patch to import Jon Smirl's work from Bitkeeper

18 files changed, 6994 insertions, 0 deletions

diff --git a/src/mesa/tnl_dd/imm/NOTES.imm b/src/mesa/tnl_dd/imm/NOTES.imm
new file mode 100644
index 0000000000..9b2bd65e4b
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/NOTES.imm
@@ -0,0 +1,112 @@
+
+NOTE:
+
+These files are incomplete.  They do not yet form a working
+implementation of hte concepts discused below.
+
+
+OVERVIEW
+
+The t_dd_imm_* files form a set of templates to produce driver -
+specific software tnl modules for a small subset of transformation and
+lighting states.
+
+The approach is quite different to the large vertex buffers of the
+src/tnl module, and is based around a cache of four recent vertices
+and a 'current' vertex which is updated directly from the Color,
+Normal, Texcoord, SecondaryColor and Fog entrypoints.
+
+The current vertex is actually a composite of the ctx->Current values
+and a partial hardware vertex maintained where the hardware values
+differ from those in ctx->Current.  For example, clamped color values
+are kept in the hardware vertex, while texcoords remain in
+ctx->Current.
+
+A crude diagram:
+
+       		+--------------+	+-------------------+
+		| ctx->Current |	| Current-HW-vertex |
+		+--------------+	+-------------------+
+		       \       	       	       	  /
+		       	\      			 /
+			 \     			/
+			  \    		       /
+			   ---------   --------
+			       	   |   |      
+			       	   v   v     
+	+--------+   +--------+  +--------+  +--------+ 
+	| vert-0 |   | vert-1 |  | vert-2 |  | vert-3 |		       	 
+	+--------+   +--------+  +--------+  +--------+		       	 
+				     |
+				     |
+				     v
+				     
+				    DMA
+
+
+Here values from ctx->Current and current-HW-vertex are merged to
+build vert-2, which is then dumped to hardware (DMA).  A state machine
+determines which vertex is built in turn, and how the vertices are
+used to present primitives to hardware.  These actions all occur
+during a call to Vertex{234}f{v}.
+
+Each vert-n includes clip coordinates and a clipmask in addition to
+the hardware (window) coordinates.  This information allows clipping
+to take place directly on these vertices, if need be.
+
+t_dd_imm_capi.h
+	
+	Color{34}{fub}{v}() implementations.  These update both
+	ctx->Current (unclamped float colors) and current-HW-vertex
+	with hardware-specific color values (typically unsigned
+	bytes).
+
+	When lighting is enabled, the functions from src/api_noop.c
+	should be used, which just update ctx->Current.  (The
+	current-hw-vertex colors are produced from lighting, which is
+	keyed to Normal3f).
+
+t_dd_imm_vb.c
+
+	Support functions for clipping and fallback.  See
+	t_dd_imm_primtmp.h.
+
+t_dd_imm_napi.c
+t_dd_imm_napi.h
+
+	Versions of Normal3f{v} to perform lighting with one or more
+	infinite lights.  Updates ctx->Current.Normal and the current
+	HW colors.
+
+	When lighting is disabled, use the functions from api_noop.c
+	instead.
+
+
+t_dd_imm_primtmp.h
+
+	State machine to control emission of vertices and primitives
+	to hardware.  Called indirectly from Vertex{234}f{v}.  Capable
+	of supporting hardware strip and fan primitives, and of
+	decomposing to discreet primitives for clipping or fallback,
+	or where the native primitive is unavailable.
+
+t_dd_imm_tapi.h
+
+	Implementations of TexCoord{v} and MultiTexCoord4f{v}ARB to
+	fire a callback when transitioning to projective texture.
+	Most drivers will need to change vertex format at this point,
+	some may need to enable a software rasterization fallback.
+
+t_dd_imm_vapi.h
+
+	Implementations of Vertex{234}f{v}.  These perform
+	transformation and cliptesting on their arguments, then jump
+	into the state machine implemented in primtmp.h.
+
+t_dd_imm_vertex.h
+
+	Support functions for building and clip-interpolating hardware
+	vertices.  Called from primtmp.h.
+
+
+Keith Whitwell, June 2001.
\ No newline at end of file
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_capi.h b/src/mesa/tnl_dd/imm/t_dd_imm_capi.h
new file mode 100644
index 0000000000..77f3a4ab49
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_capi.h
@@ -0,0 +1,419 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Gareth Hughes <gareth@valinux.com>
+ */
+
+/* Template for immediate mode color functions.
+ *
+ * FIXME: Floating-point color versions of these...
+ */
+
+
+static void TAG(Color3f)( GLfloat r, GLfloat g, GLfloat b )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = CLAMP(r, 0.0f, 1.0f);
+   CURRENT_COLOR( GCOMP ) = CLAMP(g, 0.0f, 1.0f);
+   CURRENT_COLOR( BCOMP ) = CLAMP(b, 0.0f, 1.0f);
+   CURRENT_COLOR( ACOMP ) = 1.0f;
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( RCOMP ), r );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( GCOMP ), g );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( BCOMP ), b );
+   CURRENT_COLOR( ACOMP ) = 255;
+#endif
+}
+
+static void TAG(Color3fv)( const GLfloat *v )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = CLAMP(v[0], 0.0f, 1.0f);
+   CURRENT_COLOR( GCOMP ) = CLAMP(v[1], 0.0f, 1.0f);
+   CURRENT_COLOR( BCOMP ) = CLAMP(v[2], 0.0f, 1.0f);
+   CURRENT_COLOR( ACOMP ) = 1.0f;
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( RCOMP ), v[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( GCOMP ), v[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( BCOMP ), v[2] );
+   CURRENT_COLOR( ACOMP ) = 255;
+#endif
+}
+
+static void TAG(Color3ub)( GLubyte r, GLubyte g, GLubyte b )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = UBYTE_TO_FLOAT( r );
+   CURRENT_COLOR( GCOMP ) = UBYTE_TO_FLOAT( g );
+   CURRENT_COLOR( BCOMP ) = UBYTE_TO_FLOAT( b );
+   CURRENT_COLOR( ACOMP ) = 1.0f;
+#else
+   CURRENT_COLOR( RCOMP ) = r;
+   CURRENT_COLOR( GCOMP ) = g;
+   CURRENT_COLOR( BCOMP ) = b;
+   CURRENT_COLOR( ACOMP ) = 255;
+#endif
+}
+
+static void TAG(Color3ubv)( const GLubyte *v )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = UBYTE_TO_FLOAT( v[0] );
+   CURRENT_COLOR( GCOMP ) = UBYTE_TO_FLOAT( v[1] );
+   CURRENT_COLOR( BCOMP ) = UBYTE_TO_FLOAT( v[2] );
+   CURRENT_COLOR( ACOMP ) = 1.0f;
+#else
+   CURRENT_COLOR( RCOMP ) = v[0];
+   CURRENT_COLOR( GCOMP ) = v[1];
+   CURRENT_COLOR( BCOMP ) = v[2];
+   CURRENT_COLOR( ACOMP ) = 255;
+#endif
+}
+
+static void TAG(Color4f)( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = CLAMP(r, 0.0f, 1.0f);
+   CURRENT_COLOR( GCOMP ) = CLAMP(g, 0.0f, 1.0f);
+   CURRENT_COLOR( BCOMP ) = CLAMP(b, 0.0f, 1.0f);
+   CURRENT_COLOR( ACOMP ) = CLAMP(a, 0.0f, 1.0f);
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( RCOMP ), r );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( GCOMP ), g );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( BCOMP ), b );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( ACOMP ), a );
+#endif
+}
+
+static void TAG(Color4fv)( const GLfloat *v )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = CLAMP(v[0], 0.0f, 1.0f);
+   CURRENT_COLOR( GCOMP ) = CLAMP(v[1], 0.0f, 1.0f);
+   CURRENT_COLOR( BCOMP ) = CLAMP(v[2], 0.0f, 1.0f);
+   CURRENT_COLOR( ACOMP ) = CLAMP(v[3], 0.0f, 1.0f);
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( RCOMP ), v[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( GCOMP ), v[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( BCOMP ), v[2] );
+   UNCLAMPED_FLOAT_TO_UBYTE( CURRENT_COLOR( ACOMP ), v[3] );
+#endif
+}
+
+static void TAG(Color4ub)( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = UBYTE_TO_FLOAT( r );
+   CURRENT_COLOR( GCOMP ) = UBYTE_TO_FLOAT( g );
+   CURRENT_COLOR( BCOMP ) = UBYTE_TO_FLOAT( b );
+   CURRENT_COLOR( ACOMP ) = UBYTE_TO_FLOAT( a );
+#else
+   CURRENT_COLOR( RCOMP ) = r;
+   CURRENT_COLOR( GCOMP ) = g;
+   CURRENT_COLOR( BCOMP ) = b;
+   CURRENT_COLOR( ACOMP ) = a;
+#endif
+}
+
+static void TAG(Color4ubv)( const GLubyte *v )
+{
+   GET_CURRENT;
+#ifdef COLOR_IS_FLOAT
+   CURRENT_COLOR( RCOMP ) = UBYTE_TO_FLOAT( v[0] );
+   CURRENT_COLOR( GCOMP ) = UBYTE_TO_FLOAT( v[1] );
+   CURRENT_COLOR( BCOMP ) = UBYTE_TO_FLOAT( v[2] );
+   CURRENT_COLOR( ACOMP ) = UBYTE_TO_FLOAT( v[3] );
+#else
+   CURRENT_COLOR( RCOMP ) = v[0];
+   CURRENT_COLOR( GCOMP ) = v[1];
+   CURRENT_COLOR( BCOMP ) = v[2];
+   CURRENT_COLOR( ACOMP ) = v[3];
+#endif
+}
+
+
+static void TAG(ColorMaterial3f)( GLfloat r, GLfloat g, GLfloat b )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = r;
+   color[1] = g;
+   color[2] = b;
+   color[3] = 1.0;
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial3fv)( const GLfloat *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = v[0];
+   color[1] = v[1];
+   color[2] = v[2];
+   color[3] = 1.0;
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial3ub)( GLubyte r, GLubyte g, GLubyte b )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = UBYTE_TO_FLOAT( r );
+   color[1] = UBYTE_TO_FLOAT( g );
+   color[2] = UBYTE_TO_FLOAT( b );
+   color[3] = 1.0;
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial3ubv)( const GLubyte *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = UBYTE_TO_FLOAT( v[0] );
+   color[1] = UBYTE_TO_FLOAT( v[1] );
+   color[2] = UBYTE_TO_FLOAT( v[2] );
+   color[3] = 1.0;
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial4f)( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = r;
+   color[1] = g;
+   color[2] = b;
+   color[3] = a;
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial4fv)( const GLfloat *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = v[0];
+   color[1] = v[1];
+   color[2] = v[2];
+   color[3] = v[3];
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial4ub)( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = UBYTE_TO_FLOAT( r );
+   color[1] = UBYTE_TO_FLOAT( g );
+   color[2] = UBYTE_TO_FLOAT( b );
+   color[3] = UBYTE_TO_FLOAT( a );
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+static void TAG(ColorMaterial4ubv)( const GLubyte *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLfloat *color = ctx->Current.Color;
+
+   color[0] = UBYTE_TO_FLOAT( v[0] );
+   color[1] = UBYTE_TO_FLOAT( v[1] );
+   color[2] = UBYTE_TO_FLOAT( v[2] );
+   color[3] = UBYTE_TO_FLOAT( v[3] );
+
+   _mesa_update_color_material( ctx, color );
+   RECALC_BASE_COLOR( ctx );
+}
+
+
+
+
+
+/* =============================================================
+ * Color chooser functions:
+ */
+
+static void TAG(choose_Color3f)( GLfloat r, GLfloat g, GLfloat b )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color3f = TAG(ColorMaterial3f);
+      } else {
+	 ctx->Exec->Color3f = _mesa_noop_Color3f;
+      }
+   } else {
+      ctx->Exec->Color3f = TAG(Color3f);
+   }
+   glColor3f( r, g, b );
+}
+
+static void TAG(choose_Color3fv)( const GLfloat *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color3fv = TAG(ColorMaterial3fv);
+      } else {
+	 ctx->Exec->Color3fv = _mesa_noop_Color3fv;
+      }
+   } else {
+      ctx->Exec->Color3fv = TAG(Color3fv);
+   }
+   glColor3fv( v );
+}
+
+static void TAG(choose_Color3ub)( GLubyte r, GLubyte g, GLubyte b )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color3ub = TAG(ColorMaterial3ub);
+      } else {
+	 ctx->Exec->Color3ub = _mesa_noop_Color3ub;
+      }
+   } else {
+      ctx->Exec->Color3ub = TAG(Color3ub);
+   }
+   glColor3ub( r, g, b );
+}
+
+static void TAG(choose_Color3ubv)( const GLubyte *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color3ubv = TAG(ColorMaterial3ubv);
+      } else {
+	 ctx->Exec->Color3ubv = _mesa_noop_Color3ubv;
+      }
+   } else {
+      ctx->Exec->Color3ubv = TAG(Color3ubv);
+   }
+   glColor3ubv( v );
+}
+
+static void TAG(choose_Color4f)( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color4f = TAG(ColorMaterial4f);
+      } else {
+	 ctx->Exec->Color4f = _mesa_noop_Color4f;
+      }
+   } else {
+      ctx->Exec->Color4f = TAG(Color4f);
+   }
+   glColor4f( r, g, b, a );
+}
+
+static void TAG(choose_Color4fv)( const GLfloat *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color4fv = TAG(ColorMaterial4fv);
+      } else {
+	 ctx->Exec->Color4fv = _mesa_noop_Color4fv;
+      }
+   } else {
+      ctx->Exec->Color4fv = TAG(Color4fv);
+   }
+   glColor4fv( v );
+}
+
+static void TAG(choose_Color4ub)( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color4ub = TAG(ColorMaterial4ub);
+      } else {
+	 ctx->Exec->Color4ub = _mesa_noop_Color4ub;
+      }
+   } else {
+      ctx->Exec->Color4ub = TAG(Color4ub);
+   }
+   glColor4ub( r, g, b, a );
+}
+
+static void TAG(choose_Color4ubv)( const GLubyte *v )
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if ( ctx->Light.Enabled ) {
+      if ( ctx->Light.ColorMaterialEnabled ) {
+	 ctx->Exec->Color4ubv = TAG(ColorMaterial4ubv);
+      } else {
+	 ctx->Exec->Color4ubv = _mesa_noop_Color4ubv;
+      }
+   } else {
+      ctx->Exec->Color4ubv = TAG(Color4ubv);
+   }
+   glColor4ubv( v );
+}
+
+
+
+#undef GET_CURRENT
+#undef CURRENT_COLOR
+#undef CURRENT_SPECULAR
+#undef COLOR_IS_FLOAT
+#undef RECALC_BASE_COLOR
+#undef TAG
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_napi.h b/src/mesa/tnl_dd/imm/t_dd_imm_napi.h
new file mode 100644
index 0000000000..9844f615ff
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_napi.h
@@ -0,0 +1,226 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Gareth Hughes <gareth@valinux.com>
+ *    Keith Whitwell <keith_whitwell@yahoo.com>
+ */
+
+/* Template for immediate mode normal functions.  Optimize for infinite
+ * lights when doing software lighting.
+ */
+
+static void TAG(Normal3f_single)( GLfloat x, GLfloat y, GLfloat z )
+{
+   GET_CURRENT_VERTEX;
+   const struct gl_light *light = ctx->Light.EnabledList.prev;
+   GLfloat n_dot_h, n_dot_VP, spec, sum[3];
+   GLfloat *normal = ctx->Current.Normal;
+   GLfloat scale = 1.0;
+
+   ASSIGN_3V( normal, x, y, z );
+   COPY_3V( sum, BASE_COLOR );
+
+   if ( IND & NORM_RESCALE ) {
+      scale = ctx->_ModelViewInvScale;
+   } else if ( IND & NORM_NORMALIZE ) {
+      scale = LEN_3FV( normal );
+      if ( scale != 0.0 ) scale = 1.0 / scale;
+   }
+
+   n_dot_VP = DOT3( normal, light->_VP_inf_norm ) * scale;
+   if ( n_dot_VP > 0.0F ) {
+      ACC_SCALE_SCALAR_3V( sum, n_dot_VP, light->_MatDiffuse[0] );
+      n_dot_h = DOT3( normal, light->_h_inf_norm ) * scale;
+      if ( n_dot_h > 0.0F ) {
+	 GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec );
+	 ACC_SCALE_SCALAR_3V( sum, spec, light->_MatSpecular[0] );
+      }
+   }
+
+#ifdef LIT_COLOR_IS_FLOAT
+   LIT_COLOR ( RCOMP ) = CLAMP(sum[0], 0.0f, 0.1f);
+   LIT_COLOR ( GCOMP ) = CLAMP(sum[1], 0.0f, 0.1f);
+   LIT_COLOR ( BCOMP ) = CLAMP(sum[2], 0.0f, 0.1f);
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( RCOMP ), sum[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( GCOMP ), sum[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( BCOMP ), sum[2] );
+#endif
+   LIT_COLOR( ACOMP ) = LIT_ALPHA;
+}
+
+static void TAG(Normal3fv_single)( const GLfloat *normal )
+{
+   GET_CURRENT_VERTEX;
+   const struct gl_light *light = ctx->Light.EnabledList.prev;
+   GLfloat n_dot_h, n_dot_VP, spec, sum[3];
+   GLfloat scale = 1.0;
+
+   COPY_3V( ctx->Current.Normal, normal );
+   COPY_3V( sum, BASE_COLOR );
+
+   if ( IND & NORM_RESCALE ) {
+      scale = ctx->_ModelViewInvScale;
+   } else if ( IND & NORM_NORMALIZE ) {
+      scale = LEN_3FV( normal );
+      if ( scale != 0.0 ) scale = 1.0 / scale;
+   }
+
+   n_dot_VP = DOT3( normal, light->_VP_inf_norm ) * scale;
+   if ( n_dot_VP > 0.0F ) {
+      ACC_SCALE_SCALAR_3V( sum, n_dot_VP, light->_MatDiffuse[0] );
+      n_dot_h = DOT3( normal, light->_h_inf_norm ) * scale;
+      if ( n_dot_h > 0.0F ) {
+	 GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec );
+	 ACC_SCALE_SCALAR_3V( sum, spec, light->_MatSpecular[0] );
+      }
+   }
+
+#ifdef LIT_COLOR_IS_FLOAT
+   LIT_COLOR ( RCOMP ) = CLAMP(sum[0], 0.0f, 0.1f);
+   LIT_COLOR ( GCOMP ) = CLAMP(sum[1], 0.0f, 0.1f);
+   LIT_COLOR ( BCOMP ) = CLAMP(sum[2], 0.0f, 0.1f);
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( RCOMP ), sum[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( GCOMP ), sum[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( BCOMP ), sum[2] );
+#endif
+   LIT_COLOR( ACOMP ) = LIT_ALPHA;
+}
+
+
+static void TAG(Normal3f_multi)( GLfloat x, GLfloat y, GLfloat z )
+{
+   GET_CURRENT_VERTEX;
+   struct gl_light *light;
+   GLfloat n_dot_h, n_dot_VP, spec, sum[3], tmp[3];
+   GLfloat *normal;
+
+   ASSIGN_3V( ctx->Current.Normal, x, y, z );
+   COPY_3V( sum, BASE_COLOR );
+
+   if ( IND & NORM_RESCALE ) {
+      normal = tmp;
+      ASSIGN_3V( normal, x, y, z );
+      SELF_SCALE_SCALAR_3V( normal, ctx->_ModelViewInvScale );
+   } else if ( IND & NORM_NORMALIZE ) {
+      normal = tmp;
+      ASSIGN_3V( normal, x, y, z );
+      NORMALIZE_3FV( normal );
+   } else {
+      normal = ctx->Current.Normal;
+   }
+
+   foreach ( light, &ctx->Light.EnabledList ) {
+      n_dot_VP = DOT3( normal, light->_VP_inf_norm );
+      if ( n_dot_VP > 0.0F ) {
+	 ACC_SCALE_SCALAR_3V( sum, n_dot_VP, light->_MatDiffuse[0] );
+	 n_dot_h = DOT3( normal, light->_h_inf_norm );
+	 if ( n_dot_h > 0.0F ) {
+	    GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec );
+	    ACC_SCALE_SCALAR_3V( sum, spec, light->_MatSpecular[0] );
+	 }
+      }
+   }
+
+#ifdef LIT_COLOR_IS_FLOAT
+   LIT_COLOR ( RCOMP ) = CLAMP(sum[0], 0.0f, 0.1f);
+   LIT_COLOR ( GCOMP ) = CLAMP(sum[1], 0.0f, 0.1f);
+   LIT_COLOR ( BCOMP ) = CLAMP(sum[2], 0.0f, 0.1f);
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( RCOMP ), sum[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( GCOMP ), sum[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( BCOMP ), sum[2] );
+#endif
+   LIT_COLOR( ACOMP ) = LIT_ALPHA;
+}
+
+static void TAG(Normal3fv_multi)( const GLfloat *n )
+{
+   GET_CURRENT_VERTEX;
+   struct gl_light *light;
+   GLfloat n_dot_h, n_dot_VP, spec, sum[3], tmp[3];
+   GLfloat *normal;
+
+   COPY_3V( ctx->Current.Normal, n );
+   COPY_3V( sum, BASE_COLOR );
+
+   if ( IND & NORM_RESCALE ) {
+      normal = tmp;
+      COPY_3V( normal, n );
+      SELF_SCALE_SCALAR_3V( normal, ctx->_ModelViewInvScale );
+   } else if ( IND & NORM_NORMALIZE ) {
+      normal = tmp;
+      COPY_3V( normal, n );
+      NORMALIZE_3FV( normal );
+   } else {
+      normal = ctx->Current.Normal;
+   }
+
+   foreach ( light, &ctx->Light.EnabledList ) {
+      n_dot_VP = DOT3( normal, light->_VP_inf_norm );
+      if ( n_dot_VP > 0.0F ) {
+	 ACC_SCALE_SCALAR_3V( sum, n_dot_VP, light->_MatDiffuse[0] );
+	 n_dot_h = DOT3( normal, light->_h_inf_norm );
+	 if ( n_dot_h > 0.0F ) {
+	    GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec );
+	    ACC_SCALE_SCALAR_3V( sum, spec, light->_MatSpecular[0] );
+	 }
+      }
+   }
+
+#ifdef LIT_COLOR_IS_FLOAT
+   LIT_COLOR ( RCOMP ) = CLAMP(sum[0], 0.0f, 0.1f);
+   LIT_COLOR ( GCOMP ) = CLAMP(sum[1], 0.0f, 0.1f);
+   LIT_COLOR ( BCOMP ) = CLAMP(sum[2], 0.0f, 0.1f);
+#else
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( RCOMP ), sum[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( GCOMP ), sum[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( LIT_COLOR( BCOMP ), sum[2] );
+#endif
+   LIT_COLOR( ACOMP ) = LIT_ALPHA;
+}
+
+
+
+static void TAG(init_norm)( void )
+{
+   norm_tab[IND].normal3f_single = TAG(Normal3f_single);
+   norm_tab[IND].normal3fv_single = TAG(Normal3fv_single);
+   norm_tab[IND].normal3f_multi = TAG(Normal3f_multi);
+   norm_tab[IND].normal3fv_multi = TAG(Normal3fv_multi);
+}
+
+
+
+#ifndef PRESERVE_NORMAL_DEFS
+#undef GET_CURRENT
+#undef GET_CURRENT_VERTEX
+#undef LIT_COLOR
+#undef LIT_COLOR_IS_FLOAT
+#endif
+#undef PRESERVE_NORMAL_DEFS
+#undef IND
+#undef TAG
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_primtmp.h b/src/mesa/tnl_dd/imm/t_dd_imm_primtmp.h
new file mode 100644
index 0000000000..97dca3fd42
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_primtmp.h
@@ -0,0 +1,570 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@valinux.com>
+ *    Gareth Hughes <gareth@valinux.com>
+ */
+
+/* Template for immediate mode vertices. 
