Merge branch 'origin' into glsl-compiler-1

Conflicts:

	src/mesa/main/state.c
	src/mesa/shader/program.c
	src/mesa/shader/program.h
	src/mesa/shader/programopt.c
	src/mesa/shader/slang/slang_execute.c
	src/mesa/sources
	src/mesa/swrast/s_arbshader.c
	src/mesa/swrast/s_context.c
	src/mesa/swrast/s_span.c
	src/mesa/swrast/s_zoom.c
	src/mesa/tnl/t_context.c
	src/mesa/tnl/t_save_api.c
	src/mesa/tnl/t_vb_arbprogram.c
	src/mesa/tnl/t_vp_build.c
	src/mesa/tnl/t_vtx_eval.c

23 files changed, 556 insertions, 5218 deletions

diff --git a/src/mesa/tnl/sources b/src/mesa/tnl/sources
index e01f55dbaf..a0888be11d 100644
--- a/src/mesa/tnl/sources
+++ b/src/mesa/tnl/sources
@@ -1,12 +1,7 @@
 # List of source files in this directory used for X.org xserver build
 MESA_TNL_SOURCES = \
-t_array_api.c \
-t_array_import.c \
 t_context.c \
 t_pipeline.c \
-t_save_api.c \
-t_save_loopback.c \
-t_save_playback.c \
 t_vb_arbprogram.c \
 t_vb_arbprogram_sse.c \
 t_vb_arbshader.c \
@@ -23,24 +18,17 @@ t_vb_vertex.c \
 t_vertex.c \
 t_vertex_generic.c \
 t_vertex_sse.c \
-t_vp_build.c \
-t_vtx_api.c \
-t_vtx_eval.c \
-t_vtx_exec.c \
-t_vtx_generic.c \
-t_vtx_x86.c
+t_vp_build.c 
 
 MESA_TNL_HEADERS = \
 t_array_api.h \
 t_array_import.h \
 t_context.h \
 t_pipeline.h \
-t_save_api.h \
 t_vb_arbprogram.h \
 t_vb_cliptmp.h \
 t_vb_lighttmp.h \
 t_vb_rendertmp.h \
 t_vertex.h \
 t_vp_build.h \
-t_vtx_api.h \
 tnl.h
diff --git a/src/mesa/tnl/t_array_api.c b/src/mesa/tnl/t_array_api.c
deleted file mode 100644
index 6826cf7e24..0000000000
--- a/src/mesa/tnl/t_array_api.c
+++ /dev/null
@@ -1,432 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.2
- *
- * Copyright (C) 1999-2006  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.
- */
-
-/**
- * \file t_array_api.c
- * \brief Vertex array API functions (glDrawArrays, etc)
- * \author Keith Whitwell
- */
-
-#include "glheader.h"
-#include "api_validate.h"
-#include "context.h"
-#include "imports.h"
-#include "macros.h"
-#include "mtypes.h"
-#include "state.h"
-
-#include "array_cache/acache.h"
-
-#include "t_array_api.h"
-#include "t_array_import.h"
-#include "t_save_api.h"
-#include "t_context.h"
-#include "t_pipeline.h"
-#include "dispatch.h"
-
-static void fallback_drawarrays( GLcontext *ctx, GLenum mode, GLint start,
-				 GLsizei count )
-{
-   GLint i;
-
-   assert(!ctx->CompileFlag);
-   assert(ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END);
-
-   CALL_Begin(GET_DISPATCH(), (mode));
-   for (i = 0; i < count; i++) 
-       CALL_ArrayElement(GET_DISPATCH(), ( start + i ));
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-static void fallback_drawelements( GLcontext *ctx, GLenum mode, GLsizei count,
-				   const GLuint *indices)
-{
-   GLint i;
-
-   assert(!ctx->CompileFlag);
-   assert(ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END);
-
-   /* Here, indices will already reflect the buffer object if active */
-
-   CALL_Begin(GET_DISPATCH(), (mode));
-   for (i = 0 ; i < count ; i++) {
-      CALL_ArrayElement(GET_DISPATCH(), ( indices[i] ));
-   }
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-/* Note this function no longer takes a 'start' value, the range is
- * assumed to start at zero.  The old trick of subtracting 'start'
- * from each index won't work if the indices are not in writeable
- * memory.
- */
-static void _tnl_draw_range_elements( GLcontext *ctx, GLenum mode,
-				      GLuint max_index,
-				      GLsizei index_count, GLuint *indices )
-
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_prim prim;
-   FLUSH_CURRENT( ctx, 0 );
-   
-   _tnl_vb_bind_arrays( ctx, 0, max_index );
-
-   tnl->vb.Primitive = &prim;
-   tnl->vb.Primitive[0].mode = mode | PRIM_BEGIN | PRIM_END;
-   tnl->vb.Primitive[0].start = 0;
-   tnl->vb.Primitive[0].count = index_count;
-   tnl->vb.PrimitiveCount = 1;
-
-   tnl->vb.Elts = (GLuint *)indices;
-
-   tnl->Driver.RunPipeline( ctx );
-}
-
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   /* It's tempting to get rid of this threshold value because we take
-    * very different paths if 'count' is less than or greater than 'thresh'.
-    * I've found/fixed at least one bug which only occured for particular
-    * array sizes.  Also, several conformance tests use very short arrays
-    * which means the long-array path doesn't get tested.  -Brian
-    */
-   GLuint thresh = (ctx->Driver.NeedFlush & FLUSH_STORED_VERTICES) ? 30 : 10;
-   
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawArrays %d %d\n", start, count); 
-   
-   /* Check arguments, etc.
-    */
-   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-      return;
-
-   assert(!ctx->CompileFlag);
-
-   if (!ctx->Array.LockCount && (GLuint) count < thresh) {
-      /* Small primitives: attempt to share a vb (at the expense of
-       * using the immediate interface).
-      */
-      fallback_drawarrays( ctx, mode, start, count );
-   } 
-   else if (start >= (GLint) ctx->Array.LockFirst &&
-	    start + count <= (GLint)(ctx->Array.LockFirst + ctx->Array.LockCount)) {
-      
-      struct tnl_prim prim;
-
-      /* Locked primitives which can fit in a single vertex buffer:
-       */
-      FLUSH_CURRENT( ctx, 0 );
-
-      /* Locked drawarrays.  Reuse any previously transformed data.
-       */
-      _tnl_vb_bind_arrays( ctx, ctx->Array.LockFirst, 
-			   ctx->Array.LockFirst + ctx->Array.LockCount );
-
-      tnl->vb.Primitive = &prim;
-      tnl->vb.Primitive[0].mode = mode | PRIM_BEGIN | PRIM_END;
-      tnl->vb.Primitive[0].start = start;
-      tnl->vb.Primitive[0].count = count;
-      tnl->vb.PrimitiveCount = 1;
-
-      tnl->Driver.RunPipeline( ctx );
-   } 
-   else {
-      int bufsz = 256;		/* Use a small buffer for cache goodness */
-      int j, nr;
-      int minimum, modulo, skip;
-
-      /* Large primitives requiring decomposition to multiple vertex
-       * buffers:
-       */
-      switch (mode) {
-      case GL_POINTS:
-	 minimum = 0;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_LINES:
-	 minimum = 1;
-	 modulo = 2;
-	 skip = 1;
-	 break;
-      case GL_LINE_STRIP:
-	 minimum = 1;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_TRIANGLES:
-	 minimum = 2;
-	 modulo = 3;
-	 skip = 2;
-	 break;
-      case GL_TRIANGLE_STRIP:
-	 minimum = 2;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_QUADS:
-	 minimum = 3;
-	 modulo = 4;
-	 skip = 3;
-	 break;
-      case GL_QUAD_STRIP:
-	 minimum = 3;
-	 modulo = 2;
-	 skip = 0;
-	 break;
-      case GL_LINE_LOOP:
-      case GL_TRIANGLE_FAN:
-      case GL_POLYGON:
-      default:
-	 /* Primitives requiring a copied vertex (fan-like primitives)
-	  * must use the slow path if they cannot fit in a single
-	  * vertex buffer.  
-	  */
-	 if (count <= (GLint) ctx->Const.MaxArrayLockSize) {
-	    bufsz = ctx->Const.MaxArrayLockSize;
-	    minimum = 0;
-	    modulo = 1;
-	    skip = 0;
-	 }
-	 else {
-	    fallback_drawarrays( ctx, mode, start, count );
-	    return;
-	 }
-      }
-
-      FLUSH_CURRENT( ctx, 0 );
-
-      bufsz -= bufsz % modulo;
-      bufsz -= minimum;
-      count += start;
-
-      for (j = start + minimum ; j < count ; j += nr + skip ) {
-
-	 struct tnl_prim prim;
-
-	 nr = MIN2( bufsz, count - j );
-
-         /* XXX is the last parameter a count or index into the array??? */
-	 _tnl_vb_bind_arrays( ctx, j - minimum, j + nr );
-
-	 tnl->vb.Primitive = &prim;
-	 tnl->vb.Primitive[0].mode = mode;
-
-	 if (j == start + minimum)
-	    tnl->vb.Primitive[0].mode |= PRIM_BEGIN;
-
-	 if (j + nr + skip >= count)
-	    tnl->vb.Primitive[0].mode |= PRIM_END;
-
-	 tnl->vb.Primitive[0].start = 0;
-	 tnl->vb.Primitive[0].count = nr + minimum;
-	 tnl->vb.PrimitiveCount = 1;
-
-	 tnl->Driver.RunPipeline( ctx );
-      }
-   }
-}
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawRangeElements(GLenum mode,
-		       GLuint start, GLuint end,
-		       GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint *ui_indices;
-
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawRangeElements %d %d %d\n", start, end, count); 
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* use indices in the buffer object */
-      if (!ctx->Array.ElementArrayBufferObj->Data) {
-         _mesa_warning(ctx,
-                       "DrawRangeElements with empty vertex elements buffer!");
-         return;
-      }
-      /* actual address is the sum of pointers */
-      indices = (const GLvoid *)
-         ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data,
-                      (const GLubyte *) indices);
-   }
-
-   /* Check arguments, etc.
-    */
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count,
-                                          type, indices ))
-      return;
-
-   ui_indices = (GLuint *)_ac_import_elements( ctx, GL_UNSIGNED_INT,
-					       count, type, indices );
-
-#ifdef DEBUG
-   /* check that array indices really fall inside [start, end] range */
-   {
-      GLuint i;
-      for (i = 0; i < count; i++) {
-         if (ui_indices[i] < start || ui_indices[i] > end) {
-            _mesa_warning(ctx, "Invalid array index in "
-                          "glDrawRangeElements(index=%u)", ui_indices[i]);
-         }
-      }
-   }
-#endif
-
-   assert(!ctx->CompileFlag);
-
-   if (ctx->Array.LockCount) {
-      /* Are the arrays already locked?  If so we currently have to look
-       * at the whole locked range.
-       */
-
-      if (start == 0 && ctx->Array.LockFirst == 0 && 
-	  end < (ctx->Array.LockFirst + ctx->Array.LockCount))
-	 _tnl_draw_range_elements( ctx, mode,
-				   ctx->Array.LockCount,
-				   count, ui_indices );
-      else {
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-      }
-   }
-   else if (start == 0 && end < ctx->Const.MaxArrayLockSize) {
-      /* The arrays aren't locked but we can still fit them inside a
-       * single vertexbuffer.
-       */
-      _tnl_draw_range_elements( ctx, mode, end + 1, count, ui_indices );
-   }
-   else {
-      /* Range is too big to optimize:
-       */
-      fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-}
-
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawElements(GLenum mode, GLsizei count, GLenum type,
-		  const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint *ui_indices;
-
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawElements %d\n", count); 
-
-   /* Check arguments, etc. */
-   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-      return;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* actual address is the sum of pointers */
-      indices = (const GLvoid *)
-         ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data,
-                      (const GLubyte *) indices);
-   }
-
-   ui_indices = (GLuint *)_ac_import_elements( ctx, GL_UNSIGNED_INT,
-					       count, type, indices );
-
-   assert(!ctx->CompileFlag);
-
-   if (ctx->Array.LockCount) {
-      if (ctx->Array.LockFirst == 0)
-	 _tnl_draw_range_elements( ctx, mode,
-				   ctx->Array.LockCount,
-				   count, ui_indices );
-      else
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-   else {
-      /* Scan the index list and see if we can use the locked path anyway.
-       */
-      GLuint max_elt = 0;
-      GLint i;
-
-      for (i = 0 ; i < count ; i++)
-	 if (ui_indices[i] > max_elt)
-            max_elt = ui_indices[i];
-
-      if (max_elt < ctx->Const.MaxArrayLockSize && /* can we use it? */
-	  max_elt < (GLuint) count) 	           /* do we want to use it? */
-	 _tnl_draw_range_elements( ctx, mode, max_elt+1, count, ui_indices );
-      else
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-}
-
-
-/**
- * Initialize context's vertex array fields.  Called during T 'n L context
- * creation.
- */
-void _tnl_array_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_vertex_arrays *tmp = &tnl->array_inputs;
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-   GLuint i;
-
-   vfmt->DrawArrays = _tnl_DrawArrays;
-   vfmt->DrawElements = _tnl_DrawElements;
-   vfmt->DrawRangeElements = _tnl_DrawRangeElements;
-
-   /* Setup vector pointers that will be used to bind arrays to VB's.
-    */
-   _mesa_vector4f_init( &tmp->Obj, 0, NULL);
-   _mesa_vector4f_init( &tmp->Normal, 0, NULL);   
-   _mesa_vector4f_init( &tmp->FogCoord, 0, NULL);
-   _mesa_vector4f_init( &tmp->Index, 0, NULL);
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++)
-      _mesa_vector4f_init( &tmp->TexCoord[i], 0, NULL);
-}
-
-
-/**
- * Destroy the context's vertex array stuff.
- * Called during T 'n L context destruction.
- */
-void _tnl_array_destroy( GLcontext *ctx )
-{
-   (void) ctx;
-}
diff --git a/src/mesa/tnl/t_array_api.h b/src/mesa/tnl/t_array_api.h
deleted file mode 100644
index 61d1f696b0..0000000000
--- a/src/mesa/tnl/t_array_api.h
+++ /dev/null
@@ -1,46 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2001  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.
- */
-
-#ifndef _T_VARRAY_H
-#define _T_VARRAY_H
-
-#include "mtypes.h"
-#include "t_context.h"
-
-
-extern void GLAPIENTRY _tnl_DrawArrays(GLenum mode, GLint first, GLsizei count);
-
-extern void GLAPIENTRY _tnl_DrawElements(GLenum mode, GLsizei count, GLenum type,
-			      const GLvoid *indices);
-
-extern void GLAPIENTRY _tnl_DrawRangeElements(GLenum mode, GLuint start,
-				   GLuint end, GLsizei count, GLenum type,
-				   const GLvoid *indices);
-
-
-extern void _tnl_array_init( GLcontext *ctx );
-extern void _tnl_array_destroy( GLcontext *ctx );
-
-#endif
diff --git a/src/mesa/tnl/t_array_import.c b/src/mesa/tnl/t_array_import.c
deleted file mode 100644
index 13c5689ceb..0000000000
--- a/src/mesa/tnl/t_array_import.c
+++ /dev/null
@@ -1,376 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.1
- *
- * Copyright (C) 1999-2006  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.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "imports.h"
-#include "state.h"
-#include "mtypes.h"
-
-#include "array_cache/acache.h"
-
-#include "t_array_import.h"
-#include "t_context.h"
-
-
-/**
- * XXX writable and stride are always false in these functions...
- */
-static void _tnl_import_vertex( GLcontext *ctx,
-				GLboolean writable,
-				GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_vertex(ctx,
-			   GL_FLOAT,
-			   stride ? 4*sizeof(GLfloat) : 0,
-			   0,
-			   writable,
-			   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Obj.data = (GLfloat (*)[4]) data;
-   inputs->Obj.start = (GLfloat *) data;
-   inputs->Obj.stride = tmp->StrideB;
-   inputs->Obj.size = tmp->Size;
-}
-
-static void _tnl_import_normal( GLcontext *ctx,
-				GLboolean writable,
-				GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_normal(ctx, GL_FLOAT,
-			   stride ? 3*sizeof(GLfloat) : 0, writable,
-			   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Normal.data = (GLfloat (*)[4]) data;
-   inputs->Normal.start = (GLfloat *) data;
-   inputs->Normal.stride = tmp->StrideB;
-   inputs->Normal.size = 3;
-}
-
-
-static void _tnl_import_color( GLcontext *ctx,
-			       GLboolean writable,
-			       GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_color(ctx,
-			  GL_FLOAT,
-			  stride ? 4*sizeof(GLfloat) : 0,
-			  4,
-			  writable,
-			  &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Color.data = (GLfloat (*)[4]) data;
-   inputs->Color.start = (GLfloat *) data;
-   inputs->Color.stride = tmp->StrideB;
-   inputs->Color.size = tmp->Size;
-}
-
-
-static void _tnl_import_secondarycolor( GLcontext *ctx,
-					GLboolean writable,
-					GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_secondarycolor(ctx, 
-				   GL_FLOAT,
-				   stride ? 4*sizeof(GLfloat) : 0,
-				   4,
-				   writable,
-				   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->SecondaryColor.data = (GLfloat (*)[4]) data;
-   inputs->SecondaryColor.start = (GLfloat *) data;
-   inputs->SecondaryColor.stride = tmp->StrideB;
-   inputs->SecondaryColor.size = tmp->Size;
-}
-
-static void _tnl_import_fogcoord( GLcontext *ctx,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_fogcoord(ctx, GL_FLOAT,
-			     stride ? sizeof(GLfloat) : 0, writable,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->FogCoord.data = (GLfloat (*)[4]) data;
-   inputs->FogCoord.start = (GLfloat *) data;
-   inputs->FogCoord.stride = tmp->StrideB;
-}
-
-static void _tnl_import_index( GLcontext *ctx,
-			       GLboolean writable,
-			       GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_index(ctx, GL_FLOAT,
-			  stride ? sizeof(GLfloat) : 0, writable,
-			  &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Index.data = (GLfloat (*)[4]) data;
-   inputs->Index.start = (GLfloat *) data;
-   inputs->Index.stride = tmp->StrideB;
-}
-
-
-static void _tnl_import_texcoord( GLcontext *ctx,
-				  GLuint unit,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_texcoord(ctx, unit, GL_FLOAT,
-			     stride ? 4 * sizeof(GLfloat) : 0,
-			     0,
-			     writable,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->TexCoord[unit].data = (GLfloat (*)[4]) data;
-   inputs->TexCoord[unit].start = (GLfloat *) data;
-   inputs->TexCoord[unit].stride = tmp->StrideB;
-   inputs->TexCoord[unit].size = tmp->Size;
-}
-
-
-static void _tnl_import_edgeflag( GLcontext *ctx,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-   (void) writable; (void) stride;
-
-   tmp = _ac_import_edgeflag(ctx, GL_UNSIGNED_BYTE,
-			     sizeof(GLubyte),
-			     0,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->EdgeFlag = (GLubyte *) data;
-}
-
-
-
-static void _tnl_import_attrib( GLcontext *ctx,
-                                GLuint index,
-                                GLboolean writable,
-                                GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   ASSERT(index < MAX_VERTEX_PROGRAM_ATTRIBS);
-
-   tmp = _ac_import_attrib(ctx, index, GL_FLOAT,
-                           stride ? 4 * sizeof(GLfloat) : 0,
-                           4,  /* want GLfloat[4] */
-                           writable,
-                           &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Attribs[index].data = (GLfloat (*)[4]) data;
-   inputs->Attribs[index].start = (GLfloat *) data;
-   inputs->Attribs[index].stride = tmp->StrideB;
-   inputs->Attribs[index].size = tmp->Size;
-}
-
-
-static void _tnl_constant_attrib( TNLcontext *tnl,
-				  struct tnl_vertex_arrays *tmp,
-				  GLuint i )
-{
-   tmp->Attribs[i].count = 1;
-   tmp->Attribs[i].data = (GLfloat (*)[4]) tnl->vtx.current[i];
-   tmp->Attribs[i].start = tnl->vtx.current[i];
-   tmp->Attribs[i].size = 4; 
-   tmp->Attribs[i].stride = 0;
-   tnl->vb.AttribPtr[i] = &tmp->Attribs[i];
-}
-
-
-
-void _tnl_vb_bind_arrays( GLcontext *ctx, GLint start, GLint end)
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->array_inputs;
-   const struct gl_vertex_program *program
-      = ctx->VertexProgram._Enabled ? ctx->VertexProgram.Current : NULL;
-   GLuint i, index;
-
-   VB->Count = end - start;
-   VB->Elts = NULL;
-
-   _ac_import_range( ctx, start, end );
-
-   /* Note that the generic attribute arrays are treated differently
-    * depending on whether an NV or ARB vertex program is enabled
-    * (corresponding to aliasing vs. non-aliasing behaviour).
-    * Generic array 0 always aliases vertex position.
-    */
-   for (index = 0; index < VERT_ATTRIB_MAX; index++) {
-      if (ctx->VertexProgram._Enabled
-          && (program->IsNVProgram || index == 0)
-          && ctx->Array.ArrayObj->VertexAttrib[index].Enabled) {
-         /* Use generic attribute array.  If an NV vertex program is active,
-          * the generic arrays override the conventional attributes.
-          * Otherwise, if an ARB vertex program is active, we'll import the
-          * generic attributes without aliasing over conventional attribs
-          * (see below).
-          */
-         _tnl_import_attrib( ctx, index, GL_FALSE, GL_TRUE );
-         VB->AttribPtr[index] = &tmp->Attribs[index];
-      }
-      /* use conventional arrays... */
-      else if (index == VERT_ATTRIB_POS) {
-	 _tnl_import_vertex( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Obj.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_POS] = &tmp->Obj;
-      }
-      else if (index == VERT_ATTRIB_NORMAL) {
-	 _tnl_import_normal( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Normal.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_NORMAL] = &tmp->Normal;
-      }
-      else if (index == VERT_ATTRIB_COLOR0) {
-	 _tnl_import_color( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Color.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &tmp->Color;
-      }
-      else if (index == VERT_ATTRIB_COLOR1) {
-	 _tnl_import_secondarycolor( ctx, GL_FALSE, GL_FALSE );
-	 tmp->SecondaryColor.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_COLOR1] = &tmp->SecondaryColor;
-      }
-      else if (index == VERT_ATTRIB_FOG) {
-	 _tnl_import_fogcoord( ctx, GL_FALSE, GL_FALSE );
-	 tmp->FogCoord.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_FOG] = &tmp->FogCoord;
-      }
-      else if (index == VERT_ATTRIB_COLOR_INDEX) {
-         _tnl_import_index( ctx, GL_FALSE, GL_FALSE );
-         tmp->Index.count = VB->Count;
-         VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX] = &tmp->Index;
-      }
-      else if (index >= VERT_ATTRIB_TEX0 && index <= VERT_ATTRIB_TEX7) {
-	 i = index - VERT_ATTRIB_TEX0;
-	 _tnl_import_texcoord( ctx, i, GL_FALSE, GL_FALSE );
-	 tmp->TexCoord[i].count = VB->Count;
-	 VB->AttribPtr[index] = &tmp->TexCoord[i];
-      }
-      else if (index >= VERT_ATTRIB_GENERIC1 &&
-               index <= VERT_ATTRIB_GENERIC15) {
-         const GLuint arrayIndex = index - VERT_ATTRIB_GENERIC0;
-         if (program && !program->IsNVProgram &&
-	     ctx->Array.ArrayObj->VertexAttrib[arrayIndex].Enabled) {
-            /* GL_ARB_vertex_program: bind a generic attribute array */
-            _tnl_import_attrib(ctx, arrayIndex, GL_FALSE, GL_TRUE);
-            VB->AttribPtr[index] = &tmp->Attribs[arrayIndex];
-         }
-	 else {
-            _tnl_constant_attrib(tnl, tmp, index);
-         }
-      }
-      else {
-	 _tnl_constant_attrib(tnl, tmp, index);
-      }
-      assert(VB->AttribPtr[index]);
-      assert(VB->AttribPtr[index]->size);
-   }
-
-   /* odd-ball vertex attributes */
-   {
-      _tnl_import_edgeflag( ctx, GL_TRUE, sizeof(GLboolean) );
-      VB->EdgeFlag = (GLboolean *) tmp->EdgeFlag;
-   }
-
-   /* These are constant & could be precalculated:
-    */
-   for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-      _tnl_constant_attrib(tnl, tmp, i);
-   }
-
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
diff --git a/src/mesa/tnl/t_array_import.h b/src/mesa/tnl/t_array_import.h
deleted file mode 100644
index 39b77641d5..0000000000
--- a/src/mesa/tnl/t_array_import.h
+++ /dev/null
@@ -1,36 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2001  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.
- */
-
-#ifndef _T_ARRAY_IMPORT_H
-#define _T_ARRAY_IMPORT_H
-
-#include "mtypes.h"
-#include "t_context.h"
-
-extern void _tnl_vb_bind_arrays( GLcontext *ctx, GLint start, GLint end );
-
-extern void _tnl_array_import_init( GLcontext *ctx );
-
-#endif
diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index 3887f18213..3b2f91acba 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -26,38 +26,19 @@
  */
 
