIRST_INST_bit);
+
+    if(fp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */
+	{
+        SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, fp->r700Shader.uStackSize,
+                 STACK_SIZE_shift, STACK_SIZE_mask);
+    }
+
+    SETfield(r700->ps.SQ_PGM_EXPORTS_PS.u32All, fp->r700Shader.exportMode,
+             EXPORT_MODE_shift, EXPORT_MODE_mask);
+
+    if(fp->r700Shader.killIsUsed)
+    {
+	    SETbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit);
+    }
+    else
+    {
+        CLEARbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit);
+    }
+
+    if(fp->r700Shader.depthIsExported)
+    {
+	    SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); 
+    }
+    else
+    {
+        CLEARbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit);
+    }
+
+    /* sent out shader constants. */
+
+    paramList = fp->mesa_program.Base.Parameters;
+
+    if(NULL != paramList)
+    {
+        _mesa_load_state_parameters(ctx, paramList);
+
+        unNumParamData = paramList->NumParameters * 4;
+
+        BEGIN_BATCH_NO_AUTOSTATE(2 + unNumParamData);
+        
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData));
+
+        /* assembler map const from very beginning. */
+        R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4);
+
+        unNumParamData = paramList->NumParameters;
+
+        for(ui=0; ui<unNumParamData; ui++)
+        {
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0])));
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1])));
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2])));
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3])));
+        }
+        END_BATCH();
+        COMMIT_BATCH();
+    }
+
+    return GL_TRUE;
+}
+
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h
new file mode 100644
index 0000000000..9c7813e908
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.h
@@ -0,0 +1,63 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_FRAGPROG_H_
+#define _R700_FRAGPROG_H_
+
+#include "r600_context.h"
+#include "r700_assembler.h"
+
+struct r700_fragment_program
+{
+	struct gl_fragment_program mesa_program;
+
+    r700_AssemblerBase r700AsmCode;
+	R700_Shader        r700Shader;
+
+	GLboolean translated;
+    GLboolean loaded;
+	GLboolean error;
+
+    void * shaderbo;
+
+	GLboolean WritesDepth;
+	GLuint optimization;
+};
+
+/* Internal */
+void Map_Fragment_Program(r700_AssemblerBase         *pAsm,
+						  struct gl_fragment_program *mesa_fp);
+GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp,
+					                	   struct gl_fragment_program   *mesa_fp);
+
+/* Interface */
+extern GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
+							                 struct gl_fragment_program   *mesa_vp);
+extern GLboolean r700SetupFragmentProgram(GLcontext * ctx);
+
+extern void *    r700GetActiveFpShaderBo(GLcontext * ctx);
+
+#endif /*_R700_FRAGPROG_H_*/
diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.c b/src/mesa/drivers/dri/r600/r700_ioctl.c
new file mode 100644
index 0000000000..c479532001
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_ioctl.c
@@ -0,0 +1,73 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <sched.h>
+#include <errno.h>
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "swrast/swrast.h"
+
+#include "radeon_common.h"
+#include "radeon_lock.h"
+#include "r600_context.h"
+
+#include "r700_ioctl.h"
+#include "r700_clear.h"
+
+static void r700Flush(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+    context_t *     context = R700_CONTEXT(ctx);
+
+	if (RADEON_DEBUG & DEBUG_IOCTL)
+		fprintf(stderr, "%s %d\n", __FUNCTION__, radeon->cmdbuf.cs->cdw);
+
+	/* okay if we have no cmds in the buffer &&
+	   we have no DMA flush &&
+	   we have no DMA buffer allocated.
+	   then no point flushing anything at all.
+	*/
+	if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && !radeon->dma.current)
+		return;
+
+	if (radeon->dma.flush)
+		radeon->dma.flush( ctx );
+
+	r700SendContextStates(context);
+
+	if (radeon->cmdbuf.cs->cdw)
+		rcommonFlushCmdBuf(radeon, __FUNCTION__);
+}
+
+void r700InitIoctlFuncs(struct dd_function_table *functions)
+{
+	functions->Clear = r700Clear;
+	functions->Finish = radeonFinish;
+	functions->Flush = r700Flush;
+}
diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.h b/src/mesa/drivers/dri/r600/r700_ioctl.h
new file mode 100644
index 0000000000..414dc3e23e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_ioctl.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef __R700_IOCTL_H__
+#define __R700_IOCTL_H__
+
+#include "r600_context.h"
+#include "radeon_drm.h"
+
+extern void r700InitIoctlFuncs(struct dd_function_table *functions);
+
+#endif				/* __R700_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c
new file mode 100644
index 0000000000..36de143b1a
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_oglprog.c
@@ -0,0 +1,141 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <string.h>
+
+#include "main/glheader.h"
+#include "main/imports.h"
+
+#include "shader/program.h"
+#include "tnl/tnl.h"
+
+#include "r600_context.h"
+
+#include "r700_oglprog.h"
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+
+
+static struct gl_program *r700NewProgram(GLcontext * ctx, 
+                                         GLenum target,
+					                     GLuint id)
+{
+	struct gl_program *pProgram = NULL;
+
+    struct r700_vertex_program *vp;
+	struct r700_fragment_program *fp;
+
+    switch (target) 
+    {
+    case GL_VERTEX_STATE_PROGRAM_NV:
+    case GL_VERTEX_PROGRAM_ARB:	    
+        vp       = CALLOC_STRUCT(r700_vertex_program);
+	    pProgram = _mesa_init_vertex_program(ctx, 
+                                             &vp->mesa_program,
+					                         target, 
+                                             id);
+        vp->translated = GL_FALSE;
+        vp->loaded     = GL_FALSE;
+ 
+        vp->shaderbo   = NULL;
+
+	    break;
+    case GL_FRAGMENT_PROGRAM_NV:
+    case GL_FRAGMENT_PROGRAM_ARB:
+		fp       = CALLOC_STRUCT(r700_fragment_program);
+		pProgram = _mesa_init_fragment_program(ctx, 
+                                               &fp->mesa_program,
+						                       target, 
+                                               id);
+        fp->translated = GL_FALSE;
+        fp->loaded     = GL_FALSE;
+
+        fp->shaderbo   = NULL;
+
+	    break;
+    default:
+	    _mesa_problem(ctx, "Bad target in r700NewProgram");
+    }
+
+	return pProgram;
+}
+
+static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog)
+{
+    struct r700_vertex_program   * vp;
+    struct r700_fragment_program * fp;
+    context_t *context = R700_CONTEXT(ctx);
+
+    switch (prog->Target) 
+    {
+    case GL_VERTEX_STATE_PROGRAM_NV:
+    case GL_VERTEX_PROGRAM_ARB:	    
+        vp = (struct r700_vertex_program*)prog;
+        /* Release DMA region */
+
+        r600DeleteShader(ctx, vp->shaderbo);
+
+        /* Clean up */
+        Clean_Up_Assembler(&(vp->r700AsmCode));
+        Clean_Up_Shader(&(vp->r700Shader));
+	    break;
+    case GL_FRAGMENT_PROGRAM_NV:
+    case GL_FRAGMENT_PROGRAM_ARB:
+		fp = (struct r700_fragment_program*)prog;
+        /* Release DMA region */
+
+        r600DeleteShader(ctx, fp->shaderbo);
+
+        /* Clean up */
+        Clean_Up_Assembler(&(fp->r700AsmCode));
+        Clean_Up_Shader(&(fp->r700Shader));
+	    break;
+    default:
+	    _mesa_problem(ctx, "Bad target in r700NewProgram");
+    }
+
+	_mesa_delete_program(ctx, prog);
+}
+
+static void
+r700ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
+{
+
+}
+
+static GLboolean r700IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
+{
+
+	return GL_TRUE;
+}
+
+void r700InitShaderFuncs(struct dd_function_table *functions)
+{
+	functions->NewProgram = r700NewProgram;
+	functions->DeleteProgram = r700DeleteProgram;
+	functions->ProgramStringNotify = r700ProgramStringNotify;
+	functions->IsProgramNative = r700IsProgramNative;
+}
diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.h b/src/mesa/drivers/dri/r600/r700_oglprog.h
new file mode 100644
index 0000000000..fe2e9d1974
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_oglprog.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#ifndef _R700_OGLPROG_H_
+#define _R700_OGLPROG_H_
+#include "r600_context.h"
+
+extern void r700InitShaderFuncs(struct dd_function_table *functions);
+
+#endif /*_R700_OGLPROG_H_*/
diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c
new file mode 100644
index 0000000000..f1e467a317
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_render.c
@@ -0,0 +1,458 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ *   CooperYuan <cooper.yuan@amd.com>, <cooperyuan@gmail.com>
+ */
+
+#include "main/glheader.h"
+#include "main/state.h"
+#include "main/imports.h"
+#include "main/enums.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "main/dd.h"
+#include "main/simple_list.h"
+#include "main/api_arrayelt.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "vbo/vbo.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
+#include "tnl/t_context.h"
+#include "tnl/t_vertex.h"
+#include "tnl/t_pipeline.h"
+
+#include "radeon_mipmap_tree.h"
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r600_tex.h"
+
+#include "r700_vertprog.h"
+#include "r700_fragprog.h"
+#include "r700_state.h"
+
+void r700WaitForIdle(context_t *context);
+void r700WaitForIdleClean(context_t *context);
+void r700Start3D(context_t *context);
+GLboolean r700SendTextureState(context_t *context);
+unsigned int r700PrimitiveType(int prim);
+void r600UpdateTextureState(GLcontext * ctx);
+GLboolean r700SyncSurf(context_t *context,
+		       struct radeon_bo *pbo,
+		       uint32_t read_domain,
+		       uint32_t write_domain,
+		       uint32_t sync_type);
+
+void r700WaitForIdle(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+    R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
+    R600_OUT_BATCH(WAIT_3D_IDLE_bit);
+
+    END_BATCH();
+    COMMIT_BATCH();
+}
+
+void r700WaitForIdleClean(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+    BEGIN_BATCH_NO_AUTOSTATE(5);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0));
+    R600_OUT_BATCH(CACHE_FLUSH_AND_INV_EVENT);
+
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+    R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX);
+    R600_OUT_BATCH(WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit);
+
+    END_BATCH();
+    COMMIT_BATCH();
+}
+
+void r700Start3D(context_t *context)
+{
+    BATCH_LOCALS(&context->radeon);
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+        BEGIN_BATCH_NO_AUTOSTATE(2);
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 0));
+        R600_OUT_BATCH(0);
+        END_BATCH();
+    }
+
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1));
+    R600_OUT_BATCH(0x80000000);
+    R600_OUT_BATCH(0x80000000);
+    END_BATCH();
+
+    COMMIT_BATCH();
+
+    r700WaitForIdleClean(context);
+}
+
+static GLboolean r700SetupShaders(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    GLuint exportCount;
+
+    r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0;
+    r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0;
+
+    SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
+    SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
+
+    r700SetupVertexProgram(ctx);
+
+    r700SetupFragmentProgram(ctx);
+
+    exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift);
+    r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1;
+
+    return GL_TRUE;
+}
+
+GLboolean r700SendTextureState(context_t *context)
+{
+    unsigned int i;
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    offset_modifiers offset_mod = {NO_SHIFT, 0, 0xFFFFFFFF};
+    struct radeon_bo *bo = NULL;
+    BATCH_LOCALS(&context->radeon);
+
+    for (i=0; i<R700_TEXTURE_NUMBERUNITS; i++) {
+	    radeonTexObj *t = r700->textures[i];
+	    if (t) {
+		    if (!t->image_override)
+			    bo = t->mt->bo;
+		    else
+			    bo = t->bo;
+		    if (bo) {
+
+			    r700SyncSurf(context, bo,
+					 RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM,
+					 0, TC_ACTION_ENA_bit);
+
+			    BEGIN_BATCH_NO_AUTOSTATE(9);
+			    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7));
+			    R600_OUT_BATCH(i * 7);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE0);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE1);
+			    R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2,
+						 bo,
+						 0,
+						 RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0, &offset_mod);
+			    R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3,
+						 bo,
+						 0,
+						 RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0, &offset_mod);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE4);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE5);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6);
+			    END_BATCH();
+
+			    BEGIN_BATCH_NO_AUTOSTATE(5);
+			    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, 3));
+			    R600_OUT_BATCH(i * 3);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER0);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER1);
+			    R600_OUT_BATCH(r700->textures[i]->SQ_TEX_SAMPLER2);
+			    END_BATCH();
+			    COMMIT_BATCH();
+		    }
+	    }
+    }
+    return GL_TRUE;
+}
+
+GLboolean r700SyncSurf(context_t *context,
+		       struct radeon_bo *pbo,
+		       uint32_t read_domain,
+		       uint32_t write_domain,
+		       uint32_t sync_type)
+{
+    BATCH_LOCALS(&context->radeon);
+    uint32_t cp_coher_size;
+    offset_modifiers offset_mod;
+
+    if (pbo->size == 0xffffffff)
+	    cp_coher_size = 0xffffffff;
+    else
+	    cp_coher_size = ((pbo->size + 255) >> 8);
+
+    offset_mod.shift     = NO_SHIFT;
+    offset_mod.shiftbits = 0;
+    offset_mod.mask      = 0xFFFFFFFF;
+
+    BEGIN_BATCH_NO_AUTOSTATE(5);
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SURFACE_SYNC, 3));
+    R600_OUT_BATCH(sync_type);
+    R600_OUT_BATCH(cp_coher_size);
+    R600_OUT_BATCH_RELOC(0,
+			 pbo,
+			 0,
+			 read_domain, write_domain, 0, &offset_mod); // ???
+    R600_OUT_BATCH(10);
+
+    END_BATCH();
+    COMMIT_BATCH();
+
+    return GL_TRUE;
+}
+
+unsigned int r700PrimitiveType(int prim)
+{
+    switch (prim & PRIM_MODE_MASK)
+    {
+    case GL_POINTS:
+        return DI_PT_POINTLIST;
+        break;
+    case GL_LINES:
+        return DI_PT_LINELIST;
+        break;
+    case GL_LINE_STRIP:
+        return DI_PT_LINESTRIP;
+        break;
+    case GL_LINE_LOOP:
+        return DI_PT_LINELOOP;
+        break;
+    case GL_TRIANGLES:
+        return DI_PT_TRILIST;
+        break;
+    case GL_TRIANGLE_STRIP:
+        return DI_PT_TRISTRIP;
+        break;
+    case GL_TRIANGLE_FAN:
+        return DI_PT_TRIFAN;
+        break;
+    case GL_QUADS:
+        return DI_PT_QUADLIST;
+        break;
+    case GL_QUAD_STRIP:
+        return DI_PT_QUADSTRIP;
+        break;
+    case GL_POLYGON:
+        return DI_PT_POLYGON;
+        break;
+    default:
+        assert(0);
+        return -1;
+        break;
+    }
+}
+
+static GLboolean r700RunRender(GLcontext * ctx,
+			                   struct tnl_pipeline_stage *stage)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    int lastIndex = 0;
+#if 1
+    BATCH_LOCALS(&context->radeon);
+
+    unsigned int i, j;
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *vb = &tnl->vb;
+
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	                                   (ctx->FragmentProgram._Current);
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+        fp->r700AsmCode.bR6xx = 1;
+    }
+
+    r700Start3D(context); /* TODO : this is too much. */
+
+    r700SendSQConfig(context);
+
+    r700UpdateShaders(ctx);
+
+    r700SetScissor(context);
+    r700SetRenderTarget(context, 0);
+    r700SetDepthTarget(context);
+
+    if(r700SetupStreams(ctx))
+    {
+        return GL_TRUE;
+    }
+
+    r600UpdateTextureState(ctx);
+    r700SendTextureState(context);
+
+    if(GL_FALSE == fp->translated)
+    {
+        if( GL_FALSE == r700TranslateFragmentShader(fp, &(fp->mesa_program)) )
+        {
+            return GL_TRUE;
+        }
+    }
+
+    r700SetupShaders(ctx);
+
+    r700SendFSState(context); // FIXME just a place holder for now
+    r700SendPSState(context);
+    r700SendVSState(context);
+
+    r700SendContextStates(context);
+    r700SendViewportState(context, 0);
+    r700SendRenderTargetState(context, 0);
+    r700SendDepthTargetState(context);
+
+    /* richard test code */
+    for (i = 0; i < vb->PrimitiveCount; i++) 
+    {
+        GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
+        GLuint start = vb->Primitive[i].start;
+        GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
+        GLuint numIndices = vb->Primitive[i].count;
+        GLuint numEntires;
+
+        unsigned int VGT_DRAW_INITIATOR = 0;
+        unsigned int VGT_INDEX_TYPE     = 0;
+        unsigned int VGT_PRIMITIVE_TYPE = 0;
+        unsigned int VGT_NUM_INDICES    = 0;
+        
+        numEntires = 2 /* VGT_INDEX_TYPE */
+                     + 3 /* VGT_PRIMITIVE_TYPE */
+                     + numIndices + 3; /* DRAW_INDEX_IMMD */                  
+                     
+        BEGIN_BATCH_NO_AUTOSTATE(numEntires);  
+
+        VGT_INDEX_TYPE |= DI_INDEX_SIZE_32_BIT << INDEX_TYPE_shift;
+
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0));
+        R600_OUT_BATCH(VGT_INDEX_TYPE);
+
+        VGT_NUM_INDICES = numIndices;
+
+        VGT_PRIMITIVE_TYPE |= r700PrimitiveType(prim) << VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift;
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1));
+        R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX);
+        R600_OUT_BATCH(VGT_PRIMITIVE_TYPE);
+
+        VGT_DRAW_INITIATOR |= DI_SRC_SEL_IMMEDIATE << SOURCE_SELECT_shift;
+        VGT_DRAW_INITIATOR |= DI_MAJOR_MODE_0 << MAJOR_MODE_shift;
+
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1)));
+        R600_OUT_BATCH(VGT_NUM_INDICES);
+        R600_OUT_BATCH(VGT_DRAW_INITIATOR);
+
+        for (j = lastIndex; j < lastIndex + numIndices; j++)
+        {
+            R600_OUT_BATCH(j);
+        }
+        lastIndex += numIndices;
+
+        END_BATCH();
+        COMMIT_BATCH();
+    }
+
+    /* Flush render op cached for last several quads. */
+    r700WaitForIdleClean(context);
+
+    radeonReleaseArrays(ctx, 0);
+
+#endif //0
+    rcommonFlushCmdBuf( &context->radeon, __FUNCTION__ );
+
+    return GL_FALSE;
+}
+
+static GLboolean r700RunNonTCLRender(GLcontext * ctx,
+				     struct tnl_pipeline_stage *stage) /* -------------------- */
+{
+	GLboolean bRet = GL_TRUE;
+	
+	return bRet;
+}
+
+static GLboolean r700RunTCLRender(GLcontext * ctx,  /*----------------------*/
+				  struct tnl_pipeline_stage *stage)
+{
+	GLboolean bRet = GL_FALSE;
+
+    /* TODO : sw fallback */
+
+    /**
+    * Ensure all enabled and complete textures are uploaded along with any buffers being used.
+    */
+    if(!r600ValidateBuffers(ctx))
+    {
+        return GL_TRUE;
+    }
+
+    context_t *context = R700_CONTEXT(ctx);
+
+    r700UpdateShaders(ctx);
+
+    bRet = r700RunRender(ctx, stage);
+
+    return bRet;
+	//GL_FALSE will stop to do other pipe stage in _tnl_run_pipeline
+    //The render here DOES finish the whole pipe, so GL_FALSE should be returned for success.
+}
+
+const struct tnl_pipeline_stage _r700_render_stage = {
+	"r700 Hardware Rasterization",
+	NULL,
+	NULL,
+	NULL,
+	NULL,
+	r700RunNonTCLRender
+};
+
+const struct tnl_pipeline_stage _r700_tcl_stage = {
+	"r700 Hardware Transform, Clipping and Lighting",
+	NULL,
+	NULL,
+	NULL,
+	NULL,
+	r700RunTCLRender
+};
+
+const struct tnl_pipeline_stage *r700_pipeline[] = 
+{
+    &_r700_tcl_stage,
+    &_tnl_vertex_transform_stage,
+	&_tnl_normal_transform_stage,
+	&_tnl_lighting_stage,
+	&_tnl_fog_coordinate_stage,
+	&_tnl_texgen_stage,
+	&_tnl_texture_transform_stage,
+	&_tnl_vertex_program_stage,
+
+    &_r700_render_stage,
+    &_tnl_render_stage,
+    0,
+};
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c
new file mode 100644
index 0000000000..b4fd51c137
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shader.c
@@ -0,0 +1,527 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/imports.h"
+
+#include "main/glheader.h"
+
+#include "r600_context.h"
+#include "r700_debug.h"
+
+#include "r700_shader.h"
+
+void r700ShaderInit(GLcontext * ctx)
+{
+}
+
+void AddInstToList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst)
+{
+	if(NULL == plstCFInstructions->pTail)
+	{	//first
+		plstCFInstructions->pHead = pInst;
+		plstCFInstructions->pTail = pInst;
+	}
+	else
+	{
+		plstCFInstructions->pTail->pNextInst = pInst;
+		plstCFInstructions->pTail = pInst;
+	}
+	pInst->pNextInst = NULL;
+
+	plstCFInstructions->uNumOfNode++;
+}
+
+void Init_R700_Shader(R700_Shader * pShader)
+{
+	pShader->Type = R700_SHADER_INVALID;
+	pShader->pProgram = NULL;
+	pShader->bBinaryShader = GL_FALSE;
+	pShader->bFetchShaderRequired = GL_FALSE;
+	pShader->bNeedsAssembly = GL_FALSE;
+	pShader->bLinksDirty = GL_TRUE;
+	pShader->uShaderBinaryDWORDSize = 0;
+	pShader->nRegs = 0;
+	pShader->nParamExports = 0;
+	pShader->nMemExports = 0;
+	pShader->resource = 0;
+
+	pShader->exportMode = 0;
+	pShader->depthIsImported = GL_FALSE;
+
+	pShader->positionVectorIsExported = GL_FALSE;
+	pShader->miscVectorIsExported = GL_FALSE;
+	pShader->renderTargetArrayIndexIsExported = GL_FALSE;
+	pShader->ccDist0VectorIsExported = GL_FALSE;
+	pShader->ccDist1VectorIsExported = GL_FALSE; 
+
+
+	pShader->depthIsExported = GL_FALSE;
+	pShader->stencilRefIsExported = GL_FALSE;
+	pShader->coverageToMaskIsExported = GL_FALSE;
+	pShader->maskIsExported = GL_FALSE;
+	pShader->killIsUsed = GL_FALSE;
+
+	pShader->uCFOffset = 0;
+	pShader->uStackSize = 0;
+	pShader->uMaxCallDepth = 0;
+
+	pShader->bSurfAllocated = GL_FALSE;
+	
+	pShader->lstCFInstructions.pHead=NULL;  
+	pShader->lstCFInstructions.pTail=NULL;  
+	pShader->lstCFInstructions.uNumOfNode=0;
+	pShader->lstALUInstructions.pHead=NULL; 
+	pShader->lstALUInstructions.pTail=NULL; 
+	pShader->lstALUInstructions.uNumOfNode=0;
+	pShader->lstTEXInstructions.pHead=NULL; 
+	pShader->lstTEXInstructions.pTail=NULL; 
+	pShader->lstTEXInstructions.uNumOfNode=0;
+	pShader->lstVTXInstructions.pHead=NULL; 
+	pShader->lstVTXInstructions.pTail=NULL; 
+	pShader->lstVTXInstructions.uNumOfNode=0;
+}
+
+void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst)
+{
+    R700ControlFlowSXClause*  pSXClause; 
+    R700ControlFlowSMXClause* pSMXClause;
+
+    pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode;
+    AddInstToList(&(pShader->lstCFInstructions), 
+                  (R700ShaderInstruction*)pCFInst);
+    pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType);
+
+    pSXClause = NULL;
+    pSMXClause = NULL; 
+	switch (pCFInst->m_ShaderInstType)
+	{
+	case SIT_CF_ALL_EXP_SX:
+		pSXClause =  (R700ControlFlowSXClause*)pCFInst;
+		break;
+	case SIT_CF_ALL_EXP_SMX:
+		pSMXClause = (R700ControlFlowSMXClause*)pCFInst;
+		break;
+	default:
+		break;
+	};
+
+    if((pSXClause != NULL) && (pSXClause->m_Word0.f.type == SQ_EXPORT_PARAM))
+    {
+        pShader->nParamExports += pSXClause->m_Word1.f.burst_count + 1;
+    }
+    else if ((pSMXClause != NULL) && (pSMXClause->m_Word1.f.cf_inst == SQ_CF_INST_MEM_RING) &&
+            (pSMXClause->m_Word0.f.type == SQ_EXPORT_WRITE || pSMXClause->m_Word0.f.type == SQ_EXPORT_WRITE_IND))
+    {
+        pShader->nMemExports += pSMXClause->m_Word1.f.burst_count + 1;
+    }
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pCFInst->useCount++;
+}
+
+void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst)
+{
+    pVTXInst->m_uIndex = pShader->lstVTXInstructions.uNumOfNode;
+	AddInstToList(&(pShader->lstVTXInstructions), 
+                  (R700ShaderInstruction*)pVTXInst);
+	pShader->uShaderBinaryDWORDSize += GetInstructionSize(pVTXInst->m_ShaderInstType);
+
+	if(pVTXInst->m_ShaderInstType == SIT_VTX_GENERIC)
+	{
+		R700VertexGenericFetch* pVTXGenericClause = (R700VertexGenericFetch*)pVTXInst;	
+		pShader->nRegs = (pShader->nRegs < pVTXGenericClause->m_Word1_GPR.f.dst_gpr) ? pVTXGenericClause->m_Word1_GPR.f.dst_gpr : pShader->nRegs;
+	}
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pVTXInst->useCount++;
+}
+
+void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst)
+{
+    pTEXInst->m_uIndex = pShader->lstTEXInstructions.uNumOfNode;
+	AddInstToList(&(pShader->lstTEXInstructions), 
+                  (R700ShaderInstruction*)pTEXInst);
+	pShader->uShaderBinaryDWORDSize += GetInstructionSize(pTEXInst->m_ShaderInstType);
+
+    pShader->nRegs = (pShader->nRegs < pTEXInst->m_Word1.f.dst_gpr) ? pTEXInst->m_Word1.f.dst_gpr : pShader->nRegs;
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pTEXInst->useCount++;
+}
+
+void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst)
+{
+    pALUInst->m_uIndex = pShader->lstALUInstructions.uNumOfNode;
+    AddInstToList(&(pShader->lstALUInstructions), 
+                  (R700ShaderInstruction*)pALUInst);
+    pShader->uShaderBinaryDWORDSize += GetInstructionSize(pALUInst->m_ShaderInstType);
+
+    pShader->nRegs = (pShader->nRegs < pALUInst->m_Word1.f.dst_gpr) ? pALUInst->m_Word1.f.dst_gpr : pShader->nRegs;
+
+    pShader->bLinksDirty    = GL_TRUE;
+    pShader->bNeedsAssembly = GL_TRUE;
+
+    pALUInst->useCount++;
+}
+
+void ResolveLinks(R700_Shader *pShader)
+{
+    GLuint uiSize;
+    R700ShaderInstruction  *pInst;
+    R700ALUInstruction     *pALUinst;
+    R700TextureInstruction *pTEXinst;
+    R700VertexInstruction  *pVTXinst; 
+
+    GLuint vtxOffset;
+
+	GLuint cfOffset = 0x0;  
+
+    GLuint aluOffset = cfOffset + pShader->lstCFInstructions.uNumOfNode * GetInstructionSize(SIT_CF);
+
+    GLuint texOffset = aluOffset;  // + m_lstALUInstructions.size() * R700ALUInstruction::SIZE,
+
+    pInst = pShader->lstALUInstructions.pHead;
+    while(NULL != pInst)
+    {
+        texOffset += GetInstructionSize(pInst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+  
+    vtxOffset = texOffset + pShader->lstTEXInstructions.uNumOfNode * GetInstructionSize(SIT_TEX);
+
+    if ( ((pShader->lstTEXInstructions.uNumOfNode > 0) && (texOffset % 4 != 0)) || 
+         ((pShader->lstVTXInstructions.uNumOfNode > 0) && (vtxOffset % 4 != 0))    )
+    {
+        pALUinst = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction);
+        Init_R700ALUInstruction(pALUinst);
+        AddALUInstruction(pShader, pALUinst);
+        texOffset += GetInstructionSize(SIT_ALU);
+        vtxOffset += GetInstructionSize(SIT_ALU);
+    }
+
+    pInst  = pShader->lstALUInstructions.pHead;
+    uiSize = 0;
+    while(NULL != pInst)
+    {
+        pALUinst = (R700ALUInstruction*)pInst;
+
+        if(pALUinst->m_pLinkedALUClause != NULL)
+        {
+            // This address is quad-word aligned
+            pALUinst->m_pLinkedALUClause->m_Word0.f.addr = (aluOffset + uiSize) >> 1;
+        }
+
+        uiSize += GetInstructionSize(pALUinst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+
+    pInst  = pShader->lstTEXInstructions.pHead;
+    uiSize = 0;
+    while(NULL != pInst)
+    {
+        pTEXinst = (R700TextureInstruction*)pInst;
+
+        if (pTEXinst->m_pLinkedGenericClause != NULL)
+        {
+            pTEXinst->m_pLinkedGenericClause->m_Word0.f.addr = (texOffset + uiSize) >> 1;
+        }
+
+        uiSize += GetInstructionSize(pTEXinst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+
+    pInst  = pShader->lstVTXInstructions.pHead;
+    uiSize = 0;
+    while(NULL != pInst)
+    {
+        pVTXinst = (R700VertexInstruction*)pInst;
+
+        if (pVTXinst->m_pLinkedGenericClause != NULL)
+        {
+            pVTXinst->m_pLinkedGenericClause->m_Word0.f.addr = (vtxOffset + uiSize) >> 1;
+        }
+
+        uiSize += GetInstructionSize(pVTXinst->m_ShaderInstType);
+
+        pInst = pInst->pNextInst;
+    };
+
+    pShader->bLinksDirty = GL_FALSE;
+}
+
+void Assemble(R700_Shader *pShader)
+{
+	GLuint i;
+    GLuint *pShaderBinary;
+    GLuint size_of_program;
+    GLuint *pCurrPos;
+
+    GLuint end_of_cf_instructions;
+    GLuint number_of_alu_dwords;
+
+    R700ShaderInstruction  *pInst;
+
+    if(GL_TRUE == pShader->bBinaryShader)
+    {
+        return;
+    }
+
+    if(pShader->bLinksDirty == GL_TRUE) 
+    {
+        ResolveLinks(pShader);
+    }
+
+    size_of_program = pShader->uShaderBinaryDWORDSize;
+    
+    pShaderBinary = (GLuint*) MALLOC(sizeof(GLuint)*size_of_program);
+ 
+    pCurrPos = pShaderBinary;
+
+    for (i = 0; i < size_of_program; i++)
+    {
+        pShaderBinary[i] = 0;
+    }
+
+    pInst = pShader->lstCFInstructions.pHead;
+    while(NULL != pInst)
+    {
+        switch (pInst->m_ShaderInstType)
+        {
+        case SIT_CF_GENERIC: 
+            {
+                R700ControlFlowGenericClause* pCFgeneric = (R700ControlFlowGenericClause*)pInst;
+                *pCurrPos++ = pCFgeneric->m_Word0.val;
+                *pCurrPos++ = pCFgeneric->m_Word1.val;
+            }
+            break;
+        case SIT_CF_ALU: 
+            {
+                R700ControlFlowALUClause* pCFalu = (R700ControlFlowALUClause*)pInst;
+                *pCurrPos++ = pCFalu->m_Word0.val;
+                *pCurrPos++ = pCFalu->m_Word1.val;
+            }
+            break;
+        case SIT_CF_ALL_EXP_SX: 
+            {
+                R700ControlFlowSXClause* pCFsx = (R700ControlFlowSXClause*)pInst;
+                *pCurrPos++ = pCFsx->m_Word0.val;
+                *pCurrPos++ = (pCFsx->m_Word1.val | pCFsx->m_Word1_SWIZ.val);
+            }
+            break;
+        case SIT_CF_ALL_EXP_SMX: 
+            {
+                R700ControlFlowSMXClause* pCFsmx = (R700ControlFlowSMXClause*)pInst;
+                *pCurrPos++ = pCFsmx->m_Word0.val;
+                *pCurrPos++ = (pCFsmx->m_Word1.val | pCFsmx->m_Word1_BUF.val);
+            }
+            break;
+        default:
+            break;
+        }
+
+        pInst = pInst->pNextInst;
+    };
+    
+    number_of_alu_dwords = 0;
+    pInst = pShader->lstALUInstructions.pHead;
+    while(NULL != pInst)
+    {
+        switch (pInst->m_ShaderInstType)
+        {
+        case SIT_ALU: 
+            {
+                R700ALUInstruction* pALU = (R700ALUInstruction*)pInst;
+
+                *pCurrPos++ = pALU->m_Word0.val;
+                *pCurrPos++ = (pALU->m_Word1.val | pALU->m_Word1_OP2.val | pALU->m_Word1_OP3.val);
+
+                number_of_alu_dwords += 2;
+            }
+            break;
+        case SIT_ALU_HALF_LIT: 
+            {
+                R700ALUInstructionHalfLiteral* pALUhalf = (R700ALUInstructionHalfLiteral*)pInst;
+
+                *pCurrPos++ = pALUhalf->m_Word0.val;
+                *pCurrPos++ = (pALUhalf->m_Word1.val | pALUhalf->m_Word1_OP2.val | pALUhalf->m_Word1_OP3.val);
+                *pCurrPos++ = *((GLuint*)&(pALUhalf->m_fLiteralX));
+                *pCurrPos++ = *((GLuint*)&(pALUhalf->m_fLiteralY));
+
+                number_of_alu_dwords += 4;
+            }
+            break;
+        case SIT_ALU_FALL_LIT: 
+            {
+                R700ALUInstructionFullLiteral* pALUfull = (R700ALUInstructionFullLiteral*)pInst;
+
+                *pCurrPos++ = pALUfull->m_Word0.val;
+                *pCurrPos++ = (pALUfull->m_Word1.val | pALUfull->m_Word1_OP2.val | pALUfull->m_Word1_OP3.val);
+
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralX));
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralY));
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralZ));
+                *pCurrPos++ = *((GLuint*)&(pALUfull->m_fLiteralW));
+
+                number_of_alu_dwords += 6;
+            }
+            break;
+        default:
+            break;
+        }
+
+        pInst = pInst->pNextInst;
+    };
+    
+    pInst = pShader->lstTEXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        R700TextureInstruction* pTEX = (R700TextureInstruction*)pInst;
+
+        *pCurrPos++ = pTEX->m_Word0.val;
+        *pCurrPos++ = pTEX->m_Word1.val;
+        *pCurrPos++ = pTEX->m_Word2.val;
+        *pCurrPos++ = 0x0beadeaf;
+
+        pInst = pInst->pNextInst;
+    };
+    
+    pInst = pShader->lstVTXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        switch (pInst->m_ShaderInstType)
+        {
+        case SIT_VTX_SEM: //
+            {
+                R700VertexSemanticFetch* pVTXsem = (R700VertexSemanticFetch*)pInst;
+
+                *pCurrPos++ = pVTXsem->m_Word0.val;
+                *pCurrPos++ = (pVTXsem->m_Word1.val | pVTXsem->m_Word1_SEM.val);
+                *pCurrPos++ = pVTXsem->m_Word2.val;
+                *pCurrPos++ = 0x0beadeaf;
+            }
+            break;
+        case SIT_VTX_GENERIC: //
+            {
+                R700VertexGenericFetch* pVTXgeneric = (R700VertexGenericFetch*)pInst;
+
+                *pCurrPos++ = pVTXgeneric->m_Word0.val;
+                *pCurrPos++ = (pVTXgeneric->m_Word1.val | pVTXgeneric->m_Word1_GPR.val);
+                *pCurrPos++ = pVTXgeneric->m_Word2.val;
+                *pCurrPos++ = 0x0beadeaf;
+            }
+            break;
+        default:
+            break;
+        }
+
+        pInst = pInst->pNextInst;
+    };
+
+    if(NULL != pShader->pProgram)
+    {
+        FREE(pShader->pProgram);
+    }
+    pShader->pProgram = (GLubyte*)pShaderBinary;
+
+    end_of_cf_instructions = pShader->uCFOffset + pShader->lstCFInstructions.uNumOfNode * GetInstructionSize(SIT_CF);
+    
+    pShader->uEndOfCF = end_of_cf_instructions >> 1;
+
+    pShader->uEndOfALU = (end_of_cf_instructions + number_of_alu_dwords) >> 1;
+
+    pShader->uEndOfFetch = (pShader->uCFOffset + pShader->uShaderBinaryDWORDSize) >> 1;
+
+    pShader->bNeedsAssembly = GL_FALSE;
+}
+
+void LoadProgram(R700_Shader *pShader) //context
+{
+}
+
+void UpdateShaderRegisters(R700_Shader *pShader) //context
+{
+}
+
+void DeleteInstructions(R700_Shader *pShader)
+{
+}
+
+void DebugPrint(void)
+{
+}
+
+void Clean_Up_Shader(R700_Shader *pShader)
+{
+    FREE(pShader->pProgram);
+
+    R700ShaderInstruction  *pInst;
+    R700ShaderInstruction  *pInstToFree;
+
+    pInst = pShader->lstCFInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+    pInst = pShader->lstALUInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+    pInst = pShader->lstTEXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+    pInst = pShader->lstVTXInstructions.pHead;
+    while(NULL != pInst)
+    {
+        pInstToFree = pInst;
+        pInst = pInst->pNextInst;
+        FREE(pInstToFree);
+    };
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h
new file mode 100644
index 0000000000..bfd01e1a93
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shader.h
@@ -0,0 +1,150 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef __R700_SHADER_H__
+#define __R700_SHADER_H__
+
+#include "main/mtypes.h"
+
+#include "r700_shaderinst.h"
+
+
+void r700ShaderInit(GLcontext * ctx);
+
+typedef enum R700ShaderType
+{
+    R700_SHADER_FS      = 0x0,
+    R700_SHADER_ES      = 0x1,
+    R700_SHADER_GS      = 0x2,
+    R700_SHADER_VS      = 0x3,
+    R700_SHADER_PS      = 0x4,
+    R700_SHADER_INVALID = 0x5,
+} R700ShaderType;
+
+typedef struct TypedShaderList 
+{
+	R700ShaderInstruction * pHead;
+	R700ShaderInstruction * pTail;
+	GLuint  uNumOfNode;
+} TypedShaderList;
+
+typedef struct RealRegister 
+{
+    GLuint uAddr;
+    GLuint uValue;
+} RealRegister;
+
+typedef struct InstDeps
+{
+    GLint nDstDep;
+    GLint nSrcDeps[3];
+} InstDeps;
+
+typedef struct R700_Shader 
+{
+	R700ShaderType   Type;
+
+    GLubyte*  pProgram;
+
+    GLboolean bBinaryShader;
+    GLboolean bFetchShaderRequired;
+    GLboolean bNeedsAssembly;
+    GLboolean bLinksDirty;
+
+    GLuint  uShaderBinaryDWORDSize; // in DWORDS
+    GLuint  nRegs;      
+    GLuint  nParamExports;   // VS_ EXPORT_COUNT (1 based, the actual register is 0 based!)
+    GLuint  nMemExports; 
+    GLuint  resource;     // VS and PS _RESOURCE
+    GLuint  exportMode;   // VS and PS _EXPORT_MODE
+
+    GLboolean  depthIsImported;             
+
+    // Vertex program exports
+    GLboolean  positionVectorIsExported;          
+
+    GLboolean  miscVectorIsExported;               
+    GLboolean  renderTargetArrayIndexIsExported;  
+
+    GLboolean  ccDist0VectorIsExported;  
+    GLboolean  ccDist1VectorIsExported;  
+
+    // Pixel program exports
+    GLboolean  depthIsExported;             
+    GLboolean  stencilRefIsExported;        
+    GLboolean  coverageToMaskIsExported;    
+    GLboolean  maskIsExported;              
+
+    GLboolean  killIsUsed;                  
+
+    GLuint  uStartAddr;
+    GLuint  uCFOffset;
+    GLuint  uEndOfCF;
+    GLuint  uEndOfALU;
+    GLuint  uEndOfFetch;
+    GLuint  uStackSize;
+    GLuint  uMaxCallDepth;
+
+	TypedShaderList lstCFInstructions;
+	TypedShaderList lstALUInstructions;
+	TypedShaderList lstTEXInstructions;
+	TypedShaderList lstVTXInstructions;
+
+    RealRegister RegStartAddr;
+    RealRegister RegCFOffset;
+    RealRegister RegEndCF;
+    RealRegister RegEndALU;
+    RealRegister egEndFetcg;
+
+	// -------- constants
+	GLfloat   ConstantArray[SQ_ALU_CONSTANT_PS_COUNT * 4];
+	
+	GLboolean bSurfAllocated;
+} R700_Shader;
+
+//Internal
+void AddInstToList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst);
+void ResolveLinks(R700_Shader *pShader);
+void Assemble(R700_Shader *pShader);
+
+
+//Interface
+void Init_R700_Shader(R700_Shader * pShader);
+void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst);
+void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst);
+void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst);
+void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst);
+
+void LoadProgram(R700_Shader *pShader);
+void UpdateShaderRegisters(R700_Shader *pShader);
+void DeleteInstructions(R700_Shader *pShader);
+void DebugPrint(void);
+
+void Clean_Up_Shader(R700_Shader *pShader);
+
+#endif /*__R700_SHADER_H__*/
+
diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.c b/src/mesa/drivers/dri/r600/r700_shaderinst.c
new file mode 100644
index 0000000000..c1bffee91f
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shaderinst.c
@@ -0,0 +1,224 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#include "main/mtypes.h"
+
+#include "r700_debug.h"
+#include "r700_shaderinst.h"
+
+void Init_R700ControlFlowGenericClause(R700ControlFlowGenericClause* pInst)
+{
+    pInst->m_Word0.val = 0x00000000;
+    pInst->m_Word1.val = 0x00000000;
+
+    pInst->m_pLinkedVTXInstruction = 0;
+    pInst->m_pLinkedTEXInstruction = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_GENERIC;
+}
+
+void Init_R700ControlFlowALUClause(R700ControlFlowALUClause* pInst)
+{
+    pInst->m_Word0.val = 0x00000000;
+    pInst->m_Word1.val = 0x00000000;
+
+    pInst->m_pLinkedALUInstruction = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_ALU;
+}
+
+void Init_R700ControlFlowSXClause(R700ControlFlowSXClause* pInst)
+{
+    pInst->m_Word0.val      = 0x00000000;
+    pInst->m_Word1.val      = 0x00000000;
+    pInst->m_Word1_SWIZ.val = 0x00000000;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_ALL_EXP_SX;
+}
+
+void Init_R700ControlFlowSMXClause(R700ControlFlowSMXClause* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_BUF.val = 0x00000000;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_CF_ALL_EXP_SMX;
+}
+
+void Init_R700ALUInstruction(R700ALUInstruction* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_OP2.val = 0x00000000;
+    pInst->m_Word1_OP3.val = 0x00000000;
+
+    pInst->m_pLinkedALUClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_ALU;
+}
+
+void Init_R700ALUInstructionHalfLiteral(R700ALUInstructionHalfLiteral* pInst, GLfloat x, GLfloat y)
+{
+	pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_OP2.val = 0x00000000;
+    pInst->m_Word1_OP3.val = 0x00000000;
+
+	pInst->m_pLinkedALUClause = 0;
+
+    pInst->m_fLiteralX = x;
+    pInst->m_fLiteralY = y;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_ALU_HALF_LIT;
+}
+
+void Init_R700ALUInstructionFullLiteral(R700ALUInstructionFullLiteral* pInst, GLfloat x, GLfloat y, GLfloat z, GLfloat w)
+{
+	pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_OP2.val = 0x00000000;
+    pInst->m_Word1_OP3.val = 0x00000000;
+
+	pInst->m_pLinkedALUClause = 0;
+
+    pInst->m_fLiteralX = x;
+    pInst->m_fLiteralY = y;
+    pInst->m_fLiteralZ = z;
+    pInst->m_fLiteralW = w;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_ALU_FALL_LIT;
+}
+
+void Init_R700TextureInstruction(R700TextureInstruction* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word2.val     = 0x00000000;
+
+    pInst->m_pLinkedGenericClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_TEX;
+}
+
+void Init_R700VertexSemanticFetch(R700VertexSemanticFetch* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_SEM.val = 0x00000000;
+    pInst->m_Word2.val     = 0x00000000;
+
+    pInst->m_pLinkedGenericClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_VTX_SEM;
+}
+
+void Init_R700VertexGenericFetch(R700VertexGenericFetch* pInst)
+{
+    pInst->m_Word0.val     = 0x00000000;
+    pInst->m_Word1.val     = 0x00000000;
+    pInst->m_Word1_GPR.val = 0x00000000;
+    pInst->m_Word2.val     = 0x00000000;
+
+    pInst->m_pLinkedGenericClause = 0;
+
+    pInst->useCount = 0;
+
+	pInst->m_ShaderInstType = SIT_VTX_GENERIC;
+}
+
+unsigned int GetInstructionSize(ShaderInstType instType)
+{
+    switch(instType)
+    {
+    case SIT_ALU_HALF_LIT:  
+    case SIT_TEX:           
+    case SIT_VTX:           
+    case SIT_VTX_GENERIC:   
+    case SIT_VTX_SEM:       
+        return 4;
+    case SIT_ALU_FALL_LIT:
+        return 6;
+    default:
+        break;
+    }
+
+    return 2;
+}
+
+unsigned int GetCFMaxInstructions(ShaderInstType instType)
+{
+    switch (instType)
+    {
+    case SIT_CF_ALL_EXP:    
+    case SIT_CF_ALL_EXP_SX: 
+    case SIT_CF_ALL_EXP_SMX:  
+        return 0x10;
+    case SIT_CF_GENERIC:
+        return 0x8;  //For tex and vtx
+    case SIT_CF_ALU:
+        return 0x80;
+    default:
+        break;
+    }
+    return 0x10;
+}
+
+GLboolean LinkVertexInstruction(R700ControlFlowGenericClause *pCFGeneric,
+								R700VertexInstruction *pVTXInstruction)
+{
+	if (pCFGeneric->m_pLinkedTEXInstruction != 0) 
+	{
+		r700_error(ERROR_ASM_VTX_CLAUSE, "This instruction is already linked to a texture instruction");
+		return GL_FALSE;
+    }
+
+    pCFGeneric->m_pLinkedVTXInstruction     = pVTXInstruction;
+	pVTXInstruction->m_pLinkedGenericClause = pCFGeneric;
+
+	return GL_TRUE;
+}
+
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.h b/src/mesa/drivers/dri/r600/r700_shaderinst.h
new file mode 100644
index 0000000000..2829cca0a3
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_shaderinst.h
@@ -0,0 +1,321 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef _R700_SHADERINST_H_
+#define _R700_SHADERINST_H_
+
+#include "main/glheader.h"
+
+#include "defaultendian.h" 
+#include "sq_micro_reg.h"
+
+#define SQ_ALU_CONSTANT_PS_OFFSET      0x00000000
+#define SQ_ALU_CONSTANT_PS_COUNT       0x00000100
+#define SQ_ALU_CONSTANT_VS_OFFSET      0x00000100
+#define SQ_ALU_CONSTANT_VS_COUNT       0x00000100
+#define SQ_FETCH_RESOURCE_PS_OFFSET    0x00000000
+#define SQ_FETCH_RESOURCE_PS_COUNT     0x000000a0
+#define SQ_FETCH_RESOURCE_VS_OFFSET    0x000000a0
+#define SQ_FETCH_RESOURCE_VS_COUNT     0x000000b0
+
+#define SHADERINST_TYPEMASK_CF  0x10
+#define SHADERINST_TYPEMASK_ALU 0x20
+#define SHADERINST_TYPEMASK_TEX 0x40
+#define SHADERINST_TYPEMASK_VTX 0x80
+
+typedef enum ShaderInstType 
+{
+    SIT_CF = 0x10,            /*SIZE = 0x2*/
+        SIT_CF_ALL_EXP = 0x14,    /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x10;*/
+            SIT_CF_ALL_EXP_SX = 0x15, /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x10;*/
+            SIT_CF_ALL_EXP_SMX= 0x16, /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x10;*/
+        SIT_CF_GENERIC = 0x18,    /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x8;  //For tex and vtx*/
+        SIT_CF_ALU = 0x19,        /*SIZE = 0x2, MAX_INSTRUCTIONS = 0x80;*/
+    SIT_ALU = 0x20,           /*SIZE = 0x2,*/
+        SIT_ALU_HALF_LIT = 0x21,  /*SIZE = 0x4,*/
+        SIT_ALU_FALL_LIT = 0x22,  /*SIZE = 0x6,*/
+    SIT_TEX = 0x40,           /*SIZE = 0x4,*/
+    SIT_VTX = 0x80,           /*SIZE = 0x4, MEGA_FETCH_BYTES = 0x20*/
+        SIT_VTX_GENERIC = 0x81,   /*SIZE = 0x4, MEGA_FETCH_BYTES = 0x20*/
+        SIT_VTX_SEM = 0x82       /*SIZE = 0x4, MEGA_FETCH_BYTES = 0x20*/
+} ShaderInstType;
+
+typedef struct R700ShaderInstruction 
+{
+    ShaderInstType m_ShaderInstType;
+    struct R700ShaderInstruction *pNextInst;
+    GLuint m_uIndex;
+    GLuint useCount;
+} R700ShaderInstruction;
+
+// ------------------ CF insts ---------------------------
+
+typedef R700ShaderInstruction R700ControlFlowInstruction;
+
+typedef struct R700ControlFlowAllocExportClause  
+{
+    ShaderInstType          m_ShaderInstType;
+    R700ShaderInstruction * pNextInst;    
+    GLuint m_uIndex;
+    GLuint useCount;
+		
+    sq_cf_alloc_export_word0_u      m_Word0;
+    sq_cf_alloc_export_word1_u      m_Word1;
+} R700ControlFlowAllocExportClause;
+
+typedef struct R700ControlFlowSXClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ControlFlowAllocExportClause
+		//R700ControlFlowInstruction 
+			//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+			//---------------------
+		//---------------------------
+    sq_cf_alloc_export_word0_u      m_Word0;
+    sq_cf_alloc_export_word1_u      m_Word1;
+	//-------------------------------------
+
+    sq_cf_alloc_export_word1_swiz_u m_Word1_SWIZ;
+} R700ControlFlowSXClause;
+
+typedef struct R700ControlFlowSMXClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+    //R700ControlFlowAllocExportClause
+		//R700ControlFlowInstruction 
+			//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+			//---------------------
+		//---------------------------
+    sq_cf_alloc_export_word0_u      m_Word0;
+    sq_cf_alloc_export_word1_u      m_Word1;
+	//-------------------------------
+
+    sq_cf_alloc_export_word1_buf_u m_Word1_BUF;
+} R700ControlFlowSMXClause;
+
+typedef struct R700ControlFlowGenericClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ControlFlowInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	//---------------------
+
+    sq_cf_word0_u m_Word0;
+    sq_cf_word1_u m_Word1;
+
+    struct R700VertexInstruction  *m_pLinkedVTXInstruction;
+    struct R700TextureInstruction *m_pLinkedTEXInstruction;
+} R700ControlFlowGenericClause;
+
+typedef struct R700ControlFlowALUClause 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+    //R700ControlFlowInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	//---------------------
+
+    sq_cf_alu_word0_u m_Word0;
+    sq_cf_alu_word1_u m_Word1;
+    
+    struct R700ALUInstruction *m_pLinkedALUInstruction;
+} R700ControlFlowALUClause;
+
+// ------------------- End of CF Inst ------------------------
+
+// ------------------- ALU Inst ------------------------------
+typedef struct R700ALUInstruction 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+	//---------------------
+
+    sq_alu_word0_u     m_Word0;
+    sq_alu_word1_u     m_Word1;
+    sq_alu_word1_op2_v2_u m_Word1_OP2;
+    sq_alu_word1_op3_u m_Word1_OP3;
+
+    struct R700ControlFlowALUClause *m_pLinkedALUClause;
+} R700ALUInstruction;
+
+typedef struct R700ALUInstructionHalfLiteral
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ALUInstruction 
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+
+    sq_alu_word0_u     m_Word0;
+    sq_alu_word1_u     m_Word1;
+    sq_alu_word1_op2_v2_u m_Word1_OP2;
+    sq_alu_word1_op3_u m_Word1_OP3;
+
+    struct R700ControlFlowALUClause *m_pLinkedALUClause;
+	//-------------------
+
+    GLfloat m_fLiteralX,
+            m_fLiteralY;
+} R700ALUInstructionHalfLiteral;
+
+typedef struct R700ALUInstructionFullLiteral 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ALUInstruction 
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+
+    sq_alu_word0_u     m_Word0;
+    sq_alu_word1_u     m_Word1;
+    sq_alu_word1_op2_v2_u m_Word1_OP2;
+    sq_alu_word1_op3_u m_Word1_OP3;
+
+    struct R700ControlFlowALUClause *m_pLinkedALUClause;
+	//-------------------
+
+    GLfloat m_fLiteralX,
+            m_fLiteralY,
+            m_fLiteralZ,
+            m_fLiteralW;
+} R700ALUInstructionFullLiteral;
+// ------------------- End of ALU Inst -----------------------
+
+// ------------------- Textuer/Vertex Instruction --------------------
+
+typedef struct R700TextureInstruction 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+	//---------------------
+	
+    sq_tex_word0_u m_Word0;
+    sq_tex_word1_u m_Word1;
+    sq_tex_word2_u m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+} R700TextureInstruction;
+
+typedef struct R700VertexInstruction 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+	//---------------------
+	
+    sq_vtx_word0_u     m_Word0;
+    sq_vtx_word1_u     m_Word1;
+    sq_vtx_word2_u     m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+} R700VertexInstruction;
+//
+typedef struct R700VertexSemanticFetch 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700VertexInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	
+    sq_vtx_word0_u     m_Word0;
+    sq_vtx_word1_u     m_Word1;
+    sq_vtx_word2_u     m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+	//---------------------------
+
+    sq_vtx_word1_sem_u m_Word1_SEM;
+} R700VertexSemanticFetch;
+//
+typedef struct R700VertexGenericFetch 
+{
+	ShaderInstType          m_ShaderInstType;
+	R700ShaderInstruction * pNextInst;
+	//R700VertexInstruction
+		//R700ShaderInstruction
+	GLuint m_uIndex;
+    GLuint useCount;
+		//---------------------
+	
+    sq_vtx_word0_u     m_Word0;
+    sq_vtx_word1_u     m_Word1;
+    sq_vtx_word2_u     m_Word2;
+
+    struct R700ControlFlowGenericClause *m_pLinkedGenericClause;
+	//---------------------------
+
+    sq_vtx_word1_gpr_u m_Word1_GPR;
+} R700VertexGenericFetch;
+
+// ------------------- End of Texture Vertex Instruction --------------------
+
+void Init_R700ControlFlowGenericClause(R700ControlFlowGenericClause* pInst);
+void Init_R700ControlFlowALUClause(R700ControlFlowALUClause* pInst);
+void Init_R700ControlFlowSXClause(R700ControlFlowSXClause* pInst);
+void Init_R700ControlFlowSMXClause(R700ControlFlowSMXClause* pInst);
+void Init_R700ALUInstruction(R700ALUInstruction* pInst);
+void Init_R700ALUInstructionHalfLiteral(R700ALUInstructionHalfLiteral* pInst, GLfloat x, GLfloat y);
+void Init_R700ALUInstructionFullLiteral(R700ALUInstructionFullLiteral* pInst, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+void Init_R700TextureInstruction(R700TextureInstruction* pInst);
+void Init_R700VertexSemanticFetch(R700VertexSemanticFetch* pInst);
+void Init_R700VertexGenericFetch(R700VertexGenericFetch* pInst);
+
+unsigned int GetInstructionSize(ShaderInstType instType);
+unsigned int GetCFMaxInstructions(ShaderInstType instType);
+
+GLboolean LinkVertexInstruction(R700ControlFlowGenericClause *pCFGeneric,
+								R700VertexInstruction *pVTXInstruction);
+
+#endif //_R700_SHADERINST_H_
diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c
new file mode 100644
index 0000000000..dd6ddaff3e
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_state.c
@@ -0,0 +1,1226 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/state.h"
+#include "main/imports.h"
+#include "main/enums.h"
+#include "main/macros.h"
+#include "main/dd.h"
+#include "main/simple_list.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "tnl/t_vp_build.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "main/api_arrayelt.h"
+#include "main/state.h"
+#include "main/framebuffer.h"
+
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+#include "vbo/vbo.h"
+#include "main/texformat.h"
+
+#include "r600_context.h"
+
+#include "r700_state.h"
+
+#include "r700_fragprog.h"
+#include "r700_vertprog.h"
+
+
+void r700SetDefaultStates(context_t *context) //--------------------
+{
+    
+}
+
+void r700UpdateShaders (GLcontext * ctx)  //----------------------------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+
+    GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
+    GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
+
+    struct r700_vertex_program *vp;
+	int i;
+
+    if (context->radeon.NewGLState) 
+    {
+        context->radeon.NewGLState = 0;
+
+        for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
+        {
+            /* mat states from state var not array for sw */
+            dummy_attrib[i].stride = 0;
+
+            temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
+            TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &(dummy_attrib[i]);
+        }
+
+        _tnl_UpdateFixedFunctionProgram(ctx);
+
+        for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
+        {
+            TNL_CONTEXT(ctx)->vb.AttribPtr[i] = temp_attrib[i];
+        }
+
+        r700SelectVertexShader(ctx);
+        vp = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+
+        if (vp->translated == GL_FALSE) 
+        {
+            // TODO
+            //fprintf(stderr, "Failing back to sw-tcl\n");
+            //hw_tcl_on = future_hw_tcl_on = 0;
+            //r300ResetHwState(rmesa);
+            //
+            r700UpdateStateParameters(ctx, _NEW_PROGRAM);
+            return;
+        }
+    }
+
+    r700UpdateStateParameters(ctx, _NEW_PROGRAM);
+}
+
+/*
+ * To correctly position primitives:
+ */
+void r700UpdateViewportOffset(GLcontext * ctx) //------------------
+{
+
+	//radeonUpdateScissor(ctx);
+
+    return;
+}
+
+/**
+ * Tell the card where to render (offset, pitch).
+ * Effected by glDrawBuffer, etc
+ */
+void r700UpdateDrawBuffer(GLcontext * ctx) /* TODO */ //---------------------
+{
+#if 0 /* to be enabled */
+    context_t *context = R700_CONTEXT(ctx);
+
+    switch (ctx->DrawBuffer->_ColorDrawBufferIndexes[0]) 
+    {
+	case BUFFER_FRONT_LEFT:
+	    context->target.rt = context->screen->frontBuffer;
+	    break;
+	case BUFFER_BACK_LEFT:
+	    context->target.rt = context->screen->backBuffer;
+	    break;
+	default:
+	    memset (&context->target.rt, sizeof(context->target.rt), 0);
+	}
+#endif /* to be enabled */
+}
+
+static void r700FetchStateParameter(GLcontext * ctx,
+			                        const gl_state_index state[STATE_LENGTH],
+			                        GLfloat * value)
+{
+	context_t *context = R700_CONTEXT(ctx);
+
+    /* TODO */
+}
+
+void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state) //--------------------
+{
+	struct r700_fragment_program *fp;
+	struct gl_program_parameter_list *paramList;
+	GLuint i;
+
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM)))
+		return;
+
+	fp = (struct r700_fragment_program *)ctx->FragmentProgram._Current;
+	if (!fp)
+    {
+		return;
+    }
+
+	paramList = fp->mesa_program.Base.Parameters;
+
+	if (!paramList)
+    {
+		return;
+    }
+
+	for (i = 0; i < paramList->NumParameters; i++) 
+    {
+		if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) 
+        {
+			r700FetchStateParameter(ctx,
+						paramList->Parameters[i].
+						StateIndexes,
+						paramList->ParameterValues[i]);
+		}
+	}
+}
+
+/**
+ * Called by Mesa after an internal state update.
+ */
+static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //-------------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    _swrast_InvalidateState(ctx, new_state);
+	_swsetup_InvalidateState(ctx, new_state);
+	_vbo_InvalidateState(ctx, new_state);
+	_tnl_InvalidateState(ctx, new_state);
+	_ae_invalidate_state(ctx, new_state);
+
+	if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) 
+    {
+        _mesa_update_framebuffer(ctx);
+		/* this updates the DrawBuffer's Width/Height if it's a FBO */
+		_mesa_update_draw_buffer_bounds(ctx);
+
+		r700UpdateDrawBuffer(ctx);
+	}
+
+	r700UpdateStateParameters(ctx, new_state);
+
+    if(GL_TRUE == r700->bEnablePerspective)
+    {
+        /* Do scale XY and Z by 1/W0 for perspective correction on pos. For orthogonal case, set both to one. */
+        CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit);
+        CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit);
+
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit);
+
+        SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
+        CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
+    }
+    else
+    {
+        /* For orthogonal case. */
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit);
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit);
+
+        SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit);
+
+        CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
+        SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
+    }
+
+	context->radeon.NewGLState |= new_state;
+}
+
+static void r700SetDepthState(GLcontext * ctx)
+{
+	context_t *context = R700_CONTEXT(ctx);
+
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    if (ctx->Depth.Test)
+    {
+        SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit);
+        if (ctx->Depth.Mask)
+        {
+            SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit);
+        }
+        else
+        {
+            CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit);
+        }
+
+        switch (ctx->Depth.Func)
+        {
+        case GL_NEVER:            
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_NEVER, 
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_LESS:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_LESS, 
+                     ZFUNC_shift, ZFUNC_mask);            
+            break;
+        case GL_EQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_EQUAL, 
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_LEQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_LEQUAL,  
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_GREATER:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_GREATER,  
+                     ZFUNC_shift, ZFUNC_mask);           
+            break;
+        case GL_NOTEQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_NOTEQUAL,  
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_GEQUAL:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_GEQUAL,  
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        case GL_ALWAYS:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_ALWAYS,  
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        default:
+            SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_ALWAYS,  
+                     ZFUNC_shift, ZFUNC_mask);
+            break;
+        }
+    }
+    else
+    {
+        CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit);
+        CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit);
+    }
+}
+
+static void r700AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) //---------------
+{
+}
+
+
+static void r700BlendColor(GLcontext * ctx, const GLfloat cf[4]) //----------------
+{
+}
+
+static void r700BlendEquationSeparate(GLcontext * ctx,
+				                      GLenum modeRGB, GLenum modeA) //-----------------
+{
+}
+
+static void r700BlendFuncSeparate(GLcontext * ctx,
+				  GLenum sfactorRGB, GLenum dfactorRGB,
+				  GLenum sfactorA, GLenum dfactorA) //------------------------
+{
+}
+
+static void r700UpdateCulling(GLcontext * ctx)
+{
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw);
+
+    CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit);
+    CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+    CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+
+    if (ctx->Polygon.CullFlag) 
+    {
+        switch (ctx->Polygon.CullFaceMode) 
+        {
+        case GL_FRONT:            
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        case GL_BACK:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        case GL_FRONT_AND_BACK:
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        default:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit);
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_BACK_bit);
+            break;
+        }
+    }
+
+    switch (ctx->Polygon.FrontFace) 
+    {
+        case GL_CW:
+            SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit);
+            break;
+        case GL_CCW:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); 
+            break;
+        default:
+            CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); /* default: ccw */
+            break;
+    }
+}
+
+static void r700Enable(GLcontext * ctx, GLenum cap, GLboolean state) //------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+
+	switch (cap) {
+	case GL_TEXTURE_1D:
+	case GL_TEXTURE_2D:
+	case GL_TEXTURE_3D:
+		/* empty */
+		break;
+	case GL_FOG:
+		/* empty */
+		break;
+	case GL_ALPHA_TEST:
+		//r700SetAlphaState(ctx);
+		break;
+	case GL_COLOR_LOGIC_OP:
+		//r700SetLogicOpState(ctx);
+		/* fall-through, because logic op overrides blending */
+	case GL_BLEND:
+		//r700SetBlendState(ctx);
+		break;
+	case GL_CLIP_PLANE0:
+	case GL_CLIP_PLANE1:
+	case GL_CLIP_PLANE2:
+	case GL_CLIP_PLANE3:
+	case GL_CLIP_PLANE4:
+	case GL_CLIP_PLANE5:
+		//r700SetClipPlaneState(ctx, cap, state);
+		break;
+	case GL_DEPTH_TEST:
+		r700SetDepthState(ctx);
+		break;
+	case GL_STENCIL_TEST:
+		//r700SetStencilState(ctx, state);
+		break;
+	case GL_CULL_FACE:
+		r700UpdateCulling(ctx);
+		break;
+	case GL_POLYGON_OFFSET_POINT:
+	case GL_POLYGON_OFFSET_LINE:
+	case GL_POLYGON_OFFSET_FILL:
+		//r700SetPolygonOffsetState(ctx, state);
+		break;
+	case GL_SCISSOR_TEST:
+		radeon_firevertices(&context->radeon);
+		context->radeon.state.scissor.enabled = state;
+		radeonUpdateScissor(ctx);
+		break;
+	default:
+		break;
+	}
+
+}
+
+/**
+ * Handle glColorMask()
+ */
+static void r700ColorMask(GLcontext * ctx,
+			  GLboolean r, GLboolean g, GLboolean b, GLboolean a) //------------------
+{
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw);
+	unsigned int mask = ((r ? 1 : 0) |
+			     (g ? 2 : 0) |
+			     (b ? 4 : 0) |
+			     (a ? 8 : 0));
+
+	if (mask != r700->CB_SHADER_MASK.u32All)
+		SETfield(r700->CB_SHADER_MASK.u32All, mask, OUTPUT0_ENABLE_shift, OUTPUT0_ENABLE_mask);
+}
+
+/**
+ * Change the depth testing function.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700DepthFunc(GLcontext * ctx, GLenum func) //--------------------
+{
+    r700SetDepthState(ctx);
+}
+
+/**
+ * Enable/Disable depth writing.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700DepthMask(GLcontext * ctx, GLboolean mask) //------------------
+{
+    r700SetDepthState(ctx);
+}
+
+/**
+ * Change the culling mode.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700CullFace(GLcontext * ctx, GLenum mode) //-----------------
+{
+    r700UpdateCulling(ctx);
+}
+
+/* =============================================================
+ * Fog
+ */
+static void r700Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param) //--------------
+{
+}
+
+/**
+ * Change the polygon orientation.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r700FrontFace(GLcontext * ctx, GLenum mode) //------------------
+{
+    r700UpdateCulling(ctx);
+}
+
+static void r700ShadeModel(GLcontext * ctx, GLenum mode) //--------------------
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	/* also need to set/clear FLAT_SHADE bit per param in SPI_PS_INPUT_CNTL_[0-31] */
+	switch (mode) {
+	case GL_FLAT:
+		SETbit(r700->SPI_INTERP_CONTROL_0.u32All, FLAT_SHADE_ENA_bit);
+		break;
+	case GL_SMOOTH:
+		CLEARbit(r700->SPI_INTERP_CONTROL_0.u32All, FLAT_SHADE_ENA_bit);
+		break;
+	default:
+		return;
+	}
+}
+
+static void r700PointParameter(GLcontext * ctx, GLenum pname, const GLfloat * param) //---------------
+{
+}
+
+static void r700StencilFuncSeparate(GLcontext * ctx, GLenum face,
+				    GLenum func, GLint ref, GLuint mask) //---------------------
+{
+}
+
+
+static void r700StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) //--------------
+{
+}
+
+static void r700StencilOpSeparate(GLcontext * ctx, GLenum face,
+				  GLenum fail, GLenum zfail, GLenum zpass) //--------------------
+{
+}
+
+static void r700UpdateWindow(GLcontext * ctx, int id) //--------------------
+{
+
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&context->radeon);
+	GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+	GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
+	const GLfloat *v = ctx->Viewport._WindowMap.m;
+	const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+	const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
+	GLfloat y_scale, y_bias;
+
+	if (render_to_fbo) {
+		y_scale = 1.0;
+		y_bias = 0;
+	} else {
+		y_scale = -1.0;
+		y_bias = yoffset;
+	}
+
+	GLfloat sx = v[MAT_SX];
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat sy = v[MAT_SY] * y_scale;
+	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias;
+	GLfloat sz = v[MAT_SZ] * depthScale;
+	GLfloat tz = v[MAT_TZ] * depthScale;
+
+	/* TODO : Need DMA flush as well. */
+
+	r700->viewport[id].PA_CL_VPORT_XSCALE.f32All  = sx;
+	r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx;
+
+	r700->viewport[id].PA_CL_VPORT_YSCALE.f32All  = sy;
+	r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All = ty;
+
+	r700->viewport[id].PA_CL_VPORT_ZSCALE.f32All  = sz;
+	r700->viewport[id].PA_CL_VPORT_ZOFFSET.f32All = tz;
+
+	r700->viewport[id].enabled = GL_TRUE;
+
+	r700SetScissor(context);
+}
+
+
+static void r700Viewport(GLcontext * ctx,
+                         GLint x,
+                         GLint y,
+			 GLsizei width,
+                         GLsizei height) //--------------------
+{
+	r700UpdateWindow(ctx, 0);
+
+	radeon_viewport(ctx, x, y, width, height);
+}
+
+static void r700DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) //-------------
+{
+	r700UpdateWindow(ctx, 0);
+}
+
+static void r700PointSize(GLcontext * ctx, GLfloat size) //-------------------
+{
+}
+
+static void r700LineWidth(GLcontext * ctx, GLfloat widthf) //---------------
+{
+}
+
+static void r700PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) //--------------
+{
+}
+
+
+static void r700PolygonMode(GLcontext * ctx, GLenum face, GLenum mode) //------------------
+{
+}
+ 
+static void r700RenderMode(GLcontext * ctx, GLenum mode) //---------------------
+{
+}
+
+static void r700ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq ) //-----------------
+{
+}
+
+void r700SetScissor(context_t *context) //---------------
+{
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	unsigned x1, y1, x2, y2;
+	int id = 0;
+	struct radeon_renderbuffer *rrb;
+
+	rrb = radeon_get_colorbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		return;
+	}
+	if (context->radeon.state.scissor.enabled) {
+		x1 = context->radeon.state.scissor.rect.x1;
+		y1 = context->radeon.state.scissor.rect.y1;
+		x2 = context->radeon.state.scissor.rect.x2 - 1;
+		y2 = context->radeon.state.scissor.rect.y2 - 1;
+	} else {
+		x1 = rrb->dPriv->x;
+		y1 = rrb->dPriv->y;
+		x2 = rrb->dPriv->x + rrb->dPriv->w;
+		y2 = rrb->dPriv->y + rrb->dPriv->h;
+	}
+
+	/* window */
+	SETbit(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, x1,
+		 PA_SC_WINDOW_SCISSOR_TL__TL_X_shift, PA_SC_WINDOW_SCISSOR_TL__TL_X_mask);
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, y1,
+		 PA_SC_WINDOW_SCISSOR_TL__TL_Y_shift, PA_SC_WINDOW_SCISSOR_TL__TL_Y_mask);
+
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, x2,
+		 PA_SC_WINDOW_SCISSOR_BR__BR_X_shift, PA_SC_WINDOW_SCISSOR_BR__BR_X_mask);
+	SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, y2,
+		 PA_SC_WINDOW_SCISSOR_BR__BR_Y_shift, PA_SC_WINDOW_SCISSOR_BR__BR_Y_mask);
+
+
+	SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, x1,
+		 PA_SC_CLIPRECT_0_TL__TL_X_shift, PA_SC_CLIPRECT_0_TL__TL_X_mask);
+	SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, y1,
+		 PA_SC_CLIPRECT_0_TL__TL_Y_shift, PA_SC_CLIPRECT_0_TL__TL_Y_mask);
+	SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, x2,
+		 PA_SC_CLIPRECT_0_BR__BR_X_shift, PA_SC_CLIPRECT_0_BR__BR_X_mask);
+	SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, y2,
+		 PA_SC_CLIPRECT_0_BR__BR_Y_shift, PA_SC_CLIPRECT_0_BR__BR_Y_mask);
+
+	r700->PA_SC_CLIPRECT_1_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All;
+	r700->PA_SC_CLIPRECT_1_BR.u32All = r700->PA_SC_CLIPRECT_0_BR.u32All;
+	r700->PA_SC_CLIPRECT_2_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All;
+	r700->PA_SC_CLIPRECT_2_BR.u32All = r700->PA_SC_CLIPRECT_0_BR.u32All;
+	r700->PA_SC_CLIPRECT_3_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All;
+	r700->PA_SC_CLIPRECT_3_BR.u32All = r700->PA_SC_CLIPRECT_0_BR.u32All;
+
+	/* more....2d clip */
+	SETbit(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, x1,
+		 PA_SC_GENERIC_SCISSOR_TL__TL_X_shift, PA_SC_GENERIC_SCISSOR_TL__TL_X_mask);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, y1,
+		 PA_SC_GENERIC_SCISSOR_TL__TL_Y_shift, PA_SC_GENERIC_SCISSOR_TL__TL_Y_mask);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, x2,
+		 PA_SC_GENERIC_SCISSOR_BR__BR_X_shift, PA_SC_GENERIC_SCISSOR_BR__BR_X_mask);
+	SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, y2,
+		 PA_SC_GENERIC_SCISSOR_BR__BR_Y_shift, PA_SC_GENERIC_SCISSOR_BR__BR_Y_mask);
+
+	SETbit(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All, WINDOW_OFFSET_DISABLE_bit);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All, x1,
+		 PA_SC_VPORT_SCISSOR_0_TL__TL_X_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_X_mask);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_TL.u32All, y1,
+		 PA_SC_VPORT_SCISSOR_0_TL__TL_Y_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_Y_mask);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All, x2,
+		 PA_SC_VPORT_SCISSOR_0_BR__BR_X_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_X_mask);
+	SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All, y2,
+		 PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask);
+
+	r700->viewport[id].PA_SC_VPORT_ZMIN_0.u32All = 0;
+	r700->viewport[id].PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000;
+	r700->viewport[id].enabled = GL_TRUE;
+}
+
+void r700SetRenderTarget(context_t *context, int id)
+{
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    struct radeon_renderbuffer *rrb;
+    unsigned int nPitchInPixel;
+
+    /* screen/window/view */
+    SETfield(r700->CB_TARGET_MASK.u32All, 0xF, (4 * id), TARGET0_ENABLE_mask);
+
+    rrb = radeon_get_colorbuffer(&context->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return;
+	}
+
+    /* color buffer */
+    r700->render_target[id].CB_COLOR0_BASE.u32All = context->radeon.state.color.draw_offset;
+
+    nPitchInPixel = rrb->pitch/rrb->cpp;
+    SETfield(r700->render_target[id].CB_COLOR0_SIZE.u32All, (nPitchInPixel/8)-1,
+             PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask);
+    SETfield(r700->render_target[id].CB_COLOR0_SIZE.u32All, ( (nPitchInPixel * context->radeon.radeonScreen->driScreen->fbHeight)/64 )-1,
+             SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask);
+    r700->render_target[id].CB_COLOR0_BASE.u32All = 0;
+    SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, ENDIAN_NONE, ENDIAN_shift, ENDIAN_mask);
+    SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, ARRAY_LINEAR_GENERAL,
+             CB_COLOR0_INFO__ARRAY_MODE_shift, CB_COLOR0_INFO__ARRAY_MODE_mask);
+    if(4 == rrb->cpp)
+    {
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, COLOR_8_8_8_8,
+                 CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask);
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, SWAP_ALT, COMP_SWAP_shift, COMP_SWAP_mask);
+    }
+    else
+    {
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, COLOR_5_6_5,
+                 CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask);
+        SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, SWAP_ALT_REV,
+                 COMP_SWAP_shift, COMP_SWAP_mask);
+    }
+    SETbit(r700->render_target[id].CB_COLOR0_INFO.u32All, SOURCE_FORMAT_bit);
+    SETbit(r700->render_target[id].CB_COLOR0_INFO.u32All, BLEND_CLAMP_bit);
+    SETfield(r700->render_target[id].CB_COLOR0_INFO.u32All, NUMBER_UNORM, NUMBER_TYPE_shift, NUMBER_TYPE_mask);
+
+    CLEARfield(r700->render_target[id].CB_BLEND0_CONTROL.u32All, COLOR_SRCBLEND_mask); /* no dst blend */
+    CLEARfield(r700->render_target[id].CB_BLEND0_CONTROL.u32All, ALPHA_SRCBLEND_mask); /* no dst blend */
+
+    r700->render_target[id].enabled = GL_TRUE;
+}
+
+void r700SetDepthTarget(context_t *context)
+{
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    struct radeon_renderbuffer *rrb;
+    unsigned int nPitchInPixel;
+
+    /* depth buf */
+    r700->DB_DEPTH_SIZE.u32All = 0;
+    r700->DB_DEPTH_BASE.u32All = 0;
+    r700->DB_DEPTH_INFO.u32All = 0;
+
+    r700->DB_DEPTH_CLEAR.u32All     = 0x3F800000;
+    r700->DB_DEPTH_VIEW.u32All      = 0;
+    r700->DB_RENDER_CONTROL.u32All  = 0;
+    SETbit(r700->DB_RENDER_CONTROL.u32All, STENCIL_COMPRESS_DISABLE_bit);
+    SETbit(r700->DB_RENDER_CONTROL.u32All, DEPTH_COMPRESS_DISABLE_bit);
+    r700->DB_RENDER_OVERRIDE.u32All = 0;
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+	    SETbit(r700->DB_RENDER_OVERRIDE.u32All, FORCE_SHADER_Z_ORDER_bit);
+    SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIZ_ENABLE_shift, FORCE_HIZ_ENABLE_mask);
+    SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE0_shift, FORCE_HIS_ENABLE0_mask);
+    SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE1_shift, FORCE_HIS_ENABLE1_mask);
+
+    r700->DB_ALPHA_TO_MASK.u32All = 0;
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET0_shift, ALPHA_TO_MASK_OFFSET0_mask);
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET1_shift, ALPHA_TO_MASK_OFFSET1_mask);
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET2_shift, ALPHA_TO_MASK_OFFSET2_mask);
+    SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET3_shift, ALPHA_TO_MASK_OFFSET3_mask);
+
+    rrb = radeon_get_depthbuffer(&context->radeon);
+	if (!rrb)
+		return;
+
+    nPitchInPixel = rrb->pitch/rrb->cpp;
+
+    SETfield(r700->DB_DEPTH_SIZE.u32All, (nPitchInPixel/8)-1,
+             PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask);
+    SETfield(r700->DB_DEPTH_SIZE.u32All, ( (nPitchInPixel * context->radeon.radeonScreen->driScreen->fbHeight)/64 )-1,
+             SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); /* size in pixel / 64 - 1 */
+
+    if(4 == rrb->cpp)
+    {
+        switch (GL_CONTEXT(context)->Visual.depthBits)
+        {
+        case 16:
+        case 24:
+            SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_8_24,
+                     DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask);
+            break;
+        default:
+            fprintf(stderr, "Error: Unsupported depth %d... exiting\n",
+                GL_CONTEXT(context)->Visual.depthBits);
+            _mesa_exit(-1);
+        }
+    }
+    else
+    {
+        SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_16,
+                     DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask);
+    }
+    SETfield(r700->DB_DEPTH_INFO.u32All, ARRAY_2D_TILED_THIN1,
+             DB_DEPTH_INFO__ARRAY_MODE_shift, DB_DEPTH_INFO__ARRAY_MODE_mask);
+    /* r700->DB_PREFETCH_LIMIT.bits.DEPTH_HEIGHT_TILE_MAX = (context->currentDraw->h >> 3) - 1; */ /* z buffer sie may much bigger than what need, so use actual used h. */
+}
+
+static void r700InitSQConfig(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+    int ps_prio;
+    int vs_prio;
+    int gs_prio;
+    int es_prio;
+    int num_ps_gprs;
+    int num_vs_gprs;
+    int num_gs_gprs;
+    int num_es_gprs;
+    int num_temp_gprs;
+    int num_ps_threads;
+    int num_vs_threads;
+    int num_gs_threads;
+    int num_es_threads;
+    int num_ps_stack_entries;
+    int num_vs_stack_entries;
+    int num_gs_stack_entries;
+    int num_es_stack_entries;
+
+    // SQ
+    ps_prio = 0;
+    vs_prio = 1;
+    gs_prio = 2;
+    es_prio = 3;
+    switch (context->radeon.radeonScreen->chip_family) {
+    case CHIP_FAMILY_R600:
+	    num_ps_gprs = 192;
+	    num_vs_gprs = 56;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 136;
+	    num_vs_threads = 48;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 128;
+	    num_vs_stack_entries = 128;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    case CHIP_FAMILY_RV630:
+    case CHIP_FAMILY_RV635:
+	    num_ps_gprs = 84;
+	    num_vs_gprs = 36;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 144;
+	    num_vs_threads = 40;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 40;
+	    num_vs_stack_entries = 40;
+	    num_gs_stack_entries = 32;
+	    num_es_stack_entries = 16;
+	    break;
+    case CHIP_FAMILY_RV610:
+    case CHIP_FAMILY_RV620:
+    case CHIP_FAMILY_RS780:
+    default:
+	    num_ps_gprs = 84;
+	    num_vs_gprs = 36;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 136;
+	    num_vs_threads = 48;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 40;
+	    num_vs_stack_entries = 40;
+	    num_gs_stack_entries = 32;
+	    num_es_stack_entries = 16;
+	    break;
+    case CHIP_FAMILY_RV670:
+	    num_ps_gprs = 144;
+	    num_vs_gprs = 40;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 136;
+	    num_vs_threads = 48;
+	    num_gs_threads = 4;
+	    num_es_threads = 4;
+	    num_ps_stack_entries = 40;
+	    num_vs_stack_entries = 40;
+	    num_gs_stack_entries = 32;
+	    num_es_stack_entries = 16;
+	    break;
+    case CHIP_FAMILY_RV770:
+	    num_ps_gprs = 192;
+	    num_vs_gprs = 56;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 188;
+	    num_vs_threads = 60;
+	    num_gs_threads = 0;
+	    num_es_threads = 0;
+	    num_ps_stack_entries = 256;
+	    num_vs_stack_entries = 256;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    case CHIP_FAMILY_RV730:
+    case CHIP_FAMILY_RV740:
+	    num_ps_gprs = 84;
+	    num_vs_gprs = 36;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 188;
+	    num_vs_threads = 60;
+	    num_gs_threads = 0;
+	    num_es_threads = 0;
+	    num_ps_stack_entries = 128;
+	    num_vs_stack_entries = 128;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    case CHIP_FAMILY_RV710:
+	    num_ps_gprs = 192;
+	    num_vs_gprs = 56;
+	    num_temp_gprs = 4;
+	    num_gs_gprs = 0;
+	    num_es_gprs = 0;
+	    num_ps_threads = 144;
+	    num_vs_threads = 48;
+	    num_gs_threads = 0;
+	    num_es_threads = 0;
+	    num_ps_stack_entries = 128;
+	    num_vs_stack_entries = 128;
+	    num_gs_stack_entries = 0;
+	    num_es_stack_entries = 0;
+	    break;
+    }
+
+    r700->sq_config.SQ_CONFIG.u32All = 0;
+    if ((context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV610) ||
+        (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV620) ||
+	(context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RS780) ||
+        (context->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV710))
+	    CLEARbit(r700->sq_config.SQ_CONFIG.u32All, VC_ENABLE_bit);
+    else
+	    SETbit(r700->sq_config.SQ_CONFIG.u32All, VC_ENABLE_bit);
+    SETbit(r700->sq_config.SQ_CONFIG.u32All, DX9_CONSTS_bit);
+    SETbit(r700->sq_config.SQ_CONFIG.u32All, ALU_INST_PREFER_VECTOR_bit);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, PS_PRIO_shift, PS_PRIO_mask);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, VS_PRIO_shift, VS_PRIO_mask);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, GS_PRIO_shift, GS_PRIO_mask);
+    SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, ES_PRIO_shift, ES_PRIO_mask);
+
+    r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All = 0;
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_ps_gprs, NUM_PS_GPRS_shift, NUM_PS_GPRS_mask);
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_vs_gprs, NUM_VS_GPRS_shift, NUM_VS_GPRS_mask);
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_temp_gprs,
+	     NUM_CLAUSE_TEMP_GPRS_shift, NUM_CLAUSE_TEMP_GPRS_mask);
+
+    r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All = 0;
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All, num_gs_gprs, NUM_GS_GPRS_shift, NUM_GS_GPRS_mask);
+    SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_2.u32All, num_es_gprs, NUM_ES_GPRS_shift, NUM_ES_GPRS_mask);
+
+    r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All = 0;
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_ps_threads,
+	     NUM_PS_THREADS_shift, NUM_PS_THREADS_mask);
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_vs_threads,
+	     NUM_VS_THREADS_shift, NUM_VS_THREADS_mask);
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_gs_threads,
+	     NUM_GS_THREADS_shift, NUM_GS_THREADS_mask);
+    SETfield(r700->sq_config.SQ_THREAD_RESOURCE_MGMT.u32All, num_es_threads,
+	     NUM_ES_THREADS_shift, NUM_ES_THREADS_mask);
+
+    r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All = 0;
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All, num_ps_stack_entries,
+	     NUM_PS_STACK_ENTRIES_shift, NUM_PS_STACK_ENTRIES_mask);
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_1.u32All, num_vs_stack_entries,
+	     NUM_VS_STACK_ENTRIES_shift, NUM_VS_STACK_ENTRIES_mask);
+
+    r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All = 0;
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All, num_gs_stack_entries,
+	     NUM_GS_STACK_ENTRIES_shift, NUM_GS_STACK_ENTRIES_mask);
+    SETfield(r700->sq_config.SQ_STACK_RESOURCE_MGMT_2.u32All, num_es_stack_entries,
+	     NUM_ES_STACK_ENTRIES_shift, NUM_ES_STACK_ENTRIES_mask);
+
+}
+
+/**
+ * Calculate initial hardware state and register state functions.
+ * Assumes that the command buffer and state atoms have been
+ * initialized already.
+ */
+void r700InitState(GLcontext * ctx) //-------------------
+{
+    context_t *context = R700_CONTEXT(ctx);
+
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    r700->TA_CNTL_AUX.u32All = 0;
+    SETfield(r700->TA_CNTL_AUX.u32All, 28, TD_FIFO_CREDIT_shift, TD_FIFO_CREDIT_mask);
+    r700->VC_ENHANCE.u32All = 0;
+    r700->DB_WATERMARKS.u32All = 0;
+    SETfield(r700->DB_WATERMARKS.u32All, 4, DEPTH_FREE_shift, DEPTH_FREE_mask);
+    SETfield(r700->DB_WATERMARKS.u32All, 16, DEPTH_FLUSH_shift, DEPTH_FLUSH_mask);
+    SETfield(r700->DB_WATERMARKS.u32All, 0, FORCE_SUMMARIZE_shift, FORCE_SUMMARIZE_mask);
+    SETfield(r700->DB_WATERMARKS.u32All, 4, DEPTH_PENDING_FREE_shift, DEPTH_PENDING_FREE_mask);
+    r700->SQ_DYN_GPR_CNTL_PS_FLUSH_REQ.u32All = 0;
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) {
+	    SETfield(r700->TA_CNTL_AUX.u32All, 3, GRADIENT_CREDIT_shift, GRADIENT_CREDIT_mask);
+	    r700->DB_DEBUG.u32All = 0x82000000;
+	    SETfield(r700->DB_WATERMARKS.u32All, 16, DEPTH_CACHELINE_FREE_shift, DEPTH_CACHELINE_FREE_mask);
+    } else {
+	    SETfield(r700->TA_CNTL_AUX.u32All, 2, GRADIENT_CREDIT_shift, GRADIENT_CREDIT_mask);
+	    SETfield(r700->DB_WATERMARKS.u32All, 4, DEPTH_CACHELINE_FREE_shift, DEPTH_CACHELINE_FREE_mask);
+	    SETbit(r700->SQ_DYN_GPR_CNTL_PS_FLUSH_REQ.u32All, VS_PC_LIMIT_ENABLE_bit);
+    }
+
+    /* Turn off vgt reuse */
+    r700->VGT_REUSE_OFF.u32All = 0;
+    SETbit(r700->VGT_REUSE_OFF.u32All, REUSE_OFF_bit);
+
+    /* Specify offsetting and clamp values for vertices */
+    r700->VGT_MAX_VTX_INDX.u32All      = 0xFFFFFF;
+    r700->VGT_MIN_VTX_INDX.u32All      = 0;
+    r700->VGT_INDX_OFFSET.u32All    = 0;
+
+    /* Specify the number of instances */
+    r700->VGT_DMA_NUM_INSTANCES.u32All = 1;
+
+    /* not alpha blend */
+    CLEARfield(r700->SX_ALPHA_TEST_CONTROL.u32All, ALPHA_FUNC_mask);
+    CLEARbit(r700->SX_ALPHA_TEST_CONTROL.u32All, ALPHA_TEST_ENABLE_bit);
+
+    /* default shader connections. */
+    r700->SPI_VS_OUT_ID_0.u32All  = 0x03020100;
+    r700->SPI_VS_OUT_ID_1.u32All  = 0x07060504;
+
+    r700->SPI_PS_INPUT_CNTL_0.u32All  = 0x00000800;
+    r700->SPI_PS_INPUT_CNTL_1.u32All  = 0x00000801;
+    r700->SPI_PS_INPUT_CNTL_2.u32All  = 0x00000802;
+
+    r700->SPI_THREAD_GROUPING.u32All = 0;
+    if (context->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV770)
+	    SETfield(r700->SPI_THREAD_GROUPING.u32All, 1, PS_GROUPING_shift, PS_GROUPING_mask);
+
+    SETfield(r700->CB_COLOR_CONTROL.u32All, 0xCC, ROP3_shift, ROP3_mask);
+    CLEARbit(r700->CB_COLOR_CONTROL.u32All, PER_MRT_BLEND_bit);
+
+    r700->DB_SHADER_CONTROL.u32All = 0;
+    SETbit(r700->DB_SHADER_CONTROL.u32All, DUAL_EXPORT_ENABLE_bit);
+
+    /* Set up the culling control register */
+    SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_TRIANGLES,
+             POLYMODE_FRONT_PTYPE_shift, POLYMODE_FRONT_PTYPE_mask);
+    SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DRAW_TRIANGLES,
+             POLYMODE_BACK_PTYPE_shift, POLYMODE_BACK_PTYPE_mask);
+
+    /* screen */
+    r700->PA_SC_SCREEN_SCISSOR_TL.u32All = 0x0;
+
+    SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All,
+	     ((RADEONDRIPtr)(context->radeon.radeonScreen->driScreen->pDevPriv))->width,
+	     PA_SC_SCREEN_SCISSOR_BR__BR_X_shift, PA_SC_SCREEN_SCISSOR_BR__BR_X_mask);
+    SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All,
+	     ((RADEONDRIPtr)(context->radeon.radeonScreen->driScreen->pDevPriv))->height,
+	     PA_SC_SCREEN_SCISSOR_BR__BR_Y_shift, PA_SC_SCREEN_SCISSOR_BR__BR_Y_mask);
+
+    /* 4 clip rectangles */ /* TODO : set these clip rects according to context->currentDraw->numClipRects */
+    r700->PA_SC_CLIPRECT_RULE.u32All = 0;
+    SETfield(r700->PA_SC_CLIPRECT_RULE.u32All, CLIP_RULE_mask, CLIP_RULE_shift, CLIP_RULE_mask);
+
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+	    r700->PA_SC_EDGERULE.u32All = 0;
+    else
+	    r700->PA_SC_EDGERULE.u32All = 0xAAAAAAAA;
+
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) {
+	    r700->PA_SC_MODE_CNTL.u32All = 0;
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, WALK_ORDER_ENABLE_bit);
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, FORCE_EOV_CNTDWN_ENABLE_bit);
+    } else {
+	    r700->PA_SC_MODE_CNTL.u32All = 0x00500000;
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, FORCE_EOV_REZ_ENABLE_bit);
+	    SETbit(r700->PA_SC_MODE_CNTL.u32All, FORCE_EOV_CNTDWN_ENABLE_bit);
+    }
+
+    /* Do scale XY and Z by 1/W0. */
+    r700->bEnablePerspective = GL_TRUE;
+    CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit);
+    CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit);
+
+    /* Enable viewport scaling for all three axis */
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_X_SCALE_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_X_OFFSET_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Y_SCALE_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Y_OFFSET_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Z_SCALE_ENA_bit);
+    SETbit(r700->PA_CL_VTE_CNTL.u32All, VPORT_Z_OFFSET_ENA_bit);
+
+    /* Set up point sizes and min/max values */
+    SETfield(r700->PA_SU_POINT_SIZE.u32All, 0x8,
+             PA_SU_POINT_SIZE__HEIGHT_shift, PA_SU_POINT_SIZE__HEIGHT_mask);
+    SETfield(r700->PA_SU_POINT_SIZE.u32All, 0x8,
+             PA_SU_POINT_SIZE__WIDTH_shift, PA_SU_POINT_SIZE__WIDTH_mask);
+    CLEARfield(r700->PA_SU_POINT_MINMAX.u32All, MIN_SIZE_mask);
+    SETfield(r700->PA_SU_POINT_MINMAX.u32All, 0x8000, MAX_SIZE_shift, MAX_SIZE_mask);
+
+    /* Set up line control */
+    SETfield(r700->PA_SU_LINE_CNTL.u32All, 0x8,
+             PA_SU_LINE_CNTL__WIDTH_shift, PA_SU_LINE_CNTL__WIDTH_mask);
+
+    r700->PA_SC_LINE_CNTL.u32All = 0;
+    CLEARbit(r700->PA_SC_LINE_CNTL.u32All, EXPAND_LINE_WIDTH_bit);
+    SETbit(r700->PA_SC_LINE_CNTL.u32All, LAST_PIXEL_bit);
+
+    /* Set up vertex control */
+    r700->PA_SU_VTX_CNTL.u32All = 0;
+    CLEARfield(r700->PA_SU_VTX_CNTL.u32All, QUANT_MODE_mask);
+    SETbit(r700->PA_SU_VTX_CNTL.u32All, PIX_CENTER_bit);
+    SETfield(r700->PA_SU_VTX_CNTL.u32All, X_ROUND_TO_EVEN,
+             PA_SU_VTX_CNTL__ROUND_MODE_shift, PA_SU_VTX_CNTL__ROUND_MODE_mask);
+
+    /* to 1.0 = no guard band */
+    r700->PA_CL_GB_VERT_CLIP_ADJ.u32All  = 0x3F800000;  /* 1.0 */
+    r700->PA_CL_GB_VERT_DISC_ADJ.u32All  = 0x3F800000;
+    r700->PA_CL_GB_HORZ_CLIP_ADJ.u32All  = 0x3F800000;
+    r700->PA_CL_GB_HORZ_DISC_ADJ.u32All  = 0x3F800000;
+
+    /* CB */
+    r700->CB_CLEAR_RED_R6XX.f32All = 1.0; //r6xx only
+    r700->CB_CLEAR_GREEN_R6XX.f32All = 0.0; //r6xx only
+    r700->CB_CLEAR_BLUE_R6XX.f32All = 1.0; //r6xx only
+    r700->CB_CLEAR_ALPHA_R6XX.f32All = 1.0; //r6xx only
+    r700->CB_FOG_RED_R6XX.u32All = 0; //r6xx only
+    r700->CB_FOG_GREEN_R6XX.u32All = 0; //r6xx only
+    r700->CB_FOG_BLUE_R6XX.u32All = 0; //r6xx only
+
+    r700->CB_BLEND_RED.u32All = 0;
+    r700->CB_BLEND_GREEN.u32All = 0;
+    r700->CB_BLEND_BLUE.u32All = 0;
+    r700->CB_BLEND_ALPHA.u32All = 0;
+
+    r700->CB_BLEND_CONTROL.u32All = 0;
+
+    /* Disable color compares */
+    SETfield(r700->CB_CLRCMP_CONTROL.u32All, CLRCMP_DRAW_ALWAYS,
+             CLRCMP_FCN_SRC_shift, CLRCMP_FCN_SRC_mask);
+    SETfield(r700->CB_CLRCMP_CONTROL.u32All, CLRCMP_DRAW_ALWAYS,
+             CLRCMP_FCN_DST_shift, CLRCMP_FCN_DST_mask);
+    SETfield(r700->CB_CLRCMP_CONTROL.u32All, CLRCMP_SEL_SRC,
+             CLRCMP_FCN_SEL_shift, CLRCMP_FCN_SEL_mask);
+
+    /* Zero out source */
+    r700->CB_CLRCMP_SRC.u32All = 0x00000000;
+
+    /* Put a compare color in for error checking */
+    r700->CB_CLRCMP_DST.u32All = 0x000000FF;
+
+    /* Set up color compare mask */
+    r700->CB_CLRCMP_MSK.u32All = 0xFFFFFFFF;
+
+    /* default color mask */
+    SETfield(r700->CB_SHADER_MASK.u32All, 0xF, OUTPUT0_ENABLE_shift, OUTPUT0_ENABLE_mask);
+
+    /* Enable all samples for multi-sample anti-aliasing */
+    r700->PA_SC_AA_MASK.u32All = 0xFFFFFFFF;
+    /* Turn off AA */
+    r700->PA_SC_AA_CONFIG.u32All = 0;
+
+    r700->SX_MISC.u32All = 0;
+
+    r700InitSQConfig(ctx);
+}
+
+void r700InitStateFuncs(struct dd_function_table *functions) //-----------------
+{
+	functions->UpdateState = r700InvalidateState;
+	functions->AlphaFunc = r700AlphaFunc;
+	functions->BlendColor = r700BlendColor;
+	functions->BlendEquationSeparate = r700BlendEquationSeparate;
+	functions->BlendFuncSeparate = r700BlendFuncSeparate;
+	functions->Enable = r700Enable;
+	functions->ColorMask = r700ColorMask;
+	functions->DepthFunc = r700DepthFunc;
+	functions->DepthMask = r700DepthMask;
+	functions->CullFace = r700CullFace;
+	functions->Fogfv = r700Fogfv;
+	functions->FrontFace = r700FrontFace;
+	functions->ShadeModel = r700ShadeModel;
+
+	/* ARB_point_parameters */
+	functions->PointParameterfv = r700PointParameter;
+
+	/* Stencil related */
+	functions->StencilFuncSeparate = r700StencilFuncSeparate;
+	functions->StencilMaskSeparate = r700StencilMaskSeparate;
+	functions->StencilOpSeparate = r700StencilOpSeparate;
+
+	/* Viewport related */
+	functions->Viewport = r700Viewport;
+	functions->DepthRange = r700DepthRange;
+	functions->PointSize = r700PointSize;
+	functions->LineWidth = r700LineWidth;
+
+	functions->PolygonOffset = r700PolygonOffset;
+	functions->PolygonMode = r700PolygonMode;
+
+	functions->RenderMode = r700RenderMode;
+
+	functions->ClipPlane = r700ClipPlane;
+
+	functions->Scissor = radeonScissor;
+
+	functions->DrawBuffer		= radeonDrawBuffer;
+	functions->ReadBuffer		= radeonReadBuffer;
+
+}
+
diff --git a/src/mesa/drivers/dri/r600/r700_state.h b/src/mesa/drivers/dri/r600/r700_state.h
new file mode 100644
index 0000000000..23246367db
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_state.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#ifndef _R700_STATE_H
+#define _R700_STATE_H
+
+#include "main/mtypes.h"
+
+#include "r600_context.h"
+
+#include "r700_chip.h"
+
+extern void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state);
+extern void r700UpdateShaders (GLcontext * ctx);
+
+extern void r700UpdateViewportOffset(GLcontext * ctx);
+extern void r700UpdateDrawBuffer (GLcontext * ctx);
+
+extern void r700InitState (GLcontext * ctx);
+extern void r700InitStateFuncs (struct dd_function_table *functions);
+
+extern void r700SetRenderTarget(context_t *context, int id);
+extern void r700SetDefaultStates(context_t * context);
+
+void r700SetScissor(context_t *context);
+void r700SetDepthTarget(context_t *context);
+
+#endif	/* _R600_SCREEN_H */
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c
new file mode 100644
index 0000000000..cbfeaf071b
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.c
@@ -0,0 +1,430 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+#include <math.h>
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "tnl/t_context.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+
+#include "r600_context.h"
+#include "r600_cmdbuf.h"
+
+#include "r700_debug.h"
+#include "r700_vertprog.h"
+
+unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm, 
+					           struct gl_vertex_program *mesa_vp,
+					           unsigned int unStart)
+{
+    unsigned int i;
+	unsigned int unBit;
+	unsigned int unTotal = unStart;
+
+    //!!!!!!! THE ORDER MATCH FS INPUT 
+
+	unBit = 1 << VERT_RESULT_HPOS;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_HPOS] = unTotal++;
+	}
+
+	unBit = 1 << VERT_RESULT_COL0;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_COL0] = unTotal++;
+	}
+
+	unBit = 1 << VERT_RESULT_COL1;
+	if(mesa_vp->Base.OutputsWritten & unBit)
+	{
+		pAsm->ucVP_OutputMap[VERT_RESULT_COL1] = unTotal++;
+	}
+
+	//TODO : dealing back face.
+	//unBit = 1 << VERT_RESULT_BFC0;
+	//if(mesa_vp->Base.OutputsWritten & unBit)
+	//{
+	//	pAsm->ucVP_OutputMap[VERT_RESULT_COL0] = unTotal++;
+	//}
+
+	//unBit = 1 << VERT_RESULT_BFC1;
+	//if(mesa_vp->Base.OutputsWritten & unBit)
+	//{
+	//	pAsm->ucVP_OutputMap[VERT_RESULT_COL1] = unTotal++;
+	//}
+
+	//TODO : dealing fog.
+	//unBit = 1 << VERT_RESULT_FOGC;
+	//if(mesa_vp->Base.OutputsWritten & unBit)
+	//{
+	//	pAsm->ucVP_OutputMap[VERT_RESULT_FOGC] = unTotal++;
+	//}
+
+	//TODO : dealing point size.
+	//unBit = 1 << VERT_RESULT_PSIZ;
+	//if(mesa_vp->Base.OutputsWritten & unBit)
+	//{
+	//	pAsm->ucVP_OutputMap[VERT_RESULT_PSIZ] = unTotal++;
+	//}
+
+	for(i=0; i<8; i++)
+	{
+		unBit = 1 << (VERT_RESULT_TEX0 + i);
+		if(mesa_vp->Base.OutputsWritten & unBit)
+		{
+			pAsm->ucVP_OutputMap[VERT_RESULT_TEX0 + i] = unTotal++;
+		}
+	}
+
+	return (unTotal - unStart);
+}
+
+unsigned int Map_Vertex_Input(r700_AssemblerBase       *pAsm, 
+					  struct gl_vertex_program *mesa_vp,
+					  unsigned int unStart)
+{
+	int i;
+	unsigned int unBit;
+	unsigned int unTotal = unStart;
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(mesa_vp->Base.InputsRead & unBit)
+		{
+			pAsm->ucVP_AttributeMap[i] = unTotal++;
+		}
+	}
+	return (unTotal - unStart);
+}
+
+GLboolean Process_Vertex_Program_Vfetch_Instructions(
+						struct r700_vertex_program *vp,
+						struct gl_vertex_program   *mesa_vp)
+{
+	int i;
+    unsigned int unBit;
+	VTX_FETCH_METHOD vtxFetchMethod;
+	vtxFetchMethod.bEnableMini          = GL_FALSE;
+	vtxFetchMethod.mega_fetch_remainder = 0;
+
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(mesa_vp->Base.InputsRead & unBit)
+		{
+			assemble_vfetch_instruction(&vp->r700AsmCode,
+						    i,
+						    vp->r700AsmCode.ucVP_AttributeMap[i],
+						    vp->aos_desc[i].size,
+						    vp->aos_desc[i].type,
+						    &vtxFetchMethod);
+		}
+	}
+	
+	return GL_TRUE;
+}
+
+void Map_Vertex_Program(struct r700_vertex_program *vp,
+						struct gl_vertex_program   *mesa_vp)
+{
+    GLuint ui;
+    r700_AssemblerBase *pAsm = &(vp->r700AsmCode);
+	unsigned int num_inputs;
+
+	// R0 will always be used for index into vertex buffer
+	pAsm->number_used_registers = 1;
+	pAsm->starting_vfetch_register_number = pAsm->number_used_registers;
+
+    // Map Inputs: Add 1 to mapping since R0 is used for index
+	num_inputs = Map_Vertex_Input(pAsm, mesa_vp, pAsm->number_used_registers);
+	pAsm->number_used_registers += num_inputs;
+
+	// Create VFETCH instructions for inputs
+	if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions(vp, mesa_vp) ) 
+	{
+		r700_error(ERROR_ASM_VTX_CLAUSE, "Calling Process_Vertex_Program_Vfetch_Instructions return error. \n");
+		return; //error
+	}
+
+	// Map Outputs
+	pAsm->number_of_exports = Map_Vertex_Output(pAsm, mesa_vp, pAsm->number_used_registers);
+
+	pAsm->starting_export_register_number = pAsm->number_used_registers;
+
+	pAsm->number_used_registers += pAsm->number_of_exports;
+
+    pAsm->pucOutMask = (unsigned char*) MALLOC(pAsm->number_of_exports);
+    
+    for(ui=0; ui<pAsm->number_of_exports; ui++)
+    {
+        pAsm->pucOutMask[ui] = 0x0;
+    }
+
+    /* Map temporary registers (GPRs) */
+    pAsm->starting_temp_register_number = pAsm->number_used_registers;
+
+    if(mesa_vp->Base.NumNativeTemporaries >= mesa_vp->Base.NumTemporaries)
+    {   /* arb uses NumNativeTemporaries */
+        pAsm->number_used_registers += mesa_vp->Base.NumNativeTemporaries;
+    }
+    else
+    {   /* fix func t_vp uses NumTemporaries */
+        pAsm->number_used_registers += mesa_vp->Base.NumTemporaries;
+    }
+	
+    pAsm->uFirstHelpReg = pAsm->number_used_registers;
+}
+
+GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
+					                	struct gl_vertex_program   *mesa_vp)
+{
+    GLuint i, j;
+    GLint * puiTEMPwrites;
+    struct prog_instruction *pILInst;
+    InstDeps         *pInstDeps;
+
+    puiTEMPwrites = (GLint*) MALLOC(sizeof(GLuint)*mesa_vp->Base.NumTemporaries);
+    for(i=0; i<mesa_vp->Base.NumTemporaries; i++)
+    {
+        puiTEMPwrites[i] = -1;
+    }
+
+    pInstDeps = (InstDeps*)MALLOC(sizeof(InstDeps)*mesa_vp->Base.NumInstructions);
+
+    for(i=0; i<mesa_vp->Base.NumInstructions; i++)
+    {
+        pInstDeps[i].nDstDep = -1;
+        pILInst = &(mesa_vp->Base.Instructions[i]);
+
+        //Dst
+        if(pILInst->DstReg.File == PROGRAM_TEMPORARY)
+        {
+            //Set lastwrite for the temp
+            puiTEMPwrites[pILInst->DstReg.Index] = i;
+        }
+
+        //Src
+        for(j=0; j<3; j++)
+        {
+            if(pILInst->SrcReg[j].File == PROGRAM_TEMPORARY)
+            {
+                //Set dep.
+                pInstDeps[i].nSrcDeps[j] = puiTEMPwrites[pILInst->SrcReg[j].Index];
+            }
+            else
+            {
+                pInstDeps[i].nSrcDeps[j] = -1;
+            }
+        }
+    }
+
+    vp->r700AsmCode.pInstDeps = pInstDeps;
+
+    FREE(puiTEMPwrites);
+
+    return GL_TRUE;
+}
+
+GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp,
+							   struct gl_vertex_program   *mesa_vp)
+{
+	//Init_Program
+	Init_r700_AssemblerBase(SPT_VP, &(vp->r700AsmCode), &(vp->r700Shader) );
+	Map_Vertex_Program( vp, mesa_vp );
+
+	if(GL_FALSE == Find_Instruction_Dependencies_vp(vp, mesa_vp))
+	{
+		return GL_FALSE;
+    }
+
+	if(GL_FALSE == AssembleInstr(mesa_vp->Base.NumInstructions,
+                                 &(mesa_vp->Base.Instructions[0]), 
+                                 &(vp->r700AsmCode)) )
+	{
+		return GL_FALSE;
+	} 
+
+    if(GL_FALSE == Process_Vertex_Exports(&(vp->r700AsmCode), mesa_vp->Base.OutputsWritten) )
+    {
+        return GL_FALSE;
+    }
+
+    vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 
+                         : (vp->r700AsmCode.number_used_registers - 1);
+
+	vp->r700Shader.nParamExports = vp->r700AsmCode.number_of_exports;
+
+    vp->translated = GL_TRUE;
+
+	return GL_TRUE;
+}
+
+void r700SelectVertexShader(GLcontext *ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    struct r700_vertex_program *vpc
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+        vpc->r700AsmCode.bR6xx = 1;
+    }
+    
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+	struct vertex_buffer *vb = &tnl->vb;
+
+    unsigned int unBit;
+    unsigned int i;
+	for(i=0; i<VERT_ATTRIB_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(vpc->mesa_program.Base.InputsRead & unBit) /* ctx->Array.ArrayObj->xxxxxxx */
+		{
+			vpc->aos_desc[i].size   = vb->AttribPtr[i]->size;
+			vpc->aos_desc[i].stride = vb->AttribPtr[i]->size * sizeof(GL_FLOAT);/* when emit array, data is packed. vb->AttribPtr[i]->stride;*/
+			vpc->aos_desc[i].type   = GL_FLOAT;
+		}
+	}
+
+    if(GL_FALSE == vpc->translated)
+    {
+        r700TranslateVertexShader(vpc,
+		    					  &(vpc->mesa_program) );
+    }
+}
+
+void * r700GetActiveVpShaderBo(GLcontext * ctx)
+{
+    struct r700_vertex_program *vp
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+
+    return vp->shaderbo;
+}
+
+GLboolean r700SetupVertexProgram(GLcontext * ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+
+    BATCH_LOCALS(&context->radeon);
+
+    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+    struct r700_vertex_program *vp
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+
+    struct gl_program_parameter_list *paramList;
+    unsigned int unNumParamData;
+
+    unsigned int ui;
+
+    if(GL_FALSE == vp->loaded)
+    {
+        if(vp->r700Shader.bNeedsAssembly == GL_TRUE)
+	    {
+		    Assemble( &(vp->r700Shader) );
+	    }
+
+        /* Load vp to gpu */
+        r600EmitShader(ctx, 
+                       &(vp->shaderbo), 
+                       (GLvoid *)(vp->r700Shader.pProgram),
+                       vp->r700Shader.uShaderBinaryDWORDSize,

+                       "VS"); 
+
+        vp->loaded = GL_TRUE;
+    }
+
+    DumpHwBinary(DUMP_VERTEX_SHADER, (GLvoid *)(vp->r700Shader.pProgram),
+                 vp->r700Shader.uShaderBinaryDWORDSize);
+
+    /* TODO : enable this after MemUse fixed *=
+    (context->chipobj.MemUse)(context, vp->shadercode.buf->id);
+    */
+
+    r700->vs.SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */ 
+    
+    SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.nRegs + 1,
+             NUM_GPRS_shift, NUM_GPRS_mask);
+
+    if(vp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */
+	{
+        SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.uStackSize,
+                 STACK_SIZE_shift, STACK_SIZE_mask);
+    }
+
+    SETfield(r700->SPI_VS_OUT_CONFIG.u32All, vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0,
+             VS_EXPORT_COUNT_shift, VS_EXPORT_COUNT_mask);
+	SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports,
+             NUM_INTERP_shift, NUM_INTERP_mask);
+
+    /*
+    SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, PERSP_GRADIENT_ENA_bit);
+    CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit);
+    */
+
+    /* sent out shader constants. */
+
+    paramList = vp->mesa_program.Base.Parameters;
+
+    if(NULL != paramList)
+    {
+        _mesa_load_state_parameters(ctx, paramList);
+
+        unNumParamData = paramList->NumParameters * 4;
+
+        BEGIN_BATCH_NO_AUTOSTATE(unNumParamData + 2); 
+        
+        R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData));
+        /* assembler map const from very beginning. */
+        R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4);
+
+        unNumParamData = paramList->NumParameters;
+
+        for(ui=0; ui<unNumParamData; ui++)
+        {
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0])));
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1])));
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2])));
+            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3])));
+        }
+        END_BATCH();
+        COMMIT_BATCH();
+    }
+
+    return GL_TRUE;
+}
+
+
+
+
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.h b/src/mesa/drivers/dri/r600/r700_vertprog.h
new file mode 100644
index 0000000000..6a9726a3d0
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.h
@@ -0,0 +1,88 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Authors:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+
+#ifndef _R700_VERTPROG_H_
+#define _R700_VERTPROG_H_
+
+#include "main/glheader.h"
+#include "main/mtypes.h" 
+
+#include "r700_shader.h"
+#include "r700_assembler.h"
+
+typedef struct ArrayDesc //TEMP
+{
+	GLint size;   //number of data element
+	GLenum type;  //data element type
+	GLsizei stride;
+} ArrayDesc;
+
+struct r700_vertex_program 
+{
+    struct gl_vertex_program mesa_program; /* Must be first */
+
+	struct r700_vertex_program *next;
+
+    r700_AssemblerBase r700AsmCode;
+    R700_Shader        r700Shader;
+
+	GLboolean translated;
+    GLboolean loaded;
+	
+	/* ... */
+
+    void * shaderbo;
+
+	ArrayDesc              aos_desc[VERT_ATTRIB_MAX];
+};
+
+//Internal
+unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm, 
+					           struct gl_vertex_program *mesa_vp,
+					           unsigned int unStart);
+unsigned int Map_Vertex_Input(r700_AssemblerBase       *pAsm, 
+					          struct gl_vertex_program *mesa_vp,
+					          unsigned int unStart);
+GLboolean Process_Vertex_Program_Vfetch_Instructions(
+						struct r700_vertex_program *vp,
+						struct gl_vertex_program   *mesa_vp);
+void Map_Vertex_Program(struct r700_vertex_program *vp,
+						struct gl_vertex_program   *mesa_vp);
+GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
+					                	   struct gl_vertex_program   *mesa_vp);
+
+/* Interface */
+extern GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp,
+							   struct gl_vertex_program   *mesa_vp);
+
+extern void r700SelectVertexShader(GLcontext *ctx);
+
+extern GLboolean r700SetupVertexProgram(GLcontext * ctx);
+
+extern void *    r700GetActiveVpShaderBo(GLcontext * ctx);
+
+#endif /* _R700_VERTPROG_H_ */
diff --git a/src/mesa/drivers/dri/r300/radeon_lock.h b/src/mesa/drivers/dri/r600/radeon_context.h
index a344837f47..250570f6b8 100644
--- a/src/mesa/drivers/dri/r300/radeon_lock.h
+++ b/src/mesa/drivers/dri/r600/radeon_context.h
@@ -37,79 +37,40 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Gareth Hughes <gareth@valinux.com>
  *   Keith Whitwell <keith@tungstengraphics.com>
  *   Kevin E. Martin <martin@valinux.com>
+ *   Nicolai Haehnle <prefect_@gmx.net>
  */
 
-#ifndef __RADEON_LOCK_H__
-#define __RADEON_LOCK_H__
+#ifndef __RADEON_CONTEXT_H__
+#define __RADEON_CONTEXT_H__
 
-#include "radeon_context.h"
+#include "main/mtypes.h"
+#include "main/colormac.h"
+#include "radeon_screen.h"
+#include "drm.h"
+#include "dri_util.h"
 
-extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags);
-extern void radeonUpdatePageFlipping(radeonContextPtr rmesa);
+#include "radeon_screen.h"
 
-/* Turn DEBUG_LOCKING on to find locking conflicts.
- */
-#define DEBUG_LOCKING	0
-
-#if DEBUG_LOCKING
-extern char *prevLockFile;
-extern int prevLockLine;
-
-#define DEBUG_LOCK()							\
-   do {									\
-      prevLockFile = (__FILE__);					\
-      prevLockLine = (__LINE__);					\
-   } while (0)
-
-#define DEBUG_RESET()							\
-   do {									\
-      prevLockFile = 0;							\
-      prevLockLine = 0;							\
-   } while (0)
-
-#define DEBUG_CHECK_LOCK()						\
-   do {									\
-      if (prevLockFile) {						\
-	 fprintf(stderr,						\
-		  "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",	\
-		  prevLockFile, prevLockLine, __FILE__, __LINE__);	\
-	 exit(1);							\
-      }									\
-   } while (0)
+#if R200_MERGED
+extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
 
+#define FALLBACK( radeon, bit, mode ) do {			\
+   if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n",	\
+		     __FUNCTION__, bit, mode );			\
+   radeonFallback( (radeon)->glCtx, bit, mode );		\
+} while (0)
 #else
+#define FALLBACK( radeon, bit, mode ) fprintf(stderr, "%s:%s\n", __LINE__, __FILE__);
+#endif
 
-#define DEBUG_LOCK()
-#define DEBUG_RESET()
-#define DEBUG_CHECK_LOCK()
+/* TCL fallbacks */
+extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode);
 
+#if R200_MERGED
+#define TCL_FALLBACK( ctx, bit, mode )	radeonTclFallback( ctx, bit, mode )
+#else
+#define TCL_FALLBACK( ctx, bit, mode )	;
 #endif
 
-/*
- * !!! We may want to separate locks from locks with validation.  This
- * could be used to improve performance for those things commands that
- * do not do any drawing !!!
- */
 
-/* Lock the hardware and validate our state.
- */
-#define LOCK_HARDWARE( rmesa )						\
-	do {								\
-		char __ret = 0;						\
-		DEBUG_CHECK_LOCK();					\
-		DRM_CAS((rmesa)->dri.hwLock, (rmesa)->dri.hwContext,	\
-			(DRM_LOCK_HELD | (rmesa)->dri.hwContext), __ret); \
-		if (__ret)						\
-			radeonGetLock((rmesa), 0);			\
-		DEBUG_LOCK();						\
-	} while (0)
-
-#define UNLOCK_HARDWARE( rmesa )					\
-	do {								\
-		DRM_UNLOCK((rmesa)->dri.fd,				\
-			(rmesa)->dri.hwLock,				\
-			(rmesa)->dri.hwContext);			\
-		DEBUG_RESET();						\
-	} while (0)
-
-#endif				/* __RADEON_LOCK_H__ */
+#endif				/* __RADEON_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r600/sq_micro_reg.h b/src/mesa/drivers/dri/r600/sq_micro_reg.h
new file mode 100644
index 0000000000..bfd21cef62
--- /dev/null
+++ b/src/mesa/drivers/dri/r600/sq_micro_reg.h
@@ -0,0 +1,2008 @@
+/*
+ * Copyright (C) 2008-2009  Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * Contacts:
+ *   Richard Li <RichardZ.Li@amd.com>, <richardradeon@gmail.com>
+ */
+
+#if !defined (_SQ_MICRO_REG_H)
+#define _SQ_MICRO_REG_H
+
+#if defined(LITTLEENDIAN_CPU)
+#elif defined(BIGENDIAN_CPU)
+#else
+#error "BIGENDIAN_CPU or LITTLEENDIAN_CPU must be defined"
+#endif
+
+/*
+ * SQ_ALU_SRC_GPR_BASE value
+ */
+
+#define SQ_ALU_SRC_GPR_BASE            0x00000000
+
+/*
+ * SQ_ALU_SRC_GPR_SIZE value
+ */
+
+#define SQ_ALU_SRC_GPR_SIZE            0x00000080
+
+/*
+ * SQ_ALU_SRC_KCACHE0_BASE value
+ */
+
+#define SQ_ALU_SRC_KCACHE0_BASE        0x00000080
+
+/*
+ * SQ_ALU_SRC_KCACHE0_SIZE value
+ */
+
+#define SQ_ALU_SRC_KCACHE0_SIZE        0x00000020
+
+/*
+ * SQ_ALU_SRC_KCACHE1_BASE value
+ */
+
+#define SQ_ALU_SRC_KCACHE1_BASE        0x000000a0
+
+/*
+ * SQ_ALU_SRC_KCACHE1_SIZE value
+ */
+
+#define SQ_ALU_SRC_KCACHE1_SIZE        0x00000020
+
+/*
+ * SQ_ALU_SRC_CFILE_BASE value
+ */
+
+#define SQ_ALU_SRC_CFILE_BASE          0x00000100
+
+/*
+ * SQ_ALU_SRC_CFILE_SIZE value
+ */
+
+#define SQ_ALU_SRC_CFILE_SIZE          0x00000100
+
+/*
+ * SQ_SP_OP_REDUC_BEGIN value
+ */
+
+#define SQ_SP_OP_REDUC_BEGIN           0x00000050
+
+/*
+ * SQ_SP_OP_REDUC_END value
+ */
+
+#define SQ_SP_OP_REDUC_END             0x00000053
+
+/*
+ * SQ_SP_OP_TRANS_BEGIN value
+ */
+
+#define SQ_SP_OP_TRANS_BEGIN           0x00000060
+
+/*
+ * SQ_SP_OP_TRANS_END value
+ */
+
+#define SQ_SP_OP_TRANS_END             0x0000007f
+
+/*
+ * SQ_CF_WORD0 struct
+ */
+
+#define SQ_CF_WORD0_ADDR_SIZE          32
+
+#define SQ_CF_WORD0_ADDR_SHIFT         0
+
+#define SQ_CF_WORD0_ADDR_MASK          0xffffffff
+
+#define SQ_CF_WORD0_MASK \
+     (SQ_CF_WORD0_ADDR_MASK)
+
+#define SQ_CF_WORD0_DEFAULT            0xcdcdcdcd
+
+#define SQ_CF_WORD0_GET_ADDR(sq_cf_word0) \
+     ((sq_cf_word0 & SQ_CF_WORD0_ADDR_MASK) >> SQ_CF_WORD0_ADDR_SHIFT)
+
+#define SQ_CF_WORD0_SET_ADDR(sq_cf_word0_reg, addr) \
+     sq_cf_word0_reg = (sq_cf_word0_reg & ~SQ_CF_WORD0_ADDR_MASK) | (addr << SQ_CF_WORD0_ADDR_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_word0_t {
+          unsigned int addr                           : SQ_CF_WORD0_ADDR_SIZE;
+     } sq_cf_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_word0_t {
+          unsigned int addr                           : SQ_CF_WORD0_ADDR_SIZE;
+     } sq_cf_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_word0_t f;
+} sq_cf_word0_u;
+
+
+/*
+ * SQ_CF_WORD1 struct
+ */
+
+#define SQ_CF_WORD1_POP_COUNT_SIZE     3
+#define SQ_CF_WORD1_CF_CONST_SIZE      5
+#define SQ_CF_WORD1_COND_SIZE          2
+#define SQ_CF_WORD1_COUNT_SIZE         3
+#define SQ_CF_WORD1_CALL_COUNT_SIZE    6
+#define SQ_CF_WORD1_COUNT_3_SIZE       1
+#define SQ_CF_WORD1_END_OF_PROGRAM_SIZE 1
+#define SQ_CF_WORD1_VALID_PIXEL_MODE_SIZE 1
+#define SQ_CF_WORD1_CF_INST_SIZE       7
+#define SQ_CF_WORD1_WHOLE_QUAD_MODE_SIZE 1
+#define SQ_CF_WORD1_BARRIER_SIZE       1
+
+#define SQ_CF_WORD1_POP_COUNT_SHIFT    0
+#define SQ_CF_WORD1_CF_CONST_SHIFT     3
+#define SQ_CF_WORD1_COND_SHIFT         8
+#define SQ_CF_WORD1_COUNT_SHIFT        10
+#define SQ_CF_WORD1_CALL_COUNT_SHIFT   13
+#define SQ_CF_WORD1_COUNT_3_SHIFT      19
+#define SQ_CF_WORD1_END_OF_PROGRAM_SHIFT 21
+#define SQ_CF_WORD1_VALID_PIXEL_MODE_SHIFT 22
+#define SQ_CF_WORD1_CF_INST_SHIFT      23
+#define SQ_CF_WORD1_WHOLE_QUAD_MODE_SHIFT 30
+#define SQ_CF_WORD1_BARRIER_SHIFT      31
+
+#define SQ_CF_WORD1_POP_COUNT_MASK     0x00000007
+#define SQ_CF_WORD1_CF_CONST_MASK      0x000000f8
+#define SQ_CF_WORD1_COND_MASK          0x00000300
+#define SQ_CF_WORD1_COUNT_MASK         0x00001c00
+#define SQ_CF_WORD1_CALL_COUNT_MASK    0x0007e000
+#define SQ_CF_WORD1_COUNT_3_MASK       0x00080000
+#define SQ_CF_WORD1_END_OF_PROGRAM_MASK 0x00200000
+#define SQ_CF_WORD1_VALID_PIXEL_MODE_MASK 0x00400000
+#define SQ_CF_WORD1_CF_INST_MASK       0x3f800000
+#define SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK 0x40000000
+#define SQ_CF_WORD1_BARRIER_MASK       0x80000000
+
+#define SQ_CF_WORD1_MASK \
+     (SQ_CF_WORD1_POP_COUNT_MASK | \
+      SQ_CF_WORD1_CF_CONST_MASK | \
+      SQ_CF_WORD1_COND_MASK | \
+      SQ_CF_WORD1_COUNT_MASK | \
+      SQ_CF_WORD1_CALL_COUNT_MASK | \
+      SQ_CF_WORD1_COUNT_3_MASK | \
+      SQ_CF_WORD1_END_OF_PROGRAM_MASK | \
+      SQ_CF_WORD1_VALID_PIXEL_MODE_MASK | \
+      SQ_CF_WORD1_CF_INST_MASK | \
+      SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK | \
+      SQ_CF_WORD1_BARRIER_MASK)
+
+#define SQ_CF_WORD1_DEFAULT            0xcdcdcdcd
+
+#define SQ_CF_WORD1_GET_POP_COUNT(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_POP_COUNT_MASK) >> SQ_CF_WORD1_POP_COUNT_SHIFT)
+#define SQ_CF_WORD1_GET_CF_CONST(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_CF_CONST_MASK) >> SQ_CF_WORD1_CF_CONST_SHIFT)
+#define SQ_CF_WORD1_GET_COND(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_COND_MASK) >> SQ_CF_WORD1_COND_SHIFT)
+#define SQ_CF_WORD1_GET_COUNT(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_COUNT_MASK) >> SQ_CF_WORD1_COUNT_SHIFT)
+#define SQ_CF_WORD1_GET_CALL_COUNT(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_CALL_COUNT_MASK) >> SQ_CF_WORD1_CALL_COUNT_SHIFT)
+#define SQ_CF_WORD1_GET_COUNT_3(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_COUNT_3_MASK) >> SQ_CF_WORD1_COUNT_3_SHIFT)
+#define SQ_CF_WORD1_GET_END_OF_PROGRAM(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_END_OF_PROGRAM_MASK) >> SQ_CF_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_WORD1_GET_VALID_PIXEL_MODE(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_VALID_PIXEL_MODE_MASK) >> SQ_CF_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_WORD1_GET_CF_INST(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_CF_INST_MASK) >> SQ_CF_WORD1_CF_INST_SHIFT)
+#define SQ_CF_WORD1_GET_WHOLE_QUAD_MODE(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK) >> SQ_CF_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_WORD1_GET_BARRIER(sq_cf_word1) \
+     ((sq_cf_word1 & SQ_CF_WORD1_BARRIER_MASK) >> SQ_CF_WORD1_BARRIER_SHIFT)
+
+#define SQ_CF_WORD1_SET_POP_COUNT(sq_cf_word1_reg, pop_count) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_POP_COUNT_MASK) | (pop_count << SQ_CF_WORD1_POP_COUNT_SHIFT)
+#define SQ_CF_WORD1_SET_CF_CONST(sq_cf_word1_reg, cf_const) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_CF_CONST_MASK) | (cf_const << SQ_CF_WORD1_CF_CONST_SHIFT)
+#define SQ_CF_WORD1_SET_COND(sq_cf_word1_reg, cond) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_COND_MASK) | (cond << SQ_CF_WORD1_COND_SHIFT)
+#define SQ_CF_WORD1_SET_COUNT(sq_cf_word1_reg, count) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_COUNT_MASK) | (count << SQ_CF_WORD1_COUNT_SHIFT)
+#define SQ_CF_WORD1_SET_CALL_COUNT(sq_cf_word1_reg, call_count) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_CALL_COUNT_MASK) | (call_count << SQ_CF_WORD1_CALL_COUNT_SHIFT)
+#define SQ_CF_WORD1_SET_COUNT_3(sq_cf_word1_reg, count_3) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_COUNT_3_MASK) | (count_3 << SQ_CF_WORD1_COUNT_3_SHIFT)
+#define SQ_CF_WORD1_SET_END_OF_PROGRAM(sq_cf_word1_reg, end_of_program) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_END_OF_PROGRAM_MASK) | (end_of_program << SQ_CF_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_WORD1_SET_VALID_PIXEL_MODE(sq_cf_word1_reg, valid_pixel_mode) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_VALID_PIXEL_MODE_MASK) | (valid_pixel_mode << SQ_CF_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_WORD1_SET_CF_INST(sq_cf_word1_reg, cf_inst) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_CF_INST_MASK) | (cf_inst << SQ_CF_WORD1_CF_INST_SHIFT)
+#define SQ_CF_WORD1_SET_WHOLE_QUAD_MODE(sq_cf_word1_reg, whole_quad_mode) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_WHOLE_QUAD_MODE_MASK) | (whole_quad_mode << SQ_CF_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_WORD1_SET_BARRIER(sq_cf_word1_reg, barrier) \
+     sq_cf_word1_reg = (sq_cf_word1_reg & ~SQ_CF_WORD1_BARRIER_MASK) | (barrier << SQ_CF_WORD1_BARRIER_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_word1_t {
+          unsigned int pop_count                      : SQ_CF_WORD1_POP_COUNT_SIZE;
+          unsigned int cf_const                       : SQ_CF_WORD1_CF_CONST_SIZE;
+          unsigned int cond                           : SQ_CF_WORD1_COND_SIZE;
+          unsigned int count                          : SQ_CF_WORD1_COUNT_SIZE;
+          unsigned int call_count                     : SQ_CF_WORD1_CALL_COUNT_SIZE;
+          unsigned int count_3                        : SQ_CF_WORD1_COUNT_3_SIZE;
+          unsigned int                                : 1;
+          unsigned int end_of_program                 : SQ_CF_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_WORD1_CF_INST_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int barrier                        : SQ_CF_WORD1_BARRIER_SIZE;
+     } sq_cf_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_word1_t {
+          unsigned int barrier                        : SQ_CF_WORD1_BARRIER_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_WORD1_CF_INST_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int end_of_program                 : SQ_CF_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int                                : 1;
+          unsigned int count_3                        : SQ_CF_WORD1_COUNT_3_SIZE;
+          unsigned int call_count                     : SQ_CF_WORD1_CALL_COUNT_SIZE;
+          unsigned int count                          : SQ_CF_WORD1_COUNT_SIZE;
+          unsigned int cond                           : SQ_CF_WORD1_COND_SIZE;
+          unsigned int cf_const                       : SQ_CF_WORD1_CF_CONST_SIZE;
+          unsigned int pop_count                      : SQ_CF_WORD1_POP_COUNT_SIZE;
+     } sq_cf_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_word1_t f;
+} sq_cf_word1_u;
+
+
+/*
+ * SQ_CF_ALU_WORD0 struct
+ */
+
+#define SQ_CF_ALU_WORD0_ADDR_SIZE      22
+#define SQ_CF_ALU_WORD0_KCACHE_BANK0_SIZE 4
+#define SQ_CF_ALU_WORD0_KCACHE_BANK1_SIZE 4
+#define SQ_CF_ALU_WORD0_KCACHE_MODE0_SIZE 2
+
+#define SQ_CF_ALU_WORD0_ADDR_SHIFT     0
+#define SQ_CF_ALU_WORD0_KCACHE_BANK0_SHIFT 22
+#define SQ_CF_ALU_WORD0_KCACHE_BANK1_SHIFT 26
+#define SQ_CF_ALU_WORD0_KCACHE_MODE0_SHIFT 30
+
+#define SQ_CF_ALU_WORD0_ADDR_MASK      0x003fffff
+#define SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK 0x03c00000
+#define SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK 0x3c000000
+#define SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK 0xc0000000
+
+#define SQ_CF_ALU_WORD0_MASK \
+     (SQ_CF_ALU_WORD0_ADDR_MASK | \
+      SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK | \
+      SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK | \
+      SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK)
+
+#define SQ_CF_ALU_WORD0_DEFAULT        0xcdcdcdcd
+
+#define SQ_CF_ALU_WORD0_GET_ADDR(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_ADDR_MASK) >> SQ_CF_ALU_WORD0_ADDR_SHIFT)
+#define SQ_CF_ALU_WORD0_GET_KCACHE_BANK0(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK) >> SQ_CF_ALU_WORD0_KCACHE_BANK0_SHIFT)
+#define SQ_CF_ALU_WORD0_GET_KCACHE_BANK1(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK) >> SQ_CF_ALU_WORD0_KCACHE_BANK1_SHIFT)
+#define SQ_CF_ALU_WORD0_GET_KCACHE_MODE0(sq_cf_alu_word0) \
+     ((sq_cf_alu_word0 & SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK) >> SQ_CF_ALU_WORD0_KCACHE_MODE0_SHIFT)
+
+#define SQ_CF_ALU_WORD0_SET_ADDR(sq_cf_alu_word0_reg, addr) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_ADDR_MASK) | (addr << SQ_CF_ALU_WORD0_ADDR_SHIFT)
+#define SQ_CF_ALU_WORD0_SET_KCACHE_BANK0(sq_cf_alu_word0_reg, kcache_bank0) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_KCACHE_BANK0_MASK) | (kcache_bank0 << SQ_CF_ALU_WORD0_KCACHE_BANK0_SHIFT)
+#define SQ_CF_ALU_WORD0_SET_KCACHE_BANK1(sq_cf_alu_word0_reg, kcache_bank1) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_KCACHE_BANK1_MASK) | (kcache_bank1 << SQ_CF_ALU_WORD0_KCACHE_BANK1_SHIFT)
+#define SQ_CF_ALU_WORD0_SET_KCACHE_MODE0(sq_cf_alu_word0_reg, kcache_mode0) \
+     sq_cf_alu_word0_reg = (sq_cf_alu_word0_reg & ~SQ_CF_ALU_WORD0_KCACHE_MODE0_MASK) | (kcache_mode0 << SQ_CF_ALU_WORD0_KCACHE_MODE0_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word0_t {
+          unsigned int addr                           : SQ_CF_ALU_WORD0_ADDR_SIZE;
+          unsigned int kcache_bank0                   : SQ_CF_ALU_WORD0_KCACHE_BANK0_SIZE;
+          unsigned int kcache_bank1                   : SQ_CF_ALU_WORD0_KCACHE_BANK1_SIZE;
+          unsigned int kcache_mode0                   : SQ_CF_ALU_WORD0_KCACHE_MODE0_SIZE;
+     } sq_cf_alu_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word0_t {
+          unsigned int kcache_mode0                   : SQ_CF_ALU_WORD0_KCACHE_MODE0_SIZE;
+          unsigned int kcache_bank1                   : SQ_CF_ALU_WORD0_KCACHE_BANK1_SIZE;
+          unsigned int kcache_bank0                   : SQ_CF_ALU_WORD0_KCACHE_BANK0_SIZE;
+          unsigned int addr                           : SQ_CF_ALU_WORD0_ADDR_SIZE;
+     } sq_cf_alu_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alu_word0_t f;
+} sq_cf_alu_word0_u;
+
+
+/*
+ * SQ_CF_ALU_WORD1 struct
+ */
+
+#define SQ_CF_ALU_WORD1_KCACHE_MODE1_SIZE 2
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR0_SIZE 8
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR1_SIZE 8
+#define SQ_CF_ALU_WORD1_COUNT_SIZE     7
+#define SQ_CF_ALU_WORD1_ALT_CONST_SIZE 1
+#define SQ_CF_ALU_WORD1_CF_INST_SIZE   4
+#define SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SIZE 1
+#define SQ_CF_ALU_WORD1_BARRIER_SIZE   1
+
+#define SQ_CF_ALU_WORD1_KCACHE_MODE1_SHIFT 0
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR0_SHIFT 2
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR1_SHIFT 10
+#define SQ_CF_ALU_WORD1_COUNT_SHIFT    18
+#define SQ_CF_ALU_WORD1_ALT_CONST_SHIFT 25
+#define SQ_CF_ALU_WORD1_CF_INST_SHIFT  26
+#define SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SHIFT 30
+#define SQ_CF_ALU_WORD1_BARRIER_SHIFT  31
+
+#define SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK 0x00000003
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK 0x000003fc
+#define SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK 0x0003fc00
+#define SQ_CF_ALU_WORD1_COUNT_MASK     0x01fc0000
+#define SQ_CF_ALU_WORD1_ALT_CONST_MASK 0x02000000
+#define SQ_CF_ALU_WORD1_CF_INST_MASK   0x3c000000
+#define SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK 0x40000000
+#define SQ_CF_ALU_WORD1_BARRIER_MASK   0x80000000
+
+#define SQ_CF_ALU_WORD1_MASK \
+     (SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK | \
+      SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK | \
+      SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK | \
+      SQ_CF_ALU_WORD1_COUNT_MASK | \
+      SQ_CF_ALU_WORD1_ALT_CONST_MASK | \
+      SQ_CF_ALU_WORD1_CF_INST_MASK | \
+      SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK | \
+      SQ_CF_ALU_WORD1_BARRIER_MASK)
+
+#define SQ_CF_ALU_WORD1_DEFAULT        0xcdcdcdcd
+
+#define SQ_CF_ALU_WORD1_GET_KCACHE_MODE1(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK) >> SQ_CF_ALU_WORD1_KCACHE_MODE1_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_KCACHE_ADDR0(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK) >> SQ_CF_ALU_WORD1_KCACHE_ADDR0_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_KCACHE_ADDR1(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK) >> SQ_CF_ALU_WORD1_KCACHE_ADDR1_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_COUNT(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_COUNT_MASK) >> SQ_CF_ALU_WORD1_COUNT_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_ALT_CONST(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_ALT_CONST_MASK) >> SQ_CF_ALU_WORD1_ALT_CONST_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_CF_INST(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_CF_INST_MASK) >> SQ_CF_ALU_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_WHOLE_QUAD_MODE(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK) >> SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALU_WORD1_GET_BARRIER(sq_cf_alu_word1) \
+     ((sq_cf_alu_word1 & SQ_CF_ALU_WORD1_BARRIER_MASK) >> SQ_CF_ALU_WORD1_BARRIER_SHIFT)
+
+#define SQ_CF_ALU_WORD1_SET_KCACHE_MODE1(sq_cf_alu_word1_reg, kcache_mode1) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_KCACHE_MODE1_MASK) | (kcache_mode1 << SQ_CF_ALU_WORD1_KCACHE_MODE1_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_KCACHE_ADDR0(sq_cf_alu_word1_reg, kcache_addr0) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_KCACHE_ADDR0_MASK) | (kcache_addr0 << SQ_CF_ALU_WORD1_KCACHE_ADDR0_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_KCACHE_ADDR1(sq_cf_alu_word1_reg, kcache_addr1) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_KCACHE_ADDR1_MASK) | (kcache_addr1 << SQ_CF_ALU_WORD1_KCACHE_ADDR1_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_COUNT(sq_cf_alu_word1_reg, count) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_COUNT_MASK) | (count << SQ_CF_ALU_WORD1_COUNT_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_ALT_CONST(sq_cf_alu_word1_reg, alt_const) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_ALT_CONST_MASK) | (alt_const << SQ_CF_ALU_WORD1_ALT_CONST_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_CF_INST(sq_cf_alu_word1_reg, cf_inst) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_CF_INST_MASK) | (cf_inst << SQ_CF_ALU_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_WHOLE_QUAD_MODE(sq_cf_alu_word1_reg, whole_quad_mode) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_MASK) | (whole_quad_mode << SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALU_WORD1_SET_BARRIER(sq_cf_alu_word1_reg, barrier) \
+     sq_cf_alu_word1_reg = (sq_cf_alu_word1_reg & ~SQ_CF_ALU_WORD1_BARRIER_MASK) | (barrier << SQ_CF_ALU_WORD1_BARRIER_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word1_t {
+          unsigned int kcache_mode1                   : SQ_CF_ALU_WORD1_KCACHE_MODE1_SIZE;
+          unsigned int kcache_addr0                   : SQ_CF_ALU_WORD1_KCACHE_ADDR0_SIZE;
+          unsigned int kcache_addr1                   : SQ_CF_ALU_WORD1_KCACHE_ADDR1_SIZE;
+          unsigned int count                          : SQ_CF_ALU_WORD1_COUNT_SIZE;
+          unsigned int alt_const                      : SQ_CF_ALU_WORD1_ALT_CONST_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALU_WORD1_CF_INST_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int barrier                        : SQ_CF_ALU_WORD1_BARRIER_SIZE;
+     } sq_cf_alu_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alu_word1_t {
+          unsigned int barrier                        : SQ_CF_ALU_WORD1_BARRIER_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALU_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALU_WORD1_CF_INST_SIZE;
+          unsigned int alt_const                      : SQ_CF_ALU_WORD1_ALT_CONST_SIZE;
+          unsigned int count                          : SQ_CF_ALU_WORD1_COUNT_SIZE;
+          unsigned int kcache_addr1                   : SQ_CF_ALU_WORD1_KCACHE_ADDR1_SIZE;
+          unsigned int kcache_addr0                   : SQ_CF_ALU_WORD1_KCACHE_ADDR0_SIZE;
+          unsigned int kcache_mode1                   : SQ_CF_ALU_WORD1_KCACHE_MODE1_SIZE;
+     } sq_cf_alu_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alu_word1_t f;
+} sq_cf_alu_word1_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD0 struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SIZE 13
+#define SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SIZE 2
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SIZE 7
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SIZE 7
+#define SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SIZE 2
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SHIFT 0
+#define SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SHIFT 13
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SHIFT 15
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SHIFT 22
+#define SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SHIFT 23
+#define SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SHIFT 30
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK 0x00001fff
+#define SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK 0x00006000
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK 0x003f8000
+#define SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK 0x00400000
+#define SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK 0x3f800000
+#define SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK 0xc0000000
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_DEFAULT 0xcdcdcdcd
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_ARRAY_BASE(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_TYPE(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_RW_GPR(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_RW_REL(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_INDEX_GPR(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_GET_ELEM_SIZE(sq_cf_alloc_export_word0) \
+     ((sq_cf_alloc_export_word0 & SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_ARRAY_BASE(sq_cf_alloc_export_word0_reg, array_base) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_MASK) | (array_base << SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_TYPE(sq_cf_alloc_export_word0_reg, type) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_TYPE_MASK) | (type << SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_RW_GPR(sq_cf_alloc_export_word0_reg, rw_gpr) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_MASK) | (rw_gpr << SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_RW_REL(sq_cf_alloc_export_word0_reg, rw_rel) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_MASK) | (rw_rel << SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_INDEX_GPR(sq_cf_alloc_export_word0_reg, index_gpr) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_MASK) | (index_gpr << SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD0_SET_ELEM_SIZE(sq_cf_alloc_export_word0_reg, elem_size) \
+     sq_cf_alloc_export_word0_reg = (sq_cf_alloc_export_word0_reg & ~SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_MASK) | (elem_size << SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word0_t {
+          unsigned int array_base                     : SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SIZE;
+          unsigned int type                           : SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SIZE;
+          unsigned int rw_gpr                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SIZE;
+          unsigned int rw_rel                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SIZE;
+          unsigned int index_gpr                      : SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SIZE;
+          unsigned int elem_size                      : SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SIZE;
+     } sq_cf_alloc_export_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word0_t {
+          unsigned int elem_size                      : SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE_SIZE;
+          unsigned int index_gpr                      : SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR_SIZE;
+          unsigned int rw_rel                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_REL_SIZE;
+          unsigned int rw_gpr                         : SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR_SIZE;
+          unsigned int type                           : SQ_CF_ALLOC_EXPORT_WORD0_TYPE_SIZE;
+          unsigned int array_base                     : SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE_SIZE;
+     } sq_cf_alloc_export_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word0_t f;
+} sq_cf_alloc_export_word0_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD1 struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SIZE 4
+#define SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SIZE 7
+#define SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SIZE 1
+#define SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SIZE 1
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SHIFT 17
+#define SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SHIFT 21
+#define SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SHIFT 22
+#define SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SHIFT 23
+#define SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SHIFT 30
+#define SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SHIFT 31
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK 0x001e0000
+#define SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK 0x00200000
+#define SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK 0x00400000
+#define SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK 0x3f800000
+#define SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK 0x40000000
+#define SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK 0x80000000
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_DEFAULT 0xcdcc0000
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_BURST_COUNT(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_END_OF_PROGRAM(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_VALID_PIXEL_MODE(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_CF_INST(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_WHOLE_QUAD_MODE(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_GET_BARRIER(sq_cf_alloc_export_word1) \
+     ((sq_cf_alloc_export_word1 & SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_BURST_COUNT(sq_cf_alloc_export_word1_reg, burst_count) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_MASK) | (burst_count << SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_END_OF_PROGRAM(sq_cf_alloc_export_word1_reg, end_of_program) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_MASK) | (end_of_program << SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_VALID_PIXEL_MODE(sq_cf_alloc_export_word1_reg, valid_pixel_mode) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_MASK) | (valid_pixel_mode << SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_CF_INST(sq_cf_alloc_export_word1_reg, cf_inst) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_MASK) | (cf_inst << SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_WHOLE_QUAD_MODE(sq_cf_alloc_export_word1_reg, whole_quad_mode) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_MASK) | (whole_quad_mode << SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SET_BARRIER(sq_cf_alloc_export_word1_reg, barrier) \
+     sq_cf_alloc_export_word1_reg = (sq_cf_alloc_export_word1_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_MASK) | (barrier << SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_t {
+          unsigned int                                : 17;
+          unsigned int burst_count                    : SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SIZE;
+          unsigned int end_of_program                 : SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int barrier                        : SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SIZE;
+     } sq_cf_alloc_export_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_t {
+          unsigned int barrier                        : SQ_CF_ALLOC_EXPORT_WORD1_BARRIER_SIZE;
+          unsigned int whole_quad_mode                : SQ_CF_ALLOC_EXPORT_WORD1_WHOLE_QUAD_MODE_SIZE;
+          unsigned int cf_inst                        : SQ_CF_ALLOC_EXPORT_WORD1_CF_INST_SIZE;
+          unsigned int valid_pixel_mode               : SQ_CF_ALLOC_EXPORT_WORD1_VALID_PIXEL_MODE_SIZE;
+          unsigned int end_of_program                 : SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM_SIZE;
+          unsigned int burst_count                    : SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT_SIZE;
+          unsigned int                                : 17;
+     } sq_cf_alloc_export_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word1_t f;
+} sq_cf_alloc_export_word1_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD1_BUF struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SIZE 12
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SIZE 4
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SHIFT 0
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SHIFT 12
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK 0x00000fff
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK 0x0000f000
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_DEFAULT 0x0000cdcd
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_GET_ARRAY_SIZE(sq_cf_alloc_export_word1_buf) \
+     ((sq_cf_alloc_export_word1_buf & SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_GET_COMP_MASK(sq_cf_alloc_export_word1_buf) \
+     ((sq_cf_alloc_export_word1_buf & SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_SET_ARRAY_SIZE(sq_cf_alloc_export_word1_buf_reg, array_size) \
+     sq_cf_alloc_export_word1_buf_reg = (sq_cf_alloc_export_word1_buf_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_MASK) | (array_size << SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_BUF_SET_COMP_MASK(sq_cf_alloc_export_word1_buf_reg, comp_mask) \
+     sq_cf_alloc_export_word1_buf_reg = (sq_cf_alloc_export_word1_buf_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_MASK) | (comp_mask << SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_buf_t {
+          unsigned int array_size                     : SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SIZE;
+          unsigned int comp_mask                      : SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SIZE;
+          unsigned int                                : 16;
+     } sq_cf_alloc_export_word1_buf_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_buf_t {
+          unsigned int                                : 16;
+          unsigned int comp_mask                      : SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK_SIZE;
+          unsigned int array_size                     : SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE_SIZE;
+     } sq_cf_alloc_export_word1_buf_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word1_buf_t f;
+} sq_cf_alloc_export_word1_buf_u;
+
+
+/*
+ * SQ_CF_ALLOC_EXPORT_WORD1_SWIZ struct
+ */
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SIZE 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SIZE 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SIZE 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SIZE 3
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SHIFT 0
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SHIFT 3
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SHIFT 6
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SHIFT 9
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK 0x00000007
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK 0x00000038
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK 0x000001c0
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK 0x00000e00
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_MASK \
+     (SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK | \
+      SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_DEFAULT 0x00000dcd
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_X(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_Y(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_Z(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_GET_SEL_W(sq_cf_alloc_export_word1_swiz) \
+     ((sq_cf_alloc_export_word1_swiz & SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK) >> SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SHIFT)
+
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_X(sq_cf_alloc_export_word1_swiz_reg, sel_x) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_MASK) | (sel_x << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_Y(sq_cf_alloc_export_word1_swiz_reg, sel_y) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_MASK) | (sel_y << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_Z(sq_cf_alloc_export_word1_swiz_reg, sel_z) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_MASK) | (sel_z << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SHIFT)
+#define SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SET_SEL_W(sq_cf_alloc_export_word1_swiz_reg, sel_w) \
+     sq_cf_alloc_export_word1_swiz_reg = (sq_cf_alloc_export_word1_swiz_reg & ~SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_MASK) | (sel_w << SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_swiz_t {
+          unsigned int sel_x                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SIZE;
+          unsigned int sel_y                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SIZE;
+          unsigned int sel_z                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SIZE;
+          unsigned int sel_w                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SIZE;
+          unsigned int                                : 20;
+     } sq_cf_alloc_export_word1_swiz_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_cf_alloc_export_word1_swiz_t {
+          unsigned int                                : 20;
+          unsigned int sel_w                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W_SIZE;
+          unsigned int sel_z                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z_SIZE;
+          unsigned int sel_y                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y_SIZE;
+          unsigned int sel_x                          : SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X_SIZE;
+     } sq_cf_alloc_export_word1_swiz_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_cf_alloc_export_word1_swiz_t f;
+} sq_cf_alloc_export_word1_swiz_u;
+
+
+/*
+ * SQ_ALU_WORD0 struct
+ */
+
+#define SQ_ALU_WORD0_SRC0_SEL_SIZE     9
+#define SQ_ALU_WORD0_SRC0_REL_SIZE     1
+#define SQ_ALU_WORD0_SRC0_CHAN_SIZE    2
+#define SQ_ALU_WORD0_SRC0_NEG_SIZE     1
+#define SQ_ALU_WORD0_SRC1_SEL_SIZE     9
+#define SQ_ALU_WORD0_SRC1_REL_SIZE     1
+#define SQ_ALU_WORD0_SRC1_CHAN_SIZE    2
+#define SQ_ALU_WORD0_SRC1_NEG_SIZE     1
+#define SQ_ALU_WORD0_INDEX_MODE_SIZE   3
+#define SQ_ALU_WORD0_PRED_SEL_SIZE     2
+#define SQ_ALU_WORD0_LAST_SIZE         1
+
+#define SQ_ALU_WORD0_SRC0_SEL_SHIFT    0
+#define SQ_ALU_WORD0_SRC0_REL_SHIFT    9
+#define SQ_ALU_WORD0_SRC0_CHAN_SHIFT   10
+#define SQ_ALU_WORD0_SRC0_NEG_SHIFT    12
+#define SQ_ALU_WORD0_SRC1_SEL_SHIFT    13
+#define SQ_ALU_WORD0_SRC1_REL_SHIFT    22
+#define SQ_ALU_WORD0_SRC1_CHAN_SHIFT   23
+#define SQ_ALU_WORD0_SRC1_NEG_SHIFT    25
+#define SQ_ALU_WORD0_INDEX_MODE_SHIFT  26
+#define SQ_ALU_WORD0_PRED_SEL_SHIFT    29
+#define SQ_ALU_WORD0_LAST_SHIFT        31
+
+#define SQ_ALU_WORD0_SRC0_SEL_MASK     0x000001ff
+#define SQ_ALU_WORD0_SRC0_REL_MASK     0x00000200
+#define SQ_ALU_WORD0_SRC0_CHAN_MASK    0x00000c00
+#define SQ_ALU_WORD0_SRC0_NEG_MASK     0x00001000
+#define SQ_ALU_WORD0_SRC1_SEL_MASK     0x003fe000
+#define SQ_ALU_WORD0_SRC1_REL_MASK     0x00400000
+#define SQ_ALU_WORD0_SRC1_CHAN_MASK    0x01800000
+#define SQ_ALU_WORD0_SRC1_NEG_MASK     0x02000000
+#define SQ_ALU_WORD0_INDEX_MODE_MASK   0x1c000000
+#define SQ_ALU_WORD0_PRED_SEL_MASK     0x60000000
+#define SQ_ALU_WORD0_LAST_MASK         0x80000000
+
+#define SQ_ALU_WORD0_MASK \
+     (SQ_ALU_WORD0_SRC0_SEL_MASK | \
+      SQ_ALU_WORD0_SRC0_REL_MASK | \
+      SQ_ALU_WORD0_SRC0_CHAN_MASK | \
+      SQ_ALU_WORD0_SRC0_NEG_MASK | \
+      SQ_ALU_WORD0_SRC1_SEL_MASK | \
+      SQ_ALU_WORD0_SRC1_REL_MASK | \
+      SQ_ALU_WORD0_SRC1_CHAN_MASK | \
+      SQ_ALU_WORD0_SRC1_NEG_MASK | \
+      SQ_ALU_WORD0_INDEX_MODE_MASK | \
+      SQ_ALU_WORD0_PRED_SEL_MASK | \
+      SQ_ALU_WORD0_LAST_MASK)
+
+#define SQ_ALU_WORD0_DEFAULT           0xcdcdcdcd
+
+#define SQ_ALU_WORD0_GET_SRC0_SEL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_SEL_MASK) >> SQ_ALU_WORD0_SRC0_SEL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC0_REL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_REL_MASK) >> SQ_ALU_WORD0_SRC0_REL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC0_CHAN(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_CHAN_MASK) >> SQ_ALU_WORD0_SRC0_CHAN_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC0_NEG(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC0_NEG_MASK) >> SQ_ALU_WORD0_SRC0_NEG_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_SEL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_SEL_MASK) >> SQ_ALU_WORD0_SRC1_SEL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_REL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_REL_MASK) >> SQ_ALU_WORD0_SRC1_REL_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_CHAN(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_CHAN_MASK) >> SQ_ALU_WORD0_SRC1_CHAN_SHIFT)
+#define SQ_ALU_WORD0_GET_SRC1_NEG(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_SRC1_NEG_MASK) >> SQ_ALU_WORD0_SRC1_NEG_SHIFT)
+#define SQ_ALU_WORD0_GET_INDEX_MODE(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_INDEX_MODE_MASK) >> SQ_ALU_WORD0_INDEX_MODE_SHIFT)
+#define SQ_ALU_WORD0_GET_PRED_SEL(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_PRED_SEL_MASK) >> SQ_ALU_WORD0_PRED_SEL_SHIFT)
+#define SQ_ALU_WORD0_GET_LAST(sq_alu_word0) \
+     ((sq_alu_word0 & SQ_ALU_WORD0_LAST_MASK) >> SQ_ALU_WORD0_LAST_SHIFT)
+
+#define SQ_ALU_WORD0_SET_SRC0_SEL(sq_alu_word0_reg, src0_sel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_SEL_MASK) | (src0_sel << SQ_ALU_WORD0_SRC0_SEL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC0_REL(sq_alu_word0_reg, src0_rel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_REL_MASK) | (src0_rel << SQ_ALU_WORD0_SRC0_REL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC0_CHAN(sq_alu_word0_reg, src0_chan) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_CHAN_MASK) | (src0_chan << SQ_ALU_WORD0_SRC0_CHAN_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC0_NEG(sq_alu_word0_reg, src0_neg) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC0_NEG_MASK) | (src0_neg << SQ_ALU_WORD0_SRC0_NEG_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_SEL(sq_alu_word0_reg, src1_sel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_SEL_MASK) | (src1_sel << SQ_ALU_WORD0_SRC1_SEL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_REL(sq_alu_word0_reg, src1_rel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_REL_MASK) | (src1_rel << SQ_ALU_WORD0_SRC1_REL_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_CHAN(sq_alu_word0_reg, src1_chan) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_CHAN_MASK) | (src1_chan << SQ_ALU_WORD0_SRC1_CHAN_SHIFT)
+#define SQ_ALU_WORD0_SET_SRC1_NEG(sq_alu_word0_reg, src1_neg) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_SRC1_NEG_MASK) | (src1_neg << SQ_ALU_WORD0_SRC1_NEG_SHIFT)
+#define SQ_ALU_WORD0_SET_INDEX_MODE(sq_alu_word0_reg, index_mode) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_INDEX_MODE_MASK) | (index_mode << SQ_ALU_WORD0_INDEX_MODE_SHIFT)
+#define SQ_ALU_WORD0_SET_PRED_SEL(sq_alu_word0_reg, pred_sel) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_PRED_SEL_MASK) | (pred_sel << SQ_ALU_WORD0_PRED_SEL_SHIFT)
+#define SQ_ALU_WORD0_SET_LAST(sq_alu_word0_reg, last) \
+     sq_alu_word0_reg = (sq_alu_word0_reg & ~SQ_ALU_WORD0_LAST_MASK) | (last << SQ_ALU_WORD0_LAST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word0_t {
+          unsigned int src0_sel                       : SQ_ALU_WORD0_SRC0_SEL_SIZE;
+          unsigned int src0_rel                       : SQ_ALU_WORD0_SRC0_REL_SIZE;
+          unsigned int src0_chan                      : SQ_ALU_WORD0_SRC0_CHAN_SIZE;
+          unsigned int src0_neg                       : SQ_ALU_WORD0_SRC0_NEG_SIZE;
+          unsigned int src1_sel                       : SQ_ALU_WORD0_SRC1_SEL_SIZE;
+          unsigned int src1_rel                       : SQ_ALU_WORD0_SRC1_REL_SIZE;
+          unsigned int src1_chan                      : SQ_ALU_WORD0_SRC1_CHAN_SIZE;
+          unsigned int src1_neg                       : SQ_ALU_WORD0_SRC1_NEG_SIZE;
+          unsigned int index_mode                     : SQ_ALU_WORD0_INDEX_MODE_SIZE;
+          unsigned int pred_sel                       : SQ_ALU_WORD0_PRED_SEL_SIZE;
+          unsigned int last                           : SQ_ALU_WORD0_LAST_SIZE;
+     } sq_alu_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word0_t {
+          unsigned int last                           : SQ_ALU_WORD0_LAST_SIZE;
+          unsigned int pred_sel                       : SQ_ALU_WORD0_PRED_SEL_SIZE;
+          unsigned int index_mode                     : SQ_ALU_WORD0_INDEX_MODE_SIZE;
+          unsigned int src1_neg                       : SQ_ALU_WORD0_SRC1_NEG_SIZE;
+          unsigned int src1_chan                      : SQ_ALU_WORD0_SRC1_CHAN_SIZE;
+          unsigned int src1_rel                       : SQ_ALU_WORD0_SRC1_REL_SIZE;
+          unsigned int src1_sel                       : SQ_ALU_WORD0_SRC1_SEL_SIZE;
+          unsigned int src0_neg                       : SQ_ALU_WORD0_SRC0_NEG_SIZE;
+          unsigned int src0_chan                      : SQ_ALU_WORD0_SRC0_CHAN_SIZE;
+          unsigned int src0_rel                       : SQ_ALU_WORD0_SRC0_REL_SIZE;
+          unsigned int src0_sel                       : SQ_ALU_WORD0_SRC0_SEL_SIZE;
+     } sq_alu_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word0_t f;
+} sq_alu_word0_u;
+
+
+/*
+ * SQ_ALU_WORD1 struct
+ */
+
+#define SQ_ALU_WORD1_ENCODING_SIZE     3
+#define SQ_ALU_WORD1_BANK_SWIZZLE_SIZE 3
+#define SQ_ALU_WORD1_DST_GPR_SIZE      7
+#define SQ_ALU_WORD1_DST_REL_SIZE      1
+#define SQ_ALU_WORD1_DST_CHAN_SIZE     2
+#define SQ_ALU_WORD1_CLAMP_SIZE        1
+
+#define SQ_ALU_WORD1_ENCODING_SHIFT    15
+#define SQ_ALU_WORD1_BANK_SWIZZLE_SHIFT 18
+#define SQ_ALU_WORD1_DST_GPR_SHIFT     21
+#define SQ_ALU_WORD1_DST_REL_SHIFT     28
+#define SQ_ALU_WORD1_DST_CHAN_SHIFT    29
+#define SQ_ALU_WORD1_CLAMP_SHIFT       31
+
+#define SQ_ALU_WORD1_ENCODING_MASK     0x00038000
+#define SQ_ALU_WORD1_BANK_SWIZZLE_MASK 0x001c0000
+#define SQ_ALU_WORD1_DST_GPR_MASK      0x0fe00000
+#define SQ_ALU_WORD1_DST_REL_MASK      0x10000000
+#define SQ_ALU_WORD1_DST_CHAN_MASK     0x60000000
+#define SQ_ALU_WORD1_CLAMP_MASK        0x80000000
+
+#define SQ_ALU_WORD1_MASK \
+     (SQ_ALU_WORD1_ENCODING_MASK | \
+      SQ_ALU_WORD1_BANK_SWIZZLE_MASK | \
+      SQ_ALU_WORD1_DST_GPR_MASK | \
+      SQ_ALU_WORD1_DST_REL_MASK | \
+      SQ_ALU_WORD1_DST_CHAN_MASK | \
+      SQ_ALU_WORD1_CLAMP_MASK)
+
+#define SQ_ALU_WORD1_DEFAULT           0xcdcd8000
+
+#define SQ_ALU_WORD1_GET_ENCODING(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_ENCODING_MASK) >> SQ_ALU_WORD1_ENCODING_SHIFT)
+#define SQ_ALU_WORD1_GET_BANK_SWIZZLE(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_BANK_SWIZZLE_MASK) >> SQ_ALU_WORD1_BANK_SWIZZLE_SHIFT)
+#define SQ_ALU_WORD1_GET_DST_GPR(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_DST_GPR_MASK) >> SQ_ALU_WORD1_DST_GPR_SHIFT)
+#define SQ_ALU_WORD1_GET_DST_REL(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_DST_REL_MASK) >> SQ_ALU_WORD1_DST_REL_SHIFT)
+#define SQ_ALU_WORD1_GET_DST_CHAN(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_DST_CHAN_MASK) >> SQ_ALU_WORD1_DST_CHAN_SHIFT)
+#define SQ_ALU_WORD1_GET_CLAMP(sq_alu_word1) \
+     ((sq_alu_word1 & SQ_ALU_WORD1_CLAMP_MASK) >> SQ_ALU_WORD1_CLAMP_SHIFT)
+
+#define SQ_ALU_WORD1_SET_ENCODING(sq_alu_word1_reg, encoding) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_ENCODING_MASK) | (encoding << SQ_ALU_WORD1_ENCODING_SHIFT)
+#define SQ_ALU_WORD1_SET_BANK_SWIZZLE(sq_alu_word1_reg, bank_swizzle) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_BANK_SWIZZLE_MASK) | (bank_swizzle << SQ_ALU_WORD1_BANK_SWIZZLE_SHIFT)
+#define SQ_ALU_WORD1_SET_DST_GPR(sq_alu_word1_reg, dst_gpr) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_DST_GPR_MASK) | (dst_gpr << SQ_ALU_WORD1_DST_GPR_SHIFT)
+#define SQ_ALU_WORD1_SET_DST_REL(sq_alu_word1_reg, dst_rel) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_DST_REL_MASK) | (dst_rel << SQ_ALU_WORD1_DST_REL_SHIFT)
+#define SQ_ALU_WORD1_SET_DST_CHAN(sq_alu_word1_reg, dst_chan) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_DST_CHAN_MASK) | (dst_chan << SQ_ALU_WORD1_DST_CHAN_SHIFT)
+#define SQ_ALU_WORD1_SET_CLAMP(sq_alu_word1_reg, clamp) \
+     sq_alu_word1_reg = (sq_alu_word1_reg & ~SQ_ALU_WORD1_CLAMP_MASK) | (clamp << SQ_ALU_WORD1_CLAMP_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_t {
+          unsigned int                                : 15;
+          unsigned int encoding                       : SQ_ALU_WORD1_ENCODING_SIZE;
+          unsigned int bank_swizzle                   : SQ_ALU_WORD1_BANK_SWIZZLE_SIZE;
+          unsigned int dst_gpr                        : SQ_ALU_WORD1_DST_GPR_SIZE;
+          unsigned int dst_rel                        : SQ_ALU_WORD1_DST_REL_SIZE;
+          unsigned int dst_chan                       : SQ_ALU_WORD1_DST_CHAN_SIZE;
+          unsigned int clamp                          : SQ_ALU_WORD1_CLAMP_SIZE;
+     } sq_alu_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_t {
+          unsigned int clamp                          : SQ_ALU_WORD1_CLAMP_SIZE;
+          unsigned int dst_chan                       : SQ_ALU_WORD1_DST_CHAN_SIZE;
+          unsigned int dst_rel                        : SQ_ALU_WORD1_DST_REL_SIZE;
+          unsigned int dst_gpr                        : SQ_ALU_WORD1_DST_GPR_SIZE;
+          unsigned int bank_swizzle                   : SQ_ALU_WORD1_BANK_SWIZZLE_SIZE;
+          unsigned int encoding                       : SQ_ALU_WORD1_ENCODING_SIZE;
+          unsigned int                                : 15;
+     } sq_alu_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word1_t f;
+} sq_alu_word1_u;
+
+
+/*
+ * SQ_ALU_WORD1_OP2_V2 struct
+ */
+
+#define SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE 1
+#define SQ_ALU_WORD1_OP2_V2_OMOD_SIZE  2
+#define SQ_ALU_WORD1_OP2_V2_ALU_INST_SIZE 11
+
+#define SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SHIFT 0
+#define SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SHIFT 1
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SHIFT 2
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SHIFT 3
+#define SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SHIFT 4
+#define SQ_ALU_WORD1_OP2_V2_OMOD_SHIFT 5
+#define SQ_ALU_WORD1_OP2_V2_ALU_INST_SHIFT 7
+
+#define SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK 0x00000001
+#define SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK 0x00000002
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK 0x00000004
+#define SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK 0x00000008
+#define SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK 0x00000010
+#define SQ_ALU_WORD1_OP2_V2_OMOD_MASK  0x00000060
+#define SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK 0x0003ff80
+
+#define SQ_ALU_WORD1_OP2_V2_MASK \
+     (SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK | \
+      SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK | \
+      SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK | \
+      SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK | \
+      SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK | \
+      SQ_ALU_WORD1_OP2_V2_OMOD_MASK | \
+      SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK)
+
+#define SQ_ALU_WORD1_OP2_V2_DEFAULT    0x0001cdcd
+
+#define SQ_ALU_WORD1_OP2_V2_GET_SRC0_ABS(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK) >> SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_SRC1_ABS(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK) >> SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_UPDATE_EXECUTE_MASK(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK) >> SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_UPDATE_PRED(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK) >> SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_WRITE_MASK(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK) >> SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_OMOD(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_OMOD_MASK) >> SQ_ALU_WORD1_OP2_V2_OMOD_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_GET_ALU_INST(sq_alu_word1_op2_v2) \
+     ((sq_alu_word1_op2_v2 & SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK) >> SQ_ALU_WORD1_OP2_V2_ALU_INST_SHIFT)
+
+#define SQ_ALU_WORD1_OP2_V2_SET_SRC0_ABS(sq_alu_word1_op2_v2_reg, src0_abs) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_SRC0_ABS_MASK) | (src0_abs << SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_SRC1_ABS(sq_alu_word1_op2_v2_reg, src1_abs) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_SRC1_ABS_MASK) | (src1_abs << SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_UPDATE_EXECUTE_MASK(sq_alu_word1_op2_v2_reg, update_execute_mask) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_MASK) | (update_execute_mask << SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_UPDATE_PRED(sq_alu_word1_op2_v2_reg, update_pred) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_MASK) | (update_pred << SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_WRITE_MASK(sq_alu_word1_op2_v2_reg, write_mask) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_WRITE_MASK_MASK) | (write_mask << SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_OMOD(sq_alu_word1_op2_v2_reg, omod) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_OMOD_MASK) | (omod << SQ_ALU_WORD1_OP2_V2_OMOD_SHIFT)
+#define SQ_ALU_WORD1_OP2_V2_SET_ALU_INST(sq_alu_word1_op2_v2_reg, alu_inst) \
+     sq_alu_word1_op2_v2_reg = (sq_alu_word1_op2_v2_reg & ~SQ_ALU_WORD1_OP2_V2_ALU_INST_MASK) | (alu_inst << SQ_ALU_WORD1_OP2_V2_ALU_INST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_v2_t {
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP2_V2_ALU_INST_SIZE;
+          unsigned int                                : 14;
+     } sq_alu_word1_op2_v2_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_v2_t {
+          unsigned int                                : 14;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP2_V2_ALU_INST_SIZE;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+     } sq_alu_word1_op2_v2_t;
+
+#endif
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_r6xx_t {
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int fog_export                     : 1;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int alu_inst                       : 10;
+          unsigned int                                : 14;
+     } sq_alu_word1_op2_v1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op2_r6xx_t {
+          unsigned int                                : 14;
+          unsigned int alu_inst                       : 10;
+          unsigned int omod                           : SQ_ALU_WORD1_OP2_V2_OMOD_SIZE;
+          unsigned int fog_export                     : 1;
+          unsigned int write_mask                     : SQ_ALU_WORD1_OP2_V2_WRITE_MASK_SIZE;
+          unsigned int update_pred                    : SQ_ALU_WORD1_OP2_V2_UPDATE_PRED_SIZE;
+          unsigned int update_execute_mask            : SQ_ALU_WORD1_OP2_V2_UPDATE_EXECUTE_MASK_SIZE;
+          unsigned int src1_abs                       : SQ_ALU_WORD1_OP2_V2_SRC1_ABS_SIZE;
+          unsigned int src0_abs                       : SQ_ALU_WORD1_OP2_V2_SRC0_ABS_SIZE;
+     } sq_alu_word1_op2_v1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word1_op2_v2_t f;
+     sq_alu_word1_op2_v1_t f6;
+} sq_alu_word1_op2_v2_u;
+
+
+/*
+ * SQ_ALU_WORD1_OP3 struct
+ */
+
+#define SQ_ALU_WORD1_OP3_SRC2_SEL_SIZE 9
+#define SQ_ALU_WORD1_OP3_SRC2_REL_SIZE 1
+#define SQ_ALU_WORD1_OP3_SRC2_CHAN_SIZE 2
+#define SQ_ALU_WORD1_OP3_SRC2_NEG_SIZE 1
+#define SQ_ALU_WORD1_OP3_ALU_INST_SIZE 5
+
+#define SQ_ALU_WORD1_OP3_SRC2_SEL_SHIFT 0
+#define SQ_ALU_WORD1_OP3_SRC2_REL_SHIFT 9
+#define SQ_ALU_WORD1_OP3_SRC2_CHAN_SHIFT 10
+#define SQ_ALU_WORD1_OP3_SRC2_NEG_SHIFT 12
+#define SQ_ALU_WORD1_OP3_ALU_INST_SHIFT 13
+
+#define SQ_ALU_WORD1_OP3_SRC2_SEL_MASK 0x000001ff
+#define SQ_ALU_WORD1_OP3_SRC2_REL_MASK 0x00000200
+#define SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK 0x00000c00
+#define SQ_ALU_WORD1_OP3_SRC2_NEG_MASK 0x00001000
+#define SQ_ALU_WORD1_OP3_ALU_INST_MASK 0x0003e000
+
+#define SQ_ALU_WORD1_OP3_MASK \
+     (SQ_ALU_WORD1_OP3_SRC2_SEL_MASK | \
+      SQ_ALU_WORD1_OP3_SRC2_REL_MASK | \
+      SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK | \
+      SQ_ALU_WORD1_OP3_SRC2_NEG_MASK | \
+      SQ_ALU_WORD1_OP3_ALU_INST_MASK)
+
+#define SQ_ALU_WORD1_OP3_DEFAULT       0x0001cdcd
+
+#define SQ_ALU_WORD1_OP3_GET_SRC2_SEL(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_SEL_MASK) >> SQ_ALU_WORD1_OP3_SRC2_SEL_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_SRC2_REL(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_REL_MASK) >> SQ_ALU_WORD1_OP3_SRC2_REL_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_SRC2_CHAN(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK) >> SQ_ALU_WORD1_OP3_SRC2_CHAN_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_SRC2_NEG(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_SRC2_NEG_MASK) >> SQ_ALU_WORD1_OP3_SRC2_NEG_SHIFT)
+#define SQ_ALU_WORD1_OP3_GET_ALU_INST(sq_alu_word1_op3) \
+     ((sq_alu_word1_op3 & SQ_ALU_WORD1_OP3_ALU_INST_MASK) >> SQ_ALU_WORD1_OP3_ALU_INST_SHIFT)
+
+#define SQ_ALU_WORD1_OP3_SET_SRC2_SEL(sq_alu_word1_op3_reg, src2_sel) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_SEL_MASK) | (src2_sel << SQ_ALU_WORD1_OP3_SRC2_SEL_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_SRC2_REL(sq_alu_word1_op3_reg, src2_rel) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_REL_MASK) | (src2_rel << SQ_ALU_WORD1_OP3_SRC2_REL_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_SRC2_CHAN(sq_alu_word1_op3_reg, src2_chan) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_CHAN_MASK) | (src2_chan << SQ_ALU_WORD1_OP3_SRC2_CHAN_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_SRC2_NEG(sq_alu_word1_op3_reg, src2_neg) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_SRC2_NEG_MASK) | (src2_neg << SQ_ALU_WORD1_OP3_SRC2_NEG_SHIFT)
+#define SQ_ALU_WORD1_OP3_SET_ALU_INST(sq_alu_word1_op3_reg, alu_inst) \
+     sq_alu_word1_op3_reg = (sq_alu_word1_op3_reg & ~SQ_ALU_WORD1_OP3_ALU_INST_MASK) | (alu_inst << SQ_ALU_WORD1_OP3_ALU_INST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op3_t {
+          unsigned int src2_sel                       : SQ_ALU_WORD1_OP3_SRC2_SEL_SIZE;
+          unsigned int src2_rel                       : SQ_ALU_WORD1_OP3_SRC2_REL_SIZE;
+          unsigned int src2_chan                      : SQ_ALU_WORD1_OP3_SRC2_CHAN_SIZE;
+          unsigned int src2_neg                       : SQ_ALU_WORD1_OP3_SRC2_NEG_SIZE;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP3_ALU_INST_SIZE;
+          unsigned int                                : 14;
+     } sq_alu_word1_op3_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_alu_word1_op3_t {
+          unsigned int                                : 14;
+          unsigned int alu_inst                       : SQ_ALU_WORD1_OP3_ALU_INST_SIZE;
+          unsigned int src2_neg                       : SQ_ALU_WORD1_OP3_SRC2_NEG_SIZE;
+          unsigned int src2_chan                      : SQ_ALU_WORD1_OP3_SRC2_CHAN_SIZE;
+          unsigned int src2_rel                       : SQ_ALU_WORD1_OP3_SRC2_REL_SIZE;
+          unsigned int src2_sel                       : SQ_ALU_WORD1_OP3_SRC2_SEL_SIZE;
+     } sq_alu_word1_op3_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_alu_word1_op3_t f;
+} sq_alu_word1_op3_u;
+
+
+/*
+ * SQ_TEX_WORD0 struct
+ */
+
+#define SQ_TEX_WORD0_TEX_INST_SIZE     5
+#define SQ_TEX_WORD0_BC_FRAC_MODE_SIZE 1
+#define SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SIZE 1
+#define SQ_TEX_WORD0_RESOURCE_ID_SIZE  8
+#define SQ_TEX_WORD0_SRC_GPR_SIZE      7
+#define SQ_TEX_WORD0_SRC_REL_SIZE      1
+#define SQ_TEX_WORD0_ALT_CONST_SIZE    1
+
+#define SQ_TEX_WORD0_TEX_INST_SHIFT    0
+#define SQ_TEX_WORD0_BC_FRAC_MODE_SHIFT 5
+#define SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SHIFT 7
+#define SQ_TEX_WORD0_RESOURCE_ID_SHIFT 8
+#define SQ_TEX_WORD0_SRC_GPR_SHIFT     16
+#define SQ_TEX_WORD0_SRC_REL_SHIFT     23
+#define SQ_TEX_WORD0_ALT_CONST_SHIFT   24
+
+#define SQ_TEX_WORD0_TEX_INST_MASK     0x0000001f
+#define SQ_TEX_WORD0_BC_FRAC_MODE_MASK 0x00000020
+#define SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK 0x00000080
+#define SQ_TEX_WORD0_RESOURCE_ID_MASK  0x0000ff00
+#define SQ_TEX_WORD0_SRC_GPR_MASK      0x007f0000
+#define SQ_TEX_WORD0_SRC_REL_MASK      0x00800000
+#define SQ_TEX_WORD0_ALT_CONST_MASK    0x01000000
+
+#define SQ_TEX_WORD0_MASK \
+     (SQ_TEX_WORD0_TEX_INST_MASK | \
+      SQ_TEX_WORD0_BC_FRAC_MODE_MASK | \
+      SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK | \
+      SQ_TEX_WORD0_RESOURCE_ID_MASK | \
+      SQ_TEX_WORD0_SRC_GPR_MASK | \
+      SQ_TEX_WORD0_SRC_REL_MASK | \
+      SQ_TEX_WORD0_ALT_CONST_MASK)
+
+#define SQ_TEX_WORD0_DEFAULT           0x01cdcd8d
+
+#define SQ_TEX_WORD0_GET_TEX_INST(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_TEX_INST_MASK) >> SQ_TEX_WORD0_TEX_INST_SHIFT)
+#define SQ_TEX_WORD0_GET_BC_FRAC_MODE(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_BC_FRAC_MODE_MASK) >> SQ_TEX_WORD0_BC_FRAC_MODE_SHIFT)
+#define SQ_TEX_WORD0_GET_FETCH_WHOLE_QUAD(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK) >> SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_TEX_WORD0_GET_RESOURCE_ID(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_RESOURCE_ID_MASK) >> SQ_TEX_WORD0_RESOURCE_ID_SHIFT)
+#define SQ_TEX_WORD0_GET_SRC_GPR(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_SRC_GPR_MASK) >> SQ_TEX_WORD0_SRC_GPR_SHIFT)
+#define SQ_TEX_WORD0_GET_SRC_REL(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_SRC_REL_MASK) >> SQ_TEX_WORD0_SRC_REL_SHIFT)
+#define SQ_TEX_WORD0_GET_ALT_CONST(sq_tex_word0) \
+     ((sq_tex_word0 & SQ_TEX_WORD0_ALT_CONST_MASK) >> SQ_TEX_WORD0_ALT_CONST_SHIFT)
+
+#define SQ_TEX_WORD0_SET_TEX_INST(sq_tex_word0_reg, tex_inst) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_TEX_INST_MASK) | (tex_inst << SQ_TEX_WORD0_TEX_INST_SHIFT)
+#define SQ_TEX_WORD0_SET_BC_FRAC_MODE(sq_tex_word0_reg, bc_frac_mode) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_BC_FRAC_MODE_MASK) | (bc_frac_mode << SQ_TEX_WORD0_BC_FRAC_MODE_SHIFT)
+#define SQ_TEX_WORD0_SET_FETCH_WHOLE_QUAD(sq_tex_word0_reg, fetch_whole_quad) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_FETCH_WHOLE_QUAD_MASK) | (fetch_whole_quad << SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_TEX_WORD0_SET_RESOURCE_ID(sq_tex_word0_reg, resource_id) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_RESOURCE_ID_MASK) | (resource_id << SQ_TEX_WORD0_RESOURCE_ID_SHIFT)
+#define SQ_TEX_WORD0_SET_SRC_GPR(sq_tex_word0_reg, src_gpr) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_SRC_GPR_MASK) | (src_gpr << SQ_TEX_WORD0_SRC_GPR_SHIFT)
+#define SQ_TEX_WORD0_SET_SRC_REL(sq_tex_word0_reg, src_rel) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_SRC_REL_MASK) | (src_rel << SQ_TEX_WORD0_SRC_REL_SHIFT)
+#define SQ_TEX_WORD0_SET_ALT_CONST(sq_tex_word0_reg, alt_const) \
+     sq_tex_word0_reg = (sq_tex_word0_reg & ~SQ_TEX_WORD0_ALT_CONST_MASK) | (alt_const << SQ_TEX_WORD0_ALT_CONST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_tex_word0_t {
+          unsigned int tex_inst                       : SQ_TEX_WORD0_TEX_INST_SIZE;
+          unsigned int bc_frac_mode                   : SQ_TEX_WORD0_BC_FRAC_MODE_SIZE;
+          unsigned int                                : 1;
+          unsigned int fetch_whole_quad               : SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int resource_id                    : SQ_TEX_WORD0_RESOURCE_ID_SIZE;
+          unsigned int src_gpr                        : SQ_TEX_WORD0_SRC_GPR_SIZE;
+          unsigned int src_rel                        : SQ_TEX_WORD0_SRC_REL_SIZE;
+          unsigned int alt_const                      : SQ_TEX_WORD0_ALT_CONST_SIZE;
+          unsigned int                                : 7;
+     } sq_tex_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_tex_word0_t {
+          unsigned int                                : 7;
+          unsigned int alt_const                      : SQ_TEX_WORD0_ALT_CONST_SIZE;
+          unsigned int src_rel                        : SQ_TEX_WORD0_SRC_REL_SIZE;
+          unsigned int src_gpr                        : SQ_TEX_WORD0_SRC_GPR_SIZE;
+          unsigned int resource_id                    : SQ_TEX_WORD0_RESOURCE_ID_SIZE;
+          unsigned int fetch_whole_quad               : SQ_TEX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int                                : 1;
+          unsigned int bc_frac_mode                   : SQ_TEX_WORD0_BC_FRAC_MODE_SIZE;
+          unsigned int tex_inst                       : SQ_TEX_WORD0_TEX_INST_SIZE;
+     } sq_tex_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_tex_word0_t f;
+} sq_tex_word0_u;
+
+
+/*
+ * SQ_TEX_WORD1 struct
+ */
+
+#define SQ_TEX_WORD1_DST_GPR_SIZE      7
+#define SQ_TEX_WORD1_DST_REL_SIZE      1
+#define SQ_TEX_WORD1_DST_SEL_X_SIZE    3
+#define SQ_TEX_WORD1_DST_SEL_Y_SIZE    3
+#define SQ_TEX_WORD1_DST_SEL_Z_SIZE    3
+#define SQ_TEX_WORD1_DST_SEL_W_SIZE    3
+#define SQ_TEX_WORD1_LOD_BIAS_SIZE     7
+#define SQ_TEX_WORD1_COORD_TYPE_X_SIZE 1
+#define SQ_TEX_WORD1_COORD_TYPE_Y_SIZE 1
+#define SQ_TEX_WORD1_COORD_TYPE_Z_SIZE 1
+#define SQ_TEX_WORD1_COORD_TYPE_W_SIZE 1
+
+#define SQ_TEX_WORD1_DST_GPR_SHIFT     0
+#define SQ_TEX_WORD1_DST_REL_SHIFT     7
+#define SQ_TEX_WORD1_DST_SEL_X_SHIFT   9
+#define SQ_TEX_WORD1_DST_SEL_Y_SHIFT   12
+#define SQ_TEX_WORD1_DST_SEL_Z_SHIFT   15
+#define SQ_TEX_WORD1_DST_SEL_W_SHIFT   18
+#define SQ_TEX_WORD1_LOD_BIAS_SHIFT    21
+#define SQ_TEX_WORD1_COORD_TYPE_X_SHIFT 28
+#define SQ_TEX_WORD1_COORD_TYPE_Y_SHIFT 29
+#define SQ_TEX_WORD1_COORD_TYPE_Z_SHIFT 30
+#define SQ_TEX_WORD1_COORD_TYPE_W_SHIFT 31
+
+#define SQ_TEX_WORD1_DST_GPR_MASK      0x0000007f
+#define SQ_TEX_WORD1_DST_REL_MASK      0x00000080
+#define SQ_TEX_WORD1_DST_SEL_X_MASK    0x00000e00
+#define SQ_TEX_WORD1_DST_SEL_Y_MASK    0x00007000
+#define SQ_TEX_WORD1_DST_SEL_Z_MASK    0x00038000
+#define SQ_TEX_WORD1_DST_SEL_W_MASK    0x001c0000
+#define SQ_TEX_WORD1_LOD_BIAS_MASK     0x0fe00000
+#define SQ_TEX_WORD1_COORD_TYPE_X_MASK 0x10000000
+#define SQ_TEX_WORD1_COORD_TYPE_Y_MASK 0x20000000
+#define SQ_TEX_WORD1_COORD_TYPE_Z_MASK 0x40000000
+#define SQ_TEX_WORD1_COORD_TYPE_W_MASK 0x80000000
+
+#define SQ_TEX_WORD1_MASK \
+     (SQ_TEX_WORD1_DST_GPR_MASK | \
+      SQ_TEX_WORD1_DST_REL_MASK | \
+      SQ_TEX_WORD1_DST_SEL_X_MASK | \
+      SQ_TEX_WORD1_DST_SEL_Y_MASK | \
+      SQ_TEX_WORD1_DST_SEL_Z_MASK | \
+      SQ_TEX_WORD1_DST_SEL_W_MASK | \
+      SQ_TEX_WORD1_LOD_BIAS_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_X_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_Y_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_Z_MASK | \
+      SQ_TEX_WORD1_COORD_TYPE_W_MASK)
+
+#define SQ_TEX_WORD1_DEFAULT           0xcdcdcccd
+
+#define SQ_TEX_WORD1_GET_DST_GPR(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_GPR_MASK) >> SQ_TEX_WORD1_DST_GPR_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_REL(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_REL_MASK) >> SQ_TEX_WORD1_DST_REL_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_X(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_X_MASK) >> SQ_TEX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_Y(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_Y_MASK) >> SQ_TEX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_Z(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_Z_MASK) >> SQ_TEX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_TEX_WORD1_GET_DST_SEL_W(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_DST_SEL_W_MASK) >> SQ_TEX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_TEX_WORD1_GET_LOD_BIAS(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_LOD_BIAS_MASK) >> SQ_TEX_WORD1_LOD_BIAS_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_X(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_X_MASK) >> SQ_TEX_WORD1_COORD_TYPE_X_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_Y(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_Y_MASK) >> SQ_TEX_WORD1_COORD_TYPE_Y_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_Z(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_Z_MASK) >> SQ_TEX_WORD1_COORD_TYPE_Z_SHIFT)
+#define SQ_TEX_WORD1_GET_COORD_TYPE_W(sq_tex_word1) \
+     ((sq_tex_word1 & SQ_TEX_WORD1_COORD_TYPE_W_MASK) >> SQ_TEX_WORD1_COORD_TYPE_W_SHIFT)
+
+#define SQ_TEX_WORD1_SET_DST_GPR(sq_tex_word1_reg, dst_gpr) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_GPR_MASK) | (dst_gpr << SQ_TEX_WORD1_DST_GPR_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_REL(sq_tex_word1_reg, dst_rel) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_REL_MASK) | (dst_rel << SQ_TEX_WORD1_DST_REL_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_X(sq_tex_word1_reg, dst_sel_x) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_X_MASK) | (dst_sel_x << SQ_TEX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_Y(sq_tex_word1_reg, dst_sel_y) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_Y_MASK) | (dst_sel_y << SQ_TEX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_Z(sq_tex_word1_reg, dst_sel_z) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_Z_MASK) | (dst_sel_z << SQ_TEX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_TEX_WORD1_SET_DST_SEL_W(sq_tex_word1_reg, dst_sel_w) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_DST_SEL_W_MASK) | (dst_sel_w << SQ_TEX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_TEX_WORD1_SET_LOD_BIAS(sq_tex_word1_reg, lod_bias) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_LOD_BIAS_MASK) | (lod_bias << SQ_TEX_WORD1_LOD_BIAS_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_X(sq_tex_word1_reg, coord_type_x) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_X_MASK) | (coord_type_x << SQ_TEX_WORD1_COORD_TYPE_X_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_Y(sq_tex_word1_reg, coord_type_y) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_Y_MASK) | (coord_type_y << SQ_TEX_WORD1_COORD_TYPE_Y_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_Z(sq_tex_word1_reg, coord_type_z) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_Z_MASK) | (coord_type_z << SQ_TEX_WORD1_COORD_TYPE_Z_SHIFT)
+#define SQ_TEX_WORD1_SET_COORD_TYPE_W(sq_tex_word1_reg, coord_type_w) \
+     sq_tex_word1_reg = (sq_tex_word1_reg & ~SQ_TEX_WORD1_COORD_TYPE_W_MASK) | (coord_type_w << SQ_TEX_WORD1_COORD_TYPE_W_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_tex_word1_t {
+          unsigned int dst_gpr                        : SQ_TEX_WORD1_DST_GPR_SIZE;
+          unsigned int dst_rel                        : SQ_TEX_WORD1_DST_REL_SIZE;
+          unsigned int                                : 1;
+          unsigned int dst_sel_x                      : SQ_TEX_WORD1_DST_SEL_X_SIZE;
+          unsigned int dst_sel_y                      : SQ_TEX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_z                      : SQ_TEX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_w                      : SQ_TEX_WORD1_DST_SEL_W_SIZE;
+          unsigned int lod_bias                       : SQ_TEX_WORD1_LOD_BIAS_SIZE;
+          unsigned int coord_type_x                   : SQ_TEX_WORD1_COORD_TYPE_X_SIZE;
+          unsigned int coord_type_y                   : SQ_TEX_WORD1_COORD_TYPE_Y_SIZE;
+          unsigned int coord_type_z                   : SQ_TEX_WORD1_COORD_TYPE_Z_SIZE;
+          unsigned int coord_type_w                   : SQ_TEX_WORD1_COORD_TYPE_W_SIZE;
+     } sq_tex_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_tex_word1_t {
+          unsigned int coord_type_w                   : SQ_TEX_WORD1_COORD_TYPE_W_SIZE;
+          unsigned int coord_type_z                   : SQ_TEX_WORD1_COORD_TYPE_Z_SIZE;
+          unsigned int coord_type_y                   : SQ_TEX_WORD1_COORD_TYPE_Y_SIZE;
+          unsigned int coord_type_x                   : SQ_TEX_WORD1_COORD_TYPE_X_SIZE;
+          unsigned int lod_bias                       : SQ_TEX_WORD1_LOD_BIAS_SIZE;
+          unsigned int dst_sel_w                      : SQ_TEX_WORD1_DST_SEL_W_SIZE;
+          unsigned int dst_sel_z                      : SQ_TEX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_y                      : SQ_TEX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_x                      : SQ_TEX_WORD1_DST_SEL_X_SIZE;
+          unsigned int                                : 1;
+          unsigned int dst_rel                        : SQ_TEX_WORD1_DST_REL_SIZE;
+          unsigned int dst_gpr                        : SQ_TEX_WORD1_DST_GPR_SIZE;
+     } sq_tex_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_tex_word1_t f;
+} sq_tex_word1_u;
+
+
+/*
+ * SQ_TEX_WORD2 struct
+ */
+
+#define SQ_TEX_WORD2_OFFSET_X_SIZE     5
+#define SQ_TEX_WORD2_OFFSET_Y_SIZE     5
+#define SQ_TEX_WORD2_OFFSET_Z_SIZE     5
+#define SQ_TEX_WORD2_SAMPLER_ID_SIZE   5
+#define SQ_TEX_WORD2_SRC_SEL_X_SIZE    3
+#define SQ_TEX_WORD2_SRC_SEL_Y_SIZE    3
+#define SQ_TEX_WORD2_SRC_SEL_Z_SIZE    3
+#define SQ_TEX_WORD2_SRC_SEL_W_SIZE    3
+
+#define SQ_TEX_WORD2_OFFSET_X_SHIFT    0
+#define SQ_TEX_WORD2_OFFSET_Y_SHIFT    5
+#define SQ_TEX_WORD2_OFFSET_Z_SHIFT    10
+#define SQ_TEX_WORD2_SAMPLER_ID_SHIFT  15
+#define SQ_TEX_WORD2_SRC_SEL_X_SHIFT   20
+#define SQ_TEX_WORD2_SRC_SEL_Y_SHIFT   23
+#define SQ_TEX_WORD2_SRC_SEL_Z_SHIFT   26
+#define SQ_TEX_WORD2_SRC_SEL_W_SHIFT   29
+
+#define SQ_TEX_WORD2_OFFSET_X_MASK     0x0000001f
+#define SQ_TEX_WORD2_OFFSET_Y_MASK     0x000003e0
+#define SQ_TEX_WORD2_OFFSET_Z_MASK     0x00007c00
+#define SQ_TEX_WORD2_SAMPLER_ID_MASK   0x000f8000
+#define SQ_TEX_WORD2_SRC_SEL_X_MASK    0x00700000
+#define SQ_TEX_WORD2_SRC_SEL_Y_MASK    0x03800000
+#define SQ_TEX_WORD2_SRC_SEL_Z_MASK    0x1c000000
+#define SQ_TEX_WORD2_SRC_SEL_W_MASK    0xe0000000
+
+#define SQ_TEX_WORD2_MASK \
+     (SQ_TEX_WORD2_OFFSET_X_MASK | \
+      SQ_TEX_WORD2_OFFSET_Y_MASK | \
+      SQ_TEX_WORD2_OFFSET_Z_MASK | \
+      SQ_TEX_WORD2_SAMPLER_ID_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_X_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_Y_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_Z_MASK | \
+      SQ_TEX_WORD2_SRC_SEL_W_MASK)
+
+#define SQ_TEX_WORD2_DEFAULT           0xcdcdcdcd
+
+#define SQ_TEX_WORD2_GET_OFFSET_X(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_OFFSET_X_MASK) >> SQ_TEX_WORD2_OFFSET_X_SHIFT)
+#define SQ_TEX_WORD2_GET_OFFSET_Y(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_OFFSET_Y_MASK) >> SQ_TEX_WORD2_OFFSET_Y_SHIFT)
+#define SQ_TEX_WORD2_GET_OFFSET_Z(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_OFFSET_Z_MASK) >> SQ_TEX_WORD2_OFFSET_Z_SHIFT)
+#define SQ_TEX_WORD2_GET_SAMPLER_ID(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SAMPLER_ID_MASK) >> SQ_TEX_WORD2_SAMPLER_ID_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_X(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_X_MASK) >> SQ_TEX_WORD2_SRC_SEL_X_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_Y(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_Y_MASK) >> SQ_TEX_WORD2_SRC_SEL_Y_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_Z(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_Z_MASK) >> SQ_TEX_WORD2_SRC_SEL_Z_SHIFT)
+#define SQ_TEX_WORD2_GET_SRC_SEL_W(sq_tex_word2) \
+     ((sq_tex_word2 & SQ_TEX_WORD2_SRC_SEL_W_MASK) >> SQ_TEX_WORD2_SRC_SEL_W_SHIFT)
+
+#define SQ_TEX_WORD2_SET_OFFSET_X(sq_tex_word2_reg, offset_x) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_OFFSET_X_MASK) | (offset_x << SQ_TEX_WORD2_OFFSET_X_SHIFT)
+#define SQ_TEX_WORD2_SET_OFFSET_Y(sq_tex_word2_reg, offset_y) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_OFFSET_Y_MASK) | (offset_y << SQ_TEX_WORD2_OFFSET_Y_SHIFT)
+#define SQ_TEX_WORD2_SET_OFFSET_Z(sq_tex_word2_reg, offset_z) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_OFFSET_Z_MASK) | (offset_z << SQ_TEX_WORD2_OFFSET_Z_SHIFT)
+#define SQ_TEX_WORD2_SET_SAMPLER_ID(sq_tex_word2_reg, sampler_id) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SAMPLER_ID_MASK) | (sampler_id << SQ_TEX_WORD2_SAMPLER_ID_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_X(sq_tex_word2_reg, src_sel_x) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_X_MASK) | (src_sel_x << SQ_TEX_WORD2_SRC_SEL_X_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_Y(sq_tex_word2_reg, src_sel_y) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_Y_MASK) | (src_sel_y << SQ_TEX_WORD2_SRC_SEL_Y_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_Z(sq_tex_word2_reg, src_sel_z) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_Z_MASK) | (src_sel_z << SQ_TEX_WORD2_SRC_SEL_Z_SHIFT)
+#define SQ_TEX_WORD2_SET_SRC_SEL_W(sq_tex_word2_reg, src_sel_w) \
+     sq_tex_word2_reg = (sq_tex_word2_reg & ~SQ_TEX_WORD2_SRC_SEL_W_MASK) | (src_sel_w << SQ_TEX_WORD2_SRC_SEL_W_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_tex_word2_t {
+          unsigned int offset_x                       : SQ_TEX_WORD2_OFFSET_X_SIZE;
+          unsigned int offset_y                       : SQ_TEX_WORD2_OFFSET_Y_SIZE;
+          unsigned int offset_z                       : SQ_TEX_WORD2_OFFSET_Z_SIZE;
+          unsigned int sampler_id                     : SQ_TEX_WORD2_SAMPLER_ID_SIZE;
+          unsigned int src_sel_x                      : SQ_TEX_WORD2_SRC_SEL_X_SIZE;
+          unsigned int src_sel_y                      : SQ_TEX_WORD2_SRC_SEL_Y_SIZE;
+          unsigned int src_sel_z                      : SQ_TEX_WORD2_SRC_SEL_Z_SIZE;
+          unsigned int src_sel_w                      : SQ_TEX_WORD2_SRC_SEL_W_SIZE;
+     } sq_tex_word2_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_tex_word2_t {
+          unsigned int src_sel_w                      : SQ_TEX_WORD2_SRC_SEL_W_SIZE;
+          unsigned int src_sel_z                      : SQ_TEX_WORD2_SRC_SEL_Z_SIZE;
+          unsigned int src_sel_y                      : SQ_TEX_WORD2_SRC_SEL_Y_SIZE;
+          unsigned int src_sel_x                      : SQ_TEX_WORD2_SRC_SEL_X_SIZE;
+          unsigned int sampler_id                     : SQ_TEX_WORD2_SAMPLER_ID_SIZE;
+          unsigned int offset_z                       : SQ_TEX_WORD2_OFFSET_Z_SIZE;
+          unsigned int offset_y                       : SQ_TEX_WORD2_OFFSET_Y_SIZE;
+          unsigned int offset_x                       : SQ_TEX_WORD2_OFFSET_X_SIZE;
+     } sq_tex_word2_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_tex_word2_t f;
+} sq_tex_word2_u;
+
+
+/*
+ * SQ_VTX_WORD0 struct
+ */
+
+#define SQ_VTX_WORD0_VTX_INST_SIZE     5
+#define SQ_VTX_WORD0_FETCH_TYPE_SIZE   2
+#define SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SIZE 1
+#define SQ_VTX_WORD0_BUFFER_ID_SIZE    8
+#define SQ_VTX_WORD0_SRC_GPR_SIZE      7
+#define SQ_VTX_WORD0_SRC_REL_SIZE      1
+#define SQ_VTX_WORD0_SRC_SEL_X_SIZE    2
+#define SQ_VTX_WORD0_MEGA_FETCH_COUNT_SIZE 6
+
+#define SQ_VTX_WORD0_VTX_INST_SHIFT    0
+#define SQ_VTX_WORD0_FETCH_TYPE_SHIFT  5
+#define SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SHIFT 7
+#define SQ_VTX_WORD0_BUFFER_ID_SHIFT   8
+#define SQ_VTX_WORD0_SRC_GPR_SHIFT     16
+#define SQ_VTX_WORD0_SRC_REL_SHIFT     23
+#define SQ_VTX_WORD0_SRC_SEL_X_SHIFT   24
+#define SQ_VTX_WORD0_MEGA_FETCH_COUNT_SHIFT 26
+
+#define SQ_VTX_WORD0_VTX_INST_MASK     0x0000001f
+#define SQ_VTX_WORD0_FETCH_TYPE_MASK   0x00000060
+#define SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK 0x00000080
+#define SQ_VTX_WORD0_BUFFER_ID_MASK    0x0000ff00
+#define SQ_VTX_WORD0_SRC_GPR_MASK      0x007f0000
+#define SQ_VTX_WORD0_SRC_REL_MASK      0x00800000
+#define SQ_VTX_WORD0_SRC_SEL_X_MASK    0x03000000
+#define SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK 0xfc000000
+
+#define SQ_VTX_WORD0_MASK \
+     (SQ_VTX_WORD0_VTX_INST_MASK | \
+      SQ_VTX_WORD0_FETCH_TYPE_MASK | \
+      SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK | \
+      SQ_VTX_WORD0_BUFFER_ID_MASK | \
+      SQ_VTX_WORD0_SRC_GPR_MASK | \
+      SQ_VTX_WORD0_SRC_REL_MASK | \
+      SQ_VTX_WORD0_SRC_SEL_X_MASK | \
+      SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK)
+
+#define SQ_VTX_WORD0_DEFAULT           0xcdcdcdcd
+
+#define SQ_VTX_WORD0_GET_VTX_INST(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_VTX_INST_MASK) >> SQ_VTX_WORD0_VTX_INST_SHIFT)
+#define SQ_VTX_WORD0_GET_FETCH_TYPE(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_FETCH_TYPE_MASK) >> SQ_VTX_WORD0_FETCH_TYPE_SHIFT)
+#define SQ_VTX_WORD0_GET_FETCH_WHOLE_QUAD(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK) >> SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_VTX_WORD0_GET_BUFFER_ID(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_BUFFER_ID_MASK) >> SQ_VTX_WORD0_BUFFER_ID_SHIFT)
+#define SQ_VTX_WORD0_GET_SRC_GPR(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_SRC_GPR_MASK) >> SQ_VTX_WORD0_SRC_GPR_SHIFT)
+#define SQ_VTX_WORD0_GET_SRC_REL(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_SRC_REL_MASK) >> SQ_VTX_WORD0_SRC_REL_SHIFT)
+#define SQ_VTX_WORD0_GET_SRC_SEL_X(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_SRC_SEL_X_MASK) >> SQ_VTX_WORD0_SRC_SEL_X_SHIFT)
+#define SQ_VTX_WORD0_GET_MEGA_FETCH_COUNT(sq_vtx_word0) \
+     ((sq_vtx_word0 & SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK) >> SQ_VTX_WORD0_MEGA_FETCH_COUNT_SHIFT)
+
+#define SQ_VTX_WORD0_SET_VTX_INST(sq_vtx_word0_reg, vtx_inst) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_VTX_INST_MASK) | (vtx_inst << SQ_VTX_WORD0_VTX_INST_SHIFT)
+#define SQ_VTX_WORD0_SET_FETCH_TYPE(sq_vtx_word0_reg, fetch_type) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_FETCH_TYPE_MASK) | (fetch_type << SQ_VTX_WORD0_FETCH_TYPE_SHIFT)
+#define SQ_VTX_WORD0_SET_FETCH_WHOLE_QUAD(sq_vtx_word0_reg, fetch_whole_quad) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_FETCH_WHOLE_QUAD_MASK) | (fetch_whole_quad << SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SHIFT)
+#define SQ_VTX_WORD0_SET_BUFFER_ID(sq_vtx_word0_reg, buffer_id) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_BUFFER_ID_MASK) | (buffer_id << SQ_VTX_WORD0_BUFFER_ID_SHIFT)
+#define SQ_VTX_WORD0_SET_SRC_GPR(sq_vtx_word0_reg, src_gpr) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_SRC_GPR_MASK) | (src_gpr << SQ_VTX_WORD0_SRC_GPR_SHIFT)
+#define SQ_VTX_WORD0_SET_SRC_REL(sq_vtx_word0_reg, src_rel) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_SRC_REL_MASK) | (src_rel << SQ_VTX_WORD0_SRC_REL_SHIFT)
+#define SQ_VTX_WORD0_SET_SRC_SEL_X(sq_vtx_word0_reg, src_sel_x) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_SRC_SEL_X_MASK) | (src_sel_x << SQ_VTX_WORD0_SRC_SEL_X_SHIFT)
+#define SQ_VTX_WORD0_SET_MEGA_FETCH_COUNT(sq_vtx_word0_reg, mega_fetch_count) \
+     sq_vtx_word0_reg = (sq_vtx_word0_reg & ~SQ_VTX_WORD0_MEGA_FETCH_COUNT_MASK) | (mega_fetch_count << SQ_VTX_WORD0_MEGA_FETCH_COUNT_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word0_t {
+          unsigned int vtx_inst                       : SQ_VTX_WORD0_VTX_INST_SIZE;
+          unsigned int fetch_type                     : SQ_VTX_WORD0_FETCH_TYPE_SIZE;
+          unsigned int fetch_whole_quad               : SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int buffer_id                      : SQ_VTX_WORD0_BUFFER_ID_SIZE;
+          unsigned int src_gpr                        : SQ_VTX_WORD0_SRC_GPR_SIZE;
+          unsigned int src_rel                        : SQ_VTX_WORD0_SRC_REL_SIZE;
+          unsigned int src_sel_x                      : SQ_VTX_WORD0_SRC_SEL_X_SIZE;
+          unsigned int mega_fetch_count               : SQ_VTX_WORD0_MEGA_FETCH_COUNT_SIZE;
+     } sq_vtx_word0_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word0_t {
+          unsigned int mega_fetch_count               : SQ_VTX_WORD0_MEGA_FETCH_COUNT_SIZE;
+          unsigned int src_sel_x                      : SQ_VTX_WORD0_SRC_SEL_X_SIZE;
+          unsigned int src_rel                        : SQ_VTX_WORD0_SRC_REL_SIZE;
+          unsigned int src_gpr                        : SQ_VTX_WORD0_SRC_GPR_SIZE;
+          unsigned int buffer_id                      : SQ_VTX_WORD0_BUFFER_ID_SIZE;
+          unsigned int fetch_whole_quad               : SQ_VTX_WORD0_FETCH_WHOLE_QUAD_SIZE;
+          unsigned int fetch_type                     : SQ_VTX_WORD0_FETCH_TYPE_SIZE;
+          unsigned int vtx_inst                       : SQ_VTX_WORD0_VTX_INST_SIZE;
+     } sq_vtx_word0_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word0_t f;
+} sq_vtx_word0_u;
+
+
+/*
+ * SQ_VTX_WORD1 struct
+ */
+
+#define SQ_VTX_WORD1_DST_SEL_X_SIZE    3
+#define SQ_VTX_WORD1_DST_SEL_Y_SIZE    3
+#define SQ_VTX_WORD1_DST_SEL_Z_SIZE    3
+#define SQ_VTX_WORD1_DST_SEL_W_SIZE    3
+#define SQ_VTX_WORD1_USE_CONST_FIELDS_SIZE 1
+#define SQ_VTX_WORD1_DATA_FORMAT_SIZE  6
+#define SQ_VTX_WORD1_NUM_FORMAT_ALL_SIZE 2
+#define SQ_VTX_WORD1_FORMAT_COMP_ALL_SIZE 1
+#define SQ_VTX_WORD1_SRF_MODE_ALL_SIZE 1
+
+#define SQ_VTX_WORD1_DST_SEL_X_SHIFT   9
+#define SQ_VTX_WORD1_DST_SEL_Y_SHIFT   12
+#define SQ_VTX_WORD1_DST_SEL_Z_SHIFT   15
+#define SQ_VTX_WORD1_DST_SEL_W_SHIFT   18
+#define SQ_VTX_WORD1_USE_CONST_FIELDS_SHIFT 21
+#define SQ_VTX_WORD1_DATA_FORMAT_SHIFT 22
+#define SQ_VTX_WORD1_NUM_FORMAT_ALL_SHIFT 28
+#define SQ_VTX_WORD1_FORMAT_COMP_ALL_SHIFT 30
+#define SQ_VTX_WORD1_SRF_MODE_ALL_SHIFT 31
+
+#define SQ_VTX_WORD1_DST_SEL_X_MASK    0x00000e00
+#define SQ_VTX_WORD1_DST_SEL_Y_MASK    0x00007000
+#define SQ_VTX_WORD1_DST_SEL_Z_MASK    0x00038000
+#define SQ_VTX_WORD1_DST_SEL_W_MASK    0x001c0000
+#define SQ_VTX_WORD1_USE_CONST_FIELDS_MASK 0x00200000
+#define SQ_VTX_WORD1_DATA_FORMAT_MASK  0x0fc00000
+#define SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK 0x30000000
+#define SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK 0x40000000
+#define SQ_VTX_WORD1_SRF_MODE_ALL_MASK 0x80000000
+
+#define SQ_VTX_WORD1_MASK \
+     (SQ_VTX_WORD1_DST_SEL_X_MASK | \
+      SQ_VTX_WORD1_DST_SEL_Y_MASK | \
+      SQ_VTX_WORD1_DST_SEL_Z_MASK | \
+      SQ_VTX_WORD1_DST_SEL_W_MASK | \
+      SQ_VTX_WORD1_USE_CONST_FIELDS_MASK | \
+      SQ_VTX_WORD1_DATA_FORMAT_MASK | \
+      SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK | \
+      SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK | \
+      SQ_VTX_WORD1_SRF_MODE_ALL_MASK)
+
+#define SQ_VTX_WORD1_DEFAULT           0xcdcdcc00
+
+#define SQ_VTX_WORD1_GET_DST_SEL_X(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_X_MASK) >> SQ_VTX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_VTX_WORD1_GET_DST_SEL_Y(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_Y_MASK) >> SQ_VTX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_VTX_WORD1_GET_DST_SEL_Z(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_Z_MASK) >> SQ_VTX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_VTX_WORD1_GET_DST_SEL_W(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DST_SEL_W_MASK) >> SQ_VTX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_VTX_WORD1_GET_USE_CONST_FIELDS(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_USE_CONST_FIELDS_MASK) >> SQ_VTX_WORD1_USE_CONST_FIELDS_SHIFT)
+#define SQ_VTX_WORD1_GET_DATA_FORMAT(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_DATA_FORMAT_MASK) >> SQ_VTX_WORD1_DATA_FORMAT_SHIFT)
+#define SQ_VTX_WORD1_GET_NUM_FORMAT_ALL(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK) >> SQ_VTX_WORD1_NUM_FORMAT_ALL_SHIFT)
+#define SQ_VTX_WORD1_GET_FORMAT_COMP_ALL(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK) >> SQ_VTX_WORD1_FORMAT_COMP_ALL_SHIFT)
+#define SQ_VTX_WORD1_GET_SRF_MODE_ALL(sq_vtx_word1) \
+     ((sq_vtx_word1 & SQ_VTX_WORD1_SRF_MODE_ALL_MASK) >> SQ_VTX_WORD1_SRF_MODE_ALL_SHIFT)
+
+#define SQ_VTX_WORD1_SET_DST_SEL_X(sq_vtx_word1_reg, dst_sel_x) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_X_MASK) | (dst_sel_x << SQ_VTX_WORD1_DST_SEL_X_SHIFT)
+#define SQ_VTX_WORD1_SET_DST_SEL_Y(sq_vtx_word1_reg, dst_sel_y) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_Y_MASK) | (dst_sel_y << SQ_VTX_WORD1_DST_SEL_Y_SHIFT)
+#define SQ_VTX_WORD1_SET_DST_SEL_Z(sq_vtx_word1_reg, dst_sel_z) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_Z_MASK) | (dst_sel_z << SQ_VTX_WORD1_DST_SEL_Z_SHIFT)
+#define SQ_VTX_WORD1_SET_DST_SEL_W(sq_vtx_word1_reg, dst_sel_w) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DST_SEL_W_MASK) | (dst_sel_w << SQ_VTX_WORD1_DST_SEL_W_SHIFT)
+#define SQ_VTX_WORD1_SET_USE_CONST_FIELDS(sq_vtx_word1_reg, use_const_fields) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_USE_CONST_FIELDS_MASK) | (use_const_fields << SQ_VTX_WORD1_USE_CONST_FIELDS_SHIFT)
+#define SQ_VTX_WORD1_SET_DATA_FORMAT(sq_vtx_word1_reg, data_format) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_DATA_FORMAT_MASK) | (data_format << SQ_VTX_WORD1_DATA_FORMAT_SHIFT)
+#define SQ_VTX_WORD1_SET_NUM_FORMAT_ALL(sq_vtx_word1_reg, num_format_all) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_NUM_FORMAT_ALL_MASK) | (num_format_all << SQ_VTX_WORD1_NUM_FORMAT_ALL_SHIFT)
+#define SQ_VTX_WORD1_SET_FORMAT_COMP_ALL(sq_vtx_word1_reg, format_comp_all) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_FORMAT_COMP_ALL_MASK) | (format_comp_all << SQ_VTX_WORD1_FORMAT_COMP_ALL_SHIFT)
+#define SQ_VTX_WORD1_SET_SRF_MODE_ALL(sq_vtx_word1_reg, srf_mode_all) \
+     sq_vtx_word1_reg = (sq_vtx_word1_reg & ~SQ_VTX_WORD1_SRF_MODE_ALL_MASK) | (srf_mode_all << SQ_VTX_WORD1_SRF_MODE_ALL_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_t {
+          unsigned int                                : 9;
+          unsigned int dst_sel_x                      : SQ_VTX_WORD1_DST_SEL_X_SIZE;
+          unsigned int dst_sel_y                      : SQ_VTX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_z                      : SQ_VTX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_w                      : SQ_VTX_WORD1_DST_SEL_W_SIZE;
+          unsigned int use_const_fields               : SQ_VTX_WORD1_USE_CONST_FIELDS_SIZE;
+          unsigned int data_format                    : SQ_VTX_WORD1_DATA_FORMAT_SIZE;
+          unsigned int num_format_all                 : SQ_VTX_WORD1_NUM_FORMAT_ALL_SIZE;
+          unsigned int format_comp_all                : SQ_VTX_WORD1_FORMAT_COMP_ALL_SIZE;
+          unsigned int srf_mode_all                   : SQ_VTX_WORD1_SRF_MODE_ALL_SIZE;
+     } sq_vtx_word1_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_t {
+          unsigned int srf_mode_all                   : SQ_VTX_WORD1_SRF_MODE_ALL_SIZE;
+          unsigned int format_comp_all                : SQ_VTX_WORD1_FORMAT_COMP_ALL_SIZE;
+          unsigned int num_format_all                 : SQ_VTX_WORD1_NUM_FORMAT_ALL_SIZE;
+          unsigned int data_format                    : SQ_VTX_WORD1_DATA_FORMAT_SIZE;
+          unsigned int use_const_fields               : SQ_VTX_WORD1_USE_CONST_FIELDS_SIZE;
+          unsigned int dst_sel_w                      : SQ_VTX_WORD1_DST_SEL_W_SIZE;
+          unsigned int dst_sel_z                      : SQ_VTX_WORD1_DST_SEL_Z_SIZE;
+          unsigned int dst_sel_y                      : SQ_VTX_WORD1_DST_SEL_Y_SIZE;
+          unsigned int dst_sel_x                      : SQ_VTX_WORD1_DST_SEL_X_SIZE;
+          unsigned int                                : 9;
+     } sq_vtx_word1_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word1_t f;
+} sq_vtx_word1_u;
+
+
+/*
+ * SQ_VTX_WORD1_GPR struct
+ */
+
+#define SQ_VTX_WORD1_GPR_DST_GPR_SIZE  7
+#define SQ_VTX_WORD1_GPR_DST_REL_SIZE  1
+
+#define SQ_VTX_WORD1_GPR_DST_GPR_SHIFT 0
+#define SQ_VTX_WORD1_GPR_DST_REL_SHIFT 7
+
+#define SQ_VTX_WORD1_GPR_DST_GPR_MASK  0x0000007f
+#define SQ_VTX_WORD1_GPR_DST_REL_MASK  0x00000080
+
+#define SQ_VTX_WORD1_GPR_MASK \
+     (SQ_VTX_WORD1_GPR_DST_GPR_MASK | \
+      SQ_VTX_WORD1_GPR_DST_REL_MASK)
+
+#define SQ_VTX_WORD1_GPR_DEFAULT       0x000000cd
+
+#define SQ_VTX_WORD1_GPR_GET_DST_GPR(sq_vtx_word1_gpr) \
+     ((sq_vtx_word1_gpr & SQ_VTX_WORD1_GPR_DST_GPR_MASK) >> SQ_VTX_WORD1_GPR_DST_GPR_SHIFT)
+#define SQ_VTX_WORD1_GPR_GET_DST_REL(sq_vtx_word1_gpr) \
+     ((sq_vtx_word1_gpr & SQ_VTX_WORD1_GPR_DST_REL_MASK) >> SQ_VTX_WORD1_GPR_DST_REL_SHIFT)
+
+#define SQ_VTX_WORD1_GPR_SET_DST_GPR(sq_vtx_word1_gpr_reg, dst_gpr) \
+     sq_vtx_word1_gpr_reg = (sq_vtx_word1_gpr_reg & ~SQ_VTX_WORD1_GPR_DST_GPR_MASK) | (dst_gpr << SQ_VTX_WORD1_GPR_DST_GPR_SHIFT)
+#define SQ_VTX_WORD1_GPR_SET_DST_REL(sq_vtx_word1_gpr_reg, dst_rel) \
+     sq_vtx_word1_gpr_reg = (sq_vtx_word1_gpr_reg & ~SQ_VTX_WORD1_GPR_DST_REL_MASK) | (dst_rel << SQ_VTX_WORD1_GPR_DST_REL_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_gpr_t {
+          unsigned int dst_gpr                        : SQ_VTX_WORD1_GPR_DST_GPR_SIZE;
+          unsigned int dst_rel                        : SQ_VTX_WORD1_GPR_DST_REL_SIZE;
+          unsigned int                                : 24;
+     } sq_vtx_word1_gpr_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_gpr_t {
+          unsigned int                                : 24;
+          unsigned int dst_rel                        : SQ_VTX_WORD1_GPR_DST_REL_SIZE;
+          unsigned int dst_gpr                        : SQ_VTX_WORD1_GPR_DST_GPR_SIZE;
+     } sq_vtx_word1_gpr_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word1_gpr_t f;
+} sq_vtx_word1_gpr_u;
+
+
+/*
+ * SQ_VTX_WORD1_SEM struct
+ */
+
+#define SQ_VTX_WORD1_SEM_SEMANTIC_ID_SIZE 8
+
+#define SQ_VTX_WORD1_SEM_SEMANTIC_ID_SHIFT 0
+
+#define SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK 0x000000ff
+
+#define SQ_VTX_WORD1_SEM_MASK \
+     (SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK)
+
+#define SQ_VTX_WORD1_SEM_DEFAULT       0x000000cd
+
+#define SQ_VTX_WORD1_SEM_GET_SEMANTIC_ID(sq_vtx_word1_sem) \
+     ((sq_vtx_word1_sem & SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK) >> SQ_VTX_WORD1_SEM_SEMANTIC_ID_SHIFT)
+
+#define SQ_VTX_WORD1_SEM_SET_SEMANTIC_ID(sq_vtx_word1_sem_reg, semantic_id) \
+     sq_vtx_word1_sem_reg = (sq_vtx_word1_sem_reg & ~SQ_VTX_WORD1_SEM_SEMANTIC_ID_MASK) | (semantic_id << SQ_VTX_WORD1_SEM_SEMANTIC_ID_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_sem_t {
+          unsigned int semantic_id                    : SQ_VTX_WORD1_SEM_SEMANTIC_ID_SIZE;
+          unsigned int                                : 24;
+     } sq_vtx_word1_sem_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word1_sem_t {
+          unsigned int                                : 24;
+          unsigned int semantic_id                    : SQ_VTX_WORD1_SEM_SEMANTIC_ID_SIZE;
+     } sq_vtx_word1_sem_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word1_sem_t f;
+} sq_vtx_word1_sem_u;
+
+
+/*
+ * SQ_VTX_WORD2 struct
+ */
+
+#define SQ_VTX_WORD2_OFFSET_SIZE       16
+#define SQ_VTX_WORD2_ENDIAN_SWAP_SIZE  2
+#define SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SIZE 1
+#define SQ_VTX_WORD2_MEGA_FETCH_SIZE   1
+#define SQ_VTX_WORD2_ALT_CONST_SIZE    1
+
+#define SQ_VTX_WORD2_OFFSET_SHIFT      0
+#define SQ_VTX_WORD2_ENDIAN_SWAP_SHIFT 16
+#define SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SHIFT 18
+#define SQ_VTX_WORD2_MEGA_FETCH_SHIFT  19
+#define SQ_VTX_WORD2_ALT_CONST_SHIFT   20
+
+#define SQ_VTX_WORD2_OFFSET_MASK       0x0000ffff
+#define SQ_VTX_WORD2_ENDIAN_SWAP_MASK  0x00030000
+#define SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK 0x00040000
+#define SQ_VTX_WORD2_MEGA_FETCH_MASK   0x00080000
+#define SQ_VTX_WORD2_ALT_CONST_MASK    0x00100000
+
+#define SQ_VTX_WORD2_MASK \
+     (SQ_VTX_WORD2_OFFSET_MASK | \
+      SQ_VTX_WORD2_ENDIAN_SWAP_MASK | \
+      SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK | \
+      SQ_VTX_WORD2_MEGA_FETCH_MASK | \
+      SQ_VTX_WORD2_ALT_CONST_MASK)
+
+#define SQ_VTX_WORD2_DEFAULT           0x000dcdcd
+
+#define SQ_VTX_WORD2_GET_OFFSET(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_OFFSET_MASK) >> SQ_VTX_WORD2_OFFSET_SHIFT)
+#define SQ_VTX_WORD2_GET_ENDIAN_SWAP(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_ENDIAN_SWAP_MASK) >> SQ_VTX_WORD2_ENDIAN_SWAP_SHIFT)
+#define SQ_VTX_WORD2_GET_CONST_BUF_NO_STRIDE(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK) >> SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SHIFT)
+#define SQ_VTX_WORD2_GET_MEGA_FETCH(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_MEGA_FETCH_MASK) >> SQ_VTX_WORD2_MEGA_FETCH_SHIFT)
+#define SQ_VTX_WORD2_GET_ALT_CONST(sq_vtx_word2) \
+     ((sq_vtx_word2 & SQ_VTX_WORD2_ALT_CONST_MASK) >> SQ_VTX_WORD2_ALT_CONST_SHIFT)
+
+#define SQ_VTX_WORD2_SET_OFFSET(sq_vtx_word2_reg, offset) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_OFFSET_MASK) | (offset << SQ_VTX_WORD2_OFFSET_SHIFT)
+#define SQ_VTX_WORD2_SET_ENDIAN_SWAP(sq_vtx_word2_reg, endian_swap) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_ENDIAN_SWAP_MASK) | (endian_swap << SQ_VTX_WORD2_ENDIAN_SWAP_SHIFT)
+#define SQ_VTX_WORD2_SET_CONST_BUF_NO_STRIDE(sq_vtx_word2_reg, const_buf_no_stride) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_MASK) | (const_buf_no_stride << SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SHIFT)
+#define SQ_VTX_WORD2_SET_MEGA_FETCH(sq_vtx_word2_reg, mega_fetch) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_MEGA_FETCH_MASK) | (mega_fetch << SQ_VTX_WORD2_MEGA_FETCH_SHIFT)
+#define SQ_VTX_WORD2_SET_ALT_CONST(sq_vtx_word2_reg, alt_const) \
+     sq_vtx_word2_reg = (sq_vtx_word2_reg & ~SQ_VTX_WORD2_ALT_CONST_MASK) | (alt_const << SQ_VTX_WORD2_ALT_CONST_SHIFT)
+
+#if		defined(LITTLEENDIAN_CPU)
+
+     typedef struct _sq_vtx_word2_t {
+          unsigned int offset                         : SQ_VTX_WORD2_OFFSET_SIZE;
+          unsigned int endian_swap                    : SQ_VTX_WORD2_ENDIAN_SWAP_SIZE;
+          unsigned int const_buf_no_stride            : SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SIZE;
+          unsigned int mega_fetch                     : SQ_VTX_WORD2_MEGA_FETCH_SIZE;
+          unsigned int alt_const                      : SQ_VTX_WORD2_ALT_CONST_SIZE;
+          unsigned int                                : 11;
+     } sq_vtx_word2_t;
+
+#elif		defined(BIGENDIAN_CPU)
+
+     typedef struct _sq_vtx_word2_t {
+          unsigned int                                : 11;
+          unsigned int alt_const                      : SQ_VTX_WORD2_ALT_CONST_SIZE;
+          unsigned int mega_fetch                     : SQ_VTX_WORD2_MEGA_FETCH_SIZE;
+          unsigned int const_buf_no_stride            : SQ_VTX_WORD2_CONST_BUF_NO_STRIDE_SIZE;
+          unsigned int endian_swap                    : SQ_VTX_WORD2_ENDIAN_SWAP_SIZE;
+          unsigned int offset                         : SQ_VTX_WORD2_OFFSET_SIZE;
+     } sq_vtx_word2_t;
+
+#endif
+
+typedef union {
+     unsigned int val : 32;
+     sq_vtx_word2_t f;
+} sq_vtx_word2_u;
+
+#endif /* _SQ_MICRO_REG_H */
+
+
diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile
index f223b2d922..b59ad68f44 100644
--- a/src/mesa/drivers/dri/radeon/Makefile
+++ b/src/mesa/drivers/dri/radeon/Makefile
@@ -4,32 +4,51 @@
 TOP = ../../../../..
 include $(TOP)/configs/current
 