+ *
+ * Probably instantiate once for each vertex format used:
+ *   - TINY_VERTICES
+ *   - TEX0_VERTICES
+ *   - TEX1_VERTICES
+ *   - PTEX_VERTICES
+ *
+ * Have to handle TEX->PTEX transition somehow.
+ */
+
+#define DBG 0
+
+
+
+/* =============================================================
+ * GL_POINTS
+ */
+
+static void TAG(flush_point_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   if ( !v0->mask ) {
+      LOCAL_VARS;
+      DRAW_POINT( v0 );
+   }
+}
+
+
+/* =============================================================
+ * GL_LINES
+ */
+
+static void TAG(flush_line_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_line_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   FLUSH_VERTEX = TAG(flush_line_1);
+   ACTIVE_VERTEX = IMM_VERTICES( 1 );
+}
+
+static void TAG(flush_line_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v1 = v0 - 1;
+   ACTIVE_VERTEX = IMM_VERTICES( 0 );
+   FLUSH_VERTEX = TAG(flush_line_0);
+   if (FALLBACK_OR_CLIPPING) 
+      CLIP_OR_DRAW_LINE( ctx, v1, v0 ); 
+   else
+      DRAW_LINE( ctx, v1, v0 ); 
+}
+
+
+/* =============================================================
+ * GL_LINE_LOOP
+ */
+
+static void TAG(flush_line_loop_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_line_loop_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_line_loop_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   ACTIVE_VERTEX = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_line_loop_1);
+}
+
+#define DRAW_LINELOOP_LINE( a, b )			\
+   if (!HAVE_LINE_STRIP || FALLBACK_OR_CLIPPING) {	\
+      CLIP_OR_DRAW_LINE( ctx, a, b );			\
+   } else if (EXTEND_PRIM( 1 )) {			\
+      EMIT_VERTEX( b );					\
+   } else {						\
+      BEGIN_PRIM( GL_LINE_STRIP, 2 ); 			\
+      EMIT_VERTEX( a );					\
+      EMIT_VERTEX( b );      				\
+   }
+
+static void TAG(flush_line_loop_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v1 = v0 - 1;
+   ACTIVE_VERTEX = v1;
+   FLUSH_VERTEX = TAG(flush_line_loop_2);
+   DRAW_LINELOOP_LINE( v1, v0 );
+}
+
+static void TAG(flush_line_loop_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v1 = v0 + 1;
+   ACTIVE_VERTEX = v1;
+   FLUSH_VERTEX = TAG(flush_line_loop_1);
+   DRAW_LINELOOP_LINE( v1, v0 );
+}
+
+static void TAG(end_line_loop)( GLcontext *ctx )
+{
+   LOCAL_VARS;
+
+   if ( FLUSH_VERTEX != TAG(flush_line_loop_0) ) {
+      TNL_VERTEX *v1 = ACTIVE_VERTEX;
+      TNL_VERTEX *v0 = IMM_VERTICES( 0 );
+      DRAW_LINELOOP_LINE( v1, v0 );
+   }
+}
+
+
+
+/* =============================================================
+ * GL_LINE_STRIP
+ */
+
+static void TAG(flush_line_strip_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_line_strip_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_line_strip_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   ACTIVE_VERTEX = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_line_strip_0b);
+}
+
+
+static void TAG(flush_line_strip_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v1 = v0 - 1;
+
+   ACTIVE_VERTEX = v1;
+   FLUSH_VERTEX = TAG(flush_line_strip_2);
+
+   if (!HAVE_LINE_STRIP || FALLBACK_OR_CLIPPING)
+      CLIP_OR_DRAW_LINE( ctx, v1, v0 );
+   else if (EXTEND_PRIM( 1 )) {
+      EMIT_VERTEX( v0 );
+   } else {
+      BEGIN_PRIM( GL_LINE_STRIP, 2 ); 
+      EMIT_VERTEX( v1 );
+      EMIT_VERTEX( v0 );      
+   }      
+}
+
+static void TAG(flush_line_strip_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v1 = v0 + 1;
+
+   ACTIVE_VERTEX = v1;
+   FLUSH_VERTEX = TAG(flush_line_strip_1);
+
+   if (!HAVE_LINE_STRIP || FALLBACK_OR_CLIPPING)
+      CLIP_OR_DRAW_LINE( ctx, v1, v0 );
+   else if (EXTEND_PRIM( 1 )) {
+      EMIT_VERTEX( v0 );
+   } else {
+      BEGIN_PRIM( GL_LINE_STRIP, 2 ); 
+      EMIT_VERTEX( v1 );
+      EMIT_VERTEX( v0 );      
+   }      
+}
+
+
+
+/* =============================================================
+ * GL_TRIANGLES
+ */
+
+static void TAG(flush_triangle_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_triangle_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_triangle_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   if ( DBG ) fprintf( stderr, __FUNCTION__ "\n" );
+
+   ACTIVE_VERTEX = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_triangle_1);
+   BEGIN_PRIM( GL_TRIANGLES, 0 );
+}
+
+static void TAG(flush_triangle_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   if ( DBG ) fprintf( stderr, __FUNCTION__ "\n" );
+
+   ACTIVE_VERTEX = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_triangle_2);
+}
+
+static void TAG(flush_triangle_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v2 = v0 - 2;
+   TNL_VERTEX *v1 = v0 - 1;
+
+   if ( DBG ) fprintf( stderr, __FUNCTION__ "\n" );
+
+   ACTIVE_VERTEX = v2;
+   FLUSH_VERTEX = TAG(flush_triangle_0);
+
+   /* nothing gained by trying to emit as hw primitives -- that
+    * happens normally in this case.  
+    */
+   if (FALLBACK_OR_CLIPPING)
+      CLIP_OR_DRAW_TRI( ctx, v2, v1, v0 );
+   else
+      DRAW_TRI( ctx, v2, v1, v0 );
+}
+
+
+
+
+/* =============================================================
+ * GL_TRIANGLE_STRIP
+ */
+
+static void TAG(flush_tri_strip_3)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_tri_strip_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_tri_strip_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_tri_strip_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 1 );
+   FLUSH_VERTEX = TAG(flush_tri_strip_1);
+}
+
+static void TAG(flush_tri_strip_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 2 );
+   FLUSH_VERTEX = TAG(flush_tri_strip_2);
+}
+
+#define DO_TRISTRIP_TRI( vert0, vert1 )				\
+   if (!HAVE_TRI_STRIP || FALLBACK_OR_CLIPPING) {	\
+      TNL_VERTEX *v2 = IMM_VERTICES( vert0 );			\
+      TNL_VERTEX *v1 = IMM_VERTICES( vert1 );			\
+      TAG(draw_tri)( ctx, v2, v1, v0 );				\
+   } else if (EXTEND_PRIM( 1 )) {				\
+      EMIT_VERTEX( v0 );					\
+   } else {							\
+      TNL_VERTEX *v2 = IMM_VERTICES( vert0 );			\
+      TNL_VERTEX *v1 = IMM_VERTICES( vert1 );			\
+      BEGIN_PRIM( GL_TRIANGLE_STRIP, 3 ); 			\
+      EMIT_VERTEX( v2 );					\
+      EMIT_VERTEX( v1 );      					\
+      EMIT_VERTEX( v0 );      					\
+   }      
+
+static void TAG(flush_tri_strip_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   FLUSH_VERTEX = TAG(flush_tri_strip_3);
+   ACTIVE_VERTEX = IMM_VERTICES( 3 );
+   DO_TRISTRIP_TRI( 0, 1 );
+}
+
+static void TAG(flush_tri_strip_3)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   FLUSH_VERTEX = TAG(flush_tri_strip_4);
+   ACTIVE_VERTEX = IMM_VERTICES( 0 );
+   DO_TRISTRIP_TRI( 1, 2 );
+}
+
+static void TAG(flush_tri_strip_4)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   FLUSH_VERTEX = TAG(flush_tri_strip_5);
+   ACTIVE_VERTEX = IMM_VERTICES( 1 );
+   DO_TRISTRIP_TRI( 2, 3 );
+}
+
+static void TAG(flush_tri_strip_5)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   FLUSH_VERTEX = TAG(flush_tri_strip_2);
+   ACTIVE_VERTEX = IMM_VERTICES( 2 );
+   DO_TRISTRIP_TRI( 0, 3 );
+}
+
+
+
+/* =============================================================
+ * GL_TRIANGLE_FAN
+ */
+
+static void TAG(flush_tri_fan_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_tri_fan_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_tri_fan_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   ACTIVE_VERTEX = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_tri_fan_1);
+}
+
+static void TAG(flush_tri_fan_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   ACTIVE_VERTEX = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_tri_fan_2);
+}
+
+#define DO_TRIFAN_TRI( vert0, vert1 )				\
+   if (!HAVE_TRI_FAN || FALLBACK_OR_CLIPPING) {	\
+      TNL_VERTEX *v2 = IMM_VERTICES( vert0 );			\
+      TNL_VERTEX *v1 = IMM_VERTICES( vert1 );			\
+      TAG(draw_tri)( ctx, v2, v1, v0 );				\
+   } else if (EXTEND_PRIM( 1 )) {				\
+      EMIT_VERTEX( v0 );					\
+   } else {							\
+      TNL_VERTEX *v2 = IMM_VERTICES( vert0 );			\
+      TNL_VERTEX *v1 = IMM_VERTICES( vert1 );			\
+      BEGIN_PRIM( GL_TRIANGLE_FAN, 3 ); 			\
+      EMIT_VERTEX( v2 );					\
+      EMIT_VERTEX( v1 );      					\
+      EMIT_VERTEX( v0 );      					\
+   }      
+
+static void TAG(flush_tri_fan_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 1 );
+   FLUSH_VERTEX = TAG(flush_tri_fan_3 );
+   DO_TRIFAN_TRI( 0, 1 );
+}
+
+static void TAG(flush_tri_fan_3)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 2 );
+   FLUSH_VERTEX = TAG(flush_tri_fan_2 );
+   DO_TRIFAN_TRI( 0, 2 );
+}
+
+
+
+/* =============================================================
+ * GL_QUADS
+ */
+
+static void TAG(flush_quad_3)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_quad_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_quad_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_quad_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   IMM_VERTEX( v0 ) = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_quad_1);
+}
+
+static void TAG(flush_quad_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   IMM_VERTEX( v0 ) = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_quad_2);
+}
+
+static void TAG(flush_quad_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   IMM_VERTEX( v0 ) = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_quad_3);
+}
+
+static void TAG(flush_quad_3)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v3 = v0 - 3;
+   TNL_VERTEX *v2 = v0 - 2;
+   TNL_VERTEX *v1 = v0 - 1;
+
+   IMM_VERTEX( v0 ) = v3;
+   FLUSH_VERTEX = TAG(flush_quad_0);
+
+   if (!HAVE_HW_QUADS || FALLBACK_OR_CLIPPING) {
+      CLIP_OR_DRAW_TRI( ctx, v3, v2, v0 );
+      CLIP_OR_DRAW_TRI( ctx, v2, v1, v0 );
+   } else {
+      EXTEND_PRIM_NF( GL_QUADS, 4 );
+      EMIT_VERTEX( v3 );
+      EMIT_VERTEX( v2 );
+      EMIT_VERTEX( v1 );      
+      EMIT_VERTEX( v0 );      
+   }
+}
+
+
+
+/* =============================================================
+ * GL_QUAD_STRIP
+ */
+
+static void TAG(flush_quad_strip_3)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_quad_strip_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_quad_strip_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_quad_strip_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   IMM_VERTEX( v3 ) = v0;
+   IMM_VERTEX( v0 ) = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_quad_strip_1);
+}
+
+static void TAG(flush_quad_strip_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   IMM_VERTEX( v2 ) = v0;
+   IMM_VERTEX( v0 ) = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_quad_strip_2);
+}
+
+static void TAG(flush_quad_strip_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+
+   IMM_VERTEX( v1 ) = v0;
+   IMM_VERTEX( v0 ) = v0 + 1;
+   FLUSH_VERTEX = TAG(flush_quad_strip_3);
+}
+
+static void TAG(flush_quad_strip_3)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   TNL_VERTEX *v3 = IMM_VERTEX( v3 );
+   TNL_VERTEX *v2 = IMM_VERTEX( v2 );
+   TNL_VERTEX *v1 = IMM_VERTEX( v1 );
+
+   IMM_VERTEX( v0 ) = v3;
+   IMM_VERTEX( v2 ) = v0;
+   IMM_VERTEX( v3 ) = v1;
+   FLUSH_VERTEX = TAG(flush_quad_strip_2);
+
+   if (FALLBACK_OR_CLIPPING) {
+      CLIP_OR_DRAW_TRI( ctx, v3, v2, v0 );
+      CLIP_OR_DRAW_TRI( ctx, v2, v1, v0 );
+   } else {
+      DRAW_TRI( ctx, v3, v2, v0 );
+      DRAW_TRI( ctx, v2, v1, v0 );
+   }
+}
+
+
+
+/* =============================================================
+ * GL_POLYGON
+ */
+
+static void TAG(flush_poly_2)( GLcontext *ctx, TNL_VERTEX *v0 );
+static void TAG(flush_poly_1)( GLcontext *ctx, TNL_VERTEX *v0 );
+
+static void TAG(flush_poly_0)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 1 );
+   FLUSH_VERTEX = TAG(flush_poly_1);
+}
+
+static void TAG(flush_poly_1)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 2 );
+   FLUSH_VERTEX = TAG(flush_poly_2);
+}
+
+#define DO_POLY_TRI( vert0, vert1 )				\
+   if (!HAVE_POLYGONS || FALLBACK_OR_CLIPPING) {	\
+      TNL_VERTEX *v2 = IMM_VERTICES( vert0 );			\
+      TNL_VERTEX *v1 = IMM_VERTICES( vert1 );			\
+      TAG(draw_tri)( ctx, v1, v0, v2 );				\
+   } else if (EXTEND_PRIM( 1 )) {				\
+      EMIT_VERTEX( v0 );					\
+   } else {							\
+      TNL_VERTEX *v2 = IMM_VERTICES( vert0 );			\
+      TNL_VERTEX *v1 = IMM_VERTICES( vert1 );			\
+      BEGIN_PRIM( GL_POLYGON, 3 );				\
+      EMIT_VERTEX( v2 );					\
+      EMIT_VERTEX( v1 );					\
+      EMIT_VERTEX( v0 );					\
+   }      
+
+static void TAG(flush_poly_2)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 1 );
+   FLUSH_VERTEX = TAG(flush_poly_3);
+   DO_POLY_TRI( 0, 1 );
+}
+
+static void TAG(flush_poly_3)( GLcontext *ctx, TNL_VERTEX *v0 )
+{
+   LOCAL_VARS;
+   ACTIVE_VERTEX = IMM_VERTICES( 2 );
+   FLUSH_VERTEX = TAG(flush_poly_2);
+   DO_POLY_TRI( 0, 2 );
+}
+
+
+void (*TAG(flush_tab)[GL_POLYGON+1])( GLcontext *, TNL_VERTEX * ) =
+{
+   TAG(flush_point),
+   TAG(flush_line_0),
+   TAG(flush_line_loop_0),
+   TAG(flush_line_strip_0),
+   TAG(flush_triangle_0),
+   TAG(flush_tri_strip_0),
+   TAG(flush_tri_fan_0),
+   TAG(flush_quad_0),
+   TAG(flush_quad_strip_0),
+   TAG(flush_poly_0),
+};
+
+
+#ifndef PRESERVE_PRIM_DEFS
+#undef LOCAL_VARS
+#undef GET_INTERP_FUNC
+#undef IMM_VERTEX
+#undef IMM_VERTICES
+#undef FLUSH_VERTEX
+#endif
+#undef PRESERVE_PRIM_DEFS
+#undef EXTEND_PRIM
+#undef EMIT_VERTEX
+#undef EMIT_VERTEX_TRI
+#undef EMIT_VERTEX_LINE
+#undef EMIT_VERTEX_POINT
+#undef TAG
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_tapi.h b/src/mesa/tnl_dd/imm/t_dd_imm_tapi.h
new file mode 100644
index 0000000000..b82a9abebf
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_tapi.h
@@ -0,0 +1,75 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Gareth Hughes <gareth@valinux.com>
+ *    Keith Whitwell <keith_whitwell@yahoo.com>
+ */
+
+/* Template for immediate mode texture coordinate functions.