 
-#include "api_arrayelt.h"
 #include "glheader.h"
 #include "imports.h"
 #include "context.h"
 #include "macros.h"
 #include "mtypes.h"
-#include "dlist.h"
 #include "light.h"
-#include "vtxfmt.h"
 
 #include "tnl.h"
-#include "t_array_api.h"
 #include "t_context.h"
 #include "t_pipeline.h"
-#include "t_save_api.h"
 #include "t_vp_build.h"
-#include "t_vtx_api.h"
-
-
-
-static void
-install_driver_callbacks( GLcontext *ctx )
-{
-   ctx->Driver.NewList = _tnl_NewList;
-   ctx->Driver.EndList = _tnl_EndList;
-   ctx->Driver.FlushVertices = _tnl_FlushVertices;
-   ctx->Driver.SaveFlushVertices = _tnl_SaveFlushVertices;
-   ctx->Driver.BeginCallList = _tnl_BeginCallList;
-   ctx->Driver.EndCallList = _tnl_EndCallList;
-}
-
 
+#include "vbo/vbo.h"
 
 GLboolean
 _tnl_CreateContext( GLcontext *ctx )
@@ -72,20 +53,13 @@ _tnl_CreateContext( GLcontext *ctx )
       return GL_FALSE;
    }
 
-   if (_mesa_getenv("MESA_CODEGEN"))
-      tnl->AllowCodegen = GL_TRUE;
-
    /* Initialize the VB.
     */
    tnl->vb.Size = ctx->Const.MaxArrayLockSize + MAX_CLIPPED_VERTICES;
 
 
-   /* Initialize tnl state and tnl->vtxfmt.
+   /* Initialize tnl state.
     */
-   _tnl_save_init( ctx );
-   _tnl_array_init( ctx );
-   _tnl_vtx_init( ctx );
-
    if (ctx->VertexProgram._MaintainTnlProgram) {
       _tnl_ProgramCacheInit( ctx );
       _tnl_install_pipeline( ctx, _tnl_vp_pipeline );
@@ -93,34 +67,18 @@ _tnl_CreateContext( GLcontext *ctx )
       _tnl_install_pipeline( ctx, _tnl_default_pipeline );
    }
 
-   /* Initialize the arrayelt helper
-    */
-   if (!_ae_create_context( ctx ))
-      return GL_FALSE;
-
-
    tnl->NeedNdcCoords = GL_TRUE;
-   tnl->LoopbackDListCassettes = GL_FALSE;
-   tnl->CalcDListNormalLengths = GL_TRUE;
    tnl->AllowVertexFog = GL_TRUE;
    tnl->AllowPixelFog = GL_TRUE;
 
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
-
    /* Set a few default values in the driver struct.
     */
-   install_driver_callbacks(ctx);
-   ctx->Driver.NeedFlush = 0;
-   ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
-
    tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
    tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
    tnl->Driver.NotifyMaterialChange = _mesa_validate_all_lighting_tables;
 
+   tnl->nr_blocks = 0;
+
    return GL_TRUE;
 }
 
@@ -130,11 +88,7 @@ _tnl_DestroyContext( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
 
-   _tnl_array_destroy( ctx );
-   _tnl_vtx_destroy( ctx );
-   _tnl_save_destroy( ctx );
    _tnl_destroy_pipeline( ctx );
-   _ae_destroy_context( ctx );
 
    if (ctx->VertexProgram._MaintainTnlProgram)
       _tnl_ProgramCacheDestroy( ctx );
@@ -155,10 +109,7 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
          || !tnl->AllowPixelFog;
    }
 
-   _ae_invalidate_state(ctx, new_state);
-
    tnl->pipeline.new_state |= new_state;
-   tnl->vtx.eval.new_state |= new_state;
 
    /* Calculate tnl->render_inputs:
     */
@@ -210,36 +161,21 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
 
 
 void
-_tnl_wakeup_exec( GLcontext *ctx )
+_tnl_wakeup( GLcontext *ctx )
 {
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   install_driver_callbacks(ctx);
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
    /* Assume we haven't been getting state updates either:
     */
    _tnl_InvalidateState( ctx, ~0 );
 
+#if 0
    if (ctx->Light.ColorMaterialEnabled) {
       _mesa_update_color_material( ctx, 
 				   ctx->Current.Attrib[VERT_ATTRIB_COLOR0] );
    }
+#endif
 }
 
 
-void
-_tnl_wakeup_save_exec( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   _tnl_wakeup_exec( ctx );
-   _mesa_install_save_vtxfmt( ctx, &tnl->save_vtxfmt );
-}
 
 
 /**
@@ -255,27 +191,6 @@ _tnl_need_projected_coords( GLcontext *ctx, GLboolean mode )
 }
 
 void
-_tnl_need_dlist_loopback( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->LoopbackDListCassettes = mode;
-}
-
-void
-_tnl_need_dlist_norm_lengths( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->CalcDListNormalLengths = mode;
-}
-
-void
-_tnl_isolate_materials( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->IsolateMaterials = mode;
-}
-
-void
 _tnl_allow_vertex_fog( GLcontext *ctx, GLboolean value )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h
index ea2a6bed54..733b3a5fe1 100644
--- a/src/mesa/tnl/t_context.h
+++ b/src/mesa/tnl/t_context.h
@@ -43,29 +43,6 @@
  * stages to the vertex_buffer TNLcontext::vb, where the vertex data
  * is stored. The last stage in the pipeline is the rasterizer.
  *
- * The initial vertex_buffer data may either come from an ::immediate
- * structure or client vertex_arrays or display lists:
- *
- *
- * - The ::immediate structure records all the GL commands issued between
- * glBegin and glEnd.  \n
- * The structure accumulates data, until it is either full or it is
- * flushed (usually by a state change). Before starting then the pipeline,
- * the collected vertex data in ::immediate has to be pushed into
- * TNLcontext::vb.
- * This happens in ::_tnl_vb_bind_immediate. The pipeline is then run by
- * calling tnl_device_driver::RunPipeline = ::_tnl_run_pipeline, which
- * is stored in TNLcontext::Driver.   \n
- * An ::immediate does (for performance reasons) usually not finish with a
- * glEnd, and hence it also does not need to start with a glBegin.
- * This means that the last vertices of one ::immediate may need to be
- * saved for the next one.
- *
- *
- * - NOT SURE ABOUT THIS: The vertex_arrays structure is used to handle
- * glDrawArrays etc.  \n
- * Here, the data of the vertex_arrays is copied by ::_tnl_vb_bind_arrays
- * into TNLcontext::vb, so that the pipeline can be started.
  */
 
 
@@ -79,6 +56,7 @@
 #include "math/m_vector.h"
 #include "math/m_xform.h"
 
+#include "vbo/vbo.h"
 
 #define MAX_PIPELINE_STAGES     30
 
@@ -106,7 +84,7 @@ enum {
 	_TNL_ATTRIB_COLOR1 = 4,
 	_TNL_ATTRIB_FOG = 5,
 	_TNL_ATTRIB_COLOR_INDEX = 6,
-	_TNL_ATTRIB_SEVEN = 7,
+	_TNL_ATTRIB_EDGEFLAG = 7,
 	_TNL_ATTRIB_TEX0 = 8,
 	_TNL_ATTRIB_TEX1 = 9,
 	_TNL_ATTRIB_TEX2 = 10,
@@ -115,7 +93,8 @@ enum {
 	_TNL_ATTRIB_TEX5 = 13,
 	_TNL_ATTRIB_TEX6 = 14,
 	_TNL_ATTRIB_TEX7 = 15,
-	_TNL_ATTRIB_GENERIC0 = 16,
+
+	_TNL_ATTRIB_GENERIC0 = 16, /* doesn't really exist! */
 	_TNL_ATTRIB_GENERIC1 = 17,
 	_TNL_ATTRIB_GENERIC2 = 18,
 	_TNL_ATTRIB_GENERIC3 = 19,
@@ -131,21 +110,36 @@ enum {
 	_TNL_ATTRIB_GENERIC13 = 29,
 	_TNL_ATTRIB_GENERIC14 = 30,
 	_TNL_ATTRIB_GENERIC15 = 31,
-	_TNL_ATTRIB_MAT_FRONT_AMBIENT = 32,
-	_TNL_ATTRIB_MAT_BACK_AMBIENT = 33,
-	_TNL_ATTRIB_MAT_FRONT_DIFFUSE = 34,
-	_TNL_ATTRIB_MAT_BACK_DIFFUSE = 35,
-	_TNL_ATTRIB_MAT_FRONT_SPECULAR = 36,
-	_TNL_ATTRIB_MAT_BACK_SPECULAR = 37,
-	_TNL_ATTRIB_MAT_FRONT_EMISSION = 38,
-	_TNL_ATTRIB_MAT_BACK_EMISSION = 39,
-	_TNL_ATTRIB_MAT_FRONT_SHININESS = 40,
-	_TNL_ATTRIB_MAT_BACK_SHININESS = 41,
-	_TNL_ATTRIB_MAT_FRONT_INDEXES = 42,
-	_TNL_ATTRIB_MAT_BACK_INDEXES = 43,
-	_TNL_ATTRIB_EDGEFLAG = 44,
-	_TNL_ATTRIB_POINTSIZE = 45,
-	_TNL_ATTRIB_MAX = 46
+
+	/* These alias with the generics, but they are not active
+	 * concurrently, so it's not a problem.  The TNL module
+	 * doesn't have to do anything about this as this is how they
+	 * are passed into the _draw_prims callback.
+	 *
+	 * When we generate fixed-function replacement programs (in
+	 * t_vp_build.c currently), they refer to the appropriate
+	 * generic attribute in order to pick up per-vertex material
+	 * data.
+	 */
+	_TNL_ATTRIB_MAT_FRONT_AMBIENT = 16,
+	_TNL_ATTRIB_MAT_BACK_AMBIENT = 17,
+	_TNL_ATTRIB_MAT_FRONT_DIFFUSE = 18,
+	_TNL_ATTRIB_MAT_BACK_DIFFUSE = 19,
+	_TNL_ATTRIB_MAT_FRONT_SPECULAR = 20,
+	_TNL_ATTRIB_MAT_BACK_SPECULAR = 21,
+	_TNL_ATTRIB_MAT_FRONT_EMISSION = 22,
+	_TNL_ATTRIB_MAT_BACK_EMISSION = 23,
+	_TNL_ATTRIB_MAT_FRONT_SHININESS = 24,
+	_TNL_ATTRIB_MAT_BACK_SHININESS = 25,
+	_TNL_ATTRIB_MAT_FRONT_INDEXES = 26,
+	_TNL_ATTRIB_MAT_BACK_INDEXES = 27,
+
+	/* This is really a VERT_RESULT, not an attrib.  Need to fix
+	 * tnl to understand the difference.
+	 */
+	_TNL_ATTRIB_POINTSIZE = 16,
+
+	_TNL_ATTRIB_MAX = 32
 } ;
 
 #define _TNL_ATTRIB_TEX(u)       (_TNL_ATTRIB_TEX0 + (u))
@@ -166,8 +160,8 @@ enum {
 #define _TNL_FIRST_GENERIC _TNL_ATTRIB_GENERIC0
 #define _TNL_LAST_GENERIC  _TNL_ATTRIB_GENERIC15
 
-#define _TNL_FIRST_MAT       _TNL_ATTRIB_MAT_FRONT_AMBIENT
-#define _TNL_LAST_MAT        _TNL_ATTRIB_MAT_BACK_INDEXES
+#define _TNL_FIRST_MAT       _TNL_ATTRIB_MAT_FRONT_AMBIENT /* GENERIC0 */
+#define _TNL_LAST_MAT        _TNL_ATTRIB_MAT_BACK_INDEXES  /* GENERIC11 */
 
 /* Number of available generic attributes */
 #define _TNL_NUM_GENERIC 16
@@ -175,221 +169,22 @@ enum {
 /* Number of attributes used for evaluators */
 #define _TNL_NUM_EVAL 16
 
+
 #define PRIM_BEGIN     0x10
 #define PRIM_END       0x20
-#define PRIM_WEAK      0x40
 #define PRIM_MODE_MASK 0x0f
 
-/* 
- */
-struct tnl_prim {
-   GLuint mode;
-   GLuint start;
-   GLuint count;
-};
-
-
-
-struct tnl_eval1_map {
-   struct gl_1d_map *map;
-   GLuint sz;
-};
-
-struct tnl_eval2_map {
-   struct gl_2d_map *map;
-   GLuint sz;
-};
-
-struct tnl_eval {
-   GLuint new_state;
-   struct tnl_eval1_map map1[_TNL_NUM_EVAL];
-   struct tnl_eval2_map map2[_TNL_NUM_EVAL];
-};
-
-
-#define TNL_MAX_PRIM 16
-#define TNL_MAX_COPIED_VERTS 3
-
-struct tnl_copied_vtx {
-   GLfloat buffer[_TNL_ATTRIB_MAX * 4 * TNL_MAX_COPIED_VERTS];
-   GLuint nr;
-};
-
-#define VERT_BUFFER_SIZE 2048	/* 8kbytes */
-
-
-typedef void (*tnl_attrfv_func)( const GLfloat * );
-
-struct _tnl_dynfn {
-   struct _tnl_dynfn *next, *prev;
-   GLuint key;
-   char *code;
-};
-
-struct _tnl_dynfn_lists {
-   struct _tnl_dynfn Vertex[4];
-   struct _tnl_dynfn Attribute[4];
-};
-
-struct _tnl_dynfn_generators {
-   struct _tnl_dynfn *(*Vertex[4])( GLcontext *ctx, int key );
-   struct _tnl_dynfn *(*Attribute[4])( GLcontext *ctx, int key );
-};
-
-#define _TNL_MAX_ATTR_CODEGEN 32
-
-
-/**
- * The assembly of vertices in immediate mode is separated from
- * display list compilation.  This allows a simpler immediate mode
- * treatment and a display list compiler better suited to
- * hardware-acceleration.
- */
-struct tnl_vtx {
-   GLfloat buffer[VERT_BUFFER_SIZE];
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLubyte active_sz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-   struct tnl_prim prim[TNL_MAX_PRIM];
-   GLuint prim_count;
-   GLfloat *vbptr;		      /* cursor, points into buffer */
-   GLfloat vertex[_TNL_ATTRIB_MAX*4]; /* current vertex */
-   GLfloat *attrptr[_TNL_ATTRIB_MAX]; /* points into vertex */
-   GLfloat *current[_TNL_ATTRIB_MAX]; /* points into ctx->Current, etc */
-   GLfloat CurrentFloatEdgeFlag;
-   GLuint counter, initial_counter;
-   struct tnl_copied_vtx copied;
-
-   /** Note extra space for error handler: */
-   tnl_attrfv_func tabfv[_TNL_ATTRIB_ERROR+1][4];
-
-   struct _tnl_dynfn_lists cache;
-   struct _tnl_dynfn_generators gen;
-
-   struct tnl_eval eval;
-   GLboolean *edgeflag_tmp;
-   GLboolean have_materials;
-};
-
-
-
-
-/* For display lists, this structure holds a run of vertices of the
- * same format, and a strictly well-formed set of begin/end pairs,
- * starting on the first vertex and ending at the last.  Vertex
- * copying on buffer breaks is precomputed according to these
- * primitives, though there are situations where the copying will need
- * correction at execute-time, perhaps by replaying the list as
- * immediate mode commands.
- *
- * On executing this list, the 'current' values may be updated with
- * the values of the final vertex, and often no fixup of the start of
- * the vertex list is required.
- *
- * Eval and other commands that don't fit into these vertex lists are
- * compiled using the fallback opcode mechanism provided by dlist.c.
- */
-struct tnl_vertex_list {
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-
-   GLfloat *buffer;
-   GLuint count;
-   GLuint wrap_count;		/* number of copied vertices at start */
-   GLboolean have_materials;	/* bit of a hack - quick check for materials */
-   GLboolean dangling_attr_ref;	/* current attr implicitly referenced 
-				   outside the list */
-
-   GLfloat *normal_lengths;
-   struct tnl_prim *prim;
-   GLuint prim_count;
-
-   struct tnl_vertex_store *vertex_store;
-   struct tnl_primitive_store *prim_store;
-};
-
-/* These buffers should be a reasonable size to support upload to
- * hardware?  Maybe drivers should stitch them back together, or
- * specify a desired size?
- */
-#define SAVE_BUFFER_SIZE (16*1024)
-#define SAVE_PRIM_SIZE   128
-
-/* Storage to be shared among several vertex_lists.
- */
-struct tnl_vertex_store {
-   GLfloat buffer[SAVE_BUFFER_SIZE];
-   GLuint used;
-   GLuint refcount;
-};
-
-struct tnl_primitive_store {
-   struct tnl_prim buffer[SAVE_PRIM_SIZE];
-   GLuint used;
-   GLuint refcount;
-};
-
-
-struct tnl_save {
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-
-   GLfloat *buffer;
-   GLuint count;
-   GLuint wrap_count;
-   GLuint replay_flags;
-
-   struct tnl_prim *prim;
-   GLuint prim_count, prim_max;
-
-   struct tnl_vertex_store *vertex_store;
-   struct tnl_primitive_store *prim_store;
-
-   GLfloat *vbptr;		   /* cursor, points into buffer */
-   GLfloat vertex[_TNL_ATTRIB_MAX*4];	   /* current values */
-   GLfloat *attrptr[_TNL_ATTRIB_MAX];
-   GLuint counter, initial_counter;
-   GLboolean dangling_attr_ref;
-   GLboolean have_materials;
-
-   GLuint opcode_vertex_list;
-
-   struct tnl_copied_vtx copied;
-   
-   GLfloat CurrentFloatEdgeFlag;
-
-   GLfloat *current[_TNL_ATTRIB_MAX]; /* points into ctx->ListState */
-   GLubyte *currentsz[_TNL_ATTRIB_MAX];
+static INLINE GLuint _tnl_translate_prim( const struct _mesa_prim *prim )
+{
+   GLuint flag;
+   flag = prim->mode;
+   if (prim->begin) flag |= PRIM_BEGIN;
+   if (prim->end) flag |= PRIM_END;
+   return flag;
+}
 