+CFLAGS += $(RADEON_CFLAGS)
+
 LIBNAME = radeon_dri.so
 
 MINIGLX_SOURCES = server/radeon_dri.c 
 
+ifeq ($(RADEON_LDFLAGS),)
+CS_SOURCES = radeon_cs_space_drm.c
+endif
+
+RADEON_COMMON_SOURCES = \
+	radeon_texture.c \
+	radeon_common_context.c \
+	radeon_common.c \
+	radeon_dma.c \
+	radeon_lock.c \
+	radeon_bo_legacy.c \
+	radeon_cs_legacy.c \
+	radeon_mipmap_tree.c \
+	radeon_span.c \
+	radeon_fbo.c
+
 DRIVER_SOURCES = \
 	radeon_context.c \
 	radeon_ioctl.c \
-	radeon_lock.c \
 	radeon_screen.c \
 	radeon_state.c \
 	radeon_state_init.c \
 	radeon_tex.c \
-	radeon_texmem.c \
 	radeon_texstate.c \
 	radeon_tcl.c \
 	radeon_swtcl.c \
-	radeon_span.c \
 	radeon_maos.c \
-	radeon_sanity.c 
+	radeon_sanity.c \
+	$(RADEON_COMMON_SOURCES)
 
 C_SOURCES = \
 	$(COMMON_SOURCES) \
-	$(DRIVER_SOURCES) 
+	$(DRIVER_SOURCES) \
+	$(CS_SOURCES)
 
 DRIVER_DEFINES = -DRADEON_COMMON=0
 
+DRI_LIB_DEPS += $(RADEON_LDFLAGS)
+
 X86_SOURCES = 
 
 include ../Makefile.template
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
new file mode 100644
index 0000000000..8eeaea1cb2
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h
@@ -0,0 +1,217 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_BO_H
+#define RADEON_BO_H
+
+#include <stdio.h>
+#include <stdint.h>
+//#include "radeon_track.h"
+
+#ifndef RADEON_DEBUG_BO
+#define RADEON_DEBUG_BO 0
+#endif
+
+/* bo object */
+#define RADEON_BO_FLAGS_MACRO_TILE  1
+#define RADEON_BO_FLAGS_MICRO_TILE  2
+
+struct radeon_bo_manager;
+
+struct radeon_bo {
+    uint32_t                    alignment;
+    uint32_t                    handle;
+    uint32_t                    size;
+    uint32_t                    domains;
+    uint32_t                    flags;
+    unsigned                    cref;
+#ifdef RADEON_BO_TRACK
+    struct radeon_track         *track;
+#endif
+    void                        *ptr;
+    struct radeon_bo_manager    *bom;
+    uint32_t                    space_accounted;
+};
+
+/* bo functions */
+struct radeon_bo_funcs {
+#ifdef RADEON_DEBUG_BO
+    struct radeon_bo *(*bo_open)(struct radeon_bo_manager *bom,
+                                 uint32_t handle,
+                                 uint32_t size,
+                                 uint32_t alignment,
+                                 uint32_t domains,
+                                 uint32_t flags,
+                                 char * szBufUsage);
+#else
+    struct radeon_bo *(*bo_open)(struct radeon_bo_manager *bom,
+                                 uint32_t handle,
+                                 uint32_t size,
+                                 uint32_t alignment,
+                                 uint32_t domains,
+                                 uint32_t flags);
+#endif /* RADEON_DEBUG_BO */
+    void (*bo_ref)(struct radeon_bo *bo);
+    struct radeon_bo *(*bo_unref)(struct radeon_bo *bo);
+    int (*bo_map)(struct radeon_bo *bo, int write);
+    int (*bo_unmap)(struct radeon_bo *bo);
+    int (*bo_wait)(struct radeon_bo *bo);
+    int (*bo_is_static)(struct radeon_bo *bo);
+};
+
+struct radeon_bo_manager {
+    struct radeon_bo_funcs  *funcs;
+    int                     fd;
+
+#ifdef RADEON_BO_TRACK
+    struct radeon_tracker   tracker;
+#endif
+};
+    
+static inline void _radeon_bo_debug(struct radeon_bo *bo,
+                                    const char *op,
+                                    const char *file,
+                                    const char *func,
+                                    int line)
+{
+    fprintf(stderr, "%s %p 0x%08X 0x%08X 0x%08X [%s %s %d]\n",
+            op, bo, bo->handle, bo->size, bo->cref, file, func, line);
+}
+
+static inline struct radeon_bo *_radeon_bo_open(struct radeon_bo_manager *bom,
+                                                uint32_t handle,
+                                                uint32_t size,
+                                                uint32_t alignment,
+                                                uint32_t domains,
+                                                uint32_t flags,
+#ifdef RADEON_DEBUG_BO
+                                                char * szBufUsage,
+#endif /* RADEON_DEBUG_BO */
+                                                const char *file,
+                                                const char *func,
+                                                int line)
+{
+    struct radeon_bo *bo;
+
+#ifdef RADEON_DEBUG_BO
+    bo = bom->funcs->bo_open(bom, handle, size, alignment, domains, flags, szBufUsage);
+#else
+    bo = bom->funcs->bo_open(bom, handle, size, alignment, domains, flags);
+#endif /* RADEON_DEBUG_BO */
+
+#ifdef RADEON_BO_TRACK
+    if (bo) {
+        bo->track = radeon_tracker_add_track(&bom->tracker, bo->handle);
+        radeon_track_add_event(bo->track, file, func, "open", line);
+    }
+#endif
+    return bo;
+}
+
+static inline void _radeon_bo_ref(struct radeon_bo *bo,
+                                  const char *file,
+                                  const char *func,
+                                  int line)
+{
+    bo->cref++;
+#ifdef RADEON_BO_TRACK
+    radeon_track_add_event(bo->track, file, func, "ref", line); 
+#endif
+    bo->bom->funcs->bo_ref(bo);
+}
+
+static inline struct radeon_bo *_radeon_bo_unref(struct radeon_bo *bo,
+                                                 const char *file,
+                                                 const char *func,
+                                                 int line)
+{
+    bo->cref--;
+#ifdef RADEON_BO_TRACK
+    radeon_track_add_event(bo->track, file, func, "unref", line);
+    if (bo->cref <= 0) {
+        radeon_tracker_remove_track(&bo->bom->tracker, bo->track);
+        bo->track = NULL;
+    }
+#endif
+    return bo->bom->funcs->bo_unref(bo);
+}
+
+static inline int _radeon_bo_map(struct radeon_bo *bo,
+                                 int write,
+                                 const char *file,
+                                 const char *func,
+                                 int line)
+{
+    return bo->bom->funcs->bo_map(bo, write);
+}
+
+static inline int _radeon_bo_unmap(struct radeon_bo *bo,
+                                   const char *file,
+                                   const char *func,
+                                   int line)
+{
+    return bo->bom->funcs->bo_unmap(bo);
+}
+
+static inline int _radeon_bo_wait(struct radeon_bo *bo,
+                                  const char *file,
+                                  const char *func,
+                                  int line)
+{
+    return bo->bom->funcs->bo_wait(bo);
+}
+
+static inline int radeon_bo_is_static(struct radeon_bo *bo)
+{
+	if (bo->bom->funcs->bo_is_static)
+		return bo->bom->funcs->bo_is_static(bo);
+	return 0;
+}
+
+#ifdef RADEON_DEBUG_BO
+#define radeon_bo_open(bom, h, s, a, d, f, u)\
+    _radeon_bo_open(bom, h, s, a, d, f, u, __FILE__, __FUNCTION__, __LINE__)
+#else
+#define radeon_bo_open(bom, h, s, a, d, f)\
+    _radeon_bo_open(bom, h, s, a, d, f, __FILE__, __FUNCTION__, __LINE__)
+#endif /* RADEON_DEBUG_BO */
+#define radeon_bo_ref(bo)\
+    _radeon_bo_ref(bo, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_unref(bo)\
+    _radeon_bo_unref(bo, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_map(bo, w)\
+    _radeon_bo_map(bo, w, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_unmap(bo)\
+    _radeon_bo_unmap(bo, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_debug(bo, opcode)\
+    _radeon_bo_debug(bo, opcode, __FILE__, __FUNCTION__, __LINE__)
+#define radeon_bo_wait(bo) \
+    _radeon_bo_wait(bo, __FILE__, __func__, __LINE__)
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
new file mode 100644
index 0000000000..992eb4611b
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c
@@ -0,0 +1,906 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Dave Airlie
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Dave Airlie
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <stdio.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <sys/mman.h>
+#include <sys/ioctl.h>
+#include "xf86drm.h"
+#include "texmem.h"
+#include "main/simple_list.h"
+
+#include "drm.h"
+#include "radeon_drm.h"
+#include "radeon_common.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_macros.h"
+
+/* no seriously texmem.c is this screwed up */
+struct bo_legacy_texture_object {
+    driTextureObject    base;
+    struct bo_legacy *parent;
+};
+
+struct bo_legacy {
+    struct radeon_bo    base;
+    int                 map_count;
+    uint32_t            pending;
+    int                 is_pending;
+    int                 static_bo;
+    uint32_t            offset;
+    struct bo_legacy_texture_object *tobj;
+    int                 validated;
+    int                 dirty;
+    void                *ptr;
+    struct bo_legacy    *next, *prev;
+    struct bo_legacy    *pnext, *pprev;
+#ifdef RADEON_DEBUG_BO
+    char                szBufUsage[16];
+#endif /* RADEON_DEBUG_BO */
+};
+
+struct bo_manager_legacy {
+    struct radeon_bo_manager    base;
+    unsigned                    nhandle;
+    unsigned                    nfree_handles;
+    unsigned                    cfree_handles;
+    uint32_t                    current_age;
+    struct bo_legacy            bos;
+    struct bo_legacy            pending_bos;
+    uint32_t                    fb_location;
+    uint32_t                    texture_offset;
+    unsigned                    dma_alloc_size;
+    uint32_t                    dma_buf_count;
+    unsigned                    cpendings;
+    driTextureObject            texture_swapped;
+    driTexHeap                  *texture_heap;
+    struct radeon_screen        *screen;
+    unsigned                    *free_handles;
+};
+
+static void bo_legacy_tobj_destroy(void *data, driTextureObject *t)
+{
+    struct bo_legacy_texture_object *tobj = (struct bo_legacy_texture_object *)t;
+    
+    if (tobj->parent) {
+        tobj->parent->tobj = NULL;
+        tobj->parent->validated = 0;
+    }
+}
+
+static void inline clean_handles(struct bo_manager_legacy *bom)
+{
+  while (bom->cfree_handles > 0 &&
+	 !bom->free_handles[bom->cfree_handles - 1])
+    bom->cfree_handles--;
+
+}
+static int legacy_new_handle(struct bo_manager_legacy *bom, uint32_t *handle)
+{
+    uint32_t tmp;
+
+    *handle = 0;
+    if (bom->nhandle == 0xFFFFFFFF) {
+        return -EINVAL;
+    }
+    if (bom->cfree_handles > 0) {
+        tmp = bom->free_handles[--bom->cfree_handles];
+	clean_handles(bom);
+    } else {
+        bom->cfree_handles = 0;
+        tmp = bom->nhandle++;
+    }
+    assert(tmp);
+    *handle = tmp;
+    return 0;
+}
+
+static int legacy_free_handle(struct bo_manager_legacy *bom, uint32_t handle)
+{
+    uint32_t *handles;
+
+    if (!handle) {
+        return 0;
+    }
+    if (handle == (bom->nhandle - 1)) {
+        int i;
+
+        bom->nhandle--;
+        for (i = bom->cfree_handles - 1; i >= 0; i--) {
+            if (bom->free_handles[i] == (bom->nhandle - 1)) {
+                bom->nhandle--;
+                bom->free_handles[i] = 0;
+            }
+        }
+        clean_handles(bom);
+        return 0;
+    }
+    if (bom->cfree_handles < bom->nfree_handles) {
+        bom->free_handles[bom->cfree_handles++] = handle;
+        return 0;
+    }
+    bom->nfree_handles += 0x100;
+    handles = (uint32_t*)realloc(bom->free_handles, bom->nfree_handles * 4);
+    if (handles == NULL) {
+        bom->nfree_handles -= 0x100;
+        return -ENOMEM;
+    }
+    bom->free_handles = handles;
+    bom->free_handles[bom->cfree_handles++] = handle;
+    return 0;
+}
+
+static void legacy_get_current_age(struct bo_manager_legacy *boml)
+{
+    drm_radeon_getparam_t gp;
+    unsigned char *RADEONMMIO = NULL;
+    int r;
+
+    if (   IS_R300_CLASS(boml->screen) 
+        || IS_R600_CLASS(boml->screen) ) 
+    {
+    	gp.param = RADEON_PARAM_LAST_CLEAR;
+    	gp.value = (int *)&boml->current_age;
+    	r = drmCommandWriteRead(boml->base.fd, DRM_RADEON_GETPARAM,
+       	                     &gp, sizeof(gp));
+    	if (r) {
+       	 fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, r);
+         exit(1);
+       }
+    } 
+    else {
+        RADEONMMIO = boml->screen->mmio.map;
+        boml->current_age = boml->screen->scratch[3];
+        boml->current_age = INREG(RADEON_GUI_SCRATCH_REG3);
+    }
+}
+
+static int legacy_is_pending(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (bo_legacy->is_pending <= 0) {
+        bo_legacy->is_pending = 0;
+        return 0;
+    }
+    if (boml->current_age >= bo_legacy->pending) {
+        if (boml->pending_bos.pprev == bo_legacy) {
+            boml->pending_bos.pprev = bo_legacy->pprev;
+        }
+        bo_legacy->pprev->pnext = bo_legacy->pnext;
+        if (bo_legacy->pnext) {
+            bo_legacy->pnext->pprev = bo_legacy->pprev;
+        }
+	assert(bo_legacy->is_pending <= bo->cref);
+        while (bo_legacy->is_pending--) {
+	    bo = radeon_bo_unref(bo);
+	    if (!bo)
+	      break;
+        }
+	if (bo)
+	  bo_legacy->is_pending = 0;
+        boml->cpendings--;
+        return 0;
+    }
+    return 1;
+}
+
+static int legacy_wait_pending(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (!bo_legacy->is_pending) {
+        return 0;
+    }
+    /* FIXME: lockup and userspace busy looping that's all the folks */
+    legacy_get_current_age(boml);
+    while (legacy_is_pending(bo)) {
+        usleep(10);
+        legacy_get_current_age(boml);
+    }
+    return 0;
+}
+
+static void legacy_track_pending(struct bo_manager_legacy *boml, int debug)
+{
+    struct bo_legacy *bo_legacy;
+    struct bo_legacy *next;
+
+    legacy_get_current_age(boml);
+    bo_legacy = boml->pending_bos.pnext;
+    while (bo_legacy) {
+        if (debug)
+	  fprintf(stderr,"pending %p %d %d %d\n", bo_legacy, bo_legacy->base.size,
+		  boml->current_age, bo_legacy->pending);
+        next = bo_legacy->pnext;
+        if (legacy_is_pending(&(bo_legacy->base))) {
+        }
+        bo_legacy = next;
+    } 
+}
+
+static int legacy_wait_any_pending(struct bo_manager_legacy *boml)
+{
+    struct bo_legacy *bo_legacy;
+
+    legacy_get_current_age(boml);
+    bo_legacy = boml->pending_bos.pnext;
+    if (!bo_legacy)
+      return -1;
+    legacy_wait_pending(&bo_legacy->base);
+    return 0;
+}
+
+static void legacy_kick_all_buffers(struct bo_manager_legacy *boml)
+{
+    struct bo_legacy *legacy;
+
+    legacy = boml->bos.next;
+    while (legacy != &boml->bos) {
+	if (legacy->tobj) {
+	    if (legacy->validated) {
+		driDestroyTextureObject(&legacy->tobj->base);
+		legacy->tobj = 0;
+		legacy->validated = 0;
+	    }
+	}
+	legacy = legacy->next;
+    }
+}
+
+static struct bo_legacy *bo_allocate(struct bo_manager_legacy *boml,
+                                     uint32_t size,
+                                     uint32_t alignment,
+                                     uint32_t domains,
+#ifdef RADEON_DEBUG_BO
+                                     uint32_t flags,
+                                     char * szBufUsage)
+#else
+                                     uint32_t flags)
+#endif /* RADEON_DEBUG_BO */
+{
+    struct bo_legacy *bo_legacy;
+    static int pgsize;
+
+    if (pgsize == 0)
+        pgsize = getpagesize() - 1;
+
+    size = (size + pgsize) & ~pgsize;
+
+    bo_legacy = (struct bo_legacy*)calloc(1, sizeof(struct bo_legacy));
+    if (bo_legacy == NULL) {
+        return NULL;
+    }
+    bo_legacy->base.bom = (struct radeon_bo_manager*)boml;
+    bo_legacy->base.handle = 0;
+    bo_legacy->base.size = size;
+    bo_legacy->base.alignment = alignment;
+    bo_legacy->base.domains = domains;
+    bo_legacy->base.flags = flags;
+    bo_legacy->base.ptr = NULL;
+    bo_legacy->map_count = 0;
+    bo_legacy->next = NULL;
+    bo_legacy->prev = NULL;
+    bo_legacy->pnext = NULL;
+    bo_legacy->pprev = NULL;
+    bo_legacy->next = boml->bos.next;
+    bo_legacy->prev = &boml->bos;
+    boml->bos.next = bo_legacy;
+    if (bo_legacy->next) {
+        bo_legacy->next->prev = bo_legacy;
+    }
+
+#ifdef RADEON_DEBUG_BO
+    sprintf(bo_legacy->szBufUsage, "%s", szBufUsage); 
+#endif /* RADEON_DEBUG_BO */
+
+    return bo_legacy;
+}
+
+static int bo_dma_alloc(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    drm_radeon_mem_alloc_t alloc;
+    unsigned size;
+    int base_offset;
+    int r;
+
+    /* align size on 4Kb */
+    size = (((4 * 1024) - 1) + bo->size) & ~((4 * 1024) - 1);
+    alloc.region = RADEON_MEM_REGION_GART;
+    alloc.alignment = bo_legacy->base.alignment;
+    alloc.size = size;
+    alloc.region_offset = &base_offset;
+    r = drmCommandWriteRead(bo->bom->fd,
+                            DRM_RADEON_ALLOC,
+                            &alloc,
+                            sizeof(alloc));
+    if (r) {
+        /* ptr is set to NULL if dma allocation failed */
+        bo_legacy->ptr = NULL;
+        return r;
+    }
+    bo_legacy->ptr = boml->screen->gartTextures.map + base_offset;
+    bo_legacy->offset = boml->screen->gart_texture_offset + base_offset;
+    bo->size = size;
+    boml->dma_alloc_size += size;
+    boml->dma_buf_count++;
+    return 0;
+}
+
+static int bo_dma_free(struct radeon_bo *bo)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    drm_radeon_mem_free_t memfree;
+    int r;
+
+    if (bo_legacy->ptr == NULL) {
+        /* ptr is set to NULL if dma allocation failed */
+        return 0;
+    }
+    legacy_get_current_age(boml);
+    memfree.region = RADEON_MEM_REGION_GART;
+    memfree.region_offset  = bo_legacy->offset;
+    memfree.region_offset -= boml->screen->gart_texture_offset;
+    r = drmCommandWrite(boml->base.fd,
+                        DRM_RADEON_FREE,
+                        &memfree,
+                        sizeof(memfree));
+    if (r) {
+        fprintf(stderr, "Failed to free bo[%p] at %08x\n",
+                &bo_legacy->base, memfree.region_offset);
+        fprintf(stderr, "ret = %s\n", strerror(-r));
+        return r;
+    }
+    boml->dma_alloc_size -= bo_legacy->base.size;
+    boml->dma_buf_count--;
+    return 0;
+}
+
+static void bo_free(struct bo_legacy *bo_legacy)
+{
+    struct bo_manager_legacy *boml;
+
+    if (bo_legacy == NULL) {
+        return;
+    }
+    boml = (struct bo_manager_legacy *)bo_legacy->base.bom;
+    bo_legacy->prev->next = bo_legacy->next;
+    if (bo_legacy->next) {
+        bo_legacy->next->prev = bo_legacy->prev;
+    }
+    if (!bo_legacy->static_bo) {
+        legacy_free_handle(boml, bo_legacy->base.handle);
+        if (bo_legacy->base.domains & RADEON_GEM_DOMAIN_GTT) {
+            /* dma buffers */
+            bo_dma_free(&bo_legacy->base);
+        } else {
+  	    driDestroyTextureObject(&bo_legacy->tobj->base);
+	    bo_legacy->tobj = NULL;
+            /* free backing store */
+            free(bo_legacy->ptr);
+        }
+    }
+    memset(bo_legacy, 0 , sizeof(struct bo_legacy));
+    free(bo_legacy);
+}
+
+static struct radeon_bo *bo_open(struct radeon_bo_manager *bom,
+                                 uint32_t handle,
+                                 uint32_t size,
+                                 uint32_t alignment,
+                                 uint32_t domains,
+#ifdef RADEON_DEBUG_BO
+                                 uint32_t flags,
+                                 char *   szBufUsage)
+#else
+                                 uint32_t flags)
+#endif /* RADEON_DEBUG_BO */
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    struct bo_legacy *bo_legacy;
+    int r;
+
+    if (handle) {
+        bo_legacy = boml->bos.next;
+        while (bo_legacy) {
+            if (bo_legacy->base.handle == handle) {
+                radeon_bo_ref(&(bo_legacy->base));
+                return (struct radeon_bo*)bo_legacy;
+            }
+            bo_legacy = bo_legacy->next;
+        }
+        return NULL;
+    }
+#ifdef RADEON_DEBUG_BO
+    bo_legacy = bo_allocate(boml, size, alignment, domains, flags, szBufUsage);
+#else
+    bo_legacy = bo_allocate(boml, size, alignment, domains, flags);
+#endif /* RADEON_DEBUG_BO */
+    bo_legacy->static_bo = 0;
+    r = legacy_new_handle(boml, &bo_legacy->base.handle);
+    if (r) {
+        bo_free(bo_legacy);
+        return NULL;
+    }
+    if (bo_legacy->base.domains & RADEON_GEM_DOMAIN_GTT) 
+    {
+retry:
+        legacy_track_pending(boml, 0);
+        /* dma buffers */
+
+        r = bo_dma_alloc(&(bo_legacy->base));
+        if (r) 
+        {
+	         if (legacy_wait_any_pending(boml) == -1) 
+             {
+                  bo_free(bo_legacy);
+	              return NULL;
+             }
+	         goto retry;
+	         return NULL;
+        }
+    } 
+    else 
+    {
+        bo_legacy->ptr = malloc(bo_legacy->base.size);
+        if (bo_legacy->ptr == NULL) {
+            bo_free(bo_legacy);
+            return NULL;
+        }
+    }
+    radeon_bo_ref(&(bo_legacy->base));
+
+    return (struct radeon_bo*)bo_legacy;
+}
+
+static void bo_ref(struct radeon_bo *bo)
+{
+}
+
+static struct radeon_bo *bo_unref(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (bo->cref <= 0) {
+        bo_legacy->prev->next = bo_legacy->next;
+        if (bo_legacy->next) {
+            bo_legacy->next->prev = bo_legacy->prev;
+        }
+        if (!bo_legacy->is_pending) {
+            bo_free(bo_legacy);
+        }
+        return NULL;
+    }
+    return bo;
+}
+
+static int bo_map(struct radeon_bo *bo, int write)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    legacy_wait_pending(bo);
+    bo_legacy->validated = 0;
+    bo_legacy->dirty = 1;
+    bo_legacy->map_count++;
+    bo->ptr = bo_legacy->ptr;
+    /* Read the first pixel in the frame buffer.  This should
+     * be a noop, right?  In fact without this conform fails as reading
+     * from the framebuffer sometimes produces old results -- the
+     * on-card read cache gets mixed up and doesn't notice that the
+     * framebuffer has been updated.
+     *
+     * Note that we should probably be reading some otherwise unused
+     * region of VRAM, otherwise we might get incorrect results when
+     * reading pixels from the top left of the screen.
+     *
+     * I found this problem on an R420 with glean's texCube test.
+     * Note that the R200 span code also *writes* the first pixel in the
+     * framebuffer, but I've found this to be unnecessary.
+     *  -- Nicolai Hähnle, June 2008
+     */
+    if (!(bo->domains & RADEON_GEM_DOMAIN_GTT)) {
+        int p;
+        volatile int *buf = (int*)boml->screen->driScreen->pFB;
+        p = *buf;
+    }
+
+    return 0;
+}
+
+static int bo_unmap(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (--bo_legacy->map_count > 0) 
+    {
+        return 0;
+    }
+    
+    bo->ptr = NULL;
+
+    return 0;
+}
+
+
+static int bo_is_static(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    return bo_legacy->static_bo;
+}
+
+static struct radeon_bo_funcs bo_legacy_funcs = {
+    bo_open,
+    bo_ref,
+    bo_unref,
+    bo_map,
+    bo_unmap,
+    NULL,
+    bo_is_static,
+};
+
+static int bo_vram_validate(struct radeon_bo *bo,
+                            uint32_t *soffset,
+                            uint32_t *eoffset)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    int r;
+    int retry_count = 0, pending_retry = 0;
+    
+    if (!bo_legacy->tobj) {
+	bo_legacy->tobj = CALLOC(sizeof(struct bo_legacy_texture_object));
+	bo_legacy->tobj->parent = bo_legacy;
+	make_empty_list(&bo_legacy->tobj->base);
+	bo_legacy->tobj->base.totalSize = bo->size;
+    retry:
+        r = driAllocateTexture(&boml->texture_heap, 1,
+                               &bo_legacy->tobj->base);
+        if (r) {
+		pending_retry = 0;
+		while(boml->cpendings && pending_retry++ < 10000) {
+			legacy_track_pending(boml, 0);
+			retry_count++;
+			if (retry_count > 2) {
+				free(bo_legacy->tobj);
+				bo_legacy->tobj = NULL;
+				fprintf(stderr, "Ouch! vram_validate failed %d\n", r);
+				return -1;
+			}
+			goto retry;
+		}
+	}
+        bo_legacy->offset = boml->texture_offset +
+                            bo_legacy->tobj->base.memBlock->ofs;
+        bo_legacy->dirty = 1;
+    }
+
+    assert(bo_legacy->tobj->base.memBlock);
+
+    if (bo_legacy->tobj)
+	driUpdateTextureLRU(&bo_legacy->tobj->base);
+
+    if (bo_legacy->dirty || bo_legacy->tobj->base.dirty_images[0]) {
+	    if (IS_R600_CLASS(boml->screen)) {
+		    char *src = bo_legacy->ptr;
+		    char *dst = (char *) boml->screen->driScreen->pFB +
+			    (bo_legacy->offset - boml->fb_location);
+
+		    /* FIXME: alignment, pitch, etc. */
+		    memcpy(dst, src, bo->size);
+	    } else {
+		    /* Copy to VRAM using a blit.
+		     * All memory is 4K aligned. We're using 1024 pixels wide blits.
+		     */
+		    drm_radeon_texture_t tex;
+		    drm_radeon_tex_image_t tmp;
+		    int ret;
+
+		    tex.offset = bo_legacy->offset;
+		    tex.image = &tmp;
+		    assert(!(tex.offset & 1023));
+
+		    tmp.x = 0;
+		    tmp.y = 0;
+		    if (bo->size < 4096) {
+			    tmp.width = (bo->size + 3) / 4;
+			    tmp.height = 1;
+		    } else {
+			    tmp.width = 1024;
+			    tmp.height = (bo->size + 4095) / 4096;
+		    }
+		    tmp.data = bo_legacy->ptr;
+		    tex.format = RADEON_TXFORMAT_ARGB8888;
+		    tex.width = tmp.width;
+		    tex.height = tmp.height;
+		    tex.pitch = MAX2(tmp.width / 16, 1);
+		    do {
+			    ret = drmCommandWriteRead(bo->bom->fd,
+						      DRM_RADEON_TEXTURE,
+						      &tex,
+						      sizeof(drm_radeon_texture_t));
+			    if (ret) {
+				    if (RADEON_DEBUG & DEBUG_IOCTL)
+					    fprintf(stderr, "DRM_RADEON_TEXTURE:  again!\n");
+				    usleep(1);
+			    }
+		    } while (ret == -EAGAIN);
+	    }
+	    bo_legacy->dirty = 0;
+	    bo_legacy->tobj->base.dirty_images[0] = 0;
+    }
+    return 0;
+}
+
+/* 
+ *  radeon_bo_legacy_validate -
+ *  returns:
+ *  0 - all good
+ *  -EINVAL - mapped buffer can't be validated
+ *  -EAGAIN - restart validation we've kicked all the buffers out
+ */
+int radeon_bo_legacy_validate(struct radeon_bo *bo,
+                              uint32_t *soffset,
+                              uint32_t *eoffset)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+    int r;
+    int retries = 0;
+
+    if (bo_legacy->map_count) {
+#ifdef RADEON_DEBUG_BO
+        fprintf(stderr, "bo(%p, %d, %s) is mapped (%d) can't valide it.\n",
+                bo, bo->size, bo_legacy->szBufUsage, bo_legacy->map_count);
+#else
+        fprintf(stderr, "bo(%p, %d) is mapped (%d) can't valide it.\n",
+                bo, bo->size, bo_legacy->map_count);
+#endif /* RADEON_DEBUG_BO */
+
+        return -EINVAL;
+    }
+    if (bo_legacy->static_bo || bo_legacy->validated) {
+        *soffset = bo_legacy->offset;
+        *eoffset = bo_legacy->offset + bo->size;
+
+        return 0;
+    }
+    if (!(bo->domains & RADEON_GEM_DOMAIN_GTT)) {
+
+        r = bo_vram_validate(bo, soffset, eoffset);
+        if (r) {
+	    legacy_track_pending(boml, 0);
+	    legacy_kick_all_buffers(boml);
+	    retries++;
+	    if (retries == 2) {
+		fprintf(stderr,"legacy bo: failed to get relocations into aperture\n");
+		assert(0);
+		exit(-1);
+	    }
+	    return -EAGAIN;
+        }
+    }
+    *soffset = bo_legacy->offset;
+    *eoffset = bo_legacy->offset + bo->size;
+    bo_legacy->validated = 1;
+
+    return 0;
+}
+
+void radeon_bo_legacy_pending(struct radeon_bo *bo, uint32_t pending)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bo->bom;
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    bo_legacy->pending = pending;
+    bo_legacy->is_pending++;
+    /* add to pending list */
+    radeon_bo_ref(bo);
+    if (bo_legacy->is_pending > 1) {
+        return;    
+    }
+    bo_legacy->pprev = boml->pending_bos.pprev;
+    bo_legacy->pnext = NULL;
+    bo_legacy->pprev->pnext = bo_legacy;
+    boml->pending_bos.pprev = bo_legacy;
+    boml->cpendings++;
+}
+
+void radeon_bo_manager_legacy_dtor(struct radeon_bo_manager *bom)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    struct bo_legacy *bo_legacy;
+
+    if (bom == NULL) {
+        return;
+    }
+    bo_legacy = boml->bos.next;
+    while (bo_legacy) {
+        struct bo_legacy *next;
+
+        next = bo_legacy->next;
+        bo_free(bo_legacy);
+        bo_legacy = next;
+    }
+    driDestroyTextureHeap(boml->texture_heap);
+    free(boml->free_handles);
+    free(boml);
+}
+
+static struct bo_legacy *radeon_legacy_bo_alloc_static(struct bo_manager_legacy *bom,
+						       int size, 
+#ifdef RADEON_DEBUG_BO
+                               uint32_t offset,
+                               char * szBufUsage)
+#else
+                               uint32_t offset)
+#endif /* RADEON_DEBUG_BO */
+{
+    struct bo_legacy *bo;
+
+#ifdef RADEON_DEBUG_BO
+    bo = bo_allocate(bom, size, 0, RADEON_GEM_DOMAIN_VRAM, 0, szBufUsage);
+#else
+    bo = bo_allocate(bom, size, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+#endif /* RADEON_DEBUG_BO */
+    if (bo == NULL)
+	return NULL;
+    bo->static_bo = 1;
+    bo->offset = offset + bom->fb_location;
+    bo->base.handle = bo->offset;
+    bo->ptr = bom->screen->driScreen->pFB + offset;
+    if (bo->base.handle > bom->nhandle) {
+        bom->nhandle = bo->base.handle + 1;
+    }
+    radeon_bo_ref(&(bo->base));
+    return bo;
+}
+
+struct radeon_bo_manager *radeon_bo_manager_legacy_ctor(struct radeon_screen *scrn)
+{
+    struct bo_manager_legacy *bom;
+    struct bo_legacy *bo;
+    unsigned size;
+
+    bom = (struct bo_manager_legacy*)
+          calloc(1, sizeof(struct bo_manager_legacy));
+    if (bom == NULL) {
+        return NULL;
+    }
+
+    make_empty_list(&bom->texture_swapped);
+
+    bom->texture_heap = driCreateTextureHeap(0,
+                                             bom,
+                                             scrn->texSize[0],
+                                             12,
+                                             RADEON_NR_TEX_REGIONS,
+                                             (drmTextureRegionPtr)scrn->sarea->tex_list[0],
+                                             &scrn->sarea->tex_age[0],
+                                             &bom->texture_swapped,
+                                             sizeof(struct bo_legacy_texture_object),
+                                             &bo_legacy_tobj_destroy);
+    bom->texture_offset = scrn->texOffset[0];
+
+    bom->base.funcs = &bo_legacy_funcs;
+    bom->base.fd = scrn->driScreen->fd;
+    bom->bos.next = NULL;
+    bom->bos.prev = NULL;
+    bom->pending_bos.pprev = &bom->pending_bos;
+    bom->pending_bos.pnext = NULL;
+    bom->screen = scrn;
+    bom->fb_location = scrn->fbLocation;
+    bom->nhandle = 1;
+    bom->cfree_handles = 0;
+    bom->nfree_handles = 0x400;
+    bom->free_handles = (uint32_t*)malloc(bom->nfree_handles * 4);
+    if (bom->free_handles == NULL) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+
+    /* biggest framebuffer size */
+    size = 4096*4096*4; 
+
+    /* allocate front */
+#ifdef RADEON_DEBUG_BO
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->frontOffset, "FRONT BUF");
+#else
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->frontOffset);
+#endif /* RADEON_DEBUG_BO */
+    if (!bo) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+    if (scrn->sarea->tiling_enabled) {
+        bo->base.flags = RADEON_BO_FLAGS_MACRO_TILE;
+    }
+
+    /* allocate back */
+#ifdef RADEON_DEBUG_BO
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->backOffset, "BACK BUF");
+#else
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->backOffset);
+#endif /* RADEON_DEBUG_BO */
+    if (!bo) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+    if (scrn->sarea->tiling_enabled) {
+        bo->base.flags = RADEON_BO_FLAGS_MACRO_TILE;
+    }
+
+    /* allocate depth */
+#ifdef RADEON_DEBUG_BO
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->depthOffset, "Z BUF");
+#else
+    bo = radeon_legacy_bo_alloc_static(bom, size, bom->screen->depthOffset);
+#endif /* RADEON_DEBUG_BO */
+    if (!bo) {
+        radeon_bo_manager_legacy_dtor((struct radeon_bo_manager*)bom);
+        return NULL;
+    }
+    bo->base.flags = 0;
+    if (scrn->sarea->tiling_enabled) {
+        bo->base.flags |= RADEON_BO_FLAGS_MACRO_TILE;
+        bo->base.flags |= RADEON_BO_FLAGS_MICRO_TILE;
+    }
+    return (struct radeon_bo_manager*)bom;
+}
+
+void radeon_bo_legacy_texture_age(struct radeon_bo_manager *bom)
+{
+    struct bo_manager_legacy *boml = (struct bo_manager_legacy *)bom;
+    DRI_AGE_TEXTURES(boml->texture_heap);
+}
+
+unsigned radeon_bo_legacy_relocs_size(struct radeon_bo *bo)
+{
+    struct bo_legacy *bo_legacy = (struct bo_legacy*)bo;
+
+    if (bo_legacy->static_bo || (bo->domains & RADEON_GEM_DOMAIN_GTT)) {
+        return 0;
+    }
+    return bo->size;
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.h b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.h
new file mode 100644
index 0000000000..0db817cab0
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.h
@@ -0,0 +1,46 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_BO_LEGACY_H
+#define RADEON_BO_LEGACY_H
+
+#include "radeon_screen.h"
+
+void radeon_bo_legacy_pending(struct radeon_bo *bo, uint32_t pending);
+int radeon_bo_legacy_validate(struct radeon_bo *bo,
+                              uint32_t *soffset,
+                              uint32_t *eoffset);
+struct radeon_bo_manager *radeon_bo_manager_legacy_ctor(struct radeon_screen *scrn);
+void radeon_bo_manager_legacy_dtor(struct radeon_bo_manager *bom);
+void radeon_bo_legacy_texture_age(struct radeon_bo_manager *bom);
+unsigned radeon_bo_legacy_relocs_size(struct radeon_bo *bo);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
new file mode 100644
index 0000000000..e0c70dd9a1
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
@@ -0,0 +1,85 @@
+#ifndef RADEON_CS_WRAPPER_H
+#define RADEON_CS_WRAPPER_H
+
+#ifdef HAVE_LIBDRM_RADEON
+
+#include "radeon_bo.h"
+#include "radeon_bo_gem.h"
+#include "radeon_cs.h"
+#include "radeon_cs_gem.h"
+
+#else
+#include <stdint.h>
+
+#define RADEON_GEM_DOMAIN_CPU 0x1   // Cached CPU domain
+#define RADEON_GEM_DOMAIN_GTT 0x2   // GTT or cache flushed
+#define RADEON_GEM_DOMAIN_VRAM 0x4  // VRAM domain
+
+/* to be used to build locally in mesa with no libdrm bits */
+#include "../radeon/radeon_bo_drm.h"
+#include "../radeon/radeon_cs_drm.h"
+
+#ifndef DRM_RADEON_GEM_INFO
+#define DRM_RADEON_GEM_INFO 0x1c
+
+struct drm_radeon_gem_info {
+        uint64_t gart_size;
+        uint64_t vram_size;
+        uint64_t vram_visible;
+};
+
+struct drm_radeon_info {
+	uint32_t request;
+	uint32_t pad;
+	uint32_t value;
+};
+#endif
+
+#ifndef RADEON_PARAM_DEVICE_ID
+#define RADEON_PARAM_DEVICE_ID 16
+#endif
+
+#ifndef RADEON_INFO_DEVICE_ID
+#define RADEON_INFO_DEVICE_ID 0
+#endif
+#ifndef RADEON_INFO_NUM_GB_PIPES
+#define RADEON_INFO_NUM_GB_PIPES 0
+#endif
+
+#ifndef DRM_RADEON_INFO
+#define DRM_RADEON_INFO 0x1
+#endif
+
+
+static inline uint32_t radeon_gem_name_bo(struct radeon_bo *dummy)
+{
+  return 0;
+}
+
+static inline void *radeon_bo_manager_gem_ctor(int fd)
+{
+  return NULL;
+}
+
+static inline void radeon_bo_manager_gem_dtor(void *dummy)
+{
+}
+
+static inline void *radeon_cs_manager_gem_ctor(int fd)
+{
+  return NULL;
+}
+
+static inline void radeon_cs_manager_gem_dtor(void *dummy)
+{
+}
+
+static inline void radeon_tracker_print(void *ptr, int io)
+{
+}
+#endif
+
+#include "radeon_bo_legacy.h"
+#include "radeon_cs_legacy.h"
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_chipset.h b/src/mesa/drivers/dri/radeon/radeon_chipset.h
index f6bd1eb83f..0a6a2df35b 100644
--- a/src/mesa/drivers/dri/radeon/radeon_chipset.h
+++ b/src/mesa/drivers/dri/radeon/radeon_chipset.h
@@ -255,6 +255,131 @@
 #define PCI_CHIP_RS740_796E             0x796E
 #define PCI_CHIP_RS740_796F             0x796F
 
+#define PCI_CHIP_R600_9400              0x9400
+#define PCI_CHIP_R600_9401              0x9401
+#define PCI_CHIP_R600_9402              0x9402
+#define PCI_CHIP_R600_9403              0x9403
+#define PCI_CHIP_R600_9405              0x9405
+#define PCI_CHIP_R600_940A              0x940A
+#define PCI_CHIP_R600_940B              0x940B
+#define PCI_CHIP_R600_940F              0x940F
+
+#define PCI_CHIP_RV610_94C0             0x94C0
+#define PCI_CHIP_RV610_94C1             0x94C1
+#define PCI_CHIP_RV610_94C3             0x94C3
+#define PCI_CHIP_RV610_94C4             0x94C4
+#define PCI_CHIP_RV610_94C5             0x94C5
+#define PCI_CHIP_RV610_94C6             0x94C6
+#define PCI_CHIP_RV610_94C7             0x94C7
+#define PCI_CHIP_RV610_94C8             0x94C8
+#define PCI_CHIP_RV610_94C9             0x94C9
+#define PCI_CHIP_RV610_94CB             0x94CB
+#define PCI_CHIP_RV610_94CC             0x94CC
+#define PCI_CHIP_RV610_94CD             0x94CD
+
+#define PCI_CHIP_RV630_9580             0x9580
+#define PCI_CHIP_RV630_9581             0x9581
+#define PCI_CHIP_RV630_9583             0x9583
+#define PCI_CHIP_RV630_9586             0x9586
+#define PCI_CHIP_RV630_9587             0x9587
+#define PCI_CHIP_RV630_9588             0x9588
+#define PCI_CHIP_RV630_9589             0x9589
+#define PCI_CHIP_RV630_958A             0x958A
+#define PCI_CHIP_RV630_958B             0x958B
+#define PCI_CHIP_RV630_958C             0x958C
+#define PCI_CHIP_RV630_958D             0x958D
+#define PCI_CHIP_RV630_958E             0x958E
+#define PCI_CHIP_RV630_958F             0x958F
+
+#define PCI_CHIP_RV670_9500             0x9500
+#define PCI_CHIP_RV670_9501             0x9501
+#define PCI_CHIP_RV670_9504             0x9504
+#define PCI_CHIP_RV670_9505             0x9505
+#define PCI_CHIP_RV670_9506             0x9506
+#define PCI_CHIP_RV670_9507             0x9507
+#define PCI_CHIP_RV670_9508             0x9508
+#define PCI_CHIP_RV670_9509             0x9509
+#define PCI_CHIP_RV670_950F             0x950F
+#define PCI_CHIP_RV670_9511             0x9511
+#define PCI_CHIP_RV670_9515             0x9515
+#define PCI_CHIP_RV670_9517             0x9517
+#define PCI_CHIP_RV670_9519             0x9519
+
+#define PCI_CHIP_RV620_95C0             0x95C0
+#define PCI_CHIP_RV620_95C2             0x95C2
+#define PCI_CHIP_RV620_95C4             0x95C4
+#define PCI_CHIP_RV620_95C5             0x95C5
+#define PCI_CHIP_RV620_95C6             0x95C6
+#define PCI_CHIP_RV620_95C7             0x95C7
+#define PCI_CHIP_RV620_95C9             0x95C9
+#define PCI_CHIP_RV620_95CC             0x95CC
+#define PCI_CHIP_RV620_95CD             0x95CD
+#define PCI_CHIP_RV620_95CE             0x95CE
+#define PCI_CHIP_RV620_95CF             0x95CF
+
+#define PCI_CHIP_RV635_9590             0x9590
+#define PCI_CHIP_RV635_9591             0x9591
+#define PCI_CHIP_RV635_9593             0x9593
+#define PCI_CHIP_RV635_9595             0x9595
+#define PCI_CHIP_RV635_9596             0x9596
+#define PCI_CHIP_RV635_9597             0x9597
+#define PCI_CHIP_RV635_9598             0x9598
+#define PCI_CHIP_RV635_9599             0x9599
+#define PCI_CHIP_RV635_959B             0x959B
+
+#define PCI_CHIP_RS780_9610             0x9610
+#define PCI_CHIP_RS780_9611             0x9611
+#define PCI_CHIP_RS780_9612             0x9612
+#define PCI_CHIP_RS780_9613             0x9613
+#define PCI_CHIP_RS780_9614             0x9614
+#define PCI_CHIP_RS780_9615             0x9615
+#define PCI_CHIP_RS780_9616             0x9616
+
+#define PCI_CHIP_RV770_9440             0x9440
+#define PCI_CHIP_RV770_9441             0x9441
+#define PCI_CHIP_RV770_9442             0x9442
+#define PCI_CHIP_RV770_9444             0x9444
+#define PCI_CHIP_RV770_9446             0x9446
+#define PCI_CHIP_RV770_944A             0x944A
+#define PCI_CHIP_RV770_944B             0x944B
+#define PCI_CHIP_RV770_944C             0x944C
+#define PCI_CHIP_RV770_944E             0x944E
+#define PCI_CHIP_RV770_9450             0x9450
+#define PCI_CHIP_RV770_9452             0x9452
+#define PCI_CHIP_RV770_9456             0x9456
+#define PCI_CHIP_RV770_945A             0x945A
+#define PCI_CHIP_RV770_945B             0x945B
+#define PCI_CHIP_RV790_9460             0x9460
+#define PCI_CHIP_RV790_9462             0x9462
+#define PCI_CHIP_RV770_946A             0x946A
+#define PCI_CHIP_RV770_946B             0x946B
+#define PCI_CHIP_RV770_947A             0x947A
+#define PCI_CHIP_RV770_947B             0x947B
+
+#define PCI_CHIP_RV730_9487             0x9487
+#define PCI_CHIP_RV730_9489             0x9489
+#define PCI_CHIP_RV730_948F             0x948F
+#define PCI_CHIP_RV730_9490             0x9490
+#define PCI_CHIP_RV730_9491             0x9491
+#define PCI_CHIP_RV730_9498             0x9498
+#define PCI_CHIP_RV730_949C             0x949C
+#define PCI_CHIP_RV730_949E             0x949E
+#define PCI_CHIP_RV730_949F             0x949F
+
+#define PCI_CHIP_RV710_9540             0x9540
+#define PCI_CHIP_RV710_9541             0x9541
+#define PCI_CHIP_RV710_9542             0x9542
+#define PCI_CHIP_RV710_954E             0x954E
+#define PCI_CHIP_RV710_954F             0x954F
+#define PCI_CHIP_RV710_9552             0x9552
+#define PCI_CHIP_RV710_9553             0x9553
+#define PCI_CHIP_RV710_9555             0x9555
+
+#define PCI_CHIP_RV740_94A0             0x94A0
+#define PCI_CHIP_RV740_94A1             0x94A1
+#define PCI_CHIP_RV740_94B1             0x94B1
+#define PCI_CHIP_RV740_94B3             0x94B3
+#define PCI_CHIP_RV740_94B5             0x94B5
 
 enum {
    CHIP_FAMILY_R100,
@@ -282,6 +407,17 @@ enum {
    CHIP_FAMILY_R580,
    CHIP_FAMILY_RV560,
    CHIP_FAMILY_RV570,
+   CHIP_FAMILY_R600,
+   CHIP_FAMILY_RV610,
+   CHIP_FAMILY_RV630,
+   CHIP_FAMILY_RV670,
+   CHIP_FAMILY_RV620,
+   CHIP_FAMILY_RV635,
+   CHIP_FAMILY_RS780,
+   CHIP_FAMILY_RV770,
+   CHIP_FAMILY_RV730,
+   CHIP_FAMILY_RV710,
+   CHIP_FAMILY_RV740,
    CHIP_FAMILY_LAST
 };
 
@@ -289,6 +425,7 @@ enum {
 #define RADEON_CLASS_R100		(0 << 0)
 #define RADEON_CLASS_R200		(1 << 0)
 #define RADEON_CLASS_R300		(2 << 0)
+#define RADEON_CLASS_R600		(3 << 0)
 #define RADEON_CLASS_MASK		(3 << 0)
 
 #define RADEON_CHIPSET_TCL		(1 << 2)	/* tcl support - any radeon */
diff --git a/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h
new file mode 100644
index 0000000000..abb023c7de
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h
@@ -0,0 +1,123 @@
+#ifndef COMMON_CMDBUF_H
+#define COMMON_CMDBUF_H
+
+#include "radeon_bocs_wrapper.h"
+
+void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller);
+int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller);
+int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller);
+void rcommonInitCmdBuf(radeonContextPtr rmesa);
+void rcommonDestroyCmdBuf(radeonContextPtr rmesa);
+
+void rcommonBeginBatch(radeonContextPtr rmesa,
+		       int n,
+		       int dostate,
+		       const char *file,
+		       const char *function,
+		       int line);
+
+/* +r6/r7 : code here moved */
+
+#define CP_PACKET2  (2 << 30)
+#define CP_PACKET0(reg, n)	(RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2))
+#define CP_PACKET0_ONE(reg, n)	(RADEON_CP_PACKET0 | RADEON_CP_PACKET0_ONE_REG_WR | ((n)<<16) | ((reg)>>2))
+#define CP_PACKET3(pkt, n)	(RADEON_CP_PACKET3 | (pkt) | ((n) << 16))
+
+/**
+ * Every function writing to the command buffer needs to declare this
+ * to get the necessary local variables.
+ */
+#define BATCH_LOCALS(rmesa) \
+	const radeonContextPtr b_l_rmesa = rmesa
+
+/**
+ * Prepare writing n dwords to the command buffer,
+ * including producing any necessary state emits on buffer wraparound.
+ */
+#define BEGIN_BATCH(n) rcommonBeginBatch(b_l_rmesa, n, 1, __FILE__, __FUNCTION__, __LINE__)
+
+/**
+ * Same as BEGIN_BATCH, but do not cause automatic state emits.
+ */
+#define BEGIN_BATCH_NO_AUTOSTATE(n) rcommonBeginBatch(b_l_rmesa, n, 0, __FILE__, __FUNCTION__, __LINE__)
+
+/**
+ * Write one dword to the command buffer.
+ */
+#define OUT_BATCH(data) \
+	do { \
+        radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);\
+	} while(0)
+
+/**
+ * Write a relocated dword to the command buffer.
+ */
+#define OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags) 	\
+	do { 							\
+        if (0 && offset) {					\
+            fprintf(stderr, "(%s:%s:%d) offset : %d\n",		\
+            __FILE__, __FUNCTION__, __LINE__, offset);		\
+        }							\
+        radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, offset);	\
+        radeon_cs_write_reloc(b_l_rmesa->cmdbuf.cs, 		\
+                              bo, rd, wd, flags);		\
+	if (!b_l_rmesa->radeonScreen->kernel_mm) 		\
+		b_l_rmesa->cmdbuf.cs->section_cdw += 2;		\
+	} while(0)
+
+
+/**
+ * Write n dwords from ptr to the command buffer.
+ */
+#define OUT_BATCH_TABLE(ptr,n) \
+	do { \
+		int _i; \
+        for (_i=0; _i < n; _i++) {\
+            radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, ptr[_i]);\
+        }\
+	} while(0)
+
+/**
+ * Finish writing dwords to the command buffer.
+ * The number of (direct or indirect) OUT_BATCH calls between the previous
+ * BEGIN_BATCH and END_BATCH must match the number specified at BEGIN_BATCH time.
+ */
+#define END_BATCH() \
+	do { \
+        radeon_cs_end(b_l_rmesa->cmdbuf.cs, __FILE__, __FUNCTION__, __LINE__);\
+	} while(0)
+
+/**
+ * After the last END_BATCH() of rendering, this indicates that flushing
+ * the command buffer now is okay.
+ */
+#define COMMIT_BATCH() \
+	do { \
+	} while(0)
+
+
+/** Single register write to command buffer; requires 2 dwords. */
+#define OUT_BATCH_REGVAL(reg, val) \
+	OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), 1)); \
+	OUT_BATCH((val))
+
+/** Continuous register range write to command buffer; requires 1 dword,
+ * expects count dwords afterwards for register contents. */
+#define OUT_BATCH_REGSEQ(reg, count) \
+	OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), (count)))
+
+/** Write a 32 bit float to the ring; requires 1 dword. */
+#define OUT_BATCH_FLOAT32(f) \
+	OUT_BATCH(radeonPackFloat32((f)))
+
+/* +r6/r7 : code here moved */
+
+/* Fire the buffered vertices no matter what.
+ */
+static INLINE void radeon_firevertices(radeonContextPtr radeon)
+{
+   if (radeon->cmdbuf.cs->cdw || radeon->dma.flush )
+      radeon->glCtx->Driver.Flush(radeon->glCtx); /* +r6/r7 */
+}
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
new file mode 100644
index 0000000000..dde615a4d9
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -0,0 +1,1292 @@
+/**************************************************************************
+
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/*
+   - Scissor implementation
+   - buffer swap/copy ioctls
+   - finish/flush
+   - state emission
+   - cmdbuffer management
+*/
+
+#include <errno.h>
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/arrayobj.h"
+#include "main/api_arrayelt.h"
+#include "main/enums.h"
+#include "main/colormac.h"
+#include "main/light.h"
+#include "main/framebuffer.h"
+#include "main/simple_list.h"
+#include "main/renderbuffer.h"
+#include "swrast/swrast.h"
+#include "vbo/vbo.h"
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "swrast_setup/swrast_setup.h"
+
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/polygon.h"
+#include "main/shaders.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+#include "main/stencil.h"
+#include "main/matrix.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/viewport.h"
+
+#include "dri_util.h"
+#include "vblank.h"
+
+#include "radeon_common.h"
+#include "radeon_bocs_wrapper.h"
+#include "radeon_lock.h"
+#include "radeon_drm.h"
+#include "radeon_mipmap_tree.h"
+
+#define DEBUG_CMDBUF         0
+
+/* =============================================================
+ * Scissoring
+ */
+
+static GLboolean intersect_rect(drm_clip_rect_t * out,
+				drm_clip_rect_t * a, drm_clip_rect_t * b)
+{
+	*out = *a;
+	if (b->x1 > out->x1)
+		out->x1 = b->x1;
+	if (b->y1 > out->y1)
+		out->y1 = b->y1;
+	if (b->x2 < out->x2)
+		out->x2 = b->x2;
+	if (b->y2 < out->y2)
+		out->y2 = b->y2;
+	if (out->x1 >= out->x2)
+		return GL_FALSE;
+	if (out->y1 >= out->y2)
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+void radeonRecalcScissorRects(radeonContextPtr radeon)
+{
+	drm_clip_rect_t *out;
+	int i;
+
+	/* Grow cliprect store?
+	 */
+	if (radeon->state.scissor.numAllocedClipRects < radeon->numClipRects) {
+		while (radeon->state.scissor.numAllocedClipRects <
+		       radeon->numClipRects) {
+			radeon->state.scissor.numAllocedClipRects += 1;	/* zero case */
+			radeon->state.scissor.numAllocedClipRects *= 2;
+		}
+
+		if (radeon->state.scissor.pClipRects)
+			FREE(radeon->state.scissor.pClipRects);
+
+		radeon->state.scissor.pClipRects =
+			MALLOC(radeon->state.scissor.numAllocedClipRects *
+			       sizeof(drm_clip_rect_t));
+
+		if (radeon->state.scissor.pClipRects == NULL) {
+			radeon->state.scissor.numAllocedClipRects = 0;
+			return;
+		}
+	}
+
+	out = radeon->state.scissor.pClipRects;
+	radeon->state.scissor.numClipRects = 0;
+
+	for (i = 0; i < radeon->numClipRects; i++) {
+		if (intersect_rect(out,
+				   &radeon->pClipRects[i],
+				   &radeon->state.scissor.rect)) {
+			radeon->state.scissor.numClipRects++;
+			out++;
+		}
+	}
+}
+
+void radeon_get_cliprects(radeonContextPtr radeon,
+			  struct drm_clip_rect **cliprects,
+			  unsigned int *num_cliprects,
+			  int *x_off, int *y_off)
+{
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(radeon);
+	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+
+	if (radeon->constant_cliprect) {
+		radeon->fboRect.x1 = 0;
+		radeon->fboRect.y1 = 0;
+		radeon->fboRect.x2 = radeon->glCtx->DrawBuffer->Width;
+		radeon->fboRect.y2 = radeon->glCtx->DrawBuffer->Height;
+
+		*cliprects = &radeon->fboRect;
+		*num_cliprects = 1;
+		*x_off = 0;
+		*y_off = 0;
+	} else if (radeon->front_cliprects ||
+		   rfb->pf_active || dPriv->numBackClipRects == 0) {
+		*cliprects = dPriv->pClipRects;
+		*num_cliprects = dPriv->numClipRects;
+		*x_off = dPriv->x;
+		*y_off = dPriv->y;
+	} else {
+		*num_cliprects = dPriv->numBackClipRects;
+		*cliprects = dPriv->pBackClipRects;
+		*x_off = dPriv->backX;
+		*y_off = dPriv->backY;
+	}
+}
+
+/**
+ * Update cliprects and scissors.
+ */
+void radeonSetCliprects(radeonContextPtr radeon)
+{
+	__DRIdrawablePrivate *const drawable = radeon_get_drawable(radeon);
+	__DRIdrawablePrivate *const readable = radeon_get_readable(radeon);
+	struct radeon_framebuffer *const draw_rfb = drawable->driverPrivate;
+	struct radeon_framebuffer *const read_rfb = readable->driverPrivate;
+	int x_off, y_off;
+
+	radeon_get_cliprects(radeon, &radeon->pClipRects,
+			     &radeon->numClipRects, &x_off, &y_off);
+
+	if ((draw_rfb->base.Width != drawable->w) ||
+	    (draw_rfb->base.Height != drawable->h)) {
+		_mesa_resize_framebuffer(radeon->glCtx, &draw_rfb->base,
+					 drawable->w, drawable->h);
+		draw_rfb->base.Initialized = GL_TRUE;
+	}
+
+	if (drawable != readable) {
+		if ((read_rfb->base.Width != readable->w) ||
+		    (read_rfb->base.Height != readable->h)) {
+			_mesa_resize_framebuffer(radeon->glCtx, &read_rfb->base,
+						 readable->w, readable->h);
+			read_rfb->base.Initialized = GL_TRUE;
+		}
+	}
+
+	if (radeon->state.scissor.enabled)
+		radeonRecalcScissorRects(radeon);
+
+}
+
+
+
+void radeonUpdateScissor( GLcontext *ctx )
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+
+	if ( !ctx->DrawBuffer->Name ) {
+		__DRIdrawablePrivate *dPriv = radeon_get_drawable(rmesa);
+
+		int x = ctx->Scissor.X;
+		int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height;
+		int w = ctx->Scissor.X + ctx->Scissor.Width - 1;
+		int h = dPriv->h - ctx->Scissor.Y - 1;
+
+		rmesa->state.scissor.rect.x1 = x + dPriv->x;
+		rmesa->state.scissor.rect.y1 = y + dPriv->y;
+		rmesa->state.scissor.rect.x2 = w + dPriv->x + 1;
+		rmesa->state.scissor.rect.y2 = h + dPriv->y + 1;
+	} else {
+		rmesa->state.scissor.rect.x1 = ctx->Scissor.X;
+		rmesa->state.scissor.rect.y1 = ctx->Scissor.Y;
+		rmesa->state.scissor.rect.x2 = ctx->Scissor.X + ctx->Scissor.Width;
+		rmesa->state.scissor.rect.y2 = ctx->Scissor.Y + ctx->Scissor.Height;
+	}
+
+	radeonRecalcScissorRects( rmesa );
+}
+
+/* =============================================================
+ * Scissoring
+ */
+
+void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	if (ctx->Scissor.Enabled) {
+		/* We don't pipeline cliprect changes */
+		radeon_firevertices(radeon);
+		radeonUpdateScissor(ctx);
+	}
+}
+
+
+/* ================================================================
+ * SwapBuffers with client-side throttling
+ */
+
+static uint32_t radeonGetLastFrame(radeonContextPtr radeon)
+{
+	drm_radeon_getparam_t gp;
+	int ret;
+	uint32_t frame = 0;
+
+	gp.param = RADEON_PARAM_LAST_FRAME;
+	gp.value = (int *)&frame;
+	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
+				  &gp, sizeof(gp));
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+
+	return frame;
+}
+
+uint32_t radeonGetAge(radeonContextPtr radeon)
+{
+	drm_radeon_getparam_t gp;
+	int ret;
+	uint32_t age;
+
+	gp.param = RADEON_PARAM_LAST_CLEAR;
+	gp.value = (int *)&age;
+	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM,
+				  &gp, sizeof(gp));
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+
+	return age;
+}
+
+static void radeonEmitIrqLocked(radeonContextPtr radeon)
+{
+	drm_radeon_irq_emit_t ie;
+	int ret;
+
+	ie.irq_seq = &radeon->iw.irq_seq;
+	ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT,
+				  &ie, sizeof(ie));
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+}
+
+static void radeonWaitIrq(radeonContextPtr radeon)
+{
+	int ret;
+
+	do {
+		ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT,
+				      &radeon->iw, sizeof(radeon->iw));
+	} while (ret && (errno == EINTR || errno == EBUSY));
+
+	if (ret) {
+		fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__,
+			ret);
+		exit(1);
+	}
+}
+
+static void radeonWaitForFrameCompletion(radeonContextPtr radeon)
+{
+	drm_radeon_sarea_t *sarea = radeon->sarea;
+
+	if (radeon->do_irqs) {
+		if (radeonGetLastFrame(radeon) < sarea->last_frame) {
+			if (!radeon->irqsEmitted) {
+				while (radeonGetLastFrame(radeon) <
+				       sarea->last_frame) ;
+			} else {
+				UNLOCK_HARDWARE(radeon);
+				radeonWaitIrq(radeon);
+				LOCK_HARDWARE(radeon);
+			}
+			radeon->irqsEmitted = 10;
+		}
+
+		if (radeon->irqsEmitted) {
+			radeonEmitIrqLocked(radeon);
+			radeon->irqsEmitted--;
+		}
+	} else {
+		while (radeonGetLastFrame(radeon) < sarea->last_frame) {
+			UNLOCK_HARDWARE(radeon);
+			if (radeon->do_usleeps)
+				DO_USLEEP(1);
+			LOCK_HARDWARE(radeon);
+		}
+	}
+}
+
+/* wait for idle */
+void radeonWaitForIdleLocked(radeonContextPtr radeon)
+{
+	int ret;
+	int i = 0;
+
+	do {
+		ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE);
+		if (ret)
+			DO_USLEEP(1);
+	} while (ret && ++i < 100);
+
+	if (ret < 0) {
+		UNLOCK_HARDWARE(radeon);
+		fprintf(stderr, "Error: R300 timed out... exiting\n");
+		exit(-1);
+	}
+}
+
+static void radeonWaitForIdle(radeonContextPtr radeon)
+{
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+        LOCK_HARDWARE(radeon);
+	    radeonWaitForIdleLocked(radeon);
+	    UNLOCK_HARDWARE(radeon);
+    }
+}
+
+static void radeon_flip_renderbuffers(struct radeon_framebuffer *rfb)
+{
+	int current_page = rfb->pf_current_page;
+	int next_page = (current_page + 1) % rfb->pf_num_pages;
+	struct gl_renderbuffer *tmp_rb;
+
+	/* Exchange renderbuffers if necessary but make sure their
+	 * reference counts are preserved.
+	 */
+	if (rfb->color_rb[current_page] &&
+	    rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer !=
+	    &rfb->color_rb[current_page]->base) {
+		tmp_rb = NULL;
+		_mesa_reference_renderbuffer(&tmp_rb,
+					     rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+		tmp_rb = &rfb->color_rb[current_page]->base;
+		_mesa_reference_renderbuffer(&rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb);
+		_mesa_reference_renderbuffer(&tmp_rb, NULL);
+	}
+
+	if (rfb->color_rb[next_page] &&
+	    rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer !=
+	    &rfb->color_rb[next_page]->base) {
+		tmp_rb = NULL;
+		_mesa_reference_renderbuffer(&tmp_rb,
+					     rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+		tmp_rb = &rfb->color_rb[next_page]->base;
+		_mesa_reference_renderbuffer(&rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb);
+		_mesa_reference_renderbuffer(&tmp_rb, NULL);
+	}
+}
+
+/* Copy the back color buffer to the front color buffer.
+ */
+void radeonCopyBuffer( __DRIdrawablePrivate *dPriv,
+		       const drm_clip_rect_t	  *rect)
+{
+	radeonContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	GLint nbox, i, ret;
+
+	assert(dPriv);
+	assert(dPriv->driContextPriv);
+	assert(dPriv->driContextPriv->driverPrivate);
+
+	rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+
+	LOCK_HARDWARE(rmesa);
+
+	rfb = dPriv->driverPrivate;
+
+	if ( RADEON_DEBUG & DEBUG_IOCTL ) {
+		fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx );
+	}
+
+	nbox = dPriv->numClipRects; /* must be in locked region */
+
+	for ( i = 0 ; i < nbox ; ) {
+		GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
+		drm_clip_rect_t *box = dPriv->pClipRects;
+		drm_clip_rect_t *b = rmesa->sarea->boxes;
+		GLint n = 0;
+
+		for ( ; i < nr ; i++ ) {
+
+			*b = box[i];
+
+			if (rect)
+			{
+				if (rect->x1 > b->x1)
+					b->x1 = rect->x1;
+				if (rect->y1 > b->y1)
+					b->y1 = rect->y1;
+				if (rect->x2 < b->x2)
+					b->x2 = rect->x2;
+				if (rect->y2 < b->y2)
+					b->y2 = rect->y2;
+
+				if (b->x1 >= b->x2 || b->y1 >= b->y2)
+					continue;
+			}
+
+			b++;
+			n++;
+		}
+		rmesa->sarea->nbox = n;
+
+		if (!n)
+			continue;
+
+		if (IS_R600_CLASS(rmesa->radeonScreen)) {
+			int cpp = rmesa->radeonScreen->cpp;
+			int src_pitch = rmesa->radeonScreen->backPitch * cpp;
+			int dst_pitch = rmesa->radeonScreen->frontPitch * cpp;
+			char *src = (char *)rmesa->radeonScreen->driScreen->pFB + rmesa->radeonScreen->backOffset;
+			char *dst = (char *)rmesa->radeonScreen->driScreen->pFB + rmesa->radeonScreen->frontOffset;
+			int j;
+			drm_clip_rect_t *pb = rmesa->sarea->boxes;
+
+			for (j = 0; j < n; j++) {
+				int x = pb[j].x1;
+				int y = pb[j].y1;
+				int w = pb[j].x2 - x;
+				int h = pb[j].y2 - y;
+
+				src += (y * src_pitch) + (x * cpp);
+				dst += (y * dst_pitch) + (x * cpp);
+
+				while (h--) {
+					memcpy(dst, src, w * cpp);
+					src += src_pitch;
+					dst += dst_pitch;
+				}
+			}
+		}
+
+		ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
+
+		if ( ret ) {
+			fprintf( stderr, "DRM_RADEON_SWAP_BUFFERS: return = %d\n", ret );
+			UNLOCK_HARDWARE( rmesa );
+			exit( 1 );
+		}
+	}
+
+	UNLOCK_HARDWARE( rmesa );
+}
+
+static int radeonScheduleSwap(__DRIdrawablePrivate *dPriv, GLboolean *missed_target)
+{
+	radeonContextPtr rmesa;
+
+	rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+	radeon_firevertices(rmesa);
+
+	LOCK_HARDWARE( rmesa );
+
+	if (!dPriv->numClipRects) {
+		UNLOCK_HARDWARE(rmesa);
+		usleep(10000);	/* throttle invisible client 10ms */
+		return 0;
+	}
+
+	radeonWaitForFrameCompletion(rmesa);
+
+	UNLOCK_HARDWARE(rmesa);
+	driWaitForVBlank(dPriv, missed_target);
+
+	return 0;
+}
+
+static GLboolean radeonPageFlip( __DRIdrawablePrivate *dPriv )
+{
+	radeonContextPtr radeon;
+	GLint ret;
+	__DRIscreenPrivate *psp;
+	struct radeon_renderbuffer *rrb;
+	struct radeon_framebuffer *rfb;
+
+	assert(dPriv);
+	assert(dPriv->driContextPriv);
+	assert(dPriv->driContextPriv->driverPrivate);
+
+	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+	rfb = dPriv->driverPrivate;
+	rrb = (void *)rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+
+	psp = dPriv->driScreenPriv;
+
+	LOCK_HARDWARE(radeon);
+
+	if ( RADEON_DEBUG & DEBUG_IOCTL ) {
+		fprintf(stderr, "%s: pfCurrentPage: %d %d\n", __FUNCTION__,
+			radeon->sarea->pfCurrentPage, radeon->sarea->pfState);
+	}
+	drm_clip_rect_t *box = dPriv->pClipRects;
+	drm_clip_rect_t *b = radeon->sarea->boxes;
+	b[0] = box[0];
+	radeon->sarea->nbox = 1;
+
+	ret = drmCommandNone( radeon->dri.fd, DRM_RADEON_FLIP );
+
+	UNLOCK_HARDWARE(radeon);
+
+	if ( ret ) {
+		fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret );
+		return GL_FALSE;
+	}
+
+	if (!rfb->pf_active)
+		return GL_FALSE;
+
+	rfb->pf_current_page = radeon->sarea->pfCurrentPage;
+	radeon_flip_renderbuffers(rfb);
+	radeon_draw_buffer(radeon->glCtx, &rfb->base);
+
+	return GL_TRUE;
+}
+
+
+/**
+ * Swap front and back buffer.
+ */
+void radeonSwapBuffers(__DRIdrawablePrivate * dPriv)
+{
+	int64_t ust;
+	__DRIscreenPrivate *psp;
+
+	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+		radeonContextPtr radeon;
+		GLcontext *ctx;
+
+		radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+		ctx = radeon->glCtx;
+
+		if (ctx->Visual.doubleBufferMode) {
+			GLboolean missed_target;
+			struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+			_mesa_notifySwapBuffers(ctx);/* flush pending rendering comands */
+
+			radeonScheduleSwap(dPriv, &missed_target);
+
+			if (rfb->pf_active) {
+				radeonPageFlip(dPriv);
+			} else {
+				radeonCopyBuffer(dPriv, NULL);
+			}
+
+			psp = dPriv->driScreenPriv;
+
+			rfb->swap_count++;
+			(*psp->systemTime->getUST)( & ust );
+			if ( missed_target ) {
+				rfb->swap_missed_count++;
+				rfb->swap_missed_ust = ust - rfb->swap_ust;
+			}
+
+			rfb->swap_ust = ust;
+			radeon->hw.all_dirty = GL_TRUE;
+		}
+	} else {
+		/* XXX this shouldn't be an error but we can't handle it for now */
+		_mesa_problem(NULL, "%s: drawable has no context!",
+			      __FUNCTION__);
+	}
+}
+
+void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
+			 int x, int y, int w, int h )
+{
+	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+		radeonContextPtr radeon;
+		GLcontext *ctx;
+
+		radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
+		ctx = radeon->glCtx;
+
+		if (ctx->Visual.doubleBufferMode) {
+			drm_clip_rect_t rect;
+			rect.x1 = x + dPriv->x;
+			rect.y1 = (dPriv->h - y - h) + dPriv->y;
+			rect.x2 = rect.x1 + w;
+			rect.y2 = rect.y1 + h;
+			_mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
+			radeonCopyBuffer(dPriv, &rect);
+		}
+	} else {
+		/* XXX this shouldn't be an error but we can't handle it for now */
+		_mesa_problem(NULL, "%s: drawable has no context!",
+			      __FUNCTION__);
+	}
+}
+
+void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct radeon_renderbuffer *rrbDepth = NULL, *rrbStencil = NULL,
+		*rrbColor = NULL;
+	uint32_t offset = 0;
+
+
+	if (!fb) {
+		/* this can happen during the initial context initialization */
+		return;
+	}
+
+	/* radeons only handle 1 color draw so far */
+	if (fb->_NumColorDrawBuffers != 1) {
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE);
+		return;
+	}
+
+	/* Do this here, note core Mesa, since this function is called from
+	 * many places within the driver.
+	 */
+	if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+		/* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */
+		_mesa_update_framebuffer(ctx);
+		/* this updates the DrawBuffer's Width/Height if it's a FBO */
+		_mesa_update_draw_buffer_bounds(ctx);
+	}
+
+	if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+		/* this may occur when we're called by glBindFrameBuffer() during
+		 * the process of someone setting up renderbuffers, etc.
+		 */
+		/*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/
+		return;
+	}
+
+	if (fb->Name)
+		;/* do something depthy/stencily TODO */
+
+
+		/* none */
+	if (fb->Name == 0) {
+		if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
+			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+			radeon->front_cliprects = GL_TRUE;
+			radeon->front_buffer_dirty = GL_TRUE;
+		} else {
+			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+			radeon->front_cliprects = GL_FALSE;
+		}
+	} else {
+		/* user FBO in theory */
+		struct radeon_renderbuffer *rrb;
+		rrb = radeon_renderbuffer(fb->_ColorDrawBuffers[0]);
+		if (rrb) {
+			offset = rrb->draw_offset;
+			rrbColor = rrb;
+		}
+		radeon->constant_cliprect = GL_TRUE;
+	}
+
+	if (rrbColor == NULL)
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE);
+	else
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_FALSE);
+
+
+	if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) {
+		rrbDepth = radeon_renderbuffer(fb->_DepthBuffer->Wrapped);
+		if (rrbDepth && rrbDepth->bo) {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_FALSE);
+		} else {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_TRUE);
+		}
+	} else {
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_FALSE);
+		rrbDepth = NULL;
+	}
+
+	if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) {
+		rrbStencil = radeon_renderbuffer(fb->_StencilBuffer->Wrapped);
+		if (rrbStencil && rrbStencil->bo) {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_FALSE);
+			/* need to re-compute stencil hw state */
+			if (!rrbDepth)
+				rrbDepth = rrbStencil;
+		} else {
+			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_TRUE);
+		}
+	} else {
+		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_FALSE);
+		if (ctx->Driver.Enable != NULL)
+			ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
+		else
+			ctx->NewState |= _NEW_STENCIL;
+	}
+
+	/* Update culling direction which changes depending on the
+	 * orientation of the buffer:
+	 */
+	if (ctx->Driver.FrontFace)
+		ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+	else
+		ctx->NewState |= _NEW_POLYGON;
+
+	/*
+	 * Update depth test state
+	 */
+	if (ctx->Driver.Enable) {
+		ctx->Driver.Enable(ctx, GL_DEPTH_TEST,
+				   (ctx->Depth.Test && fb->Visual.depthBits > 0));
+		/* Need to update the derived ctx->Stencil._Enabled first */
+		ctx->Driver.Enable(ctx, GL_STENCIL_TEST,
+				   (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0));
+	} else {
+		ctx->NewState |= (_NEW_DEPTH | _NEW_STENCIL);
+	}
+
+	_mesa_reference_renderbuffer(&radeon->state.depth.rb, &rrbDepth->base);
+	_mesa_reference_renderbuffer(&radeon->state.color.rb, &rrbColor->base);
+	radeon->state.color.draw_offset = offset;
+
+#if 0
+	/* update viewport since it depends on window size */
+	if (ctx->Driver.Viewport) {
+		ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y,
+				     ctx->Viewport.Width, ctx->Viewport.Height);
+	} else {
+
+	}
+#endif
+	ctx->NewState |= _NEW_VIEWPORT;
+
+	/* Set state we know depends on drawable parameters:
+	 */
+	radeonUpdateScissor(ctx);
+	radeon->NewGLState |= _NEW_SCISSOR;
+
+	if (ctx->Driver.DepthRange)
+		ctx->Driver.DepthRange(ctx,
+				       ctx->Viewport.Near,
+				       ctx->Viewport.Far);
+
+	/* Update culling direction which changes depending on the
+	 * orientation of the buffer:
+	 */
+	if (ctx->Driver.FrontFace)
+		ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+	else
+		ctx->NewState |= _NEW_POLYGON;
+}
+
+/**
+ * Called via glDrawBuffer.
+ */
+void radeonDrawBuffer( GLcontext *ctx, GLenum mode )
+{
+	if (RADEON_DEBUG & DEBUG_DRI)
+		fprintf(stderr, "%s %s\n", __FUNCTION__,
+			_mesa_lookup_enum_by_nr( mode ));
+
+	if (ctx->DrawBuffer->Name == 0) {
+		radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+
+		const GLboolean was_front_buffer_rendering =
+			radeon->is_front_buffer_rendering;
+
+		radeon->is_front_buffer_rendering = (mode == GL_FRONT_LEFT) ||
+                                            (mode == GL_FRONT);
+
+      /* If we weren't front-buffer rendering before but we are now, make sure
+       * that the front-buffer has actually been allocated.
+       */
+		if (!was_front_buffer_rendering && radeon->is_front_buffer_rendering) {
+			radeon_update_renderbuffers(radeon->dri.context,
+				radeon->dri.context->driDrawablePriv);
+      }
+	}
+
+	radeon_draw_buffer(ctx, ctx->DrawBuffer);
+}
+
+void radeonReadBuffer( GLcontext *ctx, GLenum mode )
+{
+	if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) {
+		struct radeon_context *const rmesa = RADEON_CONTEXT(ctx);
+		const GLboolean was_front_buffer_reading = rmesa->is_front_buffer_reading;
+		rmesa->is_front_buffer_reading = (mode == GL_FRONT_LEFT)
+					|| (mode == GL_FRONT);
+
+		if (!was_front_buffer_reading && rmesa->is_front_buffer_reading) {
+			radeon_update_renderbuffers(rmesa->dri.context,
+						    rmesa->dri.context->driReadablePriv);
+	 	}
+	}
+	/* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
+	if (ctx->ReadBuffer == ctx->DrawBuffer) {
+		/* This will update FBO completeness status.
+		 * A framebuffer will be incomplete if the GL_READ_BUFFER setting
+		 * refers to a missing renderbuffer.  Calling glReadBuffer can set
+		 * that straight and can make the drawing buffer complete.
+		 */
+		radeon_draw_buffer(ctx, ctx->DrawBuffer);
+	}
+}
+
+
+/* Turn on/off page flipping according to the flags in the sarea:
+ */
+void radeonUpdatePageFlipping(radeonContextPtr radeon)
+{
+	struct radeon_framebuffer *rfb = radeon_get_drawable(radeon)->driverPrivate;
+
+	rfb->pf_active = radeon->sarea->pfState;
+	rfb->pf_current_page = radeon->sarea->pfCurrentPage;
+	rfb->pf_num_pages = 2;
+	radeon_flip_renderbuffers(rfb);
+	radeon_draw_buffer(radeon->glCtx, radeon->glCtx->DrawBuffer);
+}
+
+void radeon_window_moved(radeonContextPtr radeon)
+{
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+		radeonUpdatePageFlipping(radeon);
+	}
+	radeonSetCliprects(radeon);
+}
+
+void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	__DRIcontext *driContext = radeon->dri.context;
+	void (*old_viewport)(GLcontext *ctx, GLint x, GLint y,
+			     GLsizei w, GLsizei h);
+
+	if (!driContext->driScreenPriv->dri2.enabled)
+		return;
+
+	if (!radeon->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+		if (radeon->is_front_buffer_rendering) {
+			radeonFlush(ctx);
+		}
+		radeon_update_renderbuffers(driContext, driContext->driDrawablePriv);
+		if (driContext->driDrawablePriv != driContext->driReadablePriv)
+			radeon_update_renderbuffers(driContext, driContext->driReadablePriv);
+	}
+
+	old_viewport = ctx->Driver.Viewport;
+	ctx->Driver.Viewport = NULL;
+	radeon_window_moved(radeon);
+	radeon_draw_buffer(ctx, radeon->glCtx->DrawBuffer);
+	ctx->Driver.Viewport = old_viewport;
+}
+
+static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state_atom *state)
+{
+	int i, j, reg;
+	int dwords = (*state->check) (radeon->glCtx, state);
+	drm_r300_cmd_header_t cmd;
+
+	fprintf(stderr, "  emit %s %d/%d\n", state->name, dwords, state->cmd_size);
+
+	if (RADEON_DEBUG & DEBUG_VERBOSE) {
+		for (i = 0; i < dwords;) {
+			cmd = *((drm_r300_cmd_header_t *) &state->cmd[i]);
+			reg = (cmd.packet0.reghi << 8) | cmd.packet0.reglo;
+			fprintf(stderr, "      %s[%d]: cmdpacket0 (first reg=0x%04x, count=%d)\n",
+					state->name, i, reg, cmd.packet0.count);
+			++i;
+			for (j = 0; j < cmd.packet0.count && i < dwords; j++) {
+				fprintf(stderr, "      %s[%d]: 0x%04x = %08x\n",
+						state->name, i, reg, state->cmd[i]);
+				reg += 4;
+				++i;
+			}
+		}
+	}
+}
+
+static void radeon_print_state_atom_kmm(radeonContextPtr radeon, struct radeon_state_atom *state)
+{
+	int i, j, reg, count;
+	int dwords = (*state->check) (radeon->glCtx, state);
+	uint32_t packet0;
+
+	fprintf(stderr, "  emit %s %d/%d\n", state->name, dwords, state->cmd_size);
+
+	if (RADEON_DEBUG & DEBUG_VERBOSE) {
+		for (i = 0; i < dwords;) {
+			packet0 = state->cmd[i];
+			reg = (packet0 & 0x1FFF) << 2;
+			count = ((packet0 & 0x3FFF0000) >> 16) + 1;
+			fprintf(stderr, "      %s[%d]: cmdpacket0 (first reg=0x%04x, count=%d)\n",
+					state->name, i, reg, count);
+			++i;
+			for (j = 0; j < count && i < dwords; j++) {
+				fprintf(stderr, "      %s[%d]: 0x%04x = %08x\n",
+						state->name, i, reg, state->cmd[i]);
+				reg += 4;
+				++i;
+			}
+		}
+	}
+}
+
+static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean dirty)
+{
+	BATCH_LOCALS(radeon);
+	struct radeon_state_atom *atom;
+	int dwords;
+
+	if (radeon->vtbl.pre_emit_atoms)
+		radeon->vtbl.pre_emit_atoms(radeon);
+
+	/* Emit actual atoms */
+	foreach(atom, &radeon->hw.atomlist) {
+		if ((atom->dirty || radeon->hw.all_dirty) == dirty) {
+			dwords = (*atom->check) (radeon->glCtx, atom);
+			if (dwords) {
+				if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) {
+					if (radeon->radeonScreen->kernel_mm)
+						radeon_print_state_atom_kmm(radeon, atom);
+					else
+						radeon_print_state_atom(radeon, atom);
+				}
+				if (atom->emit) {
+					(*atom->emit)(radeon->glCtx, atom);
+				} else {
+					BEGIN_BATCH_NO_AUTOSTATE(dwords);
+					OUT_BATCH_TABLE(atom->cmd, dwords);
+					END_BATCH();
+				}
+				atom->dirty = GL_FALSE;
+			} else {
+				if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) {
+					fprintf(stderr, "  skip state %s\n",
+						atom->name);
+				}
+			}
+		}
+	}
+
+	COMMIT_BATCH();
+}
+
+static GLboolean radeon_revalidate_bos(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	int ret;
+
+	ret = radeon_cs_space_check(radeon->cmdbuf.cs);
+	if (ret == RADEON_CS_SPACE_FLUSH)
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+void radeonEmitState(radeonContextPtr radeon)
+{
+	if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
+		fprintf(stderr, "%s\n", __FUNCTION__);
+
+	if (radeon->vtbl.pre_emit_state)
+		radeon->vtbl.pre_emit_state(radeon);
+
+	/* this code used to return here but now it emits zbs */
+	if (radeon->cmdbuf.cs->cdw && !radeon->hw.is_dirty && !radeon->hw.all_dirty)
+		return;
+
+	/* To avoid going across the entire set of states multiple times, just check
+	 * for enough space for the case of emitting all state, and inline the
+	 * radeonAllocCmdBuf code here without all the checks.
+	 */
+	rcommonEnsureCmdBufSpace(radeon, radeon->hw.max_state_size, __FUNCTION__);
+
+	if (!radeon->cmdbuf.cs->cdw) {
+		if (RADEON_DEBUG & DEBUG_STATE)
+			fprintf(stderr, "Begin reemit state\n");
+
+		radeonEmitAtoms(radeon, GL_FALSE);
+	}
+
+	if (RADEON_DEBUG & DEBUG_STATE)
+		fprintf(stderr, "Begin dirty state\n");
+
+	radeonEmitAtoms(radeon, GL_TRUE);
+	radeon->hw.is_dirty = GL_FALSE;
+	radeon->hw.all_dirty = GL_FALSE;
+
+}
+
+
+void radeonFlush(GLcontext *ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	if (RADEON_DEBUG & DEBUG_IOCTL)
+		fprintf(stderr, "%s %d\n", __FUNCTION__, radeon->cmdbuf.cs->cdw);
+
+	/* okay if we have no cmds in the buffer &&
+	   we have no DMA flush &&
+	   we have no DMA buffer allocated.
+	   then no point flushing anything at all.
+	*/
+	if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && !radeon->dma.current)
+		return;
+
+	if (radeon->dma.flush)
+		radeon->dma.flush( ctx );
+
+	radeonEmitState(radeon);
+
+	if (radeon->cmdbuf.cs->cdw)
+		rcommonFlushCmdBuf(radeon, __FUNCTION__);
+
+	if ((ctx->DrawBuffer->Name == 0) && radeon->front_buffer_dirty) {
+		__DRIscreen *const screen = radeon->radeonScreen->driScreen;
+
+		if (screen->dri2.loader && (screen->dri2.loader->base.version >= 2)
+			&& (screen->dri2.loader->flushFrontBuffer != NULL)) {
+			__DRIdrawablePrivate * drawable = radeon_get_drawable(radeon);
+			(*screen->dri2.loader->flushFrontBuffer)(drawable, drawable->loaderPrivate);
+
+			/* Only clear the dirty bit if front-buffer rendering is no longer
+			 * enabled.  This is done so that the dirty bit can only be set in
+			 * glDrawBuffer.  Otherwise the dirty bit would have to be set at
+			 * each of N places that do rendering.  This has worse performances,
+			 * but it is much easier to get correct.
+			 */
+			if (radeon->is_front_buffer_rendering) {
+				radeon->front_buffer_dirty = GL_FALSE;
+			}
+		}
+	}
+}
+
+/* Make sure all commands have been sent to the hardware and have
+ * completed processing.
+ */
+void radeonFinish(GLcontext * ctx)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
+	int i;
+
+	if (ctx->Driver.Flush)
+		ctx->Driver.Flush(ctx); /* +r6/r7 */
+
+	if (radeon->radeonScreen->kernel_mm) {
+		for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
+			struct radeon_renderbuffer *rrb;
+			rrb = radeon_renderbuffer(fb->_ColorDrawBuffers[i]);
+			if (rrb && rrb->bo)
+				radeon_bo_wait(rrb->bo);
+		}
+		{
+			struct radeon_renderbuffer *rrb;
+			rrb = radeon_get_depthbuffer(radeon);
+			if (rrb && rrb->bo)
+				radeon_bo_wait(rrb->bo);
+		}
+	} else if (radeon->do_irqs) {
+		LOCK_HARDWARE(radeon);
+		radeonEmitIrqLocked(radeon);
+		UNLOCK_HARDWARE(radeon);
+		radeonWaitIrq(radeon);
+	} else {
+		radeonWaitForIdle(radeon);
+	}
+}
+
+/* cmdbuffer */
+/**
+ * Send the current command buffer via ioctl to the hardware.
+ */
+int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller)
+{
+	int ret = 0;
+
+	if (rmesa->cmdbuf.flushing) {
+		fprintf(stderr, "Recursive call into r300FlushCmdBufLocked!\n");
+		exit(-1);
+	}
+	rmesa->cmdbuf.flushing = 1;
+
+	if (RADEON_DEBUG & DEBUG_IOCTL) {
+		fprintf(stderr, "%s from %s - %i cliprects\n",
+			__FUNCTION__, caller, rmesa->numClipRects);
+	}
+
+	if (rmesa->cmdbuf.cs->cdw) {
+		ret = radeon_cs_emit(rmesa->cmdbuf.cs);
+		rmesa->hw.all_dirty = GL_TRUE;
+	}
+	radeon_cs_erase(rmesa->cmdbuf.cs);
+	rmesa->cmdbuf.flushing = 0;
+
+	if (radeon_revalidate_bos(rmesa->glCtx) == GL_FALSE) {
+		fprintf(stderr,"failed to revalidate buffers\n");
+	}
+
+	return ret;
+}
+
+int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller)
+{
+	int ret;
+
+	radeonReleaseDmaRegion(rmesa);
+
+	LOCK_HARDWARE(rmesa);
+	ret = rcommonFlushCmdBufLocked(rmesa, caller);
+	UNLOCK_HARDWARE(rmesa);
+
+	if (ret) {
+		fprintf(stderr, "drmRadeonCmdBuffer: %d\n", ret);
+		_mesa_exit(ret);
+	}
+
+	return ret;
+}
+
+/**
+ * Make sure that enough space is available in the command buffer
+ * by flushing if necessary.
+ *
+ * \param dwords The number of dwords we need to be free on the command buffer
+ */
+void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller)
+{
+	if ((rmesa->cmdbuf.cs->cdw + dwords + 128) > rmesa->cmdbuf.size ||
+	    radeon_cs_need_flush(rmesa->cmdbuf.cs)) {
+		rcommonFlushCmdBuf(rmesa, caller);
+	}
+}
+
+void rcommonInitCmdBuf(radeonContextPtr rmesa)
+{
+	GLuint size;
+	/* Initialize command buffer */
+	size = 256 * driQueryOptioni(&rmesa->optionCache,
+				     "command_buffer_size");
+	if (size < 2 * rmesa->hw.max_state_size) {
+		size = 2 * rmesa->hw.max_state_size + 65535;
+	}
+	if (size > 64 * 256)
+		size = 64 * 256;
+
+	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) {
+		fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n",
+			sizeof(drm_r300_cmd_header_t));
+		fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%zd\n",
+			sizeof(drm_radeon_cmd_buffer_t));
+		fprintf(stderr,
+			"Allocating %d bytes command buffer (max state is %d bytes)\n",
+			size * 4, rmesa->hw.max_state_size * 4);
+	}
+
+	if (rmesa->radeonScreen->kernel_mm) {
+		int fd = rmesa->radeonScreen->driScreen->fd;
+		rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd);
+	} else {
+		rmesa->cmdbuf.csm = radeon_cs_manager_legacy_ctor(rmesa);
+	}
+	if (rmesa->cmdbuf.csm == NULL) {
+		/* FIXME: fatal error */
+		return;
+	}
+	rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size);
+	assert(rmesa->cmdbuf.cs != NULL);
+	rmesa->cmdbuf.size = size;
+
+	radeon_cs_space_set_flush(rmesa->cmdbuf.cs,
+				  (void (*)(void *))radeonFlush, rmesa->glCtx);
+
+	if (!rmesa->radeonScreen->kernel_mm) {
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]);
+		radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size);
+	} else {
+		struct drm_radeon_gem_info mminfo = { 0 };
+
+		if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo)))
+		{
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_visible);
+			radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size);
+		}
+	}
+
+}
+/**
+ * Destroy the command buffer
+ */
+void rcommonDestroyCmdBuf(radeonContextPtr rmesa)
+{
+	radeon_cs_destroy(rmesa->cmdbuf.cs);
+	if (rmesa->radeonScreen->driScreen->dri2.enabled || rmesa->radeonScreen->kernel_mm) {
+		radeon_cs_manager_gem_dtor(rmesa->cmdbuf.csm);
+	} else {
+		radeon_cs_manager_legacy_dtor(rmesa->cmdbuf.csm);
+	}
+}
+
+void rcommonBeginBatch(radeonContextPtr rmesa, int n,
+		       int dostate,
+		       const char *file,
+		       const char *function,
+		       int line)
+{
+	rcommonEnsureCmdBufSpace(rmesa, n, function);
+	if (!rmesa->cmdbuf.cs->cdw && dostate) {
+		if (RADEON_DEBUG & DEBUG_IOCTL)
+			fprintf(stderr, "Reemit state after flush (from %s)\n", function);
+		radeonEmitState(rmesa);
+	}
+	radeon_cs_begin(rmesa->cmdbuf.cs, n, file, function, line);
+
+        if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_IOCTL)
+                fprintf(stderr, "BEGIN_BATCH(%d) at %d, from %s:%i\n",
+                        n, rmesa->cmdbuf.cs->cdw, function, line);
+
+}
+
+void radeonUserClear(GLcontext *ctx, GLuint mask)
+{
+   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   meta_clear_tris(&rmesa->meta, mask);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h
new file mode 100644
index 0000000000..cebae18b2d
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common.h
@@ -0,0 +1,85 @@
+#ifndef COMMON_MISC_H
+#define COMMON_MISC_H
+
+#include "radeon_common_context.h"
+#include "radeon_dma.h"
+#include "radeon_texture.h"
+
+void radeonUserClear(GLcontext *ctx, GLuint mask);
+void radeonRecalcScissorRects(radeonContextPtr radeon);
+void radeonSetCliprects(radeonContextPtr radeon);
+void radeonUpdateScissor( GLcontext *ctx );
+void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h);
+
+void radeonWaitForIdleLocked(radeonContextPtr radeon);
+extern uint32_t radeonGetAge(radeonContextPtr radeon);
+void radeonCopyBuffer( __DRIdrawablePrivate *dPriv,
+		       const drm_clip_rect_t	  *rect);
+void radeonSwapBuffers(__DRIdrawablePrivate * dPriv);
+void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
+			 int x, int y, int w, int h );
+
+void radeonUpdatePageFlipping(radeonContextPtr rmesa);
+
+void radeonFlush(GLcontext *ctx);
+void radeonFinish(GLcontext * ctx);
+void radeonEmitState(radeonContextPtr radeon);
+
+void radeon_clear_tris(GLcontext *ctx, GLbitfield mask);
+
+void radeon_window_moved(radeonContextPtr radeon);
+void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb);
+void radeonDrawBuffer( GLcontext *ctx, GLenum mode );
+void radeonReadBuffer( GLcontext *ctx, GLenum mode );
+void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height);
+void radeon_get_cliprects(radeonContextPtr radeon,
+			  struct drm_clip_rect **cliprects,
+			  unsigned int *num_cliprects,
+			  int *x_off, int *y_off);
+void radeon_fbo_init(struct radeon_context *radeon);
+void
+radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
+			   struct radeon_bo *bo);
+struct radeon_renderbuffer *
+radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv);
+static inline struct radeon_renderbuffer *radeon_renderbuffer(struct gl_renderbuffer *rb)
+{
+	struct radeon_renderbuffer *rrb = (struct radeon_renderbuffer *)rb;
+	if (rrb && rrb->base.ClassID == RADEON_RB_CLASS)
+		return rrb;
+	else
+		return NULL;
+}
+
+static inline struct radeon_renderbuffer *radeon_get_renderbuffer(struct gl_framebuffer *fb, int att_index)
+{
+	if (att_index >= 0)
+		return radeon_renderbuffer(fb->Attachment[att_index].Renderbuffer);
+	else
+		return NULL;
+}
+
+static inline struct radeon_renderbuffer *radeon_get_depthbuffer(radeonContextPtr rmesa)
+{
+	struct radeon_renderbuffer *rrb;
+	rrb = radeon_renderbuffer(rmesa->state.depth.rb);
+	if (!rrb)
+		return NULL;
+
+	return rrb;
+}
+
+static inline struct radeon_renderbuffer *radeon_get_colorbuffer(radeonContextPtr rmesa)
+{
+	struct radeon_renderbuffer *rrb;
+
+	rrb = radeon_renderbuffer(rmesa->state.color.rb);
+	if (!rrb)
+		return NULL;
+	return rrb;
+}
+
+#include "radeon_cmdbuf.h"
+
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
new file mode 100644
index 0000000000..285e015c92
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -0,0 +1,864 @@
+/**************************************************************************
+
+Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
+                     VA Linux Systems Inc., Fremont, California.
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial
+portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+#include "radeon_common.h"
+#include "xmlpool.h"		/* for symbolic values of enum-type options */
+#include "utils.h"
+#include "vblank.h"
+#include "drirenderbuffer.h"
+#include "main/context.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/state.h"
+#include "main/simple_list.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/tnl.h"
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */
+#include "r600_context.h"
+#endif
+
+#define DRIVER_DATE "20090101"
+
+#ifndef RADEON_DEBUG
+int RADEON_DEBUG = (0);
+#endif
+
+
+static const char* get_chip_family_name(int chip_family)
+{
+	switch(chip_family) {
+	case CHIP_FAMILY_R100: return "R100";
+	case CHIP_FAMILY_RV100: return "RV100";
+	case CHIP_FAMILY_RS100: return "RS100";
+	case CHIP_FAMILY_RV200: return "RV200";
+	case CHIP_FAMILY_RS200: return "RS200";
+	case CHIP_FAMILY_R200: return "R200";
+	case CHIP_FAMILY_RV250: return "RV250";
+	case CHIP_FAMILY_RS300: return "RS300";
+	case CHIP_FAMILY_RV280: return "RV280";
+	case CHIP_FAMILY_R300: return "R300";
+	case CHIP_FAMILY_R350: return "R350";
+	case CHIP_FAMILY_RV350: return "RV350";
+	case CHIP_FAMILY_RV380: return "RV380";
+	case CHIP_FAMILY_R420: return "R420";
+	case CHIP_FAMILY_RV410: return "RV410";
+	case CHIP_FAMILY_RS400: return "RS400";
+	case CHIP_FAMILY_RS600: return "RS600";
+	case CHIP_FAMILY_RS690: return "RS690";
+	case CHIP_FAMILY_RS740: return "RS740";
+	case CHIP_FAMILY_RV515: return "RV515";
+	case CHIP_FAMILY_R520: return "R520";
+	case CHIP_FAMILY_RV530: return "RV530";
+	case CHIP_FAMILY_R580: return "R580";
+	case CHIP_FAMILY_RV560: return "RV560";
+	case CHIP_FAMILY_RV570: return "RV570";
+	default: return "unknown";
+	}
+}
+
+
+/* Return various strings for glGetString().
+ */
+static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name)
+{
+	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+	static char buffer[128];
+
+	switch (name) {
+	case GL_VENDOR:
+		if (IS_R600_CLASS(radeon->radeonScreen))
+			return (GLubyte *) "Advanced Micro Devices, Inc.";
+		else if (IS_R300_CLASS(radeon->radeonScreen))
+			return (GLubyte *) "DRI R300 Project";
+		else
+			return (GLubyte *) "Tungsten Graphics, Inc.";
+
+	case GL_RENDERER:
+	{
+		unsigned offset;
+		GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
+			radeon->radeonScreen->AGPMode;
+		const char* chipclass;
+		char hardwarename[32];
+
+		if (IS_R600_CLASS(radeon->radeonScreen))
+			chipclass = "R600";
+		else if (IS_R300_CLASS(radeon->radeonScreen))
+			chipclass = "R300";
+		else if (IS_R200_CLASS(radeon->radeonScreen))
+			chipclass = "R200";
+		else
+			chipclass = "R100";
+
+		sprintf(hardwarename, "%s (%s %04X)",
+		        chipclass,
+		        get_chip_family_name(radeon->radeonScreen->chip_family),
+		        radeon->radeonScreen->device_id);
+
+		offset = driGetRendererString(buffer, hardwarename, DRIVER_DATE,
+					      agp_mode);
+
+		if (IS_R600_CLASS(radeon->radeonScreen)) {
+			sprintf(&buffer[offset], " TCL");
+		} else if (IS_R300_CLASS(radeon->radeonScreen)) {
+			sprintf(&buffer[offset], " %sTCL",
+				(radeon->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)
+				? "" : "NO-");
+		} else {
+			sprintf(&buffer[offset], " %sTCL",
+				!(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)
+				? "" : "NO-");
+		}
+
+		if (radeon->radeonScreen->driScreen->dri2.enabled)
+			strcat(buffer, " DRI2");
+
+		return (GLubyte *) buffer;
+	}
+
+	default:
+		return NULL;
+	}
+}
+
+/* Initialize the driver's misc functions.
+ */
+static void radeonInitDriverFuncs(struct dd_function_table *functions)
+{
+	functions->GetString = radeonGetString;
+}
+
+/**
+ * Create and initialize all common fields of the context,
+ * including the Mesa context itself.
+ */
+GLboolean radeonInitContext(radeonContextPtr radeon,
+			    struct dd_function_table* functions,
+			    const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate)
+{
+	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+	GLcontext* ctx;
+	GLcontext* shareCtx;
+	int fthrottle_mode;
+
+	/* Fill in additional standard functions. */
+	radeonInitDriverFuncs(functions);
+
+	radeon->radeonScreen = screen;
+	/* Allocate and initialize the Mesa context */
+	if (sharedContextPrivate)
+		shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx;
+	else
+		shareCtx = NULL;
+	radeon->glCtx = _mesa_create_context(glVisual, shareCtx,
+					    functions, (void *)radeon);
+	if (!radeon->glCtx)
+		return GL_FALSE;
+
+	ctx = radeon->glCtx;
+	driContextPriv->driverPrivate = radeon;
+
+	meta_init_metaops(ctx, &radeon->meta);
+	/* DRI fields */
+	radeon->dri.context = driContextPriv;
+	radeon->dri.screen = sPriv;
+	radeon->dri.hwContext = driContextPriv->hHWContext;
+	radeon->dri.hwLock = &sPriv->pSAREA->lock;
+	radeon->dri.fd = sPriv->fd;
+	radeon->dri.drmMinor = sPriv->drm_version.minor;
+
+	radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA +
+					       screen->sarea_priv_offset);
+
+	/* Setup IRQs */
+	fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode");
+	radeon->iw.irq_seq = -1;
+	radeon->irqsEmitted = 0;
+	if (IS_R600_CLASS(radeon->radeonScreen))
+		radeon->do_irqs = 0;
+	else
+		radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS &&
+				   radeon->radeonScreen->irq);
+
+	radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
+
+	if (!radeon->do_irqs)
+		fprintf(stderr,
+			"IRQ's not enabled, falling back to %s: %d %d\n",
+			radeon->do_usleeps ? "usleeps" : "busy waits",
+			fthrottle_mode, radeon->radeonScreen->irq);
+
+        radeon->texture_depth = driQueryOptioni (&radeon->optionCache,
+					        "texture_depth");
+        if (radeon->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
+                radeon->texture_depth = ( glVisual->rgbBits > 16 ) ?
+	        DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
+
+	radeon->texture_row_align = 32;
+
+	return GL_TRUE;
+}
+
+
+
+/**
+ * Destroy the command buffer and state atoms.
+ */
+static void radeon_destroy_atom_list(radeonContextPtr radeon)
+{
+	struct radeon_state_atom *atom;
+
+	foreach(atom, &radeon->hw.atomlist) {
+		FREE(atom->cmd);
+		if (atom->lastcmd)
+			FREE(atom->lastcmd);
+	}
+
+}
+
+/**
+ * Cleanup common context fields.
+ * Called by r200DestroyContext/r300DestroyContext
+ */
+void radeonDestroyContext(__DRIcontextPrivate *driContextPriv )
+{
+#ifdef RADEON_BO_TRACK
+	FILE *track;
+#endif
+	GET_CURRENT_CONTEXT(ctx);
+	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
+	radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
+
+    /* +r6/r7 */
+    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+    /* --------- */
+
+	if (radeon == current) {
+		radeon_firevertices(radeon);
+		_mesa_make_current(NULL, NULL, NULL);
+	}
+
+	assert(radeon);
+	if (radeon) 
+    {
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */
+	    if (IS_R600_CLASS(screen))
+        {
+		r600DestroyContext(driContextPriv);
+        }
+#endif
+
+		if (radeon->dma.current) {
+			rcommonFlushCmdBuf( radeon, __FUNCTION__ );
+		}
+
+		radeonReleaseArrays(radeon->glCtx, ~0);
+		meta_destroy_metaops(&radeon->meta);
+		if (radeon->vtbl.free_context)
+			radeon->vtbl.free_context(radeon->glCtx);
+		_swsetup_DestroyContext( radeon->glCtx );
+		_tnl_DestroyContext( radeon->glCtx );
+		_vbo_DestroyContext( radeon->glCtx );
+		_swrast_DestroyContext( radeon->glCtx );
+
+		/* free atom list */
+		/* free the Mesa context */
+		_mesa_destroy_context(radeon->glCtx);
+
+		/* _mesa_destroy_context() might result in calls to functions that
+		 * depend on the DriverCtx, so don't set it to NULL before.
+		 *
+		 * radeon->glCtx->DriverCtx = NULL;
+		 */
+		/* free the option cache */
+		driDestroyOptionCache(&radeon->optionCache);
+
+		rcommonDestroyCmdBuf(radeon);
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */
+	    if (!IS_R600_CLASS(screen))
+#endif
+		radeon_destroy_atom_list(radeon);
+
+		if (radeon->state.scissor.pClipRects) {
+			FREE(radeon->state.scissor.pClipRects);
+			radeon->state.scissor.pClipRects = 0;
+		}
+	}
+#ifdef RADEON_BO_TRACK
+	track = fopen("/tmp/tracklog", "w");
+	if (track) {
+		radeon_tracker_print(&radeon->radeonScreen->bom->tracker, track);
+		fclose(track);
+	}
+#endif
+	FREE(radeon);
+}
+
+/* Force the context `c' to be unbound from its buffer.
+ */
+GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv)
+{
+	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
+
+	if (RADEON_DEBUG & DEBUG_DRI)
+		fprintf(stderr, "%s ctx %p\n", __FUNCTION__,
+			radeon->glCtx);
+
+	return GL_TRUE;
+}
+
+
+static void
+radeon_make_kernel_renderbuffer_current(radeonContextPtr radeon,
+					struct radeon_framebuffer *draw)
+{
+	/* if radeon->fake */
+	struct radeon_renderbuffer *rb;
+
+	if ((rb = (void *)draw->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->frontOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Front Buf");
+#else
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->frontOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->backOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Back Buf");
+#else
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->backOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->backPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_DEPTH].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Z Buf");
+#else
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_STENCIL].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Stencil Buf");
+#else
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+}
+
+static void
+radeon_make_renderbuffer_current(radeonContextPtr radeon,
+				 struct radeon_framebuffer *draw)
+{
+	int size = 4096*4096*4;
+	/* if radeon->fake */
+	struct radeon_renderbuffer *rb;
+
+	if (radeon->radeonScreen->kernel_mm) {
+		radeon_make_kernel_renderbuffer_current(radeon, draw);
+		return;
+	}
+
+
+	if ((rb = (void *)draw->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->frontOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Front Buf");
+#else
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->frontOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->backOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Back Buf");
+#else
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->backOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->backPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_DEPTH].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Z Buf");
+#else
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+	if ((rb = (void *)draw->base.Attachment[BUFFER_STENCIL].Renderbuffer)) {
+		if (!rb->bo) {
+#ifdef RADEON_DEBUG_BO
+			rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0,
+                        "Stencil Buf");
+#else
+            rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+						radeon->radeonScreen->depthOffset +
+						radeon->radeonScreen->fbLocation,
+						size,
+						4096,
+						RADEON_GEM_DOMAIN_VRAM,
+						0);
+#endif /* RADEON_DEBUG_BO */
+		}
+		rb->cpp = radeon->radeonScreen->cpp;
+		rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp;
+	}
+}
+
+static unsigned
+radeon_bits_per_pixel(const struct radeon_renderbuffer *rb)
+{
+   switch (rb->base._ActualFormat) {
+   case GL_RGB5:
+   case GL_DEPTH_COMPONENT16:
+      return 16;
+   case GL_RGB8:
+   case GL_RGBA8:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH24_STENCIL8_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+      return 32;
+   default:
+      return 0;
+   }
+}
+
+void
+radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
+{
+	unsigned int attachments[10];
+	__DRIbuffer *buffers = NULL;
+	__DRIscreen *screen;
+	struct radeon_renderbuffer *rb;
+	int i, count;
+	struct radeon_framebuffer *draw;
+	radeonContextPtr radeon;
+	char *regname;
+	struct radeon_bo *depth_bo = NULL, *bo;
+
+	if (RADEON_DEBUG & DEBUG_DRI)
+	    fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);
+
+	draw = drawable->driverPrivate;
+	screen = context->driScreenPriv;
+	radeon = (radeonContextPtr) context->driverPrivate;
+
+	if (screen->dri2.loader
+	   && (screen->dri2.loader->base.version > 2)
+	   && (screen->dri2.loader->getBuffersWithFormat != NULL)) {
+		struct radeon_renderbuffer *depth_rb;
+		struct radeon_renderbuffer *stencil_rb;
+
+		i = 0;
+		if ((radeon->is_front_buffer_rendering ||
+		     radeon->is_front_buffer_reading ||
+		     !draw->color_rb[1])
+		    && draw->color_rb[0]) {
+			attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+			attachments[i++] = radeon_bits_per_pixel(draw->color_rb[0]);
+		}
+
+		if (draw->color_rb[1]) {
+			attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+			attachments[i++] = radeon_bits_per_pixel(draw->color_rb[1]);
+		}
+
+		depth_rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+		stencil_rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
+
+		if ((depth_rb != NULL) && (stencil_rb != NULL)) {
+			attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL;
+			attachments[i++] = radeon_bits_per_pixel(depth_rb);
+		} else if (depth_rb != NULL) {
+			attachments[i++] = __DRI_BUFFER_DEPTH;
+			attachments[i++] = radeon_bits_per_pixel(depth_rb);
+		} else if (stencil_rb != NULL) {
+			attachments[i++] = __DRI_BUFFER_STENCIL;
+			attachments[i++] = radeon_bits_per_pixel(stencil_rb);
+		}
+
+		buffers = (*screen->dri2.loader->getBuffersWithFormat)(drawable,
+								&drawable->w,
+								&drawable->h,
+								attachments, i / 2,
+								&count,
+								drawable->loaderPrivate);
+	} else if (screen->dri2.loader) {
+		i = 0;
+		if (draw->color_rb[0])
+			attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+		if (draw->color_rb[1])
+			attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+		if (radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH))
+			attachments[i++] = __DRI_BUFFER_DEPTH;
+		if (radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL))
+			attachments[i++] = __DRI_BUFFER_STENCIL;
+
+		buffers = (*screen->dri2.loader->getBuffers)(drawable,
+								 &drawable->w,
+								 &drawable->h,
+								 attachments, i,
+								 &count,
+								 drawable->loaderPrivate);
+	}
+
+	if (buffers == NULL)
+		return;
+
+	/* set one cliprect to cover the whole drawable */
+	drawable->x = 0;
+	drawable->y = 0;
+	drawable->backX = 0;
+	drawable->backY = 0;
+	drawable->numClipRects = 1;
+	drawable->pClipRects[0].x1 = 0;
+	drawable->pClipRects[0].y1 = 0;
+	drawable->pClipRects[0].x2 = drawable->w;
+	drawable->pClipRects[0].y2 = drawable->h;
+	drawable->numBackClipRects = 1;
+	drawable->pBackClipRects[0].x1 = 0;
+	drawable->pBackClipRects[0].y1 = 0;
+	drawable->pBackClipRects[0].x2 = drawable->w;
+	drawable->pBackClipRects[0].y2 = drawable->h;
+	for (i = 0; i < count; i++) {
+		switch (buffers[i].attachment) {
+		case __DRI_BUFFER_FRONT_LEFT:
+			rb = draw->color_rb[0];
+			regname = "dri2 front buffer";
+			break;
+		case __DRI_BUFFER_FAKE_FRONT_LEFT:
+			rb = draw->color_rb[0];
+			regname = "dri2 fake front buffer";
+			break;
+		case __DRI_BUFFER_BACK_LEFT:
+			rb = draw->color_rb[1];
+			regname = "dri2 back buffer";
+			break;
+		case __DRI_BUFFER_DEPTH:
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+			regname = "dri2 depth buffer";
+			break;
+		case __DRI_BUFFER_DEPTH_STENCIL:
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+			regname = "dri2 depth / stencil buffer";
+			break;
+		case __DRI_BUFFER_STENCIL:
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
+			regname = "dri2 stencil buffer";
+			break;
+		case __DRI_BUFFER_ACCUM:
+		default:
+			fprintf(stderr,
+				"unhandled buffer attach event, attacment type %d\n",
+				buffers[i].attachment);
+			return;
+		}
+
+		if (rb == NULL)
+			continue;
+
+		if (rb->bo) {
+			uint32_t name = radeon_gem_name_bo(rb->bo);
+			if (name == buffers[i].name)
+				continue;
+		}
+
+		if (RADEON_DEBUG & DEBUG_DRI)
+			fprintf(stderr,
+				"attaching buffer %s, %d, at %d, cpp %d, pitch %d\n",
+				regname, buffers[i].name, buffers[i].attachment,
+				buffers[i].cpp, buffers[i].pitch);
+
+		rb->cpp = buffers[i].cpp;
+		rb->pitch = buffers[i].pitch;
+		rb->base.Width = drawable->w;
+		rb->base.Height = drawable->h;
+		rb->has_surface = 0;
+
+		if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) {
+			if (RADEON_DEBUG & DEBUG_DRI)
+				fprintf(stderr, "(reusing depth buffer as stencil)\n");
+			bo = depth_bo;
+			radeon_bo_ref(bo);
+		} else {
+#ifdef RADEON_DEBUG_BO
+            bo = radeon_bo_open(radeon->radeonScreen->bom,
+						buffers[i].name,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						buffers[i].flags,
+                        regname);
+#else
+			bo = radeon_bo_open(radeon->radeonScreen->bom,
+						buffers[i].name,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						buffers[i].flags);
+#endif /* RADEON_DEBUG_BO */
+			if (bo == NULL) {
+
+				fprintf(stderr, "failed to attach %s %d\n",
+					regname, buffers[i].name);
+
+			}
+		}
+
+		if (buffers[i].attachment == __DRI_BUFFER_DEPTH) {
+			if (draw->base.Visual.depthBits == 16)
+				rb->cpp = 2;
+			depth_bo = bo;
+		}
+
+		radeon_renderbuffer_set_bo(rb, bo);
+		radeon_bo_unref(bo);
+
+		if (buffers[i].attachment == __DRI_BUFFER_DEPTH_STENCIL) {
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
+			if (rb != NULL) {
+				struct radeon_bo *stencil_bo = NULL;
+
+				if (rb->bo) {
+					uint32_t name = radeon_gem_name_bo(rb->bo);
+					if (name == buffers[i].name)
+						continue;
+				}
+
+				stencil_bo = bo;
+				radeon_bo_ref(stencil_bo);
+				radeon_renderbuffer_set_bo(rb, stencil_bo);
+				radeon_bo_unref(stencil_bo);
+			}
+		}
+	}
+
+	driUpdateFramebufferSize(radeon->glCtx, drawable);
+}
+
+/* Force the context `c' to be the current context and associate with it
+ * buffer `b'.
+ */
+GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
+			    __DRIdrawablePrivate * driDrawPriv,
+			    __DRIdrawablePrivate * driReadPriv)
+{
+	radeonContextPtr radeon;
+	struct radeon_framebuffer *drfb;
+	struct gl_framebuffer *readfb;
+
+	if (!driContextPriv) {
+		if (RADEON_DEBUG & DEBUG_DRI)
+			fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
+		_mesa_make_current(NULL, NULL, NULL);
+		return GL_TRUE;
+	}
+
+	radeon = (radeonContextPtr) driContextPriv->driverPrivate;
+	drfb = driDrawPriv->driverPrivate;
+	readfb = driReadPriv->driverPrivate;
+
+	if (driContextPriv->driScreenPriv->dri2.enabled) {
+		radeon_update_renderbuffers(driContextPriv, driDrawPriv);
+		if (driDrawPriv != driReadPriv)
+			radeon_update_renderbuffers(driContextPriv, driReadPriv);
+		_mesa_reference_renderbuffer(&radeon->state.color.rb,
+			&(radeon_get_renderbuffer(&drfb->base, BUFFER_BACK_LEFT)->base));
+		_mesa_reference_renderbuffer(&radeon->state.depth.rb,
+			&(radeon_get_renderbuffer(&drfb->base, BUFFER_DEPTH)->base));
+	} else {
+		radeon_make_renderbuffer_current(radeon, drfb);
+	}
+
+	if (RADEON_DEBUG & DEBUG_DRI)
+	     fprintf(stderr, "%s ctx %p dfb %p rfb %p\n", __FUNCTION__, radeon->glCtx, drfb, readfb);
+
+	driUpdateFramebufferSize(radeon->glCtx, driDrawPriv);
+	if (driReadPriv != driDrawPriv)
+		driUpdateFramebufferSize(radeon->glCtx, driReadPriv);
+
+	_mesa_make_current(radeon->glCtx, &drfb->base, readfb);
+
+	_mesa_update_state(radeon->glCtx);
+
+	if (radeon->glCtx->DrawBuffer == &drfb->base) {
+		if (driDrawPriv->swap_interval == (unsigned)-1) {
+			int i;
+			driDrawPriv->vblFlags =
+				(radeon->radeonScreen->irq != 0)
+				? driGetDefaultVBlankFlags(&radeon->
+							   optionCache)
+				: VBLANK_FLAG_NO_IRQ;
+
+			driDrawableInitVBlank(driDrawPriv);
+			drfb->vbl_waited = driDrawPriv->vblSeq;
+
+			for (i = 0; i < 2; i++) {
+				if (drfb->color_rb[i])
+					drfb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq;
+			}
+
+		}
+
+		radeon_window_moved(radeon);
+		radeon_draw_buffer(radeon->glCtx, &drfb->base);
+	}
+
+
+	if (RADEON_DEBUG & DEBUG_DRI)
+		fprintf(stderr, "End %s\n", __FUNCTION__);
+
+	return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
new file mode 100644
index 0000000000..e4a8da0596
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -0,0 +1,596 @@
+
+#ifndef COMMON_CONTEXT_H
+#define COMMON_CONTEXT_H
+
+#include "main/mm.h"
+#include "math/m_vector.h"
+#include "texmem.h"
+#include "tnl/t_context.h"
+#include "main/colormac.h"
+
+#include "radeon_screen.h"
+#include "radeon_drm.h"
+#include "dri_util.h"
+#include "tnl/t_vertex.h"
+
+#include "dri_metaops.h"
+struct radeon_context;
+
+#include "radeon_bocs_wrapper.h"
+
+/* This union is used to avoid warnings/miscompilation
+   with float to uint32_t casts due to strict-aliasing */
+typedef union { GLfloat f; uint32_t ui32; } float_ui32_type;
+
+struct radeon_context;
+typedef struct radeon_context radeonContextRec;
+typedef struct radeon_context *radeonContextPtr;
+
+
+#define TEX_0   0x1
+#define TEX_1   0x2
+#define TEX_2   0x4
+#define TEX_3	0x8
+#define TEX_4	0x10
+#define TEX_5	0x20
+
+/* Rasterizing fallbacks */
+/* See correponding strings in r200_swtcl.c */
+#define RADEON_FALLBACK_TEXTURE		0x0001
+#define RADEON_FALLBACK_DRAW_BUFFER	0x0002
+#define RADEON_FALLBACK_STENCIL		0x0004
+#define RADEON_FALLBACK_RENDER_MODE	0x0008
+#define RADEON_FALLBACK_BLEND_EQ	0x0010
+#define RADEON_FALLBACK_BLEND_FUNC	0x0020
+#define RADEON_FALLBACK_DISABLE 	0x0040
+#define RADEON_FALLBACK_BORDER_MODE	0x0080
+#define RADEON_FALLBACK_DEPTH_BUFFER	0x0100
+#define RADEON_FALLBACK_STENCIL_BUFFER  0x0200
+
+#define R200_FALLBACK_TEXTURE           0x01
+#define R200_FALLBACK_DRAW_BUFFER       0x02
+#define R200_FALLBACK_STENCIL           0x04
+#define R200_FALLBACK_RENDER_MODE       0x08
+#define R200_FALLBACK_DISABLE           0x10
+#define R200_FALLBACK_BORDER_MODE       0x20
+
+#define RADEON_TCL_FALLBACK_RASTER            0x1 /* rasterization */
+#define RADEON_TCL_FALLBACK_UNFILLED          0x2 /* unfilled tris */
+#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE     0x4 /* twoside tris */
+#define RADEON_TCL_FALLBACK_MATERIAL          0x8 /* material in vb */
+#define RADEON_TCL_FALLBACK_TEXGEN_0          0x10 /* texgen, unit 0 */
+#define RADEON_TCL_FALLBACK_TEXGEN_1          0x20 /* texgen, unit 1 */
+#define RADEON_TCL_FALLBACK_TEXGEN_2          0x40 /* texgen, unit 2 */
+#define RADEON_TCL_FALLBACK_TCL_DISABLE       0x80 /* user disable */
+#define RADEON_TCL_FALLBACK_FOGCOORDSPEC      0x100 /* fogcoord, sep. spec light */
+
+/* The blit width for texture uploads
+ */
+#define BLIT_WIDTH_BYTES 1024
+
+/* Use the templated vertex format:
+ */
+#define COLOR_IS_RGBA
+#define TAG(x) radeon##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+#define RADEON_RB_CLASS 0xdeadbeef
+
+struct radeon_renderbuffer
+{
+	struct gl_renderbuffer base;
+	struct radeon_bo *bo;
+	unsigned int cpp;
+	/* unsigned int offset; */
+	unsigned int pitch;
+
+	uint32_t draw_offset; /* FBO */
+	/* boo Xorg 6.8.2 compat */
+	int has_surface;
+
+	GLuint pf_pending;  /**< sequence number of pending flip */
+	GLuint vbl_pending;   /**< vblank sequence number of pending flip */
+	__DRIdrawablePrivate *dPriv;
+};
+
+struct radeon_framebuffer
+{
+	struct gl_framebuffer base;
+
+	struct radeon_renderbuffer *color_rb[2];
+
+	GLuint vbl_waited;
+
+	/* buffer swap */
+	int64_t swap_ust;
+	int64_t swap_missed_ust;
+
+	GLuint swap_count;
+	GLuint swap_missed_count;
+
+	/* Drawable page flipping state */
+	GLboolean pf_active;
+	GLint pf_current_page;
+	GLint pf_num_pages;
+
+};
+
+
+struct radeon_colorbuffer_state {
+	GLuint clear;
+	int roundEnable;
+	struct gl_renderbuffer *rb;
+	uint32_t draw_offset; /* offset into color renderbuffer - FBOs */
+};
+
+struct radeon_depthbuffer_state {
+	GLuint clear;
+	struct gl_renderbuffer *rb;
+};
+
+struct radeon_scissor_state {
+	drm_clip_rect_t rect;
+	GLboolean enabled;
+
+	GLuint numClipRects;	/* Cliprects active */
+	GLuint numAllocedClipRects;	/* Cliprects available */
+	drm_clip_rect_t *pClipRects;
+};
+
+struct radeon_stencilbuffer_state {
+	GLuint clear;		/* rb3d_stencilrefmask value */
+};
+
+struct radeon_stipple_state {
+	GLuint mask[32];
+};
+
+struct radeon_state_atom {
+	struct radeon_state_atom *next, *prev;
+	const char *name;	/* for debug */
+	int cmd_size;		/* size in bytes */
+        GLuint idx;
+	GLuint is_tcl;
+        GLuint *cmd;		/* one or more cmd's */
+	GLuint *lastcmd;		/* one or more cmd's */
+	GLboolean dirty;	/* dirty-mark in emit_state_list */
+        int (*check) (GLcontext *, struct radeon_state_atom *atom); /* is this state active? */
+        void (*emit) (GLcontext *, struct radeon_state_atom *atom);
+};
+
+struct radeon_hw_state {
+  	/* Head of the linked list of state atoms. */
+	struct radeon_state_atom atomlist;
+	int max_state_size;	/* Number of bytes necessary for a full state emit. */
+	GLboolean is_dirty, all_dirty;
+};
+
+
+/* Texture related */
+typedef struct _radeon_texture_image radeon_texture_image;
+
+struct _radeon_texture_image {
+	struct gl_texture_image base;
+
+	/**
+	 * If mt != 0, the image is stored in hardware format in the
+	 * given mipmap tree. In this case, base.Data may point into the
+	 * mapping of the buffer object that contains the mipmap tree.
+	 *
+	 * If mt == 0, the image is stored in normal memory pointed to
+	 * by base.Data.
+	 */
+	struct _radeon_mipmap_tree *mt;
+	struct radeon_bo *bo;
+
+	int mtlevel; /** if mt != 0, this is the image's level in the mipmap tree */
+	int mtface; /** if mt != 0, this is the image's face in the mipmap tree */
+};
+
+
+static INLINE radeon_texture_image *get_radeon_texture_image(struct gl_texture_image *image)
+{
+	return (radeon_texture_image*)image;
+}
+
+
+typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr;
+
+#define RADEON_TXO_MICRO_TILE               (1 << 3)
+
+/* Texture object in locally shared texture space.
+ */
+struct radeon_tex_obj {
+	struct gl_texture_object base;
+	struct _radeon_mipmap_tree *mt;
+
+	/**
+	 * This is true if we've verified that the mipmap tree above is complete
+	 * and so on.
+	 */
+	GLboolean validated;
+
+	GLuint override_offset;
+	GLboolean image_override; /* Image overridden by GLX_EXT_tfp */
+	GLuint tile_bits;	/* hw texture tile bits used on this texture */
+        struct radeon_bo *bo;
+
+	GLuint pp_txfilter;	/* hardware register values */
+	GLuint pp_txformat;
+	GLuint pp_txformat_x;
+	GLuint pp_txsize;	/* npot only */
+	GLuint pp_txpitch;	/* npot only */
+	GLuint pp_border_color;
+	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
+
+        GLuint pp_txfilter_1;	/*  r300 */
+
+	/* r700 texture states */
+	GLuint SQ_TEX_RESOURCE0;
+	GLuint SQ_TEX_RESOURCE1;
+	GLuint SQ_TEX_RESOURCE2;
+	GLuint SQ_TEX_RESOURCE3;
+	GLuint SQ_TEX_RESOURCE4;
+	GLuint SQ_TEX_RESOURCE5;
+	GLuint SQ_TEX_RESOURCE6;
+
+	GLuint SQ_TEX_SAMPLER0;
+	GLuint SQ_TEX_SAMPLER1;
+	GLuint SQ_TEX_SAMPLER2;
+
+	GLboolean border_fallback;
+
+
+};
+
+static INLINE radeonTexObj* radeon_tex_obj(struct gl_texture_object *texObj)
+{
+	return (radeonTexObj*)texObj;
+}
+
+/* Need refcounting on dma buffers:
+ */
+struct radeon_dma_buffer {
+	int refcount;		/* the number of retained regions in buf */
+	drmBufPtr buf;
+};
+
+struct radeon_aos {
+	struct radeon_bo *bo; /** Buffer object where vertex data is stored */
+	int offset; /** Offset into buffer object, in bytes */
+	int components; /** Number of components per vertex */
+	int stride; /** Stride in dwords (may be 0 for repeating) */
+	int count; /** Number of vertices */
+};
+
+struct radeon_dma {
+        /* Active dma region.  Allocations for vertices and retained
+         * regions come from here.  Also used for emitting random vertices,
+         * these may be flushed by calling flush_current();
+         */
+        struct radeon_bo *current; /** Buffer that DMA memory is allocated from */
+        int current_used; /** Number of bytes allocated and forgotten about */
+        int current_vertexptr; /** End of active vertex region */
+
+        /**
+         * If current_vertexptr != current_used then flush must be non-zero.
+         * flush must be called before non-active vertex allocations can be
+         * performed.
+         */
+        void (*flush) (GLcontext *);
+
+        /* Number of "in-flight" DMA buffers, i.e. the number of buffers
+         * for which a DISCARD command is currently queued in the command buffer
+.
+         */
+        GLuint nr_released_bufs;
+};
+
+/* radeon_swtcl.c
+ */
+struct radeon_swtcl_info {
+
+	GLuint RenderIndex;
+	GLuint vertex_size;
+	GLubyte *verts;
+
+	/* Fallback rasterization functions
+	 */
+	GLuint hw_primitive;
+	GLenum render_primitive;
+	GLuint numverts;
+
+	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
+	GLuint vertex_attr_count;
+
+};
+
+#define RADEON_MAX_AOS_ARRAYS		16
+struct radeon_tcl_info {
+	struct radeon_aos aos[RADEON_MAX_AOS_ARRAYS];
+	GLuint aos_count;
+	struct radeon_bo *elt_dma_bo; /** Buffer object that contains element indices */
+	int elt_dma_offset; /** Offset into this buffer object, in bytes */
+};
+
+struct radeon_ioctl {
+	GLuint vertex_offset;
+        struct radeon_bo *bo;
+	GLuint vertex_size;
+};
+
+#define RADEON_MAX_PRIMS 64
+
+struct radeon_prim {
+	GLuint start;
+	GLuint end;
+	GLuint prim;
+};
+
+static INLINE GLuint radeonPackColor(GLuint cpp,
+                                     GLubyte r, GLubyte g,
+                                     GLubyte b, GLubyte a)
+{
+	switch (cpp) {
+	case 2:
+		return PACK_COLOR_565(r, g, b);
+	case 4:
+		return PACK_COLOR_8888(a, r, g, b);
+	default:
+		return 0;
+	}
+}
+
+#define MAX_CMD_BUF_SZ (16*1024)
+
+#define MAX_DMA_BUF_SZ (64*1024)
+
+struct radeon_store {
+	GLuint statenr;
+	GLuint primnr;
+	char cmd_buf[MAX_CMD_BUF_SZ];
+	int cmd_used;
+	int elts_start;
+};
+
+struct radeon_dri_mirror {
+	__DRIcontextPrivate *context;	/* DRI context */
+	__DRIscreenPrivate *screen;	/* DRI screen */
+
+	drm_context_t hwContext;
+	drm_hw_lock_t *hwLock;
+	int fd;
+	int drmMinor;
+};
+
+#define DEBUG_TEXTURE	0x001
+#define DEBUG_STATE	0x002
+#define DEBUG_IOCTL	0x004
+#define DEBUG_PRIMS	0x008
+#define DEBUG_VERTS	0x010
+#define DEBUG_FALLBACKS	0x020
+#define DEBUG_VFMT	0x040
+#define DEBUG_CODEGEN	0x080
+#define DEBUG_VERBOSE	0x100
+#define DEBUG_DRI       0x200
+#define DEBUG_DMA       0x400
+#define DEBUG_SANITY    0x800
+#define DEBUG_SYNC      0x1000
+#define DEBUG_PIXEL     0x2000
+#define DEBUG_MEMORY    0x4000
+
+
+typedef void (*radeon_tri_func) (radeonContextPtr,
+				 radeonVertex *,
+				 radeonVertex *, radeonVertex *);
+
+typedef void (*radeon_line_func) (radeonContextPtr,
+				  radeonVertex *, radeonVertex *);
+
+typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *);
+
+#define RADEON_MAX_BOS 32
+struct radeon_state {
+	struct radeon_colorbuffer_state color;
+	struct radeon_depthbuffer_state depth;
+	struct radeon_scissor_state scissor;
+	struct radeon_stencilbuffer_state stencil;
+
+	struct radeon_cs_space_check bos[RADEON_MAX_BOS];
+	int validated_bo_count;
+};
+
+/**
+ * This structure holds the command buffer while it is being constructed.
+ *
+ * The first batch of commands in the buffer is always the state that needs
+ * to be re-emitted when the context is lost. This batch can be skipped
+ * otherwise.
+ */
+struct radeon_cmdbuf {
+	struct radeon_cs_manager    *csm;
+	struct radeon_cs            *cs;
+	int size; /** # of dwords total */
+	unsigned int flushing:1; /** whether we're currently in FlushCmdBufLocked */
+};
+
+struct radeon_context {
+   GLcontext *glCtx;
+   radeonScreenPtr radeonScreen;	/* Screen private DRI data */
+
+   /* Texture object bookkeeping
+    */
+   int                   texture_depth;
+   float                 initialMaxAnisotropy;
+   uint32_t              texture_row_align;
+
+  struct radeon_dma dma;
+  struct radeon_hw_state hw;
+   /* Rasterization and vertex state:
+    */
+   GLuint TclFallback;
+   GLuint Fallback;
+   GLuint NewGLState;
+   DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
+
+   /* Drawable, cliprect and scissor information */
+   GLuint numClipRects;	/* Cliprects for the draw buffer */
+   drm_clip_rect_t *pClipRects;
+   unsigned int lastStamp;
+   GLboolean lost_context;
+   drm_radeon_sarea_t *sarea;	/* Private SAREA data */
+
+   /* Mirrors of some DRI state */
+   struct radeon_dri_mirror dri;
+
+   /* Busy waiting */
+   GLuint do_usleeps;
+   GLuint do_irqs;
+   GLuint irqsEmitted;
+   drm_radeon_irq_wait_t iw;
+
+   /* Derived state - for r300 only */
+   struct radeon_state state;
+
+   struct radeon_swtcl_info swtcl;
+   struct radeon_tcl_info tcl;
+   /* Configuration cache
+    */
+   driOptionCache optionCache;
+
+   struct radeon_cmdbuf cmdbuf;
+
+  drm_clip_rect_t fboRect;
+  GLboolean constant_cliprect; /* use for FBO or DRI2 rendering */
+  GLboolean front_cliprects;
+
+   /**
+    * Set if rendering has occured to the drawable's front buffer.
+    *
+    * This is used in the DRI2 case to detect that glFlush should also copy
+    * the contents of the fake front buffer to the real front buffer.
+    */
+   GLboolean front_buffer_dirty;
+
+   /**
+    * Track whether front-buffer rendering is currently enabled
+    *
+    * A separate flag is used to track this in order to support MRT more
+    * easily.
+    */
+   GLboolean is_front_buffer_rendering;
+
+   /**
+    * Track whether front-buffer is the current read target.
+    *
+    * This is closely associated with is_front_buffer_rendering, but may
+    * be set separately.  The DRI2 fake front buffer must be referenced
+    * either way.
+    */
+   GLboolean is_front_buffer_reading;
+
+   struct dri_metaops meta;
+
+   struct {
+	   void (*get_lock)(radeonContextPtr radeon);
+	   void (*update_viewport_offset)(GLcontext *ctx);
+	   void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa);
+	   void (*swtcl_flush)(GLcontext *ctx, uint32_t offset);
+	   void (*pre_emit_atoms)(radeonContextPtr rmesa);
+	   void (*pre_emit_state)(radeonContextPtr rmesa);
+	   void (*fallback)(GLcontext *ctx, GLuint bit, GLboolean mode);
+	   void (*free_context)(GLcontext *ctx);
+   } vtbl;
+};
+
+#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx))
+
+static inline __DRIdrawablePrivate* radeon_get_drawable(radeonContextPtr radeon)
+{
+	return radeon->dri.context->driDrawablePriv;
+}
+
+static inline __DRIdrawablePrivate* radeon_get_readable(radeonContextPtr radeon)
+{
+	return radeon->dri.context->driReadablePriv;
+}
+
+
+/**
+ * This function takes a float and packs it into a uint32_t
+ */
+static INLINE uint32_t radeonPackFloat32(float fl)
+{
+	union {
+		float fl;
+		uint32_t u;
+	} u;
+
+	u.fl = fl;
+	return u.u;
+}
+
+/* This is probably wrong for some values, I need to test this
+ * some more.  Range checking would be a good idea also..
+ *
+ * But it works for most things.  I'll fix it later if someone
+ * else with a better clue doesn't
+ */
+static INLINE uint32_t radeonPackFloat24(float f)
+{
+	float mantissa;
+	int exponent;
+	uint32_t float24 = 0;
+
+	if (f == 0.0)
+		return 0;
+
+	mantissa = frexpf(f, &exponent);
+
+	/* Handle -ve */
+	if (mantissa < 0) {
+		float24 |= (1 << 23);
+		mantissa = mantissa * -1.0;
+	}
+	/* Handle exponent, bias of 63 */
+	exponent += 62;
+	float24 |= (exponent << 16);
+	/* Kill 7 LSB of mantissa */
+	float24 |= (radeonPackFloat32(mantissa) & 0x7FFFFF) >> 7;
+
+	return float24;
+}
+
+GLboolean radeonInitContext(radeonContextPtr radeon,
+			    struct dd_function_table* functions,
+			    const __GLcontextModes * glVisual,
+			    __DRIcontextPrivate * driContextPriv,
+			    void *sharedContextPrivate);
+
+void radeonCleanupContext(radeonContextPtr radeon);
+GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
+void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable);
+GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
+			    __DRIdrawablePrivate * driDrawPriv,
+			    __DRIdrawablePrivate * driReadPriv);
+extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv);
+
+/* ================================================================
+ * Debugging:
+ */
+#define DO_DEBUG		1
+
+#if DO_DEBUG
+extern int RADEON_DEBUG;
+#else
+#define RADEON_DEBUG		0
+#endif
+
+#ifndef HAVE_LIBDRM_RADEON
+#ifndef RADEON_DEBUG_BO
+#define RADEON_DEBUG_BO 1
+#endif
+#endif
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_compat.c b/src/mesa/drivers/dri/radeon/radeon_compat.c
deleted file mode 100644
index 46b490d61f..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_compat.c
+++ /dev/null
@@ -1,301 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 ATI Technologies Inc., Ontario, Canada, and
-               Tungsten Graphics Inc., Austin, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#include "main/glheader.h"
-#include "main/imports.h"
-
-#include "radeon_context.h"
-#include "radeon_state.h"
-#include "radeon_ioctl.h"
-
-
-static struct { 
-	int start; 
-	int len; 
-	const char *name;
-} packet[RADEON_MAX_STATE_PACKETS] = {
-	{ RADEON_PP_MISC,7,"RADEON_PP_MISC" },
-	{ RADEON_PP_CNTL,3,"RADEON_PP_CNTL" },
-	{ RADEON_RB3D_COLORPITCH,1,"RADEON_RB3D_COLORPITCH" },
-	{ RADEON_RE_LINE_PATTERN,2,"RADEON_RE_LINE_PATTERN" },
-	{ RADEON_SE_LINE_WIDTH,1,"RADEON_SE_LINE_WIDTH" },
-	{ RADEON_PP_LUM_MATRIX,1,"RADEON_PP_LUM_MATRIX" },
-	{ RADEON_PP_ROT_MATRIX_0,2,"RADEON_PP_ROT_MATRIX_0" },
-	{ RADEON_RB3D_STENCILREFMASK,3,"RADEON_RB3D_STENCILREFMASK" },
-	{ RADEON_SE_VPORT_XSCALE,6,"RADEON_SE_VPORT_XSCALE" },
-	{ RADEON_SE_CNTL,2,"RADEON_SE_CNTL" },
-	{ RADEON_SE_CNTL_STATUS,1,"RADEON_SE_CNTL_STATUS" },
-	{ RADEON_RE_MISC,1,"RADEON_RE_MISC" },
-	{ RADEON_PP_TXFILTER_0,6,"RADEON_PP_TXFILTER_0" },
-	{ RADEON_PP_BORDER_COLOR_0,1,"RADEON_PP_BORDER_COLOR_0" },
-	{ RADEON_PP_TXFILTER_1,6,"RADEON_PP_TXFILTER_1" },
-	{ RADEON_PP_BORDER_COLOR_1,1,"RADEON_PP_BORDER_COLOR_1" },
-	{ RADEON_PP_TXFILTER_2,6,"RADEON_PP_TXFILTER_2" },
-	{ RADEON_PP_BORDER_COLOR_2,1,"RADEON_PP_BORDER_COLOR_2" },
-	{ RADEON_SE_ZBIAS_FACTOR,2,"RADEON_SE_ZBIAS_FACTOR" },
-	{ RADEON_SE_TCL_OUTPUT_VTX_FMT,11,"RADEON_SE_TCL_OUTPUT_VTX_FMT" },
-	{ RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED,17,"RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED" },
-};
-
-
-static void radeonCompatEmitPacket( radeonContextPtr rmesa, 
-				    struct radeon_state_atom *state )
-{
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-   drm_radeon_context_regs_t *ctx = &sarea->context_state;
-   drm_radeon_texture_regs_t *tex0 = &sarea->tex_state[0];
-   drm_radeon_texture_regs_t *tex1 = &sarea->tex_state[1];
-   int i;
-   int *buf = state->cmd;
-
-   for ( i = 0 ; i < state->cmd_size ; ) {
-      drm_radeon_cmd_header_t *header = (drm_radeon_cmd_header_t *)&buf[i++];
-
-      if (RADEON_DEBUG & DEBUG_STATE)
-	 fprintf(stderr, "%s %d: %s\n", __FUNCTION__, header->packet.packet_id,
-		 packet[(int)header->packet.packet_id].name);
-
-      switch (header->packet.packet_id) {
-      case RADEON_EMIT_PP_MISC:
-	 ctx->pp_misc = buf[i++]; 
-	 ctx->pp_fog_color = buf[i++];
-	 ctx->re_solid_color = buf[i++];
-	 ctx->rb3d_blendcntl = buf[i++];
-	 ctx->rb3d_depthoffset = buf[i++];
-	 ctx->rb3d_depthpitch = buf[i++];
-	 ctx->rb3d_zstencilcntl = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT;
-	 break;
-      case RADEON_EMIT_PP_CNTL:
-	 ctx->pp_cntl = buf[i++];
-	 ctx->rb3d_cntl = buf[i++];
-	 ctx->rb3d_coloroffset = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT;
-	 break;
-      case RADEON_EMIT_RB3D_COLORPITCH:
-	 ctx->rb3d_colorpitch = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT;
-	 break;
-      case RADEON_EMIT_RE_LINE_PATTERN:
-	 ctx->re_line_pattern = buf[i++];
-	 ctx->re_line_state = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_LINE;
-	 break;
-      case RADEON_EMIT_SE_LINE_WIDTH:
-	 ctx->se_line_width = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_LINE;
-	 break;
-      case RADEON_EMIT_PP_LUM_MATRIX:
-	 ctx->pp_lum_matrix = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_BUMPMAP;
-	 break;
-      case RADEON_EMIT_PP_ROT_MATRIX_0:
-	 ctx->pp_rot_matrix_0 = buf[i++];
-	 ctx->pp_rot_matrix_1 = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_BUMPMAP;
-	 break;
-      case RADEON_EMIT_RB3D_STENCILREFMASK:
-	 ctx->rb3d_stencilrefmask = buf[i++];
-	 ctx->rb3d_ropcntl = buf[i++];
-	 ctx->rb3d_planemask = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_MASKS;
-	 break;
-      case RADEON_EMIT_SE_VPORT_XSCALE:
-	 ctx->se_vport_xscale = buf[i++];
-	 ctx->se_vport_xoffset = buf[i++];
-	 ctx->se_vport_yscale = buf[i++];
-	 ctx->se_vport_yoffset = buf[i++];
-	 ctx->se_vport_zscale = buf[i++];
-	 ctx->se_vport_zoffset = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_VIEWPORT;
-	 break;
-      case RADEON_EMIT_SE_CNTL:
-	 ctx->se_cntl = buf[i++];
-	 ctx->se_coord_fmt = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_CONTEXT | RADEON_UPLOAD_VERTFMT;
-	 break;
-      case RADEON_EMIT_SE_CNTL_STATUS:
-	 ctx->se_cntl_status = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_SETUP;
-	 break;
-      case RADEON_EMIT_RE_MISC:
-	 ctx->re_misc = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_MISC;
-	 break;
-      case RADEON_EMIT_PP_TXFILTER_0:
-	 tex0->pp_txfilter = buf[i++];
-	 tex0->pp_txformat = buf[i++];
-	 tex0->pp_txoffset = buf[i++];
-	 tex0->pp_txcblend = buf[i++];
-	 tex0->pp_txablend = buf[i++];
-	 tex0->pp_tfactor = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX0;
-	 break;
-      case RADEON_EMIT_PP_BORDER_COLOR_0:
-	 tex0->pp_border_color = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX0;
-	 break;
-      case RADEON_EMIT_PP_TXFILTER_1:
-	 tex1->pp_txfilter = buf[i++];
-	 tex1->pp_txformat = buf[i++];
-	 tex1->pp_txoffset = buf[i++];
-	 tex1->pp_txcblend = buf[i++];
-	 tex1->pp_txablend = buf[i++];
-	 tex1->pp_tfactor = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX1;
-	 break;
-      case RADEON_EMIT_PP_BORDER_COLOR_1:
-	 tex1->pp_border_color = buf[i++];
-	 sarea->dirty |= RADEON_UPLOAD_TEX1;
-	 break;
-
-      case RADEON_EMIT_SE_ZBIAS_FACTOR:
-	 i++;
-	 i++;
-	 break;
-
-      case RADEON_EMIT_PP_TXFILTER_2:
-      case RADEON_EMIT_PP_BORDER_COLOR_2:
-      case RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT:
-      case RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED:
-      default:
-	 /* These states aren't understood by radeon drm 1.1 */
-	 fprintf(stderr, "Tried to emit unsupported state\n");
-	 return;
-      }
-   }
-}
-
-
-
-static void radeonCompatEmitStateLocked( radeonContextPtr rmesa )
-{
-   struct radeon_state_atom *atom;
-
-   if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (!rmesa->hw.is_dirty && !rmesa->hw.all_dirty)
-      return;
-
-   foreach(atom, &rmesa->hw.atomlist) {
-      if (rmesa->hw.all_dirty)
-	 atom->dirty = GL_TRUE;
-      if (atom->is_tcl)
-	 atom->dirty = GL_FALSE;
-      if (atom->dirty)
-	 radeonCompatEmitPacket(rmesa, atom);
-   }
- 
-   rmesa->hw.is_dirty = GL_FALSE;
-   rmesa->hw.all_dirty = GL_FALSE;
-}
-
-
-static void radeonCompatEmitPrimitiveLocked( radeonContextPtr rmesa,
-					     GLuint hw_primitive,
-					     GLuint nverts,
-					     drm_clip_rect_t *pbox,
-					     GLuint nbox )
-{
-   int i;
-
-   for ( i = 0 ; i < nbox ; ) {
-      int nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, nbox );
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      drm_radeon_vertex_t vtx;
-      
-      rmesa->sarea->dirty |= RADEON_UPLOAD_CLIPRECTS;
-      rmesa->sarea->nbox = nr - i;
-
-      for ( ; i < nr ; i++) 
-	 *b++ = pbox[i];
-      
-      if (RADEON_DEBUG & DEBUG_IOCTL)
-	 fprintf(stderr, 
-		 "RadeonFlushVertexBuffer: prim %x buf %d verts %d "
-		 "disc %d nbox %d\n",
-		 hw_primitive, 
-		 rmesa->dma.current.buf->buf->idx, 
-		 nverts, 
-		 nr == nbox,
-		 rmesa->sarea->nbox );
-
-      vtx.prim = hw_primitive;
-      vtx.idx = rmesa->dma.current.buf->buf->idx;
-      vtx.count = nverts;
-      vtx.discard = (nr == nbox);      
-
-      drmCommandWrite( rmesa->dri.fd, 
-		       DRM_RADEON_VERTEX,
-		       &vtx, sizeof(vtx));
-   }
-}
-
-
-
-/* No 'start' for 1.1 vertices ioctl: only one vertex prim/buffer!  
- */
-void radeonCompatEmitPrimitive( radeonContextPtr rmesa,
-				GLuint vertex_format,
-				GLuint hw_primitive,
-				GLuint nrverts )
-{
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   LOCK_HARDWARE( rmesa );
-
-   radeonCompatEmitStateLocked( rmesa );
-   rmesa->sarea->vc_format = vertex_format;
-   
-   if (rmesa->state.scissor.enabled) {
-      radeonCompatEmitPrimitiveLocked( rmesa, 
-				       hw_primitive,
-				       nrverts,
-				       rmesa->state.scissor.pClipRects,
-				       rmesa->state.scissor.numClipRects );
-   }
-   else {
-      radeonCompatEmitPrimitiveLocked( rmesa, 
-				       hw_primitive,
-				       nrverts,
-				       rmesa->pClipRects,
-				       rmesa->numClipRects );
-   }
-
-
-   UNLOCK_HARDWARE( rmesa );
-}
-
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index b67bda7d06..c457fb654e 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -53,6 +53,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "drivers/common/driverfuncs.h"
 
+#include "radeon_common.h"
 #include "radeon_context.h"
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
@@ -65,6 +66,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
+#define need_GL_EXT_framebuffer_object
 #include "extension_helper.h"
 
 #define DRIVER_DATE	"20061018"
@@ -72,40 +74,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h" /* for symbolic values of enum-type options */
-#ifndef RADEON_DEBUG
-int RADEON_DEBUG = (0);
-#endif
-
-
-/* Return various strings for glGetString().
- */
-static const GLubyte *radeonGetString( GLcontext *ctx, GLenum name )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   static char buffer[128];
-   unsigned   offset;
-   GLuint agp_mode = (rmesa->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
-      rmesa->radeonScreen->AGPMode;
-
-   switch ( name ) {
-   case GL_VENDOR:
-      return (GLubyte *)"Tungsten Graphics, Inc.";
-
-   case GL_RENDERER:
-      offset = driGetRendererString( buffer, "Radeon", DRIVER_DATE,
-				     agp_mode );
-
-      sprintf( & buffer[ offset ], " %sTCL",
-	       !(rmesa->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)
-	       ? "" : "NO-" );
-
-      return (GLubyte *)buffer;
-
-   default:
-      return NULL;
-   }
-}
-
 
 /* Extension strings exported by the R100 driver.
  */
@@ -121,6 +89,7 @@ const struct dri_extension card_extensions[] =
     { "GL_EXT_blend_logic_op",             NULL },
     { "GL_EXT_blend_subtract",             GL_EXT_blend_minmax_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_packed_depth_stencil",	   NULL},
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -137,6 +106,11 @@ const struct dri_extension card_extensions[] =
     { NULL,                                NULL }
 };
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 extern const struct tnl_pipeline_stage _radeon_render_stage;
 extern const struct tnl_pipeline_stage _radeon_tcl_stage;
 
@@ -160,15 +134,6 @@ static const struct tnl_pipeline_stage *radeon_pipeline[] = {
    NULL,
 };
 
-
-
-/* Initialize the driver's misc functions.
- */
-static void radeonInitDriverFuncs( struct dd_function_table *functions )
-{
-    functions->GetString	= radeonGetString;
-}
-
 static const struct dri_debug_control debug_control[] =
 {
     { "fall",  DEBUG_FALLBACKS },
@@ -188,19 +153,69 @@ static const struct dri_debug_control debug_control[] =
     { NULL,    0 }
 };
 
+static void r100_get_lock(radeonContextPtr radeon)
+{
+   r100ContextPtr rmesa = (r100ContextPtr)radeon;
+   drm_radeon_sarea_t *sarea = radeon->sarea;
+
+   RADEON_STATECHANGE(rmesa, ctx);
+   if (rmesa->radeon.sarea->tiling_enabled) {
+      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |=
+	 RADEON_COLOR_TILE_ENABLE;
+   } else {
+      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &=
+	 ~RADEON_COLOR_TILE_ENABLE;
+   }
+   
+   if (sarea->ctx_owner != rmesa->radeon.dri.hwContext) {
+      sarea->ctx_owner = rmesa->radeon.dri.hwContext;
+      
+      if (!radeon->radeonScreen->kernel_mm)
+         radeon_bo_legacy_texture_age(radeon->radeonScreen->bom);
+   }
+}
+
+static void r100_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
+{
+}
+
+static void r100_vtbl_pre_emit_state(radeonContextPtr radeon)
+{
+   r100ContextPtr rmesa = (r100ContextPtr)radeon;
+   
+   /* r100 always needs to emit ZBS to avoid TCL lockups */
+   rmesa->hw.zbs.dirty = 1;
+   radeon->hw.is_dirty = 1;
+}
+
+static void r100_vtbl_free_context(GLcontext *ctx)
+{
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   _mesa_vector4f_free( &rmesa->tcl.ObjClean );
+}
+
+static void r100_init_vtbl(radeonContextPtr radeon)
+{
+   radeon->vtbl.get_lock = r100_get_lock;
+   radeon->vtbl.update_viewport_offset = radeonUpdateViewportOffset;
+   radeon->vtbl.emit_cs_header = r100_vtbl_emit_cs_header;
+   radeon->vtbl.swtcl_flush = r100_swtcl_flush;
+   radeon->vtbl.pre_emit_state = r100_vtbl_pre_emit_state;
+   radeon->vtbl.fallback = radeonFallback;
+}
 
 /* Create the device specific context.
  */
 GLboolean
-radeonCreateContext( const __GLcontextModes *glVisual,
+r100CreateContext( const __GLcontextModes *glVisual,
                      __DRIcontextPrivate *driContextPriv,
                      void *sharedContextPrivate)
 {
    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
    radeonScreenPtr screen = (radeonScreenPtr)(sPriv->private);
    struct dd_function_table functions;
-   radeonContextPtr rmesa;
-   GLcontext *ctx, *shareCtx;
+   r100ContextPtr rmesa;
+   GLcontext *ctx;
    int i;
    int tcl_mode, fthrottle_mode;
 
@@ -209,10 +224,12 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    assert(screen);
 
    /* Allocate the Radeon context */
-   rmesa = (radeonContextPtr) CALLOC( sizeof(*rmesa) );
+   rmesa = (r100ContextPtr) CALLOC( sizeof(*rmesa) );
    if ( !rmesa )
       return GL_FALSE;
 
+   r100_init_vtbl(&rmesa->radeon);
+
    /* init exp fog table data */
    radeonInitStaticFogData();
    
@@ -220,12 +237,12 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     * Do this here so that initialMaxAnisotropy is set before we create
     * the default textures.
     */
-   driParseConfigFiles (&rmesa->optionCache, &screen->optionCache,
+   driParseConfigFiles (&rmesa->radeon.optionCache, &screen->optionCache,
 			screen->driScreen->myNum, "radeon");
-   rmesa->initialMaxAnisotropy = driQueryOptionf(&rmesa->optionCache,
+   rmesa->radeon.initialMaxAnisotropy = driQueryOptionf(&rmesa->radeon.optionCache,
                                                  "def_max_anisotropy");
 
-   if ( driQueryOptionb( &rmesa->optionCache, "hyperz" ) ) {
+   if ( driQueryOptionb( &rmesa->radeon.optionCache, "hyperz" ) ) {
       if ( sPriv->drm_version.minor < 13 )
 	 fprintf( stderr, "DRM version 1.%d too old to support HyperZ, "
 			  "disabling.\n", sPriv->drm_version.minor );
@@ -240,65 +257,17 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     * (the texture functions are especially important)
     */
    _mesa_init_driver_functions( &functions );
-   radeonInitDriverFuncs( &functions );
    radeonInitTextureFuncs( &functions );
 
-   /* Allocate the Mesa context */
-   if (sharedContextPrivate)
-      shareCtx = ((radeonContextPtr) sharedContextPrivate)->glCtx;
-   else
-      shareCtx = NULL;
-   rmesa->glCtx = _mesa_create_context(glVisual, shareCtx,
-                                       &functions, (void *) rmesa);
-   if (!rmesa->glCtx) {
-      FREE(rmesa);
-      return GL_FALSE;
-   }
-   driContextPriv->driverPrivate = rmesa;
-
-   /* Init radeon context data */
-   rmesa->dri.context = driContextPriv;
-   rmesa->dri.screen = sPriv;
-   rmesa->dri.drawable = NULL;
-   rmesa->dri.readable = NULL;
-   rmesa->dri.hwContext = driContextPriv->hHWContext;
-   rmesa->dri.hwLock = &sPriv->pSAREA->lock;
-   rmesa->dri.fd = sPriv->fd;
-   rmesa->dri.drmMinor = sPriv->drm_version.minor;
-
-   rmesa->radeonScreen = screen;
-   rmesa->sarea = (drm_radeon_sarea_t *)((GLubyte *)sPriv->pSAREA +
-				       screen->sarea_priv_offset);
-
-
-   rmesa->dma.buf0_address = rmesa->radeonScreen->buffers->list[0].address;
-
-   (void) memset( rmesa->texture_heaps, 0, sizeof( rmesa->texture_heaps ) );
-   make_empty_list( & rmesa->swapped );
-
-   rmesa->nr_heaps = screen->numTexHeaps;
-   for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-      rmesa->texture_heaps[i] = driCreateTextureHeap( i, rmesa,
-	    screen->texSize[i],
-	    12,
-	    RADEON_NR_TEX_REGIONS,
-	    (drmTextureRegionPtr)rmesa->sarea->tex_list[i],
-	    & rmesa->sarea->tex_age[i],
-	    & rmesa->swapped,
-	    sizeof( radeonTexObj ),
-	    (destroy_texture_object_t *) radeonDestroyTexObj );
-
-      driSetTextureSwapCounterLocation( rmesa->texture_heaps[i],
-					& rmesa->c_textureSwaps );
+   if (!radeonInitContext(&rmesa->radeon, &functions,
+			  glVisual, driContextPriv,
+			  sharedContextPrivate)) {
+     FREE(rmesa);
+     return GL_FALSE;
    }
-   rmesa->texture_depth = driQueryOptioni (&rmesa->optionCache,
-					   "texture_depth");
-   if (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
-      rmesa->texture_depth = ( screen->cpp == 4 ) ?
-	 DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16;
 
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->hw.all_dirty = GL_TRUE;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.hw.all_dirty = GL_TRUE;
 
    /* Set the maximum texture size small enough that we can guarentee that
     * all texture units can bind a maximal texture and have all of them in
@@ -306,26 +275,20 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     * setting allow larger textures.
     */
 
-   ctx = rmesa->glCtx;
-   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->optionCache,
+   ctx = rmesa->radeon.glCtx;
+   ctx->Const.MaxTextureUnits = driQueryOptioni (&rmesa->radeon.optionCache,
 						 "texture_units");
    ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
    ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
 
-   i = driQueryOptioni( &rmesa->optionCache, "allow_large_textures");
-
-   driCalculateMaxTextureLevels( rmesa->texture_heaps,
-				 rmesa->nr_heaps,
-				 & ctx->Const,
-				 4,
-				 11, /* max 2D texture size is 2048x2048 */
-				 8,  /* 256^3 */
-				 9,  /* \todo: max cube texture size seems to be 512x512(x6) */
-				 11, /* max rect texture size is 2048x2048. */
-				 12,
-				 GL_FALSE,
-				 i );
+   i = driQueryOptioni( &rmesa->radeon.optionCache, "allow_large_textures");
 
+   /* FIXME: When no memory manager is available we should set this 
+    * to some reasonable value based on texture memory pool size */
+   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.Max3DTextureLevels = 9;
+   ctx->Const.MaxCubeTextureLevels = 12;
+   ctx->Const.MaxTextureRectSize = 2048;
 
    ctx->Const.MaxTextureMaxAnisotropy = 16.0;
 
@@ -390,38 +353,39 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    }
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
-   if (rmesa->radeonScreen->drmSupportsCubeMapsR100)
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     driInitExtensions(ctx, mm_extensions, GL_FALSE);
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100)
       _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
-   if (rmesa->glCtx->Mesa_DXTn) {
+   if (rmesa->radeon.glCtx->Mesa_DXTn) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
       _mesa_enable_extension( ctx, "GL_S3_s3tc" );
    }
-   else if (driQueryOptionb (&rmesa->optionCache, "force_s3tc_enable")) {
+   else if (driQueryOptionb (&rmesa->radeon.optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
    }
 
-   if (rmesa->dri.drmMinor >= 9)
+   if (rmesa->radeon.radeonScreen->kernel_mm || rmesa->radeon.dri.drmMinor >= 9)
       _mesa_enable_extension( ctx, "GL_NV_texture_rectangle");
 
    /* XXX these should really go right after _mesa_init_driver_functions() */
+   radeon_fbo_init(&rmesa->radeon);
+   radeonInitSpanFuncs( ctx );
    radeonInitIoctlFuncs( ctx );
    radeonInitStateFuncs( ctx );
-   radeonInitSpanFuncs( ctx );
    radeonInitState( rmesa );
    radeonInitSwtcl( ctx );
 
    _mesa_vector4f_alloc( &rmesa->tcl.ObjClean, 0, 
 			 ctx->Const.MaxArrayLockSize, 32 );
 
-   fthrottle_mode = driQueryOptioni(&rmesa->optionCache, "fthrottle_mode");
-   rmesa->iw.irq_seq = -1;
-   rmesa->irqsEmitted = 0;
-   rmesa->do_irqs = (rmesa->radeonScreen->irq != 0 &&
-		     fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
-
-   rmesa->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
+   fthrottle_mode = driQueryOptioni(&rmesa->radeon.optionCache, "fthrottle_mode");
+   rmesa->radeon.iw.irq_seq = -1;
+   rmesa->radeon.irqsEmitted = 0;
+   rmesa->radeon.do_irqs = (rmesa->radeon.radeonScreen->irq != 0 &&
+			    fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS);
 
-   (*sPriv->systemTime->getUST)( & rmesa->swap_ust );
+   rmesa->radeon.do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
 
 
 #if DO_DEBUG
@@ -429,206 +393,21 @@ radeonCreateContext( const __GLcontextModes *glVisual,
 				       debug_control );
 #endif
 
-   tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-   if (driQueryOptionb(&rmesa->optionCache, "no_rast")) {
+   tcl_mode = driQueryOptioni(&rmesa->radeon.optionCache, "tcl_mode");
+   if (driQueryOptionb(&rmesa->radeon.optionCache, "no_rast")) {
       fprintf(stderr, "disabling 3D acceleration\n");
       FALLBACK(rmesa, RADEON_FALLBACK_DISABLE, 1);
    } else if (tcl_mode == DRI_CONF_TCL_SW ||
-	      !(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
-      if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-	 rmesa->radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
+	      !(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+      if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+	 rmesa->radeon.radeonScreen->chip_flags &= ~RADEON_CHIPSET_TCL;
 	 fprintf(stderr, "Disabling HW TCL support\n");
       }
-      TCL_FALLBACK(rmesa->glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
+      TCL_FALLBACK(rmesa->radeon.glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
    }
 
-   if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+   if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
 /*       _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
    }
    return GL_TRUE;
 }
-
-
-/* Destroy the device specific context.
- */
-/* Destroy the Mesa and driver specific context data.
- */
-void radeonDestroyContext( __DRIcontextPrivate *driContextPriv )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = (radeonContextPtr) driContextPriv->driverPrivate;
-   radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL;
-
-   /* check if we're deleting the currently bound context */
-   if (rmesa == current) {
-      RADEON_FIREVERTICES( rmesa );
-      _mesa_make_current(NULL, NULL, NULL);
-   }
-
-   /* Free radeon context resources */
-   assert(rmesa); /* should never be null */
-   if ( rmesa ) {
-      GLboolean   release_texture_heaps;
-
-
-      release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
-      _swsetup_DestroyContext( rmesa->glCtx );
-      _tnl_DestroyContext( rmesa->glCtx );
-      _vbo_DestroyContext( rmesa->glCtx );
-      _swrast_DestroyContext( rmesa->glCtx );
-
-      radeonDestroySwtcl( rmesa->glCtx );
-      radeonReleaseArrays( rmesa->glCtx, ~0 );
-      if (rmesa->dma.current.buf) {
-	 radeonReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-	 radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-      }
-
-      _mesa_vector4f_free( &rmesa->tcl.ObjClean );
-
-      if (rmesa->state.scissor.pClipRects) {
-	 FREE(rmesa->state.scissor.pClipRects);
-	 rmesa->state.scissor.pClipRects = NULL;
-      }
-
-      if ( release_texture_heaps ) {
-         /* This share group is about to go away, free our private
-          * texture object data.
-          */
-         int i;
-
-         for ( i = 0 ; i < rmesa->nr_heaps ; i++ ) {
-	    driDestroyTextureHeap( rmesa->texture_heaps[ i ] );
-	    rmesa->texture_heaps[ i ] = NULL;
-         }
-
-	 assert( is_empty_list( & rmesa->swapped ) );
-      }
-
-      /* free the Mesa context */
-      rmesa->glCtx->DriverCtx = NULL;
-      _mesa_destroy_context( rmesa->glCtx );
-
-      /* free the option cache */
-      driDestroyOptionCache (&rmesa->optionCache);
-
-      FREE( rmesa );
-   }
-}
-
-
-
-
-void
-radeonSwapBuffers( __DRIdrawablePrivate *dPriv )
-{
-
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      radeonContextPtr rmesa;
-      GLcontext *ctx;
-      rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = rmesa->glCtx;
-      if (ctx->Visual.doubleBufferMode) {
-         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-
-         if ( rmesa->doPageFlip ) {
-            radeonPageFlip( dPriv );
-         }
-         else {
-	     radeonCopyBuffer( dPriv, NULL );
-         }
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      _mesa_problem(NULL, "%s: drawable has no context!", __FUNCTION__);
-   }
-}
-
-void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-			 int x, int y, int w, int h )
-{
-    if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-	radeonContextPtr radeon;
-	GLcontext *ctx;
-
-	radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-	ctx = radeon->glCtx;
-
-	if (ctx->Visual.doubleBufferMode) {
-	    drm_clip_rect_t rect;
-	    rect.x1 = x + dPriv->x;
-	    rect.y1 = (dPriv->h - y - h) + dPriv->y;
-	    rect.x2 = rect.x1 + w;
-	    rect.y2 = rect.y1 + h;
-	    _mesa_notifySwapBuffers(ctx);	/* flush pending rendering comands */
-	    radeonCopyBuffer(dPriv, &rect);
-	}
-    } else {
-	/* XXX this shouldn't be an error but we can't handle it for now */
-	_mesa_problem(NULL, "%s: drawable has no context!",
-		      __FUNCTION__);
-    }
-}
-
-/* Make context `c' the current context and bind it to the given
- * drawing and reading surfaces.
- */
-GLboolean
-radeonMakeCurrent( __DRIcontextPrivate *driContextPriv,
-                   __DRIdrawablePrivate *driDrawPriv,
-                   __DRIdrawablePrivate *driReadPriv )
-{
-   if ( driContextPriv ) {
-      radeonContextPtr newCtx = 
-	 (radeonContextPtr) driContextPriv->driverPrivate;
-
-      if (RADEON_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *) newCtx->glCtx);
-
-      newCtx->dri.readable = driReadPriv;
-
-      if ( (newCtx->dri.drawable != driDrawPriv) ||
-           newCtx->lastStamp != driDrawPriv->lastStamp ) {
-	 if (driDrawPriv->swap_interval == (unsigned)-1) {
-	    driDrawPriv->vblFlags = (newCtx->radeonScreen->irq != 0)
-	       ? driGetDefaultVBlankFlags(&newCtx->optionCache)
-	       : VBLANK_FLAG_NO_IRQ;
-
-	    driDrawableInitVBlank( driDrawPriv );
-	 }
-
-	 newCtx->dri.drawable = driDrawPriv;
-
-	 radeonSetCliprects(newCtx);
-	 radeonUpdateViewportOffset( newCtx->glCtx );
-      }
-
-      _mesa_make_current( newCtx->glCtx,
-			  (GLframebuffer *) driDrawPriv->driverPrivate,
-			  (GLframebuffer *) driReadPriv->driverPrivate );
-
-      _mesa_update_state( newCtx->glCtx );
-   } else {
-      if (RADEON_DEBUG & DEBUG_DRI)
-	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
-      _mesa_make_current( NULL, NULL, NULL );
-   }
-
-   if (RADEON_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "End %s\n", __FUNCTION__);
-   return GL_TRUE;
-}
-
-/* Force the context `c' to be unbound from its buffer.
- */
-GLboolean
-radeonUnbindContext( __DRIcontextPrivate *driContextPriv )
-{
-   radeonContextPtr rmesa = (radeonContextPtr) driContextPriv->driverPrivate;
-
-   if (RADEON_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *) rmesa->glCtx);
-
-   return GL_TRUE;
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h
index 53df766f8c..1795d8bdb6 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_context.h
@@ -48,91 +48,23 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drm.h"
 #include "radeon_drm.h"
 #include "texmem.h"
-
 #include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/colormac.h"
-
-struct radeon_context;
-typedef struct radeon_context radeonContextRec;
-typedef struct radeon_context *radeonContextPtr;
-
-/* This union is used to avoid warnings/miscompilation
-   with float to uint32_t casts due to strict-aliasing */
-typedef union {
-	GLfloat f;
-	uint32_t ui32;
-} float_ui32_type;
-
-#include "radeon_lock.h"
 #include "radeon_screen.h"
-#include "main/mm.h"
-
-#include "math/m_vector.h"
-
-#define TEX_0   0x1
-#define TEX_1   0x2
-#define TEX_2   0x4
-#define TEX_ALL 0x7
-
-/* Rasterizing fallbacks */
-/* See correponding strings in r200_swtcl.c */
-#define RADEON_FALLBACK_TEXTURE		0x0001
-#define RADEON_FALLBACK_DRAW_BUFFER	0x0002
-#define RADEON_FALLBACK_STENCIL		0x0004
-#define RADEON_FALLBACK_RENDER_MODE	0x0008
-#define RADEON_FALLBACK_BLEND_EQ	0x0010
-#define RADEON_FALLBACK_BLEND_FUNC	0x0020
-#define RADEON_FALLBACK_DISABLE 	0x0040
-#define RADEON_FALLBACK_BORDER_MODE	0x0080
-
-/* The blit width for texture uploads
- */
-#define BLIT_WIDTH_BYTES 1024
 
-/* Use the templated vertex format:
- */
-#define COLOR_IS_RGBA
-#define TAG(x) radeon##x
-#include "tnl_dd/t_dd_vertex.h"
-#undef TAG
-
-typedef void (*radeon_tri_func) (radeonContextPtr,
-				 radeonVertex *,
-				 radeonVertex *, radeonVertex *);
-
-typedef void (*radeon_line_func) (radeonContextPtr,
-				  radeonVertex *, radeonVertex *);
+#include "radeon_common.h"
 
-typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *);
-
-struct radeon_colorbuffer_state {
-	GLuint clear;
-	int roundEnable;
-};
 
-struct radeon_depthbuffer_state {
-	GLuint clear;
-	GLfloat scale;
-};
+struct r100_context;
+typedef struct r100_context r100ContextRec;
+typedef struct r100_context *r100ContextPtr;
 
-struct radeon_scissor_state {
-	drm_clip_rect_t rect;
-	GLboolean enabled;
+#include "radeon_lock.h"
 
-	GLuint numClipRects;	/* Cliprects active */
-	GLuint numAllocedClipRects;	/* Cliprects available */
-	drm_clip_rect_t *pClipRects;
-};
 
-struct radeon_stencilbuffer_state {
-	GLboolean hwBuffer;
-	GLuint clear;		/* rb3d_stencilrefmask value */
-};
 
-struct radeon_stipple_state {
-	GLuint mask[32];
-};
+#define R100_TEX_ALL 0x7
 
 /* used for both tcl_vtx and vc_frmt tex bits (they are identical) */
 #define RADEON_ST_BIT(unit) \
@@ -141,42 +73,6 @@ struct radeon_stipple_state {
 #define RADEON_Q_BIT(unit) \
 (unit == 0 ? RADEON_CP_VC_FRMT_Q0 : (RADEON_CP_VC_FRMT_Q1 >> 2) << (2 * unit))
 
-typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr;
-
-/* Texture object in locally shared texture space.
- */
-struct radeon_tex_obj {
-	driTextureObject base;
-
-	GLuint bufAddr;		/* Offset to start of locally
-				   shared texture block */
-
-	GLuint dirty_state;	/* Flags (1 per texunit) for
-				   whether or not this texobj
-				   has dirty hardware state
-				   (pp_*) that needs to be
-				   brought into the
-				   texunit. */
-
-	drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS];
-	/* Six, for the cube faces */
-
-	GLboolean image_override; /* Image overridden by GLX_EXT_tfp */
-
-	GLuint pp_txfilter;	/* hardware register values */
-	GLuint pp_txformat;
-	GLuint pp_txoffset;	/* Image location in texmem.
-				   All cube faces follow. */
-	GLuint pp_txsize;	/* npot only */
-	GLuint pp_txpitch;	/* npot only */
-	GLuint pp_border_color;
-	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
-
-	GLboolean border_fallback;
-
-	GLuint tile_bits;	/* hw texture tile bits used on this texture */
-};
-
 struct radeon_texture_env_state {
 	radeonTexObjPtr texobj;
 	GLenum format;
@@ -187,17 +83,6 @@ struct radeon_texture_state {
 	struct radeon_texture_env_state unit[RADEON_MAX_TEXTURE_UNITS];
 };
 
-struct radeon_state_atom {
-	struct radeon_state_atom *next, *prev;
-	const char *name;	/* for debug */
-	int cmd_size;		/* size in bytes */
-	GLuint is_tcl;
-	int *cmd;		/* one or more cmd's */
-	int *lastcmd;		/* one or more cmd's */
-	GLboolean dirty;	/* dirty-mark in emit_state_list */
-	 GLboolean(*check) (GLcontext *);	/* is this state active? */
-};
-
 /* Trying to keep these relatively short as the variables are becoming
  * extravagently long.  Drop the driver name prefix off the front of
  * everything - I think we know which driver we're in by now, and keep the
@@ -410,10 +295,7 @@ struct radeon_state_atom {
 #define SHN_SHININESS      1
 #define SHN_STATE_SIZE     2
 
-struct radeon_hw_state {
-	/* Head of the linked list of state atoms. */
-	struct radeon_state_atom atomlist;
-
+struct r100_hw_state {
 	/* Hardware state, stored as cmdbuf commands:  
 	 *   -- Need to doublebuffer for
 	 *           - eliding noop statechange loops? (except line stipple count)
@@ -438,89 +320,19 @@ struct radeon_hw_state {
 	struct radeon_state_atom glt;
 	struct radeon_state_atom txr[3];	/* for NPOT */
 
-	int max_state_size;	/* Number of bytes necessary for a full state emit. */
-	GLboolean is_dirty, all_dirty;
 };
 
-struct radeon_state {
-	/* Derived state for internal purposes:
-	 */
-	struct radeon_colorbuffer_state color;
-	struct radeon_depthbuffer_state depth;
-	struct radeon_scissor_state scissor;
-	struct radeon_stencilbuffer_state stencil;
+
+struct r100_state {
 	struct radeon_stipple_state stipple;
 	struct radeon_texture_state texture;
 };
 
-/* Need refcounting on dma buffers:
- */
-struct radeon_dma_buffer {
-	int refcount;		/* the number of retained regions in buf */
-	drmBufPtr buf;
-};
-
-#define GET_START(rvb) (rmesa->radeonScreen->gart_buffer_offset +			\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-
-/* A retained region, eg vertices for indexed vertices.
- */
-struct radeon_dma_region {
-	struct radeon_dma_buffer *buf;
-	char *address;		/* == buf->address */
-	int start, end, ptr;	/* offsets from start of buf */
-	int aos_start;
-	int aos_stride;
-	int aos_size;
-};
-
-struct radeon_dma {
-	/* Active dma region.  Allocations for vertices and retained
-	 * regions come from here.  Also used for emitting random vertices,
-	 * these may be flushed by calling flush_current();
-	 */
-	struct radeon_dma_region current;
-
-	void (*flush) (radeonContextPtr);
-
-	char *buf0_address;	/* start of buf[0], for index calcs */
-	GLuint nr_released_bufs;	/* flush after so many buffers released */
-};
-
-struct radeon_dri_mirror {
-	__DRIcontextPrivate *context;	/* DRI context */
-	__DRIscreenPrivate *screen;	/* DRI screen */
-
-   /**
-    * DRI drawable bound to this context for drawing.
-    */
-	__DRIdrawablePrivate *drawable;
-
-   /**
-    * DRI drawable bound to this context for reading.
-    */
-	__DRIdrawablePrivate *readable;
-
-	drm_context_t hwContext;
-	drm_hw_lock_t *hwLock;
-	int fd;
-	int drmMinor;
-};
-
 #define RADEON_CMD_BUF_SZ  (8*1024)
-
-struct radeon_store {
-	GLuint statenr;
-	GLuint primnr;
-	char cmd_buf[RADEON_CMD_BUF_SZ];
-	int cmd_used;
-	int elts_start;
-};
-
+#define R200_ELT_BUF_SZ  (8*1024)
 /* radeon_tcl.c
  */
-struct radeon_tcl_info {
+struct r100_tcl_info {
 	GLuint vertex_format;
 	GLuint hw_primitive;
 
@@ -529,30 +341,18 @@ struct radeon_tcl_info {
 	 */
 	GLvector4f ObjClean;
 
-	struct radeon_dma_region *aos_components[8];
-	GLuint nr_aos_components;
-
 	GLuint *Elts;
 
-	struct radeon_dma_region indexed_verts;
-	struct radeon_dma_region obj;
-	struct radeon_dma_region rgba;
-	struct radeon_dma_region spec;
-	struct radeon_dma_region fog;
-	struct radeon_dma_region tex[RADEON_MAX_TEXTURE_UNITS];
-	struct radeon_dma_region norm;
+        int elt_cmd_offset;
+	int elt_cmd_start;
+        int elt_used;
 };
 
 /* radeon_swtcl.c
  */
-struct radeon_swtcl_info {
-	GLuint RenderIndex;
-	GLuint vertex_size;
+struct r100_swtcl_info {
 	GLuint vertex_format;
 
-	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-	GLuint vertex_attr_count;
-
 	GLubyte *verts;
 
 	/* Fallback rasterization functions
@@ -561,10 +361,6 @@ struct radeon_swtcl_info {
 	radeon_line_func draw_line;
 	radeon_tri_func draw_tri;
 
-	GLuint hw_primitive;
-	GLenum render_primitive;
-	GLuint numverts;
-
    /**
     * Offset of the 4UB color data within a hardware (swtcl) vertex.
     */
@@ -576,22 +372,9 @@ struct radeon_swtcl_info {
 	GLuint specoffset;
 
 	GLboolean needproj;
-
-	struct radeon_dma_region indexed_verts;
 };
 
-struct radeon_ioctl {
-	GLuint vertex_offset;
-	GLuint vertex_size;
-};
-
-#define RADEON_MAX_PRIMS 64
 
-struct radeon_prim {
-	GLuint start;
-	GLuint end;
-	GLuint prim;
-};
 
 /* A maximum total of 20 elements per vertex:  3 floats for position, 3
  * floats for normal, 4 floats for color, 4 bytes for secondary color,
@@ -602,59 +385,18 @@ struct radeon_prim {
  */
 #define RADEON_MAX_VERTEX_SIZE 20
 
-struct radeon_context {
-	GLcontext *glCtx;	/* Mesa context */
+struct r100_context {
+        struct radeon_context radeon;
 
 	/* Driver and hardware state management
 	 */
-	struct radeon_hw_state hw;
-	struct radeon_state state;
-
-	/* Texture object bookkeeping
-	 */
-	unsigned nr_heaps;
-	driTexHeap *texture_heaps[RADEON_NR_TEX_HEAPS];
-	driTextureObject swapped;
-	int texture_depth;
-	float initialMaxAnisotropy;
-
-	/* Rasterization and vertex state:
-	 */
-	GLuint TclFallback;
-	GLuint Fallback;
-	GLuint NewGLState;
-	 DECLARE_RENDERINPUTS(tnl_index_bitset);	/* index of bits for last tnl_install_attrs */
+	struct r100_hw_state hw;
+	struct r100_state state;
 
 	/* Vertex buffers
 	 */
 	struct radeon_ioctl ioctl;
-	struct radeon_dma dma;
 	struct radeon_store store;
-	/* A full state emit as of the first state emit in the main store, in case
-	 * the context is lost.
-	 */
-	struct radeon_store backup_store;
-
-	/* Page flipping
-	 */
-	GLuint doPageFlip;
-
-	/* Busy waiting
-	 */
-	GLuint do_usleeps;
-	GLuint do_irqs;
-	GLuint irqsEmitted;
-	drm_radeon_irq_wait_t iw;
-
-	/* Drawable, cliprect and scissor information
-	 */
-	GLuint numClipRects;	/* Cliprects for the draw buffer */
-	drm_clip_rect_t *pClipRects;
-	unsigned int lastStamp;
-	GLboolean lost_context;
-	GLboolean save_on_next_emit;
-	radeonScreenPtr radeonScreen;	/* Screen private DRI data */
-	drm_radeon_sarea_t *sarea;	/* Private SAREA data */
 
 	/* TCL stuff
 	 */
@@ -667,29 +409,13 @@ struct radeon_context {
 	GLmatrix tmpmat[RADEON_MAX_TEXTURE_UNITS];
 	GLuint last_ReallyEnabled;
 
-	/* VBI
-	 */
-	int64_t swap_ust;
-	int64_t swap_missed_ust;
-
-	GLuint swap_count;
-	GLuint swap_missed_count;
-
 	/* radeon_tcl.c
 	 */
-	struct radeon_tcl_info tcl;
+	struct r100_tcl_info tcl;
 
 	/* radeon_swtcl.c
 	 */
-	struct radeon_swtcl_info swtcl;
-
-	/* Mirrors of some DRI state
-	 */
-	struct radeon_dri_mirror dri;
-
-	/* Configuration cache
-	 */
-	driOptionCache optionCache;
+	struct r100_swtcl_info swtcl;
 
 	GLboolean using_hyperz;
 	GLboolean texmicrotile;
@@ -703,61 +429,19 @@ struct radeon_context {
 	GLuint c_textureSwaps;
 	GLuint c_textureBytes;
 	GLuint c_vertexBuffers;
+
 };
 
-#define RADEON_CONTEXT(ctx)		((radeonContextPtr)(ctx->DriverCtx))
-
-static INLINE GLuint radeonPackColor(GLuint cpp,
-                                     GLubyte r, GLubyte g,
-                                     GLubyte b, GLubyte a)
-{
-	switch (cpp) {
-	case 2:
-		return PACK_COLOR_565(r, g, b);
-	case 4:
-		return PACK_COLOR_8888(a, r, g, b);
-	default:
-		return 0;
-	}
-}
+
+#define R100_CONTEXT(ctx)		((r100ContextPtr)(ctx->DriverCtx))
+
 
 #define RADEON_OLD_PACKETS 1
 
-extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv);
-extern GLboolean radeonCreateContext(const __GLcontextModes * glVisual,
-				     __DRIcontextPrivate * driContextPriv,
-				     void *sharedContextPrivate);
-extern void radeonSwapBuffers(__DRIdrawablePrivate * dPriv);
-extern void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv,
-				int x, int y, int w, int h);
-extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv,
-				   __DRIdrawablePrivate * driDrawPriv,
-				   __DRIdrawablePrivate * driReadPriv);
-extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv);
-
-/* ================================================================
- * Debugging:
- */
-#define DO_DEBUG		1
-
-#if DO_DEBUG
-extern int RADEON_DEBUG;
-#else
-#define RADEON_DEBUG		0
-#endif
-
-#define DEBUG_TEXTURE	0x0001
-#define DEBUG_STATE	0x0002
-#define DEBUG_IOCTL	0x0004
-#define DEBUG_PRIMS	0x0008
-#define DEBUG_VERTS	0x0010
-#define DEBUG_FALLBACKS	0x0020
-#define DEBUG_VFMT	0x0040
-#define DEBUG_CODEGEN	0x0080
-#define DEBUG_VERBOSE	0x0100
-#define DEBUG_DRI       0x0200
-#define DEBUG_DMA       0x0400
-#define DEBUG_SANITY    0x0800
-#define DEBUG_SYNC      0x1000
+extern GLboolean r100CreateContext( const __GLcontextModes *glVisual,
+				    __DRIcontextPrivate *driContextPriv,
+				    void *sharedContextPrivate);
+  
+
 
 #endif				/* __RADEON_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_drm.h b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
new file mode 100644
index 0000000000..ee403d173c
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_drm.h
@@ -0,0 +1,237 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_CS_H
+#define RADEON_CS_H
+
+#include <stdint.h>
+#include <string.h>
+#include "drm.h"
+#include "radeon_drm.h"
+
+struct radeon_cs_reloc {
+    struct radeon_bo    *bo;
+    uint32_t            read_domain;
+    uint32_t            write_domain;
+    uint32_t            flags;
+};
+
+
+#define RADEON_CS_SPACE_OK 0
+#define RADEON_CS_SPACE_OP_TO_BIG 1
+#define RADEON_CS_SPACE_FLUSH 2
+
+struct radeon_cs_space_check {
+    struct radeon_bo *bo;
+    uint32_t read_domains;
+    uint32_t write_domain;
+    uint32_t new_accounted;
+};
+
+#define MAX_SPACE_BOS (32)
+
+struct radeon_cs_manager;
+
+struct radeon_cs {
+    struct radeon_cs_manager    *csm;
+    void                        *relocs;
+    uint32_t                    *packets;
+    unsigned                    crelocs;
+    unsigned                    relocs_total_size;
+    unsigned                    cdw;
+    unsigned                    ndw;
+    int                         section;
+    unsigned                    section_ndw;
+    unsigned                    section_cdw;
+    const char                  *section_file;
+    const char                  *section_func;
+    int                         section_line;
+    struct radeon_cs_space_check bos[MAX_SPACE_BOS];
+    int                         bo_count;
+    void                        (*space_flush_fn)(void *);
+    void                        *space_flush_data;
+};
+
+/* cs functions */
+struct radeon_cs_funcs {
+    struct radeon_cs *(*cs_create)(struct radeon_cs_manager *csm,
+                                   uint32_t ndw);
+    int (*cs_write_reloc)(struct radeon_cs *cs,
+                          struct radeon_bo *bo,
+                          uint32_t read_domain,
+                          uint32_t write_domain,
+                          uint32_t flags);
+    int (*cs_begin)(struct radeon_cs *cs,
+                    uint32_t ndw,
+                    const char *file,
+                    const char *func,
+                    int line);
+    int (*cs_end)(struct radeon_cs *cs,
+                  const char *file,
+                  const char *func,
+                  int line);
+    int (*cs_emit)(struct radeon_cs *cs);
+    int (*cs_destroy)(struct radeon_cs *cs);
+    int (*cs_erase)(struct radeon_cs *cs);
+    int (*cs_need_flush)(struct radeon_cs *cs);
+    void (*cs_print)(struct radeon_cs *cs, FILE *file);
+};
+
+struct radeon_cs_manager {
+    struct radeon_cs_funcs  *funcs;
+    int                     fd;
+    int32_t vram_limit, gart_limit;
+    int32_t vram_write_used, gart_write_used;
+    int32_t read_used;
+};
+
+static inline struct radeon_cs *radeon_cs_create(struct radeon_cs_manager *csm,
+                                                 uint32_t ndw)
+{
+    return csm->funcs->cs_create(csm, ndw);
+}
+
+static inline int radeon_cs_write_reloc(struct radeon_cs *cs,
+                                        struct radeon_bo *bo,
+                                        uint32_t read_domain,
+                                        uint32_t write_domain,
+                                        uint32_t flags)
+{
+    return cs->csm->funcs->cs_write_reloc(cs,
+                                          bo,
+                                          read_domain,
+                                          write_domain,
+                                          flags);
+}
+
+static inline int radeon_cs_begin(struct radeon_cs *cs,
+                                  uint32_t ndw,
+                                  const char *file,
+                                  const char *func,
+                                  int line)
+{
+    return cs->csm->funcs->cs_begin(cs, ndw, file, func, line);
+}
+
+static inline int radeon_cs_end(struct radeon_cs *cs,
+                                const char *file,
+                                const char *func,
+                                int line)
+{
+    return cs->csm->funcs->cs_end(cs, file, func, line);
+}
+
+static inline int radeon_cs_emit(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_emit(cs);
+}
+
+static inline int radeon_cs_destroy(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_destroy(cs);
+}
+
+static inline int radeon_cs_erase(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_erase(cs);
+}
+
+static inline int radeon_cs_need_flush(struct radeon_cs *cs)
+{
+    return cs->csm->funcs->cs_need_flush(cs);
+}
+
+static inline void radeon_cs_print(struct radeon_cs *cs, FILE *file)
+{
+    cs->csm->funcs->cs_print(cs, file);
+}
+
+static inline void radeon_cs_set_limit(struct radeon_cs *cs, uint32_t domain, uint32_t limit)
+{
+    
+    if (domain == RADEON_GEM_DOMAIN_VRAM)
+	cs->csm->vram_limit = limit;
+    else
+	cs->csm->gart_limit = limit;
+}
+
+static inline void radeon_cs_write_dword(struct radeon_cs *cs, uint32_t dword)
+{
+    cs->packets[cs->cdw++] = dword;
+    if (cs->section) {
+        cs->section_cdw++;
+    }
+}
+
+static inline void radeon_cs_write_qword(struct radeon_cs *cs, uint64_t qword)
+{
+
+    memcpy(cs->packets + cs->cdw, &qword, sizeof(qword));
+    cs->cdw+=2;
+    if (cs->section) {
+        cs->section_cdw+=2;
+    }
+}
+
+static inline void radeon_cs_space_set_flush(struct radeon_cs *cs, void (*fn)(void *), void *data)
+{
+    cs->space_flush_fn = fn;
+    cs->space_flush_data = data;
+}
+
+
+/*
+ * add a persistent BO to the list
+ * a persistent BO is one that will be referenced across flushes,
+ * i.e. colorbuffer, textures etc.
+ * They get reset when a new "operation" happens, where an operation
+ * is a state emission with a color/textures etc followed by a bunch of vertices.
+ */
+void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs,
+				       struct radeon_bo *bo,
+				       uint32_t read_domains,
+				       uint32_t write_domain);
+
+/* reset the persistent BO list */
+void radeon_cs_space_reset_bos(struct radeon_cs *cs);
+
+/* do a space check with the current persistent BO list */
+int radeon_cs_space_check(struct radeon_cs *cs);
+
+/* do a space check with the current persistent BO list and a temporary BO
+ * a temporary BO is like a DMA buffer, which  gets flushed with the
+ * command buffer */
+int radeon_cs_space_check_with_bo(struct radeon_cs *cs,
+				  struct radeon_bo *bo,
+				  uint32_t read_domains,
+				  uint32_t write_domain);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
new file mode 100644
index 0000000000..4f1065ebcf
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c
@@ -0,0 +1,408 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <errno.h>
+
+#include "radeon_bocs_wrapper.h"
+
+struct cs_manager_legacy {
+    struct radeon_cs_manager    base;
+    struct radeon_context       *ctx;
+    /* hack for scratch stuff */
+    uint32_t                    pending_age;
+    uint32_t                    pending_count;
+
+
+};
+
+struct cs_reloc_legacy {
+    struct radeon_cs_reloc  base;
+    uint32_t                cindices;
+    uint32_t                *indices;
+};
+
+
+static struct radeon_cs *cs_create(struct radeon_cs_manager *csm,
+                                   uint32_t ndw)
+{
+    struct radeon_cs *cs;
+
+    cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs));
+    if (cs == NULL) {
+        return NULL;
+    }
+    cs->csm = csm;
+    cs->ndw = (ndw + 0x3FF) & (~0x3FF);
+    cs->packets = (uint32_t*)malloc(4*cs->ndw);
+    if (cs->packets == NULL) {
+        free(cs);
+        return NULL;
+    }
+    cs->relocs_total_size = 0;
+    return cs;
+}
+
+static int cs_write_reloc(struct radeon_cs *cs,
+                          struct radeon_bo *bo,
+                          uint32_t read_domain,
+                          uint32_t write_domain,
+                          uint32_t flags)
+{
+    struct cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct cs_reloc_legacy *)cs->relocs;
+    /* check domains */
+    if ((read_domain && write_domain) || (!read_domain && !write_domain)) {
+        /* in one CS a bo can only be in read or write domain but not
+         * in read & write domain at the same sime
+         */
+        return -EINVAL;
+    }
+    if (read_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    if (write_domain == RADEON_GEM_DOMAIN_CPU) {
+        return -EINVAL;
+    }
+    /* check if bo is already referenced */
+    for(i = 0; i < cs->crelocs; i++) {
+        uint32_t *indices;
+
+        if (relocs[i].base.bo->handle == bo->handle) {
+            /* Check domains must be in read or write. As we check already
+             * checked that in argument one of the read or write domain was
+             * set we only need to check that if previous reloc as the read
+             * domain set then the read_domain should also be set for this
+             * new relocation.
+             */
+            if (relocs[i].base.read_domain && !read_domain) {
+                return -EINVAL;
+            }
+            if (relocs[i].base.write_domain && !write_domain) {
+                return -EINVAL;
+            }
+            relocs[i].base.read_domain |= read_domain;
+            relocs[i].base.write_domain |= write_domain;
+            /* save indice */
+            relocs[i].cindices++;
+            indices = (uint32_t*)realloc(relocs[i].indices,
+                                         relocs[i].cindices * 4);
+            if (indices == NULL) {
+                relocs[i].cindices -= 1;
+                return -ENOMEM;
+            }
+            relocs[i].indices = indices;
+            relocs[i].indices[relocs[i].cindices - 1] = cs->cdw - 1;
+            return 0;
+        }
+    }
+    /* add bo to reloc */
+    relocs = (struct cs_reloc_legacy*)
+             realloc(cs->relocs,
+                     sizeof(struct cs_reloc_legacy) * (cs->crelocs + 1));
+    if (relocs == NULL) {
+        return -ENOMEM;
+    }
+    cs->relocs = relocs;
+    relocs[cs->crelocs].base.bo = bo;
+    relocs[cs->crelocs].base.read_domain = read_domain;
+    relocs[cs->crelocs].base.write_domain = write_domain;
+    relocs[cs->crelocs].base.flags = flags;
+    relocs[cs->crelocs].indices = (uint32_t*)malloc(4);
+    if (relocs[cs->crelocs].indices == NULL) {
+        return -ENOMEM;
+    }
+    relocs[cs->crelocs].indices[0] = cs->cdw - 1;
+    relocs[cs->crelocs].cindices = 1;
+    cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo);
+    cs->crelocs++;
+    radeon_bo_ref(bo);
+    return 0;
+}
+
+static int cs_begin(struct radeon_cs *cs,
+                    uint32_t ndw,
+                    const char *file,
+                    const char *func,
+                    int line)
+{
+    if (cs->section) {
+        fprintf(stderr, "CS already in a section(%s,%s,%d)\n",
+                cs->section_file, cs->section_func, cs->section_line);
+        fprintf(stderr, "CS can't start section(%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    cs->section = 1;
+    cs->section_ndw = ndw;
+    cs->section_cdw = 0;
+    cs->section_file = file;
+    cs->section_func = func;
+    cs->section_line = line;
+
+
+    if (cs->cdw + ndw > cs->ndw) {
+        uint32_t tmp, *ptr;
+	int num = (ndw > 0x3FF) ? ndw : 0x3FF;
+
+        tmp = (cs->cdw + 1 + num) & (~num);
+        ptr = (uint32_t*)realloc(cs->packets, 4 * tmp);
+        if (ptr == NULL) {
+            return -ENOMEM;
+        }
+        cs->packets = ptr;
+        cs->ndw = tmp;
+    }
+
+    return 0;
+}
+
+static int cs_end(struct radeon_cs *cs,
+                  const char *file,
+                  const char *func,
+                  int line)
+
+{
+    if (!cs->section) {
+        fprintf(stderr, "CS no section to end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    cs->section = 0;
+    if (cs->section_ndw != cs->section_cdw) {
+        fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n",
+                cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw);
+        fprintf(stderr, "CS section end at (%s,%s,%d)\n",
+                file, func, line);
+        return -EPIPE;
+    }
+    return 0;
+}
+
+static int cs_process_relocs(struct radeon_cs *cs)
+{
+    struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm;
+    struct cs_reloc_legacy *relocs;
+    int i, j, r;
+
+    csm = (struct cs_manager_legacy*)cs->csm;
+    relocs = (struct cs_reloc_legacy *)cs->relocs;
+restart:
+    for (i = 0; i < cs->crelocs; i++) 
+    {
+        for (j = 0; j < relocs[i].cindices; j++) 
+        {
+            uint32_t soffset, eoffset;
+
+            r = radeon_bo_legacy_validate(relocs[i].base.bo,
+                                           &soffset, &eoffset);
+	        if (r == -EAGAIN)
+            {
+	             goto restart;
+            }
+            if (r) 
+            {
+                fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+                        relocs[i].base.bo, soffset, eoffset);
+                return r;
+            }
+            cs->packets[relocs[i].indices[j]] += soffset;
+            if (cs->packets[relocs[i].indices[j]] >= eoffset) 
+            {
+	      /*                radeon_bo_debug(relocs[i].base.bo, 12); */
+                fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n",
+                        relocs[i].base.bo, soffset, eoffset);
+                fprintf(stderr, "above end: %p 0x%08X 0x%08X\n",
+                        relocs[i].base.bo,
+                        cs->packets[relocs[i].indices[j]],
+                        eoffset);
+                exit(0);
+                return -EINVAL;
+            }
+        }
+    }
+    return 0;
+}
+
+static int cs_set_age(struct radeon_cs *cs)
+{
+    struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm;
+    struct cs_reloc_legacy *relocs;
+    int i;
+
+    relocs = (struct cs_reloc_legacy *)cs->relocs;
+    for (i = 0; i < cs->crelocs; i++) {
+        radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age);
+        radeon_bo_unref(relocs[i].base.bo);
+    }
+    return 0;
+}
+
+static int cs_emit(struct radeon_cs *cs)
+{
+    struct cs_manager_legacy *csm = (struct cs_manager_legacy*)cs->csm;
+    drm_radeon_cmd_buffer_t cmd;
+    drm_r300_cmd_header_t age;
+    uint64_t ull;
+    int r;
+
+    csm->ctx->vtbl.emit_cs_header(cs, csm->ctx);
+
+    /* append buffer age */
+    if ( IS_R300_CLASS(csm->ctx->radeonScreen) )
+    { 
+      age.scratch.cmd_type = R300_CMD_SCRATCH;
+      /* Scratch register 2 corresponds to what radeonGetAge polls */
+      csm->pending_age = 0;
+      csm->pending_count = 1;
+      ull = (uint64_t) (intptr_t) &csm->pending_age;
+      age.scratch.reg = 2;
+      age.scratch.n_bufs = 1;
+      age.scratch.flags = 0;
+      radeon_cs_write_dword(cs, age.u);
+      radeon_cs_write_qword(cs, ull);
+      radeon_cs_write_dword(cs, 0);
+    }
+
+    r = cs_process_relocs(cs);
+    if (r) {
+        return 0;
+    }
+
+    cmd.buf = (char *)cs->packets;
+    cmd.bufsz = cs->cdw * 4;
+    if (csm->ctx->state.scissor.enabled) {
+        cmd.nbox = csm->ctx->state.scissor.numClipRects;
+        cmd.boxes = (drm_clip_rect_t *) csm->ctx->state.scissor.pClipRects;
+    } else {
+        cmd.nbox = csm->ctx->numClipRects;
+        cmd.boxes = (drm_clip_rect_t *) csm->ctx->pClipRects;
+    }
+
+    //dump_cmdbuf(cs);
+
+    r = drmCommandWrite(cs->csm->fd, DRM_RADEON_CMDBUF, &cmd, sizeof(cmd));
+    if (r) {
+        return r;
+    }
+    if ((!IS_R300_CLASS(csm->ctx->radeonScreen)) &&
+        (!IS_R600_CLASS(csm->ctx->radeonScreen))) { /* +r6/r7 : No irq for r6/r7 yet. */
+	drm_radeon_irq_emit_t emit_cmd;
+	emit_cmd.irq_seq = &csm->pending_age;
+	r = drmCommandWrite(cs->csm->fd, DRM_RADEON_IRQ_EMIT, &emit_cmd, sizeof(emit_cmd));
+	if (r) {
+		return r;
+	}
+    }
+    cs_set_age(cs);
+
+    cs->csm->read_used = 0;
+    cs->csm->vram_write_used = 0;
+    cs->csm->gart_write_used = 0;
+    return 0;
+}
+
+static void inline cs_free_reloc(void *relocs_p, int crelocs)
+{
+    struct cs_reloc_legacy *relocs = relocs_p;
+    int i;
+    if (!relocs_p)
+      return;
+    for (i = 0; i < crelocs; i++)
+      free(relocs[i].indices);
+}
+
+static int cs_destroy(struct radeon_cs *cs)
+{
+    cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    free(cs->packets);
+    free(cs);
+    return 0;
+}
+
+static int cs_erase(struct radeon_cs *cs)
+{
+    cs_free_reloc(cs->relocs, cs->crelocs);
+    free(cs->relocs);
+    cs->relocs_total_size = 0;
+    cs->relocs = NULL;
+    cs->crelocs = 0;
+    cs->cdw = 0;
+    cs->section = 0;
+    return 0;
+}
+
+static int cs_need_flush(struct radeon_cs *cs)
+{
+    /* this function used to flush when the BO usage got to
+     * a certain size, now the higher levels handle this better */
+    return 0;
+}
+
+static void cs_print(struct radeon_cs *cs, FILE *file)
+{
+}
+
+static struct radeon_cs_funcs  radeon_cs_legacy_funcs = {
+    cs_create,
+    cs_write_reloc,
+    cs_begin,
+    cs_end,
+    cs_emit,
+    cs_destroy,
+    cs_erase,
+    cs_need_flush,
+    cs_print,
+};
+
+struct radeon_cs_manager *radeon_cs_manager_legacy_ctor(struct radeon_context *ctx)
+{
+    struct cs_manager_legacy *csm;
+
+    csm = (struct cs_manager_legacy*)
+          calloc(1, sizeof(struct cs_manager_legacy));
+    if (csm == NULL) {
+        return NULL;
+    }
+    csm->base.funcs = &radeon_cs_legacy_funcs;
+    csm->base.fd = ctx->dri.fd;
+    csm->ctx = ctx;
+    csm->pending_age = 1;
+    return (struct radeon_cs_manager*)csm;
+}
+
+void radeon_cs_manager_legacy_dtor(struct radeon_cs_manager *csm)
+{
+    free(csm);
+}
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h
new file mode 100644
index 0000000000..e177b4bafe
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h
@@ -0,0 +1,40 @@
+/* 
+ * Copyright © 2008 Nicolai Haehnle
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Aapo Tahkola <aet@rasterburn.org>
+ *      Nicolai Haehnle <prefect_@gmx.net>
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_CS_LEGACY_H
+#define RADEON_CS_LEGACY_H
+
+#include "radeon_common.h"
+
+struct radeon_cs_manager *radeon_cs_manager_legacy_ctor(struct radeon_context *ctx);
+void radeon_cs_manager_legacy_dtor(struct radeon_cs_manager *csm);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c
new file mode 100644
index 0000000000..5a8df7bb8c
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_cs_space_drm.c
@@ -0,0 +1,234 @@
+/* 
+ * Copyright © 2009 Red Hat Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ */
+#include <assert.h>
+#include <errno.h>
+#include <stdlib.h>
+#include "radeon_bocs_wrapper.h"
+
+struct rad_sizes {
+    int32_t op_read;
+    int32_t op_gart_write;
+    int32_t op_vram_write;
+};
+
+static inline int radeon_cs_setup_bo(struct radeon_cs_space_check *sc, struct rad_sizes *sizes)
+{
+    uint32_t read_domains, write_domain;
+    struct radeon_bo *bo;
+
+    bo = sc->bo;
+    sc->new_accounted = 0;
+    read_domains = sc->read_domains;
+    write_domain = sc->write_domain;
+
+    /* legacy needs a static check */
+    if (radeon_bo_is_static(bo)) {
+	bo->space_accounted = sc->new_accounted = (read_domains << 16) | write_domain;
+	return 0;
+    }
+
+    /* already accounted this bo */
+    if (write_domain && (write_domain == bo->space_accounted)) {
+	sc->new_accounted = bo->space_accounted;
+	return 0;
+    }
+    if (read_domains && ((read_domains << 16) == bo->space_accounted)) {
+	sc->new_accounted = bo->space_accounted;
+	return 0;
+    }
+
+    if (bo->space_accounted == 0) {
+	if (write_domain == RADEON_GEM_DOMAIN_VRAM)
+	    sizes->op_vram_write += bo->size;
+	else if (write_domain == RADEON_GEM_DOMAIN_GTT)
+	  sizes->op_gart_write += bo->size;
+	else
+	    sizes->op_read += bo->size;
+	sc->new_accounted = (read_domains << 16) | write_domain;
+    } else {
+	uint16_t old_read, old_write;
+	
+	old_read = bo->space_accounted >> 16;
+	old_write = bo->space_accounted & 0xffff;
+	
+	if (write_domain && (old_read & write_domain)) {
+	    sc->new_accounted = write_domain;
+	    /* moving from read to a write domain */
+	    if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
+		sizes->op_read -= bo->size;
+		sizes->op_vram_write += bo->size;
+	    } else if (write_domain == RADEON_GEM_DOMAIN_VRAM) {
+		sizes->op_read -= bo->size;
+		sizes->op_gart_write += bo->size;
+	    }
+	} else if (read_domains & old_write) {
+	    sc->new_accounted = bo->space_accounted & 0xffff;
+	} else {
+	    /* rewrite the domains */
+	    if (write_domain != old_write)
+		fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write);
+	    if (read_domains != old_read)
+		fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read);
+	    return RADEON_CS_SPACE_FLUSH;
+	}
+    }
+    return 0;
+}
+
+static int radeon_cs_do_space_check(struct radeon_cs *cs, struct radeon_cs_space_check *new_tmp)
+{
+    struct radeon_cs_manager *csm = cs->csm;
+    int i;
+    struct radeon_bo *bo;
+    struct rad_sizes sizes;
+    int ret;
+
+    /* check the totals for this operation */
+
+    if (cs->bo_count == 0 && !new_tmp)
+	return 0;
+
+    memset(&sizes, 0, sizeof(struct rad_sizes));
+
+    /* prepare */
+    for (i = 0; i < cs->bo_count; i++) {
+	ret = radeon_cs_setup_bo(&cs->bos[i], &sizes);
+	if (ret)
+	    return ret;
+    }
+
+    if (new_tmp) {
+	ret = radeon_cs_setup_bo(new_tmp, &sizes);
+	if (ret)
+	    return ret;
+    }
+	
+    if (sizes.op_read < 0)
+	    sizes.op_read = 0;
+
+    /* check sizes - operation first */
+    if ((sizes.op_read + sizes.op_gart_write > csm->gart_limit) ||
+	(sizes.op_vram_write > csm->vram_limit)) {
+	    return RADEON_CS_SPACE_OP_TO_BIG;
+    }
+    
+    if (((csm->vram_write_used + sizes.op_vram_write) > csm->vram_limit) ||
+	((csm->read_used + csm->gart_write_used + sizes.op_gart_write + sizes.op_read) > csm->gart_limit)) {
+	    return RADEON_CS_SPACE_FLUSH;
+    }
+    
+    csm->gart_write_used += sizes.op_gart_write;
+    csm->vram_write_used += sizes.op_vram_write;
+    csm->read_used += sizes.op_read;
+    /* commit */
+    for (i = 0; i < cs->bo_count; i++) {
+	    bo = cs->bos[i].bo;
+	    bo->space_accounted = cs->bos[i].new_accounted;
+    }
+    if (new_tmp)
+	new_tmp->bo->space_accounted = new_tmp->new_accounted;
+    
+    return RADEON_CS_SPACE_OK;
+}
+
+void radeon_cs_space_add_persistent_bo(struct radeon_cs *cs, struct radeon_bo *bo, uint32_t read_domains, uint32_t write_domain)
+{
+    int i;
+    for (i = 0; i < cs->bo_count; i++) {
+	if (cs->bos[i].bo == bo &&
+	    cs->bos[i].read_domains == read_domains &&
+	    cs->bos[i].write_domain == write_domain)
+	    return;
+    }
+    radeon_bo_ref(bo);
+    i = cs->bo_count;
+    cs->bos[i].bo = bo;
+    cs->bos[i].read_domains = read_domains;
+    cs->bos[i].write_domain = write_domain;
+    cs->bos[i].new_accounted = 0;
+    cs->bo_count++;
+
+    assert(cs->bo_count < MAX_SPACE_BOS);
+}
+
+static int radeon_cs_check_space_internal(struct radeon_cs *cs, struct radeon_cs_space_check *tmp_bo)
+{
+    int ret;
+    int flushed = 0;
+
+again:
+    ret = radeon_cs_do_space_check(cs, tmp_bo);
+    if (ret == RADEON_CS_SPACE_OP_TO_BIG)
+	return -1;
+    if (ret == RADEON_CS_SPACE_FLUSH) {
+	(*cs->space_flush_fn)(cs->space_flush_data);
+	if (flushed)
+	    return -1;
+	flushed = 1;
+	goto again;
+    }
+    return 0;
+}
+
+int radeon_cs_space_check_with_bo(struct radeon_cs *cs,
+				  struct radeon_bo *bo,
+				  uint32_t read_domains, uint32_t write_domain)
+{									
+    struct radeon_cs_space_check temp_bo;
+    int ret = 0;
+
+    if (bo) {
+	temp_bo.bo = bo;
+	temp_bo.read_domains = read_domains;
+	temp_bo.write_domain = write_domain;
+	temp_bo.new_accounted = 0;
+    }
+
+    ret = radeon_cs_check_space_internal(cs, bo ? &temp_bo : NULL);
+    return ret;
+}
+
+int radeon_cs_space_check(struct radeon_cs *cs)
+{
+    return radeon_cs_check_space_internal(cs, NULL);
+}
+
+void radeon_cs_space_reset_bos(struct radeon_cs *cs)
+{
+    int i;
+    for (i = 0; i < cs->bo_count; i++) {
+	radeon_bo_unref(cs->bos[i].bo);
+	cs->bos[i].bo = NULL;
+	cs->bos[i].read_domains = 0;
+	cs->bos[i].write_domain = 0;
+	cs->bos[i].new_accounted = 0;
+    }
+    cs->bo_count = 0;
+}
+
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
new file mode 100644
index 0000000000..48114a0012
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -0,0 +1,347 @@
+/**************************************************************************
+
+Copyright (C) 2004 Nicolai Haehnle.
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+#include "radeon_common.h"
+
+#if defined(USE_X86_ASM)
+#define COPY_DWORDS( dst, src, nr )					\
+do {									\
+	int __tmp;							\
+	__asm__ __volatile__( "rep ; movsl"				\
+			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
+			      : "0" (nr),				\
+			        "D" ((long)dst),			\
+			        "S" ((long)src) );			\
+} while (0)
+#else
+#define COPY_DWORDS( dst, src, nr )		\
+do {						\
+   int j;					\
+   for ( j = 0 ; j < nr ; j++ )			\
+      dst[j] = ((int *)src)[j];			\
+   dst += nr;					\
+} while (0)
+#endif
+
+static void radeonEmitVec4(uint32_t *out, GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & DEBUG_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 4)
+		COPY_DWORDS(out, data, count);
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out++;
+			data += stride;
+		}
+}
+
+void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & DEBUG_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 8)
+		COPY_DWORDS(out, data, count * 2);
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out[1] = *(int *)(data + 4);
+			out += 2;
+			data += stride;
+		}
+}
+
+void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & DEBUG_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 12) {
+		COPY_DWORDS(out, data, count * 3);
+    }
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out[1] = *(int *)(data + 4);
+			out[2] = *(int *)(data + 8);
+			out += 3;
+			data += stride;
+		}
+}
+
+static void radeonEmitVec16(uint32_t *out, GLvoid * data, int stride, int count)
+{
+	int i;
+
+	if (RADEON_DEBUG & DEBUG_VERTS)
+		fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+			__FUNCTION__, count, stride, (void *)out, (void *)data);
+
+	if (stride == 16)
+		COPY_DWORDS(out, data, count * 4);
+	else
+		for (i = 0; i < count; i++) {
+			out[0] = *(int *)data;
+			out[1] = *(int *)(data + 4);
+			out[2] = *(int *)(data + 8);
+			out[3] = *(int *)(data + 12);
+			out += 4;
+			data += stride;
+		}
+}
+
+void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos,
+			 GLvoid * data, int size, int stride, int count)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	uint32_t *out;
+
+	if (stride == 0) {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		count = 1;
+		aos->stride = 0;
+	} else {
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32);
+		aos->stride = size;
+	}
+
+	aos->components = size;
+	aos->count = count;
+
+	out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
+	switch (size) {
+	case 1: radeonEmitVec4(out, data, stride, count); break;
+	case 2: radeonEmitVec8(out, data, stride, count); break;
+	case 3: radeonEmitVec12(out, data, stride, count); break;
+	case 4: radeonEmitVec16(out, data, stride, count); break;
+	default:
+		assert(0);
+		break;
+	}
+}
+
+void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size)
+{
+
+	size = MAX2(size, MAX_DMA_BUF_SZ);
+
+	if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA))
+		fprintf(stderr, "%s\n", __FUNCTION__);
+
+	if (rmesa->dma.flush) {
+		rmesa->dma.flush(rmesa->glCtx);
+	}
+
+	if (rmesa->dma.nr_released_bufs > 4) {
+		rcommonFlushCmdBuf(rmesa, __FUNCTION__);
+		rmesa->dma.nr_released_bufs = 0;
+	}
+
+	if (rmesa->dma.current) {
+		radeon_bo_unmap(rmesa->dma.current);
+		radeon_bo_unref(rmesa->dma.current);
+		rmesa->dma.current = 0;
+	}
+
+again_alloc:
+#ifdef RADEON_DEBUG_BO
+    rmesa->dma.current = radeon_bo_open(rmesa->radeonScreen->bom,
+					    0, size, 4, RADEON_GEM_DOMAIN_GTT,
+					    0, "dma.current");
+#else    
+	rmesa->dma.current = radeon_bo_open(rmesa->radeonScreen->bom,
+					    0, size, 4, RADEON_GEM_DOMAIN_GTT,
+					    0);
+#endif /* RADEON_DEBUG_BO */
+
+	if (!rmesa->dma.current) {
+		rcommonFlushCmdBuf(rmesa, __FUNCTION__);
+		rmesa->dma.nr_released_bufs = 0;
+		goto again_alloc;
+	}
+
+	rmesa->dma.current_used = 0;
+	rmesa->dma.current_vertexptr = 0;
+	
+	if (radeon_cs_space_check_with_bo(rmesa->cmdbuf.cs,
+					  rmesa->dma.current,
+					  RADEON_GEM_DOMAIN_GTT, 0))
+		fprintf(stderr,"failure to revalidate BOs - badness\n");
+
+	if (!rmesa->dma.current) {
+        /* Cmd buff have been flushed in radeon_revalidate_bos */
+		rmesa->dma.nr_released_bufs = 0;
+		goto again_alloc;
+	}
+
+	radeon_bo_map(rmesa->dma.current, 1);
+}
+
+/* Allocates a region from rmesa->dma.current.  If there isn't enough
+ * space in current, grab a new buffer (and discard what was left of current)
+ */
+void radeonAllocDmaRegion(radeonContextPtr rmesa,
+			  struct radeon_bo **pbo, int *poffset,
+			  int bytes, int alignment)
+{
+	if (RADEON_DEBUG & DEBUG_IOCTL)
+		fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
+
+	if (rmesa->dma.flush)
+		rmesa->dma.flush(rmesa->glCtx);
+
+	assert(rmesa->dma.current_used == rmesa->dma.current_vertexptr);
+
+	alignment--;
+	rmesa->dma.current_used = (rmesa->dma.current_used + alignment) & ~alignment;
+
+	if (!rmesa->dma.current || rmesa->dma.current_used + bytes > rmesa->dma.current->size)
+		radeonRefillCurrentDmaRegion(rmesa, (bytes + 15) & ~15);
+
+	*poffset = rmesa->dma.current_used;
+	*pbo = rmesa->dma.current;
+	radeon_bo_ref(*pbo);
+
+	/* Always align to at least 16 bytes */
+	rmesa->dma.current_used = (rmesa->dma.current_used + bytes + 15) & ~15;
+	rmesa->dma.current_vertexptr = rmesa->dma.current_used;
+
+	assert(rmesa->dma.current_used <= rmesa->dma.current->size);
+}
+
+void radeonReleaseDmaRegion(radeonContextPtr rmesa)
+{
+	if (RADEON_DEBUG & DEBUG_IOCTL)
+		fprintf(stderr, "%s %p\n", __FUNCTION__, rmesa->dma.current);
+	if (rmesa->dma.current) {
+		rmesa->dma.nr_released_bufs++;
+		radeon_bo_unmap(rmesa->dma.current);
+	        radeon_bo_unref(rmesa->dma.current);
+	}
+	rmesa->dma.current = NULL;
+}
+
+
+/* Flush vertices in the current dma region.
+ */
+void rcommon_flush_last_swtcl_prim( GLcontext *ctx  )
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	struct radeon_dma *dma = &rmesa->dma;
+		
+
+	if (RADEON_DEBUG & DEBUG_IOCTL)
+		fprintf(stderr, "%s %p\n", __FUNCTION__, dma->current);
+	dma->flush = NULL;
+
+	if (dma->current) {
+	    GLuint current_offset = dma->current_used;
+
+	    assert (dma->current_used +
+		    rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
+		    dma->current_vertexptr);
+
+	    if (dma->current_used != dma->current_vertexptr) {
+		    dma->current_used = dma->current_vertexptr;
+
+		    rmesa->vtbl.swtcl_flush(ctx, current_offset);
+	    }
+	    rmesa->swtcl.numverts = 0;
+	}
+}
+/* Alloc space in the current dma region.
+ */
+void *
+rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
+{
+	GLuint bytes = vsize * nverts;
+	void *head;
+restart:
+	if (!rmesa->dma.current || rmesa->dma.current_vertexptr + bytes > rmesa->dma.current->size) {
+                radeonRefillCurrentDmaRegion(rmesa, bytes);
+	}
+
+        if (!rmesa->dma.flush) {
+		/* make sure we have enough space to use this in cmdbuf */
+   		rcommonEnsureCmdBufSpace(rmesa,
+			      rmesa->hw.max_state_size + (20*sizeof(int)),
+			      __FUNCTION__);
+		/* if cmdbuf flushed DMA restart */
+		if (!rmesa->dma.current)
+			goto restart;
+                rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
+                rmesa->dma.flush = rcommon_flush_last_swtcl_prim;
+        }
+
+	ASSERT( vsize == rmesa->swtcl.vertex_size * 4 );
+        ASSERT( rmesa->dma.flush == rcommon_flush_last_swtcl_prim );
+        ASSERT( rmesa->dma.current_used +
+                rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
+                rmesa->dma.current_vertexptr );
+
+	head = (rmesa->dma.current->ptr + rmesa->dma.current_vertexptr);
+	rmesa->dma.current_vertexptr += bytes;
+	rmesa->swtcl.numverts += nverts;
+	return head;
+}
+
+void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs )
+{
+   radeonContextPtr radeon = RADEON_CONTEXT( ctx );
+   int i;
+
+   if (radeon->dma.flush) {
+       radeon->dma.flush(radeon->glCtx);
+   }
+   if (radeon->tcl.elt_dma_bo) {
+	   radeon_bo_unref(radeon->tcl.elt_dma_bo);
+	   radeon->tcl.elt_dma_bo = NULL;
+   }
+   for (i = 0; i < radeon->tcl.aos_count; i++) {
+      if (radeon->tcl.aos[i].bo) {
+         radeon_bo_unref(radeon->tcl.aos[i].bo);
+         radeon->tcl.aos[i].bo = NULL;
+      }
+   }
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.h b/src/mesa/drivers/dri/radeon/radeon_dma.h
new file mode 100644
index 0000000000..06e388fc1d
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.h
@@ -0,0 +1,52 @@
+/**************************************************************************
+
+Copyright (C) 2004 Nicolai Haehnle.
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+#ifndef RADEON_DMA_H
+#define RADEON_DMA_H
+
+void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count);
+void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count);
+
+void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos,
+			 GLvoid * data, int size, int stride, int count);
+
+void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size);
+void radeonAllocDmaRegion(radeonContextPtr rmesa,
+			  struct radeon_bo **pbo, int *poffset,
+			  int bytes, int alignment);
+void radeonReleaseDmaRegion(radeonContextPtr rmesa);
+
+void rcommon_flush_last_swtcl_prim(GLcontext *ctx);
+
+void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize);
+void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs );
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c
new file mode 100644
index 0000000000..f28efa33e9
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c
@@ -0,0 +1,615 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Red Hat Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/mtypes.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/context.h"
+#include "main/texformat.h"
+#include "main/texrender.h"
+
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
+
+#define FILE_DEBUG_FLAG DEBUG_TEXTURE
+#define DBG(...) do {                                           \
+        if (RADEON_DEBUG & FILE_DEBUG_FLAG)                      \
+                _mesa_printf(__VA_ARGS__);                      \
+} while(0)
+
+static struct gl_framebuffer *
+radeon_new_framebuffer(GLcontext *ctx, GLuint name)
+{
+  return _mesa_new_framebuffer(ctx, name);
+}
+
+static void
+radeon_delete_renderbuffer(struct gl_renderbuffer *rb)
+{
+  struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+
+  ASSERT(rrb);
+
+  if (rrb && rrb->bo) {
+    radeon_bo_unref(rrb->bo);
+  }
+  _mesa_free(rrb);
+}
+
+static void *
+radeon_get_pointer(GLcontext *ctx, struct gl_renderbuffer *rb,
+		   GLint x, GLint y)
+{
+  return NULL;
+}
+
+/**
+ * Called via glRenderbufferStorageEXT() to set the format and allocate
+ * storage for a user-created renderbuffer.
+ */
+static GLboolean
+radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                                 GLenum internalFormat,
+                                 GLuint width, GLuint height)
+{
+  struct radeon_context *radeon = RADEON_CONTEXT(ctx);
+  struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+  GLboolean software_buffer = GL_FALSE;
+  int cpp;
+
+   ASSERT(rb->Name != 0);
+  switch (internalFormat) {
+   case GL_R3_G3_B2:
+   case GL_RGB4:
+   case GL_RGB5:
+      rb->_ActualFormat = GL_RGB5;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 5;
+      rb->GreenBits = 6;
+      rb->BlueBits = 5;
+      cpp = 2;
+      break;
+   case GL_RGB:
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      rb->_ActualFormat = GL_RGB8;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 8;
+      rb->GreenBits = 8;
+      rb->BlueBits = 8;
+      rb->AlphaBits = 0;
+      cpp = 4;
+      break;
+   case GL_RGBA:
+   case GL_RGBA2:
+   case GL_RGBA4:
+   case GL_RGB5_A1:
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+   case GL_RGBA16:
+      rb->_ActualFormat = GL_RGBA8;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 8;
+      rb->GreenBits = 8;
+      rb->BlueBits = 8;
+      rb->AlphaBits = 8;
+      cpp = 4;
+      break;
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+      /* alloc a depth+stencil buffer */
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->StencilBits = 8;
+      cpp = 4;
+      break;
+   case GL_DEPTH_COMPONENT16:
+      rb->_ActualFormat = GL_DEPTH_COMPONENT16;
+      rb->DataType = GL_UNSIGNED_SHORT;
+      rb->DepthBits = 16;
+      cpp = 2;
+      break;
+   case GL_DEPTH_COMPONENT:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH_COMPONENT32:
+      rb->_ActualFormat = GL_DEPTH_COMPONENT24;
+      rb->DataType = GL_UNSIGNED_INT;
+      rb->DepthBits = 24;
+      cpp = 4;
+      break;
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->DepthBits = 24;
+      rb->StencilBits = 8;
+      cpp = 4;
+      break;
+   default:
+      _mesa_problem(ctx,
+                    "Unexpected format in intel_alloc_renderbuffer_storage");
+      return GL_FALSE;
+   }
+
+  if (ctx->Driver.Flush)
+	  ctx->Driver.Flush(ctx); /* +r6/r7 */
+
+  if (rrb->bo)
+    radeon_bo_unref(rrb->bo);
+  
+    
+   if (software_buffer) {
+      return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat,
+                                             width, height);
+   }
+   else {
+     uint32_t size = width * height * cpp;
+     uint32_t pitch = ((cpp * width + 63) & ~63) / cpp;
+
+     fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width,
+	  height, pitch);
+
+     rrb->pitch = pitch * cpp;
+     rrb->cpp = cpp;
+#ifdef RADEON_DEBUG_BO
+     rrb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+			      0,
+			      size,
+			      0,
+			      RADEON_GEM_DOMAIN_VRAM,
+			      0,
+                  "Radeon RBO");
+#else
+     rrb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+			      0,
+			      size,
+			      0,
+			      RADEON_GEM_DOMAIN_VRAM,
+			      0);
+#endif /* RADEON_DEBUG_BO */
+     rb->Width = width;
+     rb->Height = height;
+       return GL_TRUE;
+   }    
+   
+}
+
+
+/**
+ * Called for each hardware renderbuffer when a _window_ is resized.
+ * Just update fields.
+ * Not used for user-created renderbuffers!
+ */
+static GLboolean
+radeon_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                           GLenum internalFormat, GLuint width, GLuint height)
+{
+   ASSERT(rb->Name == 0);
+   rb->Width = width;
+   rb->Height = height;
+   rb->_ActualFormat = internalFormat;
+
+   return GL_TRUE;
+}
+
+
+static void
+radeon_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
+		     GLuint width, GLuint height)
+{
+     struct radeon_framebuffer *radeon_fb = (struct radeon_framebuffer*)fb;
+   int i;
+
+   _mesa_resize_framebuffer(ctx, fb, width, height);
+
+   fb->Initialized = GL_TRUE; /* XXX remove someday */
+
+   if (fb->Name != 0) {
+      return;
+   }
+
+   /* Make sure all window system renderbuffers are up to date */
+   for (i = 0; i < 2; i++) {
+      struct gl_renderbuffer *rb = &radeon_fb->color_rb[i]->base;
+
+      /* only resize if size is changing */
+      if (rb && (rb->Width != width || rb->Height != height)) {
+	 rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height);
+      }
+   }
+}
+
+
+/** Dummy function for gl_renderbuffer::AllocStorage() */
+static GLboolean
+radeon_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+			 GLenum internalFormat, GLuint width, GLuint height)
+{
+   _mesa_problem(ctx, "radeon_op_alloc_storage should never be called.");
+   return GL_FALSE;
+}
+
+struct radeon_renderbuffer *
+radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv)
+{
+    struct radeon_renderbuffer *rrb;
+
+    rrb = CALLOC_STRUCT(radeon_renderbuffer);
+    if (!rrb)
+	return NULL;
+
+    _mesa_init_renderbuffer(&rrb->base, 0);
+    rrb->base.ClassID = RADEON_RB_CLASS;
+
+    /* XXX format junk */
+    switch (format) {
+	case GL_RGB5:
+	    rrb->base._ActualFormat = GL_RGB5;
+	    rrb->base._BaseFormat = GL_RGBA;
+	    rrb->base.RedBits = 5;
+	    rrb->base.GreenBits = 6;
+	    rrb->base.BlueBits = 5;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_RGB8:
+	    rrb->base._ActualFormat = GL_RGB8;
+	    rrb->base._BaseFormat = GL_RGB;
+	    rrb->base.RedBits = 8;
+	    rrb->base.GreenBits = 8;
+	    rrb->base.BlueBits = 8;
+	    rrb->base.AlphaBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_RGBA8:
+	    rrb->base._ActualFormat = GL_RGBA8;
+	    rrb->base._BaseFormat = GL_RGBA;
+	    rrb->base.RedBits = 8;
+	    rrb->base.GreenBits = 8;
+	    rrb->base.BlueBits = 8;
+	    rrb->base.AlphaBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_STENCIL_INDEX8_EXT:
+	    rrb->base._ActualFormat = GL_STENCIL_INDEX8_EXT;
+	    rrb->base._BaseFormat = GL_STENCIL_INDEX;
+	    rrb->base.StencilBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_DEPTH_COMPONENT16:
+	    rrb->base._ActualFormat = GL_DEPTH_COMPONENT16;
+	    rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+	    rrb->base.DepthBits = 16;
+	    rrb->base.DataType = GL_UNSIGNED_SHORT;
+	    break;
+	case GL_DEPTH_COMPONENT24:
+	    rrb->base._ActualFormat = GL_DEPTH_COMPONENT24;
+	    rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+	    rrb->base.DepthBits = 24;
+	    rrb->base.DataType = GL_UNSIGNED_INT;
+	    break;
+	case GL_DEPTH24_STENCIL8_EXT:
+	    rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+	    rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+	    rrb->base.DepthBits = 24;
+	    rrb->base.StencilBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+	    break;
+	default:
+	    fprintf(stderr, "%s: Unknown format 0x%04x\n", __FUNCTION__, format);
+	    _mesa_delete_renderbuffer(&rrb->base);
+	    return NULL;
+    }
+
+    rrb->dPriv = driDrawPriv;
+    rrb->base.InternalFormat = format;
+
+    rrb->base.Delete = radeon_delete_renderbuffer;
+    rrb->base.AllocStorage = radeon_alloc_window_storage;
+    rrb->base.GetPointer = radeon_get_pointer;
+
+    rrb->bo = NULL;
+    return rrb;
+}
+
+static struct gl_renderbuffer *
+radeon_new_renderbuffer(GLcontext * ctx, GLuint name)
+{
+  struct radeon_renderbuffer *rrb;
+
+  rrb = CALLOC_STRUCT(radeon_renderbuffer);
+  if (!rrb)
+    return NULL;
+
+  _mesa_init_renderbuffer(&rrb->base, name);
+  rrb->base.ClassID = RADEON_RB_CLASS;
+
+  rrb->base.Delete = radeon_delete_renderbuffer;
+  rrb->base.AllocStorage = radeon_alloc_renderbuffer_storage;
+  rrb->base.GetPointer = radeon_get_pointer;
+
+  return &rrb->base;
+}
+
+static void
+radeon_bind_framebuffer(GLcontext * ctx, GLenum target,
+                       struct gl_framebuffer *fb, struct gl_framebuffer *fbread)
+{
+   if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) {
+      radeon_draw_buffer(ctx, fb);
+   }
+   else {
+      /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */
+   }
+}
+
+static void
+radeon_framebuffer_renderbuffer(GLcontext * ctx,
+                               struct gl_framebuffer *fb,
+                               GLenum attachment, struct gl_renderbuffer *rb)
+{
+
+	if (ctx->Driver.Flush)
+		ctx->Driver.Flush(ctx); /* +r6/r7 */
+
+   _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb);
+   radeon_draw_buffer(ctx, fb);
+}
+
+
+static GLboolean
+radeon_update_wrapper(GLcontext *ctx, struct radeon_renderbuffer *rrb, 
+		     struct gl_texture_image *texImage)
+{
+	int retry = 0;
+restart:
+	if (texImage->TexFormat == &_mesa_texformat_argb8888) {
+		rrb->cpp = 4;
+		rrb->base._ActualFormat = GL_RGBA8;
+		rrb->base._BaseFormat = GL_RGBA;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to RGBA8 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_rgb565) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_RGB5;
+		rrb->base._BaseFormat = GL_RGB;
+		rrb->base.DataType = GL_UNSIGNED_SHORT;
+		DBG("Render to RGB5 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_argb1555) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_RGB5_A1;
+		rrb->base._BaseFormat = GL_RGBA;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to ARGB1555 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_argb4444) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_RGBA4;
+		rrb->base._BaseFormat = GL_RGBA;
+		rrb->base.DataType = GL_UNSIGNED_BYTE;
+		DBG("Render to ARGB1555 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_z16) {
+		rrb->cpp = 2;
+		rrb->base._ActualFormat = GL_DEPTH_COMPONENT16;
+		rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+		rrb->base.DataType = GL_UNSIGNED_SHORT;
+		DBG("Render to DEPTH16 texture OK\n");
+	}
+	else if (texImage->TexFormat == &_mesa_texformat_s8_z24) {
+		rrb->cpp = 4;
+		rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+		rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+		rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+		DBG("Render to DEPTH_STENCIL texture OK\n");
+	}
+	else {
+		/* try redoing the FBO */
+		if (retry == 1) {
+			DBG("Render to texture BAD FORMAT %d\n",
+			    texImage->TexFormat->MesaFormat);
+			return GL_FALSE;
+		}
+		texImage->TexFormat = radeonChooseTextureFormat(ctx, texImage->InternalFormat, 0,
+								texImage->TexFormat->DataType,
+								1);
+
+		retry++;
+		goto restart;
+	}
+	
+	rrb->pitch = texImage->Width * rrb->cpp;
+	rrb->base.InternalFormat = rrb->base._ActualFormat;
+	rrb->base.Width = texImage->Width;
+	rrb->base.Height = texImage->Height;
+	rrb->base.RedBits = texImage->TexFormat->RedBits;
+	rrb->base.GreenBits = texImage->TexFormat->GreenBits;
+	rrb->base.BlueBits = texImage->TexFormat->BlueBits;
+	rrb->base.AlphaBits = texImage->TexFormat->AlphaBits;
+	rrb->base.DepthBits = texImage->TexFormat->DepthBits;
+	rrb->base.StencilBits = texImage->TexFormat->StencilBits;
+	
+	rrb->base.Delete = radeon_delete_renderbuffer;
+	rrb->base.AllocStorage = radeon_nop_alloc_storage;
+	
+	return GL_TRUE;
+}
+
+
+static struct radeon_renderbuffer *
+radeon_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage)
+{
+  const GLuint name = ~0;   /* not significant, but distinct for debugging */
+  struct radeon_renderbuffer *rrb;
+
+   /* make an radeon_renderbuffer to wrap the texture image */
+   rrb = CALLOC_STRUCT(radeon_renderbuffer);
+   if (!rrb) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture");
+      return NULL;
+   }
+
+   _mesa_init_renderbuffer(&rrb->base, name);
+   rrb->base.ClassID = RADEON_RB_CLASS;
+
+   if (!radeon_update_wrapper(ctx, rrb, texImage)) {
+      _mesa_free(rrb);
+      return NULL;
+   }
+
+   return rrb;
+  
+}
+static void
+radeon_render_texture(GLcontext * ctx,
+                     struct gl_framebuffer *fb,
+                     struct gl_renderbuffer_attachment *att)
+{
+   struct gl_texture_image *newImage
+      = att->Texture->Image[att->CubeMapFace][att->TextureLevel];
+   struct radeon_renderbuffer *rrb = radeon_renderbuffer(att->Renderbuffer);
+   radeon_texture_image *radeon_image;
+   GLuint imageOffset;
+
+   (void) fb;
+
+   ASSERT(newImage);
+
+   if (newImage->Border != 0) {
+      /* Fallback on drawing to a texture with a border, which won't have a
+       * miptree.
+       */
+      _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+      _mesa_render_texture(ctx, fb, att);
+      return;
+   }
+   else if (!rrb) {
+      rrb = radeon_wrap_texture(ctx, newImage);
+      if (rrb) {
+         /* bind the wrapper to the attachment point */
+         _mesa_reference_renderbuffer(&att->Renderbuffer, &rrb->base);
+      }
+      else {
+         /* fallback to software rendering */
+         _mesa_render_texture(ctx, fb, att);
+         return;
+      }
+   }
+
+   if (!radeon_update_wrapper(ctx, rrb, newImage)) {
+       _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+       _mesa_render_texture(ctx, fb, att);
+       return;
+   }
+
+   DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n",
+       _glthread_GetID(),
+       att->Texture->Name, newImage->Width, newImage->Height,
+       rrb->base.RefCount);
+
+   /* point the renderbufer's region to the texture image region */
+   radeon_image = (radeon_texture_image *)newImage;
+   if (rrb->bo != radeon_image->mt->bo) {
+      if (rrb->bo)
+  	radeon_bo_unref(rrb->bo);
+      rrb->bo = radeon_image->mt->bo;
+      radeon_bo_ref(rrb->bo);
+   }
+
+   /* compute offset of the particular 2D image within the texture region */
+   imageOffset = radeon_miptree_image_offset(radeon_image->mt,
+                                            att->CubeMapFace,
+                                            att->TextureLevel);
+
+   if (att->Texture->Target == GL_TEXTURE_3D) {
+      GLuint offsets[6];
+      radeon_miptree_depth_offsets(radeon_image->mt, att->TextureLevel,
+				   offsets);
+      imageOffset += offsets[att->Zoffset];
+   }
+
+   /* store that offset in the region */
+   rrb->draw_offset = imageOffset;
+
+   /* update drawing region, etc */
+   radeon_draw_buffer(ctx, fb);
+}
+
+static void
+radeon_finish_render_texture(GLcontext * ctx,
+                            struct gl_renderbuffer_attachment *att)
+{
+
+}
+static void
+radeon_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+}
+
+static void
+radeon_blit_framebuffer(GLcontext *ctx,
+                       GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                       GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                       GLbitfield mask, GLenum filter)
+{
+}
+
+void radeon_fbo_init(struct radeon_context *radeon)
+{
+  radeon->glCtx->Driver.NewFramebuffer = radeon_new_framebuffer;
+  radeon->glCtx->Driver.NewRenderbuffer = radeon_new_renderbuffer;
+  radeon->glCtx->Driver.BindFramebuffer = radeon_bind_framebuffer;
+  radeon->glCtx->Driver.FramebufferRenderbuffer = radeon_framebuffer_renderbuffer;
+  radeon->glCtx->Driver.RenderTexture = radeon_render_texture;
+  radeon->glCtx->Driver.FinishRenderTexture = radeon_finish_render_texture;
+  radeon->glCtx->Driver.ResizeBuffers = radeon_resize_buffers;
+  radeon->glCtx->Driver.ValidateFramebuffer = radeon_validate_framebuffer;
+  radeon->glCtx->Driver.BlitFramebuffer = radeon_blit_framebuffer;
+}
+
+  
+void radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
+				struct radeon_bo *bo)
+{
+  struct radeon_bo *old;
+  old = rb->bo;
+  rb->bo = bo;
+  radeon_bo_ref(bo);
+  if (old)
+    radeon_bo_unref(old);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
index 09acf6b4f8..a5e4df7941 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
@@ -35,7 +35,21 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
 #include <sched.h>
-#include <errno.h> 
+#include <errno.h>
+
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/shaders.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+#include "main/stencil.h"
+#include "main/matrix.h"
 