+ */
+
+#ifndef DO_PROJ_TEX
+#error "Need to define DO_PROJ_TEX"
+#endif
+
+
+/* =============================================================
+ * Notify on calls to texture4f, to allow switch to projected texture
+ * vertex format:
+ */
+
+static void TAG(TexCoord4f)( GLfloat s, GLfloat t, GLfloat r, GLfloat q )
+{
+   GET_CURRENT;
+   DO_PROJ_TEX;
+   TEXCOORD4( s, t, r, q );
+}
+
+static void TAG(TexCoord4fv)( const GLfloat *v )
+{
+   GET_CURRENT;
+   DO_PROJ_TEX;
+   TEXCOORD4( v[0], v[1], v[2], v[3] );
+}
+
+static void TAG(MultiTexCoord4fARB)( GLenum target, GLfloat s,
+				     GLfloat t, GLfloat r, GLfloat q )
+{
+   GET_CURRENT;
+   DO_PROJ_TEX;
+   MULTI_TEXCOORD4( unit, s, t, r, q );
+}
+
+static void TAG(MultiTexCoord4fvARB)( GLenum target, const GLfloat *v )
+{
+   GET_CURRENT;
+   DO_PROJ_TEX;
+   MULTI_TEXCOORD4( unit, v[0], v[1], v[2], v[3] );
+}
+
+
+
+#undef DO_PROJ_TEX
+#undef TAG
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_vapi.h b/src/mesa/tnl_dd/imm/t_dd_imm_vapi.h
new file mode 100644
index 0000000000..95e93b1c83
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_vapi.h
@@ -0,0 +1,159 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Gareth Hughes <gareth@valinux.com>
+ *    Keith Whitwell <keithw@valinux.com>
+ */
+
+/* Template for immediate mode vertex functions.
+ */
+
+#define DBG 0
+
+#define VERTEX( ox, oy, oz, ow )
+do {
+   GET_CURRENT_VERTEX;
+   GLfloat w;
+   GLuint mask;
+   const GLfloat * const m = ctx->_ModelProjectMatrix.m;
+
+   if (DO_FULL_MATRIX) {
+      VERTEX_CLIP(0) = m[0] * ox + m[4] * oy + m[8]  * oz + m[12] * ow;
+      VERTEX_CLIP(1) = m[1] * ox + m[5] * oy + m[9]  * oz + m[13] * ow;
+      VERTEX_CLIP(2) = m[2] * ox + m[6] * oy + m[10] * oz + m[14] * ow;
+      VERTEX_CLIP(3) = m[3] * ox + m[7] * oy + m[11] * oz + m[15] * ow;
+      w = VERTEX_CLIP(3);
+   }
+   else if (DO_NOROT_MATRIX) {
+      VERTEX_CLIP(0) = m[0] * ox                          + m[12] * ow;
+      VERTEX_CLIP(1) =             m[5] * oy              + m[13] * ow;
+      VERTEX_CLIP(2) =                         m[10] * oz + m[14] * ow;
+      VERTEX_CLIP(3) =                                              ow;
+      w = ow;
+   } 
+   else {
+      ASSERT (DO_IDENTITY_MATRIX);
+      VERTEX_CLIP(0) = ox;
+      VERTEX_CLIP(1) = oy;
+      VERTEX_CLIP(2) = oz;
+      VERTEX_CLIP(3) = ow;
+      w = ow;
+   }
+
+   mask = 0;
+   if (DO_CLIP_TEST) {
+      if ( VERTEX_CLIP(0) >  w ) mask |= CLIP_RIGHT_BIT;
+      if ( VERTEX_CLIP(0) < -w ) mask |= CLIP_LEFT_BIT;
+      if ( VERTEX_CLIP(1) >  w ) mask |= CLIP_TOP_BIT;
+      if ( VERTEX_CLIP(1) < -w ) mask |= CLIP_BOTTOM_BIT;
+      if ( VERTEX_CLIP(2) >  w ) mask |= CLIP_FAR_BIT;
+      if ( VERTEX_CLIP(2) < -w ) mask |= CLIP_NEAR_BIT;
+      VERTEX_MASK(v) = mask;
+   }
+
+   if (!mask) {
+      if (HAVE_VERTEX_WIN) {
+	 if (!HAVE_HW_VIEWPORT) {
+	    const GLfloat *s = GET_VIEWPORT_MATRIX();
+	    if (HAVE_W && HAVE_HW_DIVIDE) {
+	       VERTEX_WIN( 0 ) = s[0]  * VERTEX_CLIP( 0 ) + s[12];
+	       VERTEX_WIN( 1 ) = s[5]  * VERTEX_CLIP( 1 ) + s[13];
+	       VERTEX_WIN( 2 ) = s[10] * VERTEX_CLIP( 2 ) + s[14];
+	       VERTEX_WIN( 3 ) = w;
+	    }
+	    else {
+	       const GLfloat oow = 1.0/w; /* possibly opt away */
+	       VERTEX_WIN( 0 ) = s[0]  * VERTEX_CLIP( 0 ) * oow + s[12];
+	       VERTEX_WIN( 1 ) = s[5]  * VERTEX_CLIP( 1 ) * oow + s[13];
+	       VERTEX_WIN( 2 ) = s[10] * VERTEX_CLIP( 2 ) * oow + s[14];
+	       if (HAVE_W)
+		  VERTEX_WIN( 3 ) = oow;
+	    }
+	 }
+	 else if (HAVE_W && HAVE_HW_DIVIDE) {
+	    if (!VERTEX_WIN_IS_VERTEX_CLIP) {
+	       VERTEX_WIN( 0 ) = VERTEX_CLIP( 0 );
+	       VERTEX_WIN( 1 ) = VERTEX_CLIP( 1 );
+	       VERTEX_WIN( 2 ) = VERTEX_CLIP( 2 );
+	       VERTEX_WIN( 3 ) = w;
+	    }
+	 }
+	 else {
+	    const GLfloat oow = 1.0/w; /* possibly opt away */
+	    VERTEX_WIN( 0 ) = VERTEX_CLIP( 0 ) * oow;
+	    VERTEX_WIN( 1 ) = VERTEX_CLIP( 1 ) * oow;
+	    VERTEX_WIN( 2 ) = VERTEX_CLIP( 2 ) * oow;
+	    if (HAVE_W)
+	       VERTEX_WIN( 3 ) = oow;
+	 }
+      }
+   } else if (!FALLBACK_OR_CLIPPING) {
+      SET_CLIPPING();		/* transition to clipping */
+   }
+
+   COPY_VERTEX_FROM_CURRENT;
+   BUILD_PRIM_FROM_VERTEX;
+}
+
+/* Let the compiler optimize away the constant operations:
+ */
+static void VTAG(Vertex2f)( GLfloat ox, GLfloat oy )
+{
+   /* Cliptest on clip[2] could also be eliminated...
+    */
+   VERTEX( ox, oy, 0, 1 );
+}
+
+static void VTAG(Vertex2fv)( const GLfloat *obj )
+{
+   /* Cliptest on clip[2] could also be eliminated...
+    */
+   VERTEX( obj[0], obj[1], 0, 1 );
+}
+
+static void VTAG(Vertex3f)( GLfloat ox, GLfloat oy, GLfloat oz )
+{
+   VERTEX( ox, oy, oz, 1 );
+}
+
+static void VTAG(Vertex3fv)( const GLfloat *obj )
+{
+   VERTEX( obj[0], obj[1], obj[2], 1 );
+}
+
+static void VTAG(Vertex4f)( GLfloat ox, GLfloat oy, GLfloat oz, GLfloat ow )
+{
+   VERTEX( ox, oy, oz, ow );
+}
+
+static void VTAG(Vertex4fv)( const GLfloat *obj )
+{
+   VERTEX( obj[0], obj[1], obj[2], obj[3] );
+}
+
+
+#undef DO_FULL_MATRIX
+#undef VTAG
+#undef VERTEX
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_vb.c b/src/mesa/tnl_dd/imm/t_dd_imm_vb.c
new file mode 100644
index 0000000000..0c4462f556
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_vb.c
@@ -0,0 +1,204 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith_whitwell@yahoo.com>
+ */
+
+/* Template to build clipping routines to support t_dd_imm_primtmp.h.
+ *
+ * The TAG(draw_line) and TAG(draw_triangle) routines are called in
+ * clipping and fallback scenarios, and when the native hardware
+ * primitive (eg polygons) is unavailable.
+ */
+
+
+#define CLIP_DOTPROD(K, A, B, C, D)		\
+   (CLIP_X(K)*A + CLIP_Y(K)*B + 	\
+    CLIP_Z(K)*C + CLIP_W(K)*D)
+
+#define POLY_CLIP( PLANE, A, B, C, D )					\
+do {									\
+   if (mask & PLANE) {							\
+      TNL_VERTEX **indata = inlist[in];					\
+      TNL_VERTEX **outdata = inlist[in ^= 1];				\
+      TNL_VERTEX *J = indata[0];					\
+      GLfloat dpJ = CLIP_DOTPROD(J, A, B, C, D );			\
+      GLuint outcount = 0;						\
+      GLuint i;								\
+									\
+      indata[n] = indata[0]; /* prevent rotation of vertices */		\
+      for (i = 1; i <= n; i++) {					\
+	 TNL_VERTEX *I = indata[i];					\
+	 GLfloat dpI = CLIP_DOTPROD(idx, A, B, C, D );			\
+									\
+	 if (!NEGATIVE(dpPrev)) {					\
+	    outdata[outcount++] = J;					\
+	 }								\
+									\
+	 if (DIFFERENT_SIGNS(dpI, dpJ)) {				\
+            TNL_VERTEX *O = verts++;					\
+            outdata[outcount++] = O;					\
+	    if (NEGATIVE(dpI)) {					\
+	       /* Going out of bounds.  Avoid division by zero as we	\
+		* know dp != dpPrev from DIFFERENT_SIGNS, above.	\
+		*/							\
+	       GLfloat t = dpI / (dpI - dpJ);				\
+      	       INTERP( ctx, t, O, I, J );				\
+	    } else {							\
+	       /* Coming back in.					\
+		*/							\
+	       GLfloat t = dpJ / (dpJ - dpI);				\
+	       INTERP( ctx, t, O, J, I );				\
+	    }								\
+	 }								\
+									\
+	 J = I;								\
+	 dpJ = dpI;							\
+      }									\
+									\
+      if (outcount < 3)							\
+	 return;							\
+									\
+      nr = outcount;							\
+   }									\
+} while (0)
+
+
+#define LINE_CLIP(PLANE, A, B, C, D )			\
+do {							\
+   if (mask & PLANE) {					\
+      GLfloat dpI = CLIP_DOTPROD( I, A, B, C, D );	\
+      GLfloat dpJ = CLIP_DOTPROD( J, A, B, C, D );	\
+							\
+      if (DIFFERENT_SIGNS(dpI, dpJ)) {			\
+         TNL_VERTEX *O = verts++;			\
+	 if (NEGATIVE(dpJ)) {				\
+	    GLfloat t = dpI / (dpI - dpJ);		\
+	    INTERP( ctx, t, O, I, J );	\
+            J = O;					\
+	 } else {					\
+  	    GLfloat t = dpJ / (dpJ - dpI);		\
+	    INTERP( ctx, t, O, J, I );	\
+            I = O;					\
+	 }						\
+      }							\
+      else if (NEGATIVE(dpI))				\
+	 return;					\
+  }							\
+} while (0)
+
+
+
+/* Clip a line against the viewport and user clip planes.
+ */
+static void TAG(clip_draw_line)( GLcontext *ctx,
+				 TNL_VERTEX *I,
+				 TNL_VERTEX *J,
+				 GLuint mask )
+{
+   LOCAL_VARS;
+   GET_INTERP_FUNC;
+   TNL_VERTEX tmp[MAX_CLIPPED_VERTICES];
+   TNL_VERTEX *verts = tmp;
+   TNL_VERTEX *pv = J;
+
+   LINE_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
+   LINE_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
+   LINE_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
+   LINE_CLIP( CLIP_BOTTOM_BIT,  0,  1,  0, 1 );
+   LINE_CLIP( CLIP_FAR_BIT,     0,  0, -1, 1 );
+   LINE_CLIP( CLIP_NEAR_BIT,    0,  0,  1, 1 );
+
+   if ((ctx->_TriangleCaps & DD_FLATSHADE) && J != pv)
+      COPY_PV( ctx, J, pv );
+
+   DRAW_LINE( I, J );
+}
+
+
+/* Clip a triangle against the viewport and user clip planes.
+ */
+static void TAG(clip_draw_triangle)( GLcontext *ctx,
+				     TNL_VERTEX *v0,
+				     TNL_VERTEX *v1,
+				     TNL_VERTEX *v2,
+				     GLuint mask )
+{
+   LOCAL_VARS;
+   GET_INTERP_FUNC;
+   TNL_VERTEX tmp[MAX_CLIPPED_VERTICES];
+   TNL_VERTEX *verts = tmp;
+   TNL_VERTEX *(inlist[2][MAX_CLIPPED_VERTICES]);
+   TNL_VERTEX **out;
+   GLuint in = 0;
+   GLuint n = 3;
+   GLuint i;
+
+   ASSIGN_3V(inlist, v2, v0, v1 ); /* pv rotated to slot zero */
+
+   POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
+   POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
+   POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
+   POLY_CLIP( CLIP_BOTTOM_BIT,  0,  1,  0, 1 );
+   POLY_CLIP( CLIP_FAR_BIT,     0,  0, -1, 1 );
+   POLY_CLIP( CLIP_NEAR_BIT,    0,  0,  1, 1 );
+
+   if ((ctx->_TriangleCaps & DD_FLATSHADE) && v2 != inlist[0]) 
+      COPY_PV( ctx, inlist[0], v2 );
+
+   out = inlist[in];
+   DRAW_POLYGON( out, n );
+}
+
+
+static __inline void TAG(draw_triangle)( GLcontext *ctx,
+					 TNL_VERTEX *v0,
+					 TNL_VERTEX *v1,
+					 TNL_VERTEX *v2 )
+{
+   LOCAL_VARS;
+   GLubyte ormask = (v0->mask | v1->mask | v2->mask);
+
+   if ( !ormask ) {
+      DRAW_TRI( v0, v1, v2 );
+   } else if ( !(v0->mask & v1->mask & v2->mask) ) {
+      TAG(clip_draw_triangle)( ctx, v0, v1, v2, ormask );
+   }
+}
+
+static __inline void TAG(draw_line)( GLcontext *ctx,
+				     TNL_VERTEX *v0,
+				     TNL_VERTEX *v1 )
+{
+   LOCAL_VARS;
+   GLubyte ormask = (v0->mask | v1->mask);
+
+   if ( !ormask ) {
+      DRAW_LINE( v0, v1 );
+   } else if ( !(v0->mask & v1->mask) ) {
+      TAG(clip_draw_line)( ctx, v0, v1, ormask );
+   }
+}
+
diff --git a/src/mesa/tnl_dd/imm/t_dd_imm_vbtmp.h b/src/mesa/tnl_dd/imm/t_dd_imm_vbtmp.h
new file mode 100644
index 0000000000..2f76553cff
--- /dev/null
+++ b/src/mesa/tnl_dd/imm/t_dd_imm_vbtmp.h
@@ -0,0 +1,268 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith_whitwell@yahoo.com>
+ */
+
+/* Template to build support for t_dd_imm_* tnl module using vertices
+ * as defined in t_dd_vertex.h.
+ *
+ * See t_dd_vbtmp.h for definitions of arguments to this file.
+ * Unfortunately it seems necessary to duplicate a lot of that code.
+ */
+
+#ifndef LOCALVARS
+#define LOCALVARS
+#endif
+
+
+
+/* COPY_VERTEX_FROM_CURRENT in t_dd_imm_vapi.c
+ */
+static void TAG(emit_vfmt)( GLcontext *ctx, VERTEX *v )
+{
+   LOCALVARS
+      ;
+
+   /* This template assumes (like t_dd_vbtmp.h) that color is ubyte.
+    */
+   if (DO_TEX0 || DO_TEX1 || !HAVE_TINY_VERTICES)
+   {
+      const GLubyte *col = GET_HARDWARE_COLOR();
+      if (HAVE_RGBA_COLOR) {
+	 v->v.ui[4] = *(GLuint *)&col;
+      } else {
+	 v->v.color.blue  = col[2];
+	 v->v.color.green = col[1];
+	 v->v.color.red   = col[0];
+	 v->v.color.alpha = col[3];
+      }
+   }
+   else {
+      if (HAVE_RGBA_COLOR) {
+	 v->v.ui[3] = *(GLuint *)col;
+      }
+      else {
+	 v->tv.color.blue  = col[2];
+	 v->tv.color.green = col[1];
+	 v->tv.color.red   = col[0];
+	 v->tv.color.alpha = col[3];
+      }
+   }
+
+   if (DO_TEX0) {
+      GLfloat *tc = ctx->Current.Texture[0];
+      v->v.u0 = tc[0];
+      v->v.v0 = tc[1];
+      if (DO_PTEX) {
+	 if (HAVE_PTEX_VERTICES) {
+	    v->pv.q0 = tc[3];
+	 } 
+	 else {
+	    float rhw = 1.0 / tc[3];
+	    v->v.w *= tc[3];
+	    v->v.u0 *= rhw;
+	    v->v.v0 *= rhw;
+	 } 
+      } 
+   }
+   if (DO_TEX1) {
+      GLfloat *tc = ctx->Current.Texture[1];
+      if (DO_PTEX) {
+	 v->pv.u1 = tc[0];
+	 v->pv.v1 = tc[1];
+	 v->pv.q1 = tc[3];
+      } 
+      else {
+	 v->v.u1 = tc[0];
+	 v->v.v1 = tc[1];
+      }
+   } 
+   else if (DO_PTEX) {
+      *(GLuint *)&v->pv.q1 = 0;	/* avoid culling on radeon */
+   }
+   if (DO_TEX2) {
+      GLfloat *tc = ctx->Current.Texture[2];
+      if (DO_PTEX) {
+	 v->pv.u2 = tc[0];
+	 v->pv.v2 = tc[1];
+	 v->pv.q2 = tc[3];
+      } 
+      else {
+	 v->v.u2 = tc[0];
+	 v->v.v2 = tc[1];
+      }
+   } 
+   if (DO_TEX3) {
+      GLfloat *tc = ctx->Current.Texture[3];
+      if (DO_PTEX) {
+	 v->pv.u3 = tc[0];
+	 v->pv.v3 = tc[1];
+	 v->pv.q3 = tc[3];
+      } 
+      else {
+	 v->v.u3 = tc[0];
+	 v->v.v3 = tc[1];
+      }
+   } 
+}
+
+
+
+
+static void TAG(interp)( GLcontext *ctx,
+			 GLfloat t,
+			 TNL_VERTEX *dst,
+			 TNL_VERTEX *in,
+			 TNL_VERTEX *out )
+{
+   LOCALVARS
+   const GLfloat *s = GET_VIEWPORT_MAT();
+   GLfloat w;
+
+   (void)s;
+
+   if (HAVE_HW_DIVIDE) {
+      VIEWPORT_X( dst->v.v.x, dst->clip[0] );
+      VIEWPORT_Y( dst->v.v.y, dst->clip[1] );
+      VIEWPORT_Z( dst->v.v.z, dst->clip[2] );
+      w = dstclip[3];
+   }
+   else {
+      w = 1.0 / dst->clip[3];
+      VIEWPORT_X( dst->v.v.x, dst->clip[0] * w );
+      VIEWPORT_Y( dst->v.v.y, dst->clip[1] * w );
+      VIEWPORT_Z( dst->v.v.z, dst->clip[2] * w );
+   }
+
+   if (HAVE_HW_DIVIDE || DO_TEX0) {
+
+      dst->v.v.w = w;
+
+      INTERP_UB( t, dst->v.ub4[4][0], out->v.ub4[4][0], in->v.ub4[4][0] );
+      INTERP_UB( t, dst->v.ub4[4][1], out->v.ub4[4][1], in->v.ub4[4][1] );
+      INTERP_UB( t, dst->v.ub4[4][2], out->v.ub4[4][2], in->v.ub4[4][2] );
+      INTERP_UB( t, dst->v.ub4[4][3], out->v.ub4[4][3], in->v.ub4[4][3] );
+
+      if (DO_TEX0) {
+	 if (DO_PTEX) {
+	    if (HAVE_PTEX_VERTICES) {
+	       INTERP_F( t, dst->v.pv.u0, out->v.pv.u0, in->v.pv.u0 );
+	       INTERP_F( t, dst->v.pv.v0, out->v.pv.v0, in->v.pv.v0 );
+	       INTERP_F( t, dst->v.pv.q0, out->v.pv.q0, in->v.pv.q0 );
+	    } else {
+	       GLfloat wout = out->clip[3]; /* projected clip */
+	       GLfloat win = in->clip[3]; /* projected clip */
+	       GLfloat qout = out->v.pv.w / wout;
+	       GLfloat qin = in->v.pv.w / win;
+	       GLfloat qdst, rqdst;
+
+	       ASSERT( !HAVE_HW_DIVIDE ); /* assert win, wout projected clip */
+
+	       INTERP_F( t, dst->v.v.u0, out->v.v.u0 * qout, in->v.v.u0 * qin );
+	       INTERP_F( t, dst->v.v.v0, out->v.v.v0 * qout, in->v.v.v0 * qin );
+	       INTERP_F( t, qdst, qout, qin );
+
+	       rqdst = 1.0 / qdst;
+	       dst->v.v.u0 *= rqdst;
+	       dst->v.v.v0 *= rqdst;
+	       dst->v.v.w *= rqdst;
+	    }
+	 }
+	 else {
+	    INTERP_F( t, dst->v.v.u0, out->v.v.u0, in->v.v.u0 );
+	    INTERP_F( t, dst->v.v.v0, out->v.v.v0, in->v.v.v0 );
+	 }
+      }
+      if (DO_TEX1) {
+	 if (DO_PTEX) {
+	    INTERP_F( t, dst->v.pv.u1, out->v.pv.u1, in->v.pv.u1 );
+	    INTERP_F( t, dst->v.pv.v1, out->v.pv.v1, in->v.pv.v1 );
+	    INTERP_F( t, dst->v.pv.q1, out->v.pv.q1, in->v.pv.q1 );
+	 } else {
+	    INTERP_F( t, dst->v.v.u1, out->v.v.u1, in->v.v.u1 );
+	    INTERP_F( t, dst->v.v.v1, out->v.v.v1, in->v.v.v1 );
+	 }
+      }
+      else if (DO_PTEX) {
+	 dst->v.pv.q0 = 0.0;	/* must be a valid float on radeon */
+      }
+      if (DO_TEX2) {
+	 if (DO_PTEX) {
+	    INTERP_F( t, dst->v.pv.u2, out->v.pv.u2, in->v.pv.u2 );
+	    INTERP_F( t, dst->v.pv.v2, out->v.pv.v2, in->v.pv.v2 );
+	    INTERP_F( t, dst->v.pv.q2, out->v.pv.q2, in->v.pv.q2 );
+	 } else {
+	    INTERP_F( t, dst->v.v.u2, out->v.v.u2, in->v.v.u2 );
+	    INTERP_F( t, dst->v.v.v2, out->v.v.v2, in->v.v.v2 );
+	 }
+      }
+      if (DO_TEX3) {
+	 if (DO_PTEX) {
+	    INTERP_F( t, dst->v.pv.u3, out->v.pv.u3, in->v.pv.u3 );
+	    INTERP_F( t, dst->v.pv.v3, out->v.pv.v3, in->v.pv.v3 );
+	    INTERP_F( t, dst->v.pv.q3, out->v.pv.q3, in->v.pv.q3 );
+	 } else {
+	    INTERP_F( t, dst->v.v.u3, out->v.v.u3, in->v.v.u3 );
+	    INTERP_F( t, dst->v.v.v3, out->v.v.v3, in->v.v.v3 );
+	 }
+      }
+   } else {
+      /* 4-dword vertex.  Color is in v[3] and there is no oow coordinate.