-   void (*tabfv[_TNL_ATTRIB_MAX][4])( const GLfloat * );
-};
 
 
-/**
- * A collection of vertex arrays.
- */
-struct tnl_vertex_arrays
-{
-   /* Conventional vertex attribute arrays */
-   GLvector4f  Obj;
-   GLvector4f  Normal;
-   GLvector4f  Color;
-   GLvector4f  SecondaryColor;
-   GLvector4f  FogCoord;
-   GLvector4f  TexCoord[MAX_TEXTURE_COORD_UNITS];
-   GLvector4f  Index;
-
-   GLubyte     *EdgeFlag;
-   GLuint      *Elt;
-
-   /* These attributes don't alias with the conventional attributes.
-    * The GL_NV_vertex_program extension defines 16 extra sets of vertex
-    * arrays which have precedent over the conventional arrays when enabled.
-    */
-   /* XXX I think the array size is wronge (47 vs. 16) */
-   GLvector4f  Attribs[_TNL_ATTRIB_MAX];
-};
-
 
 /**
  * Contains the current state of a running pipeline.
@@ -424,11 +219,10 @@ struct vertex_buffer
    GLvector4f  *FogCoordPtr;	                /* _TNL_BIT_FOG */
    GLvector4f  *VaryingPtr[MAX_VARYING];
 
-   struct tnl_prim  *Primitive;	              
+   const struct _mesa_prim  *Primitive;	              
    GLuint      PrimitiveCount;	      
 
    /* Inputs to the vertex program stage */
-   /* XXX This array may be too large (47 vs. 16) */
    GLvector4f *AttribPtr[_TNL_ATTRIB_MAX];      /* GL_NV_vertex_program */
 };
 
@@ -626,12 +420,6 @@ struct tnl_device_driver
     * arrays.
     */
 
-   GLboolean (*NotifyBegin)(GLcontext *ctx, GLenum p);
-   /* Allow drivers to hook in optimized begin/end engines.
-    * Return value:  GL_TRUE - driver handled the begin
-    *                GL_FALSE - driver didn't handle the begin
-    */
-
    /***
     *** Rendering -- These functions called only from t_vb_render.c
     ***/
@@ -737,26 +525,11 @@ typedef struct
     */
    struct tnl_device_driver Driver;
 
-   /* Execute:
-    */
-   struct tnl_vtx vtx;
-   
-   /* Compile:
-    */
-   struct tnl_save save;
-
    /* Pipeline
     */
    struct tnl_pipeline pipeline;
    struct vertex_buffer vb;
 
-   /* GLvectors for binding to vb:
-    */
-   struct tnl_vertex_arrays vtx_inputs;
-   struct tnl_vertex_arrays save_inputs;
-   struct tnl_vertex_arrays current;
-   struct tnl_vertex_arrays array_inputs;
-
    /* Clipspace/ndc/window vertex managment:
     */
    struct tnl_clipspace clipspace;
@@ -764,26 +537,21 @@ typedef struct
    /* Probably need a better configuration mechanism:
     */
    GLboolean NeedNdcCoords;
-   GLboolean LoopbackDListCassettes;
-   GLboolean CalcDListNormalLengths;
-   GLboolean IsolateMaterials;
    GLboolean AllowVertexFog;
    GLboolean AllowPixelFog;
-   GLboolean AllowCodegen;
-
    GLboolean _DoVertexFog;  /* eval fog function at each vertex? */
 
-   /* If True, it means we started a glBegin/End primtive with an invalid
-    * vertex/fragment program or incomplete framebuffer.  In that case,
-    * discard any buffered vertex data.
-    */
-   GLboolean DiscardPrimitive;
-
    DECLARE_RENDERINPUTS(render_inputs_bitset);
 
-   GLvertexformat exec_vtxfmt;
-   GLvertexformat save_vtxfmt;
+   GLvector4f tmp_inputs[VERT_ATTRIB_MAX];
 
+   /* Temp storage for t_draw.c: 
+    */
+   GLubyte *block[VERT_ATTRIB_MAX];
+   GLuint nr_blocks;
+
+   /* Cache of fixed-function-replacing vertex programs:
+    */
    struct tnl_cache *vp_cache;
 
 } TNLcontext;
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
new file mode 100644
index 0000000000..c97cf5f7b2
--- /dev/null
+++ b/src/mesa/tnl/t_draw.c
@@ -0,0 +1,406 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "context.h"
+#include "imports.h"
+#include "state.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "enums.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+#include "t_vp_build.h"
+#include "t_vertex.h"
+#include "tnl.h"
+
+
+
+static GLubyte *get_space(GLcontext *ctx, GLuint bytes)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLubyte *space = _mesa_malloc(bytes);
+   
+   tnl->block[tnl->nr_blocks++] = space;
+   return space;
+}
+
+
+static void free_space(GLcontext *ctx)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+   for (i = 0; i < tnl->nr_blocks; i++)
+      _mesa_free(tnl->block[i]);
+   tnl->nr_blocks = 0;
+}
+
+
+/* Convert the incoming array to GLfloats.  Understands the
+ * array->Normalized flag and selects the correct conversion method.
+ */
+#define CONVERT( TYPE, MACRO ) do {		\
+   GLuint i, j;					\
+   if (input->Normalized) {			\
+      for (i = 0; i < count; i++) {		\
+	 const TYPE *in = (TYPE *)ptr;		\
+	 for (j = 0; j < sz; j++) {		\
+	    *fptr++ = MACRO(*in);		\
+	    in++;				\
+	 }					\
+	 ptr += input->StrideB;			\
+      }						\
+   } else {					\
+      for (i = 0; i < count; i++) {		\
+	 const TYPE *in = (TYPE *)ptr;		\
+	 for (j = 0; j < sz; j++) {		\
+	    *fptr++ = (GLfloat)(*in);		\
+	    in++;				\
+	 }					\
+	 ptr += input->StrideB;			\
+      }						\
+   }						\
+} while (0)
+
+
+
+/* Adjust pointer to point at first requested element, convert to
+ * floating point, populate VB->AttribPtr[].
+ */
+static void _tnl_import_array( GLcontext *ctx,
+			       GLuint attrib,
+			       GLuint count,
+			       const struct gl_client_array *input,
+			       const GLubyte *ptr )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint stride = input->StrideB;
+
+   if (input->Type != GL_FLOAT) {
+      const GLuint sz = input->Size;
+      GLubyte *buf = get_space(ctx, count * sz * sizeof(GLfloat));
+      GLfloat *fptr = (GLfloat *)buf;
+
+      switch (input->Type) {
+      case GL_BYTE: 
+	 CONVERT(GLbyte, BYTE_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_BYTE: 
+	 CONVERT(GLubyte, UBYTE_TO_FLOAT); 
+	 break;
+      case GL_SHORT: 
+	 CONVERT(GLshort, SHORT_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_SHORT: 
+	 CONVERT(GLushort, USHORT_TO_FLOAT); 
+	 break;
+      case GL_INT: 
+	 CONVERT(GLint, INT_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_INT: 
+	 CONVERT(GLuint, UINT_TO_FLOAT); 
+	 break;
+      case GL_DOUBLE: 
+	 CONVERT(GLdouble, (GLfloat)); 
+	 break;
+      default:
+	 assert(0);
+	 break;
+      }
+
+      ptr = buf;
+      stride = sz * sizeof(GLfloat);
+   }
+
+   VB->AttribPtr[attrib] = &tnl->tmp_inputs[attrib];
+   VB->AttribPtr[attrib]->data = (GLfloat (*)[4])ptr;
+   VB->AttribPtr[attrib]->start = (GLfloat *)ptr;
+   VB->AttribPtr[attrib]->count = count;
+   VB->AttribPtr[attrib]->stride = stride;
+   VB->AttribPtr[attrib]->size = input->Size;
+
+   /* This should die, but so should the whole GLvector4f concept: 
+    */
+   VB->AttribPtr[attrib]->flags = (((1<<input->Size)-1) | 
+				   VEC_NOT_WRITEABLE |
+				   (stride == 4*sizeof(GLfloat) ? 0 : VEC_BAD_STRIDE));
+   
+   VB->AttribPtr[attrib]->storage = NULL;
+}
+
+#define CLIPVERTS  ((6 + MAX_CLIP_PLANES) * 2)
+
+
+static GLboolean *_tnl_import_edgeflag( GLcontext *ctx,
+					const GLvector4f *input,
+					GLuint count)
+{
+   const GLubyte *ptr = (const GLubyte *)input->data;
+   const GLuint stride = input->stride;
+   GLboolean *space = (GLboolean *)get_space(ctx, count + CLIPVERTS);
+   GLboolean *bptr = space;
+   GLuint i;
+
+   for (i = 0; i < count; i++) {
+      *bptr++ = ((GLfloat *)ptr)[0] == 1.0;
+      ptr += stride;
+   }
+
+   return space;
+}
+
+
+static void bind_inputs( GLcontext *ctx, 
+			 const struct gl_client_array *inputs[],
+			 GLint count,
+			 struct gl_buffer_object **bo,
+			 GLuint *nr_bo )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+
+   /* Map all the VBOs
+    */
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      const void *ptr;
+
+      if (inputs[i]->BufferObj->Name) { 
+	 if (!inputs[i]->BufferObj->Pointer) {
+	    bo[*nr_bo] = inputs[i]->BufferObj;
+	    (*nr_bo)++;
+	    ctx->Driver.MapBuffer(ctx, 
+				  GL_ARRAY_BUFFER,
+				  GL_READ_ONLY_ARB,
+				  inputs[i]->BufferObj);
+	    
+	    assert(inputs[i]->BufferObj->Pointer);
+	 }
+	 
+	 ptr = ADD_POINTERS(inputs[i]->BufferObj->Pointer,
+			    inputs[i]->Ptr);
+      }
+      else
+	 ptr = inputs[i]->Ptr;
+
+      /* Just make sure the array is floating point, otherwise convert to
+       * temporary storage.  
+       *
+       * XXX: remove the GLvector4f type at some stage and just use
+       * client arrays.
+       */
+      _tnl_import_array(ctx, i, count, inputs[i], ptr);
+   }
+
+   /* We process only the vertices between min & max index:
+    */
+   VB->Count = count;
+
+
+   /* Legacy pointers -- remove one day.
+    */
+   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
+   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
+   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
+   VB->ColorPtr[1] = NULL;
+   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
+   VB->IndexPtr[1] = NULL;
+   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
+   VB->SecondaryColorPtr[1] = NULL;
+   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
+   }
+
+   /* Clipping and drawing code still requires this to be a packed
+    * array of ubytes which can be written into.  TODO: Fix and
+    * remove.
+    */
+   if (ctx->Polygon.FrontMode != GL_FILL ||
+       ctx->Polygon.BackMode != GL_FILL)
+   {
+      VB->EdgeFlag = _tnl_import_edgeflag( ctx, 
+					   VB->AttribPtr[_TNL_ATTRIB_EDGEFLAG],
+					   VB->Count );
+   }
+
+}
+
+
+/* Translate indices to GLuints and store in VB->Elts.
+ */
+static void bind_indices( GLcontext *ctx,
+			  const struct _mesa_index_buffer *ib,
+			  struct gl_buffer_object **bo,
+			  GLuint *nr_bo)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+   void *ptr;
+
+   if (!ib) {
+      VB->Elts = NULL;
+      return;
+   }
+
+   if (ib->obj->Name && !ib->obj->Pointer) {
+      bo[*nr_bo] = ib->obj;
+      (*nr_bo)++;
+      ctx->Driver.MapBuffer(ctx, 
+			    GL_ELEMENT_ARRAY_BUFFER,
+			    GL_READ_ONLY_ARB,
+			    ib->obj);
+
+      assert(ib->obj->Pointer);
+   }
+
+   ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);
+
+   if (ib->type == GL_UNSIGNED_INT) {
+      VB->Elts = (GLuint *) ptr;
+   }
+   else {
+      GLuint *elts = (GLuint *)get_space(ctx, ib->count * sizeof(GLuint));
+      VB->Elts = elts;
+
+      if (ib->type == GL_UNSIGNED_SHORT) {
+	 const GLushort *in = (GLushort *)ptr;
+	 for (i = 0; i < ib->count; i++) 
+	    *elts++ = (GLuint)(*in++);
+      }
+      else {
+	 const GLubyte *in = (GLubyte *)ptr;
+	 for (i = 0; i < ib->count; i++) 
+	    *elts++ = (GLuint)(*in++);
+      }
+   }
+}
+
+static void bind_prims( GLcontext *ctx,
+			const struct _mesa_prim *prim,
+			GLuint nr_prims )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+
+   VB->Primitive = prim;
+   VB->PrimitiveCount = nr_prims;
+}
+
+static void unmap_vbos( GLcontext *ctx,
+			struct gl_buffer_object **bo,
+			GLuint nr_bo )
+{
+   GLuint i;
+   for (i = 0; i < nr_bo; i++) { 
+      ctx->Driver.UnmapBuffer(ctx, 
+			      0, /* target -- I don't see why this would be needed */
+			      bo[i]);
+   }
+}
+
+
+
+/* This is the main entrypoint into the slimmed-down software tnl
+ * module.  In a regular swtnl driver, this can be plugged straight
+ * into the vbo->Driver.DrawPrims() callback.
+ */
+void _tnl_draw_prims( 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)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   const GLuint TEST_SPLIT = 0;
+   const GLint max = TEST_SPLIT ? 8 : tnl->vb.Size - MAX_CLIPPED_VERTICES;
+
+   if (0)
+   {
+      GLuint i;
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
+      for (i = 0; i < nr_prims; i++)
+	 _mesa_printf("prim %d: %s start %d count %d\n", i, 
+		      _mesa_lookup_enum_by_nr(prim[i].mode),
+		      prim[i].start,
+		      prim[i].count);
+   }
+
+   if (min_index) {
+      /* We always translate away calls with min_index != 0. 
+       */
+      vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, 
+			min_index, max_index,
+			_tnl_draw_prims );
+      return;
+   }
+   else if (max_index >= max) {
+      /* The software TNL pipeline has a fixed amount of storage for
+       * vertices and it is necessary to split incoming drawing commands
+       * if they exceed that limit.
+       */
+      struct split_limits limits;
+      limits.max_verts = max;
+      limits.max_vb_size = ~0;
+      limits.max_indices = ~0;
+
+      /* This will split the buffers one way or another and
+       * recursively call back into this function.
+       */
+      vbo_split_prims( ctx, arrays, prim, nr_prims, ib, 
+		       0, max_index,
+		       _tnl_draw_prims,
+		       &limits );
+   }
+   else {
+      /* May need to map a vertex buffer object for every attribute plus
+       * one for the index buffer.
+       */
+      struct gl_buffer_object *bo[VERT_ATTRIB_MAX + 1];
+      GLuint nr_bo = 0;
+
+      /* Binding inputs may imply mapping some vertex buffer objects.
+       * They will need to be unmapped below.
+       */
+      bind_inputs(ctx, arrays, max_index+1, bo, &nr_bo);
+      bind_indices(ctx, ib, bo, &nr_bo);
+      bind_prims(ctx, prim, nr_prims );
+
+      TNL_CONTEXT(ctx)->Driver.RunPipeline(ctx);
+
+      unmap_vbos(ctx, bo, nr_bo);
+      free_space(ctx);
+   }
+}
+
diff --git a/src/mesa/tnl/t_save_api.h b/src/mesa/tnl/t_save_api.h
deleted file mode 100644
index c30e0e8ecd..0000000000
--- a/src/mesa/tnl/t_save_api.h
+++ /dev/null
@@ -1,57 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 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
-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
-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>
- *
- */
-
-#ifndef __T_SAVE_API_H__
-#define __T_SAVE_API_H__
-
-#include "t_context.h"
-
-extern GLboolean _tnl_weak_begin( GLcontext *ctx, GLenum mode );
-
-extern void _tnl_EndList( GLcontext *ctx );
-extern void _tnl_NewList( GLcontext *ctx, GLuint list, GLenum mode );
-
-extern void _tnl_EndCallList( GLcontext *ctx );
-extern void _tnl_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
-
-extern void _tnl_SaveFlushVertices( GLcontext *ctx );
-
-extern void _tnl_save_init( GLcontext *ctx );
-extern void _tnl_save_destroy( GLcontext *ctx );
-
-extern void _tnl_loopback_vertex_list( GLcontext *ctx,
-				       const struct tnl_vertex_list *list );
-
-extern void _tnl_playback_vertex_list( GLcontext *ctx, void *data );
-
-#endif
diff --git a/src/mesa/tnl/t_save_loopback.c b/src/mesa/tnl/t_save_loopback.c
deleted file mode 100644
index b96bb4ec55..0000000000
--- a/src/mesa/tnl/t_save_loopback.c
+++ /dev/null
@@ -1,330 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.3
- *
- * Copyright (C) 1999-2004  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.
- */
-
-/* Author:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "enums.h"
-#include "glapi.h"
-#include "imports.h"
-#include "macros.h"
-#include "mtypes.h"
-#include "t_context.h"
-#include "t_save_api.h"
-#include "dispatch.h"
-
-/* If someone compiles a display list like:
- *      glBegin(Triangles)
- *      glVertex() 
- *      ... lots of vertices ...
- *      glEnd()
- *
- * or: 
- *      glDrawArrays(...)
- * 
- * and then tries to execute it like this:
- *
- *      glBegin(Lines)
- *      glCallList()
- *      glEnd()
- *
- * it will wind up in here, as the vertex copying used when wrapping
- * buffers in list compilation (Triangles) won't be right for how the
- * list is being executed (as Lines). 
- *
- * This could be avoided by not compiling as vertex_lists until after
- * the first glEnd() has been seen.  However, that would miss an
- * important category of display lists, for the sake of a degenerate
- * usage. 
- *
- * Further, replaying degenerately-called lists in this fashion is
- * probably still faster than the replay using opcodes.
- */
-
-typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
-
-
-/* Wrapper functions in case glVertexAttrib*fvNV doesn't exist */
-static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc[4] = {
-   VertexAttrib1fvNV,
-   VertexAttrib2fvNV,
-   VertexAttrib3fvNV,
-   VertexAttrib4fvNV
-};
-
-
-static void VertexAttrib1fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvARB(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc_arb[4] = {
-   VertexAttrib1fvARB,
-   VertexAttrib2fvARB,
-   VertexAttrib3fvARB,
-   VertexAttrib4fvARB
-};
-
-
-
-
-
-
-
-static void mat_attr1fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SHININESS, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SHININESS, v ));
-      break;
-   }
-}
-
-
-static void mat_attr3fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_COLOR_INDEXES, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_COLOR_INDEXES, v ));
-      break;
-   }
-}
-
-
-static void mat_attr4fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_EMISSION, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_EMISSION, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_AMBIENT, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_AMBIENT, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_DIFFUSE, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_DIFFUSE, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SPECULAR, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SPECULAR, v ));
-      break;
-   }
-}
-
-
-static attr_func mat_attrfunc[4] = {
-   mat_attr1fv,
-   NULL,
-   mat_attr3fv,
-   mat_attr4fv
-};
-
-
-static void edgeflag_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_EdgeFlag(ctx->Exec, ((GLboolean)(v[0] == 1.0)));
-}
-
-struct loopback_attr {
-   GLint target;
-   GLint sz;
-   attr_func func;
-};
-
-/* Don't emit ends and begins on wrapped primitives.  Don't replay
- * wrapped vertices.  If we get here, it's probably because the the
- * precalculated wrapping is wrong.
- */
-static void loopback_prim( GLcontext *ctx,
-			   const struct tnl_vertex_list *list, GLuint i,
-			   const struct loopback_attr *la, GLuint nr )
-{
-   struct tnl_prim *prim = &list->prim[i];
-   GLint begin = prim->start;
-   GLint end = begin + prim->count;
-   GLfloat *data;
-   GLint j;
-   GLuint k;
-
-   if (prim->mode & PRIM_BEGIN) {
-      CALL_Begin(GET_DISPATCH(), ( prim->mode & PRIM_MODE_MASK ));
-   }
-   else {
-      assert(i == 0);
-      assert(begin == 0);
-      begin += list->wrap_count;
-   }
-
-   data = list->buffer + begin * list->vertex_size;
-
-   for (j = begin ; j < end ; j++) {
-      GLfloat *tmp = data + la[0].sz;
-
-      for (k = 1 ; k < nr ; k++) {
-	 la[k].func( ctx, la[k].target, tmp );
-	 tmp += la[k].sz;
-      }
-	 
-      /* Fire the vertex
-       */
-      la[0].func( ctx, VERT_ATTRIB_POS, data );
-      data = tmp;
-   }
-
-   if (prim->mode & PRIM_END) {
-      CALL_End(GET_DISPATCH(), ());
-   }
-   else {
-      assert (i == list->prim_count-1);
-   }
-}
-
-/* Primitives generated by DrawArrays/DrawElements/Rectf may be
- * caught here.  If there is no primitive in progress, execute them
- * normally, otherwise need to track and discard the generated
- * primitives.
- */
-static void loopback_weak_prim( GLcontext *ctx,
-				const struct tnl_vertex_list *list, GLuint i,
-				const struct loopback_attr *la, GLuint nr )
-{
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) 
-      loopback_prim( ctx, list, i, la, nr );
-   else {
-      struct tnl_prim *prim = &list->prim[i];
-
-      /* Use the prim_weak flag to ensure that if this primitive
-       * wraps, we don't mistake future vertex_lists for part of the
-       * surrounding primitive.
-       *
-       * While this flag is set, we are simply disposing of data
-       * generated by an operation now known to be a noop.
-       */
-      if (prim->mode & PRIM_BEGIN)
-	 ctx->Driver.CurrentExecPrimitive |= PRIM_WEAK;
-      if (prim->mode & PRIM_END)
-	 ctx->Driver.CurrentExecPrimitive &= ~PRIM_WEAK;
-   }
-}
-
-
-
-void _tnl_loopback_vertex_list( GLcontext *ctx,
-                                const struct tnl_vertex_list *list )
-{
-   struct loopback_attr la[_TNL_ATTRIB_MAX];
-   GLuint i, nr = 0;
-
-   /* conventional + generic attributes */
-   for (i = 0 ; i <= _TNL_ATTRIB_GENERIC15 ; i++) {
-      if (list->attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = list->attrsz[i];
-	 la[nr].func = vert_attrfunc[list->attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   /* material attributes */
-   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT ; 
-	i <= _TNL_ATTRIB_MAT_BACK_INDEXES ; 
-	i++) {
-      if (list->attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = list->attrsz[i];
-	 la[nr].func = mat_attrfunc[list->attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   /* special-case: edgeflag */
-   if (list->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      la[nr].target = _TNL_ATTRIB_EDGEFLAG;
-      la[nr].sz = list->attrsz[_TNL_ATTRIB_EDGEFLAG];
-      la[nr].func = edgeflag_attr1fv;
-      nr++;
-   }
-
-   for (i = 0 ; i < list->prim_count ; i++) {
-      if (list->prim[i].mode & PRIM_WEAK)
-	 loopback_weak_prim( ctx, list, i, la, nr );
-      else
-	 loopback_prim( ctx, list, i, la, nr );
-   }
-}
diff --git a/src/mesa/tnl/t_save_playback.c b/src/mesa/tnl/t_save_playback.c
deleted file mode 100644
index 426e94372e..0000000000
--- a/src/mesa/tnl/t_save_playback.c
+++ /dev/null
@@ -1,216 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.1
- *
- * Copyright (C) 1999-2004  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.
- */
-
-/* Author:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "imports.h"
-#include "mtypes.h"
-#include "macros.h"
-#include "light.h"
-#include "state.h"
-#include "t_pipeline.h"
-#include "t_save_api.h"
-#include "t_vtx_api.h"
-
-static INLINE GLint get_size( const GLfloat *f )
-{
-   if (f[3] != 1.0) return 4;
-   if (f[2] != 0.0) return 3;
-   return 2;
-}
-
-
-/* Some nasty stuff still hanging on here.  
- *
- * TODO - remove VB->ColorPtr, etc and just use the AttrPtr's.
- */
-static void _tnl_bind_vertex_list( GLcontext *ctx,
-                                   const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->save_inputs;
-   GLfloat *data = node->buffer;
-   GLuint attr, i;
-
-   /* Setup constant data in the VB.
-    */
-   VB->Count = node->count;
-   VB->Primitive = node->prim;
-   VB->PrimitiveCount = node->prim_count;
-   VB->Elts = NULL;
-   VB->NormalLengthPtr = node->normal_lengths;
-
-   for (attr = 0; attr <= _TNL_ATTRIB_EDGEFLAG; attr++) {
-      if (node->attrsz[attr]) {
-	 tmp->Attribs[attr].count = node->count;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) data;
-	 tmp->Attribs[attr].start = data;
-	 tmp->Attribs[attr].size = node->attrsz[attr];
-	 tmp->Attribs[attr].stride = node->vertex_size * sizeof(GLfloat);
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-	 data += node->attrsz[attr];
-      }
-      else {
-	 tmp->Attribs[attr].count = 1;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) tnl->vtx.current[attr];
-	 tmp->Attribs[attr].start = tnl->vtx.current[attr];
-	 tmp->Attribs[attr].size = get_size( tnl->vtx.current[attr] );
-	 tmp->Attribs[attr].stride = 0;
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-      }
-   }
-
-   
-   /* Copy edgeflag to a contiguous array
-    */
-   if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL) {
-      if (node->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-	 VB->EdgeFlag = _tnl_translate_edgeflag( ctx, data, 
-						 node->count,
-						 node->vertex_size );
-	 data++;
-      }
-      else 
-	 VB->EdgeFlag = _tnl_import_current_edgeflag( ctx, node->count );
-   }
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
-
-static void _playback_copy_to_current( GLcontext *ctx,
-				       const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   const GLfloat *data;
-   GLuint i;
-
-   if (node->count)
-      data = node->buffer + (node->count-1) * node->vertex_size;
-   else
-      data = node->buffer;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i <= _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (node->attrsz[i]) {
-	 COPY_CLEAN_4V(tnl->vtx.current[i], node->attrsz[i], data);
-	 data += node->attrsz[i];
-      }
-   }
-
-   /* Edgeflag requires special treatment:
-    */
-   if (node->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (data[0] == 1.0);
-   }
-
-   /* Colormaterial -- this kindof sucks.
-    */
-   if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
-   }
-
-   if (node->have_materials) {
-      tnl->Driver.NotifyMaterialChange( ctx );
-      ctx->NewState |= _NEW_LIGHT;
-   }
-
-   /* CurrentExecPrimitive
-    */
-   if (node->prim_count) {
-      GLenum mode = node->prim[node->prim_count - 1].mode;
-      if (mode & PRIM_END)
-	 ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-      else
-	 ctx->Driver.CurrentExecPrimitive = (mode & PRIM_MODE_MASK);
-   }
-}
-
-
-/**
- * Execute the buffer and save copied verts.
- */
-void _tnl_playback_vertex_list( GLcontext *ctx, void *data )
-{
-   const struct tnl_vertex_list *node = (const struct tnl_vertex_list *) data;
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   FLUSH_CURRENT(ctx, 0);
-
-   if (node->prim_count > 0 && node->count > 0) {
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END &&
-	  (node->prim[0].mode & PRIM_BEGIN)) {
-
-	 /* Degenerate case: list is called inside begin/end pair and
-	  * includes operations such as glBegin or glDrawArrays.
-	  */
-	 _mesa_error( ctx, GL_INVALID_OPERATION, "displaylist recursive begin");
-	 _tnl_loopback_vertex_list( ctx, node );
-	 return;
-      }
-      else if (tnl->save.replay_flags) {
-	 /* Various degnerate cases: translate into immediate mode
-	  * calls rather than trying to execute in place.
-	  */
-	 _tnl_loopback_vertex_list( ctx, node );
-	 return;
-      }
-      
-      if (ctx->NewState)
-	 _mesa_update_state( ctx );
-
-      if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-          (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glBegin (invalid vertex/fragment program)");
-         return;
-      }
-
-      _tnl_bind_vertex_list( ctx, node );
-
-      tnl->Driver.RunPipeline( ctx );
-   }
-
-   /* Copy to current?
-    */
-   _playback_copy_to_current( ctx, node );
-}
diff --git a/src/mesa/tnl/t_vb_arbprogram.c b/src/mesa/tnl/t_vb_arbprogram.c
index 0a443b3e01..97ff4c7f10 100644
--- a/src/mesa/tnl/t_vb_arbprogram.c
+++ b/src/mesa/tnl/t_vb_arbprogram.c
@@ -1096,7 +1096,7 @@ static void compile_vertex_program( struct gl_vertex_program *program,
 { 
    struct compilation cp;
    struct tnl_compiled_program *p = CALLOC_STRUCT(tnl_compiled_program);
-   GLuint i;
+   GLint i;
 