 #include "main/glheader.h"
 #include "main/imports.h"
@@ -43,6 +57,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 
 #include "radeon_context.h"
+#include "radeon_common.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
 #include "radeon_tcl.h"
@@ -58,75 +73,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define RADEON_IDLE_RETRY           16
 
 
-static void radeonWaitForIdle( radeonContextPtr rmesa );
-static int radeonFlushCmdBufLocked( radeonContextPtr rmesa, 
-				    const char * caller );
-
-static void print_state_atom( struct radeon_state_atom *state )
-{
-   int i;
-
-   fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size);
-
-   if (RADEON_DEBUG & DEBUG_VERBOSE) 
-      for (i = 0 ; i < state->cmd_size ; i++) 
-	 fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]);
-
-}
-
-static void radeonSaveHwState( radeonContextPtr rmesa )
-{
-   struct radeon_state_atom *atom;
-   char * dest = rmesa->backup_store.cmd_buf;
-
-   if (RADEON_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-   
-   rmesa->backup_store.cmd_used = 0;
-
-   foreach( atom, &rmesa->hw.atomlist ) {
-      if ( atom->check( rmesa->glCtx ) ) {
-	 int size = atom->cmd_size * 4;
-	 memcpy( dest, atom->cmd, size);
-	 dest += size;
-	 rmesa->backup_store.cmd_used += size;
-	 if (RADEON_DEBUG & DEBUG_STATE)
-	    print_state_atom( atom );
-      }
-   }
-
-   assert( rmesa->backup_store.cmd_used <= RADEON_CMD_BUF_SZ );
-   if (RADEON_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Returning to radeonEmitState\n");
-}
-
-/* At this point we were in FlushCmdBufLocked but we had lost our context, so
- * we need to unwire our current cmdbuf, hook the one with the saved state in
- * it, flush it, and then put the current one back.  This is so commands at the
- * start of a cmdbuf can rely on the state being kept from the previous one.
- */
-static void radeonBackUpAndEmitLostStateLocked( radeonContextPtr rmesa )
-{
-   GLuint nr_released_bufs;
-   struct radeon_store saved_store;
-
-   if (rmesa->backup_store.cmd_used == 0)
-      return;
-
-   if (RADEON_DEBUG & DEBUG_STATE)
-      fprintf(stderr, "Emitting backup state on lost context\n");
-
-   rmesa->lost_context = GL_FALSE;
-
-   nr_released_bufs = rmesa->dma.nr_released_bufs;
-   saved_store = rmesa->store;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->store = rmesa->backup_store;
-   radeonFlushCmdBufLocked( rmesa, __FUNCTION__ );
-   rmesa->dma.nr_released_bufs = nr_released_bufs;
-   rmesa->store = saved_store;
-}
-
 /* =============================================================
  * Kernel command buffer handling
  */
@@ -134,965 +80,381 @@ static void radeonBackUpAndEmitLostStateLocked( radeonContextPtr rmesa )
 /* The state atoms will be emitted in the order they appear in the atom list,
  * so this step is important.
  */
-void radeonSetUpAtomList( radeonContextPtr rmesa )
+void radeonSetUpAtomList( r100ContextPtr rmesa )
 {
-   int i, mtu = rmesa->glCtx->Const.MaxTextureUnits;
-
-   make_empty_list(&rmesa->hw.atomlist);
-   rmesa->hw.atomlist.name = "atom-list";
-
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.ctx);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.set);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.lin);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.msk);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.vpt);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tcl);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.msc);
+   int i, mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits;
+
+   make_empty_list(&rmesa->radeon.hw.atomlist);
+   rmesa->radeon.hw.atomlist.name = "atom-list";
+
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.ctx);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.set);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.lin);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.msk);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.vpt);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.tcl);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.msc);
    for (i = 0; i < mtu; ++i) {
-       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tex[i]);
-       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.txr[i]);
-       insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.cube[i]);
+       insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i]);
+       insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.txr[i]);
+       insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i]);
    }
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.zbs);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mtl);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.zbs);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.mtl);
    for (i = 0; i < 3 + mtu; ++i)
-      insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mat[i]);
+      insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i]);
    for (i = 0; i < 8; ++i)
-      insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.lit[i]);
+      insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i]);
    for (i = 0; i < 6; ++i)
-      insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.ucp[i]);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.eye);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.grd);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.fog);
-   insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.glt);
+      insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i]);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.eye);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.grd);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.fog);
+   insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.glt);
 }
 
-void radeonEmitState( radeonContextPtr rmesa )
+void radeonEmitScissor(r100ContextPtr rmesa)
 {
-   struct radeon_state_atom *atom;
-   char *dest;
-
-   if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->save_on_next_emit) {
-      radeonSaveHwState(rmesa);
-      rmesa->save_on_next_emit = GL_FALSE;
-   }
-
-   /* this code used to return here but now it emits zbs */
-
-   /* To avoid going across the entire set of states multiple times, just check
-    * for enough space for the case of emitting all state, and inline the
-    * radeonAllocCmdBuf code here without all the checks.
-    */
-   radeonEnsureCmdBufSpace(rmesa, rmesa->hw.max_state_size);
-   dest = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-
-   /* We always always emit zbs, this is due to a bug found by keithw in
-      the hardware and rediscovered after Erics changes by me.
-      if you ever touch this code make sure you emit zbs otherwise
-      you get tcl lockups on at least M7/7500 class of chips - airlied */
-   rmesa->hw.zbs.dirty=1;
-
-   if (RADEON_DEBUG & DEBUG_STATE) {
-      foreach(atom, &rmesa->hw.atomlist) {
-	 if (atom->dirty || rmesa->hw.all_dirty) {
-	    if (atom->check(rmesa->glCtx))
-	       print_state_atom(atom);
-	    else
-	       fprintf(stderr, "skip state %s\n", atom->name);
-	 }
-      }
-   }
-
-   foreach(atom, &rmesa->hw.atomlist) {
-      if (rmesa->hw.all_dirty)
-	 atom->dirty = GL_TRUE;
-      if (!(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) &&
-	   atom->is_tcl)
-	 atom->dirty = GL_FALSE;
-      if (atom->dirty) {
-	 if (atom->check(rmesa->glCtx)) {
-	    int size = atom->cmd_size * 4;
-	    memcpy(dest, atom->cmd, size);
-	    dest += size;
-	    rmesa->store.cmd_used += size;
-	    atom->dirty = GL_FALSE;
-	 }
-      }
-   }
-
-   assert(rmesa->store.cmd_used <= RADEON_CMD_BUF_SZ);
- 
-   rmesa->hw.is_dirty = GL_FALSE;
-   rmesa->hw.all_dirty = GL_FALSE;
+    BATCH_LOCALS(&rmesa->radeon);
+    if (!rmesa->radeon.radeonScreen->kernel_mm) {
+       return;
+    }
+    if (rmesa->radeon.state.scissor.enabled) {
+        BEGIN_BATCH(6);
+        OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 0));
+        OUT_BATCH(rmesa->hw.ctx.cmd[CTX_PP_CNTL] | RADEON_SCISSOR_ENABLE);
+        OUT_BATCH(CP_PACKET0(RADEON_RE_TOP_LEFT, 0));
+        OUT_BATCH((rmesa->radeon.state.scissor.rect.y1 << 16) |
+                  rmesa->radeon.state.scissor.rect.x1);
+        OUT_BATCH(CP_PACKET0(RADEON_RE_WIDTH_HEIGHT, 0));
+        OUT_BATCH(((rmesa->radeon.state.scissor.rect.y2 - 1) << 16) |
+                  (rmesa->radeon.state.scissor.rect.x2 - 1));
+        END_BATCH();
+    } else {
+        BEGIN_BATCH(2);
+        OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 0));
+        OUT_BATCH(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & ~RADEON_SCISSOR_ENABLE);
+        END_BATCH();
+    }
 }
 
 /* Fire a section of the retained (indexed_verts) buffer as a regular
- * primtive.  
+ * primtive.
  */
-extern void radeonEmitVbufPrim( radeonContextPtr rmesa,
+extern void radeonEmitVbufPrim( r100ContextPtr rmesa,
 				GLuint vertex_format,
 				GLuint primitive,
 				GLuint vertex_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
-
+   BATCH_LOCALS(&rmesa->radeon);
 
    assert(!(primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
-   
-   radeonEmitState( rmesa );
 
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s cmd_used/4: %d\n", __FUNCTION__,
-	      rmesa->store.cmd_used/4);
-   
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, VBUF_BUFSZ,
-						       __FUNCTION__ );
+   radeonEmitState(&rmesa->radeon);
+   radeonEmitScissor(rmesa);
+
 #if RADEON_OLD_PACKETS
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM | (3 << 16);
-   cmd[2].i = rmesa->ioctl.vertex_offset;
-   cmd[3].i = vertex_nr;
-   cmd[4].i = vertex_format;
-   cmd[5].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
-	       (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
+   BEGIN_BATCH(8);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM, 3);
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+     OUT_BATCH_RELOC(rmesa->ioctl.vertex_offset, rmesa->ioctl.bo, rmesa->ioctl.vertex_offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   } else {
+     OUT_BATCH(rmesa->ioctl.vertex_offset);
+   }
 
-   if (RADEON_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x offt 0x%x vfmt 0x%x vfcntl %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, cmd[2].i, cmd[4].i, cmd[5].i);
-#else
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_DRAW_VBUF | (1 << 16);
-   cmd[2].i = vertex_format;
-   cmd[3].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_MAOS_ENABLE |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
-	       (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
+   OUT_BATCH(vertex_nr);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |  RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
+	     (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
+
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			   rmesa->ioctl.bo,
+			   RADEON_GEM_DOMAIN_GTT,
+			   0, 0);
+   }
 
+   END_BATCH();
 
-   if (RADEON_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x vfmt 0x%x vfcntl %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, cmd[2].i, cmd[3].i);
+#else
+   BEGIN_BATCH(4);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_DRAW_VBUF, 1);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |
+	     RADEON_CP_VC_CNTL_PRIM_WALK_LIST |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_MAOS_ENABLE |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE |
+	     (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT));
+   END_BATCH();
 #endif
 }
 
-
-void radeonFlushElts( radeonContextPtr rmesa )
+void radeonFlushElts( GLcontext *ctx )
 {
-   int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start);
-   int dwords;
-#if RADEON_OLD_PACKETS
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 24)) / 2;
-#else
-   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 16)) / 2;
-#endif
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&rmesa->radeon);
+   int nr;
+   uint32_t *cmd = (uint32_t *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_start);
+   int dwords = (rmesa->radeon.cmdbuf.cs->section_ndw - rmesa->radeon.cmdbuf.cs->section_cdw);
 
    if (RADEON_DEBUG & DEBUG_IOCTL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-   assert( rmesa->dma.flush == radeonFlushElts );
-   rmesa->dma.flush = NULL;
+   assert( rmesa->radeon.dma.flush == radeonFlushElts );
+   rmesa->radeon.dma.flush = NULL;
 
-   /* Cope with odd number of elts:
-    */
-   rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2;
-   dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4;
+   nr = rmesa->tcl.elt_used;
+
+#if RADEON_OLD_PACKETS
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     dwords -= 2;
+   }
+#endif
 
 #if RADEON_OLD_PACKETS
-   cmd[1] |= (dwords - 3) << 16;
+   cmd[1] |= (dwords + 3) << 16;
    cmd[5] |= nr << RADEON_CP_VC_CNTL_NUM_SHIFT;
 #else
-   cmd[1] |= (dwords - 3) << 16;
+   cmd[1] |= (dwords + 2) << 16;
    cmd[3] |= nr << RADEON_CP_VC_CNTL_NUM_SHIFT;
 #endif
 
+   rmesa->radeon.cmdbuf.cs->cdw += dwords;
+   rmesa->radeon.cmdbuf.cs->section_cdw += dwords;
+
+#if RADEON_OLD_PACKETS
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			    rmesa->ioctl.bo,
+			    RADEON_GEM_DOMAIN_GTT,
+			    0, 0);
+   }
+#endif
+
+   END_BATCH();
+
    if (RADEON_DEBUG & DEBUG_SYNC) {
       fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
-      radeonFinish( rmesa->glCtx );
+      radeonFinish( rmesa->radeon.glCtx );
    }
-}
 
+}
 
-GLushort *radeonAllocEltsOpenEnded( radeonContextPtr rmesa,
+GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa,
 				    GLuint vertex_format,
 				    GLuint primitive,
 				    GLuint min_nr )
 {
-   drm_radeon_cmd_header_t *cmd;
    GLushort *retval;
+   int align_min_nr;
+   BATCH_LOCALS(&rmesa->radeon);
 
    if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, min_nr);
+      fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
 
    assert((primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
-   
-   radeonEmitState( rmesa );
-   
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa,
-						       ELTS_BUFSZ(min_nr),
-						       __FUNCTION__ );
+
+   radeonEmitState(&rmesa->radeon);
+   radeonEmitScissor(rmesa);
+
+   rmesa->tcl.elt_cmd_start = rmesa->radeon.cmdbuf.cs->cdw;
+
+   /* round up min_nr to align the state */
+   align_min_nr = (min_nr + 1) & ~1;
+
 #if RADEON_OLD_PACKETS
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM;
-   cmd[2].i = rmesa->ioctl.vertex_offset;
-   cmd[3].i = 0xffff;
-   cmd[4].i = vertex_format;
-   cmd[5].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_IND |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
-
-   retval = (GLushort *)(cmd+6);
-#else   
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
-   cmd[1].i = RADEON_CP_PACKET3_3D_DRAW_INDX;
-   cmd[2].i = vertex_format;
-   cmd[3].i = (primitive | 
-	       RADEON_CP_VC_CNTL_PRIM_WALK_IND |
-	       RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
-	       RADEON_CP_VC_CNTL_MAOS_ENABLE |
-	       RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
-
-   retval = (GLushort *)(cmd+4);
+   BEGIN_BATCH_NO_AUTOSTATE(2+ELTS_BUFSZ(align_min_nr)/4);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM, 0);
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+     OUT_BATCH_RELOC(rmesa->ioctl.vertex_offset, rmesa->ioctl.bo, rmesa->ioctl.vertex_offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   } else {
+     OUT_BATCH(rmesa->ioctl.vertex_offset);
+   }
+   OUT_BATCH(0xffff);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |
+	     RADEON_CP_VC_CNTL_PRIM_WALK_IND |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
+#else
+   BEGIN_BATCH_NO_AUTOSTATE(ELTS_BUFSZ(align_min_nr)/4);
+   OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_DRAW_INDX, 0);
+   OUT_BATCH(vertex_format);
+   OUT_BATCH(primitive |
+	     RADEON_CP_VC_CNTL_PRIM_WALK_IND |
+	     RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA |
+	     RADEON_CP_VC_CNTL_MAOS_ENABLE |
+	     RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE);
 #endif
 
-   if (RADEON_DEBUG & DEBUG_PRIMS)
-      fprintf(stderr, "%s: header 0x%x vfmt 0x%x prim %x \n",
-	      __FUNCTION__,
-	      cmd[1].i, vertex_format, primitive);
 
-   assert(!rmesa->dma.flush);
-   rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   rmesa->dma.flush = radeonFlushElts;
+   rmesa->tcl.elt_cmd_offset = rmesa->radeon.cmdbuf.cs->cdw;
+   rmesa->tcl.elt_used = min_nr;
 
-   rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf;
+   retval = (GLushort *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_offset);
 
-   return retval;
-}
+   if (RADEON_DEBUG & DEBUG_PRIMS)
+      fprintf(stderr, "%s: header prim %x \n",
+	      __FUNCTION__, primitive);
 
+   assert(!rmesa->radeon.dma.flush);
+   rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
+   rmesa->radeon.dma.flush = radeonFlushElts;
 
+   return retval;
+}
 
-void radeonEmitVertexAOS( radeonContextPtr rmesa,
+void radeonEmitVertexAOS( r100ContextPtr rmesa,
 			  GLuint vertex_size,
+			  struct radeon_bo *bo,
 			  GLuint offset )
 {
 #if RADEON_OLD_PACKETS
-   rmesa->ioctl.vertex_size = vertex_size;
    rmesa->ioctl.vertex_offset = offset;
+   rmesa->ioctl.bo = bo;
 #else
-   drm_radeon_cmd_header_t *cmd;
+   BATCH_LOCALS(&rmesa->radeon);
 
    if (RADEON_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL))
       fprintf(stderr, "%s:  vertex_size 0x%x offset 0x%x \n",
 	      __FUNCTION__, vertex_size, offset);
 
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, VERT_AOS_BUFSZ,
-						  __FUNCTION__ );
+   BEGIN_BATCH(7);
+   OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, 2);
+   OUT_BATCH(1);
+   OUT_BATCH(vertex_size | (vertex_size << 8));
+   OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+   END_BATCH();
 
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = RADEON_CP_PACKET3_3D_LOAD_VBPNTR | (2 << 16);
-   cmd[2].i = 1;
-   cmd[3].i = vertex_size | (vertex_size << 8);
-   cmd[4].i = offset;
 #endif
 }
-		       
 
-void radeonEmitAOS( radeonContextPtr rmesa,
-		    struct radeon_dma_region **component,
+
+void radeonEmitAOS( r100ContextPtr rmesa,
 		    GLuint nr,
 		    GLuint offset )
 {
 #if RADEON_OLD_PACKETS
    assert( nr == 1 );
-   assert( component[0]->aos_size == component[0]->aos_stride );
-   rmesa->ioctl.vertex_size = component[0]->aos_size;
-   rmesa->ioctl.vertex_offset = 
-      (component[0]->aos_start + offset * component[0]->aos_stride * 4);
+   rmesa->ioctl.bo = rmesa->radeon.tcl.aos[0].bo;
+   rmesa->ioctl.vertex_offset =
+     (rmesa->radeon.tcl.aos[0].offset + offset * rmesa->radeon.tcl.aos[0].stride * 4);
 #else
-   drm_radeon_cmd_header_t *cmd;
-   int sz = AOS_BUFSZ(nr);
+   BATCH_LOCALS(&rmesa->radeon);
+   uint32_t voffset;
+   //   int sz = AOS_BUFSZ(nr);
+   int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
    int i;
-   int *tmp;
 
    if (RADEON_DEBUG & DEBUG_IOCTL)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, sz,
-						  __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = RADEON_CP_PACKET3_3D_LOAD_VBPNTR | (((sz / sizeof(int))-3) << 16);
-   cmd[2].i = nr;
-   tmp = &cmd[0].i;
-   cmd += 3;
-
-   for (i = 0 ; i < nr ; i++) {
-      if (i & 1) {
-	 cmd[0].i |= ((component[i]->aos_stride << 24) | 
-		      (component[i]->aos_size << 16));
-	 cmd[2].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-	 cmd += 3;
+   BEGIN_BATCH(sz+2+(nr * 2));
+   OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, sz - 1);
+   OUT_BATCH(nr);
+
+   if (!rmesa->radeon.radeonScreen->kernel_mm) {
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[i+1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
       }
-      else {
-	 cmd[0].i = ((component[i]->aos_stride << 8) | 
-		     (component[i]->aos_size << 0));
-	 cmd[1].i = (component[i]->aos_start + 
-		     offset * component[i]->aos_stride * 4);
-      }
-   }
 
-   if (RADEON_DEBUG & DEBUG_VERTS) {
-      fprintf(stderr, "%s:\n", __FUNCTION__);
-      for (i = 0 ; i < sz ; i++)
-	 fprintf(stderr, "   %d: %x\n", i, tmp[i]);
-   }
-#endif
-}
-
-/* using already shifted color_fmt! */
-void radeonEmitBlit( radeonContextPtr rmesa, /* FIXME: which drmMinor is required? */
-		   GLuint color_fmt,
-		   GLuint src_pitch,
-		   GLuint src_offset,
-		   GLuint dst_pitch,
-		   GLuint dst_offset,
-		   GLint srcx, GLint srcy,
-		   GLint dstx, GLint dsty,
-		   GLuint w, GLuint h )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
-	      __FUNCTION__, 
-	      src_pitch, src_offset, srcx, srcy,
-	      dst_pitch, dst_offset, dstx, dsty,
-	      w, h);
-
-   assert( (src_pitch & 63) == 0 );
-   assert( (dst_pitch & 63) == 0 );
-   assert( (src_offset & 1023) == 0 ); 
-   assert( (dst_offset & 1023) == 0 ); 
-   assert( w < (1<<16) );
-   assert( h < (1<<16) );
-
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, 8 * sizeof(int),
-						  __FUNCTION__ );
-
-
-   cmd[0].i = 0;
-   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
-   cmd[1].i = RADEON_CP_PACKET3_CNTL_BITBLT_MULTI | (5 << 16);
-   cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_DST_PITCH_OFFSET_CNTL |
-	       RADEON_GMC_BRUSH_NONE |
-	       color_fmt |
-	       RADEON_GMC_SRC_DATATYPE_COLOR |
-	       RADEON_ROP3_S |
-	       RADEON_DP_SRC_SOURCE_MEMORY |
-	       RADEON_GMC_CLR_CMP_CNTL_DIS |
-	       RADEON_GMC_WR_MSK_DIS );
-
-   cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10);
-   cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10);
-   cmd[5].i = (srcx << 16) | srcy;
-   cmd[6].i = (dstx << 16) | dsty; /* dst */
-   cmd[7].i = (w << 16) | h;
-}
-
-
-void radeonEmitWait( radeonContextPtr rmesa, GLuint flags )
-{
-   drm_radeon_cmd_header_t *cmd;
-
-   assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) );
-
-   cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, 1 * sizeof(int),
-					   __FUNCTION__ );
-   cmd[0].i = 0;
-   cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
-   cmd[0].wait.flags = flags;
-}
-
-
-static int radeonFlushCmdBufLocked( radeonContextPtr rmesa, 
-				    const char * caller )
-{
-   int ret, i;
-   drm_radeon_cmd_buffer_t cmd;
-
-   if (rmesa->lost_context)
-      radeonBackUpAndEmitLostStateLocked(rmesa);
-
-   if (RADEON_DEBUG & DEBUG_IOCTL) {
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-
-      if (RADEON_DEBUG & DEBUG_VERBOSE) 
-	 for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 )
-	    fprintf(stderr, "%d: %x\n", i/4, 
-		    *(int *)(&rmesa->store.cmd_buf[i]));
-   }
-
-   if (RADEON_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__,
-	      rmesa->dma.nr_released_bufs);
-
-
-   if (RADEON_DEBUG & DEBUG_SANITY) {
-      if (rmesa->state.scissor.enabled) 
-	 ret = radeonSanityCmdBuffer( rmesa, 
-				      rmesa->state.scissor.numClipRects,
-				      rmesa->state.scissor.pClipRects);
-      else
-	 ret = radeonSanityCmdBuffer( rmesa, 
-				      rmesa->numClipRects,
-				      rmesa->pClipRects);
-      if (ret) {
-	 fprintf(stderr, "drmSanityCommandWrite: %d\n", ret);	 
-	 goto out;
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH_RELOC(voffset,
+			 rmesa->radeon.tcl.aos[nr - 1].bo,
+			 voffset,
+			 RADEON_GEM_DOMAIN_GTT,
+			 0, 0);
       }
-   }
-
-
-   cmd.bufsz = rmesa->store.cmd_used;
-   cmd.buf = rmesa->store.cmd_buf;
-
-   if (rmesa->state.scissor.enabled) {
-      cmd.nbox = rmesa->state.scissor.numClipRects;
-      cmd.boxes = rmesa->state.scissor.pClipRects;
    } else {
-      cmd.nbox = rmesa->numClipRects;
-      cmd.boxes = rmesa->pClipRects;
-   }
-
-   ret = drmCommandWrite( rmesa->dri.fd,
-			  DRM_RADEON_CMDBUF,
-			  &cmd, sizeof(cmd) );
-
-   if (ret)
-      fprintf(stderr, "drmCommandWrite: %d\n", ret);
-
-   if (RADEON_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
-      radeonWaitForIdleLocked( rmesa );
-   }
-
- out:
-   rmesa->store.primnr = 0;
-   rmesa->store.statenr = 0;
-   rmesa->store.cmd_used = 0;
-   rmesa->dma.nr_released_bufs = 0;
-   rmesa->save_on_next_emit = 1;
-
-   return ret;
-}
-
-
-/* Note: does not emit any commands to avoid recursion on
- * radeonAllocCmdBuf.
- */
-void radeonFlushCmdBuf( radeonContextPtr rmesa, const char *caller )
-{
-   int ret;
-
-	      
-   LOCK_HARDWARE( rmesa );
-
-   ret = radeonFlushCmdBufLocked( rmesa, caller );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if (ret) {
-      fprintf(stderr, "drm_radeon_cmd_buffer_t: %d (exiting)\n", ret);
-      exit(ret);
-   }
-}
-
-/* =============================================================
- * Hardware vertex buffer handling
- */
-
-
-void radeonRefillCurrentDmaRegion( radeonContextPtr rmesa )
-{
-   struct radeon_dma_buffer *dmabuf;
-   int fd = rmesa->dri.fd;
-   int index = 0;
-   int size = 0;
-   drmDMAReq dma;
-   int ret;
-
-   if (RADEON_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-      fprintf(stderr, "%s\n", __FUNCTION__);  
-
-   if (rmesa->dma.flush) {
-      rmesa->dma.flush( rmesa );
-   }
-
-   if (rmesa->dma.current.buf)
-      radeonReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
-
-   if (rmesa->dma.nr_released_bufs > 4)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-
-   dma.context = rmesa->dri.hwContext;
-   dma.send_count = 0;
-   dma.send_list = NULL;
-   dma.send_sizes = NULL;
-   dma.flags = 0;
-   dma.request_count = 1;
-   dma.request_size = RADEON_BUFFER_SIZE;
-   dma.request_list = &index;
-   dma.request_sizes = &size;
-   dma.granted_count = 0;
-
-   LOCK_HARDWARE(rmesa);	/* no need to validate */
-
-   ret = drmDMA( fd, &dma );
-      
-   if (ret != 0) {
-      /* Free some up this way?
-       */
-      if (rmesa->dma.nr_released_bufs) {
-	 radeonFlushCmdBufLocked( rmesa, __FUNCTION__ );
+      for (i = 0; i + 1 < nr; i += 2) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) |
+		   (rmesa->radeon.tcl.aos[i].stride << 8) |
+		   (rmesa->radeon.tcl.aos[i + 1].components << 16) |
+		   (rmesa->radeon.tcl.aos[i + 1].stride << 24));
+
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 OUT_BATCH(voffset);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 OUT_BATCH(voffset);
       }
-      
-      if (RADEON_DEBUG & DEBUG_DMA)
-	 fprintf(stderr, "Waiting for buffers\n");
-
-      radeonWaitForIdleLocked( rmesa );
-      ret = drmDMA( fd, &dma );
 
-      if ( ret != 0 ) {
-	 UNLOCK_HARDWARE( rmesa );
-	 fprintf( stderr, "Error: Could not get dma buffer... exiting\n" );
-	 exit( -1 );
+      if (nr & 1) {
+	 OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
+		   (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 OUT_BATCH(voffset);
       }
-   }
-
-   UNLOCK_HARDWARE(rmesa);
-
-   if (RADEON_DEBUG & DEBUG_DMA)
-      fprintf(stderr, "Allocated buffer %d\n", index);
-
-   dmabuf = CALLOC_STRUCT( radeon_dma_buffer );
-   dmabuf->buf = &rmesa->radeonScreen->buffers->list[index];
-   dmabuf->refcount = 1;
-
-   rmesa->dma.current.buf = dmabuf;
-   rmesa->dma.current.address = dmabuf->buf->address;
-   rmesa->dma.current.end = dmabuf->buf->total;
-   rmesa->dma.current.start = 0;
-   rmesa->dma.current.ptr = 0;
-
-   rmesa->c_vertexBuffers++;
-}
-
-void radeonReleaseDmaRegion( radeonContextPtr rmesa,
-			     struct radeon_dma_region *region,
-			     const char *caller )
-{
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
-   
-   if (!region->buf)
-      return;
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (--region->buf->refcount == 0) {
-      drm_radeon_cmd_header_t *cmd;
-
-      if (RADEON_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
-	 fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__,
-		 region->buf->buf->idx);  
-      
-      cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, sizeof(*cmd), 
-						     __FUNCTION__ );
-      cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD;
-      cmd->dma.buf_idx = region->buf->buf->idx;
-      FREE(region->buf);
-      rmesa->dma.nr_released_bufs++;
-   }
-
-   region->buf = NULL;
-   region->start = 0;
-}
-
-/* Allocates a region from rmesa->dma.current.  If there isn't enough
- * space in current, grab a new buffer (and discard what was left of current)
- */
-void radeonAllocDmaRegion( radeonContextPtr rmesa, 
-			   struct radeon_dma_region *region,
-			   int bytes,
-			   int alignment )
-{
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
-
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
-
-   if (region->buf)
-      radeonReleaseDmaRegion( rmesa, region, __FUNCTION__ );
-
-   alignment--;
-   rmesa->dma.current.start = rmesa->dma.current.ptr = 
-      (rmesa->dma.current.ptr + alignment) & ~alignment;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      radeonRefillCurrentDmaRegion( rmesa );
-
-   region->start = rmesa->dma.current.start;
-   region->ptr = rmesa->dma.current.start;
-   region->end = rmesa->dma.current.start + bytes;
-   region->address = rmesa->dma.current.address;
-   region->buf = rmesa->dma.current.buf;
-   region->buf->refcount++;
-
-   rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */
-   rmesa->dma.current.start = 
-      rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;  
-}
-
-/* ================================================================
- * SwapBuffers with client-side throttling
- */
-
-static uint32_t radeonGetLastFrame (radeonContextPtr rmesa) 
-{
-   drm_radeon_getparam_t gp;
-   int ret;
-   uint32_t frame;
-
-   gp.param = RADEON_PARAM_LAST_FRAME;
-   gp.value = (int *)&frame;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_GETPARAM,
-			      &gp, sizeof(gp) );
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drm_radeon_getparam_t: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-
-   return frame;
-}
-
-static void radeonEmitIrqLocked( radeonContextPtr rmesa )
-{
-   drm_radeon_irq_emit_t ie;
-   int ret;
-
-   ie.irq_seq = &rmesa->iw.irq_seq;
-   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_IRQ_EMIT, 
-			      &ie, sizeof(ie) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drm_radeon_irq_emit_t: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void radeonWaitIrq( radeonContextPtr rmesa )
-{
-   int ret;
-
-   do {
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_IRQ_WAIT,
-			     &rmesa->iw, sizeof(rmesa->iw) );
-   } while (ret && (errno == EINTR || errno == EBUSY));
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-}
-
-
-static void radeonWaitForFrameCompletion( radeonContextPtr rmesa )
-{
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
-
-   if (rmesa->do_irqs) {
-      if (radeonGetLastFrame(rmesa) < sarea->last_frame) {
-	 if (!rmesa->irqsEmitted) {
-	    while (radeonGetLastFrame (rmesa) < sarea->last_frame)
-	       ;
-	 }
-	 else {
-	    UNLOCK_HARDWARE( rmesa ); 
-	    radeonWaitIrq( rmesa );	
-	    LOCK_HARDWARE( rmesa ); 
-	 }
-	 rmesa->irqsEmitted = 10;
-      }
-
-      if (rmesa->irqsEmitted) {
-	 radeonEmitIrqLocked( rmesa );
-	 rmesa->irqsEmitted--;
+      for (i = 0; i + 1 < nr; i += 2) {
+	 voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+0].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
+	 voffset =  rmesa->radeon.tcl.aos[i + 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[i+1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
       }
-   } 
-   else {
-      while (radeonGetLastFrame (rmesa) < sarea->last_frame) {
-	 UNLOCK_HARDWARE( rmesa ); 
-	 if (rmesa->do_usleeps) 
-	    DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa ); 
+      if (nr & 1) {
+	 voffset =  rmesa->radeon.tcl.aos[nr - 1].offset +
+	    offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride;
+	 radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+			       rmesa->radeon.tcl.aos[nr-1].bo,
+			       RADEON_GEM_DOMAIN_GTT,
+			       0, 0);
       }
    }
-}
+   END_BATCH();
 
-/* Copy the back color buffer to the front color buffer.
- */
-void radeonCopyBuffer( __DRIdrawablePrivate *dPriv,
-		       const drm_clip_rect_t	  *rect)
-{
-   radeonContextPtr rmesa;
-   GLint nbox, i, ret;
-   GLboolean   missed_target;
-   int64_t ust;
-   __DRIscreenPrivate *psp;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx );
-   }
-
-   RADEON_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   /* Throttle the frame rate -- only allow one pending swap buffers
-    * request at a time.
-    */
-   radeonWaitForFrameCompletion( rmesa );
-   if (!rect)
-   {
-       UNLOCK_HARDWARE( rmesa );
-       driWaitForVBlank( dPriv, & missed_target );
-       LOCK_HARDWARE( rmesa );
-   }
-
-   nbox = dPriv->numClipRects; /* must be in locked region */
-
-   for ( i = 0 ; i < nbox ; ) {
-      GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      GLint n = 0;
-
-      for ( ; i < nr ; i++ ) {
-
-	  *b = box[i];
-
-	  if (rect)
-	  {
-	      if (rect->x1 > b->x1)
-		  b->x1 = rect->x1;
-	      if (rect->y1 > b->y1)
-		  b->y1 = rect->y1;
-	      if (rect->x2 < b->x2)
-		  b->x2 = rect->x2;
-	      if (rect->y2 < b->y2)
-		  b->y2 = rect->y2;
-
-	      if (b->x1 >= b->x2 || b->y1 >= b->y2)
-		  continue;
-	  }
-
-	  b++;
-	  n++;
-      }
-      rmesa->sarea->nbox = n;
-
-      if (!n)
-	 continue;
-
-      ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
-
-      if ( ret ) {
-	 fprintf( stderr, "DRM_RADEON_SWAP_BUFFERS: return = %d\n", ret );
-	 UNLOCK_HARDWARE( rmesa );
-	 exit( 1 );
-      }
-   }
-
-   UNLOCK_HARDWARE( rmesa );
-   if (!rect)
-   {
-       psp = dPriv->driScreenPriv;
-       rmesa->swap_count++;
-       (*psp->systemTime->getUST)( & ust );
-       if ( missed_target ) {
-	   rmesa->swap_missed_count++;
-	   rmesa->swap_missed_ust = ust - rmesa->swap_ust;
-       }
-
-       rmesa->swap_ust = ust;
-       rmesa->hw.all_dirty = GL_TRUE;
-   }
-}
-
-void radeonPageFlip( __DRIdrawablePrivate *dPriv )
-{
-   radeonContextPtr rmesa;
-   GLint ret;
-   GLboolean   missed_target;
-   __DRIscreenPrivate *psp;
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate;
-   psp = dPriv->driScreenPriv;
-
-   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
-      fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
-	      rmesa->sarea->pfCurrentPage);
-   }
-
-   RADEON_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-
-   /* Need to do this for the perf box placement:
-    */
-   if (dPriv->numClipRects)
-   {
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
-      b[0] = box[0];
-      rmesa->sarea->nbox = 1;
-   }
-
-   /* Throttle the frame rate -- only allow a few pending swap buffers
-    * request at a time.
-    */
-   radeonWaitForFrameCompletion( rmesa );
-   UNLOCK_HARDWARE( rmesa );
-   driWaitForVBlank( dPriv, & missed_target );
-   if ( missed_target ) {
-      rmesa->swap_missed_count++;
-      (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust );
-   }
-   LOCK_HARDWARE( rmesa );
-
-   ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret );
-      exit( 1 );
-   }
-
-   rmesa->swap_count++;
-   (void) (*psp->systemTime->getUST)( & rmesa->swap_ust );
-
-   /* Get ready for drawing next frame.  Update the renderbuffers'
-    * flippedOffset/Pitch fields so we draw into the right place.
-    */
-   driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-                        rmesa->sarea->pfCurrentPage);
-
-   radeonUpdateDrawBuffer(rmesa->glCtx);
+#endif
 }
 
-
 /* ================================================================
  * Buffer clear
  */
 #define RADEON_MAX_CLEARS	256
 
-static void radeonClear( GLcontext *ctx, GLbitfield mask )
+static void radeonKernelClear(GLcontext *ctx, GLuint flags)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   drm_radeon_sarea_t *sarea = rmesa->sarea;
+     r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   drm_radeon_sarea_t *sarea = rmesa->radeon.sarea;
    uint32_t clear;
-   GLuint flags = 0;
-   GLuint color_mask = 0;
    GLint ret, i;
    GLint cx, cy, cw, ch;
 
-   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
-      fprintf( stderr, "radeonClear\n");
-   }
-
-   {
-      LOCK_HARDWARE( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      if ( dPriv->numClipRects == 0 ) 
-	 return;
-   }
-   
-   radeonFlush( ctx ); 
-
-   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
-      flags |= RADEON_FRONT;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_FRONT_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_BACK_LEFT ) {
-      flags |= RADEON_BACK;
-      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      mask &= ~BUFFER_BIT_BACK_LEFT;
-   }
-
-   if ( mask & BUFFER_BIT_DEPTH ) {
-      flags |= RADEON_DEPTH;
-      mask &= ~BUFFER_BIT_DEPTH;
-   }
-
-   if ( (mask & BUFFER_BIT_STENCIL) && rmesa->state.stencil.hwBuffer ) {
-      flags |= RADEON_STENCIL;
-      mask &= ~BUFFER_BIT_STENCIL;
-   }
-
-   if ( mask ) {
-      if (RADEON_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
-      _swrast_Clear( ctx, mask );
-   }
-
-   if ( !flags ) 
-      return;
-
-   if (rmesa->using_hyperz) {
-      flags |= RADEON_USE_COMP_ZBUF;
-/*      if (rmesa->radeonScreen->chipset & RADEON_CHIPSET_TCL) 
-         flags |= RADEON_USE_HIERZ; */
-      if (!(rmesa->state.stencil.hwBuffer) ||
-	 ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
-	    ((rmesa->state.stencil.clear & RADEON_STENCIL_WRITE_MASK) == RADEON_STENCIL_WRITE_MASK))) {
-	  flags |= RADEON_CLEAR_FASTZ;
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
+   LOCK_HARDWARE( &rmesa->radeon );
 
    /* compute region after locking: */
    cx = ctx->DrawBuffer->_Xmin;
@@ -1112,7 +474,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
 
       gp.param = RADEON_PARAM_LAST_CLEAR;
       gp.value = (int *)&clear;
-      ret = drmCommandWriteRead( rmesa->dri.fd,
+      ret = drmCommandWriteRead( rmesa->radeon.dri.fd,
 				 DRM_RADEON_GETPARAM, &gp, sizeof(gp) );
 
       if ( ret ) {
@@ -1124,20 +486,20 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
 	 break;
       }
 
-      if ( rmesa->do_usleeps ) {
-	 UNLOCK_HARDWARE( rmesa );
+      if ( rmesa->radeon.do_usleeps ) {
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 DO_USLEEP( 1 );
-	 LOCK_HARDWARE( rmesa );
+	 LOCK_HARDWARE( &rmesa->radeon );
       }
    }
 
    /* Send current state to the hardware */
-   radeonFlushCmdBufLocked( rmesa, __FUNCTION__ );
+   rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ );
 
    for ( i = 0 ; i < dPriv->numClipRects ; ) {
       GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, dPriv->numClipRects );
       drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t *b = rmesa->sarea->boxes;
+      drm_clip_rect_t *b = rmesa->radeon.sarea->boxes;
       drm_radeon_clear_t clear;
       drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
       GLint n = 0;
@@ -1172,105 +534,107 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
 	 }
       }
 
-      rmesa->sarea->nbox = n;
+      rmesa->radeon.sarea->nbox = n;
 
       clear.flags       = flags;
-      clear.clear_color = rmesa->state.color.clear;
-      clear.clear_depth = rmesa->state.depth.clear;
+      clear.clear_color = rmesa->radeon.state.color.clear;
+      clear.clear_depth = rmesa->radeon.state.depth.clear;
       clear.color_mask  = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
-      clear.depth_mask  = rmesa->state.stencil.clear;
+      clear.depth_mask  = rmesa->radeon.state.stencil.clear;
       clear.depth_boxes = depth_boxes;
 
       n--;
-      b = rmesa->sarea->boxes;
+      b = rmesa->radeon.sarea->boxes;
       for ( ; n >= 0 ; n-- ) {
 	 depth_boxes[n].f[CLEAR_X1] = (float)b[n].x1;
 	 depth_boxes[n].f[CLEAR_Y1] = (float)b[n].y1;
 	 depth_boxes[n].f[CLEAR_X2] = (float)b[n].x2;
 	 depth_boxes[n].f[CLEAR_Y2] = (float)b[n].y2;
-	 depth_boxes[n].f[CLEAR_DEPTH] = 
-	    (float)rmesa->state.depth.clear;
+	 depth_boxes[n].f[CLEAR_DEPTH] =
+	    (float)rmesa->radeon.state.depth.clear;
       }
 
-      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_CLEAR,
+      ret = drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_CLEAR,
 			     &clear, sizeof(drm_radeon_clear_t));
 
       if ( ret ) {
-	 UNLOCK_HARDWARE( rmesa );
+	 UNLOCK_HARDWARE( &rmesa->radeon );
 	 fprintf( stderr, "DRM_RADEON_CLEAR: return = %d\n", ret );
 	 exit( 1 );
       }
    }
-
-   UNLOCK_HARDWARE( rmesa );
-   rmesa->hw.all_dirty = GL_TRUE;
+   UNLOCK_HARDWARE( &rmesa->radeon );
 }
 
-
-void radeonWaitForIdleLocked( radeonContextPtr rmesa )
+static void radeonClear( GLcontext *ctx, GLbitfield mask )
 {
-    int fd = rmesa->dri.fd;
-    int to = 0;
-    int ret, i = 0;
-
-    rmesa->c_drawWaits++;
-
-    do {
-        do {
-            ret = drmCommandNone( fd, DRM_RADEON_CP_IDLE);
-        } while ( ret && errno == EBUSY && i++ < RADEON_IDLE_RETRY );
-    } while ( ( ret == -EBUSY ) && ( to++ < RADEON_TIMEOUT ) );
-
-    if ( ret < 0 ) {
-	UNLOCK_HARDWARE( rmesa );
-	fprintf( stderr, "Error: Radeon timed out... exiting\n" );
-	exit( -1 );
-    }
-}
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLuint flags = 0;
+   GLuint color_mask = 0;
+   GLuint orig_mask = mask;
 
+   if ( RADEON_DEBUG & DEBUG_IOCTL ) {
+      fprintf( stderr, "radeonClear\n");
+   }
 
-static void radeonWaitForIdle( radeonContextPtr rmesa )
-{
-   LOCK_HARDWARE(rmesa);
-   radeonWaitForIdleLocked( rmesa );
-   UNLOCK_HARDWARE(rmesa);
-}
+   {
+      LOCK_HARDWARE( &rmesa->radeon );
+      UNLOCK_HARDWARE( &rmesa->radeon );
+      if ( dPriv->numClipRects == 0 )
+	 return;
+   }
 
+   radeon_firevertices(&rmesa->radeon);
 
-void radeonFlush( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   if ( mask & BUFFER_BIT_FRONT_LEFT ) {
+      flags |= RADEON_FRONT;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_FRONT_LEFT;
+   }
 
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
+   if ( mask & BUFFER_BIT_BACK_LEFT ) {
+      flags |= RADEON_BACK;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~BUFFER_BIT_BACK_LEFT;
+   }
 
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
+   if ( mask & BUFFER_BIT_DEPTH ) {
+      flags |= RADEON_DEPTH;
+      mask &= ~BUFFER_BIT_DEPTH;
+   }
 
-   radeonEmitState( rmesa );
-   
-   if (rmesa->store.cmd_used)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-}
+   if ( (mask & BUFFER_BIT_STENCIL) ) {
+      flags |= RADEON_STENCIL;
+      mask &= ~BUFFER_BIT_STENCIL;
+   }
 
-/* Make sure all commands have been sent to the hardware and have
- * completed processing.
- */
-void radeonFinish( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeonFlush( ctx );
-
-   if (rmesa->do_irqs) {
-      LOCK_HARDWARE( rmesa );
-      radeonEmitIrqLocked( rmesa );
-      UNLOCK_HARDWARE( rmesa );
-      radeonWaitIrq( rmesa );
+   if ( mask ) {
+      if (RADEON_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask);
+      _swrast_Clear( ctx, mask );
    }
-   else
-      radeonWaitForIdle( rmesa );
-}
 
+   if ( !flags )
+      return;
+
+   if (rmesa->using_hyperz) {
+      flags |= RADEON_USE_COMP_ZBUF;
+/*      if (rmesa->radeon.radeonScreen->chipset & RADEON_CHIPSET_TCL)
+         flags |= RADEON_USE_HIERZ; */
+      if (((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
+	    ((rmesa->radeon.state.stencil.clear & RADEON_STENCIL_WRITE_MASK) == RADEON_STENCIL_WRITE_MASK))) {
+	  flags |= RADEON_CLEAR_FASTZ;
+      }
+   }
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     radeonUserClear(ctx, orig_mask);
+   else {
+      radeonKernelClear(ctx, flags);
+      rmesa->radeon.hw.all_dirty = GL_TRUE;
+   }
+}
 
 void radeonInitIoctlFuncs( GLcontext *ctx )
 {
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h
index 4e3a44df07..18805d4c57 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h
@@ -38,31 +38,32 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "main/simple_list.h"
 #include "radeon_lock.h"
+#include "radeon_bocs_wrapper.h"
 
-
-extern void radeonEmitState( radeonContextPtr rmesa );
-extern void radeonEmitVertexAOS( radeonContextPtr rmesa,
+extern void radeonEmitVertexAOS( r100ContextPtr rmesa,
 				 GLuint vertex_size,
+				 struct radeon_bo *bo,
 				 GLuint offset );
 
-extern void radeonEmitVbufPrim( radeonContextPtr rmesa,
+extern void radeonEmitVbufPrim( r100ContextPtr rmesa,
 				GLuint vertex_format,
 				GLuint primitive,
 				GLuint vertex_nr );
 
-extern void radeonFlushElts( radeonContextPtr rmesa );
+extern void radeonFlushElts( GLcontext *ctx );
+			    
 
-extern GLushort *radeonAllocEltsOpenEnded( radeonContextPtr rmesa,
+extern GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa,
 					   GLuint vertex_format,
 					   GLuint primitive,
 					   GLuint min_nr );
 
-extern void radeonEmitAOS( radeonContextPtr rmesa,
-			   struct radeon_dma_region **regions,
+
+extern void radeonEmitAOS( r100ContextPtr rmesa,
 			   GLuint n,
 			   GLuint offset );
 
-extern void radeonEmitBlit( radeonContextPtr rmesa,
+extern void radeonEmitBlit( r100ContextPtr rmesa,
 			    GLuint color_fmt,
 			    GLuint src_pitch,
 			    GLuint src_offset,
@@ -72,30 +73,15 @@ extern void radeonEmitBlit( radeonContextPtr rmesa,
 			    GLint dstx, GLint dsty,
 			    GLuint w, GLuint h );
 
-extern void radeonEmitWait( radeonContextPtr rmesa, GLuint flags );
-
-extern void radeonFlushCmdBuf( radeonContextPtr rmesa, const char * );
-extern void radeonRefillCurrentDmaRegion( radeonContextPtr rmesa );
+extern void radeonEmitWait( r100ContextPtr rmesa, GLuint flags );
 
-extern void radeonAllocDmaRegion( radeonContextPtr rmesa,
-				  struct radeon_dma_region *region,
-				  int bytes, 
-				  int alignment );
+extern void radeonFlushCmdBuf( r100ContextPtr rmesa, const char * );
 
-extern void radeonReleaseDmaRegion( radeonContextPtr rmesa,
-				    struct radeon_dma_region *region,
-				    const char *caller );
-
-extern void radeonCopyBuffer( __DRIdrawablePrivate *drawable,
-			      const drm_clip_rect_t	 *rect);
-extern void radeonPageFlip( __DRIdrawablePrivate *drawable );
 extern void radeonFlush( GLcontext *ctx );
 extern void radeonFinish( GLcontext *ctx );
-extern void radeonWaitForIdleLocked( radeonContextPtr rmesa );
-extern void radeonWaitForVBlank( radeonContextPtr rmesa );
 extern void radeonInitIoctlFuncs( GLcontext *ctx );
-extern void radeonGetAllParams( radeonContextPtr rmesa );
-extern void radeonSetUpAtomList( radeonContextPtr rmesa );
+extern void radeonGetAllParams( r100ContextPtr rmesa );
+extern void radeonSetUpAtomList( r100ContextPtr rmesa );
 
 /* ================================================================
  * Helper macros:
@@ -105,33 +91,33 @@ extern void radeonSetUpAtomList( radeonContextPtr rmesa );
  */
 #define RADEON_NEWPRIM( rmesa )			\
 do {						\
-   if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );	\
+   if ( rmesa->radeon.dma.flush )			\
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	\
 } while (0)
 
 /* Can accomodate several state changes and primitive changes without
  * actually firing the buffer.
  */
+
 #define RADEON_STATECHANGE( rmesa, ATOM )			\
 do {								\
    RADEON_NEWPRIM( rmesa );					\
    rmesa->hw.ATOM.dirty = GL_TRUE;				\
-   rmesa->hw.is_dirty = GL_TRUE;				\
+   rmesa->radeon.hw.is_dirty = GL_TRUE;				\
 } while (0)
 
-#define RADEON_DB_STATE( ATOM )			        \
+#define RADEON_DB_STATE( ATOM )				\
    memcpy( rmesa->hw.ATOM.lastcmd, rmesa->hw.ATOM.cmd,	\
 	   rmesa->hw.ATOM.cmd_size * 4)
 
-static INLINE int RADEON_DB_STATECHANGE( 
-   radeonContextPtr rmesa,
-   struct radeon_state_atom *atom )
+static INLINE int RADEON_DB_STATECHANGE(r100ContextPtr rmesa,
+					struct radeon_state_atom *atom )
 {
    if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) {
-      int *tmp;
+      GLuint *tmp;
       RADEON_NEWPRIM( rmesa );
       atom->dirty = GL_TRUE;
-      rmesa->hw.is_dirty = GL_TRUE;
+      rmesa->radeon.hw.is_dirty = GL_TRUE;
       tmp = atom->cmd; 
       atom->cmd = atom->lastcmd;
       atom->lastcmd = tmp;
@@ -141,16 +127,6 @@ static INLINE int RADEON_DB_STATECHANGE(
       return 0;
 }
 
-
-/* Fire the buffered vertices no matter what.
- */
-#define RADEON_FIREVERTICES( rmesa )			\
-do {							\
-   if ( rmesa->store.cmd_used || rmesa->dma.flush ) {	\
-      radeonFlush( rmesa->glCtx );			\
-   }							\
-} while (0)
-
 /* Command lengths.  Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ
  * are available, you will also be adding an rmesa->state.max_state_size because
  * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts.
@@ -167,36 +143,37 @@ do {							\
 #define VBUF_BUFSZ	(4 * sizeof(int))
 #endif
 
-/* Ensure that a minimum amount of space is available in the command buffer.
- * This is used to ensure atomicity of state updates with the rendering requests
- * that rely on them.
- *
- * An alternative would be to implement a "soft lock" such that when the buffer
- * wraps at an inopportune time, we grab the lock, flush the current buffer,
- * and hang on to the lock until the critical section is finished and we flush
- * the buffer again and unlock.
- */
-static INLINE void radeonEnsureCmdBufSpace( radeonContextPtr rmesa,
-					      int bytes )
-{
-   if (rmesa->store.cmd_used + bytes > RADEON_CMD_BUF_SZ)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
-   assert( bytes <= RADEON_CMD_BUF_SZ );
-}
 
-/* Alloc space in the command buffer
- */
-static INLINE char *radeonAllocCmdBuf( radeonContextPtr rmesa,
-					 int bytes, const char *where )
+static inline uint32_t cmdpacket3(int cmd_type)
 {
-   if (rmesa->store.cmd_used + bytes > RADEON_CMD_BUF_SZ)
-      radeonFlushCmdBuf( rmesa, __FUNCTION__ );
+  drm_radeon_cmd_header_t cmd;
+
+  cmd.i = 0;
+  cmd.header.cmd_type = cmd_type;
+
+  return (uint32_t)cmd.i;
 
-   {
-      char *head = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-      rmesa->store.cmd_used += bytes;
-      return head;
-   }
 }
 
+#define OUT_BATCH_PACKET3(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3));				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+#define OUT_BATCH_PACKET3_CLIP(packet, num_extra) do {	      \
+    if (!b_l_rmesa->radeonScreen->kernel_mm) {		      \
+      OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3_CLIP));	      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    } else {						      \
+      OUT_BATCH(CP_PACKET2);				      \
+      OUT_BATCH(CP_PACKET3((packet), (num_extra)));	      \
+    }							      \
+  } while(0)
+
+
 #endif /* __RADEON_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c
index 64bb3ca103..5774f7ebcf 100644
--- a/src/mesa/drivers/dri/radeon/radeon_lock.c
+++ b/src/mesa/drivers/dri/radeon/radeon_lock.c
@@ -41,30 +41,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "main/glheader.h"
 #include "main/mtypes.h"
-#include "radeon_context.h"
+#include "main/colormac.h"
+#include "dri_util.h"
+#include "radeon_screen.h"
+#include "radeon_common.h"
 #include "radeon_lock.h"
-#include "radeon_tex.h"
-#include "radeon_state.h"
-#include "radeon_ioctl.h"
-
 #include "drirenderbuffer.h"
 
-#if DEBUG_LOCKING
-char *prevLockFile = NULL;
-int prevLockLine = 0;
-#endif
-
-/* Turn on/off page flipping according to the flags in the sarea:
- */
-static void radeonUpdatePageFlipping(radeonContextPtr rmesa)
-{
-	rmesa->doPageFlip = rmesa->sarea->pfState;
-	if (rmesa->glCtx->WinSysDrawBuffer) {
-		driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer,
-				     rmesa->sarea->pfCurrentPage);
-	}
-}
-
 /* Update the hardware state.  This is called if another context has
  * grabbed the hardware lock, which includes the X server.  This
  * function also updates the driver's window state after the X server
@@ -75,10 +58,11 @@ static void radeonUpdatePageFlipping(radeonContextPtr rmesa)
  */
 void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
 {
-	__DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-	__DRIdrawablePrivate *const readable = rmesa->dri.readable;
+	__DRIdrawablePrivate *const drawable = radeon_get_drawable(rmesa);
+	__DRIdrawablePrivate *const readable = radeon_get_readable(rmesa);
 	__DRIscreenPrivate *sPriv = rmesa->dri.screen;
-	drm_radeon_sarea_t *sarea = rmesa->sarea;
+
+	assert(drawable != NULL);
 
 	drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags);
 
@@ -96,29 +80,42 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
 	}
 
 	if (rmesa->lastStamp != drawable->lastStamp) {
-		radeonUpdatePageFlipping(rmesa);
-		radeonSetCliprects(rmesa);
-		radeonUpdateViewportOffset(rmesa->glCtx);
-		driUpdateFramebufferSize(rmesa->glCtx, drawable);
+		radeon_window_moved(rmesa);
+		rmesa->lastStamp = drawable->lastStamp;
 	}
 
-	RADEON_STATECHANGE(rmesa, ctx);
-	if (rmesa->sarea->tiling_enabled) {
-		rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |=
-		    RADEON_COLOR_TILE_ENABLE;
-	} else {
-		rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] &=
-		    ~RADEON_COLOR_TILE_ENABLE;
-	}
+	rmesa->vtbl.get_lock(rmesa);
+
+	rmesa->lost_context = GL_TRUE;
+}
 
-	if (sarea->ctx_owner != rmesa->dri.hwContext) {
-		int i;
-		sarea->ctx_owner = rmesa->dri.hwContext;
+void radeon_lock_hardware(radeonContextPtr radeon)
+{
+	char ret = 0;
+	struct radeon_framebuffer *rfb = NULL;
+	struct radeon_renderbuffer *rrb = NULL;
 
-		for (i = 0; i < rmesa->nr_heaps; i++) {
-			DRI_AGE_TEXTURES(rmesa->texture_heaps[i]);
-		}
+	if (radeon_get_drawable(radeon)) {
+		rfb = radeon_get_drawable(radeon)->driverPrivate;
+
+		if (rfb)
+			rrb = radeon_get_renderbuffer(&rfb->base,
+						      rfb->base._ColorDrawBufferIndexes[0]);
 	}
 
-	rmesa->lost_context = GL_TRUE;
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+		DRM_CAS(radeon->dri.hwLock, radeon->dri.hwContext,
+			 (DRM_LOCK_HELD | radeon->dri.hwContext), ret );
+		if (ret)
+			radeonGetLock(radeon, 0);
+	}
+}
+
+void radeon_unlock_hardware(radeonContextPtr radeon)
+{
+	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
+		DRM_UNLOCK( radeon->dri.fd,
+			    radeon->dri.hwLock,
+			    radeon->dri.hwContext );
+	}
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.h b/src/mesa/drivers/dri/radeon/radeon_lock.h
index 86e96aa7d2..2817709eed 100644
--- a/src/mesa/drivers/dri/radeon/radeon_lock.h
+++ b/src/mesa/drivers/dri/radeon/radeon_lock.h
@@ -39,74 +39,21 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Kevin E. Martin <martin@valinux.com>
  */
 
-#ifndef __RADEON_LOCK_H__
-#define __RADEON_LOCK_H__
+#ifndef COMMON_LOCK_H
+#define COMMON_LOCK_H
 
-extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags);
-
-/* Turn DEBUG_LOCKING on to find locking conflicts.
- */
-#define DEBUG_LOCKING	0
-
-#if DEBUG_LOCKING
-extern char *prevLockFile;
-extern int prevLockLine;
-
-#define DEBUG_LOCK()							\
-   do {									\
-      prevLockFile = (__FILE__);					\
-      prevLockLine = (__LINE__);					\
-   } while (0)
-
-#define DEBUG_RESET()							\
-   do {									\
-      prevLockFile = 0;							\
-      prevLockLine = 0;							\
-   } while (0)
-
-#define DEBUG_CHECK_LOCK()						\
-   do {									\
-      if ( prevLockFile ) {						\
-	 fprintf( stderr,						\
-		  "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",	\
-		  prevLockFile, prevLockLine, __FILE__, __LINE__ );	\
-	 exit( 1 );							\
-      }									\
-   } while (0)
-
-#else
+#include "main/colormac.h"
+#include "radeon_screen.h"
+#include "radeon_common.h"
 
-#define DEBUG_LOCK()
-#define DEBUG_RESET()
-#define DEBUG_CHECK_LOCK()
+extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags);
 
-#endif
-
-/*
- * !!! We may want to separate locks from locks with validation.  This
- * could be used to improve performance for those things commands that
- * do not do any drawing !!!
- */
+void radeon_lock_hardware(radeonContextPtr rmesa);
+void radeon_unlock_hardware(radeonContextPtr rmesa);
 
 /* Lock the hardware and validate our state.
  */
-#define LOCK_HARDWARE( rmesa )					\
-   do {								\
-      char __ret = 0;						\
-      DEBUG_CHECK_LOCK();					\
-      DRM_CAS( (rmesa)->dri.hwLock, (rmesa)->dri.hwContext,		\
-	       (DRM_LOCK_HELD | (rmesa)->dri.hwContext), __ret );	\
-      if ( __ret )						\
-	 radeonGetLock( (rmesa), 0 );				\
-      DEBUG_LOCK();						\
-   } while (0)
-
-#define UNLOCK_HARDWARE( rmesa )					\
-   do {									\
-      DRM_UNLOCK( (rmesa)->dri.fd,					\
-		  (rmesa)->dri.hwLock,					\
-		  (rmesa)->dri.hwContext );				\
-      DEBUG_RESET();							\
-   } while (0)
+#define LOCK_HARDWARE( rmesa )	radeon_lock_hardware(rmesa)
+#define UNLOCK_HARDWARE( rmesa )  radeon_unlock_hardware(rmesa)
 
-#endif				/* __RADEON_LOCK_H__ */
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos.h b/src/mesa/drivers/dri/radeon/radeon_maos.h
index b8935e84a0..b88eb198d5 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos.h
+++ b/src/mesa/drivers/dri/radeon/radeon_maos.h
@@ -38,6 +38,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_context.h"
 
 extern void radeonEmitArrays( GLcontext *ctx, GLuint inputs );
-extern void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs );
 
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c
index 31eea13f4e..7c6ea0530e 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c
@@ -48,160 +48,35 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_maos.h"
 #include "radeon_tcl.h"
 
-#if 0
-/* Usage:
- *   - from radeon_tcl_render
- *   - call radeonEmitArrays to ensure uptodate arrays in dma
- *   - emit primitives (new type?) which reference the data
- *       -- need to use elts for lineloop, quads, quadstrip/flat
- *       -- other primitives are all well-formed (need tristrip-1,fake-poly)
- *
- */
-static void emit_ubyte_rgba3( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
+static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos,
+			GLvoid *data, int stride, int count)
 {
    int i;
-   radeon_color_t *out = (radeon_color_t *)(rvb->start + rvb->address);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p\n",
-	      __FUNCTION__, count, stride, (void *)out);
-
-   for (i = 0; i < count; i++) {
-      out->red   = *data;
-      out->green = *(data+1);
-      out->blue  = *(data+2);
-      out->alpha = 0xFF;
-      out++;
-      data += stride;
-   }
-}
-
-static void emit_ubyte_rgba4( GLcontext *ctx,
-			      struct radeon_dma_region *rvb,
-			      char *data,
-			      int stride,
-			      int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
+   uint32_t *out;
+   int size = 1;
+   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
 
    if (RADEON_DEBUG & DEBUG_VERTS)
       fprintf(stderr, "%s count %d stride %d\n",
 	      __FUNCTION__, count, stride);
 
-   if (stride == 4)
-       COPY_DWORDS( out, data, count );
-   else
-      for (i = 0; i < count; i++) {
-	 *out++ = LE32_TO_CPU(*(int *)data);
-	 data += stride;
-      }
-}
-
-
-static void emit_ubyte_rgba( GLcontext *ctx,
-			     struct radeon_dma_region *rvb,
-			     char *data,
-			     int size,
-			     int stride,
-			     int count )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
-
-   assert (!rvb->buf);
-
    if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, 4, 4 );
+      radeonAllocDmaRegion( rmesa, &aos->bo, &aos->offset, size * 4, 32 );
       count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
+      aos->stride = 0;
    }
    else {
-      radeonAllocDmaRegion( rmesa, rvb, 4 * count, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
+      radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+      aos->stride = size;
    }
 
-   /* Emit the data
-    */
-   switch (size) {
-   case 3:
-      emit_ubyte_rgba3( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_ubyte_rgba4( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-}
-#endif
-
-#if defined(USE_X86_ASM)
-#define COPY_DWORDS( dst, src, nr )					\
-do {									\
-	int __tmp;							\
-	__asm__ __volatile__( "rep ; movsl"				\
-			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
-			      : "0" (nr),				\
-			        "D" ((long)dst),			\
-			        "S" ((long)src) );			\
-} while (0)
-#else
-#define COPY_DWORDS( dst, src, nr )		\
-do {						\
-   int j;					\
-   for ( j = 0 ; j < nr ; j++ )			\
-      dst[j] = ((int *)src)[j];			\
-   dst += nr;					\
-} while (0)
-#endif
-
-static void emit_vecfog( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
-{
-   int i;
-   GLfloat *out;
+   aos->components = size;
+   aos->count = count;
 
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = 1;
-   }
-   else {
-      radeonAllocDmaRegion( rmesa, rvb, count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 1;
-      rvb->aos_size = 1;
-   }
 
    /* Emit the data
     */
-   out = (GLfloat *)(rvb->address + rvb->start);
+   out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
    for (i = 0; i < count; i++) {
       out[0] = radeonComputeFogBlendFactor( ctx, *(GLfloat *)data );
       out++;
@@ -209,169 +84,9 @@ static void emit_vecfog( GLcontext *ctx,
    }
 }
 
-static void emit_vec4( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 4)
-      COPY_DWORDS( out, data, count );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out++;
-	 data += stride;
-      }
-}
-
-
-static void emit_vec8( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 8)
-      COPY_DWORDS( out, data, count*2 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out += 2;
-	 data += stride;
-      }
-}
-
-static void emit_vec12( GLcontext *ctx,
-		       struct radeon_dma_region *rvb,
-		       char *data,
-		       int stride,
-		       int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d out %p data %p\n",
-	      __FUNCTION__, count, stride, (void *)out, (void *)data);
-
-   if (stride == 12)
-      COPY_DWORDS( out, data, count*3 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out += 3;
-	 data += stride;
-      }
-}
-
-static void emit_vec16( GLcontext *ctx,
-			struct radeon_dma_region *rvb,
-			char *data,
-			int stride,
-			int count )
-{
-   int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d stride %d\n",
-	      __FUNCTION__, count, stride);
-
-   if (stride == 16)
-      COPY_DWORDS( out, data, count*4 );
-   else
-      for (i = 0; i < count; i++) {
-	 out[0] = *(int *)data;
-	 out[1] = *(int *)(data+4);
-	 out[2] = *(int *)(data+8);
-	 out[3] = *(int *)(data+12);
-	 out += 4;
-	 data += stride;
-      }
-}
-
-
-static void emit_vector( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int size,
-			 int stride,
-			 int count )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VERTS)
-      fprintf(stderr, "%s count %d size %d stride %d\n",
-	      __FUNCTION__, count, size, stride);
-
-   assert (!rvb->buf);
-
-   if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, size * 4, 4 );
-      count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = size;
-   }
-   else {
-      radeonAllocDmaRegion( rmesa, rvb, size * count * 4, 4 );	/* alignment? */
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = size;
-      rvb->aos_size = size;
-   }
-
-   /* Emit the data
-    */
-   switch (size) {
-   case 1:
-      emit_vec4( ctx, rvb, data, stride, count );
-      break;
-   case 2:
-      emit_vec8( ctx, rvb, data, stride, count );
-      break;
-   case 3:
-      emit_vec12( ctx, rvb, data, stride, count );
-      break;
-   case 4:
-      emit_vec16( ctx, rvb, data, stride, count );
-      break;
-   default:
-      assert(0);
-      exit(1);
-      break;
-   }
-
-}
-
-
-
-static void emit_s0_vec( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
+static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count)
 {
    int i;
-   int *out = (int *)(rvb->address + rvb->start);
-
    if (RADEON_DEBUG & DEBUG_VERTS)
       fprintf(stderr, "%s count %d stride %d\n",
 	      __FUNCTION__, count, stride);
@@ -384,14 +99,9 @@ static void emit_s0_vec( GLcontext *ctx,
    }
 }
 
-static void emit_stq_vec( GLcontext *ctx,
-			 struct radeon_dma_region *rvb,
-			 char *data,
-			 int stride,
-			 int count )
+static void emit_stq_vec(uint32_t *out, GLvoid *data, int stride, int count)
 {
    int i;
-   int *out = (int *)(rvb->address + rvb->start);
 
    if (RADEON_DEBUG & DEBUG_VERTS)
       fprintf(stderr, "%s count %d stride %d\n",
@@ -409,21 +119,16 @@ static void emit_stq_vec( GLcontext *ctx,
 
 
 
-static void emit_tex_vector( GLcontext *ctx,
-			     struct radeon_dma_region *rvb,
-			     char *data,
-			     int size,
-			     int stride,
-			     int count )
+static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos,
+			    GLvoid *data, int size, int stride, int count)
 {
    radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
    int emitsize;
+   uint32_t *out;
 
    if (RADEON_DEBUG & DEBUG_VERTS)
       fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
 
-   assert (!rvb->buf);
-
    switch (size) {
    case 4: emitsize = 3; break;
    case 3: emitsize = 3; break;
@@ -432,34 +137,33 @@ static void emit_tex_vector( GLcontext *ctx,
 
 
    if (stride == 0) {
-      radeonAllocDmaRegion( rmesa, rvb, 4 * emitsize, 4 );
+      radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, emitsize * 4, 32);
       count = 1;
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = 0;
-      rvb->aos_size = emitsize;
+      aos->stride = 0;
    }
    else {
-      radeonAllocDmaRegion( rmesa, rvb, 4 * emitsize * count, 4 );
-      rvb->aos_start = GET_START(rvb);
-      rvb->aos_stride = emitsize;
-      rvb->aos_size = emitsize;
+      radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, emitsize * count * 4, 32);
+      aos->stride = emitsize;
    }
 
+   aos->components = emitsize;
+   aos->count = count;
 
    /* Emit the data
     */
+   out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
    switch (size) {
    case 1:
-      emit_s0_vec( ctx, rvb, data, stride, count ); 
+      emit_s0_vec( out, data, stride, count );
       break;
    case 2:
-      emit_vec8( ctx, rvb, data, stride, count );
+      radeonEmitVec8( out, data, stride, count );
       break;
    case 3:
-      emit_vec12( ctx, rvb, data, stride, count );
+      radeonEmitVec12( out, data, stride, count );
       break;
    case 4:
-      emit_stq_vec( ctx, rvb, data, stride, count );
+      emit_stq_vec( out, data, stride, count );
       break;
    default:
       assert(0);
@@ -476,9 +180,8 @@ static void emit_tex_vector( GLcontext *ctx,
  */
 void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
-   struct radeon_dma_region **component = rmesa->tcl.aos_components;
    GLuint nr = 0;
    GLuint vfmt = 0;
    GLuint count = VB->Count;
@@ -491,12 +194,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 
    if (1) {
       if (!rmesa->tcl.obj.buf) 
-	 emit_vector( ctx, 
-		      &rmesa->tcl.obj, 
-		      (char *)VB->ObjPtr->data,
-		      VB->ObjPtr->size,
-		      VB->ObjPtr->stride,
-		      count);
+	rcommon_emit_vector( ctx, 
+			     &(rmesa->tcl.aos[nr]),
+			     (char *)VB->ObjPtr->data,
+			     VB->ObjPtr->size,
+			     VB->ObjPtr->stride,
+			     count);
 
       switch( VB->ObjPtr->size ) {
       case 4: vfmt |= RADEON_CP_VC_FRMT_W0;
@@ -505,21 +208,21 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
       default:
          break;
       }
-      component[nr++] = &rmesa->tcl.obj;
+      nr++;
    }
    
 
    if (inputs & VERT_BIT_NORMAL) {
       if (!rmesa->tcl.norm.buf)
-	 emit_vector( ctx, 
-		      &(rmesa->tcl.norm), 
-		      (char *)VB->NormalPtr->data,
-		      3,
-		      VB->NormalPtr->stride,
-		      count);
+	 rcommon_emit_vector( ctx, 
+			      &(rmesa->tcl.aos[nr]),
+			      (char *)VB->NormalPtr->data,
+			      3,
+			      VB->NormalPtr->stride,
+			      count);
 
       vfmt |= RADEON_CP_VC_FRMT_N0;
-      component[nr++] = &rmesa->tcl.norm;
+      nr++;
    }
 
    if (inputs & VERT_BIT_COLOR0) {
@@ -537,31 +240,30 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
       }
 
       if (!rmesa->tcl.rgba.buf)
-	 emit_vector( ctx,
-		      &(rmesa->tcl.rgba),
-		      (char *)VB->ColorPtr[0]->data,
-		      emitsize,
-		      VB->ColorPtr[0]->stride,
-		      count);
-
-
-      component[nr++] = &rmesa->tcl.rgba;
+	rcommon_emit_vector( ctx,
+			     &(rmesa->tcl.aos[nr]),
+			     (char *)VB->ColorPtr[0]->data,
+			     emitsize,
+			     VB->ColorPtr[0]->stride,
+			     count);
+
+      nr++;
    }
 
 
    if (inputs & VERT_BIT_COLOR1) {
       if (!rmesa->tcl.spec.buf) {
 
-	 emit_vector( ctx,
-		      &rmesa->tcl.spec,
-		      (char *)VB->SecondaryColorPtr[0]->data,
-		      3,
-		      VB->SecondaryColorPtr[0]->stride,
-		      count);
+	rcommon_emit_vector( ctx,
+			     &(rmesa->tcl.aos[nr]),
+			     (char *)VB->SecondaryColorPtr[0]->data,
+			     3,
+			     VB->SecondaryColorPtr[0]->stride,
+			     count);
       }
 
       vfmt |= RADEON_CP_VC_FRMT_FPSPEC;
-      component[nr++] = &rmesa->tcl.spec;
+      nr++;
    }
 
 /* FIXME: not sure if this is correct. May need to stitch this together with
@@ -570,13 +272,13 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
    if (inputs & VERT_BIT_FOG) {
       if (!rmesa->tcl.fog.buf)
 	 emit_vecfog( ctx,
-		      &(rmesa->tcl.fog),
+		      &(rmesa->tcl.aos[nr]),
 		      (char *)VB->FogCoordPtr->data,
 		      VB->FogCoordPtr->stride,
 		      count);
 
       vfmt |= RADEON_CP_VC_FRMT_FPFOG;
-      component[nr++] = &rmesa->tcl.fog;
+      nr++;
    }
 
 
@@ -587,11 +289,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
       if (inputs & VERT_BIT_TEX(unit)) {
 	 if (!rmesa->tcl.tex[unit].buf)
 	    emit_tex_vector( ctx,
-			     &(rmesa->tcl.tex[unit]),
+			     &(rmesa->tcl.aos[nr]),
 			     (char *)VB->TexCoordPtr[unit]->data,
 			     VB->TexCoordPtr[unit]->size,
 			     VB->TexCoordPtr[unit]->stride,
 			     count );
+	 nr++;
 
 	 vfmt |= RADEON_ST_BIT(unit);
          /* assume we need the 3rd coord if texgen is active for r/q OR at least
@@ -609,7 +312,6 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 		 (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1)))
 	       radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ;
 	 }
-	 component[nr++] = &rmesa->tcl.tex[unit];
       }
    }
 
@@ -622,34 +324,3 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
    rmesa->tcl.vertex_format = vfmt;
 }
 
-
-void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-   GLuint unit;
-
-#if 0
-   if (RADEON_DEBUG & DEBUG_VERTS) 
-      _tnl_print_vert_flags( __FUNCTION__, newinputs );
-#endif
-
-   if (newinputs & VERT_BIT_POS) 
-     radeonReleaseDmaRegion( rmesa, &rmesa->tcl.obj, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_NORMAL) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.norm, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_COLOR0) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.rgba, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_COLOR1) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.spec, __FUNCTION__ );
-      
-   if (newinputs & VERT_BIT_FOG)
-      radeonReleaseDmaRegion( rmesa, &rmesa->tcl.fog, __FUNCTION__ );
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (newinputs & VERT_BIT_TEX(unit))
-         radeonReleaseDmaRegion( rmesa, &rmesa->tcl.tex[unit], __FUNCTION__ );
-   }
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
index 126d0727c6..78ec119302 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
@@ -310,7 +310,7 @@ static void init_tcl_verts( void )
 
 void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
    GLuint req = 0;
    GLuint unit;
@@ -374,14 +374,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 	 break;
 
    if (rmesa->tcl.vertex_format == setup_tab[i].vertex_format &&
-       rmesa->tcl.indexed_verts.buf)
+       rmesa->radeon.tcl.aos[0].bo)
       return;
 
-   if (rmesa->tcl.indexed_verts.buf)
+   if (rmesa->radeon.tcl.aos[0].bo)
       radeonReleaseArrays( ctx, ~0 );
 
-   radeonAllocDmaRegion( rmesa,
-			 &rmesa->tcl.indexed_verts, 
+   radeonAllocDmaRegion( &rmesa->radeon,
+			 &rmesa->radeon.tcl.aos[0].bo,
+			 &rmesa->radeon.tcl.aos[0].offset,
 			 VB->Count * setup_tab[i].vertex_size * 4, 
 			 4);
 
@@ -421,29 +422,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs )
 
 
    setup_tab[i].emit( ctx, 0, VB->Count, 
-		      rmesa->tcl.indexed_verts.address + 
-		      rmesa->tcl.indexed_verts.start );
+		      rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset);
 
+   //   rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size;
+   rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size;
    rmesa->tcl.vertex_format = setup_tab[i].vertex_format;
-   rmesa->tcl.indexed_verts.aos_start = GET_START( &rmesa->tcl.indexed_verts );
-   rmesa->tcl.indexed_verts.aos_size = setup_tab[i].vertex_size;
-   rmesa->tcl.indexed_verts.aos_stride = setup_tab[i].vertex_size;
-
-   rmesa->tcl.aos_components[0] = &rmesa->tcl.indexed_verts;
-   rmesa->tcl.nr_aos_components = 1;
+   rmesa->radeon.tcl.aos_count = 1;
 }
 
 
-
-void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-#if 0
-   if (RADEON_DEBUG & DEBUG_VERTS) 
-      _tnl_print_vert_flags( __FUNCTION__, newinputs );
-#endif
-
-   if (newinputs) 
-     radeonReleaseDmaRegion( rmesa, &rmesa->tcl.indexed_verts, __FUNCTION__ );
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
new file mode 100644
index 0000000000..f04a07fecd
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
@@ -0,0 +1,423 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "radeon_mipmap_tree.h"
+
+#include <errno.h>
+#include <unistd.h>
+
+#include "main/simple_list.h"
+#include "main/texcompress.h"
+#include "main/texformat.h"
+
+static GLuint radeon_compressed_texture_size(GLcontext *ctx,
+		GLsizei width, GLsizei height, GLsizei depth,
+		GLuint mesaFormat)
+{
+	GLuint size = _mesa_compressed_texture_size(ctx, width, height, depth, mesaFormat);
+
+	if (mesaFormat == MESA_FORMAT_RGB_DXT1 ||
+	    mesaFormat == MESA_FORMAT_RGBA_DXT1) {
+		if (width + 3 < 8)	/* width one block */
+			size = size * 4;
+		else if (width + 3 < 16)
+			size = size * 2;
+	} else {
+		/* DXT3/5, 16 bytes per block */
+	  //		WARN_ONCE("DXT 3/5 suffers from multitexturing problems!\n");
+		if (width + 3 < 8)
+			size = size * 2;
+	}
+
+	return size;
+}
+
+
+static int radeon_compressed_num_bytes(GLuint mesaFormat)
+{
+   int bytes = 0;
+   switch(mesaFormat) {
+     
+   case MESA_FORMAT_RGB_FXT1:
+   case MESA_FORMAT_RGBA_FXT1:
+   case MESA_FORMAT_RGB_DXT1:
+   case MESA_FORMAT_RGBA_DXT1:
+     bytes = 2;
+     break;
+     
+   case MESA_FORMAT_RGBA_DXT3:
+   case MESA_FORMAT_RGBA_DXT5:
+     bytes = 4;
+   default:
+     break;
+   }
+   
+   return bytes;
+}
+
+/**
+ * Compute sizes and fill in offset and blit information for the given
+ * image (determined by \p face and \p level).
+ *
+ * \param curOffset points to the offset at which the image is to be stored
+ * and is updated by this function according to the size of the image.
+ */
+static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree *mt,
+	GLuint face, GLuint level, GLuint* curOffset)
+{
+	radeon_mipmap_level *lvl = &mt->levels[level];
+	uint32_t row_align = rmesa->texture_row_align - 1;
+
+	/* Find image size in bytes */
+	if (mt->compressed) {
+		/* TODO: Is this correct? Need test cases for compressed textures! */
+		lvl->rowstride = (lvl->width * mt->bpp + 63) & ~63;
+		lvl->size = radeon_compressed_texture_size(mt->radeon->glCtx,
+							   lvl->width, lvl->height, lvl->depth, mt->compressed);
+	} else if (mt->target == GL_TEXTURE_RECTANGLE_NV) {
+		lvl->rowstride = (lvl->width * mt->bpp + 63) & ~63;
+		lvl->size = lvl->rowstride * lvl->height;
+	} else if (mt->tilebits & RADEON_TXO_MICRO_TILE) {
+		/* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
+		 * though the actual offset may be different (if texture is less than
+		 * 32 bytes width) to the untiled case */
+		lvl->rowstride = (lvl->width * mt->bpp * 2 + 31) & ~31;
+		lvl->size = lvl->rowstride * ((lvl->height + 1) / 2) * lvl->depth;
+	} else {
+		lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align;
+		lvl->size = lvl->rowstride * lvl->height * lvl->depth;
+	}
+	assert(lvl->size > 0);
+
+	/* All images are aligned to a 32-byte offset */
+	*curOffset = (*curOffset + 0x1f) & ~0x1f;
+	lvl->faces[face].offset = *curOffset;
+	*curOffset += lvl->size;
+
+	if (RADEON_DEBUG & DEBUG_TEXTURE)
+	  fprintf(stderr,
+		  "level %d, face %d: rs:%d %dx%d at %d\n",
+		  level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset);
+}
+
+static GLuint minify(GLuint size, GLuint levels)
+{
+	size = size >> levels;
+	if (size < 1)
+		size = 1;
+	return size;
+}
+
+
+static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_tree *mt)
+{
+	GLuint curOffset;
+	GLuint numLevels;
+	GLuint i;
+	GLuint face;
+
+	numLevels = mt->lastLevel - mt->firstLevel + 1;
+	assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels);
+
+	curOffset = 0;
+	for(face = 0; face < mt->faces; face++) {
+
+		for(i = 0; i < numLevels; i++) {
+			mt->levels[i].width = minify(mt->width0, i);
+			mt->levels[i].height = minify(mt->height0, i);
+			mt->levels[i].depth = minify(mt->depth0, i);
+			compute_tex_image_offset(rmesa, mt, face, i, &curOffset);
+		}
+	}
+
+	/* Note the required size in memory */
+	mt->totalsize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
+}
+
+static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_tree *mt)
+{
+	GLuint curOffset;
+	GLuint numLevels;
+	GLuint i;
+
+	numLevels = mt->lastLevel - mt->firstLevel + 1;
+	assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels);
+
+	curOffset = 0;
+	for(i = 0; i < numLevels; i++) {
+		GLuint face;
+
+		mt->levels[i].width = minify(mt->width0, i);
+		mt->levels[i].height = minify(mt->height0, i);
+		mt->levels[i].depth = minify(mt->depth0, i);
+
+		for(face = 0; face < mt->faces; face++)
+			compute_tex_image_offset(rmesa, mt, face, i, &curOffset);
+	}
+
+	/* Note the required size in memory */
+	mt->totalsize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
+}
+
+/**
+ * Create a new mipmap tree, calculate its layout and allocate memory.
+ */
+radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t,
+		GLenum target, GLuint firstLevel, GLuint lastLevel,
+		GLuint width0, GLuint height0, GLuint depth0,
+		GLuint bpp, GLuint tilebits, GLuint compressed)
+{
+	radeon_mipmap_tree *mt = CALLOC_STRUCT(_radeon_mipmap_tree);
+
+	mt->radeon = rmesa;
+	mt->refcount = 1;
+	mt->t = t;
+	mt->target = target;
+	mt->faces = (target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+	mt->firstLevel = firstLevel;
+	mt->lastLevel = lastLevel;
+	mt->width0 = width0;
+	mt->height0 = height0;
+	mt->depth0 = depth0;
+	mt->bpp = compressed ? radeon_compressed_num_bytes(compressed) : bpp;
+	mt->tilebits = tilebits;
+	mt->compressed = compressed;
+
+	if (rmesa->radeonScreen->chip_family >= CHIP_FAMILY_R300)
+		calculate_miptree_layout_r300(rmesa, mt);
+	else
+		calculate_miptree_layout_r100(rmesa, mt);
+
+#ifdef RADEON_DEBUG_BO
+    mt->bo = radeon_bo_open(rmesa->radeonScreen->bom,
+                            0, mt->totalsize, 1024,
+                            RADEON_GEM_DOMAIN_VRAM,
+                            0,
+                            "MIPMAP TREE");
+#else
+	mt->bo = radeon_bo_open(rmesa->radeonScreen->bom,
+                            0, mt->totalsize, 1024,
+                            RADEON_GEM_DOMAIN_VRAM,
+                            0);
+#endif /* RADEON_DEBUG_BO */
+
+	return mt;
+}
+
+void radeon_miptree_reference(radeon_mipmap_tree *mt)
+{
+	mt->refcount++;
+	assert(mt->refcount > 0);
+}
+
+void radeon_miptree_unreference(radeon_mipmap_tree *mt)
+{
+	if (!mt)
+		return;
+
+	assert(mt->refcount > 0);
+	mt->refcount--;
+	if (!mt->refcount) {
+		radeon_bo_unref(mt->bo);
+		free(mt);
+	}
+}
+
+
+/**
+ * Calculate first and last mip levels for the given texture object,
+ * where the dimensions are taken from the given texture image at
+ * the given level.
+ *
+ * Note: level is the OpenGL level number, which is not necessarily the same
+ * as the first level that is actually present.
+ *
+ * The base level image of the given texture face must be non-null,
+ * or this will fail.
+ */
+static void calculate_first_last_level(struct gl_texture_object *tObj,
+				       GLuint *pfirstLevel, GLuint *plastLevel,
+				       GLuint face, GLuint level)
+{
+	const struct gl_texture_image * const baseImage =
+		tObj->Image[face][level];
+
+	assert(baseImage);
+	
+	/* These must be signed values.  MinLod and MaxLod can be negative numbers,
+	* and having firstLevel and lastLevel as signed prevents the need for
+	* extra sign checks.
+	*/
+	int   firstLevel;
+	int   lastLevel;
+
+	/* Yes, this looks overly complicated, but it's all needed.
+	*/
+	switch (tObj->Target) {
+	case GL_TEXTURE_1D:
+	case GL_TEXTURE_2D:
+	case GL_TEXTURE_3D:
+	case GL_TEXTURE_CUBE_MAP:
+		if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) {
+			/* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL.
+			*/
+			firstLevel = lastLevel = tObj->BaseLevel;
+		} else {
+			firstLevel = tObj->BaseLevel + (GLint)(tObj->MinLod + 0.5);
+			firstLevel = MAX2(firstLevel, tObj->BaseLevel);
+			firstLevel = MIN2(firstLevel, level + baseImage->MaxLog2);
+			lastLevel = tObj->BaseLevel + (GLint)(tObj->MaxLod + 0.5);
+			lastLevel = MAX2(lastLevel, tObj->BaseLevel);
+			lastLevel = MIN2(lastLevel, level + baseImage->MaxLog2);
+			lastLevel = MIN2(lastLevel, tObj->MaxLevel);
+			lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */
+		}
+		break;
+	case GL_TEXTURE_RECTANGLE_NV:
+	case GL_TEXTURE_4D_SGIS:
+		firstLevel = lastLevel = 0;
+		break;
+	default:
+		return;
+	}
+
+	/* save these values */
+	*pfirstLevel = firstLevel;
+	*plastLevel = lastLevel;
+}
+
+
+/**
+ * Checks whether the given miptree can hold the given texture image at the
+ * given face and level.
+ */
+GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt,
+		struct gl_texture_image *texImage, GLuint face, GLuint level)
+{
+	radeon_mipmap_level *lvl;
+
+	if (face >= mt->faces || level < mt->firstLevel || level > mt->lastLevel)
+		return GL_FALSE;
+
+	if (texImage->IsCompressed != mt->compressed)
+		return GL_FALSE;
+
+	if (!texImage->IsCompressed &&
+	    !mt->compressed &&
+	    texImage->TexFormat->TexelBytes != mt->bpp)
+		return GL_FALSE;
+
+	lvl = &mt->levels[level - mt->firstLevel];
+	if (lvl->width != texImage->Width ||
+	    lvl->height != texImage->Height ||
+	    lvl->depth != texImage->Depth)
+		return GL_FALSE;
+
+	return GL_TRUE;
+}
+
+
+/**
+ * Checks whether the given miptree has the right format to store the given texture object.
+ */
+GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj)
+{
+	struct gl_texture_image *firstImage;
+	GLuint compressed;
+	GLuint numfaces = 1;
+	GLuint firstLevel, lastLevel;
+
+	calculate_first_last_level(texObj, &firstLevel, &lastLevel, 0, texObj->BaseLevel);
+	if (texObj->Target == GL_TEXTURE_CUBE_MAP)
+		numfaces = 6;
+
+	firstImage = texObj->Image[0][firstLevel];
+	compressed = firstImage->IsCompressed ? firstImage->TexFormat->MesaFormat : 0;
+
+	return (mt->firstLevel == firstLevel &&
+	        mt->lastLevel == lastLevel &&
+	        mt->width0 == firstImage->Width &&
+	        mt->height0 == firstImage->Height &&
+	        mt->depth0 == firstImage->Depth &&
+	        mt->bpp == firstImage->TexFormat->TexelBytes &&
+	        mt->compressed == compressed);
+}
+
+
+/**
+ * Try to allocate a mipmap tree for the given texture that will fit the
+ * given image in the given position.
+ */
+void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t,
+		struct gl_texture_image *texImage, GLuint face, GLuint level)
+{
+	GLuint compressed = texImage->IsCompressed ? texImage->TexFormat->MesaFormat : 0;
+	GLuint numfaces = 1;
+	GLuint firstLevel, lastLevel;
+
+	assert(!t->mt);
+
+	calculate_first_last_level(&t->base, &firstLevel, &lastLevel, face, level);
+	if (t->base.Target == GL_TEXTURE_CUBE_MAP)
+		numfaces = 6;
+
+	if (level != firstLevel || face >= numfaces)
+		return;
+
+	t->mt = radeon_miptree_create(rmesa, t, t->base.Target,
+		firstLevel, lastLevel,
+		texImage->Width, texImage->Height, texImage->Depth,
+		texImage->TexFormat->TexelBytes, t->tile_bits, compressed);
+}
+
+/* Although we use the image_offset[] array to store relative offsets
+ * to cube faces, Mesa doesn't know anything about this and expects
+ * each cube face to be treated as a separate image.
+ *
+ * These functions present that view to mesa:
+ */
+void
+radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets)
+{
+     if (mt->target != GL_TEXTURE_3D || mt->faces == 1)
+        offsets[0] = 0;
+     else {
+	int i;
+	for (i = 0; i < 6; i++)
+		offsets[i] = mt->levels[level].faces[i].offset;
+     }
+}
+
+GLuint
+radeon_miptree_image_offset(radeon_mipmap_tree *mt,
+			    GLuint face, GLuint level)
+{
+   if (mt->target == GL_TEXTURE_CUBE_MAP_ARB)
+      return (mt->levels[level].faces[face].offset);
+   else
+      return mt->levels[level].faces[0].offset;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
new file mode 100644
index 0000000000..7ece688493
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
@@ -0,0 +1,100 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ *
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __RADEON_MIPMAP_TREE_H_
+#define __RADEON_MIPMAP_TREE_H_
+
+#include "radeon_common.h"
+
+typedef struct _radeon_mipmap_tree radeon_mipmap_tree;
+typedef struct _radeon_mipmap_level radeon_mipmap_level;
+typedef struct _radeon_mipmap_image radeon_mipmap_image;
+
+struct _radeon_mipmap_image {
+	GLuint offset; /** Offset of this image from the start of mipmap tree buffer, in bytes */
+};
+
+struct _radeon_mipmap_level {
+	GLuint width;
+	GLuint height;
+	GLuint depth;
+	GLuint size; /** Size of each image, in bytes */
+	GLuint rowstride; /** in bytes */
+	radeon_mipmap_image faces[6];
+};
+
+/* store the max possible in the miptree */
+#define RADEON_MIPTREE_MAX_TEXTURE_LEVELS 13
+
+/**
+ * A mipmap tree contains texture images in the layout that the hardware
+ * expects.
+ *
+ * The meta-data of mipmap trees is immutable, i.e. you cannot change the
+ * layout on-the-fly; however, the texture contents (i.e. texels) can be
+ * changed.
+ */
+struct _radeon_mipmap_tree {
+	radeonContextPtr radeon;
+	radeonTexObj *t;
+	struct radeon_bo *bo;
+	GLuint refcount;
+
+	GLuint totalsize; /** total size of the miptree, in bytes */
+
+	GLenum target; /** GL_TEXTURE_xxx */
+	GLuint faces; /** # of faces: 6 for cubemaps, 1 otherwise */
+	GLuint firstLevel; /** First mip level stored in this mipmap tree */
+	GLuint lastLevel; /** Last mip level stored in this mipmap tree */
+
+	GLuint width0; /** Width of firstLevel image */
+	GLuint height0; /** Height of firstLevel image */
+	GLuint depth0; /** Depth of firstLevel image */
+
+	GLuint bpp; /** Bytes per texel */
+	GLuint tilebits; /** RADEON_TXO_xxx_TILE */
+	GLuint compressed; /** MESA_FORMAT_xxx indicating a compressed format, or 0 if uncompressed */
+
+	radeon_mipmap_level levels[RADEON_MIPTREE_MAX_TEXTURE_LEVELS];
+};
+
+radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t,
+		GLenum target, GLuint firstLevel, GLuint lastLevel,
+		GLuint width0, GLuint height0, GLuint depth0,
+		GLuint bpp, GLuint tilebits, GLuint compressed);
+void radeon_miptree_reference(radeon_mipmap_tree *mt);
+void radeon_miptree_unreference(radeon_mipmap_tree *mt);
+
+GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt,
+		struct gl_texture_image *texImage, GLuint face, GLuint level);
+GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj);
+void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t,
+			      struct gl_texture_image *texImage, GLuint face, GLuint level);
+GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt,
+				   GLuint face, GLuint level);
+void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets);
+#endif /* __RADEON_MIPMAP_TREE_H_ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.c b/src/mesa/drivers/dri/radeon/radeon_sanity.c
index 6613757fce..bbed838b59 100644
--- a/src/mesa/drivers/dri/radeon/radeon_sanity.c
+++ b/src/mesa/drivers/dri/radeon/radeon_sanity.c
@@ -973,7 +973,7 @@ static int radeon_emit_packet3_cliprect( drm_radeon_cmd_buffer_t *cmdbuf )
 }
 
 
-int radeonSanityCmdBuffer( radeonContextPtr rmesa,
+int radeonSanityCmdBuffer( r100ContextPtr rmesa,
 			   int nbox,
 			   drm_clip_rect_t *boxes )
 {
diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.h b/src/mesa/drivers/dri/radeon/radeon_sanity.h
index 1ec06bc586..f30eb1c4f1 100644
--- a/src/mesa/drivers/dri/radeon/radeon_sanity.h
+++ b/src/mesa/drivers/dri/radeon/radeon_sanity.h
@@ -1,7 +1,7 @@
 #ifndef RADEON_SANITY_H
 #define RADEON_SANITY_H
 
-extern int radeonSanityCmdBuffer( radeonContextPtr rmesa,
+extern int radeonSanityCmdBuffer( r100ContextPtr rmesa,
 				  int nbox,
 				  drm_clip_rect_t *boxes );
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index 791f59826b..355fa8683a 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -35,6 +35,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \author  Gareth Hughes <gareth@valinux.com>
  */
 
+#include <errno.h>
 #include "main/glheader.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
@@ -45,32 +46,43 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_chipset.h"
 #include "radeon_macros.h"
 #include "radeon_screen.h"
+#include "radeon_common.h"
+#include "radeon_span.h"
 #if !RADEON_COMMON
 #include "radeon_context.h"
-#include "radeon_span.h"
 #include "radeon_tex.h"
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
 #include "r200_context.h"
 #include "r200_ioctl.h"
-#include "r200_span.h"
 #include "r200_tex.h"
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
 #include "r300_context.h"
 #include "r300_fragprog.h"
 #include "r300_tex.h"
-#include "radeon_span.h"
+#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+#include "r600_context.h"
+#include "r700_driconf.h" /* +r6/r7 */
+#include "r600_tex.h"     /* +r6/r7 */
 #endif
 
 #include "utils.h"
 #include "vblank.h"
 #include "drirenderbuffer.h"
 
+#include "radeon_bocs_wrapper.h"
+
 #include "GL/internal/dri_interface.h"
 
 /* Radeon configuration
  */
 #include "xmlpool.h"
 
+#define DRI_CONF_COMMAND_BUFFER_SIZE(def,min,max) \
+DRI_CONF_OPT_BEGIN_V(command_buffer_size,int,def, # min ":" # max ) \
+        DRI_CONF_DESC(en,"Size of command buffer (in KB)") \
+        DRI_CONF_DESC(de,"Grösse des Befehlspuffers (in KB)") \
+DRI_CONF_OPT_END
+
 #if !RADEON_COMMON	/* R100 */
 PUBLIC const char __driConfigOptions[] =
 DRI_CONF_BEGIN
@@ -80,6 +92,7 @@ DRI_CONF_BEGIN
         DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
         DRI_CONF_MAX_TEXTURE_UNITS(3,2,3)
         DRI_CONF_HYPERZ(false)
+        DRI_CONF_COMMAND_BUFFER_SIZE(8, 8, 32)
     DRI_CONF_SECTION_END
     DRI_CONF_SECTION_QUALITY
         DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB)
@@ -95,7 +108,7 @@ DRI_CONF_BEGIN
         DRI_CONF_NO_RAST(false)
     DRI_CONF_SECTION_END
 DRI_CONF_END;
-static const GLuint __driNConfigOptions = 14;
+static const GLuint __driNConfigOptions = 15;
 
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
 
@@ -107,6 +120,7 @@ DRI_CONF_BEGIN
         DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
         DRI_CONF_MAX_TEXTURE_UNITS(6,2,6)
         DRI_CONF_HYPERZ(false)
+        DRI_CONF_COMMAND_BUFFER_SIZE(8, 8, 32)
     DRI_CONF_SECTION_END
     DRI_CONF_SECTION_QUALITY
         DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB)
@@ -126,7 +140,7 @@ DRI_CONF_BEGIN
         DRI_CONF_NV_VERTEX_PROGRAM(false)
     DRI_CONF_SECTION_END
 DRI_CONF_END;
-static const GLuint __driNConfigOptions = 16;
+static const GLuint __driNConfigOptions = 17;
 
 extern const struct dri_extension blend_extensions[];
 extern const struct dri_extension ARB_vp_extension[];
@@ -134,7 +148,7 @@ extern const struct dri_extension NV_vp_extension[];
 extern const struct dri_extension ATI_fs_extension[];
 extern const struct dri_extension point_extensions[];
 
-#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+#elif RADEON_COMMON && (defined(RADEON_COMMON_FOR_R300) || defined(RADEON_COMMON_FOR_R600))
 
 /* TODO: integrate these into xmlpool.h! */
 #define DRI_CONF_MAX_TEXTURE_IMAGE_UNITS(def,min,max) \
@@ -149,11 +163,7 @@ DRI_CONF_OPT_BEGIN_V(texture_coord_units,int,def, # min ":" # max ) \
         DRI_CONF_DESC(de,"Anzahl der Texturkoordinateneinheiten") \
 DRI_CONF_OPT_END
 
-#define DRI_CONF_COMMAND_BUFFER_SIZE(def,min,max) \
-DRI_CONF_OPT_BEGIN_V(command_buffer_size,int,def, # min ":" # max ) \
-        DRI_CONF_DESC(en,"Size of command buffer (in KB)") \
-        DRI_CONF_DESC(de,"Grösse des Befehlspuffers (in KB)") \
-DRI_CONF_OPT_END
+
 
 #define DRI_CONF_DISABLE_S3TC(def) \
 DRI_CONF_OPT_BEGIN(disable_s3tc,bool,def) \
@@ -209,7 +219,6 @@ static const GLuint __driNConfigOptions = 17;
 extern const struct dri_extension gl_20_extension[];
 
 #ifndef RADEON_DEBUG
-int RADEON_DEBUG = 0;
 
 static const struct dri_debug_control debug_control[] = {
 	{"fall", DEBUG_FALLBACKS},
@@ -236,19 +245,36 @@ static const struct dri_debug_control debug_control[] = {
 #endif /* RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) */
 
 extern const struct dri_extension card_extensions[];
+extern const struct dri_extension mm_extensions[];
 
 static int getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo );
 
 static int
-radeonGetParam(int fd, int param, void *value)
+radeonGetParam(__DRIscreenPrivate *sPriv, int param, void *value)
 {
   int ret;
-  drm_radeon_getparam_t gp;
-
-  gp.param = param;
-  gp.value = value;
+  drm_radeon_getparam_t gp = { 0 };
+  struct drm_radeon_info info = { 0 };
+
+  if (sPriv->drm_version.major >= 2) {
+      info.value = (uint64_t)value;
+      switch (param) {
+      case RADEON_PARAM_DEVICE_ID:
+          info.request = RADEON_INFO_DEVICE_ID;
+          break;
+      case RADEON_PARAM_NUM_GB_PIPES:
+          info.request = RADEON_INFO_NUM_GB_PIPES;
+          break;
+      default:
+          return -EINVAL;
+      }
+      ret = drmCommandWriteRead(sPriv->fd, DRM_RADEON_INFO, &info, sizeof(info));
+  } else {
+      gp.param = param;
+      gp.value = value;
 
-  ret = drmCommandWriteRead( fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+      ret = drmCommandWriteRead(sPriv->fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+  }
   return ret;
 }
 
@@ -335,6 +361,12 @@ static const __DRItexOffsetExtension radeonTexOffsetExtension = {
     { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
     radeonSetTexOffset,
 };
+
+static const __DRItexBufferExtension radeonTexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   radeonSetTexBuffer,
+   radeonSetTexBuffer2,
+};
 #endif
 
 #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
@@ -349,6 +381,12 @@ static const __DRItexOffsetExtension r200texOffsetExtension = {
     { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
    r200SetTexOffset,
 };
+
+static const __DRItexBufferExtension r200TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   r200SetTexBuffer,
+   r200SetTexBuffer2,
+};
 #endif
 
 #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
@@ -356,137 +394,32 @@ static const __DRItexOffsetExtension r300texOffsetExtension = {
     { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
    r300SetTexOffset,
 };
-#endif
 
-/* Create the device specific screen private data struct.
- */
-static radeonScreenPtr
-radeonCreateScreen( __DRIscreenPrivate *sPriv )
-{
-   radeonScreenPtr screen;
-   RADEONDRIPtr dri_priv = (RADEONDRIPtr)sPriv->pDevPriv;
-   unsigned char *RADEONMMIO;
-   int i;
-   int ret;
-   uint32_t temp = 0;
-
-   if (sPriv->devPrivSize != sizeof(RADEONDRIRec)) {
-      fprintf(stderr,"\nERROR!  sizeof(RADEONDRIRec) does not match passed size from device driver\n");
-      return GL_FALSE;
-   }
-
-   /* Allocate the private area */
-   screen = (radeonScreenPtr) CALLOC( sizeof(*screen) );
-   if ( !screen ) {
-      __driUtilMessage("%s: Could not allocate memory for screen structure",
-		       __FUNCTION__);
-      return NULL;
-   }
-
-#if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
-	RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control);
+static const __DRItexBufferExtension r300TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   r300SetTexBuffer,
+   r300SetTexBuffer2,
+};
 #endif
 
-   /* parse information in __driConfigOptions */
-   driParseOptionInfo (&screen->optionCache,
-		       __driConfigOptions, __driNConfigOptions);
-
-   /* This is first since which regions we map depends on whether or
-    * not we are using a PCI card.
-    */
-   screen->card_type = (dri_priv->IsPCI ? RADEON_CARD_PCI : RADEON_CARD_AGP);
-   {
-      int ret;
-      ret = radeonGetParam( sPriv->fd, RADEON_PARAM_GART_BUFFER_OFFSET,
-			    &screen->gart_buffer_offset);
-
-      if (ret) {
-	 FREE( screen );
-	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BUFFER_OFFSET): %d\n", ret);
-	 return NULL;
-      }
-
-      ret = radeonGetParam( sPriv->fd, RADEON_PARAM_GART_BASE,
-			    &screen->gart_base);
-      if (ret) {
-	 FREE( screen );
-	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BASE): %d\n", ret);
-	 return NULL;
-      }
-
-      ret = radeonGetParam( sPriv->fd, RADEON_PARAM_IRQ_NR,
-			    &screen->irq);
-      if (ret) {
-	 FREE( screen );
-	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_IRQ_NR): %d\n", ret);
-	 return NULL;
-      }
-      screen->drmSupportsCubeMapsR200 = (sPriv->drm_version.minor >= 7);
-      screen->drmSupportsBlendColor = (sPriv->drm_version.minor >= 11);
-      screen->drmSupportsTriPerf = (sPriv->drm_version.minor >= 16);
-      screen->drmSupportsFragShader = (sPriv->drm_version.minor >= 18);
-      screen->drmSupportsPointSprites = (sPriv->drm_version.minor >= 13);
-      screen->drmSupportsCubeMapsR100 = (sPriv->drm_version.minor >= 15);
-      screen->drmSupportsVertexProgram = (sPriv->drm_version.minor >= 25);
-   }
-
-   screen->mmio.handle = dri_priv->registerHandle;
-   screen->mmio.size   = dri_priv->registerSize;
-   if ( drmMap( sPriv->fd,
-		screen->mmio.handle,
-		screen->mmio.size,
-		&screen->mmio.map ) ) {
-      FREE( screen );
-      __driUtilMessage("%s: drmMap failed\n", __FUNCTION__ );
-      return NULL;
-   }
-
-   RADEONMMIO = screen->mmio.map;
-
-   screen->status.handle = dri_priv->statusHandle;
-   screen->status.size   = dri_priv->statusSize;
-   if ( drmMap( sPriv->fd,
-		screen->status.handle,
-		screen->status.size,
-		&screen->status.map ) ) {
-      drmUnmap( screen->mmio.map, screen->mmio.size );
-      FREE( screen );
-      __driUtilMessage("%s: drmMap (2) failed\n", __FUNCTION__ );
-      return NULL;
-   }
-   screen->scratch = (__volatile__ uint32_t *)
-      ((GLubyte *)screen->status.map + RADEON_SCRATCH_REG_OFFSET);
-
-   screen->buffers = drmMapBufs( sPriv->fd );
-   if ( !screen->buffers ) {
-      drmUnmap( screen->status.map, screen->status.size );
-      drmUnmap( screen->mmio.map, screen->mmio.size );
-      FREE( screen );
-      __driUtilMessage("%s: drmMapBufs failed\n", __FUNCTION__ );
-      return NULL;
-   }
-
-   if ( dri_priv->gartTexHandle && dri_priv->gartTexMapSize ) {
-      screen->gartTextures.handle = dri_priv->gartTexHandle;
-      screen->gartTextures.size   = dri_priv->gartTexMapSize;
-      if ( drmMap( sPriv->fd,
-		   screen->gartTextures.handle,
-		   screen->gartTextures.size,
-		   (drmAddressPtr)&screen->gartTextures.map ) ) {
-	 drmUnmapBufs( screen->buffers );
-	 drmUnmap( screen->status.map, screen->status.size );
-	 drmUnmap( screen->mmio.map, screen->mmio.size );
-	 FREE( screen );
-	 __driUtilMessage("%s: drmMap failed for GART texture area\n", __FUNCTION__);
-	 return NULL;
-      }
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+static const __DRItexOffsetExtension r600texOffsetExtension = {
+    { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION },
+   r600SetTexOffset, /* +r6/r7 */
+};
 
-      screen->gart_texture_offset = dri_priv->gartTexOffset + screen->gart_base;
-   }
+static const __DRItexBufferExtension r600TexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   r600SetTexBuffer,  /* +r6/r7 */
+   r600SetTexBuffer2, /* +r6/r7 */
+};
+#endif
 
+static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id)
+{
+   screen->device_id = device_id;
    screen->chip_flags = 0;
-   /* XXX: add more chipsets */
-   switch ( dri_priv->deviceID ) {
+   switch ( device_id ) {
    case PCI_CHIP_RADEON_LY:
    case PCI_CHIP_RADEON_LZ:
    case PCI_CHIP_RADEON_QY:
@@ -561,7 +494,6 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
       screen->chip_family = CHIP_FAMILY_RS300;
       break;
 
-
    case PCI_CHIP_R300_AD:
    case PCI_CHIP_R300_AE:
    case PCI_CHIP_R300_AF:
@@ -819,11 +751,309 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
       screen->chip_flags = RADEON_CHIPSET_TCL;
       break;
 
+   case PCI_CHIP_R600_9400:
+   case PCI_CHIP_R600_9401:
+   case PCI_CHIP_R600_9402:
+   case PCI_CHIP_R600_9403:
+   case PCI_CHIP_R600_9405:
+   case PCI_CHIP_R600_940A:
+   case PCI_CHIP_R600_940B:
+   case PCI_CHIP_R600_940F:
+      screen->chip_family = CHIP_FAMILY_R600;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV610_94C0:
+   case PCI_CHIP_RV610_94C1:
+   case PCI_CHIP_RV610_94C3:
+   case PCI_CHIP_RV610_94C4:
+   case PCI_CHIP_RV610_94C5:
+   case PCI_CHIP_RV610_94C6:
+   case PCI_CHIP_RV610_94C7:
+   case PCI_CHIP_RV610_94C8:
+   case PCI_CHIP_RV610_94C9:
+   case PCI_CHIP_RV610_94CB:
+   case PCI_CHIP_RV610_94CC:
+   case PCI_CHIP_RV610_94CD:
+      screen->chip_family = CHIP_FAMILY_RV610;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV630_9580:
+   case PCI_CHIP_RV630_9581:
+   case PCI_CHIP_RV630_9583:
+   case PCI_CHIP_RV630_9586:
+   case PCI_CHIP_RV630_9587:
+   case PCI_CHIP_RV630_9588:
+   case PCI_CHIP_RV630_9589:
+   case PCI_CHIP_RV630_958A:
+   case PCI_CHIP_RV630_958B:
+   case PCI_CHIP_RV630_958C:
+   case PCI_CHIP_RV630_958D:
+   case PCI_CHIP_RV630_958E:
+   case PCI_CHIP_RV630_958F:
+      screen->chip_family = CHIP_FAMILY_RV630;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV670_9500:
+   case PCI_CHIP_RV670_9501:
+   case PCI_CHIP_RV670_9504:
+   case PCI_CHIP_RV670_9505:
+   case PCI_CHIP_RV670_9506:
+   case PCI_CHIP_RV670_9507:
+   case PCI_CHIP_RV670_9508:
+   case PCI_CHIP_RV670_9509:
+   case PCI_CHIP_RV670_950F:
+   case PCI_CHIP_RV670_9511:
+   case PCI_CHIP_RV670_9515:
+   case PCI_CHIP_RV670_9517:
+   case PCI_CHIP_RV670_9519:
+      screen->chip_family = CHIP_FAMILY_RV670;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV620_95C0:
+   case PCI_CHIP_RV620_95C2:
+   case PCI_CHIP_RV620_95C4:
+   case PCI_CHIP_RV620_95C5:
+   case PCI_CHIP_RV620_95C6:
+   case PCI_CHIP_RV620_95C7:
+   case PCI_CHIP_RV620_95C9:
+   case PCI_CHIP_RV620_95CC:
+   case PCI_CHIP_RV620_95CD:
+   case PCI_CHIP_RV620_95CE:
+   case PCI_CHIP_RV620_95CF:
+      screen->chip_family = CHIP_FAMILY_RV620;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV635_9590:
+   case PCI_CHIP_RV635_9591:
+   case PCI_CHIP_RV635_9593:
+   case PCI_CHIP_RV635_9595:
+   case PCI_CHIP_RV635_9596:
+   case PCI_CHIP_RV635_9597:
+   case PCI_CHIP_RV635_9598:
+   case PCI_CHIP_RV635_9599:
+   case PCI_CHIP_RV635_959B:
+      screen->chip_family = CHIP_FAMILY_RV635;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RS780_9610:
+   case PCI_CHIP_RS780_9611:
+   case PCI_CHIP_RS780_9612:
+   case PCI_CHIP_RS780_9613:
+   case PCI_CHIP_RS780_9614:
+   case PCI_CHIP_RS780_9615:
+   case PCI_CHIP_RS780_9616:
+      screen->chip_family = CHIP_FAMILY_RS780;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV770_9440:
+   case PCI_CHIP_RV770_9441:
+   case PCI_CHIP_RV770_9442:
+   case PCI_CHIP_RV770_9444:
+   case PCI_CHIP_RV770_9446:
+   case PCI_CHIP_RV770_944A:
+   case PCI_CHIP_RV770_944B:
+   case PCI_CHIP_RV770_944C:
+   case PCI_CHIP_RV770_944E:
+   case PCI_CHIP_RV770_9450:
+   case PCI_CHIP_RV770_9452:
+   case PCI_CHIP_RV770_9456:
+   case PCI_CHIP_RV770_945A:
+   case PCI_CHIP_RV770_945B:
+   case PCI_CHIP_RV790_9460:
+   case PCI_CHIP_RV790_9462:
+   case PCI_CHIP_RV770_946A:
+   case PCI_CHIP_RV770_946B:
+   case PCI_CHIP_RV770_947A:
+   case PCI_CHIP_RV770_947B:
+      screen->chip_family = CHIP_FAMILY_RV770;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV730_9487:
+   case PCI_CHIP_RV730_9489:
+   case PCI_CHIP_RV730_948F:
+   case PCI_CHIP_RV730_9490:
+   case PCI_CHIP_RV730_9491:
+   case PCI_CHIP_RV730_9498:
+   case PCI_CHIP_RV730_949C:
+   case PCI_CHIP_RV730_949E:
+   case PCI_CHIP_RV730_949F:
+      screen->chip_family = CHIP_FAMILY_RV730;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV710_9540:
+   case PCI_CHIP_RV710_9541:
+   case PCI_CHIP_RV710_9542:
+   case PCI_CHIP_RV710_954E:
+   case PCI_CHIP_RV710_954F:
+   case PCI_CHIP_RV710_9552:
+   case PCI_CHIP_RV710_9553:
+   case PCI_CHIP_RV710_9555:
+      screen->chip_family = CHIP_FAMILY_RV710;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
+   case PCI_CHIP_RV740_94A0:
+   case PCI_CHIP_RV740_94A1:
+   case PCI_CHIP_RV740_94B1:
+   case PCI_CHIP_RV740_94B3:
+   case PCI_CHIP_RV740_94B5:
+      screen->chip_family = CHIP_FAMILY_RV740;
+      screen->chip_flags = RADEON_CHIPSET_TCL;
+      break;
+
    default:
       fprintf(stderr, "unknown chip id 0x%x, can't guess.\n",
-	      dri_priv->deviceID);
+	      device_id);
+      return -1;
+   }
+
+   return 0;
+}
+
+
+/* Create the device specific screen private data struct.
+ */
+static radeonScreenPtr
+radeonCreateScreen( __DRIscreenPrivate *sPriv )
+{
+   radeonScreenPtr screen;
+   RADEONDRIPtr dri_priv = (RADEONDRIPtr)sPriv->pDevPriv;
+   unsigned char *RADEONMMIO = NULL;
+   int i;
+   int ret;
+   uint32_t temp = 0;
+
+   if (sPriv->devPrivSize != sizeof(RADEONDRIRec)) {
+      fprintf(stderr,"\nERROR!  sizeof(RADEONDRIRec) does not match passed size from device driver\n");
+      return GL_FALSE;
+   }
+
+   /* Allocate the private area */
+   screen = (radeonScreenPtr) CALLOC( sizeof(*screen) );
+   if ( !screen ) {
+      __driUtilMessage("%s: Could not allocate memory for screen structure",
+		       __FUNCTION__);
       return NULL;
    }
+
+#if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+	RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control);
+#endif
+   /* parse information in __driConfigOptions */
+   driParseOptionInfo (&screen->optionCache,
+		       __driConfigOptions, __driNConfigOptions);
+
+   /* This is first since which regions we map depends on whether or
+    * not we are using a PCI card.
+    */
+   screen->card_type = (dri_priv->IsPCI ? RADEON_CARD_PCI : RADEON_CARD_AGP);
+   {
+      int ret;
+
+      ret = radeonGetParam(sPriv, RADEON_PARAM_GART_BUFFER_OFFSET,
+			    &screen->gart_buffer_offset);
+
+      if (ret) {
+	 FREE( screen );
+	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BUFFER_OFFSET): %d\n", ret);
+	 return NULL;
+      }
+
+      ret = radeonGetParam(sPriv, RADEON_PARAM_GART_BASE,
+			    &screen->gart_base);
+      if (ret) {
+	 FREE( screen );
+	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_GART_BASE): %d\n", ret);
+	 return NULL;
+      }
+
+      ret = radeonGetParam(sPriv, RADEON_PARAM_IRQ_NR,
+			    &screen->irq);
+      if (ret) {
+	 FREE( screen );
+	 fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_IRQ_NR): %d\n", ret);
+	 return NULL;
+      }
+      screen->drmSupportsCubeMapsR200 = (sPriv->drm_version.minor >= 7);
+      screen->drmSupportsBlendColor = (sPriv->drm_version.minor >= 11);
+      screen->drmSupportsTriPerf = (sPriv->drm_version.minor >= 16);
+      screen->drmSupportsFragShader = (sPriv->drm_version.minor >= 18);
+      screen->drmSupportsPointSprites = (sPriv->drm_version.minor >= 13);
+      screen->drmSupportsCubeMapsR100 = (sPriv->drm_version.minor >= 15);
+      screen->drmSupportsVertexProgram = (sPriv->drm_version.minor >= 25);
+   }
+
+   ret = radeon_set_screen_flags(screen, dri_priv->deviceID);
+   if (ret == -1)
+     return NULL;
+
+   screen->mmio.handle = dri_priv->registerHandle;
+   screen->mmio.size   = dri_priv->registerSize;
+   if ( drmMap( sPriv->fd,
+		screen->mmio.handle,
+		screen->mmio.size,
+		&screen->mmio.map ) ) {
+     FREE( screen );
+     __driUtilMessage("%s: drmMap failed\n", __FUNCTION__ );
+     return NULL;
+   }
+
+   RADEONMMIO = screen->mmio.map;
+
+   screen->status.handle = dri_priv->statusHandle;
+   screen->status.size   = dri_priv->statusSize;
+   if ( drmMap( sPriv->fd,
+		screen->status.handle,
+		screen->status.size,
+		&screen->status.map ) ) {
+     drmUnmap( screen->mmio.map, screen->mmio.size );
+     FREE( screen );
+     __driUtilMessage("%s: drmMap (2) failed\n", __FUNCTION__ );
+     return NULL;
+   }
+   if (screen->chip_family < CHIP_FAMILY_R600)
+	   screen->scratch = (__volatile__ uint32_t *)
+		   ((GLubyte *)screen->status.map + RADEON_SCRATCH_REG_OFFSET);
+   else
+	   screen->scratch = (__volatile__ uint32_t *)
+		   ((GLubyte *)screen->status.map + RADEON_SCRATCH_REG_OFFSET);
+
+   screen->buffers = drmMapBufs( sPriv->fd );
+   if ( !screen->buffers ) {
+     drmUnmap( screen->status.map, screen->status.size );
+     drmUnmap( screen->mmio.map, screen->mmio.size );
+     FREE( screen );
+     __driUtilMessage("%s: drmMapBufs failed\n", __FUNCTION__ );
+     return NULL;
+   }
+
+   if ( dri_priv->gartTexHandle && dri_priv->gartTexMapSize ) {
+     screen->gartTextures.handle = dri_priv->gartTexHandle;
+     screen->gartTextures.size   = dri_priv->gartTexMapSize;
+     if ( drmMap( sPriv->fd,
+		  screen->gartTextures.handle,
+		  screen->gartTextures.size,
+		  (drmAddressPtr)&screen->gartTextures.map ) ) {
+       drmUnmapBufs( screen->buffers );
+       drmUnmap( screen->status.map, screen->status.size );
+       drmUnmap( screen->mmio.map, screen->mmio.size );
+       FREE( screen );
+       __driUtilMessage("%s: drmMap failed for GART texture area\n", __FUNCTION__);
+       return NULL;
+    }
+
+     screen->gart_texture_offset = dri_priv->gartTexOffset + screen->gart_base;
+   }
+
    if ((screen->chip_family == CHIP_FAMILY_R350 || screen->chip_family == CHIP_FAMILY_R300) &&
        sPriv->ddx_version.minor < 2) {
       fprintf(stderr, "xf86-video-ati-6.6.2 or newer needed for Radeon 9500/9700/9800 cards.\n");
@@ -836,35 +1066,57 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
    }
 
    if (getenv("R300_NO_TCL"))
-     screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+	   screen->chip_flags &= ~RADEON_CHIPSET_TCL;
 
    if (screen->chip_family <= CHIP_FAMILY_RS200)
-      screen->chip_flags |= RADEON_CLASS_R100;
+	   screen->chip_flags |= RADEON_CLASS_R100;
    else if (screen->chip_family <= CHIP_FAMILY_RV280)
-      screen->chip_flags |= RADEON_CLASS_R200;
+	   screen->chip_flags |= RADEON_CLASS_R200;
+   else if (screen->chip_family <= CHIP_FAMILY_RV570)
+	   screen->chip_flags |= RADEON_CLASS_R300;
    else
-      screen->chip_flags |= RADEON_CLASS_R300;
+	   screen->chip_flags |= RADEON_CLASS_R600;
 
    screen->cpp = dri_priv->bpp / 8;
    screen->AGPMode = dri_priv->AGPMode;
 
-   ret = radeonGetParam( sPriv->fd, RADEON_PARAM_FB_LOCATION,
-                         &temp);
-   if (ret) {
-       if (screen->chip_family < CHIP_FAMILY_RS600)
-	   screen->fbLocation      = ( INREG( RADEON_MC_FB_LOCATION ) & 0xffff) << 16;
-       else {
-           FREE( screen );
-           fprintf(stderr, "Unable to get fb location need newer drm\n");
-           return NULL;
+   ret = radeonGetParam(sPriv, RADEON_PARAM_FB_LOCATION, &temp);
+
+   /* +r6/r7 */
+   if(screen->chip_family >= CHIP_FAMILY_R600)
+   {
+       if (ret) 
+       {
+            FREE( screen );
+            fprintf(stderr, "Unable to get fb location need newer drm\n");
+            return NULL;
        }
-   } else {
-       screen->fbLocation = (temp & 0xffff) << 16;
+       else
+       {
+            screen->fbLocation = (temp & 0xffff) << 24;
+       }
+   }
+   else
+   {
+        if (ret) 
+        {
+            if (screen->chip_family < CHIP_FAMILY_RS600 && !screen->kernel_mm)
+	            screen->fbLocation      = ( INREG( RADEON_MC_FB_LOCATION ) & 0xffff) << 16;
+            else 
+            {
+                FREE( screen );
+                fprintf(stderr, "Unable to get fb location need newer drm\n");
+                return NULL;
+            }
+        } 
+        else 
+        {
+            screen->fbLocation = (temp & 0xffff) << 16;
+        }
    }
 
-   if (screen->chip_family >= CHIP_FAMILY_R300) {
-       ret = radeonGetParam( sPriv->fd, RADEON_PARAM_NUM_GB_PIPES,
-			     &temp);
+   if (IS_R300_CLASS(screen)) {
+       ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_GB_PIPES, &temp);
        if (ret) {
 	   fprintf(stderr, "Unable to get num_pipes, need newer drm\n");
 	   switch (screen->chip_family) {
@@ -901,7 +1153,6 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
        default:
 	   break;
        }
-
    }
 
    if ( sPriv->drm_version.minor >= 10 ) {
@@ -971,7 +1222,7 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
 
 #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
    if (IS_R200_CLASS(screen))
-       screen->extensions[i++] = &r200AllocateExtension.base;
+      screen->extensions[i++] = &r200AllocateExtension.base;
 
    screen->extensions[i++] = &r200texOffsetExtension.base;
 #endif
@@ -980,11 +1231,168 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
    screen->extensions[i++] = &r300texOffsetExtension.base;
 #endif
 
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   screen->extensions[i++] = &r600texOffsetExtension.base;
+#endif
+
    screen->extensions[i++] = NULL;
    sPriv->extensions = screen->extensions;
 
    screen->driScreen = sPriv;
    screen->sarea_priv_offset = dri_priv->sarea_priv_offset;
+   screen->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA +
+					       screen->sarea_priv_offset);
+
+   screen->bom = radeon_bo_manager_legacy_ctor(screen);
+   if (screen->bom == NULL) {
+     free(screen);
+     return NULL;
+   }
+
+   return screen;
+}
+
+static radeonScreenPtr
+radeonCreateScreen2(__DRIscreenPrivate *sPriv)
+{
+   radeonScreenPtr screen;
+   int i;
+   int ret;
+   uint32_t device_id = 0;
+   uint32_t temp = 0;
+
+   /* Allocate the private area */
+   screen = (radeonScreenPtr) CALLOC( sizeof(*screen) );
+   if ( !screen ) {
+      __driUtilMessage("%s: Could not allocate memory for screen structure",
+		       __FUNCTION__);
+      fprintf(stderr, "leaving here\n");
+      return NULL;
+   }
+
+#if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+	RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control);
+#endif
+
+   /* parse information in __driConfigOptions */
+   driParseOptionInfo (&screen->optionCache,
+		       __driConfigOptions, __driNConfigOptions);
+
+   screen->kernel_mm = 1;
+   screen->chip_flags = 0;
+
+   /* if we have kms we can support all of these */
+   screen->drmSupportsCubeMapsR200 = 1;
+   screen->drmSupportsBlendColor = 1;
+   screen->drmSupportsTriPerf = 1;
+   screen->drmSupportsFragShader = 1;
+   screen->drmSupportsPointSprites = 1;
+   screen->drmSupportsCubeMapsR100 = 1;
+   screen->drmSupportsVertexProgram = 1;
+   screen->irq = 1;
+
+   ret = radeonGetParam(sPriv, RADEON_PARAM_DEVICE_ID, &device_id);
+   if (ret) {
+     FREE( screen );
+     fprintf(stderr, "drm_radeon_getparam_t (RADEON_PARAM_DEVICE_ID): %d\n", ret);
+     return NULL;
+   }
+
+   ret = radeon_set_screen_flags(screen, device_id);
+   if (ret == -1)
+     return NULL;
+
+   if (getenv("R300_NO_TCL"))
+	   screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+
+   if (screen->chip_family <= CHIP_FAMILY_RS200)
+	   screen->chip_flags |= RADEON_CLASS_R100;
+   else if (screen->chip_family <= CHIP_FAMILY_RV280)
+	   screen->chip_flags |= RADEON_CLASS_R200;
+   else if (screen->chip_family <= CHIP_FAMILY_RV570)
+	   screen->chip_flags |= RADEON_CLASS_R300;
+   else
+	   screen->chip_flags |= RADEON_CLASS_R600;
+
+   if (IS_R300_CLASS(screen)) {
+       ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_GB_PIPES, &temp);
+       if (ret) {
+	   fprintf(stderr, "Unable to get num_pipes, need newer drm\n");
+	   switch (screen->chip_family) {
+	   case CHIP_FAMILY_R300:
+	   case CHIP_FAMILY_R350:
+	       screen->num_gb_pipes = 2;
+	       break;
+	   case CHIP_FAMILY_R420:
+	   case CHIP_FAMILY_R520:
+	   case CHIP_FAMILY_R580:
+	   case CHIP_FAMILY_RV560:
+	   case CHIP_FAMILY_RV570:
+	       screen->num_gb_pipes = 4;
+	       break;
+	   case CHIP_FAMILY_RV350:
+	   case CHIP_FAMILY_RV515:
+	   case CHIP_FAMILY_RV530:
+	   case CHIP_FAMILY_RV410:
+	   default:
+	       screen->num_gb_pipes = 1;
+	       break;
+	   }
+       } else {
+	   screen->num_gb_pipes = temp;
+       }
+
+       /* pipe overrides */
+       switch (device_id) {
+       case PCI_CHIP_R300_AD: /* 9500 with 1 quadpipe verified by: Reid Linnemann <lreid@cs.okstate.edu> */
+       case PCI_CHIP_RV410_5E4C: /* RV410 SE only have 1 quadpipe */
+       case PCI_CHIP_RV410_5E4F: /* RV410 SE only have 1 quadpipe */
+	   screen->num_gb_pipes = 1;
+	   break;
+       default:
+	   break;
+       }
+
+   }
+
+   i = 0;
+   screen->extensions[i++] = &driCopySubBufferExtension.base;
+   screen->extensions[i++] = &driFrameTrackingExtension.base;
+   screen->extensions[i++] = &driReadDrawableExtension;
+
+   if ( screen->irq != 0 ) {
+       screen->extensions[i++] = &driSwapControlExtension.base;
+       screen->extensions[i++] = &driMediaStreamCounterExtension.base;
+   }
+
+#if !RADEON_COMMON
+   screen->extensions[i++] = &radeonTexBufferExtension.base;
+#endif
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
+   if (IS_R200_CLASS(screen))
+       screen->extensions[i++] = &r200AllocateExtension.base;
+
+   screen->extensions[i++] = &r200TexBufferExtension.base;
+#endif
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
+   screen->extensions[i++] = &r300TexBufferExtension.base;
+#endif
+
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   screen->extensions[i++] = &r600TexBufferExtension.base;
+#endif
+
+   screen->extensions[i++] = NULL;
+   sPriv->extensions = screen->extensions;
+
+   screen->driScreen = sPriv;
+   screen->bom = radeon_bo_manager_gem_ctor(sPriv->fd);
+   if (screen->bom == NULL) {
+       free(screen);
+       return NULL;
+   }
    return screen;
 }
 
@@ -993,23 +1401,32 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv )
 static void
 radeonDestroyScreen( __DRIscreenPrivate *sPriv )
 {
-   radeonScreenPtr screen = (radeonScreenPtr)sPriv->private;
+    radeonScreenPtr screen = (radeonScreenPtr)sPriv->private;
 
-   if (!screen)
-      return;
+    if (!screen)
+        return;
 
-   if ( screen->gartTextures.map ) {
-      drmUnmap( screen->gartTextures.map, screen->gartTextures.size );
-   }
-   drmUnmapBufs( screen->buffers );
-   drmUnmap( screen->status.map, screen->status.size );
-   drmUnmap( screen->mmio.map, screen->mmio.size );
+    if (screen->kernel_mm) {
+#ifdef RADEON_BO_TRACK
+        radeon_tracker_print(&screen->bom->tracker, stderr);
+#endif
+        radeon_bo_manager_gem_dtor(screen->bom);
+    } else {
+        radeon_bo_manager_legacy_dtor(screen->bom);
+
+        if ( screen->gartTextures.map ) {
+            drmUnmap( screen->gartTextures.map, screen->gartTextures.size );
+        }
+        drmUnmapBufs( screen->buffers );
+        drmUnmap( screen->status.map, screen->status.size );
+        drmUnmap( screen->mmio.map, screen->mmio.size );
+    }
 
-   /* free all option information */
-   driDestroyOptionInfo (&screen->optionCache);
+    /* free all option information */
+    driDestroyOptionInfo (&screen->optionCache);
 
-   FREE( screen );
-   sPriv->private = NULL;
+    FREE( screen );
+    sPriv->private = NULL;
 }
 
 
@@ -1018,16 +1435,21 @@ radeonDestroyScreen( __DRIscreenPrivate *sPriv )
 static GLboolean
 radeonInitDriver( __DRIscreenPrivate *sPriv )
 {
-   sPriv->private = (void *) radeonCreateScreen( sPriv );
-   if ( !sPriv->private ) {
-      radeonDestroyScreen( sPriv );
-      return GL_FALSE;
-   }
+    if (sPriv->dri2.enabled) {
+        sPriv->private = (void *) radeonCreateScreen2( sPriv );
+    } else {
+        sPriv->private = (void *) radeonCreateScreen( sPriv );
+    }
+    if ( !sPriv->private ) {
+        radeonDestroyScreen( sPriv );
+        return GL_FALSE;
+    }
 
-   return GL_TRUE;
+    return GL_TRUE;
 }
 
 
+
 /**
  * Create the Mesa framebuffer and renderbuffers for a given window/drawable.
  *
@@ -1040,101 +1462,111 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv,
                     const __GLcontextModes *mesaVis,
                     GLboolean isPixmap )
 {
-   radeonScreenPtr screen = (radeonScreenPtr) driScrnPriv->private;
+    radeonScreenPtr screen = (radeonScreenPtr) driScrnPriv->private;
 
-   if (isPixmap) {
+    const GLboolean swDepth = GL_FALSE;
+    const GLboolean swAlpha = GL_FALSE;
+    const GLboolean swAccum = mesaVis->accumRedBits > 0;
+    const GLboolean swStencil = mesaVis->stencilBits > 0 &&
+	mesaVis->depthBits != 24;
+    GLenum rgbFormat;
+    struct radeon_framebuffer *rfb;
+
+    if (isPixmap)
       return GL_FALSE; /* not implemented */
-   }
-   else {
-      const GLboolean swDepth = GL_FALSE;
-      const GLboolean swAlpha = GL_FALSE;
-      const GLboolean swAccum = mesaVis->accumRedBits > 0;
-      const GLboolean swStencil = mesaVis->stencilBits > 0 &&
-         mesaVis->depthBits != 24;
-      struct gl_framebuffer *fb = _mesa_create_framebuffer(mesaVis);
-
-      /* front color renderbuffer */
-      {
-         driRenderbuffer *frontRb
-            = driNewRenderbuffer(GL_RGBA,
-                                 driScrnPriv->pFB + screen->frontOffset,
-                                 screen->cpp,
-                                 screen->frontOffset, screen->frontPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(frontRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base);
-      }
 
-      /* back color renderbuffer */
-      if (mesaVis->doubleBufferMode) {
-         driRenderbuffer *backRb
-            = driNewRenderbuffer(GL_RGBA,
-                                 driScrnPriv->pFB + screen->backOffset,
-                                 screen->cpp,
-                                 screen->backOffset, screen->backPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(backRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base);
-      }
+    rfb = CALLOC_STRUCT(radeon_framebuffer);
+    if (!rfb)
+      return GL_FALSE;
 
-      /* depth renderbuffer */
-      if (mesaVis->depthBits == 16) {
-         driRenderbuffer *depthRb
-            = driNewRenderbuffer(GL_DEPTH_COMPONENT16,
-                                 driScrnPriv->pFB + screen->depthOffset,
-                                 screen->cpp,
-                                 screen->depthOffset, screen->depthPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(depthRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
-	 depthRb->depthHasSurface = screen->depthHasSurface;
-      }
-      else if (mesaVis->depthBits == 24) {
-         driRenderbuffer *depthRb
-            = driNewRenderbuffer(GL_DEPTH_COMPONENT24,
-                                 driScrnPriv->pFB + screen->depthOffset,
-                                 screen->cpp,
-                                 screen->depthOffset, screen->depthPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(depthRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
-	 depthRb->depthHasSurface = screen->depthHasSurface;
-      }
+    _mesa_initialize_framebuffer(&rfb->base, mesaVis);
+
+    if (mesaVis->redBits == 5)
+        rgbFormat = GL_RGB5;
+    else if (mesaVis->alphaBits == 0)
+        rgbFormat = GL_RGB8;
+    else
+        rgbFormat = GL_RGBA8;
+
+    /* front color renderbuffer */
+    rfb->color_rb[0] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
+    _mesa_add_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT, &rfb->color_rb[0]->base);
+    rfb->color_rb[0]->has_surface = 1;
+
+    /* back color renderbuffer */
+    if (mesaVis->doubleBufferMode) {
+      rfb->color_rb[1] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_BACK_LEFT, &rfb->color_rb[1]->base);
+	rfb->color_rb[1]->has_surface = 1;
+    }
 
-      /* stencil renderbuffer */
-      if (mesaVis->stencilBits > 0 && !swStencil) {
-         driRenderbuffer *stencilRb
-            = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT,
-                                 driScrnPriv->pFB + screen->depthOffset,
-                                 screen->cpp,
-                                 screen->depthOffset, screen->depthPitch,
-                                 driDrawPriv);
-         radeonSetSpanFunctions(stencilRb, mesaVis);
-         _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base);
-	 stencilRb->depthHasSurface = screen->depthHasSurface;
+    if (mesaVis->depthBits == 24) {
+      if (mesaVis->stencilBits == 8) {
+	struct radeon_renderbuffer *depthStencilRb = radeon_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depthStencilRb->base);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_STENCIL, &depthStencilRb->base);
+	depthStencilRb->has_surface = screen->depthHasSurface;
+      } else {
+	/* depth renderbuffer */
+	struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT24, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base);
+	depth->has_surface = screen->depthHasSurface;
       }
+    } else if (mesaVis->depthBits == 16) {
+      /* just 16-bit depth buffer, no hw stencil */
+	struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT16, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base);
+	depth->has_surface = screen->depthHasSurface;
+    }
 
-      _mesa_add_soft_renderbuffers(fb,
-                                   GL_FALSE, /* color */
-                                   swDepth,
-                                   swStencil,
-                                   swAccum,
-                                   swAlpha,
-                                   GL_FALSE /* aux */);
-      driDrawPriv->driverPrivate = (void *) fb;
+    _mesa_add_soft_renderbuffers(&rfb->base,
+	    GL_FALSE, /* color */
+	    swDepth,
+	    swStencil,
+	    swAccum,
+	    swAlpha,
+	    GL_FALSE /* aux */);
+    driDrawPriv->driverPrivate = (void *) rfb;
 
-      return (driDrawPriv->driverPrivate != NULL);
-   }
+    return (driDrawPriv->driverPrivate != NULL);
 }
 
 
-static void
+static void radeon_cleanup_renderbuffers(struct radeon_framebuffer *rfb)
+{
+	struct radeon_renderbuffer *rb;
+
+	rb = rfb->color_rb[0];
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[1];
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+}
+
+void
 radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv)
 {
-   _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL);
+    struct radeon_framebuffer *rfb;
+    if (!driDrawPriv)
+	return;
+
+    rfb = (void*)driDrawPriv->driverPrivate;
+    if (!rfb)
+	return;
+    radeon_cleanup_renderbuffers(rfb);
+    _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL);
 }
 
-#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
 /**
  * Choose the appropriate CreateContext function based on the chipset.
  * Eventually, all drivers will go through this process.
@@ -1145,25 +1577,27 @@ static GLboolean radeonCreateContext(const __GLcontextModes * glVisual,
 {
 	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
 	radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private);
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+	if (IS_R600_CLASS(screen))
+		return r600CreateContext(glVisual, driContextPriv, sharedContextPriv);
+#endif
 
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
 	if (IS_R300_CLASS(screen))
 		return r300CreateContext(glVisual, driContextPriv, sharedContextPriv);
-        return GL_FALSE;
-}
-
-/**
- * Choose the appropriate DestroyContext function based on the chipset.
- */
-static void radeonDestroyContext(__DRIcontextPrivate * driContextPriv)
-{
-	radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate;
-
-	if (IS_R300_CLASS(radeon->radeonScreen))
-		return r300DestroyContext(driContextPriv);
-}
+#endif
 
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
+	if (IS_R200_CLASS(screen))
+		return r200CreateContext(glVisual, driContextPriv, sharedContextPriv);
+#endif
 
+#if !RADEON_COMMON
+	(void)screen;
+	return r100CreateContext(glVisual, driContextPriv, sharedContextPriv);
 #endif
+	return GL_FALSE;
+}
 
 
 /**
@@ -1191,6 +1625,11 @@ radeonInitScreen(__DRIscreenPrivate *psp)
    static const __DRIutilversion2 ddx_expected = { 4, 5, 0, 0 };
    static const __DRIversion dri_expected = { 4, 0, 0 };
    static const __DRIversion drm_expected = { 1, 24, 0 };
+#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600)
+   static const char *driver_name = "R600";
+   static const __DRIutilversion2 ddx_expected = { 4, 5, 0, 0 };
+   static const __DRIversion dri_expected = { 4, 0, 0 };
+   static const __DRIversion drm_expected = { 1, 24, 0 };
 #endif
    RADEONDRIPtr dri_priv = (RADEONDRIPtr) psp->pDevPriv;
 
@@ -1218,20 +1657,110 @@ radeonInitScreen(__DRIscreenPrivate *psp)
    driInitSingleExtension( NULL, NV_vp_extension );
    driInitSingleExtension( NULL, ATI_fs_extension );
    driInitExtensions( NULL, point_extensions, GL_FALSE );
-#elif defined(RADEON_COMMON_FOR_R300)
+#elif (defined(RADEON_COMMON_FOR_R300) || defined(RADEON_COMMON_FOR_R600))
    driInitSingleExtension( NULL, gl_20_extension );
 #endif
 
    if (!radeonInitDriver(psp))
        return NULL;
 
+   /* for now fill in all modes */
    return radeonFillInModes( psp,
 			     dri_priv->bpp,
 			     (dri_priv->bpp == 16) ? 16 : 24,
-			     (dri_priv->bpp == 16) ? 0  : 8,
-			     (dri_priv->backOffset != dri_priv->depthOffset) );
+			     (dri_priv->bpp == 16) ? 0  : 8, 1);
 }
+#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
+
+/**
+ * This is the driver specific part of the createNewScreen entry point.
+ * Called when using DRI2.
+ *
+ * \return the __GLcontextModes supported by this driver
+ */
+static const
+__DRIconfig **radeonInitScreen2(__DRIscreenPrivate *psp)
+{
+   GLenum fb_format[3];
+   GLenum fb_type[3];
+   /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
+    * support pageflipping at all.
+    */
+   static const GLenum back_buffer_modes[] = {
+     GLX_NONE, GLX_SWAP_UNDEFINED_OML, /*, GLX_SWAP_COPY_OML*/
+   };
+   uint8_t depth_bits[4], stencil_bits[4], msaa_samples_array[1];
+   int color;
+   __DRIconfig **configs = NULL;
+
+   /* Calling driInitExtensions here, with a NULL context pointer,
+    * does not actually enable the extensions.  It just makes sure
+    * that all the dispatch offsets for all the extensions that
+    * *might* be enables are known.  This is needed because the
+    * dispatch offsets need to be known when _mesa_context_create
+    * is called, but we can't enable the extensions until we have a
+    * context pointer.
+    *
+    * Hello chicken.  Hello egg.  How are you two today?
+    */
+   driInitExtensions( NULL, card_extensions, GL_FALSE );
+   driInitExtensions( NULL, mm_extensions, GL_FALSE );
+#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
+   driInitExtensions( NULL, blend_extensions, GL_FALSE );
+   driInitSingleExtension( NULL, ARB_vp_extension );
+   driInitSingleExtension( NULL, NV_vp_extension );
+   driInitSingleExtension( NULL, ATI_fs_extension );
+   driInitExtensions( NULL, point_extensions, GL_FALSE );
+#endif
 
+   if (!radeonInitDriver(psp)) {
+       return NULL;
+    }
+   depth_bits[0] = 0;
+   stencil_bits[0] = 0;
+   depth_bits[1] = 16;
+   stencil_bits[1] = 0;
+   depth_bits[2] = 24;
+   stencil_bits[2] = 0;
+   depth_bits[3] = 24;
+   stencil_bits[3] = 8;
+
+   msaa_samples_array[0] = 0;
+
+   fb_format[0] = GL_RGB;
+   fb_type[0] = GL_UNSIGNED_SHORT_5_6_5;
+
+   fb_format[1] = GL_BGR;
+   fb_type[1] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   fb_format[2] = GL_BGRA;
+   fb_type[2] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   for (color = 0; color < ARRAY_SIZE(fb_format); color++) {
+      __DRIconfig **new_configs;
+
+      new_configs = driCreateConfigs(fb_format[color], fb_type[color],
+				     depth_bits,
+				     stencil_bits,
+				     ARRAY_SIZE(depth_bits),
+				     back_buffer_modes,
+				     ARRAY_SIZE(back_buffer_modes),
+				     msaa_samples_array,
+				     ARRAY_SIZE(msaa_samples_array));
+      if (configs == NULL)
+	 configs = new_configs;
+      else
+	 configs = driConcatConfigs(configs, new_configs);
+   }
+
+   if (configs == NULL) {
+      fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+              __LINE__);
+      return NULL;
+   }
+
+   return (const __DRIconfig **)configs;
+}
 
 /**
  * Get information about previous buffer swaps.
@@ -1239,31 +1768,26 @@ radeonInitScreen(__DRIscreenPrivate *psp)
 static int
 getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo )
 {
-#if !RADEON_COMMON || (RADEON_COMMON && defined(RADEON_COMMON_FOR_R300))
-   radeonContextPtr  rmesa;
-#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R200)
-   r200ContextPtr  rmesa;
-#endif
+    struct radeon_framebuffer *rfb;
 
-   if ( (dPriv == NULL) || (dPriv->driContextPriv == NULL)
-	|| (dPriv->driContextPriv->driverPrivate == NULL)
-	|| (sInfo == NULL) ) {
-      return -1;
+    if ( (dPriv == NULL) || (dPriv->driContextPriv == NULL)
+	 || (dPriv->driContextPriv->driverPrivate == NULL)
+	 || (sInfo == NULL) ) {
+	return -1;
    }
 
-   rmesa = dPriv->driContextPriv->driverPrivate;
-   sInfo->swap_count = rmesa->swap_count;
-   sInfo->swap_ust = rmesa->swap_ust;
-   sInfo->swap_missed_count = rmesa->swap_missed_count;
+    rfb = dPriv->driverPrivate;
+    sInfo->swap_count = rfb->swap_count;
+    sInfo->swap_ust = rfb->swap_ust;
+    sInfo->swap_missed_count = rfb->swap_missed_count;
 
    sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0)
-       ? driCalculateSwapUsage( dPriv, 0, rmesa->swap_missed_ust )
+       ? driCalculateSwapUsage( dPriv, 0, rfb->swap_missed_ust )
        : 0.0;
 
    return 0;
 }
 
-#if !RADEON_COMMON || (RADEON_COMMON && defined(RADEON_COMMON_FOR_R300))
 const struct __DriverAPIRec driDriverAPI = {
    .InitScreen      = radeonInitScreen,
    .DestroyScreen   = radeonDestroyScreen,
@@ -1280,23 +1804,7 @@ const struct __DriverAPIRec driDriverAPI = {
    .WaitForSBC      = NULL,
    .SwapBuffersMSC  = NULL,
    .CopySubBuffer   = radeonCopySubBuffer,
+    /* DRI2 */
+   .InitScreen2     = radeonInitScreen2,
 };
-#else
-const struct __DriverAPIRec driDriverAPI = {
-   .InitScreen      = radeonInitScreen,
-   .DestroyScreen   = radeonDestroyScreen,
-   .CreateContext   = r200CreateContext,
-   .DestroyContext  = r200DestroyContext,
-   .CreateBuffer    = radeonCreateBuffer,
-   .DestroyBuffer   = radeonDestroyBuffer,
-   .SwapBuffers     = r200SwapBuffers,
-   .MakeCurrent     = r200MakeCurrent,
-   .UnbindContext   = r200UnbindContext,
-   .GetSwapInfo     = getSwapInfo,
-   .GetDrawableMSC  = driDrawableGetMSC32,
-   .WaitForMSC      = driWaitForMSC32,
-   .WaitForSBC      = NULL,
-   .SwapBuffersMSC  = NULL,
-   .CopySubBuffer   = r200CopySubBuffer,
-};
-#endif
+
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h
index b84c70bfae..2a2f6b1b0b 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.h
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.h
@@ -54,11 +54,12 @@ typedef struct {
    drmAddress map;			/* Mapping of the DRM region */
 } radeonRegionRec, *radeonRegionPtr;
 
-typedef struct {
+typedef struct radeon_screen {
    int chip_family;
    int chip_flags;
    int cpp;
    int card_type;
+   int device_id; /* PCI ID */
    int AGPMode;
    unsigned int irq;			/* IRQ number (0 means none) */
 
@@ -103,9 +104,12 @@ typedef struct {
    /* Configuration cache with default values for all contexts */
    driOptionCache optionCache;
 
-   const __DRIextension *extensions[8];
+   const __DRIextension *extensions[16];
 
    int num_gb_pipes;
+   int kernel_mm;
+   drm_radeon_sarea_t *sarea;	/* Private SAREA data */
+   struct radeon_bo_manager *bom;
 } radeonScreenRec, *radeonScreenPtr;
 
 #define IS_R100_CLASS(screen) \
@@ -114,5 +118,8 @@ typedef struct {
 	((screen->chip_flags & RADEON_CLASS_MASK) == RADEON_CLASS_R200)
 #define IS_R300_CLASS(screen) \
 	((screen->chip_flags & RADEON_CLASS_MASK) == RADEON_CLASS_R300)
+#define IS_R600_CLASS(screen) \
+	((screen->chip_flags & RADEON_CLASS_MASK) == RADEON_CLASS_R600)
 
+extern void radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv);
 #endif /* __RADEON_SCREEN_H__ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c
index 12051ff1c8..b2a468b4fd 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.c
+++ b/src/mesa/drivers/dri/radeon/radeon_span.c
@@ -43,46 +43,168 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/glheader.h"
 #include "swrast/swrast.h"
 
-#include "radeon_context.h"
-#include "radeon_ioctl.h"
-#include "radeon_state.h"
+#include "radeon_common.h"
+#include "radeon_lock.h"
 #include "radeon_span.h"
-#include "radeon_tex.h"
-
-#include "drirenderbuffer.h"
 
 #define DBG 0
 
+static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb);
+
+/* radeon tiling on r300-r500 has 4 states,
+   macro-linear/micro-linear
+   macro-linear/micro-tiled
+   macro-tiled /micro-linear
+   macro-tiled /micro-tiled
+   1 byte surface 
+   2 byte surface - two types - we only provide 8x2 microtiling
+   4 byte surface
+   8/16 byte (unused)
+*/
+   
+static GLubyte *radeon_ptr_4byte(const struct radeon_renderbuffer * rrb,
+			     GLint x, GLint y)
+{
+    GLubyte *ptr = rrb->bo->ptr;
+    uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
+    GLint offset;
+
+    if (rrb->has_surface || !(rrb->bo->flags & mask)) {
+        offset = x * rrb->cpp + y * rrb->pitch;
+    } else {
+        offset = 0;
+        if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) {
+		offset = ((y >> 4) * (rrb->pitch >> 7) + (x >> 5)) << 11;
+		offset += (((y >> 3) ^ (x >> 5)) & 0x1) << 10;
+		offset += (((y >> 4) ^ (x >> 4)) & 0x1) << 9;
+		offset += (((y >> 2) ^ (x >> 4)) & 0x1) << 8;
+		offset += (((y >> 3) ^ (x >> 3)) & 0x1) << 7;
+		offset += ((y >> 1) & 0x1) << 6;
+		offset += ((x >> 2) & 0x1) << 5;
+		offset += (y & 1) << 4;
+		offset += (x & 3) << 2;
+            } else {
+		offset = ((y >> 3) * (rrb->pitch >> 8) + (x >> 6)) << 11;
+		offset += (((y >> 2) ^ (x >> 6)) & 0x1) << 10;
+		offset += (((y >> 3) ^ (x >> 5)) & 0x1) << 9;
+		offset += (((y >> 1) ^ (x >> 5)) & 0x1) << 8;
+		offset += (((y >> 2) ^ (x >> 4)) & 0x1) << 7;
+		offset += (y & 1) << 6;
+		offset += (x & 15) << 2;
+            }
+        } else {
+	    offset = ((y >> 1) * (rrb->pitch >> 4) + (x >> 2)) << 5;
+	    offset += (y & 1) << 4;
+	    offset += (x & 3) << 2;
+        }
+    }
+    return &ptr[offset];
+}
+
+static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb,
+				     GLint x, GLint y)
+{
+    GLubyte *ptr = rrb->bo->ptr;
+    uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
+    GLint offset;
+
+    if (rrb->has_surface || !(rrb->bo->flags & mask)) {
+        offset = x * rrb->cpp + y * rrb->pitch;
+    } else {
+        offset = 0;
+        if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) {
+            if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) {
+		offset = ((y >> 4) * (rrb->pitch >> 7) + (x >> 6)) << 11;
+		offset += (((y >> 3) ^ (x >> 6)) & 0x1) << 10;
+		offset += (((y >> 4) ^ (x >> 5)) & 0x1) << 9;
+		offset += (((y >> 2) ^ (x >> 5)) & 0x1) << 8;
+		offset += (((y >> 3) ^ (x >> 4)) & 0x1) << 7;
+		offset += ((y >> 1) & 0x1) << 6;
+		offset += ((x >> 3) & 0x1) << 5;
+		offset += (y & 1) << 4;
+		offset += (x & 3) << 2;
+            } else {
+		offset = ((y >> 3) * (rrb->pitch >> 8) + (x >> 7)) << 11;
+		offset += (((y >> 2) ^ (x >> 7)) & 0x1) << 10;
+		offset += (((y >> 3) ^ (x >> 6)) & 0x1) << 9;
+		offset += (((y >> 1) ^ (x >> 6)) & 0x1) << 8;
+		offset += (((y >> 2) ^ (x >> 5)) & 0x1) << 7;
+		offset += (y & 1) << 6;
+		offset += ((x >> 4) & 0x1) << 5;
+                offset += (x & 15) << 2;
+            }
+        } else {
+	    offset = ((y >> 1) * (rrb->pitch >> 4) + (x >> 3)) << 5;
+	    offset += (y & 0x1) << 4;
+	    offset += (x & 0x7) << 1;
+        }
+    }
+    return &ptr[offset];
+}
+
+#ifndef COMPILE_R300
+static uint32_t
+z24s8_to_s8z24(uint32_t val)
+{
+   return (val << 24) | (val >> 8);
+}
+
+static uint32_t
+s8z24_to_z24s8(uint32_t val)
+{
+   return (val >> 24) | (val << 8);
+}
+#endif
+
 /*
  * Note that all information needed to access pixels in a renderbuffer
  * should be obtained through the gl_renderbuffer parameter, not per-context
  * information.
  */
 #define LOCAL_VARS						\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;		\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
-   GLubyte *buf = (GLubyte *) drb->flippedData			\
-      + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp;	\
-   GLuint p;							\
-   (void) p;
+   struct radeon_context *radeon = RADEON_CONTEXT(ctx);			\
+   struct radeon_renderbuffer *rrb = (void *) rb;		\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : rrb->base.Height - 1;\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
+   GLuint p;						\
+   (void)p;						\
+   radeon_get_cliprects(radeon, &cliprects, &num_cliprects, &x_off, &y_off);
 
 #define LOCAL_DEPTH_VARS				\
-   driRenderbuffer *drb = (driRenderbuffer *) rb;	\
-   const __DRIdrawablePrivate *dPriv = drb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;			\
-   GLuint xo = dPriv->x;				\
-   GLuint yo = dPriv->y;				\
-   GLubyte *buf = (GLubyte *) drb->Base.Data;
+   struct radeon_context *radeon = RADEON_CONTEXT(ctx);			\
+   struct radeon_renderbuffer *rrb = (void *) rb;	\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : rrb->base.Height - 1;\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
+  radeon_get_cliprects(radeon, &cliprects, &num_cliprects, &x_off, &y_off);
 
 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
 
-#define Y_FLIP(Y) (bottom - (Y))
+#define Y_FLIP(_y) ((_y) * yScale + yBias)
 
 #define HW_LOCK()
 
 #define HW_UNLOCK()
 
+/* XXX FBO: this is identical to the macro in spantmp2.h except we get
+ * the cliprect info from the context, not the driDrawable.
+ * Move this into spantmp2.h someday.
+ */
+#define HW_CLIPLOOP()							\
+   do {									\
+      int _nc = num_cliprects;						\
+      while ( _nc-- ) {							\
+	 int minx = cliprects[_nc].x1 - x_off;				\
+	 int miny = cliprects[_nc].y1 - y_off;				\
+	 int maxx = cliprects[_nc].x2 - x_off;				\
+	 int maxy = cliprects[_nc].y2 - y_off;
+
 /* ================================================================
  * Color buffer
  */
@@ -94,7 +216,41 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define TAG(x)    radeon##x##_RGB565
 #define TAG2(x,y) radeon##x##_RGB565##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
+#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 16 bit, ARGB1555 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5_REV
+
+#define TAG(x)    radeon##x##_ARGB1555
+#define TAG2(x,y) radeon##x##_ARGB1555##y
+#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 16 bit, RGBA4 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4_REV
+
+#define TAG(x)    radeon##x##_ARGB4444
+#define TAG2(x,y) radeon##x##_ARGB4444##y
+#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 32 bit, xRGB8888 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+
+#define TAG(x)    radeon##x##_xRGB8888
+#define TAG2(x,y) radeon##x##_xRGB8888##y
+#define GET_VALUE(_x, _y) ((*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) | 0xff000000))
+#define PUT_VALUE(_x, _y, d) { \
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   *_ptr = d;								\
+} while (0)
 #include "spantmp2.h"
 
 /* 32 bit, ARGB8888 color spanline and pixel functions
@@ -104,7 +260,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define TAG(x)    radeon##x##_ARGB8888
 #define TAG2(x,y) radeon##x##_ARGB8888##y
-#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
+#define GET_VALUE(_x, _y) (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)))
+#define PUT_VALUE(_x, _y, d) { \
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   *_ptr = d;								\
+} while (0)
 #include "spantmp2.h"
 
 /* ================================================================
@@ -121,70 +281,20 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * too...
  */
 
-static GLuint radeon_mba_z32(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 4 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0..1] = 0           */
-
-#ifdef COMPILE_R300
-		ba = (y / 8) * (pitch / 8) + (x / 8);
-#else
-		ba = (y / 16) * (pitch / 16) + (x / 16);
-#endif
-
-		address |= (x & 0x7) << 2;	/* a[2..4] = x[0..2]     */
-		address |= (y & 0x3) << 5;	/* a[5..6] = y[0..1]     */
-		address |= (((x & 0x10) >> 2) ^ (y & 0x4)) << 5;	/* a[7]    = x[4] ^ y[2] */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= (((x & 0x8) << 1) ^ (y & 0x10)) << 7;	/* a[11]   = x[3] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
-static INLINE GLuint
-radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y)
-{
-	GLuint pitch = drb->pitch;
-	if (drb->depthHasSurface) {
-		return 2 * (x + y * pitch);
-	} else {
-		GLuint ba, address = 0;	/* a[0]    = 0           */
-
-		ba = (y / 16) * (pitch / 32) + (x / 32);
-
-		address |= (x & 0x7) << 1;	/* a[1..3] = x[0..2]     */
-		address |= (y & 0x7) << 4;	/* a[4..6] = y[0..2]     */
-		address |= (x & 0x8) << 4;	/* a[7]    = x[3]        */
-		address |= (ba & 0x3) << 8;	/* a[8..9] = ba[0..1]    */
-		address |= (y & 0x8) << 7;	/* a[10]   = y[3]        */
-		address |= ((x & 0x10) ^ (y & 0x10)) << 7;	/* a[11]   = x[4] ^ y[4] */
-		address |= (ba & ~0x3) << 10;	/* a[12..] = ba[2..]     */
-
-		return address;
-	}
-}
-
 /* 16-bit depth buffer functions
  */
 #define VALUE_TYPE GLushort
 
 #define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )) = d;
+   *(GLushort *)radeon_ptr_2byte_8x2(rrb, _x + x_off, _y + y_off) = d
 
 #define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo ));
+   d = *(GLushort *)radeon_ptr_2byte_8x2(rrb, _x + x_off, _y + y_off)
 
 #define TAG(x) radeon##x##_z16
 #include "depthtmp.h"
 
-/* 24 bit depth, 8 bit stencil depthbuffer functions
+/* 24 bit depth
  *
  * Careful: It looks like the R300 uses ZZZS byte order while the R200
  * uses SZZZ for 24 bit depth, 8 bit stencil mode.
@@ -194,35 +304,76 @@ radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y)
 #ifdef COMPILE_R300
 #define WRITE_DEPTH( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
    tmp &= 0x000000ff;							\
    tmp |= ((d << 8) & 0xffffff00);					\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
 } while (0)
 #else
 #define WRITE_DEPTH( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );	\
+   GLuint tmp = *_ptr;							\
    tmp &= 0xff000000;							\
    tmp |= ((d) & 0x00ffffff);						\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
 } while (0)
 #endif
 
 #ifdef COMPILE_R300
 #define READ_DEPTH( d, _x, _y )						\
-  do { \
-    d = (*(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,		\
-					 _y + yo )) & 0xffffff00) >> 8; \
+  do {									\
+    d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) & 0xffffff00) >> 8; \
   }while(0)
 #else
-#define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint *)(buf + radeon_mba_z32( drb, _x + xo,			\
-					 _y + yo )) & 0x00ffffff;
+#define READ_DEPTH( d, _x, _y )	\
+  d = *(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off,	_y + y_off)) & 0x00ffffff;
+#endif
+/*
+    fprintf(stderr, "dval(%d, %d, %d, %d)=0x%08X\n", _x, xo, _y, yo, d);\
+   d = *(GLuint*)(radeon_ptr(rrb, _x,	_y )) & 0x00ffffff;
+*/
+#define TAG(x) radeon##x##_z24
+#include "depthtmp.h"
+
+/* 24 bit depth, 8 bit stencil depthbuffer functions
+ * EXT_depth_stencil
+ *
+ * Careful: It looks like the R300 uses ZZZS byte order while the R200
+ * uses SZZZ for 24 bit depth, 8 bit stencil mode.
+ */
+#define VALUE_TYPE GLuint
+
+#ifdef COMPILE_R300
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   *_ptr = d;								\
+} while (0)
+#else
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );	\
+   GLuint tmp = z24s8_to_s8z24(d);					\
+   *_ptr = tmp;					\
+} while (0)
 #endif
 
+#ifdef COMPILE_R300
+#define READ_DEPTH( d, _x, _y )						\
+  do { \
+    d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)));	\
+  }while(0)
+#else
+#define READ_DEPTH( d, _x, _y )	do {					\
+    d = s8z24_to_z24s8(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off,	_y + y_off ))); \
+  } while (0)
+#endif
+/*
+    fprintf(stderr, "dval(%d, %d, %d, %d)=0x%08X\n", _x, xo, _y, yo, d);\
+   d = *(GLuint*)(radeon_ptr(rrb, _x,	_y )) & 0x00ffffff;
+*/
 #define TAG(x) radeon##x##_z24_s8
 #include "depthtmp.h"
 
@@ -235,35 +386,35 @@ do {									\
 #ifdef COMPILE_R300
 #define WRITE_STENCIL( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off);		\
+   GLuint tmp = *_ptr;				\
    tmp &= 0xffffff00;							\
    tmp |= (d) & 0xff;							\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
 } while (0)
 #else
 #define WRITE_STENCIL( _x, _y, d )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off);		\
+   GLuint tmp = *_ptr;				\
    tmp &= 0x00ffffff;							\
    tmp |= (((d) & 0xff) << 24);						\
-   *(GLuint *)(buf + offset) = tmp;					\
+   *_ptr = tmp;					\
 } while (0)
 #endif
 
 #ifdef COMPILE_R300
 #define READ_STENCIL( d, _x, _y )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
    d = tmp & 0x000000ff;						\
 } while (0)
 #else
 #define READ_STENCIL( d, _x, _y )					\
 do {									\
-   GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo );		\
-   GLuint tmp = *(GLuint *)(buf + offset);				\
+   GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off );		\
+   GLuint tmp = *_ptr;				\
    d = (tmp & 0xff000000) >> 24;					\
 } while (0)
 #endif
@@ -271,29 +422,110 @@ do {									\
 #define TAG(x) radeon##x##_z24_s8
 #include "stenciltmp.h"
 
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
+
+static void map_unmap_rb(struct gl_renderbuffer *rb, int flag)
+{
+	struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+	int r;
+
+	if (rrb == NULL || !rrb->bo)
+		return;
+
+	if (flag) {
+		if (rrb->bo->bom->funcs->bo_wait)
+			radeon_bo_wait(rrb->bo);
+		r = radeon_bo_map(rrb->bo, 1);
+		if (r) {
+			fprintf(stderr, "(%s) error(%d) mapping buffer.\n",
+				__FUNCTION__, r);
+		}
+
+		radeonSetSpanFunctions(rrb);
+	} else {
+		radeon_bo_unmap(rrb->bo);
+		rb->GetRow = NULL;
+		rb->PutRow = NULL;
+	}
+}
+
+static void
+radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map)
+{
+	GLuint i, j;
+
+	/* color draw buffers */
+	for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++)
+		map_unmap_rb(ctx->DrawBuffer->_ColorDrawBuffers[j], map);
+
+	/* check for render to textures */
+	for (i = 0; i < BUFFER_COUNT; i++) {
+		struct gl_renderbuffer_attachment *att =
+			ctx->DrawBuffer->Attachment + i;
+		struct gl_texture_object *tex = att->Texture;
+		if (tex) {
+			/* Render to texture. Note that a mipmapped texture need not
+			 * be complete for render to texture, so we must restrict to
+			 * mapping only the attached image.
+			 */
+			radeon_texture_image *image = get_radeon_texture_image(tex->Image[att->CubeMapFace][att->TextureLevel]);
+			ASSERT(att->Renderbuffer);
+
+			if (map)
+				radeon_teximage_map(image, GL_TRUE);
+			else
+				radeon_teximage_unmap(image);
+		}
+	}
+
+	map_unmap_rb(ctx->ReadBuffer->_ColorReadBuffer, map);
+
+	/* depth buffer (Note wrapper!) */
+	if (ctx->DrawBuffer->_DepthBuffer)
+		map_unmap_rb(ctx->DrawBuffer->_DepthBuffer->Wrapped, map);
+
+	if (ctx->DrawBuffer->_StencilBuffer)
+		map_unmap_rb(ctx->DrawBuffer->_StencilBuffer->Wrapped, map);
+}
 
 static void radeonSpanRenderStart(GLcontext * ctx)
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-#ifdef COMPILE_R300
-	r300ContextPtr r300 = (r300ContextPtr) rmesa;
-	R300_FIREVERTICES(r300);
-#else
-	RADEON_FIREVERTICES(rmesa);
-#endif
-	LOCK_HARDWARE(rmesa);
-	radeonWaitForIdleLocked(rmesa);
+	int i;
+
+	radeon_firevertices(rmesa);
+
+	/* The locking and wait for idle should really only be needed in classic mode.
+	 * In a future memory manager based implementation, this should become
+	 * unnecessary due to the fact that mapping our buffers, textures, etc.
+	 * should implicitly wait for any previous rendering commands that must
+	 * be waited on. */
+	if (!rmesa->radeonScreen->driScreen->dri2.enabled) {
+		LOCK_HARDWARE(rmesa);
+		radeonWaitForIdleLocked(rmesa);
+	}
+
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
+		if (ctx->Texture.Unit[i]._ReallyEnabled)
+			ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current);
+	}
+
+	radeon_map_unmap_buffers(ctx, 1);
 }
 
 static void radeonSpanRenderFinish(GLcontext * ctx)
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	int i;
 	_swrast_flush(ctx);
-	UNLOCK_HARDWARE(rmesa);
+	if (!rmesa->radeonScreen->driScreen->dri2.enabled) {
+		UNLOCK_HARDWARE(rmesa);
+	}
+	for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
+		if (ctx->Texture.Unit[i]._ReallyEnabled)
+			ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current);
+	}
+
+	radeon_map_unmap_buffers(ctx, 0);
 }
 
 void radeonInitSpanFuncs(GLcontext * ctx)
@@ -307,20 +539,27 @@ void radeonInitSpanFuncs(GLcontext * ctx)
 /**
  * Plug in the Get/Put routines for the given driRenderbuffer.
  */
-void radeonSetSpanFunctions(driRenderbuffer * drb, const GLvisual * vis)
+static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb)
 {
-	if (drb->Base.InternalFormat == GL_RGBA) {
-		if (vis->redBits == 5 && vis->greenBits == 6
-		    && vis->blueBits == 5) {
-			radeonInitPointers_RGB565(&drb->Base);
-		} else {
-			radeonInitPointers_ARGB8888(&drb->Base);
-		}
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
-		radeonInitDepthPointers_z16(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
-		radeonInitDepthPointers_z24_s8(&drb->Base);
-	} else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
-		radeonInitStencilPointers_z24_s8(&drb->Base);
+	if (rrb->base._ActualFormat == GL_RGB5) {
+		radeonInitPointers_RGB565(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGB8) {
+		radeonInitPointers_xRGB8888(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGBA8) {
+		radeonInitPointers_ARGB8888(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGBA4) {
+		radeonInitPointers_ARGB4444(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGB5_A1) {
+		radeonInitPointers_ARGB1555(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT16) {
+		radeonInitDepthPointers_z16(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT24) {
+		radeonInitDepthPointers_z24(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_DEPTH24_STENCIL8_EXT) {
+		radeonInitDepthPointers_z24_s8(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_STENCIL_INDEX8_EXT) {
+		radeonInitStencilPointers_z24_s8(&rrb->base);
+	} else {
+		fprintf(stderr, "radeonSetSpanFunctions: bad actual format: 0x%04X\n", rrb->base._ActualFormat);
 	}
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.h b/src/mesa/drivers/dri/radeon/radeon_span.h
index 9abe0864b1..ea6a2e7fb4 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.h
+++ b/src/mesa/drivers/dri/radeon/radeon_span.h
@@ -42,9 +42,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __RADEON_SPAN_H__
 #define __RADEON_SPAN_H__
 
-#include "drirenderbuffer.h"
-
 extern void radeonInitSpanFuncs(GLcontext * ctx);
-extern void radeonSetSpanFunctions(driRenderbuffer * rb, const GLvisual * vis);
 
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index b6561001e7..0d1728b747 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -47,6 +47,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast_setup/swrast_setup.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
 #include "radeon_tcl.h"
@@ -62,7 +63,7 @@ static void radeonUpdateSpecular( GLcontext *ctx );
 
 static void radeonAlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    int pp_misc = rmesa->hw.ctx.cmd[CTX_PP_MISC];
    GLubyte refByte;
 
@@ -106,7 +107,7 @@ static void radeonAlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref )
 static void radeonBlendEquationSeparate( GLcontext *ctx,
 					 GLenum modeRGB, GLenum modeA )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint b = rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] & ~RADEON_COMB_FCN_MASK;
    GLboolean fallback = GL_FALSE;
 
@@ -147,8 +148,8 @@ static void radeonBlendFuncSeparate( GLcontext *ctx,
 				     GLenum sfactorRGB, GLenum dfactorRGB,
 				     GLenum sfactorA, GLenum dfactorA )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint b = rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] & 
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint b = rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] &
       ~(RADEON_SRC_BLEND_MASK | RADEON_DST_BLEND_MASK);
    GLboolean fallback = GL_FALSE;
 
@@ -257,7 +258,7 @@ static void radeonBlendFuncSeparate( GLcontext *ctx,
 
 static void radeonDepthFunc( GLcontext *ctx, GLenum func )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, ctx );
    rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_Z_TEST_MASK;
@@ -293,7 +294,7 @@ static void radeonDepthFunc( GLcontext *ctx, GLenum func )
 
 static void radeonDepthMask( GLcontext *ctx, GLboolean flag )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    RADEON_STATECHANGE( rmesa, ctx );
 
    if ( ctx->Depth.Mask ) {
@@ -305,16 +306,16 @@ static void radeonDepthMask( GLcontext *ctx, GLboolean flag )
 
 static void radeonClearDepth( GLcontext *ctx, GLclampd d )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint format = (rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] &
 		    RADEON_DEPTH_FORMAT_MASK);
 
    switch ( format ) {
    case RADEON_DEPTH_FORMAT_16BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x0000ffff;
+      rmesa->radeon.state.depth.clear = d * 0x0000ffff;
       break;
    case RADEON_DEPTH_FORMAT_24BIT_INT_Z:
-      rmesa->state.depth.clear = d * 0x00ffffff;
+      rmesa->radeon.state.depth.clear = d * 0x00ffffff;
       break;
    }
 }
@@ -327,7 +328,7 @@ static void radeonClearDepth( GLcontext *ctx, GLclampd d )
 
 static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    union { int i; float f; } c, d;
    GLchan col[4];
 
@@ -391,7 +392,7 @@ static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
 	 rmesa->hw.fog.cmd[FOG_D] = d.i;
       }
       break;
-   case GL_FOG_COLOR: 
+   case GL_FOG_COLOR:
       RADEON_STATECHANGE( rmesa, ctx );
       UNCLAMPED_FLOAT_TO_RGB_CHAN( col, ctx->Fog.Color );
       rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~RADEON_FOG_COLOR_MASK;
@@ -406,109 +407,13 @@ static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
    }
 }
 
-
-/* =============================================================
- * Scissoring
- */
-
-
-static GLboolean intersect_rect( drm_clip_rect_t *out,
-				 drm_clip_rect_t *a,
-				 drm_clip_rect_t *b )
-{
-   *out = *a;
-   if ( b->x1 > out->x1 ) out->x1 = b->x1;
-   if ( b->y1 > out->y1 ) out->y1 = b->y1;
-   if ( b->x2 < out->x2 ) out->x2 = b->x2;
-   if ( b->y2 < out->y2 ) out->y2 = b->y2;
-   if ( out->x1 >= out->x2 ) return GL_FALSE;
-   if ( out->y1 >= out->y2 ) return GL_FALSE;
-   return GL_TRUE;
-}
-
-
-void radeonRecalcScissorRects( radeonContextPtr rmesa )
-{
-   drm_clip_rect_t *out;
-   int i;
-
-   /* Grow cliprect store?
-    */
-   if (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-      while (rmesa->state.scissor.numAllocedClipRects < rmesa->numClipRects) {
-	 rmesa->state.scissor.numAllocedClipRects += 1;	/* zero case */
-	 rmesa->state.scissor.numAllocedClipRects *= 2;
-      }
-
-      if (rmesa->state.scissor.pClipRects)
-	 FREE(rmesa->state.scissor.pClipRects);
-
-      rmesa->state.scissor.pClipRects = 
-	 MALLOC( rmesa->state.scissor.numAllocedClipRects * 
-		 sizeof(drm_clip_rect_t) );
-
-      if ( rmesa->state.scissor.pClipRects == NULL ) {
-	 rmesa->state.scissor.numAllocedClipRects = 0;
-	 return;
-      }
-   }
-   
-   out = rmesa->state.scissor.pClipRects;
-   rmesa->state.scissor.numClipRects = 0;
-
-   for ( i = 0 ; i < rmesa->numClipRects ;  i++ ) {
-      if ( intersect_rect( out, 
-			   &rmesa->pClipRects[i], 
-			   &rmesa->state.scissor.rect ) ) {
-	 rmesa->state.scissor.numClipRects++;
-	 out++;
-      }
-   }
-}
-
-
-static void radeonUpdateScissor( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if ( rmesa->dri.drawable ) {
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-
-      int x = ctx->Scissor.X;
-      int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height;
-      int w = ctx->Scissor.X + ctx->Scissor.Width - 1;
-      int h = dPriv->h - ctx->Scissor.Y - 1;
-
-      rmesa->state.scissor.rect.x1 = x + dPriv->x;
-      rmesa->state.scissor.rect.y1 = y + dPriv->y;
-      rmesa->state.scissor.rect.x2 = w + dPriv->x + 1;
-      rmesa->state.scissor.rect.y2 = h + dPriv->y + 1;
-
-      radeonRecalcScissorRects( rmesa );
-   }
-}
-
-
-static void radeonScissor( GLcontext *ctx,
-			   GLint x, GLint y, GLsizei w, GLsizei h )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if ( ctx->Scissor.Enabled ) {
-      RADEON_FIREVERTICES( rmesa );	/* don't pipeline cliprect changes */
-      radeonUpdateScissor( ctx );
-   }
-
-}
-
-
 /* =============================================================
  * Culling
  */
 
 static void radeonCullFace( GLcontext *ctx, GLenum unused )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint s = rmesa->hw.set.cmd[SET_SE_CNTL];
    GLuint t = rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL];
 
@@ -545,7 +450,7 @@ static void radeonCullFace( GLcontext *ctx, GLenum unused )
 
 static void radeonFrontFace( GLcontext *ctx, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, set );
    rmesa->hw.set.cmd[SET_SE_CNTL] &= ~RADEON_FFACE_CULL_DIR_MASK;
@@ -553,6 +458,10 @@ static void radeonFrontFace( GLcontext *ctx, GLenum mode )
    RADEON_STATECHANGE( rmesa, tcl );
    rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] &= ~RADEON_CULL_FRONT_IS_CCW;
 
+   /* Winding is inverted when rendering to FBO */
+   if (ctx->DrawBuffer && ctx->DrawBuffer->Name)
+      mode = (mode == GL_CW) ? GL_CCW : GL_CW;
+
    switch ( mode ) {
    case GL_CW:
       rmesa->hw.set.cmd[SET_SE_CNTL] |= RADEON_FFACE_CULL_CW;
@@ -570,7 +479,7 @@ static void radeonFrontFace( GLcontext *ctx, GLenum mode )
  */
 static void radeonLineWidth( GLcontext *ctx, GLfloat widthf )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, lin );
    RADEON_STATECHANGE( rmesa, set );
@@ -587,10 +496,10 @@ static void radeonLineWidth( GLcontext *ctx, GLfloat widthf )
 
 static void radeonLineStipple( GLcontext *ctx, GLint factor, GLushort pattern )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, lin );
-   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] = 
+   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] =
       ((((GLuint)factor & 0xff) << 16) | ((GLuint)pattern));
 }
 
@@ -602,12 +511,19 @@ static void radeonColorMask( GLcontext *ctx,
 			     GLboolean r, GLboolean g,
 			     GLboolean b, GLboolean a )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint mask = radeonPackColor( rmesa->radeonScreen->cpp,
-				  ctx->Color.ColorMask[RCOMP],
-				  ctx->Color.ColorMask[GCOMP],
-				  ctx->Color.ColorMask[BCOMP],
-				  ctx->Color.ColorMask[ACOMP] );
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb;
+   GLuint mask;
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
+
+   mask = radeonPackColor( rrb->cpp,
+			   ctx->Color.ColorMask[RCOMP],
+			   ctx->Color.ColorMask[GCOMP],
+			   ctx->Color.ColorMask[BCOMP],
+			   ctx->Color.ColorMask[ACOMP] );
 
    if ( rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK] != mask ) {
       RADEON_STATECHANGE( rmesa, msk );
@@ -623,8 +539,9 @@ static void radeonColorMask( GLcontext *ctx,
 static void radeonPolygonOffset( GLcontext *ctx,
 				 GLfloat factor, GLfloat units )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   float_ui32_type constant =  { units * rmesa->state.depth.scale };
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   float_ui32_type constant =  { units * depthScale };
    float_ui32_type factoru = { factor };
 
    RADEON_STATECHANGE( rmesa, zbs );
@@ -634,7 +551,7 @@ static void radeonPolygonOffset( GLcontext *ctx,
 
 static void radeonPolygonStipple( GLcontext *ctx, const GLubyte *mask )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint i;
    drm_radeon_stipple_t stipple;
 
@@ -646,27 +563,27 @@ static void radeonPolygonStipple( GLcontext *ctx, const GLubyte *mask )
 
    /* TODO: push this into cmd mechanism
     */
-   RADEON_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
+   radeon_firevertices(&rmesa->radeon);
+   LOCK_HARDWARE( &rmesa->radeon );
 
    /* FIXME: Use window x,y offsets into stipple RAM.
     */
    stipple.mask = rmesa->state.stipple.mask;
-   drmCommandWrite( rmesa->dri.fd, DRM_RADEON_STIPPLE, 
+   drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_STIPPLE,
                     &stipple, sizeof(drm_radeon_stipple_t) );
-   UNLOCK_HARDWARE( rmesa );
+   UNLOCK_HARDWARE( &rmesa->radeon );
 }
 
 static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean flag = (ctx->_TriangleCaps & DD_TRI_UNFILLED) != 0;
 
    /* Can't generally do unfilled via tcl, but some good special
-    * cases work. 
+    * cases work.
     */
    TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_UNFILLED, flag);
-   if (rmesa->TclFallback) {
+   if (rmesa->radeon.TclFallback) {
       radeonChooseRenderState( ctx );
       radeonChooseVertexState( ctx );
    }
@@ -686,7 +603,7 @@ static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode )
  */
 static void radeonUpdateSpecular( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    uint32_t p = rmesa->hw.ctx.cmd[CTX_PP_CNTL];
    GLuint flag = 0;
 
@@ -711,7 +628,7 @@ static void radeonUpdateSpecular( GLcontext *ctx )
       rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] |= RADEON_TCL_VTX_PK_DIFFUSE;
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] |= RADEON_LIGHTING_ENABLE;
       p |=  RADEON_SPECULAR_ENABLE;
-      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &= 
+      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &=
 	 ~RADEON_DIFFUSE_SPECULAR_COMBINE;
    }
    else if (ctx->Light.Enabled) {
@@ -741,7 +658,7 @@ static void radeonUpdateSpecular( GLcontext *ctx )
 	    RADEON_TCL_COMPUTE_SPECULAR) != 0;
       }
    }
- 
+
    TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_FOGCOORDSPEC, flag);
 
    if (NEED_SECONDARY_COLOR(ctx)) {
@@ -757,7 +674,7 @@ static void radeonUpdateSpecular( GLcontext *ctx )
 
    /* Update vertex/render formats
     */
-   if (rmesa->TclFallback) { 
+   if (rmesa->radeon.TclFallback) {
       radeonChooseRenderState( ctx );
       radeonChooseVertexState( ctx );
    }
@@ -769,12 +686,12 @@ static void radeonUpdateSpecular( GLcontext *ctx )
  */
 
 
-/* Update on colormaterial, material emmissive/ambient, 
+/* Update on colormaterial, material emmissive/ambient,
  * lightmodel.globalambient
  */
 static void update_global_ambient( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    float *fcmd = (float *)RADEON_DB_STATE( glt );
 
    /* Need to do more if both emmissive & ambient are PREMULT:
@@ -782,23 +699,23 @@ static void update_global_ambient( GLcontext *ctx )
     */
    if ((rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &
        ((3 << RADEON_EMISSIVE_SOURCE_SHIFT) |
-	(3 << RADEON_AMBIENT_SOURCE_SHIFT))) == 0) 
+	(3 << RADEON_AMBIENT_SOURCE_SHIFT))) == 0)
    {
-      COPY_3V( &fcmd[GLT_RED], 
+      COPY_3V( &fcmd[GLT_RED],
 	       ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_EMISSION]);
       ACC_SCALE_3V( &fcmd[GLT_RED],
 		   ctx->Light.Model.Ambient,
 		   ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT]);
-   } 
+   }
    else
    {
       COPY_3V( &fcmd[GLT_RED], ctx->Light.Model.Ambient );
    }
-   
+
    RADEON_DB_STATECHANGE(rmesa, &rmesa->hw.glt);
 }
 
-/* Update on change to 
+/* Update on change to
  *    - light[p].colors
  *    - light[p].enabled
  */
@@ -809,13 +726,13 @@ static void update_light_colors( GLcontext *ctx, GLuint p )
 /*     fprintf(stderr, "%s\n", __FUNCTION__); */
 
    if (l->Enabled) {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+      r100ContextPtr rmesa = R100_CONTEXT(ctx);
       float *fcmd = (float *)RADEON_DB_STATE( lit[p] );
 
-      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );	 
+      COPY_4V( &fcmd[LIT_AMBIENT_RED], l->Ambient );
       COPY_4V( &fcmd[LIT_DIFFUSE_RED], l->Diffuse );
       COPY_4V( &fcmd[LIT_SPECULAR_RED], l->Specular );
-      
+
       RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.lit[p] );
    }
 }
@@ -829,7 +746,7 @@ static void check_twoside_fallback( GLcontext *ctx )
 
    if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
       if (ctx->Light.ColorMaterialEnabled &&
-	  (ctx->Light.ColorMaterialBitmask & BACK_MATERIAL_BITS) != 
+	  (ctx->Light.ColorMaterialBitmask & BACK_MATERIAL_BITS) !=
 	  ((ctx->Light.ColorMaterialBitmask & FRONT_MATERIAL_BITS)<<1))
 	 fallback = GL_TRUE;
       else {
@@ -837,7 +754,7 @@ static void check_twoside_fallback( GLcontext *ctx )
 	    if (memcmp( ctx->Light.Material.Attrib[i],
 			ctx->Light.Material.Attrib[i+1],
 			sizeof(GLfloat)*4) != 0) {
-	       fallback = GL_TRUE;  
+	       fallback = GL_TRUE;
 	       break;
 	    }
       }
@@ -849,14 +766,14 @@ static void check_twoside_fallback( GLcontext *ctx )
 
 static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+      r100ContextPtr rmesa = R100_CONTEXT(ctx);
       GLuint light_model_ctl1 = rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL];
 
       light_model_ctl1 &= ~((3 << RADEON_EMISSIVE_SOURCE_SHIFT) |
 			   (3 << RADEON_AMBIENT_SOURCE_SHIFT) |
 			   (3 << RADEON_DIFFUSE_SOURCE_SHIFT) |
-			   (3 << RADEON_SPECULAR_SOURCE_SHIFT)); 
-   
+			   (3 << RADEON_SPECULAR_SOURCE_SHIFT));
+
    if (ctx->Light.ColorMaterialEnabled) {
       GLuint mask = ctx->Light.ColorMaterialBitmask;
 
@@ -877,7 +794,7 @@ static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_STATE_MULT <<
 			     RADEON_AMBIENT_SOURCE_SHIFT);
       }
-	 
+
       if (mask & MAT_BIT_FRONT_DIFFUSE) {
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_VERTEX_DIFFUSE <<
 			     RADEON_DIFFUSE_SOURCE_SHIFT);
@@ -886,7 +803,7 @@ static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_STATE_MULT <<
 			     RADEON_DIFFUSE_SOURCE_SHIFT);
       }
-   
+
       if (mask & MAT_BIT_FRONT_SPECULAR) {
 	 light_model_ctl1 |= (RADEON_LM_SOURCE_VERTEX_DIFFUSE <<
 			     RADEON_SPECULAR_SOURCE_SHIFT);
@@ -904,27 +821,27 @@ static void radeonColorMaterial( GLcontext *ctx, GLenum face, GLenum mode )
 		   (RADEON_LM_SOURCE_STATE_MULT << RADEON_DIFFUSE_SOURCE_SHIFT) |
 		   (RADEON_LM_SOURCE_STATE_MULT << RADEON_SPECULAR_SOURCE_SHIFT);
    }
-   
+
       if (light_model_ctl1 != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]) {
 	 RADEON_STATECHANGE( rmesa, tcl );
-	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] = light_model_ctl1;      
+	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] = light_model_ctl1;
    }
 }
 
 void radeonUpdateMaterial( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLfloat (*mat)[4] = ctx->Light.Material.Attrib;
    GLfloat *fcmd = (GLfloat *)RADEON_DB_STATE( mtl );
    GLuint mask = ~0;
-   
+
    if (ctx->Light.ColorMaterialEnabled)
       mask &= ~ctx->Light.ColorMaterialBitmask;
 
    if (RADEON_DEBUG & DEBUG_STATE)
       fprintf(stderr, "%s\n", __FUNCTION__);
 
-      
+
    if (mask & MAT_BIT_FRONT_EMISSION) {
       fcmd[MTL_EMMISSIVE_RED]   = mat[MAT_ATTRIB_FRONT_EMISSION][0];
       fcmd[MTL_EMMISSIVE_GREEN] = mat[MAT_ATTRIB_FRONT_EMISSION][1];
@@ -974,11 +891,11 @@ void radeonUpdateMaterial( GLcontext *ctx )
  *
  * which are calculated in light.c and are correct for the current
  * lighting space (model or eye), hence dependencies on _NEW_MODELVIEW
- * and _MESA_NEW_NEED_EYE_COORDS.  
+ * and _MESA_NEW_NEED_EYE_COORDS.
  */
 static void update_light( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    /* Have to check these, or have an automatic shortcircuit mechanism
     * to remove noop statechanges. (Or just do a better job on the
@@ -991,12 +908,12 @@ static void update_light( GLcontext *ctx )
 	 tmp &= ~RADEON_LIGHT_IN_MODELSPACE;
       else
 	 tmp |= RADEON_LIGHT_IN_MODELSPACE;
-      
+
 
       /* Leave this test disabled: (unexplained q3 lockup) (even with
          new packets)
       */
-      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]) 
+      if (tmp != rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL])
       {
 	 RADEON_STATECHANGE( rmesa, tcl );
 	 rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] = tmp;
@@ -1020,10 +937,10 @@ static void update_light( GLcontext *ctx )
 	 if (ctx->Light.Light[p].Enabled) {
 	    struct gl_light *l = &ctx->Light.Light[p];
 	    GLfloat *fcmd = (GLfloat *)RADEON_DB_STATE( lit[p] );
-	    
+
 	    if (l->EyePosition[3] == 0.0) {
-	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm ); 
-	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm ); 
+	       COPY_3FV( &fcmd[LIT_POSITION_X], l->_VP_inf_norm );
+	       COPY_3FV( &fcmd[LIT_DIRECTION_X], l->_h_inf_norm );
 	       fcmd[LIT_POSITION_W] = 0;
 	       fcmd[LIT_DIRECTION_W] = 0;
 	    } else {
@@ -1043,30 +960,30 @@ static void update_light( GLcontext *ctx )
 static void radeonLightfv( GLcontext *ctx, GLenum light,
 			   GLenum pname, const GLfloat *params )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLint p = light - GL_LIGHT0;
    struct gl_light *l = &ctx->Light.Light[p];
    GLfloat *fcmd = (GLfloat *)rmesa->hw.lit[p].cmd;
-   
+
 
    switch (pname) {
-   case GL_AMBIENT:		
+   case GL_AMBIENT:
    case GL_DIFFUSE:
    case GL_SPECULAR:
       update_light_colors( ctx, p );
       break;
 
-   case GL_SPOT_DIRECTION: 
-      /* picked up in update_light */	
+   case GL_SPOT_DIRECTION:
+      /* picked up in update_light */
       break;
 
    case GL_POSITION: {
-      /* positions picked up in update_light, but can do flag here */	
+      /* positions picked up in update_light, but can do flag here */
       GLuint flag;
       GLuint idx = TCL_PER_LIGHT_CTL_0 + p/2;
 
       /* FIXME: Set RANGE_ATTEN only when needed */
-      if (p&1) 
+      if (p&1)
 	 flag = RADEON_LIGHT_1_IS_LOCAL;
       else
 	 flag = RADEON_LIGHT_0_IS_LOCAL;
@@ -1158,16 +1075,16 @@ static void radeonLightfv( GLcontext *ctx, GLenum light,
    }
 }
 
-		  
+
 
 
 static void radeonLightModelfv( GLcontext *ctx, GLenum pname,
 				const GLfloat *param )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    switch (pname) {
-      case GL_LIGHT_MODEL_AMBIENT: 
+      case GL_LIGHT_MODEL_AMBIENT:
 	 update_global_ambient( ctx );
 	 break;
 
@@ -1188,7 +1105,7 @@ static void radeonLightModelfv( GLcontext *ctx, GLenum pname,
 
 	 check_twoside_fallback( ctx );
 
-	 if (rmesa->TclFallback) {
+	 if (rmesa->radeon.TclFallback) {
 	    radeonChooseRenderState( ctx );
 	    radeonChooseVertexState( ctx );
 	 }
@@ -1205,7 +1122,7 @@ static void radeonLightModelfv( GLcontext *ctx, GLenum pname,
 
 static void radeonShadeModel( GLcontext *ctx, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint s = rmesa->hw.set.cmd[SET_SE_CNTL];
 
    s &= ~(RADEON_DIFFUSE_SHADE_MASK |
@@ -1244,7 +1161,7 @@ static void radeonShadeModel( GLcontext *ctx, GLenum mode )
 static void radeonClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 {
    GLint p = (GLint) plane - (GLint) GL_CLIP_PLANE0;
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLint *ip = (GLint *)ctx->Transform._ClipUserPlane[p];
 
    RADEON_STATECHANGE( rmesa, ucp[p] );
@@ -1256,7 +1173,7 @@ static void radeonClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 
 static void radeonUpdateClipPlanes( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint p;
 
    for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
@@ -1281,7 +1198,7 @@ static void
 radeonStencilFuncSeparate( GLcontext *ctx, GLenum face, GLenum func,
                            GLint ref, GLuint mask )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint refmask = (((ctx->Stencil.Ref[0] & 0xff) << RADEON_STENCIL_REF_SHIFT) |
 		     ((ctx->Stencil.ValueMask[0] & 0xff) << RADEON_STENCIL_MASK_SHIFT));
 