+       */
+      INTERP_UB( t, dst->v.ub4[3][0], out->v.ub4[3][0], in->v.ub4[3][0] );
+      INTERP_UB( t, dst->v.ub4[3][1], out->v.ub4[3][1], in->v.ub4[3][1] );
+      INTERP_UB( t, dst->v.ub4[3][2], out->v.ub4[3][2], in->v.ub4[3][2] );
+      INTERP_UB( t, dst->v.ub4[3][3], out->v.ub4[3][3], in->v.ub4[3][3] );
+   }
+}
+
+
+static __inline void TAG(copy_pv)( GLcontext *ctx,
+				   TNL_VERTEX *dst, 
+				   TNL_VERTEX *src )
+{
+   if (DO_TEX0 || DO_TEX1 || !HAVE_TINY_VERTICES) {
+      dst->v.v.ui[4] = src->v.v.ui[4];
+   }
+   else {
+      dst->v.v.ui[3] = src->v.v.ui[3];
+   }
+}
+
+
+
+static void TAG(init)( void )
+{
+   setup_tab[IND].emit = TAG(emit_vfmt);
+   setup_tab[IND].interp = TAG(interp_vfmt);
+}
+
+
+#undef IND
+#undef TAG
+
+
+
diff --git a/src/mesa/tnl_dd/t_dd.c b/src/mesa/tnl_dd/t_dd.c
new file mode 100644
index 0000000000..731da5c320
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd.c
@@ -0,0 +1,58 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+static void copy_pv_rgba4_spec5( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   GLubyte *i810verts = (GLubyte *)imesa->verts;
+   GLuint shift = imesa->vertex_stride_shift;
+   i810Vertex *dst = (i810Vertex *)(i810verts + (edst << shift));
+   i810Vertex *src = (i810Vertex *)(i810verts + (esrc << shift));
+   dst->ui[4] = src->ui[4];
+   dst->ui[5] = src->ui[5];
+}
+
+static void copy_pv_rgba4( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   GLubyte *i810verts = (GLubyte *)imesa->verts;
+   GLuint shift = imesa->vertex_stride_shift;
+   i810Vertex *dst = (i810Vertex *)(i810verts + (edst << shift));
+   i810Vertex *src = (i810Vertex *)(i810verts + (esrc << shift));
+   dst->ui[4] = src->ui[4];
+}
+
+static void copy_pv_rgba3( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   GLubyte *i810verts = (GLubyte *)imesa->verts;
+   GLuint shift = imesa->vertex_stride_shift;
+   i810Vertex *dst = (i810Vertex *)(i810verts + (edst << shift));
+   i810Vertex *src = (i810Vertex *)(i810verts + (esrc << shift));
+   dst->ui[3] = src->ui[3];
+}
diff --git a/src/mesa/tnl_dd/t_dd_dmatmp.h b/src/mesa/tnl_dd/t_dd_dmatmp.h
new file mode 100644
index 0000000000..d9f709389d
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_dmatmp.h
@@ -0,0 +1,1105 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+/* Template for render stages which build and emit vertices directly
+ * to fixed-size dma buffers.  Useful for rendering strips and other
+ * native primitives where clipping and per-vertex tweaks such as
+ * those in t_dd_tritmp.h are not required.
+ *
+ * Produces code for both inline triangles and indexed triangles.
+ * Where various primitive types are unaccelerated by hardware, the
+ * code attempts to fallback to other primitive types (quadstrips to
+ * tristrips, lineloops to linestrips), or to indexed vertices.
+ * Ultimately, a FALLBACK() macro is invoked if there is no way to
+ * render the primitive natively.
+ */
+
+#if !defined(HAVE_TRIANGLES)
+#error "must have at least triangles to use render template"
+#endif
+
+#if !HAVE_ELTS
+#define ELTS_VARS
+#define ALLOC_ELTS( nr )
+#define EMIT_ELT( offset, elt )
+#define EMIT_TWO_ELTS( offset, elt0, elt1 )
+#define INCR_ELTS( nr )
+#define ELT_INIT(prim)
+#define GET_CURRENT_VB_MAX_ELTS() 0
+#define GET_SUBSEQUENT_VB_MAX_ELTS() 0
+#define ALLOC_ELTS_NEW_PRIMITIVE(nr)
+#define RELEASE_ELT_VERTS()
+#define EMIT_INDEXED_VERTS( ctx, start, count )
+#endif
+
+#ifndef EMIT_TWO_ELTS
+#define EMIT_TWO_ELTS( offset, elt0, elt1 )	\
+do { 						\
+   EMIT_ELT( offset, elt0 ); 			\
+   EMIT_ELT( offset+1, elt1 ); 			\
+} while (0)
+#endif
+
+#ifndef FINISH
+#define FINISH
+#endif
+
+/**********************************************************************/
+/*                  Render whole begin/end objects                    */
+/**********************************************************************/
+
+
+
+static GLboolean TAG(emit_elt_verts)( GLcontext *ctx,
+				      GLuint start, GLuint count )
+{
+   if (HAVE_ELTS) {
+      LOCAL_VARS;
+      GLuint nr = count - start;
+
+      if ( nr >= GET_SUBSEQUENT_VB_MAX_VERTS() ) /* assumes same packing for
+						  * indexed and regualar verts
+						  */
+	 return GL_FALSE;
+
+      NEW_PRIMITIVE(); /* finish last prim */
+      EMIT_INDEXED_VERTS( ctx, start, count );
+      return GL_TRUE;
+   } else {
+      return GL_FALSE;
+   }
+}
+
+#if (HAVE_ELTS)
+static void TAG(emit_elts)( GLcontext *ctx, GLuint *elts, GLuint nr )
+{
+   GLint i;
+   LOCAL_VARS;
+   ELTS_VARS;
+
+   ALLOC_ELTS( nr );
+
+   for ( i = 0 ; i < nr ; i+=2, elts += 2 ) {
+      EMIT_TWO_ELTS( 0, elts[0], elts[1] );
+      INCR_ELTS( 2 );
+   }
+}
+#endif
+
+
+/***********************************************************************
+ *                    Render non-indexed primitives.
+ ***********************************************************************/
+
+
+
+static void TAG(render_points_verts)( GLcontext *ctx,
+				      GLuint start,
+				      GLuint count,
+				      GLuint flags )
+{
+   if (HAVE_POINTS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+      GLuint j, nr;
+
+      INIT( GL_POINTS );
+
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j < count; j += nr ) {
+	 nr = MIN2( currentsz, count - j );
+	 EMIT_VERTS( ctx, j, nr );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+static void TAG(render_lines_verts)( GLcontext *ctx,
+				     GLuint start,
+				     GLuint count,
+				     GLuint flags )
+{
+   if (HAVE_LINES) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+      GLuint j, nr;
+
+      INIT( GL_LINES );
+
+      /* Emit whole number of lines in total and in each buffer:
+       */
+      count -= (count-start) & 1;
+      currentsz -= currentsz & 1;
+      dmasz -= dmasz & 1;
+
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j < count; j += nr ) {
+	 nr = MIN2( currentsz, count - j );
+	 EMIT_VERTS( ctx, j, nr );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_line_strip_verts)( GLcontext *ctx,
+					  GLuint start,
+					  GLuint count,
+					  GLuint flags )
+{
+   if (HAVE_LINE_STRIPS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+      GLuint j, nr;
+
+      NEW_PRIMITIVE(); /* always a new primitive */
+      INIT( GL_LINE_STRIP );
+
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j + 1 < count; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j );
+	 EMIT_VERTS( ctx, j, nr );
+	 currentsz = dmasz;
+      }
+ 
+      FINISH;
+
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_line_loop_verts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   if (HAVE_LINE_STRIPS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+      GLuint j, nr;
+
+      NEW_PRIMITIVE();
+      INIT( GL_LINE_STRIP );
+
+      if (flags & PRIM_BEGIN)
+	 j = start;
+      else
+	 j = start + 1;
+
+      /* Ensure last vertex won't wrap buffers:
+       */
+      currentsz--;
+      dmasz--;
+
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      if (j + 1 < count) {
+	 for ( ; j + 1 < count; j += nr - 1 ) {
+	    nr = MIN2( currentsz, count - j );
+	    EMIT_VERTS( ctx, j, nr );
+	    currentsz = dmasz;
+	 }
+
+	 if (start < count - 1 && (flags & PRIM_END))
+	    EMIT_VERTS( ctx, start, 1 );
+      }
+      else if (start + 1 < count && (flags & PRIM_END)) {
+	 EMIT_VERTS( ctx, start+1, 1 );
+	 EMIT_VERTS( ctx, start, 1 );
+      }
+
+      FINISH;
+
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_triangles_verts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   LOCAL_VARS;
+   int dmasz = (GET_SUBSEQUENT_VB_MAX_VERTS()/3) * 3;
+   int currentsz = (GET_CURRENT_VB_MAX_VERTS()/3) * 3;
+   GLuint j, nr;
+
+   INIT(GL_TRIANGLES);
+
+   /* Emit whole number of tris in total.  dmasz is already a multiple
+    * of 3.
+    */
+   count -= (count-start)%3;
+
+   if (currentsz < 8)
+      currentsz = dmasz;
+
+   for (j = start; j < count; j += nr) {
+      nr = MIN2( currentsz, count - j );
+      EMIT_VERTS( ctx, j, nr );
+      currentsz = dmasz;
+   }
+   FINISH;
+}
+
+
+
+static void TAG(render_tri_strip_verts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   if (HAVE_TRI_STRIPS) {
+      LOCAL_VARS;
+      GLuint j, nr;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz;
+
+      INIT(GL_TRIANGLE_STRIP);
+      NEW_PRIMITIVE();
+
+      currentsz = GET_CURRENT_VB_MAX_VERTS();
+
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      if ((flags & PRIM_PARITY) && count - start > 2) {
+	 if (HAVE_TRI_STRIP_1 && 0) {
+	 } else {
+	    EMIT_VERTS( ctx, start, 1 );
+	    currentsz--;
+	 }
+      }
+
+      /* From here on emit even numbers of tris when wrapping over buffers:
+       */
+      dmasz -= (dmasz & 1);
+      currentsz -= (currentsz & 1);
+
+      for (j = start ; j + 2 < count; j += nr - 2 ) {
+	 nr = MIN2( currentsz, count - j );
+	 EMIT_VERTS( ctx, j, nr );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+static void TAG(render_tri_fan_verts)( GLcontext *ctx,
+				       GLuint start,
+				       GLuint count,
+				       GLuint flags )
+{
+   if (HAVE_TRI_FANS) {
+      LOCAL_VARS;
+      GLuint j, nr;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+
+      NEW_PRIMITIVE();
+      INIT(GL_TRIANGLE_FAN);
+
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      for (j = start + 1 ; j + 1 < count; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j + 1 );
+	 EMIT_VERTS( ctx, start, 1 );
+	 EMIT_VERTS( ctx, j, nr - 1 );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   }
+   else {
+      /* Could write code to emit these as indexed vertices (for the
+       * g400, for instance).
+       */
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_poly_verts)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   if (HAVE_POLYGONS) {
+      LOCAL_VARS;
+      GLuint j, nr;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+
+      NEW_PRIMITIVE();
+      INIT(GL_POLYGON);
+
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      for (j = start + 1 ; j + 1 < count ; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j + 1 );
+	 EMIT_VERTS( ctx, start, 1 );
+	 EMIT_VERTS( ctx, j, nr - 1 );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   }
+   else if (HAVE_TRI_FANS && !(ctx->_TriangleCaps & DD_FLATSHADE)) {
+      TAG(render_tri_fan_verts)( ctx, start, count, flags );
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+static void TAG(render_quad_strip_verts)( GLcontext *ctx,
+					  GLuint start,
+					  GLuint count,
+					  GLuint flags )
+{
+   GLuint j, nr;
+
+   if (HAVE_QUAD_STRIPS) {
+      LOCAL_VARS;
+      GLuint j, nr;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz;
+
+      INIT(GL_QUAD_STRIP);
+      NEW_PRIMITIVE();
+
+      currentsz = GET_CURRENT_VB_MAX_VERTS();
+
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      dmasz -= (dmasz & 2);
+      currentsz -= (currentsz & 2);
+
+      for (j = start ; j + 3 < count; j += nr - 2 ) {
+	 nr = MIN2( currentsz, count - j );
+	 EMIT_VERTS( ctx, j, nr );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   } else if (HAVE_TRI_STRIPS && (ctx->_TriangleCaps & DD_FLATSHADE)) {
+      if (TAG(emit_elt_verts)( ctx, start, count )) {
+	 LOCAL_VARS;
+	 int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+	 int currentsz;
+	 GLuint j, nr;
+
+	 /* Simulate flat-shaded quadstrips using indexed vertices:
+	  */
+	 NEW_PRIMITIVE();
+	 ELT_INIT( GL_TRIANGLES );
+
+	 currentsz = GET_CURRENT_VB_MAX_ELTS();
+
+	 /* Emit whole number of quads in total, and in each buffer.
+	  */
+	 dmasz -= dmasz & 1;
+	 count -= (count-start) & 1;
+	 currentsz -= currentsz & 1;
+
+	 if (currentsz < 12)
+	    currentsz = dmasz;
+
+	 currentsz = currentsz/6*2;
+	 dmasz = dmasz/6*2;
+
+	 for (j = start; j + 3 < count; j += nr - 2 ) {
+	    nr = MIN2( currentsz, count - j );
+	    if (nr >= 4) {
+	       GLint quads = (nr/2)-1;
+	       GLint i;
+	       ELTS_VARS;
+
+	       NEW_PRIMITIVE();
+	       ALLOC_ELTS_NEW_PRIMITIVE( quads*6 );
+
+	       for ( i = j-start ; i < j-start+quads*2 ; i+=2 ) {
+		  EMIT_TWO_ELTS( 0, (i+0), (i+1) );
+		  EMIT_TWO_ELTS( 2, (i+2), (i+1) );
+		  EMIT_TWO_ELTS( 4, (i+3), (i+2) );
+		  INCR_ELTS( 6 );
+	       }
+
+	       NEW_PRIMITIVE();
+	    }
+	    currentsz = dmasz;
+	 }
+
+	 RELEASE_ELT_VERTS();
+      }
+      else {
+	 /* Vertices won't fit in a single buffer or elts not available,
+	  * VERT_FALLBACK.
+	  */
+	 VERT_FALLBACK( ctx, start, count, flags );
+      }
+   }
+   else if (HAVE_TRI_STRIPS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_VERTS();
+      int currentsz = GET_CURRENT_VB_MAX_VERTS();
+
+      /* Emit smooth-shaded quadstrips as tristrips:
+       */
+      NEW_PRIMITIVE();
+      INIT( GL_TRIANGLE_STRIP );
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz -= dmasz & 1;
+      currentsz -= currentsz & 1;
+      count -= (count-start) & 1;
+
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      for (j = start; j + 3 < count; j += nr - 2 ) {
+	 nr = MIN2( currentsz, count - j );
+	 EMIT_VERTS( ctx, j, nr );
+	 currentsz = dmasz;
+      }
+
+      FINISH;
+
+   } else {
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_quads_verts)( GLcontext *ctx,
+				     GLuint start,
+				     GLuint count,
+				     GLuint flags )
+{
+   if (HAVE_QUADS) {
+      LOCAL_VARS;
+      int dmasz = (GET_SUBSEQUENT_VB_MAX_VERTS()/4) * 4;
+      int currentsz = (GET_CURRENT_VB_MAX_VERTS()/4) * 4;
+      GLuint j, nr;
+
+      INIT(GL_QUADS);
+
+      /* Emit whole number of quads in total.  dmasz is already a multiple
+       * of 4.
+       */
+      count -= (count-start)%4;
+
+      if (currentsz < 8)
+         currentsz = dmasz;
+
+      for (j = start; j < count; j += nr) {
+         nr = MIN2( currentsz, count - j );
+         EMIT_VERTS( ctx, j, nr );
+         currentsz = dmasz;
+      }
+      FINISH;
+   } else if (TAG(emit_elt_verts)( ctx, start, count )) {
+      /* Hardware doesn't have a quad primitive type -- try to
+       * simulate it using indexed vertices and the triangle
+       * primitive:
+       */
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint j, nr;
+
+      NEW_PRIMITIVE();
+      ELT_INIT( GL_TRIANGLES );
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz -= dmasz & 3;
+      count -= (count-start) & 3;
+      currentsz -= currentsz & 3;
+
+      /* Adjust for rendering as triangles:
+       */
+      currentsz = currentsz/6*4;
+      dmasz = dmasz/6*4;
+
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j < count; j += nr ) {
+	 nr = MIN2( currentsz, count - j );
+	 if (nr >= 4) {
+	    GLint quads = nr/4;
+	    GLint i;
+	    ELTS_VARS;
+
+	    NEW_PRIMITIVE();
+	    ALLOC_ELTS_NEW_PRIMITIVE( quads*6 );
+
+	    for ( i = j-start ; i < j-start+quads*4 ; i+=4 ) {
+	       EMIT_TWO_ELTS( 0, (i+0), (i+1) );
+	       EMIT_TWO_ELTS( 2, (i+3), (i+1) );
+	       EMIT_TWO_ELTS( 4, (i+2), (i+3) );
+	       INCR_ELTS( 6 );
+	    }
+
+	    NEW_PRIMITIVE();
+	 }
+	 currentsz = dmasz;
+      }
+
+      RELEASE_ELT_VERTS();
+   }
+   else {
+      /* Vertices won't fit in a single buffer, fallback.