 #if 1
    if (!program->IsNVProgram && program->IsPositionInvariant) {
@@ -1293,8 +1293,7 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
    m->nr_inputs = m->nr_outputs = 0;
 
    for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      if (program->Base.InputsRead & (1<<i) ||
-	  (i == VERT_ATTRIB_POS && program->IsPositionInvariant)) {
+      if (program->Base.InputsRead & (1<<i)) {
 	 GLuint j = m->nr_inputs++;
 	 m->input[j].idx = i;
 	 m->input[j].data = (GLfloat *)m->VB->AttribPtr[i]->data;
@@ -1302,16 +1301,15 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 	 m->input[j].size = m->VB->AttribPtr[i]->size;
 	 ASSIGN_4V(m->File[0][REG_IN0 + i], 0, 0, 0, 1);
       }
-   }     
+   }
 
    for (i = 0; i < VERT_RESULT_MAX; i++) {
-      if (program->Base.OutputsWritten & (1 << i) ||
-	  (i == VERT_RESULT_HPOS && program->IsPositionInvariant)) {
+      if (program->Base.OutputsWritten & (1 << i)) {
 	 GLuint j = m->nr_outputs++;
 	 m->output[j].idx = i;
 	 m->output[j].data = (GLfloat *)m->attribs[i].data;
       }
-   }     
+   }
 
 
    /* Run the actual program:
@@ -1325,7 +1323,13 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 	 case 2: m->File[0][idx][1] = m->input[j].data[1];
 	 case 1: m->File[0][idx][0] = m->input[j].data[0];
 	 }
-
+#if 0
+         printf("  attr %d/%d: %g %g %g %g\n", j, idx-REG_IN0,
+                m->input[j].data[0],
+                m->input[j].data[1],
+                m->input[j].data[2],
+                m->input[j].data[3]);
+#endif
 	 STRIDE_F(m->input[j].data, m->input[j].stride);
       }
 
@@ -1334,22 +1338,13 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 	 call_func( p, m );
       }
       else {
+         GLint j;
 	 for (j = 0; j < p->nr_instructions; j++) {
 	    union instruction inst = p->instructions[j];	 
 	    opcode_func[inst.alu.opcode]( m, inst );
 	 }
       }
 
-      /* If the program is position invariant, multiply the input position
-       * by the MVP matrix and store in the vertex position result register.
-       */
-#if 0
-      if (program->IsPositionInvariant) {
-	 TRANSFORM_POINT( m->File[0][REG_OUT0+0], 
-			  ctx->_ModelProjectMatrix.m, 
-			  m->File[0][REG_IN0+0]);
-      }
-#endif
       for (j = 0; j < m->nr_outputs; j++) {
 	 GLuint idx = REG_OUT0 + m->output[j].idx;
 	 m->output[j].data[0] = m->File[0][idx][0];
@@ -1371,15 +1366,39 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
     * TODO: 2) Integrate t_vertex.c so that we just go straight ahead
     * and build machine vertices here.
     */
-   VB->ClipPtr = &m->attribs[VERT_RESULT_HPOS];
-   VB->ClipPtr->count = VB->Count;
 
    /* XXX There seems to be confusion between using the VERT_ATTRIB_*
     * values vs _TNL_ATTRIB_* tokens here:
     */
    outputs = program->Base.OutputsWritten;
-   if (program->IsPositionInvariant) 
-      outputs |= (1<<VERT_RESULT_HPOS);
+
+   if (program->IsPositionInvariant) {
+      /* We need the exact same transform as in the fixed function path here
+         to guarantee invariance, depending on compiler optimization flags results
+         could be different otherwise */
+      VB->ClipPtr = TransformRaw( &m->attribs[0],
+				  &ctx->_ModelProjectMatrix,
+				  m->VB->AttribPtr[0] );
+
+      /* Drivers expect this to be clean to element 4...
+       */
+      switch (VB->ClipPtr->size) {
+      case 1:
+	 /* impossible */
+      case 2:
+	 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 2 );
+	 /* fall-through */
+      case 3:
+	 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 3 );
+	 /* fall-through */
+      case 4:
+	 break;
+      }
+   }
+   else {
+      VB->ClipPtr = &m->attribs[VERT_RESULT_HPOS];
+      VB->ClipPtr->count = VB->Count;
+   }
 
    if (outputs & (1<<VERT_RESULT_COL0)) {
       VB->ColorPtr[0] =
diff --git a/src/mesa/tnl/t_vb_render.c b/src/mesa/tnl/t_vb_render.c
index 163fa31257..c38f0745e1 100644
--- a/src/mesa/tnl/t_vb_render.c
+++ b/src/mesa/tnl/t_vb_render.c
@@ -305,7 +305,7 @@ static GLboolean run_render( GLcontext *ctx,
 
       for (i = 0 ; i < VB->PrimitiveCount ; i++)
       {
-	 GLuint prim = VB->Primitive[i].mode;
+	 GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 	 GLuint start = VB->Primitive[i].start;
 	 GLuint length = VB->Primitive[i].count;
 
diff --git a/src/mesa/tnl/t_vb_rendertmp.h b/src/mesa/tnl/t_vb_rendertmp.h
index 90319eddfc..2b5f4e93b2 100644
--- a/src/mesa/tnl/t_vb_rendertmp.h
+++ b/src/mesa/tnl/t_vb_rendertmp.h
@@ -45,8 +45,8 @@
 #endif
 
 #ifndef TEST_PRIM_END
-#define TEST_PRIM_END(flags) (flags & PRIM_END)
-#define TEST_PRIM_BEGIN(flags) (flags & PRIM_BEGIN)
+#define TEST_PRIM_END(prim) (flags & PRIM_END)
+#define TEST_PRIM_BEGIN(prim) (flags & PRIM_BEGIN)
 #endif
 
 #ifndef ELT
diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c
index 8b75dff96a..283782588a 100644
--- a/src/mesa/tnl/t_vp_build.c
+++ b/src/mesa/tnl/t_vp_build.c
@@ -812,14 +812,13 @@ static struct ureg calculate_light_attenuation( struct tnl_program *p,
    /* Calculate spot attenuation:
     */
    if (!p->state->unit[i].light_spotcutoff_is_180) {
-      struct ureg spot_dir = register_param3(p, STATE_LIGHT, i,
-					     STATE_SPOT_DIRECTION);
+      struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
+						  STATE_SPOT_DIR_NORMALIZED, i);
       struct ureg spot = get_temp(p);
       struct ureg slt = get_temp(p);
-	       
-      emit_normalize_vec3( p, spot, spot_dir ); /* XXX: precompute! */
-      emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot);
-      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir,W), spot);
+
+      emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm);
+      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
       emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
       emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
 
@@ -1109,29 +1108,27 @@ static void build_fog( struct tnl_program *p )
    }
 
    if (p->state->tnl_do_vertex_fog) {
-      struct ureg params = register_param2(p, STATE_FOG, STATE_FOG_PARAMS);
+      struct ureg params = register_param2(p, STATE_INTERNAL,
+					   STATE_FOG_PARAMS_OPTIMIZED);
       struct ureg tmp = get_temp(p);
 
       switch (p->state->fog_mode) {
       case FOG_LINEAR: {
 	 struct ureg id = get_identity_param(p);
-	 emit_op2(p, OPCODE_SUB, tmp, 0, swizzle1(params,Z), input); 
-	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,W)); 
+	 emit_op3(p, OPCODE_MAD, tmp, 0, input, swizzle1(params,X), swizzle1(params,Y));
 	 emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
 	 emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
 	 break;
       }
       case FOG_EXP:
 	 emit_op1(p, OPCODE_ABS, tmp, 0, input); 
-	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,X)); 
-	 emit_op2(p, OPCODE_POW, fog, WRITEMASK_X, 
-		  register_const1f(p, M_E), negate(tmp)); 
+	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,Z));
+	 emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
 	 break;
       case FOG_EXP2:
-	 emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,X)); 
+	 emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
 	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp); 
-	 emit_op2(p, OPCODE_POW, fog, WRITEMASK_X, 
-		  register_const1f(p, M_E), negate(tmp)); 
+	 emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
 	 break;
       }
       
@@ -1323,8 +1320,6 @@ static void build_texture_transform( struct tnl_program *p )
 }
 
 
-/* Seems like it could be tighter:
- */
 static void build_pointsize( struct tnl_program *p )
 {
    struct ureg eye = get_eye_position(p);
@@ -1333,20 +1328,25 @@ static void build_pointsize( struct tnl_program *p )
    struct ureg out = register_output(p, VERT_RESULT_PSIZ);
    struct ureg ut = get_temp(p);
 