@@ -1325,7 +1242,7 @@ radeonStencilFuncSeparate( GLcontext *ctx, GLenum face, GLenum func,
 static void
 radeonStencilMaskSeparate( GLcontext *ctx, GLenum face, GLuint mask )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    RADEON_STATECHANGE( rmesa, msk );
    rmesa->hw.msk.cmd[MSK_RB3D_STENCILREFMASK] &= ~RADEON_STENCIL_WRITE_MASK;
@@ -1336,20 +1253,20 @@ radeonStencilMaskSeparate( GLcontext *ctx, GLenum face, GLuint mask )
 static void radeonStencilOpSeparate( GLcontext *ctx, GLenum face, GLenum fail,
                                      GLenum zfail, GLenum zpass )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
    /* radeon 7200 have stencil bug, DEC and INC_WRAP will actually both do DEC_WRAP,
       and DEC_WRAP (and INVERT) will do INVERT. No way to get correct INC_WRAP and DEC,
       but DEC_WRAP can be fixed by using DEC and INC_WRAP at least use INC. */
-   
+
    GLuint tempRADEON_STENCIL_FAIL_DEC_WRAP;
    GLuint tempRADEON_STENCIL_FAIL_INC_WRAP;
    GLuint tempRADEON_STENCIL_ZFAIL_DEC_WRAP;
    GLuint tempRADEON_STENCIL_ZFAIL_INC_WRAP;
    GLuint tempRADEON_STENCIL_ZPASS_DEC_WRAP;
    GLuint tempRADEON_STENCIL_ZPASS_INC_WRAP;
-   
-   if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_BROKEN_STENCIL) {
+
+   if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_BROKEN_STENCIL) {
       tempRADEON_STENCIL_FAIL_DEC_WRAP = RADEON_STENCIL_FAIL_DEC;
       tempRADEON_STENCIL_FAIL_INC_WRAP = RADEON_STENCIL_FAIL_INC;
       tempRADEON_STENCIL_ZFAIL_DEC_WRAP = RADEON_STENCIL_ZFAIL_DEC;
@@ -1365,7 +1282,7 @@ static void radeonStencilOpSeparate( GLcontext *ctx, GLenum face, GLenum fail,
       tempRADEON_STENCIL_ZPASS_DEC_WRAP = RADEON_STENCIL_ZPASS_DEC_WRAP;
       tempRADEON_STENCIL_ZPASS_INC_WRAP = RADEON_STENCIL_ZPASS_INC_WRAP;
    }
-   
+
    RADEON_STATECHANGE( rmesa, ctx );
    rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] &= ~(RADEON_STENCIL_FAIL_MASK |
 					       RADEON_STENCIL_ZFAIL_MASK |
@@ -1455,9 +1372,9 @@ static void radeonStencilOpSeparate( GLcontext *ctx, GLenum face, GLenum fail,
 
 static void radeonClearStencil( GLcontext *ctx, GLint s )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   rmesa->state.stencil.clear = 
+   rmesa->radeon.state.stencil.clear =
       ((GLuint) (ctx->Stencil.Clear & 0xff) |
        (0xff << RADEON_STENCIL_MASK_SHIFT) |
        ((ctx->Stencil.WriteMask[0] & 0xff) << RADEON_STENCIL_WRITEMASK_SHIFT));
@@ -1481,20 +1398,30 @@ static void radeonClearStencil( GLcontext *ctx, GLint s )
  */
 void radeonUpdateWindow( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
-   GLfloat xoffset = (GLfloat)dPriv->x;
-   GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
+   GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+   GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer ? (ctx->DrawBuffer->Name != 0) : 0);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat y_scale, y_bias;
+
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   } else {
+      y_scale = -1.0;
+      y_bias = yoffset;
+   }
 
    float_ui32_type sx = { v[MAT_SX] };
    float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X };
-   float_ui32_type sy = { - v[MAT_SY] };
-   float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y };
-   float_ui32_type sz = { v[MAT_SZ] * rmesa->state.depth.scale };
-   float_ui32_type tz = { v[MAT_TZ] * rmesa->state.depth.scale };
+   float_ui32_type sy = { v[MAT_SY] * y_scale };
+   float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y };
+   float_ui32_type sz = { v[MAT_SZ] * depthScale };
+   float_ui32_type tz = { v[MAT_TZ] * depthScale };
 
-   RADEON_FIREVERTICES( rmesa );
    RADEON_STATECHANGE( rmesa, vpt );
 
    rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE]  = sx.ui32;
@@ -1514,6 +1441,8 @@ static void radeonViewport( GLcontext *ctx, GLint x, GLint y,
     * values, or keep the originals hanging around.
     */
    radeonUpdateWindow( ctx );
+
+   radeon_viewport(ctx, x, y, width, height);
 }
 
 static void radeonDepthRange( GLcontext *ctx, GLclampd nearval,
@@ -1524,8 +1453,8 @@ static void radeonDepthRange( GLcontext *ctx, GLclampd nearval,
 
 void radeonUpdateViewportOffset( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
    GLfloat xoffset = (GLfloat)dPriv->x;
    GLfloat yoffset = (GLfloat)dPriv->y + dPriv->h;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
@@ -1555,8 +1484,8 @@ void radeonUpdateViewportOffset( GLcontext *ctx )
                 RADEON_STIPPLE_Y_OFFSET_MASK);
 
          /* add magic offsets, then invert */
-         stx = 31 - ((rmesa->dri.drawable->x - 1) & RADEON_STIPPLE_COORD_MASK);
-         sty = 31 - ((rmesa->dri.drawable->y + rmesa->dri.drawable->h - 1)
+         stx = 31 - ((dPriv->x - 1) & RADEON_STIPPLE_COORD_MASK);
+         sty = 31 - ((dPriv->y + dPriv->h - 1)
                      & RADEON_STIPPLE_COORD_MASK);
 
          m |= ((stx << RADEON_STIPPLE_X_OFFSET_SHIFT) |
@@ -1580,20 +1509,26 @@ void radeonUpdateViewportOffset( GLcontext *ctx )
 
 static void radeonClearColor( GLcontext *ctx, const GLfloat color[4] )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLubyte c[4];
+   struct radeon_renderbuffer *rrb;
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   if (!rrb)
+     return;
+     
    CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]);
    CLAMPED_FLOAT_TO_UBYTE(c[1], color[1]);
    CLAMPED_FLOAT_TO_UBYTE(c[2], color[2]);
    CLAMPED_FLOAT_TO_UBYTE(c[3], color[3]);
-   rmesa->state.color.clear = radeonPackColor( rmesa->radeonScreen->cpp,
+   rmesa->radeon.state.color.clear = radeonPackColor( rrb->cpp,
 					       c[0], c[1], c[2], c[3] );
 }
 
 
 static void radeonRenderMode( GLcontext *ctx, GLenum mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    FALLBACK( rmesa, RADEON_FALLBACK_RENDER_MODE, (mode != GL_RENDER) );
 }
 
@@ -1619,7 +1554,7 @@ static GLuint radeon_rop_tab[] = {
 
 static void radeonLogicOpCode( GLcontext *ctx, GLenum opcode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint rop = (GLuint)opcode - GL_CLEAR;
 
    ASSERT( rop < 16 );
@@ -1628,105 +1563,13 @@ static void radeonLogicOpCode( GLcontext *ctx, GLenum opcode )
    rmesa->hw.msk.cmd[MSK_RB3D_ROPCNTL] = radeon_rop_tab[rop];
 }
 
-
-/**
- * Set up the cliprects for either front or back-buffer drawing.
- */
-void radeonSetCliprects( radeonContextPtr rmesa )
-{
-   __DRIdrawablePrivate *const drawable = rmesa->dri.drawable;
-   __DRIdrawablePrivate *const readable = rmesa->dri.readable;
-   GLframebuffer *const draw_fb = (GLframebuffer*) drawable->driverPrivate;
-   GLframebuffer *const read_fb = (GLframebuffer*) readable->driverPrivate;
-
-   if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-      /* Can't ignore 2d windows if we are page flipping.
-       */
-      if ( drawable->numBackClipRects == 0 || rmesa->doPageFlip ) {
-	 rmesa->numClipRects = drawable->numClipRects;
-	 rmesa->pClipRects = drawable->pClipRects;
-      }
-      else {
-	 rmesa->numClipRects = drawable->numBackClipRects;
-	 rmesa->pClipRects = drawable->pBackClipRects;
-      }
-   }
-   else {
-      /* front buffer (or none, or multiple buffers */
-      rmesa->numClipRects = drawable->numClipRects;
-      rmesa->pClipRects = drawable->pClipRects;
-   }
-
-   if ((draw_fb->Width != drawable->w) || (draw_fb->Height != drawable->h)) {
-      _mesa_resize_framebuffer(rmesa->glCtx, draw_fb,
-			       drawable->w, drawable->h);
-      draw_fb->Initialized = GL_TRUE;
-   }
-
-   if (drawable != readable) {
-      if ((read_fb->Width != readable->w) || (read_fb->Height != readable->h)) {
-	 _mesa_resize_framebuffer(rmesa->glCtx, read_fb,
-				  readable->w, readable->h);
-	 read_fb->Initialized = GL_TRUE;
-      }
-   }
-
-   if (rmesa->state.scissor.enabled)
-      radeonRecalcScissorRects( rmesa );
-
-   rmesa->lastStamp = drawable->lastStamp;
-}
-
-
-/**
- * Called via glDrawBuffer.
- */
-static void radeonDrawBuffer( GLcontext *ctx, GLenum mode )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_DRI)
-      fprintf(stderr, "%s %s\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   RADEON_FIREVERTICES(rmesa);	/* don't pipeline cliprect changes */
-
-   if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) {
-      /* 0 (GL_NONE) buffers or multiple color drawing buffers */
-      FALLBACK( rmesa, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   switch ( ctx->DrawBuffer->_ColorDrawBufferIndexes[0] ) {
-   case BUFFER_FRONT_LEFT:
-   case BUFFER_BACK_LEFT:
-      FALLBACK( rmesa, RADEON_FALLBACK_DRAW_BUFFER, GL_FALSE );
-      break;
-   default:
-      FALLBACK( rmesa, RADEON_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   radeonSetCliprects( rmesa );
-
-   /* We'll set the drawing engine's offset/pitch parameters later
-    * when we update other state.
-    */
-}
-
-static void radeonReadBuffer( GLcontext *ctx, GLenum mode )
-{
-   /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
-}
-
-
 /* =============================================================
  * State enable/disable
  */
 
 static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint p, flag;
 
    if ( RADEON_DEBUG & DEBUG_STATE )
@@ -1787,7 +1630,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_CLIP_PLANE2:
    case GL_CLIP_PLANE3:
    case GL_CLIP_PLANE4:
-   case GL_CLIP_PLANE5: 
+   case GL_CLIP_PLANE5:
       p = cap-GL_CLIP_PLANE0;
       RADEON_STATECHANGE( rmesa, tcl );
       if (state) {
@@ -1821,10 +1664,10 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
       RADEON_STATECHANGE(rmesa, ctx );
       if ( state ) {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~rmesa->radeon.state.color.roundEnable;
       } else {
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_DITHER_ENABLE;
-	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->state.color.roundEnable;
+	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  rmesa->radeon.state.color.roundEnable;
       }
       break;
 
@@ -1852,13 +1695,13 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_LIGHT7:
       RADEON_STATECHANGE(rmesa, tcl);
       p = cap - GL_LIGHT0;
-      if (p&1) 
+      if (p&1)
 	 flag = (RADEON_LIGHT_1_ENABLE |
-		 RADEON_LIGHT_1_ENABLE_AMBIENT | 
+		 RADEON_LIGHT_1_ENABLE_AMBIENT |
 		 RADEON_LIGHT_1_ENABLE_SPECULAR);
       else
 	 flag = (RADEON_LIGHT_0_ENABLE |
-		 RADEON_LIGHT_0_ENABLE_AMBIENT | 
+		 RADEON_LIGHT_0_ENABLE_AMBIENT |
 		 RADEON_LIGHT_0_ENABLE_SPECULAR);
 
       if (state)
@@ -1866,7 +1709,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
       else
 	 rmesa->hw.tcl.cmd[p/2 + TCL_PER_LIGHT_CTL_0] &= ~flag;
 
-      /* 
+      /*
        */
       update_light_colors( ctx, p );
       break;
@@ -1904,7 +1747,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
 	 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_ROP_ENABLE;
       }
       break;
-      
+
    case GL_NORMALIZE:
       RADEON_STATECHANGE( rmesa, tcl );
       if ( state ) {
@@ -1971,21 +1814,30 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    }
 
    case GL_SCISSOR_TEST:
-      RADEON_FIREVERTICES( rmesa );
-      rmesa->state.scissor.enabled = state;
+      radeon_firevertices(&rmesa->radeon);
+      rmesa->radeon.state.scissor.enabled = state;
       radeonUpdateScissor( ctx );
       break;
 
    case GL_STENCIL_TEST:
-      if ( rmesa->state.stencil.hwBuffer ) {
-	 RADEON_STATECHANGE( rmesa, ctx );
-	 if ( state ) {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_STENCIL_ENABLE;
+      {
+	 GLboolean hw_stencil = GL_FALSE;
+	 if (ctx->DrawBuffer) {
+	    struct radeon_renderbuffer *rrbStencil
+	       = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+	    hw_stencil = (rrbStencil && rrbStencil->bo);
+	 }
+
+	 if (hw_stencil) {
+	    RADEON_STATECHANGE( rmesa, ctx );
+	    if ( state ) {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_STENCIL_ENABLE;
+	    } else {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_STENCIL_ENABLE;
+	    }
 	 } else {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_STENCIL_ENABLE;
+	    FALLBACK( rmesa, RADEON_FALLBACK_STENCIL, state );
 	 }
-      } else {
-	 FALLBACK( rmesa, RADEON_FALLBACK_STENCIL, state );
       }
       break;
 
@@ -1995,7 +1847,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
    case GL_TEXTURE_GEN_T:
       /* Picked up in radeonUpdateTextureState.
        */
-      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE; 
+      rmesa->recheck_texgen[ctx->Texture.CurrentUnit] = GL_TRUE;
       break;
 
    case GL_COLOR_SUM_EXT:
@@ -2010,7 +1862,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
 
 static void radeonLightingSpaceChange( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean tmp;
    RADEON_STATECHANGE( rmesa, tcl );
 
@@ -2029,7 +1881,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx )
       rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &= ~RADEON_RESCALE_NORMALS;
    }
 
-   if (RADEON_DEBUG & DEBUG_STATE) 
+   if (RADEON_DEBUG & DEBUG_STATE)
       fprintf(stderr, "%s %d AFTER %x\n", __FUNCTION__, ctx->_NeedEyeCoords,
 	      rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]);
 }
@@ -2039,7 +1891,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx )
  */
 
 
-void radeonUploadTexMatrix( radeonContextPtr rmesa,
+void radeonUploadTexMatrix( r100ContextPtr rmesa,
 			    int unit, GLboolean swapcols )
 {
 /* Here's how this works: on r100, only 3 tex coords can be submitted, so the
@@ -2065,7 +1917,7 @@ void radeonUploadTexMatrix( radeonContextPtr rmesa,
    int idx = TEXMAT_0 + unit;
    float *dest = ((float *)RADEON_DB_STATE( mat[idx] )) + MAT_ELT_0;
    int i;
-   struct gl_texture_unit tUnit = rmesa->glCtx->Texture.Unit[unit];
+   struct gl_texture_unit tUnit = rmesa->radeon.glCtx->Texture.Unit[unit];
    GLfloat *src = rmesa->tmpmat[unit].m;
 
    rmesa->TexMatColSwap &= ~(1 << unit);
@@ -2119,7 +1971,7 @@ void radeonUploadTexMatrix( radeonContextPtr rmesa,
 }
 
 
-static void upload_matrix( radeonContextPtr rmesa, GLfloat *src, int idx )
+static void upload_matrix( r100ContextPtr rmesa, GLfloat *src, int idx )
 {
    float *dest = ((float *)RADEON_DB_STATE( mat[idx] ))+MAT_ELT_0;
    int i;
@@ -2135,7 +1987,7 @@ static void upload_matrix( radeonContextPtr rmesa, GLfloat *src, int idx )
    RADEON_DB_STATECHANGE( rmesa, &rmesa->hw.mat[idx] );
 }
 
-static void upload_matrix_t( radeonContextPtr rmesa, GLfloat *src, int idx )
+static void upload_matrix_t( r100ContextPtr rmesa, GLfloat *src, int idx )
 {
    float *dest = ((float *)RADEON_DB_STATE( mat[idx] ))+MAT_ELT_0;
    memcpy(dest, src, 16*sizeof(float));
@@ -2145,7 +1997,7 @@ static void upload_matrix_t( radeonContextPtr rmesa, GLfloat *src, int idx )
 
 static void update_texturematrix( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    GLuint tpc = rmesa->hw.tcl.cmd[TCL_TEXTURE_PROC_CTL];
    GLuint vs = rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXSEL];
    int unit;
@@ -2209,64 +2061,74 @@ static void update_texturematrix( GLcontext *ctx )
    }
 }
 
-
-/**
- * Tell the card where to render (offset, pitch).
- * Effected by glDrawBuffer, etc
- */
-void
-radeonUpdateDrawBuffer(GLcontext *ctx)
+static GLboolean r100ValidateBuffers(GLcontext *ctx)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   driRenderbuffer *drb;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   struct radeon_renderbuffer *rrb;
+   int i, ret;
 
-   if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-      /* draw to front */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
-   }
-   else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) {
-      /* draw to back */
-      drb = (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+   radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs);
+
+   rrb = radeon_get_colorbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
-   else {
-      /* drawing to multiple buffers, or none */
-      return;
+
+   /* depth buffer */
+   rrb = radeon_get_depthbuffer(&rmesa->radeon);
+   /* color buffer */
+   if (rrb && rrb->bo) {
+     radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, rrb->bo,
+				       0, RADEON_GEM_DOMAIN_VRAM);
    }
 
-   assert(drb);
-   assert(drb->flippedPitch);
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; ++i) {
+      radeonTexObj *t;
 
-   RADEON_STATECHANGE( rmesa, ctx );
+      if (!ctx->Texture.Unit[i]._ReallyEnabled)
+	 continue;
 
-   /* Note: we used the (possibly) page-flipped values */
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET]
-     = ((drb->flippedOffset + rmesa->radeonScreen->fbLocation)
-	& RADEON_COLOROFFSET_MASK);
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch;
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= RADEON_COLOR_TILE_ENABLE;
+      t = rmesa->state.texture.unit[i].texobj;
+      if (t->image_override && t->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
+      else if (t->mt->bo)
+	radeon_cs_space_add_persistent_bo(rmesa->radeon.cmdbuf.cs, t->mt->bo,
+			   RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0);
    }
-}
 
+   ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0);
+   if (ret)
+       return GL_FALSE;
+   return GL_TRUE;
+}
 
-void radeonValidateState( GLcontext *ctx )
+GLboolean radeonValidateState( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint new_state = rmesa->NewGLState;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint new_state = rmesa->radeon.NewGLState;
 
-   if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
-     radeonUpdateDrawBuffer(ctx);
+   if (new_state & _NEW_BUFFERS) {
+     _mesa_update_framebuffer(ctx);
+     /* this updates the DrawBuffer's Width/Height if it's a FBO */
+     _mesa_update_draw_buffer_bounds(ctx);
+     RADEON_STATECHANGE(rmesa, ctx);
    }
 
    if (new_state & _NEW_TEXTURE) {
       radeonUpdateTextureState( ctx );
-      new_state |= rmesa->NewGLState; /* may add TEXTURE_MATRIX */
+      new_state |= rmesa->radeon.NewGLState; /* may add TEXTURE_MATRIX */
    }
 
+   /* we need to do a space check here */
+   if (!r100ValidateBuffers(ctx))
+     return GL_FALSE;
+
    /* Need an event driven matrix update?
     */
-   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION)) 
+   if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION))
       upload_matrix( rmesa, ctx->_ModelProjectMatrix.m, MODEL_PROJ );
 
    /* Need these for lighting (shouldn't upload otherwise)
@@ -2290,12 +2152,14 @@ void radeonValidateState( GLcontext *ctx )
    /* emit all active clip planes if projection matrix changes.
     */
    if (new_state & (_NEW_PROJECTION)) {
-      if (ctx->Transform.ClipPlanesEnabled) 
+      if (ctx->Transform.ClipPlanesEnabled)
 	 radeonUpdateClipPlanes( ctx );
    }
 
 
-   rmesa->NewGLState = 0;
+   rmesa->radeon.NewGLState = 0;
+
+   return GL_TRUE;
 }
 
 
@@ -2306,7 +2170,7 @@ static void radeonInvalidateState( GLcontext *ctx, GLuint new_state )
    _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
-   RADEON_CONTEXT(ctx)->NewGLState |= new_state;
+   R100_CONTEXT(ctx)->radeon.NewGLState |= new_state;
 }
 
 
@@ -2317,8 +2181,8 @@ static GLboolean check_material( GLcontext *ctx )
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLint i;
 
-   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT; 
-	i < _TNL_ATTRIB_MAT_BACK_INDEXES; 
+   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT;
+	i < _TNL_ATTRIB_MAT_BACK_INDEXES;
 	i++)
       if (tnl->vb.AttribPtr[i] &&
 	  tnl->vb.AttribPtr[i]->stride)
@@ -2326,20 +2190,21 @@ static GLboolean check_material( GLcontext *ctx )
 
    return GL_FALSE;
 }
-      
+
 
 static void radeonWrapRunPipeline( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean has_material;
 
    if (0)
-      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->NewGLState);
+      fprintf(stderr, "%s, newstate: %x\n", __FUNCTION__, rmesa->radeon.NewGLState);
 
    /* Validate state:
     */
-   if (rmesa->NewGLState)
-      radeonValidateState( ctx );
+   if (rmesa->radeon.NewGLState)
+      if (!radeonValidateState( ctx ))
+	 FALLBACK(rmesa, RADEON_FALLBACK_TEXTURE, GL_TRUE);
 
    has_material = (ctx->Light.Enabled && check_material( ctx ));
 
@@ -2348,7 +2213,7 @@ static void radeonWrapRunPipeline( GLcontext *ctx )
    }
 
    /* Run the pipeline.
-    */ 
+    */
    _tnl_run_pipeline( ctx );
 
    if (has_material) {
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.h b/src/mesa/drivers/dri/radeon/radeon_state.h
index 2171879f75..a7c8eef32a 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.h
+++ b/src/mesa/drivers/dri/radeon/radeon_state.h
@@ -39,30 +39,25 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "radeon_context.h"
 
-extern void radeonInitState( radeonContextPtr rmesa );
+extern void radeonInitState( r100ContextPtr rmesa );
 extern void radeonInitStateFuncs( GLcontext *ctx );
 
 extern void radeonUpdateMaterial( GLcontext *ctx );
 
-extern void radeonSetCliprects( radeonContextPtr rmesa );
-extern void radeonRecalcScissorRects( radeonContextPtr rmesa );
 extern void radeonUpdateViewportOffset( GLcontext *ctx );
 extern void radeonUpdateWindow( GLcontext *ctx );
 extern void radeonUpdateDrawBuffer( GLcontext *ctx );
-extern void radeonUploadTexMatrix( radeonContextPtr rmesa,
+extern void radeonUploadTexMatrix( r100ContextPtr rmesa,
 				   int unit, GLboolean swapcols );
 
-extern void radeonValidateState( GLcontext *ctx );
-
-extern void radeonPrintDirty( radeonContextPtr rmesa,
-			      const char *msg );
+extern GLboolean radeonValidateState( GLcontext *ctx );
 
 
 extern void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode );
 #define FALLBACK( rmesa, bit, mode ) do {				\
    if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n",		\
 		     __FUNCTION__, bit, mode );				\
-   radeonFallback( rmesa->glCtx, bit, mode );				\
+   radeonFallback( rmesa->radeon.glCtx, bit, mode );				\
 } while (0)
 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index 57dc380050..aaa82b1d6a 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -38,39 +38,140 @@
 #include "swrast_setup/swrast_setup.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
 #include "radeon_tcl.h"
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
 
+#include "../r200/r200_reg.h"
+
 #include "xmlpool.h"
 
+/* New (1.3) state mechanism.  3 commands (packet, scalar, vector) in
+ * 1.3 cmdbuffers allow all previous state to be updated as well as
+ * the tcl scalar and vector areas.
+ */
+static struct {
+	int start;
+	int len;
+	const char *name;
+} packet[RADEON_MAX_STATE_PACKETS] = {
+	{RADEON_PP_MISC, 7, "RADEON_PP_MISC"},
+	{RADEON_PP_CNTL, 3, "RADEON_PP_CNTL"},
+	{RADEON_RB3D_COLORPITCH, 1, "RADEON_RB3D_COLORPITCH"},
+	{RADEON_RE_LINE_PATTERN, 2, "RADEON_RE_LINE_PATTERN"},
+	{RADEON_SE_LINE_WIDTH, 1, "RADEON_SE_LINE_WIDTH"},
+	{RADEON_PP_LUM_MATRIX, 1, "RADEON_PP_LUM_MATRIX"},
+	{RADEON_PP_ROT_MATRIX_0, 2, "RADEON_PP_ROT_MATRIX_0"},
+	{RADEON_RB3D_STENCILREFMASK, 3, "RADEON_RB3D_STENCILREFMASK"},
+	{RADEON_SE_VPORT_XSCALE, 6, "RADEON_SE_VPORT_XSCALE"},
+	{RADEON_SE_CNTL, 2, "RADEON_SE_CNTL"},
+	{RADEON_SE_CNTL_STATUS, 1, "RADEON_SE_CNTL_STATUS"},
+	{RADEON_RE_MISC, 1, "RADEON_RE_MISC"},
+	{RADEON_PP_TXFILTER_0, 6, "RADEON_PP_TXFILTER_0"},
+	{RADEON_PP_BORDER_COLOR_0, 1, "RADEON_PP_BORDER_COLOR_0"},
+	{RADEON_PP_TXFILTER_1, 6, "RADEON_PP_TXFILTER_1"},
+	{RADEON_PP_BORDER_COLOR_1, 1, "RADEON_PP_BORDER_COLOR_1"},
+	{RADEON_PP_TXFILTER_2, 6, "RADEON_PP_TXFILTER_2"},
+	{RADEON_PP_BORDER_COLOR_2, 1, "RADEON_PP_BORDER_COLOR_2"},
+	{RADEON_SE_ZBIAS_FACTOR, 2, "RADEON_SE_ZBIAS_FACTOR"},
+	{RADEON_SE_TCL_OUTPUT_VTX_FMT, 11, "RADEON_SE_TCL_OUTPUT_VTX_FMT"},
+	{RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED, 17,
+		    "RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED"},
+	{R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0"},
+	{R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1"},
+	{R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2"},
+	{R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3"},
+	{R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4"},
+	{R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5"},
+	{R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6"},
+	{R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7"},
+	{R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0"},
+	{R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0"},
+	{R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0"},
+	{R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL"},
+	{R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0"},
+	{R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2"},
+	{R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL"},
+	{R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0"},
+	{R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1"},
+	{R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2"},
+	{R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3"},
+	{R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4"},
+	{R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5"},
+	{R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0"},
+	{R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1"},
+	{R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2"},
+	{R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3"},
+	{R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4"},
+	{R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5"},
+	{R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL"},
+	{R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1,
+	 "R200_SE_TCL_OUTPUT_VTX_COMP_SEL"},
+	{R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3"},
+	{R200_PP_CNTL_X, 1, "R200_PP_CNTL_X"},
+	{R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET"},
+	{R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL"},
+	{R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0"},
+	{R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1"},
+	{R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2"},
+	{R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS"},
+	{R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL"},
+	{R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE"},
+	{R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4,
+		    "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0"},
+	{R200_PP_CUBIC_FACES_0, 1, "R200_PP_CUBIC_FACES_0"},	/* 61 */
+	{R200_PP_CUBIC_OFFSET_F1_0, 5, "R200_PP_CUBIC_OFFSET_F1_0"}, /* 62 */
+	{R200_PP_CUBIC_FACES_1, 1, "R200_PP_CUBIC_FACES_1"},
+	{R200_PP_CUBIC_OFFSET_F1_1, 5, "R200_PP_CUBIC_OFFSET_F1_1"},
+	{R200_PP_CUBIC_FACES_2, 1, "R200_PP_CUBIC_FACES_2"},
+	{R200_PP_CUBIC_OFFSET_F1_2, 5, "R200_PP_CUBIC_OFFSET_F1_2"},
+	{R200_PP_CUBIC_FACES_3, 1, "R200_PP_CUBIC_FACES_3"},
+	{R200_PP_CUBIC_OFFSET_F1_3, 5, "R200_PP_CUBIC_OFFSET_F1_3"},
+	{R200_PP_CUBIC_FACES_4, 1, "R200_PP_CUBIC_FACES_4"},
+	{R200_PP_CUBIC_OFFSET_F1_4, 5, "R200_PP_CUBIC_OFFSET_F1_4"},
+	{R200_PP_CUBIC_FACES_5, 1, "R200_PP_CUBIC_FACES_5"},
+	{R200_PP_CUBIC_OFFSET_F1_5, 5, "R200_PP_CUBIC_OFFSET_F1_5"},
+	{RADEON_PP_TEX_SIZE_0, 2, "RADEON_PP_TEX_SIZE_0"},
+	{RADEON_PP_TEX_SIZE_1, 2, "RADEON_PP_TEX_SIZE_1"},
+	{RADEON_PP_TEX_SIZE_2, 2, "RADEON_PP_TEX_SIZE_2"},
+	{R200_RB3D_BLENDCOLOR, 3, "R200_RB3D_BLENDCOLOR"},
+	{R200_SE_TCL_POINT_SPRITE_CNTL, 1, "R200_SE_TCL_POINT_SPRITE_CNTL"},
+	{RADEON_PP_CUBIC_FACES_0, 1, "RADEON_PP_CUBIC_FACES_0"},
+	{RADEON_PP_CUBIC_OFFSET_T0_0, 5, "RADEON_PP_CUBIC_OFFSET_T0_0"},
+	{RADEON_PP_CUBIC_FACES_1, 1, "RADEON_PP_CUBIC_FACES_1"},
+	{RADEON_PP_CUBIC_OFFSET_T1_0, 5, "RADEON_PP_CUBIC_OFFSET_T1_0"},
+	{RADEON_PP_CUBIC_FACES_2, 1, "RADEON_PP_CUBIC_FACES_2"},
+	{RADEON_PP_CUBIC_OFFSET_T2_0, 5, "RADEON_PP_CUBIC_OFFSET_T2_0"},
+	{R200_PP_TRI_PERF, 2, "R200_PP_TRI_PERF"},
+	{R200_PP_TXCBLEND_8, 32, "R200_PP_AFS_0"},     /* 85 */
+	{R200_PP_TXCBLEND_0, 32, "R200_PP_AFS_1"},
+	{R200_PP_TFACTOR_0, 8, "R200_ATF_TFACTOR"},
+	{R200_PP_TXFILTER_0, 8, "R200_PP_TXCTLALL_0"},
+	{R200_PP_TXFILTER_1, 8, "R200_PP_TXCTLALL_1"},
+	{R200_PP_TXFILTER_2, 8, "R200_PP_TXCTLALL_2"},
+	{R200_PP_TXFILTER_3, 8, "R200_PP_TXCTLALL_3"},
+	{R200_PP_TXFILTER_4, 8, "R200_PP_TXCTLALL_4"},
+	{R200_PP_TXFILTER_5, 8, "R200_PP_TXCTLALL_5"},
+	{R200_VAP_PVS_CNTL_1, 2, "R200_VAP_PVS_CNTL"},
+};
+
 /* =============================================================
  * State initialization
  */
-
-void radeonPrintDirty( radeonContextPtr rmesa, const char *msg )
+static int cmdpkt( r100ContextPtr rmesa, int id ) 
 {
-   struct radeon_state_atom *l;
-
-   fprintf(stderr, msg);
-   fprintf(stderr, ": ");
+   drm_radeon_cmd_header_t h;
 
-   foreach(l, &rmesa->hw.atomlist) {
-      if (l->dirty || rmesa->hw.all_dirty)
-	 fprintf(stderr, "%s, ", l->name);
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+     return CP_PACKET0(packet[id].start, packet[id].len - 1);
+   } else {
+     h.i = 0;
+     h.packet.cmd_type = RADEON_CMD_PACKET;
+     h.packet.packet_id = id;
    }
-
-   fprintf(stderr, "\n");
-}
-
-static int cmdpkt( int id ) 
-{
-   drm_radeon_cmd_header_t h;
-   h.i = 0;
-   h.packet.cmd_type = RADEON_CMD_PACKET;
-   h.packet.packet_id = id;
    return h.i;
 }
 
@@ -96,17 +197,17 @@ static int cmdscl( int offset, int stride, int count )
    return h.i;
 }
 
-#define CHECK( NM, FLAG )			\
-static GLboolean check_##NM( GLcontext *ctx )	\
-{						\
-   return FLAG;					\
+#define CHECK( NM, FLAG )				\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom )	\
+{							\
+   return FLAG ? atom->cmd_size : 0;			\
 }
 
 #define TCL_CHECK( NM, FLAG )				\
-static GLboolean check_##NM( GLcontext *ctx )		\
+static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom )	\
 {							\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);	\
-   return !rmesa->TclFallback && (FLAG);		\
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);	\
+   return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size : 0;	\
 }
 
 
@@ -146,81 +247,429 @@ CHECK( txr0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT))
 CHECK( txr1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT))
 CHECK( txr2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_RECT_BIT))
 
+#define OUT_VEC(hdr, data) do {			\
+    drm_radeon_cmd_header_t h;					\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0));		\
+    OUT_BATCH(0);							\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0));		\
+    OUT_BATCH(h.vectors.offset | (h.vectors.stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, h.vectors.count - 1));	\
+    OUT_BATCH_TABLE((data), h.vectors.count);				\
+  } while(0)
+
+#define OUT_SCL(hdr, data) do {					\
+    drm_radeon_cmd_header_t h;						\
+    h.i = hdr;								\
+    OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0));		\
+    OUT_BATCH((h.scalars.offset) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \
+    OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1));	\
+    OUT_BATCH_TABLE((data), h.scalars.count);				\
+  } while(0)
+
+static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+   
+   dwords += 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_SCL(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
 
 
-/* Initialize the context's hardware state.
- */
-void radeonInitState( radeonContextPtr rmesa )
+static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom)
 {
-   GLcontext *ctx = rmesa->glCtx;
-   GLuint color_fmt, depth_fmt, i;
-   GLint drawPitch, drawOffset;
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 4;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[0], atom->cmd+1);
+   END_BATCH();
+}
 
-   switch ( rmesa->radeonScreen->cpp ) {
-   case 2:
-      color_fmt = RADEON_COLOR_FORMAT_RGB565;
-      break;
-   case 4:
-      color_fmt = RADEON_COLOR_FORMAT_ARGB8888;
-      break;
-   default:
-      fprintf( stderr, "Error: Unsupported pixel depth... exiting\n" );
-      exit( -1 );
+
+static void lit_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+
+   dwords += 6;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+   OUT_VEC(atom->cmd[LIT_CMD_0], atom->cmd+1);
+   OUT_SCL(atom->cmd[LIT_CMD_1], atom->cmd+LIT_CMD_1+1);
+   END_BATCH();
+}
+
+static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   struct radeon_renderbuffer *rrb;
+   uint32_t cbpitch;
+   uint32_t zbpitch, depth_fmt;
+   uint32_t dwords = atom->cmd_size;
+
+   /* output the first 7 bytes of context */
+   BEGIN_BATCH_NO_AUTOSTATE(dwords + 4);
+   OUT_BATCH_TABLE(atom->cmd, 5);
+
+   rrb = radeon_get_depthbuffer(&r100->radeon);
+   if (!rrb) {
+     OUT_BATCH(0);
+     OUT_BATCH(0);
+   } else {
+     zbpitch = (rrb->pitch / rrb->cpp);
+     if (r100->using_hyperz)
+       zbpitch |= RADEON_DEPTH_HYPERZ;
+
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+     OUT_BATCH(zbpitch);
+     if (rrb->cpp == 4)
+        depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
+     else
+        depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
+   }
+     
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(atom->cmd[CTX_CMD_1]);
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+
+   rrb = radeon_get_colorbuffer(&r100->radeon);
+   if (!rrb || !rrb->bo) {
+      OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+      OUT_BATCH(atom->cmd[CTX_RB3D_COLOROFFSET]);
+   } else {
+      atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
+      if (rrb->cpp == 4)
+         atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
+      else
+         atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+
+      OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+      OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+   }
+
+   OUT_BATCH(atom->cmd[CTX_CMD_2]);
+
+   if (!rrb || !rrb->bo) {
+     OUT_BATCH(atom->cmd[CTX_RB3D_COLORPITCH]);
+   } else {
+     cbpitch = (rrb->pitch / rrb->cpp);
+     if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= RADEON_COLOR_TILE_ENABLE;
+     OUT_BATCH(cbpitch);
+   }
+
+   END_BATCH();
+}
+
+static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   struct radeon_renderbuffer *rrb, *drb;
+   uint32_t cbpitch = 0;
+   uint32_t zbpitch = 0;
+   uint32_t dwords = atom->cmd_size;
+   uint32_t depth_fmt;
+
+   rrb = radeon_get_colorbuffer(&r100->radeon);
+   if (!rrb || !rrb->bo) {
+      fprintf(stderr, "no rrb\n");
+      return;
+   }
+
+   atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
+   if (rrb->cpp == 4)
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
+   else switch (rrb->base._ActualFormat) {
+   case GL_RGB5:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
+	break;
+   case GL_RGBA4:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
+	break;
+   case GL_RGB5_A1:
+	atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
+	break;
+   }
+
+   cbpitch = (rrb->pitch / rrb->cpp);
+   if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+       cbpitch |= R200_COLOR_TILE_ENABLE;
+
+   drb = radeon_get_depthbuffer(&r100->radeon);
+   if (drb) {
+     zbpitch = (drb->pitch / drb->cpp);
+     if (drb->cpp == 4)
+        depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
+     else
+        depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
+     atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
+     
+   }
+
+   /* output the first 7 bytes of context */
+   dwords = 10;
+   if (drb)
+     dwords += 6;
+   if (rrb)
+     dwords += 6;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   /* In the CS case we need to split this up */
+   OUT_BATCH(CP_PACKET0(packet[0].start, 3));
+   OUT_BATCH_TABLE((atom->cmd + 1), 4);
+
+   if (drb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0));
+     OUT_BATCH_RELOC(0, drb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0));
+     OUT_BATCH(zbpitch);
+   }
+
+   OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0));
+   OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
+   OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1));
+   OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
+   OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
+
+   if (rrb) {
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0));
+     OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+     OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0));
+     OUT_BATCH(cbpitch);
+   }
+
+   // if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) {
+   //   OUT_BATCH_TABLE((atom->cmd + 14), 4);
+   // }
+
+   END_BATCH();
+   BEGIN_BATCH_NO_AUTOSTATE(4);
+   OUT_BATCH(CP_PACKET0(RADEON_RE_TOP_LEFT, 0));
+   OUT_BATCH(0);
+   OUT_BATCH(CP_PACKET0(RADEON_RE_WIDTH_HEIGHT, 0));
+   if (rrb) {
+       OUT_BATCH(((rrb->base.Width - 1) << RADEON_RE_WIDTH_SHIFT) |
+                 ((rrb->base.Height - 1) << RADEON_RE_HEIGHT_SHIFT));
+   } else {
+       OUT_BATCH(0);
+   }
+   END_BATCH();
+}
+
+static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = 3;
+   int i = atom->idx, j;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   if (!(ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT))
+	return;
+
+   if (!t)
+	return;
+
+   if (!t->mt)
+	return;
+
+   BEGIN_BATCH_NO_AUTOSTATE(dwords + (5 * 3));
+   OUT_BATCH_TABLE(atom->cmd, 3);
+   lvl = &t->mt->levels[0];
+   for (j = 0; j < 5; j++) {
+	OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_VRAM, 0, 0);
+   }
+   END_BATCH();
+}
+
+static void cube_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = 2;
+   int i = atom->idx, j;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+   uint32_t base_reg;
+
+   if (!(ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT))
+	return;
+
+   if (!t)
+	return;
+
+   if (!t->mt)
+	return;
+
+   switch(i) {
+	case 1: base_reg = RADEON_PP_CUBIC_OFFSET_T1_0; break;
+	case 2: base_reg = RADEON_PP_CUBIC_OFFSET_T2_0; break;
+	default:
+	case 0: base_reg = RADEON_PP_CUBIC_OFFSET_T0_0; break;
+   };
+   BEGIN_BATCH_NO_AUTOSTATE(dwords + (5 * 4));
+   OUT_BATCH_TABLE(atom->cmd, 2);
+   lvl = &t->mt->levels[0];
+   for (j = 0; j < 5; j++) {
+	OUT_BATCH(CP_PACKET0(base_reg + (4 * j), 0));
+	OUT_BATCH_RELOC(lvl->faces[j].offset, t->mt->bo, lvl->faces[j].offset,
+			RADEON_GEM_DOMAIN_VRAM, 0, 0);
+   }
+   END_BATCH();
+}
+
+static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+   int i = atom->idx;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+
+   if (t && t->mt && !t->image_override)
+     dwords += 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   OUT_BATCH_TABLE(atom->cmd, 3);
+   if (t && t->mt && !t->image_override) {
+     if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) {
+   	lvl = &t->mt->levels[0];
+	OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset,
+			RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     } else {
+        OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+     }
+   } else if (!t) {
+     /* workaround for old CS mechanism */
+     OUT_BATCH(r100->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]);
+     //     OUT_BATCH(r100->radeon.radeonScreen);
+   } else {
+     OUT_BATCH(t->override_offset);
    }
 
-   rmesa->state.color.clear = 0x00000000;
+   OUT_BATCH_TABLE((atom->cmd+4), 5);
+   END_BATCH();
+}
+
+static void tex_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+   r100ContextPtr r100 = R100_CONTEXT(ctx);
+   BATCH_LOCALS(&r100->radeon);
+   uint32_t dwords = atom->cmd_size;
+   int i = atom->idx;
+   radeonTexObj *t = r100->state.texture.unit[i].texobj;
+   radeon_mipmap_level *lvl;
+   int hastexture = 1;
+
+   if (!t)
+	hastexture = 0;
+   else {
+	if (!t->mt && !t->bo)
+		hastexture = 0;
+   }
+   dwords += 1;
+   if (hastexture)
+     dwords += 2;
+   else
+     dwords -= 2;
+   BEGIN_BATCH_NO_AUTOSTATE(dwords);
+
+   OUT_BATCH(CP_PACKET0(RADEON_PP_TXFILTER_0 + (24 * i), 1));
+   OUT_BATCH_TABLE((atom->cmd + 1), 2);
+
+   if (hastexture) {
+     OUT_BATCH(CP_PACKET0(RADEON_PP_TXOFFSET_0 + (24 * i), 0));
+     if (t->mt && !t->image_override) {
+        if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) {
+            lvl = &t->mt->levels[0];
+	    OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset,
+			RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+        } else {
+           OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
+		     RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+        }
+      } else {
+	if (t->bo)
+            OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
+                            RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+      }
+   }
+
+   OUT_BATCH(CP_PACKET0(RADEON_PP_TXCBLEND_0 + (i * 24), 1));
+   OUT_BATCH_TABLE((atom->cmd+4), 2);
+   OUT_BATCH(CP_PACKET0(RADEON_PP_BORDER_COLOR_0 + (i * 4), 0));
+   OUT_BATCH((atom->cmd[TEX_PP_BORDER_COLOR]));
+   END_BATCH();
+}
+
+/* Initialize the context's hardware state.
+ */
+void radeonInitState( r100ContextPtr rmesa )
+{
+   GLcontext *ctx = rmesa->radeon.glCtx;
+   GLuint i;
+
+   rmesa->radeon.state.color.clear = 0x00000000;
 
    switch ( ctx->Visual.depthBits ) {
    case 16:
-      rmesa->state.depth.clear = 0x0000ffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff;
-      depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
-      rmesa->state.stencil.clear = 0x00000000;
+      rmesa->radeon.state.depth.clear = 0x0000ffff;
+      rmesa->radeon.state.stencil.clear = 0x00000000;
       break;
    case 24:
-      rmesa->state.depth.clear = 0x00ffffff;
-      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff;
-      depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
-      rmesa->state.stencil.clear = 0xffff0000;
+      rmesa->radeon.state.depth.clear = 0x00ffffff;
+      rmesa->radeon.state.stencil.clear = 0xffff0000;
       break;
    default:
-      fprintf( stderr, "Error: Unsupported depth %d... exiting\n",
-	       ctx->Visual.depthBits );
-      exit( -1 );
+      break;
    }
 
-   /* Only have hw stencil when depth buffer is 24 bits deep */
-   rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
-				     ctx->Visual.depthBits == 24 );
+   rmesa->radeon.Fallback = 0;
 
-   rmesa->Fallback = 0;
 
-   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
-      drawOffset = rmesa->radeonScreen->backOffset;
-      drawPitch  = rmesa->radeonScreen->backPitch;
-   } else {
-      drawOffset = rmesa->radeonScreen->frontOffset;
-      drawPitch  = rmesa->radeonScreen->frontPitch;
-   }
+   rmesa->radeon.hw.max_state_size = 0;
 
-   rmesa->hw.max_state_size = 0;
-
-#define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG )				\
+#define ALLOC_STATE_IDX( ATOM, CHK, SZ, NM, FLAG, IDX )		\
    do {								\
       rmesa->hw.ATOM.cmd_size = SZ;				\
-      rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int));	\
-      rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int));	\
-      rmesa->hw.ATOM.name = NM;					\
+      rmesa->hw.ATOM.cmd = (GLuint *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.lastcmd = (GLuint *)CALLOC(SZ * sizeof(int)); \
+      rmesa->hw.ATOM.name = NM;						\
       rmesa->hw.ATOM.is_tcl = FLAG;					\
       rmesa->hw.ATOM.check = check_##CHK;				\
-      rmesa->hw.ATOM.dirty = GL_TRUE;				\
-      rmesa->hw.max_state_size += SZ * sizeof(int);		\
+      rmesa->hw.ATOM.dirty = GL_TRUE;					\
+      rmesa->hw.ATOM.idx = IDX;					\
+      rmesa->radeon.hw.max_state_size += SZ * sizeof(int);		\
    } while (0)
-      
-      
+
+#define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG )		\
+   ALLOC_STATE_IDX(ATOM, CHK, SZ, NM, FLAG, 0)
+
    /* Allocate state buffers:
     */
    ALLOC_STATE( ctx, always, CTX_STATE_SIZE, "CTX/context", 0 );
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     rmesa->hw.ctx.emit = ctx_emit_cs;
+   else
+     rmesa->hw.ctx.emit = ctx_emit;
    ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 );
    ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 );
    ALLOC_STATE( vpt, always, VPT_STATE_SIZE, "VPT/viewport", 0 );
@@ -233,20 +682,32 @@ void radeonInitState( radeonContextPtr rmesa )
    ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 1 );
    ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 1 );
    ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 );
-   ALLOC_STATE( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0 );
-   ALLOC_STATE( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0 );
-   ALLOC_STATE( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0 );
-   if (rmesa->radeonScreen->drmSupportsCubeMapsR100)
+   ALLOC_STATE_IDX( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0, 0);
+   ALLOC_STATE_IDX( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0, 1);
+   ALLOC_STATE_IDX( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0, 2);
+
+   for (i = 0; i < 3; i++) {
+      if (rmesa->radeon.radeonScreen->kernel_mm)
+          rmesa->hw.tex[i].emit = tex_emit_cs;
+      else
+          rmesa->hw.tex[i].emit = tex_emit;
+   }
+   if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100)
    {
-      ALLOC_STATE( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0 );
-      ALLOC_STATE( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0 );
-      ALLOC_STATE( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0 );
+      ALLOC_STATE_IDX( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 );
+      ALLOC_STATE_IDX( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 );
+      ALLOC_STATE_IDX( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 );
+      for (i = 0; i < 3; i++)
+          if (rmesa->radeon.radeonScreen->kernel_mm)
+              rmesa->hw.cube[i].emit = cube_emit_cs;
+          else
+              rmesa->hw.cube[i].emit = cube_emit;
    }
    else
    {
-      ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0 );
-      ALLOC_STATE( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0 );
-      ALLOC_STATE( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0 );
+      ALLOC_STATE_IDX( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 );
+      ALLOC_STATE_IDX( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 );
+      ALLOC_STATE_IDX( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 );
    }
    ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 );
    ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 );
@@ -268,43 +729,43 @@ void radeonInitState( radeonContextPtr rmesa )
    ALLOC_STATE( lit[5], tcl_lit5, LIT_STATE_SIZE, "LIT/light-5", 1 );
    ALLOC_STATE( lit[6], tcl_lit6, LIT_STATE_SIZE, "LIT/light-6", 1 );
    ALLOC_STATE( lit[7], tcl_lit7, LIT_STATE_SIZE, "LIT/light-7", 1 );
-   ALLOC_STATE( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0 );
-   ALLOC_STATE( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0 );
-   ALLOC_STATE( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0 );
+   ALLOC_STATE_IDX( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0, 0 );
+   ALLOC_STATE_IDX( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0, 1 );
+   ALLOC_STATE_IDX( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0, 2 );
 
    radeonSetUpAtomList( rmesa );
 
    /* Fill in the packet headers:
     */
-   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC);
-   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL);
-   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH);
-   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN);
-   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH);
-   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK);
-   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE);
-   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL);
-   rmesa->hw.set.cmd[SET_CMD_1] = cmdpkt(RADEON_EMIT_SE_CNTL_STATUS);
-   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC);
-   rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_0);
-   rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_0);
-   rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_1);
-   rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_1);
-   rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_2);
-   rmesa->hw.tex[2].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_2);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_0);
-   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T0);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_1);
-   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T1);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_2);
-   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T2);
-   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR);
-   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT);
+   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_MISC);
+   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CNTL);
+   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(rmesa, RADEON_EMIT_RB3D_COLORPITCH);
+   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_LINE_PATTERN);
+   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_LINE_WIDTH);
+   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RB3D_STENCILREFMASK);
+   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_VPORT_XSCALE);
+   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL);
+   rmesa->hw.set.cmd[SET_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL_STATUS);
+   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_MISC);
+   rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_0);
+   rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_0);
+   rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_1);
+   rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_1);
+   rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_2);
+   rmesa->hw.tex[2].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_2);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_0);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T0);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_1);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T1);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_2);
+   rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T2);
+   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_ZBIAS_FACTOR);
+   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT);
    rmesa->hw.mtl.cmd[MTL_CMD_0] = 
-      cmdpkt(RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED);
-   rmesa->hw.txr[0].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_0);
-   rmesa->hw.txr[1].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_1);
-   rmesa->hw.txr[2].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_2);
+      cmdpkt(rmesa, RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED);
+   rmesa->hw.txr[0].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_0);
+   rmesa->hw.txr[1].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_1);
+   rmesa->hw.txr[2].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_2);
    rmesa->hw.grd.cmd[GRD_CMD_0] = 
       cmdscl( RADEON_SS_VERT_GUARD_CLIP_ADJ_ADDR, 1, 4 );
    rmesa->hw.fog.cmd[FOG_CMD_0] = 
@@ -331,6 +792,22 @@ void radeonInitState( radeonContextPtr rmesa )
 	 cmdvec( RADEON_VS_UCP_ADDR + i, 1, 4 );
    }
 
+   if (rmesa->radeon.radeonScreen->kernel_mm) {
+      rmesa->hw.grd.emit = scl_emit;
+      rmesa->hw.fog.emit = vec_emit;
+      rmesa->hw.glt.emit = vec_emit;
+      rmesa->hw.eye.emit = vec_emit;
+      
+      for (i = 0; i <= 6; i++)
+	 rmesa->hw.mat[i].emit = vec_emit;
+
+      for (i = 0; i < 8; i++)
+	 rmesa->hw.lit[i].emit = lit_emit;
+
+      for (i = 0; i < 6; i++)
+	 rmesa->hw.ucp[i].emit = vec_emit;
+   }
+
    rmesa->last_ReallyEnabled = -1;
 
    /* Initial Harware state:
@@ -352,19 +829,7 @@ void radeonInitState( radeonContextPtr rmesa )
 					    RADEON_SRC_BLEND_GL_ONE |
 					    RADEON_DST_BLEND_GL_ZERO );
 
-   rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] =
-      rmesa->radeonScreen->depthOffset + rmesa->radeonScreen->fbLocation;
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = 
-      ((rmesa->radeonScreen->depthPitch &
-	RADEON_DEPTHPITCH_MASK) |
-       RADEON_DEPTH_ENDIAN_NO_SWAP);
-       
-   if (rmesa->using_hyperz)
-       rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] |= RADEON_DEPTH_HYPERZ;
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (depth_fmt |
-					       RADEON_Z_TEST_LESS |
+   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (RADEON_Z_TEST_LESS |
 					       RADEON_STENCIL_TEST_ALWAYS |
 					       RADEON_STENCIL_FAIL_KEEP |
 					       RADEON_STENCIL_ZPASS_KEEP |
@@ -374,7 +839,7 @@ void radeonInitState( radeonContextPtr rmesa )
    if (rmesa->using_hyperz) {
        rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_COMPRESSION_ENABLE |
 						   RADEON_Z_DECOMPRESSION_ENABLE;
-      if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+      if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
 	 /* works for q3, but slight rendering errors with glxgears ? */
 /*	 rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_HIERARCHY_ENABLE;*/
 	 /* need this otherwise get lots of lockups with q3 ??? */
@@ -386,10 +851,9 @@ void radeonInitState( radeonContextPtr rmesa )
 				     RADEON_ANTI_ALIAS_NONE);
 
    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = (RADEON_PLANE_MASK_ENABLE |
-				       color_fmt |
 				       RADEON_ZBLOCK16);
 
-   switch ( driQueryOptioni( &rmesa->optionCache, "dither_mode" ) ) {
+   switch ( driQueryOptioni( &rmesa->radeon.optionCache, "dither_mode" ) ) {
    case DRI_CONF_DITHER_XERRORDIFFRESET:
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_DITHER_INIT;
       break;
@@ -397,31 +861,18 @@ void radeonInitState( radeonContextPtr rmesa )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_SCALE_DITHER_ENABLE;
       break;
    }
-   if ( driQueryOptioni( &rmesa->optionCache, "round_mode" ) ==
+   if ( driQueryOptioni( &rmesa->radeon.optionCache, "round_mode" ) ==
 	DRI_CONF_ROUND_ROUND )
-      rmesa->state.color.roundEnable = RADEON_ROUND_ENABLE;
+      rmesa->radeon.state.color.roundEnable = RADEON_ROUND_ENABLE;
    else
-      rmesa->state.color.roundEnable = 0;
-   if ( driQueryOptioni (&rmesa->optionCache, "color_reduction" ) ==
+      rmesa->radeon.state.color.roundEnable = 0;
+   if ( driQueryOptioni (&rmesa->radeon.optionCache, "color_reduction" ) ==
 	DRI_CONF_COLOR_REDUCTION_DITHER )
       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_DITHER_ENABLE;
    else
-      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable;
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((drawOffset +
-					       rmesa->radeonScreen->fbLocation)
-					      & RADEON_COLOROFFSET_MASK);
-
-   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((drawPitch &
-					      RADEON_COLORPITCH_MASK) |
-					     RADEON_COLOR_ENDIAN_NO_SWAP);
+      rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable;
 
 
-   /* (fixed size) sarea is initialized to zero afaics so can omit version check. Phew! */
-   if (rmesa->sarea->tiling_enabled) {
-      rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= RADEON_COLOR_TILE_ENABLE;
-   }
-
    rmesa->hw.set.cmd[SET_SE_CNTL] = (RADEON_FFACE_CULL_CCW |
 				     RADEON_BFACE_SOLID |
 				     RADEON_FFACE_SOLID |
@@ -444,7 +895,7 @@ void radeonInitState( radeonContextPtr rmesa )
   					    RADEON_VC_NO_SWAP;
 #endif
 
-   if (!(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
+   if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
      rmesa->hw.set.cmd[SET_SE_CNTL_STATUS] |= RADEON_TCL_BYPASS;
    }
 
@@ -491,8 +942,8 @@ void radeonInitState( radeonContextPtr rmesa )
 	   (2 << RADEON_TXFORMAT_HEIGHT_SHIFT));
 
       /* Initialize the texture offset to the start of the card texture heap */
-      rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+      //      rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] =
+      //	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
 
       rmesa->hw.tex[i].cmd[TEX_PP_BORDER_COLOR] = 0;
       rmesa->hw.tex[i].cmd[TEX_PP_TXCBLEND] =  
@@ -513,15 +964,15 @@ void radeonInitState( radeonContextPtr rmesa )
 
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0;
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_0] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_1] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_2] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_3] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
       rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_4] =
-	  rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
+	  rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP];
    }
 
    /* Can only add ST1 at the time of doing some multitex but can keep
@@ -613,5 +1064,7 @@ void radeonInitState( radeonContextPtr rmesa )
    rmesa->hw.eye.cmd[EYE_Z] = IEEE_ONE;
    rmesa->hw.eye.cmd[EYE_RESCALE_FACTOR] = IEEE_ONE;
    
-   rmesa->hw.all_dirty = GL_TRUE;
+   rmesa->radeon.hw.all_dirty = GL_TRUE;
+
+   rcommonInitCmdBuf(&rmesa->radeon);
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
index ebea1fecdc..e31f045991 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
@@ -52,8 +52,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tcl.h"
 
 
-static void flush_last_swtcl_prim( radeonContextPtr rmesa  );
-
 /* R100: xyzw, c0, c1/fog, stq[0..2]  = 4+1+1+3*3 = 15  right? */
 /* R200: xyzw, c0, c1/fog, strq[0..5] = 4+1+1+4*6 = 30 */
 #define RADEON_MAX_TNL_VERTEX_SIZE (15 * sizeof(GLfloat))	/* for mesa _tnl stage */
@@ -64,18 +62,18 @@ static void flush_last_swtcl_prim( radeonContextPtr rmesa  );
 
 #define EMIT_ATTR( ATTR, STYLE, F0 )					\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR);	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE);	\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
    fmt_0 |= F0;								\
 } while (0)
 
 #define EMIT_PAD( N )							\
 do {									\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
-   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
-   rmesa->swtcl.vertex_attr_count++;					\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0;		\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD;	\
+   rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N);		\
+   rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
 static GLuint radeon_cp_vc_frmts[3][2] =
@@ -87,7 +85,7 @@ static GLuint radeon_cp_vc_frmts[3][2] =
 
 static void radeonSetVertexFormat( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    DECLARE_RENDERINPUTS(index_bitset);
@@ -106,7 +104,7 @@ static void radeonSetVertexFormat( GLcontext *ctx )
    }
 
    assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-   rmesa->swtcl.vertex_attr_count = 0;
+   rmesa->radeon.swtcl.vertex_attr_count = 0;
 
    /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
     * build up a hardware vertex.
@@ -204,33 +202,33 @@ static void radeonSetVertexFormat( GLcontext *ctx )
       }
    }
 
-   if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) ||
+   if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) ||
 	fmt_0 != rmesa->swtcl.vertex_format) {
       RADEON_NEWPRIM(rmesa);
       rmesa->swtcl.vertex_format = fmt_0;
-      rmesa->swtcl.vertex_size =
+      rmesa->radeon.swtcl.vertex_size =
 	  _tnl_install_attrs( ctx,
-			      rmesa->swtcl.vertex_attrs, 
-			      rmesa->swtcl.vertex_attr_count,
+			      rmesa->radeon.swtcl.vertex_attrs, 
+			      rmesa->radeon.swtcl.vertex_attr_count,
 			      NULL, 0 );
-      rmesa->swtcl.vertex_size /= 4;
-      RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
+      rmesa->radeon.swtcl.vertex_size /= 4;
+      RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset );
       if (RADEON_DEBUG & DEBUG_VERTS)
 	 fprintf( stderr, "%s: vertex_size= %d floats\n",
-		  __FUNCTION__, rmesa->swtcl.vertex_size);
+		  __FUNCTION__, rmesa->radeon.swtcl.vertex_size);
    }
 }
 
 
 static void radeonRenderStart( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
 
    radeonSetVertexFormat( ctx );
    
-   if (rmesa->dma.flush != 0 && 
-       rmesa->dma.flush != flush_last_swtcl_prim)
-      rmesa->dma.flush( rmesa );
+   if (rmesa->radeon.dma.flush != 0 && 
+       rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim)
+      rmesa->radeon.dma.flush( ctx );
 }
 
 
@@ -241,7 +239,7 @@ static void radeonRenderStart( GLcontext *ctx )
  */
 void radeonChooseVertexState( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    TNLcontext *tnl = TNL_CONTEXT(ctx);
 
    GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
@@ -254,7 +252,7 @@ void radeonChooseVertexState( GLcontext *ctx )
     * rasterization fallback.  As this function will be called again when we
     * leave a rasterization fallback, we can just skip it for now.
     */
-   if (rmesa->Fallback != 0)
+   if (rmesa->radeon.Fallback != 0)
       return;
 
    /* HW perspective divide is a win, but tiny vertex formats are a
@@ -281,80 +279,29 @@ void radeonChooseVertexState( GLcontext *ctx )
    }
 }
 
-
-/* Flush vertices in the current dma region.
- */
-static void flush_last_swtcl_prim( radeonContextPtr rmesa  )
+void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
 {
-   if (RADEON_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   rmesa->dma.flush = NULL;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   if (rmesa->dma.current.buf) {
-      struct radeon_dma_region *current = &rmesa->dma.current;
-      GLuint current_offset = (rmesa->radeonScreen->gart_buffer_offset +
-			       current->buf->buf->idx * RADEON_BUFFER_SIZE + 
-			       current->start);
+   rcommonEnsureCmdBufSpace(&rmesa->radeon,
+			    rmesa->radeon.hw.max_state_size + (12*sizeof(int)),
+			    __FUNCTION__);
 
-      assert (!(rmesa->swtcl.hw_primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));
 
-      assert (current->start + 
-	      rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	      current->ptr);
+   radeonEmitState(&rmesa->radeon);
+   radeonEmitVertexAOS( rmesa,
+			rmesa->radeon.swtcl.vertex_size,
+			rmesa->radeon.dma.current,
+			current_offset);
 
-      if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
-	 radeonEnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ +
-			          rmesa->hw.max_state_size + VBUF_BUFSZ );
-
-	 radeonEmitVertexAOS( rmesa,
-			      rmesa->swtcl.vertex_size,
-			      current_offset);
-
-	 radeonEmitVbufPrim( rmesa,
-			     rmesa->swtcl.vertex_format,
-			     rmesa->swtcl.hw_primitive,
-			     rmesa->swtcl.numverts);
-      }
+		      
+   radeonEmitVbufPrim( rmesa,
+		       rmesa->swtcl.vertex_format,
+		       rmesa->radeon.swtcl.hw_primitive,
+		       rmesa->radeon.swtcl.numverts);
 
-      rmesa->swtcl.numverts = 0;
-      current->start = current->ptr;
-   }
 }
 
-
-/* Alloc space in the current dma region.
- */
-static INLINE void *
-radeonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
-{
-   GLuint bytes = vsize * nverts;
-
-   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
-      radeonRefillCurrentDmaRegion( rmesa );
-
-   if (!rmesa->dma.flush) {
-      rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-      rmesa->dma.flush = flush_last_swtcl_prim;
-   }
-
-   assert( vsize == rmesa->swtcl.vertex_size * 4 );
-   assert( rmesa->dma.flush == flush_last_swtcl_prim );
-   assert (rmesa->dma.current.start + 
-	   rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
-	   rmesa->dma.current.ptr);
-
-
-   {
-      GLubyte *head = (GLubyte *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->dma.current.ptr += bytes;
-      rmesa->swtcl.numverts += nverts;
-      return head;
-   }
-
-}
-
-
 /*
  * Render unclipped vertex buffers by emitting vertices directly to
  * dma buffers.  Use strip/fan hardware primitives where possible.
@@ -387,22 +334,22 @@ static const GLuint hw_prim[GL_POLYGON+1] = {
 };
 
 static INLINE void
-radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim )
+radeonDmaPrimitive( r100ContextPtr rmesa, GLenum prim )
 {
    RADEON_NEWPRIM( rmesa );
-   rmesa->swtcl.hw_primitive = hw_prim[prim];
-   assert(rmesa->dma.current.ptr == rmesa->dma.current.start);
+   rmesa->radeon.swtcl.hw_primitive = hw_prim[prim];
+   //   assert(rmesa->radeon.dma.current.ptr == rmesa->radeon.dma.current.start);
 }
 
-#define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx)
+#define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx)
 #define INIT( prim ) radeonDmaPrimitive( rmesa, prim )
 #define FLUSH()  RADEON_NEWPRIM( rmesa )
-#define GET_CURRENT_VB_MAX_VERTS() \
-  (((int)rmesa->dma.current.end - (int)rmesa->dma.current.ptr) / (rmesa->swtcl.vertex_size*4))
+#define GET_CURRENT_VB_MAX_VERTS()					10\
+//  (((int)rmesa->radeon.dma.current.end - (int)rmesa->radeon.dma.current.ptr) / (rmesa->radeon.swtcl.vertex_size*4))
 #define GET_SUBSEQUENT_VB_MAX_VERTS() \
-  ((RADEON_BUFFER_SIZE) / (rmesa->swtcl.vertex_size*4))
+  ((RADEON_BUFFER_SIZE) / (rmesa->radeon.swtcl.vertex_size*4))
 #define ALLOC_VERTS( nr ) \
-  radeonAllocDmaLowVerts( rmesa, nr, rmesa->swtcl.vertex_size * 4 )
+  rcommonAllocDmaLowVerts( &rmesa->radeon, nr, rmesa->radeon.swtcl.vertex_size * 4 )
 #define EMIT_VERTS( ctx, j, nr, buf ) \
   _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf)
 
@@ -418,16 +365,13 @@ radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim )
 static GLboolean radeon_run_render( GLcontext *ctx,
 				    struct tnl_pipeline_stage *stage )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    tnl_render_func *tab = TAG(render_tab_verts);
    GLuint i;
 
-   if (rmesa->swtcl.indexed_verts.buf) 
-      RELEASE_ELT_VERTS();
-   	
-   if (rmesa->swtcl.RenderIndex != 0 ||   
+   if (rmesa->radeon.swtcl.RenderIndex != 0 ||   
        !radeon_dma_validate_render( ctx, VB ))
       return GL_TRUE;		
 
@@ -496,13 +440,13 @@ static void radeonResetLineStipple( GLcontext *ctx );
 
 #undef LOCAL_VARS
 #undef ALLOC_VERTS
-#define CTX_ARG radeonContextPtr rmesa
-#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size
-#define ALLOC_VERTS( n, size ) radeonAllocDmaLowVerts( rmesa, n, (size) * 4 )
+#define CTX_ARG r100ContextPtr rmesa
+#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size
+#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, (size) * 4 )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);		\
-   const char *radeonverts = (char *)rmesa->swtcl.verts;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);		\
+   const char *radeonverts = (char *)rmesa->radeon.swtcl.verts;
 #define VERT(x) (radeonVertex *)(radeonverts + ((x) * (vertsize) * sizeof(int)))
 #define VERTEX radeonVertex 
 #undef TAG
@@ -560,7 +504,7 @@ static struct {
 #define VERT_Y(_v) _v->v.y
 #define VERT_Z(_v) _v->v.z
 #define AREA_IS_CCW( a ) (a < 0)
-#define GET_VERTEX(e) (rmesa->swtcl.verts + ((e) * rmesa->swtcl.vertex_size * sizeof(int)))
+#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + ((e) * rmesa->radeon.swtcl.vertex_size * sizeof(int)))
 
 #define VERT_SET_RGBA( v, c )  					\
 do {								\
@@ -606,7 +550,7 @@ do {							\
 #undef INIT
 
 #define LOCAL_VARS(n)							\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);			\
    GLuint color[n], spec[n];						\
    GLuint coloroffset = rmesa->swtcl.coloroffset;	\
    GLuint specoffset = rmesa->swtcl.specoffset;			\
@@ -617,7 +561,7 @@ do {							\
  ***********************************************************************/
 
 #define RASTERIZE(x) radeonRasterPrimitive( ctx, reduced_hw_prim[x] )
-#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive
+#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive
 #undef TAG
 #define TAG(x) x
 #include "tnl_dd/t_dd_unfilled.h"
@@ -673,9 +617,9 @@ static void init_rast_tab( void )
 } while (0)
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);		\
-   const GLuint vertsize = rmesa->swtcl.vertex_size;		\
-   const char *radeonverts = (char *)rmesa->swtcl.verts;		\
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);		\
+   const GLuint vertsize = rmesa->radeon.swtcl.vertex_size;		\
+   const char *radeonverts = (char *)rmesa->radeon.swtcl.verts;		\
    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
    const GLboolean stipple = ctx->Line.StippleFlag;		\
    (void) elt; (void) stipple;
@@ -700,17 +644,17 @@ static void init_rast_tab( void )
 void radeonChooseRenderState( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint index = 0;
    GLuint flags = ctx->_TriangleCaps;
 
-   if (!rmesa->TclFallback || rmesa->Fallback) 
+   if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback) 
       return;
 
    if (flags & DD_TRI_LIGHT_TWOSIDE) index |= RADEON_TWOSIDE_BIT;
    if (flags & DD_TRI_UNFILLED)      index |= RADEON_UNFILLED_BIT;
 
-   if (index != rmesa->swtcl.RenderIndex) {
+   if (index != rmesa->radeon.swtcl.RenderIndex) {
       tnl->Driver.Render.Points = rast_tab[index].points;
       tnl->Driver.Render.Line = rast_tab[index].line;
       tnl->Driver.Render.ClippedLine = rast_tab[index].line;
@@ -727,7 +671,7 @@ void radeonChooseRenderState( GLcontext *ctx )
 	 tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon;
       }
 
-      rmesa->swtcl.RenderIndex = index;
+      rmesa->radeon.swtcl.RenderIndex = index;
    }
 }
 
@@ -739,18 +683,18 @@ void radeonChooseRenderState( GLcontext *ctx )
 
 static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   if (rmesa->swtcl.hw_primitive != hwprim) {
+   if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
       RADEON_NEWPRIM( rmesa );
-      rmesa->swtcl.hw_primitive = hwprim;
+      rmesa->radeon.swtcl.hw_primitive = hwprim;
    }
 }
 
 static void radeonRenderPrimitive( GLcontext *ctx, GLenum prim )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   rmesa->swtcl.render_primitive = prim;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   rmesa->radeon.swtcl.render_primitive = prim;
    if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) 
       radeonRasterPrimitive( ctx, reduced_hw_prim[prim] );
 }
@@ -761,7 +705,7 @@ static void radeonRenderFinish( GLcontext *ctx )
 
 static void radeonResetLineStipple( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    RADEON_STATECHANGE( rmesa, lin );
 }
 
@@ -795,17 +739,17 @@ static const char *getFallbackString(GLuint bit)
 
 void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->Fallback;
+   GLuint oldfallback = rmesa->radeon.Fallback;
 
    if (mode) {
-      rmesa->Fallback |= bit;
+      rmesa->radeon.Fallback |= bit;
       if (oldfallback == 0) {
-	 RADEON_FIREVERTICES( rmesa );
+	 radeon_firevertices(&rmesa->radeon);
 	 TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE );
 	 _swsetup_Wakeup( ctx );
-	 rmesa->swtcl.RenderIndex = ~0;
+	 rmesa->radeon.swtcl.RenderIndex = ~0;
          if (RADEON_DEBUG & DEBUG_FALLBACKS) {
             fprintf(stderr, "Radeon begin rasterization fallback: 0x%x %s\n",
                     bit, getFallbackString(bit));
@@ -813,7 +757,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
       }
    }
    else {
-      rmesa->Fallback &= ~bit;
+      rmesa->radeon.Fallback &= ~bit;
       if (oldfallback == bit) {
 	 _swrast_flush( ctx );
 	 tnl->Driver.Render.Start = radeonRenderStart;
@@ -826,14 +770,14 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 
 	 tnl->Driver.Render.ResetLineStipple = radeonResetLineStipple;
 	 TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_FALSE );
-	 if (rmesa->TclFallback) {
-	    /* These are already done if rmesa->TclFallback goes to
+	 if (rmesa->radeon.TclFallback) {
+	    /* These are already done if rmesa->radeon.TclFallback goes to
 	     * zero above. But not if it doesn't (RADEON_NO_TCL for
 	     * example?)
 	     */
 	    _tnl_invalidate_vertex_state( ctx, ~0 );
 	    _tnl_invalidate_vertices( ctx, ~0 );
-	    RENDERINPUTS_ZERO( rmesa->tnl_index_bitset );
+	    RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset );
 	    radeonChooseVertexState( ctx );
 	    radeonChooseRenderState( ctx );
 	 }
@@ -853,7 +797,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 void radeonInitSwtcl( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    static int firsttime = 1;
 
    if (firsttime) {
@@ -872,18 +816,9 @@ void radeonInitSwtcl( GLcontext *ctx )
    _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
 		       RADEON_MAX_TNL_VERTEX_SIZE);
    
-   rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
-   rmesa->swtcl.RenderIndex = ~0;
-   rmesa->swtcl.render_primitive = GL_TRIANGLES;
-   rmesa->swtcl.hw_primitive = 0;
+   rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf;
+   rmesa->radeon.swtcl.RenderIndex = ~0;
+   rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES;
+   rmesa->radeon.swtcl.hw_primitive = 0;
 }
 
-
-void radeonDestroySwtcl( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (rmesa->swtcl.indexed_verts.buf) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
-			      __FUNCTION__ );
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
index e485052ad7..da89158eeb 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
@@ -40,7 +40,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_context.h"
 
 extern void radeonInitSwtcl( GLcontext *ctx );
-extern void radeonDestroySwtcl( GLcontext *ctx );
 
 extern void radeonChooseRenderState( GLcontext *ctx );
 extern void radeonChooseVertexState( GLcontext *ctx );
@@ -63,5 +62,5 @@ extern void radeon_translate_vertex( GLcontext *ctx,
 
 extern void radeon_print_vertex( GLcontext *ctx, const radeonVertex *v );
 
-
+extern void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c
index 779e9ae5df..df6708f05e 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c
@@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
+#include "radeon_common.h"
 #include "radeon_context.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
@@ -104,7 +105,7 @@ static GLboolean discrete_prim[0x10] = {
 };
    
 
-#define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx)
+#define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx)
 #define ELT_TYPE  GLushort
 
 #define ELT_INIT(prim, hw_prim) \
@@ -125,7 +126,7 @@ static GLboolean discrete_prim[0x10] = {
 
 #define RESET_STIPPLE() do {			\
    RADEON_STATECHANGE( rmesa, lin );		\
-   radeonEmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);			\
 } while (0)
 
 #define AUTO_STIPPLE( mode )  do {		\
@@ -136,31 +137,29 @@ static GLboolean discrete_prim[0x10] = {
    else						\
       rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &=	\
 	 ~RADEON_LINE_PATTERN_AUTO_RESET;	\
-   radeonEmitState( rmesa );			\
+   radeonEmitState(&rmesa->radeon);		\
 } while (0)
 
 
 
 #define ALLOC_ELTS(nr)	radeonAllocElts( rmesa, nr )
 
-static GLushort *radeonAllocElts( radeonContextPtr rmesa, GLuint nr ) 
+static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) 
 {
-   if (rmesa->dma.flush)
-      rmesa->dma.flush( rmesa );
+      if (rmesa->radeon.dma.flush)
+	 rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
 
-   radeonEnsureCmdBufSpace(rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			   rmesa->hw.max_state_size + ELTS_BUFSZ(nr));
+      rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->radeon.hw.max_state_size + ELTS_BUFSZ(nr) + 
+			       AOS_BUFSZ(rmesa->radeon.tcl.aos_count), __FUNCTION__);
 
-   radeonEmitAOS( rmesa,
-		rmesa->tcl.aos_components,
-		rmesa->tcl.nr_aos_components, 0 );
+      radeonEmitAOS( rmesa,
+		     rmesa->radeon.tcl.aos_count, 0 );
 
-   return radeonAllocEltsOpenEnded( rmesa,
-				    rmesa->tcl.vertex_format, 
-				    rmesa->tcl.hw_primitive, nr );
+      return radeonAllocEltsOpenEnded( rmesa, rmesa->tcl.vertex_format,
+				       rmesa->tcl.hw_primitive, nr );
 }
 
-#define CLOSE_ELTS()  RADEON_NEWPRIM( rmesa )
+#define CLOSE_ELTS() if (0)  RADEON_NEWPRIM( rmesa )
 
 
 
@@ -174,15 +173,15 @@ static void radeonEmitPrim( GLcontext *ctx,
 		       GLuint start, 
 		       GLuint count)	
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r100ContextPtr rmesa = R100_CONTEXT( ctx );
    radeonTclPrimitive( ctx, prim, hwprim );
    
-   radeonEnsureCmdBufSpace( rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) +
-			    rmesa->hw.max_state_size + VBUF_BUFSZ );
+   rcommonEnsureCmdBufSpace( &rmesa->radeon,
+			     AOS_BUFSZ(rmesa->radeon.tcl.aos_count) +
+			     rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ );
 
    radeonEmitAOS( rmesa,
-		  rmesa->tcl.aos_components,
-		  rmesa->tcl.nr_aos_components,
+		  rmesa->radeon.tcl.aos_count,
 		  start );
    
    /* Why couldn't this packet have taken an offset param?
@@ -254,7 +253,7 @@ void radeonTclPrimitive( GLcontext *ctx,
 			 GLenum prim,
 			 int hw_prim )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint se_cntl;
    GLuint newprim = hw_prim | RADEON_CP_VC_CNTL_TCL_ENABLE;
 
@@ -371,7 +370,7 @@ radeonComputeFogBlendFactor( GLcontext *ctx, GLfloat fogcoord )
 static GLboolean radeon_run_tcl_render( GLcontext *ctx,
 					struct tnl_pipeline_stage *stage )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    GLuint inputs = VERT_BIT_POS | VERT_BIT_COLOR0;
@@ -379,7 +378,7 @@ static GLboolean radeon_run_tcl_render( GLcontext *ctx,
 
    /* TODO: separate this from the swtnl pipeline 
     */
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       return GL_TRUE;	/* fallback to software t&l */
 
    if (VB->Count == 0)
@@ -461,7 +460,7 @@ const struct tnl_pipeline_stage _radeon_tcl_stage =
 
 static void transition_to_swtnl( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLuint se_cntl;
 
@@ -490,7 +489,7 @@ static void transition_to_swtnl( GLcontext *ctx )
 
 static void transition_to_hwtnl( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
 
@@ -509,15 +508,15 @@ static void transition_to_hwtnl( GLcontext *ctx )
 
    tnl->Driver.NotifyMaterialChange = radeonUpdateMaterial;
 
-   if ( rmesa->dma.flush )			
-      rmesa->dma.flush( rmesa );	
+   if ( rmesa->radeon.dma.flush )			
+      rmesa->radeon.dma.flush( rmesa->radeon.glCtx );	
 
-   rmesa->dma.flush = NULL;
+   rmesa->radeon.dma.flush = NULL;
    rmesa->swtcl.vertex_format = 0;
    
-   if (rmesa->swtcl.indexed_verts.buf) 
-      radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
-			      __FUNCTION__ );
+   //   if (rmesa->swtcl.indexed_verts.buf) 
+   //      radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
+   //			      __FUNCTION__ );
 
    if (RADEON_DEBUG & DEBUG_FALLBACKS) 
       fprintf(stderr, "Radeon end tcl fallback\n");
@@ -550,11 +549,11 @@ static char *getFallbackString(GLuint bit)
 
 void radeonTclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->TclFallback;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   GLuint oldfallback = rmesa->radeon.TclFallback;
 
    if (mode) {
-      rmesa->TclFallback |= bit;
+      rmesa->radeon.TclFallback |= bit;
       if (oldfallback == 0) {
 	 if (RADEON_DEBUG & DEBUG_FALLBACKS) 
 	    fprintf(stderr, "Radeon begin tcl fallback %s\n",
@@ -563,7 +562,7 @@ void radeonTclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
       }
    }
    else {
-      rmesa->TclFallback &= ~bit;
+      rmesa->radeon.TclFallback &= ~bit;
       if (oldfallback == bit) {
 	 if (RADEON_DEBUG & DEBUG_FALLBACKS) 
 	    fprintf(stderr, "Radeon end tcl fallback %s\n",
diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c
index f2b6deb9c0..2549d5cb5c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tex.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tex.c
@@ -44,6 +44,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/texobj.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
 #include "radeon_swtcl.h"
@@ -170,10 +171,13 @@ static void radeonSetTexFilter( radeonTexObjPtr t, GLenum minf, GLenum magf )
 {
    GLuint anisotropy = (t->pp_txfilter & RADEON_MAX_ANISO_MASK);
 
+   /* Force revalidation to account for switches from/to mipmapping. */
+   t->validated = GL_FALSE;
+
    t->pp_txfilter &= ~(RADEON_MIN_FILTER_MASK | RADEON_MAG_FILTER_MASK);
 
    /* r100 chips can't handle mipmaps/aniso for cubemap/volume textures */
-   if ( t->base.tObj->Target == GL_TEXTURE_CUBE_MAP ) {
+   if ( t->base.Target == GL_TEXTURE_CUBE_MAP ) {
       switch ( minf ) {
       case GL_NEAREST:
       case GL_NEAREST_MIPMAP_NEAREST:
@@ -249,433 +253,13 @@ static void radeonSetTexBorderColor( radeonTexObjPtr t, const GLfloat color[4] )
    t->pp_border_color = radeonPackColor( 4, c[0], c[1], c[2], c[3] );
 }
 
-
-/**
- * Allocate space for and load the mesa images into the texture memory block.
- * This will happen before drawing with a new texture, or drawing with a
- * texture after it was swapped out or teximaged again.
- */
-
-static radeonTexObjPtr radeonAllocTexObj( struct gl_texture_object *texObj )
-{
-   radeonTexObjPtr t;
-
-   t = CALLOC_STRUCT( radeon_tex_obj );
-   texObj->DriverData = t;
-   if ( t != NULL ) {
-      if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
-	 fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)texObj, (void *)t );
-      }
-
-      /* Initialize non-image-dependent parts of the state:
-       */
-      t->base.tObj = texObj;
-      t->border_fallback = GL_FALSE;
-
-      t->pp_txfilter = RADEON_BORDER_MODE_OGL;
-      t->pp_txformat = (RADEON_TXFORMAT_ENDIAN_NO_SWAP |
-			RADEON_TXFORMAT_PERSPECTIVE_ENABLE);
-
-      make_empty_list( & t->base );
-
-      radeonSetTexWrap( t, texObj->WrapS, texObj->WrapT );
-      radeonSetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
-      radeonSetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
-      radeonSetTexBorderColor( t, texObj->BorderColor );
-   }
-
-   return t;
-}
-
-
-static const struct gl_texture_format *
-radeonChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
-                           GLenum format, GLenum type )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   const GLboolean do32bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32 );
-   const GLboolean force16bpt =
-       ( rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16 );
-   (void) format;
-
-   switch ( internalFormat ) {
-   case 4:
-   case GL_RGBA:
-   case GL_COMPRESSED_RGBA:
-      switch ( type ) {
-      case GL_UNSIGNED_INT_10_10_10_2:
-      case GL_UNSIGNED_INT_2_10_10_10_REV:
-	 return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      default:
-         return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_argb4444;
-      }
-
-   case 3:
-   case GL_RGB:
-   case GL_COMPRESSED_RGB:
-      switch ( type ) {
-      case GL_UNSIGNED_SHORT_4_4_4_4:
-      case GL_UNSIGNED_SHORT_4_4_4_4_REV:
-	 return _dri_texformat_argb4444;
-      case GL_UNSIGNED_SHORT_5_5_5_1:
-      case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-	 return _dri_texformat_argb1555;
-      case GL_UNSIGNED_SHORT_5_6_5:
-      case GL_UNSIGNED_SHORT_5_6_5_REV:
-	 return _dri_texformat_rgb565;
-      default:
-         return do32bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-      }
-
-   case GL_RGBA8:
-   case GL_RGB10_A2:
-   case GL_RGBA12:
-   case GL_RGBA16:
-      return !force16bpt ?
-	  _dri_texformat_argb8888 : _dri_texformat_argb4444;
-
-   case GL_RGBA4:
-   case GL_RGBA2:
-      return _dri_texformat_argb4444;
-
-   case GL_RGB5_A1:
-      return _dri_texformat_argb1555;
-
-   case GL_RGB8:
-   case GL_RGB10:
-   case GL_RGB12:
-   case GL_RGB16:
-      return !force16bpt ? _dri_texformat_argb8888 : _dri_texformat_rgb565;
-
-   case GL_RGB5:
-   case GL_RGB4:
-   case GL_R3_G3_B2:
-      return _dri_texformat_rgb565;
-
-   case GL_ALPHA:
-   case GL_ALPHA4:
-   case GL_ALPHA8:
-   case GL_ALPHA12:
-   case GL_ALPHA16:
-   case GL_COMPRESSED_ALPHA:
-      return _dri_texformat_a8;
-
-   case 1:
-   case GL_LUMINANCE:
-   case GL_LUMINANCE4:
-   case GL_LUMINANCE8:
-   case GL_LUMINANCE12:
-   case GL_LUMINANCE16:
-   case GL_COMPRESSED_LUMINANCE:
-      return _dri_texformat_l8;
-
-   case 2:
-   case GL_LUMINANCE_ALPHA:
-   case GL_LUMINANCE4_ALPHA4:
-   case GL_LUMINANCE6_ALPHA2:
-   case GL_LUMINANCE8_ALPHA8:
-   case GL_LUMINANCE12_ALPHA4:
-   case GL_LUMINANCE12_ALPHA12:
-   case GL_LUMINANCE16_ALPHA16:
-   case GL_COMPRESSED_LUMINANCE_ALPHA:
-      return _dri_texformat_al88;
-
-   case GL_INTENSITY:
-   case GL_INTENSITY4:
-   case GL_INTENSITY8:
-   case GL_INTENSITY12:
-   case GL_INTENSITY16:
-   case GL_COMPRESSED_INTENSITY:
-      return _dri_texformat_i8;
-
-   case GL_YCBCR_MESA:
-      if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
-          type == GL_UNSIGNED_BYTE)
-         return &_mesa_texformat_ycbcr;
-      else
-         return &_mesa_texformat_ycbcr_rev;
-
-   case GL_RGB_S3TC:
-   case GL_RGB4_S3TC:
-   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgb_dxt1;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
-      return &_mesa_texformat_rgba_dxt1;
-
-   case GL_RGBA_S3TC:
-   case GL_RGBA4_S3TC:
-   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
-      return &_mesa_texformat_rgba_dxt3;
-
-   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
-      return &_mesa_texformat_rgba_dxt5;
-
-   default:
-      _mesa_problem(ctx, "unexpected texture format in %s", __FUNCTION__);
-      return NULL;
-   }
-
-   return NULL; /* never get here */
-}
-
-
-static void radeonTexImage1D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_teximage1d(ctx, target, level, internalFormat,
-                          width, border, format, type, pixels,
-                          &ctx->Unpack, texObj, texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void radeonTexSubImage1D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset,
-                                 GLsizei width,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
-			     format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[0] |= (1 << level);
-}
-
-
-static void radeonTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLenum format, GLenum type, const GLvoid *pixels,
-                              const struct gl_pixelstore_attrib *packing,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_teximage2d(ctx, target, level, internalFormat,
-                          width, height, border, format, type, pixels,
-                          &ctx->Unpack, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-static void radeonTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format, GLenum type,
-                                 const GLvoid *pixels,
-                                 const struct gl_pixelstore_attrib *packing,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-			     height, format, type, pixels, packing, texObj,
-			     texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-static void radeonCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-                              GLint internalFormat,
-                              GLint width, GLint height, GLint border,
-                              GLsizei imageSize, const GLvoid *data,
-                              struct gl_texture_object *texObj,
-                              struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   if ( t != NULL ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexImage2D");
-         return;
-      }
-   }
-
-   /* Note, this will call ChooseTextureFormat */
-   _mesa_store_compressed_teximage2d(ctx, target, level, internalFormat, width,
-                                 height, border, imageSize, data, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
-
-static void radeonCompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
-                                 GLint xoffset, GLint yoffset,
-                                 GLsizei width, GLsizei height,
-                                 GLenum format,
-                                 GLsizei imageSize, const GLvoid *data,
-                                 struct gl_texture_object *texObj,
-                                 struct gl_texture_image *texImage )
-{
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
-   GLuint face;
-
-
-   /* which cube face or ordinary 2D image */
-   switch (target) {
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
-      ASSERT(face < 6);
-      break;
-   default:
-      face = 0;
-   }
-
-   assert( t ); /* this _should_ be true */
-   if ( t ) {
-      driSwapOutTextureObject( t );
-   }
-   else {
-      t = (driTextureObject *) radeonAllocTexObj( texObj );
-      if (!t) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexSubImage2D");
-         return;
-      }
-   }
-
-   _mesa_store_compressed_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
-                                 height, format, imageSize, data, texObj, texImage);
-
-   t->dirty_images[face] |= (1 << level);
-}
-
 #define SCALED_FLOAT_TO_BYTE( x, scale ) \
 		(((GLuint)((255.0F / scale) * (x))) / 2)
 
 static void radeonTexEnv( GLcontext *ctx, GLenum target,
 			  GLenum pname, const GLfloat *param )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint unit = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
 
@@ -706,7 +290,7 @@ static void radeonTexEnv( GLcontext *ctx, GLenum target,
        * functions, one mapping [-1.0,0.0] to [-128,0] and one mapping
        * [0.0,4.0] to [0,127].
        */
-      min = driQueryOptionb (&rmesa->optionCache, "no_neg_lod_bias") ?
+      min = driQueryOptionb (&rmesa->radeon.optionCache, "no_neg_lod_bias") ?
 	  0.0 : -1.0;
       bias = CLAMP( *param, min, 4.0 );
       if ( bias == 0 ) {
@@ -739,7 +323,7 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target,
 				struct gl_texture_object *texObj,
 				GLenum pname, const GLfloat *params )
 {
-   radeonTexObjPtr t = (radeonTexObjPtr) texObj->DriverData;
+   radeonTexObj* t = radeon_tex_obj(texObj);
 
    if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
       fprintf( stderr, "%s( %s )\n", __FUNCTION__,
@@ -767,57 +351,51 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target,
    case GL_TEXTURE_MAX_LEVEL:
    case GL_TEXTURE_MIN_LOD:
    case GL_TEXTURE_MAX_LOD:
+
       /* This isn't the most efficient solution but there doesn't appear to
        * be a nice alternative.  Since there's no LOD clamping,
        * we just have to rely on loading the right subset of mipmap levels
        * to simulate a clamped LOD.
        */
-      driSwapOutTextureObject( (driTextureObject *) t );
+      if (t->mt) {
+         radeon_miptree_unreference(t->mt);
+	 t->mt = 0;
+	 t->validated = GL_FALSE;
+      }
       break;
 
    default:
       return;
    }
-
-   /* Mark this texobj as dirty (one bit per tex unit)
-    */
-   t->dirty_state = TEX_ALL;
-}
-
-
-static void radeonBindTexture( GLcontext *ctx, GLenum target,
-			       struct gl_texture_object *texObj )
-{
-   if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
-      fprintf( stderr, "%s( %p ) unit=%d\n", __FUNCTION__, (void *)texObj,
-	       ctx->Texture.CurrentUnit );
-   }
-
-   assert( (target != GL_TEXTURE_1D && target != GL_TEXTURE_2D &&
-            target != GL_TEXTURE_RECTANGLE_NV && target != GL_TEXTURE_CUBE_MAP) ||
-           (texObj->DriverData != NULL) );
 }
 
-
 static void radeonDeleteTexture( GLcontext *ctx,
 				 struct gl_texture_object *texObj )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   driTextureObject * t = (driTextureObject *) texObj->DriverData;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   radeonTexObj* t = radeon_tex_obj(texObj);
+   int i;
 
    if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
       fprintf( stderr, "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj,
 	       _mesa_lookup_enum_by_nr( texObj->Target ) );
    }
 
-   if ( t != NULL ) {
-      if ( rmesa ) {
-         RADEON_FIREVERTICES( rmesa );
-      }
-
-      driDestroyTextureObject( t );
+   if ( rmesa ) {
+     radeon_firevertices(&rmesa->radeon);
+     for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) {
+       if ( t == rmesa->state.texture.unit[i].texobj ) {
+	 rmesa->state.texture.unit[i].texobj = NULL;
+	 rmesa->hw.tex[i].dirty = GL_FALSE;
+	 rmesa->hw.cube[i].dirty = GL_FALSE;
+       }
+     }
    }
 
+   if (t->mt) {
+      radeon_miptree_unreference(t->mt);
+      t->mt = 0;
+   }
    /* Free mipmap images and the texture object itself */
    _mesa_delete_texture_object(ctx, texObj);
 }
@@ -837,7 +415,7 @@ static void radeonTexGen( GLcontext *ctx,
 			  GLenum pname,
 			  const GLfloat *params )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLuint unit = ctx->Texture.CurrentUnit;
    rmesa->recheck_texgen[unit] = GL_TRUE;
 }
@@ -851,29 +429,40 @@ static void radeonTexGen( GLcontext *ctx,
 static struct gl_texture_object *
 radeonNewTextureObject( GLcontext *ctx, GLuint name, GLenum target )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_object *obj;
-   obj = _mesa_new_texture_object(ctx, name, target);
-   if (!obj)
-      return NULL;
-   obj->MaxAnisotropy = rmesa->initialMaxAnisotropy;
-   radeonAllocTexObj( obj );
-   return obj;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj);
+
+   _mesa_initialize_texture_object(&t->base, name, target);
+   t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy;
+
+   t->border_fallback = GL_FALSE;
+
+   t->pp_txfilter = RADEON_BORDER_MODE_OGL;
+   t->pp_txformat = (RADEON_TXFORMAT_ENDIAN_NO_SWAP |
+		     RADEON_TXFORMAT_PERSPECTIVE_ENABLE);
+   
+   radeonSetTexWrap( t, t->base.WrapS, t->base.WrapT );
+   radeonSetTexMaxAnisotropy( t, t->base.MaxAnisotropy );
+   radeonSetTexFilter( t, t->base.MinFilter, t->base.MagFilter );
+   radeonSetTexBorderColor( t, t->base.BorderColor );
+   return &t->base;
 }
 
 
+
 void radeonInitTextureFuncs( struct dd_function_table *functions )
 {
-   functions->ChooseTextureFormat	= radeonChooseTextureFormat;
+   functions->ChooseTextureFormat	= radeonChooseTextureFormat_mesa;
    functions->TexImage1D		= radeonTexImage1D;
    functions->TexImage2D		= radeonTexImage2D;
    functions->TexSubImage1D		= radeonTexSubImage1D;
    functions->TexSubImage2D		= radeonTexSubImage2D;
+   functions->GetTexImage               = radeonGetTexImage;
+   functions->GetCompressedTexImage     = radeonGetCompressedTexImage;
 
    functions->NewTextureObject		= radeonNewTextureObject;
-   functions->BindTexture		= radeonBindTexture;
+   //   functions->BindTexture		= radeonBindTexture;
    functions->DeleteTexture		= radeonDeleteTexture;
-   functions->IsTextureResident		= driIsTextureResident;
 
    functions->TexEnv			= radeonTexEnv;
    functions->TexParameter		= radeonTexParameter;
@@ -882,5 +471,12 @@ void radeonInitTextureFuncs( struct dd_function_table *functions )
    functions->CompressedTexImage2D	= radeonCompressedTexImage2D;
    functions->CompressedTexSubImage2D	= radeonCompressedTexSubImage2D;
 
+   functions->GenerateMipmap = radeonGenerateMipmap;
+
+   functions->NewTextureImage = radeonNewTextureImage;
+   functions->FreeTexImageData = radeonFreeTexImageData;
+   functions->MapTexture = radeonMapTexture;
+   functions->UnmapTexture = radeonUnmapTexture;
+
    driInitTextureFormats();
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.h b/src/mesa/drivers/dri/radeon/radeon_tex.h
index 8000880828..a4aaddc74f 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tex.h
+++ b/src/mesa/drivers/dri/radeon/radeon_tex.h
@@ -41,12 +41,16 @@ extern void radeonSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
                                unsigned long long offset, GLint depth,
                                GLuint pitch);
 
+extern void radeonSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv);
+extern void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+			       __DRIdrawable *dPriv);
+
 extern void radeonUpdateTextureState( GLcontext *ctx );
 
-extern int radeonUploadTexImages( radeonContextPtr rmesa, radeonTexObjPtr t,
+extern int radeonUploadTexImages( r100ContextPtr rmesa, radeonTexObjPtr t,
 				  GLuint face );
 
-extern void radeonDestroyTexObj( radeonContextPtr rmesa, radeonTexObjPtr t );
+extern void radeonDestroyTexObj( r100ContextPtr rmesa, radeonTexObjPtr t );
 
 extern void radeonInitTextureFuncs( struct dd_function_table *functions );
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_texmem.c b/src/mesa/drivers/dri/radeon/radeon_texmem.c
deleted file mode 100644
index 5f7bbe6a4c..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_texmem.c
+++ /dev/null
@@ -1,404 +0,0 @@
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     VA Linux Systems Inc., Fremont, California.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation on the rights to use, copy, modify, merge, publish,
-distribute, sub license, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
-SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Kevin E. Martin <martin@valinux.com>
- *   Gareth Hughes <gareth@valinux.com>
- *
- */
-#include <errno.h> 
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/macros.h"
-
-#include "radeon_context.h"
-#include "radeon_ioctl.h"
-#include "radeon_tex.h"
-
-#include <unistd.h>  /* for usleep() */
-
-
-/**
- * Destroy any device-dependent state associated with the texture.  This may
- * include NULLing out hardware state that points to the texture.
- */
-void
-radeonDestroyTexObj( radeonContextPtr rmesa, radeonTexObjPtr t )
-{
-   if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, (void *)t, (void *)t->base.tObj );
-   }
-
-   if ( rmesa != NULL ) {
-      unsigned   i;
-
-
-      for ( i = 0 ; i < rmesa->glCtx->Const.MaxTextureUnits ; i++ ) {
-	 if ( t == rmesa->state.texture.unit[i].texobj ) {
-	    rmesa->state.texture.unit[i].texobj = NULL;
-	 }
-      }
-   }
-}
-
-
-/* ------------------------------------------------------------
- * Texture image conversions
- */
-
-
-static void radeonUploadRectSubImage( radeonContextPtr rmesa,
-				      radeonTexObjPtr t, 
-				      struct gl_texture_image *texImage,
-				      GLint x, GLint y, 
-				      GLint width, GLint height )
-{
-   const struct gl_texture_format *texFormat = texImage->TexFormat;
-   int blit_format, dstPitch, done;
-
-   switch ( texFormat->TexelBytes ) {
-   case 1:
-      blit_format = RADEON_GMC_DST_8BPP_CI;
-      break;
-   case 2:
-      blit_format = RADEON_GMC_DST_16BPP;
-      break;
-   case 4:
-      blit_format = RADEON_GMC_DST_32BPP;
-      break;
-   default:
-      fprintf( stderr, "radeonUploadRectSubImage: unknown blit_format (texelbytes=%d)\n", 
-      	       texFormat->TexelBytes);
-      return;
-   }
-
-   t->image[0][0].data = texImage->Data;
-
-   /* Currently don't need to cope with small pitches.
-    */
-   width = texImage->Width;
-   height = texImage->Height;
-   dstPitch = t->pp_txpitch + 32;
-
-   {	/* FIXME: prefer GART-texturing if possible */
-      /* Data not in GART memory, or bad pitch.
-       */
-      for (done = 0; done < height ; ) {
-	 struct radeon_dma_region region;
-	 int lines = MIN2( height - done, RADEON_BUFFER_SIZE / dstPitch );
-	 int src_pitch;
-	 char *tex;
-
-         src_pitch = texImage->RowStride * texFormat->TexelBytes;
-
-	 tex = (char *)texImage->Data + done * src_pitch;
-
-	 memset(&region, 0, sizeof(region));
-	 radeonAllocDmaRegion( rmesa, &region, lines * dstPitch, 1024 );
-
-	 /* Copy texdata to dma:
-	  */
-	 if (0)
-	    fprintf(stderr, "%s: src_pitch %d dst_pitch %d\n",
-		    __FUNCTION__, src_pitch, dstPitch);
-
-	 if (src_pitch == dstPitch) {
-	    memcpy( region.address + region.start, tex, lines * src_pitch );
-	 } 
-	 else {
-	    char *buf = region.address + region.start;
-	    int i;
-	    for (i = 0 ; i < lines ; i++) {
-	       memcpy( buf, tex, src_pitch );
-	       buf += dstPitch;
-	       tex += src_pitch;
-	    }
-	 }
-
-	 radeonEmitWait( rmesa, RADEON_WAIT_3D );
-
-	 
-
-	 /* Blit to framebuffer
-	  */
-	 radeonEmitBlit( rmesa,
-		       blit_format,
-		       dstPitch, GET_START( &region ),
-		       dstPitch, t->bufAddr,
-		       0, 0,
-		       0, done,
-		       width, lines );
-	 
-	 radeonEmitWait( rmesa, RADEON_WAIT_2D );
-
-	 radeonReleaseDmaRegion( rmesa, &region, __FUNCTION__ );
-	 done += lines;
-      }
-   }
-}
-
-
-/**
- * Upload the texture image associated with texture \a t at the specified
- * level at the address relative to \a start.
- */
-static void uploadSubImage( radeonContextPtr rmesa, radeonTexObjPtr t, 
-			    GLint hwlevel,
-			    GLint x, GLint y, GLint width, GLint height,
-			    GLuint face )
-{
-   struct gl_texture_image *texImage = NULL;
-   GLuint offset;
-   GLint imageWidth, imageHeight;
-   GLint ret;
-   drm_radeon_texture_t tex;
-   drm_radeon_tex_image_t tmp;
-   const int level = hwlevel + t->base.firstLevel;
-
-   if ( RADEON_DEBUG & DEBUG_TEXTURE ) {
-      fprintf( stderr, "%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n", 
-	       __FUNCTION__, (void *)t, (void *)t->base.tObj, level, width, height, face );
-   }
-
-   ASSERT(face < 6);
-
-   /* Ensure we have a valid texture to upload */
-   if ( ( hwlevel < 0 ) || ( hwlevel >= RADEON_MAX_TEXTURE_LEVELS ) ) {
-      _mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
-      return;
-   }
-
-   texImage = t->base.tObj->Image[face][level];
-
-   if ( !texImage ) {
-      if ( RADEON_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: texImage %d is NULL!\n", __FUNCTION__, level );
-      return;
-   }
-   if ( !texImage->Data ) {
-      if ( RADEON_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is NULL!\n", __FUNCTION__ );
-      return;
-   }
-
-
-   if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-      assert(level == 0);
-      assert(hwlevel == 0);
-      if ( RADEON_DEBUG & DEBUG_TEXTURE )
-	 fprintf( stderr, "%s: image data is rectangular\n", __FUNCTION__);
-      radeonUploadRectSubImage( rmesa, t, texImage, x, y, width, height );
-      return;
-   }
-
-   imageWidth = texImage->Width;
-   imageHeight = texImage->Height;
-
-   offset = t->bufAddr + t->base.totalSize * face / 6;
-
-   if ( RADEON_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      GLint imageX = 0;
-      GLint imageY = 0;
-      GLint blitX = t->image[face][hwlevel].x;
-      GLint blitY = t->image[face][hwlevel].y;
-      GLint blitWidth = t->image[face][hwlevel].width;
-      GLint blitHeight = t->image[face][hwlevel].height;
-      fprintf( stderr, "   upload image: %d,%d at %d,%d\n",
-	       imageWidth, imageHeight, imageX, imageY );
-      fprintf( stderr, "   upload  blit: %d,%d at %d,%d\n",
-	       blitWidth, blitHeight, blitX, blitY );
-      fprintf( stderr, "       blit ofs: 0x%07x level: %d/%d\n",
-	       (GLuint)offset, hwlevel, level );
-   }
-
-   t->image[face][hwlevel].data = texImage->Data;
-
-   /* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
-    * NOTE: we're always use a 1KB-wide blit and I8 texture format.
-    * We used to use 1, 2 and 4-byte texels and used to use the texture
-    * width to dictate the blit width - but that won't work for compressed
-    * textures. (Brian)
-    * NOTE: can't do that with texture tiling. (sroland)
-    */
-   tex.offset = offset;
-   tex.image = &tmp;
-   /* copy (x,y,width,height,data) */
-   memcpy( &tmp, &t->image[face][hwlevel], sizeof(drm_radeon_tex_image_t) );
-
-   if (texImage->TexFormat->TexelBytes) {
-      /* use multi-byte upload scheme */
-      tex.height = imageHeight;
-      tex.width = imageWidth;
-      tex.format = t->pp_txformat & RADEON_TXFORMAT_FORMAT_MASK;
-      tex.pitch = MAX2((texImage->Width * texImage->TexFormat->TexelBytes) / 64, 1);
-      tex.offset += tmp.x & ~1023;
-      tmp.x = tmp.x % 1024;
-      if (t->tile_bits & RADEON_TXO_MICRO_TILE_X2) {
-	 /* need something like "tiled coordinates" ? */
-	 tmp.y = tmp.x / (tex.pitch * 128) * 2;
-	 tmp.x = tmp.x % (tex.pitch * 128) / 2 / texImage->TexFormat->TexelBytes;
-	 tex.pitch |= RADEON_DST_TILE_MICRO >> 22;
-      }
-      else {
-	 tmp.x = tmp.x >> (texImage->TexFormat->TexelBytes >> 1);
-      }
-      if ((t->tile_bits & RADEON_TXO_MACRO_TILE) &&
-	 (texImage->Width * texImage->TexFormat->TexelBytes >= 256)) {
-	 /* radeon switches off macro tiling for small textures/mipmaps it seems */
-	 tex.pitch |= RADEON_DST_TILE_MACRO >> 22;
-      }
-   }
-   else {
-      /* In case of for instance 8x8 texture (2x2 dxt blocks), padding after the first two blocks is
-         needed (only with dxt1 since 2 dxt3/dxt5 blocks already use 32 Byte). */
-      /* set tex.height to 1/4 since 1 "macropixel" (dxt-block) has 4 real pixels. Needed
-         so the kernel module reads the right amount of data. */
-      tex.format = RADEON_TXFORMAT_I8; /* any 1-byte texel format */
-      tex.pitch = (BLIT_WIDTH_BYTES / 64);
-      tex.height = (imageHeight + 3) / 4;
-      tex.width = (imageWidth + 3) / 4;
-      switch (t->pp_txformat & RADEON_TXFORMAT_FORMAT_MASK) {
-      case RADEON_TXFORMAT_DXT1:
-         tex.width *= 8;
-         break;
-      case RADEON_TXFORMAT_DXT23:
-      case RADEON_TXFORMAT_DXT45:
-         tex.width *= 16;
-         break;
-      }
-   }
-
-   LOCK_HARDWARE( rmesa );
-   do {
-      ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_TEXTURE,
-                                 &tex, sizeof(drm_radeon_texture_t) );
-   } while ( ret == -EAGAIN );
-
-   UNLOCK_HARDWARE( rmesa );
-
-   if ( ret ) {
-      fprintf( stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret );
-      fprintf( stderr, "   offset=0x%08x\n",
-	       offset );
-      fprintf( stderr, "   image width=%d height=%d\n",
-	       imageWidth, imageHeight );
-      fprintf( stderr, "    blit width=%d height=%d data=%p\n",
-	       t->image[face][hwlevel].width, t->image[face][hwlevel].height,
-	       t->image[face][hwlevel].data );
-      exit( 1 );
-   }
-}
-
-
-/**
- * Upload the texture images associated with texture \a t.  This might
- * require the allocation of texture memory.
- * 
- * \param rmesa Context pointer
- * \param t Texture to be uploaded
- * \param face Cube map face to be uploaded.  Zero for non-cube maps.
- */
-
-int radeonUploadTexImages( radeonContextPtr rmesa, radeonTexObjPtr t, GLuint face )
-{
-   int numLevels;
-
-   if ( !t || t->base.totalSize == 0 || t->image_override )
-      return 0;
-
-   if ( RADEON_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
-      fprintf( stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
-	       (void *)rmesa->glCtx, (void *)t->base.tObj, t->base.totalSize,
-	       t->base.firstLevel, t->base.lastLevel );
-   }
-
-   numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   if (RADEON_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      radeonFinish( rmesa->glCtx );
-   }
-
-   LOCK_HARDWARE( rmesa );
-
-   if ( t->base.memBlock == NULL ) {
-      int heap;
-
-      heap = driAllocateTexture( rmesa->texture_heaps, rmesa->nr_heaps,
-				 (driTextureObject *) t );
-      if ( heap == -1 ) {
-	 UNLOCK_HARDWARE( rmesa );
-	 return -1;
-      }
-
-      /* Set the base offset of the texture image */
-      t->bufAddr = rmesa->radeonScreen->texOffset[heap] 
-	   + t->base.memBlock->ofs;
-      t->pp_txoffset = t->bufAddr;
-
-      if (!(t->base.tObj->Image[0][0]->IsClientData)) {
-	 /* hope it's safe to add that here... */
-	 t->pp_txoffset |= t->tile_bits;
-      }
-
-      /* Mark this texobj as dirty on all units:
-       */
-      t->dirty_state = TEX_ALL;
-   }
-
-
-   /* Let the world know we've used this memory recently.
-    */
-   driUpdateTextureLRU( (driTextureObject *) t );
-   UNLOCK_HARDWARE( rmesa );
-
-
-   /* Upload any images that are new */
-   if (t->base.dirty_images[face]) {
-      int i;
-      for ( i = 0 ; i < numLevels ; i++ ) {
-         if ( (t->base.dirty_images[face] & (1 << (i+t->base.firstLevel))) != 0 ) {
-            uploadSubImage( rmesa, t, i, 0, 0, t->image[face][i].width,
-			    t->image[face][i].height, face );
-         }
-      }
-      t->base.dirty_images[face] = 0;
-   }
-
-   if (RADEON_DEBUG & DEBUG_SYNC) {
-      fprintf(stderr, "%s: Syncing\n", __FUNCTION__ );
-      radeonFinish( rmesa->glCtx );
-   }
-
-   return 0;
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c
index b165205c09..c29105d7b8 100644
--- a/src/mesa/drivers/dri/radeon/radeon_texstate.c
+++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c
@@ -39,10 +39,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "main/macros.h"
 #include "main/texformat.h"
+#include "main/teximage.h"
 #include "main/texobj.h"
 #include "main/enums.h"
 
 #include "radeon_context.h"
+#include "radeon_mipmap_tree.h"
 #include "radeon_state.h"
 #include "radeon_ioctl.h"
 #include "radeon_swtcl.h"
@@ -75,10 +77,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5) \
 			     && (tx_table[f].format != 0xffffffff) )
 
-static const struct {
+struct tx_table {
    GLuint format, filter;
-}
-tx_table[] =
+};
+
+static const struct tx_table tx_table[] =
 {
    _ALPHA(RGBA8888),
    _ALPHA_REV(RGBA8888),
@@ -111,252 +114,6 @@ tx_table[] =
 #undef _ALPHA
 #undef _INVALID
 
-/**
- * This function computes the number of bytes of storage needed for
- * the given texture object (all mipmap levels, all cube faces).
- * The \c image[face][level].x/y/width/height parameters for upload/blitting
- * are computed here.  \c pp_txfilter, \c pp_txformat, etc. will be set here
- * too.
- * 
- * \param rmesa Context pointer
- * \param tObj GL texture object whose images are to be posted to
- *                 hardware state.
- */
-static void radeonSetTexImages( radeonContextPtr rmesa,
-				struct gl_texture_object *tObj )
-{
-   radeonTexObjPtr t = (radeonTexObjPtr)tObj->DriverData;
-   const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel];
-   GLint curOffset, blitWidth;
-   GLint i, texelBytes;
-   GLint numLevels;
-   GLint log2Width, log2Height, log2Depth;
-
-   /* Set the hardware texture format
-    */
-   if ( !t->image_override ) {
-      t->pp_txformat &= ~(RADEON_TXFORMAT_FORMAT_MASK |
-                          RADEON_TXFORMAT_ALPHA_IN_MAP);
-      t->pp_txfilter &= ~RADEON_YUV_TO_RGB;
-
-      if ( VALID_FORMAT( baseImage->TexFormat->MesaFormat ) ) {
-         t->pp_txformat |= tx_table[ baseImage->TexFormat->MesaFormat ].format;
-         t->pp_txfilter |= tx_table[ baseImage->TexFormat->MesaFormat ].filter;
-      }
-      else {
-         _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__);
-         return;
-      }
-   }
-
-   texelBytes = baseImage->TexFormat->TexelBytes;
-
-   /* Compute which mipmap levels we really want to send to the hardware.
-    */
-
-   if (tObj->Target != GL_TEXTURE_CUBE_MAP)
-      driCalculateTextureFirstLastLevel( (driTextureObject *) t );
-   else {
-      /* r100 can't handle mipmaps for cube/3d textures, so don't waste
-         memory for them */
-      t->base.firstLevel = t->base.lastLevel = tObj->BaseLevel;
-   }
-   log2Width  = tObj->Image[0][t->base.firstLevel]->WidthLog2;
-   log2Height = tObj->Image[0][t->base.firstLevel]->HeightLog2;
-   log2Depth  = tObj->Image[0][t->base.firstLevel]->DepthLog2;
-
-   numLevels = t->base.lastLevel - t->base.firstLevel + 1;
-
-   assert(numLevels <= RADEON_MAX_TEXTURE_LEVELS);
-
-   /* Calculate mipmap offsets and dimensions for blitting (uploading)
-    * The idea is that we lay out the mipmap levels within a block of
-    * memory organized as a rectangle of width BLIT_WIDTH_BYTES.
-    */
-   curOffset = 0;
-   blitWidth = BLIT_WIDTH_BYTES;
-   t->tile_bits = 0;
-
-   /* figure out if this texture is suitable for tiling. */
-   if (texelBytes && (tObj->Target != GL_TEXTURE_RECTANGLE_NV)) {
-      if (rmesa->texmicrotile && (baseImage->Height > 1)) {
-	 /* allow 32 (bytes) x 1 mip (which will use two times the space
-	    the non-tiled version would use) max if base texture is large enough */
-	 if ((numLevels == 1) ||
-	   (((baseImage->Width * texelBytes / baseImage->Height) <= 32) &&
-	       (baseImage->Width * texelBytes > 64)) ||
-	    ((baseImage->Width * texelBytes / baseImage->Height) <= 16)) {
-	    /* R100 has two microtile bits (only the txoffset reg, not the blitter)
-	       weird: X2 + OPT: 32bit correct, 16bit completely hosed
-		      X2: 32bit correct, 16bit correct
-		      OPT: 32bit large mips correct, small mips hosed, 16bit completely hosed */
-	    t->tile_bits |= RADEON_TXO_MICRO_TILE_X2 /*| RADEON_TXO_MICRO_TILE_OPT*/;
-	 }
-      }
-      if ((baseImage->Width * texelBytes >= 256) && (baseImage->Height >= 16)) {
-	 /* R100 disables macro tiling only if mip width is smaller than 256 bytes, and not
-	    in the case if height is smaller than 16 (not 100% sure), as does the r200,
-	    so need to disable macro tiling in that case */
-	 if ((numLevels == 1) || ((baseImage->Width * texelBytes / baseImage->Height) <= 4)) {
-	    t->tile_bits |= RADEON_TXO_MACRO_TILE;
-	 }
-      }
-   }
-
-   for (i = 0; i < numLevels; i++) {
-      const struct gl_texture_image *texImage;
-      GLuint size;
-
-      texImage = tObj->Image[0][i + t->base.firstLevel];
-      if ( !texImage )
-	 break;
-
-      /* find image size in bytes */
-      if (texImage->IsCompressed) {
-      /* need to calculate the size AFTER padding even though the texture is
-         submitted without padding.
-         Only handle pot textures currently - don't know if npot is even possible,
-         size calculation would certainly need (trivial) adjustments.
-         Align (and later pad) to 32byte, not sure what that 64byte blit width is
-         good for? */
-         if ((t->pp_txformat & RADEON_TXFORMAT_FORMAT_MASK) == RADEON_TXFORMAT_DXT1) {
-            /* RGB_DXT1/RGBA_DXT1, 8 bytes per block */
-            if ((texImage->Width + 3) < 8) /* width one block */
-               size = texImage->CompressedSize * 4;
-            else if ((texImage->Width + 3) < 16)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-         }
-         else /* DXT3/5, 16 bytes per block */
-            if ((texImage->Width + 3) < 8)
-               size = texImage->CompressedSize * 2;
-            else size = texImage->CompressedSize;
-      }
-      else if (tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
-	 size = ((texImage->Width * texelBytes + 63) & ~63) * texImage->Height;
-      }
-      else if (t->tile_bits & RADEON_TXO_MICRO_TILE_X2) {
-	 /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned,
-	    though the actual offset may be different (if texture is less than
-	    32 bytes width) to the untiled case */
-	 int w = (texImage->Width * texelBytes * 2 + 31) & ~31;
-	 size = (w * ((texImage->Height + 1) / 2)) * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      else {
-	 int w = (texImage->Width * texelBytes + 31) & ~31;
-	 size = w * texImage->Height * texImage->Depth;
-	 blitWidth = MAX2(texImage->Width, 64 / texelBytes);
-      }
-      assert(size > 0);
-
-      /* Align to 32-byte offset.  It is faster to do this unconditionally
-       * (no branch penalty).
-       */
-
-      curOffset = (curOffset + 0x1f) & ~0x1f;
-
-      if (texelBytes) {
-	 t->image[0][i].x = curOffset; /* fix x and y coords up later together with offset */
-	 t->image[0][i].y = 0;
-	 t->image[0][i].width = MIN2(size / texelBytes, blitWidth);
-	 t->image[0][i].height = (size / texelBytes) / t->image[0][i].width;
-      }
-      else {
-         t->image[0][i].x = curOffset % BLIT_WIDTH_BYTES;
-         t->image[0][i].y = curOffset / BLIT_WIDTH_BYTES;
-         t->image[0][i].width  = MIN2(size, BLIT_WIDTH_BYTES);
-         t->image[0][i].height = size / t->image[0][i].width;     
-      }
-
-#if 0
-      /* for debugging only and only  applicable to non-rectangle targets */
-      assert(size % t->image[0][i].width == 0);
-      assert(t->image[0][i].x == 0
-             || (size < BLIT_WIDTH_BYTES && t->image[0][i].height == 1));
-#endif
-
-      if (0)
-         fprintf(stderr,
-                 "level %d: %dx%d x=%d y=%d w=%d h=%d size=%d at %d\n",
-                 i, texImage->Width, texImage->Height,
-                 t->image[0][i].x, t->image[0][i].y,
-                 t->image[0][i].width, t->image[0][i].height, size, curOffset);
-
-      curOffset += size;
-
-   }
-
-   /* Align the total size of texture memory block.
-    */
-   t->base.totalSize = (curOffset + RADEON_OFFSET_MASK) & ~RADEON_OFFSET_MASK;
-
-   /* Setup remaining cube face blits, if needed */
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      const GLuint faceSize = t->base.totalSize;
-      GLuint face;
-      /* reuse face 0 x/y/width/height - just update the offset when uploading */
-      for (face = 1; face < 6; face++) {
-         for (i = 0; i < numLevels; i++) {
-            t->image[face][i].x =  t->image[0][i].x;
-            t->image[face][i].y =  t->image[0][i].y;
-            t->image[face][i].width  = t->image[0][i].width;
-            t->image[face][i].height = t->image[0][i].height;
-         }
-      }
-      t->base.totalSize = 6 * faceSize; /* total texmem needed */
-   }
-
-   /* Hardware state:
-    */
-   t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK;
-   t->pp_txfilter |= (numLevels - 1) << RADEON_MAX_MIP_LEVEL_SHIFT;
-
-   t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK |
-		       RADEON_TXFORMAT_HEIGHT_MASK |
-                       RADEON_TXFORMAT_CUBIC_MAP_ENABLE |
-                       RADEON_TXFORMAT_F5_WIDTH_MASK |
-                       RADEON_TXFORMAT_F5_HEIGHT_MASK);
-   t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) |
-		      (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT));
-
-   if (tObj->Target == GL_TEXTURE_CUBE_MAP) {
-      assert(log2Width == log2Height);
-      t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) |
-                         (log2Height << RADEON_TXFORMAT_F5_HEIGHT_SHIFT) |
-                         (RADEON_TXFORMAT_CUBIC_MAP_ENABLE));
-      t->pp_cubic_faces = ((log2Width << RADEON_FACE_WIDTH_1_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_1_SHIFT) |
-                           (log2Width << RADEON_FACE_WIDTH_2_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_2_SHIFT) |
-                           (log2Width << RADEON_FACE_WIDTH_3_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_3_SHIFT) |
-                           (log2Width << RADEON_FACE_WIDTH_4_SHIFT) |
-                           (log2Height << RADEON_FACE_HEIGHT_4_SHIFT));
-   }
-
-   t->pp_txsize = (((tObj->Image[0][t->base.firstLevel]->Width - 1) << 0) |
-                   ((tObj->Image[0][t->base.firstLevel]->Height - 1) << 16));
-
-   /* Only need to round to nearest 32 for textures, but the blitter
-    * requires 64-byte aligned pitches, and we may/may not need the
-    * blitter.   NPOT only!
-    */
-   if ( !t->image_override ) {
-      if (baseImage->IsCompressed)
-         t->pp_txpitch = (tObj->Image[0][t->base.firstLevel]->Width + 63) & ~(63);
-      else
-         t->pp_txpitch = ((tObj->Image[0][t->base.firstLevel]->Width * texelBytes) + 63) & ~(63);
-      t->pp_txpitch -= 32;
-   }
-
-   t->dirty_state = TEX_ALL;
-
-   /* FYI: radeonUploadTexImages( rmesa, t ); used to be called here */
-}
-
-
-
 /* ================================================================
  * Texture combine functions
  */
@@ -503,7 +260,7 @@ do {							\
 
 static GLboolean radeonUpdateTextureEnv( GLcontext *ctx, int unit )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
    GLuint color_combine, alpha_combine;
    const GLuint color_combine0 = RADEON_COLOR_ARG_A_ZERO | RADEON_COLOR_ARG_B_ZERO
@@ -846,22 +603,21 @@ static GLboolean radeonUpdateTextureEnv( GLcontext *ctx, int unit )
 void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname,
                         unsigned long long offset, GLint depth, GLuint pitch)
 {
-	radeonContextPtr rmesa = pDRICtx->driverPrivate;
+	r100ContextPtr rmesa = pDRICtx->driverPrivate;
 	struct gl_texture_object *tObj =
-	    _mesa_lookup_texture(rmesa->glCtx, texname);
-	radeonTexObjPtr t;
+	    _mesa_lookup_texture(rmesa->radeon.glCtx, texname);
+	radeonTexObjPtr t = radeon_tex_obj(tObj);
 
 	if (tObj == NULL)
 		return;
 
-	t = (radeonTexObjPtr) tObj->DriverData;
-
 	t->image_override = GL_TRUE;
 
 	if (!offset)
 		return;
-
-	t->pp_txoffset = offset;
+	
+	t->bo = NULL;
+	t->override_offset = offset;
 	t->pp_txpitch = pitch - 32;
 
 	switch (depth) {
@@ -881,6 +637,125 @@ void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname,
 	}
 }
 
+void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_format,
+			 __DRIdrawable *dPriv)
+{
+	struct gl_texture_unit *texUnit;
+	struct gl_texture_object *texObj;
+	struct gl_texture_image *texImage;
+	struct radeon_renderbuffer *rb;
+	radeon_texture_image *rImage;
+	radeonContextPtr radeon;
+	r100ContextPtr rmesa;
+	struct radeon_framebuffer *rfb;
+	radeonTexObjPtr t;
+	uint32_t pitch_val;
+	uint32_t internalFormat, type, format;
+
+	type = GL_BGRA;
+	format = GL_UNSIGNED_BYTE;
+	internalFormat = (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT ? 3 : 4);
+
+	radeon = pDRICtx->driverPrivate;
+	rmesa = pDRICtx->driverPrivate;
+
+	rfb = dPriv->driverPrivate;
+        texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit];
+	texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target);
+        texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0);
+
+	rImage = get_radeon_texture_image(texImage);
+	t = radeon_tex_obj(texObj);
+        if (t == NULL) {
+    	    return;
+    	}
+
+	radeon_update_renderbuffers(pDRICtx, dPriv);
+	/* back & depth buffer are useless free them right away */
+	rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+        rb->bo = NULL;
+	}
+	rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+	if (rb && rb->bo) {
+		radeon_bo_unref(rb->bo);
+		rb->bo = NULL;
+	}
+	rb = rfb->color_rb[0];
+	if (rb->bo == NULL) {
+		/* Failed to BO for the buffer */
+		return;
+	}
+	
+	_mesa_lock_texture(radeon->glCtx, texObj);
+	if (t->bo) {
+		radeon_bo_unref(t->bo);
+		t->bo = NULL;
+	}
+	if (rImage->bo) {
+		radeon_bo_unref(rImage->bo);
+		rImage->bo = NULL;
+	}
+	if (t->mt) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = NULL;
+	}
+	if (rImage->mt) {
+		radeon_miptree_unreference(rImage->mt);
+		rImage->mt = NULL;
+	}
+	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
+				   rb->base.Width, rb->base.Height, 1, 0, rb->cpp);
+	texImage->RowStride = rb->pitch / rb->cpp;
+	texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx,
+							internalFormat,
+							type, format, 0);
+	rImage->bo = rb->bo;
+	radeon_bo_ref(rImage->bo);
+	t->bo = rb->bo;
+	radeon_bo_ref(t->bo);
+	t->tile_bits = 0;
+	t->image_override = GL_TRUE;
+	t->override_offset = 0;
+	t->pp_txpitch &= (1 << 13) -1;
+	pitch_val = rb->pitch;
+	switch (rb->cpp) {
+	case 4:
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = tx_table[MESA_FORMAT_RGB888].format;
+		else
+			t->pp_txformat = tx_table[MESA_FORMAT_ARGB8888].format;
+		t->pp_txfilter |= tx_table[MESA_FORMAT_ARGB8888].filter;
+		break;
+	case 3:
+	default:
+		t->pp_txformat = tx_table[MESA_FORMAT_RGB888].format;
+		t->pp_txfilter |= tx_table[MESA_FORMAT_RGB888].filter;
+		break;
+	case 2:
+		t->pp_txformat = tx_table[MESA_FORMAT_RGB565].format;
+		t->pp_txfilter |= tx_table[MESA_FORMAT_RGB565].filter;
+		break;
+	}
+        t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT);
+        t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
+	t->pp_txpitch = pitch_val;
+        t->pp_txpitch -= 32;
+
+	t->validated = GL_TRUE;
+	_mesa_unlock_texture(radeon->glCtx, texObj);
+	return;
+}
+
+
+void radeonSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+        radeonSetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
+
+
 #define TEXOBJ_TXFILTER_MASK (RADEON_MAX_MIP_LEVEL_MASK |	\
 			      RADEON_MIN_FILTER_MASK | 		\
 			      RADEON_MAG_FILTER_MASK |		\
@@ -901,12 +776,53 @@ void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname,
                               RADEON_TXFORMAT_NON_POWER2)
 
 
-static void import_tex_obj_state( radeonContextPtr rmesa,
+static void disable_tex_obj_state( r100ContextPtr rmesa, 
+				   int unit )
+{
+   RADEON_STATECHANGE( rmesa, tex[unit] );
+
+   RADEON_STATECHANGE( rmesa, tcl );
+   rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] &= ~(RADEON_ST_BIT(unit) |
+					     RADEON_Q_BIT(unit));
+   
+   if (rmesa->radeon.TclFallback & (RADEON_TCL_FALLBACK_TEXGEN_0<<unit)) {
+     TCL_FALLBACK( rmesa->radeon.glCtx, (RADEON_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
+     rmesa->recheck_texgen[unit] = GL_TRUE;
+   }
+
+   if (rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) {
+     /* this seems to be a genuine (r100 only?) hw bug. Need to remove the
+	cubic_map bit on unit 2 when the unit is disabled, otherwise every
+	2nd (2d) mipmap on unit 0 will be broken (may not be needed for other
+	units, better be safe than sorry though).*/
+     RADEON_STATECHANGE( rmesa, tex[unit] );
+     rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] &= ~RADEON_TXFORMAT_CUBIC_MAP_ENABLE;
+   }
+
+   {
+      GLuint inputshift = RADEON_TEXGEN_0_INPUT_SHIFT + unit*4;
+      GLuint tmp = rmesa->TexGenEnabled;
+
+      rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenEnabled &= ~(RADEON_TEXMAT_0_ENABLE<<unit);
+      rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_INPUT_MASK<<inputshift);
+      rmesa->TexGenNeedNormals[unit] = 0;
+      rmesa->TexGenEnabled |= 
+	(RADEON_TEXGEN_INPUT_TEXCOORD_0+unit) << inputshift;
+
+      if (tmp != rmesa->TexGenEnabled) {
+	rmesa->recheck_texgen[unit] = GL_TRUE;
+	rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
+      }
+   }
+}
+
+static void import_tex_obj_state( r100ContextPtr rmesa,
 				  int unit,
 				  radeonTexObjPtr texobj )
 {
 /* do not use RADEON_DB_STATE to avoid stale texture caches */
-   int *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
+   uint32_t *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0];
    GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
 
    RADEON_STATECHANGE( rmesa, tex[unit] );
@@ -915,10 +831,9 @@ static void import_tex_obj_state( radeonContextPtr rmesa,
    cmd[TEX_PP_TXFILTER] |= texobj->pp_txfilter & TEXOBJ_TXFILTER_MASK;
    cmd[TEX_PP_TXFORMAT] &= ~TEXOBJ_TXFORMAT_MASK;
    cmd[TEX_PP_TXFORMAT] |= texobj->pp_txformat & TEXOBJ_TXFORMAT_MASK;
-   cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset;
    cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color;
 
-   if (texobj->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
+   if (texobj->base.Target == GL_TEXTURE_RECTANGLE_NV) {
       GLuint *txr_cmd = RADEON_DB_STATE( txr[unit] );
       txr_cmd[TXR_PP_TEX_SIZE] = texobj->pp_txsize; /* NPOT only! */
       txr_cmd[TXR_PP_TEX_PITCH] = texobj->pp_txpitch; /* NPOT only! */
@@ -928,22 +843,12 @@ static void import_tex_obj_state( radeonContextPtr rmesa,
    else {
       se_coord_fmt &= ~(RADEON_VTX_ST0_NONPARAMETRIC << unit);
 
-      if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) {
-	 int *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
-	 GLuint bytesPerFace = texobj->base.totalSize / 6;
-	 ASSERT(texobj->base.totalSize % 6 == 0);
+      if (texobj->base.Target == GL_TEXTURE_CUBE_MAP) {
+	 uint32_t *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0];
 
 	 RADEON_STATECHANGE( rmesa, cube[unit] );
 	 cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces;
-	 /* dont know if this setup conforms to OpenGL.. 
-	  * at least it matches the behavior of mesa software renderer
-	  */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_0] = texobj->pp_txoffset; /* right */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_1] = texobj->pp_txoffset + 1 * bytesPerFace; /* left */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_2] = texobj->pp_txoffset + 2 * bytesPerFace; /* top */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_3] = texobj->pp_txoffset + 3 * bytesPerFace; /* bottom */
-	 cube_cmd[CUBE_PP_CUBIC_OFFSET_4] = texobj->pp_txoffset + 4 * bytesPerFace; /* front */
-	 cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset + 5 * bytesPerFace; /* back */
+	 /* state filled out in the cube_emit */
       }
    }
 
@@ -952,13 +857,11 @@ static void import_tex_obj_state( radeonContextPtr rmesa,
       rmesa->hw.set.cmd[SET_SE_COORDFMT] = se_coord_fmt;
    }
 
-   texobj->dirty_state &= ~(1<<unit);
+   rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
 }
 
 
-
-
-static void set_texgen_matrix( radeonContextPtr rmesa, 
+static void set_texgen_matrix( r100ContextPtr rmesa, 
 			       GLuint unit,
 			       const GLfloat *s_plane,
 			       const GLfloat *t_plane,
@@ -986,14 +889,14 @@ static void set_texgen_matrix( radeonContextPtr rmesa,
    rmesa->TexGenMatrix[unit].m[15] = q_plane[3];
 
    rmesa->TexGenEnabled |= RADEON_TEXMAT_0_ENABLE << unit;
-   rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+   rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
 }
 
 /* Returns GL_FALSE if fallback required.
  */
 static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
    GLuint inputshift = RADEON_TEXGEN_0_INPUT_SHIFT + unit*4;
    GLuint tmp = rmesa->TexGenEnabled;
@@ -1094,283 +997,189 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit )
    }
 
    if (tmp != rmesa->TexGenEnabled) {
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 
    return GL_TRUE;
 }
 
-
-static void disable_tex( GLcontext *ctx, int unit )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (RADEON_TEX_0_ENABLE<<unit)) {
-      /* Texture unit disabled */
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
-
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= ~(1UL << unit);
-	 rmesa->state.texture.unit[unit].texobj = NULL;
-      }
-
-      RADEON_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= 
-	  ~((RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE) << unit);
-
-      RADEON_STATECHANGE( rmesa, tcl );
-      rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] &= ~(RADEON_ST_BIT(unit) |
-						RADEON_Q_BIT(unit));
-
-      if (rmesa->TclFallback & (RADEON_TCL_FALLBACK_TEXGEN_0<<unit)) {
-	 TCL_FALLBACK( ctx, (RADEON_TCL_FALLBACK_TEXGEN_0<<unit), GL_FALSE);
-	 rmesa->recheck_texgen[unit] = GL_TRUE;
-      }
-
-      if (rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) {
-      /* this seems to be a genuine (r100 only?) hw bug. Need to remove the
-         cubic_map bit on unit 2 when the unit is disabled, otherwise every
-	 2nd (2d) mipmap on unit 0 will be broken (may not be needed for other
-	 units, better be safe than sorry though).*/
-	 RADEON_STATECHANGE( rmesa, tex[unit] );
-	 rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT] &= ~RADEON_TXFORMAT_CUBIC_MAP_ENABLE;
-      }
-
-      {
-	 GLuint inputshift = RADEON_TEXGEN_0_INPUT_SHIFT + unit*4;
-	 GLuint tmp = rmesa->TexGenEnabled;
-
-	 rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenEnabled &= ~(RADEON_TEXMAT_0_ENABLE<<unit);
-	 rmesa->TexGenEnabled &= ~(RADEON_TEXGEN_INPUT_MASK<<inputshift);
-	 rmesa->TexGenNeedNormals[unit] = 0;
-	 rmesa->TexGenEnabled |= 
-	     (RADEON_TEXGEN_INPUT_TEXCOORD_0+unit) << inputshift;
-
-	 if (tmp != rmesa->TexGenEnabled) {
-	    rmesa->recheck_texgen[unit] = GL_TRUE;
-	    rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
-	 }
-      }
-   }
-}
-
-static GLboolean enable_tex_2d( GLcontext *ctx, int unit )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
-
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & RADEON_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~RADEON_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
-   }
-
-   ASSERT(tObj->Target == GL_TEXTURE_2D || tObj->Target == GL_TEXTURE_1D);
-
-   if ( t->base.dirty_images[0] ) {
-      RADEON_FIREVERTICES( rmesa );
-      radeonSetTexImages( rmesa, tObj );
-      radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock && !t->image_override ) 
-	return GL_FALSE;
-   }
-
-   return GL_TRUE;
-}
-
-static GLboolean enable_tex_cube( GLcontext *ctx, int unit )
+/**
+ * Compute the cached hardware register values for the given texture object.
+ *
+ * \param rmesa Context pointer
+ * \param t the r300 texture object
+ */
+static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int unit)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
-   GLuint face;
+   const struct gl_texture_image *firstImage;
+   GLint log2Width, log2Height, log2Depth, texelBytes;
 
-   /* Need to load the 2d images associated with this unit.
-    */
-   if (t->pp_txformat & RADEON_TXFORMAT_NON_POWER2) {
-      t->pp_txformat &= ~RADEON_TXFORMAT_NON_POWER2;
-      for (face = 0; face < 6; face++)
-         t->base.dirty_images[face] = ~0;
+   if ( t->bo ) {
+	return GL_TRUE;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_CUBE_MAP);
+   firstImage = t->base.Image[0][t->mt->firstLevel];   
 
-   if ( t->base.dirty_images[0] || t->base.dirty_images[1] ||
-        t->base.dirty_images[2] || t->base.dirty_images[3] ||
-        t->base.dirty_images[4] || t->base.dirty_images[5] ) {
-      /* flush */
-      RADEON_FIREVERTICES( rmesa );
-      /* layout memory space, once for all faces */
-      radeonSetTexImages( rmesa, tObj );
+   if (firstImage->Border > 0) {
+      fprintf(stderr, "%s: border\n", __FUNCTION__);
+      return GL_FALSE;
    }
 
-   /* upload (per face) */
-   for (face = 0; face < 6; face++) {
-      if (t->base.dirty_images[face]) {
-         radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, face );
+   log2Width  = firstImage->WidthLog2;
+   log2Height = firstImage->HeightLog2;
+   log2Depth  = firstImage->DepthLog2;
+   texelBytes = firstImage->TexFormat->TexelBytes;
+
+   if (!t->image_override) {
+      if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) {
+	const struct tx_table *table = tx_table;
+
+	 t->pp_txformat &= ~(RADEON_TXFORMAT_FORMAT_MASK |
+			     RADEON_TXFORMAT_ALPHA_IN_MAP);
+	 t->pp_txfilter &= ~RADEON_YUV_TO_RGB;	 
+	 
+	 t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format;
+	 t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter;
+      } else {
+	 _mesa_problem(NULL, "unexpected texture format in %s",
+		       __FUNCTION__);
+	 return GL_FALSE;
       }
    }
-      
-   if ( !t->base.memBlock ) {
-      /* texmem alloc failed, use s/w fallback */
-      return GL_FALSE;
+   
+   t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK;
+   t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << RADEON_MAX_MIP_LEVEL_SHIFT;
+	
+   t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK |
+		       RADEON_TXFORMAT_HEIGHT_MASK |
+		       RADEON_TXFORMAT_CUBIC_MAP_ENABLE |
+		       RADEON_TXFORMAT_F5_WIDTH_MASK |
+		       RADEON_TXFORMAT_F5_HEIGHT_MASK);
+   t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) |
+		      (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT));
+   
+   t->tile_bits = 0;
+   
+   if (t->base.Target == GL_TEXTURE_CUBE_MAP) {
+      ASSERT(log2Width == log2Height);
+      t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) |
+			 (log2Height << RADEON_TXFORMAT_F5_HEIGHT_SHIFT) |
+			 /* don't think we need this bit, if it exists at all - fglrx does not set it */
+			 (RADEON_TXFORMAT_CUBIC_MAP_ENABLE));
+      t->pp_cubic_faces = ((log2Width << RADEON_FACE_WIDTH_1_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_1_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_2_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_2_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_3_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_3_SHIFT) |
+                           (log2Width << RADEON_FACE_WIDTH_4_SHIFT) |
+                           (log2Height << RADEON_FACE_HEIGHT_4_SHIFT));
    }
 
-   return GL_TRUE;
-}
-
-static GLboolean enable_tex_rect( GLcontext *ctx, int unit )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
+   t->pp_txsize = (((firstImage->Width - 1) << RADEON_TEX_USIZE_SHIFT)
+		   | ((firstImage->Height - 1) << RADEON_TEX_VSIZE_SHIFT));
 
-   if (!(t->pp_txformat & RADEON_TXFORMAT_NON_POWER2)) {
-      t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
-      t->base.dirty_images[0] = ~0;
+   if ( !t->image_override ) {
+      if (firstImage->IsCompressed)
+         t->pp_txpitch = (firstImage->Width + 63) & ~(63);
+      else
+         t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63);
+      t->pp_txpitch -= 32;
    }
 
-   ASSERT(tObj->Target == GL_TEXTURE_RECTANGLE_NV);
-
-   if ( t->base.dirty_images[0] ) {
-      RADEON_FIREVERTICES( rmesa );
-      radeonSetTexImages( rmesa, tObj );
-      radeonUploadTexImages( rmesa, (radeonTexObjPtr) tObj->DriverData, 0 );
-      if ( !t->base.memBlock &&
-           !t->image_override /* && !rmesa->prefer_gart_client_texturing  FIXME */ ) {
-	 fprintf(stderr, "%s: upload failed\n", __FUNCTION__);
-	 return GL_FALSE;
-      }
+   if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) {
+      t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2;
    }
 
    return GL_TRUE;
 }
 
-
-static GLboolean update_tex_common( GLcontext *ctx, int unit )
+static GLboolean radeon_validate_texture(GLcontext *ctx, struct gl_texture_object *texObj, int unit)
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-   struct gl_texture_object *tObj = texUnit->_Current;
-   radeonTexObjPtr t = (radeonTexObjPtr) tObj->DriverData;
-   GLenum format;
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
+   radeonTexObj *t = radeon_tex_obj(texObj);
+   int ret;
 
-   /* Fallback if there's a texture border */
-   if ( tObj->Image[0][tObj->BaseLevel]->Border > 0 ) {
-      fprintf(stderr, "%s: border\n", __FUNCTION__);
+   if (!radeon_validate_texture_miptree(ctx, texObj))
       return GL_FALSE;
-   }
+
+   ret = setup_hardware_state(rmesa, t, unit);
+   if (ret == GL_FALSE)
+     return GL_FALSE;
+
    /* yuv conversion only works in first unit */
    if (unit != 0 && (t->pp_txfilter & RADEON_YUV_TO_RGB))
       return GL_FALSE;
 
-   /* Update state if this is a different texture object to last
-    * time.
-    */
-   if ( rmesa->state.texture.unit[unit].texobj != t ) {
-      if ( rmesa->state.texture.unit[unit].texobj != NULL ) {
-	 /* The old texture is no longer bound to this texture unit.
-	  * Mark it as such.
-	  */
-
-	 rmesa->state.texture.unit[unit].texobj->base.bound &= 
-	     ~(1UL << unit);
-      }
-
-      rmesa->state.texture.unit[unit].texobj = t;
-      t->base.bound |= (1UL << unit);
-      t->dirty_state |= 1<<unit;
-      driUpdateTextureLRU( (driTextureObject *) t ); /* XXX: should be locked! */
-   }
+   RADEON_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= 
+     (RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE) << unit;
 
+   RADEON_STATECHANGE( rmesa, tcl );
+   rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] |= RADEON_ST_BIT(unit);
 
-   /* Newly enabled?
-    */
-   if ( !(rmesa->hw.ctx.cmd[CTX_PP_CNTL] & (RADEON_TEX_0_ENABLE<<unit))) {
-      RADEON_STATECHANGE( rmesa, ctx );
-      rmesa->hw.ctx.cmd[CTX_PP_CNTL] |= 
-	  (RADEON_TEX_0_ENABLE | RADEON_TEX_BLEND_0_ENABLE) << unit;
+   rmesa->recheck_texgen[unit] = GL_TRUE;
 
-      RADEON_STATECHANGE( rmesa, tcl );
-
-      rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] |= RADEON_ST_BIT(unit);
-
-      rmesa->recheck_texgen[unit] = GL_TRUE;
-   }
-
-   if (t->dirty_state & (1<<unit)) {
-      import_tex_obj_state( rmesa, unit, t );
-      /* may need to update texture matrix (for texrect adjustments) */
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
-   }
+   import_tex_obj_state( rmesa, unit, t );
 
    if (rmesa->recheck_texgen[unit]) {
       GLboolean fallback = !radeon_validate_texgen( ctx, unit );
       TCL_FALLBACK( ctx, (RADEON_TCL_FALLBACK_TEXGEN_0<<unit), fallback);
       rmesa->recheck_texgen[unit] = 0;
-      rmesa->NewGLState |= _NEW_TEXTURE_MATRIX;
+      rmesa->radeon.NewGLState |= _NEW_TEXTURE_MATRIX;
    }
 
-   format = tObj->Image[0][tObj->BaseLevel]->_BaseFormat;
-   if ( rmesa->state.texture.unit[unit].format != format ||
-	rmesa->state.texture.unit[unit].envMode != texUnit->EnvMode ) {
-      rmesa->state.texture.unit[unit].format = format;
-      rmesa->state.texture.unit[unit].envMode = texUnit->EnvMode;
-      if ( ! radeonUpdateTextureEnv( ctx, unit ) ) {
-	 return GL_FALSE;
-      }
+   if ( ! radeonUpdateTextureEnv( ctx, unit ) ) {
+     return GL_FALSE;
    }
-
    FALLBACK( rmesa, RADEON_FALLBACK_BORDER_MODE, t->border_fallback );
+
+   t->validated = GL_TRUE;
    return !t->border_fallback;
 }
 
-
-
 static GLboolean radeonUpdateTextureUnit( GLcontext *ctx, int unit )
 {
-   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
 
-   if ( texUnit->_ReallyEnabled & (TEXTURE_RECT_BIT) ) {
-      return (enable_tex_rect( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( texUnit->_ReallyEnabled & (TEXTURE_1D_BIT | TEXTURE_2D_BIT) ) {
-      return (enable_tex_2d( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
-   }
-   else if ( texUnit->_ReallyEnabled & (TEXTURE_CUBE_BIT) ) {
-      return (enable_tex_cube( ctx, unit ) &&
-	      update_tex_common( ctx, unit ));
+   if (ctx->Texture.Unit[unit]._ReallyEnabled & TEXTURE_3D_BIT) {
+     rmesa->state.texture.unit[unit].texobj = NULL;
+     return GL_FALSE;
    }
-   else if ( texUnit->_ReallyEnabled ) {
-      return GL_FALSE;
+
+   if (!ctx->Texture.Unit[unit]._ReallyEnabled) {
+     /* disable the unit */
+     disable_tex_obj_state(rmesa, unit);
+     rmesa->state.texture.unit[unit].texobj = NULL;
+     return GL_TRUE;
    }
-   else {
-      disable_tex( ctx, unit );
-      return GL_TRUE;
+
+   if (!radeon_validate_texture(ctx, ctx->Texture.Unit[unit]._Current, unit)) {
+    _mesa_warning(ctx,
+		  "failed to validate texture for unit %d.\n",
+		  unit);
+     rmesa->state.texture.unit[unit].texobj = NULL;
+     return GL_FALSE;
    }
+   rmesa->state.texture.unit[unit].texobj = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
+   return GL_TRUE;
 }
 
 void radeonUpdateTextureState( GLcontext *ctx )
 {
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   r100ContextPtr rmesa = R100_CONTEXT(ctx);
    GLboolean ok;
 
+   /* set the ctx all textures off */
+   RADEON_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] &= ~((RADEON_TEX_ENABLE_MASK) | (RADEON_TEX_BLEND_ENABLE_MASK));
+
    ok = (radeonUpdateTextureUnit( ctx, 0 ) &&
 	 radeonUpdateTextureUnit( ctx, 1 ) &&
 	 radeonUpdateTextureUnit( ctx, 2 ));
 
    FALLBACK( rmesa, RADEON_FALLBACK_TEXTURE, !ok );
 
-   if (rmesa->TclFallback)
+   if (rmesa->radeon.TclFallback)
       radeonChooseVertexState( ctx );
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c
new file mode 100644
index 0000000000..ad501c454c
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.c
@@ -0,0 +1,1033 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ * Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+ *
+ * The Weather Channel (TM) funded Tungsten Graphics to develop the
+ * initial release of the Radeon 8500 driver under the XFree86 license.
+ * This notice must be preserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/convolve.h"
+#include "main/mipmap.h"
+#include "main/texcompress.h"
+#include "main/texformat.h"
+#include "main/texstore.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+#include "main/texgetimage.h"
+
+#include "xmlpool.h"		/* for symbolic values of enum-type options */
+
+#include "radeon_common.h"
+
+#include "radeon_mipmap_tree.h"
+
+
+static void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride,
+	GLuint numrows, GLuint rowsize)
+{
+	assert(rowsize <= dststride);
+	assert(rowsize <= srcstride);
+
+	if (rowsize == srcstride && rowsize == dststride) {
+		memcpy(dst, src, numrows*rowsize);
+	} else {
+		GLuint i;
+		for(i = 0; i < numrows; ++i) {
+			memcpy(dst, src, rowsize);
+			dst += dststride;
+			src += srcstride;
+		}
+	}
+}
+
+/* textures */
+/**
+ * Allocate an empty texture image object.
+ */
+struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx)
+{
+	return CALLOC(sizeof(radeon_texture_image));
+}
+
+/**
+ * Free memory associated with this texture image.
+ */
+void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage)
+{
+	radeon_texture_image* image = get_radeon_texture_image(timage);
+
+	if (image->mt) {
+		radeon_miptree_unreference(image->mt);
+		image->mt = 0;
+		assert(!image->base.Data);
+	} else {
+		_mesa_free_texture_image_data(ctx, timage);
+	}
+	if (image->bo) {
+		radeon_bo_unref(image->bo);
+		image->bo = NULL;
+	}
+	if (timage->Data) {
+		_mesa_free_texmemory(timage->Data);
+		timage->Data = NULL;
+	}
+}
+
+/* Set Data pointer and additional data for mapped texture image */
+static void teximage_set_map_data(radeon_texture_image *image)
+{
+	radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel];
+
+	image->base.Data = image->mt->bo->ptr + lvl->faces[image->mtface].offset;
+	image->base.RowStride = lvl->rowstride / image->mt->bpp;
+}
+
+
+/**
+ * Map a single texture image for glTexImage and friends.
+ */
+void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable)
+{
+	if (image->mt) {
+		assert(!image->base.Data);
+
+		radeon_bo_map(image->mt->bo, write_enable);
+		teximage_set_map_data(image);
+	}
+}
+
+
+void radeon_teximage_unmap(radeon_texture_image *image)
+{
+	if (image->mt) {
+		assert(image->base.Data);
+
+		image->base.Data = 0;
+		radeon_bo_unmap(image->mt->bo);
+	}
+}
+
+static void map_override(GLcontext *ctx, radeonTexObj *t)
+{
+	radeon_texture_image *img = get_radeon_texture_image(t->base.Image[0][0]);
+
+	radeon_bo_map(t->bo, GL_FALSE);
+
+	img->base.Data = t->bo->ptr;
+	_mesa_set_fetch_functions(&img->base, 2);
+}
+
+static void unmap_override(GLcontext *ctx, radeonTexObj *t)
+{
+	radeon_texture_image *img = get_radeon_texture_image(t->base.Image[0][0]);
+
+	radeon_bo_unmap(t->bo);
+
+	img->base.Data = NULL;
+}
+
+/**
+ * Map a validated texture for reading during software rendering.
+ */
+void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	int face, level;
+
+	if (!radeon_validate_texture_miptree(ctx, texObj))
+	  return;
+
+	/* for r100 3D sw fallbacks don't have mt */
+	if (t->image_override && t->bo)
+		map_override(ctx, t);
+
+	if (!t->mt)
+		return;
+
+	radeon_bo_map(t->mt->bo, GL_FALSE);
+	for(face = 0; face < t->mt->faces; ++face) {
+		for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level)
+			teximage_set_map_data(get_radeon_texture_image(texObj->Image[face][level]));
+	}
+}
+
+void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	int face, level;
+
+	if (t->image_override && t->bo)
+		unmap_override(ctx, t);
+	/* for r100 3D sw fallbacks don't have mt */
+	if (!t->mt)
+	  return;
+
+	for(face = 0; face < t->mt->faces; ++face) {
+		for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level)
+			texObj->Image[face][level]->Data = 0;
+	}
+	radeon_bo_unmap(t->mt->bo);
+}
+
+GLuint radeon_face_for_target(GLenum target)
+{
+	switch (target) {
+	case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
+	case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
+	case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
+	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
+	case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
+	case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
+		return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
+	default:
+		return 0;
+	}
+}
+
+/**
+ * Wraps Mesa's implementation to ensure that the base level image is mapped.
+ *
+ * This relies on internal details of _mesa_generate_mipmap, in particular
+ * the fact that the memory for recreated texture images is always freed.
+ */
+static void radeon_generate_mipmap(GLcontext *ctx, GLenum target,
+				   struct gl_texture_object *texObj)
+{
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	GLuint nr_faces = (t->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+	int i, face;
+
+
+	_mesa_generate_mipmap(ctx, target, texObj);
+
+	for (face = 0; face < nr_faces; face++) {
+		for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
+			radeon_texture_image *image;
+
+			image = get_radeon_texture_image(texObj->Image[face][i]);
+
+			if (image == NULL)
+				break;
+
+			image->mtlevel = i;
+			image->mtface = face;
+
+			radeon_miptree_unreference(image->mt);
+			image->mt = NULL;
+		}
+	}
+	
+}
+
+void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj)
+{
+	GLuint face = radeon_face_for_target(target);
+	radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[face][texObj->BaseLevel]);
+
+	radeon_teximage_map(baseimage, GL_FALSE);
+	radeon_generate_mipmap(ctx, target, texObj);
+	radeon_teximage_unmap(baseimage);
+}
+
+
+/* try to find a format which will only need a memcopy */
+static const struct gl_texture_format *radeonChoose8888TexFormat(radeonContextPtr rmesa,
+								 GLenum srcFormat,
+								 GLenum srcType, GLboolean fbo)
+{
+	const GLuint ui = 1;
+	const GLubyte littleEndian = *((const GLubyte *)&ui);
+
+	/* r100 can only do this */
+	if (IS_R100_CLASS(rmesa->radeonScreen) || fbo)
+	  return _dri_texformat_argb8888;
+
+	if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
+	    (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
+	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
+	    (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) {
+		return &_mesa_texformat_rgba8888;
+	} else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) ||
+		   (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) ||
+		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) ||
+		   (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) {
+		return &_mesa_texformat_rgba8888_rev;
+	} else if (IS_R200_CLASS(rmesa->radeonScreen)) {
+		return _dri_texformat_argb8888;
+	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) ||
+					    srcType == GL_UNSIGNED_INT_8_8_8_8)) {
+		return &_mesa_texformat_argb8888_rev;
+	} else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) ||
+					    srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) {
+		return &_mesa_texformat_argb8888;
+	} else
+		return _dri_texformat_argb8888;
+}
+
+const struct gl_texture_format *radeonChooseTextureFormat_mesa(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type)
+{
+	return radeonChooseTextureFormat(ctx, internalFormat, format,
+					 type, 0);
+}
+
+const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type, GLboolean fbo)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	const GLboolean do32bpt =
+	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32);
+	const GLboolean force16bpt =
+	    (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16);
+	(void)format;
+
+#if 0
+	fprintf(stderr, "InternalFormat=%s(%d) type=%s format=%s\n",
+		_mesa_lookup_enum_by_nr(internalFormat), internalFormat,
+		_mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format));
+	fprintf(stderr, "do32bpt=%d force16bpt=%d\n", do32bpt, force16bpt);
+#endif
+
+	switch (internalFormat) {
+	case 4:
+	case GL_RGBA:
+	case GL_COMPRESSED_RGBA:
+		switch (type) {
+		case GL_UNSIGNED_INT_10_10_10_2:
+		case GL_UNSIGNED_INT_2_10_10_10_REV:
+			return do32bpt ? _dri_texformat_argb8888 :
+			    _dri_texformat_argb1555;
+		case GL_UNSIGNED_SHORT_4_4_4_4:
+		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
+			return _dri_texformat_argb4444;
+		case GL_UNSIGNED_SHORT_5_5_5_1:
+		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+			return _dri_texformat_argb1555;
+		default:
+			return do32bpt ? radeonChoose8888TexFormat(rmesa, format, type, fbo) :
+			    _dri_texformat_argb4444;
+		}
+
+	case 3:
+	case GL_RGB:
+	case GL_COMPRESSED_RGB:
+		switch (type) {
+		case GL_UNSIGNED_SHORT_4_4_4_4:
+		case GL_UNSIGNED_SHORT_4_4_4_4_REV:
+			return _dri_texformat_argb4444;
+		case GL_UNSIGNED_SHORT_5_5_5_1:
+		case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+			return _dri_texformat_argb1555;
+		case GL_UNSIGNED_SHORT_5_6_5:
+		case GL_UNSIGNED_SHORT_5_6_5_REV:
+			return _dri_texformat_rgb565;
+		default:
+			return do32bpt ? _dri_texformat_argb8888 :
+			    _dri_texformat_rgb565;
+		}
+
+	case GL_RGBA8:
+	case GL_RGB10_A2:
+	case GL_RGBA12:
+	case GL_RGBA16:
+		return !force16bpt ?
+			radeonChoose8888TexFormat(rmesa, format, type, fbo) :
+			_dri_texformat_argb4444;
+
+	case GL_RGBA4:
+	case GL_RGBA2:
+		return _dri_texformat_argb4444;
+
+	case GL_RGB5_A1:
+		return _dri_texformat_argb1555;
+
+	case GL_RGB8:
+	case GL_RGB10:
+	case GL_RGB12:
+	case GL_RGB16:
+		return !force16bpt ? _dri_texformat_argb8888 :
+		    _dri_texformat_rgb565;
+
+	case GL_RGB5:
+	case GL_RGB4:
+	case GL_R3_G3_B2:
+		return _dri_texformat_rgb565;
+
+	case GL_ALPHA:
+	case GL_ALPHA4:
+	case GL_ALPHA8:
+	case GL_ALPHA12:
+	case GL_ALPHA16:
+	case GL_COMPRESSED_ALPHA:
+		/* r200: can't use a8 format since interpreting hw I8 as a8 would result
+		   in wrong rgb values (same as alpha value instead of 0). */
+		if (IS_R200_CLASS(rmesa->radeonScreen))
+			return _dri_texformat_al88;
+		else
+			return _dri_texformat_a8;
+	case 1:
+	case GL_LUMINANCE:
+	case GL_LUMINANCE4:
+	case GL_LUMINANCE8:
+	case GL_LUMINANCE12:
+	case GL_LUMINANCE16:
+	case GL_COMPRESSED_LUMINANCE:
+		return _dri_texformat_l8;
+
+	case 2:
+	case GL_LUMINANCE_ALPHA:
+	case GL_LUMINANCE4_ALPHA4:
+	case GL_LUMINANCE6_ALPHA2:
+	case GL_LUMINANCE8_ALPHA8:
+	case GL_LUMINANCE12_ALPHA4:
+	case GL_LUMINANCE12_ALPHA12:
+	case GL_LUMINANCE16_ALPHA16:
+	case GL_COMPRESSED_LUMINANCE_ALPHA:
+		return _dri_texformat_al88;
+
+	case GL_INTENSITY:
+	case GL_INTENSITY4:
+	case GL_INTENSITY8:
+	case GL_INTENSITY12:
+	case GL_INTENSITY16:
+	case GL_COMPRESSED_INTENSITY:
+		return _dri_texformat_i8;
+
+	case GL_YCBCR_MESA:
+		if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
+		    type == GL_UNSIGNED_BYTE)
+			return &_mesa_texformat_ycbcr;
+		else
+			return &_mesa_texformat_ycbcr_rev;
+
+	case GL_RGB_S3TC:
+	case GL_RGB4_S3TC:
+	case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+		return &_mesa_texformat_rgb_dxt1;
+
+	case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+		return &_mesa_texformat_rgba_dxt1;
+
+	case GL_RGBA_S3TC:
+	case GL_RGBA4_S3TC:
+	case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+		return &_mesa_texformat_rgba_dxt3;
+
+	case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+		return &_mesa_texformat_rgba_dxt5;
+
+	case GL_ALPHA16F_ARB:
+		return &_mesa_texformat_alpha_float16;
+	case GL_ALPHA32F_ARB:
+		return &_mesa_texformat_alpha_float32;
+	case GL_LUMINANCE16F_ARB:
+		return &_mesa_texformat_luminance_float16;
+	case GL_LUMINANCE32F_ARB:
+		return &_mesa_texformat_luminance_float32;
+	case GL_LUMINANCE_ALPHA16F_ARB:
+		return &_mesa_texformat_luminance_alpha_float16;
+	case GL_LUMINANCE_ALPHA32F_ARB:
+		return &_mesa_texformat_luminance_alpha_float32;
+	case GL_INTENSITY16F_ARB:
+		return &_mesa_texformat_intensity_float16;
+	case GL_INTENSITY32F_ARB:
+		return &_mesa_texformat_intensity_float32;
+	case GL_RGB16F_ARB:
+		return &_mesa_texformat_rgba_float16;
+	case GL_RGB32F_ARB:
+		return &_mesa_texformat_rgba_float32;
+	case GL_RGBA16F_ARB:
+		return &_mesa_texformat_rgba_float16;
+	case GL_RGBA32F_ARB:
+		return &_mesa_texformat_rgba_float32;
+
+	case GL_DEPTH_COMPONENT:
+	case GL_DEPTH_COMPONENT16:
+	case GL_DEPTH_COMPONENT24:
+	case GL_DEPTH_COMPONENT32:
+	case GL_DEPTH_STENCIL_EXT:
+	case GL_DEPTH24_STENCIL8_EXT:
+		return &_mesa_texformat_s8_z24;
+
+	/* EXT_texture_sRGB */
+	case GL_SRGB:
+	case GL_SRGB8:
+	case GL_SRGB_ALPHA:
+	case GL_SRGB8_ALPHA8:
+	case GL_COMPRESSED_SRGB:
+	case GL_COMPRESSED_SRGB_ALPHA:
+		return &_mesa_texformat_srgba8;
+
+	case GL_SLUMINANCE:
+	case GL_SLUMINANCE8:
+	case GL_COMPRESSED_SLUMINANCE:
+		return &_mesa_texformat_sl8;
+
+	case GL_SLUMINANCE_ALPHA:
+	case GL_SLUMINANCE8_ALPHA8:
+	case GL_COMPRESSED_SLUMINANCE_ALPHA:
+		return &_mesa_texformat_sla8;
+
+	default:
+		_mesa_problem(ctx,
+			      "unexpected internalFormat 0x%x in %s",
+			      (int)internalFormat, __func__);
+		return NULL;
+	}
+
+	return NULL;		/* never get here */
+}
+
+/**
+ * All glTexImage calls go through this function.
+ */
+static void radeon_teximage(
+	GLcontext *ctx, int dims,
+	GLenum target, GLint level,
+	GLint internalFormat,
+	GLint width, GLint height, GLint depth,
+	GLsizei imageSize,
+	GLenum format, GLenum type, const GLvoid * pixels,
+	const struct gl_pixelstore_attrib *packing,
+	struct gl_texture_object *texObj,
+	struct gl_texture_image *texImage,
+	int compressed)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	radeon_texture_image* image = get_radeon_texture_image(texImage);
+	GLuint dstRowStride;
+	GLint postConvWidth = width;
+	GLint postConvHeight = height;
+	GLuint texelBytes;
+	GLuint face = radeon_face_for_target(target);
+
+	radeon_firevertices(rmesa);
+
+	t->validated = GL_FALSE;
+
+	if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) {
+	       _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth,
+						  &postConvHeight);
+	}
+
+	/* Choose and fill in the texture format for this image */
+	texImage->TexFormat = radeonChooseTextureFormat(ctx, internalFormat, format, type, 0);
+	_mesa_set_fetch_functions(texImage, dims);
+
+	if (texImage->TexFormat->TexelBytes == 0) {
+		texelBytes = 0;
+		texImage->IsCompressed = GL_TRUE;
+		texImage->CompressedSize =
+			ctx->Driver.CompressedTextureSize(ctx, texImage->Width,
+					   texImage->Height, texImage->Depth,
+					   texImage->TexFormat->MesaFormat);
+	} else {
+		texImage->IsCompressed = GL_FALSE;
+		texImage->CompressedSize = 0;
+
+		texelBytes = texImage->TexFormat->TexelBytes;
+		/* Minimum pitch of 32 bytes */
+		if (postConvWidth * texelBytes < 32) {
+		  postConvWidth = 32 / texelBytes;
+		  texImage->RowStride = postConvWidth;
+		}
+		if (!image->mt) {      
+			assert(texImage->RowStride == postConvWidth);
+		}
+	}
+
+	/* Allocate memory for image */
+	radeonFreeTexImageData(ctx, texImage); /* Mesa core only clears texImage->Data but not image->mt */
+
+	if (t->mt &&
+	    t->mt->firstLevel == level &&
+	    t->mt->lastLevel == level &&
+	    t->mt->target != GL_TEXTURE_CUBE_MAP_ARB &&
+	    !radeon_miptree_matches_image(t->mt, texImage, face, level)) {
+	  radeon_miptree_unreference(t->mt);
+	  t->mt = NULL;
+	}
+
+	if (!t->mt)
+		radeon_try_alloc_miptree(rmesa, t, texImage, face, level);
+	if (t->mt && radeon_miptree_matches_image(t->mt, texImage, face, level)) {
+		radeon_mipmap_level *lvl;
+		image->mt = t->mt;
+		image->mtlevel = level - t->mt->firstLevel;
+		image->mtface = face;
+		radeon_miptree_reference(t->mt);
+		lvl = &image->mt->levels[image->mtlevel];
+		dstRowStride = lvl->rowstride;
+	} else {
+		int size;
+		if (texImage->IsCompressed) {
+			size = texImage->CompressedSize;
+		} else {
+			size = texImage->Width * texImage->Height * texImage->Depth * texImage->TexFormat->TexelBytes;
+		}
+		texImage->Data = _mesa_alloc_texmemory(size);
+	}
+
+	/* Upload texture image; note that the spec allows pixels to be NULL */
+	if (compressed) {
+		pixels = _mesa_validate_pbo_compressed_teximage(
+			ctx, imageSize, pixels, packing, "glCompressedTexImage");
+	} else {
+		pixels = _mesa_validate_pbo_teximage(
+			ctx, dims, width, height, depth,
+			format, type, pixels, packing, "glTexImage");
+	}
+
+	if (pixels) {
+		radeon_teximage_map(image, GL_TRUE);
+
+		if (compressed) {
+			memcpy(texImage->Data, pixels, imageSize);
+		} else {
+			GLuint dstRowStride;
+			GLuint *dstImageOffsets;
+
+			if (image->mt) {
+				radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel];
+				dstRowStride = lvl->rowstride;
+			} else {
+				dstRowStride = texImage->Width * texImage->TexFormat->TexelBytes;
+			}
+
+			if (dims == 3) {
+				int i;
+
+				dstImageOffsets = _mesa_malloc(depth * sizeof(GLuint)) ;
+				if (!dstImageOffsets)
+					_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+
+				for (i = 0; i < depth; ++i) {
+					dstImageOffsets[i] = dstRowStride/texImage->TexFormat->TexelBytes * height * i;
+				}
+			} else {
+				dstImageOffsets = texImage->ImageOffsets;
+			}
+
+			if (!texImage->TexFormat->StoreImage(ctx, dims,
+						texImage->_BaseFormat,
+						texImage->TexFormat,
+						texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */
+						dstRowStride,
+						dstImageOffsets,
+						width, height, depth,
+						format, type, pixels, packing))
+				_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+
+			if (dims == 3)
+				_mesa_free(dstImageOffsets);
+		}
+
+		/* SGIS_generate_mipmap */
+		if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+			radeon_generate_mipmap(ctx, target, texObj);
+		}
+	}
+
+	_mesa_unmap_teximage_pbo(ctx, packing);
+
+	if (pixels)
+	  radeon_teximage_unmap(image);
+
+
+}
+
+void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage)
+{
+	radeon_teximage(ctx, 1, target, level, internalFormat, width, 1, 1,
+		0, format, type, pixels, packing, texObj, texImage, 0);
+}
+
+void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level,
+			   GLint internalFormat,
+			   GLint width, GLint height, GLint border,
+			   GLenum format, GLenum type, const GLvoid * pixels,
+			   const struct gl_pixelstore_attrib *packing,
+			   struct gl_texture_object *texObj,
+			   struct gl_texture_image *texImage)
+
+{
+	radeon_teximage(ctx, 2, target, level, internalFormat, width, height, 1,
+		0, format, type, pixels, packing, texObj, texImage, 0);
+}
+
+void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target,
+				     GLint level, GLint internalFormat,
+				     GLint width, GLint height, GLint border,
+				     GLsizei imageSize, const GLvoid * data,
+				     struct gl_texture_object *texObj,
+				     struct gl_texture_image *texImage)
+{
+	radeon_teximage(ctx, 2, target, level, internalFormat, width, height, 1,
+		imageSize, 0, 0, data, &ctx->Unpack, texObj, texImage, 1);
+}
+
+void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint height, GLint depth,
+		      GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage)
+{
+	radeon_teximage(ctx, 3, target, level, internalFormat, width, height, depth,
+		0, format, type, pixels, packing, texObj, texImage, 0);
+}
+
+/**
+ * Update a subregion of the given texture image.
+ */
+static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int level,
+		GLint xoffset, GLint yoffset, GLint zoffset,
+		GLsizei width, GLsizei height, GLsizei depth,
+		GLsizei imageSize,
+		GLenum format, GLenum type,
+		const GLvoid * pixels,
+		const struct gl_pixelstore_attrib *packing,
+		struct gl_texture_object *texObj,
+		struct gl_texture_image *texImage,
+		int compressed)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	radeonTexObj* t = radeon_tex_obj(texObj);
+	radeon_texture_image* image = get_radeon_texture_image(texImage);
+
+	radeon_firevertices(rmesa);
+
+	t->validated = GL_FALSE;
+	if (compressed) {
+		pixels = _mesa_validate_pbo_compressed_teximage(
+			ctx, imageSize, pixels, packing, "glCompressedTexImage");
+	} else {
+		pixels = _mesa_validate_pbo_teximage(ctx, dims,
+			width, height, depth, format, type, pixels, packing, "glTexSubImage1D");
+	}
+
+	if (pixels) {
+		GLint dstRowStride;
+		radeon_teximage_map(image, GL_TRUE);
+
+		if (image->mt) {
+			radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel];
+			dstRowStride = lvl->rowstride;
+		} else {
+			dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes;
+		}
+
+		if (compressed) {
+			uint32_t srcRowStride, bytesPerRow, rows; 
+			dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, texImage->Width);
+			srcRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width);
+			bytesPerRow = srcRowStride;
+			rows = height / 4;
+
+			copy_rows(texImage->Data, dstRowStride,  image->base.Data, srcRowStride, rows,
+				  bytesPerRow);
+			
+		} else {
+			if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
+							     texImage->TexFormat, texImage->Data,
+							     xoffset, yoffset, zoffset,
+							     dstRowStride,
+							     texImage->ImageOffsets,
+							     width, height, depth,
+							     format, type, pixels, packing))
+				_mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+		}
+
+		/* GL_SGIS_generate_mipmap */
+		if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+			radeon_generate_mipmap(ctx, target, texObj);
+		}
+	}
+
+	radeon_teximage_unmap(image);
+
+	_mesa_unmap_teximage_pbo(ctx, packing);
+
+
+}
+
+void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset,
+			 GLsizei width,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 1, target, level, xoffset, 0, 0, width, 1, 1, 0,
+		format, type, pixels, packing, texObj, texImage, 0);
+}
+
+void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset, GLint yoffset,
+			 GLsizei width, GLsizei height,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 2, target, level, xoffset, yoffset, 0, width, height, 1,
+			   0, format, type, pixels, packing, texObj, texImage,
+			   0);
+}
+
+void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target,
+				   GLint level, GLint xoffset,
+				   GLint yoffset, GLsizei width,
+				   GLsizei height, GLenum format,
+				   GLsizei imageSize, const GLvoid * data,
+				   struct gl_texture_object *texObj,
+				   struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 2, target, level, xoffset, yoffset, 0, width, height, 1,
+		imageSize, format, 0, data, &ctx->Unpack, texObj, texImage, 1);
+}
+
+
+void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset, GLint yoffset, GLint zoffset,
+			 GLsizei width, GLsizei height, GLsizei depth,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage)
+{
+	radeon_texsubimage(ctx, 3, target, level, xoffset, yoffset, zoffset, width, height, depth, 0,
+		format, type, pixels, packing, texObj, texImage, 0);
+}
+
+
+
+/**
+ * Ensure that the given image is stored in the given miptree from now on.
+ */
+static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_texture_image *image, int face, int level)
+{
+	radeon_mipmap_level *dstlvl = &mt->levels[level - mt->firstLevel];
+	unsigned char *dest;
+
+	assert(image->mt != mt);
+	assert(dstlvl->width == image->base.Width);
+	assert(dstlvl->height == image->base.Height);
+	assert(dstlvl->depth == image->base.Depth);
+
+
+	radeon_bo_map(mt->bo, GL_TRUE);
+	dest = mt->bo->ptr + dstlvl->faces[face].offset;
+
+	if (image->mt) {
+		/* Format etc. should match, so we really just need a memcpy().
+		 * In fact, that memcpy() could be done by the hardware in many
+		 * cases, provided that we have a proper memory manager.
+		 */
+		radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel-image->mt->firstLevel];
+
+		assert(srclvl->size == dstlvl->size);
+		assert(srclvl->rowstride == dstlvl->rowstride);
+
+		radeon_bo_map(image->mt->bo, GL_FALSE);
+
+		memcpy(dest,
+			image->mt->bo->ptr + srclvl->faces[face].offset,
+			dstlvl->size);
+		radeon_bo_unmap(image->mt->bo);
+
+		radeon_miptree_unreference(image->mt);
+	} else {
+		uint32_t srcrowstride;
+		uint32_t height;
+		/* need to confirm this value is correct */
+		if (mt->compressed) {
+			height = image->base.Height / 4;
+			srcrowstride = image->base.RowStride * mt->bpp;
+		} else {
+			height = image->base.Height * image->base.Depth;
+			srcrowstride = image->base.Width * image->base.TexFormat->TexelBytes;
+		}
+
+//		if (mt->tilebits)
+//			WARN_ONCE("%s: tiling not supported yet", __FUNCTION__);
+
+		copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride,
+			  height, srcrowstride);
+
+		_mesa_free_texmemory(image->base.Data);
+		image->base.Data = 0;
+	}
+
+	radeon_bo_unmap(mt->bo);
+
+	image->mt = mt;
+	image->mtface = face;
+	image->mtlevel = level;
+	radeon_miptree_reference(image->mt);
+}
+
+int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj)
+{
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	radeonTexObj *t = radeon_tex_obj(texObj);
+	radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[0][texObj->BaseLevel]);
+	int face, level;
+
+	if (t->validated || t->image_override)
+		return GL_TRUE;
+
+	if (RADEON_DEBUG & DEBUG_TEXTURE)
+		fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj);
+
+	if (baseimage->base.Border > 0)
+		return GL_FALSE;
+
+	/* Ensure a matching miptree exists.
+	 *
+	 * Differing mipmap trees can result when the app uses TexImage to
+	 * change texture dimensions.
+	 *
+	 * Prefer to use base image's miptree if it
+	 * exists, since that most likely contains more valid data (remember
+	 * that the base level is usually significantly larger than the rest
+	 * of the miptree, so cubemaps are the only possible exception).
+	 */
+	if (baseimage->mt &&
+	    baseimage->mt != t->mt &&
+	    radeon_miptree_matches_texture(baseimage->mt, &t->base)) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = baseimage->mt;
+		radeon_miptree_reference(t->mt);
+	} else if (t->mt && !radeon_miptree_matches_texture(t->mt, &t->base)) {
+		radeon_miptree_unreference(t->mt);
+		t->mt = 0;
+	}
+
+	if (!t->mt) {
+		if (RADEON_DEBUG & DEBUG_TEXTURE)
+			fprintf(stderr, " Allocate new miptree\n");
+		radeon_try_alloc_miptree(rmesa, t, &baseimage->base, 0, texObj->BaseLevel);
+		if (!t->mt) {
+			_mesa_problem(ctx, "r300_validate_texture failed to alloc miptree");
+			return GL_FALSE;
+		}
+	}
+
+	/* Ensure all images are stored in the single main miptree */
+	for(face = 0; face < t->mt->faces; ++face) {
+		for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) {
+			radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]);
+			if (RADEON_DEBUG & DEBUG_TEXTURE)
+				fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt);
+			if (t->mt == image->mt) {
+				if (RADEON_DEBUG & DEBUG_TEXTURE)
+					fprintf(stderr, "OK\n");
+
+				continue;
+			}
+
+			if (RADEON_DEBUG & DEBUG_TEXTURE)
+				fprintf(stderr, "migrating\n");
+			migrate_image_to_miptree(t->mt, image, face, level);
+		}
+	}
+
+	return GL_TRUE;
+}
+
+
+/**
+ * Need to map texture image into memory before copying image data,
+ * then unmap it.
+ */
+static void
+radeon_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
+		     GLenum format, GLenum type, GLvoid * pixels,
+		     struct gl_texture_object *texObj,
+		     struct gl_texture_image *texImage, int compressed)
+{
+	radeon_texture_image *image = get_radeon_texture_image(texImage);
+
+	if (image->mt) {
+		/* Map the texture image read-only */
+		radeon_teximage_map(image, GL_FALSE);
+	} else {
+		/* Image hasn't been uploaded to a miptree yet */
+		assert(image->base.Data);
+	}
+
+	if (compressed) {
+		_mesa_get_compressed_teximage(ctx, target, level, pixels,
+					      texObj, texImage);
+	} else {
+		_mesa_get_teximage(ctx, target, level, format, type, pixels,
+				   texObj, texImage);
+	}
+     
+	if (image->mt) {
+		radeon_teximage_unmap(image);
+	}
+}
+
+void
+radeonGetTexImage(GLcontext * ctx, GLenum target, GLint level,
+		  GLenum format, GLenum type, GLvoid * pixels,
+		  struct gl_texture_object *texObj,
+		  struct gl_texture_image *texImage)
+{
+	radeon_get_tex_image(ctx, target, level, format, type, pixels,
+			     texObj, texImage, 0);
+}
+
+void
+radeonGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
+			    GLvoid *pixels,
+			    struct gl_texture_object *texObj,
+			    struct gl_texture_image *texImage)
+{
+	radeon_get_tex_image(ctx, target, level, 0, 0, pixels,
+			     texObj, texImage, 1);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.h b/src/mesa/drivers/dri/radeon/radeon_texture.h
new file mode 100644
index 0000000000..888a55ba91
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.h
@@ -0,0 +1,122 @@
+/*
+ * Copyright (C) 2008 Nicolai Haehnle.
+ * Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+ *
+ * The Weather Channel (TM) funded Tungsten Graphics to develop the
+ * initial release of the Radeon 8500 driver under the XFree86 license.
+ * This notice must be preserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial
+ * portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef RADEON_TEXTURE_H
+#define RADEON_TEXTURE_H
+struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx);
+void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage);
+
+void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable);
+void radeon_teximage_unmap(radeon_texture_image *image);
+void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj);
+void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj);
+void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj);
+int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj);
+GLuint radeon_face_for_target(GLenum target);
+const struct gl_texture_format *radeonChooseTextureFormat_mesa(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type);
+const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx,
+							  GLint internalFormat,
+							  GLenum format,
+							  GLenum type, GLboolean fbo);
+
+void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage);
+void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint height, GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage);
+void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target,
+				GLint level, GLint internalFormat,
+				GLint width, GLint height, GLint border,
+				GLsizei imageSize, const GLvoid * data,
+				struct gl_texture_object *texObj,
+				struct gl_texture_image *texImage);
+void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level,
+		      GLint internalFormat,
+		      GLint width, GLint height, GLint depth,
+		      GLint border,
+		      GLenum format, GLenum type, const GLvoid * pixels,
+		      const struct gl_pixelstore_attrib *packing,
+		      struct gl_texture_object *texObj,
+		      struct gl_texture_image *texImage);
+void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset,
+			 GLsizei width,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage);
+void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
+				GLint xoffset, GLint yoffset,
+				GLsizei width, GLsizei height,
+				GLenum format, GLenum type,
+				const GLvoid * pixels,
+				const struct gl_pixelstore_attrib *packing,
+				struct gl_texture_object *texObj,
+				struct gl_texture_image *texImage);
+void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target,
+				   GLint level, GLint xoffset,
+				   GLint yoffset, GLsizei width,
+				   GLsizei height, GLenum format,
+				   GLsizei imageSize, const GLvoid * data,
+				   struct gl_texture_object *texObj,
+				   struct gl_texture_image *texImage);
+
+void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
+			 GLint xoffset, GLint yoffset, GLint zoffset,
+			 GLsizei width, GLsizei height, GLsizei depth,
+			 GLenum format, GLenum type,
+			 const GLvoid * pixels,
+			 const struct gl_pixelstore_attrib *packing,
+			 struct gl_texture_object *texObj,
+			 struct gl_texture_image *texImage);
+
+void radeonGetTexImage(GLcontext * ctx, GLenum target, GLint level,
+		       GLenum format, GLenum type, GLvoid * pixels,
+		       struct gl_texture_object *texObj,
+		       struct gl_texture_image *texImage);
+void radeonGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
+				 GLvoid *pixels,
+				 struct gl_texture_object *texObj,
+				 struct gl_texture_image *texImage);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/server/radeon_reg.h b/src/mesa/drivers/dri/radeon/server/radeon_reg.h
index ae2ccdf292..866807462a 100644
--- a/src/mesa/drivers/dri/radeon/server/radeon_reg.h
+++ b/src/mesa/drivers/dri/radeon/server/radeon_reg.h
@@ -2031,6 +2031,9 @@
 #define RADEON_CP_PACKET3_3D_DRAW_INDX              0xC0002A00
 #define RADEON_CP_PACKET3_LOAD_PALETTE              0xC0002C00
 #define RADEON_CP_PACKET3_3D_LOAD_VBPNTR            0xC0002F00
+#define R200_CP_CMD_3D_DRAW_VBUF_2      0xC0003400
+#define R200_CP_CMD_3D_DRAW_IMMD_2      0xC0003500
+#define R200_CP_CMD_3D_DRAW_INDX_2      0xC0003600
 #define RADEON_CP_PACKET3_CNTL_PAINT                0xC0009100
 #define RADEON_CP_PACKET3_CNTL_BITBLT               0xC0009200
 #define RADEON_CP_PACKET3_CNTL_SMALLTEXT            0xC0009300
diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c
index 305df548fa..9a01465bdf 100644
--- a/src/mesa/drivers/x11/xm_dd.c
+++ b/src/mesa/drivers/x11/xm_dd.c
@@ -912,8 +912,9 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state )
    /*
     * GL_DITHER, GL_READ/DRAW_BUFFER, buffer binding state, etc. effect
     * renderbuffer span/clear funcs.
+    * Check _NEW_COLOR to detect dither enable/disable.
     */
-   if (new_state & (_NEW_COLOR | _NEW_PIXEL | _NEW_BUFFERS)) {
+   if (new_state & (_NEW_COLOR | _NEW_BUFFERS)) {
       XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer);
       struct xmesa_renderbuffer *front_xrb, *back_xrb;
 
diff --git a/src/mesa/glapi/ARB_copy_buffer.xml b/src/mesa/glapi/ARB_copy_buffer.xml
new file mode 100644
index 0000000000..719816d817
--- /dev/null
+++ b/src/mesa/glapi/ARB_copy_buffer.xml
@@ -0,0 +1,24 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_ARB_copy_buffer" number="59">
+
+    <enum name="COPY_READ_BUFFER"   value="0x8F36"/>
+    <enum name="COPY_WRITE_BUFFER"  value="0x8F37"/>
+
+    <function name="CopyBufferSubData" offset="assign">
+        <param name="readTarget" type="GLenum"/>
+        <param name="writeTarget" type="GLenum"/>
+        <param name="readOffset" type="GLintptr"/>
+        <param name="writeOffset" type="GLintptr"/>
+        <param name="size" type="GLsizeiptr"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/ARB_map_buffer_range.xml b/src/mesa/glapi/ARB_map_buffer_range.xml
new file mode 100644
index 0000000000..afcb9b6ba6
--- /dev/null
+++ b/src/mesa/glapi/ARB_map_buffer_range.xml
@@ -0,0 +1,34 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_ARB_map_buffer_range" number="50">
+
+    <enum name="MAP_READ_BIT"                value="0x0001"/>
+    <enum name="MAP_WRITE_BIT"               value="0x0002"/>
+    <enum name="MAP_INVALIDATE_RANGE_BIT"    value="0x0004"/>
+    <enum name="MAP_INVALIDATE_BUFFER_BIT"   value="0x0008"/>
+    <enum name="MAP_FLUSH_EXPLICIT_BIT"      value="0x0010"/>
+    <enum name="MAP_UNSYNCHRONIZED_BIT"      value="0x0020"/>
+
+    <function name="MapBufferRange" offset="assign">
+        <param name="target" type="GLenum"/>
+        <param name="offset" type="GLintptr"/>
+        <param name="length" type="GLsizeiptr"/>
+        <param name="access" type="GLbitfield"/>
+        <return type="GLvoid *"/>
+    </function>
+
+    <function name="FlushMappedBufferRange" offset="assign">
+        <param name="target" type="GLenum"/>
+        <param name="offset" type="GLintptr"/>
+        <param name="length" type="GLsizeiptr"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/ARB_vertex_array_object.xml b/src/mesa/glapi/ARB_vertex_array_object.xml
new file mode 100644
index 0000000000..3b4ab64f55
--- /dev/null
+++ b/src/mesa/glapi/ARB_vertex_array_object.xml
@@ -0,0 +1,34 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_ARB_vertex_array_object" number="54">
+
+    <enum name="VERTEX_ARRAY_BINDING" value="0x85B5"/>
+
+    <function name="BindVertexArray" offset="assign">
+        <param name="array" type="GLuint"/>
+    </function>
+
+    <function name="DeleteVertexArrays" alias="DeleteVertexArraysAPPLE">
+        <param name="n" type="GLsizei"/>
+        <param name="arrays" type="const GLuint *"/>
+    </function>
+
+    <function name="GenVertexArrays" offset="assign">
+        <param name="n" type="GLsizei"/>
+        <param name="arrays" type="GLuint *"/>
+    </function>
+
+    <function name="IsVertexArray" alias="IsVertexArrayAPPLE">
+        <param name="array" type="GLuint"/>
+        <return type="GLboolean"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/EXT_provoking_vertex.xml b/src/mesa/glapi/EXT_provoking_vertex.xml
new file mode 100644
index 0000000000..f528a2c7d3
--- /dev/null
+++ b/src/mesa/glapi/EXT_provoking_vertex.xml
@@ -0,0 +1,22 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+
+<category name="GL_EXT_provoking_vertex" number="364">
+
+    <enum name="FIRST_VERTEX_CONVENTION_EXT"                  value="0x8E4D"/>
+    <enum name="LAST_VERTEX_CONVENTION_EXT"                   value="0x8E4E"/>
+    <enum name="PROVOKING_VERTEX_EXT"                         value="0x8E4F"/>
+    <enum name="QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT" value="0x8E4C"/>
+
+    <function name="ProvokingVertexEXT" offset="assign">
+        <param name="mode" type="GLenum"/>
+    </function>
+
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/Makefile b/src/mesa/glapi/Makefile
index f524167a47..2178eacef5 100644
--- a/src/mesa/glapi/Makefile
+++ b/src/mesa/glapi/Makefile
@@ -47,8 +47,12 @@ SERVER_OUTPUTS = \
 
 API_XML = gl_API.xml \
 	EXT_framebuffer_object.xml \
+	ARB_copy_buffer.xml \
 	ARB_framebuffer_object.xml \
-	APPLE_vertex_array_object.xml
+	ARB_map_buffer_range.xml \
+	ARB_vertex_array_object.xml \
+	APPLE_vertex_array_object.xml \
+	EXT_provoking_vertex.xml
 
 COMMON = gl_XML.py glX_XML.py license.py $(API_XML) typeexpr.py
 COMMON_GLX = $(COMMON) glX_API.xml glX_XML.py glX_proto_common.py
diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h
index 45b2fa077a..b3f1aed1a4 100644
--- a/src/mesa/glapi/dispatch.h
+++ b/src/mesa/glapi/dispatch.h
@@ -1746,6 +1746,21 @@
 #define CALL_RenderbufferStorageMultisample(disp, parameters) (*((disp)->RenderbufferStorageMultisample)) parameters
 #define GET_RenderbufferStorageMultisample(disp) ((disp)->RenderbufferStorageMultisample)
 #define SET_RenderbufferStorageMultisample(disp, fn) ((disp)->RenderbufferStorageMultisample = fn)
+#define CALL_FlushMappedBufferRange(disp, parameters) (*((disp)->FlushMappedBufferRange)) parameters
+#define GET_FlushMappedBufferRange(disp) ((disp)->FlushMappedBufferRange)
+#define SET_FlushMappedBufferRange(disp, fn) ((disp)->FlushMappedBufferRange = fn)
+#define CALL_MapBufferRange(disp, parameters) (*((disp)->MapBufferRange)) parameters
+#define GET_MapBufferRange(disp) ((disp)->MapBufferRange)
+#define SET_MapBufferRange(disp, fn) ((disp)->MapBufferRange = fn)
+#define CALL_BindVertexArray(disp, parameters) (*((disp)->BindVertexArray)) parameters
+#define GET_BindVertexArray(disp) ((disp)->BindVertexArray)
+#define SET_BindVertexArray(disp, fn) ((disp)->BindVertexArray = fn)
+#define CALL_GenVertexArrays(disp, parameters) (*((disp)->GenVertexArrays)) parameters
+#define GET_GenVertexArrays(disp) ((disp)->GenVertexArrays)
+#define SET_GenVertexArrays(disp, fn) ((disp)->GenVertexArrays = fn)
+#define CALL_CopyBufferSubData(disp, parameters) (*((disp)->CopyBufferSubData)) parameters
+#define GET_CopyBufferSubData(disp) ((disp)->CopyBufferSubData)
+#define SET_CopyBufferSubData(disp, fn) ((disp)->CopyBufferSubData = fn)
 #define CALL_PolygonOffsetEXT(disp, parameters) (*((disp)->PolygonOffsetEXT)) parameters
 #define GET_PolygonOffsetEXT(disp) ((disp)->PolygonOffsetEXT)
 #define SET_PolygonOffsetEXT(disp, fn) ((disp)->PolygonOffsetEXT = fn)
@@ -2370,6 +2385,9 @@
 #define CALL_FramebufferTextureLayerEXT(disp, parameters) (*((disp)->FramebufferTextureLayerEXT)) parameters
 #define GET_FramebufferTextureLayerEXT(disp) ((disp)->FramebufferTextureLayerEXT)
 #define SET_FramebufferTextureLayerEXT(disp, fn) ((disp)->FramebufferTextureLayerEXT = fn)
+#define CALL_ProvokingVertexEXT(disp, parameters) (*((disp)->ProvokingVertexEXT)) parameters
+#define GET_ProvokingVertexEXT(disp) ((disp)->ProvokingVertexEXT)
+#define SET_ProvokingVertexEXT(disp, fn) ((disp)->ProvokingVertexEXT = fn)
 #define CALL_StencilFuncSeparateATI(disp, parameters) (*((disp)->StencilFuncSeparateATI)) parameters
 #define GET_StencilFuncSeparateATI(disp) ((disp)->StencilFuncSeparateATI)
 #define SET_StencilFuncSeparateATI(disp, fn) ((disp)->StencilFuncSeparateATI = fn)
@@ -2388,7 +2406,7 @@
 
 #else
 
-#define driDispatchRemapTable_size 367
+#define driDispatchRemapTable_size 373
 extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 
 #define AttachShader_remap_index 0
@@ -2545,219 +2563,225 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define GetAttribLocationARB_remap_index 151
 #define DrawBuffersARB_remap_index 152
 #define RenderbufferStorageMultisample_remap_index 153
-#define PolygonOffsetEXT_remap_index 154
-#define GetPixelTexGenParameterfvSGIS_remap_index 155
-#define GetPixelTexGenParameterivSGIS_remap_index 156
-#define PixelTexGenParameterfSGIS_remap_index 157
-#define PixelTexGenParameterfvSGIS_remap_index 158
-#define PixelTexGenParameteriSGIS_remap_index 159
-#define PixelTexGenParameterivSGIS_remap_index 160
-#define SampleMaskSGIS_remap_index 161
-#define SamplePatternSGIS_remap_index 162
-#define ColorPointerEXT_remap_index 163
-#define EdgeFlagPointerEXT_remap_index 164
-#define IndexPointerEXT_remap_index 165
-#define NormalPointerEXT_remap_index 166
-#define TexCoordPointerEXT_remap_index 167
-#define VertexPointerEXT_remap_index 168
-#define PointParameterfEXT_remap_index 169
-#define PointParameterfvEXT_remap_index 170
-#define LockArraysEXT_remap_index 171
-#define UnlockArraysEXT_remap_index 172
-#define CullParameterdvEXT_remap_index 173
-#define CullParameterfvEXT_remap_index 174
-#define SecondaryColor3bEXT_remap_index 175
-#define SecondaryColor3bvEXT_remap_index 176
-#define SecondaryColor3dEXT_remap_index 177
-#define SecondaryColor3dvEXT_remap_index 178
-#define SecondaryColor3fEXT_remap_index 179
-#define SecondaryColor3fvEXT_remap_index 180
-#define SecondaryColor3iEXT_remap_index 181
-#define SecondaryColor3ivEXT_remap_index 182
-#define SecondaryColor3sEXT_remap_index 183
-#define SecondaryColor3svEXT_remap_index 184
-#define SecondaryColor3ubEXT_remap_index 185
-#define SecondaryColor3ubvEXT_remap_index 186
-#define SecondaryColor3uiEXT_remap_index 187
-#define SecondaryColor3uivEXT_remap_index 188
-#define SecondaryColor3usEXT_remap_index 189
-#define SecondaryColor3usvEXT_remap_index 190
-#define SecondaryColorPointerEXT_remap_index 191
-#define MultiDrawArraysEXT_remap_index 192
-#define MultiDrawElementsEXT_remap_index 193
-#define FogCoordPointerEXT_remap_index 194
-#define FogCoorddEXT_remap_index 195
-#define FogCoorddvEXT_remap_index 196
-#define FogCoordfEXT_remap_index 197
-#define FogCoordfvEXT_remap_index 198
-#define PixelTexGenSGIX_remap_index 199
-#define BlendFuncSeparateEXT_remap_index 200
-#define FlushVertexArrayRangeNV_remap_index 201
-#define VertexArrayRangeNV_remap_index 202
-#define CombinerInputNV_remap_index 203
-#define CombinerOutputNV_remap_index 204
-#define CombinerParameterfNV_remap_index 205
-#define CombinerParameterfvNV_remap_index 206
-#define CombinerParameteriNV_remap_index 207
-#define CombinerParameterivNV_remap_index 208
-#define FinalCombinerInputNV_remap_index 209
-#define GetCombinerInputParameterfvNV_remap_index 210
-#define GetCombinerInputParameterivNV_remap_index 211
-#define GetCombinerOutputParameterfvNV_remap_index 212
-#define GetCombinerOutputParameterivNV_remap_index 213
-#define GetFinalCombinerInputParameterfvNV_remap_index 214
-#define GetFinalCombinerInputParameterivNV_remap_index 215
-#define ResizeBuffersMESA_remap_index 216
-#define WindowPos2dMESA_remap_index 217
-#define WindowPos2dvMESA_remap_index 218
-#define WindowPos2fMESA_remap_index 219
-#define WindowPos2fvMESA_remap_index 220
-#define WindowPos2iMESA_remap_index 221
-#define WindowPos2ivMESA_remap_index 222
-#define WindowPos2sMESA_remap_index 223
-#define WindowPos2svMESA_remap_index 224
-#define WindowPos3dMESA_remap_index 225
-#define WindowPos3dvMESA_remap_index 226
-#define WindowPos3fMESA_remap_index 227
-#define WindowPos3fvMESA_remap_index 228
-#define WindowPos3iMESA_remap_index 229
-#define WindowPos3ivMESA_remap_index 230
-#define WindowPos3sMESA_remap_index 231
-#define WindowPos3svMESA_remap_index 232
-#define WindowPos4dMESA_remap_index 233
-#define WindowPos4dvMESA_remap_index 234
-#define WindowPos4fMESA_remap_index 235
-#define WindowPos4fvMESA_remap_index 236
-#define WindowPos4iMESA_remap_index 237
-#define WindowPos4ivMESA_remap_index 238
-#define WindowPos4sMESA_remap_index 239
-#define WindowPos4svMESA_remap_index 240
-#define MultiModeDrawArraysIBM_remap_index 241
-#define MultiModeDrawElementsIBM_remap_index 242
-#define DeleteFencesNV_remap_index 243
-#define FinishFenceNV_remap_index 244
-#define GenFencesNV_remap_index 245
-#define GetFenceivNV_remap_index 246
-#define IsFenceNV_remap_index 247
-#define SetFenceNV_remap_index 248
-#define TestFenceNV_remap_index 249
-#define AreProgramsResidentNV_remap_index 250
-#define BindProgramNV_remap_index 251
-#define DeleteProgramsNV_remap_index 252
-#define ExecuteProgramNV_remap_index 253
-#define GenProgramsNV_remap_index 254
-#define GetProgramParameterdvNV_remap_index 255
-#define GetProgramParameterfvNV_remap_index 256
-#define GetProgramStringNV_remap_index 257
-#define GetProgramivNV_remap_index 258
-#define GetTrackMatrixivNV_remap_index 259
-#define GetVertexAttribPointervNV_remap_index 260
-#define GetVertexAttribdvNV_remap_index 261
-#define GetVertexAttribfvNV_remap_index 262
-#define GetVertexAttribivNV_remap_index 263
-#define IsProgramNV_remap_index 264
-#define LoadProgramNV_remap_index 265
-#define ProgramParameters4dvNV_remap_index 266
-#define ProgramParameters4fvNV_remap_index 267
-#define RequestResidentProgramsNV_remap_index 268
-#define TrackMatrixNV_remap_index 269
-#define VertexAttrib1dNV_remap_index 270
-#define VertexAttrib1dvNV_remap_index 271
-#define VertexAttrib1fNV_remap_index 272
-#define VertexAttrib1fvNV_remap_index 273
-#define VertexAttrib1sNV_remap_index 274
-#define VertexAttrib1svNV_remap_index 275
-#define VertexAttrib2dNV_remap_index 276
-#define VertexAttrib2dvNV_remap_index 277
-#define VertexAttrib2fNV_remap_index 278
-#define VertexAttrib2fvNV_remap_index 279
-#define VertexAttrib2sNV_remap_index 280
-#define VertexAttrib2svNV_remap_index 281
-#define VertexAttrib3dNV_remap_index 282
-#define VertexAttrib3dvNV_remap_index 283
-#define VertexAttrib3fNV_remap_index 284
-#define VertexAttrib3fvNV_remap_index 285
-#define VertexAttrib3sNV_remap_index 286
-#define VertexAttrib3svNV_remap_index 287
-#define VertexAttrib4dNV_remap_index 288
-#define VertexAttrib4dvNV_remap_index 289
-#define VertexAttrib4fNV_remap_index 290
-#define VertexAttrib4fvNV_remap_index 291
-#define VertexAttrib4sNV_remap_index 292
-#define VertexAttrib4svNV_remap_index 293
-#define VertexAttrib4ubNV_remap_index 294
-#define VertexAttrib4ubvNV_remap_index 295
-#define VertexAttribPointerNV_remap_index 296
-#define VertexAttribs1dvNV_remap_index 297
-#define VertexAttribs1fvNV_remap_index 298
-#define VertexAttribs1svNV_remap_index 299
-#define VertexAttribs2dvNV_remap_index 300
-#define VertexAttribs2fvNV_remap_index 301
-#define VertexAttribs2svNV_remap_index 302
-#define VertexAttribs3dvNV_remap_index 303
-#define VertexAttribs3fvNV_remap_index 304
-#define VertexAttribs3svNV_remap_index 305
-#define VertexAttribs4dvNV_remap_index 306
-#define VertexAttribs4fvNV_remap_index 307
-#define VertexAttribs4svNV_remap_index 308
-#define VertexAttribs4ubvNV_remap_index 309
-#define GetTexBumpParameterfvATI_remap_index 310
-#define GetTexBumpParameterivATI_remap_index 311
-#define TexBumpParameterfvATI_remap_index 312
-#define TexBumpParameterivATI_remap_index 313
-#define AlphaFragmentOp1ATI_remap_index 314
-#define AlphaFragmentOp2ATI_remap_index 315
-#define AlphaFragmentOp3ATI_remap_index 316
-#define BeginFragmentShaderATI_remap_index 317
-#define BindFragmentShaderATI_remap_index 318
-#define ColorFragmentOp1ATI_remap_index 319
-#define ColorFragmentOp2ATI_remap_index 320
-#define ColorFragmentOp3ATI_remap_index 321
-#define DeleteFragmentShaderATI_remap_index 322
-#define EndFragmentShaderATI_remap_index 323
-#define GenFragmentShadersATI_remap_index 324
-#define PassTexCoordATI_remap_index 325
-#define SampleMapATI_remap_index 326
-#define SetFragmentShaderConstantATI_remap_index 327
-#define PointParameteriNV_remap_index 328
-#define PointParameterivNV_remap_index 329
-#define ActiveStencilFaceEXT_remap_index 330
-#define BindVertexArrayAPPLE_remap_index 331
-#define DeleteVertexArraysAPPLE_remap_index 332
-#define GenVertexArraysAPPLE_remap_index 333
-#define IsVertexArrayAPPLE_remap_index 334
-#define GetProgramNamedParameterdvNV_remap_index 335
-#define GetProgramNamedParameterfvNV_remap_index 336
-#define ProgramNamedParameter4dNV_remap_index 337
-#define ProgramNamedParameter4dvNV_remap_index 338
-#define ProgramNamedParameter4fNV_remap_index 339
-#define ProgramNamedParameter4fvNV_remap_index 340
-#define DepthBoundsEXT_remap_index 341
-#define BlendEquationSeparateEXT_remap_index 342
-#define BindFramebufferEXT_remap_index 343
-#define BindRenderbufferEXT_remap_index 344
-#define CheckFramebufferStatusEXT_remap_index 345
-#define DeleteFramebuffersEXT_remap_index 346
-#define DeleteRenderbuffersEXT_remap_index 347
-#define FramebufferRenderbufferEXT_remap_index 348
-#define FramebufferTexture1DEXT_remap_index 349
-#define FramebufferTexture2DEXT_remap_index 350
-#define FramebufferTexture3DEXT_remap_index 351
-#define GenFramebuffersEXT_remap_index 352
-#define GenRenderbuffersEXT_remap_index 353
-#define GenerateMipmapEXT_remap_index 354
-#define GetFramebufferAttachmentParameterivEXT_remap_index 355
-#define GetRenderbufferParameterivEXT_remap_index 356
-#define IsFramebufferEXT_remap_index 357
-#define IsRenderbufferEXT_remap_index 358
-#define RenderbufferStorageEXT_remap_index 359
-#define BlitFramebufferEXT_remap_index 360
-#define FramebufferTextureLayerEXT_remap_index 361
-#define StencilFuncSeparateATI_remap_index 362
-#define ProgramEnvParameters4fvEXT_remap_index 363
-#define ProgramLocalParameters4fvEXT_remap_index 364
-#define GetQueryObjecti64vEXT_remap_index 365
-#define GetQueryObjectui64vEXT_remap_index 366
+#define FlushMappedBufferRange_remap_index 154
+#define MapBufferRange_remap_index 155
+#define BindVertexArray_remap_index 156
+#define GenVertexArrays_remap_index 157
+#define CopyBufferSubData_remap_index 158
+#define PolygonOffsetEXT_remap_index 159
+#define GetPixelTexGenParameterfvSGIS_remap_index 160
+#define GetPixelTexGenParameterivSGIS_remap_index 161
+#define PixelTexGenParameterfSGIS_remap_index 162
+#define PixelTexGenParameterfvSGIS_remap_index 163
+#define PixelTexGenParameteriSGIS_remap_index 164
+#define PixelTexGenParameterivSGIS_remap_index 165
+#define SampleMaskSGIS_remap_index 166
+#define SamplePatternSGIS_remap_index 167
+#define ColorPointerEXT_remap_index 168
+#define EdgeFlagPointerEXT_remap_index 169
+#define IndexPointerEXT_remap_index 170
+#define NormalPointerEXT_remap_index 171
+#define TexCoordPointerEXT_remap_index 172
+#define VertexPointerEXT_remap_index 173
+#define PointParameterfEXT_remap_index 174
+#define PointParameterfvEXT_remap_index 175
+#define LockArraysEXT_remap_index 176
+#define UnlockArraysEXT_remap_index 177
+#define CullParameterdvEXT_remap_index 178
+#define CullParameterfvEXT_remap_index 179
+#define SecondaryColor3bEXT_remap_index 180
+#define SecondaryColor3bvEXT_remap_index 181
+#define SecondaryColor3dEXT_remap_index 182
+#define SecondaryColor3dvEXT_remap_index 183
+#define SecondaryColor3fEXT_remap_index 184
+#define SecondaryColor3fvEXT_remap_index 185
+#define SecondaryColor3iEXT_remap_index 186
+#define SecondaryColor3ivEXT_remap_index 187
+#define SecondaryColor3sEXT_remap_index 188
+#define SecondaryColor3svEXT_remap_index 189
+#define SecondaryColor3ubEXT_remap_index 190
+#define SecondaryColor3ubvEXT_remap_index 191
+#define SecondaryColor3uiEXT_remap_index 192
+#define SecondaryColor3uivEXT_remap_index 193
+#define SecondaryColor3usEXT_remap_index 194
+#define SecondaryColor3usvEXT_remap_index 195
+#define SecondaryColorPointerEXT_remap_index 196
+#define MultiDrawArraysEXT_remap_index 197
+#define MultiDrawElementsEXT_remap_index 198
+#define FogCoordPointerEXT_remap_index 199
+#define FogCoorddEXT_remap_index 200
+#define FogCoorddvEXT_remap_index 201
+#define FogCoordfEXT_remap_index 202
+#define FogCoordfvEXT_remap_index 203
+#define PixelTexGenSGIX_remap_index 204
+#define BlendFuncSeparateEXT_remap_index 205
+#define FlushVertexArrayRangeNV_remap_index 206
+#define VertexArrayRangeNV_remap_index 207
+#define CombinerInputNV_remap_index 208
+#define CombinerOutputNV_remap_index 209
+#define CombinerParameterfNV_remap_index 210
+#define CombinerParameterfvNV_remap_index 211
+#define CombinerParameteriNV_remap_index 212
+#define CombinerParameterivNV_remap_index 213
+#define FinalCombinerInputNV_remap_index 214
+#define GetCombinerInputParameterfvNV_remap_index 215
+#define GetCombinerInputParameterivNV_remap_index 216
+#define GetCombinerOutputParameterfvNV_remap_index 217
+#define GetCombinerOutputParameterivNV_remap_index 218
+#define GetFinalCombinerInputParameterfvNV_remap_index 219
+#define GetFinalCombinerInputParameterivNV_remap_index 220
+#define ResizeBuffersMESA_remap_index 221
+#define WindowPos2dMESA_remap_index 222
+#define WindowPos2dvMESA_remap_index 223
+#define WindowPos2fMESA_remap_index 224
+#define WindowPos2fvMESA_remap_index 225
+#define WindowPos2iMESA_remap_index 226
+#define WindowPos2ivMESA_remap_index 227
+#define WindowPos2sMESA_remap_index 228
+#define WindowPos2svMESA_remap_index 229
+#define WindowPos3dMESA_remap_index 230
+#define WindowPos3dvMESA_remap_index 231
+#define WindowPos3fMESA_remap_index 232
+#define WindowPos3fvMESA_remap_index 233
+#define WindowPos3iMESA_remap_index 234
+#define WindowPos3ivMESA_remap_index 235
+#define WindowPos3sMESA_remap_index 236
+#define WindowPos3svMESA_remap_index 237
+#define WindowPos4dMESA_remap_index 238
+#define WindowPos4dvMESA_remap_index 239
+#define WindowPos4fMESA_remap_index 240
+#define WindowPos4fvMESA_remap_index 241
+#define WindowPos4iMESA_remap_index 242
+#define WindowPos4ivMESA_remap_index 243
+#define WindowPos4sMESA_remap_index 244
+#define WindowPos4svMESA_remap_index 245
+#define MultiModeDrawArraysIBM_remap_index 246
+#define MultiModeDrawElementsIBM_remap_index 247
+#define DeleteFencesNV_remap_index 248
+#define FinishFenceNV_remap_index 249
+#define GenFencesNV_remap_index 250
+#define GetFenceivNV_remap_index 251
+#define IsFenceNV_remap_index 252
+#define SetFenceNV_remap_index 253
+#define TestFenceNV_remap_index 254
+#define AreProgramsResidentNV_remap_index 255
+#define BindProgramNV_remap_index 256
+#define DeleteProgramsNV_remap_index 257
+#define ExecuteProgramNV_remap_index 258
+#define GenProgramsNV_remap_index 259
+#define GetProgramParameterdvNV_remap_index 260
+#define GetProgramParameterfvNV_remap_index 261
+#define GetProgramStringNV_remap_index 262
+#define GetProgramivNV_remap_index 263
+#define GetTrackMatrixivNV_remap_index 264
+#define GetVertexAttribPointervNV_remap_index 265
+#define GetVertexAttribdvNV_remap_index 266
+#define GetVertexAttribfvNV_remap_index 267
+#define GetVertexAttribivNV_remap_index 268
+#define IsProgramNV_remap_index 269
+#define LoadProgramNV_remap_index 270
+#define ProgramParameters4dvNV_remap_index 271
+#define ProgramParameters4fvNV_remap_index 272
+#define RequestResidentProgramsNV_remap_index 273
+#define TrackMatrixNV_remap_index 274
+#define VertexAttrib1dNV_remap_index 275
+#define VertexAttrib1dvNV_remap_index 276
+#define VertexAttrib1fNV_remap_index 277
+#define VertexAttrib1fvNV_remap_index 278
+#define VertexAttrib1sNV_remap_index 279
+#define VertexAttrib1svNV_remap_index 280
+#define VertexAttrib2dNV_remap_index 281
+#define VertexAttrib2dvNV_remap_index 282
+#define VertexAttrib2fNV_remap_index 283
+#define VertexAttrib2fvNV_remap_index 284
+#define VertexAttrib2sNV_remap_index 285
+#define VertexAttrib2svNV_remap_index 286
+#define VertexAttrib3dNV_remap_index 287
+#define VertexAttrib3dvNV_remap_index 288
+#define VertexAttrib3fNV_remap_index 289
+#define VertexAttrib3fvNV_remap_index 290
+#define VertexAttrib3sNV_remap_index 291
+#define VertexAttrib3svNV_remap_index 292
+#define VertexAttrib4dNV_remap_index 293
+#define VertexAttrib4dvNV_remap_index 294
+#define VertexAttrib4fNV_remap_index 295
+#define VertexAttrib4fvNV_remap_index 296
+#define VertexAttrib4sNV_remap_index 297
+#define VertexAttrib4svNV_remap_index 298
+#define VertexAttrib4ubNV_remap_index 299
+#define VertexAttrib4ubvNV_remap_index 300
+#define VertexAttribPointerNV_remap_index 301
+#define VertexAttribs1dvNV_remap_index 302
+#define VertexAttribs1fvNV_remap_index 303
+#define VertexAttribs1svNV_remap_index 304
+#define VertexAttribs2dvNV_remap_index 305
+#define VertexAttribs2fvNV_remap_index 306
+#define VertexAttribs2svNV_remap_index 307
+#define VertexAttribs3dvNV_remap_index 308
+#define VertexAttribs3fvNV_remap_index 309
+#define VertexAttribs3svNV_remap_index 310
+#define VertexAttribs4dvNV_remap_index 311
+#define VertexAttribs4fvNV_remap_index 312
+#define VertexAttribs4svNV_remap_index 313
+#define VertexAttribs4ubvNV_remap_index 314
+#define GetTexBumpParameterfvATI_remap_index 315
+#define GetTexBumpParameterivATI_remap_index 316
+#define TexBumpParameterfvATI_remap_index 317
+#define TexBumpParameterivATI_remap_index 318
+#define AlphaFragmentOp1ATI_remap_index 319
+#define AlphaFragmentOp2ATI_remap_index 320
+#define AlphaFragmentOp3ATI_remap_index 321
+#define BeginFragmentShaderATI_remap_index 322
+#define BindFragmentShaderATI_remap_index 323
+#define ColorFragmentOp1ATI_remap_index 324
+#define ColorFragmentOp2ATI_remap_index 325
+#define ColorFragmentOp3ATI_remap_index 326
+#define DeleteFragmentShaderATI_remap_index 327
+#define EndFragmentShaderATI_remap_index 328
+#define GenFragmentShadersATI_remap_index 329
+#define PassTexCoordATI_remap_index 330
+#define SampleMapATI_remap_index 331
+#define SetFragmentShaderConstantATI_remap_index 332
+#define PointParameteriNV_remap_index 333
+#define PointParameterivNV_remap_index 334
+#define ActiveStencilFaceEXT_remap_index 335
+#define BindVertexArrayAPPLE_remap_index 336
+#define DeleteVertexArraysAPPLE_remap_index 337
+#define GenVertexArraysAPPLE_remap_index 338
+#define IsVertexArrayAPPLE_remap_index 339
+#define GetProgramNamedParameterdvNV_remap_index 340
+#define GetProgramNamedParameterfvNV_remap_index 341
+#define ProgramNamedParameter4dNV_remap_index 342
+#define ProgramNamedParameter4dvNV_remap_index 343
+#define ProgramNamedParameter4fNV_remap_index 344
+#define ProgramNamedParameter4fvNV_remap_index 345
+#define DepthBoundsEXT_remap_index 346
+#define BlendEquationSeparateEXT_remap_index 347
+#define BindFramebufferEXT_remap_index 348
+#define BindRenderbufferEXT_remap_index 349
+#define CheckFramebufferStatusEXT_remap_index 350
+#define DeleteFramebuffersEXT_remap_index 351
+#define DeleteRenderbuffersEXT_remap_index 352
+#define FramebufferRenderbufferEXT_remap_index 353
+#define FramebufferTexture1DEXT_remap_index 354
+#define FramebufferTexture2DEXT_remap_index 355
+#define FramebufferTexture3DEXT_remap_index 356
+#define GenFramebuffersEXT_remap_index 357
+#define GenRenderbuffersEXT_remap_index 358
+#define GenerateMipmapEXT_remap_index 359
+#define GetFramebufferAttachmentParameterivEXT_remap_index 360
+#define GetRenderbufferParameterivEXT_remap_index 361
+#define IsFramebufferEXT_remap_index 362
+#define IsRenderbufferEXT_remap_index 363
+#define RenderbufferStorageEXT_remap_index 364
+#define BlitFramebufferEXT_remap_index 365
+#define FramebufferTextureLayerEXT_remap_index 366
+#define ProvokingVertexEXT_remap_index 367
+#define StencilFuncSeparateATI_remap_index 368
+#define ProgramEnvParameters4fvEXT_remap_index 369
+#define ProgramLocalParameters4fvEXT_remap_index 370
+#define GetQueryObjecti64vEXT_remap_index 371
+#define GetQueryObjectui64vEXT_remap_index 372
 
 #define CALL_AttachShader(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint, GLuint)), driDispatchRemapTable[AttachShader_remap_index], parameters)
 #define GET_AttachShader(disp) GET_by_offset(disp, driDispatchRemapTable[AttachShader_remap_index])
@@ -3221,6 +3245,21 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define CALL_RenderbufferStorageMultisample(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLsizei, GLenum, GLsizei, GLsizei)), driDispatchRemapTable[RenderbufferStorageMultisample_remap_index], parameters)
 #define GET_RenderbufferStorageMultisample(disp) GET_by_offset(disp, driDispatchRemapTable[RenderbufferStorageMultisample_remap_index])
 #define SET_RenderbufferStorageMultisample(disp, fn) SET_by_offset(disp, driDispatchRemapTable[RenderbufferStorageMultisample_remap_index], fn)
+#define CALL_FlushMappedBufferRange(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLintptr, GLsizeiptr)), driDispatchRemapTable[FlushMappedBufferRange_remap_index], parameters)
+#define GET_FlushMappedBufferRange(disp) GET_by_offset(disp, driDispatchRemapTable[FlushMappedBufferRange_remap_index])
+#define SET_FlushMappedBufferRange(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FlushMappedBufferRange_remap_index], fn)
+#define CALL_MapBufferRange(disp, parameters) CALL_by_offset(disp, (GLvoid * (GLAPIENTRYP)(GLenum, GLintptr, GLsizeiptr, GLbitfield)), driDispatchRemapTable[MapBufferRange_remap_index], parameters)
+#define GET_MapBufferRange(disp) GET_by_offset(disp, driDispatchRemapTable[MapBufferRange_remap_index])
+#define SET_MapBufferRange(disp, fn) SET_by_offset(disp, driDispatchRemapTable[MapBufferRange_remap_index], fn)
+#define CALL_BindVertexArray(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint)), driDispatchRemapTable[BindVertexArray_remap_index], parameters)
+#define GET_BindVertexArray(disp) GET_by_offset(disp, driDispatchRemapTable[BindVertexArray_remap_index])
+#define SET_BindVertexArray(disp, fn) SET_by_offset(disp, driDispatchRemapTable[BindVertexArray_remap_index], fn)
+#define CALL_GenVertexArrays(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLsizei, GLuint *)), driDispatchRemapTable[GenVertexArrays_remap_index], parameters)
+#define GET_GenVertexArrays(disp) GET_by_offset(disp, driDispatchRemapTable[GenVertexArrays_remap_index])
+#define SET_GenVertexArrays(disp, fn) SET_by_offset(disp, driDispatchRemapTable[GenVertexArrays_remap_index], fn)
+#define CALL_CopyBufferSubData(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLintptr, GLintptr, GLsizeiptr)), driDispatchRemapTable[CopyBufferSubData_remap_index], parameters)
+#define GET_CopyBufferSubData(disp) GET_by_offset(disp, driDispatchRemapTable[CopyBufferSubData_remap_index])
+#define SET_CopyBufferSubData(disp, fn) SET_by_offset(disp, driDispatchRemapTable[CopyBufferSubData_remap_index], fn)
 #define CALL_PolygonOffsetEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLfloat, GLfloat)), driDispatchRemapTable[PolygonOffsetEXT_remap_index], parameters)
 #define GET_PolygonOffsetEXT(disp) GET_by_offset(disp, driDispatchRemapTable[PolygonOffsetEXT_remap_index])
 #define SET_PolygonOffsetEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[PolygonOffsetEXT_remap_index], fn)
@@ -3845,6 +3884,9 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define CALL_FramebufferTextureLayerEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLuint, GLint, GLint)), driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index], parameters)
 #define GET_FramebufferTextureLayerEXT(disp) GET_by_offset(disp, driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index])
 #define SET_FramebufferTextureLayerEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index], fn)
+#define CALL_ProvokingVertexEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum)), driDispatchRemapTable[ProvokingVertexEXT_remap_index], parameters)
+#define GET_ProvokingVertexEXT(disp) GET_by_offset(disp, driDispatchRemapTable[ProvokingVertexEXT_remap_index])
+#define SET_ProvokingVertexEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProvokingVertexEXT_remap_index], fn)
 #define CALL_StencilFuncSeparateATI(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLint, GLuint)), driDispatchRemapTable[StencilFuncSeparateATI_remap_index], parameters)
 #define GET_StencilFuncSeparateATI(disp) GET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index])
 #define SET_StencilFuncSeparateATI(disp, fn) SET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index], fn)
diff --git a/src/mesa/glapi/glX_proto_send.py b/src/mesa/glapi/glX_proto_send.py
index 501706acc7..daca1b767a 100644
--- a/src/mesa/glapi/glX_proto_send.py
+++ b/src/mesa/glapi/glX_proto_send.py
@@ -373,9 +373,13 @@ const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 				print '{'
 				print '    __GLXcontext * const gc = __glXGetCurrentContext();'
 				print ''
+				print '#ifdef GLX_DIRECT_RENDERING'
 				print '    if (gc->driContext) {'
 				print '    %sCALL_%s(GET_DISPATCH(), (%s));' % (ret_string, func.name, func.get_called_parameter_string())
-				print '    } else {'
+				print '    } else'
+				print '#endif'
+				print '    {'
+
 				footer = '}\n}\n'
 			else:
 				print '#define %s %d' % (func.opcode_name(), func.opcode_value())
diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml
index 4b66793e1c..06560d175a 100644
--- a/src/mesa/glapi/gl_API.xml
+++ b/src/mesa/glapi/gl_API.xml
@@ -7948,6 +7948,12 @@
 