+       */
+      VERT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+static void TAG(render_noop)( GLcontext *ctx,
+			      GLuint start,
+			      GLuint count,
+			      GLuint flags )
+{
+}
+
+
+
+
+static render_func TAG(render_tab_verts)[GL_POLYGON+2] =
+{
+   TAG(render_points_verts),
+   TAG(render_lines_verts),
+   TAG(render_line_loop_verts),
+   TAG(render_line_strip_verts),
+   TAG(render_triangles_verts),
+   TAG(render_tri_strip_verts),
+   TAG(render_tri_fan_verts),
+   TAG(render_quads_verts),
+   TAG(render_quad_strip_verts),
+   TAG(render_poly_verts),
+   TAG(render_noop),
+};
+
+
+/****************************************************************************
+ *                 Render elts using hardware indexed verts                 *
+ ****************************************************************************/
+
+#if (HAVE_ELTS)
+static void TAG(render_points_elts)( GLcontext *ctx,
+				     GLuint start,
+				     GLuint count,
+				     GLuint flags )
+{
+   if (HAVE_POINTS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      GLuint j, nr;
+
+      ELT_INIT( GL_POINTS );
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j < count; j += nr ) {
+	 nr = MIN2( currentsz, count - j );
+	 TAG(emit_elts)( ctx, elts+j, nr );
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   } else {
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+
+static void TAG(render_lines_elts)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   if (HAVE_LINES) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      GLuint j, nr;
+
+      ELT_INIT( GL_LINES );
+
+      /* Emit whole number of lines in total and in each buffer:
+       */
+      count -= (count-start) & 1;
+      currentsz -= currentsz & 1;
+      dmasz -= dmasz & 1;
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j < count; j += nr ) {
+	 nr = MIN2( currentsz, count - j );
+	 TAG(emit_elts)( ctx, elts+j, nr );
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   } else {
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_line_strip_elts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   if (HAVE_LINE_STRIPS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      GLuint j, nr;
+
+      NEW_PRIMITIVE(); /* always a new primitive */
+      ELT_INIT( GL_LINE_STRIP );
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j + 1 < count; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j );
+	 TAG(emit_elts)( ctx, elts+j, nr );
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   } else {
+      /* TODO: Try to emit as indexed lines.
+       */
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_line_loop_elts)( GLcontext *ctx,
+					GLuint start,
+					GLuint count,
+					GLuint flags )
+{
+   if (HAVE_LINE_STRIPS) {
+      LOCAL_VARS;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      GLuint j, nr;
+
+      NEW_PRIMITIVE();
+      ELT_INIT( GL_LINE_STRIP );
+
+      if (flags & PRIM_BEGIN)
+	 j = start;
+      else
+	 j = start + 1;
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      /* Ensure last vertex doesn't wrap:
+       */
+      currentsz--;
+      dmasz--;
+
+      for ( ; j + 1 < count; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j );
+/*  	 NEW_PRIMITIVE(); */
+	 TAG(emit_elts)( ctx, elts+j, nr );
+	 currentsz = dmasz;
+      }
+
+      if (flags & PRIM_END)
+	 TAG(emit_elts)( ctx, elts+start, 1 );
+
+      NEW_PRIMITIVE();
+   } else {
+      /* TODO: Try to emit as indexed lines */
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+/* For verts, we still eliminate the copy from main memory to dma
+ * buffers.  For elts, this is probably no better (worse?) than the
+ * standard path.
+ */
+static void TAG(render_triangles_elts)( GLcontext *ctx,
+					GLuint start,
+					GLuint count,
+					GLuint flags )
+{
+   LOCAL_VARS;
+   GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+   int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS()/3*3;
+   int currentsz;
+   GLuint j, nr;
+
+   NEW_PRIMITIVE();
+   ELT_INIT( GL_TRIANGLES );
+
+   currentsz = GET_CURRENT_VB_MAX_ELTS();
+
+   /* Emit whole number of tris in total.  dmasz is already a multiple
+    * of 3.
+    */
+   count -= (count-start)%3;
+   currentsz -= currentsz%3;
+   if (currentsz < 8)
+      currentsz = dmasz;
+
+   for (j = start; j < count; j += nr) {
+      nr = MIN2( currentsz, count - j );
+      TAG(emit_elts)( ctx, elts+j, nr );
+      NEW_PRIMITIVE();
+      currentsz = dmasz;
+   }
+}
+
+
+
+static void TAG(render_tri_strip_elts)( GLcontext *ctx,
+					GLuint start,
+					GLuint count,
+					GLuint flags )
+{
+   if (HAVE_TRI_STRIPS) {
+      LOCAL_VARS;
+      GLuint j, nr;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+
+      NEW_PRIMITIVE();
+      ELT_INIT( GL_TRIANGLE_STRIP );
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      if ((flags & PRIM_PARITY) && count - start > 2) {
+	 TAG(emit_elts)( ctx, elts+start, 1 );
+      }
+
+      /* Keep the same winding over multiple buffers:
+       */
+      dmasz -= (dmasz & 1);
+      currentsz -= (currentsz & 1);
+
+      for (j = start ; j + 2 < count; j += nr - 2 ) {
+	 nr = MIN2( currentsz, count - j );
+	 TAG(emit_elts)( ctx, elts+j, nr );
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   } else {
+      /* TODO: try to emit as indexed triangles */
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+static void TAG(render_tri_fan_elts)( GLcontext *ctx,
+				      GLuint start,
+				      GLuint count,
+				      GLuint flags )
+{
+   if (HAVE_TRI_FANS) {
+      LOCAL_VARS;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      GLuint j, nr;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+
+      NEW_PRIMITIVE();
+      ELT_INIT( GL_TRIANGLE_FAN );
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      for (j = start + 1 ; j + 1 < count; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j + 1 );
+	 TAG(emit_elts)( ctx, elts+start, 1 );
+	 TAG(emit_elts)( ctx, elts+j, nr - 1 );
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   } else {
+      /* TODO: try to emit as indexed triangles */
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+
+static void TAG(render_poly_elts)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   if (HAVE_POLYGONS && 0) {
+   } else if (HAVE_TRI_FANS && !(ctx->_TriangleCaps & DD_FLATSHADE)) {
+      LOCAL_VARS;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      GLuint j, nr;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+
+      NEW_PRIMITIVE();
+      ELT_INIT( GL_TRIANGLE_FAN );
+
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+      if (currentsz < 8) {
+	 NEW_BUFFER();
+	 currentsz = dmasz;
+      }
+
+      for (j = start + 1 ; j + 1 < count ; j += nr - 1 ) {
+	 nr = MIN2( currentsz, count - j + 1 );
+	 TAG(emit_elts)( ctx, elts+start, 1 );
+	 TAG(emit_elts)( ctx, elts+j, nr - 1 );
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   } else {
+      ELT_FALLBACK( ctx, start, count, flags );
+   }
+}
+
+static void TAG(render_quad_strip_elts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   if (HAVE_QUAD_STRIPS && 0) {
+   }
+   else if (HAVE_TRI_STRIPS) {
+      LOCAL_VARS;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint j, nr;
+
+      NEW_PRIMITIVE();
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz -= dmasz & 1;
+      count -= (count-start) & 1;
+      currentsz -= currentsz & 1;
+
+      if (currentsz < 12)
+	 currentsz = dmasz;
+
+      if (ctx->_TriangleCaps & DD_FLATSHADE) {
+	 ELT_INIT( GL_TRIANGLES );
+
+	 currentsz = currentsz/6*2;
+	 dmasz = dmasz/6*2;
+
+	 for (j = start; j + 3 < count; j += nr - 2 ) {
+	    nr = MIN2( currentsz, count - j );
+
+	    if (nr >= 4)
+	    {
+	       GLint i;
+	       GLint quads = (nr/2)-1;
+	       ELTS_VARS;
+
+
+	       NEW_PRIMITIVE();
+	       ALLOC_ELTS_NEW_PRIMITIVE( quads*6 );
+
+	       for ( i = j-start ; i < j-start+quads ; i++, elts += 2 ) {
+		  EMIT_TWO_ELTS( 0, elts[0], elts[1] );
+		  EMIT_TWO_ELTS( 2, elts[2], elts[1] );
+		  EMIT_TWO_ELTS( 4, elts[3], elts[2] );
+		  INCR_ELTS( 6 );
+	       }
+
+	       NEW_PRIMITIVE();
+	    }
+
+	    currentsz = dmasz;
+	 }
+      }
+      else {
+	 ELT_INIT( GL_TRIANGLE_STRIP );
+
+	 for (j = start; j + 3 < count; j += nr - 2 ) {
+	    nr = MIN2( currentsz, count - j );
+	    TAG(emit_elts)( ctx, elts+j, nr );
+	    NEW_PRIMITIVE();
+	    currentsz = dmasz;
+	 }
+      }
+   }
+}
+
+
+static void TAG(render_quads_elts)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   if (HAVE_QUADS && 0) {
+   } else {
+      LOCAL_VARS;
+      GLuint *elts = TNL_CONTEXT(ctx)->vb.Elts;
+      int dmasz = GET_SUBSEQUENT_VB_MAX_ELTS();
+      int currentsz;
+      GLuint j, nr;
+
+      ELT_INIT( GL_TRIANGLES );
+      currentsz = GET_CURRENT_VB_MAX_ELTS();
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz -= dmasz & 3;
+      count -= (count-start) & 3;
+      currentsz -= currentsz & 3;
+
+      /* Adjust for rendering as triangles:
+       */
+      currentsz = currentsz/6*4;
+      dmasz = dmasz/6*4;
+
+      if (currentsz < 8)
+	 currentsz = dmasz;
+
+      for (j = start; j + 3 < count; j += nr - 2 ) {
+	 nr = MIN2( currentsz, count - j );
+
+	 if (nr >= 4)
+	 {
+	    GLint quads = nr/4;
+	    GLint i;
+	    ELTS_VARS;
+	    NEW_PRIMITIVE();
+	    ALLOC_ELTS_NEW_PRIMITIVE( quads * 6 );
+
+	    for ( i = j-start ; i < j-start+quads ; i++, elts += 4 ) {
+	       EMIT_TWO_ELTS( 0, elts[0], elts[1] );
+	       EMIT_TWO_ELTS( 2, elts[3], elts[1] );
+	       EMIT_TWO_ELTS( 4, elts[2], elts[3] );
+	       INCR_ELTS( 6 );
+	    }
+	 }
+
+	 NEW_PRIMITIVE();
+	 currentsz = dmasz;
+      }
+   }
+}
+
+
+
+static render_func TAG(render_tab_elts)[GL_POLYGON+2] =
+{
+   TAG(render_points_elts),
+   TAG(render_lines_elts),
+   TAG(render_line_loop_elts),
+   TAG(render_line_strip_elts),
+   TAG(render_triangles_elts),
+   TAG(render_tri_strip_elts),
+   TAG(render_tri_fan_elts),
+   TAG(render_quads_elts),
+   TAG(render_quad_strip_elts),
+   TAG(render_poly_elts),
+   TAG(render_noop),
+};
+#endif
diff --git a/src/mesa/tnl_dd/t_dd_dmatmp2.h b/src/mesa/tnl_dd/t_dd_dmatmp2.h
new file mode 100644
index 0000000000..3f58d24c95
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_dmatmp2.h
@@ -0,0 +1,949 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  4.0.3
+ *
+ * Copyright (C) 1999-2002  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>
+ */
+
+
+/* Template for render stages which build and emit vertices directly
+ * to fixed-size dma buffers.  Useful for rendering strips and other
+ * native primitives where clipping and per-vertex tweaks such as
+ * those in t_dd_tritmp.h are not required.
+ *
+ */
+
+#if !HAVE_TRIANGLES || !HAVE_POINTS || !HAVE_LINES
+#error "must have points, lines & triangles to use render template"
+#endif
+
+#if !HAVE_TRI_STRIPS || !HAVE_TRI_FANS
+#error "must have tri strip and fans to use render template"
+#endif
+
+#if !HAVE_LINE_STRIPS
+#error "must have line strips to use render template"
+#endif
+
+#if !HAVE_POLYGONS
+#error "must have polygons to use render template"
+#endif
+
+#if !HAVE_ELTS
+#error "must have elts to use render template"
+#endif
+
+
+#ifndef EMIT_TWO_ELTS
+#define EMIT_TWO_ELTS( dest, offset, elt0, elt1 )	\
+do { 						\
+   (dest)[offset] = (elt0); 			\
+   (dest)[offset+1] = (elt1); 			\
+} while (0)
+#endif
+
+
+/**********************************************************************/
+/*                  Render whole begin/end objects                    */
+/**********************************************************************/
+
+
+static ELT_TYPE *TAG(emit_elts)( GLcontext *ctx, 
+			    ELT_TYPE *dest,
+			    GLuint *elts, GLuint nr )
+{
+   GLint i;
+   LOCAL_VARS;
+
+   for ( i = 0 ; i+1 < nr ; i+=2, elts += 2 ) {
+      EMIT_TWO_ELTS( dest, 0, elts[0], elts[1] );
+      dest += 2;
+   }
+   if (i < nr) {
+      EMIT_ELT( dest, 0, elts[0] );
+      dest += 1;
+   }
+   
+   return dest;
+}
+
+static ELT_TYPE *TAG(emit_consecutive_elts)( GLcontext *ctx, 
+					ELT_TYPE *dest,
+					GLuint start, GLuint nr )
+{
+   GLint i;
+   LOCAL_VARS;
+
+   for ( i = 0 ; i+1 < nr ; i+=2, start += 2 ) {
+      EMIT_TWO_ELTS( dest, 0, start, start+1 );
+      dest += 2;
+   }
+   if (i < nr) {
+      EMIT_ELT( dest, 0, start );
+      dest += 1;
+   }
+
+   return dest;
+}
+
+/***********************************************************************
+ *                    Render non-indexed primitives.
+ ***********************************************************************/
+
+
+
+static void TAG(render_points_verts)( GLcontext *ctx,
+				      GLuint start,
+				      GLuint count,
+				      GLuint flags )
+{
+   if (start < count) {
+      LOCAL_VARS;
+      if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+      EMIT_PRIM( ctx, GL_POINTS, HW_POINTS, start, count );
+   }
+}
+
+static void TAG(render_lines_verts)( GLcontext *ctx,
+				     GLuint start,
+				     GLuint count,
+				     GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+   count -= (count-start) & 1;
+
+   if (start+1 >= count)
+      return;
+
+   if ((flags & PRIM_BEGIN) && ctx->Line.StippleFlag) {
+      RESET_STIPPLE();
+      AUTO_STIPPLE( GL_TRUE );
+   }
+      
+   EMIT_PRIM( ctx, GL_LINES, HW_LINES, start, count );
+
+   if ((flags & PRIM_END) && ctx->Line.StippleFlag)
+      AUTO_STIPPLE( GL_FALSE );
+}
+
+
+static void TAG(render_line_strip_verts)( GLcontext *ctx,
+					  GLuint start,
+					  GLuint count,
+					  GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (start+1 >= count)
+      return;
+
+   if ((flags & PRIM_BEGIN) && ctx->Line.StippleFlag)
+      RESET_STIPPLE();
+
+
+   if (PREFER_DISCRETE_ELT_PRIM( count-start, HW_LINES ))
+   {   
+      int dmasz = GET_MAX_HW_ELTS();
+      GLuint j, nr;
+
+      ELT_INIT( GL_LINES, HW_LINES );
+
+      /* Emit whole number of lines in each full buffer.
+       */
+      dmasz = dmasz/2;
+
+
+      for (j = start; j + 1 < count; j += nr - 1 ) {
+	 ELT_TYPE *dest;
+	 GLint i;
+
+	 nr = MIN2( dmasz, count - j );
+	 dest = ALLOC_ELTS( (nr-1)*2 );
+	    
+	 for ( i = j ; i+1 < j+nr ; i+=1 ) {
+	    EMIT_TWO_ELTS( dest, 0, (i+0), (i+1) );
+	    dest += 2;
+	 }
+
+	 CLOSE_ELTS();
+      }
+   }
+   else
+      EMIT_PRIM( ctx, GL_LINE_STRIP, HW_LINE_STRIP, start, count );
+}
+
+
+static void TAG(render_line_loop_verts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   LOCAL_VARS;
+   GLuint j, nr;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (flags & PRIM_BEGIN) {
+      j = start;
+      if (ctx->Line.StippleFlag)
+	 RESET_STIPPLE( );
+   }
+   else
+      j = start + 1;
+
+   if (flags & PRIM_END) {
+
+      if (start+1 >= count)
+	 return;
+
+      if (PREFER_DISCRETE_ELT_PRIM( count-start, HW_LINES )) {
+	 int dmasz = GET_MAX_HW_ELTS();
+
+	 ELT_INIT( GL_LINES, HW_LINES );
+
+	 /* Emit whole number of lines in each full buffer.
+	  */
+	 dmasz = dmasz/2;
+
+	 /* Ensure last vertex doesn't wrap:
+	  */
+	 dmasz--;
+
+	 for (; j + 1 < count;  ) {
+	    GLint i;
+	    ELT_TYPE *dest;
+
+	    nr = MIN2( dmasz, count - j );
+	    dest = ALLOC_ELTS( nr*2 );	/* allocs room for 1 more line */
+
+	    for ( i = 0 ; i < nr - 1 ; i+=1 ) {
+	       EMIT_TWO_ELTS( dest, 0, (j+i), (j+i+1) );
+	       dest += 2;
+	    }
+
+	    j += nr - 1;
+
+	    /* Emit 1 more line into space alloced above */
+	    if (j + 1 >= count) {
+ 	       EMIT_TWO_ELTS( dest, 0, (j), (start) ); 
+ 	       dest += 2; 
+ 	    }
+ 
+	    CLOSE_ELTS();
+	 }
+      }
+      else
+      {
+	 int dmasz = GET_MAX_HW_ELTS() - 1;
+
+	 ELT_INIT( GL_LINE_STRIP, HW_LINE_STRIP );
+
+	 for ( ; j + 1 < count;  ) {
+	    nr = MIN2( dmasz, count - j );
+	    if (j + nr < count) {
+	       ELT_TYPE *dest = ALLOC_ELTS( nr );
+	       dest = TAG(emit_consecutive_elts)( ctx, dest, j, nr );
+	       j += nr - 1;
+	       CLOSE_ELTS();
+	    }
+	    else if (nr) {
+	       ELT_TYPE *dest = ALLOC_ELTS( nr + 1 );
+	       dest = TAG(emit_consecutive_elts)( ctx, dest, j, nr );
+	       dest = TAG(emit_consecutive_elts)( ctx, dest, start, 1 );
+	       j += nr;
+	       CLOSE_ELTS();
+	    }
+	 }   
+      }
+   } else {
+      TAG(render_line_strip_verts)( ctx, j, count, flags );
+   }
+}
+
+
+static void TAG(render_triangles_verts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   count -= (count-start)%3;
+
+   if (start+2 >= count) {
+      return;
+   }
+
+   /* need a PREFER_DISCRETE_ELT_PRIM here too..
+    */
+   EMIT_PRIM( ctx, GL_TRIANGLES, HW_TRIANGLES, start, count );
+}
+
+
+
+static void TAG(render_tri_strip_verts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (start + 2 >= count)
+      return;
+
+   if (PREFER_DISCRETE_ELT_PRIM( count-start, HW_TRIANGLES ))
+   {   
+      int dmasz = GET_MAX_HW_ELTS();
+      int parity = 0;
+      GLuint j, nr;
+
+      ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+      if (flags & PRIM_PARITY)
+	 parity = 1;
+
+      /* Emit even number of tris in each full buffer.
+       */
+      dmasz = dmasz/3;
+      dmasz -= dmasz & 1;
+
+      for (j = start; j + 2 < count; j += nr - 2 ) {
+	 ELT_TYPE *dest;
+	 GLint i;
+
+	 nr = MIN2( dmasz, count - j );
+	 dest = ALLOC_ELTS( (nr-2)*3 );
+	    
+	 for ( i = j ; i+2 < j+nr ; i++, parity^=1 ) {
+	    EMIT_ELT( dest, 0, (i+0+parity) );
+	    EMIT_ELT( dest, 1, (i+1-parity) );
+	    EMIT_ELT( dest, 2, (i+2) );
+	    dest += 3;
+	 }
+
+	 CLOSE_ELTS();
+      }
+   }
+   else if ((flags & PRIM_PARITY) == 0)  
+      EMIT_PRIM( ctx, GL_TRIANGLE_STRIP, HW_TRIANGLE_STRIP_0, start, count );
+   else if (HAVE_TRI_STRIP_1)
+      EMIT_PRIM( ctx, GL_TRIANGLE_STRIP, HW_TRIANGLE_STRIP_1, start, count );
+   else {
+      /* Emit the first triangle with elts, then the rest as a regular strip.