-   /* 1, -Z, Z * Z, 1 */      
-   emit_op1(p, OPCODE_MOV, ut, 0, swizzle1(get_identity_param(p), W));
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_YZ, ut, negate(swizzle1(eye, Z)));
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_Z, ut, negate(swizzle1(eye, Z)));
-
-
-   /* p1 +  p2 * dist + p3 * dist * dist, 0 */
-   emit_op2(p, OPCODE_DP3, ut, 0, ut, state_attenuation);
-
-   /* 1 / factor */
-   emit_op1(p, OPCODE_RCP, ut, 0, ut ); 
-
-   /* out = pointSize / factor */
-   emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size); 
+   /* p1 + dist * (p2 + dist * p3); */
+   emit_op3(p, OPCODE_MAD, ut, 0, negate(swizzle1(eye, Z)),
+		swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
+   emit_op3(p, OPCODE_MAD, ut, 0, negate(swizzle1(eye, Z)),
+		ut, swizzle1(state_attenuation, X));
+
+   /* 1 / sqrt(factor) */
+   emit_op1(p, OPCODE_RSQ, ut, 0, ut );
+
+#if 1
+   /* out = pointSize / sqrt(factor) */
+   emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
+#else
+   /* not sure, might make sense to do clamping here,
+      but it's not done in t_vb_points neither */
+   emit_op2(p, OPCODE_MUL, ut, 0, ut, state_size);
+   emit_op2(p, OPCODE_MAX, ut, 0, ut, swizzle1(state_size, Y));
+   emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
+#endif
 