 <xi:include href="ARB_framebuffer_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
 
+<xi:include href="ARB_copy_buffer.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<xi:include href="ARB_map_buffer_range.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+<xi:include href="ARB_vertex_array_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
 
 <!-- Non-ARB extensions sorted by extension number. -->
 
@@ -12224,6 +12230,8 @@
 
 <xi:include href="EXT_framebuffer_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
 
+<xi:include href="EXT_provoking_vertex.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
 <category name="GL_EXT_framebuffer_blit" number="316">
     <enum name="READ_FRAMEBUFFER_EXT" value="0x8CA8"/>
     <enum name="DRAW_FRAMEBUFFER_EXT" value="0x8CA9"/>
diff --git a/src/mesa/glapi/glapioffsets.h b/src/mesa/glapi/glapioffsets.h
index 14f305f025..9ed67a645f 100644
--- a/src/mesa/glapi/glapioffsets.h
+++ b/src/mesa/glapi/glapioffsets.h
@@ -594,220 +594,226 @@
 #define _gloffset_GetAttribLocationARB 559
 #define _gloffset_DrawBuffersARB 560
 #define _gloffset_RenderbufferStorageMultisample 561
-#define _gloffset_PolygonOffsetEXT 562
-#define _gloffset_GetPixelTexGenParameterfvSGIS 563
-#define _gloffset_GetPixelTexGenParameterivSGIS 564
-#define _gloffset_PixelTexGenParameterfSGIS 565
-#define _gloffset_PixelTexGenParameterfvSGIS 566
-#define _gloffset_PixelTexGenParameteriSGIS 567
-#define _gloffset_PixelTexGenParameterivSGIS 568
-#define _gloffset_SampleMaskSGIS 569
-#define _gloffset_SamplePatternSGIS 570
-#define _gloffset_ColorPointerEXT 571
-#define _gloffset_EdgeFlagPointerEXT 572
-#define _gloffset_IndexPointerEXT 573
-#define _gloffset_NormalPointerEXT 574
-#define _gloffset_TexCoordPointerEXT 575
-#define _gloffset_VertexPointerEXT 576
-#define _gloffset_PointParameterfEXT 577
-#define _gloffset_PointParameterfvEXT 578
-#define _gloffset_LockArraysEXT 579
-#define _gloffset_UnlockArraysEXT 580
-#define _gloffset_CullParameterdvEXT 581
-#define _gloffset_CullParameterfvEXT 582
-#define _gloffset_SecondaryColor3bEXT 583
-#define _gloffset_SecondaryColor3bvEXT 584
-#define _gloffset_SecondaryColor3dEXT 585
-#define _gloffset_SecondaryColor3dvEXT 586
-#define _gloffset_SecondaryColor3fEXT 587
-#define _gloffset_SecondaryColor3fvEXT 588
-#define _gloffset_SecondaryColor3iEXT 589
-#define _gloffset_SecondaryColor3ivEXT 590
-#define _gloffset_SecondaryColor3sEXT 591
-#define _gloffset_SecondaryColor3svEXT 592
-#define _gloffset_SecondaryColor3ubEXT 593
-#define _gloffset_SecondaryColor3ubvEXT 594
-#define _gloffset_SecondaryColor3uiEXT 595
-#define _gloffset_SecondaryColor3uivEXT 596
-#define _gloffset_SecondaryColor3usEXT 597
-#define _gloffset_SecondaryColor3usvEXT 598
-#define _gloffset_SecondaryColorPointerEXT 599
-#define _gloffset_MultiDrawArraysEXT 600
-#define _gloffset_MultiDrawElementsEXT 601
-#define _gloffset_FogCoordPointerEXT 602
-#define _gloffset_FogCoorddEXT 603
-#define _gloffset_FogCoorddvEXT 604
-#define _gloffset_FogCoordfEXT 605
-#define _gloffset_FogCoordfvEXT 606
-#define _gloffset_PixelTexGenSGIX 607
-#define _gloffset_BlendFuncSeparateEXT 608
-#define _gloffset_FlushVertexArrayRangeNV 609
-#define _gloffset_VertexArrayRangeNV 610
-#define _gloffset_CombinerInputNV 611
-#define _gloffset_CombinerOutputNV 612
-#define _gloffset_CombinerParameterfNV 613
-#define _gloffset_CombinerParameterfvNV 614
-#define _gloffset_CombinerParameteriNV 615
-#define _gloffset_CombinerParameterivNV 616
-#define _gloffset_FinalCombinerInputNV 617
-#define _gloffset_GetCombinerInputParameterfvNV 618
-#define _gloffset_GetCombinerInputParameterivNV 619
-#define _gloffset_GetCombinerOutputParameterfvNV 620
-#define _gloffset_GetCombinerOutputParameterivNV 621
-#define _gloffset_GetFinalCombinerInputParameterfvNV 622
-#define _gloffset_GetFinalCombinerInputParameterivNV 623
-#define _gloffset_ResizeBuffersMESA 624
-#define _gloffset_WindowPos2dMESA 625
-#define _gloffset_WindowPos2dvMESA 626
-#define _gloffset_WindowPos2fMESA 627
-#define _gloffset_WindowPos2fvMESA 628
-#define _gloffset_WindowPos2iMESA 629
-#define _gloffset_WindowPos2ivMESA 630
-#define _gloffset_WindowPos2sMESA 631
-#define _gloffset_WindowPos2svMESA 632
-#define _gloffset_WindowPos3dMESA 633
-#define _gloffset_WindowPos3dvMESA 634
-#define _gloffset_WindowPos3fMESA 635
-#define _gloffset_WindowPos3fvMESA 636
-#define _gloffset_WindowPos3iMESA 637
-#define _gloffset_WindowPos3ivMESA 638
-#define _gloffset_WindowPos3sMESA 639
-#define _gloffset_WindowPos3svMESA 640
-#define _gloffset_WindowPos4dMESA 641
-#define _gloffset_WindowPos4dvMESA 642
-#define _gloffset_WindowPos4fMESA 643
-#define _gloffset_WindowPos4fvMESA 644
-#define _gloffset_WindowPos4iMESA 645
-#define _gloffset_WindowPos4ivMESA 646
-#define _gloffset_WindowPos4sMESA 647
-#define _gloffset_WindowPos4svMESA 648
-#define _gloffset_MultiModeDrawArraysIBM 649
-#define _gloffset_MultiModeDrawElementsIBM 650
-#define _gloffset_DeleteFencesNV 651
-#define _gloffset_FinishFenceNV 652
-#define _gloffset_GenFencesNV 653
-#define _gloffset_GetFenceivNV 654
-#define _gloffset_IsFenceNV 655
-#define _gloffset_SetFenceNV 656
-#define _gloffset_TestFenceNV 657
-#define _gloffset_AreProgramsResidentNV 658
-#define _gloffset_BindProgramNV 659
-#define _gloffset_DeleteProgramsNV 660
-#define _gloffset_ExecuteProgramNV 661
-#define _gloffset_GenProgramsNV 662
-#define _gloffset_GetProgramParameterdvNV 663
-#define _gloffset_GetProgramParameterfvNV 664
-#define _gloffset_GetProgramStringNV 665
-#define _gloffset_GetProgramivNV 666
-#define _gloffset_GetTrackMatrixivNV 667
-#define _gloffset_GetVertexAttribPointervNV 668
-#define _gloffset_GetVertexAttribdvNV 669
-#define _gloffset_GetVertexAttribfvNV 670
-#define _gloffset_GetVertexAttribivNV 671
-#define _gloffset_IsProgramNV 672
-#define _gloffset_LoadProgramNV 673
-#define _gloffset_ProgramParameters4dvNV 674
-#define _gloffset_ProgramParameters4fvNV 675
-#define _gloffset_RequestResidentProgramsNV 676
-#define _gloffset_TrackMatrixNV 677
-#define _gloffset_VertexAttrib1dNV 678
-#define _gloffset_VertexAttrib1dvNV 679
-#define _gloffset_VertexAttrib1fNV 680
-#define _gloffset_VertexAttrib1fvNV 681
-#define _gloffset_VertexAttrib1sNV 682
-#define _gloffset_VertexAttrib1svNV 683
-#define _gloffset_VertexAttrib2dNV 684
-#define _gloffset_VertexAttrib2dvNV 685
-#define _gloffset_VertexAttrib2fNV 686
-#define _gloffset_VertexAttrib2fvNV 687
-#define _gloffset_VertexAttrib2sNV 688
-#define _gloffset_VertexAttrib2svNV 689
-#define _gloffset_VertexAttrib3dNV 690
-#define _gloffset_VertexAttrib3dvNV 691
-#define _gloffset_VertexAttrib3fNV 692
-#define _gloffset_VertexAttrib3fvNV 693
-#define _gloffset_VertexAttrib3sNV 694
-#define _gloffset_VertexAttrib3svNV 695
-#define _gloffset_VertexAttrib4dNV 696
-#define _gloffset_VertexAttrib4dvNV 697
-#define _gloffset_VertexAttrib4fNV 698
-#define _gloffset_VertexAttrib4fvNV 699
-#define _gloffset_VertexAttrib4sNV 700
-#define _gloffset_VertexAttrib4svNV 701
-#define _gloffset_VertexAttrib4ubNV 702
-#define _gloffset_VertexAttrib4ubvNV 703
-#define _gloffset_VertexAttribPointerNV 704
-#define _gloffset_VertexAttribs1dvNV 705
-#define _gloffset_VertexAttribs1fvNV 706
-#define _gloffset_VertexAttribs1svNV 707
-#define _gloffset_VertexAttribs2dvNV 708
-#define _gloffset_VertexAttribs2fvNV 709
-#define _gloffset_VertexAttribs2svNV 710
-#define _gloffset_VertexAttribs3dvNV 711
-#define _gloffset_VertexAttribs3fvNV 712
-#define _gloffset_VertexAttribs3svNV 713
-#define _gloffset_VertexAttribs4dvNV 714
-#define _gloffset_VertexAttribs4fvNV 715
-#define _gloffset_VertexAttribs4svNV 716
-#define _gloffset_VertexAttribs4ubvNV 717
-#define _gloffset_GetTexBumpParameterfvATI 718
-#define _gloffset_GetTexBumpParameterivATI 719
-#define _gloffset_TexBumpParameterfvATI 720
-#define _gloffset_TexBumpParameterivATI 721
-#define _gloffset_AlphaFragmentOp1ATI 722
-#define _gloffset_AlphaFragmentOp2ATI 723
-#define _gloffset_AlphaFragmentOp3ATI 724
-#define _gloffset_BeginFragmentShaderATI 725
-#define _gloffset_BindFragmentShaderATI 726
-#define _gloffset_ColorFragmentOp1ATI 727
-#define _gloffset_ColorFragmentOp2ATI 728
-#define _gloffset_ColorFragmentOp3ATI 729
-#define _gloffset_DeleteFragmentShaderATI 730
-#define _gloffset_EndFragmentShaderATI 731
-#define _gloffset_GenFragmentShadersATI 732
-#define _gloffset_PassTexCoordATI 733
-#define _gloffset_SampleMapATI 734
-#define _gloffset_SetFragmentShaderConstantATI 735
-#define _gloffset_PointParameteriNV 736
-#define _gloffset_PointParameterivNV 737
-#define _gloffset_ActiveStencilFaceEXT 738
-#define _gloffset_BindVertexArrayAPPLE 739
-#define _gloffset_DeleteVertexArraysAPPLE 740
-#define _gloffset_GenVertexArraysAPPLE 741
-#define _gloffset_IsVertexArrayAPPLE 742
-#define _gloffset_GetProgramNamedParameterdvNV 743
-#define _gloffset_GetProgramNamedParameterfvNV 744
-#define _gloffset_ProgramNamedParameter4dNV 745
-#define _gloffset_ProgramNamedParameter4dvNV 746
-#define _gloffset_ProgramNamedParameter4fNV 747
-#define _gloffset_ProgramNamedParameter4fvNV 748
-#define _gloffset_DepthBoundsEXT 749
-#define _gloffset_BlendEquationSeparateEXT 750
-#define _gloffset_BindFramebufferEXT 751
-#define _gloffset_BindRenderbufferEXT 752
-#define _gloffset_CheckFramebufferStatusEXT 753
-#define _gloffset_DeleteFramebuffersEXT 754
-#define _gloffset_DeleteRenderbuffersEXT 755
-#define _gloffset_FramebufferRenderbufferEXT 756
-#define _gloffset_FramebufferTexture1DEXT 757
-#define _gloffset_FramebufferTexture2DEXT 758
-#define _gloffset_FramebufferTexture3DEXT 759
-#define _gloffset_GenFramebuffersEXT 760
-#define _gloffset_GenRenderbuffersEXT 761
-#define _gloffset_GenerateMipmapEXT 762
-#define _gloffset_GetFramebufferAttachmentParameterivEXT 763
-#define _gloffset_GetRenderbufferParameterivEXT 764
-#define _gloffset_IsFramebufferEXT 765
-#define _gloffset_IsRenderbufferEXT 766
-#define _gloffset_RenderbufferStorageEXT 767
-#define _gloffset_BlitFramebufferEXT 768
-#define _gloffset_FramebufferTextureLayerEXT 769
-#define _gloffset_StencilFuncSeparateATI 770
-#define _gloffset_ProgramEnvParameters4fvEXT 771
-#define _gloffset_ProgramLocalParameters4fvEXT 772
-#define _gloffset_GetQueryObjecti64vEXT 773
-#define _gloffset_GetQueryObjectui64vEXT 774
-#define _gloffset_FIRST_DYNAMIC 775
+#define _gloffset_FlushMappedBufferRange 562
+#define _gloffset_MapBufferRange 563
+#define _gloffset_BindVertexArray 564
+#define _gloffset_GenVertexArrays 565
+#define _gloffset_CopyBufferSubData 566
+#define _gloffset_PolygonOffsetEXT 567
+#define _gloffset_GetPixelTexGenParameterfvSGIS 568
+#define _gloffset_GetPixelTexGenParameterivSGIS 569
+#define _gloffset_PixelTexGenParameterfSGIS 570
+#define _gloffset_PixelTexGenParameterfvSGIS 571
+#define _gloffset_PixelTexGenParameteriSGIS 572
+#define _gloffset_PixelTexGenParameterivSGIS 573
+#define _gloffset_SampleMaskSGIS 574
+#define _gloffset_SamplePatternSGIS 575
+#define _gloffset_ColorPointerEXT 576
+#define _gloffset_EdgeFlagPointerEXT 577
+#define _gloffset_IndexPointerEXT 578
+#define _gloffset_NormalPointerEXT 579
+#define _gloffset_TexCoordPointerEXT 580
+#define _gloffset_VertexPointerEXT 581
+#define _gloffset_PointParameterfEXT 582
+#define _gloffset_PointParameterfvEXT 583
+#define _gloffset_LockArraysEXT 584
+#define _gloffset_UnlockArraysEXT 585
+#define _gloffset_CullParameterdvEXT 586
+#define _gloffset_CullParameterfvEXT 587
+#define _gloffset_SecondaryColor3bEXT 588
+#define _gloffset_SecondaryColor3bvEXT 589
+#define _gloffset_SecondaryColor3dEXT 590
+#define _gloffset_SecondaryColor3dvEXT 591
+#define _gloffset_SecondaryColor3fEXT 592
+#define _gloffset_SecondaryColor3fvEXT 593
+#define _gloffset_SecondaryColor3iEXT 594
+#define _gloffset_SecondaryColor3ivEXT 595
+#define _gloffset_SecondaryColor3sEXT 596
+#define _gloffset_SecondaryColor3svEXT 597
+#define _gloffset_SecondaryColor3ubEXT 598
+#define _gloffset_SecondaryColor3ubvEXT 599
+#define _gloffset_SecondaryColor3uiEXT 600
+#define _gloffset_SecondaryColor3uivEXT 601
+#define _gloffset_SecondaryColor3usEXT 602
+#define _gloffset_SecondaryColor3usvEXT 603
+#define _gloffset_SecondaryColorPointerEXT 604
+#define _gloffset_MultiDrawArraysEXT 605
+#define _gloffset_MultiDrawElementsEXT 606
+#define _gloffset_FogCoordPointerEXT 607
+#define _gloffset_FogCoorddEXT 608
+#define _gloffset_FogCoorddvEXT 609
+#define _gloffset_FogCoordfEXT 610
+#define _gloffset_FogCoordfvEXT 611
+#define _gloffset_PixelTexGenSGIX 612
+#define _gloffset_BlendFuncSeparateEXT 613
+#define _gloffset_FlushVertexArrayRangeNV 614
+#define _gloffset_VertexArrayRangeNV 615
+#define _gloffset_CombinerInputNV 616
+#define _gloffset_CombinerOutputNV 617
+#define _gloffset_CombinerParameterfNV 618
+#define _gloffset_CombinerParameterfvNV 619
+#define _gloffset_CombinerParameteriNV 620
+#define _gloffset_CombinerParameterivNV 621
+#define _gloffset_FinalCombinerInputNV 622
+#define _gloffset_GetCombinerInputParameterfvNV 623
+#define _gloffset_GetCombinerInputParameterivNV 624
+#define _gloffset_GetCombinerOutputParameterfvNV 625
+#define _gloffset_GetCombinerOutputParameterivNV 626
+#define _gloffset_GetFinalCombinerInputParameterfvNV 627
+#define _gloffset_GetFinalCombinerInputParameterivNV 628
+#define _gloffset_ResizeBuffersMESA 629
+#define _gloffset_WindowPos2dMESA 630
+#define _gloffset_WindowPos2dvMESA 631
+#define _gloffset_WindowPos2fMESA 632
+#define _gloffset_WindowPos2fvMESA 633
+#define _gloffset_WindowPos2iMESA 634
+#define _gloffset_WindowPos2ivMESA 635
+#define _gloffset_WindowPos2sMESA 636
+#define _gloffset_WindowPos2svMESA 637
+#define _gloffset_WindowPos3dMESA 638
+#define _gloffset_WindowPos3dvMESA 639
+#define _gloffset_WindowPos3fMESA 640
+#define _gloffset_WindowPos3fvMESA 641
+#define _gloffset_WindowPos3iMESA 642
+#define _gloffset_WindowPos3ivMESA 643
+#define _gloffset_WindowPos3sMESA 644
+#define _gloffset_WindowPos3svMESA 645
+#define _gloffset_WindowPos4dMESA 646
+#define _gloffset_WindowPos4dvMESA 647
+#define _gloffset_WindowPos4fMESA 648
+#define _gloffset_WindowPos4fvMESA 649
+#define _gloffset_WindowPos4iMESA 650
+#define _gloffset_WindowPos4ivMESA 651
+#define _gloffset_WindowPos4sMESA 652
+#define _gloffset_WindowPos4svMESA 653
+#define _gloffset_MultiModeDrawArraysIBM 654
+#define _gloffset_MultiModeDrawElementsIBM 655
+#define _gloffset_DeleteFencesNV 656
+#define _gloffset_FinishFenceNV 657
+#define _gloffset_GenFencesNV 658
+#define _gloffset_GetFenceivNV 659
+#define _gloffset_IsFenceNV 660
+#define _gloffset_SetFenceNV 661
+#define _gloffset_TestFenceNV 662
+#define _gloffset_AreProgramsResidentNV 663
+#define _gloffset_BindProgramNV 664
+#define _gloffset_DeleteProgramsNV 665
+#define _gloffset_ExecuteProgramNV 666
+#define _gloffset_GenProgramsNV 667
+#define _gloffset_GetProgramParameterdvNV 668
+#define _gloffset_GetProgramParameterfvNV 669
+#define _gloffset_GetProgramStringNV 670
+#define _gloffset_GetProgramivNV 671
+#define _gloffset_GetTrackMatrixivNV 672
+#define _gloffset_GetVertexAttribPointervNV 673
+#define _gloffset_GetVertexAttribdvNV 674
+#define _gloffset_GetVertexAttribfvNV 675
+#define _gloffset_GetVertexAttribivNV 676
+#define _gloffset_IsProgramNV 677
+#define _gloffset_LoadProgramNV 678
+#define _gloffset_ProgramParameters4dvNV 679
+#define _gloffset_ProgramParameters4fvNV 680
+#define _gloffset_RequestResidentProgramsNV 681
+#define _gloffset_TrackMatrixNV 682
+#define _gloffset_VertexAttrib1dNV 683
+#define _gloffset_VertexAttrib1dvNV 684
+#define _gloffset_VertexAttrib1fNV 685
+#define _gloffset_VertexAttrib1fvNV 686
+#define _gloffset_VertexAttrib1sNV 687
+#define _gloffset_VertexAttrib1svNV 688
+#define _gloffset_VertexAttrib2dNV 689
+#define _gloffset_VertexAttrib2dvNV 690
+#define _gloffset_VertexAttrib2fNV 691
+#define _gloffset_VertexAttrib2fvNV 692
+#define _gloffset_VertexAttrib2sNV 693
+#define _gloffset_VertexAttrib2svNV 694
+#define _gloffset_VertexAttrib3dNV 695
+#define _gloffset_VertexAttrib3dvNV 696
+#define _gloffset_VertexAttrib3fNV 697
+#define _gloffset_VertexAttrib3fvNV 698
+#define _gloffset_VertexAttrib3sNV 699
+#define _gloffset_VertexAttrib3svNV 700
+#define _gloffset_VertexAttrib4dNV 701
+#define _gloffset_VertexAttrib4dvNV 702
+#define _gloffset_VertexAttrib4fNV 703
+#define _gloffset_VertexAttrib4fvNV 704
+#define _gloffset_VertexAttrib4sNV 705
+#define _gloffset_VertexAttrib4svNV 706
+#define _gloffset_VertexAttrib4ubNV 707
+#define _gloffset_VertexAttrib4ubvNV 708
+#define _gloffset_VertexAttribPointerNV 709
+#define _gloffset_VertexAttribs1dvNV 710
+#define _gloffset_VertexAttribs1fvNV 711
+#define _gloffset_VertexAttribs1svNV 712
+#define _gloffset_VertexAttribs2dvNV 713
+#define _gloffset_VertexAttribs2fvNV 714
+#define _gloffset_VertexAttribs2svNV 715
+#define _gloffset_VertexAttribs3dvNV 716
+#define _gloffset_VertexAttribs3fvNV 717
+#define _gloffset_VertexAttribs3svNV 718
+#define _gloffset_VertexAttribs4dvNV 719
+#define _gloffset_VertexAttribs4fvNV 720
+#define _gloffset_VertexAttribs4svNV 721
+#define _gloffset_VertexAttribs4ubvNV 722
+#define _gloffset_GetTexBumpParameterfvATI 723
+#define _gloffset_GetTexBumpParameterivATI 724
+#define _gloffset_TexBumpParameterfvATI 725
+#define _gloffset_TexBumpParameterivATI 726
+#define _gloffset_AlphaFragmentOp1ATI 727
+#define _gloffset_AlphaFragmentOp2ATI 728
+#define _gloffset_AlphaFragmentOp3ATI 729
+#define _gloffset_BeginFragmentShaderATI 730
+#define _gloffset_BindFragmentShaderATI 731
+#define _gloffset_ColorFragmentOp1ATI 732
+#define _gloffset_ColorFragmentOp2ATI 733
+#define _gloffset_ColorFragmentOp3ATI 734
+#define _gloffset_DeleteFragmentShaderATI 735
+#define _gloffset_EndFragmentShaderATI 736
+#define _gloffset_GenFragmentShadersATI 737
+#define _gloffset_PassTexCoordATI 738
+#define _gloffset_SampleMapATI 739
+#define _gloffset_SetFragmentShaderConstantATI 740
+#define _gloffset_PointParameteriNV 741
+#define _gloffset_PointParameterivNV 742
+#define _gloffset_ActiveStencilFaceEXT 743
+#define _gloffset_BindVertexArrayAPPLE 744
+#define _gloffset_DeleteVertexArraysAPPLE 745
+#define _gloffset_GenVertexArraysAPPLE 746
+#define _gloffset_IsVertexArrayAPPLE 747
+#define _gloffset_GetProgramNamedParameterdvNV 748
+#define _gloffset_GetProgramNamedParameterfvNV 749
+#define _gloffset_ProgramNamedParameter4dNV 750
+#define _gloffset_ProgramNamedParameter4dvNV 751
+#define _gloffset_ProgramNamedParameter4fNV 752
+#define _gloffset_ProgramNamedParameter4fvNV 753
+#define _gloffset_DepthBoundsEXT 754
+#define _gloffset_BlendEquationSeparateEXT 755
+#define _gloffset_BindFramebufferEXT 756
+#define _gloffset_BindRenderbufferEXT 757
+#define _gloffset_CheckFramebufferStatusEXT 758
+#define _gloffset_DeleteFramebuffersEXT 759
+#define _gloffset_DeleteRenderbuffersEXT 760
+#define _gloffset_FramebufferRenderbufferEXT 761
+#define _gloffset_FramebufferTexture1DEXT 762
+#define _gloffset_FramebufferTexture2DEXT 763
+#define _gloffset_FramebufferTexture3DEXT 764
+#define _gloffset_GenFramebuffersEXT 765
+#define _gloffset_GenRenderbuffersEXT 766
+#define _gloffset_GenerateMipmapEXT 767
+#define _gloffset_GetFramebufferAttachmentParameterivEXT 768
+#define _gloffset_GetRenderbufferParameterivEXT 769
+#define _gloffset_IsFramebufferEXT 770
+#define _gloffset_IsRenderbufferEXT 771
+#define _gloffset_RenderbufferStorageEXT 772
+#define _gloffset_BlitFramebufferEXT 773
+#define _gloffset_FramebufferTextureLayerEXT 774
+#define _gloffset_ProvokingVertexEXT 775
+#define _gloffset_StencilFuncSeparateATI 776
+#define _gloffset_ProgramEnvParameters4fvEXT 777
+#define _gloffset_ProgramLocalParameters4fvEXT 778
+#define _gloffset_GetQueryObjecti64vEXT 779
+#define _gloffset_GetQueryObjectui64vEXT 780
+#define _gloffset_FIRST_DYNAMIC 781
 
 #else
 
@@ -965,6 +971,11 @@
 #define _gloffset_GetAttribLocationARB driDispatchRemapTable[GetAttribLocationARB_remap_index]
 #define _gloffset_DrawBuffersARB driDispatchRemapTable[DrawBuffersARB_remap_index]
 #define _gloffset_RenderbufferStorageMultisample driDispatchRemapTable[RenderbufferStorageMultisample_remap_index]
+#define _gloffset_FlushMappedBufferRange driDispatchRemapTable[FlushMappedBufferRange_remap_index]
+#define _gloffset_MapBufferRange driDispatchRemapTable[MapBufferRange_remap_index]
+#define _gloffset_BindVertexArray driDispatchRemapTable[BindVertexArray_remap_index]
+#define _gloffset_GenVertexArrays driDispatchRemapTable[GenVertexArrays_remap_index]
+#define _gloffset_CopyBufferSubData driDispatchRemapTable[CopyBufferSubData_remap_index]
 #define _gloffset_PolygonOffsetEXT driDispatchRemapTable[PolygonOffsetEXT_remap_index]
 #define _gloffset_GetPixelTexGenParameterfvSGIS driDispatchRemapTable[GetPixelTexGenParameterfvSGIS_remap_index]
 #define _gloffset_GetPixelTexGenParameterivSGIS driDispatchRemapTable[GetPixelTexGenParameterivSGIS_remap_index]
@@ -1173,6 +1184,7 @@
 #define _gloffset_RenderbufferStorageEXT driDispatchRemapTable[RenderbufferStorageEXT_remap_index]
 #define _gloffset_BlitFramebufferEXT driDispatchRemapTable[BlitFramebufferEXT_remap_index]
 #define _gloffset_FramebufferTextureLayerEXT driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index]
+#define _gloffset_ProvokingVertexEXT driDispatchRemapTable[ProvokingVertexEXT_remap_index]
 #define _gloffset_StencilFuncSeparateATI driDispatchRemapTable[StencilFuncSeparateATI_remap_index]
 #define _gloffset_ProgramEnvParameters4fvEXT driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index]
 #define _gloffset_ProgramLocalParameters4fvEXT driDispatchRemapTable[ProgramLocalParameters4fvEXT_remap_index]
diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h
index 9ecb036461..308de4facc 100644
--- a/src/mesa/glapi/glapitable.h
+++ b/src/mesa/glapi/glapitable.h
@@ -602,219 +602,225 @@ struct _glapi_table
    GLint (GLAPIENTRYP GetAttribLocationARB)(GLhandleARB program, const GLcharARB * name); /* 559 */
    void (GLAPIENTRYP DrawBuffersARB)(GLsizei n, const GLenum * bufs); /* 560 */
    void (GLAPIENTRYP RenderbufferStorageMultisample)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height); /* 561 */
-   void (GLAPIENTRYP PolygonOffsetEXT)(GLfloat factor, GLfloat bias); /* 562 */
-   void (GLAPIENTRYP GetPixelTexGenParameterfvSGIS)(GLenum pname, GLfloat * params); /* 563 */
-   void (GLAPIENTRYP GetPixelTexGenParameterivSGIS)(GLenum pname, GLint * params); /* 564 */
-   void (GLAPIENTRYP PixelTexGenParameterfSGIS)(GLenum pname, GLfloat param); /* 565 */
-   void (GLAPIENTRYP PixelTexGenParameterfvSGIS)(GLenum pname, const GLfloat * params); /* 566 */
-   void (GLAPIENTRYP PixelTexGenParameteriSGIS)(GLenum pname, GLint param); /* 567 */
-   void (GLAPIENTRYP PixelTexGenParameterivSGIS)(GLenum pname, const GLint * params); /* 568 */
-   void (GLAPIENTRYP SampleMaskSGIS)(GLclampf value, GLboolean invert); /* 569 */
-   void (GLAPIENTRYP SamplePatternSGIS)(GLenum pattern); /* 570 */
-   void (GLAPIENTRYP ColorPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 571 */
-   void (GLAPIENTRYP EdgeFlagPointerEXT)(GLsizei stride, GLsizei count, const GLboolean * pointer); /* 572 */
-   void (GLAPIENTRYP IndexPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 573 */
-   void (GLAPIENTRYP NormalPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 574 */
-   void (GLAPIENTRYP TexCoordPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 575 */
-   void (GLAPIENTRYP VertexPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 576 */
-   void (GLAPIENTRYP PointParameterfEXT)(GLenum pname, GLfloat param); /* 577 */
-   void (GLAPIENTRYP PointParameterfvEXT)(GLenum pname, const GLfloat * params); /* 578 */
-   void (GLAPIENTRYP LockArraysEXT)(GLint first, GLsizei count); /* 579 */
-   void (GLAPIENTRYP UnlockArraysEXT)(void); /* 580 */
-   void (GLAPIENTRYP CullParameterdvEXT)(GLenum pname, GLdouble * params); /* 581 */
-   void (GLAPIENTRYP CullParameterfvEXT)(GLenum pname, GLfloat * params); /* 582 */
-   void (GLAPIENTRYP SecondaryColor3bEXT)(GLbyte red, GLbyte green, GLbyte blue); /* 583 */
-   void (GLAPIENTRYP SecondaryColor3bvEXT)(const GLbyte * v); /* 584 */
-   void (GLAPIENTRYP SecondaryColor3dEXT)(GLdouble red, GLdouble green, GLdouble blue); /* 585 */
-   void (GLAPIENTRYP SecondaryColor3dvEXT)(const GLdouble * v); /* 586 */
-   void (GLAPIENTRYP SecondaryColor3fEXT)(GLfloat red, GLfloat green, GLfloat blue); /* 587 */
-   void (GLAPIENTRYP SecondaryColor3fvEXT)(const GLfloat * v); /* 588 */
-   void (GLAPIENTRYP SecondaryColor3iEXT)(GLint red, GLint green, GLint blue); /* 589 */
-   void (GLAPIENTRYP SecondaryColor3ivEXT)(const GLint * v); /* 590 */
-   void (GLAPIENTRYP SecondaryColor3sEXT)(GLshort red, GLshort green, GLshort blue); /* 591 */
-   void (GLAPIENTRYP SecondaryColor3svEXT)(const GLshort * v); /* 592 */
-   void (GLAPIENTRYP SecondaryColor3ubEXT)(GLubyte red, GLubyte green, GLubyte blue); /* 593 */
-   void (GLAPIENTRYP SecondaryColor3ubvEXT)(const GLubyte * v); /* 594 */
-   void (GLAPIENTRYP SecondaryColor3uiEXT)(GLuint red, GLuint green, GLuint blue); /* 595 */
-   void (GLAPIENTRYP SecondaryColor3uivEXT)(const GLuint * v); /* 596 */
-   void (GLAPIENTRYP SecondaryColor3usEXT)(GLushort red, GLushort green, GLushort blue); /* 597 */
-   void (GLAPIENTRYP SecondaryColor3usvEXT)(const GLushort * v); /* 598 */
-   void (GLAPIENTRYP SecondaryColorPointerEXT)(GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 599 */
-   void (GLAPIENTRYP MultiDrawArraysEXT)(GLenum mode, GLint * first, GLsizei * count, GLsizei primcount); /* 600 */
-   void (GLAPIENTRYP MultiDrawElementsEXT)(GLenum mode, const GLsizei * count, GLenum type, const GLvoid ** indices, GLsizei primcount); /* 601 */
-   void (GLAPIENTRYP FogCoordPointerEXT)(GLenum type, GLsizei stride, const GLvoid * pointer); /* 602 */
-   void (GLAPIENTRYP FogCoorddEXT)(GLdouble coord); /* 603 */
-   void (GLAPIENTRYP FogCoorddvEXT)(const GLdouble * coord); /* 604 */
-   void (GLAPIENTRYP FogCoordfEXT)(GLfloat coord); /* 605 */
-   void (GLAPIENTRYP FogCoordfvEXT)(const GLfloat * coord); /* 606 */
-   void (GLAPIENTRYP PixelTexGenSGIX)(GLenum mode); /* 607 */
-   void (GLAPIENTRYP BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha); /* 608 */
-   void (GLAPIENTRYP FlushVertexArrayRangeNV)(void); /* 609 */
-   void (GLAPIENTRYP VertexArrayRangeNV)(GLsizei length, const GLvoid * pointer); /* 610 */
-   void (GLAPIENTRYP CombinerInputNV)(GLenum stage, GLenum portion, GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 611 */
-   void (GLAPIENTRYP CombinerOutputNV)(GLenum stage, GLenum portion, GLenum abOutput, GLenum cdOutput, GLenum sumOutput, GLenum scale, GLenum bias, GLboolean abDotProduct, GLboolean cdDotProduct, GLboolean muxSum); /* 612 */
-   void (GLAPIENTRYP CombinerParameterfNV)(GLenum pname, GLfloat param); /* 613 */
-   void (GLAPIENTRYP CombinerParameterfvNV)(GLenum pname, const GLfloat * params); /* 614 */
-   void (GLAPIENTRYP CombinerParameteriNV)(GLenum pname, GLint param); /* 615 */
-   void (GLAPIENTRYP CombinerParameterivNV)(GLenum pname, const GLint * params); /* 616 */
-   void (GLAPIENTRYP FinalCombinerInputNV)(GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 617 */
-   void (GLAPIENTRYP GetCombinerInputParameterfvNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLfloat * params); /* 618 */
-   void (GLAPIENTRYP GetCombinerInputParameterivNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLint * params); /* 619 */
-   void (GLAPIENTRYP GetCombinerOutputParameterfvNV)(GLenum stage, GLenum portion, GLenum pname, GLfloat * params); /* 620 */
-   void (GLAPIENTRYP GetCombinerOutputParameterivNV)(GLenum stage, GLenum portion, GLenum pname, GLint * params); /* 621 */
-   void (GLAPIENTRYP GetFinalCombinerInputParameterfvNV)(GLenum variable, GLenum pname, GLfloat * params); /* 622 */
-   void (GLAPIENTRYP GetFinalCombinerInputParameterivNV)(GLenum variable, GLenum pname, GLint * params); /* 623 */
-   void (GLAPIENTRYP ResizeBuffersMESA)(void); /* 624 */
-   void (GLAPIENTRYP WindowPos2dMESA)(GLdouble x, GLdouble y); /* 625 */
-   void (GLAPIENTRYP WindowPos2dvMESA)(const GLdouble * v); /* 626 */
-   void (GLAPIENTRYP WindowPos2fMESA)(GLfloat x, GLfloat y); /* 627 */
-   void (GLAPIENTRYP WindowPos2fvMESA)(const GLfloat * v); /* 628 */
-   void (GLAPIENTRYP WindowPos2iMESA)(GLint x, GLint y); /* 629 */
-   void (GLAPIENTRYP WindowPos2ivMESA)(const GLint * v); /* 630 */
-   void (GLAPIENTRYP WindowPos2sMESA)(GLshort x, GLshort y); /* 631 */
-   void (GLAPIENTRYP WindowPos2svMESA)(const GLshort * v); /* 632 */
-   void (GLAPIENTRYP WindowPos3dMESA)(GLdouble x, GLdouble y, GLdouble z); /* 633 */
-   void (GLAPIENTRYP WindowPos3dvMESA)(const GLdouble * v); /* 634 */
-   void (GLAPIENTRYP WindowPos3fMESA)(GLfloat x, GLfloat y, GLfloat z); /* 635 */
-   void (GLAPIENTRYP WindowPos3fvMESA)(const GLfloat * v); /* 636 */
-   void (GLAPIENTRYP WindowPos3iMESA)(GLint x, GLint y, GLint z); /* 637 */
-   void (GLAPIENTRYP WindowPos3ivMESA)(const GLint * v); /* 638 */
-   void (GLAPIENTRYP WindowPos3sMESA)(GLshort x, GLshort y, GLshort z); /* 639 */
-   void (GLAPIENTRYP WindowPos3svMESA)(const GLshort * v); /* 640 */
-   void (GLAPIENTRYP WindowPos4dMESA)(GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 641 */
-   void (GLAPIENTRYP WindowPos4dvMESA)(const GLdouble * v); /* 642 */
-   void (GLAPIENTRYP WindowPos4fMESA)(GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 643 */
-   void (GLAPIENTRYP WindowPos4fvMESA)(const GLfloat * v); /* 644 */
-   void (GLAPIENTRYP WindowPos4iMESA)(GLint x, GLint y, GLint z, GLint w); /* 645 */
-   void (GLAPIENTRYP WindowPos4ivMESA)(const GLint * v); /* 646 */
-   void (GLAPIENTRYP WindowPos4sMESA)(GLshort x, GLshort y, GLshort z, GLshort w); /* 647 */
-   void (GLAPIENTRYP WindowPos4svMESA)(const GLshort * v); /* 648 */
-   void (GLAPIENTRYP MultiModeDrawArraysIBM)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); /* 649 */
-   void (GLAPIENTRYP MultiModeDrawElementsIBM)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); /* 650 */
-   void (GLAPIENTRYP DeleteFencesNV)(GLsizei n, const GLuint * fences); /* 651 */
-   void (GLAPIENTRYP FinishFenceNV)(GLuint fence); /* 652 */
-   void (GLAPIENTRYP GenFencesNV)(GLsizei n, GLuint * fences); /* 653 */
-   void (GLAPIENTRYP GetFenceivNV)(GLuint fence, GLenum pname, GLint * params); /* 654 */
-   GLboolean (GLAPIENTRYP IsFenceNV)(GLuint fence); /* 655 */
-   void (GLAPIENTRYP SetFenceNV)(GLuint fence, GLenum condition); /* 656 */
-   GLboolean (GLAPIENTRYP TestFenceNV)(GLuint fence); /* 657 */
-   GLboolean (GLAPIENTRYP AreProgramsResidentNV)(GLsizei n, const GLuint * ids, GLboolean * residences); /* 658 */
-   void (GLAPIENTRYP BindProgramNV)(GLenum target, GLuint program); /* 659 */
-   void (GLAPIENTRYP DeleteProgramsNV)(GLsizei n, const GLuint * programs); /* 660 */
-   void (GLAPIENTRYP ExecuteProgramNV)(GLenum target, GLuint id, const GLfloat * params); /* 661 */
-   void (GLAPIENTRYP GenProgramsNV)(GLsizei n, GLuint * programs); /* 662 */
-   void (GLAPIENTRYP GetProgramParameterdvNV)(GLenum target, GLuint index, GLenum pname, GLdouble * params); /* 663 */
-   void (GLAPIENTRYP GetProgramParameterfvNV)(GLenum target, GLuint index, GLenum pname, GLfloat * params); /* 664 */
-   void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 665 */
-   void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 666 */
-   void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 667 */
-   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 668 */
-   void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 669 */
-   void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 670 */
-   void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 671 */
-   GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 672 */
-   void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 673 */
-   void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 674 */
-   void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 675 */
-   void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 676 */
-   void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 677 */
-   void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 678 */
-   void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 679 */
-   void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 680 */
-   void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 681 */
-   void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 682 */
-   void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 683 */
-   void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 684 */
-   void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 685 */
-   void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 686 */
-   void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 687 */
-   void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 688 */
-   void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 689 */
-   void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 690 */
-   void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 691 */
-   void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 692 */
-   void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 693 */
-   void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 694 */
-   void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 695 */
-   void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 696 */
-   void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 697 */
-   void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 698 */
-   void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 699 */
-   void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 700 */
-   void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 701 */
-   void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 702 */
-   void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 703 */
-   void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 704 */
-   void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 705 */
-   void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 706 */
-   void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 707 */
-   void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 708 */
-   void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 709 */
-   void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 710 */
-   void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 711 */
-   void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 712 */
-   void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 713 */
-   void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 714 */
-   void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 715 */
-   void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 716 */
-   void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 717 */
-   void (GLAPIENTRYP GetTexBumpParameterfvATI)(GLenum pname, GLfloat * param); /* 718 */
-   void (GLAPIENTRYP GetTexBumpParameterivATI)(GLenum pname, GLint * param); /* 719 */
-   void (GLAPIENTRYP TexBumpParameterfvATI)(GLenum pname, const GLfloat * param); /* 720 */
-   void (GLAPIENTRYP TexBumpParameterivATI)(GLenum pname, const GLint * param); /* 721 */
-   void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 722 */
-   void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 723 */
-   void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 724 */
-   void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 725 */
-   void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 726 */
-   void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 727 */
-   void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 728 */
-   void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 729 */
-   void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 730 */
-   void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 731 */
-   GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 732 */
-   void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 733 */
-   void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 734 */
-   void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 735 */
-   void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 736 */
-   void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 737 */
-   void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 738 */
-   void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 739 */
-   void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 740 */
-   void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 741 */
-   GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 742 */
-   void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 743 */
-   void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 744 */
-   void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 745 */
-   void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 746 */
-   void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 747 */
-   void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 748 */
-   void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 749 */
-   void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 750 */
-   void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 751 */
-   void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 752 */
-   GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 753 */
-   void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 754 */
-   void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 755 */
-   void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 756 */
-   void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 757 */
-   void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 758 */
-   void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 759 */
-   void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 760 */
-   void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 761 */
-   void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 762 */
-   void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 763 */
-   void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 764 */
-   GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 765 */
-   GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 766 */
-   void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 767 */
-   void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 768 */
-   void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 769 */
-   void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 770 */
-   void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 771 */
-   void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 772 */
-   void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 773 */
-   void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 774 */
+   void (GLAPIENTRYP FlushMappedBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length); /* 562 */
+   GLvoid * (GLAPIENTRYP MapBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access); /* 563 */
+   void (GLAPIENTRYP BindVertexArray)(GLuint array); /* 564 */
+   void (GLAPIENTRYP GenVertexArrays)(GLsizei n, GLuint * arrays); /* 565 */
+   void (GLAPIENTRYP CopyBufferSubData)(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size); /* 566 */
+   void (GLAPIENTRYP PolygonOffsetEXT)(GLfloat factor, GLfloat bias); /* 567 */
+   void (GLAPIENTRYP GetPixelTexGenParameterfvSGIS)(GLenum pname, GLfloat * params); /* 568 */
+   void (GLAPIENTRYP GetPixelTexGenParameterivSGIS)(GLenum pname, GLint * params); /* 569 */
+   void (GLAPIENTRYP PixelTexGenParameterfSGIS)(GLenum pname, GLfloat param); /* 570 */
+   void (GLAPIENTRYP PixelTexGenParameterfvSGIS)(GLenum pname, const GLfloat * params); /* 571 */
+   void (GLAPIENTRYP PixelTexGenParameteriSGIS)(GLenum pname, GLint param); /* 572 */
+   void (GLAPIENTRYP PixelTexGenParameterivSGIS)(GLenum pname, const GLint * params); /* 573 */
+   void (GLAPIENTRYP SampleMaskSGIS)(GLclampf value, GLboolean invert); /* 574 */
+   void (GLAPIENTRYP SamplePatternSGIS)(GLenum pattern); /* 575 */
+   void (GLAPIENTRYP ColorPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 576 */
+   void (GLAPIENTRYP EdgeFlagPointerEXT)(GLsizei stride, GLsizei count, const GLboolean * pointer); /* 577 */
+   void (GLAPIENTRYP IndexPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 578 */
+   void (GLAPIENTRYP NormalPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 579 */
+   void (GLAPIENTRYP TexCoordPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 580 */
+   void (GLAPIENTRYP VertexPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 581 */
+   void (GLAPIENTRYP PointParameterfEXT)(GLenum pname, GLfloat param); /* 582 */
+   void (GLAPIENTRYP PointParameterfvEXT)(GLenum pname, const GLfloat * params); /* 583 */
+   void (GLAPIENTRYP LockArraysEXT)(GLint first, GLsizei count); /* 584 */
+   void (GLAPIENTRYP UnlockArraysEXT)(void); /* 585 */
+   void (GLAPIENTRYP CullParameterdvEXT)(GLenum pname, GLdouble * params); /* 586 */
+   void (GLAPIENTRYP CullParameterfvEXT)(GLenum pname, GLfloat * params); /* 587 */
+   void (GLAPIENTRYP SecondaryColor3bEXT)(GLbyte red, GLbyte green, GLbyte blue); /* 588 */
+   void (GLAPIENTRYP SecondaryColor3bvEXT)(const GLbyte * v); /* 589 */
+   void (GLAPIENTRYP SecondaryColor3dEXT)(GLdouble red, GLdouble green, GLdouble blue); /* 590 */
+   void (GLAPIENTRYP SecondaryColor3dvEXT)(const GLdouble * v); /* 591 */
+   void (GLAPIENTRYP SecondaryColor3fEXT)(GLfloat red, GLfloat green, GLfloat blue); /* 592 */
+   void (GLAPIENTRYP SecondaryColor3fvEXT)(const GLfloat * v); /* 593 */
+   void (GLAPIENTRYP SecondaryColor3iEXT)(GLint red, GLint green, GLint blue); /* 594 */
+   void (GLAPIENTRYP SecondaryColor3ivEXT)(const GLint * v); /* 595 */
+   void (GLAPIENTRYP SecondaryColor3sEXT)(GLshort red, GLshort green, GLshort blue); /* 596 */
+   void (GLAPIENTRYP SecondaryColor3svEXT)(const GLshort * v); /* 597 */
+   void (GLAPIENTRYP SecondaryColor3ubEXT)(GLubyte red, GLubyte green, GLubyte blue); /* 598 */
+   void (GLAPIENTRYP SecondaryColor3ubvEXT)(const GLubyte * v); /* 599 */
+   void (GLAPIENTRYP SecondaryColor3uiEXT)(GLuint red, GLuint green, GLuint blue); /* 600 */
+   void (GLAPIENTRYP SecondaryColor3uivEXT)(const GLuint * v); /* 601 */
+   void (GLAPIENTRYP SecondaryColor3usEXT)(GLushort red, GLushort green, GLushort blue); /* 602 */
+   void (GLAPIENTRYP SecondaryColor3usvEXT)(const GLushort * v); /* 603 */
+   void (GLAPIENTRYP SecondaryColorPointerEXT)(GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 604 */
+   void (GLAPIENTRYP MultiDrawArraysEXT)(GLenum mode, GLint * first, GLsizei * count, GLsizei primcount); /* 605 */
+   void (GLAPIENTRYP MultiDrawElementsEXT)(GLenum mode, const GLsizei * count, GLenum type, const GLvoid ** indices, GLsizei primcount); /* 606 */
+   void (GLAPIENTRYP FogCoordPointerEXT)(GLenum type, GLsizei stride, const GLvoid * pointer); /* 607 */
+   void (GLAPIENTRYP FogCoorddEXT)(GLdouble coord); /* 608 */
+   void (GLAPIENTRYP FogCoorddvEXT)(const GLdouble * coord); /* 609 */
+   void (GLAPIENTRYP FogCoordfEXT)(GLfloat coord); /* 610 */
+   void (GLAPIENTRYP FogCoordfvEXT)(const GLfloat * coord); /* 611 */
+   void (GLAPIENTRYP PixelTexGenSGIX)(GLenum mode); /* 612 */
+   void (GLAPIENTRYP BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha); /* 613 */
+   void (GLAPIENTRYP FlushVertexArrayRangeNV)(void); /* 614 */
+   void (GLAPIENTRYP VertexArrayRangeNV)(GLsizei length, const GLvoid * pointer); /* 615 */
+   void (GLAPIENTRYP CombinerInputNV)(GLenum stage, GLenum portion, GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 616 */
+   void (GLAPIENTRYP CombinerOutputNV)(GLenum stage, GLenum portion, GLenum abOutput, GLenum cdOutput, GLenum sumOutput, GLenum scale, GLenum bias, GLboolean abDotProduct, GLboolean cdDotProduct, GLboolean muxSum); /* 617 */
+   void (GLAPIENTRYP CombinerParameterfNV)(GLenum pname, GLfloat param); /* 618 */
+   void (GLAPIENTRYP CombinerParameterfvNV)(GLenum pname, const GLfloat * params); /* 619 */
+   void (GLAPIENTRYP CombinerParameteriNV)(GLenum pname, GLint param); /* 620 */
+   void (GLAPIENTRYP CombinerParameterivNV)(GLenum pname, const GLint * params); /* 621 */
+   void (GLAPIENTRYP FinalCombinerInputNV)(GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 622 */
+   void (GLAPIENTRYP GetCombinerInputParameterfvNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLfloat * params); /* 623 */
+   void (GLAPIENTRYP GetCombinerInputParameterivNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLint * params); /* 624 */
+   void (GLAPIENTRYP GetCombinerOutputParameterfvNV)(GLenum stage, GLenum portion, GLenum pname, GLfloat * params); /* 625 */
+   void (GLAPIENTRYP GetCombinerOutputParameterivNV)(GLenum stage, GLenum portion, GLenum pname, GLint * params); /* 626 */
+   void (GLAPIENTRYP GetFinalCombinerInputParameterfvNV)(GLenum variable, GLenum pname, GLfloat * params); /* 627 */
+   void (GLAPIENTRYP GetFinalCombinerInputParameterivNV)(GLenum variable, GLenum pname, GLint * params); /* 628 */
+   void (GLAPIENTRYP ResizeBuffersMESA)(void); /* 629 */
+   void (GLAPIENTRYP WindowPos2dMESA)(GLdouble x, GLdouble y); /* 630 */
+   void (GLAPIENTRYP WindowPos2dvMESA)(const GLdouble * v); /* 631 */
+   void (GLAPIENTRYP WindowPos2fMESA)(GLfloat x, GLfloat y); /* 632 */
+   void (GLAPIENTRYP WindowPos2fvMESA)(const GLfloat * v); /* 633 */
+   void (GLAPIENTRYP WindowPos2iMESA)(GLint x, GLint y); /* 634 */
+   void (GLAPIENTRYP WindowPos2ivMESA)(const GLint * v); /* 635 */
+   void (GLAPIENTRYP WindowPos2sMESA)(GLshort x, GLshort y); /* 636 */
+   void (GLAPIENTRYP WindowPos2svMESA)(const GLshort * v); /* 637 */
+   void (GLAPIENTRYP WindowPos3dMESA)(GLdouble x, GLdouble y, GLdouble z); /* 638 */
+   void (GLAPIENTRYP WindowPos3dvMESA)(const GLdouble * v); /* 639 */
+   void (GLAPIENTRYP WindowPos3fMESA)(GLfloat x, GLfloat y, GLfloat z); /* 640 */
+   void (GLAPIENTRYP WindowPos3fvMESA)(const GLfloat * v); /* 641 */
+   void (GLAPIENTRYP WindowPos3iMESA)(GLint x, GLint y, GLint z); /* 642 */
+   void (GLAPIENTRYP WindowPos3ivMESA)(const GLint * v); /* 643 */
+   void (GLAPIENTRYP WindowPos3sMESA)(GLshort x, GLshort y, GLshort z); /* 644 */
+   void (GLAPIENTRYP WindowPos3svMESA)(const GLshort * v); /* 645 */
+   void (GLAPIENTRYP WindowPos4dMESA)(GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 646 */
+   void (GLAPIENTRYP WindowPos4dvMESA)(const GLdouble * v); /* 647 */
+   void (GLAPIENTRYP WindowPos4fMESA)(GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 648 */
+   void (GLAPIENTRYP WindowPos4fvMESA)(const GLfloat * v); /* 649 */
+   void (GLAPIENTRYP WindowPos4iMESA)(GLint x, GLint y, GLint z, GLint w); /* 650 */
+   void (GLAPIENTRYP WindowPos4ivMESA)(const GLint * v); /* 651 */
+   void (GLAPIENTRYP WindowPos4sMESA)(GLshort x, GLshort y, GLshort z, GLshort w); /* 652 */
+   void (GLAPIENTRYP WindowPos4svMESA)(const GLshort * v); /* 653 */
+   void (GLAPIENTRYP MultiModeDrawArraysIBM)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); /* 654 */
+   void (GLAPIENTRYP MultiModeDrawElementsIBM)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); /* 655 */
+   void (GLAPIENTRYP DeleteFencesNV)(GLsizei n, const GLuint * fences); /* 656 */
+   void (GLAPIENTRYP FinishFenceNV)(GLuint fence); /* 657 */
+   void (GLAPIENTRYP GenFencesNV)(GLsizei n, GLuint * fences); /* 658 */
+   void (GLAPIENTRYP GetFenceivNV)(GLuint fence, GLenum pname, GLint * params); /* 659 */
+   GLboolean (GLAPIENTRYP IsFenceNV)(GLuint fence); /* 660 */
+   void (GLAPIENTRYP SetFenceNV)(GLuint fence, GLenum condition); /* 661 */
+   GLboolean (GLAPIENTRYP TestFenceNV)(GLuint fence); /* 662 */
+   GLboolean (GLAPIENTRYP AreProgramsResidentNV)(GLsizei n, const GLuint * ids, GLboolean * residences); /* 663 */
+   void (GLAPIENTRYP BindProgramNV)(GLenum target, GLuint program); /* 664 */
+   void (GLAPIENTRYP DeleteProgramsNV)(GLsizei n, const GLuint * programs); /* 665 */
+   void (GLAPIENTRYP ExecuteProgramNV)(GLenum target, GLuint id, const GLfloat * params); /* 666 */
+   void (GLAPIENTRYP GenProgramsNV)(GLsizei n, GLuint * programs); /* 667 */
+   void (GLAPIENTRYP GetProgramParameterdvNV)(GLenum target, GLuint index, GLenum pname, GLdouble * params); /* 668 */
+   void (GLAPIENTRYP GetProgramParameterfvNV)(GLenum target, GLuint index, GLenum pname, GLfloat * params); /* 669 */
+   void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 670 */
+   void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 671 */
+   void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 672 */
+   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 673 */
+   void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 674 */
+   void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 675 */
+   void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 676 */
+   GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 677 */
+   void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 678 */
+   void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 679 */
+   void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 680 */
+   void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 681 */
+   void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 682 */
+   void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 683 */
+   void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 684 */
+   void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 685 */
+   void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 686 */
+   void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 687 */
+   void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 688 */
+   void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 689 */
+   void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 690 */
+   void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 691 */
+   void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 692 */
+   void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 693 */
+   void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 694 */
+   void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 695 */
+   void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 696 */
+   void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 697 */
+   void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 698 */
+   void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 699 */
+   void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 700 */
+   void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 701 */
+   void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 702 */
+   void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 703 */
+   void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 704 */
+   void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 705 */
+   void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 706 */
+   void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 707 */
+   void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 708 */
+   void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 709 */
+   void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 710 */
+   void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 711 */
+   void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 712 */
+   void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 713 */
+   void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 714 */
+   void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 715 */
+   void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 716 */
+   void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 717 */
+   void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 718 */
+   void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 719 */
+   void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 720 */
+   void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 721 */
+   void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 722 */
+   void (GLAPIENTRYP GetTexBumpParameterfvATI)(GLenum pname, GLfloat * param); /* 723 */
+   void (GLAPIENTRYP GetTexBumpParameterivATI)(GLenum pname, GLint * param); /* 724 */
+   void (GLAPIENTRYP TexBumpParameterfvATI)(GLenum pname, const GLfloat * param); /* 725 */
+   void (GLAPIENTRYP TexBumpParameterivATI)(GLenum pname, const GLint * param); /* 726 */
+   void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 727 */
+   void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 728 */
+   void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 729 */
+   void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 730 */
+   void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 731 */
+   void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 732 */
+   void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 733 */
+   void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 734 */
+   void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 735 */
+   void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 736 */
+   GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 737 */
+   void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 738 */
+   void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 739 */
+   void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 740 */
+   void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 741 */
+   void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 742 */
+   void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 743 */
+   void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 744 */
+   void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 745 */
+   void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 746 */
+   GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 747 */
+   void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 748 */
+   void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 749 */
+   void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 750 */
+   void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 751 */
+   void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 752 */
+   void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 753 */
+   void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 754 */
+   void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 755 */
+   void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 756 */
+   void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 757 */
+   GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 758 */
+   void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 759 */
+   void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 760 */
+   void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 761 */
+   void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 762 */
+   void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 763 */
+   void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 764 */
+   void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 765 */
+   void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 766 */
+   void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 767 */
+   void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 768 */
+   void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 769 */
+   GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 770 */
+   GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 771 */
+   void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 772 */
+   void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 773 */
+   void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 774 */
+   void (GLAPIENTRYP ProvokingVertexEXT)(GLenum mode); /* 775 */
+   void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 776 */
+   void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 777 */
+   void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 778 */
+   void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 779 */
+   void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 780 */
 };
 
 #endif /* !defined( _GLAPI_TABLE_H_ ) */
diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h
index 7ccd9707c3..86fda613f0 100644
--- a/src/mesa/glapi/glapitemp.h
+++ b/src/mesa/glapi/glapitemp.h
@@ -4011,63 +4011,88 @@ KEYWORD1 void KEYWORD2 NAME(RenderbufferStorageMultisample)(GLenum target, GLsiz
    DISPATCH(RenderbufferStorageMultisample, (target, samples, internalformat, width, height), (F, "glRenderbufferStorageMultisample(0x%x, %d, 0x%x, %d, %d);\n", target, samples, internalformat, width, height));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FlushMappedBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length)
+{
+   DISPATCH(FlushMappedBufferRange, (target, offset, length), (F, "glFlushMappedBufferRange(0x%x, %d, %d);\n", target, offset, length));
+}
+
+KEYWORD1 GLvoid * KEYWORD2 NAME(MapBufferRange)(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access)
+{
+   RETURN_DISPATCH(MapBufferRange, (target, offset, length, access), (F, "glMapBufferRange(0x%x, %d, %d, %d);\n", target, offset, length, access));
+}
+
+KEYWORD1 void KEYWORD2 NAME(BindVertexArray)(GLuint array)
+{
+   DISPATCH(BindVertexArray, (array), (F, "glBindVertexArray(%d);\n", array));
+}
+
+KEYWORD1 void KEYWORD2 NAME(GenVertexArrays)(GLsizei n, GLuint * arrays)
+{
+   DISPATCH(GenVertexArrays, (n, arrays), (F, "glGenVertexArrays(%d, %p);\n", n, (const void *) arrays));
+}
+
+KEYWORD1 void KEYWORD2 NAME(CopyBufferSubData)(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size)
+{
+   DISPATCH(CopyBufferSubData, (readTarget, writeTarget, readOffset, writeOffset, size), (F, "glCopyBufferSubData(0x%x, 0x%x, %d, %d, %d);\n", readTarget, writeTarget, readOffset, writeOffset, size));
+}
+
 KEYWORD1 void KEYWORD2 NAME(PolygonOffsetEXT)(GLfloat factor, GLfloat bias)
 {
    DISPATCH(PolygonOffsetEXT, (factor, bias), (F, "glPolygonOffsetEXT(%f, %f);\n", factor, bias));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, GLfloat * params)
 {
    DISPATCH(GetPixelTexGenParameterfvSGIS, (pname, params), (F, "glGetPixelTexGenParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLenum pname, GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLenum pname, GLint * params)
 {
    DISPATCH(GetPixelTexGenParameterivSGIS, (pname, params), (F, "glGetPixelTexGenParameterivSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, GLfloat param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pname, GLfloat param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, GLfloat param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pname, GLfloat param)
 {
    DISPATCH(PixelTexGenParameterfSGIS, (pname, param), (F, "glPixelTexGenParameterfSGIS(0x%x, %f);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_571)(GLenum pname, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_571)(GLenum pname, const GLfloat * params)
 {
    DISPATCH(PixelTexGenParameterfvSGIS, (pname, params), (F, "glPixelTexGenParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, GLint param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_572)(GLenum pname, GLint param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, GLint param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_572)(GLenum pname, GLint param)
 {
    DISPATCH(PixelTexGenParameteriSGIS, (pname, param), (F, "glPixelTexGenParameteriSGIS(0x%x, %d);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, const GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_573)(GLenum pname, const GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, const GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_573)(GLenum pname, const GLint * params)
 {
    DISPATCH(PixelTexGenParameterivSGIS, (pname, params), (F, "glPixelTexGenParameterivSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLclampf value, GLboolean invert);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_574)(GLclampf value, GLboolean invert);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLclampf value, GLboolean invert)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_574)(GLclampf value, GLboolean invert)
 {
    DISPATCH(SampleMaskSGIS, (value, invert), (F, "glSampleMaskSGIS(%f, %d);\n", value, invert));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pattern);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_575)(GLenum pattern);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pattern)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_575)(GLenum pattern)
 {
    DISPATCH(SamplePatternSGIS, (pattern), (F, "glSamplePatternSGIS(0x%x);\n", pattern));
 }
@@ -4117,9 +4142,9 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterfEXT)(GLenum pname, GLfloat param)
    DISPATCH(PointParameterfEXT, (pname, param), (F, "glPointParameterfEXT(0x%x, %f);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat param)
 {
    DISPATCH(PointParameterfEXT, (pname, param), (F, "glPointParameterfSGIS(0x%x, %f);\n", pname, param));
 }
@@ -4139,9 +4164,9 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterfvEXT)(GLenum pname, const GLfloat * p
    DISPATCH(PointParameterfvEXT, (pname, params), (F, "glPointParameterfvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_583)(GLenum pname, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_583)(GLenum pname, const GLfloat * params)
 {
    DISPATCH(PointParameterfvEXT, (pname, params), (F, "glPointParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
@@ -4156,16 +4181,16 @@ KEYWORD1 void KEYWORD2 NAME(UnlockArraysEXT)(void)
    DISPATCH(UnlockArraysEXT, (), (F, "glUnlockArraysEXT();\n"));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLdouble * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_586)(GLenum pname, GLdouble * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLdouble * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_586)(GLenum pname, GLdouble * params)
 {
    DISPATCH(CullParameterdvEXT, (pname, params), (F, "glCullParameterdvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_587)(GLenum pname, GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_587)(GLenum pname, GLfloat * params)
 {
    DISPATCH(CullParameterfvEXT, (pname, params), (F, "glCullParameterfvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
@@ -4410,9 +4435,9 @@ KEYWORD1 void KEYWORD2 NAME(FogCoordfvEXT)(const GLfloat * coord)
    DISPATCH(FogCoordfvEXT, (coord), (F, "glFogCoordfvEXT(%p);\n", (const void *) coord));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum mode);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_612)(GLenum mode);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum mode)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_612)(GLenum mode)
 {
    DISPATCH(PixelTexGenSGIX, (mode), (F, "glPixelTexGenSGIX(0x%x);\n", mode));
 }
@@ -4427,9 +4452,9 @@ KEYWORD1 void KEYWORD2 NAME(BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfac
    DISPATCH(BlendFuncSeparateEXT, (sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha), (F, "glBlendFuncSeparateEXT(0x%x, 0x%x, 0x%x, 0x%x);\n", sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_608)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_613)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_608)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_613)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)
 {
    DISPATCH(BlendFuncSeparateEXT, (sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha), (F, "glBlendFuncSeparateINGR(0x%x, 0x%x, 0x%x, 0x%x);\n", sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha));
 }
@@ -4794,65 +4819,65 @@ KEYWORD1 void KEYWORD2 NAME(WindowPos4svMESA)(const GLshort * v)
    DISPATCH(WindowPos4svMESA, (v), (F, "glWindowPos4svMESA(%p);\n", (const void *) v));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride)
 {
    DISPATCH(MultiModeDrawArraysIBM, (mode, first, count, primcount, modestride), (F, "glMultiModeDrawArraysIBM(%p, %p, %p, %d, %d);\n", (const void *) mode, (const void *) first, (const void *) count, primcount, modestride));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_655)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_655)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride)
 {
    DISPATCH(MultiModeDrawElementsIBM, (mode, count, type, indices, primcount, modestride), (F, "glMultiModeDrawElementsIBM(%p, %p, 0x%x, %p, %d, %d);\n", (const void *) mode, (const void *) count, type, (const void *) indices, primcount, modestride));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLsizei n, const GLuint * fences);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLsizei n, const GLuint * fences);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLsizei n, const GLuint * fences)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLsizei n, const GLuint * fences)
 {
    DISPATCH(DeleteFencesNV, (n, fences), (F, "glDeleteFencesNV(%d, %p);\n", n, (const void *) fences));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLuint fence);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLuint fence)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence)
 {
    DISPATCH(FinishFenceNV, (fence), (F, "glFinishFenceNV(%d);\n", fence));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLsizei n, GLuint * fences);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_658)(GLsizei n, GLuint * fences);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLsizei n, GLuint * fences)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_658)(GLsizei n, GLuint * fences)
 {
    DISPATCH(GenFencesNV, (n, fences), (F, "glGenFencesNV(%d, %p);\n", n, (const void *) fences));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence, GLenum pname, GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_659)(GLuint fence, GLenum pname, GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence, GLenum pname, GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_659)(GLuint fence, GLenum pname, GLint * params)
 {
    DISPATCH(GetFenceivNV, (fence, pname, params), (F, "glGetFenceivNV(%d, 0x%x, %p);\n", fence, pname, (const void *) params));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_660)(GLuint fence);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_660)(GLuint fence)
 {
    RETURN_DISPATCH(IsFenceNV, (fence), (F, "glIsFenceNV(%d);\n", fence));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence, GLenum condition);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_661)(GLuint fence, GLenum condition);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence, GLenum condition)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_661)(GLuint fence, GLenum condition)
 {
    DISPATCH(SetFenceNV, (fence, condition), (F, "glSetFenceNV(%d, 0x%x);\n", fence, condition));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_662)(GLuint fence);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_662)(GLuint fence)
 {
    RETURN_DISPATCH(TestFenceNV, (fence), (F, "glTestFenceNV(%d);\n", fence));
 }
@@ -5297,37 +5322,47 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterivNV)(GLenum pname, const GLint * para
    DISPATCH(PointParameterivNV, (pname, params), (F, "glPointParameterivNV(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_738)(GLenum face);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_743)(GLenum face);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_738)(GLenum face)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_743)(GLenum face)
 {
    DISPATCH(ActiveStencilFaceEXT, (face), (F, "glActiveStencilFaceEXT(0x%x);\n", face));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_739)(GLuint array);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_744)(GLuint array);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_739)(GLuint array)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_744)(GLuint array)
 {
    DISPATCH(BindVertexArrayAPPLE, (array), (F, "glBindVertexArrayAPPLE(%d);\n", array));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_740)(GLsizei n, const GLuint * arrays);
+KEYWORD1 void KEYWORD2 NAME(DeleteVertexArrays)(GLsizei n, const GLuint * arrays)
+{
+   DISPATCH(DeleteVertexArraysAPPLE, (n, arrays), (F, "glDeleteVertexArrays(%d, %p);\n", n, (const void *) arrays));
+}
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_740)(GLsizei n, const GLuint * arrays)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLsizei n, const GLuint * arrays);
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLsizei n, const GLuint * arrays)
 {
    DISPATCH(DeleteVertexArraysAPPLE, (n, arrays), (F, "glDeleteVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_741)(GLsizei n, GLuint * arrays);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_746)(GLsizei n, GLuint * arrays);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_741)(GLsizei n, GLuint * arrays)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_746)(GLsizei n, GLuint * arrays)
 {
    DISPATCH(GenVertexArraysAPPLE, (n, arrays), (F, "glGenVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_742)(GLuint array);
+KEYWORD1 GLboolean KEYWORD2 NAME(IsVertexArray)(GLuint array)
+{
+   RETURN_DISPATCH(IsVertexArrayAPPLE, (array), (F, "glIsVertexArray(%d);\n", array));
+}
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_742)(GLuint array)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_747)(GLuint array);
+
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_747)(GLuint array)
 {
    RETURN_DISPATCH(IsVertexArrayAPPLE, (array), (F, "glIsVertexArrayAPPLE(%d);\n", array));
 }
@@ -5362,9 +5397,9 @@ KEYWORD1 void KEYWORD2 NAME(ProgramNamedParameter4fvNV)(GLuint id, GLsizei len,
    DISPATCH(ProgramNamedParameter4fvNV, (id, len, name, v), (F, "glProgramNamedParameter4fvNV(%d, %d, %p, %p);\n", id, len, (const void *) name, (const void *) v));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_749)(GLclampd zmin, GLclampd zmax);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_754)(GLclampd zmin, GLclampd zmax);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_749)(GLclampd zmin, GLclampd zmax)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_754)(GLclampd zmin, GLclampd zmax)
 {
    DISPATCH(DepthBoundsEXT, (zmin, zmax), (F, "glDepthBoundsEXT(%f, %f);\n", zmin, zmax));
 }
@@ -5374,9 +5409,9 @@ KEYWORD1 void KEYWORD2 NAME(BlendEquationSeparate)(GLenum modeRGB, GLenum modeA)
    DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparate(0x%x, 0x%x);\n", modeRGB, modeA));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_750)(GLenum modeRGB, GLenum modeA);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_755)(GLenum modeRGB, GLenum modeA);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_750)(GLenum modeRGB, GLenum modeA)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_755)(GLenum modeRGB, GLenum modeA)
 {
    DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparateEXT(0x%x, 0x%x);\n", modeRGB, modeA));
 }
@@ -5556,9 +5591,9 @@ KEYWORD1 void KEYWORD2 NAME(BlitFramebuffer)(GLint srcX0, GLint srcY0, GLint src
    DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebuffer(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
 {
    DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebufferEXT(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter));
 }
@@ -5573,37 +5608,42 @@ KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayerEXT)(GLenum target, GLenum at
    DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayerEXT(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
+KEYWORD1 void KEYWORD2 NAME(ProvokingVertexEXT)(GLenum mode)
+{
+   DISPATCH(ProvokingVertexEXT, (mode), (F, "glProvokingVertexEXT(0x%x);\n", mode));
+}
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_776)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_776)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask)
 {
    DISPATCH(StencilFuncSeparateATI, (frontfunc, backfunc, ref, mask), (F, "glStencilFuncSeparateATI(0x%x, 0x%x, %d, %d);\n", frontfunc, backfunc, ref, mask));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_777)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_777)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
 {
    DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_778)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_778)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
 {
    DISPATCH(ProgramLocalParameters4fvEXT, (target, index, count, params), (F, "glProgramLocalParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_779)(GLuint id, GLenum pname, GLint64EXT * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_779)(GLuint id, GLenum pname, GLint64EXT * params)
 {
    DISPATCH(GetQueryObjecti64vEXT, (id, pname, params), (F, "glGetQueryObjecti64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_780)(GLuint id, GLenum pname, GLuint64EXT * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_780)(GLuint id, GLenum pname, GLuint64EXT * params)
 {
    DISPATCH(GetQueryObjectui64vEXT, (id, pname, params), (F, "glGetQueryObjectui64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params));
 }
@@ -6184,15 +6224,20 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(GetAttribLocationARB),
    TABLE_ENTRY(DrawBuffersARB),
    TABLE_ENTRY(RenderbufferStorageMultisample),
+   TABLE_ENTRY(FlushMappedBufferRange),
+   TABLE_ENTRY(MapBufferRange),
+   TABLE_ENTRY(BindVertexArray),
+   TABLE_ENTRY(GenVertexArrays),
+   TABLE_ENTRY(CopyBufferSubData),
    TABLE_ENTRY(PolygonOffsetEXT),
-   TABLE_ENTRY(_dispatch_stub_563),
-   TABLE_ENTRY(_dispatch_stub_564),
-   TABLE_ENTRY(_dispatch_stub_565),
-   TABLE_ENTRY(_dispatch_stub_566),
-   TABLE_ENTRY(_dispatch_stub_567),
    TABLE_ENTRY(_dispatch_stub_568),
    TABLE_ENTRY(_dispatch_stub_569),
    TABLE_ENTRY(_dispatch_stub_570),
+   TABLE_ENTRY(_dispatch_stub_571),
+   TABLE_ENTRY(_dispatch_stub_572),
+   TABLE_ENTRY(_dispatch_stub_573),
+   TABLE_ENTRY(_dispatch_stub_574),
+   TABLE_ENTRY(_dispatch_stub_575),
    TABLE_ENTRY(ColorPointerEXT),
    TABLE_ENTRY(EdgeFlagPointerEXT),
    TABLE_ENTRY(IndexPointerEXT),
@@ -6203,8 +6248,8 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(PointParameterfvEXT),
    TABLE_ENTRY(LockArraysEXT),
    TABLE_ENTRY(UnlockArraysEXT),
-   TABLE_ENTRY(_dispatch_stub_581),
-   TABLE_ENTRY(_dispatch_stub_582),
+   TABLE_ENTRY(_dispatch_stub_586),
+   TABLE_ENTRY(_dispatch_stub_587),
    TABLE_ENTRY(SecondaryColor3bEXT),
    TABLE_ENTRY(SecondaryColor3bvEXT),
    TABLE_ENTRY(SecondaryColor3dEXT),
@@ -6229,7 +6274,7 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(FogCoorddvEXT),
    TABLE_ENTRY(FogCoordfEXT),
    TABLE_ENTRY(FogCoordfvEXT),
-   TABLE_ENTRY(_dispatch_stub_607),
+   TABLE_ENTRY(_dispatch_stub_612),
    TABLE_ENTRY(BlendFuncSeparateEXT),
    TABLE_ENTRY(FlushVertexArrayRangeNV),
    TABLE_ENTRY(VertexArrayRangeNV),
@@ -6271,15 +6316,15 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(WindowPos4ivMESA),
    TABLE_ENTRY(WindowPos4sMESA),
    TABLE_ENTRY(WindowPos4svMESA),
-   TABLE_ENTRY(_dispatch_stub_649),
-   TABLE_ENTRY(_dispatch_stub_650),
-   TABLE_ENTRY(_dispatch_stub_651),
-   TABLE_ENTRY(_dispatch_stub_652),
-   TABLE_ENTRY(_dispatch_stub_653),
    TABLE_ENTRY(_dispatch_stub_654),
    TABLE_ENTRY(_dispatch_stub_655),
    TABLE_ENTRY(_dispatch_stub_656),
    TABLE_ENTRY(_dispatch_stub_657),
+   TABLE_ENTRY(_dispatch_stub_658),
+   TABLE_ENTRY(_dispatch_stub_659),
+   TABLE_ENTRY(_dispatch_stub_660),
+   TABLE_ENTRY(_dispatch_stub_661),
+   TABLE_ENTRY(_dispatch_stub_662),
    TABLE_ENTRY(AreProgramsResidentNV),
    TABLE_ENTRY(BindProgramNV),
    TABLE_ENTRY(DeleteProgramsNV),
@@ -6360,19 +6405,19 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(SetFragmentShaderConstantATI),
    TABLE_ENTRY(PointParameteriNV),
    TABLE_ENTRY(PointParameterivNV),
-   TABLE_ENTRY(_dispatch_stub_738),
-   TABLE_ENTRY(_dispatch_stub_739),
-   TABLE_ENTRY(_dispatch_stub_740),
-   TABLE_ENTRY(_dispatch_stub_741),
-   TABLE_ENTRY(_dispatch_stub_742),
+   TABLE_ENTRY(_dispatch_stub_743),
+   TABLE_ENTRY(_dispatch_stub_744),
+   TABLE_ENTRY(_dispatch_stub_745),
+   TABLE_ENTRY(_dispatch_stub_746),
+   TABLE_ENTRY(_dispatch_stub_747),
    TABLE_ENTRY(GetProgramNamedParameterdvNV),
    TABLE_ENTRY(GetProgramNamedParameterfvNV),
    TABLE_ENTRY(ProgramNamedParameter4dNV),
    TABLE_ENTRY(ProgramNamedParameter4dvNV),
    TABLE_ENTRY(ProgramNamedParameter4fNV),
    TABLE_ENTRY(ProgramNamedParameter4fvNV),
-   TABLE_ENTRY(_dispatch_stub_749),
-   TABLE_ENTRY(_dispatch_stub_750),
+   TABLE_ENTRY(_dispatch_stub_754),
+   TABLE_ENTRY(_dispatch_stub_755),
    TABLE_ENTRY(BindFramebufferEXT),
    TABLE_ENTRY(BindRenderbufferEXT),
    TABLE_ENTRY(CheckFramebufferStatusEXT),
@@ -6390,13 +6435,14 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(IsFramebufferEXT),
    TABLE_ENTRY(IsRenderbufferEXT),
    TABLE_ENTRY(RenderbufferStorageEXT),
-   TABLE_ENTRY(_dispatch_stub_768),
-   TABLE_ENTRY(FramebufferTextureLayerEXT),
-   TABLE_ENTRY(_dispatch_stub_770),
-   TABLE_ENTRY(_dispatch_stub_771),
-   TABLE_ENTRY(_dispatch_stub_772),
    TABLE_ENTRY(_dispatch_stub_773),
-   TABLE_ENTRY(_dispatch_stub_774),
+   TABLE_ENTRY(FramebufferTextureLayerEXT),
+   TABLE_ENTRY(ProvokingVertexEXT),
+   TABLE_ENTRY(_dispatch_stub_776),
+   TABLE_ENTRY(_dispatch_stub_777),
+   TABLE_ENTRY(_dispatch_stub_778),
+   TABLE_ENTRY(_dispatch_stub_779),
+   TABLE_ENTRY(_dispatch_stub_780),
    /* A whole bunch of no-op functions.  These might be called
     * when someone tries to call a dynamically-registered
     * extension function without a current rendering context.
@@ -6763,6 +6809,8 @@ static _glapi_proc UNUSED_TABLE_NAME[] = {
    TABLE_ENTRY(IsProgramARB),
    TABLE_ENTRY(PointParameteri),
    TABLE_ENTRY(PointParameteriv),
+   TABLE_ENTRY(DeleteVertexArrays),
+   TABLE_ENTRY(IsVertexArray),
    TABLE_ENTRY(BlendEquationSeparate),
    TABLE_ENTRY(BindFramebuffer),
    TABLE_ENTRY(BindRenderbuffer),
diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h
index 680893cfc4..bd3ebc8872 100644
--- a/src/mesa/glapi/glprocs.h
+++ b/src/mesa/glapi/glprocs.h
@@ -614,6 +614,11 @@ static const char gl_string_table[] =
     "glGetAttribLocationARB\0"
     "glDrawBuffersARB\0"
     "glRenderbufferStorageMultisample\0"
+    "glFlushMappedBufferRange\0"
+    "glMapBufferRange\0"
+    "glBindVertexArray\0"
+    "glGenVertexArrays\0"
+    "glCopyBufferSubData\0"
     "glPolygonOffsetEXT\0"
     "glGetPixelTexGenParameterfvSGIS\0"
     "glGetPixelTexGenParameterivSGIS\0"
@@ -822,6 +827,7 @@ static const char gl_string_table[] =
     "glRenderbufferStorageEXT\0"
     "glBlitFramebufferEXT\0"
     "glFramebufferTextureLayerEXT\0"
+    "glProvokingVertexEXT\0"
     "glStencilFuncSeparateATI\0"
     "glProgramEnvParameters4fvEXT\0"
     "glProgramLocalParameters4fvEXT\0"
@@ -1104,6 +1110,8 @@ static const char gl_string_table[] =
     "glIsProgramARB\0"
     "glPointParameteri\0"
     "glPointParameteriv\0"
+    "glDeleteVertexArrays\0"
+    "glIsVertexArray\0"
     "glBlendEquationSeparate\0"
     "glBlendEquationSeparateATI\0"
     "glBindFramebuffer\0"
@@ -1142,39 +1150,39 @@ static const char gl_string_table[] =
 #define gl_dispatch_stub_364 mgl_dispatch_stub_364
 #define gl_dispatch_stub_365 mgl_dispatch_stub_365
 #define gl_dispatch_stub_366 mgl_dispatch_stub_366
-#define gl_dispatch_stub_563 mgl_dispatch_stub_563
-#define gl_dispatch_stub_564 mgl_dispatch_stub_564
-#define gl_dispatch_stub_565 mgl_dispatch_stub_565
-#define gl_dispatch_stub_566 mgl_dispatch_stub_566
-#define gl_dispatch_stub_567 mgl_dispatch_stub_567
 #define gl_dispatch_stub_568 mgl_dispatch_stub_568
 #define gl_dispatch_stub_569 mgl_dispatch_stub_569
 #define gl_dispatch_stub_570 mgl_dispatch_stub_570
-#define gl_dispatch_stub_581 mgl_dispatch_stub_581
-#define gl_dispatch_stub_582 mgl_dispatch_stub_582
-#define gl_dispatch_stub_607 mgl_dispatch_stub_607
-#define gl_dispatch_stub_649 mgl_dispatch_stub_649
-#define gl_dispatch_stub_650 mgl_dispatch_stub_650
-#define gl_dispatch_stub_651 mgl_dispatch_stub_651
-#define gl_dispatch_stub_652 mgl_dispatch_stub_652
-#define gl_dispatch_stub_653 mgl_dispatch_stub_653
+#define gl_dispatch_stub_571 mgl_dispatch_stub_571
+#define gl_dispatch_stub_572 mgl_dispatch_stub_572
+#define gl_dispatch_stub_573 mgl_dispatch_stub_573
+#define gl_dispatch_stub_574 mgl_dispatch_stub_574
+#define gl_dispatch_stub_575 mgl_dispatch_stub_575
+#define gl_dispatch_stub_586 mgl_dispatch_stub_586
+#define gl_dispatch_stub_587 mgl_dispatch_stub_587
+#define gl_dispatch_stub_612 mgl_dispatch_stub_612
 #define gl_dispatch_stub_654 mgl_dispatch_stub_654
 #define gl_dispatch_stub_655 mgl_dispatch_stub_655
 #define gl_dispatch_stub_656 mgl_dispatch_stub_656
 #define gl_dispatch_stub_657 mgl_dispatch_stub_657
-#define gl_dispatch_stub_738 mgl_dispatch_stub_738
-#define gl_dispatch_stub_739 mgl_dispatch_stub_739
-#define gl_dispatch_stub_740 mgl_dispatch_stub_740
-#define gl_dispatch_stub_741 mgl_dispatch_stub_741
-#define gl_dispatch_stub_742 mgl_dispatch_stub_742
-#define gl_dispatch_stub_749 mgl_dispatch_stub_749
-#define gl_dispatch_stub_750 mgl_dispatch_stub_750
-#define gl_dispatch_stub_768 mgl_dispatch_stub_768
-#define gl_dispatch_stub_770 mgl_dispatch_stub_770
-#define gl_dispatch_stub_771 mgl_dispatch_stub_771
-#define gl_dispatch_stub_772 mgl_dispatch_stub_772
+#define gl_dispatch_stub_658 mgl_dispatch_stub_658
+#define gl_dispatch_stub_659 mgl_dispatch_stub_659
+#define gl_dispatch_stub_660 mgl_dispatch_stub_660
+#define gl_dispatch_stub_661 mgl_dispatch_stub_661
+#define gl_dispatch_stub_662 mgl_dispatch_stub_662
+#define gl_dispatch_stub_743 mgl_dispatch_stub_743
+#define gl_dispatch_stub_744 mgl_dispatch_stub_744
+#define gl_dispatch_stub_745 mgl_dispatch_stub_745
+#define gl_dispatch_stub_746 mgl_dispatch_stub_746
+#define gl_dispatch_stub_747 mgl_dispatch_stub_747
+#define gl_dispatch_stub_754 mgl_dispatch_stub_754
+#define gl_dispatch_stub_755 mgl_dispatch_stub_755
 #define gl_dispatch_stub_773 mgl_dispatch_stub_773
-#define gl_dispatch_stub_774 mgl_dispatch_stub_774
+#define gl_dispatch_stub_776 mgl_dispatch_stub_776
+#define gl_dispatch_stub_777 mgl_dispatch_stub_777
+#define gl_dispatch_stub_778 mgl_dispatch_stub_778
+#define gl_dispatch_stub_779 mgl_dispatch_stub_779
+#define gl_dispatch_stub_780 mgl_dispatch_stub_780
 #endif /* USE_MGL_NAMESPACE */
 
 
@@ -1192,39 +1200,39 @@ void GLAPIENTRY gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params
 void GLAPIENTRY gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format, GLenum type, GLvoid * values);
 void GLAPIENTRY gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params);
 void GLAPIENTRY gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params);
-void GLAPIENTRY gl_dispatch_stub_563(GLenum pname, GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_564(GLenum pname, GLint * params);
-void GLAPIENTRY gl_dispatch_stub_565(GLenum pname, GLfloat param);
-void GLAPIENTRY gl_dispatch_stub_566(GLenum pname, const GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_567(GLenum pname, GLint param);
-void GLAPIENTRY gl_dispatch_stub_568(GLenum pname, const GLint * params);
-void GLAPIENTRY gl_dispatch_stub_569(GLclampf value, GLboolean invert);
-void GLAPIENTRY gl_dispatch_stub_570(GLenum pattern);
-void GLAPIENTRY gl_dispatch_stub_581(GLenum pname, GLdouble * params);
-void GLAPIENTRY gl_dispatch_stub_582(GLenum pname, GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_607(GLenum mode);
-void GLAPIENTRY gl_dispatch_stub_649(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
-void GLAPIENTRY gl_dispatch_stub_650(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
-void GLAPIENTRY gl_dispatch_stub_651(GLsizei n, const GLuint * fences);
-void GLAPIENTRY gl_dispatch_stub_652(GLuint fence);
-void GLAPIENTRY gl_dispatch_stub_653(GLsizei n, GLuint * fences);
-void GLAPIENTRY gl_dispatch_stub_654(GLuint fence, GLenum pname, GLint * params);
-GLboolean GLAPIENTRY gl_dispatch_stub_655(GLuint fence);
-void GLAPIENTRY gl_dispatch_stub_656(GLuint fence, GLenum condition);
-GLboolean GLAPIENTRY gl_dispatch_stub_657(GLuint fence);
-void GLAPIENTRY gl_dispatch_stub_738(GLenum face);
-void GLAPIENTRY gl_dispatch_stub_739(GLuint array);
-void GLAPIENTRY gl_dispatch_stub_740(GLsizei n, const GLuint * arrays);
-void GLAPIENTRY gl_dispatch_stub_741(GLsizei n, GLuint * arrays);
-GLboolean GLAPIENTRY gl_dispatch_stub_742(GLuint array);
-void GLAPIENTRY gl_dispatch_stub_749(GLclampd zmin, GLclampd zmax);
-void GLAPIENTRY gl_dispatch_stub_750(GLenum modeRGB, GLenum modeA);
-void GLAPIENTRY gl_dispatch_stub_768(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
-void GLAPIENTRY gl_dispatch_stub_770(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
-void GLAPIENTRY gl_dispatch_stub_771(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_772(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
-void GLAPIENTRY gl_dispatch_stub_773(GLuint id, GLenum pname, GLint64EXT * params);
-void GLAPIENTRY gl_dispatch_stub_774(GLuint id, GLenum pname, GLuint64EXT * params);
+void GLAPIENTRY gl_dispatch_stub_568(GLenum pname, GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_569(GLenum pname, GLint * params);
+void GLAPIENTRY gl_dispatch_stub_570(GLenum pname, GLfloat param);
+void GLAPIENTRY gl_dispatch_stub_571(GLenum pname, const GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_572(GLenum pname, GLint param);
+void GLAPIENTRY gl_dispatch_stub_573(GLenum pname, const GLint * params);
+void GLAPIENTRY gl_dispatch_stub_574(GLclampf value, GLboolean invert);
+void GLAPIENTRY gl_dispatch_stub_575(GLenum pattern);
+void GLAPIENTRY gl_dispatch_stub_586(GLenum pname, GLdouble * params);
+void GLAPIENTRY gl_dispatch_stub_587(GLenum pname, GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_612(GLenum mode);
+void GLAPIENTRY gl_dispatch_stub_654(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
+void GLAPIENTRY gl_dispatch_stub_655(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
+void GLAPIENTRY gl_dispatch_stub_656(GLsizei n, const GLuint * fences);
+void GLAPIENTRY gl_dispatch_stub_657(GLuint fence);
+void GLAPIENTRY gl_dispatch_stub_658(GLsizei n, GLuint * fences);
+void GLAPIENTRY gl_dispatch_stub_659(GLuint fence, GLenum pname, GLint * params);
+GLboolean GLAPIENTRY gl_dispatch_stub_660(GLuint fence);
+void GLAPIENTRY gl_dispatch_stub_661(GLuint fence, GLenum condition);
+GLboolean GLAPIENTRY gl_dispatch_stub_662(GLuint fence);
+void GLAPIENTRY gl_dispatch_stub_743(GLenum face);
+void GLAPIENTRY gl_dispatch_stub_744(GLuint array);
+void GLAPIENTRY gl_dispatch_stub_745(GLsizei n, const GLuint * arrays);
+void GLAPIENTRY gl_dispatch_stub_746(GLsizei n, GLuint * arrays);
+GLboolean GLAPIENTRY gl_dispatch_stub_747(GLuint array);
+void GLAPIENTRY gl_dispatch_stub_754(GLclampd zmin, GLclampd zmax);
+void GLAPIENTRY gl_dispatch_stub_755(GLenum modeRGB, GLenum modeA);
+void GLAPIENTRY gl_dispatch_stub_773(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+void GLAPIENTRY gl_dispatch_stub_776(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
+void GLAPIENTRY gl_dispatch_stub_777(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_778(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+void GLAPIENTRY gl_dispatch_stub_779(GLuint id, GLenum pname, GLint64EXT * params);
+void GLAPIENTRY gl_dispatch_stub_780(GLuint id, GLenum pname, GLuint64EXT * params);
 #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */
 
 static const glprocs_table_t static_functions[] = {
@@ -1790,517 +1798,525 @@ static const glprocs_table_t static_functions[] = {
     NAME_FUNC_OFFSET( 8911, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
     NAME_FUNC_OFFSET( 8934, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
     NAME_FUNC_OFFSET( 8951, glRenderbufferStorageMultisample, glRenderbufferStorageMultisample, NULL, _gloffset_RenderbufferStorageMultisample),
-    NAME_FUNC_OFFSET( 8984, glPolygonOffsetEXT, glPolygonOffsetEXT, NULL, _gloffset_PolygonOffsetEXT),
-    NAME_FUNC_OFFSET( 9003, gl_dispatch_stub_563, gl_dispatch_stub_563, NULL, _gloffset_GetPixelTexGenParameterfvSGIS),
-    NAME_FUNC_OFFSET( 9035, gl_dispatch_stub_564, gl_dispatch_stub_564, NULL, _gloffset_GetPixelTexGenParameterivSGIS),
-    NAME_FUNC_OFFSET( 9067, gl_dispatch_stub_565, gl_dispatch_stub_565, NULL, _gloffset_PixelTexGenParameterfSGIS),
-    NAME_FUNC_OFFSET( 9095, gl_dispatch_stub_566, gl_dispatch_stub_566, NULL, _gloffset_PixelTexGenParameterfvSGIS),
-    NAME_FUNC_OFFSET( 9124, gl_dispatch_stub_567, gl_dispatch_stub_567, NULL, _gloffset_PixelTexGenParameteriSGIS),
-    NAME_FUNC_OFFSET( 9152, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_PixelTexGenParameterivSGIS),
-    NAME_FUNC_OFFSET( 9181, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SampleMaskSGIS),
-    NAME_FUNC_OFFSET( 9198, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_SamplePatternSGIS),
-    NAME_FUNC_OFFSET( 9218, glColorPointerEXT, glColorPointerEXT, NULL, _gloffset_ColorPointerEXT),
-    NAME_FUNC_OFFSET( 9236, glEdgeFlagPointerEXT, glEdgeFlagPointerEXT, NULL, _gloffset_EdgeFlagPointerEXT),
-    NAME_FUNC_OFFSET( 9257, glIndexPointerEXT, glIndexPointerEXT, NULL, _gloffset_IndexPointerEXT),
-    NAME_FUNC_OFFSET( 9275, glNormalPointerEXT, glNormalPointerEXT, NULL, _gloffset_NormalPointerEXT),
-    NAME_FUNC_OFFSET( 9294, glTexCoordPointerEXT, glTexCoordPointerEXT, NULL, _gloffset_TexCoordPointerEXT),
-    NAME_FUNC_OFFSET( 9315, glVertexPointerEXT, glVertexPointerEXT, NULL, _gloffset_VertexPointerEXT),
-    NAME_FUNC_OFFSET( 9334, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET( 9355, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET( 9377, glLockArraysEXT, glLockArraysEXT, NULL, _gloffset_LockArraysEXT),
-    NAME_FUNC_OFFSET( 9393, glUnlockArraysEXT, glUnlockArraysEXT, NULL, _gloffset_UnlockArraysEXT),
-    NAME_FUNC_OFFSET( 9411, gl_dispatch_stub_581, gl_dispatch_stub_581, NULL, _gloffset_CullParameterdvEXT),
-    NAME_FUNC_OFFSET( 9432, gl_dispatch_stub_582, gl_dispatch_stub_582, NULL, _gloffset_CullParameterfvEXT),
-    NAME_FUNC_OFFSET( 9453, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
-    NAME_FUNC_OFFSET( 9475, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
-    NAME_FUNC_OFFSET( 9498, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
-    NAME_FUNC_OFFSET( 9520, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
-    NAME_FUNC_OFFSET( 9543, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
-    NAME_FUNC_OFFSET( 9565, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
-    NAME_FUNC_OFFSET( 9588, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
-    NAME_FUNC_OFFSET( 9610, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
-    NAME_FUNC_OFFSET( 9633, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
-    NAME_FUNC_OFFSET( 9655, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
-    NAME_FUNC_OFFSET( 9678, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
-    NAME_FUNC_OFFSET( 9701, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
-    NAME_FUNC_OFFSET( 9725, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
-    NAME_FUNC_OFFSET( 9748, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
-    NAME_FUNC_OFFSET( 9772, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
-    NAME_FUNC_OFFSET( 9795, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
-    NAME_FUNC_OFFSET( 9819, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
-    NAME_FUNC_OFFSET( 9846, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
-    NAME_FUNC_OFFSET( 9867, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
-    NAME_FUNC_OFFSET( 9890, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
-    NAME_FUNC_OFFSET( 9911, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
-    NAME_FUNC_OFFSET( 9926, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
-    NAME_FUNC_OFFSET( 9942, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
-    NAME_FUNC_OFFSET( 9957, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
-    NAME_FUNC_OFFSET( 9973, gl_dispatch_stub_607, gl_dispatch_stub_607, NULL, _gloffset_PixelTexGenSGIX),
-    NAME_FUNC_OFFSET( 9991, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(10014, glFlushVertexArrayRangeNV, glFlushVertexArrayRangeNV, NULL, _gloffset_FlushVertexArrayRangeNV),
-    NAME_FUNC_OFFSET(10040, glVertexArrayRangeNV, glVertexArrayRangeNV, NULL, _gloffset_VertexArrayRangeNV),
-    NAME_FUNC_OFFSET(10061, glCombinerInputNV, glCombinerInputNV, NULL, _gloffset_CombinerInputNV),
-    NAME_FUNC_OFFSET(10079, glCombinerOutputNV, glCombinerOutputNV, NULL, _gloffset_CombinerOutputNV),
-    NAME_FUNC_OFFSET(10098, glCombinerParameterfNV, glCombinerParameterfNV, NULL, _gloffset_CombinerParameterfNV),
-    NAME_FUNC_OFFSET(10121, glCombinerParameterfvNV, glCombinerParameterfvNV, NULL, _gloffset_CombinerParameterfvNV),
-    NAME_FUNC_OFFSET(10145, glCombinerParameteriNV, glCombinerParameteriNV, NULL, _gloffset_CombinerParameteriNV),
-    NAME_FUNC_OFFSET(10168, glCombinerParameterivNV, glCombinerParameterivNV, NULL, _gloffset_CombinerParameterivNV),
-    NAME_FUNC_OFFSET(10192, glFinalCombinerInputNV, glFinalCombinerInputNV, NULL, _gloffset_FinalCombinerInputNV),
-    NAME_FUNC_OFFSET(10215, glGetCombinerInputParameterfvNV, glGetCombinerInputParameterfvNV, NULL, _gloffset_GetCombinerInputParameterfvNV),
-    NAME_FUNC_OFFSET(10247, glGetCombinerInputParameterivNV, glGetCombinerInputParameterivNV, NULL, _gloffset_GetCombinerInputParameterivNV),
-    NAME_FUNC_OFFSET(10279, glGetCombinerOutputParameterfvNV, glGetCombinerOutputParameterfvNV, NULL, _gloffset_GetCombinerOutputParameterfvNV),
-    NAME_FUNC_OFFSET(10312, glGetCombinerOutputParameterivNV, glGetCombinerOutputParameterivNV, NULL, _gloffset_GetCombinerOutputParameterivNV),
-    NAME_FUNC_OFFSET(10345, glGetFinalCombinerInputParameterfvNV, glGetFinalCombinerInputParameterfvNV, NULL, _gloffset_GetFinalCombinerInputParameterfvNV),
-    NAME_FUNC_OFFSET(10382, glGetFinalCombinerInputParameterivNV, glGetFinalCombinerInputParameterivNV, NULL, _gloffset_GetFinalCombinerInputParameterivNV),
-    NAME_FUNC_OFFSET(10419, glResizeBuffersMESA, glResizeBuffersMESA, NULL, _gloffset_ResizeBuffersMESA),
-    NAME_FUNC_OFFSET(10439, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(10457, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(10476, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(10494, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(10513, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(10531, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(10550, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(10568, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(10587, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(10605, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(10624, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(10642, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(10661, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(10679, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(10698, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(10716, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(10735, glWindowPos4dMESA, glWindowPos4dMESA, NULL, _gloffset_WindowPos4dMESA),
-    NAME_FUNC_OFFSET(10753, glWindowPos4dvMESA, glWindowPos4dvMESA, NULL, _gloffset_WindowPos4dvMESA),
-    NAME_FUNC_OFFSET(10772, glWindowPos4fMESA, glWindowPos4fMESA, NULL, _gloffset_WindowPos4fMESA),
-    NAME_FUNC_OFFSET(10790, glWindowPos4fvMESA, glWindowPos4fvMESA, NULL, _gloffset_WindowPos4fvMESA),
-    NAME_FUNC_OFFSET(10809, glWindowPos4iMESA, glWindowPos4iMESA, NULL, _gloffset_WindowPos4iMESA),
-    NAME_FUNC_OFFSET(10827, glWindowPos4ivMESA, glWindowPos4ivMESA, NULL, _gloffset_WindowPos4ivMESA),
-    NAME_FUNC_OFFSET(10846, glWindowPos4sMESA, glWindowPos4sMESA, NULL, _gloffset_WindowPos4sMESA),
-    NAME_FUNC_OFFSET(10864, glWindowPos4svMESA, glWindowPos4svMESA, NULL, _gloffset_WindowPos4svMESA),
-    NAME_FUNC_OFFSET(10883, gl_dispatch_stub_649, gl_dispatch_stub_649, NULL, _gloffset_MultiModeDrawArraysIBM),
-    NAME_FUNC_OFFSET(10908, gl_dispatch_stub_650, gl_dispatch_stub_650, NULL, _gloffset_MultiModeDrawElementsIBM),
-    NAME_FUNC_OFFSET(10935, gl_dispatch_stub_651, gl_dispatch_stub_651, NULL, _gloffset_DeleteFencesNV),
-    NAME_FUNC_OFFSET(10952, gl_dispatch_stub_652, gl_dispatch_stub_652, NULL, _gloffset_FinishFenceNV),
-    NAME_FUNC_OFFSET(10968, gl_dispatch_stub_653, gl_dispatch_stub_653, NULL, _gloffset_GenFencesNV),
-    NAME_FUNC_OFFSET(10982, gl_dispatch_stub_654, gl_dispatch_stub_654, NULL, _gloffset_GetFenceivNV),
-    NAME_FUNC_OFFSET(10997, gl_dispatch_stub_655, gl_dispatch_stub_655, NULL, _gloffset_IsFenceNV),
-    NAME_FUNC_OFFSET(11009, gl_dispatch_stub_656, gl_dispatch_stub_656, NULL, _gloffset_SetFenceNV),
-    NAME_FUNC_OFFSET(11022, gl_dispatch_stub_657, gl_dispatch_stub_657, NULL, _gloffset_TestFenceNV),
-    NAME_FUNC_OFFSET(11036, glAreProgramsResidentNV, glAreProgramsResidentNV, NULL, _gloffset_AreProgramsResidentNV),
-    NAME_FUNC_OFFSET(11060, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
-    NAME_FUNC_OFFSET(11076, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
-    NAME_FUNC_OFFSET(11095, glExecuteProgramNV, glExecuteProgramNV, NULL, _gloffset_ExecuteProgramNV),
-    NAME_FUNC_OFFSET(11114, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
-    NAME_FUNC_OFFSET(11130, glGetProgramParameterdvNV, glGetProgramParameterdvNV, NULL, _gloffset_GetProgramParameterdvNV),
-    NAME_FUNC_OFFSET(11156, glGetProgramParameterfvNV, glGetProgramParameterfvNV, NULL, _gloffset_GetProgramParameterfvNV),
-    NAME_FUNC_OFFSET(11182, glGetProgramStringNV, glGetProgramStringNV, NULL, _gloffset_GetProgramStringNV),
-    NAME_FUNC_OFFSET(11203, glGetProgramivNV, glGetProgramivNV, NULL, _gloffset_GetProgramivNV),
-    NAME_FUNC_OFFSET(11220, glGetTrackMatrixivNV, glGetTrackMatrixivNV, NULL, _gloffset_GetTrackMatrixivNV),
-    NAME_FUNC_OFFSET(11241, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(11269, glGetVertexAttribdvNV, glGetVertexAttribdvNV, NULL, _gloffset_GetVertexAttribdvNV),
-    NAME_FUNC_OFFSET(11291, glGetVertexAttribfvNV, glGetVertexAttribfvNV, NULL, _gloffset_GetVertexAttribfvNV),
-    NAME_FUNC_OFFSET(11313, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV),
-    NAME_FUNC_OFFSET(11335, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(11349, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV),
-    NAME_FUNC_OFFSET(11365, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV),
-    NAME_FUNC_OFFSET(11390, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV),
-    NAME_FUNC_OFFSET(11415, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV),
-    NAME_FUNC_OFFSET(11443, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV),
-    NAME_FUNC_OFFSET(11459, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV),
-    NAME_FUNC_OFFSET(11478, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV),
-    NAME_FUNC_OFFSET(11498, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV),
-    NAME_FUNC_OFFSET(11517, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV),
-    NAME_FUNC_OFFSET(11537, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV),
-    NAME_FUNC_OFFSET(11556, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV),
-    NAME_FUNC_OFFSET(11576, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV),
-    NAME_FUNC_OFFSET(11595, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV),
-    NAME_FUNC_OFFSET(11615, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV),
-    NAME_FUNC_OFFSET(11634, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV),
-    NAME_FUNC_OFFSET(11654, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV),
-    NAME_FUNC_OFFSET(11673, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV),
-    NAME_FUNC_OFFSET(11693, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV),
-    NAME_FUNC_OFFSET(11712, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV),
-    NAME_FUNC_OFFSET(11732, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV),
-    NAME_FUNC_OFFSET(11751, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV),
-    NAME_FUNC_OFFSET(11771, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV),
-    NAME_FUNC_OFFSET(11790, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV),
-    NAME_FUNC_OFFSET(11810, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV),
-    NAME_FUNC_OFFSET(11829, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV),
-    NAME_FUNC_OFFSET(11849, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV),
-    NAME_FUNC_OFFSET(11868, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV),
-    NAME_FUNC_OFFSET(11888, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV),
-    NAME_FUNC_OFFSET(11907, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV),
-    NAME_FUNC_OFFSET(11927, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV),
-    NAME_FUNC_OFFSET(11947, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV),
-    NAME_FUNC_OFFSET(11968, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV),
-    NAME_FUNC_OFFSET(11992, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV),
-    NAME_FUNC_OFFSET(12013, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV),
-    NAME_FUNC_OFFSET(12034, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV),
-    NAME_FUNC_OFFSET(12055, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV),
-    NAME_FUNC_OFFSET(12076, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV),
-    NAME_FUNC_OFFSET(12097, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV),
-    NAME_FUNC_OFFSET(12118, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV),
-    NAME_FUNC_OFFSET(12139, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV),
-    NAME_FUNC_OFFSET(12160, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV),
-    NAME_FUNC_OFFSET(12181, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV),
-    NAME_FUNC_OFFSET(12202, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV),
-    NAME_FUNC_OFFSET(12223, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV),
-    NAME_FUNC_OFFSET(12244, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV),
-    NAME_FUNC_OFFSET(12266, glGetTexBumpParameterfvATI, glGetTexBumpParameterfvATI, NULL, _gloffset_GetTexBumpParameterfvATI),
-    NAME_FUNC_OFFSET(12293, glGetTexBumpParameterivATI, glGetTexBumpParameterivATI, NULL, _gloffset_GetTexBumpParameterivATI),
-    NAME_FUNC_OFFSET(12320, glTexBumpParameterfvATI, glTexBumpParameterfvATI, NULL, _gloffset_TexBumpParameterfvATI),
-    NAME_FUNC_OFFSET(12344, glTexBumpParameterivATI, glTexBumpParameterivATI, NULL, _gloffset_TexBumpParameterivATI),
-    NAME_FUNC_OFFSET(12368, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI),
-    NAME_FUNC_OFFSET(12390, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI),
-    NAME_FUNC_OFFSET(12412, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI),
-    NAME_FUNC_OFFSET(12434, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI),
-    NAME_FUNC_OFFSET(12459, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI),
-    NAME_FUNC_OFFSET(12483, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI),
-    NAME_FUNC_OFFSET(12505, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI),
-    NAME_FUNC_OFFSET(12527, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI),
-    NAME_FUNC_OFFSET(12549, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI),
-    NAME_FUNC_OFFSET(12575, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI),
-    NAME_FUNC_OFFSET(12598, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI),
-    NAME_FUNC_OFFSET(12622, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI),
-    NAME_FUNC_OFFSET(12640, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI),
-    NAME_FUNC_OFFSET(12655, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI),
-    NAME_FUNC_OFFSET(12686, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(12706, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(12727, gl_dispatch_stub_738, gl_dispatch_stub_738, NULL, _gloffset_ActiveStencilFaceEXT),
-    NAME_FUNC_OFFSET(12750, gl_dispatch_stub_739, gl_dispatch_stub_739, NULL, _gloffset_BindVertexArrayAPPLE),
-    NAME_FUNC_OFFSET(12773, gl_dispatch_stub_740, gl_dispatch_stub_740, NULL, _gloffset_DeleteVertexArraysAPPLE),
-    NAME_FUNC_OFFSET(12799, gl_dispatch_stub_741, gl_dispatch_stub_741, NULL, _gloffset_GenVertexArraysAPPLE),
-    NAME_FUNC_OFFSET(12822, gl_dispatch_stub_742, gl_dispatch_stub_742, NULL, _gloffset_IsVertexArrayAPPLE),
-    NAME_FUNC_OFFSET(12843, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV),
-    NAME_FUNC_OFFSET(12874, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV),
-    NAME_FUNC_OFFSET(12905, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV),
-    NAME_FUNC_OFFSET(12933, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV),
-    NAME_FUNC_OFFSET(12962, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV),
-    NAME_FUNC_OFFSET(12990, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV),
-    NAME_FUNC_OFFSET(13019, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_DepthBoundsEXT),
-    NAME_FUNC_OFFSET(13036, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(13063, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
-    NAME_FUNC_OFFSET(13084, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
-    NAME_FUNC_OFFSET(13106, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
-    NAME_FUNC_OFFSET(13134, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
-    NAME_FUNC_OFFSET(13158, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
-    NAME_FUNC_OFFSET(13183, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
-    NAME_FUNC_OFFSET(13212, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
-    NAME_FUNC_OFFSET(13238, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
-    NAME_FUNC_OFFSET(13264, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
-    NAME_FUNC_OFFSET(13290, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
-    NAME_FUNC_OFFSET(13311, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
-    NAME_FUNC_OFFSET(13333, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
-    NAME_FUNC_OFFSET(13353, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
-    NAME_FUNC_OFFSET(13394, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
-    NAME_FUNC_OFFSET(13426, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
-    NAME_FUNC_OFFSET(13445, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
-    NAME_FUNC_OFFSET(13465, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
-    NAME_FUNC_OFFSET(13490, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_BlitFramebufferEXT),
-    NAME_FUNC_OFFSET(13511, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
-    NAME_FUNC_OFFSET(13540, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_StencilFuncSeparateATI),
-    NAME_FUNC_OFFSET(13565, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_ProgramEnvParameters4fvEXT),
-    NAME_FUNC_OFFSET(13594, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_ProgramLocalParameters4fvEXT),
-    NAME_FUNC_OFFSET(13625, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_GetQueryObjecti64vEXT),
-    NAME_FUNC_OFFSET(13649, gl_dispatch_stub_774, gl_dispatch_stub_774, NULL, _gloffset_GetQueryObjectui64vEXT),
-    NAME_FUNC_OFFSET(13674, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement),
-    NAME_FUNC_OFFSET(13692, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture),
-    NAME_FUNC_OFFSET(13709, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays),
-    NAME_FUNC_OFFSET(13725, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident),
-    NAME_FUNC_OFFSET(13750, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D),
-    NAME_FUNC_OFFSET(13770, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D),
-    NAME_FUNC_OFFSET(13790, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D),
-    NAME_FUNC_OFFSET(13813, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D),
-    NAME_FUNC_OFFSET(13836, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures),
-    NAME_FUNC_OFFSET(13856, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures),
-    NAME_FUNC_OFFSET(13873, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv),
-    NAME_FUNC_OFFSET(13890, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture),
-    NAME_FUNC_OFFSET(13905, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures),
-    NAME_FUNC_OFFSET(13929, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D),
-    NAME_FUNC_OFFSET(13948, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D),
-    NAME_FUNC_OFFSET(13967, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor),
-    NAME_FUNC_OFFSET(13983, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation),
-    NAME_FUNC_OFFSET(14002, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements),
-    NAME_FUNC_OFFSET(14025, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
-    NAME_FUNC_OFFSET(14041, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
-    NAME_FUNC_OFFSET(14057, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv),
-    NAME_FUNC_OFFSET(14084, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv),
-    NAME_FUNC_OFFSET(14111, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable),
-    NAME_FUNC_OFFSET(14131, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
-    NAME_FUNC_OFFSET(14150, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
-    NAME_FUNC_OFFSET(14169, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
-    NAME_FUNC_OFFSET(14199, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
-    NAME_FUNC_OFFSET(14229, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
-    NAME_FUNC_OFFSET(14259, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
-    NAME_FUNC_OFFSET(14289, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable),
-    NAME_FUNC_OFFSET(14308, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable),
-    NAME_FUNC_OFFSET(14331, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D),
-    NAME_FUNC_OFFSET(14356, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D),
-    NAME_FUNC_OFFSET(14381, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf),
-    NAME_FUNC_OFFSET(14408, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv),
-    NAME_FUNC_OFFSET(14436, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri),
-    NAME_FUNC_OFFSET(14463, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv),
-    NAME_FUNC_OFFSET(14491, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D),
-    NAME_FUNC_OFFSET(14520, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D),
-    NAME_FUNC_OFFSET(14549, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter),
-    NAME_FUNC_OFFSET(14575, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv),
-    NAME_FUNC_OFFSET(14606, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv),
-    NAME_FUNC_OFFSET(14637, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter),
-    NAME_FUNC_OFFSET(14661, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D),
-    NAME_FUNC_OFFSET(14684, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram),
-    NAME_FUNC_OFFSET(14702, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv),
-    NAME_FUNC_OFFSET(14731, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv),
-    NAME_FUNC_OFFSET(14760, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax),
-    NAME_FUNC_OFFSET(14775, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv),
-    NAME_FUNC_OFFSET(14801, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv),
-    NAME_FUNC_OFFSET(14827, glHistogram, glHistogram, NULL, _gloffset_Histogram),
-    NAME_FUNC_OFFSET(14842, glMinmax, glMinmax, NULL, _gloffset_Minmax),
-    NAME_FUNC_OFFSET(14854, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram),
-    NAME_FUNC_OFFSET(14874, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax),
-    NAME_FUNC_OFFSET(14891, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D),
-    NAME_FUNC_OFFSET(14907, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D),
-    NAME_FUNC_OFFSET(14926, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D),
-    NAME_FUNC_OFFSET(14949, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB),
-    NAME_FUNC_OFFSET(14965, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB),
-    NAME_FUNC_OFFSET(14987, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB),
-    NAME_FUNC_OFFSET(15005, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB),
-    NAME_FUNC_OFFSET(15024, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB),
-    NAME_FUNC_OFFSET(15042, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB),
-    NAME_FUNC_OFFSET(15061, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB),
-    NAME_FUNC_OFFSET(15079, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB),
-    NAME_FUNC_OFFSET(15098, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB),
-    NAME_FUNC_OFFSET(15116, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB),
-    NAME_FUNC_OFFSET(15135, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB),
-    NAME_FUNC_OFFSET(15153, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB),
-    NAME_FUNC_OFFSET(15172, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB),
-    NAME_FUNC_OFFSET(15190, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB),
-    NAME_FUNC_OFFSET(15209, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB),
-    NAME_FUNC_OFFSET(15227, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB),
-    NAME_FUNC_OFFSET(15246, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB),
-    NAME_FUNC_OFFSET(15264, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB),
-    NAME_FUNC_OFFSET(15283, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB),
-    NAME_FUNC_OFFSET(15301, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB),
-    NAME_FUNC_OFFSET(15320, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB),
-    NAME_FUNC_OFFSET(15338, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB),
-    NAME_FUNC_OFFSET(15357, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB),
-    NAME_FUNC_OFFSET(15375, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB),
-    NAME_FUNC_OFFSET(15394, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB),
-    NAME_FUNC_OFFSET(15412, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB),
-    NAME_FUNC_OFFSET(15431, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB),
-    NAME_FUNC_OFFSET(15449, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB),
-    NAME_FUNC_OFFSET(15468, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB),
-    NAME_FUNC_OFFSET(15486, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB),
-    NAME_FUNC_OFFSET(15505, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB),
-    NAME_FUNC_OFFSET(15523, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB),
-    NAME_FUNC_OFFSET(15542, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB),
-    NAME_FUNC_OFFSET(15560, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB),
-    NAME_FUNC_OFFSET(15579, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate),
-    NAME_FUNC_OFFSET(15602, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB),
-    NAME_FUNC_OFFSET(15625, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB),
-    NAME_FUNC_OFFSET(15648, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB),
-    NAME_FUNC_OFFSET(15671, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB),
-    NAME_FUNC_OFFSET(15694, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB),
-    NAME_FUNC_OFFSET(15711, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB),
-    NAME_FUNC_OFFSET(15734, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB),
-    NAME_FUNC_OFFSET(15757, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB),
-    NAME_FUNC_OFFSET(15780, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB),
-    NAME_FUNC_OFFSET(15806, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB),
-    NAME_FUNC_OFFSET(15832, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB),
-    NAME_FUNC_OFFSET(15858, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB),
-    NAME_FUNC_OFFSET(15882, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB),
-    NAME_FUNC_OFFSET(15909, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB),
-    NAME_FUNC_OFFSET(15935, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB),
-    NAME_FUNC_OFFSET(15955, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB),
-    NAME_FUNC_OFFSET(15975, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB),
-    NAME_FUNC_OFFSET(15995, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB),
-    NAME_FUNC_OFFSET(16018, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB),
-    NAME_FUNC_OFFSET(16042, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB),
-    NAME_FUNC_OFFSET(16065, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB),
-    NAME_FUNC_OFFSET(16089, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB),
-    NAME_FUNC_OFFSET(16106, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB),
-    NAME_FUNC_OFFSET(16124, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB),
-    NAME_FUNC_OFFSET(16141, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB),
-    NAME_FUNC_OFFSET(16159, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB),
-    NAME_FUNC_OFFSET(16176, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB),
-    NAME_FUNC_OFFSET(16194, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB),
-    NAME_FUNC_OFFSET(16211, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB),
-    NAME_FUNC_OFFSET(16229, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB),
-    NAME_FUNC_OFFSET(16246, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB),
-    NAME_FUNC_OFFSET(16264, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB),
-    NAME_FUNC_OFFSET(16281, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB),
-    NAME_FUNC_OFFSET(16299, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB),
-    NAME_FUNC_OFFSET(16316, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB),
-    NAME_FUNC_OFFSET(16334, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB),
-    NAME_FUNC_OFFSET(16351, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB),
-    NAME_FUNC_OFFSET(16369, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB),
-    NAME_FUNC_OFFSET(16386, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB),
-    NAME_FUNC_OFFSET(16404, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB),
-    NAME_FUNC_OFFSET(16423, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB),
-    NAME_FUNC_OFFSET(16442, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB),
-    NAME_FUNC_OFFSET(16461, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB),
-    NAME_FUNC_OFFSET(16480, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB),
-    NAME_FUNC_OFFSET(16500, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB),
-    NAME_FUNC_OFFSET(16520, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB),
-    NAME_FUNC_OFFSET(16540, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB),
-    NAME_FUNC_OFFSET(16558, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB),
-    NAME_FUNC_OFFSET(16575, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB),
-    NAME_FUNC_OFFSET(16593, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB),
-    NAME_FUNC_OFFSET(16610, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB),
-    NAME_FUNC_OFFSET(16628, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB),
-    NAME_FUNC_OFFSET(16646, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB),
-    NAME_FUNC_OFFSET(16663, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB),
-    NAME_FUNC_OFFSET(16681, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB),
-    NAME_FUNC_OFFSET(16700, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB),
-    NAME_FUNC_OFFSET(16719, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB),
-    NAME_FUNC_OFFSET(16738, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB),
-    NAME_FUNC_OFFSET(16760, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB),
-    NAME_FUNC_OFFSET(16773, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB),
-    NAME_FUNC_OFFSET(16786, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB),
-    NAME_FUNC_OFFSET(16802, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB),
-    NAME_FUNC_OFFSET(16818, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB),
-    NAME_FUNC_OFFSET(16831, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB),
-    NAME_FUNC_OFFSET(16854, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB),
-    NAME_FUNC_OFFSET(16874, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB),
-    NAME_FUNC_OFFSET(16893, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB),
-    NAME_FUNC_OFFSET(16904, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB),
-    NAME_FUNC_OFFSET(16916, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB),
-    NAME_FUNC_OFFSET(16930, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB),
-    NAME_FUNC_OFFSET(16943, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB),
-    NAME_FUNC_OFFSET(16959, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB),
-    NAME_FUNC_OFFSET(16970, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB),
-    NAME_FUNC_OFFSET(16983, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB),
-    NAME_FUNC_OFFSET(17002, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB),
-    NAME_FUNC_OFFSET(17022, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB),
-    NAME_FUNC_OFFSET(17035, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB),
-    NAME_FUNC_OFFSET(17045, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB),
-    NAME_FUNC_OFFSET(17061, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB),
-    NAME_FUNC_OFFSET(17080, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB),
-    NAME_FUNC_OFFSET(17098, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB),
-    NAME_FUNC_OFFSET(17119, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB),
-    NAME_FUNC_OFFSET(17134, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB),
-    NAME_FUNC_OFFSET(17149, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB),
-    NAME_FUNC_OFFSET(17163, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB),
-    NAME_FUNC_OFFSET(17178, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB),
-    NAME_FUNC_OFFSET(17190, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB),
-    NAME_FUNC_OFFSET(17203, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB),
-    NAME_FUNC_OFFSET(17215, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB),
-    NAME_FUNC_OFFSET(17228, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB),
-    NAME_FUNC_OFFSET(17240, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB),
-    NAME_FUNC_OFFSET(17253, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB),
-    NAME_FUNC_OFFSET(17265, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB),
-    NAME_FUNC_OFFSET(17278, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB),
-    NAME_FUNC_OFFSET(17290, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB),
-    NAME_FUNC_OFFSET(17303, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB),
-    NAME_FUNC_OFFSET(17315, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB),
-    NAME_FUNC_OFFSET(17328, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB),
-    NAME_FUNC_OFFSET(17340, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB),
-    NAME_FUNC_OFFSET(17353, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB),
-    NAME_FUNC_OFFSET(17365, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB),
-    NAME_FUNC_OFFSET(17378, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB),
-    NAME_FUNC_OFFSET(17397, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB),
-    NAME_FUNC_OFFSET(17416, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB),
-    NAME_FUNC_OFFSET(17435, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB),
-    NAME_FUNC_OFFSET(17448, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB),
-    NAME_FUNC_OFFSET(17466, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB),
-    NAME_FUNC_OFFSET(17487, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
-    NAME_FUNC_OFFSET(17505, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
-    NAME_FUNC_OFFSET(17525, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET(17539, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET(17556, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SampleMaskSGIS),
-    NAME_FUNC_OFFSET(17572, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_SamplePatternSGIS),
-    NAME_FUNC_OFFSET(17591, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17609, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17630, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17652, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17671, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17693, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17716, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
-    NAME_FUNC_OFFSET(17735, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
-    NAME_FUNC_OFFSET(17755, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
-    NAME_FUNC_OFFSET(17774, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
-    NAME_FUNC_OFFSET(17794, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
-    NAME_FUNC_OFFSET(17813, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
-    NAME_FUNC_OFFSET(17833, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
-    NAME_FUNC_OFFSET(17852, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
-    NAME_FUNC_OFFSET(17872, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
-    NAME_FUNC_OFFSET(17891, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
-    NAME_FUNC_OFFSET(17911, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
-    NAME_FUNC_OFFSET(17931, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
-    NAME_FUNC_OFFSET(17952, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
-    NAME_FUNC_OFFSET(17972, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
-    NAME_FUNC_OFFSET(17993, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
-    NAME_FUNC_OFFSET(18013, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
-    NAME_FUNC_OFFSET(18034, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
-    NAME_FUNC_OFFSET(18058, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
-    NAME_FUNC_OFFSET(18076, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
-    NAME_FUNC_OFFSET(18096, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
-    NAME_FUNC_OFFSET(18114, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
-    NAME_FUNC_OFFSET(18126, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
-    NAME_FUNC_OFFSET(18139, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
-    NAME_FUNC_OFFSET(18151, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
-    NAME_FUNC_OFFSET(18164, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(18184, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(18208, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(18222, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(18239, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(18254, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(18272, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(18286, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(18303, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(18318, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(18336, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(18350, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(18367, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(18382, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(18400, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(18414, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(18431, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(18446, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(18464, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(18478, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(18495, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(18510, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(18528, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(18542, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(18559, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(18574, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(18592, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(18606, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(18623, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(18638, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(18656, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(18670, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(18687, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(18702, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(18720, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
-    NAME_FUNC_OFFSET(18737, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
-    NAME_FUNC_OFFSET(18757, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
-    NAME_FUNC_OFFSET(18774, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18800, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18829, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(18844, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(18862, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(18881, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(18905, gl_dispatch_stub_750, gl_dispatch_stub_750, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(18932, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
-    NAME_FUNC_OFFSET(18950, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
-    NAME_FUNC_OFFSET(18969, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
-    NAME_FUNC_OFFSET(18994, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
-    NAME_FUNC_OFFSET(19015, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
-    NAME_FUNC_OFFSET(19037, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
-    NAME_FUNC_OFFSET(19063, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
-    NAME_FUNC_OFFSET(19086, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
-    NAME_FUNC_OFFSET(19109, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
-    NAME_FUNC_OFFSET(19132, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
-    NAME_FUNC_OFFSET(19150, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
-    NAME_FUNC_OFFSET(19169, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
-    NAME_FUNC_OFFSET(19186, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
-    NAME_FUNC_OFFSET(19224, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
-    NAME_FUNC_OFFSET(19253, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
-    NAME_FUNC_OFFSET(19269, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
-    NAME_FUNC_OFFSET(19286, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
-    NAME_FUNC_OFFSET(19308, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_BlitFramebufferEXT),
-    NAME_FUNC_OFFSET(19326, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
+    NAME_FUNC_OFFSET( 8984, glFlushMappedBufferRange, glFlushMappedBufferRange, NULL, _gloffset_FlushMappedBufferRange),
+    NAME_FUNC_OFFSET( 9009, glMapBufferRange, glMapBufferRange, NULL, _gloffset_MapBufferRange),
+    NAME_FUNC_OFFSET( 9026, glBindVertexArray, glBindVertexArray, NULL, _gloffset_BindVertexArray),
+    NAME_FUNC_OFFSET( 9044, glGenVertexArrays, glGenVertexArrays, NULL, _gloffset_GenVertexArrays),
+    NAME_FUNC_OFFSET( 9062, glCopyBufferSubData, glCopyBufferSubData, NULL, _gloffset_CopyBufferSubData),
+    NAME_FUNC_OFFSET( 9082, glPolygonOffsetEXT, glPolygonOffsetEXT, NULL, _gloffset_PolygonOffsetEXT),
+    NAME_FUNC_OFFSET( 9101, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_GetPixelTexGenParameterfvSGIS),
+    NAME_FUNC_OFFSET( 9133, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_GetPixelTexGenParameterivSGIS),
+    NAME_FUNC_OFFSET( 9165, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_PixelTexGenParameterfSGIS),
+    NAME_FUNC_OFFSET( 9193, gl_dispatch_stub_571, gl_dispatch_stub_571, NULL, _gloffset_PixelTexGenParameterfvSGIS),
+    NAME_FUNC_OFFSET( 9222, gl_dispatch_stub_572, gl_dispatch_stub_572, NULL, _gloffset_PixelTexGenParameteriSGIS),
+    NAME_FUNC_OFFSET( 9250, gl_dispatch_stub_573, gl_dispatch_stub_573, NULL, _gloffset_PixelTexGenParameterivSGIS),
+    NAME_FUNC_OFFSET( 9279, gl_dispatch_stub_574, gl_dispatch_stub_574, NULL, _gloffset_SampleMaskSGIS),
+    NAME_FUNC_OFFSET( 9296, gl_dispatch_stub_575, gl_dispatch_stub_575, NULL, _gloffset_SamplePatternSGIS),
+    NAME_FUNC_OFFSET( 9316, glColorPointerEXT, glColorPointerEXT, NULL, _gloffset_ColorPointerEXT),
+    NAME_FUNC_OFFSET( 9334, glEdgeFlagPointerEXT, glEdgeFlagPointerEXT, NULL, _gloffset_EdgeFlagPointerEXT),
+    NAME_FUNC_OFFSET( 9355, glIndexPointerEXT, glIndexPointerEXT, NULL, _gloffset_IndexPointerEXT),
+    NAME_FUNC_OFFSET( 9373, glNormalPointerEXT, glNormalPointerEXT, NULL, _gloffset_NormalPointerEXT),
+    NAME_FUNC_OFFSET( 9392, glTexCoordPointerEXT, glTexCoordPointerEXT, NULL, _gloffset_TexCoordPointerEXT),
+    NAME_FUNC_OFFSET( 9413, glVertexPointerEXT, glVertexPointerEXT, NULL, _gloffset_VertexPointerEXT),
+    NAME_FUNC_OFFSET( 9432, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET( 9453, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET( 9475, glLockArraysEXT, glLockArraysEXT, NULL, _gloffset_LockArraysEXT),
+    NAME_FUNC_OFFSET( 9491, glUnlockArraysEXT, glUnlockArraysEXT, NULL, _gloffset_UnlockArraysEXT),
+    NAME_FUNC_OFFSET( 9509, gl_dispatch_stub_586, gl_dispatch_stub_586, NULL, _gloffset_CullParameterdvEXT),
+    NAME_FUNC_OFFSET( 9530, gl_dispatch_stub_587, gl_dispatch_stub_587, NULL, _gloffset_CullParameterfvEXT),
+    NAME_FUNC_OFFSET( 9551, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
+    NAME_FUNC_OFFSET( 9573, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
+    NAME_FUNC_OFFSET( 9596, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
+    NAME_FUNC_OFFSET( 9618, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
+    NAME_FUNC_OFFSET( 9641, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
+    NAME_FUNC_OFFSET( 9663, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
+    NAME_FUNC_OFFSET( 9686, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
+    NAME_FUNC_OFFSET( 9708, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
+    NAME_FUNC_OFFSET( 9731, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
+    NAME_FUNC_OFFSET( 9753, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
+    NAME_FUNC_OFFSET( 9776, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
+    NAME_FUNC_OFFSET( 9799, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
+    NAME_FUNC_OFFSET( 9823, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
+    NAME_FUNC_OFFSET( 9846, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
+    NAME_FUNC_OFFSET( 9870, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
+    NAME_FUNC_OFFSET( 9893, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
+    NAME_FUNC_OFFSET( 9917, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
+    NAME_FUNC_OFFSET( 9944, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
+    NAME_FUNC_OFFSET( 9965, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
+    NAME_FUNC_OFFSET( 9988, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
+    NAME_FUNC_OFFSET(10009, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
+    NAME_FUNC_OFFSET(10024, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
+    NAME_FUNC_OFFSET(10040, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
+    NAME_FUNC_OFFSET(10055, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
+    NAME_FUNC_OFFSET(10071, gl_dispatch_stub_612, gl_dispatch_stub_612, NULL, _gloffset_PixelTexGenSGIX),
+    NAME_FUNC_OFFSET(10089, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(10112, glFlushVertexArrayRangeNV, glFlushVertexArrayRangeNV, NULL, _gloffset_FlushVertexArrayRangeNV),
+    NAME_FUNC_OFFSET(10138, glVertexArrayRangeNV, glVertexArrayRangeNV, NULL, _gloffset_VertexArrayRangeNV),
+    NAME_FUNC_OFFSET(10159, glCombinerInputNV, glCombinerInputNV, NULL, _gloffset_CombinerInputNV),
+    NAME_FUNC_OFFSET(10177, glCombinerOutputNV, glCombinerOutputNV, NULL, _gloffset_CombinerOutputNV),
+    NAME_FUNC_OFFSET(10196, glCombinerParameterfNV, glCombinerParameterfNV, NULL, _gloffset_CombinerParameterfNV),
+    NAME_FUNC_OFFSET(10219, glCombinerParameterfvNV, glCombinerParameterfvNV, NULL, _gloffset_CombinerParameterfvNV),
+    NAME_FUNC_OFFSET(10243, glCombinerParameteriNV, glCombinerParameteriNV, NULL, _gloffset_CombinerParameteriNV),
+    NAME_FUNC_OFFSET(10266, glCombinerParameterivNV, glCombinerParameterivNV, NULL, _gloffset_CombinerParameterivNV),
+    NAME_FUNC_OFFSET(10290, glFinalCombinerInputNV, glFinalCombinerInputNV, NULL, _gloffset_FinalCombinerInputNV),
+    NAME_FUNC_OFFSET(10313, glGetCombinerInputParameterfvNV, glGetCombinerInputParameterfvNV, NULL, _gloffset_GetCombinerInputParameterfvNV),
+    NAME_FUNC_OFFSET(10345, glGetCombinerInputParameterivNV, glGetCombinerInputParameterivNV, NULL, _gloffset_GetCombinerInputParameterivNV),
+    NAME_FUNC_OFFSET(10377, glGetCombinerOutputParameterfvNV, glGetCombinerOutputParameterfvNV, NULL, _gloffset_GetCombinerOutputParameterfvNV),
+    NAME_FUNC_OFFSET(10410, glGetCombinerOutputParameterivNV, glGetCombinerOutputParameterivNV, NULL, _gloffset_GetCombinerOutputParameterivNV),
+    NAME_FUNC_OFFSET(10443, glGetFinalCombinerInputParameterfvNV, glGetFinalCombinerInputParameterfvNV, NULL, _gloffset_GetFinalCombinerInputParameterfvNV),
+    NAME_FUNC_OFFSET(10480, glGetFinalCombinerInputParameterivNV, glGetFinalCombinerInputParameterivNV, NULL, _gloffset_GetFinalCombinerInputParameterivNV),
+    NAME_FUNC_OFFSET(10517, glResizeBuffersMESA, glResizeBuffersMESA, NULL, _gloffset_ResizeBuffersMESA),
+    NAME_FUNC_OFFSET(10537, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(10555, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(10574, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(10592, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(10611, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(10629, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(10648, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(10666, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(10685, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(10703, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(10722, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(10740, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(10759, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(10777, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(10796, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(10814, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(10833, glWindowPos4dMESA, glWindowPos4dMESA, NULL, _gloffset_WindowPos4dMESA),
+    NAME_FUNC_OFFSET(10851, glWindowPos4dvMESA, glWindowPos4dvMESA, NULL, _gloffset_WindowPos4dvMESA),
+    NAME_FUNC_OFFSET(10870, glWindowPos4fMESA, glWindowPos4fMESA, NULL, _gloffset_WindowPos4fMESA),
+    NAME_FUNC_OFFSET(10888, glWindowPos4fvMESA, glWindowPos4fvMESA, NULL, _gloffset_WindowPos4fvMESA),
+    NAME_FUNC_OFFSET(10907, glWindowPos4iMESA, glWindowPos4iMESA, NULL, _gloffset_WindowPos4iMESA),
+    NAME_FUNC_OFFSET(10925, glWindowPos4ivMESA, glWindowPos4ivMESA, NULL, _gloffset_WindowPos4ivMESA),
+    NAME_FUNC_OFFSET(10944, glWindowPos4sMESA, glWindowPos4sMESA, NULL, _gloffset_WindowPos4sMESA),
+    NAME_FUNC_OFFSET(10962, glWindowPos4svMESA, glWindowPos4svMESA, NULL, _gloffset_WindowPos4svMESA),
+    NAME_FUNC_OFFSET(10981, gl_dispatch_stub_654, gl_dispatch_stub_654, NULL, _gloffset_MultiModeDrawArraysIBM),
+    NAME_FUNC_OFFSET(11006, gl_dispatch_stub_655, gl_dispatch_stub_655, NULL, _gloffset_MultiModeDrawElementsIBM),
+    NAME_FUNC_OFFSET(11033, gl_dispatch_stub_656, gl_dispatch_stub_656, NULL, _gloffset_DeleteFencesNV),
+    NAME_FUNC_OFFSET(11050, gl_dispatch_stub_657, gl_dispatch_stub_657, NULL, _gloffset_FinishFenceNV),
+    NAME_FUNC_OFFSET(11066, gl_dispatch_stub_658, gl_dispatch_stub_658, NULL, _gloffset_GenFencesNV),
+    NAME_FUNC_OFFSET(11080, gl_dispatch_stub_659, gl_dispatch_stub_659, NULL, _gloffset_GetFenceivNV),
+    NAME_FUNC_OFFSET(11095, gl_dispatch_stub_660, gl_dispatch_stub_660, NULL, _gloffset_IsFenceNV),
+    NAME_FUNC_OFFSET(11107, gl_dispatch_stub_661, gl_dispatch_stub_661, NULL, _gloffset_SetFenceNV),
+    NAME_FUNC_OFFSET(11120, gl_dispatch_stub_662, gl_dispatch_stub_662, NULL, _gloffset_TestFenceNV),
+    NAME_FUNC_OFFSET(11134, glAreProgramsResidentNV, glAreProgramsResidentNV, NULL, _gloffset_AreProgramsResidentNV),
+    NAME_FUNC_OFFSET(11158, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
+    NAME_FUNC_OFFSET(11174, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
+    NAME_FUNC_OFFSET(11193, glExecuteProgramNV, glExecuteProgramNV, NULL, _gloffset_ExecuteProgramNV),
+    NAME_FUNC_OFFSET(11212, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
+    NAME_FUNC_OFFSET(11228, glGetProgramParameterdvNV, glGetProgramParameterdvNV, NULL, _gloffset_GetProgramParameterdvNV),
+    NAME_FUNC_OFFSET(11254, glGetProgramParameterfvNV, glGetProgramParameterfvNV, NULL, _gloffset_GetProgramParameterfvNV),
+    NAME_FUNC_OFFSET(11280, glGetProgramStringNV, glGetProgramStringNV, NULL, _gloffset_GetProgramStringNV),
+    NAME_FUNC_OFFSET(11301, glGetProgramivNV, glGetProgramivNV, NULL, _gloffset_GetProgramivNV),
+    NAME_FUNC_OFFSET(11318, glGetTrackMatrixivNV, glGetTrackMatrixivNV, NULL, _gloffset_GetTrackMatrixivNV),
+    NAME_FUNC_OFFSET(11339, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(11367, glGetVertexAttribdvNV, glGetVertexAttribdvNV, NULL, _gloffset_GetVertexAttribdvNV),
+    NAME_FUNC_OFFSET(11389, glGetVertexAttribfvNV, glGetVertexAttribfvNV, NULL, _gloffset_GetVertexAttribfvNV),
+    NAME_FUNC_OFFSET(11411, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV),
+    NAME_FUNC_OFFSET(11433, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(11447, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV),
+    NAME_FUNC_OFFSET(11463, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV),
+    NAME_FUNC_OFFSET(11488, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV),
+    NAME_FUNC_OFFSET(11513, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV),
+    NAME_FUNC_OFFSET(11541, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV),
+    NAME_FUNC_OFFSET(11557, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV),
+    NAME_FUNC_OFFSET(11576, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV),
+    NAME_FUNC_OFFSET(11596, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV),
+    NAME_FUNC_OFFSET(11615, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV),
+    NAME_FUNC_OFFSET(11635, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV),
+    NAME_FUNC_OFFSET(11654, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV),
+    NAME_FUNC_OFFSET(11674, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV),
+    NAME_FUNC_OFFSET(11693, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV),
+    NAME_FUNC_OFFSET(11713, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV),
+    NAME_FUNC_OFFSET(11732, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV),
+    NAME_FUNC_OFFSET(11752, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV),
+    NAME_FUNC_OFFSET(11771, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV),
+    NAME_FUNC_OFFSET(11791, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV),
+    NAME_FUNC_OFFSET(11810, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV),
+    NAME_FUNC_OFFSET(11830, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV),
+    NAME_FUNC_OFFSET(11849, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV),
+    NAME_FUNC_OFFSET(11869, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV),
+    NAME_FUNC_OFFSET(11888, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV),
+    NAME_FUNC_OFFSET(11908, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV),
+    NAME_FUNC_OFFSET(11927, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV),
+    NAME_FUNC_OFFSET(11947, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV),
+    NAME_FUNC_OFFSET(11966, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV),
+    NAME_FUNC_OFFSET(11986, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV),
+    NAME_FUNC_OFFSET(12005, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV),
+    NAME_FUNC_OFFSET(12025, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV),
+    NAME_FUNC_OFFSET(12045, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV),
+    NAME_FUNC_OFFSET(12066, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV),
+    NAME_FUNC_OFFSET(12090, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV),
+    NAME_FUNC_OFFSET(12111, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV),
+    NAME_FUNC_OFFSET(12132, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV),
+    NAME_FUNC_OFFSET(12153, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV),
+    NAME_FUNC_OFFSET(12174, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV),
+    NAME_FUNC_OFFSET(12195, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV),
+    NAME_FUNC_OFFSET(12216, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV),
+    NAME_FUNC_OFFSET(12237, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV),
+    NAME_FUNC_OFFSET(12258, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV),
+    NAME_FUNC_OFFSET(12279, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV),
+    NAME_FUNC_OFFSET(12300, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV),
+    NAME_FUNC_OFFSET(12321, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV),
+    NAME_FUNC_OFFSET(12342, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV),
+    NAME_FUNC_OFFSET(12364, glGetTexBumpParameterfvATI, glGetTexBumpParameterfvATI, NULL, _gloffset_GetTexBumpParameterfvATI),
+    NAME_FUNC_OFFSET(12391, glGetTexBumpParameterivATI, glGetTexBumpParameterivATI, NULL, _gloffset_GetTexBumpParameterivATI),
+    NAME_FUNC_OFFSET(12418, glTexBumpParameterfvATI, glTexBumpParameterfvATI, NULL, _gloffset_TexBumpParameterfvATI),
+    NAME_FUNC_OFFSET(12442, glTexBumpParameterivATI, glTexBumpParameterivATI, NULL, _gloffset_TexBumpParameterivATI),
+    NAME_FUNC_OFFSET(12466, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI),
+    NAME_FUNC_OFFSET(12488, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI),
+    NAME_FUNC_OFFSET(12510, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI),
+    NAME_FUNC_OFFSET(12532, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI),
+    NAME_FUNC_OFFSET(12557, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI),
+    NAME_FUNC_OFFSET(12581, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI),
+    NAME_FUNC_OFFSET(12603, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI),
+    NAME_FUNC_OFFSET(12625, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI),
+    NAME_FUNC_OFFSET(12647, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI),
+    NAME_FUNC_OFFSET(12673, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI),
+    NAME_FUNC_OFFSET(12696, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI),
+    NAME_FUNC_OFFSET(12720, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI),
+    NAME_FUNC_OFFSET(12738, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI),
+    NAME_FUNC_OFFSET(12753, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI),
+    NAME_FUNC_OFFSET(12784, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(12804, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(12825, gl_dispatch_stub_743, gl_dispatch_stub_743, NULL, _gloffset_ActiveStencilFaceEXT),
+    NAME_FUNC_OFFSET(12848, gl_dispatch_stub_744, gl_dispatch_stub_744, NULL, _gloffset_BindVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(12871, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_DeleteVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(12897, gl_dispatch_stub_746, gl_dispatch_stub_746, NULL, _gloffset_GenVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(12920, gl_dispatch_stub_747, gl_dispatch_stub_747, NULL, _gloffset_IsVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(12941, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV),
+    NAME_FUNC_OFFSET(12972, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV),
+    NAME_FUNC_OFFSET(13003, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV),
+    NAME_FUNC_OFFSET(13031, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV),
+    NAME_FUNC_OFFSET(13060, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV),
+    NAME_FUNC_OFFSET(13088, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV),
+    NAME_FUNC_OFFSET(13117, gl_dispatch_stub_754, gl_dispatch_stub_754, NULL, _gloffset_DepthBoundsEXT),
+    NAME_FUNC_OFFSET(13134, gl_dispatch_stub_755, gl_dispatch_stub_755, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(13161, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
+    NAME_FUNC_OFFSET(13182, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
+    NAME_FUNC_OFFSET(13204, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
+    NAME_FUNC_OFFSET(13232, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
+    NAME_FUNC_OFFSET(13256, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
+    NAME_FUNC_OFFSET(13281, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
+    NAME_FUNC_OFFSET(13310, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
+    NAME_FUNC_OFFSET(13336, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
+    NAME_FUNC_OFFSET(13362, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
+    NAME_FUNC_OFFSET(13388, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
+    NAME_FUNC_OFFSET(13409, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
+    NAME_FUNC_OFFSET(13431, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
+    NAME_FUNC_OFFSET(13451, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
+    NAME_FUNC_OFFSET(13492, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
+    NAME_FUNC_OFFSET(13524, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
+    NAME_FUNC_OFFSET(13543, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
+    NAME_FUNC_OFFSET(13563, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
+    NAME_FUNC_OFFSET(13588, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_BlitFramebufferEXT),
+    NAME_FUNC_OFFSET(13609, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
+    NAME_FUNC_OFFSET(13638, glProvokingVertexEXT, glProvokingVertexEXT, NULL, _gloffset_ProvokingVertexEXT),
+    NAME_FUNC_OFFSET(13659, gl_dispatch_stub_776, gl_dispatch_stub_776, NULL, _gloffset_StencilFuncSeparateATI),
+    NAME_FUNC_OFFSET(13684, gl_dispatch_stub_777, gl_dispatch_stub_777, NULL, _gloffset_ProgramEnvParameters4fvEXT),
+    NAME_FUNC_OFFSET(13713, gl_dispatch_stub_778, gl_dispatch_stub_778, NULL, _gloffset_ProgramLocalParameters4fvEXT),
+    NAME_FUNC_OFFSET(13744, gl_dispatch_stub_779, gl_dispatch_stub_779, NULL, _gloffset_GetQueryObjecti64vEXT),
+    NAME_FUNC_OFFSET(13768, gl_dispatch_stub_780, gl_dispatch_stub_780, NULL, _gloffset_GetQueryObjectui64vEXT),
+    NAME_FUNC_OFFSET(13793, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement),
+    NAME_FUNC_OFFSET(13811, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture),
+    NAME_FUNC_OFFSET(13828, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays),
+    NAME_FUNC_OFFSET(13844, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident),
+    NAME_FUNC_OFFSET(13869, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D),
+    NAME_FUNC_OFFSET(13889, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D),
+    NAME_FUNC_OFFSET(13909, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D),
+    NAME_FUNC_OFFSET(13932, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D),
+    NAME_FUNC_OFFSET(13955, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures),
+    NAME_FUNC_OFFSET(13975, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures),
+    NAME_FUNC_OFFSET(13992, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv),
+    NAME_FUNC_OFFSET(14009, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture),
+    NAME_FUNC_OFFSET(14024, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures),
+    NAME_FUNC_OFFSET(14048, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D),
+    NAME_FUNC_OFFSET(14067, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D),
+    NAME_FUNC_OFFSET(14086, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor),
+    NAME_FUNC_OFFSET(14102, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation),
+    NAME_FUNC_OFFSET(14121, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements),
+    NAME_FUNC_OFFSET(14144, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
+    NAME_FUNC_OFFSET(14160, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
+    NAME_FUNC_OFFSET(14176, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv),
+    NAME_FUNC_OFFSET(14203, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv),
+    NAME_FUNC_OFFSET(14230, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable),
+    NAME_FUNC_OFFSET(14250, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
+    NAME_FUNC_OFFSET(14269, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
+    NAME_FUNC_OFFSET(14288, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
+    NAME_FUNC_OFFSET(14318, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
+    NAME_FUNC_OFFSET(14348, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
+    NAME_FUNC_OFFSET(14378, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
+    NAME_FUNC_OFFSET(14408, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable),
+    NAME_FUNC_OFFSET(14427, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable),
+    NAME_FUNC_OFFSET(14450, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D),
+    NAME_FUNC_OFFSET(14475, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D),
+    NAME_FUNC_OFFSET(14500, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf),
+    NAME_FUNC_OFFSET(14527, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv),
+    NAME_FUNC_OFFSET(14555, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri),
+    NAME_FUNC_OFFSET(14582, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv),
+    NAME_FUNC_OFFSET(14610, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D),
+    NAME_FUNC_OFFSET(14639, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D),
+    NAME_FUNC_OFFSET(14668, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter),
+    NAME_FUNC_OFFSET(14694, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv),
+    NAME_FUNC_OFFSET(14725, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv),
+    NAME_FUNC_OFFSET(14756, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter),
+    NAME_FUNC_OFFSET(14780, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D),
+    NAME_FUNC_OFFSET(14803, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram),
+    NAME_FUNC_OFFSET(14821, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv),
+    NAME_FUNC_OFFSET(14850, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv),
+    NAME_FUNC_OFFSET(14879, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax),
+    NAME_FUNC_OFFSET(14894, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv),
+    NAME_FUNC_OFFSET(14920, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv),
+    NAME_FUNC_OFFSET(14946, glHistogram, glHistogram, NULL, _gloffset_Histogram),
+    NAME_FUNC_OFFSET(14961, glMinmax, glMinmax, NULL, _gloffset_Minmax),
+    NAME_FUNC_OFFSET(14973, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram),
+    NAME_FUNC_OFFSET(14993, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax),
+    NAME_FUNC_OFFSET(15010, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D),
+    NAME_FUNC_OFFSET(15026, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D),
+    NAME_FUNC_OFFSET(15045, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D),
+    NAME_FUNC_OFFSET(15068, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB),
+    NAME_FUNC_OFFSET(15084, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB),
+    NAME_FUNC_OFFSET(15106, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB),
+    NAME_FUNC_OFFSET(15124, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB),
+    NAME_FUNC_OFFSET(15143, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB),
+    NAME_FUNC_OFFSET(15161, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB),
+    NAME_FUNC_OFFSET(15180, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB),
+    NAME_FUNC_OFFSET(15198, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB),
+    NAME_FUNC_OFFSET(15217, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB),
+    NAME_FUNC_OFFSET(15235, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB),
+    NAME_FUNC_OFFSET(15254, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB),
+    NAME_FUNC_OFFSET(15272, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB),
+    NAME_FUNC_OFFSET(15291, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB),
+    NAME_FUNC_OFFSET(15309, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB),
+    NAME_FUNC_OFFSET(15328, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB),
+    NAME_FUNC_OFFSET(15346, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB),
+    NAME_FUNC_OFFSET(15365, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB),
+    NAME_FUNC_OFFSET(15383, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB),
+    NAME_FUNC_OFFSET(15402, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB),
+    NAME_FUNC_OFFSET(15420, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB),
+    NAME_FUNC_OFFSET(15439, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB),
+    NAME_FUNC_OFFSET(15457, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB),
+    NAME_FUNC_OFFSET(15476, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB),
+    NAME_FUNC_OFFSET(15494, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB),
+    NAME_FUNC_OFFSET(15513, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB),
+    NAME_FUNC_OFFSET(15531, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB),
+    NAME_FUNC_OFFSET(15550, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB),
+    NAME_FUNC_OFFSET(15568, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB),
+    NAME_FUNC_OFFSET(15587, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB),
+    NAME_FUNC_OFFSET(15605, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB),
+    NAME_FUNC_OFFSET(15624, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB),
+    NAME_FUNC_OFFSET(15642, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB),
+    NAME_FUNC_OFFSET(15661, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB),
+    NAME_FUNC_OFFSET(15679, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB),
+    NAME_FUNC_OFFSET(15698, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate),
+    NAME_FUNC_OFFSET(15721, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB),
+    NAME_FUNC_OFFSET(15744, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB),
+    NAME_FUNC_OFFSET(15767, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB),
+    NAME_FUNC_OFFSET(15790, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB),
+    NAME_FUNC_OFFSET(15813, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB),
+    NAME_FUNC_OFFSET(15830, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB),
+    NAME_FUNC_OFFSET(15853, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB),
+    NAME_FUNC_OFFSET(15876, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB),
+    NAME_FUNC_OFFSET(15899, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB),
+    NAME_FUNC_OFFSET(15925, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB),
+    NAME_FUNC_OFFSET(15951, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB),
+    NAME_FUNC_OFFSET(15977, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB),
+    NAME_FUNC_OFFSET(16001, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB),
+    NAME_FUNC_OFFSET(16028, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB),
+    NAME_FUNC_OFFSET(16054, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB),
+    NAME_FUNC_OFFSET(16074, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB),
+    NAME_FUNC_OFFSET(16094, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB),
+    NAME_FUNC_OFFSET(16114, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB),
+    NAME_FUNC_OFFSET(16137, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB),
+    NAME_FUNC_OFFSET(16161, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB),
+    NAME_FUNC_OFFSET(16184, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB),
+    NAME_FUNC_OFFSET(16208, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB),
+    NAME_FUNC_OFFSET(16225, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB),
+    NAME_FUNC_OFFSET(16243, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB),
+    NAME_FUNC_OFFSET(16260, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB),
+    NAME_FUNC_OFFSET(16278, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB),
+    NAME_FUNC_OFFSET(16295, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB),
+    NAME_FUNC_OFFSET(16313, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB),
+    NAME_FUNC_OFFSET(16330, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB),
+    NAME_FUNC_OFFSET(16348, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB),
+    NAME_FUNC_OFFSET(16365, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB),
+    NAME_FUNC_OFFSET(16383, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB),
+    NAME_FUNC_OFFSET(16400, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB),
+    NAME_FUNC_OFFSET(16418, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB),
+    NAME_FUNC_OFFSET(16435, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB),
+    NAME_FUNC_OFFSET(16453, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB),
+    NAME_FUNC_OFFSET(16470, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB),
+    NAME_FUNC_OFFSET(16488, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB),
+    NAME_FUNC_OFFSET(16505, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB),
+    NAME_FUNC_OFFSET(16523, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB),
+    NAME_FUNC_OFFSET(16542, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB),
+    NAME_FUNC_OFFSET(16561, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB),
+    NAME_FUNC_OFFSET(16580, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB),
+    NAME_FUNC_OFFSET(16599, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB),
+    NAME_FUNC_OFFSET(16619, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB),
+    NAME_FUNC_OFFSET(16639, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB),
+    NAME_FUNC_OFFSET(16659, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB),
+    NAME_FUNC_OFFSET(16677, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB),
+    NAME_FUNC_OFFSET(16694, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB),
+    NAME_FUNC_OFFSET(16712, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB),
+    NAME_FUNC_OFFSET(16729, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB),
+    NAME_FUNC_OFFSET(16747, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB),
+    NAME_FUNC_OFFSET(16765, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB),
+    NAME_FUNC_OFFSET(16782, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB),
+    NAME_FUNC_OFFSET(16800, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB),
+    NAME_FUNC_OFFSET(16819, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB),
+    NAME_FUNC_OFFSET(16838, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB),
+    NAME_FUNC_OFFSET(16857, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB),
+    NAME_FUNC_OFFSET(16879, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB),
+    NAME_FUNC_OFFSET(16892, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB),
+    NAME_FUNC_OFFSET(16905, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB),
+    NAME_FUNC_OFFSET(16921, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB),
+    NAME_FUNC_OFFSET(16937, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB),
+    NAME_FUNC_OFFSET(16950, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB),
+    NAME_FUNC_OFFSET(16973, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB),
+    NAME_FUNC_OFFSET(16993, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB),
+    NAME_FUNC_OFFSET(17012, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB),
+    NAME_FUNC_OFFSET(17023, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB),
+    NAME_FUNC_OFFSET(17035, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB),
+    NAME_FUNC_OFFSET(17049, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB),
+    NAME_FUNC_OFFSET(17062, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB),
+    NAME_FUNC_OFFSET(17078, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB),
+    NAME_FUNC_OFFSET(17089, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB),
+    NAME_FUNC_OFFSET(17102, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB),
+    NAME_FUNC_OFFSET(17121, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB),
+    NAME_FUNC_OFFSET(17141, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB),
+    NAME_FUNC_OFFSET(17154, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB),
+    NAME_FUNC_OFFSET(17164, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB),
+    NAME_FUNC_OFFSET(17180, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB),
+    NAME_FUNC_OFFSET(17199, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB),
+    NAME_FUNC_OFFSET(17217, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB),
+    NAME_FUNC_OFFSET(17238, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB),
+    NAME_FUNC_OFFSET(17253, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB),
+    NAME_FUNC_OFFSET(17268, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB),
+    NAME_FUNC_OFFSET(17282, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB),
+    NAME_FUNC_OFFSET(17297, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB),
+    NAME_FUNC_OFFSET(17309, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB),
+    NAME_FUNC_OFFSET(17322, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB),
+    NAME_FUNC_OFFSET(17334, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB),
+    NAME_FUNC_OFFSET(17347, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB),
+    NAME_FUNC_OFFSET(17359, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB),
+    NAME_FUNC_OFFSET(17372, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB),
+    NAME_FUNC_OFFSET(17384, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB),
+    NAME_FUNC_OFFSET(17397, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB),
+    NAME_FUNC_OFFSET(17409, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB),
+    NAME_FUNC_OFFSET(17422, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB),
+    NAME_FUNC_OFFSET(17434, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB),
+    NAME_FUNC_OFFSET(17447, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB),
+    NAME_FUNC_OFFSET(17459, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB),
+    NAME_FUNC_OFFSET(17472, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB),
+    NAME_FUNC_OFFSET(17484, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB),
+    NAME_FUNC_OFFSET(17497, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB),
+    NAME_FUNC_OFFSET(17516, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB),
+    NAME_FUNC_OFFSET(17535, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB),
+    NAME_FUNC_OFFSET(17554, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB),
+    NAME_FUNC_OFFSET(17567, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB),
+    NAME_FUNC_OFFSET(17585, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB),
+    NAME_FUNC_OFFSET(17606, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
+    NAME_FUNC_OFFSET(17624, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
+    NAME_FUNC_OFFSET(17644, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
+    NAME_FUNC_OFFSET(17658, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
+    NAME_FUNC_OFFSET(17675, gl_dispatch_stub_574, gl_dispatch_stub_574, NULL, _gloffset_SampleMaskSGIS),
+    NAME_FUNC_OFFSET(17691, gl_dispatch_stub_575, gl_dispatch_stub_575, NULL, _gloffset_SamplePatternSGIS),
+    NAME_FUNC_OFFSET(17710, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17728, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17749, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17771, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17790, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17812, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17835, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
+    NAME_FUNC_OFFSET(17854, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
+    NAME_FUNC_OFFSET(17874, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
+    NAME_FUNC_OFFSET(17893, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
+    NAME_FUNC_OFFSET(17913, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
+    NAME_FUNC_OFFSET(17932, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
+    NAME_FUNC_OFFSET(17952, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
+    NAME_FUNC_OFFSET(17971, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
+    NAME_FUNC_OFFSET(17991, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
+    NAME_FUNC_OFFSET(18010, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
+    NAME_FUNC_OFFSET(18030, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
+    NAME_FUNC_OFFSET(18050, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
+    NAME_FUNC_OFFSET(18071, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
+    NAME_FUNC_OFFSET(18091, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
+    NAME_FUNC_OFFSET(18112, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
+    NAME_FUNC_OFFSET(18132, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
+    NAME_FUNC_OFFSET(18153, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
+    NAME_FUNC_OFFSET(18177, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
+    NAME_FUNC_OFFSET(18195, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
+    NAME_FUNC_OFFSET(18215, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
+    NAME_FUNC_OFFSET(18233, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
+    NAME_FUNC_OFFSET(18245, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
+    NAME_FUNC_OFFSET(18258, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
+    NAME_FUNC_OFFSET(18270, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
+    NAME_FUNC_OFFSET(18283, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(18303, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(18327, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(18341, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(18358, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(18373, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(18391, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(18405, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(18422, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(18437, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(18455, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(18469, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(18486, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(18501, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(18519, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(18533, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(18550, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(18565, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(18583, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(18597, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(18614, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(18629, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(18647, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(18661, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(18678, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(18693, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(18711, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(18725, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(18742, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(18757, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(18775, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(18789, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(18806, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(18821, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(18839, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
+    NAME_FUNC_OFFSET(18856, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
+    NAME_FUNC_OFFSET(18876, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
+    NAME_FUNC_OFFSET(18893, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(18919, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(18948, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(18963, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(18981, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(19000, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_DeleteVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(19021, gl_dispatch_stub_747, gl_dispatch_stub_747, NULL, _gloffset_IsVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(19037, gl_dispatch_stub_755, gl_dispatch_stub_755, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(19061, gl_dispatch_stub_755, gl_dispatch_stub_755, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(19088, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
+    NAME_FUNC_OFFSET(19106, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
+    NAME_FUNC_OFFSET(19125, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
+    NAME_FUNC_OFFSET(19150, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
+    NAME_FUNC_OFFSET(19171, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
+    NAME_FUNC_OFFSET(19193, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
+    NAME_FUNC_OFFSET(19219, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
+    NAME_FUNC_OFFSET(19242, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
+    NAME_FUNC_OFFSET(19265, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
+    NAME_FUNC_OFFSET(19288, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
+    NAME_FUNC_OFFSET(19306, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
+    NAME_FUNC_OFFSET(19325, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
+    NAME_FUNC_OFFSET(19342, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
+    NAME_FUNC_OFFSET(19380, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
+    NAME_FUNC_OFFSET(19409, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
+    NAME_FUNC_OFFSET(19425, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
+    NAME_FUNC_OFFSET(19442, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
+    NAME_FUNC_OFFSET(19464, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_BlitFramebufferEXT),
+    NAME_FUNC_OFFSET(19482, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
     NAME_FUNC_OFFSET(-1, NULL, NULL, NULL, 0)
 };
 
diff --git a/src/mesa/main/api_arrayelt.c b/src/mesa/main/api_arrayelt.c
index d124c724c9..f5b7d1e138 100644
--- a/src/mesa/main/api_arrayelt.c
+++ b/src/mesa/main/api_arrayelt.c
@@ -1094,48 +1094,49 @@ static void _ae_update_state( GLcontext *ctx )
    AEarray *aa = actx->arrays;
    AEattrib *at = actx->attribs;
    GLuint i;
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
 
    actx->nr_vbos = 0;
 
    /* conventional vertex arrays */
-  if (ctx->Array.ArrayObj->Index.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Index;
+   if (arrayObj->Index.Enabled) {
+      aa->array = &arrayObj->Index;
       aa->offset = IndexFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->EdgeFlag;
+   if (arrayObj->EdgeFlag.Enabled) {
+      aa->array = &arrayObj->EdgeFlag;
       aa->offset = _gloffset_EdgeFlagv;
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->Normal.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Normal;
+   if (arrayObj->Normal.Enabled) {
+      aa->array = &arrayObj->Normal;
       aa->offset = NormalFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->Color.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Color;
+   if (arrayObj->Color.Enabled) {
+      aa->array = &arrayObj->Color;
       aa->offset = ColorFuncs[aa->array->Size-3][TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->SecondaryColor;
+   if (arrayObj->SecondaryColor.Enabled) {
+      aa->array = &arrayObj->SecondaryColor;
       aa->offset = SecondaryColorFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->FogCoord;
+   if (arrayObj->FogCoord.Enabled) {
+      aa->array = &arrayObj->FogCoord;
       aa->offset = FogCoordFuncs[TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
    for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      struct gl_client_array *attribArray = &ctx->Array.ArrayObj->TexCoord[i];
+      struct gl_client_array *attribArray = &arrayObj->TexCoord[i];
       if (attribArray->Enabled) {
          /* NOTE: we use generic glVertexAttribNV functions here.
           * If we ever remove GL_NV_vertex_program this will have to change.
@@ -1152,8 +1153,8 @@ static void _ae_update_state( GLcontext *ctx )
    }
 
    /* generic vertex attribute arrays */   
-   for (i = 1; i < VERT_ATTRIB_MAX; i++) {  /* skip zero! */
-      struct gl_client_array *attribArray = &ctx->Array.ArrayObj->VertexAttrib[i];
+   for (i = 1; i < Elements(arrayObj->VertexAttrib); i++) {  /* skip zero! */
+      struct gl_client_array *attribArray = &arrayObj->VertexAttrib[i];
       if (attribArray->Enabled) {
          at->array = attribArray;
          /* Note: we can't grab the _glapi_Dispatch->VertexAttrib1fvNV
@@ -1179,18 +1180,18 @@ static void _ae_update_state( GLcontext *ctx )
    }
 
    /* finally, vertex position */
-   if (ctx->Array.ArrayObj->VertexAttrib[0].Enabled) {
+   if (arrayObj->VertexAttrib[0].Enabled) {
       /* Use glVertex(v) instead of glVertexAttrib(0, v) to be sure it's
        * issued as the last (provoking) attribute).
        */
-      aa->array = &ctx->Array.ArrayObj->VertexAttrib[0];
+      aa->array = &arrayObj->VertexAttrib[0];
       assert(aa->array->Size >= 2); /* XXX fix someday? */
       aa->offset = VertexFuncs[aa->array->Size-2][TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
    }
-   else if (ctx->Array.ArrayObj->Vertex.Enabled) {
-      aa->array = &ctx->Array.ArrayObj->Vertex;
+   else if (arrayObj->Vertex.Enabled) {
+      aa->array = &arrayObj->Vertex;
       aa->offset = VertexFuncs[aa->array->Size-2][TYPE_IDX(aa->array->Type)];
       check_vbo(actx, aa->array->BufferObj);
       aa++;
diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c
index 6f66ff47a0..e49cd041a6 100644
--- a/src/mesa/main/api_exec.c
+++ b/src/mesa/main/api_exec.c
@@ -641,6 +641,8 @@ _mesa_init_exec_table(struct _glapi_table *exec)
    /* ???. GL_EXT_depth_bounds_test */
    SET_DepthBoundsEXT(exec, _mesa_DepthBoundsEXT);
 
+   SET_ProvokingVertexEXT(exec, _mesa_ProvokingVertexEXT);
+
    /* ARB 1. GL_ARB_multitexture */
 #if _HAVE_FULL_GL
    SET_ActiveTextureARB(exec, _mesa_ActiveTextureARB);
@@ -894,5 +896,16 @@ _mesa_init_exec_table(struct _glapi_table *exec)
     */
    SET_RenderbufferStorageMultisample(exec, _mesa_RenderbufferStorageMultisample);
 #endif
-}
 
+#if FEATURE_ARB_map_buffer_range
+   SET_MapBufferRange(exec, _mesa_MapBufferRange);
+   SET_FlushMappedBufferRange(exec, _mesa_FlushMappedBufferRange);
+#endif
+
+   /* GL_ARB_copy_buffer */
+   SET_CopyBufferSubData(exec, _mesa_CopyBufferSubData);
+
+   /* GL_ARB_vertex_array_object */
+   SET_BindVertexArray(exec, _mesa_BindVertexArray);
+   SET_GenVertexArrays(exec, _mesa_GenVertexArrays);
+}
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index 2c6f370df9..b2f11ffbfe 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -30,6 +30,26 @@
 #include "state.h"
 
 
+
+/**
+ * \return  number of bytes in array [count] of type.
+ */
+static GLsizei
+index_bytes(GLenum type, GLsizei count)
+{
+   if (type == GL_UNSIGNED_INT) {
+      return count * sizeof(GLuint);
+   }
+   else if (type == GL_UNSIGNED_BYTE) {
+      return count * sizeof(GLubyte);
+   }
+   else {
+      ASSERT(type == GL_UNSIGNED_SHORT);
+      return count * sizeof(GLushort);
+   }
+}
+
+
 /**
  * Find the max index in the given element/index buffer
  */
@@ -44,10 +64,8 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
 
    if (elementBuf->Name) {
       /* elements are in a user-defined buffer object.  need to map it */
-      map = ctx->Driver.MapBuffer(ctx,
-                                  GL_ELEMENT_ARRAY_BUFFER_ARB,
-                                  GL_READ_ONLY,
-                                  elementBuf);
+      map = ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER,
+                                  GL_READ_ONLY, elementBuf);
       /* Actual address is the sum of pointers */
       indices = (const GLvoid *) ADD_POINTERS(map, (const GLubyte *) indices);
    }
@@ -70,14 +88,16 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
    }
 
    if (map) {
-      ctx->Driver.UnmapBuffer(ctx,
-                              GL_ELEMENT_ARRAY_BUFFER_ARB,
-                              ctx->Array.ElementArrayBufferObj);
+      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB, elementBuf);
    }
 
    return max;
 }
 
+
+/**
+ * Check if OK to render by examining framebuffer status and vertex arrays.
+ */
 static GLboolean
 check_valid_to_render(GLcontext *ctx, char *function)
 {
@@ -105,6 +125,12 @@ check_valid_to_render(GLcontext *ctx, char *function)
    return GL_TRUE;
 }
 
+
+/**
+ * Error checking for glDrawElements().  Includes parameter checking
+ * and VBO bounds checking.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
+ */
 GLboolean
 _mesa_validate_DrawElements(GLcontext *ctx,
 			    GLenum mode, GLsizei count, GLenum type,
@@ -140,27 +166,8 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    /* Vertex buffer object tests */
    if (ctx->Array.ElementArrayBufferObj->Name) {
       /* use indices in the buffer object */
-      GLuint indexBytes;
-
-      if (!ctx->Array.ElementArrayBufferObj->Size) {
-         _mesa_warning(ctx,
-                       "glDrawElements called with empty array elements buffer");
-         return GL_FALSE;
-      }
-
-      if (type == GL_UNSIGNED_INT) {
-         indexBytes = count * sizeof(GLuint);
-      }
-      else if (type == GL_UNSIGNED_BYTE) {
-         indexBytes = count * sizeof(GLubyte);
-      }
-      else {
-         ASSERT(type == GL_UNSIGNED_SHORT);
-         indexBytes = count * sizeof(GLushort);
-      }
-
       /* make sure count doesn't go outside buffer bounds */
-      if (indexBytes > (GLuint) ctx->Array.ElementArrayBufferObj->Size) {
+      if (index_bytes(type, count) > ctx->Array.ElementArrayBufferObj->Size) {
          _mesa_warning(ctx, "glDrawElements index out of buffer bounds");
          return GL_FALSE;
       }
@@ -175,8 +182,10 @@ _mesa_validate_DrawElements(GLcontext *ctx,
       /* find max array index */
       GLuint max = max_buffer_index(ctx, count, type, indices,
                                     ctx->Array.ElementArrayBufferObj);
-      if (max >= ctx->Array._MaxElement) {
+      if (max >= ctx->Array.ArrayObj->_MaxElement) {
          /* the max element is out of bounds of one or more enabled arrays */
+         _mesa_warning(ctx, "glDrawElements() index=%u is "
+                       "out of bounds (max=%u)", max, ctx->Array.ArrayObj->_MaxElement);
          return GL_FALSE;
       }
    }
@@ -184,6 +193,12 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    return GL_TRUE;
 }
 
+
+/**
+ * Error checking for glDrawRangeElements().  Includes parameter checking
+ * and VBO bounds checking.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
+ */
 GLboolean
 _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 				 GLuint start, GLuint end,
@@ -224,21 +239,8 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
    /* Vertex buffer object tests */
    if (ctx->Array.ElementArrayBufferObj->Name) {
       /* use indices in the buffer object */
-      GLsizei indexBytes;
-
-      if (type == GL_UNSIGNED_INT) {
-         indexBytes = count * sizeof(GLuint);
-      }
-      else if (type == GL_UNSIGNED_BYTE) {
-         indexBytes = count * sizeof(GLubyte);
-      }
-      else {
-         ASSERT(type == GL_UNSIGNED_SHORT);
-         indexBytes = count * sizeof(GLushort);
-      }
-
       /* make sure count doesn't go outside buffer bounds */
-      if (indexBytes > ctx->Array.ElementArrayBufferObj->Size) {
+      if (index_bytes(type, count) > ctx->Array.ElementArrayBufferObj->Size) {
          _mesa_warning(ctx, "glDrawRangeElements index out of buffer bounds");
          return GL_FALSE;
       }
@@ -252,7 +254,7 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
    if (ctx->Const.CheckArrayBounds) {
       GLuint max = max_buffer_index(ctx, count, type, indices,
                                     ctx->Array.ElementArrayBufferObj);
-      if (max >= ctx->Array._MaxElement) {
+      if (max >= ctx->Array.ArrayObj->_MaxElement) {
          /* the max element is out of bounds of one or more enabled arrays */
          return GL_FALSE;
       }
@@ -265,6 +267,7 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 /**
  * Called from the tnl module to error check the function parameters and
  * verify that we really can draw something.
+ * \return GL_TRUE if OK to render, GL_FALSE if error found
  */
 GLboolean
 _mesa_validate_DrawArrays(GLcontext *ctx,
@@ -290,7 +293,7 @@ _mesa_validate_DrawArrays(GLcontext *ctx,
       return GL_FALSE;
 
    if (ctx->Const.CheckArrayBounds) {
-      if (start + count > (GLint) ctx->Array._MaxElement)
+      if (start + count > (GLint) ctx->Array.ArrayObj->_MaxElement)
          return GL_FALSE;
    }
 
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index 0fa5f0de55..fd35d4e38c 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -1,9 +1,10 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  * (C) Copyright IBM Corporation 2006
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -46,6 +47,7 @@
 #include "bufferobj.h"
 #endif
 #include "arrayobj.h"
+#include "macros.h"
 #include "glapi/dispatch.h"
 
 
@@ -61,10 +63,11 @@
 static INLINE struct gl_array_object *
 lookup_arrayobj(GLcontext *ctx, GLuint id)
 {
-   return (id == 0) 
-     ? NULL 
-     : (struct gl_array_object *) _mesa_HashLookup(ctx->Shared->ArrayObjects,
-						   id);
+   if (id == 0)
+      return NULL;
+   else
+      return (struct gl_array_object *)
+         _mesa_HashLookup(ctx->Array.Objects, id);
 }
 
 
@@ -79,6 +82,7 @@ unbind_array_object_vbos(GLcontext *ctx, struct gl_array_object *obj)
    GLuint i;
 
    _mesa_reference_buffer_object(ctx, &obj->Vertex.BufferObj, NULL);
+   _mesa_reference_buffer_object(ctx, &obj->Weight.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->Normal.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->Color.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->SecondaryColor.BufferObj, NULL);
@@ -86,10 +90,10 @@ unbind_array_object_vbos(GLcontext *ctx, struct gl_array_object *obj)
    _mesa_reference_buffer_object(ctx, &obj->Index.BufferObj, NULL);
    _mesa_reference_buffer_object(ctx, &obj->EdgeFlag.BufferObj, NULL);
 
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
+   for (i = 0; i < Elements(obj->TexCoord); i++)
       _mesa_reference_buffer_object(ctx, &obj->TexCoord[i].BufferObj, NULL);
 
-   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+   for (i = 0; i < Elements(obj->VertexAttrib); i++)
       _mesa_reference_buffer_object(ctx, &obj->VertexAttrib[i].BufferObj,NULL);
 }
 
@@ -198,7 +202,8 @@ init_array(GLcontext *ctx,
    array->Normalized = GL_FALSE;
 #if FEATURE_ARB_vertex_buffer_object
    /* Vertex array buffers */
-   array->BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &array->BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 }
 
@@ -220,16 +225,17 @@ _mesa_initialize_array_object( GLcontext *ctx,
 
    /* Init the individual arrays */
    init_array(ctx, &obj->Vertex, 4, GL_FLOAT);
+   init_array(ctx, &obj->Weight, 1, GL_FLOAT);
    init_array(ctx, &obj->Normal, 3, GL_FLOAT);
    init_array(ctx, &obj->Color, 4, GL_FLOAT);
    init_array(ctx, &obj->SecondaryColor, 4, GL_FLOAT);
    init_array(ctx, &obj->FogCoord, 1, GL_FLOAT);
    init_array(ctx, &obj->Index, 1, GL_FLOAT);
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+   for (i = 0; i < Elements(obj->TexCoord); i++) {
       init_array(ctx, &obj->TexCoord[i], 4, GL_FLOAT);
    }
    init_array(ctx, &obj->EdgeFlag, 1, GL_BOOL);
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(obj->VertexAttrib); i++) {
       init_array(ctx, &obj->VertexAttrib[i], 4, GL_FLOAT);
    }
 
@@ -242,12 +248,12 @@ _mesa_initialize_array_object( GLcontext *ctx,
 /**
  * Add the given array object to the array object pool.
  */
-void
-_mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj )
+static void
+save_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    if (obj->Name > 0) {
       /* insert into hash table */
-      _mesa_HashInsert(ctx->Shared->ArrayObjects, obj->Name, obj);
+      _mesa_HashInsert(ctx->Array.Objects, obj->Name, obj);
    }
 }
 
@@ -256,13 +262,90 @@ _mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj )
  * Remove the given array object from the array object pool.
  * Do not deallocate the array object though.
  */
-void
-_mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
+static void
+remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 {
    if (obj->Name > 0) {
       /* remove from hash table */
-      _mesa_HashRemove(ctx->Shared->ArrayObjects, obj->Name);
+      _mesa_HashRemove(ctx->Array.Objects, obj->Name);
+   }
+}
+
+
+
+/**
+ * Compute the index of the last array element that can be safely accessed
+ * in a vertex array.  We can really only do this when the array lives in
+ * a VBO.
+ * The array->_MaxElement field will be updated.
+ * Later in glDrawArrays/Elements/etc we can do some bounds checking.
+ */
+static void
+compute_max_element(struct gl_client_array *array)
+{
+   if (array->BufferObj->Name) {
+      /* Compute the max element we can access in the VBO without going
+       * out of bounds.
+       */
+      array->_MaxElement = ((GLsizeiptrARB) array->BufferObj->Size
+                            - (GLsizeiptrARB) array->Ptr + array->StrideB
+                            - array->_ElementSize) / array->StrideB;
+      if (0)
+         _mesa_printf("%s Object %u  Size %u  MaxElement %u\n",
+                      __FUNCTION__,
+                      array->BufferObj->Name,
+                      (GLuint) array->BufferObj->Size,
+                      array->_MaxElement);
    }
+   else {
+      /* user-space array, no idea how big it is */
+      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
+   }
+}
+
+
+/**
+ * Helper for update_arrays().
+ * \return  min(current min, array->_MaxElement).
+ */
+static GLuint
+update_min(GLuint min, struct gl_client_array *array)
+{
+   compute_max_element(array);
+   if (array->Enabled)
+      return MIN2(min, array->_MaxElement);
+   else
+      return min;
+}
+
+
+/**
+ * Examine vertex arrays to update the gl_array_object::_MaxElement field.
+ */
+void
+_mesa_update_array_object_max_element(GLcontext *ctx,
+                                      struct gl_array_object *arrayObj)
+{
+   GLuint i, min = ~0;
+
+   min = update_min(min, &arrayObj->Vertex);
+   min = update_min(min, &arrayObj->Weight);
+   min = update_min(min, &arrayObj->Normal);
+   min = update_min(min, &arrayObj->Color);
+   min = update_min(min, &arrayObj->SecondaryColor);
+   min = update_min(min, &arrayObj->FogCoord);
+   min = update_min(min, &arrayObj->Index);
+   min = update_min(min, &arrayObj->EdgeFlag);
+#if FEATURE_point_size_array
+   min = update_min(min, &arrayObj->PointSize);
+#endif
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++)
+      min = update_min(min, &arrayObj->TexCoord[i]);
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
+      min = update_min(min, &arrayObj->VertexAttrib[i]);
+
+   /* _MaxElement is one past the last legal array element */
+   arrayObj->_MaxElement = min;
 }
 
 
@@ -270,18 +353,15 @@ _mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 /* API Functions                                                      */
 /**********************************************************************/
 
+
 /**
- * Bind a new array.
- *
- * \todo
- * The binding could be done more efficiently by comparing the non-NULL
- * pointers in the old and new objects.  The only arrays that are "dirty" are
- * the ones that are non-NULL in either object.
+ * Helper for _mesa_BindVertexArray() and _mesa_BindVertexArrayAPPLE().
+ * \param genRequired  specifies behavour when id was not generated with
+ *                     glGenVertexArrays().
  */
-void GLAPIENTRY
-_mesa_BindVertexArrayAPPLE( GLuint id )
+static void
+bind_vertex_array(GLcontext *ctx, GLuint id, GLboolean genRequired)
 {
-   GET_CURRENT_CONTEXT(ctx);
    struct gl_array_object * const oldObj = ctx->Array.ArrayObj;
    struct gl_array_object *newObj = NULL;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
@@ -304,14 +384,18 @@ _mesa_BindVertexArrayAPPLE( GLuint id )
       /* non-default array object */
       newObj = lookup_arrayobj(ctx, id);
       if (!newObj) {
-         /* If this is a new array object id, allocate an array object now.
-	  */
+         if (genRequired) {
+            _mesa_error(ctx, GL_INVALID_OPERATION, "glBindVertexArray(id)");
+            return;
+         }
+
+         /* For APPLE version, generate a new array object now */
 	 newObj = (*ctx->Driver.NewArrayObject)(ctx, id);
          if (!newObj) {
             _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindVertexArrayAPPLE");
             return;
          }
-         _mesa_save_array_object(ctx, newObj);
+         save_array_object(ctx, newObj);
       }
    }
 
@@ -321,7 +405,37 @@ _mesa_BindVertexArrayAPPLE( GLuint id )
 
    /* Pass BindVertexArray call to device driver */
    if (ctx->Driver.BindArrayObject && newObj)
-      (*ctx->Driver.BindArrayObject)( ctx, newObj );
+      ctx->Driver.BindArrayObject(ctx, newObj);
+}
+
+
+/**
+ * ARB version of glBindVertexArray()
+ * This function behaves differently from glBindVertexArrayAPPLE() in
+ * that this function requires all ids to have been previously generated
+ * by glGenVertexArrays[APPLE]().
+ */
+void GLAPIENTRY
+_mesa_BindVertexArray( GLuint id )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   bind_vertex_array(ctx, id, GL_TRUE);
+}
+
+
+/**
+ * Bind a new array.
+ *
+ * \todo
+ * The binding could be done more efficiently by comparing the non-NULL
+ * pointers in the old and new objects.  The only arrays that are "dirty" are
+ * the ones that are non-NULL in either object.
+ */
+void GLAPIENTRY
+_mesa_BindVertexArrayAPPLE( GLuint id )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   bind_vertex_array(ctx, id, GL_FALSE);
 }
 
 
@@ -343,8 +457,6 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
       return;
    }
 
-   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
-
    for (i = 0; i < n; i++) {
       struct gl_array_object *obj = lookup_arrayobj(ctx, ids[i]);
 
@@ -360,7 +472,7 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
 	 }
 
 	 /* The ID is immediately freed for re-use */
-	 _mesa_remove_array_object(ctx, obj);
+	 remove_array_object(ctx, obj);
 
          /* Unreference the array object. 
           * If refcount hits zero, the object will be deleted.
@@ -368,21 +480,19 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
          _mesa_reference_array_object(ctx, &obj, NULL);
       }
    }
-
-   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
 }
 
 
 /**
  * Generate a set of unique array object IDs and store them in \c arrays.
- * 
+ * Helper for _mesa_GenVertexArrays[APPLE]() functions below.
  * \param n       Number of IDs to generate.
  * \param arrays  Array of \c n locations to store the IDs.
+ * \param vboOnly Will arrays have to reside in VBOs?
  */
-void GLAPIENTRY
-_mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
+static void 
+gen_vertex_arrays(GLcontext *ctx, GLsizei n, GLuint *arrays, GLboolean vboOnly)
 {
-   GET_CURRENT_CONTEXT(ctx);
    GLuint first;
    GLint i;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
@@ -396,12 +506,7 @@ _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
       return;
    }
 
-   /*
-    * This must be atomic (generation and allocation of array object IDs)
-    */
-   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
-
-   first = _mesa_HashFindFreeKeyBlock(ctx->Shared->ArrayObjects, n);
+   first = _mesa_HashFindFreeKeyBlock(ctx->Array.Objects, n);
 
    /* Allocate new, empty array objects and return identifiers */
    for (i = 0; i < n; i++) {
@@ -410,15 +515,37 @@ _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
 
       obj = (*ctx->Driver.NewArrayObject)( ctx, name );
       if (!obj) {
-         _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
          _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenVertexArraysAPPLE");
          return;
       }
-      _mesa_save_array_object(ctx, obj);
+      obj->VBOonly = vboOnly;
+      save_array_object(ctx, obj);
       arrays[i] = first + i;
    }
+}
+
+
+/**
+ * ARB version of glGenVertexArrays()
+ * All arrays will be required to live in VBOs.
+ */
+void GLAPIENTRY
+_mesa_GenVertexArrays(GLsizei n, GLuint *arrays)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   gen_vertex_arrays(ctx, n, arrays, GL_TRUE);
+}
+
 
-   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
+/**
+ * APPLE version of glGenVertexArraysAPPLE()
+ * Arrays may live in VBOs or ordinary memory.
+ */
+void GLAPIENTRY
+_mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *arrays)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   gen_vertex_arrays(ctx, n, arrays, GL_FALSE);
 }
 
 
@@ -439,9 +566,7 @@ _mesa_IsVertexArrayAPPLE( GLuint id )
    if (id == 0)
       return GL_FALSE;
 
-   _glthread_LOCK_MUTEX(ctx->Shared->Mutex);
    obj = lookup_arrayobj(ctx, id);
-   _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex);
 
    return (obj != NULL) ? GL_TRUE : GL_FALSE;
 }
diff --git a/src/mesa/main/arrayobj.h b/src/mesa/main/arrayobj.h
index 9c4036af5a..8999edc724 100644
--- a/src/mesa/main/arrayobj.h
+++ b/src/mesa/main/arrayobj.h
@@ -1,9 +1,10 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
  * (C) Copyright IBM Corporation 2006
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -56,22 +57,25 @@ extern void
 _mesa_initialize_array_object( GLcontext *ctx,
                                struct gl_array_object *obj, GLuint name );
 
-extern void
-_mesa_save_array_object( GLcontext *ctx, struct gl_array_object *obj );
 
 extern void
-_mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj );
-
+_mesa_update_array_object_max_element(GLcontext *ctx,
+                                      struct gl_array_object *arrayObj);
 
 
 /*
  * API functions
  */
 
+
+void GLAPIENTRY _mesa_BindVertexArray( GLuint id );
+
 void GLAPIENTRY _mesa_BindVertexArrayAPPLE( GLuint id );
 
 void GLAPIENTRY _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids);
 
+void GLAPIENTRY _mesa_GenVertexArrays(GLsizei n, GLuint *arrays);
+
 void GLAPIENTRY _mesa_GenVertexArraysAPPLE(GLsizei n, GLuint *buffer);
 
 GLboolean GLAPIENTRY _mesa_IsVertexArrayAPPLE( GLuint id );
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index 4d85d54bb9..cb49c4cb07 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -1324,20 +1324,22 @@ _mesa_PopAttrib(void)
  * counts when pushing/popping the GL_CLIENT_VERTEX_ARRAY_BIT attribute group.
  */
 static void
-adjust_buffer_object_ref_counts(struct gl_array_attrib *array, GLint step)
+adjust_buffer_object_ref_counts(struct gl_array_object *arrayObj, GLint step)
 {
    GLuint i;
-   array->ArrayObj->Vertex.BufferObj->RefCount += step;
-   array->ArrayObj->Normal.BufferObj->RefCount += step;
-   array->ArrayObj->Color.BufferObj->RefCount += step;
-   array->ArrayObj->SecondaryColor.BufferObj->RefCount += step;
-   array->ArrayObj->FogCoord.BufferObj->RefCount += step;
-   array->ArrayObj->Index.BufferObj->RefCount += step;
-   array->ArrayObj->EdgeFlag.BufferObj->RefCount += step;
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
-      array->ArrayObj->TexCoord[i].BufferObj->RefCount += step;
-   for (i = 0; i < VERT_ATTRIB_MAX; i++)
-      array->ArrayObj->VertexAttrib[i].BufferObj->RefCount += step;
+
+   arrayObj->Vertex.BufferObj->RefCount += step;
+   arrayObj->Weight.BufferObj->RefCount += step;
+   arrayObj->Normal.BufferObj->RefCount += step;
+   arrayObj->Color.BufferObj->RefCount += step;
+   arrayObj->SecondaryColor.BufferObj->RefCount += step;
+   arrayObj->FogCoord.BufferObj->RefCount += step;
+   arrayObj->Index.BufferObj->RefCount += step;
+   arrayObj->EdgeFlag.BufferObj->RefCount += step;
+   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
+      arrayObj->TexCoord[i].BufferObj->RefCount += step;
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
+      arrayObj->VertexAttrib[i].BufferObj->RefCount += step;
 }
 
 
@@ -1428,7 +1430,7 @@ _mesa_PushClientAttrib(GLbitfield mask)
       newnode->next = head;
       head = newnode;
       /* bump reference counts on buffer objects */
-      adjust_buffer_object_ref_counts(&ctx->Array, 1);
+      adjust_buffer_object_ref_counts(ctx->Array.ArrayObj, 1);
    }
 
    ctx->ClientAttribStack[ctx->ClientAttribStackDepth] = head;
@@ -1478,7 +1480,7 @@ _mesa_PopClientAttrib(void)
 	    struct gl_array_attrib * data =
 	      (struct gl_array_attrib *) node->data;
 
-            adjust_buffer_object_ref_counts(&ctx->Array, -1);
+            adjust_buffer_object_ref_counts(ctx->Array.ArrayObj, -1);
 	 
             ctx->Array.ActiveTexture = data->ActiveTexture;
 	    if (data->LockCount != 0)
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index c8d160baa9..d640f5358e 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -1,9 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
- * Copyright (C) 1999-2009  VMware, Inc.  All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -39,10 +39,15 @@
 #include "bufferobj.h"
 
 
+/* Debug flags */
+/*#define VBO_DEBUG*/
+/*#define BOUNDS_CHECK*/
+
+
 #ifdef FEATURE_OES_mapbuffer
-#define DEFAULT_ACCESS GL_WRITE_ONLY;
+#define DEFAULT_ACCESS GL_MAP_WRITE_BIT
 #else
-#define DEFAULT_ACCESS GL_READ_WRITE;
+#define DEFAULT_ACCESS (GL_MAP_READ_BIT | GL_MAP_WRITE_BIT)
 #endif
 
 
@@ -73,6 +78,16 @@ get_buffer(GLcontext *ctx, GLenum target)
       case GL_PIXEL_UNPACK_BUFFER_EXT:
          bufObj = ctx->Unpack.BufferObj;
          break;
+      case GL_COPY_READ_BUFFER:
+         if (ctx->Extensions.ARB_copy_buffer) {
+            bufObj = ctx->CopyReadBuffer;
+         }
+         break;
+      case GL_COPY_WRITE_BUFFER:
+         if (ctx->Extensions.ARB_copy_buffer) {
+            bufObj = ctx->CopyWriteBuffer;
+         }
+         break;
       default:
          /* error must be recorded by caller */
          return NULL;
@@ -86,6 +101,24 @@ get_buffer(GLcontext *ctx, GLenum target)
 
 
 /**
+ * Convert a GLbitfield describing the mapped buffer access flags
+ * into one of GL_READ_WRITE, GL_READ_ONLY, or GL_WRITE_ONLY.
+ */
+static GLenum
+simplified_access_mode(GLbitfield access)
+{
+   const GLbitfield rwFlags = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT;
+   if ((access & rwFlags) == rwFlags)
+      return GL_READ_WRITE;
+   if ((access & GL_MAP_READ_BIT) == GL_MAP_READ_BIT)
+      return GL_READ_ONLY;
+   if ((access & GL_MAP_WRITE_BIT) == GL_MAP_WRITE_BIT)
+      return GL_WRITE_ONLY;
+   return GL_READ_WRITE; /* this should never happen, but no big deal */
+}
+
+
+/**
  * Tests the subdata range parameters and sets the GL error code for
  * \c glBufferSubDataARB and \c glGetBufferSubDataARB.
  *
@@ -147,7 +180,7 @@ buffer_object_subdata_range_good( GLcontext * ctx, GLenum target,
  * 
  * Default callback for the \c dd_function_table::NewBufferObject() hook.
  */
-struct gl_buffer_object *
+static struct gl_buffer_object *
 _mesa_new_buffer_object( GLcontext *ctx, GLuint name, GLenum target )
 {
    struct gl_buffer_object *obj;
@@ -165,7 +198,7 @@ _mesa_new_buffer_object( GLcontext *ctx, GLuint name, GLenum target )
  * 
  * Default callback for the \c dd_function_table::DeleteBuffer() hook.
  */
-void
+static void
 _mesa_delete_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
 {
    (void) ctx;
@@ -194,7 +227,7 @@ _mesa_reference_buffer_object(GLcontext *ctx,
       return;
 
    if (*ptr) {
-      /* Unreference the old texture */
+      /* Unreference the old buffer */
       GLboolean deleteFlag = GL_FALSE;
       struct gl_buffer_object *oldObj = *ptr;
 
@@ -227,7 +260,7 @@ _mesa_reference_buffer_object(GLcontext *ctx,
    ASSERT(!*ptr);
 
    if (bufObj) {
-      /* reference new texture */
+      /* reference new buffer */
       /*_glthread_LOCK_MUTEX(tex->Mutex);*/
       if (bufObj->RefCount == 0) {
          /* this buffer's being deleted (look just above) */
@@ -261,7 +294,7 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
    obj->RefCount = 1;
    obj->Name = name;
    obj->Usage = GL_STATIC_DRAW_ARB;
-   obj->Access = DEFAULT_ACCESS;
+   obj->AccessFlags = DEFAULT_ACCESS;
 }
 
 
@@ -283,7 +316,7 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
  *
  * \sa glBufferDataARB, dd_function_table::BufferData.
  */
-void
+static void
 _mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
 		   const GLvoid * data, GLenum usage,
 		   struct gl_buffer_object * bufObj )
@@ -322,7 +355,7 @@ _mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
  *
  * \sa glBufferSubDataARB, dd_function_table::BufferSubData.
  */
-void
+static void
 _mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
 		      GLsizeiptrARB size, const GLvoid * data,
 		      struct gl_buffer_object * bufObj )
@@ -355,7 +388,7 @@ _mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
  *
  * \sa glBufferGetSubDataARB, dd_function_table::GetBufferSubData.
  */
-void
+static void
 _mesa_buffer_get_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
 			  GLsizeiptrARB size, GLvoid * data,
 			  struct gl_buffer_object * bufObj )
@@ -382,14 +415,13 @@ _mesa_buffer_get_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
  *
  * \sa glMapBufferARB, dd_function_table::MapBuffer
  */
-void *
+static void *
 _mesa_buffer_map( GLcontext *ctx, GLenum target, GLenum access,
 		  struct gl_buffer_object *bufObj )
 {
    (void) ctx;
    (void) target;
    (void) access;
-   ASSERT(!bufObj->OnCard);
    /* Just return a direct pointer to the data */
    if (bufObj->Pointer) {
       /* already mapped! */
@@ -401,19 +433,55 @@ _mesa_buffer_map( GLcontext *ctx, GLenum target, GLenum access,
 
 
 /**
+ * Default fallback for \c dd_function_table::MapBufferRange().
+ * Called via glMapBufferRange().
+ */
+static void *
+_mesa_buffer_map_range( GLcontext *ctx, GLenum target, GLintptr offset,
+                        GLsizeiptr length, GLbitfield access,
+                        struct gl_buffer_object *bufObj )
+{
+   (void) ctx;
+   (void) target;
+   (void) access;
+   (void) length;
+   assert(!bufObj->Pointer);
+   /* Just return a direct pointer to the data */
+   return bufObj->Data + offset;
+}
+
+
+/**
+ * Default fallback for \c dd_function_table::FlushMappedBufferRange().
+ * Called via glFlushMappedBufferRange().
+ */
+static void
+_mesa_buffer_flush_mapped_range( GLcontext *ctx, GLenum target, 
+                                 GLintptr offset, GLsizeiptr length,
+                                 struct gl_buffer_object *obj )
+{
+   (void) ctx;
+   (void) target;
+   (void) offset;
+   (void) length;
+   (void) obj;
+   /* no-op */
+}
+
+
+/**
  * Default callback for \c dd_function_table::MapBuffer().
  *
  * The input parameters will have been already tested for errors.
  *
  * \sa glUnmapBufferARB, dd_function_table::UnmapBuffer
  */
-GLboolean
+static GLboolean
 _mesa_buffer_unmap( GLcontext *ctx, GLenum target,
                     struct gl_buffer_object *bufObj )
 {
    (void) ctx;
    (void) target;
-   ASSERT(!bufObj->OnCard);
    /* XXX we might assert here that bufObj->Pointer is non-null */
    bufObj->Pointer = NULL;
    return GL_TRUE;
@@ -421,21 +489,51 @@ _mesa_buffer_unmap( GLcontext *ctx, GLenum target,
 
 
 /**
+ * Default fallback for \c dd_function_table::CopyBufferSubData().
+ * Called via glCopyBuffserSubData().
+ */
+static void
+_mesa_copy_buffer_subdata(GLcontext *ctx,
+                          struct gl_buffer_object *src,
+                          struct gl_buffer_object *dst,
+                          GLintptr readOffset, GLintptr writeOffset,
+                          GLsizeiptr size)
+{
+   GLubyte *srcPtr, *dstPtr;
+
+   /* buffer should not already be mapped */
+   assert(!src->Pointer);
+   assert(!dst->Pointer);
+
+   srcPtr = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_COPY_READ_BUFFER,
+                                              GL_READ_ONLY, src);
+   dstPtr = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_COPY_WRITE_BUFFER,
+                                              GL_WRITE_ONLY, dst);
+
+   if (srcPtr && dstPtr)
+      _mesa_memcpy(dstPtr + writeOffset, srcPtr + readOffset, size);
+
+   ctx->Driver.UnmapBuffer(ctx, GL_COPY_READ_BUFFER, src);
+   ctx->Driver.UnmapBuffer(ctx, GL_COPY_WRITE_BUFFER, dst);
+}
+
+
+
+/**
  * Initialize the state associated with buffer objects
  */
 void
 _mesa_init_buffer_objects( GLcontext *ctx )
 {
-   /* Allocate the default buffer object and set refcount so high that
-    * it never gets deleted.
-    * XXX with recent/improved refcounting this may not longer be needed.
-    */
-   ctx->Array.NullBufferObj = _mesa_new_buffer_object(ctx, 0, 0);
-   if (ctx->Array.NullBufferObj)
-      ctx->Array.NullBufferObj->RefCount = 1000;
-
-   ctx->Array.ArrayBufferObj = ctx->Array.NullBufferObj;
-   ctx->Array.ElementArrayBufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Array.ArrayBufferObj,
+                                 ctx->Shared->NullBufferObj);
+   _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj,
+                                 ctx->Shared->NullBufferObj);
+
+   _mesa_reference_buffer_object(ctx, &ctx->CopyReadBuffer,
+                                 ctx->Shared->NullBufferObj);
+   _mesa_reference_buffer_object(ctx, &ctx->CopyWriteBuffer,
+                                 ctx->Shared->NullBufferObj);
 }
 
 
@@ -462,8 +560,22 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
    case GL_PIXEL_UNPACK_BUFFER_EXT:
       bindTarget = &ctx->Unpack.BufferObj;
       break;
+   case GL_COPY_READ_BUFFER:
+      if (ctx->Extensions.ARB_copy_buffer) {
+         bindTarget = &ctx->CopyReadBuffer;
+      }
+      break;
+   case GL_COPY_WRITE_BUFFER:
+      if (ctx->Extensions.ARB_copy_buffer) {
+         bindTarget = &ctx->CopyWriteBuffer;
+      }
+      break;
    default:
-      _mesa_error(ctx, GL_INVALID_ENUM, "glBindBufferARB(target)");
+      ; /* no-op / we'll hit the follow error test next */
+   }
+
+   if (!bindTarget) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glBindBufferARB(target 0x%x)");
       return;
    }
 
@@ -479,7 +591,7 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
       /* The spec says there's not a buffer object named 0, but we use
        * one internally because it simplifies things.
        */
-      newBufObj = ctx->Array.NullBufferObj;
+      newBufObj = ctx->Shared->NullBufferObj;
    }
    else {
       /* non-default buffer object */
@@ -500,7 +612,7 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
    _mesa_reference_buffer_object(ctx, bindTarget, newBufObj);
 
    /* Pass BindBuffer call to device driver */
-   if (ctx->Driver.BindBuffer && newBufObj)
+   if (ctx->Driver.BindBuffer)
       ctx->Driver.BindBuffer( ctx, target, newBufObj );
 }
 
@@ -746,11 +858,37 @@ unbind(GLcontext *ctx,
        struct gl_buffer_object *obj)
 {
    if (*ptr == obj) {
-      _mesa_reference_buffer_object(ctx, ptr, ctx->Array.NullBufferObj);
+      _mesa_reference_buffer_object(ctx, ptr, ctx->Shared->NullBufferObj);
    }
 }
 
 
+/**
+ * Plug default/fallback buffer object functions into the device
+ * driver hooks.
+ */
+void
+_mesa_init_buffer_object_functions(struct dd_function_table *driver)
+{
+   /* GL_ARB_vertex/pixel_buffer_object */
+   driver->NewBufferObject = _mesa_new_buffer_object;
+   driver->DeleteBuffer = _mesa_delete_buffer_object;
+   driver->BindBuffer = NULL;
+   driver->BufferData = _mesa_buffer_data;
+   driver->BufferSubData = _mesa_buffer_subdata;
+   driver->GetBufferSubData = _mesa_buffer_get_subdata;
+   driver->MapBuffer = _mesa_buffer_map;
+   driver->UnmapBuffer = _mesa_buffer_unmap;
+
+   /* GL_ARB_map_buffer_range */
+   driver->MapBufferRange = _mesa_buffer_map_range;
+   driver->FlushMappedBufferRange = _mesa_buffer_flush_mapped_range;
+
+   /* GL_ARB_copy_buffer */
+   driver->CopyBufferSubData = _mesa_copy_buffer_subdata;
+}
+
+
 
 /**********************************************************************/
 /* API Functions                                                      */
@@ -789,6 +927,7 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
    for (i = 0; i < n; i++) {
       struct gl_buffer_object *bufObj = _mesa_lookup_bufferobj(ctx, ids[i]);
       if (bufObj) {
+         struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
          GLuint j;
 
          ASSERT(bufObj->Name == ids[i]);
@@ -796,23 +935,24 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
          if (bufObj->Pointer) {
             /* if mapped, unmap it now */
             ctx->Driver.UnmapBuffer(ctx, 0, bufObj);
-            bufObj->Access = DEFAULT_ACCESS;
+            bufObj->AccessFlags = DEFAULT_ACCESS;
             bufObj->Pointer = NULL;
          }
 
          /* unbind any vertex pointers bound to this buffer */
-         unbind(ctx, &ctx->Array.ArrayObj->Vertex.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->Normal.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->Color.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->SecondaryColor.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->FogCoord.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->Index.BufferObj, bufObj);
-         unbind(ctx, &ctx->Array.ArrayObj->EdgeFlag.BufferObj, bufObj);
-         for (j = 0; j < MAX_TEXTURE_COORD_UNITS; j++) {
-            unbind(ctx, &ctx->Array.ArrayObj->TexCoord[j].BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Vertex.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Weight.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Normal.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Color.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->SecondaryColor.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->FogCoord.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->Index.BufferObj, bufObj);
+         unbind(ctx, &arrayObj->EdgeFlag.BufferObj, bufObj);
+         for (j = 0; j < Elements(arrayObj->TexCoord); j++) {
+            unbind(ctx, &arrayObj->TexCoord[j].BufferObj, bufObj);
          }
-         for (j = 0; j < VERT_ATTRIB_MAX; j++) {
-            unbind(ctx, &ctx->Array.ArrayObj->VertexAttrib[j].BufferObj, bufObj);
+         for (j = 0; j < Elements(arrayObj->VertexAttrib); j++) {
+            unbind(ctx, &arrayObj->VertexAttrib[j].BufferObj, bufObj);
          }
 
          if (ctx->Array.ArrayBufferObj == bufObj) {
@@ -947,19 +1087,31 @@ _mesa_BufferDataARB(GLenum target, GLsizeiptrARB size,
       return;
    }
    if (bufObj->Name == 0) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glBufferDataARB" );
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glBufferDataARB(buffer 0)" );
       return;
    }
    
    if (bufObj->Pointer) {
       /* Unmap the existing buffer.  We'll replace it now.  Not an error. */
       ctx->Driver.UnmapBuffer(ctx, target, bufObj);
-      bufObj->Access = DEFAULT_ACCESS;
+      bufObj->AccessFlags = DEFAULT_ACCESS;
       bufObj->Pointer = NULL;
    }  
 
+   FLUSH_VERTICES(ctx, _NEW_BUFFER_OBJECT);
+
    ASSERT(ctx->Driver.BufferData);
 
+   bufObj->Written = GL_TRUE;
+
+#ifdef VBO_DEBUG
+   _mesa_printf("glBufferDataARB(%u, sz %ld, from %p, usage 0x%x)\n",
+                bufObj->Name, size, data, usage);
+#endif
+
+#ifdef BOUNDS_CHECK
+   size += 100;
+#endif
    /* Give the buffer object to the driver!  <data> may be null! */
    ctx->Driver.BufferData( ctx, target, size, data, usage, bufObj );
 }
@@ -980,6 +1132,8 @@ _mesa_BufferSubDataARB(GLenum target, GLintptrARB offset,
       return;
    }
 
+   bufObj->Written = GL_TRUE;
+
    ASSERT(ctx->Driver.BufferSubData);
    ctx->Driver.BufferSubData( ctx, target, offset, size, data, bufObj );
 }
@@ -1010,13 +1164,18 @@ _mesa_MapBufferARB(GLenum target, GLenum access)
 {
    GET_CURRENT_CONTEXT(ctx);
    struct gl_buffer_object * bufObj;
+   GLbitfield accessFlags;
    ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, NULL);
 
    switch (access) {
       case GL_READ_ONLY_ARB:
+         accessFlags = GL_MAP_READ_BIT;
+         break;
       case GL_WRITE_ONLY_ARB:
+         accessFlags = GL_MAP_WRITE_BIT;
+         break;
       case GL_READ_WRITE_ARB:
-         /* OK */
+         accessFlags = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT;
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glMapBufferARB(access)");
@@ -1029,7 +1188,7 @@ _mesa_MapBufferARB(GLenum target, GLenum access)
       return NULL;
    }
    if (bufObj->Name == 0) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glMapBufferARB" );
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glMapBufferARB(buffer 0)" );
       return NULL;
    }
    if (bufObj->Pointer) {
@@ -1043,7 +1202,34 @@ _mesa_MapBufferARB(GLenum target, GLenum access)
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "glMapBufferARB(access)");
    }
 
-   bufObj->Access = access;
+   bufObj->AccessFlags = accessFlags;
+   bufObj->Offset = 0;
+   bufObj->Length = bufObj->Size;
+
+   if (access == GL_WRITE_ONLY_ARB || access == GL_READ_WRITE_ARB)
+      bufObj->Written = GL_TRUE;
+
+#ifdef VBO_DEBUG
+   _mesa_printf("glMapBufferARB(%u, sz %ld, access 0x%x)\n",
+                bufObj->Name, bufObj->Size, access);
+   if (access == GL_WRITE_ONLY_ARB) {
+      GLuint i;
+      GLubyte *b = (GLubyte *) bufObj->Pointer;
+      for (i = 0; i < bufObj->Size; i++)
+         b[i] = i & 0xff;
+   }
+#endif
+
+#ifdef BOUNDS_CHECK
+   {
+      GLubyte *buf = (GLubyte *) bufObj->Pointer;
+      GLuint i;
+      /* buffer is 100 bytes larger than requested, fill with magic value */
+      for (i = 0; i < 100; i++) {
+         buf[bufObj->Size - i - 1] = 123;
+      }
+   }
+#endif
 
    return bufObj->Pointer;
 }
@@ -1071,9 +1257,47 @@ _mesa_UnmapBufferARB(GLenum target)
       return GL_FALSE;
    }
 
+#ifdef BOUNDS_CHECK
+   if (bufObj->Access != GL_READ_ONLY_ARB) {
+      GLubyte *buf = (GLubyte *) bufObj->Pointer;
+      GLuint i;
+      /* check that last 100 bytes are still = magic value */
+      for (i = 0; i < 100; i++) {
+         GLuint pos = bufObj->Size - i - 1;
+         if (buf[pos] != 123) {
+            _mesa_warning(ctx, "Out of bounds buffer object write detected"
+                          " at position %d (value = %u)\n",
+                          pos, buf[pos]);
+         }
+      }
+   }
+#endif
+
+#ifdef VBO_DEBUG
+   if (bufObj->AccessFlags & GL_MAP_WRITE_BIT) {
+      GLuint i, unchanged = 0;
+      GLubyte *b = (GLubyte *) bufObj->Pointer;
+      GLint pos = -1;
+      /* check which bytes changed */
+      for (i = 0; i < bufObj->Size - 1; i++) {
+         if (b[i] == (i & 0xff) && b[i+1] == ((i+1) & 0xff)) {
+            unchanged++;
+            if (pos == -1)
+               pos = i;
+         }
+      }
+      if (unchanged) {
+         _mesa_printf("glUnmapBufferARB(%u): %u of %ld unchanged, starting at %d\n",
+                      bufObj->Name, unchanged, bufObj->Size, pos);
+      }
+   }
+#endif
+
    status = ctx->Driver.UnmapBuffer( ctx, target, bufObj );
-   bufObj->Access = DEFAULT_ACCESS;
+   bufObj->AccessFlags = DEFAULT_ACCESS;
    bufObj->Pointer = NULL;
+   bufObj->Offset = 0;
+   bufObj->Length = 0;
 
    return status;
 }
@@ -1104,7 +1328,7 @@ _mesa_GetBufferParameterivARB(GLenum target, GLenum pname, GLint *params)
          *params = bufObj->Usage;
          break;
       case GL_BUFFER_ACCESS_ARB:
-         *params = bufObj->Access;
+         *params = simplified_access_mode(bufObj->AccessFlags);
          break;
       case GL_BUFFER_MAPPED_ARB:
          *params = (bufObj->Pointer != NULL);
@@ -1140,3 +1364,234 @@ _mesa_GetBufferPointervARB(GLenum target, GLenum pname, GLvoid **params)
 
    *params = bufObj->Pointer;
 }
+
+
+void GLAPIENTRY
+_mesa_CopyBufferSubData(GLenum readTarget, GLenum writeTarget,
+                        GLintptr readOffset, GLintptr writeOffset,
+                        GLsizeiptr size)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_buffer_object *src, *dst;
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   src = get_buffer(ctx, readTarget);
+   if (!src || src->Name == 0) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glCopyBuffserSubData(readTarget = 0x%x)", readTarget);
+      return;
+   }
+
+   dst = get_buffer(ctx, writeTarget);
+   if (!dst || dst->Name == 0) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glCopyBuffserSubData(writeTarget = 0x%x)", writeTarget);
+      return;
+   }
+
+   if (src->Pointer) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glCopyBuffserSubData(readBuffer is mapped)");
+      return;
+   }
+
+   if (dst->Pointer) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glCopyBuffserSubData(writeBuffer is mapped)");
+      return;
+   }
+
+   if (readOffset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(readOffset = %d)", readOffset);
+      return;
+   }
+
+   if (writeOffset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(writeOffset = %d)", writeOffset);
+      return;
+   }
+
+   if (readOffset + size > src->Size) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(readOffset + size = %d)",
+                  readOffset, size);
+      return;
+   }
+
+   if (writeOffset + size > dst->Size) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glCopyBuffserSubData(writeOffset + size = %d)",
+                  writeOffset, size);
+      return;
+   }
+
+   if (src == dst) {
+      if (readOffset + size <= writeOffset) {
+         /* OK */
+      }
+      else if (writeOffset + size <= readOffset) {
+         /* OK */
+      }
+      else {
+         /* overlapping src/dst is illegal */
+         _mesa_error(ctx, GL_INVALID_VALUE,
+                     "glCopyBuffserSubData(overlapping src/dst)");
+         return;
+      }
+   }
+
+   ctx->Driver.CopyBufferSubData(ctx, src, dst, readOffset, writeOffset, size);
+}
+
+
+/**
+ * See GL_ARB_map_buffer_range spec
+ */
+void * GLAPIENTRY
+_mesa_MapBufferRange(GLenum target, GLintptr offset, GLsizeiptr length,
+                     GLbitfield access)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_buffer_object *bufObj;
+   ASSERT_OUTSIDE_BEGIN_END_WITH_RETVAL(ctx, NULL);
+
+   if (!ctx->Extensions.ARB_map_buffer_range) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(extension not supported)");
+      return NULL;
+   }
+
+   if (offset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(offset = %ld)", offset);
+      return NULL;
+   }
+
+   if (length < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(length = %ld)", length);
+      return NULL;
+   }
+
+   if ((access & (GL_MAP_READ_BIT | GL_MAP_WRITE_BIT)) == 0) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(access indicates neither read or write)");
+      return NULL;
+   }
+
+   if (access & GL_MAP_READ_BIT) {
+      if ((access & GL_MAP_INVALIDATE_RANGE_BIT) ||
+          (access & GL_MAP_INVALIDATE_BUFFER_BIT) ||
+          (access & GL_MAP_UNSYNCHRONIZED_BIT)) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glMapBufferRange(invalid access flags)");
+         return NULL;
+      }
+   }
+
+   if ((access & GL_MAP_FLUSH_EXPLICIT_BIT) &&
+       ((access & GL_MAP_WRITE_BIT) == 0)) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(invalid access flags)");
+      return NULL;
+   }
+
+   bufObj = get_buffer(ctx, target);
+   if (!bufObj || bufObj->Name == 0) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glMapBufferRange(target = 0x%x)", target);
+      return NULL;
+   }
+
+   if (offset + length > bufObj->Size) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(offset + length > size)");
+      return NULL;
+   }
+
+   if (bufObj->Pointer) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(buffer already mapped)");
+      return NULL;
+   }
+      
+   ASSERT(ctx->Driver.MapBufferRange);
+   bufObj->Pointer = ctx->Driver.MapBufferRange(ctx, target, offset, length,
+                                                access, bufObj);
+
+   bufObj->Offset = offset;
+   bufObj->Length = length;
+   bufObj->AccessFlags = access;
+
+   return bufObj->Pointer;
+}
+
+
+/**
+ * See GL_ARB_map_buffer_range spec
+ */
+void GLAPIENTRY
+_mesa_FlushMappedBufferRange(GLenum target, GLintptr offset, GLsizeiptr length)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct gl_buffer_object *bufObj;
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (!ctx->Extensions.ARB_map_buffer_range) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(extension not supported)");
+      return;
+   }
+
+   if (offset < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(offset = %ld)", offset);
+      return;
+   }
+
+   if (length < 0) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glMapBufferRange(length = %ld)", length);
+      return;
+   }
+
+   bufObj = get_buffer(ctx, target);
+   if (!bufObj) {
+      _mesa_error(ctx, GL_INVALID_ENUM,
+                  "glMapBufferRange(target = 0x%x)", target);
+      return;
+   }
+
+   if (bufObj->Name == 0) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(current buffer is 0)");
+      return;
+   }
+
+   if (!bufObj->Pointer) {
+      /* buffer is not mapped */
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(buffer is not mapped)");
+      return;
+   }
+
+   if ((bufObj->AccessFlags & GL_MAP_FLUSH_EXPLICIT_BIT) == 0) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glMapBufferRange(GL_MAP_FLUSH_EXPLICIT_BIT not set)");
+      return;
+   }
+
+   if (offset + length > bufObj->Length) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+             "glMapBufferRange(offset %ld + length %ld > mapped length %ld)",
+             offset, length, bufObj->Length);
+      return;
+   }
+
+   ASSERT(bufObj->AccessFlags & GL_MAP_WRITE_BIT);
+
+   if (ctx->Driver.FlushMappedBufferRange)
+      ctx->Driver.FlushMappedBufferRange(ctx, target, offset, length, bufObj);
+}
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 3c08f0083c..ef59ff83c8 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -41,11 +42,6 @@ _mesa_init_buffer_objects( GLcontext *ctx );
 extern void
 _mesa_update_default_objects_buffer_objects(GLcontext *ctx);
 
-extern struct gl_buffer_object *
-_mesa_new_buffer_object( GLcontext *ctx, GLuint name, GLenum target );
-
-extern void
-_mesa_delete_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj );
 
 extern struct gl_buffer_object *
 _mesa_lookup_bufferobj(GLcontext *ctx, GLuint buffer);
@@ -59,29 +55,6 @@ _mesa_reference_buffer_object(GLcontext *ctx,
                               struct gl_buffer_object **ptr,
                               struct gl_buffer_object *bufObj);
 
-extern void
-_mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
-		   const GLvoid * data, GLenum usage,
-		   struct gl_buffer_object * bufObj );
-
-extern void
-_mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
-		      GLsizeiptrARB size, const GLvoid * data,
-		      struct gl_buffer_object * bufObj );
-
-extern void
-_mesa_buffer_get_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
-			  GLsizeiptrARB size, GLvoid * data,
-			  struct gl_buffer_object * bufObj );
-
-extern void *
-_mesa_buffer_map( GLcontext *ctx, GLenum target, GLenum access,
-		  struct gl_buffer_object * bufObj );
-
-extern GLboolean
-_mesa_buffer_unmap( GLcontext *ctx, GLenum target,
-                    struct gl_buffer_object * bufObj );
-
 extern GLboolean
 _mesa_validate_pbo_access(GLuint dimensions,
                           const struct gl_pixelstore_attrib *pack,
@@ -116,6 +89,9 @@ _mesa_unmap_readpix_pbo(GLcontext *ctx,
                         const struct gl_pixelstore_attrib *pack);
 
 
+extern void
+_mesa_init_buffer_object_functions(struct dd_function_table *driver);
+
 
 /*
  * API functions
@@ -154,4 +130,16 @@ _mesa_GetBufferParameterivARB(GLenum target, GLenum pname, GLint *params);
 extern void GLAPIENTRY
 _mesa_GetBufferPointervARB(GLenum target, GLenum pname, GLvoid **params);
 
+extern void GLAPIENTRY
+_mesa_CopyBufferSubData(GLenum readTarget, GLenum writeTarget,
+                        GLintptr readOffset, GLintptr writeOffset,
+                        GLsizeiptr size);
+
+extern void * GLAPIENTRY
+_mesa_MapBufferRange(GLenum target, GLintptr offset, GLsizeiptr length,
+                     GLbitfield access);
+
+extern void GLAPIENTRY
+_mesa_FlushMappedBufferRange(GLenum target, GLintptr offset, GLsizeiptr length);
+
 #endif
diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c
index c5f13345f0..d8b5f3b1f4 100644
--- a/src/mesa/main/buffers.c
+++ b/src/mesa/main/buffers.c
@@ -443,7 +443,7 @@ _mesa_readbuffer(GLcontext *ctx, GLenum buffer, GLint bufferIndex)
    fb->ColorReadBuffer = buffer;
    fb->_ColorReadBufferIndex = bufferIndex;
 