+       * TODO:  Make this unlikely in t_imm_api.c
+       */
+      ELT_TYPE *dest;
+
+      ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+      dest = ALLOC_ELTS( 3 );
+      EMIT_ELT( dest, 0, (start+1) );
+      EMIT_ELT( dest, 1, (start+0) );
+      EMIT_ELT( dest, 2, (start+2) );
+      dest += 3;
+      CLOSE_ELTS();
+
+      start++;
+      if (start + 2 >= count)
+	 return;
+
+      EMIT_PRIM( ctx, GL_TRIANGLE_STRIP, HW_TRIANGLE_STRIP_0, start, 
+		 count );
+   }
+}
+
+static void TAG(render_tri_fan_verts)( GLcontext *ctx,
+				       GLuint start,
+				       GLuint count,
+				       GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (start+2 >= count) 
+      return;
+
+   if (PREFER_DISCRETE_ELT_PRIM( count-start, HW_TRIANGLES ))
+   {   
+      int dmasz = GET_MAX_HW_ELTS();
+      GLuint j, nr;
+
+      ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+      dmasz = dmasz/3;
+
+      for (j = start + 1; j + 1 < count; j += nr - 1 ) {
+	 ELT_TYPE *dest;
+	 GLint i;
+
+	 nr = MIN2( dmasz, count - j );
+	 dest = ALLOC_ELTS( (nr-1)*3 );
+	    
+	 for ( i = j ; i+1 < j+nr ; i++ ) {
+	    EMIT_ELT( dest, 0, (start) );
+	    EMIT_ELT( dest, 1, (i) );
+	    EMIT_ELT( dest, 2, (i+1) );
+	    dest += 3;
+	 }
+	 
+	 CLOSE_ELTS();
+      }
+   }
+   else {
+      EMIT_PRIM( ctx, GL_TRIANGLE_FAN, HW_TRIANGLE_FAN, start, count );
+   }
+}
+
+
+static void TAG(render_poly_verts)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (start+2 >= count) 
+      return;
+
+   EMIT_PRIM( ctx, GL_POLYGON, HW_POLYGON, start, count );
+}
+
+static void TAG(render_quad_strip_verts)( GLcontext *ctx,
+					  GLuint start,
+					  GLuint count,
+					  GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   count -= (count-start) & 1;
+
+   if (start+3 >= count) 
+      return;
+
+   if (HAVE_QUAD_STRIPS) {
+      EMIT_PRIM( ctx, GL_QUAD_STRIP, HW_QUAD_STRIP, start, count );
+   } 
+   else if (ctx->_TriangleCaps & DD_FLATSHADE) {
+      LOCAL_VARS;
+      int dmasz = GET_MAX_HW_ELTS();
+      GLuint j, nr;
+
+      ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz = (dmasz/6)*2;
+
+      for (j = start; j + 3 < count; j += nr - 2 ) {
+	 ELT_TYPE *dest;
+	 GLint quads, i;
+
+	 nr = MIN2( dmasz, count - j );
+	 quads = (nr/2)-1;
+	 dest = ALLOC_ELTS( quads*6 );
+	    
+	 for ( i = j ; i < j+quads*2 ; i+=2 ) {
+	    EMIT_TWO_ELTS( dest, 0, (i+0), (i+1) );
+	    EMIT_TWO_ELTS( dest, 2, (i+2), (i+1) );
+	    EMIT_TWO_ELTS( dest, 4, (i+3), (i+2) );
+	    dest += 6;
+	 }
+
+	 CLOSE_ELTS();
+      }
+   }
+   else {
+      EMIT_PRIM( ctx, GL_TRIANGLE_STRIP, HW_TRIANGLE_STRIP_0, start, count );
+   }
+}
+
+
+static void TAG(render_quads_verts)( GLcontext *ctx,
+				     GLuint start,
+				     GLuint count,
+				     GLuint flags )
+{
+   LOCAL_VARS;
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+   count -= (count-start)%4;
+
+   if (start+3 >= count) 
+      return;
+
+   if (HAVE_QUADS) {
+      EMIT_PRIM( ctx, HW_QUADS, GL_QUADS, start, count );
+   } 
+   else {
+      /* Hardware doesn't have a quad primitive type -- simulate it
+       * using indexed vertices and the triangle primitive: 
+       */
+      LOCAL_VARS;
+      int dmasz = GET_MAX_HW_ELTS();
+      GLuint j, nr;
+
+      ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+      /* Adjust for rendering as triangles:
+       */
+      dmasz = (dmasz/6)*4;
+
+      for (j = start; j < count; j += nr ) {
+	 ELT_TYPE *dest;
+	 GLint quads, i;
+
+	 nr = MIN2( dmasz, count - j );
+	 quads = nr/4;
+	 dest = ALLOC_ELTS( quads*6 );
+
+	 for ( i = j ; i < j+quads*4 ; i+=4 ) {
+	    EMIT_TWO_ELTS( dest, 0, (i+0), (i+1) );
+	    EMIT_TWO_ELTS( dest, 2, (i+3), (i+1) );
+	    EMIT_TWO_ELTS( dest, 4, (i+2), (i+3) );
+	    dest += 6;
+	 }
+
+	 CLOSE_ELTS();
+      }
+   }
+}
+
+static void TAG(render_noop)( GLcontext *ctx,
+			      GLuint start,
+			      GLuint count,
+			      GLuint flags )
+{
+}
+
+
+
+
+static render_func TAG(render_tab_verts)[GL_POLYGON+2] =
+{
+   TAG(render_points_verts),
+   TAG(render_lines_verts),
+   TAG(render_line_loop_verts),
+   TAG(render_line_strip_verts),
+   TAG(render_triangles_verts),
+   TAG(render_tri_strip_verts),
+   TAG(render_tri_fan_verts),
+   TAG(render_quads_verts),
+   TAG(render_quad_strip_verts),
+   TAG(render_poly_verts),
+   TAG(render_noop),
+};
+
+
+/****************************************************************************
+ *                 Render elts using hardware indexed verts                 *
+ ****************************************************************************/
+
+static void TAG(render_points_elts)( GLcontext *ctx,
+				     GLuint start,
+				     GLuint count,
+				     GLuint flags )
+{
+   LOCAL_VARS;
+   int dmasz = GET_MAX_HW_ELTS();
+   GLuint *elts = GET_MESA_ELTS();
+   GLuint j, nr;
+   ELT_TYPE *dest;
+
+   ELT_INIT( GL_POINTS, HW_POINTS );
+
+   for (j = start; j < count; j += nr ) {
+      nr = MIN2( dmasz, count - j );
+      dest = ALLOC_ELTS( nr );
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+      CLOSE_ELTS();
+   }
+}
+
+
+
+static void TAG(render_lines_elts)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   LOCAL_VARS;
+   int dmasz = GET_MAX_HW_ELTS();
+   GLuint *elts = GET_MESA_ELTS();
+   GLuint j, nr;
+   ELT_TYPE *dest;
+
+   if (start+1 >= count)
+      return;
+
+   if ((flags & PRIM_BEGIN) && ctx->Line.StippleFlag) {
+      RESET_STIPPLE();
+      AUTO_STIPPLE( GL_TRUE );
+   }
+
+   ELT_INIT( GL_LINES, HW_LINES );
+
+   /* Emit whole number of lines in total and in each buffer:
+    */
+   count -= (count-start) & 1;
+   dmasz -= dmasz & 1;
+
+   for (j = start; j < count; j += nr ) {
+      nr = MIN2( dmasz, count - j );
+      dest = ALLOC_ELTS( nr );
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+      CLOSE_ELTS();
+   }
+
+   if ((flags & PRIM_END) && ctx->Line.StippleFlag)
+      AUTO_STIPPLE( GL_FALSE );
+}
+
+
+static void TAG(render_line_strip_elts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   LOCAL_VARS;
+   int dmasz = GET_MAX_HW_ELTS();
+   GLuint *elts = GET_MESA_ELTS();
+   GLuint j, nr;
+   ELT_TYPE *dest;
+
+   if (start+1 >= count)
+      return;
+
+   ELT_INIT( GL_LINE_STRIP, HW_LINE_STRIP );
+
+   if ((flags & PRIM_BEGIN) && ctx->Line.StippleFlag)
+      RESET_STIPPLE();
+
+   for (j = start; j + 1 < count; j += nr - 1 ) {
+      nr = MIN2( dmasz, count - j );
+      dest = ALLOC_ELTS( nr );
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+      CLOSE_ELTS();
+   }
+}
+
+
+static void TAG(render_line_loop_elts)( GLcontext *ctx,
+					GLuint start,
+					GLuint count,
+					GLuint flags )
+{
+   LOCAL_VARS;
+   int dmasz = GET_MAX_HW_ELTS();
+   GLuint *elts = GET_MESA_ELTS();
+   GLuint j, nr;
+   ELT_TYPE *dest;
+
+   if (0) fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (flags & PRIM_BEGIN)
+      j = start;
+   else
+      j = start + 1;
+
+   
+   if (flags & PRIM_END) {
+      if (start+1 >= count)
+	 return;
+   } 
+   else {
+      if (j+1 >= count)
+	 return;
+   }
+
+   ELT_INIT( GL_LINE_STRIP, HW_LINE_STRIP );
+
+   if ((flags & PRIM_BEGIN) && ctx->Line.StippleFlag)
+      RESET_STIPPLE();
+
+   
+   /* Ensure last vertex doesn't wrap:
+    */
+   dmasz--;
+
+   for ( ; j + 1 < count; ) {
+      nr = MIN2( dmasz, count - j );
+      dest = ALLOC_ELTS( nr+1 );	/* Reserve possible space for last elt */
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+      j += nr - 1;
+      if (j + 1 >= count && (flags & PRIM_END)) {
+	 dest = TAG(emit_elts)( ctx, dest, elts+start, 1 );
+      }
+      CLOSE_ELTS();
+   }
+}
+
+
+static void TAG(render_triangles_elts)( GLcontext *ctx,
+					GLuint start,
+					GLuint count,
+					GLuint flags )
+{
+   LOCAL_VARS;
+   GLuint *elts = GET_MESA_ELTS();
+   int dmasz = GET_MAX_HW_ELTS()/3*3;
+   GLuint j, nr;
+   ELT_TYPE *dest;
+
+   if (start+2 >= count)
+      return;
+
+   ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+
+   /* Emit whole number of tris in total.  dmasz is already a multiple
+    * of 3.
+    */
+   count -= (count-start)%3;
+
+   for (j = start; j < count; j += nr) {
+      nr = MIN2( dmasz, count - j );
+      dest = ALLOC_ELTS( nr );
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+      CLOSE_ELTS();
+   }
+}
+
+
+
+static void TAG(render_tri_strip_elts)( GLcontext *ctx,
+					GLuint start,
+					GLuint count,
+					GLuint flags )
+{
+   LOCAL_VARS;
+   GLuint j, nr;
+   GLuint *elts = GET_MESA_ELTS();
+   int dmasz = GET_MAX_HW_ELTS();
+   ELT_TYPE *dest;
+
+   if (start+2 >= count)
+      return;
+
+   ELT_INIT( GL_TRIANGLE_STRIP, HW_TRIANGLE_STRIP_0 );
+
+   /* Keep the same winding over multiple buffers:
+    */
+   dmasz -= (dmasz & 1);
+
+   for (j = start ; j + 2 < count; j += nr - 2 ) {
+      nr = MIN2( dmasz, count - j );
+
+      if (flags & PRIM_PARITY) {
+	 dest = ALLOC_ELTS( nr );	
+	 dest = TAG(emit_elts)( ctx, dest, elts+j, 1 );
+	 dest = TAG(emit_elts)( ctx, dest, elts+j, nr-1 );
+	 nr--; flags &= ~PRIM_PARITY;
+	 CLOSE_ELTS();
+      }
+      else {
+	 dest = ALLOC_ELTS( nr );
+	 dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+	 CLOSE_ELTS();
+      }
+   }
+}
+
+static void TAG(render_tri_fan_elts)( GLcontext *ctx,
+				      GLuint start,
+				      GLuint count,
+				      GLuint flags )
+{
+   LOCAL_VARS;
+   GLuint *elts = GET_MESA_ELTS();
+   GLuint j, nr;
+   int dmasz = GET_MAX_HW_ELTS();
+   ELT_TYPE *dest;
+
+   if (start+2 >= count)
+      return;
+
+   ELT_INIT( GL_TRIANGLE_FAN, HW_TRIANGLE_FAN );
+
+   for (j = start + 1 ; j + 1 < count; j += nr - 1 ) {
+      nr = MIN2( dmasz, count - j + 1 );
+      dest = ALLOC_ELTS( nr );
+      dest = TAG(emit_elts)( ctx, dest, elts+start, 1 );
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr - 1 );
+      CLOSE_ELTS();
+   }
+}
+
+
+static void TAG(render_poly_elts)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   LOCAL_VARS;
+   GLuint *elts = GET_MESA_ELTS();
+   GLuint j, nr;
+   int dmasz = GET_MAX_HW_ELTS();
+   ELT_TYPE *dest;
+
+   if (start+2 >= count)
+      return;
+
+   ELT_INIT( GL_POLYGON, HW_POLYGON );
+
+   for (j = start + 1 ; j + 1 < count ; j += nr - 1 ) {
+      nr = MIN2( dmasz, count - j + 1 );
+      dest = ALLOC_ELTS( nr );
+      dest = TAG(emit_elts)( ctx, dest, elts+start, 1 );
+      dest = TAG(emit_elts)( ctx, dest, elts+j, nr - 1 );
+      CLOSE_ELTS();
+   }
+}
+
+static void TAG(render_quad_strip_elts)( GLcontext *ctx,
+					 GLuint start,
+					 GLuint count,
+					 GLuint flags )
+{
+   if (start+3 >= count)
+      return;
+
+   if (HAVE_QUAD_STRIPS && 0) {
+   }
+   else {
+      LOCAL_VARS;
+      GLuint *elts = GET_MESA_ELTS();
+      int dmasz = GET_MAX_HW_ELTS();
+      GLuint j, nr;
+      ELT_TYPE *dest;
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz -= dmasz & 1;
+      count -= (count-start) & 1;
+
+      if (ctx->_TriangleCaps & DD_FLATSHADE) {
+	 ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+	 dmasz = dmasz/6*2;
+
+	 for (j = start; j + 3 < count; j += nr - 2 ) {
+	    nr = MIN2( dmasz, count - j );
+
+	    if (nr >= 4)
+	    {
+	       GLint quads = (nr/2)-1;
+	       ELT_TYPE *dest = ALLOC_ELTS( quads*6 );
+	       GLint i;
+
+	       for ( i = j-start ; i < j-start+quads ; i++, elts += 2 ) {
+		  EMIT_TWO_ELTS( dest, 0, elts[0], elts[1] );
+		  EMIT_TWO_ELTS( dest, 2, elts[2], elts[1] );
+		  EMIT_TWO_ELTS( dest, 4, elts[3], elts[2] );
+		  dest += 6;
+	       }
+
+	       CLOSE_ELTS();
+	    }
+	 }
+      }
+      else {
+	 ELT_INIT( GL_TRIANGLE_STRIP, HW_TRIANGLE_STRIP_0 );
+
+	 for (j = start; j + 3 < count; j += nr - 2 ) {
+	    nr = MIN2( dmasz, count - j );
+	    dest = ALLOC_ELTS( nr );
+	    dest = TAG(emit_elts)( ctx, dest, elts+j, nr );
+	    CLOSE_ELTS();
+	 }
+      }
+   }
+}
+
+
+static void TAG(render_quads_elts)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   if (start+3 >= count)
+      return;
+
+   if (HAVE_QUADS && 0) {
+   } else {
+      LOCAL_VARS;
+      GLuint *elts = GET_MESA_ELTS();
+      int dmasz = GET_MAX_HW_ELTS();
+      GLuint j, nr;
+
+      ELT_INIT( GL_TRIANGLES, HW_TRIANGLES );
+
+      /* Emit whole number of quads in total, and in each buffer.
+       */
+      dmasz -= dmasz & 3;
+      count -= (count-start) & 3;
+
+      /* Adjust for rendering as triangles:
+       */
+      dmasz = dmasz/6*4;
+
+      for (j = start; j + 3 < count; j += nr ) {
+	 nr = MIN2( dmasz, count - j );
+
+	 {
+	    GLint quads = nr/4;
+	    ELT_TYPE *dest = ALLOC_ELTS( quads * 6 );
+	    GLint i;
+
+	    for ( i = j-start ; i < j-start+quads ; i++, elts += 4 ) {
+	       EMIT_TWO_ELTS( dest, 0, elts[0], elts[1] );
+	       EMIT_TWO_ELTS( dest, 2, elts[3], elts[1] );
+	       EMIT_TWO_ELTS( dest, 4, elts[2], elts[3] );
+	       dest += 6;
+	    }
+
+	    CLOSE_ELTS();
+	 }
+      }
+   }
+}
+
+
+
+static render_func TAG(render_tab_elts)[GL_POLYGON+2] =
+{
+   TAG(render_points_elts),
+   TAG(render_lines_elts),
+   TAG(render_line_loop_elts),
+   TAG(render_line_strip_elts),
+   TAG(render_triangles_elts),
+   TAG(render_tri_strip_elts),
+   TAG(render_tri_fan_elts),
+   TAG(render_quads_elts),
+   TAG(render_quad_strip_elts),
+   TAG(render_poly_elts),
+   TAG(render_noop),
+};
diff --git a/src/mesa/tnl_dd/t_dd_rendertmp.h b/src/mesa/tnl_dd/t_dd_rendertmp.h
new file mode 100644
index 0000000000..fc00e76934
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_rendertmp.h
@@ -0,0 +1,437 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#ifndef POSTFIX
+#define POSTFIX
+#endif
+
+#ifndef INIT
+#define INIT(x)
+#endif
+
+#ifndef NEED_EDGEFLAG_SETUP
+#define NEED_EDGEFLAG_SETUP 0
+#define EDGEFLAG_GET(a) 0
+#define EDGEFLAG_SET(a,b) (void)b
+#endif
+
+#ifndef RESET_STIPPLE
+#define RESET_STIPPLE
+#endif
+
+#ifndef RESET_OCCLUSION
+#define RESET_OCCLUSION
+#endif
+
+#ifndef TEST_PRIM_END
+#define TEST_PRIM_END(flags) (flags & PRIM_END)
+#define TEST_PRIM_BEGIN(flags) (flags & PRIM_BEGIN)
+#define TEST_PRIM_PARITY(flags) (flags & PRIM_PARITY)
+#endif
+
+#ifndef ELT
+#define ELT(x) x
+#endif
+
+#ifndef RENDER_TAB_QUALIFIER
+#define RENDER_TAB_QUALIFIER static
+#endif
+
+static void TAG(render_points)( GLcontext *ctx,
+				GLuint start,
+				GLuint count,
+				GLuint flags )
+{
+   LOCAL_VARS;
+   (void) flags;
+
+   RESET_OCCLUSION;
+   INIT(GL_POINTS);
+   RENDER_POINTS( start, count );
+   POSTFIX;
+}
+
+static void TAG(render_lines)( GLcontext *ctx,
+			       GLuint start,
+			       GLuint count,
+			       GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   RESET_OCCLUSION;
+   INIT(GL_LINES);
+   for (j=start+1; j<count; j+=2 ) {
+      RENDER_LINE( ELT(j-1), ELT(j) );
+      RESET_STIPPLE;
+   }
+   POSTFIX;
+}
+
+
+static void TAG(render_line_strip)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   RESET_OCCLUSION;
+   INIT(GL_LINE_STRIP);
+
+   for (j=start+1; j<count; j++ )
+      RENDER_LINE( ELT(j-1), ELT(j) );
+
+   if (TEST_PRIM_END(flags))
+      RESET_STIPPLE;
+
+   POSTFIX;
+}
+
+
+static void TAG(render_line_loop)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   GLuint i;
+   LOCAL_VARS;
+
+   (void) flags;
+
+   RESET_OCCLUSION;
+   INIT(GL_LINE_LOOP);
+
+   if (start+1 < count) {
+      if (TEST_PRIM_BEGIN(flags)) {
+	 RENDER_LINE( ELT(start), ELT(start+1) );
+      }
+
+      for ( i = start+2 ; i < count ; i++) {
+	 RENDER_LINE( ELT(i-1), ELT(i) );
+      }
+
+      if ( TEST_PRIM_END(flags)) {
+	 RENDER_LINE( ELT(count-1), ELT(start) );
+	 RESET_STIPPLE;
+      }
+   }
+
+   POSTFIX;
+}
+
+
+static void TAG(render_triangles)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_TRIANGLES);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+2; j<count; j+=3) {
+	 /* Leave the edgeflags as supplied by the user.