    release_temp(p, ut);
 }
diff --git a/src/mesa/tnl/t_vtx_api.c b/src/mesa/tnl/t_vtx_api.c
deleted file mode 100644
index 77eec8b3b4..0000000000
--- a/src/mesa/tnl/t_vtx_api.c
+++ /dev/null
@@ -1,1046 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 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
-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
-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 "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-#include "simple_list.h"
-
-#include "dispatch.h"
-
-static void reset_attrfv( TNLcontext *tnl );
-
-/** Note extra space for error index: */
-static tnl_attrfv_func choose[_TNL_ATTRIB_ERROR+1][4];
-static tnl_attrfv_func generic_attr_func[_TNL_MAX_ATTR_CODEGEN][4];
-
-
-/* Close off the last primitive, execute the buffer, restart the
- * primitive.  
- */
-static void _tnl_wrap_buffers( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   
-
-   if (tnl->vtx.prim_count == 0) {
-      tnl->vtx.copied.nr = 0;
-      tnl->vtx.counter = tnl->vtx.initial_counter;
-      tnl->vtx.vbptr = tnl->vtx.buffer;
-   }
-   else {
-      GLuint last_prim = tnl->vtx.prim[tnl->vtx.prim_count-1].mode;
-      GLuint last_count;
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-	 GLint i = tnl->vtx.prim_count - 1;
-	 assert(i >= 0);
-	 tnl->vtx.prim[i].count = ((tnl->vtx.initial_counter - 
-				    tnl->vtx.counter) - 
-				   tnl->vtx.prim[i].start);
-      }
-
-      last_count = tnl->vtx.prim[tnl->vtx.prim_count-1].count;
-
-      /* Execute the buffer and save copied vertices.
-       */
-      if (tnl->vtx.counter != tnl->vtx.initial_counter)
-	 _tnl_flush_vtx( ctx );
-      else {
-	 tnl->vtx.prim_count = 0;
-	 tnl->vtx.copied.nr = 0;
-      }
-
-      /* Emit a glBegin to start the new list.
-       */
-      assert(tnl->vtx.prim_count == 0);
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-	 tnl->vtx.prim[0].mode = ctx->Driver.CurrentExecPrimitive;
-	 tnl->vtx.prim[0].start = 0;
-	 tnl->vtx.prim[0].count = 0;
-	 tnl->vtx.prim_count++;
-      
-	 if (tnl->vtx.copied.nr == last_count)
-	    tnl->vtx.prim[0].mode |= last_prim & PRIM_BEGIN;
-      }
-   }
-}
-
-
-/* Deal with buffer wrapping where provoked by the vertex buffer
- * filling up, as opposed to upgrade_vertex().
- *
- * Make it GLAPIENTRY, so we can tail from the codegen'ed Vertex*fv
- */
-void GLAPIENTRY _tnl_wrap_filled_vertex( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat *data = tnl->vtx.copied.buffer;
-   GLuint i;
-
-   /* Run pipeline on current vertices, copy wrapped vertices
-    * to tnl->copied.
-    */
-   _tnl_wrap_buffers( ctx );
-   
-   /* Copy stored stored vertices to start of new list. 
-    */
-   assert(tnl->vtx.counter > tnl->vtx.copied.nr);
-
-   for (i = 0 ; i < tnl->vtx.copied.nr ; i++) {
-      _mesa_memcpy( tnl->vtx.vbptr, data, 
-		    tnl->vtx.vertex_size * sizeof(GLfloat));
-      tnl->vtx.vbptr += tnl->vtx.vertex_size;
-      data += tnl->vtx.vertex_size;
-      tnl->vtx.counter--;
-   }
-
-   tnl->vtx.copied.nr = 0;
-}
-
-
-/*
- * Copy the active vertex's values to the ctx->Current fields.
- */
-static void _tnl_copy_to_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i < _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (tnl->vtx.attrsz[i]) {
-         /* Note: the tnl->vtx.current[i] pointers points to
-          * the ctx->Current fields.  The first 16 or so, anyway.
-          */
-	 COPY_CLEAN_4V(tnl->vtx.current[i], 
-                       tnl->vtx.attrsz[i], 
-                       tnl->vtx.attrptr[i]);
-      }
-   }
-
-   /* Edgeflag requires additional treatment:
-    */
-   if (tnl->vtx.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = 
-	 (tnl->vtx.CurrentFloatEdgeFlag == 1.0);
-   }
-
-   /* Colormaterial -- this kindof sucks.
-    */
-   if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, 
-				  ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
-   }
-
-   if (tnl->vtx.have_materials) {
-      tnl->Driver.NotifyMaterialChange( ctx );
-      ctx->NewState |= _NEW_LIGHT;
-   }
-         
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-
-static void _tnl_copy_from_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLint i;
-
-   /* Edgeflag requires additional treatment:
-    */
-   tnl->vtx.CurrentFloatEdgeFlag = (GLfloat) ctx->Current.EdgeFlag;
-   
-   for (i = _TNL_ATTRIB_POS+1 ; i < _TNL_ATTRIB_MAX ; i++) 
-      switch (tnl->vtx.attrsz[i]) {
-      case 4: tnl->vtx.attrptr[i][3] = tnl->vtx.current[i][3];
-      case 3: tnl->vtx.attrptr[i][2] = tnl->vtx.current[i][2];
-      case 2: tnl->vtx.attrptr[i][1] = tnl->vtx.current[i][1];
-      case 1: tnl->vtx.attrptr[i][0] = tnl->vtx.current[i][0];
-	 break;
-      }
-
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-}
-
-
-/* Flush existing data, set new attrib size, replay copied vertices.
- */ 
-static void _tnl_wrap_upgrade_vertex( GLcontext *ctx, 
-				      GLuint attr,
-				      GLuint newsz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint oldsz;
-   GLuint i;
-   GLfloat *tmp;
-   GLint lastcount = tnl->vtx.initial_counter - tnl->vtx.counter;
-
-   /* Run pipeline on current vertices, copy wrapped vertices
-    * to tnl->vtx.copied.
-    */
-   _tnl_wrap_buffers( ctx );
-
-
-   /* Do a COPY_TO_CURRENT to ensure back-copying works for the case
-    * when the attribute already exists in the vertex and is having
-    * its size increased.  
-    */
-   _tnl_copy_to_current( ctx );
-
-
-   /* Heuristic: Attempt to isolate attributes received outside
-    * begin/end so that they don't bloat the vertices.
-    */
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END &&
-       tnl->vtx.attrsz[attr] == 0 && 
-       lastcount > 8 &&
-       tnl->vtx.vertex_size) {
-      reset_attrfv( tnl );
-   }
-
-   /* Fix up sizes:
-    */
-   oldsz = tnl->vtx.attrsz[attr];
-   tnl->vtx.attrsz[attr] = newsz;
-
-   tnl->vtx.vertex_size += newsz - oldsz;
-   tnl->vtx.counter = MIN2( VERT_BUFFER_SIZE / tnl->vtx.vertex_size,
-			    ctx->Const.MaxArrayLockSize );
-   tnl->vtx.initial_counter = tnl->vtx.counter;
-   tnl->vtx.vbptr = tnl->vtx.buffer;
-
-
-   /* Recalculate all the attrptr[] values
-    */
-   for (i = 0, tmp = tnl->vtx.vertex ; i < _TNL_ATTRIB_MAX ; i++) {
-      if (tnl->vtx.attrsz[i]) {
-	 tnl->vtx.attrptr[i] = tmp;
-	 tmp += tnl->vtx.attrsz[i];
-      }
-      else 
-	 tnl->vtx.attrptr[i] = NULL; /* will not be dereferenced */
-   }
-
-   /* Copy from current to repopulate the vertex with correct values.
-    */
-   _tnl_copy_from_current( ctx );
-
-   /* Replay stored vertices to translate them
-    * to new format here.
-    *
-    * -- No need to replay - just copy piecewise
-    */
-   if (tnl->vtx.copied.nr)
-   {
-      const GLfloat *data = tnl->vtx.copied.buffer;
-      GLfloat *dest = tnl->vtx.buffer;
-      GLuint j;
-
-      for (i = 0 ; i < tnl->vtx.copied.nr ; i++) {
-	 for (j = 0 ; j < _TNL_ATTRIB_MAX ; j++) {
-	    if (tnl->vtx.attrsz[j]) {
-	       if (j == attr) {
-		  if (oldsz) {
-		     COPY_CLEAN_4V( dest, oldsz, data );
-		     data += oldsz;
-		     dest += newsz;
-		  } else {
-		     COPY_SZ_4V( dest, newsz, tnl->vtx.current[j] );
-		     dest += newsz;
-		  }
-	       }
-	       else {
-		  GLuint sz = tnl->vtx.attrsz[j];
-		  COPY_SZ_4V( dest, sz, data );
-		  dest += sz;
-		  data += sz;
-	       }
-	    }
-	 }
-      }
-
-      tnl->vtx.vbptr = dest;
-      tnl->vtx.counter -= tnl->vtx.copied.nr;
-      tnl->vtx.copied.nr = 0;
-   }
-
-   /* For codegen - attrptr's may have changed, so need to redo
-    * codegen.  Might be a reasonable place to try & detect attributes
-    * in the vertex which aren't being submitted any more.
-    */
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) 
-      if (tnl->vtx.attrsz[i]) {
-	 GLuint j = tnl->vtx.attrsz[i] - 1;
-
-	 if (i < _TNL_MAX_ATTR_CODEGEN)
-	    tnl->vtx.tabfv[i][j] = choose[i][j];
-      }
-
-}
-
-
-static void _tnl_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   static const GLfloat id[4] = { 0, 0, 0, 1 };
-   int i;
-
-   if (0) 
-      _mesa_printf("%s attr %d sz %d -> %d\n",
-		   __FUNCTION__, attr, tnl->vtx.attrsz[attr], sz);
-
-   if (tnl->vtx.attrsz[attr] < sz) {
-      /* New size is larger.  Need to flush existing vertices and get
-       * an enlarged vertex format.
-       */
-      _tnl_wrap_upgrade_vertex( ctx, attr, sz );
-   }
-   else if (sz < tnl->vtx.active_sz[attr]) {
-      /* New size is smaller - just need to fill in some
-       * zeros.  Don't need to flush or wrap.
-       */
-      for (i = sz ; i <= tnl->vtx.attrsz[attr] ; i++)
-	 tnl->vtx.attrptr[attr][i-1] = id[i-1];
-   }
-
-   tnl->vtx.active_sz[attr] = sz;
-
-   /* Does setting NeedFlush belong here?  Necessitates resetting
-    * vtxfmt on each flush (otherwise flags won't get reset
-    * afterwards).
-    */
-   if (attr == 0) 
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   else 
-      ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-}
-
-#ifdef USE_X86_ASM
-
-static struct _tnl_dynfn *lookup( struct _tnl_dynfn *l, GLuint key )
-{
-   struct _tnl_dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key == key) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-
-static tnl_attrfv_func do_codegen( GLcontext *ctx, GLuint attr, GLuint sz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   struct _tnl_dynfn *dfn = NULL;
-
-   if (attr == 0) {
-      GLuint key = tnl->vtx.vertex_size;
-
-      dfn = lookup( &tnl->vtx.cache.Vertex[sz-1], key );
-
-      if (!dfn)
-	 dfn = tnl->vtx.gen.Vertex[sz-1]( ctx, key );
-   }
-   else {
-      GLuint key = (GLuint) tnl->vtx.attrptr[attr];
-
-      dfn = lookup( &tnl->vtx.cache.Attribute[sz-1], key );
-
-      if (!dfn)
-	 dfn = tnl->vtx.gen.Attribute[sz-1]( ctx, key );
-   }
-
-   if (dfn) 
-      return *(tnl_attrfv_func *) &dfn->code;
-   else
-      return NULL;
-}
-
-#endif /* USE_X86_ASM */
-
-/* Helper function for 'CHOOSE' macro.  Do what's necessary when an
- * entrypoint is called for the first time.
- */
-
-static tnl_attrfv_func do_choose( GLuint attr, GLuint sz )
-{ 
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint oldsz = tnl->vtx.active_sz[attr];
-
-   assert(attr < _TNL_MAX_ATTR_CODEGEN);
-
-   if (oldsz != sz) {
-      /* Reset any active pointers for this attribute 
-       */
-      if (oldsz)
-	 tnl->vtx.tabfv[attr][oldsz-1] = choose[attr][oldsz-1];
-   
-      _tnl_fixup_vertex( ctx, attr, sz );
- 
-   }
-
-
-   /* Try to use codegen:
-    */
-#ifdef USE_X86_ASM
-   if (tnl->AllowCodegen)
-      tnl->vtx.tabfv[attr][sz-1] = do_codegen( ctx, attr, sz );
-   else
-#endif
-      tnl->vtx.tabfv[attr][sz-1] = NULL;
-
-   /* Else use generic version:
-    */
-   if (!tnl->vtx.tabfv[attr][sz-1])
-      tnl->vtx.tabfv[attr][sz-1] = generic_attr_func[attr][sz-1];
-
-   ASSERT(tnl->vtx.tabfv[attr][sz-1]);
-   return tnl->vtx.tabfv[attr][sz-1];
-}
-
-
-
-#define CHOOSE( ATTR, N )				\
-static void choose_##ATTR##_##N( const GLfloat *v )	\
-{							\
-   tnl_attrfv_func f = do_choose(ATTR, N);		\
-   ASSERT(f); \
-   f( v );						\
-}
-
-#define CHOOSERS( ATTRIB )	\
-   CHOOSE( ATTRIB, 1 )		\
-   CHOOSE( ATTRIB, 2 )		\
-   CHOOSE( ATTRIB, 3 )		\
-   CHOOSE( ATTRIB, 4 )		\
-
-
-#define INIT_CHOOSERS(ATTR)			\
-   ASSERT(ATTR <= _TNL_ATTRIB_ERROR);\
-   choose[ATTR][0] = choose_##ATTR##_1;		\
-   choose[ATTR][1] = choose_##ATTR##_2;		\
-   choose[ATTR][2] = choose_##ATTR##_3;		\
-   choose[ATTR][3] = choose_##ATTR##_4;
-
-/* conventional attributes */
-CHOOSERS( 0 )
-CHOOSERS( 1 )
-CHOOSERS( 2 )
-CHOOSERS( 3 )
-CHOOSERS( 4 )
-CHOOSERS( 5 )
-CHOOSERS( 6 )
-CHOOSERS( 7 )
-CHOOSERS( 8 )
-CHOOSERS( 9 )
-CHOOSERS( 10 )
-CHOOSERS( 11 )
-CHOOSERS( 12 )
-CHOOSERS( 13 )
-CHOOSERS( 14 )
-CHOOSERS( 15 )
-
-/* generic attributes */
-CHOOSERS( 16 )
-CHOOSERS( 17 )
-CHOOSERS( 18 )
-CHOOSERS( 19 )
-CHOOSERS( 20 )
-CHOOSERS( 21 )
-CHOOSERS( 22 )
-CHOOSERS( 23 )
-CHOOSERS( 24 )
-CHOOSERS( 25 )
-CHOOSERS( 26 )
-CHOOSERS( 27 )
-CHOOSERS( 28 )
-CHOOSERS( 29 )
-CHOOSERS( 30 )
-CHOOSERS( 31 )
-
-
-/**
- * This function will get called when glVertexAttribNV/ARB() is called
- * with an invalid index parameter.
- */
-static void
-error_attrib(const GLfloat *unused)
-{
-   GET_CURRENT_CONTEXT( ctx );
-   (void) unused;
-   _mesa_error( ctx, GL_INVALID_VALUE, "glVertexAttrib(index)" );
-}   
-
-
-
-/**
- * Reset all the per-vertex functions pointers to point to the default
- * "chooser" functions.
- */
-static void
-reset_attrfv(TNLcontext *tnl)
-{   
-   GLuint i;
-
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) 
-      if (tnl->vtx.attrsz[i]) {
-	 GLint j = tnl->vtx.attrsz[i] - 1;
-	 tnl->vtx.attrsz[i] = 0;
-	 tnl->vtx.active_sz[i] = 0;
-
-	 if (i < _TNL_MAX_ATTR_CODEGEN) {
-            while (j >= 0) {
-	       tnl->vtx.tabfv[i][j] = choose[i][j];
-               j--;
-            }
-         }
-      }
-
-   tnl->vtx.vertex_size = 0;
-   tnl->vtx.have_materials = 0;
-}
-      
-
-
-/**
- * Materials:  
- * 
- * These are treated as per-vertex attributes, at indices above where
- * the NV_vertex_program leaves off.  There are a lot of good things
- * about treating materials this way.  
- *
- * However: I don't want to double the number of generated functions
- * just to cope with this, so I unroll the 'C' varients of CHOOSE and
- * ATTRF into this function, and dispense with codegen and
- * second-level dispatch.
- *
- * There is no aliasing of material attributes with other entrypoints.
- */
-#define OTHER_ATTR( A, N, params )			\
-do {							\
-   if (tnl->vtx.active_sz[A] != N) {			\
-      _tnl_fixup_vertex( ctx, A, N );			\
-   }							\
-							\
-   {							\
-      GLfloat *dest = tnl->vtx.attrptr[A];		\
-      if (N>0) dest[0] = (params)[0];			\
-      if (N>1) dest[1] = (params)[1];			\
-      if (N>2) dest[2] = (params)[2];			\
-      if (N>3) dest[3] = (params)[3];			\
-   }							\
-} while (0)
-
-
-#define MAT( ATTR, N, face, params )			\
-do {							\
-   if (face != GL_BACK)					\
-      OTHER_ATTR( ATTR, N, params ); /* front */	\
-   if (face != GL_FRONT)				\
-      OTHER_ATTR( ATTR + 1, N, params ); /* back */	\
-} while (0)
-
-
-/**
- * Called by glMaterialfv().
- * Colormaterial is dealt with later on.
- */
-static void GLAPIENTRY
-_tnl_Materialfv( GLenum face, GLenum pname, const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (face) {
-   case GL_FRONT:
-   case GL_BACK:
-   case GL_FRONT_AND_BACK:
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-
-   switch (pname) {
-   case GL_EMISSION:
-      MAT( _TNL_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
-      break;
-   case GL_AMBIENT:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      break;
-   case GL_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   case GL_SPECULAR:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
-      break;
-   case GL_SHININESS:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
-      break;
-   case GL_COLOR_INDEXES:
-      MAT( _TNL_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
-      break;
-   case GL_AMBIENT_AND_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-
-   tnl->vtx.have_materials = GL_TRUE;
-}
-
-
-static void GLAPIENTRY _tnl_EdgeFlag( GLboolean b )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat f = (GLfloat)b;
-
-   OTHER_ATTR( _TNL_ATTRIB_EDGEFLAG, 1, &f );
-}
-
-
-/* Eval
- */
-static void GLAPIENTRY _tnl_EvalCoord1f( GLfloat u )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* TODO: use a CHOOSE() function for this: */
-   {
-      GLint i;
-      if (tnl->vtx.eval.new_state) 
-	 _tnl_update_eval( ctx );
-
-      for (i = 0; i < _TNL_NUM_EVAL; i++) {
-	 if (tnl->vtx.eval.map1[i].map) 
-	    if (tnl->vtx.active_sz[i] != tnl->vtx.eval.map1[i].sz)
-	       _tnl_fixup_vertex( ctx, i, tnl->vtx.eval.map1[i].sz );
-      }
-   }
-
-
-   _mesa_memcpy( tnl->vtx.copied.buffer, tnl->vtx.vertex, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-
-   _tnl_do_EvalCoord1f( ctx, u );
-
-   _mesa_memcpy( tnl->vtx.vertex, tnl->vtx.copied.buffer,
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-}
-
-static void GLAPIENTRY _tnl_EvalCoord2f( GLfloat u, GLfloat v )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* TODO: use a CHOOSE() function for this: */
-   {
-      GLint i;
-      if (tnl->vtx.eval.new_state) 
-	 _tnl_update_eval( ctx );
-
-      for (i = 0; i < _TNL_NUM_EVAL; i++) {
-	 if (tnl->vtx.eval.map2[i].map) 
-	    if (tnl->vtx.active_sz[i] != tnl->vtx.eval.map2[i].sz)
-	       _tnl_fixup_vertex( ctx, i, tnl->vtx.eval.map2[i].sz );
-      }
-
-      if (ctx->Eval.AutoNormal) 
-	 if (tnl->vtx.active_sz[_TNL_ATTRIB_NORMAL] != 3)
-	    _tnl_fixup_vertex( ctx, _TNL_ATTRIB_NORMAL, 3 );
-   }
-
-   _mesa_memcpy( tnl->vtx.copied.buffer, tnl->vtx.vertex, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-
-   _tnl_do_EvalCoord2f( ctx, u, v );
-
-   _mesa_memcpy( tnl->vtx.vertex, tnl->vtx.copied.buffer, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-}
-
-static void GLAPIENTRY _tnl_EvalCoord1fv( const GLfloat *u )
-{
-   _tnl_EvalCoord1f( u[0] );
-}
-
-static void GLAPIENTRY _tnl_EvalCoord2fv( const GLfloat *u )
-{
-   _tnl_EvalCoord2f( u[0], u[1] );
-}
-
-static void GLAPIENTRY _tnl_EvalPoint1( GLint i )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   GLfloat du = ((ctx->Eval.MapGrid1u2 - ctx->Eval.MapGrid1u1) /
-		 (GLfloat) ctx->Eval.MapGrid1un);
-   GLfloat u = i * du + ctx->Eval.MapGrid1u1;
-
-   _tnl_EvalCoord1f( u );
-}
-
-
-static void GLAPIENTRY _tnl_EvalPoint2( GLint i, GLint j )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   GLfloat du = ((ctx->Eval.MapGrid2u2 - ctx->Eval.MapGrid2u1) / 
-		 (GLfloat) ctx->Eval.MapGrid2un);
-   GLfloat dv = ((ctx->Eval.MapGrid2v2 - ctx->Eval.MapGrid2v1) / 
-		 (GLfloat) ctx->Eval.MapGrid2vn);
-   GLfloat u = i * du + ctx->Eval.MapGrid2u1;
-   GLfloat v = j * dv + ctx->Eval.MapGrid2v1;
-
-   _tnl_EvalCoord2f( u, v );
-}
-
-
-/**
- * Called from glBegin.
- * ctx->Driver.CurrentExecPrimitive will be set to <mode>.
- */
-static void GLAPIENTRY _tnl_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-
-   if (mode > GL_POLYGON) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glBegin(mode)");
-      return;
-   }
-
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) {
-      /* we're not inside a glBegin/End pair */
-      TNLcontext *tnl = TNL_CONTEXT(ctx); 
-      GLuint i;
-
-      if (ctx->NewState) {
-	 _mesa_update_state( ctx );
-
-         if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-            (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-            _mesa_error(ctx, GL_INVALID_OPERATION,
-                        "glBegin (invalid vertex/fragment program)");
-            tnl->DiscardPrimitive = GL_TRUE;
-            return;
-         }
-
-         if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
-            _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
-                        "glBegin(incomplete framebuffer)");
-            tnl->DiscardPrimitive = GL_TRUE;
-            return;
-         }
-
-         tnl->DiscardPrimitive = GL_FALSE;
-
-	 if (!(tnl->Driver.NotifyBegin && 
-	       tnl->Driver.NotifyBegin( ctx, mode )))
-	     CALL_Begin(ctx->Exec, (mode));
-	 return;
-      }
-
-      /* Heuristic: attempt to isolate attributes occuring outside
-       * begin/end pairs.
-       */
-      if (tnl->vtx.vertex_size && !tnl->vtx.attrsz[0]) 
-	 _tnl_FlushVertices( ctx, ~0 );
-
-      i = tnl->vtx.prim_count++;
-      tnl->vtx.prim[i].mode = mode | PRIM_BEGIN;
-      tnl->vtx.prim[i].start = tnl->vtx.initial_counter - tnl->vtx.counter;
-      tnl->vtx.prim[i].count = 0;
-
-      ctx->Driver.CurrentExecPrimitive = mode;
-   }
-   else {
-      /* already inside glBegin/End */
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-   }      
-}
-
-
-/**
- * Called from glEnd.
- */
-static void GLAPIENTRY _tnl_End( void )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-
-   if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-      /* closing an open glBegin primitive */
-      TNLcontext *tnl = TNL_CONTEXT(ctx); 
-      int idx = tnl->vtx.initial_counter - tnl->vtx.counter;
-      int i = tnl->vtx.prim_count - 1;
-
-      tnl->vtx.prim[i].mode |= PRIM_END; 
-      tnl->vtx.prim[i].count = idx - tnl->vtx.prim[i].start;
-
-      ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-
-      /* Two choices which effect the way vertex attributes are
-       * carried over (or not) between adjacent primitives.
-       */
-#if 0
-      if (tnl->vtx.prim_count == TNL_MAX_PRIM) 
-	 _tnl_FlushVertices( ctx, ~0 );
-#else
-      if (tnl->vtx.prim_count == TNL_MAX_PRIM)
-	 _tnl_flush_vtx( ctx );	
-#endif
-
-   }
-   else {
-      /* glBegin hasn't been called! */
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-   }
-}
-
-
-/**
- * XXX why aren't all members initialized here??
- */
-static void _tnl_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = _tnl_Begin;
-   vfmt->CallList = _mesa_CallList;
-   vfmt->CallLists = _mesa_CallLists;
-   vfmt->EdgeFlag = _tnl_EdgeFlag;
-   vfmt->End = _tnl_End;
-   vfmt->EvalCoord1f = _tnl_EvalCoord1f;
-   vfmt->EvalCoord1fv = _tnl_EvalCoord1fv;
-   vfmt->EvalCoord2f = _tnl_EvalCoord2f;
-   vfmt->EvalCoord2fv = _tnl_EvalCoord2fv;
-   vfmt->EvalPoint1 = _tnl_EvalPoint1;
-   vfmt->EvalPoint2 = _tnl_EvalPoint2;
-   vfmt->Materialfv = _tnl_Materialfv;
-
-   vfmt->Rectf = _mesa_noop_Rectf;
-   vfmt->EvalMesh1 = _mesa_noop_EvalMesh1;
-   vfmt->EvalMesh2 = _mesa_noop_EvalMesh2;
-}
-
-
-
-void _tnl_FlushVertices( GLcontext *ctx, GLuint flags )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   (void) flags;
-
-   if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-      /* still inside a glBegin/End pair.  How'd we get here??? */
-      return;
-   }
-
-   if (tnl->DiscardPrimitive) {
-      /* discard any primitives */
-      tnl->vtx.prim_count = 0;
-      tnl->vtx.counter = tnl->vtx.initial_counter;
-      tnl->vtx.vbptr = tnl->vtx.buffer;
-   }
-
-   if (tnl->vtx.counter != tnl->vtx.initial_counter) {
-      _tnl_flush_vtx( ctx );
-   }
-
-   if (tnl->vtx.vertex_size) {
-      _tnl_copy_to_current( ctx );
-      reset_attrfv( tnl );
-   }
-
-   ctx->Driver.NeedFlush = 0;
-}
-
-
-/**
- * Init the tnl->vtx->current[] pointers to point to the corresponding
- * fields in ctx->Current attribute group.
- */
-static void _tnl_current_init( GLcontext *ctx ) 
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLint i;
-
-   /* setup the pointers for the typical (32) vertex attributes */
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) 
-      tnl->vtx.current[i] = ctx->Current.Attrib[i];
-
-   /* setup pointers for the 12 material attributes */
-   for (i = 0; i < MAT_ATTRIB_MAX; i++)
-      tnl->vtx.current[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = 
-	 ctx->Light.Material.Attrib[i];
-
-   /* special case */
-   tnl->vtx.current[_TNL_ATTRIB_EDGEFLAG] = &tnl->vtx.CurrentFloatEdgeFlag;
-}
-
-static struct _tnl_dynfn *no_codegen( GLcontext *ctx, int key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void _tnl_vtx_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   struct tnl_vertex_arrays *tmp = &tnl->vtx_inputs;
-   GLuint i;
-   static int firsttime = 1;
-   
-   if (firsttime) {
-      firsttime = 0;
-
-      /* conventional attributes */
-      INIT_CHOOSERS( 0 );
-      INIT_CHOOSERS( 1 );
-      INIT_CHOOSERS( 2 );
-      INIT_CHOOSERS( 3 );
-      INIT_CHOOSERS( 4 );
-      INIT_CHOOSERS( 5 );
-      INIT_CHOOSERS( 6 );
-      INIT_CHOOSERS( 7 );
-      INIT_CHOOSERS( 8 );
-      INIT_CHOOSERS( 9 );
-      INIT_CHOOSERS( 10 );
-      INIT_CHOOSERS( 11 );
-      INIT_CHOOSERS( 12 );
-      INIT_CHOOSERS( 13 );
-      INIT_CHOOSERS( 14 );
-      INIT_CHOOSERS( 15 );
-
-      /* generic attributes */
-      INIT_CHOOSERS( 16 );
-      INIT_CHOOSERS( 17 );
-      INIT_CHOOSERS( 18 );
-      INIT_CHOOSERS( 19 );
-      INIT_CHOOSERS( 20 );
-      INIT_CHOOSERS( 21 );
-      INIT_CHOOSERS( 22 );
-      INIT_CHOOSERS( 23 );
-      INIT_CHOOSERS( 24 );
-      INIT_CHOOSERS( 25 );
-      INIT_CHOOSERS( 26 );
-      INIT_CHOOSERS( 27 );
-      INIT_CHOOSERS( 28 );
-      INIT_CHOOSERS( 29 );
-      INIT_CHOOSERS( 30 );
-      INIT_CHOOSERS( 31 );
-
-      choose[_TNL_ATTRIB_ERROR][0] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][1] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][2] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][3] = error_attrib;
-
-#ifdef USE_X86_ASM
-      if (tnl->AllowCodegen) {
-         _tnl_x86choosers(choose, do_choose); /* x86 INIT_CHOOSERS */
-      }
-#endif
-
-      _tnl_generic_attr_table_init( generic_attr_func );
-   }
-
-   for (i = 0; i < _TNL_ATTRIB_EDGEFLAG; i++)
-      _mesa_vector4f_init( &tmp->Attribs[i], 0, NULL);
-
-   for (i = 0; i < 4; i++) {
-      make_empty_list( &tnl->vtx.cache.Vertex[i] );
-      make_empty_list( &tnl->vtx.cache.Attribute[i] );
-      tnl->vtx.gen.Vertex[i] = no_codegen;
-      tnl->vtx.gen.Attribute[i] = no_codegen;
-   }
-
-#ifdef USE_X86_ASM
-   _tnl_InitX86Codegen( &tnl->vtx.gen );
-#endif
-
-   _tnl_current_init( ctx );
-   _tnl_exec_vtxfmt_init( ctx );
-   _tnl_generic_exec_vtxfmt_init( ctx );
-#ifdef USE_X86_ASM
-   if (tnl->AllowCodegen) {
-      _tnl_x86_exec_vtxfmt_init( ctx ); /* x86 DISPATCH_ATTRFV */
-   }
-#endif
-
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
-   _mesa_memcpy( tnl->vtx.tabfv, choose, sizeof(choose) );
-
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) {
-      tnl->vtx.attrsz[i] = 0;
-      tnl->vtx.active_sz[i] = 0;
-   }
-
-   tnl->vtx.vertex_size = 0;
-   tnl->vtx.have_materials = 0;
-}
-
-static void free_funcs( struct _tnl_dynfn *l )
-{
-   struct _tnl_dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      ALIGN_FREE( f->code );
-      FREE( f );
-   }
-}
-
-
-void _tnl_vtx_destroy( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   for (i = 0; i < 4; i++) {
-      free_funcs( &tnl->vtx.cache.Vertex[i] );
-      free_funcs( &tnl->vtx.cache.Attribute[i] ); 
-   }
-}
-
diff --git a/src/mesa/tnl/t_vtx_api.h b/src/mesa/tnl/t_vtx_api.h
deleted file mode 100644
index 5311f40a85..0000000000
--- a/src/mesa/tnl/t_vtx_api.h
+++ /dev/null