-   ctx->NewState |= _NEW_PIXEL;
+   ctx->NewState |= _NEW_BUFFERS;
 }
 
 
diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
index f7acd2f08e..f77a29a43e 100644
--- a/src/mesa/main/config.h
+++ b/src/mesa/main/config.h
@@ -168,7 +168,6 @@
 /** For GL_ARB_vertex_program */
 /*@{*/
 #define MAX_VERTEX_PROGRAM_ADDRESS_REGS 1
-#define MAX_VERTEX_PROGRAM_ATTRIBS     16
 /*@}*/
 
 /** For GL_ARB_fragment_program */
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 0a4c9acdfe..415e339cb8 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -455,7 +455,7 @@ _mesa_init_current(GLcontext *ctx)
    GLuint i;
 
    /* Init all to (0,0,0,1) */
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(ctx->Current.Attrib); i++) {
       ASSIGN_4V( ctx->Current.Attrib[i], 0.0, 0.0, 0.0, 1.0 );
    }
 
@@ -592,6 +592,9 @@ _mesa_init_constants(GLcontext *ctx)
    /* GL_ATI_envmap_bumpmap */
    ctx->Const.SupportedBumpUnits = SUPPORTED_ATI_BUMP_UNITS;
 
+   /* GL_EXT_provoking_vertex */
+   ctx->Const.QuadsFollowProvokingVertexConvention = GL_TRUE;
+
    /* sanity checks */
    ASSERT(ctx->Const.MaxTextureUnits == MIN2(ctx->Const.MaxTextureImageUnits,
                                              ctx->Const.MaxTextureCoordUnits));
@@ -602,6 +605,10 @@ _mesa_init_constants(GLcontext *ctx)
    ASSERT(MAX_NV_VERTEX_PROGRAM_TEMPS <= MAX_PROGRAM_TEMPS);
    ASSERT(MAX_NV_VERTEX_PROGRAM_INPUTS <= VERT_ATTRIB_MAX);
    ASSERT(MAX_NV_VERTEX_PROGRAM_OUTPUTS <= VERT_RESULT_MAX);
+
+   /* check that we don't exceed various 32-bit bitfields */
+   ASSERT(VERT_RESULT_MAX <= 32);
+   ASSERT(FRAG_ATTRIB_MAX <= 32);
 }
 
 
@@ -1004,6 +1011,7 @@ _mesa_free_context_data( GLcontext *ctx )
 #if FEATURE_ARB_occlusion_query
    _mesa_free_query_data(ctx);
 #endif
+   _mesa_free_varray_data(ctx);
 
    _mesa_delete_array_object(ctx, ctx->Array.DefaultArrayObj);
 
@@ -1017,10 +1025,6 @@ _mesa_free_context_data( GLcontext *ctx )
    _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj, NULL);
 #endif
 
-#if FEATURE_ARB_vertex_buffer_object
-   _mesa_delete_buffer_object(ctx, ctx->Array.NullBufferObj);
-#endif
-
    /* free dispatch tables */
    _mesa_free(ctx->Exec);
    _mesa_free(ctx->Save);
diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h
index 32b1d4e9fa..1d92e510a4 100644
--- a/src/mesa/main/dd.h
+++ b/src/mesa/main/dd.h
@@ -798,6 +798,12 @@ struct dd_function_table {
    void * (*MapBuffer)( GLcontext *ctx, GLenum target, GLenum access,
 			struct gl_buffer_object *obj );
 
+   void (*CopyBufferSubData)( GLcontext *ctx,
+                              struct gl_buffer_object *src,
+                              struct gl_buffer_object *dst,
+                              GLintptr readOffset, GLintptr writeOffset,
+                              GLsizeiptr size );
+
    /* May return NULL if MESA_MAP_NOWAIT_BIT is set in access:
     */
    void * (*MapBufferRange)( GLcontext *ctx, GLenum target,
diff --git a/src/mesa/main/debug.c b/src/mesa/main/debug.c
index b54e47919d..1c8c44fcb9 100644
--- a/src/mesa/main/debug.c
+++ b/src/mesa/main/debug.c
@@ -24,12 +24,15 @@
  */
 
 #include "mtypes.h"
+#include "attrib.h"
 #include "colormac.h"
 #include "context.h"
 #include "hash.h"
 #include "imports.h"
 #include "debug.h"
 #include "get.h"
+#include "pixelstore.h"
+#include "readpix.h"
 #include "texobj.h"
 #include "texformat.h"
 
@@ -341,3 +344,104 @@ _mesa_dump_textures(GLboolean dumpImages)
    DumpImages = dumpImages;
    _mesa_HashWalk(ctx->Shared->TexObjects, dump_texture_cb, ctx);
 }
+
+
+void
+_mesa_dump_color_buffer(const char *filename)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint w = ctx->DrawBuffer->Width;
+   const GLuint h = ctx->DrawBuffer->Height;
+   GLubyte *buf;
+
+   buf = (GLubyte *) _mesa_malloc(w * h * 4);
+
+   _mesa_PushClientAttrib(GL_CLIENT_PIXEL_STORE_BIT);
+   _mesa_PixelStorei(GL_PACK_ALIGNMENT, 1);
+   _mesa_PixelStorei(GL_PACK_INVERT_MESA, GL_TRUE);
+
+   _mesa_ReadPixels(0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, buf);
+
+   _mesa_printf("ReadBuffer %p 0x%x  DrawBuffer %p 0x%x\n",
+                ctx->ReadBuffer->_ColorReadBuffer,
+                ctx->ReadBuffer->ColorReadBuffer,
+                ctx->DrawBuffer->_ColorDrawBuffers[0],
+                ctx->DrawBuffer->ColorDrawBuffer[0]);
+   _mesa_printf("Writing %d x %d color buffer to %s\n", w, h, filename);
+   write_ppm(filename, buf, w, h, 4, 0, 1, 2);
+
+   _mesa_PopClientAttrib();
+
+   _mesa_free(buf);
+}
+
+
+void
+_mesa_dump_depth_buffer(const char *filename)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint w = ctx->DrawBuffer->Width;
+   const GLuint h = ctx->DrawBuffer->Height;
+   GLuint *buf;
+   GLubyte *buf2;
+   GLuint i;
+
+   buf = (GLuint *) _mesa_malloc(w * h * 4);  /* 4 bpp */
+   buf2 = (GLubyte *) _mesa_malloc(w * h * 3); /* 3 bpp */
+
+   _mesa_PushClientAttrib(GL_CLIENT_PIXEL_STORE_BIT);
+   _mesa_PixelStorei(GL_PACK_ALIGNMENT, 1);
+   _mesa_PixelStorei(GL_PACK_INVERT_MESA, GL_TRUE);
+
+   _mesa_ReadPixels(0, 0, w, h, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, buf);
+
+   /* spread 24 bits of Z across R, G, B */
+   for (i = 0; i < w * h; i++) {
+      buf2[i*3+0] = (buf[i] >> 24) & 0xff;
+      buf2[i*3+1] = (buf[i] >> 16) & 0xff;
+      buf2[i*3+2] = (buf[i] >>  8) & 0xff;
+   }
+
+   _mesa_printf("Writing %d x %d depth buffer to %s\n", w, h, filename);
+   write_ppm(filename, buf2, w, h, 3, 0, 1, 2);
+
+   _mesa_PopClientAttrib();
+
+   _mesa_free(buf);
+   _mesa_free(buf2);
+}
+
+
+void
+_mesa_dump_stencil_buffer(const char *filename)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint w = ctx->DrawBuffer->Width;
+   const GLuint h = ctx->DrawBuffer->Height;
+   GLubyte *buf;
+   GLubyte *buf2;
+   GLuint i;
+
+   buf = (GLubyte *) _mesa_malloc(w * h);  /* 1 bpp */
+   buf2 = (GLubyte *) _mesa_malloc(w * h * 3); /* 3 bpp */
+
+   _mesa_PushClientAttrib(GL_CLIENT_PIXEL_STORE_BIT);
+   _mesa_PixelStorei(GL_PACK_ALIGNMENT, 1);
+   _mesa_PixelStorei(GL_PACK_INVERT_MESA, GL_TRUE);
+
+   _mesa_ReadPixels(0, 0, w, h, GL_STENCIL_INDEX, GL_UNSIGNED_BYTE, buf);
+
+   for (i = 0; i < w * h; i++) {
+      buf2[i*3+0] = buf[i];
+      buf2[i*3+1] = (buf[i] & 127) * 2;
+      buf2[i*3+2] = (buf[i] - 128) * 2;
+   }
+
+   _mesa_printf("Writing %d x %d stencil buffer to %s\n", w, h, filename);
+   write_ppm(filename, buf2, w, h, 3, 0, 1, 2);
+
+   _mesa_PopClientAttrib();
+
+   _mesa_free(buf);
+   _mesa_free(buf2);
+}
diff --git a/src/mesa/main/debug.h b/src/mesa/main/debug.h
index 1862ec75b7..bb384c4324 100644
--- a/src/mesa/main/debug.h
+++ b/src/mesa/main/debug.h
@@ -60,4 +60,13 @@ extern void _mesa_init_debug( GLcontext *ctx );
 extern void
 _mesa_dump_textures(GLboolean dumpImages);
 
+extern void
+_mesa_dump_color_buffer(const char *filename);
+
+extern void
+_mesa_dump_depth_buffer(const char *filename);
+
+extern void
+_mesa_dump_stencil_buffer(const char *filename);
+
 #endif
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index c5a1c1f38f..49f202daa1 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -352,6 +352,9 @@ typedef enum
    OPCODE_EVAL_P1,
    OPCODE_EVAL_P2,
 
+   /* GL_EXT_provoking_vertex */
+   OPCODE_PROVOKING_VERTEX,
+
    /* The following three are meta instructions */
    OPCODE_ERROR,                /* raise compiled-in error */
    OPCODE_CONTINUE,
@@ -4911,7 +4914,7 @@ save_Attr1fNV(GLenum attr, GLfloat x)
       n[2].f = x;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 1;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, 0, 0, 1);
 
@@ -4933,7 +4936,7 @@ save_Attr2fNV(GLenum attr, GLfloat x, GLfloat y)
       n[3].f = y;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 2;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, y, 0, 1);
 
@@ -4956,7 +4959,7 @@ save_Attr3fNV(GLenum attr, GLfloat x, GLfloat y, GLfloat z)
       n[4].f = z;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 3;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, y, z, 1);
 
@@ -4980,7 +4983,7 @@ save_Attr4fNV(GLenum attr, GLfloat x, GLfloat y, GLfloat z, GLfloat w)
       n[5].f = w;
    }
 
-   ASSERT(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+   ASSERT(attr < MAX_VERTEX_GENERIC_ATTRIBS);
    ctx->ListState.ActiveAttribSize[attr] = 4;
    ASSIGN_4V(ctx->ListState.CurrentAttrib[attr], x, y, z, w);
 
@@ -5757,6 +5760,25 @@ save_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
 #endif
 
 
+/** GL_EXT_provoking_vertex */
+static void GLAPIENTRY
+save_ProvokingVertexEXT(GLenum mode)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   Node *n;
+   ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx);
+   n = ALLOC_INSTRUCTION(ctx, OPCODE_PROVOKING_VERTEX, 1);
+   if (n) {
+      n[1].e = mode;
+   }
+   if (ctx->ExecuteFlag) {
+      /*CALL_ProvokingVertexEXT(ctx->Exec, (mode));*/
+      _mesa_ProvokingVertexEXT(mode);
+   }
+}
+
+
+
 /**
  * Save an error-generating command into display list.
  *
@@ -6321,6 +6343,9 @@ execute_list(GLcontext *ctx, GLuint list)
          case OPCODE_SHADE_MODEL:
             CALL_ShadeModel(ctx->Exec, (n[1].e));
             break;
+         case OPCODE_PROVOKING_VERTEX:
+            CALL_ProvokingVertexEXT(ctx->Exec, (n[1].e));
+            break;
          case OPCODE_STENCIL_FUNC:
             CALL_StencilFunc(ctx->Exec, (n[1].e, n[2].i, n[3].ui));
             break;
@@ -8284,6 +8309,18 @@ _mesa_init_dlist_table(struct _glapi_table *table)
    SET_ProgramEnvParameters4fvEXT(table, save_ProgramEnvParameters4fvEXT);
    SET_ProgramLocalParameters4fvEXT(table, save_ProgramLocalParameters4fvEXT);
 #endif
+
+   /* ARB 50. GL_ARB_map_buffer_range */
+#if FEATURE_ARB_map_buffer_range
+   SET_MapBufferRange(table, _mesa_MapBufferRange); /* no dlist save */
+   SET_FlushMappedBufferRange(table, _mesa_FlushMappedBufferRange); /* no dl */
+#endif
+
+   /* ARB 59. GL_ARB_copy_buffer */
+   SET_CopyBufferSubData(table, _mesa_CopyBufferSubData); /* no dlist save */
+
+   /* 364. GL_EXT_provoking_vertex */
+   SET_ProvokingVertexEXT(table, save_ProvokingVertexEXT);
 }
 
 
@@ -8520,6 +8557,11 @@ print_list(GLcontext *ctx, GLuint list)
             _mesa_printf("EVAL_P2 %d %d\n", n[1].i, n[2].i);
             break;
 
+         case OPCODE_PROVOKING_VERTEX:
+            _mesa_printf("ProvokingVertex %s\n",
+                         _mesa_lookup_enum_by_nr(n[1].ui));
+            break;
+
             /*
              * meta opcodes/commands
              */
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index f432be183c..48268fcd27 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -119,6 +119,7 @@ client_state(GLcontext *ctx, GLenum cap, GLboolean state)
          CHECK_EXTENSION(NV_vertex_program, cap);
          {
             GLint n = (GLint) cap - GL_VERTEX_ATTRIB_ARRAY0_NV;
+            ASSERT(n < Elements(ctx->Array.ArrayObj->VertexAttrib));
             var = &ctx->Array.ArrayObj->VertexAttrib[n].Enabled;
             flag = _NEW_ARRAY_ATTRIB(n);
          }
@@ -222,14 +223,16 @@ get_texcoord_unit(GLcontext *ctx)
 
 /**
  * Helper function to enable or disable a texture target.
+ * \param bit  one of the TEXTURE_x_BIT values
+ * \return GL_TRUE if state is changing or GL_FALSE if no change
  */
 static GLboolean
-enable_texture(GLcontext *ctx, GLboolean state, GLbitfield bit)
+enable_texture(GLcontext *ctx, GLboolean state, GLbitfield texBit)
 {
    const GLuint curr = ctx->Texture.CurrentUnit;
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[curr];
-   const GLuint newenabled = (!state)
-       ? (texUnit->Enabled & ~bit) :  (texUnit->Enabled | bit);
+   const GLbitfield newenabled = state
+      ? (texUnit->Enabled | texBit) : (texUnit->Enabled & ~texBit);
 
    if (!ctx->DrawBuffer->Visual.rgbMode || texUnit->Enabled == newenabled)
        return GL_FALSE;
@@ -1314,6 +1317,7 @@ _mesa_IsEnabled( GLenum cap )
          CHECK_EXTENSION(NV_vertex_program);
          {
             GLint n = (GLint) cap - GL_VERTEX_ATTRIB_ARRAY0_NV;
+            ASSERT(n < Elements(ctx->Array.ArrayObj->VertexAttrib));
             return (ctx->Array.ArrayObj->VertexAttrib[n].Enabled != 0);
          }
       case GL_MAP1_VERTEX_ATTRIB0_4_NV:
diff --git a/src/mesa/main/enums.c b/src/mesa/main/enums.c
index dff5544086..6e2adea636 100644
--- a/src/mesa/main/enums.c
+++ b/src/mesa/main/enums.c
@@ -334,6 +334,8 @@ LONGSTRING static const char enum_string_table[] =
    "GL_COPY\0"
    "GL_COPY_INVERTED\0"
    "GL_COPY_PIXEL_TOKEN\0"
+   "GL_COPY_READ_BUFFER\0"
+   "GL_COPY_WRITE_BUFFER\0"
    "GL_CULL_FACE\0"
    "GL_CULL_FACE_MODE\0"
    "GL_CULL_VERTEX_EXT\0"
@@ -512,6 +514,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_FEEDBACK_BUFFER_SIZE\0"
    "GL_FEEDBACK_BUFFER_TYPE\0"
    "GL_FILL\0"
+   "GL_FIRST_VERTEX_CONVENTION_EXT\0"
    "GL_FLAT\0"
    "GL_FLOAT\0"
    "GL_FLOAT_MAT2\0"
@@ -694,6 +697,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_INVERSE_TRANSPOSE_NV\0"
    "GL_INVERT\0"
    "GL_KEEP\0"
+   "GL_LAST_VERTEX_CONVENTION_EXT\0"
    "GL_LEFT\0"
    "GL_LEQUAL\0"
    "GL_LESS\0"
@@ -819,7 +823,13 @@ LONGSTRING static const char enum_string_table[] =
    "GL_MAP2_VERTEX_ATTRIB8_4_NV\0"
    "GL_MAP2_VERTEX_ATTRIB9_4_NV\0"
    "GL_MAP_COLOR\0"
+   "GL_MAP_FLUSH_EXPLICIT_BIT\0"
+   "GL_MAP_INVALIDATE_BUFFER_BIT\0"
+   "GL_MAP_INVALIDATE_RANGE_BIT\0"
+   "GL_MAP_READ_BIT\0"
    "GL_MAP_STENCIL\0"
+   "GL_MAP_UNSYNCHRONIZED_BIT\0"
+   "GL_MAP_WRITE_BIT\0"
    "GL_MATRIX0_ARB\0"
    "GL_MATRIX0_NV\0"
    "GL_MATRIX10_ARB\0"
@@ -1271,6 +1281,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_PROJECTION\0"
    "GL_PROJECTION_MATRIX\0"
    "GL_PROJECTION_STACK_DEPTH\0"
+   "GL_PROVOKING_VERTEX_EXT\0"
    "GL_PROXY_COLOR_TABLE\0"
    "GL_PROXY_HISTOGRAM\0"
    "GL_PROXY_HISTOGRAM_EXT\0"
@@ -1291,6 +1302,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_Q\0"
    "GL_QUADRATIC_ATTENUATION\0"
    "GL_QUADS\0"
+   "GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT\0"
    "GL_QUAD_MESH_SUN\0"
    "GL_QUAD_STRIP\0"
    "GL_QUERY_COUNTER_BITS\0"
@@ -1787,6 +1799,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_VENDOR\0"
    "GL_VERSION\0"
    "GL_VERTEX_ARRAY\0"
+   "GL_VERTEX_ARRAY_BINDING\0"
    "GL_VERTEX_ARRAY_BINDING_APPLE\0"
    "GL_VERTEX_ARRAY_BUFFER_BINDING\0"
    "GL_VERTEX_ARRAY_BUFFER_BINDING_ARB\0"
@@ -1858,7 +1871,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_ZOOM_Y\0"
    ;
 
-static const enum_elt all_enums[1820] =
+static const enum_elt all_enums[1833] =
 {
    {     0, 0x00000600 }, /* GL_2D */
    {     6, 0x00001407 }, /* GL_2_BYTES */
@@ -2158,2131 +2171,2144 @@ static const enum_elt all_enums[1820] =
    {  6027, 0x00001503 }, /* GL_COPY */
    {  6035, 0x0000150C }, /* GL_COPY_INVERTED */
    {  6052, 0x00000706 }, /* GL_COPY_PIXEL_TOKEN */
-   {  6072, 0x00000B44 }, /* GL_CULL_FACE */
-   {  6085, 0x00000B45 }, /* GL_CULL_FACE_MODE */
-   {  6103, 0x000081AA }, /* GL_CULL_VERTEX_EXT */
-   {  6122, 0x000081AC }, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
-   {  6154, 0x000081AB }, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
-   {  6189, 0x00008626 }, /* GL_CURRENT_ATTRIB_NV */
-   {  6210, 0x00000001 }, /* GL_CURRENT_BIT */
-   {  6225, 0x00000B00 }, /* GL_CURRENT_COLOR */
-   {  6242, 0x00008453 }, /* GL_CURRENT_FOG_COORD */
-   {  6263, 0x00008453 }, /* GL_CURRENT_FOG_COORDINATE */
-   {  6289, 0x00000B01 }, /* GL_CURRENT_INDEX */
-   {  6306, 0x00008641 }, /* GL_CURRENT_MATRIX_ARB */
-   {  6328, 0x00008845 }, /* GL_CURRENT_MATRIX_INDEX_ARB */
-   {  6356, 0x00008641 }, /* GL_CURRENT_MATRIX_NV */
-   {  6377, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
-   {  6411, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_NV */
-   {  6444, 0x00000B02 }, /* GL_CURRENT_NORMAL */
-   {  6462, 0x00008843 }, /* GL_CURRENT_PALETTE_MATRIX_ARB */
-   {  6492, 0x00008B8D }, /* GL_CURRENT_PROGRAM */
-   {  6511, 0x00008865 }, /* GL_CURRENT_QUERY */
-   {  6528, 0x00008865 }, /* GL_CURRENT_QUERY_ARB */
-   {  6549, 0x00000B04 }, /* GL_CURRENT_RASTER_COLOR */
-   {  6573, 0x00000B09 }, /* GL_CURRENT_RASTER_DISTANCE */
-   {  6600, 0x00000B05 }, /* GL_CURRENT_RASTER_INDEX */
-   {  6624, 0x00000B07 }, /* GL_CURRENT_RASTER_POSITION */
-   {  6651, 0x00000B08 }, /* GL_CURRENT_RASTER_POSITION_VALID */
-   {  6684, 0x0000845F }, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
-   {  6718, 0x00000B06 }, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
-   {  6751, 0x00008459 }, /* GL_CURRENT_SECONDARY_COLOR */
-   {  6778, 0x00000B03 }, /* GL_CURRENT_TEXTURE_COORDS */
-   {  6804, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB */
-   {  6829, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB_ARB */
-   {  6858, 0x000086A8 }, /* GL_CURRENT_WEIGHT_ARB */
-   {  6880, 0x00000900 }, /* GL_CW */
-   {  6886, 0x0000875B }, /* GL_DEBUG_ASSERT_MESA */
-   {  6907, 0x00008759 }, /* GL_DEBUG_OBJECT_MESA */
-   {  6928, 0x0000875A }, /* GL_DEBUG_PRINT_MESA */
-   {  6948, 0x00002101 }, /* GL_DECAL */
-   {  6957, 0x00001E03 }, /* GL_DECR */
-   {  6965, 0x00008508 }, /* GL_DECR_WRAP */
-   {  6978, 0x00008508 }, /* GL_DECR_WRAP_EXT */
-   {  6995, 0x00008B80 }, /* GL_DELETE_STATUS */
-   {  7012, 0x00001801 }, /* GL_DEPTH */
-   {  7021, 0x000088F0 }, /* GL_DEPTH24_STENCIL8 */
-   {  7041, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT */
-   {  7061, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT_EXT */
-   {  7085, 0x00000D1F }, /* GL_DEPTH_BIAS */
-   {  7099, 0x00000D56 }, /* GL_DEPTH_BITS */
-   {  7113, 0x00008891 }, /* GL_DEPTH_BOUNDS_EXT */
-   {  7133, 0x00008890 }, /* GL_DEPTH_BOUNDS_TEST_EXT */
-   {  7158, 0x00000100 }, /* GL_DEPTH_BUFFER_BIT */
-   {  7178, 0x0000864F }, /* GL_DEPTH_CLAMP_NV */
-   {  7196, 0x00000B73 }, /* GL_DEPTH_CLEAR_VALUE */
-   {  7217, 0x00001902 }, /* GL_DEPTH_COMPONENT */
-   {  7236, 0x000081A5 }, /* GL_DEPTH_COMPONENT16 */
-   {  7257, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_ARB */
-   {  7282, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_SGIX */
-   {  7308, 0x000081A6 }, /* GL_DEPTH_COMPONENT24 */
-   {  7329, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_ARB */
-   {  7354, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_SGIX */
-   {  7380, 0x000081A7 }, /* GL_DEPTH_COMPONENT32 */
-   {  7401, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_ARB */
-   {  7426, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_SGIX */
-   {  7452, 0x00000B74 }, /* GL_DEPTH_FUNC */
-   {  7466, 0x00000B70 }, /* GL_DEPTH_RANGE */
-   {  7481, 0x00000D1E }, /* GL_DEPTH_SCALE */
-   {  7496, 0x000084F9 }, /* GL_DEPTH_STENCIL */
-   {  7513, 0x0000821A }, /* GL_DEPTH_STENCIL_ATTACHMENT */
-   {  7541, 0x000084F9 }, /* GL_DEPTH_STENCIL_NV */
-   {  7561, 0x0000886F }, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
-   {  7589, 0x0000886E }, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
-   {  7617, 0x00000B71 }, /* GL_DEPTH_TEST */
-   {  7631, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE */
-   {  7653, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE_ARB */
-   {  7679, 0x00000B72 }, /* GL_DEPTH_WRITEMASK */
-   {  7698, 0x00001201 }, /* GL_DIFFUSE */
-   {  7709, 0x00000BD0 }, /* GL_DITHER */
-   {  7719, 0x00000A02 }, /* GL_DOMAIN */
-   {  7729, 0x00001100 }, /* GL_DONT_CARE */
-   {  7742, 0x000086AE }, /* GL_DOT3_RGB */
-   {  7754, 0x000086AF }, /* GL_DOT3_RGBA */
-   {  7767, 0x000086AF }, /* GL_DOT3_RGBA_ARB */
-   {  7784, 0x00008741 }, /* GL_DOT3_RGBA_EXT */
-   {  7801, 0x000086AE }, /* GL_DOT3_RGB_ARB */
-   {  7817, 0x00008740 }, /* GL_DOT3_RGB_EXT */
-   {  7833, 0x0000140A }, /* GL_DOUBLE */
-   {  7843, 0x00000C32 }, /* GL_DOUBLEBUFFER */
-   {  7859, 0x00000C01 }, /* GL_DRAW_BUFFER */
-   {  7874, 0x00008825 }, /* GL_DRAW_BUFFER0 */
-   {  7890, 0x00008825 }, /* GL_DRAW_BUFFER0_ARB */
-   {  7910, 0x00008825 }, /* GL_DRAW_BUFFER0_ATI */
-   {  7930, 0x00008826 }, /* GL_DRAW_BUFFER1 */
-   {  7946, 0x0000882F }, /* GL_DRAW_BUFFER10 */
-   {  7963, 0x0000882F }, /* GL_DRAW_BUFFER10_ARB */
-   {  7984, 0x0000882F }, /* GL_DRAW_BUFFER10_ATI */
-   {  8005, 0x00008830 }, /* GL_DRAW_BUFFER11 */
-   {  8022, 0x00008830 }, /* GL_DRAW_BUFFER11_ARB */
-   {  8043, 0x00008830 }, /* GL_DRAW_BUFFER11_ATI */
-   {  8064, 0x00008831 }, /* GL_DRAW_BUFFER12 */
-   {  8081, 0x00008831 }, /* GL_DRAW_BUFFER12_ARB */
-   {  8102, 0x00008831 }, /* GL_DRAW_BUFFER12_ATI */
-   {  8123, 0x00008832 }, /* GL_DRAW_BUFFER13 */
-   {  8140, 0x00008832 }, /* GL_DRAW_BUFFER13_ARB */
-   {  8161, 0x00008832 }, /* GL_DRAW_BUFFER13_ATI */
-   {  8182, 0x00008833 }, /* GL_DRAW_BUFFER14 */
-   {  8199, 0x00008833 }, /* GL_DRAW_BUFFER14_ARB */
-   {  8220, 0x00008833 }, /* GL_DRAW_BUFFER14_ATI */
-   {  8241, 0x00008834 }, /* GL_DRAW_BUFFER15 */
-   {  8258, 0x00008834 }, /* GL_DRAW_BUFFER15_ARB */
-   {  8279, 0x00008834 }, /* GL_DRAW_BUFFER15_ATI */
-   {  8300, 0x00008826 }, /* GL_DRAW_BUFFER1_ARB */
-   {  8320, 0x00008826 }, /* GL_DRAW_BUFFER1_ATI */
-   {  8340, 0x00008827 }, /* GL_DRAW_BUFFER2 */
-   {  8356, 0x00008827 }, /* GL_DRAW_BUFFER2_ARB */
-   {  8376, 0x00008827 }, /* GL_DRAW_BUFFER2_ATI */
-   {  8396, 0x00008828 }, /* GL_DRAW_BUFFER3 */
-   {  8412, 0x00008828 }, /* GL_DRAW_BUFFER3_ARB */
-   {  8432, 0x00008828 }, /* GL_DRAW_BUFFER3_ATI */
-   {  8452, 0x00008829 }, /* GL_DRAW_BUFFER4 */
-   {  8468, 0x00008829 }, /* GL_DRAW_BUFFER4_ARB */
-   {  8488, 0x00008829 }, /* GL_DRAW_BUFFER4_ATI */
-   {  8508, 0x0000882A }, /* GL_DRAW_BUFFER5 */
-   {  8524, 0x0000882A }, /* GL_DRAW_BUFFER5_ARB */
-   {  8544, 0x0000882A }, /* GL_DRAW_BUFFER5_ATI */
-   {  8564, 0x0000882B }, /* GL_DRAW_BUFFER6 */
-   {  8580, 0x0000882B }, /* GL_DRAW_BUFFER6_ARB */
-   {  8600, 0x0000882B }, /* GL_DRAW_BUFFER6_ATI */
-   {  8620, 0x0000882C }, /* GL_DRAW_BUFFER7 */
-   {  8636, 0x0000882C }, /* GL_DRAW_BUFFER7_ARB */
-   {  8656, 0x0000882C }, /* GL_DRAW_BUFFER7_ATI */
-   {  8676, 0x0000882D }, /* GL_DRAW_BUFFER8 */
-   {  8692, 0x0000882D }, /* GL_DRAW_BUFFER8_ARB */
-   {  8712, 0x0000882D }, /* GL_DRAW_BUFFER8_ATI */
-   {  8732, 0x0000882E }, /* GL_DRAW_BUFFER9 */
-   {  8748, 0x0000882E }, /* GL_DRAW_BUFFER9_ARB */
-   {  8768, 0x0000882E }, /* GL_DRAW_BUFFER9_ATI */
-   {  8788, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER */
-   {  8808, 0x00008CA6 }, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
-   {  8840, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER_EXT */
-   {  8864, 0x00000705 }, /* GL_DRAW_PIXEL_TOKEN */
-   {  8884, 0x00000304 }, /* GL_DST_ALPHA */
-   {  8897, 0x00000306 }, /* GL_DST_COLOR */
-   {  8910, 0x0000877A }, /* GL_DU8DV8_ATI */
-   {  8924, 0x00008779 }, /* GL_DUDV_ATI */
-   {  8936, 0x000088EA }, /* GL_DYNAMIC_COPY */
-   {  8952, 0x000088EA }, /* GL_DYNAMIC_COPY_ARB */
-   {  8972, 0x000088E8 }, /* GL_DYNAMIC_DRAW */
-   {  8988, 0x000088E8 }, /* GL_DYNAMIC_DRAW_ARB */
-   {  9008, 0x000088E9 }, /* GL_DYNAMIC_READ */
-   {  9024, 0x000088E9 }, /* GL_DYNAMIC_READ_ARB */
-   {  9044, 0x00000B43 }, /* GL_EDGE_FLAG */
-   {  9057, 0x00008079 }, /* GL_EDGE_FLAG_ARRAY */
-   {  9076, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
-   {  9110, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB */
-   {  9148, 0x00008093 }, /* GL_EDGE_FLAG_ARRAY_POINTER */
-   {  9175, 0x0000808C }, /* GL_EDGE_FLAG_ARRAY_STRIDE */
-   {  9201, 0x00008893 }, /* GL_ELEMENT_ARRAY_BUFFER */
-   {  9225, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
-   {  9257, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB */
-   {  9293, 0x00001600 }, /* GL_EMISSION */
-   {  9305, 0x00002000 }, /* GL_ENABLE_BIT */
-   {  9319, 0x00000202 }, /* GL_EQUAL */
-   {  9328, 0x00001509 }, /* GL_EQUIV */
-   {  9337, 0x00010000 }, /* GL_EVAL_BIT */
-   {  9349, 0x00000800 }, /* GL_EXP */
-   {  9356, 0x00000801 }, /* GL_EXP2 */
-   {  9364, 0x00001F03 }, /* GL_EXTENSIONS */
-   {  9378, 0x00002400 }, /* GL_EYE_LINEAR */
-   {  9392, 0x00002502 }, /* GL_EYE_PLANE */
-   {  9405, 0x0000855C }, /* GL_EYE_PLANE_ABSOLUTE_NV */
-   {  9430, 0x0000855B }, /* GL_EYE_RADIAL_NV */
-   {  9447, 0x00000000 }, /* GL_FALSE */
-   {  9456, 0x00001101 }, /* GL_FASTEST */
-   {  9467, 0x00001C01 }, /* GL_FEEDBACK */
-   {  9479, 0x00000DF0 }, /* GL_FEEDBACK_BUFFER_POINTER */
-   {  9506, 0x00000DF1 }, /* GL_FEEDBACK_BUFFER_SIZE */
-   {  9530, 0x00000DF2 }, /* GL_FEEDBACK_BUFFER_TYPE */
-   {  9554, 0x00001B02 }, /* GL_FILL */
-   {  9562, 0x00001D00 }, /* GL_FLAT */
-   {  9570, 0x00001406 }, /* GL_FLOAT */
-   {  9579, 0x00008B5A }, /* GL_FLOAT_MAT2 */
-   {  9593, 0x00008B5A }, /* GL_FLOAT_MAT2_ARB */
-   {  9611, 0x00008B65 }, /* GL_FLOAT_MAT2x3 */
-   {  9627, 0x00008B66 }, /* GL_FLOAT_MAT2x4 */
-   {  9643, 0x00008B5B }, /* GL_FLOAT_MAT3 */
-   {  9657, 0x00008B5B }, /* GL_FLOAT_MAT3_ARB */
-   {  9675, 0x00008B67 }, /* GL_FLOAT_MAT3x2 */
-   {  9691, 0x00008B68 }, /* GL_FLOAT_MAT3x4 */
-   {  9707, 0x00008B5C }, /* GL_FLOAT_MAT4 */
-   {  9721, 0x00008B5C }, /* GL_FLOAT_MAT4_ARB */
-   {  9739, 0x00008B69 }, /* GL_FLOAT_MAT4x2 */
-   {  9755, 0x00008B6A }, /* GL_FLOAT_MAT4x3 */
-   {  9771, 0x00008B50 }, /* GL_FLOAT_VEC2 */
-   {  9785, 0x00008B50 }, /* GL_FLOAT_VEC2_ARB */
-   {  9803, 0x00008B51 }, /* GL_FLOAT_VEC3 */
-   {  9817, 0x00008B51 }, /* GL_FLOAT_VEC3_ARB */
-   {  9835, 0x00008B52 }, /* GL_FLOAT_VEC4 */
-   {  9849, 0x00008B52 }, /* GL_FLOAT_VEC4_ARB */
-   {  9867, 0x00000B60 }, /* GL_FOG */
-   {  9874, 0x00000080 }, /* GL_FOG_BIT */
-   {  9885, 0x00000B66 }, /* GL_FOG_COLOR */
-   {  9898, 0x00008451 }, /* GL_FOG_COORD */
-   {  9911, 0x00008451 }, /* GL_FOG_COORDINATE */
-   {  9929, 0x00008457 }, /* GL_FOG_COORDINATE_ARRAY */
-   {  9953, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
-   {  9992, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB */
-   { 10035, 0x00008456 }, /* GL_FOG_COORDINATE_ARRAY_POINTER */
-   { 10067, 0x00008455 }, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
-   { 10098, 0x00008454 }, /* GL_FOG_COORDINATE_ARRAY_TYPE */
-   { 10127, 0x00008450 }, /* GL_FOG_COORDINATE_SOURCE */
-   { 10152, 0x00008457 }, /* GL_FOG_COORD_ARRAY */
-   { 10171, 0x0000889D }, /* GL_FOG_COORD_ARRAY_BUFFER_BINDING */
-   { 10205, 0x00008456 }, /* GL_FOG_COORD_ARRAY_POINTER */
-   { 10232, 0x00008455 }, /* GL_FOG_COORD_ARRAY_STRIDE */
-   { 10258, 0x00008454 }, /* GL_FOG_COORD_ARRAY_TYPE */
-   { 10282, 0x00008450 }, /* GL_FOG_COORD_SRC */
-   { 10299, 0x00000B62 }, /* GL_FOG_DENSITY */
-   { 10314, 0x0000855A }, /* GL_FOG_DISTANCE_MODE_NV */
-   { 10338, 0x00000B64 }, /* GL_FOG_END */
-   { 10349, 0x00000C54 }, /* GL_FOG_HINT */
-   { 10361, 0x00000B61 }, /* GL_FOG_INDEX */
-   { 10374, 0x00000B65 }, /* GL_FOG_MODE */
-   { 10386, 0x00008198 }, /* GL_FOG_OFFSET_SGIX */
-   { 10405, 0x00008199 }, /* GL_FOG_OFFSET_VALUE_SGIX */
-   { 10430, 0x00000B63 }, /* GL_FOG_START */
-   { 10443, 0x00008452 }, /* GL_FRAGMENT_DEPTH */
-   { 10461, 0x00008804 }, /* GL_FRAGMENT_PROGRAM_ARB */
-   { 10485, 0x00008B30 }, /* GL_FRAGMENT_SHADER */
-   { 10504, 0x00008B30 }, /* GL_FRAGMENT_SHADER_ARB */
-   { 10527, 0x00008B8B }, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
-   { 10562, 0x00008D40 }, /* GL_FRAMEBUFFER */
-   { 10577, 0x00008215 }, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
-   { 10614, 0x00008214 }, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
-   { 10650, 0x00008210 }, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
-   { 10691, 0x00008211 }, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
-   { 10732, 0x00008216 }, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
-   { 10769, 0x00008213 }, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
-   { 10806, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
-   { 10844, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
-   { 10886, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
-   { 10924, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
-   { 10966, 0x00008212 }, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
-   { 11001, 0x00008217 }, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
-   { 11040, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
-   { 11089, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
-   { 11137, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
-   { 11189, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
-   { 11229, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT */
-   { 11273, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
-   { 11313, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
-   { 11357, 0x00008CA6 }, /* GL_FRAMEBUFFER_BINDING_EXT */
-   { 11384, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE */
-   { 11408, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE_EXT */
-   { 11436, 0x00008218 }, /* GL_FRAMEBUFFER_DEFAULT */
-   { 11459, 0x00008D40 }, /* GL_FRAMEBUFFER_EXT */
-   { 11478, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
-   { 11515, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
-   { 11556, 0x00008CD9 }, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
-   { 11597, 0x00008CDB }, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
-   { 11639, 0x00008CD8 }, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
-   { 11690, 0x00008CDA }, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
-   { 11728, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
-   { 11773, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
-   { 11822, 0x00008D56 }, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
-   { 11860, 0x00008CDC }, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
-   { 11902, 0x00008CDE }, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
-   { 11934, 0x00008219 }, /* GL_FRAMEBUFFER_UNDEFINED */
-   { 11959, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED */
-   { 11986, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
-   { 12017, 0x00000404 }, /* GL_FRONT */
-   { 12026, 0x00000408 }, /* GL_FRONT_AND_BACK */
-   { 12044, 0x00000B46 }, /* GL_FRONT_FACE */
-   { 12058, 0x00000400 }, /* GL_FRONT_LEFT */
-   { 12072, 0x00000401 }, /* GL_FRONT_RIGHT */
-   { 12087, 0x00008006 }, /* GL_FUNC_ADD */
-   { 12099, 0x00008006 }, /* GL_FUNC_ADD_EXT */
-   { 12115, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT */
-   { 12140, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT_EXT */
-   { 12169, 0x0000800A }, /* GL_FUNC_SUBTRACT */
-   { 12186, 0x0000800A }, /* GL_FUNC_SUBTRACT_EXT */
-   { 12207, 0x00008191 }, /* GL_GENERATE_MIPMAP */
-   { 12226, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT */
-   { 12250, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT_SGIS */
-   { 12279, 0x00008191 }, /* GL_GENERATE_MIPMAP_SGIS */
-   { 12303, 0x00000206 }, /* GL_GEQUAL */
-   { 12313, 0x00000204 }, /* GL_GREATER */
-   { 12324, 0x00001904 }, /* GL_GREEN */
-   { 12333, 0x00000D19 }, /* GL_GREEN_BIAS */
-   { 12347, 0x00000D53 }, /* GL_GREEN_BITS */
-   { 12361, 0x00000D18 }, /* GL_GREEN_SCALE */
-   { 12376, 0x00008000 }, /* GL_HINT_BIT */
-   { 12388, 0x00008024 }, /* GL_HISTOGRAM */
-   { 12401, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE */
-   { 12425, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE_EXT */
-   { 12453, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE */
-   { 12476, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE_EXT */
-   { 12503, 0x00008024 }, /* GL_HISTOGRAM_EXT */
-   { 12520, 0x00008027 }, /* GL_HISTOGRAM_FORMAT */
-   { 12540, 0x00008027 }, /* GL_HISTOGRAM_FORMAT_EXT */
-   { 12564, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE */
-   { 12588, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE_EXT */
-   { 12616, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE */
-   { 12644, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE_EXT */
-   { 12676, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE */
-   { 12698, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE_EXT */
-   { 12724, 0x0000802D }, /* GL_HISTOGRAM_SINK */
-   { 12742, 0x0000802D }, /* GL_HISTOGRAM_SINK_EXT */
-   { 12764, 0x00008026 }, /* GL_HISTOGRAM_WIDTH */
-   { 12783, 0x00008026 }, /* GL_HISTOGRAM_WIDTH_EXT */
-   { 12806, 0x0000862A }, /* GL_IDENTITY_NV */
-   { 12821, 0x00008150 }, /* GL_IGNORE_BORDER_HP */
-   { 12841, 0x00008B9B }, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
-   { 12881, 0x00008B9A }, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
-   { 12919, 0x00001E02 }, /* GL_INCR */
-   { 12927, 0x00008507 }, /* GL_INCR_WRAP */
-   { 12940, 0x00008507 }, /* GL_INCR_WRAP_EXT */
-   { 12957, 0x00008222 }, /* GL_INDEX */
-   { 12966, 0x00008077 }, /* GL_INDEX_ARRAY */
-   { 12981, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING */
-   { 13011, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING_ARB */
-   { 13045, 0x00008091 }, /* GL_INDEX_ARRAY_POINTER */
-   { 13068, 0x00008086 }, /* GL_INDEX_ARRAY_STRIDE */
-   { 13090, 0x00008085 }, /* GL_INDEX_ARRAY_TYPE */
-   { 13110, 0x00000D51 }, /* GL_INDEX_BITS */
-   { 13124, 0x00000C20 }, /* GL_INDEX_CLEAR_VALUE */
-   { 13145, 0x00000BF1 }, /* GL_INDEX_LOGIC_OP */
-   { 13163, 0x00000C30 }, /* GL_INDEX_MODE */
-   { 13177, 0x00000D13 }, /* GL_INDEX_OFFSET */
-   { 13193, 0x00000D12 }, /* GL_INDEX_SHIFT */
-   { 13208, 0x00000C21 }, /* GL_INDEX_WRITEMASK */
-   { 13227, 0x00008B84 }, /* GL_INFO_LOG_LENGTH */
-   { 13246, 0x00001404 }, /* GL_INT */
-   { 13253, 0x00008049 }, /* GL_INTENSITY */
-   { 13266, 0x0000804C }, /* GL_INTENSITY12 */
-   { 13281, 0x0000804C }, /* GL_INTENSITY12_EXT */
-   { 13300, 0x0000804D }, /* GL_INTENSITY16 */
-   { 13315, 0x0000804D }, /* GL_INTENSITY16_EXT */
-   { 13334, 0x0000804A }, /* GL_INTENSITY4 */
-   { 13348, 0x0000804A }, /* GL_INTENSITY4_EXT */
-   { 13366, 0x0000804B }, /* GL_INTENSITY8 */
-   { 13380, 0x0000804B }, /* GL_INTENSITY8_EXT */
-   { 13398, 0x00008049 }, /* GL_INTENSITY_EXT */
-   { 13415, 0x00008575 }, /* GL_INTERPOLATE */
-   { 13430, 0x00008575 }, /* GL_INTERPOLATE_ARB */
-   { 13449, 0x00008575 }, /* GL_INTERPOLATE_EXT */
-   { 13468, 0x00008B53 }, /* GL_INT_VEC2 */
-   { 13480, 0x00008B53 }, /* GL_INT_VEC2_ARB */
-   { 13496, 0x00008B54 }, /* GL_INT_VEC3 */
-   { 13508, 0x00008B54 }, /* GL_INT_VEC3_ARB */
-   { 13524, 0x00008B55 }, /* GL_INT_VEC4 */
-   { 13536, 0x00008B55 }, /* GL_INT_VEC4_ARB */
-   { 13552, 0x00000500 }, /* GL_INVALID_ENUM */
-   { 13568, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION */
-   { 13601, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
-   { 13638, 0x00000502 }, /* GL_INVALID_OPERATION */
-   { 13659, 0x00000501 }, /* GL_INVALID_VALUE */
-   { 13676, 0x0000862B }, /* GL_INVERSE_NV */
-   { 13690, 0x0000862D }, /* GL_INVERSE_TRANSPOSE_NV */
-   { 13714, 0x0000150A }, /* GL_INVERT */
-   { 13724, 0x00001E00 }, /* GL_KEEP */
-   { 13732, 0x00000406 }, /* GL_LEFT */
-   { 13740, 0x00000203 }, /* GL_LEQUAL */
-   { 13750, 0x00000201 }, /* GL_LESS */
-   { 13758, 0x00004000 }, /* GL_LIGHT0 */
-   { 13768, 0x00004001 }, /* GL_LIGHT1 */
-   { 13778, 0x00004002 }, /* GL_LIGHT2 */
-   { 13788, 0x00004003 }, /* GL_LIGHT3 */
-   { 13798, 0x00004004 }, /* GL_LIGHT4 */
-   { 13808, 0x00004005 }, /* GL_LIGHT5 */
-   { 13818, 0x00004006 }, /* GL_LIGHT6 */
-   { 13828, 0x00004007 }, /* GL_LIGHT7 */
-   { 13838, 0x00000B50 }, /* GL_LIGHTING */
-   { 13850, 0x00000040 }, /* GL_LIGHTING_BIT */
-   { 13866, 0x00000B53 }, /* GL_LIGHT_MODEL_AMBIENT */
-   { 13889, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL */
-   { 13918, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL_EXT */
-   { 13951, 0x00000B51 }, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
-   { 13979, 0x00000B52 }, /* GL_LIGHT_MODEL_TWO_SIDE */
-   { 14003, 0x00001B01 }, /* GL_LINE */
-   { 14011, 0x00002601 }, /* GL_LINEAR */
-   { 14021, 0x00001208 }, /* GL_LINEAR_ATTENUATION */
-   { 14043, 0x00008170 }, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
-   { 14073, 0x0000844F }, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
-   { 14104, 0x00002703 }, /* GL_LINEAR_MIPMAP_LINEAR */
-   { 14128, 0x00002701 }, /* GL_LINEAR_MIPMAP_NEAREST */
-   { 14153, 0x00000001 }, /* GL_LINES */
-   { 14162, 0x00000004 }, /* GL_LINE_BIT */
-   { 14174, 0x00000002 }, /* GL_LINE_LOOP */
-   { 14187, 0x00000707 }, /* GL_LINE_RESET_TOKEN */
-   { 14207, 0x00000B20 }, /* GL_LINE_SMOOTH */
-   { 14222, 0x00000C52 }, /* GL_LINE_SMOOTH_HINT */
-   { 14242, 0x00000B24 }, /* GL_LINE_STIPPLE */
-   { 14258, 0x00000B25 }, /* GL_LINE_STIPPLE_PATTERN */
-   { 14282, 0x00000B26 }, /* GL_LINE_STIPPLE_REPEAT */
-   { 14305, 0x00000003 }, /* GL_LINE_STRIP */
-   { 14319, 0x00000702 }, /* GL_LINE_TOKEN */
-   { 14333, 0x00000B21 }, /* GL_LINE_WIDTH */
-   { 14347, 0x00000B23 }, /* GL_LINE_WIDTH_GRANULARITY */
-   { 14373, 0x00000B22 }, /* GL_LINE_WIDTH_RANGE */
-   { 14393, 0x00008B82 }, /* GL_LINK_STATUS */
-   { 14408, 0x00000B32 }, /* GL_LIST_BASE */
-   { 14421, 0x00020000 }, /* GL_LIST_BIT */
-   { 14433, 0x00000B33 }, /* GL_LIST_INDEX */
-   { 14447, 0x00000B30 }, /* GL_LIST_MODE */
-   { 14460, 0x00000101 }, /* GL_LOAD */
-   { 14468, 0x00000BF1 }, /* GL_LOGIC_OP */
-   { 14480, 0x00000BF0 }, /* GL_LOGIC_OP_MODE */
-   { 14497, 0x00008CA1 }, /* GL_LOWER_LEFT */
-   { 14511, 0x00001909 }, /* GL_LUMINANCE */
-   { 14524, 0x00008041 }, /* GL_LUMINANCE12 */
-   { 14539, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12 */
-   { 14562, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12_EXT */
-   { 14589, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4 */
-   { 14611, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4_EXT */
-   { 14637, 0x00008041 }, /* GL_LUMINANCE12_EXT */
-   { 14656, 0x00008042 }, /* GL_LUMINANCE16 */
-   { 14671, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16 */
-   { 14694, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16_EXT */
-   { 14721, 0x00008042 }, /* GL_LUMINANCE16_EXT */
-   { 14740, 0x0000803F }, /* GL_LUMINANCE4 */
-   { 14754, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4 */
-   { 14775, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4_EXT */
-   { 14800, 0x0000803F }, /* GL_LUMINANCE4_EXT */
-   { 14818, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2 */
-   { 14839, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2_EXT */
-   { 14864, 0x00008040 }, /* GL_LUMINANCE8 */
-   { 14878, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8 */
-   { 14899, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8_EXT */
-   { 14924, 0x00008040 }, /* GL_LUMINANCE8_EXT */
-   { 14942, 0x0000190A }, /* GL_LUMINANCE_ALPHA */
-   { 14961, 0x00000D90 }, /* GL_MAP1_COLOR_4 */
-   { 14977, 0x00000DD0 }, /* GL_MAP1_GRID_DOMAIN */
-   { 14997, 0x00000DD1 }, /* GL_MAP1_GRID_SEGMENTS */
-   { 15019, 0x00000D91 }, /* GL_MAP1_INDEX */
-   { 15033, 0x00000D92 }, /* GL_MAP1_NORMAL */
-   { 15048, 0x00000D93 }, /* GL_MAP1_TEXTURE_COORD_1 */
-   { 15072, 0x00000D94 }, /* GL_MAP1_TEXTURE_COORD_2 */
-   { 15096, 0x00000D95 }, /* GL_MAP1_TEXTURE_COORD_3 */
-   { 15120, 0x00000D96 }, /* GL_MAP1_TEXTURE_COORD_4 */
-   { 15144, 0x00000D97 }, /* GL_MAP1_VERTEX_3 */
-   { 15161, 0x00000D98 }, /* GL_MAP1_VERTEX_4 */
-   { 15178, 0x00008660 }, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
-   { 15206, 0x0000866A }, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
-   { 15235, 0x0000866B }, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
-   { 15264, 0x0000866C }, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
-   { 15293, 0x0000866D }, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
-   { 15322, 0x0000866E }, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
-   { 15351, 0x0000866F }, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
-   { 15380, 0x00008661 }, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
-   { 15408, 0x00008662 }, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
-   { 15436, 0x00008663 }, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
-   { 15464, 0x00008664 }, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
-   { 15492, 0x00008665 }, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
-   { 15520, 0x00008666 }, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
-   { 15548, 0x00008667 }, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
-   { 15576, 0x00008668 }, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
-   { 15604, 0x00008669 }, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
-   { 15632, 0x00000DB0 }, /* GL_MAP2_COLOR_4 */
-   { 15648, 0x00000DD2 }, /* GL_MAP2_GRID_DOMAIN */
-   { 15668, 0x00000DD3 }, /* GL_MAP2_GRID_SEGMENTS */
-   { 15690, 0x00000DB1 }, /* GL_MAP2_INDEX */
-   { 15704, 0x00000DB2 }, /* GL_MAP2_NORMAL */
-   { 15719, 0x00000DB3 }, /* GL_MAP2_TEXTURE_COORD_1 */
-   { 15743, 0x00000DB4 }, /* GL_MAP2_TEXTURE_COORD_2 */
-   { 15767, 0x00000DB5 }, /* GL_MAP2_TEXTURE_COORD_3 */
-   { 15791, 0x00000DB6 }, /* GL_MAP2_TEXTURE_COORD_4 */
-   { 15815, 0x00000DB7 }, /* GL_MAP2_VERTEX_3 */
-   { 15832, 0x00000DB8 }, /* GL_MAP2_VERTEX_4 */
-   { 15849, 0x00008670 }, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
-   { 15877, 0x0000867A }, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
-   { 15906, 0x0000867B }, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
-   { 15935, 0x0000867C }, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
-   { 15964, 0x0000867D }, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
-   { 15993, 0x0000867E }, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
-   { 16022, 0x0000867F }, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
-   { 16051, 0x00008671 }, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
-   { 16079, 0x00008672 }, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
-   { 16107, 0x00008673 }, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
-   { 16135, 0x00008674 }, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
-   { 16163, 0x00008675 }, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
-   { 16191, 0x00008676 }, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
-   { 16219, 0x00008677 }, /* GL_MAP2_VERTEX_ATTRIB7_4_NV */
-   { 16247, 0x00008678 }, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
-   { 16275, 0x00008679 }, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
-   { 16303, 0x00000D10 }, /* GL_MAP_COLOR */
-   { 16316, 0x00000D11 }, /* GL_MAP_STENCIL */
-   { 16331, 0x000088C0 }, /* GL_MATRIX0_ARB */
-   { 16346, 0x00008630 }, /* GL_MATRIX0_NV */
-   { 16360, 0x000088CA }, /* GL_MATRIX10_ARB */
-   { 16376, 0x000088CB }, /* GL_MATRIX11_ARB */
-   { 16392, 0x000088CC }, /* GL_MATRIX12_ARB */
-   { 16408, 0x000088CD }, /* GL_MATRIX13_ARB */
-   { 16424, 0x000088CE }, /* GL_MATRIX14_ARB */
-   { 16440, 0x000088CF }, /* GL_MATRIX15_ARB */
-   { 16456, 0x000088D0 }, /* GL_MATRIX16_ARB */
-   { 16472, 0x000088D1 }, /* GL_MATRIX17_ARB */
-   { 16488, 0x000088D2 }, /* GL_MATRIX18_ARB */
-   { 16504, 0x000088D3 }, /* GL_MATRIX19_ARB */
-   { 16520, 0x000088C1 }, /* GL_MATRIX1_ARB */
-   { 16535, 0x00008631 }, /* GL_MATRIX1_NV */
-   { 16549, 0x000088D4 }, /* GL_MATRIX20_ARB */
-   { 16565, 0x000088D5 }, /* GL_MATRIX21_ARB */
-   { 16581, 0x000088D6 }, /* GL_MATRIX22_ARB */
-   { 16597, 0x000088D7 }, /* GL_MATRIX23_ARB */
-   { 16613, 0x000088D8 }, /* GL_MATRIX24_ARB */
-   { 16629, 0x000088D9 }, /* GL_MATRIX25_ARB */
-   { 16645, 0x000088DA }, /* GL_MATRIX26_ARB */
-   { 16661, 0x000088DB }, /* GL_MATRIX27_ARB */
-   { 16677, 0x000088DC }, /* GL_MATRIX28_ARB */
-   { 16693, 0x000088DD }, /* GL_MATRIX29_ARB */
-   { 16709, 0x000088C2 }, /* GL_MATRIX2_ARB */
-   { 16724, 0x00008632 }, /* GL_MATRIX2_NV */
-   { 16738, 0x000088DE }, /* GL_MATRIX30_ARB */
-   { 16754, 0x000088DF }, /* GL_MATRIX31_ARB */
-   { 16770, 0x000088C3 }, /* GL_MATRIX3_ARB */
-   { 16785, 0x00008633 }, /* GL_MATRIX3_NV */
-   { 16799, 0x000088C4 }, /* GL_MATRIX4_ARB */
-   { 16814, 0x00008634 }, /* GL_MATRIX4_NV */
-   { 16828, 0x000088C5 }, /* GL_MATRIX5_ARB */
-   { 16843, 0x00008635 }, /* GL_MATRIX5_NV */
-   { 16857, 0x000088C6 }, /* GL_MATRIX6_ARB */
-   { 16872, 0x00008636 }, /* GL_MATRIX6_NV */
-   { 16886, 0x000088C7 }, /* GL_MATRIX7_ARB */
-   { 16901, 0x00008637 }, /* GL_MATRIX7_NV */
-   { 16915, 0x000088C8 }, /* GL_MATRIX8_ARB */
-   { 16930, 0x000088C9 }, /* GL_MATRIX9_ARB */
-   { 16945, 0x00008844 }, /* GL_MATRIX_INDEX_ARRAY_ARB */
-   { 16971, 0x00008849 }, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
-   { 17005, 0x00008846 }, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
-   { 17036, 0x00008848 }, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
-   { 17069, 0x00008847 }, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
-   { 17100, 0x00000BA0 }, /* GL_MATRIX_MODE */
-   { 17115, 0x00008840 }, /* GL_MATRIX_PALETTE_ARB */
-   { 17137, 0x00008008 }, /* GL_MAX */
-   { 17144, 0x00008073 }, /* GL_MAX_3D_TEXTURE_SIZE */
-   { 17167, 0x000088FF }, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
-   { 17199, 0x00000D35 }, /* GL_MAX_ATTRIB_STACK_DEPTH */
-   { 17225, 0x00000D3B }, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
-   { 17258, 0x00008177 }, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
-   { 17284, 0x00008178 }, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-   { 17318, 0x00000D32 }, /* GL_MAX_CLIP_PLANES */
-   { 17337, 0x00008CDF }, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
-   { 17366, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
-   { 17398, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI */
-   { 17434, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
-   { 17470, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB */
-   { 17510, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT */
-   { 17536, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT_EXT */
-   { 17566, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH */
-   { 17591, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH_EXT */
-   { 17620, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
-   { 17649, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB */
-   { 17682, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS */
-   { 17702, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ARB */
-   { 17726, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ATI */
-   { 17750, 0x000080E9 }, /* GL_MAX_ELEMENTS_INDICES */
-   { 17774, 0x000080E8 }, /* GL_MAX_ELEMENTS_VERTICES */
-   { 17799, 0x00000D30 }, /* GL_MAX_EVAL_ORDER */
-   { 17817, 0x00008008 }, /* GL_MAX_EXT */
-   { 17828, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
-   { 17863, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB */
-   { 17902, 0x00000D31 }, /* GL_MAX_LIGHTS */
-   { 17916, 0x00000B31 }, /* GL_MAX_LIST_NESTING */
-   { 17936, 0x00008841 }, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
-   { 17974, 0x00000D36 }, /* GL_MAX_MODELVIEW_STACK_DEPTH */
-   { 18003, 0x00000D37 }, /* GL_MAX_NAME_STACK_DEPTH */
-   { 18027, 0x00008842 }, /* GL_MAX_PALETTE_MATRICES_ARB */
-   { 18055, 0x00000D34 }, /* GL_MAX_PIXEL_MAP_TABLE */
-   { 18078, 0x000088B1 }, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
-   { 18115, 0x0000880B }, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
-   { 18151, 0x000088AD }, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
-   { 18178, 0x000088F5 }, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
-   { 18207, 0x000088B5 }, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
-   { 18241, 0x000088F4 }, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
-   { 18277, 0x000088F6 }, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
-   { 18304, 0x000088A1 }, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
-   { 18336, 0x000088B4 }, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
-   { 18372, 0x000088F8 }, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
-   { 18401, 0x000088F7 }, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
-   { 18430, 0x0000862F }, /* GL_MAX_PROGRAM_MATRICES_ARB */
-   { 18458, 0x0000862E }, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
-   { 18496, 0x000088B3 }, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-   { 18540, 0x0000880E }, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-   { 18583, 0x000088AF }, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
-   { 18617, 0x000088A3 }, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-   { 18656, 0x000088AB }, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
-   { 18693, 0x000088A7 }, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
-   { 18731, 0x00008810 }, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-   { 18774, 0x0000880F }, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-   { 18817, 0x000088A9 }, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
-   { 18847, 0x000088A5 }, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
-   { 18878, 0x0000880D }, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
-   { 18914, 0x0000880C }, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
-   { 18950, 0x00000D38 }, /* GL_MAX_PROJECTION_STACK_DEPTH */
-   { 18980, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
-   { 19014, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_NV */
-   { 19047, 0x000084E8 }, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
-   { 19076, 0x00008D57 }, /* GL_MAX_SAMPLES */
-   { 19091, 0x00008504 }, /* GL_MAX_SHININESS_NV */
-   { 19111, 0x00008505 }, /* GL_MAX_SPOT_EXPONENT_NV */
-   { 19135, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS */
-   { 19157, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS_ARB */
-   { 19183, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS */
-   { 19210, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS_ARB */
-   { 19241, 0x000084FD }, /* GL_MAX_TEXTURE_LOD_BIAS */
-   { 19265, 0x000084FF }, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
-   { 19299, 0x00000D33 }, /* GL_MAX_TEXTURE_SIZE */
-   { 19319, 0x00000D39 }, /* GL_MAX_TEXTURE_STACK_DEPTH */
-   { 19346, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS */
-   { 19367, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS_ARB */
-   { 19392, 0x0000862F }, /* GL_MAX_TRACK_MATRICES_NV */
-   { 19417, 0x0000862E }, /* GL_MAX_TRACK_MATRIX_STACK_DEPTH_NV */
-   { 19452, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS */
-   { 19474, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS_ARB */
-   { 19500, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS */
-   { 19522, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS_ARB */
-   { 19548, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
-   { 19582, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB */
-   { 19620, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
-   { 19653, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB */
-   { 19690, 0x000086A4 }, /* GL_MAX_VERTEX_UNITS_ARB */
-   { 19714, 0x00000D3A }, /* GL_MAX_VIEWPORT_DIMS */
-   { 19735, 0x00008007 }, /* GL_MIN */
-   { 19742, 0x0000802E }, /* GL_MINMAX */
-   { 19752, 0x0000802E }, /* GL_MINMAX_EXT */
-   { 19766, 0x0000802F }, /* GL_MINMAX_FORMAT */
-   { 19783, 0x0000802F }, /* GL_MINMAX_FORMAT_EXT */
-   { 19804, 0x00008030 }, /* GL_MINMAX_SINK */
-   { 19819, 0x00008030 }, /* GL_MINMAX_SINK_EXT */
-   { 19838, 0x00008007 }, /* GL_MIN_EXT */
-   { 19849, 0x00008370 }, /* GL_MIRRORED_REPEAT */
-   { 19868, 0x00008370 }, /* GL_MIRRORED_REPEAT_ARB */
-   { 19891, 0x00008370 }, /* GL_MIRRORED_REPEAT_IBM */
-   { 19914, 0x00008742 }, /* GL_MIRROR_CLAMP_ATI */
-   { 19934, 0x00008742 }, /* GL_MIRROR_CLAMP_EXT */
-   { 19954, 0x00008912 }, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
-   { 19984, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_ATI */
-   { 20012, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
-   { 20040, 0x00001700 }, /* GL_MODELVIEW */
-   { 20053, 0x00001700 }, /* GL_MODELVIEW0_ARB */
-   { 20071, 0x0000872A }, /* GL_MODELVIEW10_ARB */
-   { 20090, 0x0000872B }, /* GL_MODELVIEW11_ARB */
-   { 20109, 0x0000872C }, /* GL_MODELVIEW12_ARB */
-   { 20128, 0x0000872D }, /* GL_MODELVIEW13_ARB */
-   { 20147, 0x0000872E }, /* GL_MODELVIEW14_ARB */
-   { 20166, 0x0000872F }, /* GL_MODELVIEW15_ARB */
-   { 20185, 0x00008730 }, /* GL_MODELVIEW16_ARB */
-   { 20204, 0x00008731 }, /* GL_MODELVIEW17_ARB */
-   { 20223, 0x00008732 }, /* GL_MODELVIEW18_ARB */
-   { 20242, 0x00008733 }, /* GL_MODELVIEW19_ARB */
-   { 20261, 0x0000850A }, /* GL_MODELVIEW1_ARB */
-   { 20279, 0x00008734 }, /* GL_MODELVIEW20_ARB */
-   { 20298, 0x00008735 }, /* GL_MODELVIEW21_ARB */
-   { 20317, 0x00008736 }, /* GL_MODELVIEW22_ARB */
-   { 20336, 0x00008737 }, /* GL_MODELVIEW23_ARB */
-   { 20355, 0x00008738 }, /* GL_MODELVIEW24_ARB */
-   { 20374, 0x00008739 }, /* GL_MODELVIEW25_ARB */
-   { 20393, 0x0000873A }, /* GL_MODELVIEW26_ARB */
-   { 20412, 0x0000873B }, /* GL_MODELVIEW27_ARB */
-   { 20431, 0x0000873C }, /* GL_MODELVIEW28_ARB */
-   { 20450, 0x0000873D }, /* GL_MODELVIEW29_ARB */
-   { 20469, 0x00008722 }, /* GL_MODELVIEW2_ARB */
-   { 20487, 0x0000873E }, /* GL_MODELVIEW30_ARB */
-   { 20506, 0x0000873F }, /* GL_MODELVIEW31_ARB */
-   { 20525, 0x00008723 }, /* GL_MODELVIEW3_ARB */
-   { 20543, 0x00008724 }, /* GL_MODELVIEW4_ARB */
-   { 20561, 0x00008725 }, /* GL_MODELVIEW5_ARB */
-   { 20579, 0x00008726 }, /* GL_MODELVIEW6_ARB */
-   { 20597, 0x00008727 }, /* GL_MODELVIEW7_ARB */
-   { 20615, 0x00008728 }, /* GL_MODELVIEW8_ARB */
-   { 20633, 0x00008729 }, /* GL_MODELVIEW9_ARB */
-   { 20651, 0x00000BA6 }, /* GL_MODELVIEW_MATRIX */
-   { 20671, 0x00008629 }, /* GL_MODELVIEW_PROJECTION_NV */
-   { 20698, 0x00000BA3 }, /* GL_MODELVIEW_STACK_DEPTH */
-   { 20723, 0x00002100 }, /* GL_MODULATE */
-   { 20735, 0x00008744 }, /* GL_MODULATE_ADD_ATI */
-   { 20755, 0x00008745 }, /* GL_MODULATE_SIGNED_ADD_ATI */
-   { 20782, 0x00008746 }, /* GL_MODULATE_SUBTRACT_ATI */
-   { 20807, 0x00000103 }, /* GL_MULT */
-   { 20815, 0x0000809D }, /* GL_MULTISAMPLE */
-   { 20830, 0x000086B2 }, /* GL_MULTISAMPLE_3DFX */
-   { 20850, 0x0000809D }, /* GL_MULTISAMPLE_ARB */
-   { 20869, 0x20000000 }, /* GL_MULTISAMPLE_BIT */
-   { 20888, 0x20000000 }, /* GL_MULTISAMPLE_BIT_3DFX */
-   { 20912, 0x20000000 }, /* GL_MULTISAMPLE_BIT_ARB */
-   { 20935, 0x00008534 }, /* GL_MULTISAMPLE_FILTER_HINT_NV */
-   { 20965, 0x00002A25 }, /* GL_N3F_V3F */
-   { 20976, 0x00000D70 }, /* GL_NAME_STACK_DEPTH */
-   { 20996, 0x0000150E }, /* GL_NAND */
-   { 21004, 0x00002600 }, /* GL_NEAREST */
-   { 21015, 0x0000844E }, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
-   { 21046, 0x0000844D }, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
-   { 21078, 0x00002702 }, /* GL_NEAREST_MIPMAP_LINEAR */
-   { 21103, 0x00002700 }, /* GL_NEAREST_MIPMAP_NEAREST */
-   { 21129, 0x00000200 }, /* GL_NEVER */
-   { 21138, 0x00001102 }, /* GL_NICEST */
-   { 21148, 0x00000000 }, /* GL_NONE */
-   { 21156, 0x00001505 }, /* GL_NOOP */
-   { 21164, 0x00001508 }, /* GL_NOR */
-   { 21171, 0x00000BA1 }, /* GL_NORMALIZE */
-   { 21184, 0x00008075 }, /* GL_NORMAL_ARRAY */
-   { 21200, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
-   { 21231, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING_ARB */
-   { 21266, 0x0000808F }, /* GL_NORMAL_ARRAY_POINTER */
-   { 21290, 0x0000807F }, /* GL_NORMAL_ARRAY_STRIDE */
-   { 21313, 0x0000807E }, /* GL_NORMAL_ARRAY_TYPE */
-   { 21334, 0x00008511 }, /* GL_NORMAL_MAP */
-   { 21348, 0x00008511 }, /* GL_NORMAL_MAP_ARB */
-   { 21366, 0x00008511 }, /* GL_NORMAL_MAP_NV */
-   { 21383, 0x00000205 }, /* GL_NOTEQUAL */
-   { 21395, 0x00000000 }, /* GL_NO_ERROR */
-   { 21407, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
-   { 21441, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB */
-   { 21479, 0x00008B89 }, /* GL_OBJECT_ACTIVE_ATTRIBUTES_ARB */
-   { 21511, 0x00008B8A }, /* GL_OBJECT_ACTIVE_ATTRIBUTE_MAX_LENGTH_ARB */
-   { 21553, 0x00008B86 }, /* GL_OBJECT_ACTIVE_UNIFORMS_ARB */
-   { 21583, 0x00008B87 }, /* GL_OBJECT_ACTIVE_UNIFORM_MAX_LENGTH_ARB */
-   { 21623, 0x00008B85 }, /* GL_OBJECT_ATTACHED_OBJECTS_ARB */
-   { 21654, 0x00008B81 }, /* GL_OBJECT_COMPILE_STATUS_ARB */
-   { 21683, 0x00008B80 }, /* GL_OBJECT_DELETE_STATUS_ARB */
-   { 21711, 0x00008B84 }, /* GL_OBJECT_INFO_LOG_LENGTH_ARB */
-   { 21741, 0x00002401 }, /* GL_OBJECT_LINEAR */
-   { 21758, 0x00008B82 }, /* GL_OBJECT_LINK_STATUS_ARB */
-   { 21784, 0x00002501 }, /* GL_OBJECT_PLANE */
-   { 21800, 0x00008B88 }, /* GL_OBJECT_SHADER_SOURCE_LENGTH_ARB */
-   { 21835, 0x00008B4F }, /* GL_OBJECT_SUBTYPE_ARB */
-   { 21857, 0x00008B4E }, /* GL_OBJECT_TYPE_ARB */
-   { 21876, 0x00008B83 }, /* GL_OBJECT_VALIDATE_STATUS_ARB */
-   { 21906, 0x00008165 }, /* GL_OCCLUSION_TEST_HP */
-   { 21927, 0x00008166 }, /* GL_OCCLUSION_TEST_RESULT_HP */
-   { 21955, 0x00000001 }, /* GL_ONE */
-   { 21962, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA */
-   { 21990, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA_EXT */
-   { 22022, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR */
-   { 22050, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR_EXT */
-   { 22082, 0x00000305 }, /* GL_ONE_MINUS_DST_ALPHA */
-   { 22105, 0x00000307 }, /* GL_ONE_MINUS_DST_COLOR */
-   { 22128, 0x00000303 }, /* GL_ONE_MINUS_SRC_ALPHA */
-   { 22151, 0x00000301 }, /* GL_ONE_MINUS_SRC_COLOR */
-   { 22174, 0x00008598 }, /* GL_OPERAND0_ALPHA */
-   { 22192, 0x00008598 }, /* GL_OPERAND0_ALPHA_ARB */
-   { 22214, 0x00008598 }, /* GL_OPERAND0_ALPHA_EXT */
-   { 22236, 0x00008590 }, /* GL_OPERAND0_RGB */
-   { 22252, 0x00008590 }, /* GL_OPERAND0_RGB_ARB */
-   { 22272, 0x00008590 }, /* GL_OPERAND0_RGB_EXT */
-   { 22292, 0x00008599 }, /* GL_OPERAND1_ALPHA */
-   { 22310, 0x00008599 }, /* GL_OPERAND1_ALPHA_ARB */
-   { 22332, 0x00008599 }, /* GL_OPERAND1_ALPHA_EXT */
-   { 22354, 0x00008591 }, /* GL_OPERAND1_RGB */
-   { 22370, 0x00008591 }, /* GL_OPERAND1_RGB_ARB */
-   { 22390, 0x00008591 }, /* GL_OPERAND1_RGB_EXT */
-   { 22410, 0x0000859A }, /* GL_OPERAND2_ALPHA */
-   { 22428, 0x0000859A }, /* GL_OPERAND2_ALPHA_ARB */
-   { 22450, 0x0000859A }, /* GL_OPERAND2_ALPHA_EXT */
-   { 22472, 0x00008592 }, /* GL_OPERAND2_RGB */
-   { 22488, 0x00008592 }, /* GL_OPERAND2_RGB_ARB */
-   { 22508, 0x00008592 }, /* GL_OPERAND2_RGB_EXT */
-   { 22528, 0x0000859B }, /* GL_OPERAND3_ALPHA_NV */
-   { 22549, 0x00008593 }, /* GL_OPERAND3_RGB_NV */
-   { 22568, 0x00001507 }, /* GL_OR */
-   { 22574, 0x00000A01 }, /* GL_ORDER */
-   { 22583, 0x0000150D }, /* GL_OR_INVERTED */
-   { 22598, 0x0000150B }, /* GL_OR_REVERSE */
-   { 22612, 0x00000505 }, /* GL_OUT_OF_MEMORY */
-   { 22629, 0x00000D05 }, /* GL_PACK_ALIGNMENT */
-   { 22647, 0x0000806C }, /* GL_PACK_IMAGE_HEIGHT */
-   { 22668, 0x00008758 }, /* GL_PACK_INVERT_MESA */
-   { 22688, 0x00000D01 }, /* GL_PACK_LSB_FIRST */
-   { 22706, 0x00000D02 }, /* GL_PACK_ROW_LENGTH */
-   { 22725, 0x0000806B }, /* GL_PACK_SKIP_IMAGES */
-   { 22745, 0x00000D04 }, /* GL_PACK_SKIP_PIXELS */
-   { 22765, 0x00000D03 }, /* GL_PACK_SKIP_ROWS */
-   { 22783, 0x00000D00 }, /* GL_PACK_SWAP_BYTES */
-   { 22802, 0x00008B92 }, /* GL_PALETTE4_R5_G6_B5_OES */
-   { 22827, 0x00008B94 }, /* GL_PALETTE4_RGB5_A1_OES */
-   { 22851, 0x00008B90 }, /* GL_PALETTE4_RGB8_OES */
-   { 22872, 0x00008B93 }, /* GL_PALETTE4_RGBA4_OES */
-   { 22894, 0x00008B91 }, /* GL_PALETTE4_RGBA8_OES */
-   { 22916, 0x00008B97 }, /* GL_PALETTE8_R5_G6_B5_OES */
-   { 22941, 0x00008B99 }, /* GL_PALETTE8_RGB5_A1_OES */
-   { 22965, 0x00008B95 }, /* GL_PALETTE8_RGB8_OES */
-   { 22986, 0x00008B98 }, /* GL_PALETTE8_RGBA4_OES */
-   { 23008, 0x00008B96 }, /* GL_PALETTE8_RGBA8_OES */
-   { 23030, 0x00000700 }, /* GL_PASS_THROUGH_TOKEN */
-   { 23052, 0x00000C50 }, /* GL_PERSPECTIVE_CORRECTION_HINT */
-   { 23083, 0x00000C79 }, /* GL_PIXEL_MAP_A_TO_A */
-   { 23103, 0x00000CB9 }, /* GL_PIXEL_MAP_A_TO_A_SIZE */
-   { 23128, 0x00000C78 }, /* GL_PIXEL_MAP_B_TO_B */
-   { 23148, 0x00000CB8 }, /* GL_PIXEL_MAP_B_TO_B_SIZE */
-   { 23173, 0x00000C77 }, /* GL_PIXEL_MAP_G_TO_G */
-   { 23193, 0x00000CB7 }, /* GL_PIXEL_MAP_G_TO_G_SIZE */
-   { 23218, 0x00000C75 }, /* GL_PIXEL_MAP_I_TO_A */
-   { 23238, 0x00000CB5 }, /* GL_PIXEL_MAP_I_TO_A_SIZE */
-   { 23263, 0x00000C74 }, /* GL_PIXEL_MAP_I_TO_B */
-   { 23283, 0x00000CB4 }, /* GL_PIXEL_MAP_I_TO_B_SIZE */
-   { 23308, 0x00000C73 }, /* GL_PIXEL_MAP_I_TO_G */
-   { 23328, 0x00000CB3 }, /* GL_PIXEL_MAP_I_TO_G_SIZE */
-   { 23353, 0x00000C70 }, /* GL_PIXEL_MAP_I_TO_I */
-   { 23373, 0x00000CB0 }, /* GL_PIXEL_MAP_I_TO_I_SIZE */
-   { 23398, 0x00000C72 }, /* GL_PIXEL_MAP_I_TO_R */
-   { 23418, 0x00000CB2 }, /* GL_PIXEL_MAP_I_TO_R_SIZE */
-   { 23443, 0x00000C76 }, /* GL_PIXEL_MAP_R_TO_R */
-   { 23463, 0x00000CB6 }, /* GL_PIXEL_MAP_R_TO_R_SIZE */
-   { 23488, 0x00000C71 }, /* GL_PIXEL_MAP_S_TO_S */
-   { 23508, 0x00000CB1 }, /* GL_PIXEL_MAP_S_TO_S_SIZE */
-   { 23533, 0x00000020 }, /* GL_PIXEL_MODE_BIT */
-   { 23551, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER */
-   { 23572, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING */
-   { 23601, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING_EXT */
-   { 23634, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER_EXT */
-   { 23659, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER */
-   { 23682, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
-   { 23713, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING_EXT */
-   { 23748, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER_EXT */
-   { 23775, 0x00001B00 }, /* GL_POINT */
-   { 23784, 0x00000000 }, /* GL_POINTS */
-   { 23794, 0x00000002 }, /* GL_POINT_BIT */
-   { 23807, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION */
-   { 23837, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_ARB */
-   { 23871, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_EXT */
-   { 23905, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_SGIS */
-   { 23940, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE */
-   { 23969, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_ARB */
-   { 24002, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_EXT */
-   { 24035, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_SGIS */
-   { 24069, 0x00000B11 }, /* GL_POINT_SIZE */
-   { 24083, 0x00000B13 }, /* GL_POINT_SIZE_GRANULARITY */
-   { 24109, 0x00008127 }, /* GL_POINT_SIZE_MAX */
-   { 24127, 0x00008127 }, /* GL_POINT_SIZE_MAX_ARB */
-   { 24149, 0x00008127 }, /* GL_POINT_SIZE_MAX_EXT */
-   { 24171, 0x00008127 }, /* GL_POINT_SIZE_MAX_SGIS */
-   { 24194, 0x00008126 }, /* GL_POINT_SIZE_MIN */
-   { 24212, 0x00008126 }, /* GL_POINT_SIZE_MIN_ARB */
-   { 24234, 0x00008126 }, /* GL_POINT_SIZE_MIN_EXT */
-   { 24256, 0x00008126 }, /* GL_POINT_SIZE_MIN_SGIS */
-   { 24279, 0x00000B12 }, /* GL_POINT_SIZE_RANGE */
-   { 24299, 0x00000B10 }, /* GL_POINT_SMOOTH */
-   { 24315, 0x00000C51 }, /* GL_POINT_SMOOTH_HINT */
-   { 24336, 0x00008861 }, /* GL_POINT_SPRITE */
-   { 24352, 0x00008861 }, /* GL_POINT_SPRITE_ARB */
-   { 24372, 0x00008CA0 }, /* GL_POINT_SPRITE_COORD_ORIGIN */
-   { 24401, 0x00008861 }, /* GL_POINT_SPRITE_NV */
-   { 24420, 0x00008863 }, /* GL_POINT_SPRITE_R_MODE_NV */
-   { 24446, 0x00000701 }, /* GL_POINT_TOKEN */
-   { 24461, 0x00000009 }, /* GL_POLYGON */
-   { 24472, 0x00000008 }, /* GL_POLYGON_BIT */
-   { 24487, 0x00000B40 }, /* GL_POLYGON_MODE */
-   { 24503, 0x00008039 }, /* GL_POLYGON_OFFSET_BIAS */
-   { 24526, 0x00008038 }, /* GL_POLYGON_OFFSET_FACTOR */
-   { 24551, 0x00008037 }, /* GL_POLYGON_OFFSET_FILL */
-   { 24574, 0x00002A02 }, /* GL_POLYGON_OFFSET_LINE */
-   { 24597, 0x00002A01 }, /* GL_POLYGON_OFFSET_POINT */
-   { 24621, 0x00002A00 }, /* GL_POLYGON_OFFSET_UNITS */
-   { 24645, 0x00000B41 }, /* GL_POLYGON_SMOOTH */
-   { 24663, 0x00000C53 }, /* GL_POLYGON_SMOOTH_HINT */
-   { 24686, 0x00000B42 }, /* GL_POLYGON_STIPPLE */
-   { 24705, 0x00000010 }, /* GL_POLYGON_STIPPLE_BIT */
-   { 24728, 0x00000703 }, /* GL_POLYGON_TOKEN */
-   { 24745, 0x00001203 }, /* GL_POSITION */
-   { 24757, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
-   { 24789, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI */
-   { 24825, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
-   { 24858, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI */
-   { 24895, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
-   { 24926, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI */
-   { 24961, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
-   { 24993, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI */
-   { 25029, 0x000080D2 }, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
-   { 25062, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
-   { 25094, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI */
-   { 25130, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
-   { 25163, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI */
-   { 25200, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS */
-   { 25230, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS_SGI */
-   { 25264, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE */
-   { 25295, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE_SGI */
-   { 25330, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
-   { 25361, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS_EXT */
-   { 25396, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
-   { 25428, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE_EXT */
-   { 25464, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS */
-   { 25494, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS_EXT */
-   { 25528, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE */
-   { 25559, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE_EXT */
-   { 25594, 0x000080D1 }, /* GL_POST_CONVOLUTION_COLOR_TABLE */
-   { 25626, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS */
-   { 25657, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS_EXT */
-   { 25692, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE */
-   { 25724, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE_EXT */
-   { 25760, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS */
-   { 25789, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS_EXT */
-   { 25822, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE */
-   { 25852, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE_EXT */
-   { 25886, 0x0000817B }, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
-   { 25925, 0x00008179 }, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
-   { 25958, 0x0000817C }, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
-   { 25998, 0x0000817A }, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
-   { 26032, 0x00008578 }, /* GL_PREVIOUS */
-   { 26044, 0x00008578 }, /* GL_PREVIOUS_ARB */
-   { 26060, 0x00008578 }, /* GL_PREVIOUS_EXT */
-   { 26076, 0x00008577 }, /* GL_PRIMARY_COLOR */
-   { 26093, 0x00008577 }, /* GL_PRIMARY_COLOR_ARB */
-   { 26114, 0x00008577 }, /* GL_PRIMARY_COLOR_EXT */
-   { 26135, 0x000088B0 }, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
-   { 26168, 0x00008805 }, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
-   { 26200, 0x000088AC }, /* GL_PROGRAM_ATTRIBS_ARB */
-   { 26223, 0x00008677 }, /* GL_PROGRAM_BINDING_ARB */
-   { 26246, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_ARB */
-   { 26276, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_NV */
-   { 26305, 0x00008874 }, /* GL_PROGRAM_ERROR_STRING_ARB */
-   { 26333, 0x00008876 }, /* GL_PROGRAM_FORMAT_ARB */
-   { 26355, 0x00008875 }, /* GL_PROGRAM_FORMAT_ASCII_ARB */
-   { 26383, 0x000088A0 }, /* GL_PROGRAM_INSTRUCTIONS_ARB */
-   { 26411, 0x00008627 }, /* GL_PROGRAM_LENGTH_ARB */
-   { 26433, 0x00008627 }, /* GL_PROGRAM_LENGTH_NV */
-   { 26454, 0x000088B2 }, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-   { 26494, 0x00008808 }, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-   { 26533, 0x000088AE }, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
-   { 26563, 0x000088A2 }, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-   { 26598, 0x000088AA }, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
-   { 26631, 0x000088A6 }, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
-   { 26665, 0x0000880A }, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-   { 26704, 0x00008809 }, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-   { 26743, 0x00008B40 }, /* GL_PROGRAM_OBJECT_ARB */
-   { 26765, 0x000088A8 }, /* GL_PROGRAM_PARAMETERS_ARB */
-   { 26791, 0x00008644 }, /* GL_PROGRAM_PARAMETER_NV */
-   { 26815, 0x00008647 }, /* GL_PROGRAM_RESIDENT_NV */
-   { 26838, 0x00008628 }, /* GL_PROGRAM_STRING_ARB */
-   { 26860, 0x00008628 }, /* GL_PROGRAM_STRING_NV */
-   { 26881, 0x00008646 }, /* GL_PROGRAM_TARGET_NV */
-   { 26902, 0x000088A4 }, /* GL_PROGRAM_TEMPORARIES_ARB */
-   { 26929, 0x00008807 }, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
-   { 26961, 0x00008806 }, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
-   { 26993, 0x000088B6 }, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
-   { 27028, 0x00001701 }, /* GL_PROJECTION */
-   { 27042, 0x00000BA7 }, /* GL_PROJECTION_MATRIX */
-   { 27063, 0x00000BA4 }, /* GL_PROJECTION_STACK_DEPTH */
-   { 27089, 0x000080D3 }, /* GL_PROXY_COLOR_TABLE */
-   { 27110, 0x00008025 }, /* GL_PROXY_HISTOGRAM */
-   { 27129, 0x00008025 }, /* GL_PROXY_HISTOGRAM_EXT */
-   { 27152, 0x000080D5 }, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
-   { 27191, 0x000080D4 }, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
-   { 27229, 0x00008063 }, /* GL_PROXY_TEXTURE_1D */
-   { 27249, 0x00008C19 }, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
-   { 27279, 0x00008063 }, /* GL_PROXY_TEXTURE_1D_EXT */
-   { 27303, 0x00008064 }, /* GL_PROXY_TEXTURE_2D */
-   { 27323, 0x00008C1B }, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
-   { 27353, 0x00008064 }, /* GL_PROXY_TEXTURE_2D_EXT */
-   { 27377, 0x00008070 }, /* GL_PROXY_TEXTURE_3D */
-   { 27397, 0x000080BD }, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
-   { 27430, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP */
-   { 27456, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP_ARB */
-   { 27486, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
-   { 27517, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_NV */
-   { 27547, 0x00002003 }, /* GL_Q */
-   { 27552, 0x00001209 }, /* GL_QUADRATIC_ATTENUATION */
-   { 27577, 0x00000007 }, /* GL_QUADS */
-   { 27586, 0x00008614 }, /* GL_QUAD_MESH_SUN */
-   { 27603, 0x00000008 }, /* GL_QUAD_STRIP */
-   { 27617, 0x00008864 }, /* GL_QUERY_COUNTER_BITS */
-   { 27639, 0x00008864 }, /* GL_QUERY_COUNTER_BITS_ARB */
-   { 27665, 0x00008866 }, /* GL_QUERY_RESULT */
-   { 27681, 0x00008866 }, /* GL_QUERY_RESULT_ARB */
-   { 27701, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE */
-   { 27727, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE_ARB */
-   { 27757, 0x00002002 }, /* GL_R */
-   { 27762, 0x00002A10 }, /* GL_R3_G3_B2 */
-   { 27774, 0x00019262 }, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
-   { 27807, 0x00000C02 }, /* GL_READ_BUFFER */
-   { 27822, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER */
-   { 27842, 0x00008CAA }, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
-   { 27874, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER_EXT */
-   { 27898, 0x000088B8 }, /* GL_READ_ONLY */
-   { 27911, 0x000088B8 }, /* GL_READ_ONLY_ARB */
-   { 27928, 0x000088BA }, /* GL_READ_WRITE */
-   { 27942, 0x000088BA }, /* GL_READ_WRITE_ARB */
-   { 27960, 0x00001903 }, /* GL_RED */
-   { 27967, 0x00008016 }, /* GL_REDUCE */
-   { 27977, 0x00008016 }, /* GL_REDUCE_EXT */
-   { 27991, 0x00000D15 }, /* GL_RED_BIAS */
-   { 28003, 0x00000D52 }, /* GL_RED_BITS */
-   { 28015, 0x00000D14 }, /* GL_RED_SCALE */
-   { 28028, 0x00008512 }, /* GL_REFLECTION_MAP */
-   { 28046, 0x00008512 }, /* GL_REFLECTION_MAP_ARB */
-   { 28068, 0x00008512 }, /* GL_REFLECTION_MAP_NV */
-   { 28089, 0x00001C00 }, /* GL_RENDER */
-   { 28099, 0x00008D41 }, /* GL_RENDERBUFFER */
-   { 28115, 0x00008D53 }, /* GL_RENDERBUFFER_ALPHA_SIZE */
-   { 28142, 0x00008CA7 }, /* GL_RENDERBUFFER_BINDING_EXT */
-   { 28170, 0x00008D52 }, /* GL_RENDERBUFFER_BLUE_SIZE */
-   { 28196, 0x00008D54 }, /* GL_RENDERBUFFER_DEPTH_SIZE */
-   { 28223, 0x00008D41 }, /* GL_RENDERBUFFER_EXT */
-   { 28243, 0x00008D51 }, /* GL_RENDERBUFFER_GREEN_SIZE */
-   { 28270, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT */
-   { 28293, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT_EXT */
-   { 28320, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
-   { 28352, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
-   { 28388, 0x00008D50 }, /* GL_RENDERBUFFER_RED_SIZE */
-   { 28413, 0x00008CAB }, /* GL_RENDERBUFFER_SAMPLES */
-   { 28437, 0x00008D55 }, /* GL_RENDERBUFFER_STENCIL_SIZE */
-   { 28466, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH */
-   { 28488, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH_EXT */
-   { 28514, 0x00001F01 }, /* GL_RENDERER */
-   { 28526, 0x00000C40 }, /* GL_RENDER_MODE */
-   { 28541, 0x00002901 }, /* GL_REPEAT */
-   { 28551, 0x00001E01 }, /* GL_REPLACE */
-   { 28562, 0x00008062 }, /* GL_REPLACE_EXT */
-   { 28577, 0x00008153 }, /* GL_REPLICATE_BORDER_HP */
-   { 28600, 0x0000803A }, /* GL_RESCALE_NORMAL */
-   { 28618, 0x0000803A }, /* GL_RESCALE_NORMAL_EXT */
-   { 28640, 0x00000102 }, /* GL_RETURN */
-   { 28650, 0x00001907 }, /* GL_RGB */
-   { 28657, 0x00008052 }, /* GL_RGB10 */
-   { 28666, 0x00008059 }, /* GL_RGB10_A2 */
-   { 28678, 0x00008059 }, /* GL_RGB10_A2_EXT */
-   { 28694, 0x00008052 }, /* GL_RGB10_EXT */
-   { 28707, 0x00008053 }, /* GL_RGB12 */
-   { 28716, 0x00008053 }, /* GL_RGB12_EXT */
-   { 28729, 0x00008054 }, /* GL_RGB16 */
-   { 28738, 0x00008054 }, /* GL_RGB16_EXT */
-   { 28751, 0x0000804E }, /* GL_RGB2_EXT */
-   { 28763, 0x0000804F }, /* GL_RGB4 */
-   { 28771, 0x0000804F }, /* GL_RGB4_EXT */
-   { 28783, 0x000083A1 }, /* GL_RGB4_S3TC */
-   { 28796, 0x00008050 }, /* GL_RGB5 */
-   { 28804, 0x00008057 }, /* GL_RGB5_A1 */
-   { 28815, 0x00008057 }, /* GL_RGB5_A1_EXT */
-   { 28830, 0x00008050 }, /* GL_RGB5_EXT */
-   { 28842, 0x00008051 }, /* GL_RGB8 */
-   { 28850, 0x00008051 }, /* GL_RGB8_EXT */
-   { 28862, 0x00001908 }, /* GL_RGBA */
-   { 28870, 0x0000805A }, /* GL_RGBA12 */
-   { 28880, 0x0000805A }, /* GL_RGBA12_EXT */
-   { 28894, 0x0000805B }, /* GL_RGBA16 */
-   { 28904, 0x0000805B }, /* GL_RGBA16_EXT */
-   { 28918, 0x00008055 }, /* GL_RGBA2 */
-   { 28927, 0x00008055 }, /* GL_RGBA2_EXT */
-   { 28940, 0x00008056 }, /* GL_RGBA4 */
-   { 28949, 0x000083A5 }, /* GL_RGBA4_DXT5_S3TC */
-   { 28968, 0x00008056 }, /* GL_RGBA4_EXT */
-   { 28981, 0x000083A3 }, /* GL_RGBA4_S3TC */
-   { 28995, 0x00008058 }, /* GL_RGBA8 */
-   { 29004, 0x00008058 }, /* GL_RGBA8_EXT */
-   { 29017, 0x00008F97 }, /* GL_RGBA8_SNORM */
-   { 29032, 0x000083A4 }, /* GL_RGBA_DXT5_S3TC */
-   { 29050, 0x00000C31 }, /* GL_RGBA_MODE */
-   { 29063, 0x000083A2 }, /* GL_RGBA_S3TC */
-   { 29076, 0x00008F93 }, /* GL_RGBA_SNORM */
-   { 29090, 0x000083A0 }, /* GL_RGB_S3TC */
-   { 29102, 0x00008573 }, /* GL_RGB_SCALE */
-   { 29115, 0x00008573 }, /* GL_RGB_SCALE_ARB */
-   { 29132, 0x00008573 }, /* GL_RGB_SCALE_EXT */
-   { 29149, 0x00000407 }, /* GL_RIGHT */
-   { 29158, 0x00002000 }, /* GL_S */
-   { 29163, 0x00008B5D }, /* GL_SAMPLER_1D */
-   { 29177, 0x00008B61 }, /* GL_SAMPLER_1D_SHADOW */
-   { 29198, 0x00008B5E }, /* GL_SAMPLER_2D */
-   { 29212, 0x00008B62 }, /* GL_SAMPLER_2D_SHADOW */
-   { 29233, 0x00008B5F }, /* GL_SAMPLER_3D */
-   { 29247, 0x00008B60 }, /* GL_SAMPLER_CUBE */
-   { 29263, 0x000080A9 }, /* GL_SAMPLES */
-   { 29274, 0x000086B4 }, /* GL_SAMPLES_3DFX */
-   { 29290, 0x000080A9 }, /* GL_SAMPLES_ARB */
-   { 29305, 0x00008914 }, /* GL_SAMPLES_PASSED */
-   { 29323, 0x00008914 }, /* GL_SAMPLES_PASSED_ARB */
-   { 29345, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
-   { 29373, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE_ARB */
-   { 29405, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE */
-   { 29428, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE_ARB */
-   { 29455, 0x000080A8 }, /* GL_SAMPLE_BUFFERS */
-   { 29473, 0x000086B3 }, /* GL_SAMPLE_BUFFERS_3DFX */
-   { 29496, 0x000080A8 }, /* GL_SAMPLE_BUFFERS_ARB */
-   { 29518, 0x000080A0 }, /* GL_SAMPLE_COVERAGE */
-   { 29537, 0x000080A0 }, /* GL_SAMPLE_COVERAGE_ARB */
-   { 29560, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT */
-   { 29586, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT_ARB */
-   { 29616, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE */
-   { 29641, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE_ARB */
-   { 29670, 0x00080000 }, /* GL_SCISSOR_BIT */
-   { 29685, 0x00000C10 }, /* GL_SCISSOR_BOX */
-   { 29700, 0x00000C11 }, /* GL_SCISSOR_TEST */
-   { 29716, 0x0000845E }, /* GL_SECONDARY_COLOR_ARRAY */
-   { 29741, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
-   { 29781, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB */
-   { 29825, 0x0000845D }, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
-   { 29858, 0x0000845A }, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
-   { 29888, 0x0000845C }, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
-   { 29920, 0x0000845B }, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
-   { 29950, 0x00001C02 }, /* GL_SELECT */
-   { 29960, 0x00000DF3 }, /* GL_SELECTION_BUFFER_POINTER */
-   { 29988, 0x00000DF4 }, /* GL_SELECTION_BUFFER_SIZE */
-   { 30013, 0x00008012 }, /* GL_SEPARABLE_2D */
-   { 30029, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR */
-   { 30056, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR_EXT */
-   { 30087, 0x0000150F }, /* GL_SET */
-   { 30094, 0x00008B48 }, /* GL_SHADER_OBJECT_ARB */
-   { 30115, 0x00008B88 }, /* GL_SHADER_SOURCE_LENGTH */
-   { 30139, 0x00008B4F }, /* GL_SHADER_TYPE */
-   { 30154, 0x00000B54 }, /* GL_SHADE_MODEL */
-   { 30169, 0x00008B8C }, /* GL_SHADING_LANGUAGE_VERSION */
-   { 30197, 0x000080BF }, /* GL_SHADOW_AMBIENT_SGIX */
-   { 30220, 0x000081FB }, /* GL_SHARED_TEXTURE_PALETTE_EXT */
-   { 30250, 0x00001601 }, /* GL_SHININESS */
-   { 30263, 0x00001402 }, /* GL_SHORT */
-   { 30272, 0x00008F9C }, /* GL_SIGNED_NORMALIZED */
-   { 30293, 0x000081F9 }, /* GL_SINGLE_COLOR */
-   { 30309, 0x000081F9 }, /* GL_SINGLE_COLOR_EXT */
-   { 30329, 0x000085CC }, /* GL_SLICE_ACCUM_SUN */
-   { 30348, 0x00008C46 }, /* GL_SLUMINANCE */
-   { 30362, 0x00008C47 }, /* GL_SLUMINANCE8 */
-   { 30377, 0x00008C45 }, /* GL_SLUMINANCE8_ALPHA8 */
-   { 30399, 0x00008C44 }, /* GL_SLUMINANCE_ALPHA */
-   { 30419, 0x00001D01 }, /* GL_SMOOTH */
-   { 30429, 0x00000B23 }, /* GL_SMOOTH_LINE_WIDTH_GRANULARITY */
-   { 30462, 0x00000B22 }, /* GL_SMOOTH_LINE_WIDTH_RANGE */
-   { 30489, 0x00000B13 }, /* GL_SMOOTH_POINT_SIZE_GRANULARITY */
-   { 30522, 0x00000B12 }, /* GL_SMOOTH_POINT_SIZE_RANGE */
-   { 30549, 0x00008588 }, /* GL_SOURCE0_ALPHA */
-   { 30566, 0x00008588 }, /* GL_SOURCE0_ALPHA_ARB */
-   { 30587, 0x00008588 }, /* GL_SOURCE0_ALPHA_EXT */
-   { 30608, 0x00008580 }, /* GL_SOURCE0_RGB */
-   { 30623, 0x00008580 }, /* GL_SOURCE0_RGB_ARB */
-   { 30642, 0x00008580 }, /* GL_SOURCE0_RGB_EXT */
-   { 30661, 0x00008589 }, /* GL_SOURCE1_ALPHA */
-   { 30678, 0x00008589 }, /* GL_SOURCE1_ALPHA_ARB */
-   { 30699, 0x00008589 }, /* GL_SOURCE1_ALPHA_EXT */
-   { 30720, 0x00008581 }, /* GL_SOURCE1_RGB */
-   { 30735, 0x00008581 }, /* GL_SOURCE1_RGB_ARB */
-   { 30754, 0x00008581 }, /* GL_SOURCE1_RGB_EXT */
-   { 30773, 0x0000858A }, /* GL_SOURCE2_ALPHA */
-   { 30790, 0x0000858A }, /* GL_SOURCE2_ALPHA_ARB */
-   { 30811, 0x0000858A }, /* GL_SOURCE2_ALPHA_EXT */
-   { 30832, 0x00008582 }, /* GL_SOURCE2_RGB */
-   { 30847, 0x00008582 }, /* GL_SOURCE2_RGB_ARB */
-   { 30866, 0x00008582 }, /* GL_SOURCE2_RGB_EXT */
-   { 30885, 0x0000858B }, /* GL_SOURCE3_ALPHA_NV */
-   { 30905, 0x00008583 }, /* GL_SOURCE3_RGB_NV */
-   { 30923, 0x00001202 }, /* GL_SPECULAR */
-   { 30935, 0x00002402 }, /* GL_SPHERE_MAP */
-   { 30949, 0x00001206 }, /* GL_SPOT_CUTOFF */
-   { 30964, 0x00001204 }, /* GL_SPOT_DIRECTION */
-   { 30982, 0x00001205 }, /* GL_SPOT_EXPONENT */
-   { 30999, 0x00008588 }, /* GL_SRC0_ALPHA */
-   { 31013, 0x00008580 }, /* GL_SRC0_RGB */
-   { 31025, 0x00008589 }, /* GL_SRC1_ALPHA */
-   { 31039, 0x00008581 }, /* GL_SRC1_RGB */
-   { 31051, 0x0000858A }, /* GL_SRC2_ALPHA */
-   { 31065, 0x00008582 }, /* GL_SRC2_RGB */
-   { 31077, 0x00000302 }, /* GL_SRC_ALPHA */
-   { 31090, 0x00000308 }, /* GL_SRC_ALPHA_SATURATE */
-   { 31112, 0x00000300 }, /* GL_SRC_COLOR */
-   { 31125, 0x00008C40 }, /* GL_SRGB */
-   { 31133, 0x00008C41 }, /* GL_SRGB8 */
-   { 31142, 0x00008C43 }, /* GL_SRGB8_ALPHA8 */
-   { 31158, 0x00008C42 }, /* GL_SRGB_ALPHA */
-   { 31172, 0x00000503 }, /* GL_STACK_OVERFLOW */
-   { 31190, 0x00000504 }, /* GL_STACK_UNDERFLOW */
-   { 31209, 0x000088E6 }, /* GL_STATIC_COPY */
-   { 31224, 0x000088E6 }, /* GL_STATIC_COPY_ARB */
-   { 31243, 0x000088E4 }, /* GL_STATIC_DRAW */
-   { 31258, 0x000088E4 }, /* GL_STATIC_DRAW_ARB */
-   { 31277, 0x000088E5 }, /* GL_STATIC_READ */
-   { 31292, 0x000088E5 }, /* GL_STATIC_READ_ARB */
-   { 31311, 0x00001802 }, /* GL_STENCIL */
-   { 31322, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT */
-   { 31344, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */
-   { 31370, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */
-   { 31391, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */
-   { 31416, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */
-   { 31437, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */
-   { 31462, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
-   { 31494, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */
-   { 31530, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
-   { 31562, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */
-   { 31598, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */
-   { 31618, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */
-   { 31645, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */
-   { 31671, 0x00000D57 }, /* GL_STENCIL_BITS */
-   { 31687, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */
-   { 31709, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */
-   { 31732, 0x00000B94 }, /* GL_STENCIL_FAIL */
-   { 31748, 0x00000B92 }, /* GL_STENCIL_FUNC */
-   { 31764, 0x00001901 }, /* GL_STENCIL_INDEX */
-   { 31781, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */
-   { 31804, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */
-   { 31826, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */
-   { 31848, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */
-   { 31870, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */
-   { 31891, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */
-   { 31918, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */
-   { 31945, 0x00000B97 }, /* GL_STENCIL_REF */
-   { 31960, 0x00000B90 }, /* GL_STENCIL_TEST */
-   { 31976, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
-   { 32005, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */
-   { 32027, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */
-   { 32048, 0x00000C33 }, /* GL_STEREO */
-   { 32058, 0x000088E2 }, /* GL_STREAM_COPY */
-   { 32073, 0x000088E2 }, /* GL_STREAM_COPY_ARB */
-   { 32092, 0x000088E0 }, /* GL_STREAM_DRAW */
-   { 32107, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */
-   { 32126, 0x000088E1 }, /* GL_STREAM_READ */
-   { 32141, 0x000088E1 }, /* GL_STREAM_READ_ARB */
-   { 32160, 0x00000D50 }, /* GL_SUBPIXEL_BITS */
-   { 32177, 0x000084E7 }, /* GL_SUBTRACT */
-   { 32189, 0x000084E7 }, /* GL_SUBTRACT_ARB */
-   { 32205, 0x00002001 }, /* GL_T */
-   { 32210, 0x00002A2A }, /* GL_T2F_C3F_V3F */
-   { 32225, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */
-   { 32244, 0x00002A29 }, /* GL_T2F_C4UB_V3F */
-   { 32260, 0x00002A2B }, /* GL_T2F_N3F_V3F */
-   { 32275, 0x00002A27 }, /* GL_T2F_V3F */
-   { 32286, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */
-   { 32305, 0x00002A28 }, /* GL_T4F_V4F */
-   { 32316, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */
-   { 32339, 0x00001702 }, /* GL_TEXTURE */
-   { 32350, 0x000084C0 }, /* GL_TEXTURE0 */
-   { 32362, 0x000084C0 }, /* GL_TEXTURE0_ARB */
-   { 32378, 0x000084C1 }, /* GL_TEXTURE1 */
-   { 32390, 0x000084CA }, /* GL_TEXTURE10 */
-   { 32403, 0x000084CA }, /* GL_TEXTURE10_ARB */
-   { 32420, 0x000084CB }, /* GL_TEXTURE11 */
-   { 32433, 0x000084CB }, /* GL_TEXTURE11_ARB */
-   { 32450, 0x000084CC }, /* GL_TEXTURE12 */
-   { 32463, 0x000084CC }, /* GL_TEXTURE12_ARB */
-   { 32480, 0x000084CD }, /* GL_TEXTURE13 */
-   { 32493, 0x000084CD }, /* GL_TEXTURE13_ARB */
-   { 32510, 0x000084CE }, /* GL_TEXTURE14 */
-   { 32523, 0x000084CE }, /* GL_TEXTURE14_ARB */
-   { 32540, 0x000084CF }, /* GL_TEXTURE15 */
-   { 32553, 0x000084CF }, /* GL_TEXTURE15_ARB */
-   { 32570, 0x000084D0 }, /* GL_TEXTURE16 */
-   { 32583, 0x000084D0 }, /* GL_TEXTURE16_ARB */
-   { 32600, 0x000084D1 }, /* GL_TEXTURE17 */
-   { 32613, 0x000084D1 }, /* GL_TEXTURE17_ARB */
-   { 32630, 0x000084D2 }, /* GL_TEXTURE18 */
-   { 32643, 0x000084D2 }, /* GL_TEXTURE18_ARB */
-   { 32660, 0x000084D3 }, /* GL_TEXTURE19 */
-   { 32673, 0x000084D3 }, /* GL_TEXTURE19_ARB */
-   { 32690, 0x000084C1 }, /* GL_TEXTURE1_ARB */
-   { 32706, 0x000084C2 }, /* GL_TEXTURE2 */
-   { 32718, 0x000084D4 }, /* GL_TEXTURE20 */
-   { 32731, 0x000084D4 }, /* GL_TEXTURE20_ARB */
-   { 32748, 0x000084D5 }, /* GL_TEXTURE21 */
-   { 32761, 0x000084D5 }, /* GL_TEXTURE21_ARB */
-   { 32778, 0x000084D6 }, /* GL_TEXTURE22 */
-   { 32791, 0x000084D6 }, /* GL_TEXTURE22_ARB */
-   { 32808, 0x000084D7 }, /* GL_TEXTURE23 */
-   { 32821, 0x000084D7 }, /* GL_TEXTURE23_ARB */
-   { 32838, 0x000084D8 }, /* GL_TEXTURE24 */
-   { 32851, 0x000084D8 }, /* GL_TEXTURE24_ARB */
-   { 32868, 0x000084D9 }, /* GL_TEXTURE25 */
-   { 32881, 0x000084D9 }, /* GL_TEXTURE25_ARB */
-   { 32898, 0x000084DA }, /* GL_TEXTURE26 */
-   { 32911, 0x000084DA }, /* GL_TEXTURE26_ARB */
-   { 32928, 0x000084DB }, /* GL_TEXTURE27 */
-   { 32941, 0x000084DB }, /* GL_TEXTURE27_ARB */
-   { 32958, 0x000084DC }, /* GL_TEXTURE28 */
-   { 32971, 0x000084DC }, /* GL_TEXTURE28_ARB */
-   { 32988, 0x000084DD }, /* GL_TEXTURE29 */
-   { 33001, 0x000084DD }, /* GL_TEXTURE29_ARB */
-   { 33018, 0x000084C2 }, /* GL_TEXTURE2_ARB */
-   { 33034, 0x000084C3 }, /* GL_TEXTURE3 */
-   { 33046, 0x000084DE }, /* GL_TEXTURE30 */
-   { 33059, 0x000084DE }, /* GL_TEXTURE30_ARB */
-   { 33076, 0x000084DF }, /* GL_TEXTURE31 */
-   { 33089, 0x000084DF }, /* GL_TEXTURE31_ARB */
-   { 33106, 0x000084C3 }, /* GL_TEXTURE3_ARB */
-   { 33122, 0x000084C4 }, /* GL_TEXTURE4 */
-   { 33134, 0x000084C4 }, /* GL_TEXTURE4_ARB */
-   { 33150, 0x000084C5 }, /* GL_TEXTURE5 */
-   { 33162, 0x000084C5 }, /* GL_TEXTURE5_ARB */
-   { 33178, 0x000084C6 }, /* GL_TEXTURE6 */
-   { 33190, 0x000084C6 }, /* GL_TEXTURE6_ARB */
-   { 33206, 0x000084C7 }, /* GL_TEXTURE7 */
-   { 33218, 0x000084C7 }, /* GL_TEXTURE7_ARB */
-   { 33234, 0x000084C8 }, /* GL_TEXTURE8 */
-   { 33246, 0x000084C8 }, /* GL_TEXTURE8_ARB */
-   { 33262, 0x000084C9 }, /* GL_TEXTURE9 */
-   { 33274, 0x000084C9 }, /* GL_TEXTURE9_ARB */
-   { 33290, 0x00000DE0 }, /* GL_TEXTURE_1D */
-   { 33304, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */
-   { 33328, 0x00000DE1 }, /* GL_TEXTURE_2D */
-   { 33342, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */
-   { 33366, 0x0000806F }, /* GL_TEXTURE_3D */
-   { 33380, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */
-   { 33402, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */
-   { 33428, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */
-   { 33450, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */
-   { 33472, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
-   { 33504, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */
-   { 33526, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
-   { 33558, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */
-   { 33580, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */
-   { 33608, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */
-   { 33640, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
-   { 33673, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */
-   { 33705, 0x00040000 }, /* GL_TEXTURE_BIT */
-   { 33720, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */
-   { 33741, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */
-   { 33766, 0x00001005 }, /* GL_TEXTURE_BORDER */
-   { 33784, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */
-   { 33808, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
-   { 33839, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
-   { 33869, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
-   { 33899, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
-   { 33934, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
-   { 33965, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-   { 34003, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */
-   { 34030, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
-   { 34062, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-   { 34096, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */
-   { 34120, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */
-   { 34148, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */
-   { 34172, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */
-   { 34200, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
-   { 34233, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */
-   { 34257, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */
-   { 34279, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */
-   { 34301, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */
-   { 34327, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */
-   { 34361, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
-   { 34394, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */
-   { 34431, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */
-   { 34459, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */
-   { 34491, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */
-   { 34514, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
-   { 34552, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */
-   { 34594, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */
-   { 34625, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */
-   { 34653, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
-   { 34683, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */
-   { 34711, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */
-   { 34731, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */
-   { 34755, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
-   { 34786, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */
-   { 34821, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
-   { 34852, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */
-   { 34887, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
-   { 34918, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */
-   { 34953, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
-   { 34984, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */
-   { 35019, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
-   { 35050, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */
-   { 35085, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
-   { 35116, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */
-   { 35151, 0x00008071 }, /* GL_TEXTURE_DEPTH */
-   { 35168, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */
-   { 35190, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */
-   { 35216, 0x00002300 }, /* GL_TEXTURE_ENV */
-   { 35231, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */
-   { 35252, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */
-   { 35272, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */
-   { 35298, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */
-   { 35318, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */
-   { 35335, 0x00000C62 }, /* GL_TEXTURE_GEN_R */
-   { 35352, 0x00000C60 }, /* GL_TEXTURE_GEN_S */
-   { 35369, 0x00000C61 }, /* GL_TEXTURE_GEN_T */
-   { 35386, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */
-   { 35411, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */
-   { 35433, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */
-   { 35459, 0x00001001 }, /* GL_TEXTURE_HEIGHT */
-   { 35477, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */
-   { 35503, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */
-   { 35529, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */
-   { 35559, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */
-   { 35586, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */
-   { 35611, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */
-   { 35631, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */
-   { 35655, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
-   { 35682, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
-   { 35709, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
-   { 35736, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */
-   { 35762, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */
-   { 35792, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */
-   { 35814, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */
-   { 35832, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
-   { 35862, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
-   { 35890, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
-   { 35918, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
-   { 35946, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */
-   { 35967, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */
-   { 35986, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */
-   { 36008, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */
-   { 36027, 0x00008066 }, /* GL_TEXTURE_PRIORITY */
-   { 36047, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */
-   { 36072, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */
-   { 36096, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */
-   { 36116, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */
-   { 36140, 0x00008067 }, /* GL_TEXTURE_RESIDENT */
-   { 36160, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */
-   { 36183, 0x000088F1 }, /* GL_TEXTURE_STENCIL_SIZE */
-   { 36207, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */
-   { 36232, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
-   { 36266, 0x00001000 }, /* GL_TEXTURE_WIDTH */
-   { 36283, 0x00008072 }, /* GL_TEXTURE_WRAP_R */
-   { 36301, 0x00002802 }, /* GL_TEXTURE_WRAP_S */
-   { 36319, 0x00002803 }, /* GL_TEXTURE_WRAP_T */
-   { 36337, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */
-   { 36357, 0x00008648 }, /* GL_TRACK_MATRIX_NV */
-   { 36376, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */
-   { 36405, 0x00001000 }, /* GL_TRANSFORM_BIT */
-   { 36422, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */
-   { 36448, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */
-   { 36478, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
-   { 36510, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
-   { 36540, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */
-   { 36574, 0x0000862C }, /* GL_TRANSPOSE_NV */
-   { 36590, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */
-   { 36621, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */
-   { 36656, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */
-   { 36684, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */
-   { 36716, 0x00000004 }, /* GL_TRIANGLES */
-   { 36729, 0x00000006 }, /* GL_TRIANGLE_FAN */
-   { 36745, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */
-   { 36766, 0x00000005 }, /* GL_TRIANGLE_STRIP */
-   { 36784, 0x00000001 }, /* GL_TRUE */
-   { 36792, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */
-   { 36812, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */
-   { 36835, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */
-   { 36855, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */
-   { 36876, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */
-   { 36898, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */
-   { 36920, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */
-   { 36940, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */
-   { 36961, 0x00001401 }, /* GL_UNSIGNED_BYTE */
-   { 36978, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */
-   { 37005, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */
-   { 37028, 0x00001405 }, /* GL_UNSIGNED_INT */
-   { 37044, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */
-   { 37071, 0x000084FA }, /* GL_UNSIGNED_INT_24_8 */
-   { 37092, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */
-   { 37116, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */
-   { 37147, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */
-   { 37171, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */
-   { 37199, 0x00008C17 }, /* GL_UNSIGNED_NORMALIZED */
-   { 37222, 0x00001403 }, /* GL_UNSIGNED_SHORT */
-   { 37240, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
-   { 37270, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */
-   { 37296, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
-   { 37326, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */
-   { 37352, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */
-   { 37376, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */
-   { 37404, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */
-   { 37432, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */
-   { 37459, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
-   { 37491, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */
-   { 37522, 0x00008CA2 }, /* GL_UPPER_LEFT */
-   { 37536, 0x00002A20 }, /* GL_V2F */
-   { 37543, 0x00002A21 }, /* GL_V3F */
-   { 37550, 0x00008B83 }, /* GL_VALIDATE_STATUS */
-   { 37569, 0x00001F00 }, /* GL_VENDOR */
-   { 37579, 0x00001F02 }, /* GL_VERSION */
-   { 37590, 0x00008074 }, /* GL_VERTEX_ARRAY */
-   { 37606, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */
-   { 37636, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
-   { 37667, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */
-   { 37702, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */
-   { 37726, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */
-   { 37747, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */
-   { 37770, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */
-   { 37791, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
-   { 37818, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
-   { 37846, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
-   { 37874, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
-   { 37902, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
-   { 37930, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
-   { 37958, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
-   { 37986, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
-   { 38013, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
-   { 38040, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
-   { 38067, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
-   { 38094, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
-   { 38121, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
-   { 38148, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
-   { 38175, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
-   { 38202, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
-   { 38229, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
-   { 38267, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */
-   { 38309, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
-   { 38340, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */
-   { 38375, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
-   { 38409, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */
-   { 38447, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
-   { 38478, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */
-   { 38513, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
-   { 38541, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */
-   { 38573, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
-   { 38603, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */
-   { 38637, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
-   { 38665, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */
-   { 38697, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */
-   { 38717, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */
-   { 38739, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */
-   { 38768, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */
-   { 38789, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */
-   { 38818, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */
-   { 38851, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */
-   { 38883, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */
-   { 38910, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */
-   { 38941, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */
-   { 38971, 0x00008B31 }, /* GL_VERTEX_SHADER */
-   { 38988, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */
-   { 39009, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */
-   { 39036, 0x00000BA2 }, /* GL_VIEWPORT */
-   { 39048, 0x00000800 }, /* GL_VIEWPORT_BIT */
-   { 39064, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */
-   { 39084, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
-   { 39115, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */
-   { 39150, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */
-   { 39178, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */
-   { 39203, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
-   { 39230, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */
-   { 39255, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */
-   { 39279, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */
-   { 39298, 0x000088B9 }, /* GL_WRITE_ONLY */
-   { 39312, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */
-   { 39330, 0x00001506 }, /* GL_XOR */
-   { 39337, 0x000085B9 }, /* GL_YCBCR_422_APPLE */
-   { 39356, 0x00008757 }, /* GL_YCBCR_MESA */
-   { 39370, 0x00000000 }, /* GL_ZERO */
-   { 39378, 0x00000D16 }, /* GL_ZOOM_X */
-   { 39388, 0x00000D17 }, /* GL_ZOOM_Y */
+   {  6072, 0x00008F36 }, /* GL_COPY_READ_BUFFER */
+   {  6092, 0x00008F37 }, /* GL_COPY_WRITE_BUFFER */
+   {  6113, 0x00000B44 }, /* GL_CULL_FACE */
+   {  6126, 0x00000B45 }, /* GL_CULL_FACE_MODE */
+   {  6144, 0x000081AA }, /* GL_CULL_VERTEX_EXT */
+   {  6163, 0x000081AC }, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
+   {  6195, 0x000081AB }, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
+   {  6230, 0x00008626 }, /* GL_CURRENT_ATTRIB_NV */
+   {  6251, 0x00000001 }, /* GL_CURRENT_BIT */
+   {  6266, 0x00000B00 }, /* GL_CURRENT_COLOR */
+   {  6283, 0x00008453 }, /* GL_CURRENT_FOG_COORD */
+   {  6304, 0x00008453 }, /* GL_CURRENT_FOG_COORDINATE */
+   {  6330, 0x00000B01 }, /* GL_CURRENT_INDEX */
+   {  6347, 0x00008641 }, /* GL_CURRENT_MATRIX_ARB */
+   {  6369, 0x00008845 }, /* GL_CURRENT_MATRIX_INDEX_ARB */
+   {  6397, 0x00008641 }, /* GL_CURRENT_MATRIX_NV */
+   {  6418, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
+   {  6452, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_NV */
+   {  6485, 0x00000B02 }, /* GL_CURRENT_NORMAL */
+   {  6503, 0x00008843 }, /* GL_CURRENT_PALETTE_MATRIX_ARB */
+   {  6533, 0x00008B8D }, /* GL_CURRENT_PROGRAM */
+   {  6552, 0x00008865 }, /* GL_CURRENT_QUERY */
+   {  6569, 0x00008865 }, /* GL_CURRENT_QUERY_ARB */
+   {  6590, 0x00000B04 }, /* GL_CURRENT_RASTER_COLOR */
+   {  6614, 0x00000B09 }, /* GL_CURRENT_RASTER_DISTANCE */
+   {  6641, 0x00000B05 }, /* GL_CURRENT_RASTER_INDEX */
+   {  6665, 0x00000B07 }, /* GL_CURRENT_RASTER_POSITION */
+   {  6692, 0x00000B08 }, /* GL_CURRENT_RASTER_POSITION_VALID */
+   {  6725, 0x0000845F }, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
+   {  6759, 0x00000B06 }, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
+   {  6792, 0x00008459 }, /* GL_CURRENT_SECONDARY_COLOR */
+   {  6819, 0x00000B03 }, /* GL_CURRENT_TEXTURE_COORDS */
+   {  6845, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB */
+   {  6870, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB_ARB */
+   {  6899, 0x000086A8 }, /* GL_CURRENT_WEIGHT_ARB */
+   {  6921, 0x00000900 }, /* GL_CW */
+   {  6927, 0x0000875B }, /* GL_DEBUG_ASSERT_MESA */
+   {  6948, 0x00008759 }, /* GL_DEBUG_OBJECT_MESA */
+   {  6969, 0x0000875A }, /* GL_DEBUG_PRINT_MESA */
+   {  6989, 0x00002101 }, /* GL_DECAL */
+   {  6998, 0x00001E03 }, /* GL_DECR */
+   {  7006, 0x00008508 }, /* GL_DECR_WRAP */
+   {  7019, 0x00008508 }, /* GL_DECR_WRAP_EXT */
+   {  7036, 0x00008B80 }, /* GL_DELETE_STATUS */
+   {  7053, 0x00001801 }, /* GL_DEPTH */
+   {  7062, 0x000088F0 }, /* GL_DEPTH24_STENCIL8 */
+   {  7082, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT */
+   {  7102, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT_EXT */
+   {  7126, 0x00000D1F }, /* GL_DEPTH_BIAS */
+   {  7140, 0x00000D56 }, /* GL_DEPTH_BITS */
+   {  7154, 0x00008891 }, /* GL_DEPTH_BOUNDS_EXT */
+   {  7174, 0x00008890 }, /* GL_DEPTH_BOUNDS_TEST_EXT */
+   {  7199, 0x00000100 }, /* GL_DEPTH_BUFFER_BIT */
+   {  7219, 0x0000864F }, /* GL_DEPTH_CLAMP_NV */
+   {  7237, 0x00000B73 }, /* GL_DEPTH_CLEAR_VALUE */
+   {  7258, 0x00001902 }, /* GL_DEPTH_COMPONENT */
+   {  7277, 0x000081A5 }, /* GL_DEPTH_COMPONENT16 */
+   {  7298, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_ARB */
+   {  7323, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_SGIX */
+   {  7349, 0x000081A6 }, /* GL_DEPTH_COMPONENT24 */
+   {  7370, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_ARB */
+   {  7395, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_SGIX */
+   {  7421, 0x000081A7 }, /* GL_DEPTH_COMPONENT32 */
+   {  7442, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_ARB */
+   {  7467, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_SGIX */
+   {  7493, 0x00000B74 }, /* GL_DEPTH_FUNC */
+   {  7507, 0x00000B70 }, /* GL_DEPTH_RANGE */
+   {  7522, 0x00000D1E }, /* GL_DEPTH_SCALE */
+   {  7537, 0x000084F9 }, /* GL_DEPTH_STENCIL */
+   {  7554, 0x0000821A }, /* GL_DEPTH_STENCIL_ATTACHMENT */
+   {  7582, 0x000084F9 }, /* GL_DEPTH_STENCIL_NV */
+   {  7602, 0x0000886F }, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
+   {  7630, 0x0000886E }, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
+   {  7658, 0x00000B71 }, /* GL_DEPTH_TEST */
+   {  7672, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE */
+   {  7694, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE_ARB */
+   {  7720, 0x00000B72 }, /* GL_DEPTH_WRITEMASK */
+   {  7739, 0x00001201 }, /* GL_DIFFUSE */
+   {  7750, 0x00000BD0 }, /* GL_DITHER */
+   {  7760, 0x00000A02 }, /* GL_DOMAIN */
+   {  7770, 0x00001100 }, /* GL_DONT_CARE */
+   {  7783, 0x000086AE }, /* GL_DOT3_RGB */
+   {  7795, 0x000086AF }, /* GL_DOT3_RGBA */
+   {  7808, 0x000086AF }, /* GL_DOT3_RGBA_ARB */
+   {  7825, 0x00008741 }, /* GL_DOT3_RGBA_EXT */
+   {  7842, 0x000086AE }, /* GL_DOT3_RGB_ARB */
+   {  7858, 0x00008740 }, /* GL_DOT3_RGB_EXT */
+   {  7874, 0x0000140A }, /* GL_DOUBLE */
+   {  7884, 0x00000C32 }, /* GL_DOUBLEBUFFER */
+   {  7900, 0x00000C01 }, /* GL_DRAW_BUFFER */
+   {  7915, 0x00008825 }, /* GL_DRAW_BUFFER0 */
+   {  7931, 0x00008825 }, /* GL_DRAW_BUFFER0_ARB */
+   {  7951, 0x00008825 }, /* GL_DRAW_BUFFER0_ATI */
+   {  7971, 0x00008826 }, /* GL_DRAW_BUFFER1 */
+   {  7987, 0x0000882F }, /* GL_DRAW_BUFFER10 */
+   {  8004, 0x0000882F }, /* GL_DRAW_BUFFER10_ARB */
+   {  8025, 0x0000882F }, /* GL_DRAW_BUFFER10_ATI */
+   {  8046, 0x00008830 }, /* GL_DRAW_BUFFER11 */
+   {  8063, 0x00008830 }, /* GL_DRAW_BUFFER11_ARB */
+   {  8084, 0x00008830 }, /* GL_DRAW_BUFFER11_ATI */
+   {  8105, 0x00008831 }, /* GL_DRAW_BUFFER12 */
+   {  8122, 0x00008831 }, /* GL_DRAW_BUFFER12_ARB */
+   {  8143, 0x00008831 }, /* GL_DRAW_BUFFER12_ATI */
+   {  8164, 0x00008832 }, /* GL_DRAW_BUFFER13 */
+   {  8181, 0x00008832 }, /* GL_DRAW_BUFFER13_ARB */
+   {  8202, 0x00008832 }, /* GL_DRAW_BUFFER13_ATI */
+   {  8223, 0x00008833 }, /* GL_DRAW_BUFFER14 */
+   {  8240, 0x00008833 }, /* GL_DRAW_BUFFER14_ARB */
+   {  8261, 0x00008833 }, /* GL_DRAW_BUFFER14_ATI */
+   {  8282, 0x00008834 }, /* GL_DRAW_BUFFER15 */
+   {  8299, 0x00008834 }, /* GL_DRAW_BUFFER15_ARB */
+   {  8320, 0x00008834 }, /* GL_DRAW_BUFFER15_ATI */
+   {  8341, 0x00008826 }, /* GL_DRAW_BUFFER1_ARB */
+   {  8361, 0x00008826 }, /* GL_DRAW_BUFFER1_ATI */
+   {  8381, 0x00008827 }, /* GL_DRAW_BUFFER2 */
+   {  8397, 0x00008827 }, /* GL_DRAW_BUFFER2_ARB */
+   {  8417, 0x00008827 }, /* GL_DRAW_BUFFER2_ATI */
+   {  8437, 0x00008828 }, /* GL_DRAW_BUFFER3 */
+   {  8453, 0x00008828 }, /* GL_DRAW_BUFFER3_ARB */
+   {  8473, 0x00008828 }, /* GL_DRAW_BUFFER3_ATI */
+   {  8493, 0x00008829 }, /* GL_DRAW_BUFFER4 */
+   {  8509, 0x00008829 }, /* GL_DRAW_BUFFER4_ARB */
+   {  8529, 0x00008829 }, /* GL_DRAW_BUFFER4_ATI */
+   {  8549, 0x0000882A }, /* GL_DRAW_BUFFER5 */
+   {  8565, 0x0000882A }, /* GL_DRAW_BUFFER5_ARB */
+   {  8585, 0x0000882A }, /* GL_DRAW_BUFFER5_ATI */
+   {  8605, 0x0000882B }, /* GL_DRAW_BUFFER6 */
+   {  8621, 0x0000882B }, /* GL_DRAW_BUFFER6_ARB */
+   {  8641, 0x0000882B }, /* GL_DRAW_BUFFER6_ATI */
+   {  8661, 0x0000882C }, /* GL_DRAW_BUFFER7 */
+   {  8677, 0x0000882C }, /* GL_DRAW_BUFFER7_ARB */
+   {  8697, 0x0000882C }, /* GL_DRAW_BUFFER7_ATI */
+   {  8717, 0x0000882D }, /* GL_DRAW_BUFFER8 */
+   {  8733, 0x0000882D }, /* GL_DRAW_BUFFER8_ARB */
+   {  8753, 0x0000882D }, /* GL_DRAW_BUFFER8_ATI */
+   {  8773, 0x0000882E }, /* GL_DRAW_BUFFER9 */
+   {  8789, 0x0000882E }, /* GL_DRAW_BUFFER9_ARB */
+   {  8809, 0x0000882E }, /* GL_DRAW_BUFFER9_ATI */
+   {  8829, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER */
+   {  8849, 0x00008CA6 }, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
+   {  8881, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER_EXT */
+   {  8905, 0x00000705 }, /* GL_DRAW_PIXEL_TOKEN */
+   {  8925, 0x00000304 }, /* GL_DST_ALPHA */
+   {  8938, 0x00000306 }, /* GL_DST_COLOR */
+   {  8951, 0x0000877A }, /* GL_DU8DV8_ATI */
+   {  8965, 0x00008779 }, /* GL_DUDV_ATI */
+   {  8977, 0x000088EA }, /* GL_DYNAMIC_COPY */
+   {  8993, 0x000088EA }, /* GL_DYNAMIC_COPY_ARB */
+   {  9013, 0x000088E8 }, /* GL_DYNAMIC_DRAW */
+   {  9029, 0x000088E8 }, /* GL_DYNAMIC_DRAW_ARB */
+   {  9049, 0x000088E9 }, /* GL_DYNAMIC_READ */
+   {  9065, 0x000088E9 }, /* GL_DYNAMIC_READ_ARB */
+   {  9085, 0x00000B43 }, /* GL_EDGE_FLAG */
+   {  9098, 0x00008079 }, /* GL_EDGE_FLAG_ARRAY */
+   {  9117, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
+   {  9151, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB */
+   {  9189, 0x00008093 }, /* GL_EDGE_FLAG_ARRAY_POINTER */
+   {  9216, 0x0000808C }, /* GL_EDGE_FLAG_ARRAY_STRIDE */
+   {  9242, 0x00008893 }, /* GL_ELEMENT_ARRAY_BUFFER */
+   {  9266, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
+   {  9298, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB */
+   {  9334, 0x00001600 }, /* GL_EMISSION */
+   {  9346, 0x00002000 }, /* GL_ENABLE_BIT */
+   {  9360, 0x00000202 }, /* GL_EQUAL */
+   {  9369, 0x00001509 }, /* GL_EQUIV */
+   {  9378, 0x00010000 }, /* GL_EVAL_BIT */
+   {  9390, 0x00000800 }, /* GL_EXP */
+   {  9397, 0x00000801 }, /* GL_EXP2 */
+   {  9405, 0x00001F03 }, /* GL_EXTENSIONS */
+   {  9419, 0x00002400 }, /* GL_EYE_LINEAR */
+   {  9433, 0x00002502 }, /* GL_EYE_PLANE */
+   {  9446, 0x0000855C }, /* GL_EYE_PLANE_ABSOLUTE_NV */
+   {  9471, 0x0000855B }, /* GL_EYE_RADIAL_NV */
+   {  9488, 0x00000000 }, /* GL_FALSE */
+   {  9497, 0x00001101 }, /* GL_FASTEST */
+   {  9508, 0x00001C01 }, /* GL_FEEDBACK */
+   {  9520, 0x00000DF0 }, /* GL_FEEDBACK_BUFFER_POINTER */
+   {  9547, 0x00000DF1 }, /* GL_FEEDBACK_BUFFER_SIZE */
+   {  9571, 0x00000DF2 }, /* GL_FEEDBACK_BUFFER_TYPE */
+   {  9595, 0x00001B02 }, /* GL_FILL */
+   {  9603, 0x00008E4D }, /* GL_FIRST_VERTEX_CONVENTION_EXT */
+   {  9634, 0x00001D00 }, /* GL_FLAT */
+   {  9642, 0x00001406 }, /* GL_FLOAT */
+   {  9651, 0x00008B5A }, /* GL_FLOAT_MAT2 */
+   {  9665, 0x00008B5A }, /* GL_FLOAT_MAT2_ARB */
+   {  9683, 0x00008B65 }, /* GL_FLOAT_MAT2x3 */
+   {  9699, 0x00008B66 }, /* GL_FLOAT_MAT2x4 */
+   {  9715, 0x00008B5B }, /* GL_FLOAT_MAT3 */
+   {  9729, 0x00008B5B }, /* GL_FLOAT_MAT3_ARB */
+   {  9747, 0x00008B67 }, /* GL_FLOAT_MAT3x2 */
+   {  9763, 0x00008B68 }, /* GL_FLOAT_MAT3x4 */
+   {  9779, 0x00008B5C }, /* GL_FLOAT_MAT4 */
+   {  9793, 0x00008B5C }, /* GL_FLOAT_MAT4_ARB */
+   {  9811, 0x00008B69 }, /* GL_FLOAT_MAT4x2 */
+   {  9827, 0x00008B6A }, /* GL_FLOAT_MAT4x3 */
+   {  9843, 0x00008B50 }, /* GL_FLOAT_VEC2 */
+   {  9857, 0x00008B50 }, /* GL_FLOAT_VEC2_ARB */
+   {  9875, 0x00008B51 }, /* GL_FLOAT_VEC3 */
+   {  9889, 0x00008B51 }, /* GL_FLOAT_VEC3_ARB */
+   {  9907, 0x00008B52 }, /* GL_FLOAT_VEC4 */
+   {  9921, 0x00008B52 }, /* GL_FLOAT_VEC4_ARB */
+   {  9939, 0x00000B60 }, /* GL_FOG */
+   {  9946, 0x00000080 }, /* GL_FOG_BIT */
+   {  9957, 0x00000B66 }, /* GL_FOG_COLOR */
+   {  9970, 0x00008451 }, /* GL_FOG_COORD */
+   {  9983, 0x00008451 }, /* GL_FOG_COORDINATE */
+   { 10001, 0x00008457 }, /* GL_FOG_COORDINATE_ARRAY */
+   { 10025, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
+   { 10064, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB */
+   { 10107, 0x00008456 }, /* GL_FOG_COORDINATE_ARRAY_POINTER */
+   { 10139, 0x00008455 }, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
+   { 10170, 0x00008454 }, /* GL_FOG_COORDINATE_ARRAY_TYPE */
+   { 10199, 0x00008450 }, /* GL_FOG_COORDINATE_SOURCE */
+   { 10224, 0x00008457 }, /* GL_FOG_COORD_ARRAY */
+   { 10243, 0x0000889D }, /* GL_FOG_COORD_ARRAY_BUFFER_BINDING */
+   { 10277, 0x00008456 }, /* GL_FOG_COORD_ARRAY_POINTER */
+   { 10304, 0x00008455 }, /* GL_FOG_COORD_ARRAY_STRIDE */
+   { 10330, 0x00008454 }, /* GL_FOG_COORD_ARRAY_TYPE */
+   { 10354, 0x00008450 }, /* GL_FOG_COORD_SRC */
+   { 10371, 0x00000B62 }, /* GL_FOG_DENSITY */
+   { 10386, 0x0000855A }, /* GL_FOG_DISTANCE_MODE_NV */
+   { 10410, 0x00000B64 }, /* GL_FOG_END */
+   { 10421, 0x00000C54 }, /* GL_FOG_HINT */
+   { 10433, 0x00000B61 }, /* GL_FOG_INDEX */
+   { 10446, 0x00000B65 }, /* GL_FOG_MODE */
+   { 10458, 0x00008198 }, /* GL_FOG_OFFSET_SGIX */
+   { 10477, 0x00008199 }, /* GL_FOG_OFFSET_VALUE_SGIX */
+   { 10502, 0x00000B63 }, /* GL_FOG_START */
+   { 10515, 0x00008452 }, /* GL_FRAGMENT_DEPTH */
+   { 10533, 0x00008804 }, /* GL_FRAGMENT_PROGRAM_ARB */
+   { 10557, 0x00008B30 }, /* GL_FRAGMENT_SHADER */
+   { 10576, 0x00008B30 }, /* GL_FRAGMENT_SHADER_ARB */
+   { 10599, 0x00008B8B }, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
+   { 10634, 0x00008D40 }, /* GL_FRAMEBUFFER */
+   { 10649, 0x00008215 }, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
+   { 10686, 0x00008214 }, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
+   { 10722, 0x00008210 }, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
+   { 10763, 0x00008211 }, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
+   { 10804, 0x00008216 }, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
+   { 10841, 0x00008213 }, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
+   { 10878, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
+   { 10916, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
+   { 10958, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
+   { 10996, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
+   { 11038, 0x00008212 }, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
+   { 11073, 0x00008217 }, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
+   { 11112, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
+   { 11161, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
+   { 11209, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
+   { 11261, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
+   { 11301, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT */
+   { 11345, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
+   { 11385, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
+   { 11429, 0x00008CA6 }, /* GL_FRAMEBUFFER_BINDING_EXT */
+   { 11456, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE */
+   { 11480, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE_EXT */
+   { 11508, 0x00008218 }, /* GL_FRAMEBUFFER_DEFAULT */
+   { 11531, 0x00008D40 }, /* GL_FRAMEBUFFER_EXT */
+   { 11550, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
+   { 11587, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
+   { 11628, 0x00008CD9 }, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
+   { 11669, 0x00008CDB }, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
+   { 11711, 0x00008CD8 }, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
+   { 11762, 0x00008CDA }, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
+   { 11800, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
+   { 11845, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
+   { 11894, 0x00008D56 }, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
+   { 11932, 0x00008CDC }, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
+   { 11974, 0x00008CDE }, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+   { 12006, 0x00008219 }, /* GL_FRAMEBUFFER_UNDEFINED */
+   { 12031, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED */
+   { 12058, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
+   { 12089, 0x00000404 }, /* GL_FRONT */
+   { 12098, 0x00000408 }, /* GL_FRONT_AND_BACK */
+   { 12116, 0x00000B46 }, /* GL_FRONT_FACE */
+   { 12130, 0x00000400 }, /* GL_FRONT_LEFT */
+   { 12144, 0x00000401 }, /* GL_FRONT_RIGHT */
+   { 12159, 0x00008006 }, /* GL_FUNC_ADD */
+   { 12171, 0x00008006 }, /* GL_FUNC_ADD_EXT */
+   { 12187, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT */
+   { 12212, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT_EXT */
+   { 12241, 0x0000800A }, /* GL_FUNC_SUBTRACT */
+   { 12258, 0x0000800A }, /* GL_FUNC_SUBTRACT_EXT */
+   { 12279, 0x00008191 }, /* GL_GENERATE_MIPMAP */
+   { 12298, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT */
+   { 12322, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT_SGIS */
+   { 12351, 0x00008191 }, /* GL_GENERATE_MIPMAP_SGIS */
+   { 12375, 0x00000206 }, /* GL_GEQUAL */
+   { 12385, 0x00000204 }, /* GL_GREATER */
+   { 12396, 0x00001904 }, /* GL_GREEN */
+   { 12405, 0x00000D19 }, /* GL_GREEN_BIAS */
+   { 12419, 0x00000D53 }, /* GL_GREEN_BITS */
+   { 12433, 0x00000D18 }, /* GL_GREEN_SCALE */
+   { 12448, 0x00008000 }, /* GL_HINT_BIT */
+   { 12460, 0x00008024 }, /* GL_HISTOGRAM */
+   { 12473, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE */
+   { 12497, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE_EXT */
+   { 12525, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE */
+   { 12548, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE_EXT */
+   { 12575, 0x00008024 }, /* GL_HISTOGRAM_EXT */
+   { 12592, 0x00008027 }, /* GL_HISTOGRAM_FORMAT */
+   { 12612, 0x00008027 }, /* GL_HISTOGRAM_FORMAT_EXT */
+   { 12636, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE */
+   { 12660, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE_EXT */
+   { 12688, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE */
+   { 12716, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE_EXT */
+   { 12748, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE */
+   { 12770, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE_EXT */
+   { 12796, 0x0000802D }, /* GL_HISTOGRAM_SINK */
+   { 12814, 0x0000802D }, /* GL_HISTOGRAM_SINK_EXT */
+   { 12836, 0x00008026 }, /* GL_HISTOGRAM_WIDTH */
+   { 12855, 0x00008026 }, /* GL_HISTOGRAM_WIDTH_EXT */
+   { 12878, 0x0000862A }, /* GL_IDENTITY_NV */
+   { 12893, 0x00008150 }, /* GL_IGNORE_BORDER_HP */
+   { 12913, 0x00008B9B }, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
+   { 12953, 0x00008B9A }, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
+   { 12991, 0x00001E02 }, /* GL_INCR */
+   { 12999, 0x00008507 }, /* GL_INCR_WRAP */
+   { 13012, 0x00008507 }, /* GL_INCR_WRAP_EXT */
+   { 13029, 0x00008222 }, /* GL_INDEX */
+   { 13038, 0x00008077 }, /* GL_INDEX_ARRAY */
+   { 13053, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING */
+   { 13083, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING_ARB */
+   { 13117, 0x00008091 }, /* GL_INDEX_ARRAY_POINTER */
+   { 13140, 0x00008086 }, /* GL_INDEX_ARRAY_STRIDE */
+   { 13162, 0x00008085 }, /* GL_INDEX_ARRAY_TYPE */
+   { 13182, 0x00000D51 }, /* GL_INDEX_BITS */
+   { 13196, 0x00000C20 }, /* GL_INDEX_CLEAR_VALUE */
+   { 13217, 0x00000BF1 }, /* GL_INDEX_LOGIC_OP */
+   { 13235, 0x00000C30 }, /* GL_INDEX_MODE */
+   { 13249, 0x00000D13 }, /* GL_INDEX_OFFSET */
+   { 13265, 0x00000D12 }, /* GL_INDEX_SHIFT */
+   { 13280, 0x00000C21 }, /* GL_INDEX_WRITEMASK */
+   { 13299, 0x00008B84 }, /* GL_INFO_LOG_LENGTH */
+   { 13318, 0x00001404 }, /* GL_INT */
+   { 13325, 0x00008049 }, /* GL_INTENSITY */
+   { 13338, 0x0000804C }, /* GL_INTENSITY12 */
+   { 13353, 0x0000804C }, /* GL_INTENSITY12_EXT */
+   { 13372, 0x0000804D }, /* GL_INTENSITY16 */
+   { 13387, 0x0000804D }, /* GL_INTENSITY16_EXT */
+   { 13406, 0x0000804A }, /* GL_INTENSITY4 */
+   { 13420, 0x0000804A }, /* GL_INTENSITY4_EXT */
+   { 13438, 0x0000804B }, /* GL_INTENSITY8 */
+   { 13452, 0x0000804B }, /* GL_INTENSITY8_EXT */
+   { 13470, 0x00008049 }, /* GL_INTENSITY_EXT */
+   { 13487, 0x00008575 }, /* GL_INTERPOLATE */
+   { 13502, 0x00008575 }, /* GL_INTERPOLATE_ARB */
+   { 13521, 0x00008575 }, /* GL_INTERPOLATE_EXT */
+   { 13540, 0x00008B53 }, /* GL_INT_VEC2 */
+   { 13552, 0x00008B53 }, /* GL_INT_VEC2_ARB */
+   { 13568, 0x00008B54 }, /* GL_INT_VEC3 */
+   { 13580, 0x00008B54 }, /* GL_INT_VEC3_ARB */
+   { 13596, 0x00008B55 }, /* GL_INT_VEC4 */
+   { 13608, 0x00008B55 }, /* GL_INT_VEC4_ARB */
+   { 13624, 0x00000500 }, /* GL_INVALID_ENUM */
+   { 13640, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION */
+   { 13673, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
+   { 13710, 0x00000502 }, /* GL_INVALID_OPERATION */
+   { 13731, 0x00000501 }, /* GL_INVALID_VALUE */
+   { 13748, 0x0000862B }, /* GL_INVERSE_NV */
+   { 13762, 0x0000862D }, /* GL_INVERSE_TRANSPOSE_NV */
+   { 13786, 0x0000150A }, /* GL_INVERT */
+   { 13796, 0x00001E00 }, /* GL_KEEP */
+   { 13804, 0x00008E4E }, /* GL_LAST_VERTEX_CONVENTION_EXT */
+   { 13834, 0x00000406 }, /* GL_LEFT */
+   { 13842, 0x00000203 }, /* GL_LEQUAL */
+   { 13852, 0x00000201 }, /* GL_LESS */
+   { 13860, 0x00004000 }, /* GL_LIGHT0 */
+   { 13870, 0x00004001 }, /* GL_LIGHT1 */
+   { 13880, 0x00004002 }, /* GL_LIGHT2 */
+   { 13890, 0x00004003 }, /* GL_LIGHT3 */
+   { 13900, 0x00004004 }, /* GL_LIGHT4 */
+   { 13910, 0x00004005 }, /* GL_LIGHT5 */
+   { 13920, 0x00004006 }, /* GL_LIGHT6 */
+   { 13930, 0x00004007 }, /* GL_LIGHT7 */
+   { 13940, 0x00000B50 }, /* GL_LIGHTING */
+   { 13952, 0x00000040 }, /* GL_LIGHTING_BIT */
+   { 13968, 0x00000B53 }, /* GL_LIGHT_MODEL_AMBIENT */
+   { 13991, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL */
+   { 14020, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL_EXT */
+   { 14053, 0x00000B51 }, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
+   { 14081, 0x00000B52 }, /* GL_LIGHT_MODEL_TWO_SIDE */
+   { 14105, 0x00001B01 }, /* GL_LINE */
+   { 14113, 0x00002601 }, /* GL_LINEAR */
+   { 14123, 0x00001208 }, /* GL_LINEAR_ATTENUATION */
+   { 14145, 0x00008170 }, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
+   { 14175, 0x0000844F }, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
+   { 14206, 0x00002703 }, /* GL_LINEAR_MIPMAP_LINEAR */
+   { 14230, 0x00002701 }, /* GL_LINEAR_MIPMAP_NEAREST */
+   { 14255, 0x00000001 }, /* GL_LINES */
+   { 14264, 0x00000004 }, /* GL_LINE_BIT */
+   { 14276, 0x00000002 }, /* GL_LINE_LOOP */
+   { 14289, 0x00000707 }, /* GL_LINE_RESET_TOKEN */
+   { 14309, 0x00000B20 }, /* GL_LINE_SMOOTH */
+   { 14324, 0x00000C52 }, /* GL_LINE_SMOOTH_HINT */
+   { 14344, 0x00000B24 }, /* GL_LINE_STIPPLE */
+   { 14360, 0x00000B25 }, /* GL_LINE_STIPPLE_PATTERN */
+   { 14384, 0x00000B26 }, /* GL_LINE_STIPPLE_REPEAT */
+   { 14407, 0x00000003 }, /* GL_LINE_STRIP */
+   { 14421, 0x00000702 }, /* GL_LINE_TOKEN */
+   { 14435, 0x00000B21 }, /* GL_LINE_WIDTH */
+   { 14449, 0x00000B23 }, /* GL_LINE_WIDTH_GRANULARITY */
+   { 14475, 0x00000B22 }, /* GL_LINE_WIDTH_RANGE */
+   { 14495, 0x00008B82 }, /* GL_LINK_STATUS */
+   { 14510, 0x00000B32 }, /* GL_LIST_BASE */
+   { 14523, 0x00020000 }, /* GL_LIST_BIT */
+   { 14535, 0x00000B33 }, /* GL_LIST_INDEX */
+   { 14549, 0x00000B30 }, /* GL_LIST_MODE */
+   { 14562, 0x00000101 }, /* GL_LOAD */
+   { 14570, 0x00000BF1 }, /* GL_LOGIC_OP */
+   { 14582, 0x00000BF0 }, /* GL_LOGIC_OP_MODE */
+   { 14599, 0x00008CA1 }, /* GL_LOWER_LEFT */
+   { 14613, 0x00001909 }, /* GL_LUMINANCE */
+   { 14626, 0x00008041 }, /* GL_LUMINANCE12 */
+   { 14641, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12 */
+   { 14664, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12_EXT */
+   { 14691, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4 */
+   { 14713, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4_EXT */
+   { 14739, 0x00008041 }, /* GL_LUMINANCE12_EXT */
+   { 14758, 0x00008042 }, /* GL_LUMINANCE16 */
+   { 14773, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16 */
+   { 14796, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16_EXT */
+   { 14823, 0x00008042 }, /* GL_LUMINANCE16_EXT */
+   { 14842, 0x0000803F }, /* GL_LUMINANCE4 */
+   { 14856, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4 */
+   { 14877, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4_EXT */
+   { 14902, 0x0000803F }, /* GL_LUMINANCE4_EXT */
+   { 14920, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2 */
+   { 14941, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2_EXT */
+   { 14966, 0x00008040 }, /* GL_LUMINANCE8 */
+   { 14980, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8 */
+   { 15001, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8_EXT */
+   { 15026, 0x00008040 }, /* GL_LUMINANCE8_EXT */
+   { 15044, 0x0000190A }, /* GL_LUMINANCE_ALPHA */
+   { 15063, 0x00000D90 }, /* GL_MAP1_COLOR_4 */
+   { 15079, 0x00000DD0 }, /* GL_MAP1_GRID_DOMAIN */
+   { 15099, 0x00000DD1 }, /* GL_MAP1_GRID_SEGMENTS */
+   { 15121, 0x00000D91 }, /* GL_MAP1_INDEX */
+   { 15135, 0x00000D92 }, /* GL_MAP1_NORMAL */
+   { 15150, 0x00000D93 }, /* GL_MAP1_TEXTURE_COORD_1 */
+   { 15174, 0x00000D94 }, /* GL_MAP1_TEXTURE_COORD_2 */
+   { 15198, 0x00000D95 }, /* GL_MAP1_TEXTURE_COORD_3 */
+   { 15222, 0x00000D96 }, /* GL_MAP1_TEXTURE_COORD_4 */
+   { 15246, 0x00000D97 }, /* GL_MAP1_VERTEX_3 */
+   { 15263, 0x00000D98 }, /* GL_MAP1_VERTEX_4 */
+   { 15280, 0x00008660 }, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
+   { 15308, 0x0000866A }, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
+   { 15337, 0x0000866B }, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
+   { 15366, 0x0000866C }, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
+   { 15395, 0x0000866D }, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
+   { 15424, 0x0000866E }, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
+   { 15453, 0x0000866F }, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
+   { 15482, 0x00008661 }, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
+   { 15510, 0x00008662 }, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
+   { 15538, 0x00008663 }, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
+   { 15566, 0x00008664 }, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
+   { 15594, 0x00008665 }, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
+   { 15622, 0x00008666 }, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
+   { 15650, 0x00008667 }, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
+   { 15678, 0x00008668 }, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
+   { 15706, 0x00008669 }, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
+   { 15734, 0x00000DB0 }, /* GL_MAP2_COLOR_4 */
+   { 15750, 0x00000DD2 }, /* GL_MAP2_GRID_DOMAIN */
+   { 15770, 0x00000DD3 }, /* GL_MAP2_GRID_SEGMENTS */
+   { 15792, 0x00000DB1 }, /* GL_MAP2_INDEX */
+   { 15806, 0x00000DB2 }, /* GL_MAP2_NORMAL */
+   { 15821, 0x00000DB3 }, /* GL_MAP2_TEXTURE_COORD_1 */
+   { 15845, 0x00000DB4 }, /* GL_MAP2_TEXTURE_COORD_2 */
+   { 15869, 0x00000DB5 }, /* GL_MAP2_TEXTURE_COORD_3 */
+   { 15893, 0x00000DB6 }, /* GL_MAP2_TEXTURE_COORD_4 */
+   { 15917, 0x00000DB7 }, /* GL_MAP2_VERTEX_3 */
+   { 15934, 0x00000DB8 }, /* GL_MAP2_VERTEX_4 */
+   { 15951, 0x00008670 }, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
+   { 15979, 0x0000867A }, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
+   { 16008, 0x0000867B }, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
+   { 16037, 0x0000867C }, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
+   { 16066, 0x0000867D }, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
+   { 16095, 0x0000867E }, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
+   { 16124, 0x0000867F }, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
+   { 16153, 0x00008671 }, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
+   { 16181, 0x00008672 }, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
+   { 16209, 0x00008673 }, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
+   { 16237, 0x00008674 }, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
+   { 16265, 0x00008675 }, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
+   { 16293, 0x00008676 }, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
+   { 16321, 0x00008677 }, /* GL_MAP2_VERTEX_ATTRIB7_4_NV */
+   { 16349, 0x00008678 }, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
+   { 16377, 0x00008679 }, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
+   { 16405, 0x00000D10 }, /* GL_MAP_COLOR */
+   { 16418, 0x00000010 }, /* GL_MAP_FLUSH_EXPLICIT_BIT */
+   { 16444, 0x00000008 }, /* GL_MAP_INVALIDATE_BUFFER_BIT */
+   { 16473, 0x00000004 }, /* GL_MAP_INVALIDATE_RANGE_BIT */
+   { 16501, 0x00000001 }, /* GL_MAP_READ_BIT */
+   { 16517, 0x00000D11 }, /* GL_MAP_STENCIL */
+   { 16532, 0x00000020 }, /* GL_MAP_UNSYNCHRONIZED_BIT */
+   { 16558, 0x00000002 }, /* GL_MAP_WRITE_BIT */
+   { 16575, 0x000088C0 }, /* GL_MATRIX0_ARB */
+   { 16590, 0x00008630 }, /* GL_MATRIX0_NV */
+   { 16604, 0x000088CA }, /* GL_MATRIX10_ARB */
+   { 16620, 0x000088CB }, /* GL_MATRIX11_ARB */
+   { 16636, 0x000088CC }, /* GL_MATRIX12_ARB */
+   { 16652, 0x000088CD }, /* GL_MATRIX13_ARB */
+   { 16668, 0x000088CE }, /* GL_MATRIX14_ARB */
+   { 16684, 0x000088CF }, /* GL_MATRIX15_ARB */
+   { 16700, 0x000088D0 }, /* GL_MATRIX16_ARB */
+   { 16716, 0x000088D1 }, /* GL_MATRIX17_ARB */
+   { 16732, 0x000088D2 }, /* GL_MATRIX18_ARB */
+   { 16748, 0x000088D3 }, /* GL_MATRIX19_ARB */
+   { 16764, 0x000088C1 }, /* GL_MATRIX1_ARB */
+   { 16779, 0x00008631 }, /* GL_MATRIX1_NV */
+   { 16793, 0x000088D4 }, /* GL_MATRIX20_ARB */
+   { 16809, 0x000088D5 }, /* GL_MATRIX21_ARB */
+   { 16825, 0x000088D6 }, /* GL_MATRIX22_ARB */
+   { 16841, 0x000088D7 }, /* GL_MATRIX23_ARB */
+   { 16857, 0x000088D8 }, /* GL_MATRIX24_ARB */
+   { 16873, 0x000088D9 }, /* GL_MATRIX25_ARB */
+   { 16889, 0x000088DA }, /* GL_MATRIX26_ARB */
+   { 16905, 0x000088DB }, /* GL_MATRIX27_ARB */
+   { 16921, 0x000088DC }, /* GL_MATRIX28_ARB */
+   { 16937, 0x000088DD }, /* GL_MATRIX29_ARB */
+   { 16953, 0x000088C2 }, /* GL_MATRIX2_ARB */
+   { 16968, 0x00008632 }, /* GL_MATRIX2_NV */
+   { 16982, 0x000088DE }, /* GL_MATRIX30_ARB */
+   { 16998, 0x000088DF }, /* GL_MATRIX31_ARB */
+   { 17014, 0x000088C3 }, /* GL_MATRIX3_ARB */
+   { 17029, 0x00008633 }, /* GL_MATRIX3_NV */
+   { 17043, 0x000088C4 }, /* GL_MATRIX4_ARB */
+   { 17058, 0x00008634 }, /* GL_MATRIX4_NV */
+   { 17072, 0x000088C5 }, /* GL_MATRIX5_ARB */
+   { 17087, 0x00008635 }, /* GL_MATRIX5_NV */
+   { 17101, 0x000088C6 }, /* GL_MATRIX6_ARB */
+   { 17116, 0x00008636 }, /* GL_MATRIX6_NV */
+   { 17130, 0x000088C7 }, /* GL_MATRIX7_ARB */
+   { 17145, 0x00008637 }, /* GL_MATRIX7_NV */
+   { 17159, 0x000088C8 }, /* GL_MATRIX8_ARB */
+   { 17174, 0x000088C9 }, /* GL_MATRIX9_ARB */
+   { 17189, 0x00008844 }, /* GL_MATRIX_INDEX_ARRAY_ARB */
+   { 17215, 0x00008849 }, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
+   { 17249, 0x00008846 }, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
+   { 17280, 0x00008848 }, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
+   { 17313, 0x00008847 }, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
+   { 17344, 0x00000BA0 }, /* GL_MATRIX_MODE */
+   { 17359, 0x00008840 }, /* GL_MATRIX_PALETTE_ARB */
+   { 17381, 0x00008008 }, /* GL_MAX */
+   { 17388, 0x00008073 }, /* GL_MAX_3D_TEXTURE_SIZE */
+   { 17411, 0x000088FF }, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
+   { 17443, 0x00000D35 }, /* GL_MAX_ATTRIB_STACK_DEPTH */
+   { 17469, 0x00000D3B }, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
+   { 17502, 0x00008177 }, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
+   { 17528, 0x00008178 }, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+   { 17562, 0x00000D32 }, /* GL_MAX_CLIP_PLANES */
+   { 17581, 0x00008CDF }, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
+   { 17610, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
+   { 17642, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI */
+   { 17678, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
+   { 17714, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB */
+   { 17754, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT */
+   { 17780, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT_EXT */
+   { 17810, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH */
+   { 17835, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH_EXT */
+   { 17864, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
+   { 17893, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB */
+   { 17926, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS */
+   { 17946, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ARB */
+   { 17970, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ATI */
+   { 17994, 0x000080E9 }, /* GL_MAX_ELEMENTS_INDICES */
+   { 18018, 0x000080E8 }, /* GL_MAX_ELEMENTS_VERTICES */
+   { 18043, 0x00000D30 }, /* GL_MAX_EVAL_ORDER */
+   { 18061, 0x00008008 }, /* GL_MAX_EXT */
+   { 18072, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
+   { 18107, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB */
+   { 18146, 0x00000D31 }, /* GL_MAX_LIGHTS */
+   { 18160, 0x00000B31 }, /* GL_MAX_LIST_NESTING */
+   { 18180, 0x00008841 }, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
+   { 18218, 0x00000D36 }, /* GL_MAX_MODELVIEW_STACK_DEPTH */
+   { 18247, 0x00000D37 }, /* GL_MAX_NAME_STACK_DEPTH */
+   { 18271, 0x00008842 }, /* GL_MAX_PALETTE_MATRICES_ARB */
+   { 18299, 0x00000D34 }, /* GL_MAX_PIXEL_MAP_TABLE */
+   { 18322, 0x000088B1 }, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
+   { 18359, 0x0000880B }, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
+   { 18395, 0x000088AD }, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
+   { 18422, 0x000088F5 }, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
+   { 18451, 0x000088B5 }, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
+   { 18485, 0x000088F4 }, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
+   { 18521, 0x000088F6 }, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
+   { 18548, 0x000088A1 }, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
+   { 18580, 0x000088B4 }, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
+   { 18616, 0x000088F8 }, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
+   { 18645, 0x000088F7 }, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
+   { 18674, 0x0000862F }, /* GL_MAX_PROGRAM_MATRICES_ARB */
+   { 18702, 0x0000862E }, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
+   { 18740, 0x000088B3 }, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+   { 18784, 0x0000880E }, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+   { 18827, 0x000088AF }, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
+   { 18861, 0x000088A3 }, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+   { 18900, 0x000088AB }, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
+   { 18937, 0x000088A7 }, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
+   { 18975, 0x00008810 }, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+   { 19018, 0x0000880F }, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+   { 19061, 0x000088A9 }, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
+   { 19091, 0x000088A5 }, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
+   { 19122, 0x0000880D }, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
+   { 19158, 0x0000880C }, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
+   { 19194, 0x00000D38 }, /* GL_MAX_PROJECTION_STACK_DEPTH */
+   { 19224, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
+   { 19258, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_NV */
+   { 19291, 0x000084E8 }, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
+   { 19320, 0x00008D57 }, /* GL_MAX_SAMPLES */
+   { 19335, 0x00008504 }, /* GL_MAX_SHININESS_NV */
+   { 19355, 0x00008505 }, /* GL_MAX_SPOT_EXPONENT_NV */
+   { 19379, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS */
+   { 19401, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS_ARB */
+   { 19427, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS */
+   { 19454, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS_ARB */
+   { 19485, 0x000084FD }, /* GL_MAX_TEXTURE_LOD_BIAS */
+   { 19509, 0x000084FF }, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
+   { 19543, 0x00000D33 }, /* GL_MAX_TEXTURE_SIZE */
+   { 19563, 0x00000D39 }, /* GL_MAX_TEXTURE_STACK_DEPTH */
+   { 19590, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS */
+   { 19611, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS_ARB */
+   { 19636, 0x0000862F }, /* GL_MAX_TRACK_MATRICES_NV */
+   { 19661, 0x0000862E }, /* GL_MAX_TRACK_MATRIX_STACK_DEPTH_NV */
+   { 19696, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS */
+   { 19718, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS_ARB */
+   { 19744, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS */
+   { 19766, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS_ARB */
+   { 19792, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
+   { 19826, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB */
+   { 19864, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
+   { 19897, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB */
+   { 19934, 0x000086A4 }, /* GL_MAX_VERTEX_UNITS_ARB */
+   { 19958, 0x00000D3A }, /* GL_MAX_VIEWPORT_DIMS */
+   { 19979, 0x00008007 }, /* GL_MIN */
+   { 19986, 0x0000802E }, /* GL_MINMAX */
+   { 19996, 0x0000802E }, /* GL_MINMAX_EXT */
+   { 20010, 0x0000802F }, /* GL_MINMAX_FORMAT */
+   { 20027, 0x0000802F }, /* GL_MINMAX_FORMAT_EXT */
+   { 20048, 0x00008030 }, /* GL_MINMAX_SINK */
+   { 20063, 0x00008030 }, /* GL_MINMAX_SINK_EXT */
+   { 20082, 0x00008007 }, /* GL_MIN_EXT */
+   { 20093, 0x00008370 }, /* GL_MIRRORED_REPEAT */
+   { 20112, 0x00008370 }, /* GL_MIRRORED_REPEAT_ARB */
+   { 20135, 0x00008370 }, /* GL_MIRRORED_REPEAT_IBM */
+   { 20158, 0x00008742 }, /* GL_MIRROR_CLAMP_ATI */
+   { 20178, 0x00008742 }, /* GL_MIRROR_CLAMP_EXT */
+   { 20198, 0x00008912 }, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
+   { 20228, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_ATI */
+   { 20256, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
+   { 20284, 0x00001700 }, /* GL_MODELVIEW */
+   { 20297, 0x00001700 }, /* GL_MODELVIEW0_ARB */
+   { 20315, 0x0000872A }, /* GL_MODELVIEW10_ARB */
+   { 20334, 0x0000872B }, /* GL_MODELVIEW11_ARB */
+   { 20353, 0x0000872C }, /* GL_MODELVIEW12_ARB */
+   { 20372, 0x0000872D }, /* GL_MODELVIEW13_ARB */
+   { 20391, 0x0000872E }, /* GL_MODELVIEW14_ARB */
+   { 20410, 0x0000872F }, /* GL_MODELVIEW15_ARB */
+   { 20429, 0x00008730 }, /* GL_MODELVIEW16_ARB */
+   { 20448, 0x00008731 }, /* GL_MODELVIEW17_ARB */
+   { 20467, 0x00008732 }, /* GL_MODELVIEW18_ARB */
+   { 20486, 0x00008733 }, /* GL_MODELVIEW19_ARB */
+   { 20505, 0x0000850A }, /* GL_MODELVIEW1_ARB */
+   { 20523, 0x00008734 }, /* GL_MODELVIEW20_ARB */
+   { 20542, 0x00008735 }, /* GL_MODELVIEW21_ARB */
+   { 20561, 0x00008736 }, /* GL_MODELVIEW22_ARB */
+   { 20580, 0x00008737 }, /* GL_MODELVIEW23_ARB */
+   { 20599, 0x00008738 }, /* GL_MODELVIEW24_ARB */
+   { 20618, 0x00008739 }, /* GL_MODELVIEW25_ARB */
+   { 20637, 0x0000873A }, /* GL_MODELVIEW26_ARB */
+   { 20656, 0x0000873B }, /* GL_MODELVIEW27_ARB */
+   { 20675, 0x0000873C }, /* GL_MODELVIEW28_ARB */
+   { 20694, 0x0000873D }, /* GL_MODELVIEW29_ARB */
+   { 20713, 0x00008722 }, /* GL_MODELVIEW2_ARB */
+   { 20731, 0x0000873E }, /* GL_MODELVIEW30_ARB */
+   { 20750, 0x0000873F }, /* GL_MODELVIEW31_ARB */
+   { 20769, 0x00008723 }, /* GL_MODELVIEW3_ARB */
+   { 20787, 0x00008724 }, /* GL_MODELVIEW4_ARB */
+   { 20805, 0x00008725 }, /* GL_MODELVIEW5_ARB */
+   { 20823, 0x00008726 }, /* GL_MODELVIEW6_ARB */
+   { 20841, 0x00008727 }, /* GL_MODELVIEW7_ARB */
+   { 20859, 0x00008728 }, /* GL_MODELVIEW8_ARB */
+   { 20877, 0x00008729 }, /* GL_MODELVIEW9_ARB */
+   { 20895, 0x00000BA6 }, /* GL_MODELVIEW_MATRIX */
+   { 20915, 0x00008629 }, /* GL_MODELVIEW_PROJECTION_NV */
+   { 20942, 0x00000BA3 }, /* GL_MODELVIEW_STACK_DEPTH */
+   { 20967, 0x00002100 }, /* GL_MODULATE */
+   { 20979, 0x00008744 }, /* GL_MODULATE_ADD_ATI */
+   { 20999, 0x00008745 }, /* GL_MODULATE_SIGNED_ADD_ATI */
+   { 21026, 0x00008746 }, /* GL_MODULATE_SUBTRACT_ATI */
+   { 21051, 0x00000103 }, /* GL_MULT */
+   { 21059, 0x0000809D }, /* GL_MULTISAMPLE */
+   { 21074, 0x000086B2 }, /* GL_MULTISAMPLE_3DFX */
+   { 21094, 0x0000809D }, /* GL_MULTISAMPLE_ARB */
+   { 21113, 0x20000000 }, /* GL_MULTISAMPLE_BIT */
+   { 21132, 0x20000000 }, /* GL_MULTISAMPLE_BIT_3DFX */
+   { 21156, 0x20000000 }, /* GL_MULTISAMPLE_BIT_ARB */
+   { 21179, 0x00008534 }, /* GL_MULTISAMPLE_FILTER_HINT_NV */
+   { 21209, 0x00002A25 }, /* GL_N3F_V3F */
+   { 21220, 0x00000D70 }, /* GL_NAME_STACK_DEPTH */
+   { 21240, 0x0000150E }, /* GL_NAND */
+   { 21248, 0x00002600 }, /* GL_NEAREST */
+   { 21259, 0x0000844E }, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
+   { 21290, 0x0000844D }, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
+   { 21322, 0x00002702 }, /* GL_NEAREST_MIPMAP_LINEAR */
+   { 21347, 0x00002700 }, /* GL_NEAREST_MIPMAP_NEAREST */
+   { 21373, 0x00000200 }, /* GL_NEVER */
+   { 21382, 0x00001102 }, /* GL_NICEST */
+   { 21392, 0x00000000 }, /* GL_NONE */
+   { 21400, 0x00001505 }, /* GL_NOOP */
+   { 21408, 0x00001508 }, /* GL_NOR */
+   { 21415, 0x00000BA1 }, /* GL_NORMALIZE */
+   { 21428, 0x00008075 }, /* GL_NORMAL_ARRAY */
+   { 21444, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
+   { 21475, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING_ARB */
+   { 21510, 0x0000808F }, /* GL_NORMAL_ARRAY_POINTER */
+   { 21534, 0x0000807F }, /* GL_NORMAL_ARRAY_STRIDE */
+   { 21557, 0x0000807E }, /* GL_NORMAL_ARRAY_TYPE */
+   { 21578, 0x00008511 }, /* GL_NORMAL_MAP */
+   { 21592, 0x00008511 }, /* GL_NORMAL_MAP_ARB */
+   { 21610, 0x00008511 }, /* GL_NORMAL_MAP_NV */
+   { 21627, 0x00000205 }, /* GL_NOTEQUAL */
+   { 21639, 0x00000000 }, /* GL_NO_ERROR */
+   { 21651, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
+   { 21685, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB */
+   { 21723, 0x00008B89 }, /* GL_OBJECT_ACTIVE_ATTRIBUTES_ARB */
+   { 21755, 0x00008B8A }, /* GL_OBJECT_ACTIVE_ATTRIBUTE_MAX_LENGTH_ARB */
+   { 21797, 0x00008B86 }, /* GL_OBJECT_ACTIVE_UNIFORMS_ARB */
+   { 21827, 0x00008B87 }, /* GL_OBJECT_ACTIVE_UNIFORM_MAX_LENGTH_ARB */
+   { 21867, 0x00008B85 }, /* GL_OBJECT_ATTACHED_OBJECTS_ARB */
+   { 21898, 0x00008B81 }, /* GL_OBJECT_COMPILE_STATUS_ARB */
+   { 21927, 0x00008B80 }, /* GL_OBJECT_DELETE_STATUS_ARB */
+   { 21955, 0x00008B84 }, /* GL_OBJECT_INFO_LOG_LENGTH_ARB */
+   { 21985, 0x00002401 }, /* GL_OBJECT_LINEAR */
+   { 22002, 0x00008B82 }, /* GL_OBJECT_LINK_STATUS_ARB */
+   { 22028, 0x00002501 }, /* GL_OBJECT_PLANE */
+   { 22044, 0x00008B88 }, /* GL_OBJECT_SHADER_SOURCE_LENGTH_ARB */
+   { 22079, 0x00008B4F }, /* GL_OBJECT_SUBTYPE_ARB */
+   { 22101, 0x00008B4E }, /* GL_OBJECT_TYPE_ARB */
+   { 22120, 0x00008B83 }, /* GL_OBJECT_VALIDATE_STATUS_ARB */
+   { 22150, 0x00008165 }, /* GL_OCCLUSION_TEST_HP */
+   { 22171, 0x00008166 }, /* GL_OCCLUSION_TEST_RESULT_HP */
+   { 22199, 0x00000001 }, /* GL_ONE */
+   { 22206, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA */
+   { 22234, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA_EXT */
+   { 22266, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR */
+   { 22294, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR_EXT */
+   { 22326, 0x00000305 }, /* GL_ONE_MINUS_DST_ALPHA */
+   { 22349, 0x00000307 }, /* GL_ONE_MINUS_DST_COLOR */
+   { 22372, 0x00000303 }, /* GL_ONE_MINUS_SRC_ALPHA */
+   { 22395, 0x00000301 }, /* GL_ONE_MINUS_SRC_COLOR */
+   { 22418, 0x00008598 }, /* GL_OPERAND0_ALPHA */
+   { 22436, 0x00008598 }, /* GL_OPERAND0_ALPHA_ARB */
+   { 22458, 0x00008598 }, /* GL_OPERAND0_ALPHA_EXT */
+   { 22480, 0x00008590 }, /* GL_OPERAND0_RGB */
+   { 22496, 0x00008590 }, /* GL_OPERAND0_RGB_ARB */
+   { 22516, 0x00008590 }, /* GL_OPERAND0_RGB_EXT */
+   { 22536, 0x00008599 }, /* GL_OPERAND1_ALPHA */
+   { 22554, 0x00008599 }, /* GL_OPERAND1_ALPHA_ARB */
+   { 22576, 0x00008599 }, /* GL_OPERAND1_ALPHA_EXT */
+   { 22598, 0x00008591 }, /* GL_OPERAND1_RGB */
+   { 22614, 0x00008591 }, /* GL_OPERAND1_RGB_ARB */
+   { 22634, 0x00008591 }, /* GL_OPERAND1_RGB_EXT */
+   { 22654, 0x0000859A }, /* GL_OPERAND2_ALPHA */
+   { 22672, 0x0000859A }, /* GL_OPERAND2_ALPHA_ARB */
+   { 22694, 0x0000859A }, /* GL_OPERAND2_ALPHA_EXT */
+   { 22716, 0x00008592 }, /* GL_OPERAND2_RGB */
+   { 22732, 0x00008592 }, /* GL_OPERAND2_RGB_ARB */
+   { 22752, 0x00008592 }, /* GL_OPERAND2_RGB_EXT */
+   { 22772, 0x0000859B }, /* GL_OPERAND3_ALPHA_NV */
+   { 22793, 0x00008593 }, /* GL_OPERAND3_RGB_NV */
+   { 22812, 0x00001507 }, /* GL_OR */
+   { 22818, 0x00000A01 }, /* GL_ORDER */
+   { 22827, 0x0000150D }, /* GL_OR_INVERTED */
+   { 22842, 0x0000150B }, /* GL_OR_REVERSE */
+   { 22856, 0x00000505 }, /* GL_OUT_OF_MEMORY */
+   { 22873, 0x00000D05 }, /* GL_PACK_ALIGNMENT */
+   { 22891, 0x0000806C }, /* GL_PACK_IMAGE_HEIGHT */
+   { 22912, 0x00008758 }, /* GL_PACK_INVERT_MESA */
+   { 22932, 0x00000D01 }, /* GL_PACK_LSB_FIRST */
+   { 22950, 0x00000D02 }, /* GL_PACK_ROW_LENGTH */
+   { 22969, 0x0000806B }, /* GL_PACK_SKIP_IMAGES */
+   { 22989, 0x00000D04 }, /* GL_PACK_SKIP_PIXELS */
+   { 23009, 0x00000D03 }, /* GL_PACK_SKIP_ROWS */
+   { 23027, 0x00000D00 }, /* GL_PACK_SWAP_BYTES */
+   { 23046, 0x00008B92 }, /* GL_PALETTE4_R5_G6_B5_OES */
+   { 23071, 0x00008B94 }, /* GL_PALETTE4_RGB5_A1_OES */
+   { 23095, 0x00008B90 }, /* GL_PALETTE4_RGB8_OES */
+   { 23116, 0x00008B93 }, /* GL_PALETTE4_RGBA4_OES */
+   { 23138, 0x00008B91 }, /* GL_PALETTE4_RGBA8_OES */
+   { 23160, 0x00008B97 }, /* GL_PALETTE8_R5_G6_B5_OES */
+   { 23185, 0x00008B99 }, /* GL_PALETTE8_RGB5_A1_OES */
+   { 23209, 0x00008B95 }, /* GL_PALETTE8_RGB8_OES */
+   { 23230, 0x00008B98 }, /* GL_PALETTE8_RGBA4_OES */
+   { 23252, 0x00008B96 }, /* GL_PALETTE8_RGBA8_OES */
+   { 23274, 0x00000700 }, /* GL_PASS_THROUGH_TOKEN */
+   { 23296, 0x00000C50 }, /* GL_PERSPECTIVE_CORRECTION_HINT */
+   { 23327, 0x00000C79 }, /* GL_PIXEL_MAP_A_TO_A */
+   { 23347, 0x00000CB9 }, /* GL_PIXEL_MAP_A_TO_A_SIZE */
+   { 23372, 0x00000C78 }, /* GL_PIXEL_MAP_B_TO_B */
+   { 23392, 0x00000CB8 }, /* GL_PIXEL_MAP_B_TO_B_SIZE */
+   { 23417, 0x00000C77 }, /* GL_PIXEL_MAP_G_TO_G */
+   { 23437, 0x00000CB7 }, /* GL_PIXEL_MAP_G_TO_G_SIZE */
+   { 23462, 0x00000C75 }, /* GL_PIXEL_MAP_I_TO_A */
+   { 23482, 0x00000CB5 }, /* GL_PIXEL_MAP_I_TO_A_SIZE */
+   { 23507, 0x00000C74 }, /* GL_PIXEL_MAP_I_TO_B */
+   { 23527, 0x00000CB4 }, /* GL_PIXEL_MAP_I_TO_B_SIZE */
+   { 23552, 0x00000C73 }, /* GL_PIXEL_MAP_I_TO_G */
+   { 23572, 0x00000CB3 }, /* GL_PIXEL_MAP_I_TO_G_SIZE */
+   { 23597, 0x00000C70 }, /* GL_PIXEL_MAP_I_TO_I */
+   { 23617, 0x00000CB0 }, /* GL_PIXEL_MAP_I_TO_I_SIZE */
+   { 23642, 0x00000C72 }, /* GL_PIXEL_MAP_I_TO_R */
+   { 23662, 0x00000CB2 }, /* GL_PIXEL_MAP_I_TO_R_SIZE */
+   { 23687, 0x00000C76 }, /* GL_PIXEL_MAP_R_TO_R */
+   { 23707, 0x00000CB6 }, /* GL_PIXEL_MAP_R_TO_R_SIZE */
+   { 23732, 0x00000C71 }, /* GL_PIXEL_MAP_S_TO_S */
+   { 23752, 0x00000CB1 }, /* GL_PIXEL_MAP_S_TO_S_SIZE */
+   { 23777, 0x00000020 }, /* GL_PIXEL_MODE_BIT */
+   { 23795, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER */
+   { 23816, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING */
+   { 23845, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING_EXT */
+   { 23878, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER_EXT */
+   { 23903, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER */
+   { 23926, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
+   { 23957, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING_EXT */
+   { 23992, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER_EXT */
+   { 24019, 0x00001B00 }, /* GL_POINT */
+   { 24028, 0x00000000 }, /* GL_POINTS */
+   { 24038, 0x00000002 }, /* GL_POINT_BIT */
+   { 24051, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION */
+   { 24081, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_ARB */
+   { 24115, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_EXT */
+   { 24149, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_SGIS */
+   { 24184, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE */
+   { 24213, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_ARB */
+   { 24246, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_EXT */
+   { 24279, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_SGIS */
+   { 24313, 0x00000B11 }, /* GL_POINT_SIZE */
+   { 24327, 0x00000B13 }, /* GL_POINT_SIZE_GRANULARITY */
+   { 24353, 0x00008127 }, /* GL_POINT_SIZE_MAX */
+   { 24371, 0x00008127 }, /* GL_POINT_SIZE_MAX_ARB */
+   { 24393, 0x00008127 }, /* GL_POINT_SIZE_MAX_EXT */
+   { 24415, 0x00008127 }, /* GL_POINT_SIZE_MAX_SGIS */
+   { 24438, 0x00008126 }, /* GL_POINT_SIZE_MIN */
+   { 24456, 0x00008126 }, /* GL_POINT_SIZE_MIN_ARB */
+   { 24478, 0x00008126 }, /* GL_POINT_SIZE_MIN_EXT */
+   { 24500, 0x00008126 }, /* GL_POINT_SIZE_MIN_SGIS */
+   { 24523, 0x00000B12 }, /* GL_POINT_SIZE_RANGE */
+   { 24543, 0x00000B10 }, /* GL_POINT_SMOOTH */
+   { 24559, 0x00000C51 }, /* GL_POINT_SMOOTH_HINT */
+   { 24580, 0x00008861 }, /* GL_POINT_SPRITE */
+   { 24596, 0x00008861 }, /* GL_POINT_SPRITE_ARB */
+   { 24616, 0x00008CA0 }, /* GL_POINT_SPRITE_COORD_ORIGIN */
+   { 24645, 0x00008861 }, /* GL_POINT_SPRITE_NV */
+   { 24664, 0x00008863 }, /* GL_POINT_SPRITE_R_MODE_NV */
+   { 24690, 0x00000701 }, /* GL_POINT_TOKEN */
+   { 24705, 0x00000009 }, /* GL_POLYGON */
+   { 24716, 0x00000008 }, /* GL_POLYGON_BIT */
+   { 24731, 0x00000B40 }, /* GL_POLYGON_MODE */
+   { 24747, 0x00008039 }, /* GL_POLYGON_OFFSET_BIAS */
+   { 24770, 0x00008038 }, /* GL_POLYGON_OFFSET_FACTOR */
+   { 24795, 0x00008037 }, /* GL_POLYGON_OFFSET_FILL */
+   { 24818, 0x00002A02 }, /* GL_POLYGON_OFFSET_LINE */
+   { 24841, 0x00002A01 }, /* GL_POLYGON_OFFSET_POINT */
+   { 24865, 0x00002A00 }, /* GL_POLYGON_OFFSET_UNITS */
+   { 24889, 0x00000B41 }, /* GL_POLYGON_SMOOTH */
+   { 24907, 0x00000C53 }, /* GL_POLYGON_SMOOTH_HINT */
+   { 24930, 0x00000B42 }, /* GL_POLYGON_STIPPLE */
+   { 24949, 0x00000010 }, /* GL_POLYGON_STIPPLE_BIT */
+   { 24972, 0x00000703 }, /* GL_POLYGON_TOKEN */
+   { 24989, 0x00001203 }, /* GL_POSITION */
+   { 25001, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
+   { 25033, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI */
+   { 25069, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
+   { 25102, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI */
+   { 25139, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
+   { 25170, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI */
+   { 25205, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
+   { 25237, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI */
+   { 25273, 0x000080D2 }, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
+   { 25306, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
+   { 25338, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI */
+   { 25374, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
+   { 25407, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI */
+   { 25444, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS */
+   { 25474, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS_SGI */
+   { 25508, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE */
+   { 25539, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE_SGI */
+   { 25574, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
+   { 25605, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS_EXT */
+   { 25640, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
+   { 25672, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE_EXT */
+   { 25708, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS */
+   { 25738, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS_EXT */
+   { 25772, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE */
+   { 25803, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE_EXT */
+   { 25838, 0x000080D1 }, /* GL_POST_CONVOLUTION_COLOR_TABLE */
+   { 25870, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS */
+   { 25901, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS_EXT */
+   { 25936, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE */
+   { 25968, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE_EXT */
+   { 26004, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS */
+   { 26033, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS_EXT */
+   { 26066, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE */
+   { 26096, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE_EXT */
+   { 26130, 0x0000817B }, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
+   { 26169, 0x00008179 }, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
+   { 26202, 0x0000817C }, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
+   { 26242, 0x0000817A }, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
+   { 26276, 0x00008578 }, /* GL_PREVIOUS */
+   { 26288, 0x00008578 }, /* GL_PREVIOUS_ARB */
+   { 26304, 0x00008578 }, /* GL_PREVIOUS_EXT */
+   { 26320, 0x00008577 }, /* GL_PRIMARY_COLOR */
+   { 26337, 0x00008577 }, /* GL_PRIMARY_COLOR_ARB */
+   { 26358, 0x00008577 }, /* GL_PRIMARY_COLOR_EXT */
+   { 26379, 0x000088B0 }, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
+   { 26412, 0x00008805 }, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
+   { 26444, 0x000088AC }, /* GL_PROGRAM_ATTRIBS_ARB */
+   { 26467, 0x00008677 }, /* GL_PROGRAM_BINDING_ARB */
+   { 26490, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_ARB */
+   { 26520, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_NV */
+   { 26549, 0x00008874 }, /* GL_PROGRAM_ERROR_STRING_ARB */
+   { 26577, 0x00008876 }, /* GL_PROGRAM_FORMAT_ARB */
+   { 26599, 0x00008875 }, /* GL_PROGRAM_FORMAT_ASCII_ARB */
+   { 26627, 0x000088A0 }, /* GL_PROGRAM_INSTRUCTIONS_ARB */
+   { 26655, 0x00008627 }, /* GL_PROGRAM_LENGTH_ARB */
+   { 26677, 0x00008627 }, /* GL_PROGRAM_LENGTH_NV */
+   { 26698, 0x000088B2 }, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+   { 26738, 0x00008808 }, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+   { 26777, 0x000088AE }, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
+   { 26807, 0x000088A2 }, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+   { 26842, 0x000088AA }, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
+   { 26875, 0x000088A6 }, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
+   { 26909, 0x0000880A }, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+   { 26948, 0x00008809 }, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+   { 26987, 0x00008B40 }, /* GL_PROGRAM_OBJECT_ARB */
+   { 27009, 0x000088A8 }, /* GL_PROGRAM_PARAMETERS_ARB */
+   { 27035, 0x00008644 }, /* GL_PROGRAM_PARAMETER_NV */
+   { 27059, 0x00008647 }, /* GL_PROGRAM_RESIDENT_NV */
+   { 27082, 0x00008628 }, /* GL_PROGRAM_STRING_ARB */
+   { 27104, 0x00008628 }, /* GL_PROGRAM_STRING_NV */
+   { 27125, 0x00008646 }, /* GL_PROGRAM_TARGET_NV */
+   { 27146, 0x000088A4 }, /* GL_PROGRAM_TEMPORARIES_ARB */
+   { 27173, 0x00008807 }, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
+   { 27205, 0x00008806 }, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
+   { 27237, 0x000088B6 }, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
+   { 27272, 0x00001701 }, /* GL_PROJECTION */
+   { 27286, 0x00000BA7 }, /* GL_PROJECTION_MATRIX */
+   { 27307, 0x00000BA4 }, /* GL_PROJECTION_STACK_DEPTH */
+   { 27333, 0x00008E4F }, /* GL_PROVOKING_VERTEX_EXT */
+   { 27357, 0x000080D3 }, /* GL_PROXY_COLOR_TABLE */
+   { 27378, 0x00008025 }, /* GL_PROXY_HISTOGRAM */
+   { 27397, 0x00008025 }, /* GL_PROXY_HISTOGRAM_EXT */
+   { 27420, 0x000080D5 }, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
+   { 27459, 0x000080D4 }, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
+   { 27497, 0x00008063 }, /* GL_PROXY_TEXTURE_1D */
+   { 27517, 0x00008C19 }, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
+   { 27547, 0x00008063 }, /* GL_PROXY_TEXTURE_1D_EXT */
+   { 27571, 0x00008064 }, /* GL_PROXY_TEXTURE_2D */
+   { 27591, 0x00008C1B }, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
+   { 27621, 0x00008064 }, /* GL_PROXY_TEXTURE_2D_EXT */
+   { 27645, 0x00008070 }, /* GL_PROXY_TEXTURE_3D */
+   { 27665, 0x000080BD }, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
+   { 27698, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP */
+   { 27724, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP_ARB */
+   { 27754, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
+   { 27785, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_NV */
+   { 27815, 0x00002003 }, /* GL_Q */
+   { 27820, 0x00001209 }, /* GL_QUADRATIC_ATTENUATION */
+   { 27845, 0x00000007 }, /* GL_QUADS */
+   { 27854, 0x00008E4C }, /* GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT */
+   { 27902, 0x00008614 }, /* GL_QUAD_MESH_SUN */
+   { 27919, 0x00000008 }, /* GL_QUAD_STRIP */
+   { 27933, 0x00008864 }, /* GL_QUERY_COUNTER_BITS */
+   { 27955, 0x00008864 }, /* GL_QUERY_COUNTER_BITS_ARB */
+   { 27981, 0x00008866 }, /* GL_QUERY_RESULT */
+   { 27997, 0x00008866 }, /* GL_QUERY_RESULT_ARB */
+   { 28017, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE */
+   { 28043, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE_ARB */
+   { 28073, 0x00002002 }, /* GL_R */
+   { 28078, 0x00002A10 }, /* GL_R3_G3_B2 */
+   { 28090, 0x00019262 }, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
+   { 28123, 0x00000C02 }, /* GL_READ_BUFFER */
+   { 28138, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER */
+   { 28158, 0x00008CAA }, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
+   { 28190, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER_EXT */
+   { 28214, 0x000088B8 }, /* GL_READ_ONLY */
+   { 28227, 0x000088B8 }, /* GL_READ_ONLY_ARB */
+   { 28244, 0x000088BA }, /* GL_READ_WRITE */
+   { 28258, 0x000088BA }, /* GL_READ_WRITE_ARB */
+   { 28276, 0x00001903 }, /* GL_RED */
+   { 28283, 0x00008016 }, /* GL_REDUCE */
+   { 28293, 0x00008016 }, /* GL_REDUCE_EXT */
+   { 28307, 0x00000D15 }, /* GL_RED_BIAS */
+   { 28319, 0x00000D52 }, /* GL_RED_BITS */
+   { 28331, 0x00000D14 }, /* GL_RED_SCALE */
+   { 28344, 0x00008512 }, /* GL_REFLECTION_MAP */
+   { 28362, 0x00008512 }, /* GL_REFLECTION_MAP_ARB */
+   { 28384, 0x00008512 }, /* GL_REFLECTION_MAP_NV */
+   { 28405, 0x00001C00 }, /* GL_RENDER */
+   { 28415, 0x00008D41 }, /* GL_RENDERBUFFER */
+   { 28431, 0x00008D53 }, /* GL_RENDERBUFFER_ALPHA_SIZE */
+   { 28458, 0x00008CA7 }, /* GL_RENDERBUFFER_BINDING_EXT */
+   { 28486, 0x00008D52 }, /* GL_RENDERBUFFER_BLUE_SIZE */
+   { 28512, 0x00008D54 }, /* GL_RENDERBUFFER_DEPTH_SIZE */
+   { 28539, 0x00008D41 }, /* GL_RENDERBUFFER_EXT */
+   { 28559, 0x00008D51 }, /* GL_RENDERBUFFER_GREEN_SIZE */
+   { 28586, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT */
+   { 28609, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT_EXT */
+   { 28636, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
+   { 28668, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
+   { 28704, 0x00008D50 }, /* GL_RENDERBUFFER_RED_SIZE */
+   { 28729, 0x00008CAB }, /* GL_RENDERBUFFER_SAMPLES */
+   { 28753, 0x00008D55 }, /* GL_RENDERBUFFER_STENCIL_SIZE */
+   { 28782, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH */
+   { 28804, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH_EXT */
+   { 28830, 0x00001F01 }, /* GL_RENDERER */
+   { 28842, 0x00000C40 }, /* GL_RENDER_MODE */
+   { 28857, 0x00002901 }, /* GL_REPEAT */
+   { 28867, 0x00001E01 }, /* GL_REPLACE */
+   { 28878, 0x00008062 }, /* GL_REPLACE_EXT */
+   { 28893, 0x00008153 }, /* GL_REPLICATE_BORDER_HP */
+   { 28916, 0x0000803A }, /* GL_RESCALE_NORMAL */
+   { 28934, 0x0000803A }, /* GL_RESCALE_NORMAL_EXT */
+   { 28956, 0x00000102 }, /* GL_RETURN */
+   { 28966, 0x00001907 }, /* GL_RGB */
+   { 28973, 0x00008052 }, /* GL_RGB10 */
+   { 28982, 0x00008059 }, /* GL_RGB10_A2 */
+   { 28994, 0x00008059 }, /* GL_RGB10_A2_EXT */
+   { 29010, 0x00008052 }, /* GL_RGB10_EXT */
+   { 29023, 0x00008053 }, /* GL_RGB12 */
+   { 29032, 0x00008053 }, /* GL_RGB12_EXT */
+   { 29045, 0x00008054 }, /* GL_RGB16 */
+   { 29054, 0x00008054 }, /* GL_RGB16_EXT */
+   { 29067, 0x0000804E }, /* GL_RGB2_EXT */
+   { 29079, 0x0000804F }, /* GL_RGB4 */
+   { 29087, 0x0000804F }, /* GL_RGB4_EXT */
+   { 29099, 0x000083A1 }, /* GL_RGB4_S3TC */
+   { 29112, 0x00008050 }, /* GL_RGB5 */
+   { 29120, 0x00008057 }, /* GL_RGB5_A1 */
+   { 29131, 0x00008057 }, /* GL_RGB5_A1_EXT */
+   { 29146, 0x00008050 }, /* GL_RGB5_EXT */
+   { 29158, 0x00008051 }, /* GL_RGB8 */
+   { 29166, 0x00008051 }, /* GL_RGB8_EXT */
+   { 29178, 0x00001908 }, /* GL_RGBA */
+   { 29186, 0x0000805A }, /* GL_RGBA12 */
+   { 29196, 0x0000805A }, /* GL_RGBA12_EXT */
+   { 29210, 0x0000805B }, /* GL_RGBA16 */
+   { 29220, 0x0000805B }, /* GL_RGBA16_EXT */
+   { 29234, 0x00008055 }, /* GL_RGBA2 */
+   { 29243, 0x00008055 }, /* GL_RGBA2_EXT */
+   { 29256, 0x00008056 }, /* GL_RGBA4 */
+   { 29265, 0x000083A5 }, /* GL_RGBA4_DXT5_S3TC */
+   { 29284, 0x00008056 }, /* GL_RGBA4_EXT */
+   { 29297, 0x000083A3 }, /* GL_RGBA4_S3TC */
+   { 29311, 0x00008058 }, /* GL_RGBA8 */
+   { 29320, 0x00008058 }, /* GL_RGBA8_EXT */
+   { 29333, 0x00008F97 }, /* GL_RGBA8_SNORM */
+   { 29348, 0x000083A4 }, /* GL_RGBA_DXT5_S3TC */
+   { 29366, 0x00000C31 }, /* GL_RGBA_MODE */
+   { 29379, 0x000083A2 }, /* GL_RGBA_S3TC */
+   { 29392, 0x00008F93 }, /* GL_RGBA_SNORM */
+   { 29406, 0x000083A0 }, /* GL_RGB_S3TC */
+   { 29418, 0x00008573 }, /* GL_RGB_SCALE */
+   { 29431, 0x00008573 }, /* GL_RGB_SCALE_ARB */
+   { 29448, 0x00008573 }, /* GL_RGB_SCALE_EXT */
+   { 29465, 0x00000407 }, /* GL_RIGHT */
+   { 29474, 0x00002000 }, /* GL_S */
+   { 29479, 0x00008B5D }, /* GL_SAMPLER_1D */
+   { 29493, 0x00008B61 }, /* GL_SAMPLER_1D_SHADOW */
+   { 29514, 0x00008B5E }, /* GL_SAMPLER_2D */
+   { 29528, 0x00008B62 }, /* GL_SAMPLER_2D_SHADOW */
+   { 29549, 0x00008B5F }, /* GL_SAMPLER_3D */
+   { 29563, 0x00008B60 }, /* GL_SAMPLER_CUBE */
+   { 29579, 0x000080A9 }, /* GL_SAMPLES */
+   { 29590, 0x000086B4 }, /* GL_SAMPLES_3DFX */
+   { 29606, 0x000080A9 }, /* GL_SAMPLES_ARB */
+   { 29621, 0x00008914 }, /* GL_SAMPLES_PASSED */
+   { 29639, 0x00008914 }, /* GL_SAMPLES_PASSED_ARB */
+   { 29661, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
+   { 29689, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE_ARB */
+   { 29721, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE */
+   { 29744, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE_ARB */
+   { 29771, 0x000080A8 }, /* GL_SAMPLE_BUFFERS */
+   { 29789, 0x000086B3 }, /* GL_SAMPLE_BUFFERS_3DFX */
+   { 29812, 0x000080A8 }, /* GL_SAMPLE_BUFFERS_ARB */
+   { 29834, 0x000080A0 }, /* GL_SAMPLE_COVERAGE */
+   { 29853, 0x000080A0 }, /* GL_SAMPLE_COVERAGE_ARB */
+   { 29876, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT */
+   { 29902, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT_ARB */
+   { 29932, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE */
+   { 29957, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE_ARB */
+   { 29986, 0x00080000 }, /* GL_SCISSOR_BIT */
+   { 30001, 0x00000C10 }, /* GL_SCISSOR_BOX */
+   { 30016, 0x00000C11 }, /* GL_SCISSOR_TEST */
+   { 30032, 0x0000845E }, /* GL_SECONDARY_COLOR_ARRAY */
+   { 30057, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
+   { 30097, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB */
+   { 30141, 0x0000845D }, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
+   { 30174, 0x0000845A }, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
+   { 30204, 0x0000845C }, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
+   { 30236, 0x0000845B }, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
+   { 30266, 0x00001C02 }, /* GL_SELECT */
+   { 30276, 0x00000DF3 }, /* GL_SELECTION_BUFFER_POINTER */
+   { 30304, 0x00000DF4 }, /* GL_SELECTION_BUFFER_SIZE */
+   { 30329, 0x00008012 }, /* GL_SEPARABLE_2D */
+   { 30345, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR */
+   { 30372, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR_EXT */
+   { 30403, 0x0000150F }, /* GL_SET */
+   { 30410, 0x00008B48 }, /* GL_SHADER_OBJECT_ARB */
+   { 30431, 0x00008B88 }, /* GL_SHADER_SOURCE_LENGTH */
+   { 30455, 0x00008B4F }, /* GL_SHADER_TYPE */
+   { 30470, 0x00000B54 }, /* GL_SHADE_MODEL */
+   { 30485, 0x00008B8C }, /* GL_SHADING_LANGUAGE_VERSION */
+   { 30513, 0x000080BF }, /* GL_SHADOW_AMBIENT_SGIX */
+   { 30536, 0x000081FB }, /* GL_SHARED_TEXTURE_PALETTE_EXT */
+   { 30566, 0x00001601 }, /* GL_SHININESS */
+   { 30579, 0x00001402 }, /* GL_SHORT */
+   { 30588, 0x00008F9C }, /* GL_SIGNED_NORMALIZED */
+   { 30609, 0x000081F9 }, /* GL_SINGLE_COLOR */
+   { 30625, 0x000081F9 }, /* GL_SINGLE_COLOR_EXT */
+   { 30645, 0x000085CC }, /* GL_SLICE_ACCUM_SUN */
+   { 30664, 0x00008C46 }, /* GL_SLUMINANCE */
+   { 30678, 0x00008C47 }, /* GL_SLUMINANCE8 */
+   { 30693, 0x00008C45 }, /* GL_SLUMINANCE8_ALPHA8 */
+   { 30715, 0x00008C44 }, /* GL_SLUMINANCE_ALPHA */
+   { 30735, 0x00001D01 }, /* GL_SMOOTH */
+   { 30745, 0x00000B23 }, /* GL_SMOOTH_LINE_WIDTH_GRANULARITY */
+   { 30778, 0x00000B22 }, /* GL_SMOOTH_LINE_WIDTH_RANGE */
+   { 30805, 0x00000B13 }, /* GL_SMOOTH_POINT_SIZE_GRANULARITY */
+   { 30838, 0x00000B12 }, /* GL_SMOOTH_POINT_SIZE_RANGE */
+   { 30865, 0x00008588 }, /* GL_SOURCE0_ALPHA */
+   { 30882, 0x00008588 }, /* GL_SOURCE0_ALPHA_ARB */
+   { 30903, 0x00008588 }, /* GL_SOURCE0_ALPHA_EXT */
+   { 30924, 0x00008580 }, /* GL_SOURCE0_RGB */
+   { 30939, 0x00008580 }, /* GL_SOURCE0_RGB_ARB */
+   { 30958, 0x00008580 }, /* GL_SOURCE0_RGB_EXT */
+   { 30977, 0x00008589 }, /* GL_SOURCE1_ALPHA */
+   { 30994, 0x00008589 }, /* GL_SOURCE1_ALPHA_ARB */
+   { 31015, 0x00008589 }, /* GL_SOURCE1_ALPHA_EXT */
+   { 31036, 0x00008581 }, /* GL_SOURCE1_RGB */
+   { 31051, 0x00008581 }, /* GL_SOURCE1_RGB_ARB */
+   { 31070, 0x00008581 }, /* GL_SOURCE1_RGB_EXT */
+   { 31089, 0x0000858A }, /* GL_SOURCE2_ALPHA */
+   { 31106, 0x0000858A }, /* GL_SOURCE2_ALPHA_ARB */
+   { 31127, 0x0000858A }, /* GL_SOURCE2_ALPHA_EXT */
+   { 31148, 0x00008582 }, /* GL_SOURCE2_RGB */
+   { 31163, 0x00008582 }, /* GL_SOURCE2_RGB_ARB */
+   { 31182, 0x00008582 }, /* GL_SOURCE2_RGB_EXT */
+   { 31201, 0x0000858B }, /* GL_SOURCE3_ALPHA_NV */
+   { 31221, 0x00008583 }, /* GL_SOURCE3_RGB_NV */
+   { 31239, 0x00001202 }, /* GL_SPECULAR */
+   { 31251, 0x00002402 }, /* GL_SPHERE_MAP */
+   { 31265, 0x00001206 }, /* GL_SPOT_CUTOFF */
+   { 31280, 0x00001204 }, /* GL_SPOT_DIRECTION */
+   { 31298, 0x00001205 }, /* GL_SPOT_EXPONENT */
+   { 31315, 0x00008588 }, /* GL_SRC0_ALPHA */
+   { 31329, 0x00008580 }, /* GL_SRC0_RGB */
+   { 31341, 0x00008589 }, /* GL_SRC1_ALPHA */
+   { 31355, 0x00008581 }, /* GL_SRC1_RGB */
+   { 31367, 0x0000858A }, /* GL_SRC2_ALPHA */
+   { 31381, 0x00008582 }, /* GL_SRC2_RGB */
+   { 31393, 0x00000302 }, /* GL_SRC_ALPHA */
+   { 31406, 0x00000308 }, /* GL_SRC_ALPHA_SATURATE */
+   { 31428, 0x00000300 }, /* GL_SRC_COLOR */
+   { 31441, 0x00008C40 }, /* GL_SRGB */
+   { 31449, 0x00008C41 }, /* GL_SRGB8 */
+   { 31458, 0x00008C43 }, /* GL_SRGB8_ALPHA8 */
+   { 31474, 0x00008C42 }, /* GL_SRGB_ALPHA */
+   { 31488, 0x00000503 }, /* GL_STACK_OVERFLOW */
+   { 31506, 0x00000504 }, /* GL_STACK_UNDERFLOW */
+   { 31525, 0x000088E6 }, /* GL_STATIC_COPY */
+   { 31540, 0x000088E6 }, /* GL_STATIC_COPY_ARB */
+   { 31559, 0x000088E4 }, /* GL_STATIC_DRAW */
+   { 31574, 0x000088E4 }, /* GL_STATIC_DRAW_ARB */
+   { 31593, 0x000088E5 }, /* GL_STATIC_READ */
+   { 31608, 0x000088E5 }, /* GL_STATIC_READ_ARB */
+   { 31627, 0x00001802 }, /* GL_STENCIL */
+   { 31638, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT */
+   { 31660, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */
+   { 31686, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */
+   { 31707, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */
+   { 31732, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */
+   { 31753, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */
+   { 31778, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
+   { 31810, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */
+   { 31846, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
+   { 31878, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */
+   { 31914, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */
+   { 31934, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */
+   { 31961, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */
+   { 31987, 0x00000D57 }, /* GL_STENCIL_BITS */
+   { 32003, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */
+   { 32025, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */
+   { 32048, 0x00000B94 }, /* GL_STENCIL_FAIL */
+   { 32064, 0x00000B92 }, /* GL_STENCIL_FUNC */
+   { 32080, 0x00001901 }, /* GL_STENCIL_INDEX */
+   { 32097, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */
+   { 32120, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */
+   { 32142, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */
+   { 32164, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */
+   { 32186, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */
+   { 32207, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */
+   { 32234, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */
+   { 32261, 0x00000B97 }, /* GL_STENCIL_REF */
+   { 32276, 0x00000B90 }, /* GL_STENCIL_TEST */
+   { 32292, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
+   { 32321, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */
+   { 32343, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */
+   { 32364, 0x00000C33 }, /* GL_STEREO */
+   { 32374, 0x000088E2 }, /* GL_STREAM_COPY */
+   { 32389, 0x000088E2 }, /* GL_STREAM_COPY_ARB */
+   { 32408, 0x000088E0 }, /* GL_STREAM_DRAW */
+   { 32423, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */
+   { 32442, 0x000088E1 }, /* GL_STREAM_READ */
+   { 32457, 0x000088E1 }, /* GL_STREAM_READ_ARB */
+   { 32476, 0x00000D50 }, /* GL_SUBPIXEL_BITS */
+   { 32493, 0x000084E7 }, /* GL_SUBTRACT */
+   { 32505, 0x000084E7 }, /* GL_SUBTRACT_ARB */
+   { 32521, 0x00002001 }, /* GL_T */
+   { 32526, 0x00002A2A }, /* GL_T2F_C3F_V3F */
+   { 32541, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */
+   { 32560, 0x00002A29 }, /* GL_T2F_C4UB_V3F */
+   { 32576, 0x00002A2B }, /* GL_T2F_N3F_V3F */
+   { 32591, 0x00002A27 }, /* GL_T2F_V3F */
+   { 32602, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */
+   { 32621, 0x00002A28 }, /* GL_T4F_V4F */
+   { 32632, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */
+   { 32655, 0x00001702 }, /* GL_TEXTURE */
+   { 32666, 0x000084C0 }, /* GL_TEXTURE0 */
+   { 32678, 0x000084C0 }, /* GL_TEXTURE0_ARB */
+   { 32694, 0x000084C1 }, /* GL_TEXTURE1 */
+   { 32706, 0x000084CA }, /* GL_TEXTURE10 */
+   { 32719, 0x000084CA }, /* GL_TEXTURE10_ARB */
+   { 32736, 0x000084CB }, /* GL_TEXTURE11 */
+   { 32749, 0x000084CB }, /* GL_TEXTURE11_ARB */
+   { 32766, 0x000084CC }, /* GL_TEXTURE12 */
+   { 32779, 0x000084CC }, /* GL_TEXTURE12_ARB */
+   { 32796, 0x000084CD }, /* GL_TEXTURE13 */
+   { 32809, 0x000084CD }, /* GL_TEXTURE13_ARB */
+   { 32826, 0x000084CE }, /* GL_TEXTURE14 */
+   { 32839, 0x000084CE }, /* GL_TEXTURE14_ARB */
+   { 32856, 0x000084CF }, /* GL_TEXTURE15 */
+   { 32869, 0x000084CF }, /* GL_TEXTURE15_ARB */
+   { 32886, 0x000084D0 }, /* GL_TEXTURE16 */
+   { 32899, 0x000084D0 }, /* GL_TEXTURE16_ARB */
+   { 32916, 0x000084D1 }, /* GL_TEXTURE17 */
+   { 32929, 0x000084D1 }, /* GL_TEXTURE17_ARB */
+   { 32946, 0x000084D2 }, /* GL_TEXTURE18 */
+   { 32959, 0x000084D2 }, /* GL_TEXTURE18_ARB */
+   { 32976, 0x000084D3 }, /* GL_TEXTURE19 */
+   { 32989, 0x000084D3 }, /* GL_TEXTURE19_ARB */
+   { 33006, 0x000084C1 }, /* GL_TEXTURE1_ARB */
+   { 33022, 0x000084C2 }, /* GL_TEXTURE2 */
+   { 33034, 0x000084D4 }, /* GL_TEXTURE20 */
+   { 33047, 0x000084D4 }, /* GL_TEXTURE20_ARB */
+   { 33064, 0x000084D5 }, /* GL_TEXTURE21 */
+   { 33077, 0x000084D5 }, /* GL_TEXTURE21_ARB */
+   { 33094, 0x000084D6 }, /* GL_TEXTURE22 */
+   { 33107, 0x000084D6 }, /* GL_TEXTURE22_ARB */
+   { 33124, 0x000084D7 }, /* GL_TEXTURE23 */
+   { 33137, 0x000084D7 }, /* GL_TEXTURE23_ARB */
+   { 33154, 0x000084D8 }, /* GL_TEXTURE24 */
+   { 33167, 0x000084D8 }, /* GL_TEXTURE24_ARB */
+   { 33184, 0x000084D9 }, /* GL_TEXTURE25 */
+   { 33197, 0x000084D9 }, /* GL_TEXTURE25_ARB */
+   { 33214, 0x000084DA }, /* GL_TEXTURE26 */
+   { 33227, 0x000084DA }, /* GL_TEXTURE26_ARB */
+   { 33244, 0x000084DB }, /* GL_TEXTURE27 */
+   { 33257, 0x000084DB }, /* GL_TEXTURE27_ARB */
+   { 33274, 0x000084DC }, /* GL_TEXTURE28 */
+   { 33287, 0x000084DC }, /* GL_TEXTURE28_ARB */
+   { 33304, 0x000084DD }, /* GL_TEXTURE29 */
+   { 33317, 0x000084DD }, /* GL_TEXTURE29_ARB */
+   { 33334, 0x000084C2 }, /* GL_TEXTURE2_ARB */
+   { 33350, 0x000084C3 }, /* GL_TEXTURE3 */
+   { 33362, 0x000084DE }, /* GL_TEXTURE30 */
+   { 33375, 0x000084DE }, /* GL_TEXTURE30_ARB */
+   { 33392, 0x000084DF }, /* GL_TEXTURE31 */
+   { 33405, 0x000084DF }, /* GL_TEXTURE31_ARB */
+   { 33422, 0x000084C3 }, /* GL_TEXTURE3_ARB */
+   { 33438, 0x000084C4 }, /* GL_TEXTURE4 */
+   { 33450, 0x000084C4 }, /* GL_TEXTURE4_ARB */
+   { 33466, 0x000084C5 }, /* GL_TEXTURE5 */
+   { 33478, 0x000084C5 }, /* GL_TEXTURE5_ARB */
+   { 33494, 0x000084C6 }, /* GL_TEXTURE6 */
+   { 33506, 0x000084C6 }, /* GL_TEXTURE6_ARB */
+   { 33522, 0x000084C7 }, /* GL_TEXTURE7 */
+   { 33534, 0x000084C7 }, /* GL_TEXTURE7_ARB */
+   { 33550, 0x000084C8 }, /* GL_TEXTURE8 */
+   { 33562, 0x000084C8 }, /* GL_TEXTURE8_ARB */
+   { 33578, 0x000084C9 }, /* GL_TEXTURE9 */
+   { 33590, 0x000084C9 }, /* GL_TEXTURE9_ARB */
+   { 33606, 0x00000DE0 }, /* GL_TEXTURE_1D */
+   { 33620, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */
+   { 33644, 0x00000DE1 }, /* GL_TEXTURE_2D */
+   { 33658, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */
+   { 33682, 0x0000806F }, /* GL_TEXTURE_3D */
+   { 33696, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */
+   { 33718, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */
+   { 33744, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */
+   { 33766, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */
+   { 33788, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
+   { 33820, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */
+   { 33842, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
+   { 33874, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */
+   { 33896, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */
+   { 33924, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */
+   { 33956, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
+   { 33989, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */
+   { 34021, 0x00040000 }, /* GL_TEXTURE_BIT */
+   { 34036, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */
+   { 34057, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */
+   { 34082, 0x00001005 }, /* GL_TEXTURE_BORDER */
+   { 34100, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */
+   { 34124, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
+   { 34155, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
+   { 34185, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
+   { 34215, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
+   { 34250, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
+   { 34281, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+   { 34319, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */
+   { 34346, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
+   { 34378, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
+   { 34412, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */
+   { 34436, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */
+   { 34464, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */
+   { 34488, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */
+   { 34516, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
+   { 34549, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */
+   { 34573, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */
+   { 34595, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */
+   { 34617, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */
+   { 34643, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */
+   { 34677, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
+   { 34710, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */
+   { 34747, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */
+   { 34775, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */
+   { 34807, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */
+   { 34830, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
+   { 34868, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */
+   { 34910, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */
+   { 34941, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */
+   { 34969, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
+   { 34999, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */
+   { 35027, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */
+   { 35047, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */
+   { 35071, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
+   { 35102, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */
+   { 35137, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
+   { 35168, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */
+   { 35203, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
+   { 35234, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */
+   { 35269, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
+   { 35300, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */
+   { 35335, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
+   { 35366, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */
+   { 35401, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
+   { 35432, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */
+   { 35467, 0x00008071 }, /* GL_TEXTURE_DEPTH */
+   { 35484, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */
+   { 35506, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */
+   { 35532, 0x00002300 }, /* GL_TEXTURE_ENV */
+   { 35547, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */
+   { 35568, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */
+   { 35588, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */
+   { 35614, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */
+   { 35634, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */
+   { 35651, 0x00000C62 }, /* GL_TEXTURE_GEN_R */
+   { 35668, 0x00000C60 }, /* GL_TEXTURE_GEN_S */
+   { 35685, 0x00000C61 }, /* GL_TEXTURE_GEN_T */
+   { 35702, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */
+   { 35727, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */
+   { 35749, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */
+   { 35775, 0x00001001 }, /* GL_TEXTURE_HEIGHT */
+   { 35793, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */
+   { 35819, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */
+   { 35845, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */
+   { 35875, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */
+   { 35902, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */
+   { 35927, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */
+   { 35947, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */
+   { 35971, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
+   { 35998, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
+   { 36025, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
+   { 36052, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */
+   { 36078, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */
+   { 36108, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */
+   { 36130, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */
+   { 36148, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
+   { 36178, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
+   { 36206, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
+   { 36234, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
+   { 36262, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */
+   { 36283, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */
+   { 36302, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */
+   { 36324, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */
+   { 36343, 0x00008066 }, /* GL_TEXTURE_PRIORITY */
+   { 36363, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */
+   { 36388, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */
+   { 36412, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */
+   { 36432, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */
+   { 36456, 0x00008067 }, /* GL_TEXTURE_RESIDENT */
+   { 36476, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */
+   { 36499, 0x000088F1 }, /* GL_TEXTURE_STENCIL_SIZE */
+   { 36523, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */
+   { 36548, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
+   { 36582, 0x00001000 }, /* GL_TEXTURE_WIDTH */
+   { 36599, 0x00008072 }, /* GL_TEXTURE_WRAP_R */
+   { 36617, 0x00002802 }, /* GL_TEXTURE_WRAP_S */
+   { 36635, 0x00002803 }, /* GL_TEXTURE_WRAP_T */
+   { 36653, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */
+   { 36673, 0x00008648 }, /* GL_TRACK_MATRIX_NV */
+   { 36692, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */
+   { 36721, 0x00001000 }, /* GL_TRANSFORM_BIT */
+   { 36738, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */
+   { 36764, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */
+   { 36794, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
+   { 36826, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
+   { 36856, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */
+   { 36890, 0x0000862C }, /* GL_TRANSPOSE_NV */
+   { 36906, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */
+   { 36937, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */
+   { 36972, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */
+   { 37000, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */
+   { 37032, 0x00000004 }, /* GL_TRIANGLES */
+   { 37045, 0x00000006 }, /* GL_TRIANGLE_FAN */
+   { 37061, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */
+   { 37082, 0x00000005 }, /* GL_TRIANGLE_STRIP */
+   { 37100, 0x00000001 }, /* GL_TRUE */
+   { 37108, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */
+   { 37128, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */
+   { 37151, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */
+   { 37171, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */
+   { 37192, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */
+   { 37214, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */
+   { 37236, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */
+   { 37256, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */
+   { 37277, 0x00001401 }, /* GL_UNSIGNED_BYTE */
+   { 37294, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */
+   { 37321, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */
+   { 37344, 0x00001405 }, /* GL_UNSIGNED_INT */
+   { 37360, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */
+   { 37387, 0x000084FA }, /* GL_UNSIGNED_INT_24_8 */
+   { 37408, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */
+   { 37432, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */
+   { 37463, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */
+   { 37487, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */
+   { 37515, 0x00008C17 }, /* GL_UNSIGNED_NORMALIZED */
+   { 37538, 0x00001403 }, /* GL_UNSIGNED_SHORT */
+   { 37556, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
+   { 37586, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */
+   { 37612, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
+   { 37642, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */
+   { 37668, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */
+   { 37692, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */
+   { 37720, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */
+   { 37748, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */
+   { 37775, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
+   { 37807, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */
+   { 37838, 0x00008CA2 }, /* GL_UPPER_LEFT */
+   { 37852, 0x00002A20 }, /* GL_V2F */
+   { 37859, 0x00002A21 }, /* GL_V3F */
+   { 37866, 0x00008B83 }, /* GL_VALIDATE_STATUS */
+   { 37885, 0x00001F00 }, /* GL_VENDOR */
+   { 37895, 0x00001F02 }, /* GL_VERSION */
+   { 37906, 0x00008074 }, /* GL_VERTEX_ARRAY */
+   { 37922, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING */
+   { 37946, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */
+   { 37976, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
+   { 38007, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */
+   { 38042, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */
+   { 38066, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */
+   { 38087, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */
+   { 38110, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */
+   { 38131, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
+   { 38158, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
+   { 38186, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
+   { 38214, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
+   { 38242, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
+   { 38270, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
+   { 38298, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
+   { 38326, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
+   { 38353, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
+   { 38380, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
+   { 38407, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
+   { 38434, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
+   { 38461, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
+   { 38488, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
+   { 38515, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
+   { 38542, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
+   { 38569, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
+   { 38607, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */
+   { 38649, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
+   { 38680, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */
+   { 38715, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
+   { 38749, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */
+   { 38787, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
+   { 38818, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */
+   { 38853, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
+   { 38881, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */
+   { 38913, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
+   { 38943, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */
+   { 38977, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
+   { 39005, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */
+   { 39037, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */
+   { 39057, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */
+   { 39079, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */
+   { 39108, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */
+   { 39129, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */
+   { 39158, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */
+   { 39191, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */
+   { 39223, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */
+   { 39250, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */
+   { 39281, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */
+   { 39311, 0x00008B31 }, /* GL_VERTEX_SHADER */
+   { 39328, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */
+   { 39349, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */
+   { 39376, 0x00000BA2 }, /* GL_VIEWPORT */
+   { 39388, 0x00000800 }, /* GL_VIEWPORT_BIT */
+   { 39404, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */
+   { 39424, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
+   { 39455, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */
+   { 39490, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */
+   { 39518, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */
+   { 39543, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
+   { 39570, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */
+   { 39595, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */
+   { 39619, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */
+   { 39638, 0x000088B9 }, /* GL_WRITE_ONLY */
+   { 39652, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */
+   { 39670, 0x00001506 }, /* GL_XOR */
+   { 39677, 0x000085B9 }, /* GL_YCBCR_422_APPLE */
+   { 39696, 0x00008757 }, /* GL_YCBCR_MESA */
+   { 39710, 0x00000000 }, /* GL_ZERO */
+   { 39718, 0x00000D16 }, /* GL_ZOOM_X */
+   { 39728, 0x00000D17 }, /* GL_ZOOM_Y */
 };
 