+	  */
+	 RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+	 RESET_STIPPLE;
+      }
+   } else {
+      for (j=start+2; j<count; j+=3) {
+	 RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+      }
+   }
+   POSTFIX;
+}
+
+
+
+static void TAG(render_tri_strip)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   GLuint j;
+   GLuint parity = 0;
+   LOCAL_VARS;
+
+   if (TEST_PRIM_PARITY(flags))
+      parity = 1;
+
+   INIT(GL_TRIANGLE_STRIP);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+2;j<count;j++,parity^=1) {
+	 GLuint ej2 = ELT(j-2+parity);
+	 GLuint ej1 = ELT(j-1-parity);
+	 GLuint ej = ELT(j);
+	 GLboolean ef2 = EDGEFLAG_GET( ej2 );
+	 GLboolean ef1 = EDGEFLAG_GET( ej1 );
+	 GLboolean ef = EDGEFLAG_GET( ej );
+	 EDGEFLAG_SET( ej2, GL_TRUE );
+	 EDGEFLAG_SET( ej1, GL_TRUE );
+	 EDGEFLAG_SET( ej, GL_TRUE );
+	 RENDER_TRI( ej2, ej1, ej );
+	 EDGEFLAG_SET( ej2, ef2 );
+	 EDGEFLAG_SET( ej1, ef1 );
+	 EDGEFLAG_SET( ej, ef );
+	 RESET_STIPPLE;
+      }
+   } else {
+      for (j=start+2; j<count ; j++, parity^=1) {
+	 RENDER_TRI( ELT(j-2+parity), ELT(j-1-parity), ELT(j) );
+      }
+   }
+   POSTFIX;
+}
+
+
+static void TAG(render_tri_fan)( GLcontext *ctx,
+				 GLuint start,
+				 GLuint count,
+				 GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_TRIANGLE_FAN);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+2;j<count;j++) {
+	 /* For trifans, all edges are boundary.
+	  */
+	 GLuint ejs = ELT(start);
+	 GLuint ej1 = ELT(j-1);
+	 GLuint ej = ELT(j);
+	 GLboolean efs = EDGEFLAG_GET( ejs );
+	 GLboolean ef1 = EDGEFLAG_GET( ej1 );
+	 GLboolean ef = EDGEFLAG_GET( ej );
+	 EDGEFLAG_SET( ejs, GL_TRUE );
+	 EDGEFLAG_SET( ej1, GL_TRUE );
+	 EDGEFLAG_SET( ej, GL_TRUE );
+	 RENDER_TRI( ejs, ej1, ej);
+	 EDGEFLAG_SET( ejs, efs );
+	 EDGEFLAG_SET( ej1, ef1 );
+	 EDGEFLAG_SET( ej, ef );
+	 RESET_STIPPLE;
+      }
+   } else {
+      for (j=start+2;j<count;j++) {
+	 RENDER_TRI( ELT(start), ELT(j-1), ELT(j) );
+      }
+   }
+
+   POSTFIX;
+}
+
+
+static void TAG(render_poly)( GLcontext *ctx,
+			      GLuint start,
+			      GLuint count,
+			      GLuint flags )
+{
+   GLuint j = start+2;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_POLYGON);
+   if (NEED_EDGEFLAG_SETUP) {
+      GLboolean efstart = EDGEFLAG_GET( ELT(start) );
+      GLboolean efcount = EDGEFLAG_GET( ELT(count-1) );
+
+      /* If the primitive does not begin here, the first edge
+       * is non-boundary.
+       */
+      if (!TEST_PRIM_BEGIN(flags))
+	 EDGEFLAG_SET( ELT(start), GL_FALSE );
+
+      /* If the primitive does not end here, the final edge is
+       * non-boundary.
+       */
+      if (!TEST_PRIM_END(flags))
+	 EDGEFLAG_SET( ELT(count-1), GL_FALSE );
+
+      /* Draw the first triangles (possibly zero)
+       */
+      if (j<count-1) {
+	 GLboolean ef = EDGEFLAG_GET( ELT(j) );
+	 EDGEFLAG_SET( ELT(j), GL_FALSE );
+	 RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+	 EDGEFLAG_SET( ELT(j), ef );
+	 j++;
+
+	 /* Don't render the first edge again:
+	  */
+	 EDGEFLAG_SET( ELT(start), GL_FALSE );
+
+	 for (;j<count-1;j++) {
+	    GLboolean efj = EDGEFLAG_GET( ELT(j) );
+	    EDGEFLAG_SET( ELT(j), GL_FALSE );
+	    RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+	    EDGEFLAG_SET( ELT(j), efj );
+	 }
+      }
+
+      /* Draw the last or only triangle
+       */
+      if (j < count)
+	 RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+
+      /* Restore the first and last edgeflags:
+       */
+      EDGEFLAG_SET( ELT(count-1), efcount );
+      EDGEFLAG_SET( ELT(start), efstart );
+
+      if (TEST_PRIM_END(flags)) {
+	 RESET_STIPPLE;
+      }
+   }
+   else {
+      for (j=start+2;j<count;j++) {
+	 RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+      }
+   }
+   POSTFIX;
+}
+
+static void TAG(render_quads)( GLcontext *ctx,
+			       GLuint start,
+			       GLuint count,
+			       GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_QUADS);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+3; j<count; j+=4) {
+	 /* Use user-specified edgeflags for quads.
+	  */
+	 RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+	 RESET_STIPPLE;
+      }
+   } else {
+      for (j=start+3; j<count; j+=4) {
+	 RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+      }
+   }
+   POSTFIX;
+}
+
+static void TAG(render_quad_strip)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_QUAD_STRIP);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+3;j<count;j+=2) {
+	 /* All edges are boundary.  Set edgeflags to 1, draw the
+	  * quad, and restore them to the original values.
+	  */
+	 GLboolean ef3 = EDGEFLAG_GET( ELT(j-3) );
+	 GLboolean ef2 = EDGEFLAG_GET( ELT(j-2) );
+	 GLboolean ef1 = EDGEFLAG_GET( ELT(j-1) );
+	 GLboolean ef = EDGEFLAG_GET( ELT(j) );
+	 EDGEFLAG_SET( ELT(j-3), GL_TRUE );
+	 EDGEFLAG_SET( ELT(j-2), GL_TRUE );
+	 EDGEFLAG_SET( ELT(j-1), GL_TRUE );
+	 EDGEFLAG_SET( ELT(j), GL_TRUE );
+	 RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+	 EDGEFLAG_SET( ELT(j-3), ef3 );
+	 EDGEFLAG_SET( ELT(j-2), ef2 );
+	 EDGEFLAG_SET( ELT(j-1), ef1 );
+	 EDGEFLAG_SET( ELT(j), ef );
+	 RESET_STIPPLE;
+      }
+   } else {
+      for (j=start+3;j<count;j+=2) {
+	 RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+      }
+   }
+   POSTFIX;
+}
+
+static void TAG(render_noop)( GLcontext *ctx,
+			      GLuint start,
+			      GLuint count,
+			      GLuint flags )
+{
+   (void)(ctx && start && count && flags);
+}
+
+RENDER_TAB_QUALIFIER void (*TAG(render_tab)[GL_POLYGON+2])(GLcontext *,
+							   GLuint,
+							   GLuint,
+							   GLuint) =
+{
+   TAG(render_points),
+   TAG(render_lines),
+   TAG(render_line_loop),
+   TAG(render_line_strip),
+   TAG(render_triangles),
+   TAG(render_tri_strip),
+   TAG(render_tri_fan),
+   TAG(render_quads),
+   TAG(render_quad_strip),
+   TAG(render_poly),
+   TAG(render_noop),
+};
+
+
+
+#ifndef PRESERVE_VB_DEFS
+#undef RENDER_TRI
+#undef RENDER_QUAD
+#undef RENDER_LINE
+#undef RENDER_POINTS
+#undef LOCAL_VARS
+#undef INIT
+#undef POSTFIX
+#undef RESET_STIPPLE
+#undef DBG
+#undef ELT
+#undef RENDER_TAB_QUALIFIER
+#endif
+
+#ifndef PRESERVE_TAG
+#undef TAG
+#endif
+
+#undef PRESERVE_VB_DEFS
+#undef PRESERVE_TAG
diff --git a/src/mesa/tnl_dd/t_dd_triemit.h b/src/mesa/tnl_dd/t_dd_triemit.h
new file mode 100644
index 0000000000..2f5bcee7ec
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_triemit.h
@@ -0,0 +1,157 @@
+#ifndef DO_DEBUG_VERTS
+#define DO_DEBUG_VERTS 0
+#endif 
+
+#ifndef PRINT_VERTEX
+#define PRINT_VERTEX(x) 
+#endif
+
+#if defined(USE_X86_ASM)
+#define COPY_DWORDS( j, vb, vertsize, v )				\
+do {									\
+	int __tmp;							\
+	__asm__ __volatile__( "rep ; movsl"				\
+			      : "=%c" (j), "=D" (vb), "=S" (__tmp)	\
+			      : "0" (vertsize),				\
+			        "D" ((long)vb),				\
+			        "S" ((long)v) );			\
+} while (0)
+#else
+#define COPY_DWORDS( j, vb, vertsize, v )				\
+do {									\
+   for ( j = 0 ; j < vertsize ; j++ )					\
+      vb[j] = ((GLuint *)v)[j];						\
+   vb += vertsize;							\
+} while (0)
+#endif
+
+
+
+#if HAVE_QUADS
+static __inline void TAG(quad)( CTX_ARG,
+				VERTEX *v0,
+				VERTEX *v1,
+				VERTEX *v2,
+				VERTEX *v3 )
+{
+   GLuint vertsize = GET_VERTEX_DWORDS();
+   GLuint *vb = (GLuint *)ALLOC_VERTS( 4, vertsize);
+   GLuint j;
+
+   if (DO_DEBUG_VERTS) {
+      fprintf(stderr, "%s\n", __FUNCTION__);
+      PRINT_VERTEX(v0);
+      PRINT_VERTEX(v1);
+      PRINT_VERTEX(v2);
+      PRINT_VERTEX(v3);
+   }
+      
+   COPY_DWORDS( j, vb, vertsize, v0 );
+   COPY_DWORDS( j, vb, vertsize, v1 );
+   COPY_DWORDS( j, vb, vertsize, v2 );
+   COPY_DWORDS( j, vb, vertsize, v3 );
+}
+#else
+static __inline void TAG(quad)( CTX_ARG,
+				VERTEX *v0,
+				VERTEX *v1,
+				VERTEX *v2,
+				VERTEX *v3 )
+{
+   GLuint vertsize = GET_VERTEX_DWORDS();
+   GLuint *vb = (GLuint *)ALLOC_VERTS(  6, vertsize);
+   GLuint j;
+
+   if (DO_DEBUG_VERTS) {
+      fprintf(stderr, "%s\n", __FUNCTION__);
+      PRINT_VERTEX(v0);
+      PRINT_VERTEX(v1);
+      PRINT_VERTEX(v2);
+      PRINT_VERTEX(v3);
+   }
+ 
+   COPY_DWORDS( j, vb, vertsize, v0 );
+   COPY_DWORDS( j, vb, vertsize, v1 );
+   COPY_DWORDS( j, vb, vertsize, v3 );
+   COPY_DWORDS( j, vb, vertsize, v1 );
+   COPY_DWORDS( j, vb, vertsize, v2 );
+   COPY_DWORDS( j, vb, vertsize, v3 );
+}
+#endif
+
+
+static __inline void TAG(triangle)( CTX_ARG,
+				    VERTEX *v0,
+				    VERTEX *v1,
+				    VERTEX *v2 )
+{
+   GLuint vertsize = GET_VERTEX_DWORDS();
+   GLuint *vb = (GLuint *)ALLOC_VERTS( 3, vertsize);
+   GLuint j;
+
+   if (DO_DEBUG_VERTS) {
+      fprintf(stderr, "%s\n", __FUNCTION__);
+      PRINT_VERTEX(v0);
+      PRINT_VERTEX(v1);
+      PRINT_VERTEX(v2);
+   }
+
+   COPY_DWORDS( j, vb, vertsize, v0 );
+   COPY_DWORDS( j, vb, vertsize, v1 );
+   COPY_DWORDS( j, vb, vertsize, v2 );
+}
+
+
+#if HAVE_LINES
+static __inline void TAG(line)( CTX_ARG,
+				VERTEX *v0,
+				VERTEX *v1 )
+{
+   GLuint vertsize = GET_VERTEX_DWORDS();
+   GLuint *vb = (GLuint *)ALLOC_VERTS( 2, vertsize);
+   GLuint j;
+
+   COPY_DWORDS( j, vb, vertsize, v0 );
+   COPY_DWORDS( j, vb, vertsize, v1 );
+}
+#endif
+
+#if HAVE_POINTS
+static __inline void TAG(point)( CTX_ARG,
+				 VERTEX *v0 )
+{
+   GLuint vertsize = GET_VERTEX_DWORDS();
+   GLuint *vb = (GLuint *)ALLOC_VERTS( 1, vertsize);
+   int j;
+
+   COPY_DWORDS( j, vb, vertsize, v0 );
+}
+#endif
+
+
+static void TAG(fast_clipped_poly)( GLcontext *ctx, const GLuint *elts,
+				    GLuint n )
+{
+   LOCAL_VARS
+   GLuint vertsize = GET_VERTEX_DWORDS();
+   GLuint *vb = (GLuint *)ALLOC_VERTS( (n-2) * 3, vertsize );
+   const GLuint *start = (const GLuint *)VERT(elts[0]);
+   int i,j;
+
+   if (DO_DEBUG_VERTS) {
+      fprintf(stderr, "%s\n", __FUNCTION__);
+      PRINT_VERTEX(VERT(elts[0]));
+      PRINT_VERTEX(VERT(elts[1]));
+   }
+
+   for (i = 2 ; i < n ; i++) {
+      if (DO_DEBUG_VERTS) {
+	 PRINT_VERTEX(VERT(elts[i]));
+      }
+
+      COPY_DWORDS( j, vb, vertsize, VERT(elts[i-1]) );
+      COPY_DWORDS( j, vb, vertsize, VERT(elts[i]) );
+      COPY_DWORDS( j, vb, vertsize, start );
+   }
+}
+
diff --git a/src/mesa/tnl_dd/t_dd_tritmp.h b/src/mesa/tnl_dd/t_dd_tritmp.h
new file mode 100644
index 0000000000..cc47798d5e
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_tritmp.h
@@ -0,0 +1,739 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+/* Template for building functions to plug into the driver interface
+ * of t_vb_render.c:
+ *     ctx->Driver.QuadFunc
+ *     ctx->Driver.TriangleFunc
+ *     ctx->Driver.LineFunc
+ *     ctx->Driver.PointsFunc
+ *
+ * DO_TWOSIDE:   Plug back-color values from the VB into backfacing triangles,
+ *               and restore vertices afterwards.
+ * DO_OFFSET:    Calculate offset for triangles and adjust vertices.  Restore
+ *               vertices after rendering.
+ * DO_FLAT:      For hardware without native flatshading, copy provoking colors
+ *               into the other vertices.  Restore after rendering.
+ * DO_UNFILLED:  Decompose triangles to lines and points where appropriate.
+ *
+ * HAVE_RGBA: Vertices have rgba values (otherwise index values).
+ * HAVE_SPEC: Vertices have secondary rgba values.
+ *
+ * VERT_X(v): Alias for vertex x value.
+ * VERT_Y(v): Alias for vertex y value.
+ * VERT_Z(v): Alias for vertex z value.
+ * DEPTH_SCALE: Scale for offset.
+ *
+ * VERTEX: Hardware vertex type.
+ * GET_VERTEX(n): Retreive vertex with index n.
+ * AREA_IS_CCW(a): Return true if triangle with signed area a is ccw.
+ *
+ * VERT_SET_RGBA: Assign vertex rgba from VB color.
+ * VERT_COPY_RGBA: Copy vertex rgba another vertex.
+ * VERT_SAVE_RGBA: Save vertex rgba to a local variable.
+ * VERT_RESTORE_RGBA: Restore vertex rgba from a local variable.
+ *   --> Similar for IND and SPEC.
+ *
+ * LOCAL_VARS(n): (At least) define local vars for save/restore rgba.