@@ -1,86 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 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
-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
-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>
- *
- */
-
-#ifndef __T_VTX_API_H__
-#define __T_VTX_API_H__
-
-#include "t_context.h"
-
-
-/* t_vtx_api.c:
- */
-extern void _tnl_vtx_init( GLcontext *ctx );
-extern void _tnl_vtx_destroy( GLcontext *ctx );
-
-extern void _tnl_FlushVertices( GLcontext *ctx, GLuint flags );
-extern void _tnl_flush_vtx( GLcontext *ctx );
-
-extern void GLAPIENTRY _tnl_wrap_filled_vertex( GLcontext *ctx );
-
-/* t_vtx_exec.c:
- */
-
-extern void _tnl_do_EvalCoord2f( GLcontext* ctx, GLfloat u, GLfloat v );
-extern void _tnl_do_EvalCoord1f(GLcontext* ctx, GLfloat u);
-extern void _tnl_update_eval( GLcontext *ctx );
-
-extern GLboolean *_tnl_translate_edgeflag( GLcontext *ctx, 
-					   const GLfloat *data, 
-					   GLuint count,
-					   GLuint stride );
-
-extern GLboolean *_tnl_import_current_edgeflag( GLcontext *ctx,
-						GLuint count );
-
-
-
-/* t_vtx_generic.c:
- */
-extern void _tnl_generic_exec_vtxfmt_init( GLcontext *ctx );
-
-extern void _tnl_generic_attr_table_init( tnl_attrfv_func (*tab)[4] );
-
-/* t_vtx_x86.c:
- */
-extern void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen );
-
-extern void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx );
-
-extern void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-			      tnl_attrfv_func (*do_choose)( GLuint attr,
-							GLuint sz ));
-
-
-
-
-#endif
diff --git a/src/mesa/tnl/t_vtx_eval.c b/src/mesa/tnl/t_vtx_eval.c
deleted file mode 100644
index bb621ecbe3..0000000000
--- a/src/mesa/tnl/t_vtx_eval.c
+++ /dev/null
@@ -1,264 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.1
- *
- * Copyright (C) 1999-2006  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.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "api_eval.h"
-#include "context.h"
-#include "macros.h"
-#include "math/m_eval.h"
-#include "t_vtx_api.h"
-#include "dispatch.h"
-
-
-static void clear_active_eval1( TNLcontext *tnl, GLuint attr ) 
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   tnl->vtx.eval.map1[attr].map = NULL;
-}
-
-static void clear_active_eval2( TNLcontext *tnl, GLuint attr ) 
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   tnl->vtx.eval.map2[attr].map = NULL;
-}
-
-static void set_active_eval1( TNLcontext *tnl, GLuint attr, GLuint dim, 
-			      struct gl_1d_map *map )
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   if (!tnl->vtx.eval.map1[attr].map) {
-      tnl->vtx.eval.map1[attr].map = map;
-      tnl->vtx.eval.map1[attr].sz = dim;
-   }
-} 
-
-static void set_active_eval2( TNLcontext *tnl, GLuint attr, GLuint dim, 
-			      struct gl_2d_map *map )
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   if (!tnl->vtx.eval.map2[attr].map) {
-      tnl->vtx.eval.map2[attr].map = map;
-      tnl->vtx.eval.map2[attr].sz = dim;
-   }
-} 
-
-void _tnl_update_eval( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   /* Vertex program maps have priority over conventional attribs */
-
-   for (attr = 0; attr < _TNL_NUM_EVAL; attr++) {
-      clear_active_eval1( tnl, attr );
-      clear_active_eval2( tnl, attr );
-   }
-
-   if (ctx->Eval.Map1Color4) 
-      set_active_eval1( tnl, VERT_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
-      
-   if (ctx->Eval.Map2Color4) 
-      set_active_eval2( tnl, VERT_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
-
-   if (ctx->Eval.Map1TextureCoord4) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
-   else if (ctx->Eval.Map1TextureCoord3) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
-   else if (ctx->Eval.Map1TextureCoord2) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
-   else if (ctx->Eval.Map1TextureCoord1) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
-
-   if (ctx->Eval.Map2TextureCoord4) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
-   else if (ctx->Eval.Map2TextureCoord3) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
-   else if (ctx->Eval.Map2TextureCoord2) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
-   else if (ctx->Eval.Map2TextureCoord1) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
-
-   if (ctx->Eval.Map1Normal) 
-      set_active_eval1( tnl, VERT_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
-
-   if (ctx->Eval.Map2Normal) 
-      set_active_eval2( tnl, VERT_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
-
-   if (ctx->Eval.Map1Vertex4) 
-      set_active_eval1( tnl, VERT_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
-   else if (ctx->Eval.Map1Vertex3) 
-      set_active_eval1( tnl, VERT_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
-
-   if (ctx->Eval.Map2Vertex4) 
-      set_active_eval2( tnl, VERT_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
-   else if (ctx->Eval.Map2Vertex3) 
-      set_active_eval2( tnl, VERT_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
-
-   /* Evaluators with generic attributes is only supported for NV vertex
-    * programs, not ARB vertex programs.  16 evaluator maps are supported.
-    * We do this after the conventional attributes since the spec says that
-    * these generic maps have higher priority.
-    */
-   if (ctx->VertexProgram._Current &&
-       ctx->VertexProgram._Current->IsNVProgram) {
-      for (attr = 0; attr < 16; attr++) {
-	 if (ctx->Eval.Map1Attrib[attr]) 
-	    set_active_eval1( tnl, attr, 4, &ctx->EvalMap.Map1Attrib[attr] );
-
-	 if (ctx->Eval.Map2Attrib[attr]) 
-	    set_active_eval2( tnl, attr, 4, &ctx->EvalMap.Map2Attrib[attr] );
-      }
-   }
-
-   tnl->vtx.eval.new_state = 0;
-}
-
-
-
-void _tnl_do_EvalCoord1f(GLcontext* ctx, GLfloat u)
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   for (attr = 1; attr < _TNL_NUM_EVAL; attr++) {
-      struct gl_1d_map *map = tnl->vtx.eval.map1[attr].map;
-      if (map) {
-	 GLfloat uu = (u - map->u1) * map->du;
-	 GLfloat data[4];
-
-	 ASSIGN_4V(data, 0, 0, 0, 1);
-
-	 _math_horner_bezier_curve(map->Points, data, uu, 
-				   tnl->vtx.eval.map1[attr].sz, 
-				   map->Order);
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[attr],
-		     tnl->vtx.attrsz[attr],
-		     data );
-      }
-   }
-
-   /** Vertex -- EvalCoord1f is a noop if this map not enabled:
-    **/
-   if (tnl->vtx.eval.map1[0].map) {
-      struct gl_1d_map *map = tnl->vtx.eval.map1[0].map;
-      GLfloat uu = (u - map->u1) * map->du;
-      GLfloat vertex[4];
-
-      ASSIGN_4V(vertex, 0, 0, 0, 1);
-
-      _math_horner_bezier_curve(map->Points, vertex, uu, 
-				tnl->vtx.eval.map1[0].sz, 
-				map->Order);
-
-      if (tnl->vtx.eval.map1[0].sz == 4) 
-	 CALL_Vertex4fv(GET_DISPATCH(), ( vertex ));
-      else
-	 CALL_Vertex3fv(GET_DISPATCH(), ( vertex )); 
-   }
-}
-
-
-
-void _tnl_do_EvalCoord2f( GLcontext* ctx, GLfloat u, GLfloat v )
-{   
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   for (attr = 1; attr < _TNL_NUM_EVAL; attr++) {
-      struct gl_2d_map *map = tnl->vtx.eval.map2[attr].map;
-      if (map) {
-	 GLfloat uu = (u - map->u1) * map->du;
-	 GLfloat vv = (v - map->v1) * map->dv;
-	 GLfloat data[4];
-
-	 ASSIGN_4V(data, 0, 0, 0, 1);
-
-	 _math_horner_bezier_surf(map->Points, 
-				  data, 
-				  uu, vv, 
-				  tnl->vtx.eval.map2[attr].sz, 
-				  map->Uorder, map->Vorder);
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[attr],
-		     tnl->vtx.attrsz[attr],
-		     data );
-      }
-   }
-
-   /** Vertex -- EvalCoord2f is a noop if this map not enabled:
-    **/
-   if (tnl->vtx.eval.map2[0].map) {
-      struct gl_2d_map *map = tnl->vtx.eval.map2[0].map;
-      GLfloat uu = (u - map->u1) * map->du;
-      GLfloat vv = (v - map->v1) * map->dv;
-      GLfloat vertex[4];
-
-      ASSIGN_4V(vertex, 0, 0, 0, 1);
-
-      if (ctx->Eval.AutoNormal) {
-	 GLfloat normal[4];
-         GLfloat du[4], dv[4];
-
-         _math_de_casteljau_surf(map->Points, vertex, du, dv, uu, vv, 
-				 tnl->vtx.eval.map2[0].sz,
-				 map->Uorder, map->Vorder);
-
-	 if (tnl->vtx.eval.map2[0].sz == 4) {
-	    du[0] = du[0]*vertex[3] - du[3]*vertex[0];
-	    du[1] = du[1]*vertex[3] - du[3]*vertex[1];
-	    du[2] = du[2]*vertex[3] - du[3]*vertex[2];
-	 
-	    dv[0] = dv[0]*vertex[3] - dv[3]*vertex[0];
-	    dv[1] = dv[1]*vertex[3] - dv[3]*vertex[1];
-	    dv[2] = dv[2]*vertex[3] - dv[3]*vertex[2];
-	 }
-
-
-         CROSS3(normal, du, dv);
-         NORMALIZE_3FV(normal);
-	 normal[3] = 1.0;
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[_TNL_ATTRIB_NORMAL],
-		     tnl->vtx.attrsz[_TNL_ATTRIB_NORMAL],
-		     normal );
-
-      }
-      else {
-         _math_horner_bezier_surf(map->Points, vertex, uu, vv, 
-				  tnl->vtx.eval.map2[0].sz,
-				  map->Uorder, map->Vorder);
-      }
-
-      if (tnl->vtx.attrsz[0] == 4) 
-	 CALL_Vertex4fv(GET_DISPATCH(), ( vertex ));
-      else
-	 CALL_Vertex3fv(GET_DISPATCH(), ( vertex )); 
-   }
-}
-
-
diff --git a/src/mesa/tnl/t_vtx_exec.c b/src/mesa/tnl/t_vtx_exec.c
deleted file mode 100644
index 900c4ab6cc..0000000000
--- a/src/mesa/tnl/t_vtx_exec.c
+++ /dev/null
@@ -1,293 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * Copyright (C) 1999-2003  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.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "api_eval.h"
-#include "context.h"
-#include "enums.h"
-#include "state.h"
-#include "macros.h"
-#include "math/m_eval.h"
-#include "t_vtx_api.h"
-#include "t_pipeline.h"
-
-
-static void _tnl_print_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint count = tnl->vtx.initial_counter - tnl->vtx.counter;
-   GLuint i;
-
-   _mesa_debug(ctx, "_tnl_print_vtx: %u vertices %d primitives, %d vertsize\n",
-               count,
-	       tnl->vtx.prim_count,
-	       tnl->vtx.vertex_size);
-
-   for (i = 0 ; i < tnl->vtx.prim_count ; i++) {
-      struct tnl_prim *prim = &tnl->vtx.prim[i];
-      _mesa_debug(NULL, "   prim %d: %s %d..%d %s %s\n",
-		  i, 
-		  _mesa_lookup_enum_by_nr(prim->mode & PRIM_MODE_MASK),
-		  prim->start, 
-		  prim->start + prim->count,
-		  (prim->mode & PRIM_BEGIN) ? "BEGIN" : "(wrap)",
-		  (prim->mode & PRIM_END) ? "END" : "(wrap)");
-   }
-}
-
-GLboolean *_tnl_translate_edgeflag( GLcontext *ctx, const GLfloat *data,
-				    GLuint count, GLuint stride )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLboolean *ef = tnl->vtx.edgeflag_tmp;
-   GLuint i;
-
-   if (!ef) 
-      ef = tnl->vtx.edgeflag_tmp = (GLboolean *) MALLOC( tnl->vb.Size );
-   
-   for (i = 0 ; i < count ; i++, data += stride)
-      ef[i] = (data[0] == 1.0);
-
-   return ef;
-}
-
-
-GLboolean *_tnl_import_current_edgeflag( GLcontext *ctx,
-					 GLuint count )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLboolean *ef = tnl->vtx.edgeflag_tmp;
-   GLboolean tmp = ctx->Current.EdgeFlag;
-   GLuint i;
-
-   if (!ef) 
-      ef = tnl->vtx.edgeflag_tmp = (GLboolean *) MALLOC( tnl->vb.Size );
-
-   for (i = 0 ; i < count ; i++)
-      ef[i] = tmp;
-
-   return ef;
-}
-
-static INLINE GLint get_size( const GLfloat *f )
-{
-   if (f[3] != 1.0) return 4;
-   if (f[2] != 0.0) return 3;
-   return 2;
-}
-
-/* Some nasty stuff still hanging on here.  
- *
- * TODO - remove VB->NormalPtr, etc and just use the AttrPtr's.
- */
-static void _tnl_vb_bind_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->vtx_inputs;
-   GLfloat *data = tnl->vtx.buffer;
-   GLuint count = tnl->vtx.initial_counter - tnl->vtx.counter;
-   GLuint attr, i;
-
-#undef DEBUG_VTX
-
-#ifdef DEBUG_VTX
-   fprintf(stderr, "_tnl_vb_bind_vtx(): %d verts %d vertsize\n",
-		  count, tnl->vtx.vertex_size);
-#endif
-
-
-   /* Setup constant data in the VB.
-    */
-   VB->Count = count;
-   VB->Primitive = tnl->vtx.prim;
-   VB->PrimitiveCount = tnl->vtx.prim_count;
-   VB->Elts = NULL;
-   VB->NormalLengthPtr = NULL;
-
-   for (attr = 0; attr <= _TNL_ATTRIB_EDGEFLAG ; attr++) {
-      if (tnl->vtx.attrsz[attr]) {
-	 tmp->Attribs[attr].count = count;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) data;
-	 tmp->Attribs[attr].start = data;
-	 tmp->Attribs[attr].size = tnl->vtx.attrsz[attr];
-	 tmp->Attribs[attr].stride = tnl->vtx.vertex_size * sizeof(GLfloat);
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-	 data += tnl->vtx.attrsz[attr];
-      }
-      else {
-/* 	 VB->AttribPtr[attr] = &tnl->current.Attribs[attr]; */
-
-
-	 tmp->Attribs[attr].count = 1;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) tnl->vtx.current[attr];
-	 tmp->Attribs[attr].start = tnl->vtx.current[attr];
-	 tmp->Attribs[attr].size = get_size( tnl->vtx.current[attr] );
-	 tmp->Attribs[attr].stride = 0;
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-      }
-   }
-
-   
-   /* Copy and translate EdgeFlag to a contiguous array of GLbooleans
-    */
-   if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL) {
-      if (tnl->vtx.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-	 VB->EdgeFlag = _tnl_translate_edgeflag( ctx, data, count,
-						 tnl->vtx.vertex_size );
-	 data++;
-      }
-      else 
-	 VB->EdgeFlag = _tnl_import_current_edgeflag( ctx, count );
-   }
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
-
-
-/*
- * NOTE: Need to have calculated primitives by this point -- do it on the fly.
- * NOTE: Old 'parity' issue is gone.
- */
-static GLuint _tnl_copy_vertices( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT( ctx );
-   GLuint nr = tnl->vtx.prim[tnl->vtx.prim_count-1].count;
-   GLuint ovf, i;
-   GLuint sz = tnl->vtx.vertex_size;
-   GLfloat *dst = tnl->vtx.copied.buffer;
-   GLfloat *src = (tnl->vtx.buffer + 
-		   tnl->vtx.prim[tnl->vtx.prim_count-1].start * 
-		   tnl->vtx.vertex_size);
-
-
-   switch( ctx->Driver.CurrentExecPrimitive )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      else {
-	 _mesa_memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 1;
-      }
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 return 1;
-      } else {
-	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 _mesa_memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      /* no parity issue, but need to make sure the tri is not drawn twice */
-      if (nr & 1) {
-	 tnl->vtx.prim[tnl->vtx.prim_count-1].count--;
-      }
-      /* fallthrough */
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      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) );
-      return i;
-   case PRIM_OUTSIDE_BEGIN_END:
-      return 0;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-
-/**
- * Execute the buffer and save copied verts.
- */
-void _tnl_flush_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint vertex_count = tnl->vtx.initial_counter - tnl->vtx.counter;
-
-   if (0)
-      _tnl_print_vtx( ctx );
-
-   if (tnl->vtx.prim_count && vertex_count) {
-
-      tnl->vtx.copied.nr = _tnl_copy_vertices( ctx ); 
-
-      if (tnl->vtx.copied.nr != vertex_count) {
-	 if (ctx->NewState)
-	    _mesa_update_state( ctx );
-      
-	 _tnl_vb_bind_vtx( ctx );
-
-	 tnl->Driver.RunPipeline( ctx );
-      }
-   }
-
-   tnl->vtx.prim_count = 0;
-   tnl->vtx.counter = tnl->vtx.initial_counter;
-   tnl->vtx.vbptr = tnl->vtx.buffer;
-}
diff --git a/src/mesa/tnl/t_vtx_generic.c b/src/mesa/tnl/t_vtx_generic.c
deleted file mode 100644
index 28de445ed4..0000000000
--- a/src/mesa/tnl/t_vtx_generic.c
+++ /dev/null
@@ -1,615 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 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
-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 "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-
-
-/* Versions of all the entrypoints for situations where codegen isn't
- * available.  
- *
- * Note: Only one size for each attribute may be active at once.
- * Eg. if Color3f is installed/active, then Color4f may not be, even
- * if the vertex actually contains 4 color coordinates.  This is
- * because the 3f version won't otherwise set color[3] to 1.0 -- this
- * is the job of the chooser function when switching between Color4f
- * and Color3f.
- */
-#define ATTRFV( ATTR, N )				\
-static void attrib_##ATTR##_##N( const GLfloat *v )	\
-{							\
-   GET_CURRENT_CONTEXT( ctx );				\
-   TNLcontext *tnl = TNL_CONTEXT(ctx);			\
-							\
-   if ((ATTR) == 0) {					\
-      GLuint i;						\
-							\
-      if (N>0) tnl->vtx.vbptr[0] = v[0];		\
-      if (N>1) tnl->vtx.vbptr[1] = v[1];		\
-      if (N>2) tnl->vtx.vbptr[2] = v[2];		\
-      if (N>3) tnl->vtx.vbptr[3] = v[3];		\
-							\
-      for (i = N; i < tnl->vtx.vertex_size; i++)	\
-	 tnl->vtx.vbptr[i] = tnl->vtx.vertex[i];	\
-							\
-      tnl->vtx.vbptr += tnl->vtx.vertex_size;		\
-							\
-      if (--tnl->vtx.counter == 0)			\
-	 _tnl_wrap_filled_vertex( ctx );		\
-   }							\
-   else {						\
-      GLfloat *dest = tnl->vtx.attrptr[ATTR];		\
-      if (N>0) dest[0] = v[0];				\
-      if (N>1) dest[1] = v[1];				\
-      if (N>2) dest[2] = v[2];				\
-      if (N>3) dest[3] = v[3];				\
-   }							\
-}
-
-#define INIT(TAB, ATTR)						\
-   TAB[ATTR][0] = attrib_##ATTR##_1;				\
-   TAB[ATTR][1] = attrib_##ATTR##_2;				\
-   TAB[ATTR][2] = attrib_##ATTR##_3;				\
-   TAB[ATTR][3] = attrib_##ATTR##_4;
-
-
-#define ATTRS( ATTRIB )				\
-   ATTRFV( ATTRIB, 1 )				\
-   ATTRFV( ATTRIB, 2 )				\
-   ATTRFV( ATTRIB, 3 )				\
-   ATTRFV( ATTRIB, 4 )			
-
-/* conventional attribs */
-ATTRS( 0 )
-ATTRS( 1 )
-ATTRS( 2 )
-ATTRS( 3 )
-ATTRS( 4 )
-ATTRS( 5 )
-ATTRS( 6 )
-ATTRS( 7 )
-ATTRS( 8 )
-ATTRS( 9 )
-ATTRS( 10 )
-ATTRS( 11 )
-ATTRS( 12 )
-ATTRS( 13 )
-ATTRS( 14 )
-ATTRS( 15 )
-
-/* generic attribs */
-ATTRS( 16 )
-ATTRS( 17 )
-ATTRS( 18 )
-ATTRS( 19 )
-ATTRS( 20 )
-ATTRS( 21 )
-ATTRS( 22 )
-ATTRS( 23 )
-ATTRS( 24 )
-ATTRS( 25 )
-ATTRS( 26 )
-ATTRS( 27 )
-ATTRS( 28 )
-ATTRS( 29 )
-ATTRS( 30 )
-ATTRS( 31 )
-
-
-void _tnl_generic_attr_table_init( tnl_attrfv_func (*tab)[4] )
-{
-   /* conventional attribs */
-   INIT( tab, 0 );
-   INIT( tab, 1 );
-   INIT( tab, 2 );
-   INIT( tab, 3 );
-   INIT( tab, 4 );
-   INIT( tab, 5 );
-   INIT( tab, 6 );
-   INIT( tab, 7 );
-   INIT( tab, 8 );
-   INIT( tab, 9 );
-   INIT( tab, 10 );
-   INIT( tab, 11 );
-   INIT( tab, 12 );
-   INIT( tab, 13 );
-   INIT( tab, 14 );
-   INIT( tab, 15 );
-
-   /* generic attribs */
-   INIT( tab, 16 );
-   INIT( tab, 17 );
-   INIT( tab, 18 );
-   INIT( tab, 19 );
-   INIT( tab, 20 );
-   INIT( tab, 21 );
-   INIT( tab, 22 );
-   INIT( tab, 23 );
-   INIT( tab, 24 );
-   INIT( tab, 25 );
-   INIT( tab, 26 );
-   INIT( tab, 27 );
-   INIT( tab, 28 );
-   INIT( tab, 29 );
-   INIT( tab, 30 );
-   INIT( tab, 31 );
-}
-
-/* These can be made efficient with codegen.  Further, by adding more
- * logic to do_choose(), the double-dispatch for legacy entrypoints
- * like glVertex3f() can be removed.
- */
-#define DISPATCH_ATTRFV( ATTR, COUNT, P )	\
-do {						\
-   GET_CURRENT_CONTEXT( ctx ); 			\
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 		\
-   tnl->vtx.tabfv[ATTR][COUNT-1]( P );		\
-} while (0)
-
-#define DISPATCH_ATTR1FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 1, V )
-#define DISPATCH_ATTR2FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 2, V )
-#define DISPATCH_ATTR3FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 3, V )
-#define DISPATCH_ATTR4FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 4, V )
-
-#define DISPATCH_ATTR1F( ATTR, S ) DISPATCH_ATTRFV( ATTR, 1, &(S) )
-
-#if defined(USE_X86_ASM) && 0 /* will break register calling convention */
-/* Naughty cheat:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) DISPATCH_ATTRFV( ATTR, 2, &(S) )
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) DISPATCH_ATTRFV( ATTR, 3, &(S) )
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q ) DISPATCH_ATTRFV( ATTR, 4, &(S) )
-#else
-/* Safe:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) 		\
-do { 						\
-   GLfloat v[2]; 				\
-   v[0] = S; v[1] = T;				\
-   DISPATCH_ATTR2FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) 		\
-do { 						\
-   GLfloat v[3]; 				\
-   v[0] = S; v[1] = T; v[2] = R;		\
-   DISPATCH_ATTR3FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q )	\
-do { 						\
-   GLfloat v[4]; 				\
-   v[0] = S; v[1] = T; v[2] = R; v[3] = Q;	\
-   DISPATCH_ATTR4FV( ATTR, v );			\
-} while (0)
-#endif
-
-
-static void GLAPIENTRY _tnl_Vertex2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_POS, x, y );
-}
-
-static void GLAPIENTRY _tnl_Vertex2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_POS, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Vertex3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_Vertex4f( GLfloat x, GLfloat y, GLfloat z, 
-				      GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_POS, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_Vertex4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord1f( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_TEX0, x );
-}
-
-static void GLAPIENTRY _tnl_TexCoord1fv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_TEX0, x, y );
-}
-
-static void GLAPIENTRY _tnl_TexCoord2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_TEX0, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_TexCoord3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord4f( GLfloat x, GLfloat y, GLfloat z,
-					GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_TEX0, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_TexCoord4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_Normal3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_NORMAL, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Normal3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_NORMAL, v );
-}
-
-static void GLAPIENTRY _tnl_FogCoordfEXT( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_FOG, x );
-}
-
-static void GLAPIENTRY _tnl_FogCoordfvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_FOG, v );
-}
-
-static void GLAPIENTRY _tnl_Color3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR0, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Color3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _tnl_Color4f( GLfloat x, GLfloat y, GLfloat z, 
-				     GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_COLOR0, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_Color4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _tnl_Indexf( GLfloat v )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_COLOR_INDEX, v );
-}
-
-static void GLAPIENTRY _tnl_Indexfv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_COLOR_INDEX, v );
-}
-
-static void GLAPIENTRY _tnl_SecondaryColor3fEXT( GLfloat x, GLfloat y, 
-						 GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR1, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_SecondaryColor3fvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR1, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord1f( GLenum target, GLfloat x  )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1F( attr, x );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord1fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord2f( GLenum target, GLfloat x, GLfloat y )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2F( attr, x, y );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord2fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord3f( GLenum target, GLfloat x, 
-					     GLfloat y, GLfloat z)
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3F( attr, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord3fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord4f( GLenum target, GLfloat x, 
-					     GLfloat y, GLfloat z,
-					     GLfloat w )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4F( attr, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord4fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4FV( attr, v );
-}
-
-
-/**
- * GL_NV_vertex_program Vertex Attributes
- * Note that these attributes DO alias the conventional attributes.
- * Also, calling glVertexAttribNV(0, xxx) is equivalent to glVertex(xxx).
- */
-
-static void GLAPIENTRY _tnl_VertexAttrib1fNV( GLuint index, GLfloat x )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR1F( index, x );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib1fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR1FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fNV(GLuint index, GLfloat x, GLfloat y)
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR2F( index, x, y );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR2FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fNV( GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR3F( index, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR3FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fNV( GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z,
-					      GLfloat w )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR4F( index, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR4FV( index, v );
-}
-
-
-
-/**
- * GL_ARB_vertex_program Vertex Attributes
- * Note that these attributes do NOT alias the conventional attributes.
- * Also, calling glVertexAttribARB(0, xxx) is equivalent to glVertex(xxx).
- */
-
-static void GLAPIENTRY _tnl_VertexAttrib1fARB( GLuint index, GLfloat x )
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR1F( index, x );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib1fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR1FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fARB( GLuint index, GLfloat x, 
-					      GLfloat y )
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR2F( index, x, y );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR2FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fARB(GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR3F( index, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR3FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fARB(GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z, GLfloat w)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR4F( index, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR4FV( index, v );
-}
-
-
-/* Install the generic versions of the 2nd level dispatch
- * functions.  Some of these have a codegen alternative.
- */
-void _tnl_generic_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   vfmt->Color3f = _tnl_Color3f;