-static const unsigned reduced_enums[1319] =
+static const unsigned reduced_enums[1325] =
 {
-       469, /* GL_FALSE */
-       683, /* GL_LINES */
-       685, /* GL_LINE_LOOP */
-       692, /* GL_LINE_STRIP */
-      1709, /* GL_TRIANGLES */
-      1712, /* GL_TRIANGLE_STRIP */
-      1710, /* GL_TRIANGLE_FAN */
-      1254, /* GL_QUADS */
-      1256, /* GL_QUAD_STRIP */
-      1142, /* GL_POLYGON */
-      1154, /* GL_POLYGON_STIPPLE_BIT */
-      1103, /* GL_PIXEL_MODE_BIT */
-       670, /* GL_LIGHTING_BIT */
-       497, /* GL_FOG_BIT */
+       471, /* GL_FALSE */
+       687, /* GL_LINES */
+       689, /* GL_LINE_LOOP */
+       696, /* GL_LINE_STRIP */
+      1721, /* GL_TRIANGLES */
+      1724, /* GL_TRIANGLE_STRIP */
+      1722, /* GL_TRIANGLE_FAN */
+      1265, /* GL_QUADS */
+      1268, /* GL_QUAD_STRIP */
+      1152, /* GL_POLYGON */
+      1164, /* GL_POLYGON_STIPPLE_BIT */
+      1113, /* GL_PIXEL_MODE_BIT */
+       674, /* GL_LIGHTING_BIT */
+       500, /* GL_FOG_BIT */
          8, /* GL_ACCUM */
-       702, /* GL_LOAD */
-      1308, /* GL_RETURN */
-       976, /* GL_MULT */
+       706, /* GL_LOAD */
+      1320, /* GL_RETURN */
+       986, /* GL_MULT */
         23, /* GL_ADD */
-       992, /* GL_NEVER */
-       660, /* GL_LESS */
-       459, /* GL_EQUAL */
-       659, /* GL_LEQUAL */
-       583, /* GL_GREATER */
-      1007, /* GL_NOTEQUAL */
-       582, /* GL_GEQUAL */
+      1002, /* GL_NEVER */
+       664, /* GL_LESS */
+       461, /* GL_EQUAL */
+       663, /* GL_LEQUAL */
+       586, /* GL_GREATER */
+      1017, /* GL_NOTEQUAL */
+       585, /* GL_GEQUAL */
         46, /* GL_ALWAYS */
-      1448, /* GL_SRC_COLOR */
-      1036, /* GL_ONE_MINUS_SRC_COLOR */
-      1446, /* GL_SRC_ALPHA */
-      1035, /* GL_ONE_MINUS_SRC_ALPHA */
-       438, /* GL_DST_ALPHA */
-      1033, /* GL_ONE_MINUS_DST_ALPHA */
-       439, /* GL_DST_COLOR */
-      1034, /* GL_ONE_MINUS_DST_COLOR */
-      1447, /* GL_SRC_ALPHA_SATURATE */
-       570, /* GL_FRONT_LEFT */
-       571, /* GL_FRONT_RIGHT */
+      1460, /* GL_SRC_COLOR */
+      1046, /* GL_ONE_MINUS_SRC_COLOR */
+      1458, /* GL_SRC_ALPHA */
+      1045, /* GL_ONE_MINUS_SRC_ALPHA */
+       440, /* GL_DST_ALPHA */
+      1043, /* GL_ONE_MINUS_DST_ALPHA */
+       441, /* GL_DST_COLOR */
+      1044, /* GL_ONE_MINUS_DST_COLOR */
+      1459, /* GL_SRC_ALPHA_SATURATE */
+       573, /* GL_FRONT_LEFT */
+       574, /* GL_FRONT_RIGHT */
         68, /* GL_BACK_LEFT */
         69, /* GL_BACK_RIGHT */
-       567, /* GL_FRONT */
+       570, /* GL_FRONT */
         67, /* GL_BACK */
-       658, /* GL_LEFT */
-      1350, /* GL_RIGHT */
-       568, /* GL_FRONT_AND_BACK */
+       662, /* GL_LEFT */
+      1362, /* GL_RIGHT */
+       571, /* GL_FRONT_AND_BACK */
         62, /* GL_AUX0 */
         63, /* GL_AUX1 */
         64, /* GL_AUX2 */
         65, /* GL_AUX3 */
-       649, /* GL_INVALID_ENUM */
-       653, /* GL_INVALID_VALUE */
-       652, /* GL_INVALID_OPERATION */
-      1453, /* GL_STACK_OVERFLOW */
-      1454, /* GL_STACK_UNDERFLOW */
-      1061, /* GL_OUT_OF_MEMORY */
-       650, /* GL_INVALID_FRAMEBUFFER_OPERATION */
+       652, /* GL_INVALID_ENUM */
+       656, /* GL_INVALID_VALUE */
+       655, /* GL_INVALID_OPERATION */
+      1465, /* GL_STACK_OVERFLOW */
+      1466, /* GL_STACK_UNDERFLOW */
+      1071, /* GL_OUT_OF_MEMORY */
+       653, /* GL_INVALID_FRAMEBUFFER_OPERATION */
          0, /* GL_2D */
          2, /* GL_3D */
          3, /* GL_3D_COLOR */
          4, /* GL_3D_COLOR_TEXTURE */
          6, /* GL_4D_COLOR_TEXTURE */
-      1081, /* GL_PASS_THROUGH_TOKEN */
-      1141, /* GL_POINT_TOKEN */
-       693, /* GL_LINE_TOKEN */
-      1155, /* GL_POLYGON_TOKEN */
+      1091, /* GL_PASS_THROUGH_TOKEN */
+      1151, /* GL_POINT_TOKEN */
+       697, /* GL_LINE_TOKEN */
+      1165, /* GL_POLYGON_TOKEN */
         73, /* GL_BITMAP_TOKEN */
-       437, /* GL_DRAW_PIXEL_TOKEN */
+       439, /* GL_DRAW_PIXEL_TOKEN */
        297, /* GL_COPY_PIXEL_TOKEN */
-       686, /* GL_LINE_RESET_TOKEN */
-       462, /* GL_EXP */
-       463, /* GL_EXP2 */
-       331, /* GL_CW */
+       690, /* GL_LINE_RESET_TOKEN */
+       464, /* GL_EXP */
+       465, /* GL_EXP2 */
+       333, /* GL_CW */
        122, /* GL_CCW */
        143, /* GL_COEFF */
-      1058, /* GL_ORDER */
-       375, /* GL_DOMAIN */
-       305, /* GL_CURRENT_COLOR */
-       308, /* GL_CURRENT_INDEX */
-       314, /* GL_CURRENT_NORMAL */
-       327, /* GL_CURRENT_TEXTURE_COORDS */
-       319, /* GL_CURRENT_RASTER_COLOR */
-       321, /* GL_CURRENT_RASTER_INDEX */
-       325, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
-       322, /* GL_CURRENT_RASTER_POSITION */
-       323, /* GL_CURRENT_RASTER_POSITION_VALID */
-       320, /* GL_CURRENT_RASTER_DISTANCE */
-      1134, /* GL_POINT_SMOOTH */
-      1123, /* GL_POINT_SIZE */
-      1133, /* GL_POINT_SIZE_RANGE */
-      1124, /* GL_POINT_SIZE_GRANULARITY */
-       687, /* GL_LINE_SMOOTH */
-       694, /* GL_LINE_WIDTH */
-       696, /* GL_LINE_WIDTH_RANGE */
-       695, /* GL_LINE_WIDTH_GRANULARITY */
-       689, /* GL_LINE_STIPPLE */
-       690, /* GL_LINE_STIPPLE_PATTERN */
-       691, /* GL_LINE_STIPPLE_REPEAT */
-       701, /* GL_LIST_MODE */
-       860, /* GL_MAX_LIST_NESTING */
-       698, /* GL_LIST_BASE */
-       700, /* GL_LIST_INDEX */
-      1144, /* GL_POLYGON_MODE */
-      1151, /* GL_POLYGON_SMOOTH */
-      1153, /* GL_POLYGON_STIPPLE */
-       448, /* GL_EDGE_FLAG */
-       298, /* GL_CULL_FACE */
-       299, /* GL_CULL_FACE_MODE */
-       569, /* GL_FRONT_FACE */
-       669, /* GL_LIGHTING */
-       674, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
-       675, /* GL_LIGHT_MODEL_TWO_SIDE */
-       671, /* GL_LIGHT_MODEL_AMBIENT */
-      1396, /* GL_SHADE_MODEL */
+      1068, /* GL_ORDER */
+       377, /* GL_DOMAIN */
+       307, /* GL_CURRENT_COLOR */
+       310, /* GL_CURRENT_INDEX */
+       316, /* GL_CURRENT_NORMAL */
+       329, /* GL_CURRENT_TEXTURE_COORDS */
+       321, /* GL_CURRENT_RASTER_COLOR */
+       323, /* GL_CURRENT_RASTER_INDEX */
+       327, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
+       324, /* GL_CURRENT_RASTER_POSITION */
+       325, /* GL_CURRENT_RASTER_POSITION_VALID */
+       322, /* GL_CURRENT_RASTER_DISTANCE */
+      1144, /* GL_POINT_SMOOTH */
+      1133, /* GL_POINT_SIZE */
+      1143, /* GL_POINT_SIZE_RANGE */
+      1134, /* GL_POINT_SIZE_GRANULARITY */
+       691, /* GL_LINE_SMOOTH */
+       698, /* GL_LINE_WIDTH */
+       700, /* GL_LINE_WIDTH_RANGE */
+       699, /* GL_LINE_WIDTH_GRANULARITY */
+       693, /* GL_LINE_STIPPLE */
+       694, /* GL_LINE_STIPPLE_PATTERN */
+       695, /* GL_LINE_STIPPLE_REPEAT */
+       705, /* GL_LIST_MODE */
+       870, /* GL_MAX_LIST_NESTING */
+       702, /* GL_LIST_BASE */
+       704, /* GL_LIST_INDEX */
+      1154, /* GL_POLYGON_MODE */
+      1161, /* GL_POLYGON_SMOOTH */
+      1163, /* GL_POLYGON_STIPPLE */
+       450, /* GL_EDGE_FLAG */
+       300, /* GL_CULL_FACE */
+       301, /* GL_CULL_FACE_MODE */
+       572, /* GL_FRONT_FACE */
+       673, /* GL_LIGHTING */
+       678, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
+       679, /* GL_LIGHT_MODEL_TWO_SIDE */
+       675, /* GL_LIGHT_MODEL_AMBIENT */
+      1408, /* GL_SHADE_MODEL */
        190, /* GL_COLOR_MATERIAL_FACE */
        191, /* GL_COLOR_MATERIAL_PARAMETER */
        189, /* GL_COLOR_MATERIAL */
-       496, /* GL_FOG */
-       518, /* GL_FOG_INDEX */
-       514, /* GL_FOG_DENSITY */
-       522, /* GL_FOG_START */
-       516, /* GL_FOG_END */
-       519, /* GL_FOG_MODE */
-       498, /* GL_FOG_COLOR */
-       362, /* GL_DEPTH_RANGE */
-       369, /* GL_DEPTH_TEST */
-       372, /* GL_DEPTH_WRITEMASK */
-       350, /* GL_DEPTH_CLEAR_VALUE */
-       361, /* GL_DEPTH_FUNC */
+       499, /* GL_FOG */
+       521, /* GL_FOG_INDEX */
+       517, /* GL_FOG_DENSITY */
+       525, /* GL_FOG_START */
+       519, /* GL_FOG_END */
+       522, /* GL_FOG_MODE */
+       501, /* GL_FOG_COLOR */
+       364, /* GL_DEPTH_RANGE */
+       371, /* GL_DEPTH_TEST */
+       374, /* GL_DEPTH_WRITEMASK */
+       352, /* GL_DEPTH_CLEAR_VALUE */
+       363, /* GL_DEPTH_FUNC */
         12, /* GL_ACCUM_CLEAR_VALUE */
-      1489, /* GL_STENCIL_TEST */
-      1477, /* GL_STENCIL_CLEAR_VALUE */
-      1479, /* GL_STENCIL_FUNC */
-      1491, /* GL_STENCIL_VALUE_MASK */
-      1478, /* GL_STENCIL_FAIL */
-      1486, /* GL_STENCIL_PASS_DEPTH_FAIL */
-      1487, /* GL_STENCIL_PASS_DEPTH_PASS */
-      1488, /* GL_STENCIL_REF */
-      1492, /* GL_STENCIL_WRITEMASK */
-       829, /* GL_MATRIX_MODE */
-       997, /* GL_NORMALIZE */
-      1801, /* GL_VIEWPORT */
-       971, /* GL_MODELVIEW_STACK_DEPTH */
-      1234, /* GL_PROJECTION_STACK_DEPTH */
-      1687, /* GL_TEXTURE_STACK_DEPTH */
-       969, /* GL_MODELVIEW_MATRIX */
-      1233, /* GL_PROJECTION_MATRIX */
-      1672, /* GL_TEXTURE_MATRIX */
+      1501, /* GL_STENCIL_TEST */
+      1489, /* GL_STENCIL_CLEAR_VALUE */
+      1491, /* GL_STENCIL_FUNC */
+      1503, /* GL_STENCIL_VALUE_MASK */
+      1490, /* GL_STENCIL_FAIL */
+      1498, /* GL_STENCIL_PASS_DEPTH_FAIL */
+      1499, /* GL_STENCIL_PASS_DEPTH_PASS */
+      1500, /* GL_STENCIL_REF */
+      1504, /* GL_STENCIL_WRITEMASK */
+       839, /* GL_MATRIX_MODE */
+      1007, /* GL_NORMALIZE */
+      1814, /* GL_VIEWPORT */
+       981, /* GL_MODELVIEW_STACK_DEPTH */
+      1244, /* GL_PROJECTION_STACK_DEPTH */
+      1699, /* GL_TEXTURE_STACK_DEPTH */
+       979, /* GL_MODELVIEW_MATRIX */
+      1243, /* GL_PROJECTION_MATRIX */
+      1684, /* GL_TEXTURE_MATRIX */
         60, /* GL_ATTRIB_STACK_DEPTH */
        133, /* GL_CLIENT_ATTRIB_STACK_DEPTH */
         43, /* GL_ALPHA_TEST */
         44, /* GL_ALPHA_TEST_FUNC */
         45, /* GL_ALPHA_TEST_REF */
-       374, /* GL_DITHER */
+       376, /* GL_DITHER */
         77, /* GL_BLEND_DST */
         86, /* GL_BLEND_SRC */
         74, /* GL_BLEND */
-       704, /* GL_LOGIC_OP_MODE */
-       623, /* GL_INDEX_LOGIC_OP */
+       708, /* GL_LOGIC_OP_MODE */
+       626, /* GL_INDEX_LOGIC_OP */
        188, /* GL_COLOR_LOGIC_OP */
         66, /* GL_AUX_BUFFERS */
-       385, /* GL_DRAW_BUFFER */
-      1266, /* GL_READ_BUFFER */
-      1377, /* GL_SCISSOR_BOX */
-      1378, /* GL_SCISSOR_TEST */
-       622, /* GL_INDEX_CLEAR_VALUE */
-       627, /* GL_INDEX_WRITEMASK */
+       387, /* GL_DRAW_BUFFER */
+      1278, /* GL_READ_BUFFER */
+      1389, /* GL_SCISSOR_BOX */
+      1390, /* GL_SCISSOR_TEST */
+       625, /* GL_INDEX_CLEAR_VALUE */
+       630, /* GL_INDEX_WRITEMASK */
        185, /* GL_COLOR_CLEAR_VALUE */
        227, /* GL_COLOR_WRITEMASK */
-       624, /* GL_INDEX_MODE */
-      1343, /* GL_RGBA_MODE */
-       384, /* GL_DOUBLEBUFFER */
-      1493, /* GL_STEREO */
-      1301, /* GL_RENDER_MODE */
-      1082, /* GL_PERSPECTIVE_CORRECTION_HINT */
-      1135, /* GL_POINT_SMOOTH_HINT */
-       688, /* GL_LINE_SMOOTH_HINT */
-      1152, /* GL_POLYGON_SMOOTH_HINT */
-       517, /* GL_FOG_HINT */
-      1653, /* GL_TEXTURE_GEN_S */
-      1654, /* GL_TEXTURE_GEN_T */
-      1652, /* GL_TEXTURE_GEN_R */
-      1651, /* GL_TEXTURE_GEN_Q */
-      1095, /* GL_PIXEL_MAP_I_TO_I */
-      1101, /* GL_PIXEL_MAP_S_TO_S */
-      1097, /* GL_PIXEL_MAP_I_TO_R */
-      1093, /* GL_PIXEL_MAP_I_TO_G */
-      1091, /* GL_PIXEL_MAP_I_TO_B */
-      1089, /* GL_PIXEL_MAP_I_TO_A */
-      1099, /* GL_PIXEL_MAP_R_TO_R */
-      1087, /* GL_PIXEL_MAP_G_TO_G */
-      1085, /* GL_PIXEL_MAP_B_TO_B */
-      1083, /* GL_PIXEL_MAP_A_TO_A */
-      1096, /* GL_PIXEL_MAP_I_TO_I_SIZE */
-      1102, /* GL_PIXEL_MAP_S_TO_S_SIZE */
-      1098, /* GL_PIXEL_MAP_I_TO_R_SIZE */
-      1094, /* GL_PIXEL_MAP_I_TO_G_SIZE */
-      1092, /* GL_PIXEL_MAP_I_TO_B_SIZE */
-      1090, /* GL_PIXEL_MAP_I_TO_A_SIZE */
-      1100, /* GL_PIXEL_MAP_R_TO_R_SIZE */
-      1088, /* GL_PIXEL_MAP_G_TO_G_SIZE */
-      1086, /* GL_PIXEL_MAP_B_TO_B_SIZE */
-      1084, /* GL_PIXEL_MAP_A_TO_A_SIZE */
-      1721, /* GL_UNPACK_SWAP_BYTES */
-      1716, /* GL_UNPACK_LSB_FIRST */
-      1717, /* GL_UNPACK_ROW_LENGTH */
-      1720, /* GL_UNPACK_SKIP_ROWS */
-      1719, /* GL_UNPACK_SKIP_PIXELS */
-      1714, /* GL_UNPACK_ALIGNMENT */
-      1070, /* GL_PACK_SWAP_BYTES */
-      1065, /* GL_PACK_LSB_FIRST */
-      1066, /* GL_PACK_ROW_LENGTH */
-      1069, /* GL_PACK_SKIP_ROWS */
-      1068, /* GL_PACK_SKIP_PIXELS */
-      1062, /* GL_PACK_ALIGNMENT */
-       782, /* GL_MAP_COLOR */
-       783, /* GL_MAP_STENCIL */
-       626, /* GL_INDEX_SHIFT */
-       625, /* GL_INDEX_OFFSET */
-      1279, /* GL_RED_SCALE */
-      1277, /* GL_RED_BIAS */
-      1818, /* GL_ZOOM_X */
-      1819, /* GL_ZOOM_Y */
-       587, /* GL_GREEN_SCALE */
-       585, /* GL_GREEN_BIAS */
+       627, /* GL_INDEX_MODE */
+      1355, /* GL_RGBA_MODE */
+       386, /* GL_DOUBLEBUFFER */
+      1505, /* GL_STEREO */
+      1313, /* GL_RENDER_MODE */
+      1092, /* GL_PERSPECTIVE_CORRECTION_HINT */
+      1145, /* GL_POINT_SMOOTH_HINT */
+       692, /* GL_LINE_SMOOTH_HINT */
+      1162, /* GL_POLYGON_SMOOTH_HINT */
+       520, /* GL_FOG_HINT */
+      1665, /* GL_TEXTURE_GEN_S */
+      1666, /* GL_TEXTURE_GEN_T */
+      1664, /* GL_TEXTURE_GEN_R */
+      1663, /* GL_TEXTURE_GEN_Q */
+      1105, /* GL_PIXEL_MAP_I_TO_I */
+      1111, /* GL_PIXEL_MAP_S_TO_S */
+      1107, /* GL_PIXEL_MAP_I_TO_R */
+      1103, /* GL_PIXEL_MAP_I_TO_G */
+      1101, /* GL_PIXEL_MAP_I_TO_B */
+      1099, /* GL_PIXEL_MAP_I_TO_A */
+      1109, /* GL_PIXEL_MAP_R_TO_R */
+      1097, /* GL_PIXEL_MAP_G_TO_G */
+      1095, /* GL_PIXEL_MAP_B_TO_B */
+      1093, /* GL_PIXEL_MAP_A_TO_A */
+      1106, /* GL_PIXEL_MAP_I_TO_I_SIZE */
+      1112, /* GL_PIXEL_MAP_S_TO_S_SIZE */
+      1108, /* GL_PIXEL_MAP_I_TO_R_SIZE */
+      1104, /* GL_PIXEL_MAP_I_TO_G_SIZE */
+      1102, /* GL_PIXEL_MAP_I_TO_B_SIZE */
+      1100, /* GL_PIXEL_MAP_I_TO_A_SIZE */
+      1110, /* GL_PIXEL_MAP_R_TO_R_SIZE */
+      1098, /* GL_PIXEL_MAP_G_TO_G_SIZE */
+      1096, /* GL_PIXEL_MAP_B_TO_B_SIZE */
+      1094, /* GL_PIXEL_MAP_A_TO_A_SIZE */
+      1733, /* GL_UNPACK_SWAP_BYTES */
+      1728, /* GL_UNPACK_LSB_FIRST */
+      1729, /* GL_UNPACK_ROW_LENGTH */
+      1732, /* GL_UNPACK_SKIP_ROWS */
+      1731, /* GL_UNPACK_SKIP_PIXELS */
+      1726, /* GL_UNPACK_ALIGNMENT */
+      1080, /* GL_PACK_SWAP_BYTES */
+      1075, /* GL_PACK_LSB_FIRST */
+      1076, /* GL_PACK_ROW_LENGTH */
+      1079, /* GL_PACK_SKIP_ROWS */
+      1078, /* GL_PACK_SKIP_PIXELS */
+      1072, /* GL_PACK_ALIGNMENT */
+       786, /* GL_MAP_COLOR */
+       791, /* GL_MAP_STENCIL */
+       629, /* GL_INDEX_SHIFT */
+       628, /* GL_INDEX_OFFSET */
+      1291, /* GL_RED_SCALE */
+      1289, /* GL_RED_BIAS */
+      1831, /* GL_ZOOM_X */
+      1832, /* GL_ZOOM_Y */
+       590, /* GL_GREEN_SCALE */
+       588, /* GL_GREEN_BIAS */
         92, /* GL_BLUE_SCALE */
         90, /* GL_BLUE_BIAS */
         42, /* GL_ALPHA_SCALE */
         40, /* GL_ALPHA_BIAS */
-       363, /* GL_DEPTH_SCALE */
-       344, /* GL_DEPTH_BIAS */
-       855, /* GL_MAX_EVAL_ORDER */
-       859, /* GL_MAX_LIGHTS */
-       838, /* GL_MAX_CLIP_PLANES */
-       904, /* GL_MAX_TEXTURE_SIZE */
-       865, /* GL_MAX_PIXEL_MAP_TABLE */
-       834, /* GL_MAX_ATTRIB_STACK_DEPTH */
-       862, /* GL_MAX_MODELVIEW_STACK_DEPTH */
-       863, /* GL_MAX_NAME_STACK_DEPTH */
-       891, /* GL_MAX_PROJECTION_STACK_DEPTH */
-       905, /* GL_MAX_TEXTURE_STACK_DEPTH */
-       919, /* GL_MAX_VIEWPORT_DIMS */
-       835, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
-      1500, /* GL_SUBPIXEL_BITS */
-       621, /* GL_INDEX_BITS */
-      1278, /* GL_RED_BITS */
-       586, /* GL_GREEN_BITS */
+       365, /* GL_DEPTH_SCALE */
+       346, /* GL_DEPTH_BIAS */
+       865, /* GL_MAX_EVAL_ORDER */
+       869, /* GL_MAX_LIGHTS */
+       848, /* GL_MAX_CLIP_PLANES */
+       914, /* GL_MAX_TEXTURE_SIZE */
+       875, /* GL_MAX_PIXEL_MAP_TABLE */
+       844, /* GL_MAX_ATTRIB_STACK_DEPTH */
+       872, /* GL_MAX_MODELVIEW_STACK_DEPTH */
+       873, /* GL_MAX_NAME_STACK_DEPTH */
+       901, /* GL_MAX_PROJECTION_STACK_DEPTH */
+       915, /* GL_MAX_TEXTURE_STACK_DEPTH */
+       929, /* GL_MAX_VIEWPORT_DIMS */
+       845, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
+      1512, /* GL_SUBPIXEL_BITS */
+       624, /* GL_INDEX_BITS */
+      1290, /* GL_RED_BITS */
+       589, /* GL_GREEN_BITS */
         91, /* GL_BLUE_BITS */
         41, /* GL_ALPHA_BITS */
-       345, /* GL_DEPTH_BITS */
-      1475, /* GL_STENCIL_BITS */
+       347, /* GL_DEPTH_BITS */
+      1487, /* GL_STENCIL_BITS */
         14, /* GL_ACCUM_RED_BITS */
         13, /* GL_ACCUM_GREEN_BITS */
         10, /* GL_ACCUM_BLUE_BITS */
          9, /* GL_ACCUM_ALPHA_BITS */
-       985, /* GL_NAME_STACK_DEPTH */
+       995, /* GL_NAME_STACK_DEPTH */
         61, /* GL_AUTO_NORMAL */
-       728, /* GL_MAP1_COLOR_4 */
-       731, /* GL_MAP1_INDEX */
-       732, /* GL_MAP1_NORMAL */
-       733, /* GL_MAP1_TEXTURE_COORD_1 */
-       734, /* GL_MAP1_TEXTURE_COORD_2 */
-       735, /* GL_MAP1_TEXTURE_COORD_3 */
-       736, /* GL_MAP1_TEXTURE_COORD_4 */
-       737, /* GL_MAP1_VERTEX_3 */
-       738, /* GL_MAP1_VERTEX_4 */
-       755, /* GL_MAP2_COLOR_4 */
-       758, /* GL_MAP2_INDEX */
-       759, /* GL_MAP2_NORMAL */
-       760, /* GL_MAP2_TEXTURE_COORD_1 */
-       761, /* GL_MAP2_TEXTURE_COORD_2 */
-       762, /* GL_MAP2_TEXTURE_COORD_3 */
-       763, /* GL_MAP2_TEXTURE_COORD_4 */
-       764, /* GL_MAP2_VERTEX_3 */
-       765, /* GL_MAP2_VERTEX_4 */
-       729, /* GL_MAP1_GRID_DOMAIN */
-       730, /* GL_MAP1_GRID_SEGMENTS */
-       756, /* GL_MAP2_GRID_DOMAIN */
-       757, /* GL_MAP2_GRID_SEGMENTS */
-      1577, /* GL_TEXTURE_1D */
-      1579, /* GL_TEXTURE_2D */
-       472, /* GL_FEEDBACK_BUFFER_POINTER */
-       473, /* GL_FEEDBACK_BUFFER_SIZE */
-       474, /* GL_FEEDBACK_BUFFER_TYPE */
-      1387, /* GL_SELECTION_BUFFER_POINTER */
-      1388, /* GL_SELECTION_BUFFER_SIZE */
-      1691, /* GL_TEXTURE_WIDTH */
-      1658, /* GL_TEXTURE_HEIGHT */
-      1614, /* GL_TEXTURE_COMPONENTS */
-      1598, /* GL_TEXTURE_BORDER_COLOR */
-      1597, /* GL_TEXTURE_BORDER */
-       376, /* GL_DONT_CARE */
-       470, /* GL_FASTEST */
-       993, /* GL_NICEST */
+       732, /* GL_MAP1_COLOR_4 */
+       735, /* GL_MAP1_INDEX */
+       736, /* GL_MAP1_NORMAL */
+       737, /* GL_MAP1_TEXTURE_COORD_1 */
+       738, /* GL_MAP1_TEXTURE_COORD_2 */
+       739, /* GL_MAP1_TEXTURE_COORD_3 */
+       740, /* GL_MAP1_TEXTURE_COORD_4 */
+       741, /* GL_MAP1_VERTEX_3 */
+       742, /* GL_MAP1_VERTEX_4 */
+       759, /* GL_MAP2_COLOR_4 */
+       762, /* GL_MAP2_INDEX */
+       763, /* GL_MAP2_NORMAL */
+       764, /* GL_MAP2_TEXTURE_COORD_1 */
+       765, /* GL_MAP2_TEXTURE_COORD_2 */
+       766, /* GL_MAP2_TEXTURE_COORD_3 */
+       767, /* GL_MAP2_TEXTURE_COORD_4 */
+       768, /* GL_MAP2_VERTEX_3 */
+       769, /* GL_MAP2_VERTEX_4 */
+       733, /* GL_MAP1_GRID_DOMAIN */
+       734, /* GL_MAP1_GRID_SEGMENTS */
+       760, /* GL_MAP2_GRID_DOMAIN */
+       761, /* GL_MAP2_GRID_SEGMENTS */
+      1589, /* GL_TEXTURE_1D */
+      1591, /* GL_TEXTURE_2D */
+       474, /* GL_FEEDBACK_BUFFER_POINTER */
+       475, /* GL_FEEDBACK_BUFFER_SIZE */
+       476, /* GL_FEEDBACK_BUFFER_TYPE */
+      1399, /* GL_SELECTION_BUFFER_POINTER */
+      1400, /* GL_SELECTION_BUFFER_SIZE */
+      1703, /* GL_TEXTURE_WIDTH */
+      1670, /* GL_TEXTURE_HEIGHT */
+      1626, /* GL_TEXTURE_COMPONENTS */
+      1610, /* GL_TEXTURE_BORDER_COLOR */
+      1609, /* GL_TEXTURE_BORDER */
+       378, /* GL_DONT_CARE */
+       472, /* GL_FASTEST */
+      1003, /* GL_NICEST */
         47, /* GL_AMBIENT */
-       373, /* GL_DIFFUSE */
-      1435, /* GL_SPECULAR */
-      1156, /* GL_POSITION */
-      1438, /* GL_SPOT_DIRECTION */
-      1439, /* GL_SPOT_EXPONENT */
-      1437, /* GL_SPOT_CUTOFF */
+       375, /* GL_DIFFUSE */
+      1447, /* GL_SPECULAR */
+      1166, /* GL_POSITION */
+      1450, /* GL_SPOT_DIRECTION */
+      1451, /* GL_SPOT_EXPONENT */
+      1449, /* GL_SPOT_CUTOFF */
        271, /* GL_CONSTANT_ATTENUATION */
-       678, /* GL_LINEAR_ATTENUATION */
-      1253, /* GL_QUADRATIC_ATTENUATION */
+       682, /* GL_LINEAR_ATTENUATION */
+      1264, /* GL_QUADRATIC_ATTENUATION */
        241, /* GL_COMPILE */
        242, /* GL_COMPILE_AND_EXECUTE */
        117, /* GL_BYTE */
-      1722, /* GL_UNSIGNED_BYTE */
-      1401, /* GL_SHORT */
-      1733, /* GL_UNSIGNED_SHORT */
-       629, /* GL_INT */
-      1725, /* GL_UNSIGNED_INT */
-       477, /* GL_FLOAT */
+      1734, /* GL_UNSIGNED_BYTE */
+      1413, /* GL_SHORT */
+      1745, /* GL_UNSIGNED_SHORT */
+       632, /* GL_INT */
+      1737, /* GL_UNSIGNED_INT */
+       480, /* GL_FLOAT */
          1, /* GL_2_BYTES */
          5, /* GL_3_BYTES */
          7, /* GL_4_BYTES */
-       383, /* GL_DOUBLE */
+       385, /* GL_DOUBLE */
        129, /* GL_CLEAR */
         49, /* GL_AND */
         51, /* GL_AND_REVERSE */
        295, /* GL_COPY */
         50, /* GL_AND_INVERTED */
-       995, /* GL_NOOP */
-      1814, /* GL_XOR */
-      1057, /* GL_OR */
-       996, /* GL_NOR */
-       460, /* GL_EQUIV */
-       656, /* GL_INVERT */
-      1060, /* GL_OR_REVERSE */
+      1005, /* GL_NOOP */
+      1827, /* GL_XOR */
+      1067, /* GL_OR */
+      1006, /* GL_NOR */
+       462, /* GL_EQUIV */
+       659, /* GL_INVERT */
+      1070, /* GL_OR_REVERSE */
        296, /* GL_COPY_INVERTED */
-      1059, /* GL_OR_INVERTED */
-       986, /* GL_NAND */
-      1392, /* GL_SET */
-       457, /* GL_EMISSION */
-      1400, /* GL_SHININESS */
+      1069, /* GL_OR_INVERTED */
+       996, /* GL_NAND */
+      1404, /* GL_SET */
+       459, /* GL_EMISSION */
+      1412, /* GL_SHININESS */
         48, /* GL_AMBIENT_AND_DIFFUSE */
        187, /* GL_COLOR_INDEXES */
-       936, /* GL_MODELVIEW */
-      1232, /* GL_PROJECTION */
-      1512, /* GL_TEXTURE */
+       946, /* GL_MODELVIEW */
+      1242, /* GL_PROJECTION */
+      1524, /* GL_TEXTURE */
        144, /* GL_COLOR */
-       340, /* GL_DEPTH */
-      1461, /* GL_STENCIL */
+       342, /* GL_DEPTH */
+      1473, /* GL_STENCIL */
        186, /* GL_COLOR_INDEX */
-      1480, /* GL_STENCIL_INDEX */
-       351, /* GL_DEPTH_COMPONENT */
-      1274, /* GL_RED */
-       584, /* GL_GREEN */
+      1492, /* GL_STENCIL_INDEX */
+       353, /* GL_DEPTH_COMPONENT */
+      1286, /* GL_RED */
+       587, /* GL_GREEN */
         89, /* GL_BLUE */
         31, /* GL_ALPHA */
-      1309, /* GL_RGB */
-      1328, /* GL_RGBA */
-       706, /* GL_LUMINANCE */
-       727, /* GL_LUMINANCE_ALPHA */
+      1321, /* GL_RGB */
+      1340, /* GL_RGBA */
+       710, /* GL_LUMINANCE */
+       731, /* GL_LUMINANCE_ALPHA */
         72, /* GL_BITMAP */
-      1112, /* GL_POINT */
-       676, /* GL_LINE */
-       475, /* GL_FILL */
-      1283, /* GL_RENDER */
-       471, /* GL_FEEDBACK */
-      1386, /* GL_SELECT */
-       476, /* GL_FLAT */
-      1410, /* GL_SMOOTH */
-       657, /* GL_KEEP */
-      1303, /* GL_REPLACE */
-       611, /* GL_INCR */
-       336, /* GL_DECR */
-      1748, /* GL_VENDOR */
-      1300, /* GL_RENDERER */
-      1749, /* GL_VERSION */
-       464, /* GL_EXTENSIONS */
-      1351, /* GL_S */
-      1503, /* GL_T */
-      1263, /* GL_R */
-      1252, /* GL_Q */
-       972, /* GL_MODULATE */
-       335, /* GL_DECAL */
-      1648, /* GL_TEXTURE_ENV_MODE */
-      1647, /* GL_TEXTURE_ENV_COLOR */
-      1646, /* GL_TEXTURE_ENV */
-       465, /* GL_EYE_LINEAR */
-      1019, /* GL_OBJECT_LINEAR */
-      1436, /* GL_SPHERE_MAP */
-      1650, /* GL_TEXTURE_GEN_MODE */
-      1021, /* GL_OBJECT_PLANE */
-       466, /* GL_EYE_PLANE */
-       987, /* GL_NEAREST */
-       677, /* GL_LINEAR */
-       991, /* GL_NEAREST_MIPMAP_NEAREST */
-       682, /* GL_LINEAR_MIPMAP_NEAREST */
-       990, /* GL_NEAREST_MIPMAP_LINEAR */
-       681, /* GL_LINEAR_MIPMAP_LINEAR */
-      1671, /* GL_TEXTURE_MAG_FILTER */
-      1679, /* GL_TEXTURE_MIN_FILTER */
-      1693, /* GL_TEXTURE_WRAP_S */
-      1694, /* GL_TEXTURE_WRAP_T */
+      1122, /* GL_POINT */
+       680, /* GL_LINE */
+       477, /* GL_FILL */
+      1295, /* GL_RENDER */
+       473, /* GL_FEEDBACK */
+      1398, /* GL_SELECT */
+       479, /* GL_FLAT */
+      1422, /* GL_SMOOTH */
+       660, /* GL_KEEP */
+      1315, /* GL_REPLACE */
+       614, /* GL_INCR */
+       338, /* GL_DECR */
+      1760, /* GL_VENDOR */
+      1312, /* GL_RENDERER */
+      1761, /* GL_VERSION */
+       466, /* GL_EXTENSIONS */
+      1363, /* GL_S */
+      1515, /* GL_T */
+      1275, /* GL_R */
+      1263, /* GL_Q */
+       982, /* GL_MODULATE */
+       337, /* GL_DECAL */
+      1660, /* GL_TEXTURE_ENV_MODE */
+      1659, /* GL_TEXTURE_ENV_COLOR */
+      1658, /* GL_TEXTURE_ENV */
+       467, /* GL_EYE_LINEAR */
+      1029, /* GL_OBJECT_LINEAR */
+      1448, /* GL_SPHERE_MAP */
+      1662, /* GL_TEXTURE_GEN_MODE */
+      1031, /* GL_OBJECT_PLANE */
+       468, /* GL_EYE_PLANE */
+       997, /* GL_NEAREST */
+       681, /* GL_LINEAR */
+      1001, /* GL_NEAREST_MIPMAP_NEAREST */
+       686, /* GL_LINEAR_MIPMAP_NEAREST */
+      1000, /* GL_NEAREST_MIPMAP_LINEAR */
+       685, /* GL_LINEAR_MIPMAP_LINEAR */
+      1683, /* GL_TEXTURE_MAG_FILTER */
+      1691, /* GL_TEXTURE_MIN_FILTER */
+      1705, /* GL_TEXTURE_WRAP_S */
+      1706, /* GL_TEXTURE_WRAP_T */
        123, /* GL_CLAMP */
-      1302, /* GL_REPEAT */
-      1150, /* GL_POLYGON_OFFSET_UNITS */
-      1149, /* GL_POLYGON_OFFSET_POINT */
-      1148, /* GL_POLYGON_OFFSET_LINE */
-      1264, /* GL_R3_G3_B2 */
-      1745, /* GL_V2F */
-      1746, /* GL_V3F */
+      1314, /* GL_REPEAT */
+      1160, /* GL_POLYGON_OFFSET_UNITS */
+      1159, /* GL_POLYGON_OFFSET_POINT */
+      1158, /* GL_POLYGON_OFFSET_LINE */
+      1276, /* GL_R3_G3_B2 */
+      1757, /* GL_V2F */
+      1758, /* GL_V3F */
        120, /* GL_C4UB_V2F */
        121, /* GL_C4UB_V3F */
        118, /* GL_C3F_V3F */
-       984, /* GL_N3F_V3F */
+       994, /* GL_N3F_V3F */
        119, /* GL_C4F_N3F_V3F */
-      1508, /* GL_T2F_V3F */
-      1510, /* GL_T4F_V4F */
-      1506, /* GL_T2F_C4UB_V3F */
-      1504, /* GL_T2F_C3F_V3F */
-      1507, /* GL_T2F_N3F_V3F */
-      1505, /* GL_T2F_C4F_N3F_V3F */
-      1509, /* GL_T4F_C4F_N3F_V4F */
+      1520, /* GL_T2F_V3F */
+      1522, /* GL_T4F_V4F */
+      1518, /* GL_T2F_C4UB_V3F */
+      1516, /* GL_T2F_C3F_V3F */
+      1519, /* GL_T2F_N3F_V3F */
+      1517, /* GL_T2F_C4F_N3F_V3F */
+      1521, /* GL_T4F_C4F_N3F_V4F */
        136, /* GL_CLIP_PLANE0 */
        137, /* GL_CLIP_PLANE1 */
        138, /* GL_CLIP_PLANE2 */
        139, /* GL_CLIP_PLANE3 */
        140, /* GL_CLIP_PLANE4 */
        141, /* GL_CLIP_PLANE5 */
-       661, /* GL_LIGHT0 */
-       662, /* GL_LIGHT1 */
-       663, /* GL_LIGHT2 */
-       664, /* GL_LIGHT3 */
-       665, /* GL_LIGHT4 */
-       666, /* GL_LIGHT5 */
-       667, /* GL_LIGHT6 */
-       668, /* GL_LIGHT7 */
-       588, /* GL_HINT_BIT */
+       665, /* GL_LIGHT0 */
+       666, /* GL_LIGHT1 */
+       667, /* GL_LIGHT2 */
+       668, /* GL_LIGHT3 */
+       669, /* GL_LIGHT4 */
+       670, /* GL_LIGHT5 */
+       671, /* GL_LIGHT6 */
+       672, /* GL_LIGHT7 */
+       591, /* GL_HINT_BIT */
        273, /* GL_CONSTANT_COLOR */
-      1031, /* GL_ONE_MINUS_CONSTANT_COLOR */
+      1041, /* GL_ONE_MINUS_CONSTANT_COLOR */
        268, /* GL_CONSTANT_ALPHA */
-      1029, /* GL_ONE_MINUS_CONSTANT_ALPHA */
+      1039, /* GL_ONE_MINUS_CONSTANT_ALPHA */
         75, /* GL_BLEND_COLOR */
-       572, /* GL_FUNC_ADD */
-       920, /* GL_MIN */
-       831, /* GL_MAX */
+       575, /* GL_FUNC_ADD */
+       930, /* GL_MIN */
+       841, /* GL_MAX */
         80, /* GL_BLEND_EQUATION */
-       576, /* GL_FUNC_SUBTRACT */
-       574, /* GL_FUNC_REVERSE_SUBTRACT */
+       579, /* GL_FUNC_SUBTRACT */
+       577, /* GL_FUNC_REVERSE_SUBTRACT */
        276, /* GL_CONVOLUTION_1D */
        277, /* GL_CONVOLUTION_2D */
-      1389, /* GL_SEPARABLE_2D */
+      1401, /* GL_SEPARABLE_2D */
        280, /* GL_CONVOLUTION_BORDER_MODE */
        284, /* GL_CONVOLUTION_FILTER_SCALE */
        282, /* GL_CONVOLUTION_FILTER_BIAS */
-      1275, /* GL_REDUCE */
+      1287, /* GL_REDUCE */
        286, /* GL_CONVOLUTION_FORMAT */
        290, /* GL_CONVOLUTION_WIDTH */
        288, /* GL_CONVOLUTION_HEIGHT */
-       846, /* GL_MAX_CONVOLUTION_WIDTH */
-       844, /* GL_MAX_CONVOLUTION_HEIGHT */
-      1189, /* GL_POST_CONVOLUTION_RED_SCALE */
-      1185, /* GL_POST_CONVOLUTION_GREEN_SCALE */
-      1180, /* GL_POST_CONVOLUTION_BLUE_SCALE */
-      1176, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
-      1187, /* GL_POST_CONVOLUTION_RED_BIAS */
-      1183, /* GL_POST_CONVOLUTION_GREEN_BIAS */
-      1178, /* GL_POST_CONVOLUTION_BLUE_BIAS */
-      1174, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
-       589, /* GL_HISTOGRAM */
-      1236, /* GL_PROXY_HISTOGRAM */
-       605, /* GL_HISTOGRAM_WIDTH */
-       595, /* GL_HISTOGRAM_FORMAT */
-       601, /* GL_HISTOGRAM_RED_SIZE */
-       597, /* GL_HISTOGRAM_GREEN_SIZE */
-       592, /* GL_HISTOGRAM_BLUE_SIZE */
-       590, /* GL_HISTOGRAM_ALPHA_SIZE */
-       599, /* GL_HISTOGRAM_LUMINANCE_SIZE */
-       603, /* GL_HISTOGRAM_SINK */
-       921, /* GL_MINMAX */
-       923, /* GL_MINMAX_FORMAT */
-       925, /* GL_MINMAX_SINK */
-      1511, /* GL_TABLE_TOO_LARGE_EXT */
-      1724, /* GL_UNSIGNED_BYTE_3_3_2 */
-      1735, /* GL_UNSIGNED_SHORT_4_4_4_4 */
-      1737, /* GL_UNSIGNED_SHORT_5_5_5_1 */
-      1730, /* GL_UNSIGNED_INT_8_8_8_8 */
-      1726, /* GL_UNSIGNED_INT_10_10_10_2 */
-      1147, /* GL_POLYGON_OFFSET_FILL */
-      1146, /* GL_POLYGON_OFFSET_FACTOR */
-      1145, /* GL_POLYGON_OFFSET_BIAS */
-      1306, /* GL_RESCALE_NORMAL */
+       856, /* GL_MAX_CONVOLUTION_WIDTH */
+       854, /* GL_MAX_CONVOLUTION_HEIGHT */
+      1199, /* GL_POST_CONVOLUTION_RED_SCALE */
+      1195, /* GL_POST_CONVOLUTION_GREEN_SCALE */
+      1190, /* GL_POST_CONVOLUTION_BLUE_SCALE */
+      1186, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
+      1197, /* GL_POST_CONVOLUTION_RED_BIAS */
+      1193, /* GL_POST_CONVOLUTION_GREEN_BIAS */
+      1188, /* GL_POST_CONVOLUTION_BLUE_BIAS */
+      1184, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
+       592, /* GL_HISTOGRAM */
+      1247, /* GL_PROXY_HISTOGRAM */
+       608, /* GL_HISTOGRAM_WIDTH */
+       598, /* GL_HISTOGRAM_FORMAT */
+       604, /* GL_HISTOGRAM_RED_SIZE */
+       600, /* GL_HISTOGRAM_GREEN_SIZE */
+       595, /* GL_HISTOGRAM_BLUE_SIZE */
+       593, /* GL_HISTOGRAM_ALPHA_SIZE */
+       602, /* GL_HISTOGRAM_LUMINANCE_SIZE */
+       606, /* GL_HISTOGRAM_SINK */
+       931, /* GL_MINMAX */
+       933, /* GL_MINMAX_FORMAT */
+       935, /* GL_MINMAX_SINK */
+      1523, /* GL_TABLE_TOO_LARGE_EXT */
+      1736, /* GL_UNSIGNED_BYTE_3_3_2 */
+      1747, /* GL_UNSIGNED_SHORT_4_4_4_4 */
+      1749, /* GL_UNSIGNED_SHORT_5_5_5_1 */
+      1742, /* GL_UNSIGNED_INT_8_8_8_8 */
+      1738, /* GL_UNSIGNED_INT_10_10_10_2 */
+      1157, /* GL_POLYGON_OFFSET_FILL */
+      1156, /* GL_POLYGON_OFFSET_FACTOR */
+      1155, /* GL_POLYGON_OFFSET_BIAS */
+      1318, /* GL_RESCALE_NORMAL */
         36, /* GL_ALPHA4 */
         38, /* GL_ALPHA8 */
         32, /* GL_ALPHA12 */
         34, /* GL_ALPHA16 */
-       717, /* GL_LUMINANCE4 */
-       723, /* GL_LUMINANCE8 */
-       707, /* GL_LUMINANCE12 */
-       713, /* GL_LUMINANCE16 */
-       718, /* GL_LUMINANCE4_ALPHA4 */
-       721, /* GL_LUMINANCE6_ALPHA2 */
-       724, /* GL_LUMINANCE8_ALPHA8 */
-       710, /* GL_LUMINANCE12_ALPHA4 */
-       708, /* GL_LUMINANCE12_ALPHA12 */
-       714, /* GL_LUMINANCE16_ALPHA16 */
-       630, /* GL_INTENSITY */
-       635, /* GL_INTENSITY4 */
-       637, /* GL_INTENSITY8 */
-       631, /* GL_INTENSITY12 */
-       633, /* GL_INTENSITY16 */
-      1318, /* GL_RGB2_EXT */
-      1319, /* GL_RGB4 */
-      1322, /* GL_RGB5 */
-      1326, /* GL_RGB8 */
-      1310, /* GL_RGB10 */
-      1314, /* GL_RGB12 */
-      1316, /* GL_RGB16 */
-      1333, /* GL_RGBA2 */
-      1335, /* GL_RGBA4 */
-      1323, /* GL_RGB5_A1 */
-      1339, /* GL_RGBA8 */
-      1311, /* GL_RGB10_A2 */
-      1329, /* GL_RGBA12 */
-      1331, /* GL_RGBA16 */
-      1684, /* GL_TEXTURE_RED_SIZE */
-      1656, /* GL_TEXTURE_GREEN_SIZE */
-      1595, /* GL_TEXTURE_BLUE_SIZE */
-      1582, /* GL_TEXTURE_ALPHA_SIZE */
-      1669, /* GL_TEXTURE_LUMINANCE_SIZE */
-      1660, /* GL_TEXTURE_INTENSITY_SIZE */
-      1304, /* GL_REPLACE_EXT */
-      1240, /* GL_PROXY_TEXTURE_1D */
-      1243, /* GL_PROXY_TEXTURE_2D */
-      1689, /* GL_TEXTURE_TOO_LARGE_EXT */
-      1681, /* GL_TEXTURE_PRIORITY */
-      1686, /* GL_TEXTURE_RESIDENT */
-      1585, /* GL_TEXTURE_BINDING_1D */
-      1587, /* GL_TEXTURE_BINDING_2D */
-      1589, /* GL_TEXTURE_BINDING_3D */
-      1067, /* GL_PACK_SKIP_IMAGES */
-      1063, /* GL_PACK_IMAGE_HEIGHT */
-      1718, /* GL_UNPACK_SKIP_IMAGES */
-      1715, /* GL_UNPACK_IMAGE_HEIGHT */
-      1581, /* GL_TEXTURE_3D */
-      1246, /* GL_PROXY_TEXTURE_3D */
-      1643, /* GL_TEXTURE_DEPTH */
-      1692, /* GL_TEXTURE_WRAP_R */
-       832, /* GL_MAX_3D_TEXTURE_SIZE */
-      1750, /* GL_VERTEX_ARRAY */
-       998, /* GL_NORMAL_ARRAY */
+       721, /* GL_LUMINANCE4 */
+       727, /* GL_LUMINANCE8 */
+       711, /* GL_LUMINANCE12 */
+       717, /* GL_LUMINANCE16 */
+       722, /* GL_LUMINANCE4_ALPHA4 */
+       725, /* GL_LUMINANCE6_ALPHA2 */
+       728, /* GL_LUMINANCE8_ALPHA8 */
+       714, /* GL_LUMINANCE12_ALPHA4 */
+       712, /* GL_LUMINANCE12_ALPHA12 */
+       718, /* GL_LUMINANCE16_ALPHA16 */
+       633, /* GL_INTENSITY */
+       638, /* GL_INTENSITY4 */
+       640, /* GL_INTENSITY8 */
+       634, /* GL_INTENSITY12 */
+       636, /* GL_INTENSITY16 */
+      1330, /* GL_RGB2_EXT */
+      1331, /* GL_RGB4 */
+      1334, /* GL_RGB5 */
+      1338, /* GL_RGB8 */
+      1322, /* GL_RGB10 */
+      1326, /* GL_RGB12 */
+      1328, /* GL_RGB16 */
+      1345, /* GL_RGBA2 */
+      1347, /* GL_RGBA4 */
+      1335, /* GL_RGB5_A1 */
+      1351, /* GL_RGBA8 */
+      1323, /* GL_RGB10_A2 */
+      1341, /* GL_RGBA12 */
+      1343, /* GL_RGBA16 */
+      1696, /* GL_TEXTURE_RED_SIZE */
+      1668, /* GL_TEXTURE_GREEN_SIZE */
+      1607, /* GL_TEXTURE_BLUE_SIZE */
+      1594, /* GL_TEXTURE_ALPHA_SIZE */
+      1681, /* GL_TEXTURE_LUMINANCE_SIZE */
+      1672, /* GL_TEXTURE_INTENSITY_SIZE */
+      1316, /* GL_REPLACE_EXT */
+      1251, /* GL_PROXY_TEXTURE_1D */
+      1254, /* GL_PROXY_TEXTURE_2D */
+      1701, /* GL_TEXTURE_TOO_LARGE_EXT */
+      1693, /* GL_TEXTURE_PRIORITY */
+      1698, /* GL_TEXTURE_RESIDENT */
+      1597, /* GL_TEXTURE_BINDING_1D */
+      1599, /* GL_TEXTURE_BINDING_2D */
+      1601, /* GL_TEXTURE_BINDING_3D */
+      1077, /* GL_PACK_SKIP_IMAGES */
+      1073, /* GL_PACK_IMAGE_HEIGHT */
+      1730, /* GL_UNPACK_SKIP_IMAGES */
+      1727, /* GL_UNPACK_IMAGE_HEIGHT */
+      1593, /* GL_TEXTURE_3D */
+      1257, /* GL_PROXY_TEXTURE_3D */
+      1655, /* GL_TEXTURE_DEPTH */
+      1704, /* GL_TEXTURE_WRAP_R */
+       842, /* GL_MAX_3D_TEXTURE_SIZE */
+      1762, /* GL_VERTEX_ARRAY */
+      1008, /* GL_NORMAL_ARRAY */
        145, /* GL_COLOR_ARRAY */
-       615, /* GL_INDEX_ARRAY */
-      1622, /* GL_TEXTURE_COORD_ARRAY */
-       449, /* GL_EDGE_FLAG_ARRAY */
-      1755, /* GL_VERTEX_ARRAY_SIZE */
-      1757, /* GL_VERTEX_ARRAY_TYPE */
-      1756, /* GL_VERTEX_ARRAY_STRIDE */
-      1003, /* GL_NORMAL_ARRAY_TYPE */
-      1002, /* GL_NORMAL_ARRAY_STRIDE */
+       618, /* GL_INDEX_ARRAY */
+      1634, /* GL_TEXTURE_COORD_ARRAY */
+       451, /* GL_EDGE_FLAG_ARRAY */
+      1768, /* GL_VERTEX_ARRAY_SIZE */
+      1770, /* GL_VERTEX_ARRAY_TYPE */
+      1769, /* GL_VERTEX_ARRAY_STRIDE */
+      1013, /* GL_NORMAL_ARRAY_TYPE */
+      1012, /* GL_NORMAL_ARRAY_STRIDE */
        149, /* GL_COLOR_ARRAY_SIZE */
        151, /* GL_COLOR_ARRAY_TYPE */
        150, /* GL_COLOR_ARRAY_STRIDE */
-       620, /* GL_INDEX_ARRAY_TYPE */
-       619, /* GL_INDEX_ARRAY_STRIDE */
-      1626, /* GL_TEXTURE_COORD_ARRAY_SIZE */
-      1628, /* GL_TEXTURE_COORD_ARRAY_TYPE */
-      1627, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
-       453, /* GL_EDGE_FLAG_ARRAY_STRIDE */
-      1754, /* GL_VERTEX_ARRAY_POINTER */
-      1001, /* GL_NORMAL_ARRAY_POINTER */
+       623, /* GL_INDEX_ARRAY_TYPE */
+       622, /* GL_INDEX_ARRAY_STRIDE */
+      1638, /* GL_TEXTURE_COORD_ARRAY_SIZE */
+      1640, /* GL_TEXTURE_COORD_ARRAY_TYPE */
+      1639, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
+       455, /* GL_EDGE_FLAG_ARRAY_STRIDE */
+      1767, /* GL_VERTEX_ARRAY_POINTER */
+      1011, /* GL_NORMAL_ARRAY_POINTER */
        148, /* GL_COLOR_ARRAY_POINTER */
-       618, /* GL_INDEX_ARRAY_POINTER */
-      1625, /* GL_TEXTURE_COORD_ARRAY_POINTER */
-       452, /* GL_EDGE_FLAG_ARRAY_POINTER */
-       977, /* GL_MULTISAMPLE */
-      1363, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
-      1365, /* GL_SAMPLE_ALPHA_TO_ONE */
-      1370, /* GL_SAMPLE_COVERAGE */
-      1367, /* GL_SAMPLE_BUFFERS */
-      1358, /* GL_SAMPLES */
-      1374, /* GL_SAMPLE_COVERAGE_VALUE */
-      1372, /* GL_SAMPLE_COVERAGE_INVERT */
+       621, /* GL_INDEX_ARRAY_POINTER */
+      1637, /* GL_TEXTURE_COORD_ARRAY_POINTER */
+       454, /* GL_EDGE_FLAG_ARRAY_POINTER */
+       987, /* GL_MULTISAMPLE */
+      1375, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
+      1377, /* GL_SAMPLE_ALPHA_TO_ONE */
+      1382, /* GL_SAMPLE_COVERAGE */
+      1379, /* GL_SAMPLE_BUFFERS */
+      1370, /* GL_SAMPLES */
+      1386, /* GL_SAMPLE_COVERAGE_VALUE */
+      1384, /* GL_SAMPLE_COVERAGE_INVERT */
        192, /* GL_COLOR_MATRIX */
        194, /* GL_COLOR_MATRIX_STACK_DEPTH */
-       840, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
-      1172, /* GL_POST_COLOR_MATRIX_RED_SCALE */
-      1168, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
-      1163, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
-      1159, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
-      1170, /* GL_POST_COLOR_MATRIX_RED_BIAS */
-      1166, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
-      1161, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
-      1157, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
-      1605, /* GL_TEXTURE_COLOR_TABLE_SGI */
-      1247, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
-      1607, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
+       850, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
+      1182, /* GL_POST_COLOR_MATRIX_RED_SCALE */
+      1178, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
+      1173, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
+      1169, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
+      1180, /* GL_POST_COLOR_MATRIX_RED_BIAS */
+      1176, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
+      1171, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
+      1167, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
+      1617, /* GL_TEXTURE_COLOR_TABLE_SGI */
+      1258, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
+      1619, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
         79, /* GL_BLEND_DST_RGB */
         88, /* GL_BLEND_SRC_RGB */
         78, /* GL_BLEND_DST_ALPHA */
         87, /* GL_BLEND_SRC_ALPHA */
        198, /* GL_COLOR_TABLE */
-      1182, /* GL_POST_CONVOLUTION_COLOR_TABLE */
-      1165, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
-      1235, /* GL_PROXY_COLOR_TABLE */
-      1239, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
-      1238, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
+      1192, /* GL_POST_CONVOLUTION_COLOR_TABLE */
+      1175, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
+      1246, /* GL_PROXY_COLOR_TABLE */
+      1250, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
+      1249, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
        222, /* GL_COLOR_TABLE_SCALE */
        202, /* GL_COLOR_TABLE_BIAS */
        207, /* GL_COLOR_TABLE_FORMAT */
@@ -4295,667 +4321,667 @@ static const unsigned reduced_enums[1319] =
        213, /* GL_COLOR_TABLE_INTENSITY_SIZE */
         70, /* GL_BGR */
         71, /* GL_BGRA */
-       854, /* GL_MAX_ELEMENTS_VERTICES */
-       853, /* GL_MAX_ELEMENTS_INDICES */
-      1659, /* GL_TEXTURE_INDEX_SIZE_EXT */
+       864, /* GL_MAX_ELEMENTS_VERTICES */
+       863, /* GL_MAX_ELEMENTS_INDICES */
+      1671, /* GL_TEXTURE_INDEX_SIZE_EXT */
        142, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */
-      1129, /* GL_POINT_SIZE_MIN */
-      1125, /* GL_POINT_SIZE_MAX */
-      1119, /* GL_POINT_FADE_THRESHOLD_SIZE */
-      1115, /* GL_POINT_DISTANCE_ATTENUATION */
+      1139, /* GL_POINT_SIZE_MIN */
+      1135, /* GL_POINT_SIZE_MAX */
+      1129, /* GL_POINT_FADE_THRESHOLD_SIZE */
+      1125, /* GL_POINT_DISTANCE_ATTENUATION */
        124, /* GL_CLAMP_TO_BORDER */
        127, /* GL_CLAMP_TO_EDGE */
-      1680, /* GL_TEXTURE_MIN_LOD */
-      1678, /* GL_TEXTURE_MAX_LOD */
-      1584, /* GL_TEXTURE_BASE_LEVEL */
-      1677, /* GL_TEXTURE_MAX_LEVEL */
-       608, /* GL_IGNORE_BORDER_HP */
+      1692, /* GL_TEXTURE_MIN_LOD */
+      1690, /* GL_TEXTURE_MAX_LOD */
+      1596, /* GL_TEXTURE_BASE_LEVEL */
+      1689, /* GL_TEXTURE_MAX_LEVEL */
+       611, /* GL_IGNORE_BORDER_HP */
        272, /* GL_CONSTANT_BORDER_HP */
-      1305, /* GL_REPLICATE_BORDER_HP */
+      1317, /* GL_REPLICATE_BORDER_HP */
        278, /* GL_CONVOLUTION_BORDER_COLOR */
-      1026, /* GL_OCCLUSION_TEST_HP */
-      1027, /* GL_OCCLUSION_TEST_RESULT_HP */
-       679, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
-      1599, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
-      1601, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
-      1603, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
-      1604, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-      1602, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
-      1600, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
-       836, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
-       837, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-      1192, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
-      1194, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
-      1191, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
-      1193, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
-      1667, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
-      1668, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
-      1666, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
-       578, /* GL_GENERATE_MIPMAP */
-       579, /* GL_GENERATE_MIPMAP_HINT */
-       520, /* GL_FOG_OFFSET_SGIX */
-       521, /* GL_FOG_OFFSET_VALUE_SGIX */
-      1613, /* GL_TEXTURE_COMPARE_SGIX */
-      1612, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
-      1663, /* GL_TEXTURE_LEQUAL_R_SGIX */
-      1655, /* GL_TEXTURE_GEQUAL_R_SGIX */
-       352, /* GL_DEPTH_COMPONENT16 */
-       355, /* GL_DEPTH_COMPONENT24 */
-       358, /* GL_DEPTH_COMPONENT32 */
-       300, /* GL_CULL_VERTEX_EXT */
-       302, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
-       301, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
-      1811, /* GL_WRAP_BORDER_SUN */
-      1606, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
-       672, /* GL_LIGHT_MODEL_COLOR_CONTROL */
-      1403, /* GL_SINGLE_COLOR */
-      1390, /* GL_SEPARATE_SPECULAR_COLOR */
-      1399, /* GL_SHARED_TEXTURE_PALETTE_EXT */
-       531, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
-       532, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
-       539, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
-       534, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
-       530, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
-       529, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
-       533, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
-       540, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
-       551, /* GL_FRAMEBUFFER_DEFAULT */
-       564, /* GL_FRAMEBUFFER_UNDEFINED */
-       365, /* GL_DEPTH_STENCIL_ATTACHMENT */
-       614, /* GL_INDEX */
-      1723, /* GL_UNSIGNED_BYTE_2_3_3_REV */
-      1738, /* GL_UNSIGNED_SHORT_5_6_5 */
-      1739, /* GL_UNSIGNED_SHORT_5_6_5_REV */
-      1736, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
-      1734, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
-      1731, /* GL_UNSIGNED_INT_8_8_8_8_REV */
-      1729, /* GL_UNSIGNED_INT_2_10_10_10_REV */
-      1675, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
-      1676, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
-      1674, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
-       928, /* GL_MIRRORED_REPEAT */
-      1346, /* GL_RGB_S3TC */
-      1321, /* GL_RGB4_S3TC */
-      1344, /* GL_RGBA_S3TC */
-      1338, /* GL_RGBA4_S3TC */
-      1342, /* GL_RGBA_DXT5_S3TC */
-      1336, /* GL_RGBA4_DXT5_S3TC */
+      1036, /* GL_OCCLUSION_TEST_HP */
+      1037, /* GL_OCCLUSION_TEST_RESULT_HP */
+       683, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
+      1611, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
+      1613, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
+      1615, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
+      1616, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+      1614, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
+      1612, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
+       846, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
+       847, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+      1202, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
+      1204, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
+      1201, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
+      1203, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
+      1679, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
+      1680, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
+      1678, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
+       581, /* GL_GENERATE_MIPMAP */
+       582, /* GL_GENERATE_MIPMAP_HINT */
+       523, /* GL_FOG_OFFSET_SGIX */
+       524, /* GL_FOG_OFFSET_VALUE_SGIX */
+      1625, /* GL_TEXTURE_COMPARE_SGIX */
+      1624, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
+      1675, /* GL_TEXTURE_LEQUAL_R_SGIX */
+      1667, /* GL_TEXTURE_GEQUAL_R_SGIX */
+       354, /* GL_DEPTH_COMPONENT16 */
+       357, /* GL_DEPTH_COMPONENT24 */
+       360, /* GL_DEPTH_COMPONENT32 */
+       302, /* GL_CULL_VERTEX_EXT */
+       304, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
+       303, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
+      1824, /* GL_WRAP_BORDER_SUN */
+      1618, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
+       676, /* GL_LIGHT_MODEL_COLOR_CONTROL */
+      1415, /* GL_SINGLE_COLOR */
+      1402, /* GL_SEPARATE_SPECULAR_COLOR */
+      1411, /* GL_SHARED_TEXTURE_PALETTE_EXT */
+       534, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
+       535, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
+       542, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
+       537, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
+       533, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
+       532, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
+       536, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
+       543, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
+       554, /* GL_FRAMEBUFFER_DEFAULT */
+       567, /* GL_FRAMEBUFFER_UNDEFINED */
+       367, /* GL_DEPTH_STENCIL_ATTACHMENT */
+       617, /* GL_INDEX */
+      1735, /* GL_UNSIGNED_BYTE_2_3_3_REV */
+      1750, /* GL_UNSIGNED_SHORT_5_6_5 */
+      1751, /* GL_UNSIGNED_SHORT_5_6_5_REV */
+      1748, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
+      1746, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
+      1743, /* GL_UNSIGNED_INT_8_8_8_8_REV */
+      1741, /* GL_UNSIGNED_INT_2_10_10_10_REV */
+      1687, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
+      1688, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
+      1686, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
+       938, /* GL_MIRRORED_REPEAT */
+      1358, /* GL_RGB_S3TC */
+      1333, /* GL_RGB4_S3TC */
+      1356, /* GL_RGBA_S3TC */
+      1350, /* GL_RGBA4_S3TC */
+      1354, /* GL_RGBA_DXT5_S3TC */
+      1348, /* GL_RGBA4_DXT5_S3TC */
        261, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
        256, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
        257, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
        258, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
-       989, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
-       988, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
-       680, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
-       507, /* GL_FOG_COORDINATE_SOURCE */
-       499, /* GL_FOG_COORD */
-       523, /* GL_FRAGMENT_DEPTH */
-       306, /* GL_CURRENT_FOG_COORD */
-       506, /* GL_FOG_COORDINATE_ARRAY_TYPE */
-       505, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
-       504, /* GL_FOG_COORDINATE_ARRAY_POINTER */
-       501, /* GL_FOG_COORDINATE_ARRAY */
+       999, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
+       998, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
+       684, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
+       510, /* GL_FOG_COORDINATE_SOURCE */
+       502, /* GL_FOG_COORD */
+       526, /* GL_FRAGMENT_DEPTH */
+       308, /* GL_CURRENT_FOG_COORD */
+       509, /* GL_FOG_COORDINATE_ARRAY_TYPE */
+       508, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
+       507, /* GL_FOG_COORDINATE_ARRAY_POINTER */
+       504, /* GL_FOG_COORDINATE_ARRAY */
        196, /* GL_COLOR_SUM */
-       326, /* GL_CURRENT_SECONDARY_COLOR */
-      1383, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
-      1385, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
-      1384, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
-      1382, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
-      1379, /* GL_SECONDARY_COLOR_ARRAY */
-       324, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
+       328, /* GL_CURRENT_SECONDARY_COLOR */
+      1395, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
+      1397, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
+      1396, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
+      1394, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
+      1391, /* GL_SECONDARY_COLOR_ARRAY */
+       326, /* GL_CURRENT_RASTER_SECONDARY_COLOR */
         28, /* GL_ALIASED_POINT_SIZE_RANGE */
         27, /* GL_ALIASED_LINE_WIDTH_RANGE */
-      1513, /* GL_TEXTURE0 */
-      1515, /* GL_TEXTURE1 */
-      1537, /* GL_TEXTURE2 */
-      1559, /* GL_TEXTURE3 */
-      1565, /* GL_TEXTURE4 */
-      1567, /* GL_TEXTURE5 */
-      1569, /* GL_TEXTURE6 */
-      1571, /* GL_TEXTURE7 */
-      1573, /* GL_TEXTURE8 */
-      1575, /* GL_TEXTURE9 */
-      1516, /* GL_TEXTURE10 */
-      1518, /* GL_TEXTURE11 */
-      1520, /* GL_TEXTURE12 */
-      1522, /* GL_TEXTURE13 */
-      1524, /* GL_TEXTURE14 */
-      1526, /* GL_TEXTURE15 */
-      1528, /* GL_TEXTURE16 */
-      1530, /* GL_TEXTURE17 */
-      1532, /* GL_TEXTURE18 */
-      1534, /* GL_TEXTURE19 */
-      1538, /* GL_TEXTURE20 */
-      1540, /* GL_TEXTURE21 */
-      1542, /* GL_TEXTURE22 */
-      1544, /* GL_TEXTURE23 */
-      1546, /* GL_TEXTURE24 */
-      1548, /* GL_TEXTURE25 */
-      1550, /* GL_TEXTURE26 */
-      1552, /* GL_TEXTURE27 */
-      1554, /* GL_TEXTURE28 */
-      1556, /* GL_TEXTURE29 */
-      1560, /* GL_TEXTURE30 */
-      1562, /* GL_TEXTURE31 */
+      1525, /* GL_TEXTURE0 */
+      1527, /* GL_TEXTURE1 */
+      1549, /* GL_TEXTURE2 */
+      1571, /* GL_TEXTURE3 */
+      1577, /* GL_TEXTURE4 */
+      1579, /* GL_TEXTURE5 */
+      1581, /* GL_TEXTURE6 */
+      1583, /* GL_TEXTURE7 */
+      1585, /* GL_TEXTURE8 */
+      1587, /* GL_TEXTURE9 */
+      1528, /* GL_TEXTURE10 */
+      1530, /* GL_TEXTURE11 */
+      1532, /* GL_TEXTURE12 */
+      1534, /* GL_TEXTURE13 */
+      1536, /* GL_TEXTURE14 */
+      1538, /* GL_TEXTURE15 */
+      1540, /* GL_TEXTURE16 */
+      1542, /* GL_TEXTURE17 */
+      1544, /* GL_TEXTURE18 */
+      1546, /* GL_TEXTURE19 */
+      1550, /* GL_TEXTURE20 */
+      1552, /* GL_TEXTURE21 */
+      1554, /* GL_TEXTURE22 */
+      1556, /* GL_TEXTURE23 */
+      1558, /* GL_TEXTURE24 */
+      1560, /* GL_TEXTURE25 */
+      1562, /* GL_TEXTURE26 */
+      1564, /* GL_TEXTURE27 */
+      1566, /* GL_TEXTURE28 */
+      1568, /* GL_TEXTURE29 */
+      1572, /* GL_TEXTURE30 */
+      1574, /* GL_TEXTURE31 */
         18, /* GL_ACTIVE_TEXTURE */
        130, /* GL_CLIENT_ACTIVE_TEXTURE */
-       906, /* GL_MAX_TEXTURE_UNITS */
-      1702, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
-      1705, /* GL_TRANSPOSE_PROJECTION_MATRIX */
-      1707, /* GL_TRANSPOSE_TEXTURE_MATRIX */
-      1699, /* GL_TRANSPOSE_COLOR_MATRIX */
-      1501, /* GL_SUBTRACT */
-       894, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
+       916, /* GL_MAX_TEXTURE_UNITS */
+      1714, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
+      1717, /* GL_TRANSPOSE_PROJECTION_MATRIX */
+      1719, /* GL_TRANSPOSE_TEXTURE_MATRIX */
+      1711, /* GL_TRANSPOSE_COLOR_MATRIX */
+      1513, /* GL_SUBTRACT */
+       904, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
        244, /* GL_COMPRESSED_ALPHA */
        248, /* GL_COMPRESSED_LUMINANCE */
        249, /* GL_COMPRESSED_LUMINANCE_ALPHA */
        246, /* GL_COMPRESSED_INTENSITY */
        252, /* GL_COMPRESSED_RGB */
        253, /* GL_COMPRESSED_RGBA */
-      1620, /* GL_TEXTURE_COMPRESSION_HINT */
-      1682, /* GL_TEXTURE_RECTANGLE_ARB */
-      1592, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
-      1250, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
-       892, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
-       364, /* GL_DEPTH_STENCIL */
-      1727, /* GL_UNSIGNED_INT_24_8 */
-       902, /* GL_MAX_TEXTURE_LOD_BIAS */
-      1673, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
-       903, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
-      1649, /* GL_TEXTURE_FILTER_CONTROL */
-      1664, /* GL_TEXTURE_LOD_BIAS */
+      1632, /* GL_TEXTURE_COMPRESSION_HINT */
+      1694, /* GL_TEXTURE_RECTANGLE_ARB */
+      1604, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
+      1261, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
+       902, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
+       366, /* GL_DEPTH_STENCIL */
+      1739, /* GL_UNSIGNED_INT_24_8 */
+       912, /* GL_MAX_TEXTURE_LOD_BIAS */
+      1685, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
+       913, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
+      1661, /* GL_TEXTURE_FILTER_CONTROL */
+      1676, /* GL_TEXTURE_LOD_BIAS */
        229, /* GL_COMBINE4 */
-       896, /* GL_MAX_SHININESS_NV */
-       897, /* GL_MAX_SPOT_EXPONENT_NV */
-       612, /* GL_INCR_WRAP */
-       337, /* GL_DECR_WRAP */
-       948, /* GL_MODELVIEW1_ARB */
-      1004, /* GL_NORMAL_MAP */
-      1280, /* GL_REFLECTION_MAP */
-      1629, /* GL_TEXTURE_CUBE_MAP */
-      1590, /* GL_TEXTURE_BINDING_CUBE_MAP */
-      1637, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
-      1631, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
-      1639, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
-      1633, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
-      1641, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
-      1635, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
-      1248, /* GL_PROXY_TEXTURE_CUBE_MAP */
-       848, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
-       983, /* GL_MULTISAMPLE_FILTER_HINT_NV */
-       515, /* GL_FOG_DISTANCE_MODE_NV */
-       468, /* GL_EYE_RADIAL_NV */
-       467, /* GL_EYE_PLANE_ABSOLUTE_NV */
+       906, /* GL_MAX_SHININESS_NV */
+       907, /* GL_MAX_SPOT_EXPONENT_NV */
+       615, /* GL_INCR_WRAP */
+       339, /* GL_DECR_WRAP */
+       958, /* GL_MODELVIEW1_ARB */
+      1014, /* GL_NORMAL_MAP */
+      1292, /* GL_REFLECTION_MAP */
+      1641, /* GL_TEXTURE_CUBE_MAP */
+      1602, /* GL_TEXTURE_BINDING_CUBE_MAP */
+      1649, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
+      1643, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
+      1651, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
+      1645, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
+      1653, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
+      1647, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
+      1259, /* GL_PROXY_TEXTURE_CUBE_MAP */
+       858, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
+       993, /* GL_MULTISAMPLE_FILTER_HINT_NV */
+       518, /* GL_FOG_DISTANCE_MODE_NV */
+       470, /* GL_EYE_RADIAL_NV */
+       469, /* GL_EYE_PLANE_ABSOLUTE_NV */
        228, /* GL_COMBINE */
        235, /* GL_COMBINE_RGB */
        230, /* GL_COMBINE_ALPHA */
-      1347, /* GL_RGB_SCALE */
+      1359, /* GL_RGB_SCALE */
         24, /* GL_ADD_SIGNED */
-       640, /* GL_INTERPOLATE */
+       643, /* GL_INTERPOLATE */
        267, /* GL_CONSTANT */
-      1198, /* GL_PRIMARY_COLOR */
-      1195, /* GL_PREVIOUS */
-      1418, /* GL_SOURCE0_RGB */
-      1424, /* GL_SOURCE1_RGB */
-      1430, /* GL_SOURCE2_RGB */
-      1434, /* GL_SOURCE3_RGB_NV */
-      1415, /* GL_SOURCE0_ALPHA */
-      1421, /* GL_SOURCE1_ALPHA */
-      1427, /* GL_SOURCE2_ALPHA */
-      1433, /* GL_SOURCE3_ALPHA_NV */
-      1040, /* GL_OPERAND0_RGB */
-      1046, /* GL_OPERAND1_RGB */
-      1052, /* GL_OPERAND2_RGB */
-      1056, /* GL_OPERAND3_RGB_NV */
-      1037, /* GL_OPERAND0_ALPHA */
-      1043, /* GL_OPERAND1_ALPHA */
-      1049, /* GL_OPERAND2_ALPHA */
-      1055, /* GL_OPERAND3_ALPHA_NV */
-      1751, /* GL_VERTEX_ARRAY_BINDING_APPLE */
-      1815, /* GL_YCBCR_422_APPLE */
-      1740, /* GL_UNSIGNED_SHORT_8_8_APPLE */
-      1742, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
-      1405, /* GL_SLICE_ACCUM_SUN */
-      1255, /* GL_QUAD_MESH_SUN */
-      1711, /* GL_TRIANGLE_MESH_SUN */
-      1789, /* GL_VERTEX_PROGRAM_ARB */
-      1800, /* GL_VERTEX_STATE_PROGRAM_NV */
-      1776, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
-      1782, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
-      1784, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
-      1786, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
-       328, /* GL_CURRENT_VERTEX_ATTRIB */
-      1211, /* GL_PROGRAM_LENGTH_ARB */
-      1225, /* GL_PROGRAM_STRING_ARB */
-       970, /* GL_MODELVIEW_PROJECTION_NV */
-       607, /* GL_IDENTITY_NV */
-       654, /* GL_INVERSE_NV */
-      1704, /* GL_TRANSPOSE_NV */
-       655, /* GL_INVERSE_TRANSPOSE_NV */
-       878, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
-       877, /* GL_MAX_PROGRAM_MATRICES_ARB */
-       785, /* GL_MATRIX0_NV */
-       797, /* GL_MATRIX1_NV */
-       809, /* GL_MATRIX2_NV */
-       813, /* GL_MATRIX3_NV */
-       815, /* GL_MATRIX4_NV */
-       817, /* GL_MATRIX5_NV */
-       819, /* GL_MATRIX6_NV */
-       821, /* GL_MATRIX7_NV */
-       312, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
-       309, /* GL_CURRENT_MATRIX_ARB */
-      1792, /* GL_VERTEX_PROGRAM_POINT_SIZE */
-      1795, /* GL_VERTEX_PROGRAM_TWO_SIDE */
-      1223, /* GL_PROGRAM_PARAMETER_NV */
-      1780, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
-      1227, /* GL_PROGRAM_TARGET_NV */
-      1224, /* GL_PROGRAM_RESIDENT_NV */
-      1696, /* GL_TRACK_MATRIX_NV */
-      1697, /* GL_TRACK_MATRIX_TRANSFORM_NV */
-      1790, /* GL_VERTEX_PROGRAM_BINDING_NV */
-      1205, /* GL_PROGRAM_ERROR_POSITION_ARB */
-       349, /* GL_DEPTH_CLAMP_NV */
-      1758, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
-      1765, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
-      1766, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
-      1767, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
-      1768, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
-      1769, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
-      1770, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
-      1771, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
-      1772, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
-      1773, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
-      1759, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
-      1760, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
-      1761, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
-      1762, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
-      1763, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
-      1764, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
-       739, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
-       746, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
-       747, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
-       748, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
-       749, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
-       750, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
-       751, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
-       752, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
-       753, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
-       754, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
-       740, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
-       741, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
-       742, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
-       743, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
-       744, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
-       745, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
-       766, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
-       773, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
-       774, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
-       775, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
-       776, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
-       777, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
-       778, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
-      1204, /* GL_PROGRAM_BINDING_ARB */
-       780, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
-       781, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
-       767, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
-       768, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
-       769, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
-       770, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
-       771, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
-       772, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
-      1618, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
-      1615, /* GL_TEXTURE_COMPRESSED */
-      1009, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
+      1208, /* GL_PRIMARY_COLOR */
+      1205, /* GL_PREVIOUS */
+      1430, /* GL_SOURCE0_RGB */
+      1436, /* GL_SOURCE1_RGB */
+      1442, /* GL_SOURCE2_RGB */
+      1446, /* GL_SOURCE3_RGB_NV */
+      1427, /* GL_SOURCE0_ALPHA */
+      1433, /* GL_SOURCE1_ALPHA */
+      1439, /* GL_SOURCE2_ALPHA */
+      1445, /* GL_SOURCE3_ALPHA_NV */
+      1050, /* GL_OPERAND0_RGB */
+      1056, /* GL_OPERAND1_RGB */
+      1062, /* GL_OPERAND2_RGB */
+      1066, /* GL_OPERAND3_RGB_NV */
+      1047, /* GL_OPERAND0_ALPHA */
+      1053, /* GL_OPERAND1_ALPHA */
+      1059, /* GL_OPERAND2_ALPHA */
+      1065, /* GL_OPERAND3_ALPHA_NV */
+      1763, /* GL_VERTEX_ARRAY_BINDING */
+      1828, /* GL_YCBCR_422_APPLE */
+      1752, /* GL_UNSIGNED_SHORT_8_8_APPLE */
+      1754, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
+      1417, /* GL_SLICE_ACCUM_SUN */
+      1267, /* GL_QUAD_MESH_SUN */
+      1723, /* GL_TRIANGLE_MESH_SUN */
+      1802, /* GL_VERTEX_PROGRAM_ARB */
+      1813, /* GL_VERTEX_STATE_PROGRAM_NV */
+      1789, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
+      1795, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
+      1797, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
+      1799, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
+       330, /* GL_CURRENT_VERTEX_ATTRIB */
+      1221, /* GL_PROGRAM_LENGTH_ARB */
+      1235, /* GL_PROGRAM_STRING_ARB */
+       980, /* GL_MODELVIEW_PROJECTION_NV */
+       610, /* GL_IDENTITY_NV */
+       657, /* GL_INVERSE_NV */
+      1716, /* GL_TRANSPOSE_NV */
+       658, /* GL_INVERSE_TRANSPOSE_NV */
+       888, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
+       887, /* GL_MAX_PROGRAM_MATRICES_ARB */
+       795, /* GL_MATRIX0_NV */
+       807, /* GL_MATRIX1_NV */
+       819, /* GL_MATRIX2_NV */
+       823, /* GL_MATRIX3_NV */
+       825, /* GL_MATRIX4_NV */
+       827, /* GL_MATRIX5_NV */
+       829, /* GL_MATRIX6_NV */
+       831, /* GL_MATRIX7_NV */
+       314, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
+       311, /* GL_CURRENT_MATRIX_ARB */
+      1805, /* GL_VERTEX_PROGRAM_POINT_SIZE */
+      1808, /* GL_VERTEX_PROGRAM_TWO_SIDE */
+      1233, /* GL_PROGRAM_PARAMETER_NV */
+      1793, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
+      1237, /* GL_PROGRAM_TARGET_NV */
+      1234, /* GL_PROGRAM_RESIDENT_NV */
+      1708, /* GL_TRACK_MATRIX_NV */
+      1709, /* GL_TRACK_MATRIX_TRANSFORM_NV */
+      1803, /* GL_VERTEX_PROGRAM_BINDING_NV */
+      1215, /* GL_PROGRAM_ERROR_POSITION_ARB */
+       351, /* GL_DEPTH_CLAMP_NV */
+      1771, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
+      1778, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
+      1779, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
+      1780, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
+      1781, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
+      1782, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
+      1783, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
+      1784, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
+      1785, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
+      1786, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
+      1772, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
+      1773, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
+      1774, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
+      1775, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
+      1776, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
+      1777, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
+       743, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
+       750, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
+       751, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
+       752, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
+       753, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
+       754, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
+       755, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
+       756, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
+       757, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
+       758, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
+       744, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
+       745, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
+       746, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
+       747, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
+       748, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
+       749, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
+       770, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
+       777, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
+       778, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
+       779, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
+       780, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
+       781, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
+       782, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
+      1214, /* GL_PROGRAM_BINDING_ARB */
+       784, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
+       785, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
+       771, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
+       772, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
+       773, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
+       774, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
+       775, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
+       776, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
+      1630, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
+      1627, /* GL_TEXTURE_COMPRESSED */
+      1019, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
        266, /* GL_COMPRESSED_TEXTURE_FORMATS */
-       918, /* GL_MAX_VERTEX_UNITS_ARB */
+       928, /* GL_MAX_VERTEX_UNITS_ARB */
         22, /* GL_ACTIVE_VERTEX_UNITS_ARB */
-      1810, /* GL_WEIGHT_SUM_UNITY_ARB */
-      1788, /* GL_VERTEX_BLEND_ARB */
-       330, /* GL_CURRENT_WEIGHT_ARB */
-      1809, /* GL_WEIGHT_ARRAY_TYPE_ARB */
-      1808, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
-      1807, /* GL_WEIGHT_ARRAY_SIZE_ARB */
-      1806, /* GL_WEIGHT_ARRAY_POINTER_ARB */
-      1803, /* GL_WEIGHT_ARRAY_ARB */
-       377, /* GL_DOT3_RGB */
-       378, /* GL_DOT3_RGBA */
+      1823, /* GL_WEIGHT_SUM_UNITY_ARB */
+      1801, /* GL_VERTEX_BLEND_ARB */
+       332, /* GL_CURRENT_WEIGHT_ARB */
+      1822, /* GL_WEIGHT_ARRAY_TYPE_ARB */
+      1821, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
+      1820, /* GL_WEIGHT_ARRAY_SIZE_ARB */
+      1819, /* GL_WEIGHT_ARRAY_POINTER_ARB */
+      1816, /* GL_WEIGHT_ARRAY_ARB */
+       379, /* GL_DOT3_RGB */
+       380, /* GL_DOT3_RGBA */
        260, /* GL_COMPRESSED_RGB_FXT1_3DFX */
        255, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
-       978, /* GL_MULTISAMPLE_3DFX */
-      1368, /* GL_SAMPLE_BUFFERS_3DFX */
-      1359, /* GL_SAMPLES_3DFX */
-       959, /* GL_MODELVIEW2_ARB */
-       962, /* GL_MODELVIEW3_ARB */
-       963, /* GL_MODELVIEW4_ARB */
-       964, /* GL_MODELVIEW5_ARB */
-       965, /* GL_MODELVIEW6_ARB */
-       966, /* GL_MODELVIEW7_ARB */
-       967, /* GL_MODELVIEW8_ARB */
-       968, /* GL_MODELVIEW9_ARB */
-       938, /* GL_MODELVIEW10_ARB */
-       939, /* GL_MODELVIEW11_ARB */
-       940, /* GL_MODELVIEW12_ARB */
-       941, /* GL_MODELVIEW13_ARB */
-       942, /* GL_MODELVIEW14_ARB */
-       943, /* GL_MODELVIEW15_ARB */
-       944, /* GL_MODELVIEW16_ARB */
-       945, /* GL_MODELVIEW17_ARB */
-       946, /* GL_MODELVIEW18_ARB */
-       947, /* GL_MODELVIEW19_ARB */
-       949, /* GL_MODELVIEW20_ARB */
-       950, /* GL_MODELVIEW21_ARB */
-       951, /* GL_MODELVIEW22_ARB */
-       952, /* GL_MODELVIEW23_ARB */
-       953, /* GL_MODELVIEW24_ARB */
-       954, /* GL_MODELVIEW25_ARB */
-       955, /* GL_MODELVIEW26_ARB */
-       956, /* GL_MODELVIEW27_ARB */
-       957, /* GL_MODELVIEW28_ARB */
-       958, /* GL_MODELVIEW29_ARB */
-       960, /* GL_MODELVIEW30_ARB */
-       961, /* GL_MODELVIEW31_ARB */
-       382, /* GL_DOT3_RGB_EXT */
-       380, /* GL_DOT3_RGBA_EXT */
-       932, /* GL_MIRROR_CLAMP_EXT */
-       935, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
-       973, /* GL_MODULATE_ADD_ATI */
-       974, /* GL_MODULATE_SIGNED_ADD_ATI */
-       975, /* GL_MODULATE_SUBTRACT_ATI */
-      1816, /* GL_YCBCR_MESA */
-      1064, /* GL_PACK_INVERT_MESA */
-       333, /* GL_DEBUG_OBJECT_MESA */
-       334, /* GL_DEBUG_PRINT_MESA */
-       332, /* GL_DEBUG_ASSERT_MESA */
+       988, /* GL_MULTISAMPLE_3DFX */
+      1380, /* GL_SAMPLE_BUFFERS_3DFX */
+      1371, /* GL_SAMPLES_3DFX */
+       969, /* GL_MODELVIEW2_ARB */
+       972, /* GL_MODELVIEW3_ARB */
+       973, /* GL_MODELVIEW4_ARB */
+       974, /* GL_MODELVIEW5_ARB */
+       975, /* GL_MODELVIEW6_ARB */
+       976, /* GL_MODELVIEW7_ARB */
+       977, /* GL_MODELVIEW8_ARB */
+       978, /* GL_MODELVIEW9_ARB */
+       948, /* GL_MODELVIEW10_ARB */
+       949, /* GL_MODELVIEW11_ARB */
+       950, /* GL_MODELVIEW12_ARB */
+       951, /* GL_MODELVIEW13_ARB */
+       952, /* GL_MODELVIEW14_ARB */
+       953, /* GL_MODELVIEW15_ARB */
+       954, /* GL_MODELVIEW16_ARB */
+       955, /* GL_MODELVIEW17_ARB */
+       956, /* GL_MODELVIEW18_ARB */
+       957, /* GL_MODELVIEW19_ARB */
+       959, /* GL_MODELVIEW20_ARB */
+       960, /* GL_MODELVIEW21_ARB */
+       961, /* GL_MODELVIEW22_ARB */
+       962, /* GL_MODELVIEW23_ARB */
+       963, /* GL_MODELVIEW24_ARB */
+       964, /* GL_MODELVIEW25_ARB */
+       965, /* GL_MODELVIEW26_ARB */
+       966, /* GL_MODELVIEW27_ARB */
+       967, /* GL_MODELVIEW28_ARB */
+       968, /* GL_MODELVIEW29_ARB */
+       970, /* GL_MODELVIEW30_ARB */
+       971, /* GL_MODELVIEW31_ARB */
+       384, /* GL_DOT3_RGB_EXT */
+       382, /* GL_DOT3_RGBA_EXT */
+       942, /* GL_MIRROR_CLAMP_EXT */
+       945, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
+       983, /* GL_MODULATE_ADD_ATI */
+       984, /* GL_MODULATE_SIGNED_ADD_ATI */
+       985, /* GL_MODULATE_SUBTRACT_ATI */
+      1829, /* GL_YCBCR_MESA */
+      1074, /* GL_PACK_INVERT_MESA */
+       335, /* GL_DEBUG_OBJECT_MESA */
+       336, /* GL_DEBUG_PRINT_MESA */
+       334, /* GL_DEBUG_ASSERT_MESA */
        107, /* GL_BUFFER_SIZE */
        109, /* GL_BUFFER_USAGE */
        113, /* GL_BUMP_ROT_MATRIX_ATI */
        114, /* GL_BUMP_ROT_MATRIX_SIZE_ATI */
        112, /* GL_BUMP_NUM_TEX_UNITS_ATI */
        116, /* GL_BUMP_TEX_UNITS_ATI */
-       441, /* GL_DUDV_ATI */
-       440, /* GL_DU8DV8_ATI */
+       443, /* GL_DUDV_ATI */
+       442, /* GL_DU8DV8_ATI */
        111, /* GL_BUMP_ENVMAP_ATI */
        115, /* GL_BUMP_TARGET_ATI */
-      1466, /* GL_STENCIL_BACK_FUNC */
-      1464, /* GL_STENCIL_BACK_FAIL */
-      1468, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
-      1470, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
-       524, /* GL_FRAGMENT_PROGRAM_ARB */
-      1202, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
-      1230, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
-      1229, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
-      1214, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-      1220, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-      1219, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-       867, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
-       890, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
-       889, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
-       880, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-       886, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-       885, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-       850, /* GL_MAX_DRAW_BUFFERS */
-       386, /* GL_DRAW_BUFFER0 */
-       389, /* GL_DRAW_BUFFER1 */
-       410, /* GL_DRAW_BUFFER2 */
-       413, /* GL_DRAW_BUFFER3 */
-       416, /* GL_DRAW_BUFFER4 */
-       419, /* GL_DRAW_BUFFER5 */
-       422, /* GL_DRAW_BUFFER6 */
-       425, /* GL_DRAW_BUFFER7 */
-       428, /* GL_DRAW_BUFFER8 */
-       431, /* GL_DRAW_BUFFER9 */
-       390, /* GL_DRAW_BUFFER10 */
-       393, /* GL_DRAW_BUFFER11 */
-       396, /* GL_DRAW_BUFFER12 */
-       399, /* GL_DRAW_BUFFER13 */
-       402, /* GL_DRAW_BUFFER14 */
-       405, /* GL_DRAW_BUFFER15 */
+      1478, /* GL_STENCIL_BACK_FUNC */
+      1476, /* GL_STENCIL_BACK_FAIL */
+      1480, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
+      1482, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
+       527, /* GL_FRAGMENT_PROGRAM_ARB */
+      1212, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
+      1240, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
+      1239, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
+      1224, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+      1230, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+      1229, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+       877, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
+       900, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
+       899, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
+       890, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+       896, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+       895, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+       860, /* GL_MAX_DRAW_BUFFERS */
+       388, /* GL_DRAW_BUFFER0 */
+       391, /* GL_DRAW_BUFFER1 */
+       412, /* GL_DRAW_BUFFER2 */
+       415, /* GL_DRAW_BUFFER3 */
+       418, /* GL_DRAW_BUFFER4 */
+       421, /* GL_DRAW_BUFFER5 */
+       424, /* GL_DRAW_BUFFER6 */
+       427, /* GL_DRAW_BUFFER7 */
+       430, /* GL_DRAW_BUFFER8 */
+       433, /* GL_DRAW_BUFFER9 */
+       392, /* GL_DRAW_BUFFER10 */
+       395, /* GL_DRAW_BUFFER11 */
+       398, /* GL_DRAW_BUFFER12 */
+       401, /* GL_DRAW_BUFFER13 */
+       404, /* GL_DRAW_BUFFER14 */
+       407, /* GL_DRAW_BUFFER15 */
         81, /* GL_BLEND_EQUATION_ALPHA */
-       830, /* GL_MATRIX_PALETTE_ARB */
-       861, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
-       864, /* GL_MAX_PALETTE_MATRICES_ARB */
-       315, /* GL_CURRENT_PALETTE_MATRIX_ARB */
-       824, /* GL_MATRIX_INDEX_ARRAY_ARB */
-       310, /* GL_CURRENT_MATRIX_INDEX_ARB */
-       826, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
-       828, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
-       827, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
-       825, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
-      1644, /* GL_TEXTURE_DEPTH_SIZE */
-       370, /* GL_DEPTH_TEXTURE_MODE */
-      1610, /* GL_TEXTURE_COMPARE_MODE */
-      1608, /* GL_TEXTURE_COMPARE_FUNC */
+       840, /* GL_MATRIX_PALETTE_ARB */
+       871, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
+       874, /* GL_MAX_PALETTE_MATRICES_ARB */
+       317, /* GL_CURRENT_PALETTE_MATRIX_ARB */
+       834, /* GL_MATRIX_INDEX_ARRAY_ARB */
+       312, /* GL_CURRENT_MATRIX_INDEX_ARB */
+       836, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
+       838, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
+       837, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
+       835, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
+      1656, /* GL_TEXTURE_DEPTH_SIZE */
+       372, /* GL_DEPTH_TEXTURE_MODE */
+      1622, /* GL_TEXTURE_COMPARE_MODE */
+      1620, /* GL_TEXTURE_COMPARE_FUNC */
        239, /* GL_COMPARE_R_TO_TEXTURE */
-      1136, /* GL_POINT_SPRITE */
+      1146, /* GL_POINT_SPRITE */
        292, /* GL_COORD_REPLACE */
-      1140, /* GL_POINT_SPRITE_R_MODE_NV */
-      1257, /* GL_QUERY_COUNTER_BITS */
-       317, /* GL_CURRENT_QUERY */
-      1259, /* GL_QUERY_RESULT */
-      1261, /* GL_QUERY_RESULT_AVAILABLE */
-       912, /* GL_MAX_VERTEX_ATTRIBS */
-      1778, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
-       368, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
-       367, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
-       898, /* GL_MAX_TEXTURE_COORDS */
-       900, /* GL_MAX_TEXTURE_IMAGE_UNITS */
-      1207, /* GL_PROGRAM_ERROR_STRING_ARB */
-      1209, /* GL_PROGRAM_FORMAT_ASCII_ARB */
-      1208, /* GL_PROGRAM_FORMAT_ARB */
-      1690, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
-       347, /* GL_DEPTH_BOUNDS_TEST_EXT */
-       346, /* GL_DEPTH_BOUNDS_EXT */
+      1150, /* GL_POINT_SPRITE_R_MODE_NV */
+      1269, /* GL_QUERY_COUNTER_BITS */
+       319, /* GL_CURRENT_QUERY */
+      1271, /* GL_QUERY_RESULT */
+      1273, /* GL_QUERY_RESULT_AVAILABLE */
+       922, /* GL_MAX_VERTEX_ATTRIBS */
+      1791, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
+       370, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
+       369, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
+       908, /* GL_MAX_TEXTURE_COORDS */
+       910, /* GL_MAX_TEXTURE_IMAGE_UNITS */
+      1217, /* GL_PROGRAM_ERROR_STRING_ARB */
+      1219, /* GL_PROGRAM_FORMAT_ASCII_ARB */
+      1218, /* GL_PROGRAM_FORMAT_ARB */
+      1702, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
+       349, /* GL_DEPTH_BOUNDS_TEST_EXT */
+       348, /* GL_DEPTH_BOUNDS_EXT */
         52, /* GL_ARRAY_BUFFER */
-       454, /* GL_ELEMENT_ARRAY_BUFFER */
+       456, /* GL_ELEMENT_ARRAY_BUFFER */
         53, /* GL_ARRAY_BUFFER_BINDING */
-       455, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
-      1752, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
-       999, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
+       457, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
+      1765, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
+      1009, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
        146, /* GL_COLOR_ARRAY_BUFFER_BINDING */
-       616, /* GL_INDEX_ARRAY_BUFFER_BINDING */
-      1623, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
-       450, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
-      1380, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
-       502, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
-      1804, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
-      1774, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
-      1210, /* GL_PROGRAM_INSTRUCTIONS_ARB */
-       873, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
-      1216, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-       882, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-      1228, /* GL_PROGRAM_TEMPORARIES_ARB */
-       888, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
-      1218, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
-       884, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
-      1222, /* GL_PROGRAM_PARAMETERS_ARB */
-       887, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
-      1217, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
-       883, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
-      1203, /* GL_PROGRAM_ATTRIBS_ARB */
-       868, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
-      1215, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
-       881, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
-      1201, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
-       866, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
-      1213, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-       879, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-       874, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
-       870, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
-      1231, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
-      1701, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
-      1270, /* GL_READ_ONLY */
-      1812, /* GL_WRITE_ONLY */
-      1272, /* GL_READ_WRITE */
+       619, /* GL_INDEX_ARRAY_BUFFER_BINDING */
+      1635, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
+       452, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
+      1392, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
+       505, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
+      1817, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
+      1787, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
+      1220, /* GL_PROGRAM_INSTRUCTIONS_ARB */
+       883, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
+      1226, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+       892, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+      1238, /* GL_PROGRAM_TEMPORARIES_ARB */
+       898, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
+      1228, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
+       894, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
+      1232, /* GL_PROGRAM_PARAMETERS_ARB */
+       897, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
+      1227, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
+       893, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
+      1213, /* GL_PROGRAM_ATTRIBS_ARB */
+       878, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
+      1225, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
+       891, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
+      1211, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
+       876, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
+      1223, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+       889, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+       884, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
+       880, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
+      1241, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
+      1713, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
+      1282, /* GL_READ_ONLY */
+      1825, /* GL_WRITE_ONLY */
+      1284, /* GL_READ_WRITE */
        101, /* GL_BUFFER_ACCESS */
        103, /* GL_BUFFER_MAPPED */
        105, /* GL_BUFFER_MAP_POINTER */
-      1695, /* GL_TIME_ELAPSED_EXT */
-       784, /* GL_MATRIX0_ARB */
-       796, /* GL_MATRIX1_ARB */
-       808, /* GL_MATRIX2_ARB */
-       812, /* GL_MATRIX3_ARB */
-       814, /* GL_MATRIX4_ARB */
-       816, /* GL_MATRIX5_ARB */
-       818, /* GL_MATRIX6_ARB */
-       820, /* GL_MATRIX7_ARB */
-       822, /* GL_MATRIX8_ARB */
-       823, /* GL_MATRIX9_ARB */
-       786, /* GL_MATRIX10_ARB */
-       787, /* GL_MATRIX11_ARB */
-       788, /* GL_MATRIX12_ARB */
-       789, /* GL_MATRIX13_ARB */
-       790, /* GL_MATRIX14_ARB */
-       791, /* GL_MATRIX15_ARB */
-       792, /* GL_MATRIX16_ARB */
-       793, /* GL_MATRIX17_ARB */
-       794, /* GL_MATRIX18_ARB */
-       795, /* GL_MATRIX19_ARB */
-       798, /* GL_MATRIX20_ARB */
-       799, /* GL_MATRIX21_ARB */
-       800, /* GL_MATRIX22_ARB */
-       801, /* GL_MATRIX23_ARB */
-       802, /* GL_MATRIX24_ARB */
-       803, /* GL_MATRIX25_ARB */
-       804, /* GL_MATRIX26_ARB */
-       805, /* GL_MATRIX27_ARB */
-       806, /* GL_MATRIX28_ARB */
-       807, /* GL_MATRIX29_ARB */
-       810, /* GL_MATRIX30_ARB */
-       811, /* GL_MATRIX31_ARB */
-      1496, /* GL_STREAM_DRAW */
-      1498, /* GL_STREAM_READ */
-      1494, /* GL_STREAM_COPY */
-      1457, /* GL_STATIC_DRAW */
-      1459, /* GL_STATIC_READ */
-      1455, /* GL_STATIC_COPY */
-       444, /* GL_DYNAMIC_DRAW */
-       446, /* GL_DYNAMIC_READ */
-       442, /* GL_DYNAMIC_COPY */
-      1104, /* GL_PIXEL_PACK_BUFFER */
-      1108, /* GL_PIXEL_UNPACK_BUFFER */
-      1105, /* GL_PIXEL_PACK_BUFFER_BINDING */
-      1109, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
-       341, /* GL_DEPTH24_STENCIL8 */
-      1688, /* GL_TEXTURE_STENCIL_SIZE */
-       871, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
-       869, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
-       872, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
-       876, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
-       875, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
-       833, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
-      1490, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
+      1707, /* GL_TIME_ELAPSED_EXT */
+       794, /* GL_MATRIX0_ARB */
+       806, /* GL_MATRIX1_ARB */
+       818, /* GL_MATRIX2_ARB */
+       822, /* GL_MATRIX3_ARB */
+       824, /* GL_MATRIX4_ARB */
+       826, /* GL_MATRIX5_ARB */
+       828, /* GL_MATRIX6_ARB */
+       830, /* GL_MATRIX7_ARB */
+       832, /* GL_MATRIX8_ARB */
+       833, /* GL_MATRIX9_ARB */
+       796, /* GL_MATRIX10_ARB */
+       797, /* GL_MATRIX11_ARB */
+       798, /* GL_MATRIX12_ARB */
+       799, /* GL_MATRIX13_ARB */
+       800, /* GL_MATRIX14_ARB */
+       801, /* GL_MATRIX15_ARB */
+       802, /* GL_MATRIX16_ARB */
+       803, /* GL_MATRIX17_ARB */
+       804, /* GL_MATRIX18_ARB */
+       805, /* GL_MATRIX19_ARB */
+       808, /* GL_MATRIX20_ARB */
+       809, /* GL_MATRIX21_ARB */
+       810, /* GL_MATRIX22_ARB */
+       811, /* GL_MATRIX23_ARB */
+       812, /* GL_MATRIX24_ARB */
+       813, /* GL_MATRIX25_ARB */
+       814, /* GL_MATRIX26_ARB */
+       815, /* GL_MATRIX27_ARB */
+       816, /* GL_MATRIX28_ARB */
+       817, /* GL_MATRIX29_ARB */
+       820, /* GL_MATRIX30_ARB */
+       821, /* GL_MATRIX31_ARB */
+      1508, /* GL_STREAM_DRAW */
+      1510, /* GL_STREAM_READ */
+      1506, /* GL_STREAM_COPY */
+      1469, /* GL_STATIC_DRAW */
+      1471, /* GL_STATIC_READ */
+      1467, /* GL_STATIC_COPY */
+       446, /* GL_DYNAMIC_DRAW */
+       448, /* GL_DYNAMIC_READ */
+       444, /* GL_DYNAMIC_COPY */
+      1114, /* GL_PIXEL_PACK_BUFFER */
+      1118, /* GL_PIXEL_UNPACK_BUFFER */
+      1115, /* GL_PIXEL_PACK_BUFFER_BINDING */
+      1119, /* GL_PIXEL_UNPACK_BUFFER_BINDING */
+       343, /* GL_DEPTH24_STENCIL8 */
+      1700, /* GL_TEXTURE_STENCIL_SIZE */
+       881, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
+       879, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
+       882, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
+       886, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
+       885, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
+       843, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
+      1502, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
         17, /* GL_ACTIVE_STENCIL_FACE_EXT */
-       933, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
-      1361, /* GL_SAMPLES_PASSED */
-       525, /* GL_FRAGMENT_SHADER */
-      1798, /* GL_VERTEX_SHADER */
-      1221, /* GL_PROGRAM_OBJECT_ARB */
-      1393, /* GL_SHADER_OBJECT_ARB */
-       857, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
-       916, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
-       910, /* GL_MAX_VARYING_FLOATS */
-       914, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
-       842, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
-      1024, /* GL_OBJECT_TYPE_ARB */
-      1395, /* GL_SHADER_TYPE */
-       490, /* GL_FLOAT_VEC2 */
-       492, /* GL_FLOAT_VEC3 */
-       494, /* GL_FLOAT_VEC4 */
-       643, /* GL_INT_VEC2 */
-       645, /* GL_INT_VEC3 */
-       647, /* GL_INT_VEC4 */
+       943, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
+      1373, /* GL_SAMPLES_PASSED */
+       528, /* GL_FRAGMENT_SHADER */
+      1811, /* GL_VERTEX_SHADER */
+      1231, /* GL_PROGRAM_OBJECT_ARB */
+      1405, /* GL_SHADER_OBJECT_ARB */
+       867, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
+       926, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
+       920, /* GL_MAX_VARYING_FLOATS */
+       924, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
+       852, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
+      1034, /* GL_OBJECT_TYPE_ARB */
+      1407, /* GL_SHADER_TYPE */
+       493, /* GL_FLOAT_VEC2 */
+       495, /* GL_FLOAT_VEC3 */
+       497, /* GL_FLOAT_VEC4 */
+       646, /* GL_INT_VEC2 */
+       648, /* GL_INT_VEC3 */
+       650, /* GL_INT_VEC4 */
         93, /* GL_BOOL */
         95, /* GL_BOOL_VEC2 */
         97, /* GL_BOOL_VEC3 */
         99, /* GL_BOOL_VEC4 */
-       478, /* GL_FLOAT_MAT2 */
-       482, /* GL_FLOAT_MAT3 */
-       486, /* GL_FLOAT_MAT4 */
-      1352, /* GL_SAMPLER_1D */
-      1354, /* GL_SAMPLER_2D */
-      1356, /* GL_SAMPLER_3D */
-      1357, /* GL_SAMPLER_CUBE */
-      1353, /* GL_SAMPLER_1D_SHADOW */
-      1355, /* GL_SAMPLER_2D_SHADOW */
-       480, /* GL_FLOAT_MAT2x3 */
-       481, /* GL_FLOAT_MAT2x4 */
-       484, /* GL_FLOAT_MAT3x2 */
-       485, /* GL_FLOAT_MAT3x4 */
-       488, /* GL_FLOAT_MAT4x2 */
-       489, /* GL_FLOAT_MAT4x3 */
-       339, /* GL_DELETE_STATUS */
+       481, /* GL_FLOAT_MAT2 */
+       485, /* GL_FLOAT_MAT3 */
+       489, /* GL_FLOAT_MAT4 */
+      1364, /* GL_SAMPLER_1D */
+      1366, /* GL_SAMPLER_2D */
+      1368, /* GL_SAMPLER_3D */
+      1369, /* GL_SAMPLER_CUBE */
+      1365, /* GL_SAMPLER_1D_SHADOW */
+      1367, /* GL_SAMPLER_2D_SHADOW */
+       483, /* GL_FLOAT_MAT2x3 */
+       484, /* GL_FLOAT_MAT2x4 */
+       487, /* GL_FLOAT_MAT3x2 */
+       488, /* GL_FLOAT_MAT3x4 */
+       491, /* GL_FLOAT_MAT4x2 */
+       492, /* GL_FLOAT_MAT4x3 */
+       341, /* GL_DELETE_STATUS */
        243, /* GL_COMPILE_STATUS */
-       697, /* GL_LINK_STATUS */
-      1747, /* GL_VALIDATE_STATUS */
-       628, /* GL_INFO_LOG_LENGTH */
+       701, /* GL_LINK_STATUS */
+      1759, /* GL_VALIDATE_STATUS */
+       631, /* GL_INFO_LOG_LENGTH */
         55, /* GL_ATTACHED_SHADERS */
         20, /* GL_ACTIVE_UNIFORMS */
         21, /* GL_ACTIVE_UNIFORM_MAX_LENGTH */
-      1394, /* GL_SHADER_SOURCE_LENGTH */
+      1406, /* GL_SHADER_SOURCE_LENGTH */
         15, /* GL_ACTIVE_ATTRIBUTES */
         16, /* GL_ACTIVE_ATTRIBUTE_MAX_LENGTH */
-       527, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
-      1397, /* GL_SHADING_LANGUAGE_VERSION */
-       316, /* GL_CURRENT_PROGRAM */
-      1073, /* GL_PALETTE4_RGB8_OES */
-      1075, /* GL_PALETTE4_RGBA8_OES */
-      1071, /* GL_PALETTE4_R5_G6_B5_OES */
-      1074, /* GL_PALETTE4_RGBA4_OES */
-      1072, /* GL_PALETTE4_RGB5_A1_OES */
-      1078, /* GL_PALETTE8_RGB8_OES */
-      1080, /* GL_PALETTE8_RGBA8_OES */
-      1076, /* GL_PALETTE8_R5_G6_B5_OES */
-      1079, /* GL_PALETTE8_RGBA4_OES */
-      1077, /* GL_PALETTE8_RGB5_A1_OES */
-       610, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
-       609, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
-      1732, /* GL_UNSIGNED_NORMALIZED */
-      1578, /* GL_TEXTURE_1D_ARRAY_EXT */
-      1241, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
-      1580, /* GL_TEXTURE_2D_ARRAY_EXT */
-      1244, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
-      1586, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
-      1588, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
-      1449, /* GL_SRGB */
-      1450, /* GL_SRGB8 */
-      1452, /* GL_SRGB_ALPHA */
-      1451, /* GL_SRGB8_ALPHA8 */
-      1409, /* GL_SLUMINANCE_ALPHA */
-      1408, /* GL_SLUMINANCE8_ALPHA8 */
-      1406, /* GL_SLUMINANCE */
-      1407, /* GL_SLUMINANCE8 */
+       530, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
+      1409, /* GL_SHADING_LANGUAGE_VERSION */
+       318, /* GL_CURRENT_PROGRAM */
+      1083, /* GL_PALETTE4_RGB8_OES */
+      1085, /* GL_PALETTE4_RGBA8_OES */
+      1081, /* GL_PALETTE4_R5_G6_B5_OES */
+      1084, /* GL_PALETTE4_RGBA4_OES */
+      1082, /* GL_PALETTE4_RGB5_A1_OES */
+      1088, /* GL_PALETTE8_RGB8_OES */
+      1090, /* GL_PALETTE8_RGBA8_OES */
+      1086, /* GL_PALETTE8_R5_G6_B5_OES */
+      1089, /* GL_PALETTE8_RGBA4_OES */
+      1087, /* GL_PALETTE8_RGB5_A1_OES */
+       613, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
+       612, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
+      1744, /* GL_UNSIGNED_NORMALIZED */
+      1590, /* GL_TEXTURE_1D_ARRAY_EXT */
+      1252, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
+      1592, /* GL_TEXTURE_2D_ARRAY_EXT */
+      1255, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
+      1598, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
+      1600, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
+      1461, /* GL_SRGB */
+      1462, /* GL_SRGB8 */
+      1464, /* GL_SRGB_ALPHA */
+      1463, /* GL_SRGB8_ALPHA8 */
+      1421, /* GL_SLUMINANCE_ALPHA */
+      1420, /* GL_SLUMINANCE8_ALPHA8 */
+      1418, /* GL_SLUMINANCE */
+      1419, /* GL_SLUMINANCE8 */
        264, /* GL_COMPRESSED_SRGB */
        265, /* GL_COMPRESSED_SRGB_ALPHA */
        262, /* GL_COMPRESSED_SLUMINANCE */
        263, /* GL_COMPRESSED_SLUMINANCE_ALPHA */
-      1138, /* GL_POINT_SPRITE_COORD_ORIGIN */
-       705, /* GL_LOWER_LEFT */
-      1744, /* GL_UPPER_LEFT */
-      1472, /* GL_STENCIL_BACK_REF */
-      1473, /* GL_STENCIL_BACK_VALUE_MASK */
-      1474, /* GL_STENCIL_BACK_WRITEMASK */
-       435, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
-      1286, /* GL_RENDERBUFFER_BINDING_EXT */
-      1267, /* GL_READ_FRAMEBUFFER */
-       434, /* GL_DRAW_FRAMEBUFFER */
-      1268, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
-      1296, /* GL_RENDERBUFFER_SAMPLES */
-       537, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
-       535, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
-       546, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
-       542, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
-       544, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
-       549, /* GL_FRAMEBUFFER_COMPLETE */
-       553, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
-       559, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
-       557, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
-       555, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
-       558, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
-       556, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
-       562, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
-       565, /* GL_FRAMEBUFFER_UNSUPPORTED */
-       563, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
-       839, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
+      1148, /* GL_POINT_SPRITE_COORD_ORIGIN */
+       709, /* GL_LOWER_LEFT */
+      1756, /* GL_UPPER_LEFT */
+      1484, /* GL_STENCIL_BACK_REF */
+      1485, /* GL_STENCIL_BACK_VALUE_MASK */
+      1486, /* GL_STENCIL_BACK_WRITEMASK */
+       437, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
+      1298, /* GL_RENDERBUFFER_BINDING_EXT */
+      1279, /* GL_READ_FRAMEBUFFER */
+       436, /* GL_DRAW_FRAMEBUFFER */
+      1280, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
+      1308, /* GL_RENDERBUFFER_SAMPLES */
+       540, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
+       538, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
+       549, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
+       545, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
+       547, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
+       552, /* GL_FRAMEBUFFER_COMPLETE */
+       556, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
+       562, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
+       560, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
+       558, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
+       561, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
+       559, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
+       565, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
+       568, /* GL_FRAMEBUFFER_UNSUPPORTED */
+       566, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+       849, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
        152, /* GL_COLOR_ATTACHMENT0 */
        154, /* GL_COLOR_ATTACHMENT1 */
        168, /* GL_COLOR_ATTACHMENT2 */
@@ -4972,36 +4998,42 @@ static const unsigned reduced_enums[1319] =
        161, /* GL_COLOR_ATTACHMENT13 */
        163, /* GL_COLOR_ATTACHMENT14 */
        165, /* GL_COLOR_ATTACHMENT15 */
-       342, /* GL_DEPTH_ATTACHMENT */
-      1462, /* GL_STENCIL_ATTACHMENT */
-       528, /* GL_FRAMEBUFFER */
-      1284, /* GL_RENDERBUFFER */
-      1298, /* GL_RENDERBUFFER_WIDTH */
-      1291, /* GL_RENDERBUFFER_HEIGHT */
-      1293, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
-      1485, /* GL_STENCIL_INDEX_EXT */
-      1482, /* GL_STENCIL_INDEX1_EXT */
-      1483, /* GL_STENCIL_INDEX4_EXT */
-      1484, /* GL_STENCIL_INDEX8_EXT */
-      1481, /* GL_STENCIL_INDEX16_EXT */
-      1295, /* GL_RENDERBUFFER_RED_SIZE */
-      1290, /* GL_RENDERBUFFER_GREEN_SIZE */
-      1287, /* GL_RENDERBUFFER_BLUE_SIZE */
-      1285, /* GL_RENDERBUFFER_ALPHA_SIZE */
-      1288, /* GL_RENDERBUFFER_DEPTH_SIZE */
-      1297, /* GL_RENDERBUFFER_STENCIL_SIZE */
-       561, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
-       895, /* GL_MAX_SAMPLES */
-      1345, /* GL_RGBA_SNORM */
-      1341, /* GL_RGBA8_SNORM */
-      1402, /* GL_SIGNED_NORMALIZED */
-       461, /* GL_EVAL_BIT */
-      1265, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
-       699, /* GL_LIST_BIT */
-      1594, /* GL_TEXTURE_BIT */
-      1376, /* GL_SCISSOR_BIT */
+       344, /* GL_DEPTH_ATTACHMENT */
+      1474, /* GL_STENCIL_ATTACHMENT */
+       531, /* GL_FRAMEBUFFER */
+      1296, /* GL_RENDERBUFFER */
+      1310, /* GL_RENDERBUFFER_WIDTH */
+      1303, /* GL_RENDERBUFFER_HEIGHT */
+      1305, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
+      1497, /* GL_STENCIL_INDEX_EXT */
+      1494, /* GL_STENCIL_INDEX1_EXT */
+      1495, /* GL_STENCIL_INDEX4_EXT */
+      1496, /* GL_STENCIL_INDEX8_EXT */
+      1493, /* GL_STENCIL_INDEX16_EXT */
+      1307, /* GL_RENDERBUFFER_RED_SIZE */
+      1302, /* GL_RENDERBUFFER_GREEN_SIZE */
+      1299, /* GL_RENDERBUFFER_BLUE_SIZE */
+      1297, /* GL_RENDERBUFFER_ALPHA_SIZE */
+      1300, /* GL_RENDERBUFFER_DEPTH_SIZE */
+      1309, /* GL_RENDERBUFFER_STENCIL_SIZE */
+       564, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
+       905, /* GL_MAX_SAMPLES */
+      1266, /* GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT */
+       478, /* GL_FIRST_VERTEX_CONVENTION_EXT */
+       661, /* GL_LAST_VERTEX_CONVENTION_EXT */
+      1245, /* GL_PROVOKING_VERTEX_EXT */
+       298, /* GL_COPY_READ_BUFFER */
+       299, /* GL_COPY_WRITE_BUFFER */
+      1357, /* GL_RGBA_SNORM */
+      1353, /* GL_RGBA8_SNORM */
+      1414, /* GL_SIGNED_NORMALIZED */
+       463, /* GL_EVAL_BIT */
+      1277, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
+       703, /* GL_LIST_BIT */
+      1606, /* GL_TEXTURE_BIT */
+      1388, /* GL_SCISSOR_BIT */
         29, /* GL_ALL_ATTRIB_BITS */
-       980, /* GL_MULTISAMPLE_BIT */
+       990, /* GL_MULTISAMPLE_BIT */
         30, /* GL_ALL_CLIENT_ATTRIB_BITS */
 };
 
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index 5c4bea9cf6..c60b58a492 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.3
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -44,6 +45,7 @@ static const struct {
    const char *name;
    int flag_offset;
 } default_extensions[] = {
+   { OFF, "GL_ARB_copy_buffer",                F(ARB_copy_buffer) },
    { OFF, "GL_ARB_depth_texture",              F(ARB_depth_texture) },
    { ON,  "GL_ARB_draw_buffers",               F(ARB_draw_buffers) },
    { OFF, "GL_ARB_fragment_program",           F(ARB_fragment_program) },
@@ -52,6 +54,7 @@ static const struct {
    { OFF, "GL_ARB_framebuffer_object",         F(ARB_framebuffer_object) },
    { OFF, "GL_ARB_half_float_pixel",           F(ARB_half_float_pixel) },
    { OFF, "GL_ARB_imaging",                    F(ARB_imaging) },
+   { OFF, "GL_ARB_map_buffer_range",           F(ARB_map_buffer_range) },
    { ON,  "GL_ARB_multisample",                F(ARB_multisample) },
    { OFF, "GL_ARB_multitexture",               F(ARB_multitexture) },
    { OFF, "GL_ARB_occlusion_query",            F(ARB_occlusion_query) },
@@ -75,6 +78,7 @@ static const struct {
    { OFF, "GL_ARB_texture_non_power_of_two",   F(ARB_texture_non_power_of_two)},
    { OFF, "GL_ARB_texture_rectangle",          F(NV_texture_rectangle) },
    { ON,  "GL_ARB_transpose_matrix",           F(ARB_transpose_matrix) },
+   { OFF, "GL_ARB_vertex_array_object",        F(ARB_vertex_array_object) },
    { ON,  "GL_ARB_vertex_buffer_object",       F(ARB_vertex_buffer_object) },
    { OFF, "GL_ARB_vertex_program",             F(ARB_vertex_program) },
    { OFF, "GL_ARB_vertex_shader",              F(ARB_vertex_shader) },
@@ -106,6 +110,7 @@ static const struct {
    { OFF, "GL_EXT_pixel_buffer_object",        F(EXT_pixel_buffer_object) },
    { OFF, "GL_EXT_point_parameters",           F(EXT_point_parameters) },
    { ON,  "GL_EXT_polygon_offset",             F(EXT_polygon_offset) },
+   { OFF, "GL_EXT_provoking_vertex",           F(EXT_provoking_vertex) },
    { ON,  "GL_EXT_rescale_normal",             F(EXT_rescale_normal) },
    { OFF, "GL_EXT_secondary_color",            F(EXT_secondary_color) },
    { ON,  "GL_EXT_separate_specular_color",    F(EXT_separate_specular_color) },
@@ -183,6 +188,7 @@ static const struct {
 void
 _mesa_enable_sw_extensions(GLcontext *ctx)
 {
+   ctx->Extensions.ARB_copy_buffer = GL_TRUE;
    ctx->Extensions.ARB_depth_texture = GL_TRUE;
    /*ctx->Extensions.ARB_draw_buffers = GL_TRUE;*/
 #if FEATURE_ARB_fragment_program
@@ -197,6 +203,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #endif
    ctx->Extensions.ARB_half_float_pixel = GL_TRUE;
    ctx->Extensions.ARB_imaging = GL_TRUE;
+   ctx->Extensions.ARB_map_buffer_range = GL_TRUE;
    ctx->Extensions.ARB_multitexture = GL_TRUE;
 #if FEATURE_ARB_occlusion_query
    ctx->Extensions.ARB_occlusion_query = GL_TRUE;
@@ -221,6 +228,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    /*ctx->Extensions.ARB_texture_float = GL_TRUE;*/
    ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE;
    ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE;
+   ctx->Extensions.ARB_vertex_array_object = GL_TRUE;
 #if FEATURE_ARB_vertex_program
    ctx->Extensions.ARB_vertex_program = GL_TRUE;
 #endif
@@ -261,6 +269,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE;
 #endif
    ctx->Extensions.EXT_point_parameters = GL_TRUE;
+   ctx->Extensions.EXT_provoking_vertex = GL_TRUE;
    ctx->Extensions.EXT_shadow_funcs = GL_TRUE;
    ctx->Extensions.EXT_secondary_color = GL_TRUE;
    ctx->Extensions.EXT_shared_texture_palette = GL_TRUE;
@@ -440,8 +449,9 @@ _mesa_enable_2_1_extensions(GLcontext *ctx)
 
 /**
  * Either enable or disable the named extension.
+ * \return GL_TRUE for success, GL_FALSE if invalid extension name
  */
-static void
+static GLboolean
 set_extension( GLcontext *ctx, const char *name, GLboolean state )
 {
    GLboolean *base = (GLboolean *) &ctx->Extensions;
@@ -450,7 +460,7 @@ set_extension( GLcontext *ctx, const char *name, GLboolean state )
    if (ctx->Extensions.String) {
       /* The string was already queried - can't change it now! */
       _mesa_problem(ctx, "Trying to enable/disable extension after glGetString(GL_EXTENSIONS): %s", name);
-      return;
+      return GL_FALSE;
    }
 
    for (i = 0 ; i < Elements(default_extensions) ; i++) {
@@ -459,10 +469,10 @@ set_extension( GLcontext *ctx, const char *name, GLboolean state )
             GLboolean *enabled = base + default_extensions[i].flag_offset;
             *enabled = state;
          }
-         return;
+         return GL_TRUE;
       }
    }
-   _mesa_problem(ctx, "Trying to enable unknown extension: %s", name);
+   return GL_FALSE;
 }
 
 
@@ -473,7 +483,8 @@ set_extension( GLcontext *ctx, const char *name, GLboolean state )
 void
 _mesa_enable_extension( GLcontext *ctx, const char *name )
 {
-   set_extension(ctx, name, GL_TRUE);
+   if (!set_extension(ctx, name, GL_TRUE))
+      _mesa_problem(ctx, "Trying to enable unknown extension: %s", name);
 }
 
 
@@ -484,7 +495,8 @@ _mesa_enable_extension( GLcontext *ctx, const char *name )
 void
 _mesa_disable_extension( GLcontext *ctx, const char *name )
 {
-   set_extension(ctx, name, GL_FALSE);
+   if (!set_extension(ctx, name, GL_FALSE))
+      _mesa_problem(ctx, "Trying to disable unknown extension: %s", name);
 }
 
 
@@ -509,6 +521,80 @@ _mesa_extension_is_enabled( GLcontext *ctx, const char *name )
 
 
 /**
+ * Append string 'b' onto string 'a'.  Free 'a' and return new string.
+ */
+static char *
+append(const char *a, const char *b)
+{
+   const GLuint aLen = a ? _mesa_strlen(a) : 0;
+   const GLuint bLen = b ? _mesa_strlen(b) : 0;
+   char *s = _mesa_calloc(aLen + bLen + 1);
+   if (s) {
+      if (a)
+         _mesa_memcpy(s, a, aLen);
+      if (b)
+         _mesa_memcpy(s + aLen, b, bLen);
+      s[aLen + bLen] = '\0';
+   }
+   if (a)
+      _mesa_free((void *) a);
+   return s;
+}
+
+
+/**
+ * Check the MESA_EXTENSION_OVERRIDE env var.
+ * For extension names that are recognized, turn them on.  For extension
+ * names that are recognized and prefixed with '-', turn them off.
+ * Return a string of the unknown/leftover names.
+ */
+static const char *
+get_extension_override( GLcontext *ctx )
+{
+   const char *envExt = _mesa_getenv("MESA_EXTENSION_OVERRIDE");
+   char *extraExt = NULL;
+   char ext[1000];
+   GLuint extLen = 0;
+   GLuint i;
+   GLboolean disableExt = GL_FALSE;
+
+   if (!envExt)
+      return NULL;
+
+   for (i = 0; ; i++) {
+      if (envExt[i] == '\0' || envExt[i] == ' ') {
+         /* terminate/process 'ext' if extLen > 0 */
+         if (extLen > 0) {
+            assert(extLen < sizeof(ext));
+            /* enable extension named by 'ext' */
+            ext[extLen] = 0;
+            if (!set_extension(ctx, ext, !disableExt)) {
+               /* unknown extension name, append it to extraExt */
+               if (extraExt) {
+                  extraExt = append(extraExt, " ");
+               }
+               extraExt = append(extraExt, ext);
+            }
+            extLen = 0;
+            disableExt = GL_FALSE;
+         }
+         if (envExt[i] == '\0')
+            break;
+      }
+      else if (envExt[i] == '-') {
+         disableExt = GL_TRUE;
+      }
+      else {
+         /* accumulate this non-space character */
+         ext[extLen++] = envExt[i];
+      }
+   }
+
+   return extraExt;
+}
+
+
+/**
  * Run through the default_extensions array above and set the
  * ctx->Extensions.ARB/EXT_* flags accordingly.
  * To be called during context initialization.
@@ -536,8 +622,9 @@ GLubyte *
 _mesa_make_extension_string( GLcontext *ctx )
 {
    const GLboolean *base = (const GLboolean *) &ctx->Extensions;
+   const char *extraExt = get_extension_override(ctx);
    GLuint extStrLen = 0;
-   GLubyte *s;
+   char *s;
    GLuint i;
 
    /* first, compute length of the extension string */
@@ -547,7 +634,14 @@ _mesa_make_extension_string( GLcontext *ctx )
          extStrLen += (GLuint)_mesa_strlen(default_extensions[i].name) + 1;
       }
    }
-   s = (GLubyte *) _mesa_malloc(extStrLen);
+
+   if (extraExt)
+      extStrLen += _mesa_strlen(extraExt) + 1; /* +1 for space */
+
+   /* allocate the extension string */
+   s = (char *) _mesa_malloc(extStrLen);
+   if (!s)
+      return NULL;
 
    /* second, build the extension string */
    extStrLen = 0;
@@ -557,13 +651,18 @@ _mesa_make_extension_string( GLcontext *ctx )
          GLuint len = (GLuint)_mesa_strlen(default_extensions[i].name);
          _mesa_memcpy(s + extStrLen, default_extensions[i].name, len);
          extStrLen += len;
-         s[extStrLen] = (GLubyte) ' ';
+         s[extStrLen] = ' ';
          extStrLen++;
       }
    }
    ASSERT(extStrLen > 0);
 
-   s[extStrLen - 1] = 0;
+   s[extStrLen - 1] = 0; /* -1 to overwrite trailing the ' ' */
 
-   return s;
+   if (extraExt) {
+      s = append(s, " ");
+      s = append(s, extraExt);
+   }
+
+   return (GLubyte *) s;
 }
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index 1ed6fc3383..516159bbf2 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -1835,6 +1835,14 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          CHECK_EXT1(EXT_framebuffer_blit, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->ReadBuffer->Name);
          break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetBooleanv");
+         params[0] = ctx->Light.ProvokingVertex;
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetBooleanv");
+         params[0] = ctx->Const.QuadsFollowProvokingVertexConvention;
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Const.FragmentProgram.MaxUniformComponents);
@@ -3649,6 +3657,14 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          CHECK_EXT1(EXT_framebuffer_blit, "GetFloatv");
          params[0] = (GLfloat)(ctx->ReadBuffer->Name);
          break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetFloatv");
+         params[0] = BOOLEAN_TO_FLOAT(ctx->Light.ProvokingVertex);
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetFloatv");
+         params[0] = BOOLEAN_TO_FLOAT(ctx->Const.QuadsFollowProvokingVertexConvention);
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetFloatv");
          params[0] = (GLfloat)(ctx->Const.FragmentProgram.MaxUniformComponents);
@@ -5463,6 +5479,14 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          CHECK_EXT1(EXT_framebuffer_blit, "GetIntegerv");
          params[0] = ctx->ReadBuffer->Name;
          break;
+      case GL_PROVOKING_VERTEX_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetIntegerv");
+         params[0] = BOOLEAN_TO_INT(ctx->Light.ProvokingVertex);
+         break;
+      case GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT:
+         CHECK_EXT1(EXT_provoking_vertex, "GetIntegerv");
+         params[0] = BOOLEAN_TO_INT(ctx->Const.QuadsFollowProvokingVertexConvention);
+         break;
       case GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB:
          CHECK_EXT1(ARB_fragment_shader, "GetIntegerv");
          params[0] = ctx->Const.FragmentProgram.MaxUniformComponents;
diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py
index 43ee5fff10..5666ad0e42 100644
--- a/src/mesa/main/get_gen.py
+++ b/src/mesa/main/get_gen.py
@@ -976,6 +976,13 @@ StateVars = [
 	( "GL_READ_FRAMEBUFFER_BINDING_EXT", GLint, ["ctx->ReadBuffer->Name"], "",
 	  ["EXT_framebuffer_blit"] ),
 
+	# GL_EXT_provoking_vertex
+	( "GL_PROVOKING_VERTEX_EXT", GLboolean,
+	  ["ctx->Light.ProvokingVertex"], "", ["EXT_provoking_vertex"] ),
+	( "GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION_EXT", GLboolean,
+	  ["ctx->Const.QuadsFollowProvokingVertexConvention"], "",
+	  ["EXT_provoking_vertex"] ),
+
 	# GL_ARB_fragment_shader
 	( "GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB", GLint,
 	  ["ctx->Const.FragmentProgram.MaxUniformComponents"], "",
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index 01fbe40a03..090e5eb330 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -182,6 +182,8 @@ _mesa_sizeof_type( GLenum type )
 	 return sizeof(GLint);
       case GL_FLOAT:
 	 return sizeof(GLfloat);
+      case GL_DOUBLE:
+	 return sizeof(GLdouble);
       case GL_HALF_FLOAT_ARB:
 	 return sizeof(GLhalfARB);
       default:
diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c
index ac604fd12c..0f0d831fee 100644
--- a/src/mesa/main/light.c
+++ b/src/mesa/main/light.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.0
+ * Version:  7.5
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -64,6 +65,37 @@ _mesa_ShadeModel( GLenum mode )
 
 
 /**
+ * Set the provoking vertex (the vertex which specifies the prim's
+ * color when flat shading) to either the first or last vertex of the
+ * triangle or line.
+ */
+void GLAPIENTRY
+_mesa_ProvokingVertexEXT(GLenum mode)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   if (MESA_VERBOSE&VERBOSE_API)
+      _mesa_debug(ctx, "glProvokingVertexEXT 0x%x\n", mode);
+
+   switch (mode) {
+   case GL_FIRST_VERTEX_CONVENTION_EXT:
+   case GL_LAST_VERTEX_CONVENTION_EXT:
+      break;
+   default:
+      _mesa_error(ctx, GL_INVALID_ENUM, "glProvokingVertexEXT(0x%x)", mode);
+      return;
+   }
+
+   if (ctx->Light.ProvokingVertex == mode)
+      return;
+
+   FLUSH_VERTICES(ctx, _NEW_LIGHT);
+   ctx->Light.ProvokingVertex = mode;
+}
+
+
+/**
  * Helper function called by _mesa_Lightfv and _mesa_PopAttrib to set
  * per-light state.
  * For GL_POSITION and GL_SPOT_DIRECTION the params position/direction
@@ -1348,6 +1380,7 @@ _mesa_init_lighting( GLcontext *ctx )
    init_lightmodel( &ctx->Light.Model );
    init_material( &ctx->Light.Material );
    ctx->Light.ShadeModel = GL_SMOOTH;
+   ctx->Light.ProvokingVertex = GL_LAST_VERTEX_CONVENTION_EXT;
    ctx->Light.Enabled = GL_FALSE;
    ctx->Light.ColorMaterialFace = GL_FRONT_AND_BACK;
    ctx->Light.ColorMaterialMode = GL_AMBIENT_AND_DIFFUSE;
diff --git a/src/mesa/main/light.h b/src/mesa/main/light.h
index b97e17b5be..9c1a5eefad 100644
--- a/src/mesa/main/light.h
+++ b/src/mesa/main/light.h
@@ -1,13 +1,9 @@
-/**
- * \file light.h
- * Lighting.
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  3.5
+ * Version:  7.5
  *
- * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -37,6 +33,10 @@
 extern void GLAPIENTRY
 _mesa_ShadeModel( GLenum mode );
 
+extern void GLAPIENTRY
+_mesa_ProvokingVertexEXT(GLenum mode);
+
+
 #if _HAVE_FULL_GL
 extern void GLAPIENTRY
 _mesa_ColorMaterial( GLenum face, GLenum mode );
diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h
index 62c3ead3e1..ef973314e3 100644
--- a/src/mesa/main/mfeatures.h
+++ b/src/mesa/main/mfeatures.h
@@ -61,6 +61,7 @@
 #define FEATURE_ARB_occlusion_query  _HAVE_FULL_GL
 #define FEATURE_ARB_fragment_program  _HAVE_FULL_GL
 #define FEATURE_ARB_framebuffer_object  _HAVE_FULL_GL
+#define FEATURE_ARB_map_buffer_range  _HAVE_FULL_GL
 #define FEATURE_ARB_pixel_buffer_object  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_buffer_object  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_program  _HAVE_FULL_GL
diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c
index b306700484..3dca09d9f2 100644
--- a/src/mesa/main/mipmap.c
+++ b/src/mesa/main/mipmap.c
@@ -1481,9 +1481,12 @@ next_mipmap_level_size(GLenum target, GLint border,
 
 
 /**
- * For GL_SGIX_generate_mipmap:
- * Generate a complete set of mipmaps from texObj's base-level image.
+ * Automatic mipmap generation.
+ * This is the fallback/default function for ctx->Driver.GenerateMipmap().
+ * Generate a complete set of mipmaps from texObj's BaseLevel image.
  * Stop at texObj's MaxLevel or when we get to the 1x1 texture.
+ * For cube maps, target will be one of
+ * GL_TEXTURE_CUBE_MAP_POSITIVE/NEGATIVE_X/Y/Z; never GL_TEXTURE_CUBE_MAP.
  */
 void
 _mesa_generate_mipmap(GLcontext *ctx, GLenum target,
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 3b0ebcb7ae..d0309f5e90 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -770,6 +770,7 @@ struct gl_light_attrib
 
    GLboolean Enabled;			/**< Lighting enabled flag */
    GLenum ShadeModel;			/**< GL_FLAT or GL_SMOOTH */
+   GLenum ProvokingVertex;              /**< GL_EXT_provoking_vertex */
    GLenum ColorMaterialFace;		/**< GL_FRONT, BACK or FRONT_AND_BACK */
    GLenum ColorMaterialMode;		/**< GL_AMBIENT, GL_DIFFUSE, etc */
    GLbitfield ColorMaterialBitmask;	/**< bitmask formed from Face and Mode */
@@ -1028,7 +1029,7 @@ struct gl_stencil_attrib
 
 /**
  * An index for each type of texture object.  These correspond to the GL
- * target target enums, such as GL_TEXTURE_2D, GL_TEXTURE_CUBE_MAP, etc.
+ * texture target enums, such as GL_TEXTURE_2D, GL_TEXTURE_CUBE_MAP, etc.
  * Note: the order is from highest priority to lowest priority.
  */
 typedef enum
@@ -1498,14 +1499,17 @@ struct gl_buffer_object
 {
    GLint RefCount;
    GLuint Name;
-   GLenum Usage;
-   GLenum Access;
-   GLvoid *Pointer;          /**< Only valid while buffer is mapped */
-   GLintptr Offset;          /**< mapped offset */
-   GLsizeiptr Length;        /**< mapped length */
-   GLsizeiptrARB Size;       /**< Size of storage in bytes */
-   GLubyte *Data;            /**< Location of storage either in RAM or VRAM. */
-   GLboolean OnCard;         /**< Is buffer in VRAM? (hardware drivers) */
+   GLenum Usage;        /**< GL_STREAM_DRAW_ARB, GL_STREAM_READ_ARB, etc. */
+   GLsizeiptrARB Size;  /**< Size of buffer storage in bytes */
+   GLubyte *Data;       /**< Location of storage either in RAM or VRAM. */
+   /** Fields describing a mapped buffer */
+   /*@{*/
+   GLbitfield AccessFlags; /**< Mask of GL_MAP_x_BIT flags */
+   GLvoid *Pointer;     /**< User-space address of mapping */
+   GLintptr Offset;     /**< Mapped offset */
+   GLsizeiptr Length;   /**< Mapped length */
+   /*@}*/
+   GLboolean Written;   /**< Ever written to? (for debugging) */
 };
 
 
@@ -1541,10 +1545,10 @@ struct gl_client_array
    const GLubyte *Ptr;          /**< Points to array data */
    GLboolean Enabled;		/**< Enabled flag is a boolean */
    GLboolean Normalized;        /**< GL_ARB_vertex_program */
+   GLuint _ElementSize;         /**< size of each element in bytes */
 
-   /**< GL_ARB_vertex_buffer_object */
-   struct gl_buffer_object *BufferObj;
-   GLuint _MaxElement;
+   struct gl_buffer_object *BufferObj;/**< GL_ARB_vertex_buffer_object */
+   GLuint _MaxElement;          /**< max element index into array buffer + 1 */
 };
 
 
@@ -1559,10 +1563,12 @@ struct gl_array_object
 
    GLint RefCount;
    _glthread_Mutex Mutex;
+   GLboolean VBOonly;  /**< require all arrays to live in VBOs? */
 
    /** Conventional vertex arrays */
    /*@{*/
    struct gl_client_array Vertex;
+   struct gl_client_array Weight;
    struct gl_client_array Normal;
    struct gl_client_array Color;
    struct gl_client_array SecondaryColor;
@@ -1573,11 +1579,22 @@ struct gl_array_object
    struct gl_client_array PointSize;
    /*@}*/
 
-   /** Generic arrays for vertex programs/shaders */
-   struct gl_client_array VertexAttrib[VERT_ATTRIB_MAX];
+   /**
+    * Generic arrays for vertex programs/shaders.
+    * For NV vertex programs, these attributes alias and take priority
+    * over the conventional attribs above.  For ARB vertex programs and
+    * GLSL vertex shaders, these attributes are separate.
+    */
+   struct gl_client_array VertexAttrib[MAX_VERTEX_GENERIC_ATTRIBS];
 
    /** Mask of _NEW_ARRAY_* values indicating which arrays are enabled */
    GLbitfield _Enabled;
+
+   /**
+    * Min of all enabled arrays' _MaxElement.  When arrays reside inside VBOs
+    * we can determine the max legal (in bounds) glDrawElements array index.
+    */
+   GLuint _MaxElement;
 };
 
 
@@ -1586,9 +1603,15 @@ struct gl_array_object
  */
 struct gl_array_attrib
 {
+   /** Currently bound array object. See _mesa_BindVertexArrayAPPLE() */
    struct gl_array_object *ArrayObj;
+
+   /** The default vertex array object */
    struct gl_array_object *DefaultArrayObj;
 
+   /** Array objects (GL_ARB/APPLE_vertex_array_object) */
+   struct _mesa_HashTable *Objects;
+
    GLint ActiveTexture;		/**< Client Active Texture */
    GLuint LockFirst;            /**< GL_EXT_compiled_vertex_array */
    GLuint LockCount;            /**< GL_EXT_compiled_vertex_array */
@@ -1596,11 +1619,9 @@ struct gl_array_attrib
    GLbitfield NewState;		/**< mask of _NEW_ARRAY_* values */
 
 #if FEATURE_ARB_vertex_buffer_object
-   struct gl_buffer_object *NullBufferObj;
    struct gl_buffer_object *ArrayBufferObj;
    struct gl_buffer_object *ElementArrayBufferObj;
 #endif
-   GLuint _MaxElement;          /* Min of all enabled array's maxes */
 };
 
 
@@ -2068,6 +2089,8 @@ struct gl_shared_state
    GLuint TextureStateStamp;	        /**< state notification for shared tex */
    /*@}*/
 
+   /** Default buffer object for vertex arrays that aren't in VBOs */
+   struct gl_buffer_object *NullBufferObj;
 
    /**
     * \name Vertex/fragment programs
@@ -2101,9 +2124,6 @@ struct gl_shared_state
    struct _mesa_HashTable *FrameBuffers;
 #endif
 
-   /** Objects associated with the GL_APPLE_vertex_array_object extension. */
-   struct _mesa_HashTable *ArrayObjects;
-
    void *DriverData;  /**< Device driver shared state */
 };
 
@@ -2407,6 +2427,9 @@ struct gl_constants
    GLuint MaxVarying;  /**< Number of float[4] varying parameters */
 
    GLbitfield SupportedBumpUnits; /**> units supporting GL_ATI_envmap_bumpmap as targets */
+
+   /**< GL_EXT_provoking_vertex */
+   GLboolean QuadsFollowProvokingVertexConvention;
 };
 
 
@@ -2417,6 +2440,7 @@ struct gl_constants
 struct gl_extensions
 {
    GLboolean dummy;  /* don't remove this! */
+   GLboolean ARB_copy_buffer;
    GLboolean ARB_depth_texture;
    GLboolean ARB_draw_buffers;
    GLboolean ARB_fragment_program;
@@ -2425,6 +2449,7 @@ struct gl_extensions
    GLboolean ARB_framebuffer_object;
    GLboolean ARB_half_float_pixel;
    GLboolean ARB_imaging;
+   GLboolean ARB_map_buffer_range;
    GLboolean ARB_multisample;
    GLboolean ARB_multitexture;
    GLboolean ARB_occlusion_query;
@@ -2444,6 +2469,7 @@ struct gl_extensions
    GLboolean ARB_texture_mirrored_repeat;
    GLboolean ARB_texture_non_power_of_two;
    GLboolean ARB_transpose_matrix;
+   GLboolean ARB_vertex_array_object;
    GLboolean ARB_vertex_buffer_object;
    GLboolean ARB_vertex_program;
    GLboolean ARB_vertex_shader;
@@ -2475,6 +2501,7 @@ struct gl_extensions
    GLboolean EXT_pixel_buffer_object;
    GLboolean EXT_point_parameters;
    GLboolean EXT_polygon_offset;
+   GLboolean EXT_provoking_vertex;
    GLboolean EXT_rescale_normal;
    GLboolean EXT_shadow_funcs;
    GLboolean EXT_secondary_color;
@@ -2623,6 +2650,7 @@ struct gl_matrix_stack
 #define _NEW_PROGRAM            0x8000000  /**< __GLcontextRec::VertexProgram */
 #define _NEW_CURRENT_ATTRIB     0x10000000  /**< __GLcontextRec::Current */
 #define _NEW_PROGRAM_CONSTANTS  0x20000000
+#define _NEW_BUFFER_OBJECT      0x40000000
 #define _NEW_ALL ~0
 /*@}*/
 
@@ -2948,6 +2976,9 @@ struct __GLcontextRec
    struct gl_shader_state Shader; /**< GLSL shader object state */
 
    struct gl_query_state Query;  /**< occlusion, timer queries */
+
+   struct gl_buffer_object *CopyReadBuffer; /**< GL_ARB_copy_buffer */
+   struct gl_buffer_object *CopyWriteBuffer; /**< GL_ARB_copy_buffer */
    /*@}*/
 
 #if FEATURE_EXT_framebuffer_object
diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c
index 57ae9c721a..d9f3e476e8 100644
--- a/src/mesa/main/pixel.c
+++ b/src/mesa/main/pixel.c
@@ -170,7 +170,7 @@ _mesa_PixelMapfv( GLenum map, GLsizei mapsize, const GLfloat *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
                                               ctx->Unpack.BufferObj);
@@ -229,7 +229,7 @@ _mesa_PixelMapuiv(GLenum map, GLsizei mapsize, const GLuint *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
                                               ctx->Unpack.BufferObj);
@@ -303,7 +303,7 @@ _mesa_PixelMapusv(GLenum map, GLsizei mapsize, const GLushort *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                                               GL_READ_ONLY_ARB,
                                               ctx->Unpack.BufferObj);
@@ -371,7 +371,7 @@ _mesa_GetPixelMapfv( GLenum map, GLfloat *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
                                               ctx->Pack.BufferObj);
@@ -432,7 +432,7 @@ _mesa_GetPixelMapuiv( GLenum map, GLuint *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
                                               ctx->Pack.BufferObj);
@@ -494,7 +494,7 @@ _mesa_GetPixelMapusv( GLenum map, GLushort *values )
          return;
       }
       /* restore */
-      ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+      ctx->DefaultPacking.BufferObj = ctx->Shared->NullBufferObj;
       buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
                                               GL_WRITE_ONLY_ARB,
                                               ctx->Pack.BufferObj);
diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c
index ff1a6344cc..6a641f83f2 100644
--- a/src/mesa/main/pixelstore.c
+++ b/src/mesa/main/pixelstore.c
@@ -245,7 +245,8 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->Pack.ClientStorage = GL_FALSE;
    ctx->Pack.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->Pack.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Pack.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
    ctx->Unpack.Alignment = 4;
    ctx->Unpack.RowLength = 0;
@@ -258,7 +259,8 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->Unpack.ClientStorage = GL_FALSE;
    ctx->Unpack.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->Unpack.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 
    /*
@@ -278,6 +280,7 @@ _mesa_init_pixelstore( GLcontext *ctx )
    ctx->DefaultPacking.ClientStorage = GL_FALSE;
    ctx->DefaultPacking.Invert = GL_FALSE;
 #if FEATURE_EXT_pixel_buffer_object
-   ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx, &ctx->DefaultPacking.BufferObj,
+                                 ctx->Shared->NullBufferObj);
 #endif
 }
diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c
index c25b31af02..a73c6e0508 100644
--- a/src/mesa/main/queryobj.c
+++ b/src/mesa/main/queryobj.c
@@ -38,7 +38,7 @@
  * \param id - the new object's ID
  * \return pointer to new query_object object or NULL if out of memory.
  */
-struct gl_query_object *
+static struct gl_query_object *
 _mesa_new_query_object(GLcontext *ctx, GLuint id)
 {
    struct gl_query_object *q = MALLOC_STRUCT(gl_query_object);
@@ -57,7 +57,7 @@ _mesa_new_query_object(GLcontext *ctx, GLuint id)
  * Begin a query.  Software driver fallback.
  * Called via ctx->Driver.BeginQuery().
  */
-void
+static void
 _mesa_begin_query(GLcontext *ctx, struct gl_query_object *q)
 {
    /* no-op */
@@ -68,7 +68,7 @@ _mesa_begin_query(GLcontext *ctx, struct gl_query_object *q)
  * End a query.  Software driver fallback.
  * Called via ctx->Driver.EndQuery().
  */
-void
+static void
 _mesa_end_query(GLcontext *ctx, struct gl_query_object *q)
 {
    q->Ready = GL_TRUE;
@@ -79,7 +79,7 @@ _mesa_end_query(GLcontext *ctx, struct gl_query_object *q)
  * Wait for query to complete.  Software driver fallback.
  * Called via ctx->Driver.WaitQuery().
  */
-void
+static void
 _mesa_wait_query(GLcontext *ctx, struct gl_query_object *q)
 {
    /* For software drivers, _mesa_end_query() should have completed the query.
@@ -94,7 +94,7 @@ _mesa_wait_query(GLcontext *ctx, struct gl_query_object *q)
  * Check if a query results are ready.  Software driver fallback.
  * Called via ctx->Driver.CheckQuery().
  */
-void
+static void
 _mesa_check_query(GLcontext *ctx, struct gl_query_object *q)
 {
    /* No-op for sw rendering.
@@ -107,7 +107,7 @@ _mesa_check_query(GLcontext *ctx, struct gl_query_object *q)
  * Delete a query object.  Called via ctx->Driver.DeleteQuery().
  * Not removed from hash table here.
  */
-void
+static void
 _mesa_delete_query(GLcontext *ctx, struct gl_query_object *q)
 {
    _mesa_free(q);
@@ -123,6 +123,18 @@ lookup_query_object(GLcontext *ctx, GLuint id)
 
 
 
+void
+_mesa_init_query_object_functions(struct dd_function_table *driver)
+{
+   driver->NewQueryObject = _mesa_new_query_object;
+   driver->DeleteQuery = _mesa_delete_query;
+   driver->BeginQuery = _mesa_begin_query;
+   driver->EndQuery = _mesa_end_query;
+   driver->WaitQuery = _mesa_wait_query;
+   driver->CheckQuery = _mesa_check_query;
+}
+
+
 void GLAPIENTRY
 _mesa_GenQueriesARB(GLsizei n, GLuint *ids)
 {
diff --git a/src/mesa/main/queryobj.h b/src/mesa/main/queryobj.h
index bc02b65b54..ee775ef959 100644
--- a/src/mesa/main/queryobj.h
+++ b/src/mesa/main/queryobj.h
@@ -27,9 +27,6 @@
 #define OCCLUDE_H
 
 
-extern struct gl_query_object *
-_mesa_new_query_object(GLcontext *ctx, GLuint id);
-
 extern void
 _mesa_init_query(GLcontext *ctx);
 
@@ -37,19 +34,7 @@ extern void
 _mesa_free_query_data(GLcontext *ctx);
 
 extern void
-_mesa_delete_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_begin_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_end_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_wait_query(GLcontext *ctx, struct gl_query_object *q);
-
-extern void
-_mesa_check_query(GLcontext *ctx, struct gl_query_object *q);
+_mesa_init_query_object_functions(struct dd_function_table *driver);
 
 
 extern void GLAPIENTRY
diff --git a/src/mesa/main/shared.c b/src/mesa/main/shared.c
index ae8dd2a836..ad6e6ce7cd 100644
--- a/src/mesa/main/shared.c
+++ b/src/mesa/main/shared.c
@@ -33,6 +33,7 @@
 #include "mtypes.h"
 #include "hash.h"
 #include "arrayobj.h"
+#include "bufferobj.h"
 #include "shared.h"
 #include "shader/program.h"
 #include "shader/shader_api.h"
@@ -92,7 +93,12 @@ _mesa_alloc_shared_state(GLcontext *ctx)
    shared->BufferObjects = _mesa_NewHashTable();
 #endif
 
-   shared->ArrayObjects = _mesa_NewHashTable();
+   /* Allocate the default buffer object and set refcount so high that
+    * it never gets deleted.
+    * XXX with recent/improved refcounting this may not longer be needed.
+    */
+   shared->NullBufferObj = ctx->Driver.NewBufferObject(ctx, 0, 0);
+   shared->NullBufferObj->RefCount = 1000 * 1000 * 1000;
 
    /* Create default texture objects */
    for (i = 0; i < NUM_TEXTURE_TARGETS; i++) {
@@ -199,18 +205,6 @@ delete_bufferobj_cb(GLuint id, void *data, void *userData)
 
 
 /**
- * Callback for deleting an array object.  Called by _mesa_HashDeleteAll().
- */
-static void
-delete_arrayobj_cb(GLuint id, void *data, void *userData)
-{
-   struct gl_array_object *arrayObj = (struct gl_array_object *) data;
-   GLcontext *ctx = (GLcontext *) userData;
-   _mesa_delete_array_object(ctx, arrayObj);
-}
-
-
-/**
  * Callback for freeing shader program data. Call it before delete_shader_cb
  * to avoid memory access error.
  */
@@ -312,9 +306,6 @@ _mesa_free_shared_state(GLcontext *ctx, struct gl_shared_state *shared)
    _mesa_HashDeleteAll(shared->Programs, delete_program_cb, ctx);
    _mesa_DeleteHashTable(shared->Programs);
 
-   _mesa_HashDeleteAll(shared->ArrayObjects, delete_arrayobj_cb, ctx);
-   _mesa_DeleteHashTable(shared->ArrayObjects);
-
 #if FEATURE_ARB_vertex_program
    _mesa_reference_vertprog(ctx, &shared->DefaultVertexProgram, NULL);
 #endif
@@ -341,6 +332,10 @@ _mesa_free_shared_state(GLcontext *ctx, struct gl_shared_state *shared)
    _mesa_DeleteHashTable(shared->RenderBuffers);
 #endif
 
+#if FEATURE_ARB_vertex_buffer_object
+   ctx->Driver.DeleteBuffer(ctx, shared->NullBufferObj);
+#endif
+
    /*
     * Free texture objects (after FBOs since some textures might have
     * been bound to FBOs).
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index 1d3758a302..d8191ab518 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -64,114 +64,159 @@ update_separate_specular(GLcontext *ctx)
 
 
 /**
- * Update state dependent on vertex arrays.
+ * Compute the index of the last array element that can be safely accessed
+ * in a vertex array.  We can really only do this when the array lives in
+ * a VBO.
+ * The array->_MaxElement field will be updated.
+ * Later in glDrawArrays/Elements/etc we can do some bounds checking.
+ */
+static void
+compute_max_element(struct gl_client_array *array)
+{
+   assert(array->Enabled);
+   if (array->BufferObj->Name) {
+      GLsizeiptrARB offset = (GLsizeiptrARB) array->Ptr;
+      GLsizeiptrARB obj_size = (GLsizeiptrARB) array->BufferObj->Size;
+
+      if (offset < obj_size) {
+	 array->_MaxElement = (obj_size - offset +
+			       array->StrideB -
+			       array->_ElementSize) / array->StrideB;
+      } else {
+	 array->_MaxElement = 0;
+      }
+      /* Compute the max element we can access in the VBO without going
+       * out of bounds.
+       */
+      array->_MaxElement = ((GLsizeiptrARB) array->BufferObj->Size
+                            - (GLsizeiptrARB) array->Ptr + array->StrideB
+                            - array->_ElementSize) / array->StrideB;
+   }
+   else {
+      /* user-space array, no idea how big it is */
+      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
+   }
+}
+
+
+/**
+ * Helper for update_arrays().
+ * \return  min(current min, array->_MaxElement).
+ */
+static GLuint
+update_min(GLuint min, struct gl_client_array *array)
+{
+   compute_max_element(array);
+   return MIN2(min, array->_MaxElement);
+}
+
+
+/**
+ * Update ctx->Array._MaxElement (the max legal index into all enabled arrays).
+ * Need to do this upon new array state or new buffer object state.
  */
 static void
 update_arrays( GLcontext *ctx )
 {
-   GLuint i, min;
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   GLuint i, min = ~0;
 
    /* find min of _MaxElement values for all enabled arrays */
 
    /* 0 */
    if (ctx->VertexProgram._Current
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_POS].Enabled) {
-      min = ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_POS]._MaxElement;
+       && arrayObj->VertexAttrib[VERT_ATTRIB_POS].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_POS]);
    }
-   else if (ctx->Array.ArrayObj->Vertex.Enabled) {
-      min = ctx->Array.ArrayObj->Vertex._MaxElement;
-   }
-   else {
-      /* can't draw anything without vertex positions! */
-      min = 0;
+   else if (arrayObj->Vertex.Enabled) {
+      min = update_min(min, &arrayObj->Vertex);
    }
 
    /* 1 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_WEIGHT]);
    }
    /* no conventional vertex weight array */
 
    /* 2 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_NORMAL].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_NORMAL]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_NORMAL].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_NORMAL]);
    }
-   else if (ctx->Array.ArrayObj->Normal.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Normal._MaxElement);
+   else if (arrayObj->Normal.Enabled) {
+      min = update_min(min, &arrayObj->Normal);
    }
 
    /* 3 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR0].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR0]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR0].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR0]);
    }
-   else if (ctx->Array.ArrayObj->Color.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Color._MaxElement);
+   else if (arrayObj->Color.Enabled) {
+      min = update_min(min, &arrayObj->Color);
    }
 
    /* 4 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR1].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR1]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR1].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR1]);
    }
-   else if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->SecondaryColor._MaxElement);
+   else if (arrayObj->SecondaryColor.Enabled) {
+      min = update_min(min, &arrayObj->SecondaryColor);
    }
 
    /* 5 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_FOG].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_FOG]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_FOG].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_FOG]);
    }
-   else if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->FogCoord._MaxElement);
+   else if (arrayObj->FogCoord.Enabled) {
+      min = update_min(min, &arrayObj->FogCoord);
    }
 
    /* 6 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_COLOR_INDEX]);
    }
-   else if (ctx->Array.ArrayObj->Index.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->Index._MaxElement);
+   else if (arrayObj->Index.Enabled) {
+      min = update_min(min, &arrayObj->Index);
    }
 
-
    /* 7 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]._MaxElement);
+       && arrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
+      min = update_min(min, &arrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]);
    }
 
    /* 8..15 */
    for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++) {
       if (ctx->VertexProgram._Enabled
-          && ctx->Array.ArrayObj->VertexAttrib[i].Enabled) {
-         min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[i]._MaxElement);
+          && arrayObj->VertexAttrib[i].Enabled) {
+         min = update_min(min, &arrayObj->VertexAttrib[i]);
       }
       else if (i - VERT_ATTRIB_TEX0 < ctx->Const.MaxTextureCoordUnits
-               && ctx->Array.ArrayObj->TexCoord[i - VERT_ATTRIB_TEX0].Enabled) {
-         min = MIN2(min, ctx->Array.ArrayObj->TexCoord[i - VERT_ATTRIB_TEX0]._MaxElement);
+               && arrayObj->TexCoord[i - VERT_ATTRIB_TEX0].Enabled) {
+         min = update_min(min, &arrayObj->TexCoord[i - VERT_ATTRIB_TEX0]);
       }
    }
 
    /* 16..31 */
    if (ctx->VertexProgram._Current) {
-      for (i = 0; i < Elements(ctx->Array.ArrayObj->VertexAttrib); i++) {
-         if (ctx->Array.ArrayObj->VertexAttrib[i].Enabled) {
-            min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[i]._MaxElement);
+      for (i = 0; i < Elements(arrayObj->VertexAttrib); i++) {
+         if (arrayObj->VertexAttrib[i].Enabled) {
+            min = update_min(min, &arrayObj->VertexAttrib[i]);
          }
       }
    }
 
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->EdgeFlag._MaxElement);
+   if (arrayObj->EdgeFlag.Enabled) {
+      min = update_min(min, &arrayObj->EdgeFlag);
    }
 
    /* _MaxElement is one past the last legal array element */
-   ctx->Array._MaxElement = min;
+   arrayObj->_MaxElement = min;
 }
 
 
@@ -501,7 +546,7 @@ _mesa_update_state_locked( GLcontext *ctx )
 
    /* Determine which state flags effect vertex/fragment program state */
    if (ctx->FragmentProgram._MaintainTexEnvProgram) {
-      prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR |
+      prog_flags |= (_NEW_TEXTURE | _NEW_FOG |
 		     _NEW_ARRAY | _NEW_LIGHT | _NEW_POINT | _NEW_RENDERMODE);
    }
    if (ctx->VertexProgram._MaintainTnlProgram) {
@@ -547,7 +592,7 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _DD_NEW_SEPARATE_SPECULAR)
       update_separate_specular( ctx );
 
-   if (new_state & (_NEW_ARRAY | _NEW_PROGRAM))
+   if (new_state & (_NEW_ARRAY | _NEW_PROGRAM | _NEW_BUFFER_OBJECT))
       update_arrays( ctx );
 
    if (new_state & (_NEW_BUFFERS | _NEW_VIEWPORT))
diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c
index a3f1246c98..6b090ff399 100644
--- a/src/mesa/main/texenvprogram.c
+++ b/src/mesa/main/texenvprogram.c
@@ -62,6 +62,18 @@ struct texenvprog_cache_item
    struct texenvprog_cache_item *next;
 };
 
+static GLboolean
+texenv_doing_secondary_color(GLcontext *ctx)
+{
+   if (ctx->Light.Enabled &&
+       (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR))
+      return GL_TRUE;
+
+   if (ctx->Fog.ColorSumEnabled)
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
 
 /**
  * Up to nine instructions per tex unit, plus fog, specular color.
@@ -298,7 +310,7 @@ static GLbitfield get_fp_input_mask( GLcontext *ctx )
       if (ctx->Light.Enabled) {
          fp_inputs |= FRAG_BIT_COL0;
 
-         if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+         if (texenv_doing_secondary_color(ctx))
             fp_inputs |= FRAG_BIT_COL1;
       }
 
@@ -416,8 +428,8 @@ static void make_state_key( GLcontext *ctx,  struct state_key *key )
        }
    }
 
-   /* _DD_NEW_SEPARATE_SPECULAR */
-   if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
+   /* _NEW_LIGHT | _NEW_FOG */
+   if (texenv_doing_secondary_color(ctx)) {
       key->separate_specular = 1;
       inputs_referenced |= FRAG_BIT_COL1;
    }
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index 89ee69770e..2082f945f1 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -281,7 +281,8 @@ valid_texture_object(const struct gl_texture_object *tex)
       _mesa_problem(NULL, "invalid reference to a deleted texture object");
       return GL_FALSE;
    default:
-      _mesa_problem(NULL, "invalid texture object Target value");
+      _mesa_problem(NULL, "invalid texture object Target 0x%x, Id = %u",
+                    tex->Target, tex->Name);
       return GL_FALSE;
    }
 }
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index e60ab6aa12..d27c59381c 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -469,8 +469,10 @@ set_tex_parameterf(GLcontext *ctx,
          return GL_TRUE;
       }
       else {
-         _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
+         static GLuint count = 0;
+         if (count++ < 10)
+            _mesa_error(ctx, GL_INVALID_ENUM,
+                        "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
       }
       return GL_FALSE;
 
@@ -880,7 +882,8 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
                         "glGetTexLevelParameter[if]v(pname)");
          break;
       case GL_TEXTURE_STENCIL_SIZE_EXT:
-         if (ctx->Extensions.EXT_packed_depth_stencil) {
+         if (ctx->Extensions.EXT_packed_depth_stencil ||
+             ctx->Extensions.ARB_framebuffer_object) {
             *params = img->TexFormat->StencilBits;
          }
          else {
diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c
index 8f6621ee45..3d5b8faecf 100644
--- a/src/mesa/main/varray.c
+++ b/src/mesa/main/varray.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,6 +30,7 @@
 #include "context.h"
 #include "enable.h"
 #include "enums.h"
+#include "hash.h"
 #include "mtypes.h"
 #include "varray.h"
 #include "arrayobj.h"
@@ -36,26 +38,39 @@
 
 
 /**
- * Update the fields of a vertex array object.
- * We need to do a few special things for arrays that live in
- * vertex buffer objects.
+ * Set the fields of a vertex array.
+ * Also do an error check for GL_ARB_vertex_array_object: check that
+ * all arrays reside in VBOs when using a vertex array object.
  *
  * \param array  the array to update
  * \param dirtyBit  which bit to set in ctx->Array.NewState for this array
  * \param elementSize  size of each array element, in bytes
  * \param size  components per element (1, 2, 3 or 4)
  * \param type  datatype of each component (GL_FLOAT, GL_INT, etc)
+ * \param format  either GL_RGBA or GL_BGRA
  * \param stride  stride between elements, in elements
  * \param normalized  are integer types converted to floats in [-1, 1]?
  * \param ptr  the address (or offset inside VBO) of the array data
+ * \return GL_TRUE if no error, GL_FALSE if error
  */
-static void
+static GLboolean
 update_array(GLcontext *ctx, struct gl_client_array *array,
              GLbitfield dirtyBit, GLsizei elementSize,
              GLint size, GLenum type, GLenum format,
              GLsizei stride, GLboolean normalized, const GLvoid *ptr)
 {
    ASSERT(format == GL_RGBA || format == GL_BGRA);
+
+   if (ctx->Array.ArrayObj->VBOonly &&
+       ctx->Array.ArrayBufferObj->Name == 0) {
+      /* GL_ARB_vertex_array_object requires that all arrays reside in VBOs.
+       * Generate GL_INVALID_OPERATION if that's not true.
+       */
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glVertex/Normal/EtcPointer(non-VBO array)");
+      return GL_FALSE;
+   }
+
    array->Size = size;
    array->Type = type;
    array->Format = format;
@@ -63,24 +78,15 @@ update_array(GLcontext *ctx, struct gl_client_array *array,
    array->StrideB = stride ? stride : elementSize;
    array->Normalized = normalized;
    array->Ptr = (const GLubyte *) ptr;
-#if FEATURE_ARB_vertex_buffer_object
+   array->_ElementSize = elementSize;
+
    _mesa_reference_buffer_object(ctx, &array->BufferObj,
                                  ctx->Array.ArrayBufferObj);
 
-   /* Compute the index of the last array element that's inside the buffer.
-    * Later in glDrawArrays we'll check if start + count > _MaxElement to
-    * be sure we won't go out of bounds.
-    */
-   if (ctx->Array.ArrayBufferObj->Name)
-      array->_MaxElement = ((GLsizeiptrARB) ctx->Array.ArrayBufferObj->Size
-                            - (GLsizeiptrARB) array->Ptr + array->StrideB
-                            - elementSize) / array->StrideB;
-   else
-#endif
-      array->_MaxElement = 2 * 1000 * 1000 * 1000; /* just a big number */
-
    ctx->NewState |= _NEW_ARRAY;
    ctx->Array.NewState |= dirtyBit;
+
+   return GL_TRUE;
 }
 
 
@@ -133,8 +139,9 @@ _mesa_VertexPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Vertex, _NEW_ARRAY_VERTEX,
-                elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Vertex, _NEW_ARRAY_VERTEX,
+                     elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.VertexPointer)
       ctx->Driver.VertexPointer( ctx, size, type, stride, ptr );
@@ -183,8 +190,9 @@ _mesa_NormalPointer(GLenum type, GLsizei stride, const GLvoid *ptr )
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Normal, _NEW_ARRAY_NORMAL,
-                elementSize, 3, type, GL_RGBA, stride, GL_TRUE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Normal, _NEW_ARRAY_NORMAL,
+                     elementSize, 3, type, GL_RGBA, stride, GL_TRUE, ptr))
+      return;
 
    if (ctx->Driver.NormalPointer)
       ctx->Driver.NormalPointer( ctx, type, stride, ptr );
@@ -261,8 +269,9 @@ _mesa_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Color, _NEW_ARRAY_COLOR0,
-                elementSize, size, type, format, stride, GL_TRUE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Color, _NEW_ARRAY_COLOR0,
+                     elementSize, size, type, format, stride, GL_TRUE, ptr))
+      return;
 
    if (ctx->Driver.ColorPointer)
       ctx->Driver.ColorPointer( ctx, size, type, stride, ptr );
@@ -293,8 +302,9 @@ _mesa_FogCoordPointerEXT(GLenum type, GLsizei stride, const GLvoid *ptr)
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->FogCoord, _NEW_ARRAY_FOGCOORD,
-                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->FogCoord, _NEW_ARRAY_FOGCOORD,
+                     elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.FogCoordPointer)
       ctx->Driver.FogCoordPointer( ctx, type, stride, ptr );
@@ -334,8 +344,9 @@ _mesa_IndexPointer(GLenum type, GLsizei stride, const GLvoid *ptr)
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->Index, _NEW_ARRAY_INDEX,
-                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->Index, _NEW_ARRAY_INDEX,
+                     elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.IndexPointer)
       ctx->Driver.IndexPointer( ctx, type, stride, ptr );
@@ -408,8 +419,10 @@ _mesa_SecondaryColorPointerEXT(GLint size, GLenum type,
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->SecondaryColor, _NEW_ARRAY_COLOR1,
-                elementSize, size, type, format, stride, GL_TRUE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->SecondaryColor,
+                     _NEW_ARRAY_COLOR1, elementSize, size, type,
+                     format, stride, GL_TRUE, ptr))
+      return;
 
    if (ctx->Driver.SecondaryColorPointer)
       ctx->Driver.SecondaryColorPointer( ctx, size, type, stride, ptr );
@@ -467,9 +480,10 @@ _mesa_TexCoordPointer(GLint size, GLenum type, GLsizei stride,
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->TexCoord[unit],
-                _NEW_ARRAY_TEXCOORD(unit),
-                elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->TexCoord[unit],
+                     _NEW_ARRAY_TEXCOORD(unit),
+                     elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.TexCoordPointer)
       ctx->Driver.TexCoordPointer( ctx, size, type, stride, ptr );
@@ -487,9 +501,10 @@ _mesa_EdgeFlagPointer(GLsizei stride, const GLvoid *ptr)
       return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->EdgeFlag, _NEW_ARRAY_EDGEFLAG,
-                sizeof(GLboolean), 1, GL_UNSIGNED_BYTE, GL_RGBA,
-                stride, GL_FALSE, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->EdgeFlag, _NEW_ARRAY_EDGEFLAG,
+                     sizeof(GLboolean), 1, GL_UNSIGNED_BYTE, GL_RGBA,
+                     stride, GL_FALSE, ptr))
+      return;
 
    if (ctx->Driver.EdgeFlagPointer)
       ctx->Driver.EdgeFlagPointer( ctx, stride, ptr );
@@ -528,6 +543,12 @@ _mesa_PointSizePointer(GLenum type, GLsizei stride, const GLvoid *ptr)
 
 
 #if FEATURE_NV_vertex_program
+/**
+ * Set a vertex attribute array.
+ * Note that these arrays DO alias the conventional GL vertex arrays
+ * (position, normal, color, fog, texcoord, etc).
+ * The generic attribute slots at #16 and above are not touched.
+ */
 void GLAPIENTRY
 _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
                             GLsizei stride, const GLvoid *ptr)
@@ -593,9 +614,10 @@ _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
-                _NEW_ARRAY_ATTRIB(index),
-                elementSize, size, type, format, stride, normalized, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
+                     _NEW_ARRAY_ATTRIB(index),
+                     elementSize, size, type, format, stride, normalized, ptr))
+      return;
 
    if (ctx->Driver.VertexAttribPointer)
       ctx->Driver.VertexAttribPointer( ctx, index, size, type, stride, ptr );
@@ -604,6 +626,11 @@ _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
 
 
 #if FEATURE_ARB_vertex_program
+/**
+ * Set a generic vertex attribute array.
+ * Note that these arrays DO NOT alias the conventional GL vertex arrays
+ * (position, normal, color, fog, texcoord, etc).
+ */
 void GLAPIENTRY
 _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
                              GLboolean normalized,
@@ -687,9 +714,10 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
          return;
    }
 
-   update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
-                _NEW_ARRAY_ATTRIB(index),
-                elementSize, size, type, format, stride, normalized, ptr);
+   if (!update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
+                     _NEW_ARRAY_ATTRIB(index),
+                     elementSize, size, type, format, stride, normalized, ptr))
+      return;
 
    if (ctx->Driver.VertexAttribPointer)
       ctx->Driver.VertexAttribPointer(ctx, index, size, type, stride, ptr);
@@ -1065,6 +1093,52 @@ _mesa_MultiModeDrawElementsIBM( const GLenum * mode, const GLsizei * count,
 
 
 /**
+ * Print vertex array's fields.
+ */
+static void
+print_array(const char *name, GLint index, const struct gl_client_array *array)
+{
+   if (index >= 0)
+      _mesa_printf("  %s[%d]: ", name, index);
+   else
+      _mesa_printf("  %s: ", name);
+   _mesa_printf("Ptr=%p, Type=0x%x, Size=%d, ElemSize=%u, Stride=%d, Buffer=%u(Size %u), MaxElem=%u\n",
+                array->Ptr, array->Type, array->Size,
+                array->_ElementSize, array->StrideB,
+                array->BufferObj->Name, array->BufferObj->Size,
+                array->_MaxElement);
+}
+
+
+/**
+ * Print current vertex object/array info.  For debug.
+ */
+void
+_mesa_print_arrays(GLcontext *ctx)
+{
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   GLuint i;
+
+   _mesa_update_array_object_max_element(ctx, arrayObj);
+
+   _mesa_printf("Array Object %u\n", arrayObj->Name);
+   if (arrayObj->Vertex.Enabled)
+      print_array("Vertex", -1, &arrayObj->Vertex);
+   if (arrayObj->Normal.Enabled)
+      print_array("Normal", -1, &arrayObj->Normal);
+   if (arrayObj->Color.Enabled)
+      print_array("Color", -1, &arrayObj->Color);
+   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
+      if (arrayObj->TexCoord[i].Enabled)
+         print_array("TexCoord", i, &arrayObj->TexCoord[i]);
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++)
+      if (arrayObj->VertexAttrib[i].Enabled)
+         print_array("Attrib", i, &arrayObj->VertexAttrib[i]);
+   _mesa_printf("  _MaxElement = %u\n", arrayObj->_MaxElement);
+}
+
+
+/**
  * Initialize vertex array state for given context.
  */
 void 
@@ -1074,4 +1148,29 @@ _mesa_init_varray(GLcontext *ctx)
    _mesa_reference_array_object(ctx, &ctx->Array.ArrayObj,
                                 ctx->Array.DefaultArrayObj);
    ctx->Array.ActiveTexture = 0;   /* GL_ARB_multitexture */
+
+   ctx->Array.Objects = _mesa_NewHashTable();
+}
+
+
+/**
+ * Callback for deleting an array object.  Called by _mesa_HashDeleteAll().
+ */
+static void
+delete_arrayobj_cb(GLuint id, void *data, void *userData)
+{
+   struct gl_array_object *arrayObj = (struct gl_array_object *) data;
+   GLcontext *ctx = (GLcontext *) userData;
+   _mesa_delete_array_object(ctx, arrayObj);
+}
+
+
+/**
+ * Free vertex array state for given context.
+ */
+void 
+_mesa_free_varray_data(GLcontext *ctx)
+{
+   _mesa_HashDeleteAll(ctx->Array.Objects, delete_arrayobj_cb, ctx);
+   _mesa_DeleteHashTable(ctx->Array.Objects);
 }
diff --git a/src/mesa/main/varray.h b/src/mesa/main/varray.h
index 97d5c8219d..d4d505ae04 100644
--- a/src/mesa/main/varray.h
+++ b/src/mesa/main/varray.h
@@ -1,18 +1,9 @@
-/**
- * \file varray.h
- * Vertex arrays.
- *
- * \if subset
- * (No-op)
- *
- * \endif
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  4.1
+ * Version:  7.6
  *
- * Copyright (C) 1999-2002  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -169,14 +160,20 @@ _mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
                         GLenum type, const GLvoid *indices);
 
 
+extern void
+_mesa_print_arrays(GLcontext *ctx);
 
 extern void
 _mesa_init_varray( GLcontext * ctx );
 
+extern void 
+_mesa_free_varray_data(GLcontext *ctx);
+
 #else
 
 /** No-op */
 #define _mesa_init_varray( c )  ((void)0)
+#define _mesa_free_varray_data( c )  ((void)0)
 
 #endif
 
diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h
index ba027465d4..d4d3dd1a94 100644
--- a/src/mesa/main/version.h
+++ b/src/mesa/main/version.h
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.5
+ * Version:  7.6
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,9 +30,9 @@
 
 /* Mesa version */
 #define MESA_MAJOR 7
-#define MESA_MINOR 5
+#define MESA_MINOR 6
 #define MESA_PATCH 0
-#define MESA_VERSION_STRING "7.5"
+#define MESA_VERSION_STRING "7.6-devel"
 
 /* To make version comparison easy */
 #define MESA_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
diff --git a/src/mesa/math/m_vector.c b/src/mesa/math/m_vector.c
index c5e2fd1de1..4cbab11a35 100644
--- a/src/mesa/math/m_vector.c
+++ b/src/mesa/math/m_vector.c
@@ -1,4 +1,3 @@
-
 /*
  * Mesa 3-D graphics library
  * Version:  3.5
@@ -37,11 +36,12 @@
 
 
 
-/*
+/**
  * Given a vector [count][4] of floats, set all the [][elt] values
  * to 0 (if elt = 0, 1, 2) or 1.0 (if elt = 3).
  */
-void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
+void
+_mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
 {
    static const GLubyte elem_bits[4] = {
       VEC_DIRTY_0,
@@ -54,12 +54,13 @@ void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint count, GLuint elt )
    GLfloat (*data)[4] = (GLfloat (*)[4])vec->start;
    GLuint i;
 
-   for (i = 0 ; i < count ; i++)
+   for (i = 0; i < count; i++)
       data[i][elt] = v;
 
    vec->flags &= ~elem_bits[elt];
 }
 
+
 static const GLubyte size_bits[5] = {
    0,
    VEC_SIZE_1,
@@ -69,61 +70,53 @@ static const GLubyte size_bits[5] = {
 };
 
 
-
-/*
+/**
  * Initialize GLvector objects.
- * Input: v - the vector object to initialize.
- *        flags - bitwise-OR of VEC_* flags
- *        storage - pointer to storage for the vector's data
+ * \param v  the vector object to initialize.
+ * \param flags  bitwise-OR of VEC_* flags
+ * \param storage  pointer to storage for the vector's data
  */
-
-
-void _mesa_vector4f_init( GLvector4f *v, GLuint flags, GLfloat (*storage)[4] )
+void
+_mesa_vector4f_init( GLvector4f *v, GLbitfield flags, GLfloat (*storage)[4] )
 {
    v->stride = 4 * sizeof(GLfloat);
    v->size = 2;   /* may change: 2-4 for vertices and 1-4 for texcoords */
    v->data = storage;
    v->start = (GLfloat *) storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags ;
+   v->flags = size_bits[4] | flags;
 }
 
 
-
-
-/*
+/**
  * Initialize GLvector objects and allocate storage.
- * Input: v - the vector object
- *        sz - unused????
- *        flags - bitwise-OR of VEC_* flags
- *        count - number of elements to allocate in vector
- *        alignment - desired memory alignment for the data (in bytes)
+ * \param v  the vector object
+ * \param flags  bitwise-OR of VEC_* flags
+ * \param count  number of elements to allocate in vector
+ * \param alignment  desired memory alignment for the data (in bytes)
  */
-
-
-void _mesa_vector4f_alloc( GLvector4f *v, GLuint flags, GLuint count,
-			GLuint alignment )
+void
+_mesa_vector4f_alloc( GLvector4f *v, GLbitfield flags, GLuint count,
+                      GLuint alignment )
 {
    v->stride = 4 * sizeof(GLfloat);
    v->size = 2;
    v->storage = ALIGN_MALLOC( count * 4 * sizeof(GLfloat), alignment );
+   v->storage_count = count;
    v->start = (GLfloat *) v->storage;
    v->data = (GLfloat (*)[4]) v->storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags | VEC_MALLOC ;
+   v->flags = size_bits[4] | flags | VEC_MALLOC;
 }
 
 
-
-
-/*
+/**
  * Vector deallocation.  Free whatever memory is pointed to by the
  * vector's storage field if the VEC_MALLOC flag is set.
  * DO NOT free the GLvector object itself, though.
  */
-
-
-void _mesa_vector4f_free( GLvector4f *v )
+void
+_mesa_vector4f_free( GLvector4f *v )
 {
    if (v->flags & VEC_MALLOC) {
       ALIGN_FREE( v->storage );
@@ -135,13 +128,15 @@ void _mesa_vector4f_free( GLvector4f *v )
 }
 
 
-/*
+/**
  * For debugging
  */
-void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
+void
+_mesa_vector4f_print( const GLvector4f *v, const GLubyte *cullmask,
+                      GLboolean culling )
 {
-   GLfloat c[4] = { 0, 0, 0, 1 };
-   const char *templates[5] = {
+   static const GLfloat c[4] = { 0, 0, 0, 1 };
+   static const char *templates[5] = {
       "%d:\t0, 0, 0, 1\n",
       "%d:\t%f, 0, 0, 1\n",
       "%d:\t%f, %f, 0, 1\n",
@@ -154,30 +149,32 @@ void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
    GLuint j, i = 0, count;
 
    _mesa_printf("data-start\n");
-   for ( ; d != v->start ; STRIDE_F(d, v->stride), i++)
+   for (; d != v->start; STRIDE_F(d, v->stride), i++)
       _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
 
    _mesa_printf("start-count(%u)\n", v->count);
    count = i + v->count;
 
    if (culling) {
-      for ( ; i < count ; STRIDE_F(d, v->stride), i++)
+      for (; i < count; STRIDE_F(d, v->stride), i++)
 	 if (cullmask[i])
 	    _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
    }
    else {
-      for ( ; i < count ; STRIDE_F(d, v->stride), i++)
+      for (; i < count; STRIDE_F(d, v->stride), i++)
 	 _mesa_printf(t, i, d[0], d[1], d[2], d[3]);
    }
 
-   for (j = v->size ; j < 4; j++) {
+   for (j = v->size; j < 4; j++) {
       if ((v->flags & (1<<j)) == 0) {
 
 	 _mesa_printf("checking col %u is clean as advertised ", j);
 
-	 for (i = 0, d = (GLfloat *) v->data ;
-	      i < count && d[j] == c[j] ;
-	      i++, STRIDE_F(d, v->stride)) {};
+	 for (i = 0, d = (GLfloat *) v->data;
+	      i < count && d[j] == c[j];
+	      i++, STRIDE_F(d, v->stride)) {
+            /* no-op */
+         }
 
 	 if (i == count)
 	    _mesa_printf(" --> ok\n");
@@ -186,5 +183,3 @@ void _mesa_vector4f_print( GLvector4f *v, GLubyte *cullmask, GLboolean culling )
       }
    }
 }
-
-
diff --git a/src/mesa/math/m_vector.h b/src/mesa/math/m_vector.h
index 647388ac7d..71281d5758 100644
--- a/src/mesa/math/m_vector.h
+++ b/src/mesa/math/m_vector.h
@@ -31,7 +31,6 @@
 #define _M_VECTOR_H_
 
 #include "main/glheader.h"
-#include "main/mtypes.h"		/* hack for GLchan */
 
 
 #define VEC_DIRTY_0        0x1
@@ -50,7 +49,8 @@
 
 
 
-/* Wrap all the information about vectors up in a struct.  Has
+/**
+ * Wrap all the information about vectors up in a struct.  Has
  * additional fields compared to the other vectors to help us track of
  * different vertex sizes, and whether we need to clean columns out
  * because they contain non-(0,0,0,1) values.
@@ -60,29 +60,27 @@
  * the transformation routines.
  */
 typedef struct {
-   GLfloat (*data)[4];	/* may be malloc'd or point to client data */
-   GLfloat *start;	/* points somewhere inside of <data> */
-   GLuint count;	/* size of the vector (in elements) */
-   GLuint stride;	/* stride from one element to the next (in bytes) */
-   GLuint size;		/* 2-4 for vertices and 1-4 for texcoords */
-   GLuint flags;	/* which columns are dirty */
-   void *storage;	/* self-allocated storage */
+   GLfloat (*data)[4];	/**< may be malloc'd or point to client data */
+   GLfloat *start;	/**< points somewhere inside of <data> */
+   GLuint count;	/**< size of the vector (in elements) */
+   GLuint stride;	/**< stride from one element to the next (in bytes) */
+   GLuint size;		/**< 2-4 for vertices and 1-4 for texcoords */
+   GLbitfield flags;	/**< bitmask of VEC_x flags */
+   void *storage;	/**< self-allocated storage */
+   GLuint storage_count; /**< storage size in elements */
 } GLvector4f;
 
 
-extern void _mesa_vector4f_init( GLvector4f *v, GLuint flags,
+extern void _mesa_vector4f_init( GLvector4f *v, GLbitfield flags,
 			      GLfloat (*storage)[4] );
-extern void _mesa_vector4f_alloc( GLvector4f *v, GLuint flags,
+extern void _mesa_vector4f_alloc( GLvector4f *v, GLbitfield flags,
 			       GLuint count, GLuint alignment );
 extern void _mesa_vector4f_free( GLvector4f *v );
-extern void _mesa_vector4f_print( GLvector4f *v, GLubyte *, GLboolean );
+extern void _mesa_vector4f_print( const GLvector4f *v, const GLubyte *, GLboolean );
 extern void _mesa_vector4f_clean_elem( GLvector4f *vec, GLuint nr, GLuint elt );
 
 
-
-
-
-/*
+/**
  * Given vector <v>, return a pointer (cast to <type *> to the <i>-th element.
  *
  * End up doing a lot of slow imuls if not careful.
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 0d4f13c970..bc65aba39a 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -563,6 +563,7 @@ struct var_cache
                                  * we take up with our state tokens or constants. Note that
                                  * this is _not_ the same as the number of param registers
                                  * we eventually use */
+   GLuint swizzle;              /**< swizzle to access this variable */
    struct var_cache *next;
 };
 
@@ -581,6 +582,7 @@ var_cache_create (struct var_cache **va)
       (**va).param_binding_begin = ~0;
       (**va).param_binding_length = ~0;
       (**va).alias_binding = NULL;
+      (**va).swizzle = SWIZZLE_XYZW;
       (**va).next = NULL;
    }
 }
@@ -872,15 +874,16 @@ parse_signed_float (const GLubyte ** inst, struct arb_program *Program)
  * This picks out a constant value from the parsed array. The constant vector is r
  * returned in the *values array, which should be of length 4.
  *
- * \param values - The 4 component vector with the constant value in it
+ * \param values - return the vector constant values.
+ * \param size - returns the number elements in valuesOut [1..4]
  */
 static GLvoid
-parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Program,
-                GLboolean use)
+parse_constant(const GLubyte ** inst, GLfloat *values, GLint *size,
+               struct arb_program *Program,
+               GLboolean use)
 {
    GLuint components, i;
 
-
    switch (*(*inst)++) {
       case CONSTANT_SCALAR:
          if (use == GL_TRUE) {
@@ -893,7 +896,7 @@ parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Prog
                values[1] =
                values[2] = values[3] = parse_signed_float (inst, Program);
          }
-
+         *size = 1;
          break;
       case CONSTANT_VECTOR:
          values[0] = values[1] = values[2] = 0;
@@ -902,7 +905,12 @@ parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Prog
          for (i = 0; i < components; i++) {
             values[i] = parse_signed_float (inst, Program);
          }
+         *size = 4;
          break;
+      default:
+         _mesa_problem(NULL, "unexpected case in parse_constant()");
+         values[0] = 0.0F;
+         *size = 0;
    }
 }
 
@@ -945,7 +953,7 @@ parse_generic_attrib_num(GLcontext *ctx, const GLubyte ** inst,
 {
    GLint i = parse_integer(inst, Program);
 
-   if ((i < 0) || (i >= MAX_VERTEX_PROGRAM_ATTRIBS))
+   if ((i < 0) || (i >= MAX_VERTEX_GENERIC_ATTRIBS))
    {
       program_error(ctx, Program->Position,
                     "Invalid generic vertex attribute index");
@@ -1505,10 +1513,10 @@ generic_attrib_check(struct var_cache *vc_head)
 {
    int a;
    struct var_cache *curr;
-   GLboolean explicitAttrib[MAX_VERTEX_PROGRAM_ATTRIBS],
-      genericAttrib[MAX_VERTEX_PROGRAM_ATTRIBS];
+   GLboolean explicitAttrib[MAX_VERTEX_GENERIC_ATTRIBS],
+      genericAttrib[MAX_VERTEX_GENERIC_ATTRIBS];
 
-   for (a=0; a<MAX_VERTEX_PROGRAM_ATTRIBS; a++) {
+   for (a=0; a<MAX_VERTEX_GENERIC_ATTRIBS; a++) {
       explicitAttrib[a] = GL_FALSE;
       genericAttrib[a] = GL_FALSE;
    }
@@ -1519,11 +1527,11 @@ generic_attrib_check(struct var_cache *vc_head)
          if (curr->attrib_is_generic) {
             GLuint attr = (curr->attrib_binding == 0)
                ? 0 : (curr->attrib_binding - VERT_ATTRIB_GENERIC0);
-            assert(attr < MAX_VERTEX_PROGRAM_ATTRIBS);
+            assert(attr < MAX_VERTEX_GENERIC_ATTRIBS);
             genericAttrib[attr] = GL_TRUE;
          }
          else {
-            assert(curr->attrib_binding < MAX_VERTEX_PROGRAM_ATTRIBS);
+            assert(curr->attrib_binding < MAX_VERTEX_GENERIC_ATTRIBS);
             explicitAttrib[ curr->attrib_binding ] = GL_TRUE;
          }
       }
@@ -1531,7 +1539,7 @@ generic_attrib_check(struct var_cache *vc_head)
       curr = curr->next;
    }
 
-   for (a=0; a<MAX_VERTEX_PROGRAM_ATTRIBS; a++) {
+   for (a=0; a<MAX_VERTEX_GENERIC_ATTRIBS; a++) {
       if ((explicitAttrib[a]) && (genericAttrib[a]))
          return 1;
    }
@@ -1819,7 +1827,6 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
    GLint idx;
    GLuint err = 0;
    gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0};
-   GLfloat const_values[4];
 
    GLubyte token = *(*inst)++;
 
@@ -1911,18 +1918,31 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
 
       case PARAM_CONSTANT:
          /* parsing something like {1.0, 2.0, 3.0, 4.0} */
-         parse_constant (inst, const_values, Program, use);
-         idx = _mesa_add_named_constant(Program->Base.Parameters,
-                                        (char *) param_var->name,
-                                        const_values, 4);
-         if (param_var->param_binding_begin == ~0U)
-            param_var->param_binding_begin = idx;
-         param_var->param_binding_type = PROGRAM_STATE_VAR;
-         /* Note: when we reference this parameter in an instruction later,
-          * we'll check if it's really a constant/immediate and set the
-          * instruction register type appropriately.
-          */
-         param_var->param_binding_length++;
+         {
+            GLfloat const_values[4];
+            GLint size;
+            parse_constant(inst, const_values, &size, Program, use);
+            if (param_var->name[0] == ' ') {
+               /* this is an unnamed constant */
+               idx = _mesa_add_unnamed_constant(Program->Base.Parameters,
+                                                const_values, size,
+                                                &param_var->swizzle);
+            }
+            else {
+               /* named parameter/constant */
+               idx = _mesa_add_named_constant(Program->Base.Parameters,
+                                              (char *) param_var->name,
+                                              const_values, size);
+            }
+            if (param_var->param_binding_begin == ~0U)
+               param_var->param_binding_begin = idx;
+            param_var->param_binding_type = PROGRAM_STATE_VAR;
+            /* Note: when we reference this parameter in an instruction later,
+             * we'll check if it's really a constant/immediate and set the
+             * instruction register type appropriately.
+             */
+            param_var->param_binding_length++;
+         }
          break;
 
       default:
@@ -2431,6 +2451,9 @@ parse_swizzle_mask(const GLubyte ** inst, GLubyte *swizzle, GLint len)
             return;
       }
    }
+
+   if (len == 1)
+      swizzle[1] = swizzle[2] = swizzle[3] = swizzle[0];
 }
 
 
@@ -2485,7 +2508,7 @@ static GLuint
 parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
                struct var_cache **vc_head,
                struct arb_program *Program,
-               gl_register_file * File, GLint * Index,
+               gl_register_file * File, GLint * Index, GLuint *swizzle,
                GLboolean *IsRelOffset )
 {
    struct var_cache *src;
@@ -2494,6 +2517,8 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
 
    *IsRelOffset = 0;
 
+   *swizzle = SWIZZLE_XYZW; /* default */
+
    /* And the binding for the src */
    switch (*(*inst)++) {
       case REGISTER_ATTRIB:
@@ -2549,6 +2574,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
                      }
 
                      *Index = src->param_binding_begin + offset;
+                     *swizzle = src->swizzle;
                      break;
 
                   case ARRAY_INDEX_RELATIVE:
@@ -2571,6 +2597,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
                         /* And store it properly */
                         *Index = src->param_binding_begin + rel_off;
                         *IsRelOffset = 1;
+                        *swizzle = src->swizzle;
                      }
                      break;
                }
@@ -2582,6 +2609,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
 
                *File = (gl_register_file) src->param_binding_type;
                *Index = src->param_binding_begin;
+               *swizzle = src->swizzle;
                break;
          }
          break;
@@ -2650,6 +2678,21 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
 }
 
 
+static GLuint
+swizzle_swizzle(GLuint baseSwizzle, const GLubyte swizzle[4])
+{
+   GLuint i, swz, s[4];
+   for (i = 0; i < 4; i++) {
+      GLuint c = swizzle[i];
+      if (c <= SWIZZLE_W)
+         s[i] = GET_SWZ(baseSwizzle, c);
+      else
+         s[i] = c;
+   }
+   swz = MAKE_SWIZZLE4(s[0], s[1], s[2], s[3]);
+   return swz;
+}
+
 /**
  * Parse vertex/fragment program vector source register.
  */
@@ -2664,12 +2707,14 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst,
    GLubyte negateMask;
    GLubyte swizzle[4];
    GLboolean isRelOffset;
+   GLuint baseSwizzle;
 
    /* Grab the sign */
    negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE;
 
    /* And the src reg */
-   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset))
+   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle,
+                     &isRelOffset))
       return 1;
 
    /* finally, the swizzle */
@@ -2677,7 +2722,7 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst,
 
    reg->File = file;
    reg->Index = index;
-   reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]);
+   reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle);
    reg->Negate = negateMask;
    reg->RelAddr = isRelOffset;
    return 0;
@@ -2698,12 +2743,14 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst,
    GLubyte negateMask;
    GLubyte swizzle[4];
    GLboolean isRelOffset;
+   GLuint baseSwizzle;
 
    /* Grab the sign */
    negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE;
 
    /* And the src reg */
-   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset))
+   if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle,
+                     &isRelOffset))
       return 1;
 
    /* finally, the swizzle */
@@ -2711,7 +2758,7 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst,
 
    reg->File = file;
    reg->Index = index;
-   reg->Swizzle = (swizzle[0] << 0);
+   reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle);
    reg->Negate = negateMask;
    reg->RelAddr = isRelOffset;
    return 0;
@@ -3022,8 +3069,10 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst,
 	    GLubyte negateMask;
             gl_register_file file;
 	    GLint index;
+            GLuint baseSwizzle;
 
-	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &rel))
+	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index,
+                              &baseSwizzle, &rel))
 	       return 1;
 	    parse_extended_swizzle_mask(inst, swizzle, &negateMask);
 	    fp->SrcReg[0].File = file;
@@ -3363,11 +3412,13 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst,
 	    GLboolean relAddr;
             gl_register_file file;
 	    GLint index;
+            GLuint baseSwizzle;
 
 	    if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg))
 	       return 1;
 
-	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &relAddr))
+	    if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index,
+                              &baseSwizzle, &relAddr))
 	       return 1;
 	    parse_extended_swizzle_mask (inst, swizzle, &negateMask);
 	    vp->SrcReg[0].File = file;
diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c
index 981565ab8f..39136efada 100644
--- a/src/mesa/shader/arbprogram.c
+++ b/src/mesa/shader/arbprogram.c
@@ -74,8 +74,6 @@ _mesa_BindProgram(GLenum target, GLuint id)
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-
    /* Error-check target and get curProg */
    if ((target == GL_VERTEX_PROGRAM_ARB) && /* == GL_VERTEX_PROGRAM_NV */
         (ctx->Extensions.NV_vertex_program ||
@@ -132,6 +130,9 @@ _mesa_BindProgram(GLenum target, GLuint id)
       return;
    }
 
+   /* signal new program (and its new constants) */
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+
    /* bind newProg */
    if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */
       _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
@@ -253,6 +254,8 @@ _mesa_EnableVertexAttribArrayARB(GLuint index)
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
    FLUSH_VERTICES(ctx, _NEW_ARRAY);
    ctx->Array.ArrayObj->VertexAttrib[index].Enabled = GL_TRUE;
    ctx->Array.ArrayObj->_Enabled |= _NEW_ARRAY_ATTRIB(index);
@@ -272,6 +275,8 @@ _mesa_DisableVertexAttribArrayARB(GLuint index)
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
    FLUSH_VERTICES(ctx, _NEW_ARRAY);
    ctx->Array.ArrayObj->VertexAttrib[index].Enabled = GL_FALSE;
    ctx->Array.ArrayObj->_Enabled &= ~_NEW_ARRAY_ATTRIB(index);
@@ -298,32 +303,41 @@ _mesa_GetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble *params)
 }
 
 
+/**
+ * Return info for a generic vertex attribute array (no alias with
+ * legacy vertex attributes (pos, normal, color, etc)).
+ */
 void GLAPIENTRY
 _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS) {
+   if (index >= MAX_VERTEX_GENERIC_ATTRIBS) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glGetVertexAttribfvARB(index)");
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Enabled;
+         params[0] = (GLfloat) array->Enabled;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = (GLfloat) array->Size;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = (GLfloat) array->Stride;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = (GLfloat) array->Type;
          break;
       case GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Normalized;
+         params[0] = array->Normalized;
          break;
       case GL_CURRENT_VERTEX_ATTRIB_ARB:
          if (index == 0) {
@@ -335,7 +349,7 @@ _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params)
          COPY_4V(params, ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + index]);
          break;
       case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name;
+         params[0] = (GLfloat) array->BufferObj->Name;
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribfvARB(pname)");
@@ -379,6 +393,8 @@ _mesa_GetVertexAttribPointervARB(GLuint index, GLenum pname, GLvoid **pointer)
       return;
    }
 
+   ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib));
+
    *pointer = (GLvoid *) ctx->Array.ArrayObj->VertexAttrib[index].Ptr;
 }
 
@@ -489,7 +505,7 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (target == GL_FRAGMENT_PROGRAM_ARB
        && ctx->Extensions.ARB_fragment_program) {
@@ -537,7 +553,7 @@ _mesa_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    GLfloat * dest;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (count <= 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glProgramEnvParameters4fv(count)");
@@ -631,7 +647,7 @@ _mesa_ProgramLocalParameter4fARB(GLenum target, GLuint index,
    struct gl_program *prog;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if ((target == GL_FRAGMENT_PROGRAM_NV
         && ctx->Extensions.NV_fragment_program) ||
@@ -685,7 +701,7 @@ _mesa_ProgramLocalParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    GLint i;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    if (count <= 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glProgramLocalParameters4fv(count)");
diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c
index 5142c2a4a5..d6469b17be 100644
--- a/src/mesa/shader/nvprogram.c
+++ b/src/mesa/shader/nvprogram.c
@@ -354,6 +354,7 @@ _mesa_GetTrackMatrixivNV(GLenum target, GLuint address,
 void GLAPIENTRY
 _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -362,15 +363,17 @@ _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params)
       return;
    }
 
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_ATTRIB_ARRAY_SIZE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = array->Size;
          break;
       case GL_ATTRIB_ARRAY_STRIDE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = array->Stride;
          break;
       case GL_ATTRIB_ARRAY_TYPE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = array->Type;
          break;
       case GL_CURRENT_ATTRIB_NV:
          if (index == 0) {
@@ -395,6 +398,7 @@ _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params)
 void GLAPIENTRY
 _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -403,15 +407,17 @@ _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params)
       return;
    }
 
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_ATTRIB_ARRAY_SIZE_NV:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = (GLfloat) array->Size;
          break;
       case GL_ATTRIB_ARRAY_STRIDE_NV:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = (GLfloat) array->Stride;
          break;
       case GL_ATTRIB_ARRAY_TYPE_NV:
-         params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = (GLfloat) array->Type;
          break;
       case GL_CURRENT_ATTRIB_NV:
          if (index == 0) {
@@ -436,6 +442,7 @@ _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params)
 void GLAPIENTRY
 _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
 {
+   const struct gl_client_array *array;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -444,15 +451,17 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
       return;
    }
 
+   array = &ctx->Array.ArrayObj->VertexAttrib[index];
+
    switch (pname) {
       case GL_ATTRIB_ARRAY_SIZE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Size;
+         params[0] = array->Size;
          break;
       case GL_ATTRIB_ARRAY_STRIDE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Stride;
+         params[0] = array->Stride;
          break;
       case GL_ATTRIB_ARRAY_TYPE_NV:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Type;
+         params[0] = array->Type;
          break;
       case GL_CURRENT_ATTRIB_NV:
          if (index == 0) {
@@ -467,7 +476,7 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
          params[3] = (GLint) ctx->Current.Attrib[index][3];
          break;
       case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB:
-         params[0] = ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name;
+         params[0] = array->BufferObj->Name;
          break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribdvNV");
@@ -706,7 +715,7 @@ _mesa_ProgramNamedParameter4fNV(GLuint id, GLsizei len, const GLubyte *name,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    prog = _mesa_lookup_program(ctx, id);
    if (!prog || prog->Target != GL_FRAGMENT_PROGRAM_NV) {
diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c
index 68a59350a1..f4beb9a78b 100644
--- a/src/mesa/shader/prog_execute.c
+++ b/src/mesa/shader/prog_execute.c
@@ -225,6 +225,13 @@ fetch_vector4(const struct prog_src_register *source,
       result[2] = -result[2];
       result[3] = -result[3];
    }
+
+#ifdef NAN_CHECK
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+#endif
 }
 
 
@@ -479,6 +486,13 @@ store_vector4(const struct prog_instruction *inst,
       }
    }
 
+#ifdef NAN_CHECK
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+#endif
+
    if (writeMask & WRITEMASK_X)
       dst[0] = value[0];
    if (writeMask & WRITEMASK_Y)
@@ -832,10 +846,14 @@ _mesa_execute_program(GLcontext * ctx,
          break;
       case OPCODE_EX2:         /* Exponential base 2 */
          {
-            GLfloat a[4], result[4];
+            GLfloat a[4], result[4], val;
             fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] = result[1] = result[2] = result[3] =
-               (GLfloat) _mesa_pow(2.0, a[0]);
+            val = (GLfloat) _mesa_pow(2.0, a[0]);
+            /*
+            if (IS_INF_OR_NAN(val))
+               val = 1.0e10;
+            */
+            result[0] = result[1] = result[2] = result[3] = val;
             store_vector4(inst, machine, result);
          }
          break;
@@ -911,12 +929,17 @@ _mesa_execute_program(GLcontext * ctx,
          break;
       case OPCODE_LG2:         /* log base 2 */
          {
-            GLfloat a[4], result[4];
+            GLfloat a[4], result[4], val;
             fetch_vector1(&inst->SrcReg[0], machine, a);
 	    /* The fast LOG2 macro doesn't meet the precision requirements.
 	     */
-            result[0] = result[1] = result[2] = result[3] =
-		(log(a[0]) * 1.442695F);
+            if (a[0] == 0.0F) {
+               val = 0.0F;
+            }
+            else {
+               val = log(a[0]) * 1.442695F;
+            }
+            result[0] = result[1] = result[2] = result[3] = val;
             store_vector4(inst, machine, result);
          }
          break;
diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c
index ae3a003fee..44c961927a 100644
--- a/src/mesa/shader/prog_instruction.c
+++ b/src/mesa/shader/prog_instruction.c
@@ -343,7 +343,10 @@ _mesa_opcode_string(gl_inst_opcode opcode)
 {
    if (opcode < MAX_OPCODE)
       return InstInfo[opcode].Name;
-   else
-      return "OP?";
+   else {
+      static char s[20];
+      _mesa_snprintf(s, sizeof(s), "OP%u", opcode);
+      return s;
+   }
 }
 
diff --git a/src/mesa/shader/prog_optimize.c b/src/mesa/shader/prog_optimize.c
index 6ba2e76ff9..be903106a0 100644
--- a/src/mesa/shader/prog_optimize.c
+++ b/src/mesa/shader/prog_optimize.c
@@ -547,15 +547,13 @@ update_interval(GLint intBegin[], GLint intEnd[], GLuint index, GLuint ic)
 
 
 /**
- * Find the live intervals for each temporary register in the program.
- * For register R, the interval [A,B] indicates that R is referenced
- * from instruction A through instruction B.
- * Special consideration is needed for loops and subroutines.
- * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason
+ * Find first/last instruction that references each temporary register.
  */
-static GLboolean
-find_live_intervals(struct gl_program *prog,
-                    struct interval_list *liveIntervals)
+GLboolean
+_mesa_find_temp_intervals(const struct prog_instruction *instructions,
+                          GLuint numInstructions,
+                          GLint intBegin[MAX_PROGRAM_TEMPS],
+                          GLint intEnd[MAX_PROGRAM_TEMPS])
 {
    struct loop_info
    {
@@ -563,26 +561,15 @@ find_live_intervals(struct gl_program *prog,
    };
    struct loop_info loopStack[MAX_LOOP_NESTING];
    GLuint loopStackDepth = 0;
-   GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS];
    GLuint i;
 
-   /*
-    * Note: we'll return GL_FALSE below if we find relative indexing
-    * into the TEMP register file.  We can't handle that yet.
-    * We also give up on subroutines for now.
-    */
-
-   if (dbg) {
-      _mesa_printf("Optimize: Begin find intervals\n");
-   }
-
    for (i = 0; i < MAX_PROGRAM_TEMPS; i++){
       intBegin[i] = intEnd[i] = -1;
    }
 
    /* Scan instructions looking for temporary registers */
-   for (i = 0; i < prog->NumInstructions; i++) {
-      const struct prog_instruction *inst = prog->Instructions + i;
+   for (i = 0; i < numInstructions; i++) {
+      const struct prog_instruction *inst = instructions + i;
       if (inst->Opcode == OPCODE_BGNLOOP) {
          loopStack[loopStackDepth].Start = i;
          loopStack[loopStackDepth].End = inst->BranchTarget;
@@ -595,7 +582,7 @@ find_live_intervals(struct gl_program *prog,
          return GL_FALSE;
       }
       else {
-         const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
+         const GLuint numSrc = 3;/*_mesa_num_inst_src_regs(inst->Opcode);*/
          GLuint j;
          for (j = 0; j < numSrc; j++) {
             if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) {
@@ -624,6 +611,39 @@ find_live_intervals(struct gl_program *prog,
       }
    }
 
+   return GL_TRUE;
+}
+
+
+/**
+ * Find the live intervals for each temporary register in the program.
+ * For register R, the interval [A,B] indicates that R is referenced
+ * from instruction A through instruction B.
+ * Special consideration is needed for loops and subroutines.
+ * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason
+ */
+static GLboolean
+find_live_intervals(struct gl_program *prog,
+                    struct interval_list *liveIntervals)
+{
+   GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS];
+   GLuint i;
+
+   /*
+    * Note: we'll return GL_FALSE below if we find relative indexing
+    * into the TEMP register file.  We can't handle that yet.
+    * We also give up on subroutines for now.
+    */
+
+   if (dbg) {
+      _mesa_printf("Optimize: Begin find intervals\n");
+   }
+
+   /* build intermediate arrays */
+   if (!_mesa_find_temp_intervals(prog->Instructions, prog->NumInstructions,
+                                  intBegin, intEnd))
+      return GL_FALSE;
+
    /* Build live intervals list from intermediate arrays */
    liveIntervals->Num = 0;
    for (i = 0; i < MAX_PROGRAM_TEMPS; i++) {
@@ -792,8 +812,6 @@ _mesa_reallocate_registers(struct gl_program *prog)
 }
 
 
-
-
 /**
  * Apply optimizations to the given program to eliminate unnecessary
  * instructions, temp regs, etc.
diff --git a/src/mesa/shader/prog_optimize.h b/src/mesa/shader/prog_optimize.h
index d102cfd9fc..43894a2723 100644
--- a/src/mesa/shader/prog_optimize.h
+++ b/src/mesa/shader/prog_optimize.h
@@ -25,7 +25,19 @@
 #ifndef PROG_OPT_H
 #define PROG_OPT_H
 
+
+#include "main/config.h"
+
+
 struct gl_program;
+struct prog_instruction;
+
+
+extern GLboolean
+_mesa_find_temp_intervals(const struct prog_instruction *instructions,
+                          GLuint numInstructions,
+                          GLint intBegin[MAX_PROGRAM_TEMPS],
+                          GLint intEnd[MAX_PROGRAM_TEMPS]);
 
 extern void
 _mesa_optimize_program(GLcontext *ctx, struct gl_program *program);
diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c
index 9967f2978d..de7fef1f86 100644
--- a/src/mesa/shader/prog_print.c
+++ b/src/mesa/shader/prog_print.c
@@ -75,7 +75,11 @@ file_string(gl_register_file f, gl_prog_print_mode mode)
    case PROGRAM_UNDEFINED:
       return "UNDEFINED";
    default:
-      return "Unknown program file!";
+      {
+         static char s[20];
+         _mesa_snprintf(s, sizeof(s), "FILE%u", f);
+         return s;
+      }
    }
 }
 
@@ -736,7 +740,10 @@ _mesa_fprint_instruction_opt(FILE *f,
                                 mode, prog);
       }
       else {
-         _mesa_fprintf(f, "Other opcode %d\n", inst->Opcode);
+         fprint_alu_instruction(f, inst,
+                                _mesa_opcode_string(inst->Opcode),
+                                3/*_mesa_num_inst_src_regs(inst->Opcode)*/,
+                                mode, prog);
       }
       break;
    }
@@ -941,6 +948,10 @@ _mesa_write_shader_to_file(const struct gl_shader *shader)
       fprintf(f, "/*\n");
       _mesa_fprint_program_opt(f, shader->Program, PROG_PRINT_DEBUG, GL_TRUE);
       fprintf(f, "*/\n");
+      fprintf(f, "/* Parameters / constants */\n");
+      fprintf(f, "/*\n");
+      _mesa_fprint_parameter_list(f, shader->Program->Parameters);
+      fprintf(f, "*/\n");
    }
 
    fclose(f);
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index d270bf9e1c..4623ff60de 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -820,3 +820,63 @@ _mesa_find_free_register(const struct gl_program *prog, GLuint regFile)
 
    return -1;
 }
+
+
+
+/**
+ * "Post-process" a GPU program.  This is intended to be used for debugging.
+ * Example actions include no-op'ing instructions or changing instruction
+ * behaviour.
+ */
+void
+_mesa_postprocess_program(GLcontext *ctx, struct gl_program *prog)
+{
+   static const GLfloat white[4] = { 0.5, 0.5, 0.5, 0.5 };
+   GLuint i;
+   GLuint whiteSwizzle;
+   GLint whiteIndex = _mesa_add_unnamed_constant(prog->Parameters,
+                                                 white, 4, &whiteSwizzle);
+
+   (void) whiteIndex;
+
+   for (i = 0; i < prog->NumInstructions; i++) {
+      struct prog_instruction *inst = prog->Instructions + i;
+      const GLuint n = _mesa_num_inst_src_regs(inst->Opcode);
+
+      (void) n;
+
+      if (_mesa_is_tex_instruction(inst->Opcode)) {
+#if 0
+         /* replace TEX/TXP/TXB with MOV */
+         inst->Opcode = OPCODE_MOV;
+         inst->DstReg.WriteMask = WRITEMASK_XYZW;
+         inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+         inst->SrcReg[0].Negate = NEGATE_NONE;
+#endif
+
+#if 0
+         /* disable shadow texture mode */
+         inst->TexShadow = 0;
+#endif
+      }
+
+      if (inst->Opcode == OPCODE_TXP) {
+#if 0
+         inst->Opcode = OPCODE_MOV;
+         inst->DstReg.WriteMask = WRITEMASK_XYZW;
+         inst->SrcReg[0].File = PROGRAM_CONSTANT;
+         inst->SrcReg[0].Index = whiteIndex;
+         inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+         inst->SrcReg[0].Negate = NEGATE_NONE;
+#endif
+#if 0
+         inst->TexShadow = 0;
+#endif
+#if 0
+         inst->Opcode = OPCODE_TEX;
+         inst->TexShadow = 0;
+#endif
+      }
+
+   }
+}
diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h
index 48176162c3..56a4191f57 100644
--- a/src/mesa/shader/program.h
+++ b/src/mesa/shader/program.h
@@ -122,6 +122,8 @@ _mesa_combine_programs(GLcontext *ctx,
 extern GLint
 _mesa_find_free_register(const struct gl_program *prog, GLuint regFile);
 
+extern void
+_mesa_postprocess_program(GLcontext *ctx, struct gl_program *prog);
 
 
 #endif /* PROGRAM_H */
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index cf9f8b9222..648fbc186b 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -1488,7 +1488,7 @@ _mesa_use_program(GLcontext *ctx, GLuint program)
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
 
    if (program) {
       shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram");
@@ -1510,6 +1510,10 @@ _mesa_use_program(GLcontext *ctx, GLuint program)
                          shProg->Shaders[i]->Name,
                          shProg->Shaders[i]->Type);
          }
+         if (shProg->VertexProgram)
+            printf(" vert prog %u\n", shProg->VertexProgram->Base.Id);
+         if (shProg->FragmentProgram)
+            printf(" frag prog %u\n", shProg->FragmentProgram->Base.Id);
       }
    }
    else {
@@ -1790,7 +1794,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
@@ -1930,7 +1934,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows,
       return;
    }
 
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS);
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c
index 387659ff30..f6032d1e9a 100644
--- a/src/mesa/shader/slang/slang_link.c
+++ b/src/mesa/shader/slang/slang_link.c
@@ -717,6 +717,8 @@ _slang_link(GLcontext *ctx,
       struct gl_vertex_program *linked_vprog =
          vertex_program(_mesa_clone_program(ctx, &vertProg->Base));
       shProg->VertexProgram = linked_vprog; /* refcount OK */
+      /* vertex program ID not significant; just set Id for debugging purposes */
+      shProg->VertexProgram->Base.Id = shProg->Name;
       ASSERT(shProg->VertexProgram->Base.RefCount == 1);
    }
 
@@ -725,6 +727,8 @@ _slang_link(GLcontext *ctx,
       struct gl_fragment_program *linked_fprog = 
          fragment_program(_mesa_clone_program(ctx, &fragProg->Base));
       shProg->FragmentProgram = linked_fprog; /* refcount OK */
+      /* vertex program ID not significant; just set Id for debugging purposes */
+      shProg->FragmentProgram->Base.Id = shProg->Name;
       ASSERT(shProg->FragmentProgram->Base.RefCount == 1);
    }
 