-   vfmt->Color3fv = _tnl_Color3fv;
-   vfmt->Color4f = _tnl_Color4f;
-   vfmt->Color4fv = _tnl_Color4fv;
-   vfmt->Indexf = _tnl_Indexf;
-   vfmt->Indexfv = _tnl_Indexfv;
-   vfmt->FogCoordfEXT = _tnl_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = _tnl_FogCoordfvEXT;
-   vfmt->MultiTexCoord1fARB = _tnl_MultiTexCoord1f;
-   vfmt->MultiTexCoord1fvARB = _tnl_MultiTexCoord1fv;
-   vfmt->MultiTexCoord2fARB = _tnl_MultiTexCoord2f;
-   vfmt->MultiTexCoord2fvARB = _tnl_MultiTexCoord2fv;
-   vfmt->MultiTexCoord3fARB = _tnl_MultiTexCoord3f;
-   vfmt->MultiTexCoord3fvARB = _tnl_MultiTexCoord3fv;
-   vfmt->MultiTexCoord4fARB = _tnl_MultiTexCoord4f;
-   vfmt->MultiTexCoord4fvARB = _tnl_MultiTexCoord4fv;
-   vfmt->Normal3f = _tnl_Normal3f;
-   vfmt->Normal3fv = _tnl_Normal3fv;
-   vfmt->SecondaryColor3fEXT = _tnl_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = _tnl_SecondaryColor3fvEXT;
-   vfmt->TexCoord1f = _tnl_TexCoord1f;
-   vfmt->TexCoord1fv = _tnl_TexCoord1fv;
-   vfmt->TexCoord2f = _tnl_TexCoord2f;
-   vfmt->TexCoord2fv = _tnl_TexCoord2fv;
-   vfmt->TexCoord3f = _tnl_TexCoord3f;
-   vfmt->TexCoord3fv = _tnl_TexCoord3fv;
-   vfmt->TexCoord4f = _tnl_TexCoord4f;
-   vfmt->TexCoord4fv = _tnl_TexCoord4fv;
-   vfmt->Vertex2f = _tnl_Vertex2f;
-   vfmt->Vertex2fv = _tnl_Vertex2fv;
-   vfmt->Vertex3f = _tnl_Vertex3f;
-   vfmt->Vertex3fv = _tnl_Vertex3fv;
-   vfmt->Vertex4f = _tnl_Vertex4f;
-   vfmt->Vertex4fv = _tnl_Vertex4fv;
-   vfmt->VertexAttrib1fNV = _tnl_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = _tnl_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV = _tnl_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = _tnl_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV = _tnl_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = _tnl_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV = _tnl_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = _tnl_VertexAttrib4fvNV;
-   vfmt->VertexAttrib1fARB = _tnl_VertexAttrib1fARB;
-   vfmt->VertexAttrib1fvARB = _tnl_VertexAttrib1fvARB;
-   vfmt->VertexAttrib2fARB = _tnl_VertexAttrib2fARB;
-   vfmt->VertexAttrib2fvARB = _tnl_VertexAttrib2fvARB;
-   vfmt->VertexAttrib3fARB = _tnl_VertexAttrib3fARB;
-   vfmt->VertexAttrib3fvARB = _tnl_VertexAttrib3fvARB;
-   vfmt->VertexAttrib4fARB = _tnl_VertexAttrib4fARB;
-   vfmt->VertexAttrib4fvARB = _tnl_VertexAttrib4fvARB;
-}
diff --git a/src/mesa/tnl/t_vtx_x86.c b/src/mesa/tnl/t_vtx_x86.c
deleted file mode 100644
index 99a0550932..0000000000
--- a/src/mesa/tnl/t_vtx_x86.c
+++ /dev/null
@@ -1,396 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 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
-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>
- *   Daniel Borca <dborca@yahoo.com>
- */
-
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-#include "simple_list.h"
-
-#if defined(USE_X86_ASM) && !defined(HAVE_NONSTANDARD_GLAPIENTRY)
-
-#define EXTERN( FUNC )		\
-extern const char FUNC[];	\
-extern const char FUNC##_end[]
-
-EXTERN( _tnl_x86_Attribute1fv );
-EXTERN( _tnl_x86_Attribute2fv );
-EXTERN( _tnl_x86_Attribute3fv );
-EXTERN( _tnl_x86_Attribute4fv );
-EXTERN( _tnl_x86_Vertex1fv );
-EXTERN( _tnl_x86_Vertex2fv );
-EXTERN( _tnl_x86_Vertex3fv );
-EXTERN( _tnl_x86_Vertex4fv );
-
-EXTERN( _tnl_x86_dispatch_attrf1 );
-EXTERN( _tnl_x86_dispatch_attrf2 );
-EXTERN( _tnl_x86_dispatch_attrf3 );
-EXTERN( _tnl_x86_dispatch_attrf4 );
-EXTERN( _tnl_x86_dispatch_attrfv );
-EXTERN( _tnl_x86_dispatch_multitexcoordf1 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf2 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf3 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf4 );
-EXTERN( _tnl_x86_dispatch_multitexcoordfv );
-EXTERN( _tnl_x86_dispatch_vertexattribf1 );
-EXTERN( _tnl_x86_dispatch_vertexattribf2 );
-EXTERN( _tnl_x86_dispatch_vertexattribf3 );
-EXTERN( _tnl_x86_dispatch_vertexattribf4 );
-EXTERN( _tnl_x86_dispatch_vertexattribfv );
-
-EXTERN( _tnl_x86_choose_fv );
-
-
-#define DONT_KNOW_OFFSETS 1
-
-
-#define DFN( FUNC, CACHE, KEY )				\
-   struct _tnl_dynfn *dfn = MALLOC_STRUCT( _tnl_dynfn );\
-   const char *start = FUNC;				\
-   const char *end = FUNC##_end;			\
-   int offset = 0;               			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key = KEY;					\
-   dfn->code = ALIGN_MALLOC( end - start, 16 );		\
-   _mesa_memcpy (dfn->code, start, end - start)
-
-
-
-#undef DEBUG_VTX
-
-#ifdef DEBUG_VTX
-#define FIXUP_PRINTF( offset, NEWVAL ) \
-      fprintf(stderr, "%s/%d: offset %d, new value: 0x%x\n", __FILE__, __LINE__, offset, (int)(NEWVAL))
-#define FIXUPREL_PRINTF( offset, NEWVAL, CODE ) \
-      fprintf(stderr, "%s/%d: offset %d, new value: 0x%x\n", __FILE__, __LINE__, offset, (int)(NEWVAL) - ((int)(CODE)+offset) - 4)
-#else
-#define FIXUP_PRINTF( offset, NEWVAL )
-#define FIXUPREL_PRINTF( offset, NEWVAL, CODE )
-#endif
-
-#define FIXUP( CODE, KNOWN_OFFSET, CHECKVAL, NEWVAL )	\
-do {							\
-   GLint subst = 0x10101010 + CHECKVAL;			\
-							\
-   if (DONT_KNOW_OFFSETS) {				\
-      while (*(int *)(CODE+offset) != subst) offset++;	\
-      *(int *)(CODE+offset) = (int)(NEWVAL);		\
-      FIXUP_PRINTF(offset, NEWVAL);			\
-      offset += 4;					\
-   }							\
-   else {						\
-      int *icode = (int *)(CODE+KNOWN_OFFSET);		\
-      assert (*icode == subst);				\
-      *icode = (int)NEWVAL;				\
-   }							\
-} while (0)
-
-
-
-#define FIXUPREL( CODE, KNOWN_OFFSET, CHECKVAL, NEWVAL )\
-do {							\
-   GLint subst = 0x10101010 + CHECKVAL;			\
-							\
-   if (DONT_KNOW_OFFSETS) {				\
-      while (*(int *)(CODE+offset) != subst) offset++;	\
-      *(int *)(CODE+offset) = (int)(NEWVAL) - ((int)(CODE)+offset) - 4; \
-      FIXUPREL_PRINTF(offset, NEWVAL, CODE);		\
-      offset += 4;					\
-   }							\
-   else {						\
-      int *icode = (int *)(CODE+KNOWN_OFFSET);		\
-      assert (*icode == subst);				\
-      *icode = (int)(NEWVAL) - (int)(icode) - 4;	\
-   }							\
-} while (0)
-
-
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-static struct _tnl_dynfn *makeX86Vertex1fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex1fv, tnl->vtx.cache.Vertex[1-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 1);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[1]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Vertex2fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex2fv, tnl->vtx.cache.Vertex[2-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 2);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[2]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Vertex3fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (vertex_size) {
-      default: {
-         DFN ( _tnl_x86_Vertex3fv, tnl->vtx.cache.Vertex[3-1], vertex_size );
-
-         FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-         FIXUP(dfn->code, 0, 1, vertex_size - 3);
-         FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[3]);
-         FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-         FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-         FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-         FIXUP(dfn->code, 0, 4, (int)ctx);
-         FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-         return dfn;
-      }
-   }
-}
-
-static struct _tnl_dynfn *makeX86Vertex4fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex4fv, tnl->vtx.cache.Vertex[4-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 4);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[4]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-
-static struct _tnl_dynfn *makeX86Attribute1fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute1fv, tnl->vtx.cache.Attribute[1-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute2fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute2fv, tnl->vtx.cache.Attribute[2-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute3fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute3fv, tnl->vtx.cache.Attribute[3-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-   FIXUP(dfn->code, 0, 2, 8+dest);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute4fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute4fv, tnl->vtx.cache.Attribute[4-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-   FIXUP(dfn->code, 0, 2, 8+dest);
-   FIXUP(dfn->code, 0, 3, 12+dest);
-
-   return dfn;
-}
-
-
-void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen )
-{
-   gen->Vertex[0] = makeX86Vertex1fv;
-   gen->Vertex[1] = makeX86Vertex2fv;
-   gen->Vertex[2] = makeX86Vertex3fv;
-   gen->Vertex[3] = makeX86Vertex4fv;
-   gen->Attribute[0] = makeX86Attribute1fv;
-   gen->Attribute[1] = makeX86Attribute2fv;
-   gen->Attribute[2] = makeX86Attribute3fv;
-   gen->Attribute[3] = makeX86Attribute4fv;
-}
-
-
-#define MKDISP(FUNC, SIZE, ATTR, WARP)					\
-do {									\
-   char *code;								\
-   const char *start = WARP;						\
-   const char *end = WARP##_end;					\
-   int offset = 0;							\
-   code = ALIGN_MALLOC( end - start, 16 );				\
-   _mesa_memcpy (code, start, end - start);					\
-   FIXUP(code, 0, 0, (int)&(TNL_CONTEXT(ctx)->vtx.tabfv[ATTR][SIZE-1]));\
-   *(void **)&vfmt->FUNC = code;					\
-} while (0)
-
-
-/* Install the codegen'ed versions of the 2nd level dispatch
- * functions.  We should keep a list and free them in the end...
- */
-void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   MKDISP(Color3f,             3, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrf3);
-   MKDISP(Color3fv,            3, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrfv);
-   MKDISP(Color4f,             4, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrf4);
-   MKDISP(Color4fv,            4, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrfv);
-   MKDISP(FogCoordfEXT,        1, _TNL_ATTRIB_FOG,    _tnl_x86_dispatch_attrf1);
-   MKDISP(FogCoordfvEXT,       1, _TNL_ATTRIB_FOG,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Normal3f,            3, _TNL_ATTRIB_NORMAL, _tnl_x86_dispatch_attrf3);
-   MKDISP(Normal3fv,           3, _TNL_ATTRIB_NORMAL, _tnl_x86_dispatch_attrfv);
-   MKDISP(SecondaryColor3fEXT, 3, _TNL_ATTRIB_COLOR1, _tnl_x86_dispatch_attrf3);
-   MKDISP(SecondaryColor3fvEXT,3, _TNL_ATTRIB_COLOR1, _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord1f,          1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf1);
-   MKDISP(TexCoord1fv,         1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord2f,          2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf2);
-   MKDISP(TexCoord2fv,         2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord3f,          3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf3);
-   MKDISP(TexCoord3fv,         3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord4f,          4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf4);
-   MKDISP(TexCoord4fv,         4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex2f,            2, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf2);
-   MKDISP(Vertex2fv,           2, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex3f,            3, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf3);
-   MKDISP(Vertex3fv,           3, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex4f,            4, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf4);
-   MKDISP(Vertex4fv,           4, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-
-   MKDISP(MultiTexCoord1fARB,  1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf1);
-   MKDISP(MultiTexCoord1fvARB, 1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord2fARB,  2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf2);
-   MKDISP(MultiTexCoord2fvARB, 2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord3fARB,  3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf3);
-   MKDISP(MultiTexCoord3fvARB, 3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord4fARB,  4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf4);
-   MKDISP(MultiTexCoord4fvARB, 4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-
-   MKDISP(VertexAttrib1fNV,    1, 0,                  _tnl_x86_dispatch_vertexattribf1);
-   MKDISP(VertexAttrib1fvNV,   1, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib2fNV,    2, 0,                  _tnl_x86_dispatch_vertexattribf2);
-   MKDISP(VertexAttrib2fvNV,   2, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib3fNV,    3, 0,                  _tnl_x86_dispatch_vertexattribf3);
-   MKDISP(VertexAttrib3fvNV,   3, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib4fNV,    4, 0,                  _tnl_x86_dispatch_vertexattribf4);
-   MKDISP(VertexAttrib4fvNV,   4, 0,                  _tnl_x86_dispatch_vertexattribfv);
-}
-
-
-/* Install the codegen'ed choosers.
- * We should keep a list and free them in the end...
- */
-void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-		       tnl_attrfv_func (*do_choose)( GLuint attr,
-						 GLuint sz ))
-{
-   int attr, size;
-
-   for (attr = 0; attr < _TNL_MAX_ATTR_CODEGEN; attr++) {
-      for (size = 0; size < 4; size++) {
-         char *code;
-         const char *start = _tnl_x86_choose_fv;
-         const char *end = _tnl_x86_choose_fv_end;
-         int offset = 0;
-         code = ALIGN_MALLOC( end - start, 16 );
-         _mesa_memcpy (code, start, end - start);
-         FIXUP(code, 0, 0, attr);
-         FIXUP(code, 0, 1, size + 1);
-         FIXUPREL(code, 0, 2, do_choose);
-         choose[attr][size] = (tnl_attrfv_func)code;
-      }
-   }
-}
-
-#else 
-
-void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen )
-{
-   (void) gen;
-}
-
-
-void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx )
-{
-   (void) ctx;
-}
-
-
-void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-		       tnl_attrfv_func (*do_choose)( GLuint attr,
-						 GLuint sz ))
-{
-   (void) choose;
-   (void) do_choose;
-}
-
-#endif
diff --git a/src/mesa/tnl/t_vtx_x86_gcc.S b/src/mesa/tnl/t_vtx_x86_gcc.S
deleted file mode 100644
index 93753be759..0000000000
--- a/src/mesa/tnl/t_vtx_x86_gcc.S
+++ /dev/null
@@ -1,561 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 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
-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>
- *   Daniel Borca <dborca@yahoo.com>
- */
-
-#if defined (__DJGPP__) || defined (__MINGW32__) || defined (__CYGWIN__)
-#define GLOBL( x )	\
-.globl _##x;		\
-_##x:
-#else  /* !defined (__DJGPP__) && !defined (__MINGW32__) && !defined (__CYGWIN__) */
-#define GLOBL( x )	\
-.globl x;		\
-x:
-#endif /* !defined (__DJGPP__) && !defined (__MINGW32__) && !defined (__CYGWIN__) */
-
-
-#if !defined (STDCALL_API)
-#define RETCLEAN( x )	ret
-#else
-#define RETCLEAN( x )	ret $x
-#endif
-
-
-#define _JMP(x)		\
-.byte 0xe9;		\
-.long x
-
-#define _CALL(x)	\
-.byte 0xe8;		\
-.long x
-
-
-/* Someone who knew a lot about this sort of thing would use this
- * macro to note current offsets, etc in a special region of the
- * object file & just make everything work out neat.  I don't know
- * enough to do that...
- */
-
-#define SUBST( x ) (0x10101010 + x)
-
-
-.data
-
-
-/* [dBorca] TODO
- * Unfold functions for each vertex size?
- * Build super-specialized SSE versions?
- *
- * There is a trick in Vertex*fv: under certain conditions,
- * we tail to _tnl_wrap_filled_vertex(ctx). This means that
- * if Vertex*fv is STDCALL, then _tnl_wrap_filled_vertex must
- * be STDCALL as well, because (GLcontext *) and (GLfloat *)
- * have the same size.
- */
-.align 4
-GLOBL ( _tnl_x86_Vertex1fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* 0x0 --> tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	addl	$4, %edi	/* tnl->vtx.vbptr += 1 */
-	movl	$SUBST(1), %ecx	/* 0x1 --> (tnl->vtx.vertex_size - 1) */
-	movl	$SUBST(2), %esi	/* 0x2 --> (tnl->vtx.vertex + 1) */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* 0x0 --> tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* 0x3 --> counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* 0x3 --> counter */
-	je	.0		/* if (counter == 0) goto .0 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.0:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex1fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex2fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	addl	$8, %edi	/* tnl->vtx.vbptr += 2 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 2 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 2 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.1		/* if (counter == 0) goto .1 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.1:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex2fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex3fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	8(%ecx), %esi	/* load v[2] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	movl	%esi, 8(%edi)	/* tnl->vtx.vbptr[2] = v[2] */
-	addl	$12, %edi	/* tnl->vtx.vbptr += 3 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 3 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 3 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.2		/* if (counter == 0) goto .2 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.2:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex3fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex4fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	8(%ecx), %esi	/* load v[2] */
-	movl	12(%ecx), %ecx	/* load v[3] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	movl	%esi, 8(%edi)	/* tnl->vtx.vbptr[2] = v[2] */
-	movl	%ecx, 12(%edi)	/* tnl->vtx.vbptr[3] = v[3] */
-	addl	$16, %edi	/* tnl->vtx.vbptr += 4 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 4 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 4 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.3		/* if (counter == 0) goto .3 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.3:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex4fv_end )
-
-
-/**
- * Generic handlers for vector format data. 
- */
-GLOBL( _tnl_x86_Attribute1fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute1fv_end )
-
-GLOBL( _tnl_x86_Attribute2fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute2fv_end )
-
-GLOBL( _tnl_x86_Attribute3fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	8(%ecx), %ecx	/* load v[2] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	movl	%ecx, SUBST(2)	/* store v[2] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute3fv_end )
-
-GLOBL( _tnl_x86_Attribute4fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	movl	8(%ecx), %eax	/* load v[2] */
-	movl	12(%ecx), %edx	/* load v[3] */
-	movl	%eax, SUBST(2)	/* store v[2] to current vertex */
-	movl	%edx, SUBST(3)	/* store v[3] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute4fv_end )
-
-
-/* Choosers:
- *
- * Must generate all of these ahead of first usage.  Generate at
- * compile-time?
- */
-GLOBL( _tnl_x86_choose_fv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	$SUBST(0), (%esp)	/* arg 0 - attrib */
-	movl	$SUBST(1), 4(%esp)	/* arg 1 - N */
-	_CALL	(SUBST(2))		/* call do_choose */
-	add	$12, %esp		/* tear down stack frame */
-	jmp	*%eax			/* jump to new func */
-GLOBL ( _tnl_x86_choose_fv_end )
-
-
-/* FIRST LEVEL FUNCTIONS -- these are plugged directly into GL dispatch.
- *
- * In the 1st level dispatch functions, switch to a different
- * calling convention -- (const GLfloat *v) in %ecx.
- *
- * As with regular (x86) dispatch, don't create a new stack frame -
- * just let the 'ret' in the dispatched function return straight
- * back to the original caller.
- *
- * Vertex/Normal/Color, etc: the address of the function pointer
- * is known at codegen time.
- */
-
-/* Unfortunately, have to play with the stack in the non-fv case:
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_attrf1 )
-GLOBL( _tnl_x86_dispatch_attrf2 )
-GLOBL( _tnl_x86_dispatch_attrf3 )
-GLOBL( _tnl_x86_dispatch_attrf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$12, %esp	/* tear down frame */
-	ret			/* return */
-GLOBL( _tnl_x86_dispatch_attrf4_end )
-GLOBL( _tnl_x86_dispatch_attrf3_end )
-GLOBL( _tnl_x86_dispatch_attrf2_end )
-GLOBL( _tnl_x86_dispatch_attrf1_end )
-
-#else  /* defined(STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_attrf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$4		/* return */
-GLOBL( _tnl_x86_dispatch_attrf1_end )
-
-GLOBL( _tnl_x86_dispatch_attrf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_attrf2_end )
-
-GLOBL( _tnl_x86_dispatch_attrf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_attrf3_end )
-
-GLOBL( _tnl_x86_dispatch_attrf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_attrf4_end )
-#endif /* defined(STDCALL_API) */
-
-/* The fv case is simpler:
- */
-GLOBL( _tnl_x86_dispatch_attrfv )
-	jmp	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-GLOBL( _tnl_x86_dispatch_attrfv_end )
-
-
-/* MultiTexcoord: the address of the function pointer must be
- * calculated, but can use the index argument slot to hold 'v', and
- * avoid setting up a new stack frame.
- *
- * [dBorca]
- * right, this would be the preferred approach, but gcc does not
- * clean up the stack after each function call when optimizing (-fdefer-pop);
- * can it make assumptions about what's already on the stack?  I dunno,
- * but in this case, we can't mess with the caller's stack frame, and
- * we must use a model like `_x86_dispatch_attrfv' above.  Caveat emptor!
- */
-
-/* Also, will only need a maximum of four of each of these per context:
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_multitexcoordf1 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf2 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf3 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf4 )
-	movl	4(%esp), %ecx
-	leal	8(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, 4(%esp)
-	sall	$4, %ecx
-	jmp	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-GLOBL( _tnl_x86_dispatch_multitexcoordf4_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf3_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf2_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf1_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordfv )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, 4(%esp)
-	sall	$4, %ecx
-	jmp	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
-
-#else  /* defined (STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf1_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf2_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf3_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$20		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf4_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordfv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	movl	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
-#endif /* defined (STDCALL_API) */
-
-
-/* VertexAttrib: the address of the function pointer must be
- * calculated.
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_vertexattribf1 )
-GLOBL( _tnl_x86_dispatch_vertexattribf2 )
-GLOBL( _tnl_x86_dispatch_vertexattribf3 )
-GLOBL( _tnl_x86_dispatch_vertexattribf4 )
-	movl	4(%esp), %eax
-	cmpl	$16, %eax
-	jb	.8		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.8:
-	leal	8(%esp), %ecx	/* calculate 'v' */
-	movl	%ecx, 4(%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	jmp	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-GLOBL( _tnl_x86_dispatch_vertexattribf4_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf3_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf2_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf1_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribfv )
-	movl	4(%esp), %eax
-	cmpl	$16, %eax
-	jb	.9		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.9:
-	movl	8(%esp), %ecx	/* load 'v' */
-	movl	%ecx, 4(%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	jmp	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
-
-#else  /* defined (STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_vertexattribf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.81		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.81:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf1_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.82		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.82:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf2_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.83		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.83:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf3_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.84		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.84:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$20		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf4_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribfv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.9		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.9:
-	movl	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
-#endif /* defined (STDCALL_API) */
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index dab422429a..17cb30adc7 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -50,30 +50,14 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state );
  * dispatch and/or driver callbacks.
  */
 
-/* Restore just the ctx->Exec table:
- */
-extern void
-_tnl_wakeup_exec( GLcontext *ctx );
-
-/* Restore both ctx->Exec and ctx->Save:
- */
 extern void
-_tnl_wakeup_save_exec( GLcontext *ctx );
+_tnl_wakeup( GLcontext *ctx );
 
 /* Driver configuration options:
  */
 extern void
 _tnl_need_projected_coords( GLcontext *ctx, GLboolean flag );
 
-extern void
-_tnl_need_dlist_loopback( GLcontext *ctx, GLboolean flag );
-
-extern void
-_tnl_need_dlist_norm_lengths( GLcontext *ctx, GLboolean flag );
-
-extern void
-_tnl_isolate_materials( GLcontext *ctx, GLboolean flag );
-
 
 /* Control whether T&L does per-vertex fog
  */
@@ -86,4 +70,16 @@ _tnl_allow_pixel_fog( GLcontext *ctx, GLboolean value );
 extern void
 _tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program);
 
+struct _mesa_prim;
+struct _mesa_index_buffer;
+
+void
+_tnl_draw_prims( 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);
+
 #endif