@@ -839,6 +843,14 @@ _slang_link(GLcontext *ctx,
       }
    }
 
+   /* Debug: */
+   if (0) {
+      if (shProg->VertexProgram)
+         _mesa_postprocess_program(ctx, &shProg->VertexProgram->Base);
+      if (shProg->FragmentProgram)
+         _mesa_postprocess_program(ctx, &shProg->FragmentProgram->Base);
+   }
+
    if (ctx->Shader.Flags & GLSL_DUMP) {
       _mesa_printf("Varying vars:\n");
       _mesa_print_parameter_list(shProg->Varying);
diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S
index 493e094cde..12d74fa4fb 100644
--- a/src/mesa/sparc/glapi_sparc.S
+++ b/src/mesa/sparc/glapi_sparc.S
@@ -759,23 +759,28 @@ gl_dispatch_functions_start:
 	GL_STUB(glGetAttribLocationARB, _gloffset_GetAttribLocationARB)
 	GL_STUB(glDrawBuffersARB, _gloffset_DrawBuffersARB)
 	GL_STUB(glRenderbufferStorageMultisample, _gloffset_RenderbufferStorageMultisample)
+	GL_STUB(glFlushMappedBufferRange, _gloffset_FlushMappedBufferRange)
+	GL_STUB(glMapBufferRange, _gloffset_MapBufferRange)
+	GL_STUB(glBindVertexArray, _gloffset_BindVertexArray)
+	GL_STUB(glGenVertexArrays, _gloffset_GenVertexArrays)
+	GL_STUB(glCopyBufferSubData, _gloffset_CopyBufferSubData)
 	GL_STUB(glPolygonOffsetEXT, _gloffset_PolygonOffsetEXT)
-	GL_STUB(gl_dispatch_stub_563, _gloffset_GetPixelTexGenParameterfvSGIS)
-	HIDDEN(gl_dispatch_stub_563)
-	GL_STUB(gl_dispatch_stub_564, _gloffset_GetPixelTexGenParameterivSGIS)
-	HIDDEN(gl_dispatch_stub_564)
-	GL_STUB(gl_dispatch_stub_565, _gloffset_PixelTexGenParameterfSGIS)
-	HIDDEN(gl_dispatch_stub_565)
-	GL_STUB(gl_dispatch_stub_566, _gloffset_PixelTexGenParameterfvSGIS)
-	HIDDEN(gl_dispatch_stub_566)
-	GL_STUB(gl_dispatch_stub_567, _gloffset_PixelTexGenParameteriSGIS)
-	HIDDEN(gl_dispatch_stub_567)
-	GL_STUB(gl_dispatch_stub_568, _gloffset_PixelTexGenParameterivSGIS)
+	GL_STUB(gl_dispatch_stub_568, _gloffset_GetPixelTexGenParameterfvSGIS)
 	HIDDEN(gl_dispatch_stub_568)
-	GL_STUB(gl_dispatch_stub_569, _gloffset_SampleMaskSGIS)
+	GL_STUB(gl_dispatch_stub_569, _gloffset_GetPixelTexGenParameterivSGIS)
 	HIDDEN(gl_dispatch_stub_569)
-	GL_STUB(gl_dispatch_stub_570, _gloffset_SamplePatternSGIS)
+	GL_STUB(gl_dispatch_stub_570, _gloffset_PixelTexGenParameterfSGIS)
 	HIDDEN(gl_dispatch_stub_570)
+	GL_STUB(gl_dispatch_stub_571, _gloffset_PixelTexGenParameterfvSGIS)
+	HIDDEN(gl_dispatch_stub_571)
+	GL_STUB(gl_dispatch_stub_572, _gloffset_PixelTexGenParameteriSGIS)
+	HIDDEN(gl_dispatch_stub_572)
+	GL_STUB(gl_dispatch_stub_573, _gloffset_PixelTexGenParameterivSGIS)
+	HIDDEN(gl_dispatch_stub_573)
+	GL_STUB(gl_dispatch_stub_574, _gloffset_SampleMaskSGIS)
+	HIDDEN(gl_dispatch_stub_574)
+	GL_STUB(gl_dispatch_stub_575, _gloffset_SamplePatternSGIS)
+	HIDDEN(gl_dispatch_stub_575)
 	GL_STUB(glColorPointerEXT, _gloffset_ColorPointerEXT)
 	GL_STUB(glEdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT)
 	GL_STUB(glIndexPointerEXT, _gloffset_IndexPointerEXT)
@@ -786,10 +791,10 @@ gl_dispatch_functions_start:
 	GL_STUB(glPointParameterfvEXT, _gloffset_PointParameterfvEXT)
 	GL_STUB(glLockArraysEXT, _gloffset_LockArraysEXT)
 	GL_STUB(glUnlockArraysEXT, _gloffset_UnlockArraysEXT)
-	GL_STUB(gl_dispatch_stub_581, _gloffset_CullParameterdvEXT)
-	HIDDEN(gl_dispatch_stub_581)
-	GL_STUB(gl_dispatch_stub_582, _gloffset_CullParameterfvEXT)
-	HIDDEN(gl_dispatch_stub_582)
+	GL_STUB(gl_dispatch_stub_586, _gloffset_CullParameterdvEXT)
+	HIDDEN(gl_dispatch_stub_586)
+	GL_STUB(gl_dispatch_stub_587, _gloffset_CullParameterfvEXT)
+	HIDDEN(gl_dispatch_stub_587)
 	GL_STUB(glSecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT)
 	GL_STUB(glSecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT)
 	GL_STUB(glSecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT)
@@ -814,8 +819,8 @@ gl_dispatch_functions_start:
 	GL_STUB(glFogCoorddvEXT, _gloffset_FogCoorddvEXT)
 	GL_STUB(glFogCoordfEXT, _gloffset_FogCoordfEXT)
 	GL_STUB(glFogCoordfvEXT, _gloffset_FogCoordfvEXT)
-	GL_STUB(gl_dispatch_stub_607, _gloffset_PixelTexGenSGIX)
-	HIDDEN(gl_dispatch_stub_607)
+	GL_STUB(gl_dispatch_stub_612, _gloffset_PixelTexGenSGIX)
+	HIDDEN(gl_dispatch_stub_612)
 	GL_STUB(glBlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT)
 	GL_STUB(glFlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV)
 	GL_STUB(glVertexArrayRangeNV, _gloffset_VertexArrayRangeNV)
@@ -857,24 +862,24 @@ gl_dispatch_functions_start:
 	GL_STUB(glWindowPos4ivMESA, _gloffset_WindowPos4ivMESA)
 	GL_STUB(glWindowPos4sMESA, _gloffset_WindowPos4sMESA)
 	GL_STUB(glWindowPos4svMESA, _gloffset_WindowPos4svMESA)
-	GL_STUB(gl_dispatch_stub_649, _gloffset_MultiModeDrawArraysIBM)
-	HIDDEN(gl_dispatch_stub_649)
-	GL_STUB(gl_dispatch_stub_650, _gloffset_MultiModeDrawElementsIBM)
-	HIDDEN(gl_dispatch_stub_650)
-	GL_STUB(gl_dispatch_stub_651, _gloffset_DeleteFencesNV)
-	HIDDEN(gl_dispatch_stub_651)
-	GL_STUB(gl_dispatch_stub_652, _gloffset_FinishFenceNV)
-	HIDDEN(gl_dispatch_stub_652)
-	GL_STUB(gl_dispatch_stub_653, _gloffset_GenFencesNV)
-	HIDDEN(gl_dispatch_stub_653)
-	GL_STUB(gl_dispatch_stub_654, _gloffset_GetFenceivNV)
+	GL_STUB(gl_dispatch_stub_654, _gloffset_MultiModeDrawArraysIBM)
 	HIDDEN(gl_dispatch_stub_654)
-	GL_STUB(gl_dispatch_stub_655, _gloffset_IsFenceNV)
+	GL_STUB(gl_dispatch_stub_655, _gloffset_MultiModeDrawElementsIBM)
 	HIDDEN(gl_dispatch_stub_655)
-	GL_STUB(gl_dispatch_stub_656, _gloffset_SetFenceNV)
+	GL_STUB(gl_dispatch_stub_656, _gloffset_DeleteFencesNV)
 	HIDDEN(gl_dispatch_stub_656)
-	GL_STUB(gl_dispatch_stub_657, _gloffset_TestFenceNV)
+	GL_STUB(gl_dispatch_stub_657, _gloffset_FinishFenceNV)
 	HIDDEN(gl_dispatch_stub_657)
+	GL_STUB(gl_dispatch_stub_658, _gloffset_GenFencesNV)
+	HIDDEN(gl_dispatch_stub_658)
+	GL_STUB(gl_dispatch_stub_659, _gloffset_GetFenceivNV)
+	HIDDEN(gl_dispatch_stub_659)
+	GL_STUB(gl_dispatch_stub_660, _gloffset_IsFenceNV)
+	HIDDEN(gl_dispatch_stub_660)
+	GL_STUB(gl_dispatch_stub_661, _gloffset_SetFenceNV)
+	HIDDEN(gl_dispatch_stub_661)
+	GL_STUB(gl_dispatch_stub_662, _gloffset_TestFenceNV)
+	HIDDEN(gl_dispatch_stub_662)
 	GL_STUB(glAreProgramsResidentNV, _gloffset_AreProgramsResidentNV)
 	GL_STUB(glBindProgramNV, _gloffset_BindProgramNV)
 	GL_STUB(glDeleteProgramsNV, _gloffset_DeleteProgramsNV)
@@ -955,26 +960,26 @@ gl_dispatch_functions_start:
 	GL_STUB(glSetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI)
 	GL_STUB(glPointParameteriNV, _gloffset_PointParameteriNV)
 	GL_STUB(glPointParameterivNV, _gloffset_PointParameterivNV)
-	GL_STUB(gl_dispatch_stub_738, _gloffset_ActiveStencilFaceEXT)
-	HIDDEN(gl_dispatch_stub_738)
-	GL_STUB(gl_dispatch_stub_739, _gloffset_BindVertexArrayAPPLE)
-	HIDDEN(gl_dispatch_stub_739)
-	GL_STUB(gl_dispatch_stub_740, _gloffset_DeleteVertexArraysAPPLE)
-	HIDDEN(gl_dispatch_stub_740)
-	GL_STUB(gl_dispatch_stub_741, _gloffset_GenVertexArraysAPPLE)
-	HIDDEN(gl_dispatch_stub_741)
-	GL_STUB(gl_dispatch_stub_742, _gloffset_IsVertexArrayAPPLE)
-	HIDDEN(gl_dispatch_stub_742)
+	GL_STUB(gl_dispatch_stub_743, _gloffset_ActiveStencilFaceEXT)
+	HIDDEN(gl_dispatch_stub_743)
+	GL_STUB(gl_dispatch_stub_744, _gloffset_BindVertexArrayAPPLE)
+	HIDDEN(gl_dispatch_stub_744)
+	GL_STUB(gl_dispatch_stub_745, _gloffset_DeleteVertexArraysAPPLE)
+	HIDDEN(gl_dispatch_stub_745)
+	GL_STUB(gl_dispatch_stub_746, _gloffset_GenVertexArraysAPPLE)
+	HIDDEN(gl_dispatch_stub_746)
+	GL_STUB(gl_dispatch_stub_747, _gloffset_IsVertexArrayAPPLE)
+	HIDDEN(gl_dispatch_stub_747)
 	GL_STUB(glGetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV)
 	GL_STUB(glGetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV)
 	GL_STUB(glProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV)
 	GL_STUB(glProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV)
 	GL_STUB(glProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV)
 	GL_STUB(glProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV)
-	GL_STUB(gl_dispatch_stub_749, _gloffset_DepthBoundsEXT)
-	HIDDEN(gl_dispatch_stub_749)
-	GL_STUB(gl_dispatch_stub_750, _gloffset_BlendEquationSeparateEXT)
-	HIDDEN(gl_dispatch_stub_750)
+	GL_STUB(gl_dispatch_stub_754, _gloffset_DepthBoundsEXT)
+	HIDDEN(gl_dispatch_stub_754)
+	GL_STUB(gl_dispatch_stub_755, _gloffset_BlendEquationSeparateEXT)
+	HIDDEN(gl_dispatch_stub_755)
 	GL_STUB(glBindFramebufferEXT, _gloffset_BindFramebufferEXT)
 	GL_STUB(glBindRenderbufferEXT, _gloffset_BindRenderbufferEXT)
 	GL_STUB(glCheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT)
@@ -992,19 +997,20 @@ gl_dispatch_functions_start:
 	GL_STUB(glIsFramebufferEXT, _gloffset_IsFramebufferEXT)
 	GL_STUB(glIsRenderbufferEXT, _gloffset_IsRenderbufferEXT)
 	GL_STUB(glRenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT)
-	GL_STUB(gl_dispatch_stub_768, _gloffset_BlitFramebufferEXT)
-	HIDDEN(gl_dispatch_stub_768)
-	GL_STUB(glFramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT)
-	GL_STUB(gl_dispatch_stub_770, _gloffset_StencilFuncSeparateATI)
-	HIDDEN(gl_dispatch_stub_770)
-	GL_STUB(gl_dispatch_stub_771, _gloffset_ProgramEnvParameters4fvEXT)
-	HIDDEN(gl_dispatch_stub_771)
-	GL_STUB(gl_dispatch_stub_772, _gloffset_ProgramLocalParameters4fvEXT)
-	HIDDEN(gl_dispatch_stub_772)
-	GL_STUB(gl_dispatch_stub_773, _gloffset_GetQueryObjecti64vEXT)
+	GL_STUB(gl_dispatch_stub_773, _gloffset_BlitFramebufferEXT)
 	HIDDEN(gl_dispatch_stub_773)
-	GL_STUB(gl_dispatch_stub_774, _gloffset_GetQueryObjectui64vEXT)
-	HIDDEN(gl_dispatch_stub_774)
+	GL_STUB(glFramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT)
+	GL_STUB(glProvokingVertexEXT, _gloffset_ProvokingVertexEXT)
+	GL_STUB(gl_dispatch_stub_776, _gloffset_StencilFuncSeparateATI)
+	HIDDEN(gl_dispatch_stub_776)
+	GL_STUB(gl_dispatch_stub_777, _gloffset_ProgramEnvParameters4fvEXT)
+	HIDDEN(gl_dispatch_stub_777)
+	GL_STUB(gl_dispatch_stub_778, _gloffset_ProgramLocalParameters4fvEXT)
+	HIDDEN(gl_dispatch_stub_778)
+	GL_STUB(gl_dispatch_stub_779, _gloffset_GetQueryObjecti64vEXT)
+	HIDDEN(gl_dispatch_stub_779)
+	GL_STUB(gl_dispatch_stub_780, _gloffset_GetQueryObjectui64vEXT)
+	HIDDEN(gl_dispatch_stub_780)
 	GL_STUB_ALIAS(glArrayElementEXT, glArrayElement)
 	GL_STUB_ALIAS(glBindTextureEXT, glBindTexture)
 	GL_STUB_ALIAS(glDrawArraysEXT, glDrawArrays)
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index 3ba7b26928..5d4d8eee02 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -24,11 +24,12 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * 
  **************************************************************************/
- /*
-  * Authors:
-  *   Keith Whitwell <keith@tungstengraphics.com>
-  *   Brian Paul
-  */
+
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ *   Brian Paul
+ */
 
 #include "main/imports.h"
 #include "shader/prog_parameter.h"
@@ -44,19 +45,21 @@
 #include "st_program.h"
 #include "st_inlines.h"
 
+
 /**
  * Pass the given program parameters to the graphics pipe as a
  * constant buffer.
- * \param id  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
+ * \param shader_type  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
  */
 void st_upload_constants( struct st_context *st,
                           struct gl_program_parameter_list *params,
-                          unsigned id)
+                          unsigned shader_type)
 {
    struct pipe_context *pipe = st->pipe;
-   struct pipe_constant_buffer *cbuf = &st->state.constants[id];
+   struct pipe_constant_buffer *cbuf = &st->state.constants[shader_type];
 
-   assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT);
+   assert(shader_type == PIPE_SHADER_VERTEX ||
+          shader_type == PIPE_SHADER_FRAGMENT);
 
    /* update constants */
    if (params && params->NumParameters) {
@@ -68,13 +71,14 @@ void st_upload_constants( struct st_context *st,
        * avoid gratuitous rendering synchronization.
        */
       pipe_buffer_reference(&cbuf->buffer, NULL );
-      cbuf->buffer = pipe_buffer_create(pipe->screen, 16, PIPE_BUFFER_USAGE_CONSTANT,
+      cbuf->buffer = pipe_buffer_create(pipe->screen, 16,
+                                        PIPE_BUFFER_USAGE_CONSTANT,
 					paramBytes );
 
-      if (0)
-      {
-	 printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
-                __FUNCTION__, id, params->NumParameters, params->StateFlags);
+      if (0) {
+	 debug_printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
+                      __FUNCTION__, shader_type, params->NumParameters,
+                      params->StateFlags);
          _mesa_print_parameter_list(params);
       }
 
@@ -84,15 +88,16 @@ void st_upload_constants( struct st_context *st,
 				       0, paramBytes,
 				       params->ParameterValues);
 
-      st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf);
+      st->pipe->set_constant_buffer(st->pipe, shader_type, 0, cbuf);
    }
    else {
-      st->constants.tracked_state[id].dirty.mesa = 0;
-      //  st->pipe->set_constant_buffer(st->pipe, id, 0, NULL);
+      st->constants.tracked_state[shader_type].dirty.mesa = 0x0;
    }
 }
 
-/* Vertex shader:
+
+/**
+ * Vertex shader:
  */
 static void update_vs_constants(struct st_context *st )
 {
@@ -102,6 +107,7 @@ static void update_vs_constants(struct st_context *st )
    st_upload_constants( st, params, PIPE_SHADER_VERTEX );
 }
 
+
 const struct st_tracked_state st_update_vs_constants = {
    "st_update_vs_constants",				/* name */
    {							/* dirty */
@@ -111,7 +117,10 @@ const struct st_tracked_state st_update_vs_constants = {
    update_vs_constants					/* update */
 };
 
-/* Fragment shader:
+
+
+/**
+ * Fragment shader:
  */
 static void update_fs_constants(struct st_context *st )
 {
@@ -121,6 +130,7 @@ static void update_fs_constants(struct st_context *st )
    st_upload_constants( st, params, PIPE_SHADER_FRAGMENT );
 }
 
+
 const struct st_tracked_state st_update_fs_constants = {
    "st_update_fs_constants",				/* name */
    {							/* dirty */
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index 3ef919a45e..5209a6a0c9 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -122,6 +122,7 @@ update_framebuffer_state( struct st_context *st )
                                    strb->surface);
             framebuffer->nr_cbufs++;
          }
+         strb->defined = GL_TRUE; /* we'll be drawing something */
       }
    }
    for (i = framebuffer->nr_cbufs; i < PIPE_MAX_COLOR_BUFS; i++) {
diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c
index 5c7206409c..36b28cb4df 100644
--- a/src/mesa/state_tracker/st_atom_rasterizer.c
+++ b/src/mesa/state_tracker/st_atom_rasterizer.c
@@ -102,6 +102,9 @@ static void update_raster_state( struct st_context *st )
    if (ctx->Light.ShadeModel == GL_FLAT)
       raster->flatshade = 1;
 
+   if (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION_EXT)
+      raster->flatshade_first = 1;
+
    /* _NEW_LIGHT | _NEW_PROGRAM
     *
     * Back-face colors can come from traditional lighting (when
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
index a627c0374a..8c1fd5ce02 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.c
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -26,8 +26,14 @@
  **************************************************************************/
 
 
+/**
+ * Functions for pixel buffer objects and vertex/element buffer objects.
+ */
+
+
 #include "main/imports.h"
 #include "main/mtypes.h"
+#include "main/arrayobj.h"
 #include "main/bufferobj.h"
 
 #include "st_inlines.h"
@@ -39,14 +45,6 @@
 #include "pipe/p_inlines.h"
 
 
-
-/* Pixel buffers and Vertex/index buffers are handled through these
- * mesa callbacks.  Framebuffer/Renderbuffer objects are
- * created/managed elsewhere.
- */
-
-
-
 /**
  * There is some duplication between mesa's bufferobjects and our
  * bufmgr buffers.  Both have an integer handle and a hashtable to
@@ -100,8 +98,10 @@ st_bufferobj_subdata(GLcontext *ctx,
 {
    struct st_buffer_object *st_obj = st_buffer_object(obj);
 
-   if (offset >= st_obj->size || size > (st_obj->size - offset))
-      return;
+   /* we may be called from VBO code, so double-check params here */
+   ASSERT(offset >= 0);
+   ASSERT(size >= 0);
+   ASSERT(offset + size <= obj->Size);
 
    st_cond_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer,
 				   offset, size, data);
@@ -120,8 +120,10 @@ st_bufferobj_get_subdata(GLcontext *ctx,
 {
    struct st_buffer_object *st_obj = st_buffer_object(obj);
 
-   if (offset >= st_obj->size || size > (st_obj->size - offset))
-      return;
+   /* we may be called from VBO code, so double-check params here */
+   ASSERT(offset >= 0);
+   ASSERT(size >= 0);
+   ASSERT(offset + size <= obj->Size);
 
    st_cond_flush_pipe_buffer_read(st_context(ctx), st_obj->buffer,
 				  offset, size, data);
@@ -174,8 +176,6 @@ st_bufferobj_data(GLcontext *ctx,
       return;
    }
 
-   st_obj->size = size;
-
    if (data)
       st_no_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer, 0,
 				    size, data);
@@ -190,7 +190,7 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
                  struct gl_buffer_object *obj)
 {
    struct st_buffer_object *st_obj = st_buffer_object(obj);
-   GLuint flags;
+   uint flags;
 
    switch (access) {
    case GL_WRITE_ONLY:
@@ -209,7 +209,7 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
    obj->Pointer = st_cond_flush_pipe_buffer_map(st_context(ctx),
 						st_obj->buffer,
 						flags);
-   if(obj->Pointer) {
+   if (obj->Pointer) {
       obj->Offset = 0;
       obj->Length = obj->Size;
    }
@@ -217,7 +217,6 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
 }
 
 
-
 /**
  * Called via glMapBufferRange().
  */
@@ -228,7 +227,7 @@ st_bufferobj_map_range(GLcontext *ctx, GLenum target,
 {
    struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_buffer_object *st_obj = st_buffer_object(obj);
-   GLuint flags = 0;
+   uint flags = 0x0;
    char *map;
 
    if (access & GL_MAP_WRITE_BIT)
@@ -297,6 +296,43 @@ st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
 }
 
 
+/**
+ * Called via glCopyBufferSubData().
+ */
+static void
+st_copy_buffer_subdata(GLcontext *ctx,
+                       struct gl_buffer_object *src,
+                       struct gl_buffer_object *dst,
+                       GLintptr readOffset, GLintptr writeOffset,
+                       GLsizeiptr size)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *srcObj = st_buffer_object(src);
+   struct st_buffer_object *dstObj = st_buffer_object(dst);
+   ubyte *srcPtr, *dstPtr;
+
+   /* buffer should not already be mapped */
+   assert(!src->Pointer);
+   assert(!dst->Pointer);
+
+   srcPtr = (ubyte *) pipe_buffer_map_range(pipe->screen,
+                                            srcObj->buffer,
+                                            readOffset, size,
+                                            PIPE_BUFFER_USAGE_CPU_READ);
+
+   dstPtr = (ubyte *) pipe_buffer_map_range(pipe->screen,
+                                            dstObj->buffer,
+                                            writeOffset, size,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   if (srcPtr && dstPtr)
+      _mesa_memcpy(dstPtr + writeOffset, srcPtr + readOffset, size);
+
+   pipe_buffer_unmap(pipe->screen, srcObj->buffer);
+   pipe_buffer_unmap(pipe->screen, dstObj->buffer);
+}
+
+
 void
 st_init_bufferobject_functions(struct dd_function_table *functions)
 {
@@ -309,4 +345,9 @@ st_init_bufferobject_functions(struct dd_function_table *functions)
    functions->MapBufferRange = st_bufferobj_map_range;
    functions->FlushMappedBufferRange = st_bufferobj_flush_mapped_range;
    functions->UnmapBuffer = st_bufferobj_unmap;
+   functions->CopyBufferSubData = st_copy_buffer_subdata;
+
+   /* For GL_APPLE_vertex_array_object */
+   functions->NewArrayObject = _mesa_new_array_object;
+   functions->DeleteArrayObject = _mesa_delete_array_object;
 }
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h
index dcbb5a5233..fda6d05dd3 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.h
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.h
@@ -40,22 +40,14 @@ struct st_buffer_object
 {
    struct gl_buffer_object Base;
    struct pipe_buffer *buffer;  
-   GLsizeiptrARB size;
 };
 
 
-/* Are the obj->Name tests necessary?  Unfortunately yes, mesa
- * allocates a couple of gl_buffer_object structs statically, and the
- * Name == 0 test is the only way to identify them and avoid casting
- * them erroneously to our structs.
- */
+/** cast wrapper */
 static INLINE struct st_buffer_object *
 st_buffer_object(struct gl_buffer_object *obj)
 {
-   if (obj->Name)
-      return (struct st_buffer_object *) obj;
-   else
-      return NULL;
+   return (struct st_buffer_object *) obj;
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index 21ddf2fc7a..ecdb988033 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -100,6 +100,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    strb->Base.Height = height;
    init_renderbuffer_bits(strb, format);
 
+   strb->defined = GL_FALSE;  /* undefined contents now */
+
    if(strb->software) {
       struct pipe_format_block block;
       size_t size;
@@ -463,6 +465,134 @@ st_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
 }
 
 
+/**
+ * Copy back color buffer to front color buffer.
+ */
+static void
+copy_back_to_front(struct st_context *st,
+                   struct gl_framebuffer *fb,
+                   gl_buffer_index frontIndex,
+                   gl_buffer_index backIndex)
+
+{
+   struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+   struct pipe_surface *surf_front, *surf_back;
+
+   (void) st_get_framebuffer_surface(stfb, frontIndex, &surf_front);
+   (void) st_get_framebuffer_surface(stfb, backIndex, &surf_back);
+
+   if (surf_front && surf_back) {
+      st->pipe->surface_copy(st->pipe,
+                             surf_front, 0, 0,  /* dest */
+                             surf_back, 0, 0,   /* src */
+                             fb->Width, fb->Height);
+   }
+}
+
+
+/**
+ * Check if we're drawing into, or read from, a front color buffer.  If the
+ * front buffer is missing, create it now.
+ *
+ * The back color buffer must exist since we'll use its format/samples info
+ * for creating the front buffer.
+ *
+ * \param frontIndex  either BUFFER_FRONT_LEFT or BUFFER_FRONT_RIGHT
+ * \param backIndex  either BUFFER_BACK_LEFT or BUFFER_BACK_RIGHT
+ */
+static void
+check_create_front_buffer(GLcontext *ctx, struct gl_framebuffer *fb,
+                          gl_buffer_index frontIndex,
+                          gl_buffer_index backIndex)
+{
+   if (fb->Attachment[frontIndex].Renderbuffer == NULL) {
+      GLboolean create = GL_FALSE;
+
+      /* check if drawing to or reading from front buffer */
+      if (fb->_ColorReadBufferIndex == frontIndex) {
+         create = GL_TRUE;
+      }
+      else {
+         GLuint b;
+         for (b = 0; b < fb->_NumColorDrawBuffers; b++) {
+            if (fb->_ColorDrawBufferIndexes[b] == frontIndex) {
+               create = GL_TRUE;
+               break;
+            }
+         }
+      }
+
+      if (create) {
+         struct st_renderbuffer *back;
+         struct gl_renderbuffer *front;
+         enum pipe_format colorFormat;
+         uint samples;
+
+         if (0)
+            _mesa_debug(ctx, "Allocate new front buffer\n");
+
+         /* get back renderbuffer info */
+         back = st_renderbuffer(fb->Attachment[backIndex].Renderbuffer);
+         colorFormat = back->format;
+         samples = back->Base.NumSamples;
+
+         /* create front renderbuffer */
+         front = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
+         _mesa_add_renderbuffer(fb, frontIndex, front);
+
+         /* alloc texture/surface for new front buffer */
+         front->AllocStorage(ctx, front, front->InternalFormat,
+                             fb->Width, fb->Height);
+
+         /* initialize the front color buffer contents by copying
+          * the back buffer.
+          */
+         copy_back_to_front(ctx->st, fb, frontIndex, backIndex);
+      }
+   }
+}
+
+
+/**
+ * If front left/right color buffers are missing, create them now.
+ */
+static void
+check_create_front_buffers(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   /* check if we need to create the front left buffer now */
+   check_create_front_buffer(ctx, fb, BUFFER_FRONT_LEFT, BUFFER_BACK_LEFT);
+
+   if (fb->Visual.stereoMode) {
+      check_create_front_buffer(ctx, fb, BUFFER_FRONT_RIGHT, BUFFER_BACK_RIGHT);
+   }
+
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+/**
+ * Called via glDrawBuffer.
+ */
+static void
+st_DrawBuffers(GLcontext *ctx, GLsizei count, const GLenum *buffers)
+{
+   (void) count;
+   (void) buffers;
+   check_create_front_buffers(ctx, ctx->DrawBuffer);
+}
+
+
+/**
+ * Called via glReadBuffer.
+ */
+static void
+st_ReadBuffer(GLcontext *ctx, GLenum buffer)
+{
+   (void) buffer;
+   check_create_front_buffers(ctx, ctx->ReadBuffer);
+}
+
+
 void st_init_fbo_functions(struct dd_function_table *functions)
 {
    functions->NewFramebuffer = st_new_framebuffer;
@@ -475,4 +605,7 @@ void st_init_fbo_functions(struct dd_function_table *functions)
    /* no longer needed by core Mesa, drivers handle resizes...
    functions->ResizeBuffers = st_resize_buffers;
    */
+
+   functions->DrawBuffers = st_DrawBuffers;
+   functions->ReadBuffer = st_ReadBuffer;
 }
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
index 9a199550d9..bea6eb89c3 100644
--- a/src/mesa/state_tracker/st_cb_fbo.h
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -40,6 +40,7 @@ struct st_renderbuffer
    struct pipe_texture *texture;
    struct pipe_surface *surface; /* temporary view into texture */
    enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
+   GLboolean defined;        /**< defined contents? */
 
    /**
     * Used only when hardware accumulation buffers are not supported.
diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c
index fbaffd154f..8ceeeabcd3 100644
--- a/src/mesa/state_tracker/st_cb_flush.c
+++ b/src/mesa/state_tracker/st_cb_flush.c
@@ -47,10 +47,19 @@
 #include "util/u_blit.h"
 
 
+/** Check if we have a front color buffer and if it's been drawn to. */
 static INLINE GLboolean
 is_front_buffer_dirty(struct st_context *st)
 {
-   return st->frontbuffer_status == FRONT_STATUS_DIRTY;
+   if (st->frontbuffer_status == FRONT_STATUS_DIRTY) {
+      return GL_TRUE;
+   }
+   else {
+      GLframebuffer *fb = st->ctx->DrawBuffer;
+      struct st_renderbuffer *strb
+         = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+      return strb && strb->defined;
+   }
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
index 7dd2352739..3bcccd0df4 100644
--- a/src/mesa/state_tracker/st_cb_rasterpos.c
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -194,9 +194,10 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw)
    rs->stage.destroy = rastpos_destroy;
    rs->ctx = ctx;
 
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(rs->array); i++) {
       rs->array[i].Size = 4;
       rs->array[i].Type = GL_FLOAT;
+      rs->array[i].Format = GL_RGBA;
       rs->array[i].Stride = 0;
       rs->array[i].StrideB = 0;
       rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i];
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index 6ffed56d9a..18adb35e87 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -45,6 +45,7 @@ struct blit_state;
 struct bitmap_cache;
 
 
+/** XXX we'd like to get rid of these */
 #define FRONT_STATUS_UNDEFINED    0
 #define FRONT_STATUS_DIRTY        1
 #define FRONT_STATUS_COPY_OF_BACK 2
@@ -111,7 +112,7 @@ struct st_context
       struct gl_fragment_program *fragment_program;
    } cb;
 
-   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ */
+   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ (XXX to be removed) */
 
    char vendor[100];
    char renderer[100];
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 225541a30b..914a507bef 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -159,12 +159,21 @@ static GLuint fixed_types[4] = {
  * Return a PIPE_FORMAT_x for the given GL datatype and size.
  */
 GLuint
-st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized)
+st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
+                      GLboolean normalized)
 {
    assert((type >= GL_BYTE && type <= GL_DOUBLE) ||
           type == GL_FIXED);
    assert(size >= 1);
    assert(size <= 4);
+   assert(format == GL_RGBA || format == GL_BGRA);
+
+   if (format == GL_BGRA) {
+      /* this is an odd-ball case */
+      assert(type == GL_UNSIGNED_BYTE);
+      assert(normalized);
+      return PIPE_FORMAT_B8G8R8A8_UNORM;
+   }
 
    if (normalized) {
       switch (type) {
@@ -220,8 +229,10 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count,
       struct st_buffer_object *stobj = st_buffer_object(array->BufferObj);
       ubyte *map;
 
-      if (!stobj)
+      if (!stobj || stobj->Base.Name == 0) {
+         /* edge flags are not in a VBO */
          return NULL;
+      }
 
       vec = (unsigned *) _mesa_calloc(sizeof(unsigned) * ((count + 31) / 32));
       if (!vec)
@@ -392,6 +403,7 @@ setup_interleaved_attribs(GLcontext *ctx,
       velements[attr].src_format =
          st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Format,
                                arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
    }
@@ -479,6 +491,7 @@ setup_non_interleaved_attribs(GLcontext *ctx,
       velements[attr].src_format
          = st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                  arrays[mesaAttr]->Size,
+                                 arrays[mesaAttr]->Format,
                                  arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
    }
diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h
index da04fce8e2..dcfe7e1536 100644
--- a/src/mesa/state_tracker/st_draw.h
+++ b/src/mesa/state_tracker/st_draw.h
@@ -62,7 +62,8 @@ st_feedback_draw_vbo(GLcontext *ctx,
 /* Internal function:
  */
 extern GLuint
-st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized);
+st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
+                      GLboolean normalized);
 
 
 /**
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index 32502a9cda..2712c131c0 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -178,6 +178,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
       velements[attr].src_format = 
          st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Format,
                                arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
 
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index e6728ae895..8a958e8bd8 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -143,19 +143,21 @@ void st_init_extensions(struct st_context *st)
    /*
     * Extensions that are supported by all Gallium drivers:
     */
-   ctx->Extensions.ARB_multisample = GL_TRUE;
+   ctx->Extensions.ARB_copy_buffer = GL_TRUE;
    ctx->Extensions.ARB_fragment_program = GL_TRUE;
+   ctx->Extensions.ARB_map_buffer_range = GL_TRUE;
+   ctx->Extensions.ARB_multisample = GL_TRUE;
    ctx->Extensions.ARB_texture_border_clamp = GL_TRUE; /* XXX temp */
    ctx->Extensions.ARB_texture_compression = GL_TRUE;
    ctx->Extensions.ARB_texture_cube_map = GL_TRUE;
    ctx->Extensions.ARB_texture_env_combine = GL_TRUE;
    ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE;
    ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE;
-   ctx->Extensions.ARB_vertex_program = GL_TRUE;
+   ctx->Extensions.ARB_vertex_array_object = GL_TRUE;
    ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;
+   ctx->Extensions.ARB_vertex_program = GL_TRUE;
 
    ctx->Extensions.EXT_blend_color = GL_TRUE;
-   ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
    ctx->Extensions.EXT_blend_func_separate = GL_TRUE;
    ctx->Extensions.EXT_blend_logic_op = GL_TRUE;
    ctx->Extensions.EXT_blend_minmax = GL_TRUE;
@@ -166,15 +168,20 @@ void st_init_extensions(struct st_context *st)
    ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;
    ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE;
    ctx->Extensions.EXT_point_parameters = GL_TRUE;
+   ctx->Extensions.EXT_provoking_vertex = GL_TRUE;
    ctx->Extensions.EXT_secondary_color = GL_TRUE;
    ctx->Extensions.EXT_stencil_wrap = GL_TRUE;
    ctx->Extensions.EXT_texture_env_add = GL_TRUE;
    ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
    ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
    ctx->Extensions.EXT_texture_lod_bias = GL_TRUE;
+   ctx->Extensions.EXT_vertex_array_bgra = GL_TRUE;
+
+   ctx->Extensions.APPLE_vertex_array_object = GL_TRUE;
 
    ctx->Extensions.NV_blend_square = GL_TRUE;
    ctx->Extensions.NV_texgen_reflection = GL_TRUE;
+   ctx->Extensions.NV_texture_env_combine4 = GL_TRUE;
 
    ctx->Extensions.SGI_color_matrix = GL_TRUE;
    ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
@@ -198,6 +205,10 @@ void st_init_extensions(struct st_context *st)
       ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE;
    }
 
+   if (screen->get_param(screen, PIPE_CAP_BLEND_EQUATION_SEPARATE)) {
+      ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
+   }
+
    if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_CLAMP) > 0) {
       ctx->Extensions.EXT_texture_mirror_clamp = GL_TRUE;
    }
@@ -282,4 +293,10 @@ void st_init_extensions(struct st_context *st)
                                    PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
       ctx->Extensions.MESA_ycbcr_texture = GL_TRUE;
    }
+
+   /* GL_ARB_framebuffer_object */
+   if (ctx->Extensions.EXT_packed_depth_stencil) {
+      /* we support always support GL_EXT_framebuffer_blit */
+      ctx->Extensions.ARB_framebuffer_object = GL_TRUE;
+   }
 }
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index fe99fc0892..ca32b2e573 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -58,19 +58,19 @@ st_create_framebuffer( const __GLcontextModes *visual,
 
       _mesa_initialize_framebuffer(&stfb->Base, visual);
 
-      {
-         /* fake frontbuffer */
-         /* XXX allocation should only happen in the unusual case
-            it's actually needed */
+      if (visual->doubleBufferMode) {
          struct gl_renderbuffer *rb
             = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
-         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
       }
-
-      if (visual->doubleBufferMode) {
+      else {
+         /* Only allocate front buffer right now if we're single buffered.
+          * If double-buffered, allocate front buffer on demand later.
+          * See check_create_front_buffers().
+          */
          struct gl_renderbuffer *rb
             = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
-         _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
       }
 
       if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
@@ -286,6 +286,115 @@ st_notify_swapbuffers(struct st_framebuffer *stfb)
 }
 
 
+/**
+ * Swap the front/back color buffers.  Exchange the front/back pointers
+ * and update some derived state.
+ * No need to call st_notify_swapbuffers() first.
+ *
+ * For a single-buffered framebuffer, no swap occurs, but we still return
+ * the pointer(s) to the front color buffer(s).
+ *
+ * \param front_left  returns pointer to front-left renderbuffer after swap
+ * \param front_right  returns pointer to front-right renderbuffer after swap
+ */
+void
+st_swapbuffers(struct st_framebuffer *stfb,
+               struct pipe_surface **front_left,
+               struct pipe_surface **front_right)
+{
+   struct gl_framebuffer *fb = &stfb->Base;
+
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      st_flush( ctx->st, 
+		PIPE_FLUSH_RENDER_CACHE | 
+		PIPE_FLUSH_SWAPBUFFERS |
+		PIPE_FLUSH_FRAME,
+                NULL );
+   }
+
+   if (!fb->Visual.doubleBufferMode) {
+      /* single buffer mode - return pointers to front surfaces */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+      return;
+   }
+
+   /* swap left buffers */
+   if (fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer = rbTemp;
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front buffer, display the back buffer */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+   }
+
+   /* swap right buffers (for stereo) */
+   if (fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer = rbTemp;
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front right buffer, display back right buffer (if exists) */
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+   }
+
+   /* Update the _ColorDrawBuffers[] array and _ColorReadBuffer pointer */
+   _mesa_update_framebuffer(ctx);
+
+   /* Make sure we draw into the new back surface */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
 void *st_framebuffer_private( struct st_framebuffer *stfb )
 {
    return stfb->Private;
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 18d1046831..806e0ca8f6 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -151,6 +151,7 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_ATTRIB_TEX5:
          case VERT_ATTRIB_TEX6:
          case VERT_ATTRIB_TEX7:
+            assert(slot < Elements(vs_input_semantic_name));
             vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
             vs_input_semantic_index[slot] = num_generic++;
             break;
@@ -171,6 +172,7 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_ATTRIB_GENERIC14:
          case VERT_ATTRIB_GENERIC15:
             assert(attr < VERT_ATTRIB_MAX);
+            assert(slot < Elements(vs_input_semantic_name));
             vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
             vs_input_semantic_index[slot] = num_generic++;
             break;
@@ -198,6 +200,7 @@ st_translate_vertex_program(struct st_context *st,
 
    /* initialize output semantics to defaults */
    for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) {
+      assert(i < Elements(vs_output_semantic_name));
       vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC;
       vs_output_semantic_index[i] = 0;
       output_flags[i] = 0x0;
@@ -273,6 +276,7 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_RESULT_VAR0:
             /* fall-through */
          default:
+            assert(slot < Elements(vs_output_semantic_name));
             if (outputSemanticName) {
                /* use provided semantic into */
                assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT);
@@ -286,6 +290,7 @@ st_translate_vertex_program(struct st_context *st,
             }
          }
 
+         assert(slot < Elements(output_flags));
          output_flags[slot] = stvp->Base.Base.OutputFlags[attr];
       }
    }
@@ -303,6 +308,26 @@ st_translate_vertex_program(struct st_context *st,
       outputMapping = defaultOutputMapping;
    }
 
+#if 0 /* debug */
+   {
+      GLuint i;
+      printf("outputMapping? %d\n", outputMapping ? 1 : 0);
+      if (outputMapping) {
+         printf("attr -> slot\n");
+         for (i = 0; i < 16;  i++) {
+            printf(" %2d       %3d\n", i, outputMapping[i]);
+         }
+      }
+      printf("slot    sem_name  sem_index\n");
+      for (i = 0; i < vs_num_outputs; i++) {
+         printf(" %2d         %d         %d\n",
+                i,
+                vs_output_semantic_name[i],
+                vs_output_semantic_index[i]);
+      }
+   }
+#endif
+
    /* free old shader state, if any */
    if (stvp->state.tokens) {
       _mesa_free((void *) stvp->state.tokens);
diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h
index c411687bb6..04d3a3d7c2 100644
--- a/src/mesa/state_tracker/st_public.h
+++ b/src/mesa/state_tracker/st_public.h
@@ -103,6 +103,10 @@ void st_finish( struct st_context *st );
 
 void st_notify_swapbuffers(struct st_framebuffer *stfb);
 
+void st_swapbuffers(struct st_framebuffer *stfb,
+                    struct pipe_surface **front_left,
+                    struct pipe_surface **front_right);
+
 int st_set_teximage(struct pipe_texture *pt, int target);
 
 /** Redirect rendering into stfb's surface to a texture image */
diff --git a/src/mesa/swrast/s_imaging.c b/src/mesa/swrast/s_imaging.c
index d6be3aa022..3578b713f6 100644
--- a/src/mesa/swrast/s_imaging.c
+++ b/src/mesa/swrast/s_imaging.c
@@ -60,7 +60,7 @@ _swrast_CopyColorTable( GLcontext *ctx,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ColorTable(target, internalformat, width, GL_RGBA, CHAN_TYPE, data);
 
@@ -94,7 +94,7 @@ _swrast_CopyColorSubTable( GLcontext *ctx,GLenum target, GLsizei start,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ColorSubTable(target, start, width, GL_RGBA, CHAN_TYPE, data);
 
@@ -126,7 +126,7 @@ _swrast_CopyConvolutionFilter1D(GLcontext *ctx, GLenum target,
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    /* store as convolution filter */
    _mesa_ConvolutionFilter1D(target, internalFormat, width,
@@ -178,12 +178,12 @@ _swrast_CopyConvolutionFilter2D(GLcontext *ctx, GLenum target,
    ctx->Unpack.SkipImages = 0;
    ctx->Unpack.SwapBytes = GL_FALSE;
    ctx->Unpack.LsbFirst = GL_FALSE;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
    ctx->NewState |= _NEW_PACKUNPACK;
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
-   ctx->Unpack.BufferObj = ctx->Array.NullBufferObj;
+   ctx->Unpack.BufferObj = ctx->Shared->NullBufferObj;
 
    _mesa_ConvolutionFilter2D(target, internalFormat, width, height,
                              GL_RGBA, CHAN_TYPE, rgba);
diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c
index fa8ca1d0e2..0e2793b474 100644
--- a/src/mesa/swrast/s_span.c
+++ b/src/mesa/swrast/s_span.c
@@ -1297,7 +1297,6 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
           span->primitive == GL_LINE ||
 	  span->primitive == GL_POLYGON ||
           span->primitive == GL_BITMAP);
-   ASSERT(span->end <= MAX_WIDTH);
 
    /* Fragment write masks */
    if (span->arrayMask & SPAN_MASK) {
@@ -1310,12 +1309,12 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    }
 
    /* Clip to window/scissor box */
-   if ((swrast->_RasterMask & CLIP_BIT) || (span->primitive != GL_POLYGON)) {
-      if (!clip_span(ctx, span)) {
-	 return;
-      }
+   if (!clip_span(ctx, span)) {
+      return;
    }
 
+   ASSERT(span->end <= MAX_WIDTH);
+
 #ifdef DEBUG
    /* Make sure all fragments are within window bounds */
    if (span->arrayMask & SPAN_XY) {
@@ -1356,15 +1355,6 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    if (ctx->Stencil._Enabled || ctx->Depth.Test) {
       if (!(span->arrayMask & SPAN_Z))
          _swrast_span_interpolate_z(ctx, span);
-
-      if ((span->arrayMask & SPAN_XY) == 0) {
-         if (span->x < fb->_Xmin || span->x + span->end > fb->_Xmax ||
-             span->y < fb->_Ymin || span->y >= fb->_Ymax) {
-            printf("Bad span clipping at %d, %d\n", span->x, span->y);
-            return;
-         }
-      }
-
       if (ctx->Stencil._Enabled) {
          /* Combined Z/stencil tests */
          if (!_swrast_stencil_and_ztest_span(ctx, span)) {
diff --git a/src/mesa/swrast/s_texfilter.c b/src/mesa/swrast/s_texfilter.c
index dd59314cd9..6b1f934647 100644
--- a/src/mesa/swrast/s_texfilter.c
+++ b/src/mesa/swrast/s_texfilter.c
@@ -1329,7 +1329,7 @@ static void
 opt_sample_rgb_2d(GLcontext *ctx,
                   const struct gl_texture_object *tObj,
                   GLuint n, const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLchan rgba[][4])
+                  const GLfloat lambda[], GLfloat rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
    const GLfloat width = (GLfloat) img->Width;
@@ -1351,9 +1351,9 @@ opt_sample_rgb_2d(GLcontext *ctx,
       GLint j = IFLOOR(texcoords[k][1] * height) & rowMask;
       GLint pos = (j << shift) | i;
       GLchan *texel = ((GLchan *) img->Data) + 3*pos;
-      rgba[k][RCOMP] = texel[0];
-      rgba[k][GCOMP] = texel[1];
-      rgba[k][BCOMP] = texel[2];
+      rgba[k][RCOMP] = CHAN_TO_FLOAT(texel[0]);
+      rgba[k][GCOMP] = CHAN_TO_FLOAT(texel[1]);
+      rgba[k][BCOMP] = CHAN_TO_FLOAT(texel[2]);
    }
 }
 
@@ -1370,7 +1370,7 @@ static void
 opt_sample_rgba_2d(GLcontext *ctx,
                    const struct gl_texture_object *tObj,
                    GLuint n, const GLfloat texcoords[][4],
-                   const GLfloat lambda[], GLchan rgba[][4])
+                   const GLfloat lambda[], GLfloat rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
    const GLfloat width = (GLfloat) img->Width;
@@ -1392,7 +1392,10 @@ opt_sample_rgba_2d(GLcontext *ctx,
       const GLint row = IFLOOR(texcoords[i][1] * height) & rowMask;
       const GLint pos = (row << shift) | col;
       const GLchan *texel = ((GLchan *) img->Data) + (pos << 2);    /* pos*4 */
-      COPY_4V(rgba[i], texel);
+      rgba[i][RCOMP] = CHAN_TO_FLOAT(texel[0]);
+      rgba[i][GCOMP] = CHAN_TO_FLOAT(texel[1]);
+      rgba[i][BCOMP] = CHAN_TO_FLOAT(texel[2]);
+      rgba[i][ACOMP] = CHAN_TO_FLOAT(texel[3]);
    }
 }
 
@@ -1425,7 +1428,6 @@ sample_lambda_2d(GLcontext *ctx,
       case GL_NEAREST:
          if (repeatNoBorderPOT) {
             switch (tImg->TexFormat->MesaFormat) {
-#if 0
             case MESA_FORMAT_RGB:
                opt_sample_rgb_2d(ctx, tObj, m, texcoords + minStart,
                                  NULL, rgba + minStart);
@@ -1434,7 +1436,6 @@ sample_lambda_2d(GLcontext *ctx,
 	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + minStart,
                                   NULL, rgba + minStart);
                break;
-#endif
             default:
                sample_nearest_2d(ctx, tObj, m, texcoords + minStart,
                                  NULL, rgba + minStart );
@@ -1484,7 +1485,6 @@ sample_lambda_2d(GLcontext *ctx,
       case GL_NEAREST:
          if (repeatNoBorderPOT) {
             switch (tImg->TexFormat->MesaFormat) {
-#if 0
             case MESA_FORMAT_RGB:
                opt_sample_rgb_2d(ctx, tObj, m, texcoords + magStart,
                                  NULL, rgba + magStart);
@@ -1493,7 +1493,6 @@ sample_lambda_2d(GLcontext *ctx,
 	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + magStart,
                                   NULL, rgba + magStart);
                break;
-#endif
             default:
                sample_nearest_2d(ctx, tObj, m, texcoords + magStart,
                                  NULL, rgba + magStart );
@@ -3180,7 +3179,6 @@ _swrast_choose_texture_sample_func( GLcontext *ctx,
          }
          else {
             /* check for a few optimized cases */
-#if 0
             const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
             ASSERT(t->MinFilter == GL_NEAREST);
             if (t->WrapS == GL_REPEAT &&
@@ -3197,10 +3195,6 @@ _swrast_choose_texture_sample_func( GLcontext *ctx,
                      img->TexFormat->MesaFormat == MESA_FORMAT_RGBA) {
                return &opt_sample_rgba_2d;
             }
-#else
-            if (0)
-               ;
-#endif
             else {
                return &sample_nearest_2d;
             }
diff --git a/src/mesa/tnl/t_vb_cliptmp.h b/src/mesa/tnl/t_vb_cliptmp.h
index 788fe329ed..618b8b3130 100644
--- a/src/mesa/tnl/t_vb_cliptmp.h
+++ b/src/mesa/tnl/t_vb_cliptmp.h
@@ -127,7 +127,7 @@ TAG(clip_line)( GLcontext *ctx, GLuint v0, GLuint v1, GLubyte mask )
    GLuint p;
    const GLuint v0_orig = v0;
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       LINE_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       LINE_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       LINE_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
@@ -199,7 +199,24 @@ TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
 
    ASSIGN_3V(inlist, v2, v0, v1 ); /* pv rotated to slot zero */
 
-   if (mask & 0x3f) {
+   if (0) {
+      /* print pre-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("pre clip:\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+         assert(!IS_INF_OR_NAN(coord[j][0]));
+         assert(!IS_INF_OR_NAN(coord[j][1]));
+         assert(!IS_INF_OR_NAN(coord[j][2]));
+         assert(!IS_INF_OR_NAN(coord[j][3]));
+      }
+   }
+
+
+   if (mask & CLIP_FRUSTUM_BITS) {
       POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
@@ -227,6 +244,18 @@ TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
       }
    }
 
+   if (0) {
+      /* print post-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("post clip:\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+      }
+   }
+
    tnl->Driver.Render.ClippedPolygon( ctx, inlist, n );
 }
 
@@ -250,7 +279,7 @@ TAG(clip_quad)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLuint v3,
 
    ASSIGN_4V(inlist, v3, v0, v1, v2 ); /* pv rotated to slot zero */
 
-   if (mask & 0x3f) {
+   if (mask & CLIP_FRUSTUM_BITS) {
       POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
       POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
       POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
index 1795f62c32..66c5e13729 100644
--- a/src/mesa/tnl/t_vb_program.c
+++ b/src/mesa/tnl/t_vb_program.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.6
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -46,6 +47,16 @@
 #include "tnl/t_pipeline.h"
 
 
+#ifdef NAN_CHECK
+/** Check for NaNs and very large values */
+static INLINE void
+check_float(float x)
+{
+   assert(!IS_INF_OR_NAN(x));
+   assert(1.0e-15 <= x && x <= 1.0e15);
+}
+#endif
+
 
 /*!
  * Private storage for the vertex program pipeline stage.
@@ -207,7 +218,7 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine)
 {
    /* Input registers get initialized from the current vertex attribs */
    MEMCPY(machine->VertAttribs, ctx->Current.Attrib,
-          MAX_VERTEX_PROGRAM_ATTRIBS * 4 * sizeof(GLfloat));
+          MAX_VERTEX_GENERIC_ATTRIBS * 4 * sizeof(GLfloat));
 
    if (ctx->VertexProgram._Current->IsNVProgram) {
       GLuint i;
@@ -351,6 +362,12 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
 	    const GLuint size = VB->AttribPtr[attr]->size;
 	    const GLuint stride = VB->AttribPtr[attr]->stride;
 	    const GLfloat *data = (GLfloat *) (ptr + stride * i);
+#ifdef NAN_CHECK
+            check_float(data[0]);
+            check_float(data[1]);
+            check_float(data[2]);
+            check_float(data[3]);
+#endif
 	    COPY_CLEAN_4V(machine.VertAttribs[attr], size, data);
 	 }
       }
@@ -361,6 +378,12 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
       /* copy the output registers into the VB->attribs arrays */
       for (j = 0; j < numOutputs; j++) {
          const GLuint attr = outputs[j];
+#ifdef NAN_CHECK
+         check_float(machine.Outputs[attr][0]);
+         check_float(machine.Outputs[attr][1]);
+         check_float(machine.Outputs[attr][2]);
+         check_float(machine.Outputs[attr][3]);
+#endif
          COPY_4V(store->results[attr].data[i], machine.Outputs[attr]);
       }
 #if 0
diff --git a/src/mesa/tnl/t_vb_rendertmp.h b/src/mesa/tnl/t_vb_rendertmp.h
index 2b5f4e93b2..75f6f55bdc 100644
--- a/src/mesa/tnl/t_vb_rendertmp.h
+++ b/src/mesa/tnl/t_vb_rendertmp.h
@@ -82,7 +82,10 @@ static void TAG(render_lines)( GLcontext *ctx,
    INIT(GL_LINES);
    for (j=start+1; j<count; j+=2 ) {
       RESET_STIPPLE;
-      RENDER_LINE( ELT(j-1), ELT(j) );
+      if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+         RENDER_LINE( ELT(j-1), ELT(j) );
+      else
+         RENDER_LINE( ELT(j), ELT(j-1) );
    }
    POSTFIX;
 }
@@ -103,9 +106,12 @@ static void TAG(render_line_strip)( GLcontext *ctx,
       RESET_STIPPLE;
    }
 
-   for (j=start+1; j<count; j++ )
-      RENDER_LINE( ELT(j-1), ELT(j) );
-
+   for (j=start+1; j<count; j++ ) {
+      if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+         RENDER_LINE( ELT(j-1), ELT(j) );
+      else
+         RENDER_LINE( ELT(j), ELT(j-1) );
+   }
    POSTFIX;
 }
 
@@ -125,15 +131,24 @@ static void TAG(render_line_loop)( GLcontext *ctx,
    if (start+1 < count) {
       if (TEST_PRIM_BEGIN(flags)) {
 	 RESET_STIPPLE;
-	 RENDER_LINE( ELT(start), ELT(start+1) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(start), ELT(start+1) );
+         else
+            RENDER_LINE( ELT(start+1), ELT(start) );
       }
 
       for ( i = start+2 ; i < count ; i++) {
-	 RENDER_LINE( ELT(i-1), ELT(i) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(i-1), ELT(i) );
+         else
+            RENDER_LINE( ELT(i), ELT(i-1) );
       }
 
       if ( TEST_PRIM_END(flags)) {
-	 RENDER_LINE( ELT(count-1), ELT(start) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(count-1), ELT(start) );
+         else
+            RENDER_LINE( ELT(start), ELT(count-1) );
       }
    }
 
@@ -156,11 +171,17 @@ static void TAG(render_triangles)( GLcontext *ctx,
 	 /* Leave the edgeflags as supplied by the user.
 	  */
 	 RESET_STIPPLE;
-	 RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1), ELT(j), ELT(j-2) );
       }
    } else {
       for (j=start+2; j<count; j+=3) {
-	 RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1), ELT(j), ELT(j-2) );
       }
    }
    POSTFIX;
@@ -180,26 +201,38 @@ static void TAG(render_tri_strip)( GLcontext *ctx,
    INIT(GL_TRIANGLE_STRIP);
    if (NEED_EDGEFLAG_SETUP) {
       for (j=start+2;j<count;j++,parity^=1) {
-	 GLuint ej2 = ELT(j-2+parity);
-	 GLuint ej1 = ELT(j-1-parity);
-	 GLuint ej = ELT(j);
-	 GLboolean ef2 = EDGEFLAG_GET( ej2 );
-	 GLboolean ef1 = EDGEFLAG_GET( ej1 );
-	 GLboolean ef = EDGEFLAG_GET( ej );
+         GLuint ej2, ej1, ej;
+         GLboolean ef2, ef1, ef;
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT) {
+            ej2 = ELT(j-2+parity);
+            ej1 = ELT(j-1-parity);
+            ej = ELT(j);
+         }
+         else {
+            ej2 = ELT(j-1+parity);
+            ej1 = ELT(j-parity);
+            ej = ELT(j-2);
+         }
+	 ef2 = EDGEFLAG_GET( ej2 );
+	 ef1 = EDGEFLAG_GET( ej1 );
+	 ef = EDGEFLAG_GET( ej );
 	 if (TEST_PRIM_BEGIN(flags)) {
 	    RESET_STIPPLE;
 	 }
 	 EDGEFLAG_SET( ej2, GL_TRUE );
 	 EDGEFLAG_SET( ej1, GL_TRUE );
 	 EDGEFLAG_SET( ej, GL_TRUE );
-	 RENDER_TRI( ej2, ej1, ej );
+         RENDER_TRI( ej2, ej1, ej );
 	 EDGEFLAG_SET( ej2, ef2 );
 	 EDGEFLAG_SET( ej1, ef1 );
 	 EDGEFLAG_SET( ej, ef );
       }
    } else {
       for (j=start+2; j<count ; j++, parity^=1) {
-	 RENDER_TRI( ELT(j-2+parity), ELT(j-1-parity), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2+parity), ELT(j-1-parity), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1+parity), ELT(j-parity), ELT(j-2) );
       }
    }
    POSTFIX;
@@ -232,14 +265,20 @@ static void TAG(render_tri_fan)( GLcontext *ctx,
 	 EDGEFLAG_SET( ejs, GL_TRUE );
 	 EDGEFLAG_SET( ej1, GL_TRUE );
 	 EDGEFLAG_SET( ej, GL_TRUE );
-	 RENDER_TRI( ejs, ej1, ej);
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ejs, ej1, ej);
+         else
+            RENDER_TRI( ej, ejs, ej1);
 	 EDGEFLAG_SET( ejs, efs );
 	 EDGEFLAG_SET( ej1, ef1 );
 	 EDGEFLAG_SET( ej, ef );
       }
    } else {
       for (j=start+2;j<count;j++) {
-	 RENDER_TRI( ELT(start), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(start), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j), ELT(start), ELT(j-1) );
       }
    }
 
@@ -331,11 +370,19 @@ static void TAG(render_quads)( GLcontext *ctx,
 	 /* Use user-specified edgeflags for quads.
 	  */
 	 RESET_STIPPLE;
-	 RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j-1), ELT(j), ELT(j-3) );
       }
    } else {
       for (j=start+3; j<count; j+=4) {
-	 RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j-1), ELT(j), ELT(j-3) );
       }
    }
    POSTFIX;
@@ -367,7 +414,11 @@ static void TAG(render_quad_strip)( GLcontext *ctx,
 	 EDGEFLAG_SET( ELT(j-2), GL_TRUE );
 	 EDGEFLAG_SET( ELT(j-1), GL_TRUE );
 	 EDGEFLAG_SET( ELT(j), GL_TRUE );
-	 RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j), ELT(j-1), ELT(j-3) );
 	 EDGEFLAG_SET( ELT(j-3), ef3 );
 	 EDGEFLAG_SET( ELT(j-2), ef2 );
 	 EDGEFLAG_SET( ELT(j-1), ef1 );
@@ -375,7 +426,11 @@ static void TAG(render_quad_strip)( GLcontext *ctx,
       }
    } else {
       for (j=start+3;j<count;j+=2) {
-	 RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j), ELT(j-1), ELT(j-3) );
       }
    }
    POSTFIX;
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
index ca8190fd05..75c32e0b9b 100644
--- a/src/mesa/vbo/vbo_context.c
+++ b/src/mesa/vbo/vbo_context.c
@@ -28,6 +28,7 @@
 #include "main/imports.h"
 #include "main/mtypes.h"
 #include "main/api_arrayelt.h"
+#include "main/bufferobj.h"
 #include "math/m_eval.h"
 #include "vbo.h"
 #include "vbo_context.h"
@@ -81,7 +82,8 @@ static void init_legacy_currval(GLcontext *ctx)
       cl->Type = GL_FLOAT;
       cl->Format = GL_RGBA;
       cl->Ptr = (const void *)ctx->Current.Attrib[i];
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -106,7 +108,8 @@ static void init_generic_currval(GLcontext *ctx)
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -150,7 +153,8 @@ static void init_mat_currval(GLcontext *ctx)
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
-      cl->BufferObj = ctx->Array.NullBufferObj;
+      _mesa_reference_buffer_object(ctx, &cl->BufferObj,
+                                    ctx->Shared->NullBufferObj);
    }
 }
 
@@ -211,7 +215,7 @@ GLboolean _vbo_CreateContext( GLcontext *ctx )
       for (i = 0; i < 4; i++)
 	 vbo->map_vp_none[28+i] = i;	
       
-      for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      for (i = 0; i < Elements(vbo->map_vp_arb); i++)
 	 vbo->map_vp_arb[i] = i;
    }
 
diff --git a/src/mesa/vbo/vbo_context.h b/src/mesa/vbo/vbo_context.h
index 5c85161caa..8b726dc8ac 100644
--- a/src/mesa/vbo/vbo_context.h
+++ b/src/mesa/vbo/vbo_context.h
@@ -68,8 +68,8 @@ struct vbo_context {
    struct gl_client_array *generic_currval;
    struct gl_client_array *mat_currval;
 
-   GLuint map_vp_none[32];
-   GLuint map_vp_arb[32];
+   GLuint map_vp_none[VERT_ATTRIB_MAX];
+   GLuint map_vp_arb[VERT_ATTRIB_MAX];
 
    GLfloat *current[VBO_ATTRIB_MAX]; /* points into ctx->Current, ctx->Light.Material */
    GLfloat CurrentFloatEdgeFlag;
@@ -92,13 +92,13 @@ static INLINE struct vbo_context *vbo_context(GLcontext *ctx)
    return (struct vbo_context *)(ctx->swtnl_im);
 }
 
-enum {
-   VP_NONE = 1,
-   VP_NV,
-   VP_ARB
-};
 
-static INLINE GLuint get_program_mode( GLcontext *ctx )
+/**
+ * Return VP_x token to indicate whether we're running fixed-function
+ * vertex transformation, an NV vertex program or ARB vertex program/shader.
+ */
+static INLINE enum vp_mode
+get_program_mode( GLcontext *ctx )
 {
    if (!ctx->VertexProgram._Current)
       return VP_NONE;
diff --git a/src/mesa/vbo/vbo_exec.h b/src/mesa/vbo/vbo_exec.h
index 100bb8a5de..e0f44892cf 100644
--- a/src/mesa/vbo/vbo_exec.h
+++ b/src/mesa/vbo/vbo_exec.h
@@ -48,6 +48,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define ERROR_ATTRIB 16
 
 
+/** Current vertex program mode */
+enum vp_mode {
+   VP_NONE,   /**< fixed function */
+   VP_NV,     /**< NV vertex program */
+   VP_ARB     /**< ARB vertex program or GLSL vertex shader */
+};
 
 
 struct vbo_exec_eval1_map {
@@ -103,7 +109,7 @@ struct vbo_exec_context
        * values are squashed down to the 32 attributes passed to the
        * vertex program below:
        */
-      GLuint program_mode;
+      enum vp_mode program_mode;
       GLuint enabled_flags;
       const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
    } vtx;
@@ -116,7 +122,7 @@ struct vbo_exec_context
    } eval;
 
    struct {
-      GLuint program_mode;
+      enum vp_mode program_mode;
       GLuint enabled_flags;
       GLuint array_obj;
 
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index 34e849aaab..b746a77bc1 100644
--- a/src/mesa/vbo/vbo_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -675,7 +675,7 @@ void vbo_use_buffer_objects(GLcontext *ctx)
    GLsizei size = VBO_VERT_BUFFER_SIZE;
 
    /* Make sure this func is only used once */
-   assert(exec->vtx.bufferobj == ctx->Array.NullBufferObj);
+   assert(exec->vtx.bufferobj == ctx->Shared->NullBufferObj);
    if (exec->vtx.buffer_map) {
       _mesa_align_free(exec->vtx.buffer_map);
       exec->vtx.buffer_map = NULL;
@@ -701,7 +701,7 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec )
     */
    _mesa_reference_buffer_object(ctx,
                                  &exec->vtx.bufferobj,
-                                 ctx->Array.NullBufferObj);
+                                 ctx->Shared->NullBufferObj);
 
    ASSERT(!exec->vtx.buffer_map);
    exec->vtx.buffer_map = (GLfloat *)ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE, 64);
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
index de66cdd92f..f4b9b2f744 100644
--- a/src/mesa/vbo/vbo_exec_array.c
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -1,6 +1,7 @@
 /**************************************************************************
  * 
  * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -36,12 +37,13 @@
 
 #include "vbo_context.h"
 
-/* Compute min and max elements for drawelements calls.
+
+/**
+ * Compute min and max elements for glDraw[Range]Elements() calls.
  */
-static void get_minmax_index( GLuint count, GLuint type, 
-			      const GLvoid *indices,
-			      GLuint *min_index,
-			      GLuint *max_index)
+static void
+get_minmax_index(GLuint count, GLuint type, const GLvoid *indices,
+                 GLuint *min_index, GLuint *max_index)
 {
    GLuint i;
 
@@ -89,9 +91,199 @@ static void get_minmax_index( GLuint count, GLuint type,
 }
 
 
-/* Just translate the arrayobj into a sane layout.
+/**
+ * Check that element 'j' of the array has reasonable data.
+ * Map VBO if needed.
+ */
+static void
+check_array_data(GLcontext *ctx, struct gl_client_array *array,
+                 GLuint attrib, GLuint j)
+{
+   if (array->Enabled) {
+      const void *data = array->Ptr;
+      if (array->BufferObj->Name) {
+         if (!array->BufferObj->Pointer) {
+            /* need to map now */
+            array->BufferObj->Pointer = ctx->Driver.MapBuffer(ctx,
+                                                              GL_ARRAY_BUFFER_ARB,
+                                                              GL_READ_ONLY,
+                                                              array->BufferObj);
+         }
+         data = ADD_POINTERS(data, array->BufferObj->Pointer);
+      }
+      switch (array->Type) {
+      case GL_FLOAT:
+         {
+            GLfloat *f = (GLfloat *) ((GLubyte *) data + array->StrideB * j);
+            GLuint k;
+            for (k = 0; k < array->Size; k++) {
+               if (IS_INF_OR_NAN(f[k]) ||
+                   f[k] >= 1.0e20 || f[k] <= -1.0e10) {
+                  _mesa_printf("Bad array data:\n");
+                  _mesa_printf("  Element[%u].%u = %f\n", j, k, f[k]);
+                  _mesa_printf("  Array %u at %p\n", attrib, (void* ) array);
+                  _mesa_printf("  Type 0x%x, Size %d, Stride %d\n",
+                               array->Type, array->Size, array->Stride);
+                  _mesa_printf("  Address/offset %p in Buffer Object %u\n",
+                               array->Ptr, array->BufferObj->Name);
+                  f[k] = 1.0; /* XXX replace the bad value! */
+               }
+               //assert(!IS_INF_OR_NAN(f[k]));
+            }
+         }
+         break;
+      default:
+         ;
+      }
+   }
+}
+
+
+/**
+ * Unmap the buffer object referenced by given array, if mapped.
+ */
+static void
+unmap_array_buffer(GLcontext *ctx, struct gl_client_array *array)
+{
+   if (array->Enabled &&
+       array->BufferObj->Name &&
+       array->BufferObj->Pointer) {
+      ctx->Driver.UnmapBuffer(ctx,
+                              GL_ARRAY_BUFFER_ARB,
+                              array->BufferObj);
+   }
+}
+
+
+/**
+ * Examine the array's data for NaNs, etc.
+ */
+static void
+check_draw_elements_data(GLcontext *ctx, GLsizei count, GLenum elemType,
+                         const void *elements)
+{
+   struct gl_array_object *arrayObj = ctx->Array.ArrayObj;
+   const void *elemMap;
+   GLint i, k;
+
+   if (ctx->Array.ElementArrayBufferObj->Name) {
+      elemMap = ctx->Driver.MapBuffer(ctx,
+                                      GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                      GL_READ_ONLY,
+                                      ctx->Array.ElementArrayBufferObj);
+      elements = ADD_POINTERS(elements, elemMap);
+   }
+
+   for (i = 0; i < count; i++) {
+      GLuint j;
+
+      /* j = element[i] */
+      switch (elemType) {
+      case GL_UNSIGNED_BYTE:
+         j = ((const GLubyte *) elements)[i];
+         break;
+      case GL_UNSIGNED_SHORT:
+         j = ((const GLushort *) elements)[i];
+         break;
+      case GL_UNSIGNED_INT:
+         j = ((const GLuint *) elements)[i];
+         break;
+      default:
+         assert(0);
+      }
+
+      /* check element j of each enabled array */
+      check_array_data(ctx, &arrayObj->Vertex, VERT_ATTRIB_POS, j);
+      check_array_data(ctx, &arrayObj->Normal, VERT_ATTRIB_NORMAL, j);
+      check_array_data(ctx, &arrayObj->Color, VERT_ATTRIB_COLOR0, j);
+      check_array_data(ctx, &arrayObj->SecondaryColor, VERT_ATTRIB_COLOR1, j);
+      for (k = 0; k < Elements(arrayObj->TexCoord); k++) {
+         check_array_data(ctx, &arrayObj->TexCoord[k], VERT_ATTRIB_TEX0 + k, j);
+      }
+      for (k = 0; k < Elements(arrayObj->VertexAttrib); k++) {
+         check_array_data(ctx, &arrayObj->VertexAttrib[k], VERT_ATTRIB_GENERIC0 + k, j);
+      }
+   }
+
+   if (ctx->Array.ElementArrayBufferObj->Name) {
+      ctx->Driver.UnmapBuffer(ctx,
+			      GL_ELEMENT_ARRAY_BUFFER_ARB,
+			      ctx->Array.ElementArrayBufferObj);
+   }
+
+   unmap_array_buffer(ctx, &arrayObj->Vertex);
+   unmap_array_buffer(ctx, &arrayObj->Normal);
+   unmap_array_buffer(ctx, &arrayObj->Color);
+   for (k = 0; k < Elements(arrayObj->TexCoord); k++) {
+      unmap_array_buffer(ctx, &arrayObj->TexCoord[k]);
+   }
+   for (k = 0; k < Elements(arrayObj->VertexAttrib); k++) {
+      unmap_array_buffer(ctx, &arrayObj->VertexAttrib[k]);
+   }
+}
+
+
+/**
+ * Check array data, looking for NaNs, etc.
+ */
+static void
+check_draw_arrays_data(GLcontext *ctx, GLint start, GLsizei count)
+{
+   /* TO DO */
+}
+
+
+/**
+ * Print info/data for glDrawArrays().
+ */
+static void
+print_draw_arrays(GLcontext *ctx, struct vbo_exec_context *exec,
+                  GLenum mode, GLint start, GLsizei count)
+{
+   int i;
+
+   _mesa_printf("vbo_exec_DrawArrays(mode 0x%x, start %d, count %d):\n",
+                mode, start, count);
+
+   for (i = 0; i < 32; i++) {
+      GLuint bufName = exec->array.inputs[i]->BufferObj->Name;
+      GLint stride = exec->array.inputs[i]->Stride;
+      _mesa_printf("attr %2d: size %d stride %d  enabled %d  "
+                   "ptr %p  Bufobj %u\n",
+                   i,
+                   exec->array.inputs[i]->Size,
+                   stride,
+                   /*exec->array.inputs[i]->Enabled,*/
+                   exec->array.legacy_array[i]->Enabled,
+                   exec->array.inputs[i]->Ptr,
+                   bufName);
+
+      if (bufName) {
+         struct gl_buffer_object *buf = _mesa_lookup_bufferobj(ctx, bufName);
+         GLubyte *p = ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER_ARB,
+                                            GL_READ_ONLY_ARB, buf);
+         int offset = (int) (GLintptr) exec->array.inputs[i]->Ptr;
+         float *f = (float *) (p + offset);
+         int *k = (int *) f;
+         int i;
+         int n = (count * stride) / 4;
+         if (n > 32)
+            n = 32;
+         _mesa_printf("  Data at offset %d:\n", offset);
+         for (i = 0; i < n; i++) {
+            _mesa_printf("    float[%d] = 0x%08x %f\n", i, k[i], f[i]);
+         }
+         ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, buf);
+      }
+   }
+}
+
+
+/**
+ * Just translate the arrayobj into a sane layout.
  */
-static void bind_array_obj( GLcontext *ctx )
+static void
+bind_array_obj(GLcontext *ctx)
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
@@ -103,7 +295,7 @@ static void bind_array_obj( GLcontext *ctx )
     * go away.
     */
    exec->array.legacy_array[VERT_ATTRIB_POS] = &arrayObj->Vertex;
-   exec->array.legacy_array[VERT_ATTRIB_WEIGHT] = &vbo->legacy_currval[VERT_ATTRIB_WEIGHT];
+   exec->array.legacy_array[VERT_ATTRIB_WEIGHT] = &arrayObj->Weight;
    exec->array.legacy_array[VERT_ATTRIB_NORMAL] = &arrayObj->Normal;
    exec->array.legacy_array[VERT_ATTRIB_COLOR0] = &arrayObj->Color;
    exec->array.legacy_array[VERT_ATTRIB_COLOR1] = &arrayObj->SecondaryColor;
@@ -115,11 +307,10 @@ static void bind_array_obj( GLcontext *ctx )
    }
    exec->array.legacy_array[VERT_ATTRIB_EDGEFLAG] = &arrayObj->EdgeFlag;
 
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
+   for (i = 0; i < Elements(arrayObj->TexCoord); i++)
       exec->array.legacy_array[VERT_ATTRIB_TEX0 + i] = &arrayObj->TexCoord[i];
 
-   for (i = 0; i < MAX_VERTEX_GENERIC_ATTRIBS; i++) {
-      assert(i < Elements(arrayObj->VertexAttrib));
+   for (i = 0; i < Elements(arrayObj->VertexAttrib); i++) {
       assert(i < Elements(exec->array.generic_array));
       exec->array.generic_array[i] = &arrayObj->VertexAttrib[i];
    }
@@ -127,7 +318,9 @@ static void bind_array_obj( GLcontext *ctx )
    exec->array.array_obj = arrayObj->Name;
 }
 
-static void recalculate_input_bindings( GLcontext *ctx )
+
+static void
+recalculate_input_bindings(GLcontext *ctx)
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
@@ -140,9 +333,10 @@ static void recalculate_input_bindings( GLcontext *ctx )
 
    switch (exec->array.program_mode) {
    case VP_NONE:
-      /* When no vertex program is active, we put the material values
-       * into the generic slots.  This is the only situation where
-       * material values are available as per-vertex attributes.
+      /* When no vertex program is active (or the vertex program is generated
+       * from fixed-function state).  We put the material values into the
+       * generic slots.  This is the only situation where material values
+       * are available as per-vertex attributes.
        */
       for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
@@ -165,8 +359,8 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
          const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
       }
-
       break;
+
    case VP_NV:
       /* NV_vertex_program - attribute arrays alias and override
        * conventional, legacy arrays.  No materials, and the generic
@@ -190,11 +384,11 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
          const_inputs |= 1 << i;
       }
-
       break;
+
    case VP_ARB:
-      /* ARB_vertex_program - Only the attribute zero (position) array
-       * aliases and overrides the legacy position array.  
+      /* GL_ARB_vertex_program or GLSL vertex shader - Only the generic[0]
+       * attribute array aliases and overrides the legacy position array.  
        *
        * Otherwise, legacy attributes available in the legacy slots,
        * generic attributes in the generic slots and materials are not
@@ -209,7 +403,6 @@ static void recalculate_input_bindings( GLcontext *ctx )
          const_inputs |= 1 << 0;
       }
 
-
       for (i = 1; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
@@ -234,7 +427,9 @@ static void recalculate_input_bindings( GLcontext *ctx )
    _mesa_set_varying_vp_inputs( ctx, ~const_inputs );
 }
 
-static void bind_arrays( GLcontext *ctx )
+
+static void
+bind_arrays(GLcontext *ctx)
 {
 #if 0
    if (ctx->Array.ArrayObj.Name != exec->array.array_obj) {
@@ -279,9 +474,18 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
       return;
    }
 
+#if 0
+   check_draw_arrays_data(ctx, start, count);
+#else
+   (void) check_draw_arrays_data;
+#endif
+
    bind_arrays( ctx );
 
-   /* Again...
+   /* Again... because we may have changed the bitmask of per-vertex varying
+    * attributes.  If we regenerate the fixed-function vertex program now
+    * we may be able to prune down the number of vertex attributes which we
+    * need in the shader.
     */
    if (ctx->NewState)
       _mesa_update_state( ctx );
@@ -295,50 +499,72 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
    prim[0].count = count;
    prim[0].indexed = 0;
 
-   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, NULL, start, start + count - 1 );
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, NULL,
+                    start, start + count - 1 );
 
 #if 0
-   {
-      int i;
-
-      _mesa_printf("vbo_exec_DrawArrays(mode 0x%x, start %d, count %d):\n",
-                   mode, start, count);
-
-      for (i = 0; i < 32; i++) {
-         GLuint bufName = exec->array.inputs[i]->BufferObj->Name;
-         GLint stride = exec->array.inputs[i]->Stride;
-         _mesa_printf("attr %2d: size %d stride %d  enabled %d  "
-                      "ptr %p  Bufobj %u\n",
-                      i,
-                      exec->array.inputs[i]->Size,
-                      stride,
-                      /*exec->array.inputs[i]->Enabled,*/
-                      exec->array.legacy_array[i]->Enabled,
-                      exec->array.inputs[i]->Ptr,
-                      bufName);
-         
-         if (bufName) {
-            struct gl_buffer_object *buf = _mesa_lookup_bufferobj(ctx, bufName);
-            GLubyte *p = ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER_ARB,
-                                            GL_READ_ONLY_ARB, buf);
-            int offset = (int) exec->array.inputs[i]->Ptr;
-            float *f = (float *) (p + offset);
-            int *k = (int *) f;
-            int i;
-            int n = (count * stride) / 4;
-            if (n > 32)
-               n = 32;
-            _mesa_printf("  Data at offset %d:\n", offset);
-            for (i = 0; i < n; i++) {
-               _mesa_printf("    float[%d] = 0x%08x %f\n", i, k[i], f[i]);
-            }
-            ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, buf);
+   print_draw_arrays(ctx, exec, mode, start, count);
+#else
+   (void) print_draw_arrays;
+#endif
+}
+
+
+/**
+ * Map GL_ELEMENT_ARRAY_BUFFER and print contents.
+ */
+static void
+dump_element_buffer(GLcontext *ctx, GLenum type)
+{
+   const GLvoid *map = ctx->Driver.MapBuffer(ctx,
+                                             GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                             GL_READ_ONLY,
+                                             ctx->Array.ElementArrayBufferObj);
+   switch (type) {
+   case GL_UNSIGNED_BYTE:
+      {
+         const GLubyte *us = (const GLubyte *) map;
+         GLuint i;
+         for (i = 0; i < ctx->Array.ElementArrayBufferObj->Size; i++) {
+            _mesa_printf("%02x ", us[i]);
+            if (i % 32 == 31)
+               _mesa_printf("\n");
          }
+         _mesa_printf("\n");
       }
+      break;
+   case GL_UNSIGNED_SHORT:
+      {
+         const GLushort *us = (const GLushort *) map;
+         GLuint i;
+         for (i = 0; i < ctx->Array.ElementArrayBufferObj->Size / 2; i++) {
+            _mesa_printf("%04x ", us[i]);
+            if (i % 16 == 15)
+               _mesa_printf("\n");
+         }
+         _mesa_printf("\n");
+      }
+      break;
+   case GL_UNSIGNED_INT:
+      {
+         const GLuint *us = (const GLuint *) map;
+         GLuint i;
+         for (i = 0; i < ctx->Array.ElementArrayBufferObj->Size / 4; i++) {
+            _mesa_printf("%08x ", us[i]);
+            if (i % 8 == 7)
+               _mesa_printf("\n");
+         }
+         _mesa_printf("\n");
+      }
+      break;
+   default:
+      ;
    }
-#endif
-}
 
+   ctx->Driver.UnmapBuffer(ctx,
+                           GL_ELEMENT_ARRAY_BUFFER_ARB,
+                           ctx->Array.ElementArrayBufferObj);
+}
 
 
 static void GLAPIENTRY
@@ -352,9 +578,41 @@ vbo_exec_DrawRangeElements(GLenum mode,
    struct _mesa_index_buffer ib;
    struct _mesa_prim prim[1];
 
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
+   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count,
+                                          type, indices ))
       return;
 
+   if (end >= ctx->Array.ArrayObj->_MaxElement) {
+      /* the max element is out of bounds of one or more enabled arrays */
+      _mesa_warning(ctx, "glDraw[Range]Elements(start %u, end %u, count %d, "
+                    "type 0x%x, indices=%p)\n"
+                    "\tindex=%u is out of bounds (max=%u)  "
+                    "Element Buffer %u (size %d)",
+                    start, end, count, type, indices, end,
+                    ctx->Array.ArrayObj->_MaxElement - 1,
+                    ctx->Array.ElementArrayBufferObj->Name,
+                    ctx->Array.ElementArrayBufferObj->Size);
+
+      if (0)
+         dump_element_buffer(ctx, type);
+
+      if (0)
+         _mesa_print_arrays(ctx);
+      return;
+   }
+   else if (0) {
+      _mesa_printf("glDraw[Range]Elements"
+                   "(start %u, end %u, type 0x%x, count %d) ElemBuf %u\n",
+                   start, end, type, count,
+                   ctx->Array.ElementArrayBufferObj->Name);
+   }
+
+#if 0
+   check_draw_elements_data(ctx, count, type, indices);
+#else
+   (void) check_draw_elements_data;
+#endif
+
    FLUSH_CURRENT( ctx, 0 );
 
    if (ctx->NewState)
@@ -418,8 +676,10 @@ vbo_exec_DrawRangeElements(GLenum mode,
    vbo->draw_prims( ctx, exec->array.inputs, prim, 1, &ib, start, end );
 }
 
+
 static void GLAPIENTRY
-vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices)
+vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type,
+                      const GLvoid *indices)
 {
    GET_CURRENT_CONTEXT(ctx);
    GLuint min_index = 0;
@@ -435,11 +695,12 @@ vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *ind
 
    if (ctx->Array.ElementArrayBufferObj->Name) {
       const GLvoid *map = ctx->Driver.MapBuffer(ctx,
-						 GL_ELEMENT_ARRAY_BUFFER_ARB,
-						 GL_READ_ONLY,
-						 ctx->Array.ElementArrayBufferObj);
+                                                GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                                GL_READ_ONLY,
+                                                ctx->Array.ElementArrayBufferObj);
 
-      get_minmax_index(count, type, ADD_POINTERS(map, indices), &min_index, &max_index);
+      get_minmax_index(count, type, ADD_POINTERS(map, indices),
+                       &min_index, &max_index);
 
       ctx->Driver.UnmapBuffer(ctx,
 			      GL_ELEMENT_ARRAY_BUFFER_ARB,
@@ -457,10 +718,8 @@ vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *ind
  * Initialization
  */
 
-
-
-
-void vbo_exec_array_init( struct vbo_exec_context *exec )
+void
+vbo_exec_array_init( struct vbo_exec_context *exec )
 {
 #if 1
    exec->vtxfmt.DrawArrays = vbo_exec_DrawArrays;
@@ -474,7 +733,8 @@ void vbo_exec_array_init( struct vbo_exec_context *exec )
 }
 
 
-void vbo_exec_array_destroy( struct vbo_exec_context *exec )
+void
+vbo_exec_array_destroy( struct vbo_exec_context *exec )
 {
    /* nothing to do */
 }
diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index c53a4eee00..18419928b2 100644
--- a/src/mesa/vbo/vbo_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -35,7 +35,8 @@
 #include "vbo_context.h"
 
 
-static void vbo_exec_debug_verts( struct vbo_exec_context *exec )
+static void
+vbo_exec_debug_verts( struct vbo_exec_context *exec )
 {
    GLuint count = exec->vtx.vert_count;
    GLuint i;
@@ -64,19 +65,19 @@ static void vbo_exec_debug_verts( struct vbo_exec_context *exec )
  * NOTE: Need to have calculated primitives by this point -- do it on the fly.
  * NOTE: Old 'parity' issue is gone.
  */
-static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
+static GLuint
+vbo_copy_vertices( struct vbo_exec_context *exec )
 {
    GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
    GLuint ovf, i;
    GLuint sz = exec->vtx.vertex_size;
    GLfloat *dst = exec->vtx.copied.buffer;
-   GLfloat *src = (exec->vtx.buffer_map + 
-		   exec->vtx.prim[exec->vtx.prim_count-1].start * 
-		   exec->vtx.vertex_size);
+   const GLfloat *src = (exec->vtx.buffer_map + 
+                         exec->vtx.prim[exec->vtx.prim_count-1].start * 
+                         exec->vtx.vertex_size);
 
 
-   switch( exec->ctx->Driver.CurrentExecPrimitive )
-   {
+   switch (exec->ctx->Driver.CurrentExecPrimitive) {
    case GL_POINTS:
       return 0;
    case GL_LINES:
@@ -95,8 +96,9 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
 	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
       return i;
    case GL_LINE_STRIP:
-      if (nr == 0) 
+      if (nr == 0) {
 	 return 0;
+      }
       else {
 	 _mesa_memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
 	 return 1;
@@ -104,12 +106,14 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
    case GL_LINE_LOOP:
    case GL_TRIANGLE_FAN:
    case GL_POLYGON:
-      if (nr == 0) 
+      if (nr == 0) {
 	 return 0;
+      }
       else if (nr == 1) {
 	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
 	 return 1;
-      } else {
+      }
+      else {
 	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
 	 _mesa_memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
 	 return 2;
@@ -122,9 +126,15 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
       /* fallthrough */
    case GL_QUAD_STRIP:
       switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
+      case 0:
+         ovf = 0;
+         break;
+      case 1:
+         ovf = 1;
+         break;
+      default:
+         ovf = 2 + (nr & 1);
+         break;
       }
       for (i = 0 ; i < ovf ; i++)
 	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
@@ -141,13 +151,14 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
 
 /* TODO: populate these as the vertex is defined:
  */
-static void vbo_exec_bind_arrays( GLcontext *ctx )
+static void
+vbo_exec_bind_arrays( GLcontext *ctx )
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
    struct gl_client_array *arrays = exec->vtx.arrays;
-   GLuint count = exec->vtx.vert_count;
-   GLubyte *data = (GLubyte *)exec->vtx.buffer_map;
+   const GLuint count = exec->vtx.vert_count;
+   const GLubyte *data = (GLubyte *) exec->vtx.buffer_map;
    const GLuint *map;
    GLuint attr;
    GLbitfield varying_inputs = 0x0;
@@ -227,7 +238,7 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 	 arrays[attr]._MaxElement = count; /* ??? */
 
 	 data += exec->vtx.attrsz[src] * sizeof(GLfloat);
-         varying_inputs |= 1<<attr;
+         varying_inputs |= 1 << attr;
       }
    }
 
@@ -235,18 +246,19 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 }
 
 
-static void vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
+static void
+vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
 {
    GLenum target = GL_ARRAY_BUFFER_ARB;
 
    if (exec->vtx.bufferobj->Name) {
       GLcontext *ctx = exec->ctx;
       
-      if(ctx->Driver.FlushMappedBufferRange) {
+      if (ctx->Driver.FlushMappedBufferRange) {
          GLintptr offset = exec->vtx.buffer_used - exec->vtx.bufferobj->Offset;
          GLsizeiptr length = (exec->vtx.buffer_ptr - exec->vtx.buffer_map) * sizeof(float);
 
-         if(length)
+         if (length)
             ctx->Driver.FlushMappedBufferRange(ctx, target,
                                                offset, length,
                                                exec->vtx.bufferobj);
@@ -265,7 +277,9 @@ static void vbo_exec_vtx_unmap( struct vbo_exec_context *exec )
    }
 }
 
-void vbo_exec_vtx_map( struct vbo_exec_context *exec )
+
+void
+vbo_exec_vtx_map( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
    const GLenum target = GL_ARRAY_BUFFER_ARB;
@@ -287,8 +301,7 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
    }
 
    if (VBO_VERT_BUFFER_SIZE > exec->vtx.buffer_used + 1024 &&
-       ctx->Driver.MapBufferRange)
-   {
+       ctx->Driver.MapBufferRange) {
       exec->vtx.buffer_map = 
          (GLfloat *)ctx->Driver.MapBufferRange(ctx, 
                                                target, 
@@ -314,13 +327,15 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
                                                   0, VBO_VERT_BUFFER_SIZE,
                                                   accessRange,
                                                   exec->vtx.bufferobj);
-      else
+      if (!exec->vtx.buffer_map)
          exec->vtx.buffer_map =
             (GLfloat *)ctx->Driver.MapBuffer(ctx, target, access, exec->vtx.bufferobj);
+      assert(exec->vtx.buffer_map);
       exec->vtx.buffer_ptr = exec->vtx.buffer_map;
    }
 
-   if (0) _mesa_printf("map %d..\n", exec->vtx.buffer_used);
+   if (0)
+      _mesa_printf("map %d..\n", exec->vtx.buffer_used);
 }
 
 
@@ -328,13 +343,12 @@ void vbo_exec_vtx_map( struct vbo_exec_context *exec )
 /**
  * Execute the buffer and save copied verts.
  */
-void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
-                         GLboolean unmap )
+void
+vbo_exec_vtx_flush( struct vbo_exec_context *exec, GLboolean unmap )
 {
    if (0)
       vbo_exec_debug_verts( exec );
 
-
    if (exec->vtx.prim_count && 
        exec->vtx.vert_count) {
 
@@ -355,8 +369,9 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
             vbo_exec_vtx_unmap( exec );
          }
 
-         if (0) _mesa_printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
-                      exec->vtx.vert_count);
+         if (0)
+            _mesa_printf("%s %d %d\n", __FUNCTION__, exec->vtx.prim_count,
+                         exec->vtx.vert_count);
 
 	 vbo_context(ctx)->draw_prims( ctx, 
 				       exec->vtx.inputs, 
@@ -370,7 +385,7 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
           */
          if (exec->vtx.bufferobj->Name && !unmap) {
             vbo_exec_vtx_map( exec );
-          }
+         }
       }
    }
 
@@ -382,14 +397,12 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec,
       vbo_exec_vtx_unmap( exec );
    }
 
-
    if (unmap) 
       exec->vtx.max_vert = 0;
    else
       exec->vtx.max_vert = ((VBO_VERT_BUFFER_SIZE - exec->vtx.buffer_used) / 
                             (exec->vtx.vertex_size * sizeof(GLfloat)));
 
-
    exec->vtx.buffer_ptr = exec->vtx.buffer_map;
    exec->vtx.prim_count = 0;
    exec->vtx.vert_count = 0;
diff --git a/src/mesa/vbo/vbo_rebase.c b/src/mesa/vbo/vbo_rebase.c
index dae778e741..ea87dede64 100644
--- a/src/mesa/vbo/vbo_rebase.c
+++ b/src/mesa/vbo/vbo_rebase.c
@@ -161,7 +161,7 @@ void vbo_rebase_prims( GLcontext *ctx,
 				 GL_ELEMENT_ARRAY_BUFFER,
 				 ib->obj);
 
-      tmp_ib.obj = ctx->Array.NullBufferObj;
+      tmp_ib.obj = ctx->Shared->NullBufferObj;
       tmp_ib.ptr = tmp_indices;
       tmp_ib.count = ib->count;
       tmp_ib.type = ib->type;
diff --git a/src/mesa/vbo/vbo_save_api.c b/src/mesa/vbo/vbo_save_api.c
index a7f2706206..d00d304d2e 100644
--- a/src/mesa/vbo/vbo_save_api.c
+++ b/src/mesa/vbo/vbo_save_api.c
@@ -1158,10 +1158,10 @@ static void vbo_print_vertex_list( GLcontext *ctx, void *data )
    GLuint i;
    (void) ctx;
 
-   _mesa_debug(NULL, "VBO-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
-               node->count,
-	       node->prim_count,
-	       node->vertex_size);
+   _mesa_printf("VBO-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
+		node->count,
+		node->prim_count,
+		node->vertex_size);
 
    for (i = 0 ; i < node->prim_count ; i++) {
       struct _mesa_prim *prim = &node->prim[i];
diff --git a/src/mesa/vbo/vbo_split_copy.c b/src/mesa/vbo/vbo_split_copy.c
index b9a5c56987..d7ffebf607 100644
--- a/src/mesa/vbo/vbo_split_copy.c
+++ b/src/mesa/vbo/vbo_split_copy.c
@@ -31,6 +31,7 @@
 
 #include "main/glheader.h"
 #include "main/imports.h"
+#include "main/image.h"
 #include "main/macros.h"
 #include "main/enums.h"
 #include "main/mtypes.h"
@@ -41,7 +42,8 @@
 
 #define ELT_TABLE_SIZE 16
 
-/* Used for vertex-level splitting of indexed buffers.  Note that
+/**
+ * Used for vertex-level splitting of indexed buffers.  Note that
  * non-indexed primitives may be converted to indexed in some cases
  * (eg loops, fans) in order to use this splitting path.
  */
@@ -73,23 +75,21 @@ struct copy_context {
    GLuint *translated_elt_buf;
    const GLuint *srcelt;
 
-   /* A baby hash table to avoid re-emitting (some) duplicate
+   /** A baby hash table to avoid re-emitting (some) duplicate
     * vertices when splitting indexed primitives.
     */
    struct { 
       GLuint in;
       GLuint out;
    } vert_cache[ELT_TABLE_SIZE];
-      
 
    GLuint vertex_size;
    GLubyte *dstbuf;
-   GLubyte *dstptr;		/* dstptr == dstbuf + dstelt_max * vertsize */
-   GLuint dstbuf_size;	/* in vertices */
-   GLuint dstbuf_nr;		/* count of emitted vertices, also the
-				 * largest value in dstelt.  Our
-				 * MaxIndex.
-				 */
+   GLubyte *dstptr;     /**< dstptr == dstbuf + dstelt_max * vertsize */
+   GLuint dstbuf_size;  /**< in vertices */
+   GLuint dstbuf_nr;    /**< count of emitted vertices, also the largest value
+                         * in dstelt.  Our MaxIndex.
+                         */
 
    GLuint *dstelt;
    GLuint dstelt_nr;
@@ -102,32 +102,19 @@ struct copy_context {
 };
 
 
-static GLuint type_size( GLenum type )
-{
-   switch(type) {
-   case GL_BYTE: return sizeof(GLbyte);
-   case GL_UNSIGNED_BYTE: return sizeof(GLubyte);
-   case GL_SHORT: return sizeof(GLshort);
-   case GL_UNSIGNED_SHORT: return sizeof(GLushort);
-   case GL_INT: return sizeof(GLint);
-   case GL_UNSIGNED_INT: return sizeof(GLuint);
-   case GL_FLOAT: return sizeof(GLfloat);
-   case GL_DOUBLE: return sizeof(GLdouble);
-   default: return 0;
-   }
-}
-
 static GLuint attr_size( const struct gl_client_array *array )
 {
-   return array->Size * type_size(array->Type);
+   return array->Size * _mesa_sizeof_type(array->Type);
 }
 
 
-/* Starts returning true slightly before the buffer fills, to ensure
+/**
+ * Starts returning true slightly before the buffer fills, to ensure
  * that there is sufficient room for any remaining vertices to finish
  * off the prim:
  */
-static GLboolean check_flush( struct copy_context *copy )
+static GLboolean
+check_flush( struct copy_context *copy )
 {
    GLenum mode = copy->dstprim[copy->dstprim_nr].mode; 
 
@@ -145,7 +132,44 @@ static GLboolean check_flush( struct copy_context *copy )
    return GL_FALSE;
 }
 
-static void flush( struct copy_context *copy )
+
+/**
+ * Dump the parameters/info for a vbo->draw() call.
+ */
+static void
+dump_draw_info(GLcontext *ctx,
+               const struct gl_client_array **arrays,
+               const struct _mesa_prim *prims,
+               GLuint nr_prims,
+               const struct _mesa_index_buffer *ib,
+               GLuint min_index,
+               GLuint max_index)
+{
+   GLuint i, j;
+
+   _mesa_printf("VBO Draw:\n");
+   for (i = 0; i < nr_prims; i++) {
+      _mesa_printf("Prim %u of %u\n", i, nr_prims);
+      _mesa_printf("  Prim mode 0x%x\n", prims[i].mode);
+      _mesa_printf("  IB: %p\n", (void*) ib);
+      for (j = 0; j < VERT_ATTRIB_MAX; j++) {
+         _mesa_printf("    array %d at %p:\n", j, (void*) arrays[j]);
+         _mesa_printf("      enabled %d, ptr %p, size %d, type 0x%x, stride %d\n",
+                      arrays[j]->Enabled, arrays[j]->Ptr,
+                      arrays[j]->Size, arrays[j]->Type, arrays[j]->StrideB);
+         if (0) {
+            GLint k = prims[i].start + prims[i].count - 1;
+            GLfloat *last = (GLfloat *) (arrays[j]->Ptr + arrays[j]->Stride * k);
+            _mesa_printf("        last: %f %f %f\n",
+                         last[0], last[1], last[2]);
+         }
+      }
+   }
+}
+
+
+static void
+flush( struct copy_context *copy )
 {
    GLuint i;
 
@@ -153,6 +177,18 @@ static void flush( struct copy_context *copy )
     */
    copy->dstib.count = copy->dstelt_nr;
 
+#if 0
+   dump_draw_info(copy->ctx,
+                  copy->dstarray_ptr,
+                  copy->dstprim,
+                  copy->dstprim_nr,
+                  &copy->dstib,
+                  0,
+                  copy->dstbuf_nr);
+#else
+   (void) dump_draw_info;
+#endif
+
    copy->draw( copy->ctx,
 	       copy->dstarray_ptr,
 	       copy->dstprim,
@@ -175,8 +211,11 @@ static void flush( struct copy_context *copy )
 }
 
 
-
-static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
+/**
+ * Called at begin of each primitive during replay.
+ */
+static void
+begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
 {
    struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
 
@@ -187,10 +226,12 @@ static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag
 }
 
 
-/* Use a hashtable to attempt to identify recently-emitted vertices
+/**
+ * Use a hashtable to attempt to identify recently-emitted vertices
  * and avoid re-emitting them.
  */
-static GLuint elt(struct copy_context *copy, GLuint elt_idx)
+static GLuint
+elt(struct copy_context *copy, GLuint elt_idx)
 {
    GLuint elt = copy->srcelt[elt_idx];
    GLuint slot = elt & (ELT_TABLE_SIZE-1);
@@ -213,6 +254,17 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
 	 memcpy(csr, srcptr, copy->varying[i].size);
 	 csr += copy->varying[i].size;
 
+#ifdef NAN_CHECK
+         if (srcarray->Type == GL_FLOAT) {
+            GLuint k;
+            GLfloat *f = (GLfloat *) srcptr;
+            for (k = 0; k < srcarray->Size; k++) {
+               assert(!IS_INF_OR_NAN(f[k]));
+               assert(f[k] <= 1.0e20 && f[k] >= -1.0e20);
+            }
+         }
+#endif
+
 	 if (0) 
 	 {
 	    const GLuint *f = (const GLuint *)srcptr;
@@ -222,7 +274,6 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
 	       _mesa_printf("%x ", f[j]);
 	    _mesa_printf("\n");
 	 }
-	       
       }
 
       copy->vert_cache[slot].in = elt;
@@ -230,9 +281,8 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
       copy->dstptr += copy->vertex_size;
 
       assert(csr == copy->dstptr);
-      assert(copy->dstptr == (copy->dstbuf + 
-				    copy->dstbuf_nr * 
-				    copy->vertex_size));
+      assert(copy->dstptr == (copy->dstbuf +
+                              copy->dstbuf_nr * copy->vertex_size));
    }
 /*    else */
 /*       _mesa_printf("  --> reuse vertex\n"); */
@@ -242,7 +292,12 @@ static GLuint elt(struct copy_context *copy, GLuint elt_idx)
    return check_flush(copy);
 }
 
-static void end( struct copy_context *copy, GLboolean end_flag )
+
+/**
+ * Called at end of each primitive during replay.
+ */
+static void
+end( struct copy_context *copy, GLboolean end_flag )
 {
    struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
 
@@ -257,8 +312,8 @@ static void end( struct copy_context *copy, GLboolean end_flag )
 }
 
 
-
-static void replay_elts( struct copy_context *copy )
+static void
+replay_elts( struct copy_context *copy )
 {
    GLuint i, j, k;
    GLboolean split;
@@ -362,7 +417,8 @@ static void replay_elts( struct copy_context *copy )
 }
 
 
-static void replay_init( struct copy_context *copy )
+static void
+replay_init( struct copy_context *copy )
 {
    GLcontext *ctx = copy->ctx;
    GLuint i;
@@ -388,10 +444,7 @@ static void replay_init( struct copy_context *copy )
 	 copy->vertex_size += attr_size(copy->array[i]);
       
 	 if (vbo->Name && !vbo->Pointer) 
-	    ctx->Driver.MapBuffer(ctx,
-				  GL_ARRAY_BUFFER_ARB, 
-				  GL_WRITE_ONLY, /* XXX */
-				  vbo);
+	    ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY, vbo);
 
 	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
 						 copy->array[i]->Ptr);
@@ -405,12 +458,11 @@ static void replay_init( struct copy_context *copy )
     * do it internally.
     */
    if (copy->ib->obj->Name && !copy->ib->obj->Pointer) 
-      ctx->Driver.MapBuffer(ctx, 
-			    GL_ARRAY_BUFFER_ARB, /* XXX */
-			    GL_WRITE_ONLY, /* XXX */
+      ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY,
 			    copy->ib->obj);
 
-   srcptr = (const GLubyte *)ADD_POINTERS(copy->ib->obj->Pointer, copy->ib->ptr);
+   srcptr = (const GLubyte *) ADD_POINTERS(copy->ib->obj->Pointer,
+                                           copy->ib->ptr);
 
    switch (copy->ib->type) {
    case GL_UNSIGNED_BYTE:
@@ -434,7 +486,6 @@ static void replay_init( struct copy_context *copy )
       copy->srcelt = (const GLuint *)srcptr;
       break;
    }
-   
 
    /* Figure out the maximum allowed vertex buffer size:
     */
@@ -449,8 +500,7 @@ static void replay_init( struct copy_context *copy )
     *
     * XXX:  This should be a VBO!
     */
-   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * 
-			       copy->vertex_size);   
+   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * copy->vertex_size);   
    copy->dstptr = copy->dstbuf;
 
    /* Setup new vertex arrays to point into the output buffer: 
@@ -467,7 +517,7 @@ static void replay_init( struct copy_context *copy )
       dst->Ptr = copy->dstbuf + offset;
       dst->Enabled = GL_TRUE;
       dst->Normalized = src->Normalized; 
-      dst->BufferObj = ctx->Array.NullBufferObj;
+      dst->BufferObj = ctx->Shared->NullBufferObj;
       dst->_MaxElement = copy->dstbuf_size; /* may be less! */
 
       offset += copy->varying[i].size;
@@ -487,12 +537,16 @@ static void replay_init( struct copy_context *copy )
     */
    copy->dstib.count = 0;	/* duplicates dstelt_nr */
    copy->dstib.type = GL_UNSIGNED_INT;
-   copy->dstib.obj = ctx->Array.NullBufferObj;
+   copy->dstib.obj = ctx->Shared->NullBufferObj;
    copy->dstib.ptr = copy->dstelt;
 }
 
 
-static void replay_finish( struct copy_context *copy )
+/**
+ * Free up everything allocated during split/replay.
+ */
+static void
+replay_finish( struct copy_context *copy )
 {
    GLcontext *ctx = copy->ctx;
    GLuint i;
@@ -502,25 +556,26 @@ static void replay_finish( struct copy_context *copy )
    _mesa_free(copy->translated_elt_buf);
    _mesa_free(copy->dstbuf);
    _mesa_free(copy->dstelt);
-   
+
    /* Unmap VBO's 
     */
    for (i = 0; i < copy->nr_varying; i++) {
       struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
-
       if (vbo->Name && vbo->Pointer) 
-	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, vbo);
+	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, vbo);
    }
 
    /* Unmap index buffer:
     */
    if (copy->ib->obj->Name && copy->ib->obj->Pointer) {
-      ctx->Driver.UnmapBuffer(ctx, 
-			      GL_ARRAY_BUFFER_ARB, /* XXX */
-			      copy->ib->obj);
+      ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, copy->ib->obj);
    }
 }
 
+
+/**
+ * Split VBO into smaller pieces, draw the pieces.
+ */
 void vbo_split_copy( GLcontext *ctx,
 		     const struct gl_client_array *arrays[],
 		     const struct _mesa_prim *prim,
@@ -546,13 +601,11 @@ void vbo_split_copy( GLcontext *ctx,
    copy.draw = draw;
    copy.limits = limits;
 
-
    /* Clear the vertex cache:
     */
    for (i = 0; i < ELT_TABLE_SIZE; i++)
       copy.vert_cache[i].in = ~0;
 
-
    replay_init(&copy);
    replay_elts(&copy);
    replay_finish(&copy);
diff --git a/src/mesa/vbo/vbo_split_inplace.c b/src/mesa/vbo/vbo_split_inplace.c
index fbc856e93b..266bc56c82 100644
--- a/src/mesa/vbo/vbo_split_inplace.c
+++ b/src/mesa/vbo/vbo_split_inplace.c
@@ -63,35 +63,19 @@ static void flush_vertex( struct split_context *split )
    if (!split->dstprim_nr) 
       return;
 
-   if (split->ib) {
-      /* This should basically be multipass rendering over the same
-       * unchanging set of VBO's.  Would like the driver not to
-       * re-upload the data, or swtnl not to re-transform the
-       * vertices.
-       */
-      assert(split->max_index - split->min_index < split->limits->max_verts);
-      min_index = split->min_index;
-      max_index = split->max_index;
-   }
-   else {
-      /* Non-indexed rendering.  Cannot assume that the primitives are
-       * ordered by increasing vertex, because of entrypoints like
-       * MultiDrawArrays.
-       */
-      GLuint i;
-      min_index = split->dstprim[0].start;
-      max_index = min_index + split->dstprim[0].count - 1;
+   GLuint i;
+   min_index = split->dstprim[0].start;
+   max_index = min_index + split->dstprim[0].count - 1;
 
-      for (i = 1; i < split->dstprim_nr; i++) {
-	 GLuint tmp_min = split->dstprim[i].start;
-	 GLuint tmp_max = tmp_min + split->dstprim[i].count - 1;
+   for (i = 1; i < split->dstprim_nr; i++) {
+      GLuint tmp_min = split->dstprim[i].start;
+      GLuint tmp_max = tmp_min + split->dstprim[i].count - 1;
 
-	 if (tmp_min < min_index) 
-	    min_index = tmp_min;
+      if (tmp_min < min_index)
+	 min_index = tmp_min;
 
-	 if (tmp_max > max_index) 
-	    max_index = tmp_max;
-      }
+      if (tmp_max > max_index)
+	 max_index = tmp_max;
    }
 
    assert(max_index >= min_index);
@@ -221,7 +205,7 @@ static void split_prims( struct split_context *split)
 
 	 ib.count = count;
 	 ib.type = GL_UNSIGNED_INT;
-	 ib.obj = split->ctx->Array.NullBufferObj;
+	 ib.obj = split->ctx->Shared->NullBufferObj;
 	 ib.ptr = elts;
 	    
 	 tmpprim = *prim;
diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S
index fe05549c1d..90ad36a8f3 100644
--- a/src/mesa/x86-64/glapi_x86-64.S
+++ b/src/mesa/x86-64/glapi_x86-64.S
@@ -21157,12 +21157,205 @@ GL_PREFIX(RenderbufferStorageMultisample):
 	.size	GL_PREFIX(RenderbufferStorageMultisample), .-GL_PREFIX(RenderbufferStorageMultisample)
 
 	.p2align	4,,15
+	.globl	GL_PREFIX(FlushMappedBufferRange)
+	.type	GL_PREFIX(FlushMappedBufferRange), @function
+GL_PREFIX(FlushMappedBufferRange):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4496(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4496(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4496(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4496(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(FlushMappedBufferRange), .-GL_PREFIX(FlushMappedBufferRange)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(MapBufferRange)
+	.type	GL_PREFIX(MapBufferRange), @function
+GL_PREFIX(MapBufferRange):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4504(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%rbp
+	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4504(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4504(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%rbp
+	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4504(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(MapBufferRange), .-GL_PREFIX(MapBufferRange)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(BindVertexArray)
+	.type	GL_PREFIX(BindVertexArray), @function
+GL_PREFIX(BindVertexArray):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4512(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	4512(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4512(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	4512(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(BindVertexArray), .-GL_PREFIX(BindVertexArray)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(GenVertexArrays)
+	.type	GL_PREFIX(GenVertexArrays), @function
+GL_PREFIX(GenVertexArrays):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4520(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4520(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4520(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
+	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
+	movq	4520(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(GenVertexArrays), .-GL_PREFIX(GenVertexArrays)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(CopyBufferSubData)
+	.type	GL_PREFIX(CopyBufferSubData), @function
+GL_PREFIX(CopyBufferSubData):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4528(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	call	_x86_64_get_dispatch@PLT
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4528(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4528(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	call	_glapi_get_dispatch
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
+	movq	4528(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(CopyBufferSubData), .-GL_PREFIX(CopyBufferSubData)
+
+	.p2align	4,,15
 	.globl	GL_PREFIX(PolygonOffsetEXT)
 	.type	GL_PREFIX(PolygonOffsetEXT), @function
 GL_PREFIX(PolygonOffsetEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4496(%rax), %r11
+	movq	4536(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21172,13 +21365,13 @@ GL_PREFIX(PolygonOffsetEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4496(%rax), %r11
+	movq	4536(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4496(%rax), %r11
+	movq	4536(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21188,19 +21381,19 @@ GL_PREFIX(PolygonOffsetEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4496(%rax), %r11
+	movq	4536(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PolygonOffsetEXT), .-GL_PREFIX(PolygonOffsetEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_563)
-	.type	GL_PREFIX(_dispatch_stub_563), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_563))
-GL_PREFIX(_dispatch_stub_563):
+	.globl	GL_PREFIX(_dispatch_stub_568)
+	.type	GL_PREFIX(_dispatch_stub_568), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_568))
+GL_PREFIX(_dispatch_stub_568):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4504(%rax), %r11
+	movq	4544(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21210,13 +21403,13 @@ GL_PREFIX(_dispatch_stub_563):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4504(%rax), %r11
+	movq	4544(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4504(%rax), %r11
+	movq	4544(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21226,19 +21419,19 @@ GL_PREFIX(_dispatch_stub_563):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4504(%rax), %r11
+	movq	4544(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_563), .-GL_PREFIX(_dispatch_stub_563)
+	.size	GL_PREFIX(_dispatch_stub_568), .-GL_PREFIX(_dispatch_stub_568)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_564)
-	.type	GL_PREFIX(_dispatch_stub_564), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_564))
-GL_PREFIX(_dispatch_stub_564):
+	.globl	GL_PREFIX(_dispatch_stub_569)
+	.type	GL_PREFIX(_dispatch_stub_569), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_569))
+GL_PREFIX(_dispatch_stub_569):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4512(%rax), %r11
+	movq	4552(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21248,13 +21441,13 @@ GL_PREFIX(_dispatch_stub_564):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4512(%rax), %r11
+	movq	4552(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4512(%rax), %r11
+	movq	4552(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21264,19 +21457,19 @@ GL_PREFIX(_dispatch_stub_564):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4512(%rax), %r11
+	movq	4552(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_564), .-GL_PREFIX(_dispatch_stub_564)
+	.size	GL_PREFIX(_dispatch_stub_569), .-GL_PREFIX(_dispatch_stub_569)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_565)
-	.type	GL_PREFIX(_dispatch_stub_565), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_565))
-GL_PREFIX(_dispatch_stub_565):
+	.globl	GL_PREFIX(_dispatch_stub_570)
+	.type	GL_PREFIX(_dispatch_stub_570), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_570))
+GL_PREFIX(_dispatch_stub_570):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4520(%rax), %r11
+	movq	4560(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21286,13 +21479,13 @@ GL_PREFIX(_dispatch_stub_565):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4520(%rax), %r11
+	movq	4560(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4520(%rax), %r11
+	movq	4560(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21302,19 +21495,19 @@ GL_PREFIX(_dispatch_stub_565):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4520(%rax), %r11
+	movq	4560(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_565), .-GL_PREFIX(_dispatch_stub_565)
+	.size	GL_PREFIX(_dispatch_stub_570), .-GL_PREFIX(_dispatch_stub_570)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_566)
-	.type	GL_PREFIX(_dispatch_stub_566), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_566))
-GL_PREFIX(_dispatch_stub_566):
+	.globl	GL_PREFIX(_dispatch_stub_571)
+	.type	GL_PREFIX(_dispatch_stub_571), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_571))
+GL_PREFIX(_dispatch_stub_571):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4528(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21324,13 +21517,13 @@ GL_PREFIX(_dispatch_stub_566):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4528(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4528(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21340,19 +21533,19 @@ GL_PREFIX(_dispatch_stub_566):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4528(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_566), .-GL_PREFIX(_dispatch_stub_566)
+	.size	GL_PREFIX(_dispatch_stub_571), .-GL_PREFIX(_dispatch_stub_571)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_567)
-	.type	GL_PREFIX(_dispatch_stub_567), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_567))
-GL_PREFIX(_dispatch_stub_567):
+	.globl	GL_PREFIX(_dispatch_stub_572)
+	.type	GL_PREFIX(_dispatch_stub_572), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_572))
+GL_PREFIX(_dispatch_stub_572):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4536(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21362,13 +21555,13 @@ GL_PREFIX(_dispatch_stub_567):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4536(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4536(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21378,19 +21571,19 @@ GL_PREFIX(_dispatch_stub_567):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4536(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_567), .-GL_PREFIX(_dispatch_stub_567)
+	.size	GL_PREFIX(_dispatch_stub_572), .-GL_PREFIX(_dispatch_stub_572)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_568)
-	.type	GL_PREFIX(_dispatch_stub_568), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_568))
-GL_PREFIX(_dispatch_stub_568):
+	.globl	GL_PREFIX(_dispatch_stub_573)
+	.type	GL_PREFIX(_dispatch_stub_573), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_573))
+GL_PREFIX(_dispatch_stub_573):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4544(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21400,13 +21593,13 @@ GL_PREFIX(_dispatch_stub_568):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4544(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4544(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21416,19 +21609,19 @@ GL_PREFIX(_dispatch_stub_568):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4544(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_568), .-GL_PREFIX(_dispatch_stub_568)
+	.size	GL_PREFIX(_dispatch_stub_573), .-GL_PREFIX(_dispatch_stub_573)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_569)
-	.type	GL_PREFIX(_dispatch_stub_569), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_569))
-GL_PREFIX(_dispatch_stub_569):
+	.globl	GL_PREFIX(_dispatch_stub_574)
+	.type	GL_PREFIX(_dispatch_stub_574), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_574))
+GL_PREFIX(_dispatch_stub_574):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4552(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21438,13 +21631,13 @@ GL_PREFIX(_dispatch_stub_569):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4552(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4552(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21454,40 +21647,40 @@ GL_PREFIX(_dispatch_stub_569):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4552(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_569), .-GL_PREFIX(_dispatch_stub_569)
+	.size	GL_PREFIX(_dispatch_stub_574), .-GL_PREFIX(_dispatch_stub_574)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_570)
-	.type	GL_PREFIX(_dispatch_stub_570), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_570))
-GL_PREFIX(_dispatch_stub_570):
+	.globl	GL_PREFIX(_dispatch_stub_575)
+	.type	GL_PREFIX(_dispatch_stub_575), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_575))
+GL_PREFIX(_dispatch_stub_575):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4560(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4560(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4560(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4560(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_570), .-GL_PREFIX(_dispatch_stub_570)
+	.size	GL_PREFIX(_dispatch_stub_575), .-GL_PREFIX(_dispatch_stub_575)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(ColorPointerEXT)
@@ -21495,7 +21688,7 @@ GL_PREFIX(_dispatch_stub_570):
 GL_PREFIX(ColorPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4568(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21509,13 +21702,13 @@ GL_PREFIX(ColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4568(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4568(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21529,7 +21722,7 @@ GL_PREFIX(ColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4568(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorPointerEXT), .-GL_PREFIX(ColorPointerEXT)
@@ -21540,7 +21733,7 @@ GL_PREFIX(ColorPointerEXT):
 GL_PREFIX(EdgeFlagPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4576(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21550,13 +21743,13 @@ GL_PREFIX(EdgeFlagPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4576(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4576(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21566,7 +21759,7 @@ GL_PREFIX(EdgeFlagPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4576(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(EdgeFlagPointerEXT), .-GL_PREFIX(EdgeFlagPointerEXT)
@@ -21577,7 +21770,7 @@ GL_PREFIX(EdgeFlagPointerEXT):
 GL_PREFIX(IndexPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4584(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21591,13 +21784,13 @@ GL_PREFIX(IndexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4584(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4584(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21611,7 +21804,7 @@ GL_PREFIX(IndexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4584(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IndexPointerEXT), .-GL_PREFIX(IndexPointerEXT)
@@ -21622,7 +21815,7 @@ GL_PREFIX(IndexPointerEXT):
 GL_PREFIX(NormalPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4592(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21636,13 +21829,13 @@ GL_PREFIX(NormalPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4592(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4592(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21656,7 +21849,7 @@ GL_PREFIX(NormalPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4592(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(NormalPointerEXT), .-GL_PREFIX(NormalPointerEXT)
@@ -21667,7 +21860,7 @@ GL_PREFIX(NormalPointerEXT):
 GL_PREFIX(TexCoordPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4600(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21681,13 +21874,13 @@ GL_PREFIX(TexCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4600(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4600(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21701,7 +21894,7 @@ GL_PREFIX(TexCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4600(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexCoordPointerEXT), .-GL_PREFIX(TexCoordPointerEXT)
@@ -21712,7 +21905,7 @@ GL_PREFIX(TexCoordPointerEXT):
 GL_PREFIX(VertexPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4608(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21726,13 +21919,13 @@ GL_PREFIX(VertexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4608(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4608(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21746,7 +21939,7 @@ GL_PREFIX(VertexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4608(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexPointerEXT), .-GL_PREFIX(VertexPointerEXT)
@@ -21757,7 +21950,7 @@ GL_PREFIX(VertexPointerEXT):
 GL_PREFIX(PointParameterfEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4616(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21767,13 +21960,13 @@ GL_PREFIX(PointParameterfEXT):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4616(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4616(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21783,7 +21976,7 @@ GL_PREFIX(PointParameterfEXT):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4616(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterfEXT), .-GL_PREFIX(PointParameterfEXT)
@@ -21794,7 +21987,7 @@ GL_PREFIX(PointParameterfEXT):
 GL_PREFIX(PointParameterfvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4624(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21804,13 +21997,13 @@ GL_PREFIX(PointParameterfvEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4624(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4624(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21820,7 +22013,7 @@ GL_PREFIX(PointParameterfvEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4624(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterfvEXT), .-GL_PREFIX(PointParameterfvEXT)
@@ -21831,7 +22024,7 @@ GL_PREFIX(PointParameterfvEXT):
 GL_PREFIX(LockArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4632(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21841,13 +22034,13 @@ GL_PREFIX(LockArraysEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4632(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4632(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21857,7 +22050,7 @@ GL_PREFIX(LockArraysEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4632(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(LockArraysEXT), .-GL_PREFIX(LockArraysEXT)
@@ -21868,37 +22061,37 @@ GL_PREFIX(LockArraysEXT):
 GL_PREFIX(UnlockArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4640(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4640(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4640(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4640(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(UnlockArraysEXT), .-GL_PREFIX(UnlockArraysEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_581)
-	.type	GL_PREFIX(_dispatch_stub_581), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_581))
-GL_PREFIX(_dispatch_stub_581):
+	.globl	GL_PREFIX(_dispatch_stub_586)
+	.type	GL_PREFIX(_dispatch_stub_586), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_586))
+GL_PREFIX(_dispatch_stub_586):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4648(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21908,13 +22101,13 @@ GL_PREFIX(_dispatch_stub_581):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4648(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4648(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21924,19 +22117,19 @@ GL_PREFIX(_dispatch_stub_581):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4648(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_581), .-GL_PREFIX(_dispatch_stub_581)
+	.size	GL_PREFIX(_dispatch_stub_586), .-GL_PREFIX(_dispatch_stub_586)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_582)
-	.type	GL_PREFIX(_dispatch_stub_582), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_582))
-GL_PREFIX(_dispatch_stub_582):
+	.globl	GL_PREFIX(_dispatch_stub_587)
+	.type	GL_PREFIX(_dispatch_stub_587), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_587))
+GL_PREFIX(_dispatch_stub_587):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4656(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21946,13 +22139,13 @@ GL_PREFIX(_dispatch_stub_582):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4656(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4656(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21962,10 +22155,10 @@ GL_PREFIX(_dispatch_stub_582):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4656(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_582), .-GL_PREFIX(_dispatch_stub_582)
+	.size	GL_PREFIX(_dispatch_stub_587), .-GL_PREFIX(_dispatch_stub_587)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(SecondaryColor3bEXT)
@@ -21973,7 +22166,7 @@ GL_PREFIX(_dispatch_stub_582):
 GL_PREFIX(SecondaryColor3bEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4664(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21983,13 +22176,13 @@ GL_PREFIX(SecondaryColor3bEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4664(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4664(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21999,7 +22192,7 @@ GL_PREFIX(SecondaryColor3bEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4664(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3bEXT), .-GL_PREFIX(SecondaryColor3bEXT)
@@ -22010,25 +22203,25 @@ GL_PREFIX(SecondaryColor3bEXT):
 GL_PREFIX(SecondaryColor3bvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4672(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4672(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4672(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4672(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3bvEXT), .-GL_PREFIX(SecondaryColor3bvEXT)
@@ -22039,7 +22232,7 @@ GL_PREFIX(SecondaryColor3bvEXT):
 GL_PREFIX(SecondaryColor3dEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4680(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -22051,13 +22244,13 @@ GL_PREFIX(SecondaryColor3dEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4680(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4680(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -22069,7 +22262,7 @@ GL_PREFIX(SecondaryColor3dEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4680(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3dEXT), .-GL_PREFIX(SecondaryColor3dEXT)
@@ -22080,25 +22273,25 @@ GL_PREFIX(SecondaryColor3dEXT):
 GL_PREFIX(SecondaryColor3dvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4688(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4688(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4688(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4688(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3dvEXT), .-GL_PREFIX(SecondaryColor3dvEXT)
@@ -22109,7 +22302,7 @@ GL_PREFIX(SecondaryColor3dvEXT):
 GL_PREFIX(SecondaryColor3fEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4696(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -22121,13 +22314,13 @@ GL_PREFIX(SecondaryColor3fEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4696(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4696(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -22139,7 +22332,7 @@ GL_PREFIX(SecondaryColor3fEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4696(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3fEXT), .-GL_PREFIX(SecondaryColor3fEXT)
@@ -22150,25 +22343,25 @@ GL_PREFIX(SecondaryColor3fEXT):
 GL_PREFIX(SecondaryColor3fvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4704(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4704(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4704(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4704(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3fvEXT), .-GL_PREFIX(SecondaryColor3fvEXT)
@@ -22179,7 +22372,7 @@ GL_PREFIX(SecondaryColor3fvEXT):
 GL_PREFIX(SecondaryColor3iEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4712(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22189,13 +22382,13 @@ GL_PREFIX(SecondaryColor3iEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4712(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4712(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22205,7 +22398,7 @@ GL_PREFIX(SecondaryColor3iEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4712(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3iEXT), .-GL_PREFIX(SecondaryColor3iEXT)
@@ -22216,25 +22409,25 @@ GL_PREFIX(SecondaryColor3iEXT):
 GL_PREFIX(SecondaryColor3ivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4720(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4720(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4720(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4720(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ivEXT), .-GL_PREFIX(SecondaryColor3ivEXT)
@@ -22245,7 +22438,7 @@ GL_PREFIX(SecondaryColor3ivEXT):
 GL_PREFIX(SecondaryColor3sEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4728(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22255,13 +22448,13 @@ GL_PREFIX(SecondaryColor3sEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4728(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4728(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22271,7 +22464,7 @@ GL_PREFIX(SecondaryColor3sEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4728(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3sEXT), .-GL_PREFIX(SecondaryColor3sEXT)
@@ -22282,25 +22475,25 @@ GL_PREFIX(SecondaryColor3sEXT):
 GL_PREFIX(SecondaryColor3svEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4736(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4736(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4736(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4736(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3svEXT), .-GL_PREFIX(SecondaryColor3svEXT)
@@ -22311,7 +22504,7 @@ GL_PREFIX(SecondaryColor3svEXT):
 GL_PREFIX(SecondaryColor3ubEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4744(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22321,13 +22514,13 @@ GL_PREFIX(SecondaryColor3ubEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4744(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4744(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22337,7 +22530,7 @@ GL_PREFIX(SecondaryColor3ubEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4744(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ubEXT), .-GL_PREFIX(SecondaryColor3ubEXT)
@@ -22348,25 +22541,25 @@ GL_PREFIX(SecondaryColor3ubEXT):
 GL_PREFIX(SecondaryColor3ubvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4752(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4752(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4752(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4752(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ubvEXT), .-GL_PREFIX(SecondaryColor3ubvEXT)
@@ -22377,7 +22570,7 @@ GL_PREFIX(SecondaryColor3ubvEXT):
 GL_PREFIX(SecondaryColor3uiEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4760(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22387,13 +22580,13 @@ GL_PREFIX(SecondaryColor3uiEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4760(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4760(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22403,7 +22596,7 @@ GL_PREFIX(SecondaryColor3uiEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4760(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3uiEXT), .-GL_PREFIX(SecondaryColor3uiEXT)
@@ -22414,25 +22607,25 @@ GL_PREFIX(SecondaryColor3uiEXT):
 GL_PREFIX(SecondaryColor3uivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4768(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4768(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4768(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4768(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3uivEXT), .-GL_PREFIX(SecondaryColor3uivEXT)
@@ -22443,7 +22636,7 @@ GL_PREFIX(SecondaryColor3uivEXT):
 GL_PREFIX(SecondaryColor3usEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4776(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22453,13 +22646,13 @@ GL_PREFIX(SecondaryColor3usEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4776(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4776(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22469,7 +22662,7 @@ GL_PREFIX(SecondaryColor3usEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4776(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3usEXT), .-GL_PREFIX(SecondaryColor3usEXT)
@@ -22480,25 +22673,25 @@ GL_PREFIX(SecondaryColor3usEXT):
 GL_PREFIX(SecondaryColor3usvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4784(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4784(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4784(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4784(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3usvEXT), .-GL_PREFIX(SecondaryColor3usvEXT)
@@ -22509,7 +22702,7 @@ GL_PREFIX(SecondaryColor3usvEXT):
 GL_PREFIX(SecondaryColorPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4792(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22523,13 +22716,13 @@ GL_PREFIX(SecondaryColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4792(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4792(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22543,7 +22736,7 @@ GL_PREFIX(SecondaryColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4792(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColorPointerEXT), .-GL_PREFIX(SecondaryColorPointerEXT)
@@ -22554,7 +22747,7 @@ GL_PREFIX(SecondaryColorPointerEXT):
 GL_PREFIX(MultiDrawArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4800(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22568,13 +22761,13 @@ GL_PREFIX(MultiDrawArraysEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4800(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4800(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22588,7 +22781,7 @@ GL_PREFIX(MultiDrawArraysEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4800(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(MultiDrawArraysEXT), .-GL_PREFIX(MultiDrawArraysEXT)
@@ -22599,7 +22792,7 @@ GL_PREFIX(MultiDrawArraysEXT):
 GL_PREFIX(MultiDrawElementsEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4808(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22613,13 +22806,13 @@ GL_PREFIX(MultiDrawElementsEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4808(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4808(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22633,7 +22826,7 @@ GL_PREFIX(MultiDrawElementsEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4808(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(MultiDrawElementsEXT), .-GL_PREFIX(MultiDrawElementsEXT)
@@ -22644,7 +22837,7 @@ GL_PREFIX(MultiDrawElementsEXT):
 GL_PREFIX(FogCoordPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4816(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22654,13 +22847,13 @@ GL_PREFIX(FogCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4816(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4816(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22670,7 +22863,7 @@ GL_PREFIX(FogCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4816(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordPointerEXT), .-GL_PREFIX(FogCoordPointerEXT)
@@ -22681,7 +22874,7 @@ GL_PREFIX(FogCoordPointerEXT):
 GL_PREFIX(FogCoorddEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4824(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$8, %rsp
@@ -22689,13 +22882,13 @@ GL_PREFIX(FogCoorddEXT):
 	call	_x86_64_get_dispatch@PLT
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4824(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4824(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$8, %rsp
@@ -22703,7 +22896,7 @@ GL_PREFIX(FogCoorddEXT):
 	call	_glapi_get_dispatch
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4824(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoorddEXT), .-GL_PREFIX(FogCoorddEXT)
@@ -22714,25 +22907,25 @@ GL_PREFIX(FogCoorddEXT):
 GL_PREFIX(FogCoorddvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4832(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4832(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4832(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4832(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoorddvEXT), .-GL_PREFIX(FogCoorddvEXT)
@@ -22743,7 +22936,7 @@ GL_PREFIX(FogCoorddvEXT):
 GL_PREFIX(FogCoordfEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4840(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$8, %rsp
@@ -22751,13 +22944,13 @@ GL_PREFIX(FogCoordfEXT):
 	call	_x86_64_get_dispatch@PLT
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4840(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4840(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$8, %rsp
@@ -22765,7 +22958,7 @@ GL_PREFIX(FogCoordfEXT):
 	call	_glapi_get_dispatch
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4840(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordfEXT), .-GL_PREFIX(FogCoordfEXT)
@@ -22776,58 +22969,58 @@ GL_PREFIX(FogCoordfEXT):
 GL_PREFIX(FogCoordfvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4848(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4848(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4848(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4848(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordfvEXT), .-GL_PREFIX(FogCoordfvEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_607)
-	.type	GL_PREFIX(_dispatch_stub_607), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_607))
-GL_PREFIX(_dispatch_stub_607):
+	.globl	GL_PREFIX(_dispatch_stub_612)
+	.type	GL_PREFIX(_dispatch_stub_612), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_612))
+GL_PREFIX(_dispatch_stub_612):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4856(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4856(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4856(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4856(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_607), .-GL_PREFIX(_dispatch_stub_607)
+	.size	GL_PREFIX(_dispatch_stub_612), .-GL_PREFIX(_dispatch_stub_612)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(BlendFuncSeparateEXT)
@@ -22835,7 +23028,7 @@ GL_PREFIX(_dispatch_stub_607):
 GL_PREFIX(BlendFuncSeparateEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4864(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22849,13 +23042,13 @@ GL_PREFIX(BlendFuncSeparateEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4864(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4864(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22869,7 +23062,7 @@ GL_PREFIX(BlendFuncSeparateEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4864(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BlendFuncSeparateEXT), .-GL_PREFIX(BlendFuncSeparateEXT)
@@ -22880,25 +23073,25 @@ GL_PREFIX(BlendFuncSeparateEXT):
 GL_PREFIX(FlushVertexArrayRangeNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4872(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4872(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4872(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4872(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FlushVertexArrayRangeNV), .-GL_PREFIX(FlushVertexArrayRangeNV)
@@ -22909,7 +23102,7 @@ GL_PREFIX(FlushVertexArrayRangeNV):
 GL_PREFIX(VertexArrayRangeNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4880(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22919,13 +23112,13 @@ GL_PREFIX(VertexArrayRangeNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4880(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4880(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22935,7 +23128,7 @@ GL_PREFIX(VertexArrayRangeNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4880(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexArrayRangeNV), .-GL_PREFIX(VertexArrayRangeNV)
@@ -22946,7 +23139,7 @@ GL_PREFIX(VertexArrayRangeNV):
 GL_PREFIX(CombinerInputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4888(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22964,13 +23157,13 @@ GL_PREFIX(CombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4888(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4888(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22988,7 +23181,7 @@ GL_PREFIX(CombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4888(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerInputNV), .-GL_PREFIX(CombinerInputNV)
@@ -22999,7 +23192,7 @@ GL_PREFIX(CombinerInputNV):
 GL_PREFIX(CombinerOutputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4896(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23017,13 +23210,13 @@ GL_PREFIX(CombinerOutputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4896(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4896(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23041,7 +23234,7 @@ GL_PREFIX(CombinerOutputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4896(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerOutputNV), .-GL_PREFIX(CombinerOutputNV)
@@ -23052,7 +23245,7 @@ GL_PREFIX(CombinerOutputNV):
 GL_PREFIX(CombinerParameterfNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4904(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23062,13 +23255,13 @@ GL_PREFIX(CombinerParameterfNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4904(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4904(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23078,7 +23271,7 @@ GL_PREFIX(CombinerParameterfNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4904(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterfNV), .-GL_PREFIX(CombinerParameterfNV)
@@ -23089,7 +23282,7 @@ GL_PREFIX(CombinerParameterfNV):
 GL_PREFIX(CombinerParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4912(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23099,13 +23292,13 @@ GL_PREFIX(CombinerParameterfvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4912(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4912(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23115,7 +23308,7 @@ GL_PREFIX(CombinerParameterfvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4912(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterfvNV), .-GL_PREFIX(CombinerParameterfvNV)
@@ -23126,7 +23319,7 @@ GL_PREFIX(CombinerParameterfvNV):
 GL_PREFIX(CombinerParameteriNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4920(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23136,13 +23329,13 @@ GL_PREFIX(CombinerParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4920(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4920(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23152,7 +23345,7 @@ GL_PREFIX(CombinerParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4920(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameteriNV), .-GL_PREFIX(CombinerParameteriNV)
@@ -23163,7 +23356,7 @@ GL_PREFIX(CombinerParameteriNV):
 GL_PREFIX(CombinerParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4928(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23173,13 +23366,13 @@ GL_PREFIX(CombinerParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4928(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4928(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23189,7 +23382,7 @@ GL_PREFIX(CombinerParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4928(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterivNV), .-GL_PREFIX(CombinerParameterivNV)
@@ -23200,7 +23393,7 @@ GL_PREFIX(CombinerParameterivNV):
 GL_PREFIX(FinalCombinerInputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4936(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23214,13 +23407,13 @@ GL_PREFIX(FinalCombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4936(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4936(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23234,7 +23427,7 @@ GL_PREFIX(FinalCombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4936(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FinalCombinerInputNV), .-GL_PREFIX(FinalCombinerInputNV)
@@ -23245,7 +23438,7 @@ GL_PREFIX(FinalCombinerInputNV):
 GL_PREFIX(GetCombinerInputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4944(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23259,13 +23452,13 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4944(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4944(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23279,7 +23472,7 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4944(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerInputParameterfvNV), .-GL_PREFIX(GetCombinerInputParameterfvNV)
@@ -23290,7 +23483,7 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 GL_PREFIX(GetCombinerInputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4952(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23304,13 +23497,13 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4952(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4952(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23324,7 +23517,7 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4952(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerInputParameterivNV), .-GL_PREFIX(GetCombinerInputParameterivNV)
@@ -23335,7 +23528,7 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 GL_PREFIX(GetCombinerOutputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4960(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23349,13 +23542,13 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4960(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4960(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23369,7 +23562,7 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4960(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerOutputParameterfvNV), .-GL_PREFIX(GetCombinerOutputParameterfvNV)
@@ -23380,7 +23573,7 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 GL_PREFIX(GetCombinerOutputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4968(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23394,13 +23587,13 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4968(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4968(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23414,7 +23607,7 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4968(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerOutputParameterivNV), .-GL_PREFIX(GetCombinerOutputParameterivNV)
@@ -23425,7 +23618,7 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4976(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23435,13 +23628,13 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4976(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4976(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23451,7 +23644,7 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4976(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFinalCombinerInputParameterfvNV), .-GL_PREFIX(GetFinalCombinerInputParameterfvNV)
@@ -23462,7 +23655,7 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 GL_PREFIX(GetFinalCombinerInputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4984(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23472,13 +23665,13 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4984(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4984(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23488,7 +23681,7 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4984(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFinalCombinerInputParameterivNV), .-GL_PREFIX(GetFinalCombinerInputParameterivNV)
@@ -23499,25 +23692,25 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 GL_PREFIX(ResizeBuffersMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4992(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4992(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4992(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4992(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ResizeBuffersMESA), .-GL_PREFIX(ResizeBuffersMESA)
@@ -23528,7 +23721,7 @@ GL_PREFIX(ResizeBuffersMESA):
 GL_PREFIX(WindowPos2dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5000(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23538,13 +23731,13 @@ GL_PREFIX(WindowPos2dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5000(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5000(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23554,7 +23747,7 @@ GL_PREFIX(WindowPos2dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5000(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2dMESA), .-GL_PREFIX(WindowPos2dMESA)
@@ -23565,25 +23758,25 @@ GL_PREFIX(WindowPos2dMESA):
 GL_PREFIX(WindowPos2dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5008(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5008(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5008(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5008(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2dvMESA), .-GL_PREFIX(WindowPos2dvMESA)
@@ -23594,7 +23787,7 @@ GL_PREFIX(WindowPos2dvMESA):
 GL_PREFIX(WindowPos2fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5016(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23604,13 +23797,13 @@ GL_PREFIX(WindowPos2fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5016(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5016(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23620,7 +23813,7 @@ GL_PREFIX(WindowPos2fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5016(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2fMESA), .-GL_PREFIX(WindowPos2fMESA)
@@ -23631,25 +23824,25 @@ GL_PREFIX(WindowPos2fMESA):
 GL_PREFIX(WindowPos2fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5024(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5024(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5024(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5024(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2fvMESA), .-GL_PREFIX(WindowPos2fvMESA)
@@ -23660,7 +23853,7 @@ GL_PREFIX(WindowPos2fvMESA):
 GL_PREFIX(WindowPos2iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5032(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23670,13 +23863,13 @@ GL_PREFIX(WindowPos2iMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5032(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5032(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23686,7 +23879,7 @@ GL_PREFIX(WindowPos2iMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5032(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2iMESA), .-GL_PREFIX(WindowPos2iMESA)
@@ -23697,25 +23890,25 @@ GL_PREFIX(WindowPos2iMESA):
 GL_PREFIX(WindowPos2ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5040(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5040(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5040(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5040(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2ivMESA), .-GL_PREFIX(WindowPos2ivMESA)
@@ -23726,7 +23919,7 @@ GL_PREFIX(WindowPos2ivMESA):
 GL_PREFIX(WindowPos2sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5048(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23736,13 +23929,13 @@ GL_PREFIX(WindowPos2sMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5048(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5048(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23752,7 +23945,7 @@ GL_PREFIX(WindowPos2sMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5048(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2sMESA), .-GL_PREFIX(WindowPos2sMESA)
@@ -23763,25 +23956,25 @@ GL_PREFIX(WindowPos2sMESA):
 GL_PREFIX(WindowPos2svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5056(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5056(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5056(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5056(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2svMESA), .-GL_PREFIX(WindowPos2svMESA)
@@ -23792,7 +23985,7 @@ GL_PREFIX(WindowPos2svMESA):
 GL_PREFIX(WindowPos3dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5064(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23804,13 +23997,13 @@ GL_PREFIX(WindowPos3dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5064(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5064(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23822,7 +24015,7 @@ GL_PREFIX(WindowPos3dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5064(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3dMESA), .-GL_PREFIX(WindowPos3dMESA)
@@ -23833,25 +24026,25 @@ GL_PREFIX(WindowPos3dMESA):
 GL_PREFIX(WindowPos3dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5072(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5072(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5072(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5072(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3dvMESA), .-GL_PREFIX(WindowPos3dvMESA)
@@ -23862,7 +24055,7 @@ GL_PREFIX(WindowPos3dvMESA):
 GL_PREFIX(WindowPos3fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5080(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23874,13 +24067,13 @@ GL_PREFIX(WindowPos3fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5080(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5080(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23892,7 +24085,7 @@ GL_PREFIX(WindowPos3fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5080(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3fMESA), .-GL_PREFIX(WindowPos3fMESA)
@@ -23903,25 +24096,25 @@ GL_PREFIX(WindowPos3fMESA):
 GL_PREFIX(WindowPos3fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5088(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5088(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5088(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5088(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3fvMESA), .-GL_PREFIX(WindowPos3fvMESA)
@@ -23932,7 +24125,7 @@ GL_PREFIX(WindowPos3fvMESA):
 GL_PREFIX(WindowPos3iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5096(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23942,13 +24135,13 @@ GL_PREFIX(WindowPos3iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5096(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5096(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23958,7 +24151,7 @@ GL_PREFIX(WindowPos3iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5096(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3iMESA), .-GL_PREFIX(WindowPos3iMESA)
@@ -23969,25 +24162,25 @@ GL_PREFIX(WindowPos3iMESA):
 GL_PREFIX(WindowPos3ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5104(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5104(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5104(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5104(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3ivMESA), .-GL_PREFIX(WindowPos3ivMESA)
@@ -23998,7 +24191,7 @@ GL_PREFIX(WindowPos3ivMESA):
 GL_PREFIX(WindowPos3sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5112(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24008,13 +24201,13 @@ GL_PREFIX(WindowPos3sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5112(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5112(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24024,7 +24217,7 @@ GL_PREFIX(WindowPos3sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5112(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3sMESA), .-GL_PREFIX(WindowPos3sMESA)
@@ -24035,25 +24228,25 @@ GL_PREFIX(WindowPos3sMESA):
 GL_PREFIX(WindowPos3svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5120(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5120(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5120(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5120(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3svMESA), .-GL_PREFIX(WindowPos3svMESA)
@@ -24064,7 +24257,7 @@ GL_PREFIX(WindowPos3svMESA):
 GL_PREFIX(WindowPos4dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5128(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -24078,13 +24271,13 @@ GL_PREFIX(WindowPos4dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5128(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5128(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -24098,7 +24291,7 @@ GL_PREFIX(WindowPos4dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5128(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4dMESA), .-GL_PREFIX(WindowPos4dMESA)
@@ -24109,25 +24302,25 @@ GL_PREFIX(WindowPos4dMESA):
 GL_PREFIX(WindowPos4dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5136(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5136(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5136(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5136(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4dvMESA), .-GL_PREFIX(WindowPos4dvMESA)
@@ -24138,7 +24331,7 @@ GL_PREFIX(WindowPos4dvMESA):
 GL_PREFIX(WindowPos4fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5144(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -24152,13 +24345,13 @@ GL_PREFIX(WindowPos4fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5144(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5144(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -24172,7 +24365,7 @@ GL_PREFIX(WindowPos4fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5144(%rax), %r11
+	movq	5184(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4fMESA), .-GL_PREFIX(WindowPos4fMESA)
@@ -24183,25 +24376,25 @@ GL_PREFIX(WindowPos4fMESA):
 GL_PREFIX(WindowPos4fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5152(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5152(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5152(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5152(%rax), %r11
+	movq	5192(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4fvMESA), .-GL_PREFIX(WindowPos4fvMESA)
@@ -24212,7 +24405,7 @@ GL_PREFIX(WindowPos4fvMESA):
 GL_PREFIX(WindowPos4iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5160(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24226,13 +24419,13 @@ GL_PREFIX(WindowPos4iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5160(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5160(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24246,7 +24439,7 @@ GL_PREFIX(WindowPos4iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5160(%rax), %r11
+	movq	5200(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4iMESA), .-GL_PREFIX(WindowPos4iMESA)
@@ -24257,25 +24450,25 @@ GL_PREFIX(WindowPos4iMESA):
 GL_PREFIX(WindowPos4ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5168(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5168(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5168(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5168(%rax), %r11
+	movq	5208(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4ivMESA), .-GL_PREFIX(WindowPos4ivMESA)
@@ -24286,7 +24479,7 @@ GL_PREFIX(WindowPos4ivMESA):
 GL_PREFIX(WindowPos4sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5176(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24300,13 +24493,13 @@ GL_PREFIX(WindowPos4sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5176(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5176(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24320,7 +24513,7 @@ GL_PREFIX(WindowPos4sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5176(%rax), %r11
+	movq	5216(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4sMESA), .-GL_PREFIX(WindowPos4sMESA)
@@ -24331,37 +24524,37 @@ GL_PREFIX(WindowPos4sMESA):
 GL_PREFIX(WindowPos4svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5184(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5184(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5184(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5184(%rax), %r11
+	movq	5224(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4svMESA), .-GL_PREFIX(WindowPos4svMESA)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_649)
-	.type	GL_PREFIX(_dispatch_stub_649), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_649))
-GL_PREFIX(_dispatch_stub_649):
+	.globl	GL_PREFIX(_dispatch_stub_654)
+	.type	GL_PREFIX(_dispatch_stub_654), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_654))
+GL_PREFIX(_dispatch_stub_654):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5192(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24375,13 +24568,13 @@ GL_PREFIX(_dispatch_stub_649):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5192(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5192(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24395,19 +24588,19 @@ GL_PREFIX(_dispatch_stub_649):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5192(%rax), %r11
+	movq	5232(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_649), .-GL_PREFIX(_dispatch_stub_649)
+	.size	GL_PREFIX(_dispatch_stub_654), .-GL_PREFIX(_dispatch_stub_654)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_650)
-	.type	GL_PREFIX(_dispatch_stub_650), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_650))
-GL_PREFIX(_dispatch_stub_650):
+	.globl	GL_PREFIX(_dispatch_stub_655)
+	.type	GL_PREFIX(_dispatch_stub_655), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_655))
+GL_PREFIX(_dispatch_stub_655):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5200(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24425,13 +24618,13 @@ GL_PREFIX(_dispatch_stub_650):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5200(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5200(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24449,19 +24642,19 @@ GL_PREFIX(_dispatch_stub_650):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5200(%rax), %r11
+	movq	5240(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_650), .-GL_PREFIX(_dispatch_stub_650)
+	.size	GL_PREFIX(_dispatch_stub_655), .-GL_PREFIX(_dispatch_stub_655)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_651)
-	.type	GL_PREFIX(_dispatch_stub_651), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_651))
-GL_PREFIX(_dispatch_stub_651):
+	.globl	GL_PREFIX(_dispatch_stub_656)
+	.type	GL_PREFIX(_dispatch_stub_656), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_656))
+GL_PREFIX(_dispatch_stub_656):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5208(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24471,13 +24664,13 @@ GL_PREFIX(_dispatch_stub_651):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5208(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5208(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24487,49 +24680,49 @@ GL_PREFIX(_dispatch_stub_651):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5208(%rax), %r11
+	movq	5248(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_651), .-GL_PREFIX(_dispatch_stub_651)
+	.size	GL_PREFIX(_dispatch_stub_656), .-GL_PREFIX(_dispatch_stub_656)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_652)
-	.type	GL_PREFIX(_dispatch_stub_652), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_652))
-GL_PREFIX(_dispatch_stub_652):
+	.globl	GL_PREFIX(_dispatch_stub_657)
+	.type	GL_PREFIX(_dispatch_stub_657), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_657))
+GL_PREFIX(_dispatch_stub_657):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5216(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5216(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5216(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5216(%rax), %r11
+	movq	5256(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_652), .-GL_PREFIX(_dispatch_stub_652)
+	.size	GL_PREFIX(_dispatch_stub_657), .-GL_PREFIX(_dispatch_stub_657)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_653)
-	.type	GL_PREFIX(_dispatch_stub_653), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_653))
-GL_PREFIX(_dispatch_stub_653):
+	.globl	GL_PREFIX(_dispatch_stub_658)
+	.type	GL_PREFIX(_dispatch_stub_658), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_658))
+GL_PREFIX(_dispatch_stub_658):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5224(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24539,13 +24732,13 @@ GL_PREFIX(_dispatch_stub_653):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5224(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5224(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24555,19 +24748,19 @@ GL_PREFIX(_dispatch_stub_653):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5224(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_653), .-GL_PREFIX(_dispatch_stub_653)
+	.size	GL_PREFIX(_dispatch_stub_658), .-GL_PREFIX(_dispatch_stub_658)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_654)
-	.type	GL_PREFIX(_dispatch_stub_654), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_654))
-GL_PREFIX(_dispatch_stub_654):
+	.globl	GL_PREFIX(_dispatch_stub_659)
+	.type	GL_PREFIX(_dispatch_stub_659), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_659))
+GL_PREFIX(_dispatch_stub_659):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5232(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24577,13 +24770,13 @@ GL_PREFIX(_dispatch_stub_654):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5232(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5232(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24593,49 +24786,49 @@ GL_PREFIX(_dispatch_stub_654):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5232(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_654), .-GL_PREFIX(_dispatch_stub_654)
+	.size	GL_PREFIX(_dispatch_stub_659), .-GL_PREFIX(_dispatch_stub_659)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_655)
-	.type	GL_PREFIX(_dispatch_stub_655), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_655))
-GL_PREFIX(_dispatch_stub_655):
+	.globl	GL_PREFIX(_dispatch_stub_660)
+	.type	GL_PREFIX(_dispatch_stub_660), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_660))
+GL_PREFIX(_dispatch_stub_660):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5240(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5240(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5240(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5240(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_655), .-GL_PREFIX(_dispatch_stub_655)
+	.size	GL_PREFIX(_dispatch_stub_660), .-GL_PREFIX(_dispatch_stub_660)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_656)
-	.type	GL_PREFIX(_dispatch_stub_656), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_656))
-GL_PREFIX(_dispatch_stub_656):
+	.globl	GL_PREFIX(_dispatch_stub_661)
+	.type	GL_PREFIX(_dispatch_stub_661), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_661))
+GL_PREFIX(_dispatch_stub_661):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5248(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24645,13 +24838,13 @@ GL_PREFIX(_dispatch_stub_656):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5248(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5248(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24661,40 +24854,40 @@ GL_PREFIX(_dispatch_stub_656):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5248(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_656), .-GL_PREFIX(_dispatch_stub_656)
+	.size	GL_PREFIX(_dispatch_stub_661), .-GL_PREFIX(_dispatch_stub_661)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_657)
-	.type	GL_PREFIX(_dispatch_stub_657), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_657))
-GL_PREFIX(_dispatch_stub_657):
+	.globl	GL_PREFIX(_dispatch_stub_662)
+	.type	GL_PREFIX(_dispatch_stub_662), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_662))
+GL_PREFIX(_dispatch_stub_662):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5256(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5256(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5256(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5256(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_657), .-GL_PREFIX(_dispatch_stub_657)
+	.size	GL_PREFIX(_dispatch_stub_662), .-GL_PREFIX(_dispatch_stub_662)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(AreProgramsResidentNV)
@@ -24702,7 +24895,7 @@ GL_PREFIX(_dispatch_stub_657):
 GL_PREFIX(AreProgramsResidentNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5264(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24712,13 +24905,13 @@ GL_PREFIX(AreProgramsResidentNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5264(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5264(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24728,7 +24921,7 @@ GL_PREFIX(AreProgramsResidentNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5264(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AreProgramsResidentNV), .-GL_PREFIX(AreProgramsResidentNV)
@@ -24739,7 +24932,7 @@ GL_PREFIX(AreProgramsResidentNV):
 GL_PREFIX(BindProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5272(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24749,13 +24942,13 @@ GL_PREFIX(BindProgramNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5272(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5272(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24765,7 +24958,7 @@ GL_PREFIX(BindProgramNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5272(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindProgramNV), .-GL_PREFIX(BindProgramNV)
@@ -24776,7 +24969,7 @@ GL_PREFIX(BindProgramNV):
 GL_PREFIX(DeleteProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5280(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24786,13 +24979,13 @@ GL_PREFIX(DeleteProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5280(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5280(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24802,7 +24995,7 @@ GL_PREFIX(DeleteProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5280(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteProgramsNV), .-GL_PREFIX(DeleteProgramsNV)
@@ -24813,7 +25006,7 @@ GL_PREFIX(DeleteProgramsNV):
 GL_PREFIX(ExecuteProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5288(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24823,13 +25016,13 @@ GL_PREFIX(ExecuteProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5288(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5288(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24839,7 +25032,7 @@ GL_PREFIX(ExecuteProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5288(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ExecuteProgramNV), .-GL_PREFIX(ExecuteProgramNV)
@@ -24850,7 +25043,7 @@ GL_PREFIX(ExecuteProgramNV):
 GL_PREFIX(GenProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5296(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24860,13 +25053,13 @@ GL_PREFIX(GenProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5296(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5296(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24876,7 +25069,7 @@ GL_PREFIX(GenProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5296(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenProgramsNV), .-GL_PREFIX(GenProgramsNV)
@@ -24887,7 +25080,7 @@ GL_PREFIX(GenProgramsNV):
 GL_PREFIX(GetProgramParameterdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5304(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24901,13 +25094,13 @@ GL_PREFIX(GetProgramParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5304(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5304(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24921,7 +25114,7 @@ GL_PREFIX(GetProgramParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5304(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramParameterdvNV), .-GL_PREFIX(GetProgramParameterdvNV)
@@ -24932,7 +25125,7 @@ GL_PREFIX(GetProgramParameterdvNV):
 GL_PREFIX(GetProgramParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5312(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24946,13 +25139,13 @@ GL_PREFIX(GetProgramParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5312(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5312(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24966,7 +25159,7 @@ GL_PREFIX(GetProgramParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5312(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramParameterfvNV), .-GL_PREFIX(GetProgramParameterfvNV)
@@ -24977,7 +25170,7 @@ GL_PREFIX(GetProgramParameterfvNV):
 GL_PREFIX(GetProgramStringNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5320(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24987,13 +25180,13 @@ GL_PREFIX(GetProgramStringNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5320(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5320(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25003,7 +25196,7 @@ GL_PREFIX(GetProgramStringNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5320(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramStringNV), .-GL_PREFIX(GetProgramStringNV)
@@ -25014,7 +25207,7 @@ GL_PREFIX(GetProgramStringNV):
 GL_PREFIX(GetProgramivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5328(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25024,13 +25217,13 @@ GL_PREFIX(GetProgramivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5328(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5328(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25040,7 +25233,7 @@ GL_PREFIX(GetProgramivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5328(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramivNV), .-GL_PREFIX(GetProgramivNV)
@@ -25051,7 +25244,7 @@ GL_PREFIX(GetProgramivNV):
 GL_PREFIX(GetTrackMatrixivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5336(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25065,13 +25258,13 @@ GL_PREFIX(GetTrackMatrixivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5336(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5336(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25085,7 +25278,7 @@ GL_PREFIX(GetTrackMatrixivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5336(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTrackMatrixivNV), .-GL_PREFIX(GetTrackMatrixivNV)
@@ -25096,7 +25289,7 @@ GL_PREFIX(GetTrackMatrixivNV):
 GL_PREFIX(GetVertexAttribPointervNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5344(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25106,13 +25299,13 @@ GL_PREFIX(GetVertexAttribPointervNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5344(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5344(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25122,7 +25315,7 @@ GL_PREFIX(GetVertexAttribPointervNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5344(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribPointervNV), .-GL_PREFIX(GetVertexAttribPointervNV)
@@ -25133,7 +25326,7 @@ GL_PREFIX(GetVertexAttribPointervNV):
 GL_PREFIX(GetVertexAttribdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5352(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25143,13 +25336,13 @@ GL_PREFIX(GetVertexAttribdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5352(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5352(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25159,7 +25352,7 @@ GL_PREFIX(GetVertexAttribdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5352(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribdvNV), .-GL_PREFIX(GetVertexAttribdvNV)
@@ -25170,7 +25363,7 @@ GL_PREFIX(GetVertexAttribdvNV):
 GL_PREFIX(GetVertexAttribfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5360(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25180,13 +25373,13 @@ GL_PREFIX(GetVertexAttribfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5360(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5360(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25196,7 +25389,7 @@ GL_PREFIX(GetVertexAttribfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5360(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribfvNV), .-GL_PREFIX(GetVertexAttribfvNV)
@@ -25207,7 +25400,7 @@ GL_PREFIX(GetVertexAttribfvNV):
 GL_PREFIX(GetVertexAttribivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5368(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25217,13 +25410,13 @@ GL_PREFIX(GetVertexAttribivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5368(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5368(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25233,7 +25426,7 @@ GL_PREFIX(GetVertexAttribivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5368(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribivNV), .-GL_PREFIX(GetVertexAttribivNV)
@@ -25244,25 +25437,25 @@ GL_PREFIX(GetVertexAttribivNV):
 GL_PREFIX(IsProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5376(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5376(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5376(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5376(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsProgramNV), .-GL_PREFIX(IsProgramNV)
@@ -25273,7 +25466,7 @@ GL_PREFIX(IsProgramNV):
 GL_PREFIX(LoadProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5384(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25287,13 +25480,13 @@ GL_PREFIX(LoadProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5384(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5384(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25307,7 +25500,7 @@ GL_PREFIX(LoadProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5384(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(LoadProgramNV), .-GL_PREFIX(LoadProgramNV)
@@ -25318,7 +25511,7 @@ GL_PREFIX(LoadProgramNV):
 GL_PREFIX(ProgramParameters4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5392(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25332,13 +25525,13 @@ GL_PREFIX(ProgramParameters4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5392(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5392(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25352,7 +25545,7 @@ GL_PREFIX(ProgramParameters4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5392(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramParameters4dvNV), .-GL_PREFIX(ProgramParameters4dvNV)
@@ -25363,7 +25556,7 @@ GL_PREFIX(ProgramParameters4dvNV):
 GL_PREFIX(ProgramParameters4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5400(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25377,13 +25570,13 @@ GL_PREFIX(ProgramParameters4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5400(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5400(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25397,7 +25590,7 @@ GL_PREFIX(ProgramParameters4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5400(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramParameters4fvNV), .-GL_PREFIX(ProgramParameters4fvNV)
@@ -25408,7 +25601,7 @@ GL_PREFIX(ProgramParameters4fvNV):
 GL_PREFIX(RequestResidentProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5408(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25418,13 +25611,13 @@ GL_PREFIX(RequestResidentProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5408(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5408(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25434,7 +25627,7 @@ GL_PREFIX(RequestResidentProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5408(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(RequestResidentProgramsNV), .-GL_PREFIX(RequestResidentProgramsNV)
@@ -25445,7 +25638,7 @@ GL_PREFIX(RequestResidentProgramsNV):
 GL_PREFIX(TrackMatrixNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5416(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25459,13 +25652,13 @@ GL_PREFIX(TrackMatrixNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5416(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5416(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25479,7 +25672,7 @@ GL_PREFIX(TrackMatrixNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5416(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TrackMatrixNV), .-GL_PREFIX(TrackMatrixNV)
@@ -25490,7 +25683,7 @@ GL_PREFIX(TrackMatrixNV):
 GL_PREFIX(VertexAttrib1dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5424(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25500,13 +25693,13 @@ GL_PREFIX(VertexAttrib1dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5424(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5424(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25516,7 +25709,7 @@ GL_PREFIX(VertexAttrib1dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5424(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1dNV), .-GL_PREFIX(VertexAttrib1dNV)
@@ -25527,7 +25720,7 @@ GL_PREFIX(VertexAttrib1dNV):
 GL_PREFIX(VertexAttrib1dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5432(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25537,13 +25730,13 @@ GL_PREFIX(VertexAttrib1dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5432(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5432(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25553,7 +25746,7 @@ GL_PREFIX(VertexAttrib1dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5432(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1dvNV), .-GL_PREFIX(VertexAttrib1dvNV)
@@ -25564,7 +25757,7 @@ GL_PREFIX(VertexAttrib1dvNV):
 GL_PREFIX(VertexAttrib1fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5440(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25574,13 +25767,13 @@ GL_PREFIX(VertexAttrib1fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5440(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5440(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25590,7 +25783,7 @@ GL_PREFIX(VertexAttrib1fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5440(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1fNV), .-GL_PREFIX(VertexAttrib1fNV)
@@ -25601,7 +25794,7 @@ GL_PREFIX(VertexAttrib1fNV):
 GL_PREFIX(VertexAttrib1fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5448(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25611,13 +25804,13 @@ GL_PREFIX(VertexAttrib1fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5448(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5448(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25627,7 +25820,7 @@ GL_PREFIX(VertexAttrib1fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5448(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1fvNV), .-GL_PREFIX(VertexAttrib1fvNV)
@@ -25638,7 +25831,7 @@ GL_PREFIX(VertexAttrib1fvNV):
 GL_PREFIX(VertexAttrib1sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5456(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25648,13 +25841,13 @@ GL_PREFIX(VertexAttrib1sNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5456(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5456(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25664,7 +25857,7 @@ GL_PREFIX(VertexAttrib1sNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5456(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1sNV), .-GL_PREFIX(VertexAttrib1sNV)
@@ -25675,7 +25868,7 @@ GL_PREFIX(VertexAttrib1sNV):
 GL_PREFIX(VertexAttrib1svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5464(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25685,13 +25878,13 @@ GL_PREFIX(VertexAttrib1svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5464(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5464(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25701,7 +25894,7 @@ GL_PREFIX(VertexAttrib1svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5464(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1svNV), .-GL_PREFIX(VertexAttrib1svNV)
@@ -25712,7 +25905,7 @@ GL_PREFIX(VertexAttrib1svNV):
 GL_PREFIX(VertexAttrib2dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5472(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25724,13 +25917,13 @@ GL_PREFIX(VertexAttrib2dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5472(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5472(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25742,7 +25935,7 @@ GL_PREFIX(VertexAttrib2dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5472(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2dNV), .-GL_PREFIX(VertexAttrib2dNV)
@@ -25753,7 +25946,7 @@ GL_PREFIX(VertexAttrib2dNV):
 GL_PREFIX(VertexAttrib2dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5480(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25763,13 +25956,13 @@ GL_PREFIX(VertexAttrib2dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5480(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5480(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25779,7 +25972,7 @@ GL_PREFIX(VertexAttrib2dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5480(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2dvNV), .-GL_PREFIX(VertexAttrib2dvNV)
@@ -25790,7 +25983,7 @@ GL_PREFIX(VertexAttrib2dvNV):
 GL_PREFIX(VertexAttrib2fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5488(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25802,13 +25995,13 @@ GL_PREFIX(VertexAttrib2fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5488(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5488(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25820,7 +26013,7 @@ GL_PREFIX(VertexAttrib2fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5488(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2fNV), .-GL_PREFIX(VertexAttrib2fNV)
@@ -25831,7 +26024,7 @@ GL_PREFIX(VertexAttrib2fNV):
 GL_PREFIX(VertexAttrib2fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5496(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25841,13 +26034,13 @@ GL_PREFIX(VertexAttrib2fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5496(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5496(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25857,7 +26050,7 @@ GL_PREFIX(VertexAttrib2fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5496(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2fvNV), .-GL_PREFIX(VertexAttrib2fvNV)
@@ -25868,7 +26061,7 @@ GL_PREFIX(VertexAttrib2fvNV):
 GL_PREFIX(VertexAttrib2sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5504(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25878,13 +26071,13 @@ GL_PREFIX(VertexAttrib2sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5504(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5504(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25894,7 +26087,7 @@ GL_PREFIX(VertexAttrib2sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5504(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2sNV), .-GL_PREFIX(VertexAttrib2sNV)
@@ -25905,7 +26098,7 @@ GL_PREFIX(VertexAttrib2sNV):
 GL_PREFIX(VertexAttrib2svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5512(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25915,13 +26108,13 @@ GL_PREFIX(VertexAttrib2svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5512(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5512(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25931,7 +26124,7 @@ GL_PREFIX(VertexAttrib2svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5512(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2svNV), .-GL_PREFIX(VertexAttrib2svNV)
@@ -25942,7 +26135,7 @@ GL_PREFIX(VertexAttrib2svNV):
 GL_PREFIX(VertexAttrib3dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5520(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -25956,13 +26149,13 @@ GL_PREFIX(VertexAttrib3dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5520(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5520(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -25976,7 +26169,7 @@ GL_PREFIX(VertexAttrib3dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5520(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3dNV), .-GL_PREFIX(VertexAttrib3dNV)
@@ -25987,7 +26180,7 @@ GL_PREFIX(VertexAttrib3dNV):
 GL_PREFIX(VertexAttrib3dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5528(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25997,13 +26190,13 @@ GL_PREFIX(VertexAttrib3dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5528(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5528(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26013,7 +26206,7 @@ GL_PREFIX(VertexAttrib3dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5528(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3dvNV), .-GL_PREFIX(VertexAttrib3dvNV)
@@ -26024,7 +26217,7 @@ GL_PREFIX(VertexAttrib3dvNV):
 GL_PREFIX(VertexAttrib3fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5536(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26038,13 +26231,13 @@ GL_PREFIX(VertexAttrib3fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5536(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5536(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26058,7 +26251,7 @@ GL_PREFIX(VertexAttrib3fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5536(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3fNV), .-GL_PREFIX(VertexAttrib3fNV)
@@ -26069,7 +26262,7 @@ GL_PREFIX(VertexAttrib3fNV):
 GL_PREFIX(VertexAttrib3fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5544(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26079,13 +26272,13 @@ GL_PREFIX(VertexAttrib3fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5544(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5544(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26095,7 +26288,7 @@ GL_PREFIX(VertexAttrib3fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5544(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3fvNV), .-GL_PREFIX(VertexAttrib3fvNV)
@@ -26106,7 +26299,7 @@ GL_PREFIX(VertexAttrib3fvNV):
 GL_PREFIX(VertexAttrib3sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5552(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26120,13 +26313,13 @@ GL_PREFIX(VertexAttrib3sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5552(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5552(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26140,7 +26333,7 @@ GL_PREFIX(VertexAttrib3sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5552(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3sNV), .-GL_PREFIX(VertexAttrib3sNV)
@@ -26151,7 +26344,7 @@ GL_PREFIX(VertexAttrib3sNV):
 GL_PREFIX(VertexAttrib3svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5560(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26161,13 +26354,13 @@ GL_PREFIX(VertexAttrib3svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5560(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5560(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26177,7 +26370,7 @@ GL_PREFIX(VertexAttrib3svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5560(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3svNV), .-GL_PREFIX(VertexAttrib3svNV)
@@ -26188,7 +26381,7 @@ GL_PREFIX(VertexAttrib3svNV):
 GL_PREFIX(VertexAttrib4dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5568(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26204,13 +26397,13 @@ GL_PREFIX(VertexAttrib4dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5568(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5568(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26226,7 +26419,7 @@ GL_PREFIX(VertexAttrib4dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5568(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4dNV), .-GL_PREFIX(VertexAttrib4dNV)
@@ -26237,7 +26430,7 @@ GL_PREFIX(VertexAttrib4dNV):
 GL_PREFIX(VertexAttrib4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5576(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26247,13 +26440,13 @@ GL_PREFIX(VertexAttrib4dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5576(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5576(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26263,7 +26456,7 @@ GL_PREFIX(VertexAttrib4dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5576(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4dvNV), .-GL_PREFIX(VertexAttrib4dvNV)
@@ -26274,7 +26467,7 @@ GL_PREFIX(VertexAttrib4dvNV):
 GL_PREFIX(VertexAttrib4fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5584(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26290,13 +26483,13 @@ GL_PREFIX(VertexAttrib4fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5584(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5584(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26312,7 +26505,7 @@ GL_PREFIX(VertexAttrib4fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5584(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4fNV), .-GL_PREFIX(VertexAttrib4fNV)
@@ -26323,7 +26516,7 @@ GL_PREFIX(VertexAttrib4fNV):
 GL_PREFIX(VertexAttrib4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5592(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26333,13 +26526,13 @@ GL_PREFIX(VertexAttrib4fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5592(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5592(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26349,7 +26542,7 @@ GL_PREFIX(VertexAttrib4fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5592(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4fvNV), .-GL_PREFIX(VertexAttrib4fvNV)
@@ -26360,7 +26553,7 @@ GL_PREFIX(VertexAttrib4fvNV):
 GL_PREFIX(VertexAttrib4sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5600(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26374,13 +26567,13 @@ GL_PREFIX(VertexAttrib4sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5600(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5600(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26394,7 +26587,7 @@ GL_PREFIX(VertexAttrib4sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5600(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4sNV), .-GL_PREFIX(VertexAttrib4sNV)
@@ -26405,7 +26598,7 @@ GL_PREFIX(VertexAttrib4sNV):
 GL_PREFIX(VertexAttrib4svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5608(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26415,13 +26608,13 @@ GL_PREFIX(VertexAttrib4svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5608(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5608(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26431,7 +26624,7 @@ GL_PREFIX(VertexAttrib4svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5608(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4svNV), .-GL_PREFIX(VertexAttrib4svNV)
@@ -26442,7 +26635,7 @@ GL_PREFIX(VertexAttrib4svNV):
 GL_PREFIX(VertexAttrib4ubNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5616(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26456,13 +26649,13 @@ GL_PREFIX(VertexAttrib4ubNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5616(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5616(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26476,7 +26669,7 @@ GL_PREFIX(VertexAttrib4ubNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5616(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4ubNV), .-GL_PREFIX(VertexAttrib4ubNV)
@@ -26487,7 +26680,7 @@ GL_PREFIX(VertexAttrib4ubNV):
 GL_PREFIX(VertexAttrib4ubvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5624(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26497,13 +26690,13 @@ GL_PREFIX(VertexAttrib4ubvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5624(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5624(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26513,7 +26706,7 @@ GL_PREFIX(VertexAttrib4ubvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5624(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4ubvNV), .-GL_PREFIX(VertexAttrib4ubvNV)
@@ -26524,7 +26717,7 @@ GL_PREFIX(VertexAttrib4ubvNV):
 GL_PREFIX(VertexAttribPointerNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5632(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26538,13 +26731,13 @@ GL_PREFIX(VertexAttribPointerNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5632(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5632(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26558,7 +26751,7 @@ GL_PREFIX(VertexAttribPointerNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5632(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribPointerNV), .-GL_PREFIX(VertexAttribPointerNV)
@@ -26569,7 +26762,7 @@ GL_PREFIX(VertexAttribPointerNV):
 GL_PREFIX(VertexAttribs1dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5640(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26579,13 +26772,13 @@ GL_PREFIX(VertexAttribs1dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5640(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5640(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26595,7 +26788,7 @@ GL_PREFIX(VertexAttribs1dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5640(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1dvNV), .-GL_PREFIX(VertexAttribs1dvNV)
@@ -26606,7 +26799,7 @@ GL_PREFIX(VertexAttribs1dvNV):
 GL_PREFIX(VertexAttribs1fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5648(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26616,13 +26809,13 @@ GL_PREFIX(VertexAttribs1fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5648(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5648(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26632,7 +26825,7 @@ GL_PREFIX(VertexAttribs1fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5648(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1fvNV), .-GL_PREFIX(VertexAttribs1fvNV)
@@ -26643,7 +26836,7 @@ GL_PREFIX(VertexAttribs1fvNV):
 GL_PREFIX(VertexAttribs1svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5656(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26653,13 +26846,13 @@ GL_PREFIX(VertexAttribs1svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5656(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5656(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26669,7 +26862,7 @@ GL_PREFIX(VertexAttribs1svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5656(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1svNV), .-GL_PREFIX(VertexAttribs1svNV)
@@ -26680,7 +26873,7 @@ GL_PREFIX(VertexAttribs1svNV):
 GL_PREFIX(VertexAttribs2dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5664(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26690,13 +26883,13 @@ GL_PREFIX(VertexAttribs2dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5664(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5664(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26706,7 +26899,7 @@ GL_PREFIX(VertexAttribs2dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5664(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2dvNV), .-GL_PREFIX(VertexAttribs2dvNV)
@@ -26717,7 +26910,7 @@ GL_PREFIX(VertexAttribs2dvNV):
 GL_PREFIX(VertexAttribs2fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5672(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26727,13 +26920,13 @@ GL_PREFIX(VertexAttribs2fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5672(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5672(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26743,7 +26936,7 @@ GL_PREFIX(VertexAttribs2fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5672(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2fvNV), .-GL_PREFIX(VertexAttribs2fvNV)
@@ -26754,7 +26947,7 @@ GL_PREFIX(VertexAttribs2fvNV):
 GL_PREFIX(VertexAttribs2svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5680(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26764,13 +26957,13 @@ GL_PREFIX(VertexAttribs2svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5680(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5680(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26780,7 +26973,7 @@ GL_PREFIX(VertexAttribs2svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5680(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2svNV), .-GL_PREFIX(VertexAttribs2svNV)
@@ -26791,7 +26984,7 @@ GL_PREFIX(VertexAttribs2svNV):
 GL_PREFIX(VertexAttribs3dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5688(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26801,13 +26994,13 @@ GL_PREFIX(VertexAttribs3dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5688(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5688(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26817,7 +27010,7 @@ GL_PREFIX(VertexAttribs3dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5688(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3dvNV), .-GL_PREFIX(VertexAttribs3dvNV)
@@ -26828,7 +27021,7 @@ GL_PREFIX(VertexAttribs3dvNV):
 GL_PREFIX(VertexAttribs3fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5696(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26838,13 +27031,13 @@ GL_PREFIX(VertexAttribs3fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5696(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5696(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26854,7 +27047,7 @@ GL_PREFIX(VertexAttribs3fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5696(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3fvNV), .-GL_PREFIX(VertexAttribs3fvNV)
@@ -26865,7 +27058,7 @@ GL_PREFIX(VertexAttribs3fvNV):
 GL_PREFIX(VertexAttribs3svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5704(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26875,13 +27068,13 @@ GL_PREFIX(VertexAttribs3svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5704(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5704(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26891,7 +27084,7 @@ GL_PREFIX(VertexAttribs3svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5704(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3svNV), .-GL_PREFIX(VertexAttribs3svNV)
@@ -26902,7 +27095,7 @@ GL_PREFIX(VertexAttribs3svNV):
 GL_PREFIX(VertexAttribs4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5712(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26912,13 +27105,13 @@ GL_PREFIX(VertexAttribs4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5712(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5712(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26928,7 +27121,7 @@ GL_PREFIX(VertexAttribs4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5712(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4dvNV), .-GL_PREFIX(VertexAttribs4dvNV)
@@ -26939,7 +27132,7 @@ GL_PREFIX(VertexAttribs4dvNV):
 GL_PREFIX(VertexAttribs4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5720(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26949,13 +27142,13 @@ GL_PREFIX(VertexAttribs4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5720(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5720(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26965,7 +27158,7 @@ GL_PREFIX(VertexAttribs4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5720(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4fvNV), .-GL_PREFIX(VertexAttribs4fvNV)
@@ -26976,7 +27169,7 @@ GL_PREFIX(VertexAttribs4fvNV):
 GL_PREFIX(VertexAttribs4svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5728(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26986,13 +27179,13 @@ GL_PREFIX(VertexAttribs4svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5728(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5728(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27002,7 +27195,7 @@ GL_PREFIX(VertexAttribs4svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5728(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4svNV), .-GL_PREFIX(VertexAttribs4svNV)
@@ -27013,7 +27206,7 @@ GL_PREFIX(VertexAttribs4svNV):
 GL_PREFIX(VertexAttribs4ubvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5736(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27023,13 +27216,13 @@ GL_PREFIX(VertexAttribs4ubvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5736(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5736(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27039,7 +27232,7 @@ GL_PREFIX(VertexAttribs4ubvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5736(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4ubvNV), .-GL_PREFIX(VertexAttribs4ubvNV)
@@ -27050,7 +27243,7 @@ GL_PREFIX(VertexAttribs4ubvNV):
 GL_PREFIX(GetTexBumpParameterfvATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5744(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27060,13 +27253,13 @@ GL_PREFIX(GetTexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5744(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5744(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27076,7 +27269,7 @@ GL_PREFIX(GetTexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5744(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTexBumpParameterfvATI), .-GL_PREFIX(GetTexBumpParameterfvATI)
@@ -27087,7 +27280,7 @@ GL_PREFIX(GetTexBumpParameterfvATI):
 GL_PREFIX(GetTexBumpParameterivATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5752(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27097,13 +27290,13 @@ GL_PREFIX(GetTexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5752(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5752(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27113,7 +27306,7 @@ GL_PREFIX(GetTexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5752(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTexBumpParameterivATI), .-GL_PREFIX(GetTexBumpParameterivATI)
@@ -27124,7 +27317,7 @@ GL_PREFIX(GetTexBumpParameterivATI):
 GL_PREFIX(TexBumpParameterfvATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5760(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27134,13 +27327,13 @@ GL_PREFIX(TexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5760(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5760(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27150,7 +27343,7 @@ GL_PREFIX(TexBumpParameterfvATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5760(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexBumpParameterfvATI), .-GL_PREFIX(TexBumpParameterfvATI)
@@ -27161,7 +27354,7 @@ GL_PREFIX(TexBumpParameterfvATI):
 GL_PREFIX(TexBumpParameterivATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5768(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27171,13 +27364,13 @@ GL_PREFIX(TexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5768(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5768(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27187,7 +27380,7 @@ GL_PREFIX(TexBumpParameterivATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5768(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexBumpParameterivATI), .-GL_PREFIX(TexBumpParameterivATI)
@@ -27198,7 +27391,7 @@ GL_PREFIX(TexBumpParameterivATI):
 GL_PREFIX(AlphaFragmentOp1ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5776(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27216,13 +27409,13 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5776(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5776(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27240,7 +27433,7 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5776(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp1ATI), .-GL_PREFIX(AlphaFragmentOp1ATI)
@@ -27251,7 +27444,7 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 GL_PREFIX(AlphaFragmentOp2ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5784(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27269,13 +27462,13 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5784(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5784(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27293,7 +27486,7 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5784(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp2ATI), .-GL_PREFIX(AlphaFragmentOp2ATI)
@@ -27304,7 +27497,7 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 GL_PREFIX(AlphaFragmentOp3ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5792(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27322,13 +27515,13 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5792(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5792(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27346,7 +27539,7 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5792(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp3ATI), .-GL_PREFIX(AlphaFragmentOp3ATI)
@@ -27357,25 +27550,25 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 GL_PREFIX(BeginFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5800(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	5800(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5800(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	5800(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BeginFragmentShaderATI), .-GL_PREFIX(BeginFragmentShaderATI)
@@ -27386,25 +27579,25 @@ GL_PREFIX(BeginFragmentShaderATI):
 GL_PREFIX(BindFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5808(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5808(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5808(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5808(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindFragmentShaderATI), .-GL_PREFIX(BindFragmentShaderATI)
@@ -27415,7 +27608,7 @@ GL_PREFIX(BindFragmentShaderATI):
 GL_PREFIX(ColorFragmentOp1ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5816(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27433,13 +27626,13 @@ GL_PREFIX(ColorFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5816(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5816(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27457,7 +27650,7 @@ GL_PREFIX(ColorFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5816(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp1ATI), .-GL_PREFIX(ColorFragmentOp1ATI)
@@ -27468,7 +27661,7 @@ GL_PREFIX(ColorFragmentOp1ATI):
 GL_PREFIX(ColorFragmentOp2ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5824(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27486,13 +27679,13 @@ GL_PREFIX(ColorFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5824(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5824(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27510,7 +27703,7 @@ GL_PREFIX(ColorFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5824(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp2ATI), .-GL_PREFIX(ColorFragmentOp2ATI)
@@ -27521,7 +27714,7 @@ GL_PREFIX(ColorFragmentOp2ATI):
 GL_PREFIX(ColorFragmentOp3ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5832(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27539,13 +27732,13 @@ GL_PREFIX(ColorFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5832(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5832(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27563,7 +27756,7 @@ GL_PREFIX(ColorFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5832(%rax), %r11
+	movq	5872(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp3ATI), .-GL_PREFIX(ColorFragmentOp3ATI)
@@ -27574,25 +27767,25 @@ GL_PREFIX(ColorFragmentOp3ATI):
 GL_PREFIX(DeleteFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5840(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5840(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5840(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5840(%rax), %r11
+	movq	5880(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteFragmentShaderATI), .-GL_PREFIX(DeleteFragmentShaderATI)
@@ -27603,25 +27796,25 @@ GL_PREFIX(DeleteFragmentShaderATI):
 GL_PREFIX(EndFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5848(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	5848(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5848(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	5848(%rax), %r11
+	movq	5888(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(EndFragmentShaderATI), .-GL_PREFIX(EndFragmentShaderATI)
@@ -27632,25 +27825,25 @@ GL_PREFIX(EndFragmentShaderATI):
 GL_PREFIX(GenFragmentShadersATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5856(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5856(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5856(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5856(%rax), %r11
+	movq	5896(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenFragmentShadersATI), .-GL_PREFIX(GenFragmentShadersATI)
@@ -27661,7 +27854,7 @@ GL_PREFIX(GenFragmentShadersATI):
 GL_PREFIX(PassTexCoordATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5864(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27671,13 +27864,13 @@ GL_PREFIX(PassTexCoordATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5864(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5864(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27687,7 +27880,7 @@ GL_PREFIX(PassTexCoordATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5864(%rax), %r11
+	movq	5904(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PassTexCoordATI), .-GL_PREFIX(PassTexCoordATI)
@@ -27698,7 +27891,7 @@ GL_PREFIX(PassTexCoordATI):
 GL_PREFIX(SampleMapATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5872(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27708,13 +27901,13 @@ GL_PREFIX(SampleMapATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5872(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5872(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27724,7 +27917,7 @@ GL_PREFIX(SampleMapATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5872(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SampleMapATI), .-GL_PREFIX(SampleMapATI)
@@ -27735,7 +27928,7 @@ GL_PREFIX(SampleMapATI):
 GL_PREFIX(SetFragmentShaderConstantATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5880(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27745,13 +27938,13 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5880(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5880(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27761,7 +27954,7 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5880(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SetFragmentShaderConstantATI), .-GL_PREFIX(SetFragmentShaderConstantATI)
@@ -27772,7 +27965,7 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 GL_PREFIX(PointParameteriNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5888(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27782,13 +27975,13 @@ GL_PREFIX(PointParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5888(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5888(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27798,7 +27991,7 @@ GL_PREFIX(PointParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5888(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameteriNV), .-GL_PREFIX(PointParameteriNV)
@@ -27809,7 +28002,7 @@ GL_PREFIX(PointParameteriNV):
 GL_PREFIX(PointParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5896(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27819,13 +28012,13 @@ GL_PREFIX(PointParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5896(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5896(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27835,79 +28028,79 @@ GL_PREFIX(PointParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5896(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterivNV), .-GL_PREFIX(PointParameterivNV)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_738)
-	.type	GL_PREFIX(_dispatch_stub_738), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_738))
-GL_PREFIX(_dispatch_stub_738):
+	.globl	GL_PREFIX(_dispatch_stub_743)
+	.type	GL_PREFIX(_dispatch_stub_743), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_743))
+GL_PREFIX(_dispatch_stub_743):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5904(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5904(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5904(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5904(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_738), .-GL_PREFIX(_dispatch_stub_738)
+	.size	GL_PREFIX(_dispatch_stub_743), .-GL_PREFIX(_dispatch_stub_743)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_739)
-	.type	GL_PREFIX(_dispatch_stub_739), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_739))
-GL_PREFIX(_dispatch_stub_739):
+	.globl	GL_PREFIX(_dispatch_stub_744)
+	.type	GL_PREFIX(_dispatch_stub_744), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_744))
+GL_PREFIX(_dispatch_stub_744):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5912(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5912(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5912(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5912(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_739), .-GL_PREFIX(_dispatch_stub_739)
+	.size	GL_PREFIX(_dispatch_stub_744), .-GL_PREFIX(_dispatch_stub_744)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_740)
-	.type	GL_PREFIX(_dispatch_stub_740), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_740))
-GL_PREFIX(_dispatch_stub_740):
+	.globl	GL_PREFIX(_dispatch_stub_745)
+	.type	GL_PREFIX(_dispatch_stub_745), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_745))
+GL_PREFIX(_dispatch_stub_745):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5920(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27917,13 +28110,13 @@ GL_PREFIX(_dispatch_stub_740):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5920(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5920(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27933,19 +28126,19 @@ GL_PREFIX(_dispatch_stub_740):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5920(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_740), .-GL_PREFIX(_dispatch_stub_740)
+	.size	GL_PREFIX(_dispatch_stub_745), .-GL_PREFIX(_dispatch_stub_745)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_741)
-	.type	GL_PREFIX(_dispatch_stub_741), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_741))
-GL_PREFIX(_dispatch_stub_741):
+	.globl	GL_PREFIX(_dispatch_stub_746)
+	.type	GL_PREFIX(_dispatch_stub_746), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_746))
+GL_PREFIX(_dispatch_stub_746):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5928(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27955,13 +28148,13 @@ GL_PREFIX(_dispatch_stub_741):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5928(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5928(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27971,40 +28164,40 @@ GL_PREFIX(_dispatch_stub_741):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5928(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_741), .-GL_PREFIX(_dispatch_stub_741)
+	.size	GL_PREFIX(_dispatch_stub_746), .-GL_PREFIX(_dispatch_stub_746)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_742)
-	.type	GL_PREFIX(_dispatch_stub_742), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_742))
-GL_PREFIX(_dispatch_stub_742):
+	.globl	GL_PREFIX(_dispatch_stub_747)
+	.type	GL_PREFIX(_dispatch_stub_747), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_747))
+GL_PREFIX(_dispatch_stub_747):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5936(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5936(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5936(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5936(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_742), .-GL_PREFIX(_dispatch_stub_742)
+	.size	GL_PREFIX(_dispatch_stub_747), .-GL_PREFIX(_dispatch_stub_747)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(GetProgramNamedParameterdvNV)
@@ -28012,7 +28205,7 @@ GL_PREFIX(_dispatch_stub_742):
 GL_PREFIX(GetProgramNamedParameterdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5944(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28026,13 +28219,13 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5944(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5944(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28046,7 +28239,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5944(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramNamedParameterdvNV), .-GL_PREFIX(GetProgramNamedParameterdvNV)
@@ -28057,7 +28250,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 GL_PREFIX(GetProgramNamedParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5952(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28071,13 +28264,13 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5952(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5952(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28091,7 +28284,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5952(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramNamedParameterfvNV), .-GL_PREFIX(GetProgramNamedParameterfvNV)
@@ -28102,7 +28295,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 GL_PREFIX(ProgramNamedParameter4dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5960(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$56, %rsp
@@ -28122,13 +28315,13 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5960(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5960(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$56, %rsp
@@ -28148,7 +28341,7 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5960(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4dNV), .-GL_PREFIX(ProgramNamedParameter4dNV)
@@ -28159,7 +28352,7 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 GL_PREFIX(ProgramNamedParameter4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5968(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28173,13 +28366,13 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5968(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5968(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28193,7 +28386,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5968(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4dvNV), .-GL_PREFIX(ProgramNamedParameter4dvNV)
@@ -28204,7 +28397,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 GL_PREFIX(ProgramNamedParameter4fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5976(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$56, %rsp
@@ -28224,13 +28417,13 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5976(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5976(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$56, %rsp
@@ -28250,7 +28443,7 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5976(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4fNV), .-GL_PREFIX(ProgramNamedParameter4fNV)
@@ -28261,7 +28454,7 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 GL_PREFIX(ProgramNamedParameter4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5984(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28275,13 +28468,13 @@ GL_PREFIX(ProgramNamedParameter4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5984(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5984(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28295,19 +28488,19 @@ GL_PREFIX(ProgramNamedParameter4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5984(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4fvNV), .-GL_PREFIX(ProgramNamedParameter4fvNV)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_749)
-	.type	GL_PREFIX(_dispatch_stub_749), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_749))
-GL_PREFIX(_dispatch_stub_749):
+	.globl	GL_PREFIX(_dispatch_stub_754)
+	.type	GL_PREFIX(_dispatch_stub_754), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_754))
+GL_PREFIX(_dispatch_stub_754):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5992(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28317,13 +28510,13 @@ GL_PREFIX(_dispatch_stub_749):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5992(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5992(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28333,19 +28526,19 @@ GL_PREFIX(_dispatch_stub_749):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5992(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_749), .-GL_PREFIX(_dispatch_stub_749)
+	.size	GL_PREFIX(_dispatch_stub_754), .-GL_PREFIX(_dispatch_stub_754)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_750)
-	.type	GL_PREFIX(_dispatch_stub_750), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_750))
-GL_PREFIX(_dispatch_stub_750):
+	.globl	GL_PREFIX(_dispatch_stub_755)
+	.type	GL_PREFIX(_dispatch_stub_755), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_755))
+GL_PREFIX(_dispatch_stub_755):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6000(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28355,13 +28548,13 @@ GL_PREFIX(_dispatch_stub_750):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6000(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6000(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28371,10 +28564,10 @@ GL_PREFIX(_dispatch_stub_750):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6000(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_750), .-GL_PREFIX(_dispatch_stub_750)
+	.size	GL_PREFIX(_dispatch_stub_755), .-GL_PREFIX(_dispatch_stub_755)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(BindFramebufferEXT)
@@ -28382,7 +28575,7 @@ GL_PREFIX(_dispatch_stub_750):
 GL_PREFIX(BindFramebufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6008(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28392,13 +28585,13 @@ GL_PREFIX(BindFramebufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6008(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6008(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28408,7 +28601,7 @@ GL_PREFIX(BindFramebufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6008(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindFramebufferEXT), .-GL_PREFIX(BindFramebufferEXT)
@@ -28419,7 +28612,7 @@ GL_PREFIX(BindFramebufferEXT):
 GL_PREFIX(BindRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6016(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28429,13 +28622,13 @@ GL_PREFIX(BindRenderbufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6016(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6016(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28445,7 +28638,7 @@ GL_PREFIX(BindRenderbufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6016(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindRenderbufferEXT), .-GL_PREFIX(BindRenderbufferEXT)
@@ -28456,25 +28649,25 @@ GL_PREFIX(BindRenderbufferEXT):
 GL_PREFIX(CheckFramebufferStatusEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6024(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6024(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6024(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6024(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CheckFramebufferStatusEXT), .-GL_PREFIX(CheckFramebufferStatusEXT)
@@ -28485,7 +28678,7 @@ GL_PREFIX(CheckFramebufferStatusEXT):
 GL_PREFIX(DeleteFramebuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6032(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28495,13 +28688,13 @@ GL_PREFIX(DeleteFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6032(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6032(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28511,7 +28704,7 @@ GL_PREFIX(DeleteFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6032(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteFramebuffersEXT), .-GL_PREFIX(DeleteFramebuffersEXT)
@@ -28522,7 +28715,7 @@ GL_PREFIX(DeleteFramebuffersEXT):
 GL_PREFIX(DeleteRenderbuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6040(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28532,13 +28725,13 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6040(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6040(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28548,7 +28741,7 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6040(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteRenderbuffersEXT), .-GL_PREFIX(DeleteRenderbuffersEXT)
@@ -28559,7 +28752,7 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 GL_PREFIX(FramebufferRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6048(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28573,13 +28766,13 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6048(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6048(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28593,7 +28786,7 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6048(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferRenderbufferEXT), .-GL_PREFIX(FramebufferRenderbufferEXT)
@@ -28604,7 +28797,7 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 GL_PREFIX(FramebufferTexture1DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6056(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28618,13 +28811,13 @@ GL_PREFIX(FramebufferTexture1DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6056(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6056(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28638,7 +28831,7 @@ GL_PREFIX(FramebufferTexture1DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6056(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture1DEXT), .-GL_PREFIX(FramebufferTexture1DEXT)
@@ -28649,7 +28842,7 @@ GL_PREFIX(FramebufferTexture1DEXT):
 GL_PREFIX(FramebufferTexture2DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6064(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28663,13 +28856,13 @@ GL_PREFIX(FramebufferTexture2DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6064(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6064(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28683,7 +28876,7 @@ GL_PREFIX(FramebufferTexture2DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6064(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture2DEXT), .-GL_PREFIX(FramebufferTexture2DEXT)
@@ -28694,7 +28887,7 @@ GL_PREFIX(FramebufferTexture2DEXT):
 GL_PREFIX(FramebufferTexture3DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6072(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28712,13 +28905,13 @@ GL_PREFIX(FramebufferTexture3DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6072(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6072(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28736,7 +28929,7 @@ GL_PREFIX(FramebufferTexture3DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6072(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture3DEXT), .-GL_PREFIX(FramebufferTexture3DEXT)
@@ -28747,7 +28940,7 @@ GL_PREFIX(FramebufferTexture3DEXT):
 GL_PREFIX(GenFramebuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6080(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28757,13 +28950,13 @@ GL_PREFIX(GenFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6080(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6080(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28773,7 +28966,7 @@ GL_PREFIX(GenFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6080(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenFramebuffersEXT), .-GL_PREFIX(GenFramebuffersEXT)
@@ -28784,7 +28977,7 @@ GL_PREFIX(GenFramebuffersEXT):
 GL_PREFIX(GenRenderbuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6088(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28794,13 +28987,13 @@ GL_PREFIX(GenRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6088(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6088(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28810,7 +29003,7 @@ GL_PREFIX(GenRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6088(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenRenderbuffersEXT), .-GL_PREFIX(GenRenderbuffersEXT)
@@ -28821,25 +29014,25 @@ GL_PREFIX(GenRenderbuffersEXT):
 GL_PREFIX(GenerateMipmapEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6096(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6096(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6096(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6096(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenerateMipmapEXT), .-GL_PREFIX(GenerateMipmapEXT)
@@ -28850,7 +29043,7 @@ GL_PREFIX(GenerateMipmapEXT):
 GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6104(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28864,13 +29057,13 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6104(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6104(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28884,7 +29077,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6104(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFramebufferAttachmentParameterivEXT), .-GL_PREFIX(GetFramebufferAttachmentParameterivEXT)
@@ -28895,7 +29088,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 GL_PREFIX(GetRenderbufferParameterivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6112(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28905,13 +29098,13 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6112(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6112(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28921,7 +29114,7 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6112(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetRenderbufferParameterivEXT), .-GL_PREFIX(GetRenderbufferParameterivEXT)
@@ -28932,25 +29125,25 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 GL_PREFIX(IsFramebufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6120(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6120(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6120(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6120(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsFramebufferEXT), .-GL_PREFIX(IsFramebufferEXT)
@@ -28961,25 +29154,25 @@ GL_PREFIX(IsFramebufferEXT):
 GL_PREFIX(IsRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6128(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6128(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6128(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6128(%rax), %r11
+	movq	6168(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsRenderbufferEXT), .-GL_PREFIX(IsRenderbufferEXT)
@@ -28990,7 +29183,7 @@ GL_PREFIX(IsRenderbufferEXT):
 GL_PREFIX(RenderbufferStorageEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6136(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29004,13 +29197,13 @@ GL_PREFIX(RenderbufferStorageEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6136(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6136(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29024,19 +29217,19 @@ GL_PREFIX(RenderbufferStorageEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6136(%rax), %r11
+	movq	6176(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(RenderbufferStorageEXT), .-GL_PREFIX(RenderbufferStorageEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_768)
-	.type	GL_PREFIX(_dispatch_stub_768), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_768))
-GL_PREFIX(_dispatch_stub_768):
+	.globl	GL_PREFIX(_dispatch_stub_773)
+	.type	GL_PREFIX(_dispatch_stub_773), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_773))
+GL_PREFIX(_dispatch_stub_773):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6144(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29054,13 +29247,13 @@ GL_PREFIX(_dispatch_stub_768):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6144(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6144(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29078,10 +29271,10 @@ GL_PREFIX(_dispatch_stub_768):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6144(%rax), %r11
+	movq	6184(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_768), .-GL_PREFIX(_dispatch_stub_768)
+	.size	GL_PREFIX(_dispatch_stub_773), .-GL_PREFIX(_dispatch_stub_773)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(FramebufferTextureLayerEXT)
@@ -29089,7 +29282,7 @@ GL_PREFIX(_dispatch_stub_768):
 GL_PREFIX(FramebufferTextureLayerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6152(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29103,13 +29296,13 @@ GL_PREFIX(FramebufferTextureLayerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6152(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6152(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29123,19 +29316,48 @@ GL_PREFIX(FramebufferTextureLayerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6152(%rax), %r11
+	movq	6192(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTextureLayerEXT), .-GL_PREFIX(FramebufferTextureLayerEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_770)
-	.type	GL_PREFIX(_dispatch_stub_770), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_770))
-GL_PREFIX(_dispatch_stub_770):
+	.globl	GL_PREFIX(ProvokingVertexEXT)
+	.type	GL_PREFIX(ProvokingVertexEXT), @function
+GL_PREFIX(ProvokingVertexEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6160(%rax), %r11
+	movq	6200(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	6200(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	6200(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	6200(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(ProvokingVertexEXT), .-GL_PREFIX(ProvokingVertexEXT)
+
+	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_776)
+	.type	GL_PREFIX(_dispatch_stub_776), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_776))
+GL_PREFIX(_dispatch_stub_776):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	6208(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29149,13 +29371,13 @@ GL_PREFIX(_dispatch_stub_770):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6160(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6160(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29169,19 +29391,19 @@ GL_PREFIX(_dispatch_stub_770):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6160(%rax), %r11
+	movq	6208(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_770), .-GL_PREFIX(_dispatch_stub_770)
+	.size	GL_PREFIX(_dispatch_stub_776), .-GL_PREFIX(_dispatch_stub_776)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_771)
-	.type	GL_PREFIX(_dispatch_stub_771), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_771))
-GL_PREFIX(_dispatch_stub_771):
+	.globl	GL_PREFIX(_dispatch_stub_777)
+	.type	GL_PREFIX(_dispatch_stub_777), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_777))
+GL_PREFIX(_dispatch_stub_777):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6168(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29195,13 +29417,13 @@ GL_PREFIX(_dispatch_stub_771):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6168(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6168(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29215,19 +29437,19 @@ GL_PREFIX(_dispatch_stub_771):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6168(%rax), %r11
+	movq	6216(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_771), .-GL_PREFIX(_dispatch_stub_771)
+	.size	GL_PREFIX(_dispatch_stub_777), .-GL_PREFIX(_dispatch_stub_777)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_772)
-	.type	GL_PREFIX(_dispatch_stub_772), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_772))
-GL_PREFIX(_dispatch_stub_772):
+	.globl	GL_PREFIX(_dispatch_stub_778)
+	.type	GL_PREFIX(_dispatch_stub_778), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_778))
+GL_PREFIX(_dispatch_stub_778):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6176(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29241,13 +29463,13 @@ GL_PREFIX(_dispatch_stub_772):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6176(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6176(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29261,19 +29483,19 @@ GL_PREFIX(_dispatch_stub_772):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6176(%rax), %r11
+	movq	6224(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_772), .-GL_PREFIX(_dispatch_stub_772)
+	.size	GL_PREFIX(_dispatch_stub_778), .-GL_PREFIX(_dispatch_stub_778)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_773)
-	.type	GL_PREFIX(_dispatch_stub_773), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_773))
-GL_PREFIX(_dispatch_stub_773):
+	.globl	GL_PREFIX(_dispatch_stub_779)
+	.type	GL_PREFIX(_dispatch_stub_779), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_779))
+GL_PREFIX(_dispatch_stub_779):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6184(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29283,13 +29505,13 @@ GL_PREFIX(_dispatch_stub_773):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6184(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6184(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29299,19 +29521,19 @@ GL_PREFIX(_dispatch_stub_773):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6184(%rax), %r11
+	movq	6232(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_773), .-GL_PREFIX(_dispatch_stub_773)
+	.size	GL_PREFIX(_dispatch_stub_779), .-GL_PREFIX(_dispatch_stub_779)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_774)
-	.type	GL_PREFIX(_dispatch_stub_774), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_774))
-GL_PREFIX(_dispatch_stub_774):
+	.globl	GL_PREFIX(_dispatch_stub_780)
+	.type	GL_PREFIX(_dispatch_stub_780), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_780))
+GL_PREFIX(_dispatch_stub_780):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6192(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29321,13 +29543,13 @@ GL_PREFIX(_dispatch_stub_774):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6192(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6192(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29337,10 +29559,10 @@ GL_PREFIX(_dispatch_stub_774):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6192(%rax), %r11
+	movq	6240(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_774), .-GL_PREFIX(_dispatch_stub_774)
+	.size	GL_PREFIX(_dispatch_stub_780), .-GL_PREFIX(_dispatch_stub_780)
 
 	.globl GL_PREFIX(ArrayElementEXT) ; .set GL_PREFIX(ArrayElementEXT), GL_PREFIX(ArrayElement)
 	.globl GL_PREFIX(BindTextureEXT) ; .set GL_PREFIX(BindTextureEXT), GL_PREFIX(BindTexture)
@@ -29595,7 +29817,9 @@ GL_PREFIX(_dispatch_stub_774):
 	.globl GL_PREFIX(IsProgramARB) ; .set GL_PREFIX(IsProgramARB), GL_PREFIX(IsProgramNV)
 	.globl GL_PREFIX(PointParameteri) ; .set GL_PREFIX(PointParameteri), GL_PREFIX(PointParameteriNV)
 	.globl GL_PREFIX(PointParameteriv) ; .set GL_PREFIX(PointParameteriv), GL_PREFIX(PointParameterivNV)
-	.globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_750)
+	.globl GL_PREFIX(DeleteVertexArrays) ; .set GL_PREFIX(DeleteVertexArrays), GL_PREFIX(_dispatch_stub_745)
+	.globl GL_PREFIX(IsVertexArray) ; .set GL_PREFIX(IsVertexArray), GL_PREFIX(_dispatch_stub_747)
+	.globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_755)
 	.globl GL_PREFIX(BindFramebuffer) ; .set GL_PREFIX(BindFramebuffer), GL_PREFIX(BindFramebufferEXT)
 	.globl GL_PREFIX(BindRenderbuffer) ; .set GL_PREFIX(BindRenderbuffer), GL_PREFIX(BindRenderbufferEXT)
 	.globl GL_PREFIX(CheckFramebufferStatus) ; .set GL_PREFIX(CheckFramebufferStatus), GL_PREFIX(CheckFramebufferStatusEXT)
@@ -29613,7 +29837,7 @@ GL_PREFIX(_dispatch_stub_774):
 	.globl GL_PREFIX(IsFramebuffer) ; .set GL_PREFIX(IsFramebuffer), GL_PREFIX(IsFramebufferEXT)
 	.globl GL_PREFIX(IsRenderbuffer) ; .set GL_PREFIX(IsRenderbuffer), GL_PREFIX(IsRenderbufferEXT)
 	.globl GL_PREFIX(RenderbufferStorage) ; .set GL_PREFIX(RenderbufferStorage), GL_PREFIX(RenderbufferStorageEXT)
-	.globl GL_PREFIX(BlitFramebuffer) ; .set GL_PREFIX(BlitFramebuffer), GL_PREFIX(_dispatch_stub_768)
+	.globl GL_PREFIX(BlitFramebuffer) ; .set GL_PREFIX(BlitFramebuffer), GL_PREFIX(_dispatch_stub_773)
 	.globl GL_PREFIX(FramebufferTextureLayer) ; .set GL_PREFIX(FramebufferTextureLayer), GL_PREFIX(FramebufferTextureLayerEXT)
 
 #if defined(GLX_USE_TLS) && defined(__linux__)
diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S
index 7aa344f214..40fc6f2229 100644
--- a/src/mesa/x86/glapi_x86.S
+++ b/src/mesa/x86/glapi_x86.S
@@ -713,23 +713,28 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(GetAttribLocationARB, _gloffset_GetAttribLocationARB, GetAttribLocationARB@8)
 	GL_STUB(DrawBuffersARB, _gloffset_DrawBuffersARB, DrawBuffersARB@8)
 	GL_STUB(RenderbufferStorageMultisample, _gloffset_RenderbufferStorageMultisample, RenderbufferStorageMultisample@20)
+	GL_STUB(FlushMappedBufferRange, _gloffset_FlushMappedBufferRange, FlushMappedBufferRange@12)
+	GL_STUB(MapBufferRange, _gloffset_MapBufferRange, MapBufferRange@16)
+	GL_STUB(BindVertexArray, _gloffset_BindVertexArray, BindVertexArray@4)
+	GL_STUB(GenVertexArrays, _gloffset_GenVertexArrays, GenVertexArrays@8)
+	GL_STUB(CopyBufferSubData, _gloffset_CopyBufferSubData, CopyBufferSubData@20)
 	GL_STUB(PolygonOffsetEXT, _gloffset_PolygonOffsetEXT, PolygonOffsetEXT@8)
-	GL_STUB(_dispatch_stub_563, _gloffset_GetPixelTexGenParameterfvSGIS, _dispatch_stub_563@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_563, _dispatch_stub_563@8))
-	GL_STUB(_dispatch_stub_564, _gloffset_GetPixelTexGenParameterivSGIS, _dispatch_stub_564@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_564, _dispatch_stub_564@8))
-	GL_STUB(_dispatch_stub_565, _gloffset_PixelTexGenParameterfSGIS, _dispatch_stub_565@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_565, _dispatch_stub_565@8))
-	GL_STUB(_dispatch_stub_566, _gloffset_PixelTexGenParameterfvSGIS, _dispatch_stub_566@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_566, _dispatch_stub_566@8))
-	GL_STUB(_dispatch_stub_567, _gloffset_PixelTexGenParameteriSGIS, _dispatch_stub_567@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_567, _dispatch_stub_567@8))
-	GL_STUB(_dispatch_stub_568, _gloffset_PixelTexGenParameterivSGIS, _dispatch_stub_568@8)
+	GL_STUB(_dispatch_stub_568, _gloffset_GetPixelTexGenParameterfvSGIS, _dispatch_stub_568@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_568, _dispatch_stub_568@8))
-	GL_STUB(_dispatch_stub_569, _gloffset_SampleMaskSGIS, _dispatch_stub_569@8)
+	GL_STUB(_dispatch_stub_569, _gloffset_GetPixelTexGenParameterivSGIS, _dispatch_stub_569@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_569, _dispatch_stub_569@8))
-	GL_STUB(_dispatch_stub_570, _gloffset_SamplePatternSGIS, _dispatch_stub_570@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_570, _dispatch_stub_570@4))
+	GL_STUB(_dispatch_stub_570, _gloffset_PixelTexGenParameterfSGIS, _dispatch_stub_570@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_570, _dispatch_stub_570@8))
+	GL_STUB(_dispatch_stub_571, _gloffset_PixelTexGenParameterfvSGIS, _dispatch_stub_571@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_571, _dispatch_stub_571@8))
+	GL_STUB(_dispatch_stub_572, _gloffset_PixelTexGenParameteriSGIS, _dispatch_stub_572@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_572, _dispatch_stub_572@8))
+	GL_STUB(_dispatch_stub_573, _gloffset_PixelTexGenParameterivSGIS, _dispatch_stub_573@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_573, _dispatch_stub_573@8))
+	GL_STUB(_dispatch_stub_574, _gloffset_SampleMaskSGIS, _dispatch_stub_574@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_574, _dispatch_stub_574@8))
+	GL_STUB(_dispatch_stub_575, _gloffset_SamplePatternSGIS, _dispatch_stub_575@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_575, _dispatch_stub_575@4))
 	GL_STUB(ColorPointerEXT, _gloffset_ColorPointerEXT, ColorPointerEXT@20)
 	GL_STUB(EdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT, EdgeFlagPointerEXT@12)
 	GL_STUB(IndexPointerEXT, _gloffset_IndexPointerEXT, IndexPointerEXT@16)
@@ -740,10 +745,10 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(PointParameterfvEXT, _gloffset_PointParameterfvEXT, PointParameterfvEXT@8)
 	GL_STUB(LockArraysEXT, _gloffset_LockArraysEXT, LockArraysEXT@8)
 	GL_STUB(UnlockArraysEXT, _gloffset_UnlockArraysEXT, UnlockArraysEXT@0)
-	GL_STUB(_dispatch_stub_581, _gloffset_CullParameterdvEXT, _dispatch_stub_581@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_581, _dispatch_stub_581@8))
-	GL_STUB(_dispatch_stub_582, _gloffset_CullParameterfvEXT, _dispatch_stub_582@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_582, _dispatch_stub_582@8))
+	GL_STUB(_dispatch_stub_586, _gloffset_CullParameterdvEXT, _dispatch_stub_586@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_586, _dispatch_stub_586@8))
+	GL_STUB(_dispatch_stub_587, _gloffset_CullParameterfvEXT, _dispatch_stub_587@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_587, _dispatch_stub_587@8))
 	GL_STUB(SecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT, SecondaryColor3bEXT@12)
 	GL_STUB(SecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT, SecondaryColor3bvEXT@4)
 	GL_STUB(SecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT, SecondaryColor3dEXT@24)
@@ -768,8 +773,8 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(FogCoorddvEXT, _gloffset_FogCoorddvEXT, FogCoorddvEXT@4)
 	GL_STUB(FogCoordfEXT, _gloffset_FogCoordfEXT, FogCoordfEXT@4)
 	GL_STUB(FogCoordfvEXT, _gloffset_FogCoordfvEXT, FogCoordfvEXT@4)
-	GL_STUB(_dispatch_stub_607, _gloffset_PixelTexGenSGIX, _dispatch_stub_607@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_607, _dispatch_stub_607@4))
+	GL_STUB(_dispatch_stub_612, _gloffset_PixelTexGenSGIX, _dispatch_stub_612@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_612, _dispatch_stub_612@4))
 	GL_STUB(BlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT, BlendFuncSeparateEXT@16)
 	GL_STUB(FlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV, FlushVertexArrayRangeNV@0)
 	GL_STUB(VertexArrayRangeNV, _gloffset_VertexArrayRangeNV, VertexArrayRangeNV@8)
@@ -811,24 +816,24 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(WindowPos4ivMESA, _gloffset_WindowPos4ivMESA, WindowPos4ivMESA@4)
 	GL_STUB(WindowPos4sMESA, _gloffset_WindowPos4sMESA, WindowPos4sMESA@16)
 	GL_STUB(WindowPos4svMESA, _gloffset_WindowPos4svMESA, WindowPos4svMESA@4)
-	GL_STUB(_dispatch_stub_649, _gloffset_MultiModeDrawArraysIBM, _dispatch_stub_649@20)
-	HIDDEN(GL_PREFIX(_dispatch_stub_649, _dispatch_stub_649@20))
-	GL_STUB(_dispatch_stub_650, _gloffset_MultiModeDrawElementsIBM, _dispatch_stub_650@24)
-	HIDDEN(GL_PREFIX(_dispatch_stub_650, _dispatch_stub_650@24))
-	GL_STUB(_dispatch_stub_651, _gloffset_DeleteFencesNV, _dispatch_stub_651@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_651, _dispatch_stub_651@8))
-	GL_STUB(_dispatch_stub_652, _gloffset_FinishFenceNV, _dispatch_stub_652@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_652, _dispatch_stub_652@4))
-	GL_STUB(_dispatch_stub_653, _gloffset_GenFencesNV, _dispatch_stub_653@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_653, _dispatch_stub_653@8))
-	GL_STUB(_dispatch_stub_654, _gloffset_GetFenceivNV, _dispatch_stub_654@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_654, _dispatch_stub_654@12))
-	GL_STUB(_dispatch_stub_655, _gloffset_IsFenceNV, _dispatch_stub_655@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_655, _dispatch_stub_655@4))
-	GL_STUB(_dispatch_stub_656, _gloffset_SetFenceNV, _dispatch_stub_656@8)
+	GL_STUB(_dispatch_stub_654, _gloffset_MultiModeDrawArraysIBM, _dispatch_stub_654@20)
+	HIDDEN(GL_PREFIX(_dispatch_stub_654, _dispatch_stub_654@20))
+	GL_STUB(_dispatch_stub_655, _gloffset_MultiModeDrawElementsIBM, _dispatch_stub_655@24)
+	HIDDEN(GL_PREFIX(_dispatch_stub_655, _dispatch_stub_655@24))
+	GL_STUB(_dispatch_stub_656, _gloffset_DeleteFencesNV, _dispatch_stub_656@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_656, _dispatch_stub_656@8))
-	GL_STUB(_dispatch_stub_657, _gloffset_TestFenceNV, _dispatch_stub_657@4)
+	GL_STUB(_dispatch_stub_657, _gloffset_FinishFenceNV, _dispatch_stub_657@4)
 	HIDDEN(GL_PREFIX(_dispatch_stub_657, _dispatch_stub_657@4))
+	GL_STUB(_dispatch_stub_658, _gloffset_GenFencesNV, _dispatch_stub_658@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_658, _dispatch_stub_658@8))
+	GL_STUB(_dispatch_stub_659, _gloffset_GetFenceivNV, _dispatch_stub_659@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_659, _dispatch_stub_659@12))
+	GL_STUB(_dispatch_stub_660, _gloffset_IsFenceNV, _dispatch_stub_660@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_660, _dispatch_stub_660@4))
+	GL_STUB(_dispatch_stub_661, _gloffset_SetFenceNV, _dispatch_stub_661@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_661, _dispatch_stub_661@8))
+	GL_STUB(_dispatch_stub_662, _gloffset_TestFenceNV, _dispatch_stub_662@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_662, _dispatch_stub_662@4))
 	GL_STUB(AreProgramsResidentNV, _gloffset_AreProgramsResidentNV, AreProgramsResidentNV@12)
 	GL_STUB(BindProgramNV, _gloffset_BindProgramNV, BindProgramNV@8)
 	GL_STUB(DeleteProgramsNV, _gloffset_DeleteProgramsNV, DeleteProgramsNV@8)
@@ -909,26 +914,26 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(SetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI, SetFragmentShaderConstantATI@8)
 	GL_STUB(PointParameteriNV, _gloffset_PointParameteriNV, PointParameteriNV@8)
 	GL_STUB(PointParameterivNV, _gloffset_PointParameterivNV, PointParameterivNV@8)
-	GL_STUB(_dispatch_stub_738, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_738@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_738, _dispatch_stub_738@4))
-	GL_STUB(_dispatch_stub_739, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_739@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_739, _dispatch_stub_739@4))
-	GL_STUB(_dispatch_stub_740, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_740@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_740, _dispatch_stub_740@8))
-	GL_STUB(_dispatch_stub_741, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_741@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_741, _dispatch_stub_741@8))
-	GL_STUB(_dispatch_stub_742, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_742@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_742, _dispatch_stub_742@4))
+	GL_STUB(_dispatch_stub_743, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_743@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_743, _dispatch_stub_743@4))
+	GL_STUB(_dispatch_stub_744, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_744@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_744, _dispatch_stub_744@4))
+	GL_STUB(_dispatch_stub_745, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_745@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_745, _dispatch_stub_745@8))
+	GL_STUB(_dispatch_stub_746, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_746@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_746, _dispatch_stub_746@8))
+	GL_STUB(_dispatch_stub_747, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_747@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_747, _dispatch_stub_747@4))
 	GL_STUB(GetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV, GetProgramNamedParameterdvNV@16)
 	GL_STUB(GetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV, GetProgramNamedParameterfvNV@16)
 	GL_STUB(ProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV, ProgramNamedParameter4dNV@44)
 	GL_STUB(ProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV, ProgramNamedParameter4dvNV@16)
 	GL_STUB(ProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV, ProgramNamedParameter4fNV@28)
 	GL_STUB(ProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV, ProgramNamedParameter4fvNV@16)
-	GL_STUB(_dispatch_stub_749, _gloffset_DepthBoundsEXT, _dispatch_stub_749@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_749, _dispatch_stub_749@16))
-	GL_STUB(_dispatch_stub_750, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_750@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_750, _dispatch_stub_750@8))
+	GL_STUB(_dispatch_stub_754, _gloffset_DepthBoundsEXT, _dispatch_stub_754@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_754, _dispatch_stub_754@16))
+	GL_STUB(_dispatch_stub_755, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_755@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_755, _dispatch_stub_755@8))
 	GL_STUB(BindFramebufferEXT, _gloffset_BindFramebufferEXT, BindFramebufferEXT@8)
 	GL_STUB(BindRenderbufferEXT, _gloffset_BindRenderbufferEXT, BindRenderbufferEXT@8)
 	GL_STUB(CheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT, CheckFramebufferStatusEXT@4)
@@ -946,19 +951,20 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(IsFramebufferEXT, _gloffset_IsFramebufferEXT, IsFramebufferEXT@4)
 	GL_STUB(IsRenderbufferEXT, _gloffset_IsRenderbufferEXT, IsRenderbufferEXT@4)
 	GL_STUB(RenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT, RenderbufferStorageEXT@16)
-	GL_STUB(_dispatch_stub_768, _gloffset_BlitFramebufferEXT, _dispatch_stub_768@40)
-	HIDDEN(GL_PREFIX(_dispatch_stub_768, _dispatch_stub_768@40))
+	GL_STUB(_dispatch_stub_773, _gloffset_BlitFramebufferEXT, _dispatch_stub_773@40)
+	HIDDEN(GL_PREFIX(_dispatch_stub_773, _dispatch_stub_773@40))
 	GL_STUB(FramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20)
-	GL_STUB(_dispatch_stub_770, _gloffset_StencilFuncSeparateATI, _dispatch_stub_770@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_770, _dispatch_stub_770@16))
-	GL_STUB(_dispatch_stub_771, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_771@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_771, _dispatch_stub_771@16))
-	GL_STUB(_dispatch_stub_772, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_772@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@16))
-	GL_STUB(_dispatch_stub_773, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_773@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_773, _dispatch_stub_773@12))
-	GL_STUB(_dispatch_stub_774, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_774@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_774, _dispatch_stub_774@12))
+	GL_STUB(ProvokingVertexEXT, _gloffset_ProvokingVertexEXT, ProvokingVertexEXT@4)
+	GL_STUB(_dispatch_stub_776, _gloffset_StencilFuncSeparateATI, _dispatch_stub_776@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_776, _dispatch_stub_776@16))
+	GL_STUB(_dispatch_stub_777, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_777@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_777, _dispatch_stub_777@16))
+	GL_STUB(_dispatch_stub_778, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_778@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_778, _dispatch_stub_778@16))
+	GL_STUB(_dispatch_stub_779, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_779@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_779, _dispatch_stub_779@12))
+	GL_STUB(_dispatch_stub_780, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_780@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_780, _dispatch_stub_780@12))
 	GL_STUB_ALIAS(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4, ArrayElement, ArrayElement@4)
 	GL_STUB_ALIAS(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8, BindTexture, BindTexture@8)
 	GL_STUB_ALIAS(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12, DrawArrays, DrawArrays@12)