+ *
+ */
+
+#if HAVE_RGBA
+#define VERT_SET_IND( v, c ) (void) c
+#define VERT_COPY_IND( v0, v1 )
+#define VERT_SAVE_IND( idx )
+#define VERT_RESTORE_IND( idx )
+#if HAVE_BACK_COLORS
+#define VERT_SET_RGBA( v, c )
+#endif
+#else
+#define VERT_SET_RGBA( v, c ) (void) c
+#define VERT_COPY_RGBA( v0, v1 )
+#define VERT_SAVE_RGBA( idx )
+#define VERT_RESTORE_RGBA( idx )
+#if HAVE_BACK_COLORS
+#define VERT_SET_IND( v, c )
+#endif
+#endif
+
+#if !HAVE_SPEC
+#define VERT_SET_SPEC( v, c ) (void) c
+#define VERT_COPY_SPEC( v0, v1 )
+#define VERT_SAVE_SPEC( idx )
+#define VERT_RESTORE_SPEC( idx )
+#if HAVE_BACK_COLORS
+#define VERT_COPY_SPEC1( v )
+#endif
+#else
+#if HAVE_BACK_COLORS
+#define VERT_SET_SPEC( v, c )
+#endif
+#endif
+
+#if !HAVE_BACK_COLORS
+#define VERT_COPY_SPEC1( v )
+#define VERT_COPY_IND1( v )
+#define VERT_COPY_RGBA1( v )
+#endif
+
+#ifndef INSANE_VERTICES
+#define VERT_SET_Z(v,val) VERT_Z(v) = val
+#define VERT_Z_ADD(v,val) VERT_Z(v) += val
+#endif
+
+#if DO_TRI
+static void TAG(triangle)( GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
+   VERTEX *v[3];
+   GLfloat offset;
+   GLfloat z[3];
+   GLenum mode = GL_FILL;
+   GLuint facing;
+   LOCAL_VARS(3);
+
+/*     fprintf(stderr, "%s\n", __FUNCTION__); */
+
+   v[0] = (VERTEX *)GET_VERTEX(e0);
+   v[1] = (VERTEX *)GET_VERTEX(e1);
+   v[2] = (VERTEX *)GET_VERTEX(e2);
+
+   if (DO_TWOSIDE || DO_OFFSET || DO_UNFILLED)
+   {
+      GLfloat ex = VERT_X(v[0]) - VERT_X(v[2]);
+      GLfloat ey = VERT_Y(v[0]) - VERT_Y(v[2]);
+      GLfloat fx = VERT_X(v[1]) - VERT_X(v[2]);
+      GLfloat fy = VERT_Y(v[1]) - VERT_Y(v[2]);
+      GLfloat cc = ex*fy - ey*fx;
+
+      if (DO_TWOSIDE || DO_UNFILLED)
+      {
+	 facing = AREA_IS_CCW( cc ) ^ ctx->Polygon._FrontBit;
+
+	 if (DO_UNFILLED) {
+	    if (facing) {
+	       mode = ctx->Polygon.BackMode;
+	       if (ctx->Polygon.CullFlag &&
+		   ctx->Polygon.CullFaceMode != GL_FRONT) {
+		  return;
+	       }
+	    } else {
+	       mode = ctx->Polygon.FrontMode;
+	       if (ctx->Polygon.CullFlag &&
+		   ctx->Polygon.CullFaceMode != GL_BACK) {
+		  return;
+	       }
+	    }
+	 }
+
+	 if (DO_TWOSIDE && facing == 1)
+	 {
+	    if (HAVE_RGBA) {
+	       if (HAVE_BACK_COLORS) {
+		  if (!DO_FLAT) {
+		     VERT_SAVE_RGBA( 0 );
+		     VERT_SAVE_RGBA( 1 );
+		     VERT_COPY_RGBA1( v[0] );
+		     VERT_COPY_RGBA1( v[1] );
+		  }
+		  VERT_SAVE_RGBA( 2 );
+		  VERT_COPY_RGBA1( v[2] );
+		  if (HAVE_SPEC) {
+		     if (!DO_FLAT) {
+			VERT_SAVE_SPEC( 0 );
+			VERT_SAVE_SPEC( 1 );
+			VERT_COPY_SPEC1( v[0] );
+			VERT_COPY_SPEC1( v[1] );
+		     }
+		     VERT_SAVE_SPEC( 2 );
+		     VERT_COPY_SPEC1( v[2] );
+		  }
+	       }
+	       else {
+		  GLchan (*vbcolor)[4] = VB->ColorPtr[1]->Ptr;
+		  ASSERT(VB->ColorPtr[1]->Type == CHAN_TYPE);
+		  ASSERT(VB->ColorPtr[1]->StrideB == 4*sizeof(GLchan));
+		  (void) vbcolor;
+
+		  if (!DO_FLAT) {
+		     VERT_SAVE_RGBA( 0 );
+		     VERT_SAVE_RGBA( 1 );
+		     VERT_SET_RGBA( v[0], vbcolor[e0] );
+		     VERT_SET_RGBA( v[1], vbcolor[e1] );
+		  }
+		  VERT_SAVE_RGBA( 2 );
+		  VERT_SET_RGBA( v[2], vbcolor[e2] );
+
+		  if (HAVE_SPEC && VB->SecondaryColorPtr[1]) {
+		     GLchan (*vbspec)[4] = VB->SecondaryColorPtr[1]->Ptr;
+
+		     if (!DO_FLAT) {
+			VERT_SAVE_SPEC( 0 );
+			VERT_SAVE_SPEC( 1 );
+			VERT_SET_SPEC( v[0], vbspec[e0] );
+			VERT_SET_SPEC( v[1], vbspec[e1] );
+		     }
+		     VERT_SAVE_SPEC( 2 );
+		     VERT_SET_SPEC( v[2], vbspec[e2] );
+		  }
+	       }
+	    }
+	    else {
+	       GLuint *vbindex = VB->IndexPtr[1]->data;
+	       if (!DO_FLAT) {
+		  VERT_SET_IND( v[0], vbindex[e0] );
+		  VERT_SET_IND( v[1], vbindex[e1] );
+	       }
+	       VERT_SET_IND( v[2], vbindex[e2] );
+	    }
+	 }
+      }
+
+
+      if (DO_OFFSET)
+      {
+	 offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE;
+	 z[0] = VERT_Z(v[0]);
+	 z[1] = VERT_Z(v[1]);
+	 z[2] = VERT_Z(v[2]);
+	 if (cc * cc > 1e-16) {
+	    GLfloat ic	= 1.0 / cc;
+	    GLfloat ez	= z[0] - z[2];
+	    GLfloat fz	= z[1] - z[2];
+	    GLfloat a	= ey*fz - ez*fy;
+	    GLfloat b	= ez*fx - ex*fz;
+	    GLfloat ac	= a * ic;
+	    GLfloat bc	= b * ic;
+	    if ( ac < 0.0f ) ac = -ac;
+	    if ( bc < 0.0f ) bc = -bc;
+	    offset += MAX2( ac, bc ) * ctx->Polygon.OffsetFactor;
+	 }
+	 offset *= ctx->MRD;
+      }
+   }
+
+   if (DO_FLAT) {
+      if (HAVE_RGBA) {
+	 VERT_SAVE_RGBA( 0 );
+	 VERT_SAVE_RGBA( 1 );
+	 VERT_COPY_RGBA( v[0], v[2] );
+	 VERT_COPY_RGBA( v[1], v[2] );
+	 if (HAVE_SPEC && VB->SecondaryColorPtr[0]) {
+	    VERT_SAVE_SPEC( 0 );
+	    VERT_SAVE_SPEC( 1 );
+	    VERT_COPY_SPEC( v[0], v[2] );
+	    VERT_COPY_SPEC( v[1], v[2] );
+	 }
+      }
+      else {
+	 VERT_SAVE_IND( 0 );
+	 VERT_SAVE_IND( 1 );
+	 VERT_COPY_IND( v[0], v[2] );
+	 VERT_COPY_IND( v[1], v[2] );
+      }
+   }
+
+   if (mode == GL_POINT) {
+      if (DO_OFFSET && ctx->Polygon.OffsetPoint) {
+	 VERT_Z_ADD(v[0], offset);
+	 VERT_Z_ADD(v[1], offset);
+	 VERT_Z_ADD(v[2], offset);
+      }
+      UNFILLED_TRI( ctx, GL_POINT, e0, e1, e2 );
+   } else if (mode == GL_LINE) {
+      if (DO_OFFSET && ctx->Polygon.OffsetLine) {
+	 VERT_Z_ADD(v[0], offset);
+	 VERT_Z_ADD(v[1], offset);
+	 VERT_Z_ADD(v[2], offset);
+      }
+      UNFILLED_TRI( ctx, GL_LINE, e0, e1, e2 );
+   } else {
+      if (DO_OFFSET && ctx->Polygon.OffsetFill) {
+	 VERT_Z_ADD(v[0], offset);
+	 VERT_Z_ADD(v[1], offset);
+	 VERT_Z_ADD(v[2], offset);
+      }
+      if (DO_UNFILLED)
+	 RASTERIZE( GL_TRIANGLES );
+      TRI( v[0], v[1], v[2] );
+   }
+
+   if (DO_OFFSET)
+   {
+      VERT_SET_Z(v[0], z[0]);
+      VERT_SET_Z(v[1], z[1]);
+      VERT_SET_Z(v[2], z[2]);
+   }
+
+   if (DO_TWOSIDE && facing == 1)
+   {
+      if (HAVE_RGBA) {
+	 if (!DO_FLAT) {
+	    VERT_RESTORE_RGBA( 0 );
+	    VERT_RESTORE_RGBA( 1 );
+	 }
+	 VERT_RESTORE_RGBA( 2 );
+	 if (HAVE_SPEC) {
+	    if (!DO_FLAT) {
+	       VERT_RESTORE_SPEC( 0 );
+	       VERT_RESTORE_SPEC( 1 );
+	    }
+	    VERT_RESTORE_SPEC( 2 );
+	 }
+      }
+      else {
+	 GLuint *vbindex = VB->IndexPtr[0]->data;
+	 if (!DO_FLAT) {
+	    VERT_SET_IND( v[0], vbindex[e0] );
+	    VERT_SET_IND( v[1], vbindex[e1] );
+	 }
+	 VERT_SET_IND( v[2], vbindex[e2] );
+      }
+   }
+
+
+   if (DO_FLAT) {
+      if (HAVE_RGBA) {
+	 VERT_RESTORE_RGBA( 0 );
+	 VERT_RESTORE_RGBA( 1 );
+	 if (HAVE_SPEC && VB->SecondaryColorPtr[0]) {
+	    VERT_RESTORE_SPEC( 0 );
+	    VERT_RESTORE_SPEC( 1 );
+	 }
+      }
+      else {
+	 VERT_RESTORE_IND( 0 );
+	 VERT_RESTORE_IND( 1 );
+      }
+   }
+}
+#endif
+
+#if DO_QUAD
+#if DO_FULL_QUAD
+static void TAG(quad)( GLcontext *ctx,
+		       GLuint e0, GLuint e1, GLuint e2, GLuint e3 )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
+   VERTEX *v[4];
+   GLfloat offset;
+   GLfloat z[4];
+   GLenum mode = GL_FILL;
+   GLuint facing;
+   LOCAL_VARS(4);
+
+   v[0] = (VERTEX *)GET_VERTEX(e0);
+   v[1] = (VERTEX *)GET_VERTEX(e1);
+   v[2] = (VERTEX *)GET_VERTEX(e2);
+   v[3] = (VERTEX *)GET_VERTEX(e3);
+
+   if (DO_TWOSIDE || DO_OFFSET || DO_UNFILLED)
+   {
+      GLfloat ex = VERT_X(v[2]) - VERT_X(v[0]);
+      GLfloat ey = VERT_Y(v[2]) - VERT_Y(v[0]);
+      GLfloat fx = VERT_X(v[3]) - VERT_X(v[1]);
+      GLfloat fy = VERT_Y(v[3]) - VERT_Y(v[1]);
+      GLfloat cc = ex*fy - ey*fx;
+
+      if (DO_TWOSIDE || DO_UNFILLED)
+      {
+	 facing = AREA_IS_CCW( cc ) ^ ctx->Polygon._FrontBit;
+
+	 if (DO_UNFILLED) {
+	    if (facing) {
+	       mode = ctx->Polygon.BackMode;
+	       if (ctx->Polygon.CullFlag &&
+		   ctx->Polygon.CullFaceMode != GL_FRONT) {
+		  return;
+	       }
+	    } else {
+	       mode = ctx->Polygon.FrontMode;
+	       if (ctx->Polygon.CullFlag &&
+		   ctx->Polygon.CullFaceMode != GL_BACK) {
+		  return;
+	       }
+	    }
+	 }
+
+	 if (DO_TWOSIDE && facing == 1)
+	 {
+	    if (HAVE_RGBA) {
+	       GLchan (*vbcolor)[4] = VB->ColorPtr[1]->Ptr;
+	       (void)vbcolor;
+
+	       if (HAVE_BACK_COLORS) {
+                  if (!DO_FLAT) {
+                     VERT_SAVE_RGBA( 0 );
+                     VERT_SAVE_RGBA( 1 );
+                     VERT_SAVE_RGBA( 2 );
+		     VERT_COPY_RGBA1( v[0] );
+		     VERT_COPY_RGBA1( v[1] );
+		     VERT_COPY_RGBA1( v[2] );
+		  }
+		  VERT_SAVE_RGBA( 3 );
+		  VERT_COPY_RGBA1( v[3] );
+		  if (HAVE_SPEC) {
+                     if (!DO_FLAT) {
+                        VERT_SAVE_SPEC( 0 );
+                        VERT_SAVE_SPEC( 1 );
+                        VERT_SAVE_SPEC( 2 );
+			VERT_COPY_SPEC1( v[0] );
+			VERT_COPY_SPEC1( v[1] );
+			VERT_COPY_SPEC1( v[2] );
+		     }
+		     VERT_SAVE_SPEC( 3 );
+		     VERT_COPY_SPEC1( v[3] );
+		  }
+	       }
+	       else {
+	          if (!DO_FLAT) {
+		     VERT_SAVE_RGBA( 0 );
+		     VERT_SAVE_RGBA( 1 );
+		     VERT_SAVE_RGBA( 2 );
+		     VERT_SET_RGBA( v[0], vbcolor[e0] );
+		     VERT_SET_RGBA( v[1], vbcolor[e1] );
+		     VERT_SET_RGBA( v[2], vbcolor[e2] );
+	          }
+	          VERT_SAVE_RGBA( 3 );
+	          VERT_SET_RGBA( v[3], vbcolor[e3] );
+
+	          if (HAVE_SPEC && VB->SecondaryColorPtr[1]) {
+		     GLchan (*vbspec)[4] = VB->SecondaryColorPtr[1]->Ptr;
+		     ASSERT(VB->SecondaryColorPtr[1]->StrideB==4*sizeof(GLchan));
+
+		     if (!DO_FLAT) {
+		        VERT_SAVE_SPEC( 0 );
+		        VERT_SAVE_SPEC( 1 );
+		        VERT_SAVE_SPEC( 2 );
+		        VERT_SET_SPEC( v[0], vbspec[e0] );
+		        VERT_SET_SPEC( v[1], vbspec[e1] );
+		        VERT_SET_SPEC( v[2], vbspec[e2] );
+		     }
+		     VERT_SAVE_SPEC( 3 );
+		     VERT_SET_SPEC( v[3], vbspec[e3] );
+	          }
+	       }
+	    }
+	    else {
+	       GLuint *vbindex = VB->IndexPtr[1]->data;
+	       if (!DO_FLAT) {
+		  VERT_SET_IND( v[0], vbindex[e0] );
+		  VERT_SET_IND( v[1], vbindex[e1] );
+		  VERT_SET_IND( v[2], vbindex[e2] );
+	       }
+	       VERT_SET_IND( v[3], vbindex[e3] );
+	    }
+	 }
+      }
+
+
+      if (DO_OFFSET)
+      {
+	 offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE;
+	 z[0] = VERT_Z(v[0]);
+	 z[1] = VERT_Z(v[1]);
+	 z[2] = VERT_Z(v[2]);
+	 z[3] = VERT_Z(v[3]);
+	 if (cc * cc > 1e-16) {
+	    GLfloat ez = z[2] - z[0];
+	    GLfloat fz = z[3] - z[1];
+	    GLfloat a	= ey*fz - ez*fy;
+	    GLfloat b	= ez*fx - ex*fz;
+	    GLfloat ic	= 1.0 / cc;
+	    GLfloat ac	= a * ic;
+	    GLfloat bc	= b * ic;
+	    if ( ac < 0.0f ) ac = -ac;
+	    if ( bc < 0.0f ) bc = -bc;
+	    offset += MAX2( ac, bc ) * ctx->Polygon.OffsetFactor;
+	 }
+	 offset *= ctx->MRD;
+      }
+   }
+
+   if (DO_FLAT) {
+      if (HAVE_RGBA) {
+	 VERT_SAVE_RGBA( 0 );
+	 VERT_SAVE_RGBA( 1 );
+	 VERT_SAVE_RGBA( 2 );
+	 VERT_COPY_RGBA( v[0], v[3] );
+	 VERT_COPY_RGBA( v[1], v[3] );
+	 VERT_COPY_RGBA( v[2], v[3] );
+	 if (HAVE_SPEC && VB->SecondaryColorPtr[0]) {
+	    VERT_SAVE_SPEC( 0 );
+	    VERT_SAVE_SPEC( 1 );
+	    VERT_SAVE_SPEC( 2 );
+	    VERT_COPY_SPEC( v[0], v[3] );
+	    VERT_COPY_SPEC( v[1], v[3] );
+	    VERT_COPY_SPEC( v[2], v[3] );
+	 }
+      }
+      else {
+	 VERT_SAVE_IND( 0 );
+	 VERT_SAVE_IND( 1 );
+	 VERT_SAVE_IND( 2 );
+	 VERT_COPY_IND( v[0], v[3] );
+	 VERT_COPY_IND( v[1], v[3] );
+	 VERT_COPY_IND( v[2], v[3] );
+      }
+   }
+
+   if (mode == GL_POINT) {
+      if (( DO_OFFSET) && ctx->Polygon.OffsetPoint) {
+	 VERT_Z_ADD(v[0], offset);
+	 VERT_Z_ADD(v[1], offset);
+	 VERT_Z_ADD(v[2], offset);
+	 VERT_Z_ADD(v[3], offset);
+      }
+      UNFILLED_QUAD( ctx, GL_POINT, e0, e1, e2, e3 );
+   } else if (mode == GL_LINE) {
+      if (DO_OFFSET && ctx->Polygon.OffsetLine) {
+	 VERT_Z_ADD(v[0], offset);
+	 VERT_Z_ADD(v[1], offset);
+	 VERT_Z_ADD(v[2], offset);
+	 VERT_Z_ADD(v[3], offset);
+      }
+      UNFILLED_QUAD( ctx, GL_LINE, e0, e1, e2, e3 );
+   } else {
+      if (DO_OFFSET && ctx->Polygon.OffsetFill) {
+	 VERT_Z_ADD(v[0], offset);
+	 VERT_Z_ADD(v[1], offset);
+	 VERT_Z_ADD(v[2], offset);
+	 VERT_Z_ADD(v[3], offset);
+      }
+      RASTERIZE( GL_TRIANGLES );
+      QUAD( (v[0]), (v[1]), (v[2]), (v[3]) );
+   }
+
+   if (DO_OFFSET)
+   {
+      VERT_SET_Z(v[0], z[0]);
+      VERT_SET_Z(v[1], z[1]);
+      VERT_SET_Z(v[2], z[2]);
+      VERT_SET_Z(v[3], z[3]);
+   }
+
+   if (DO_TWOSIDE && facing == 1)
+   {
+      if (HAVE_RGBA) {
+	 if (!DO_FLAT) {
+	    VERT_RESTORE_RGBA( 0 );
+	    VERT_RESTORE_RGBA( 1 );
+	    VERT_RESTORE_RGBA( 2 );
+	 }
+	 VERT_RESTORE_RGBA( 3 );
+	 if (HAVE_SPEC) {
+	    if (!DO_FLAT) {
+	       VERT_RESTORE_SPEC( 0 );
+	       VERT_RESTORE_SPEC( 1 );
+	       VERT_RESTORE_SPEC( 2 );
+	    }
+	    VERT_RESTORE_SPEC( 3 );
+	 }
+      }
+      else {
+	 GLuint *vbindex = VB->IndexPtr[0]->data;
+	 if (!DO_FLAT) {
+	    VERT_SET_IND( v[0], vbindex[e0] );
+	    VERT_SET_IND( v[1], vbindex[e1] );
+	    VERT_SET_IND( v[2], vbindex[e2] );
+	 }
+	 VERT_SET_IND( v[3], vbindex[e3] );
+      }
+   }
+
+
+   if (DO_FLAT) {
+      if (HAVE_RGBA) {
+	 VERT_RESTORE_RGBA( 0 );
+	 VERT_RESTORE_RGBA( 1 );
+	 VERT_RESTORE_RGBA( 2 );
+	 if (HAVE_SPEC && VB->SecondaryColorPtr[0]) {
+	    VERT_RESTORE_SPEC( 0 );
+	    VERT_RESTORE_SPEC( 1 );
+	    VERT_RESTORE_SPEC( 2 );
+	 }
+      }
+      else {
+	 VERT_RESTORE_IND( 0 );
+	 VERT_RESTORE_IND( 1 );
+	 VERT_RESTORE_IND( 2 );
+      }
+   }
+}
+#else
+static void TAG(quad)( GLcontext *ctx, GLuint e0,
+		       GLuint e1, GLuint e2, GLuint e3 )
+{
+   if (DO_UNFILLED) {
+      struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+      GLubyte ef1 = VB->EdgeFlag[e1];
+      GLubyte ef3 = VB->EdgeFlag[e3];
+      VB->EdgeFlag[e1] = 0;
+      TAG(triangle)( ctx, e0, e1, e3 );
+      VB->EdgeFlag[e1] = ef1;
+      VB->EdgeFlag[e3] = 0;
+      TAG(triangle)( ctx, e1, e2, e3 );
+      VB->EdgeFlag[e3] = ef3;
+   } else {
+      TAG(triangle)( ctx, e0, e1, e3 );
+      TAG(triangle)( ctx, e1, e2, e3 );
+   }
+}
+#endif
+#endif
+
+#if DO_LINE
+static void TAG(line)( GLcontext *ctx, GLuint e0, GLuint e1 )
+{
+   TNLvertexbuffer *VB = &TNL_CONTEXT(ctx)->vb;
+   VERTEX *v[2];
+   LOCAL_VARS(2);
+
+   v[0] = (VERTEX *)GET_VERTEX(e0);
+   v[1] = (VERTEX *)GET_VERTEX(e1);
+
+   if (DO_FLAT) {
+      if (HAVE_RGBA) {
+	 VERT_SAVE_RGBA( 0 );
+	 VERT_COPY_RGBA( v[0], v[1] );
+	 if (HAVE_SPEC && VB->SecondaryColorPtr[0]) {
+	    VERT_SAVE_SPEC( 0 );
+	    VERT_COPY_SPEC( v[0], v[1] );
+	 }
+      }
+      else {
+	 VERT_SAVE_IND( 0 );
+	 VERT_COPY_IND( v[0], v[1] );
+      }
+   }
+
+   LINE( v[0], v[1] );
+
+   if (DO_FLAT) {
+      if (HAVE_RGBA) {
+	 VERT_RESTORE_RGBA( 0 );
+
+	 if (HAVE_SPEC && VB->SecondaryColorPtr[0]) {
+	    VERT_RESTORE_SPEC( 0 );
+	 }
+      }
+      else {
+	 VERT_RESTORE_IND( 0 );
+      }
+   }
+}
+#endif
+
+#if DO_POINTS
+static void TAG(points)( GLcontext *ctx, GLuint first, GLuint last )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
+   int i;
+   LOCAL_VARS(1);
+
+   if (VB->Elts == 0) {
+      for ( i = first ; i < last ; i++ ) {
+	 if ( VB->ClipMask[i] == 0 ) {
+	    VERTEX *v = (VERTEX *)GET_VERTEX(i);
+	    POINT( v );
+	 }
+      }
+   } else {
+      for ( i = first ; i < last ; i++ ) {
+	 GLuint e = VB->Elts[i];
+	 if ( VB->ClipMask[e] == 0 ) {
+	    VERTEX *v = (VERTEX *)GET_VERTEX(e);
+	    POINT( v );
+	 }
+      }
+   }
+}
+#endif
+
+static void TAG(init)( void )
+{
+#if DO_QUAD
+   TAB[IND].quad = TAG(quad);
+#endif
+#if DO_TRI
+   TAB[IND].triangle = TAG(triangle);
+#endif
+#if DO_LINE
+   TAB[IND].line = TAG(line);
+#endif
+#if DO_POINTS
+   TAB[IND].points = TAG(points);
+#endif
+}
+
+#undef IND
+#undef TAG
+
+#if HAVE_RGBA
+#undef VERT_SET_IND
+#undef VERT_COPY_IND
+#undef VERT_SAVE_IND
+#undef VERT_RESTORE_IND
+#if HAVE_BACK_COLORS
+#undef VERT_SET_RGBA
+#endif
+#else
+#undef VERT_SET_RGBA
+#undef VERT_COPY_RGBA
+#undef VERT_SAVE_RGBA
+#undef VERT_RESTORE_RGBA
+#if HAVE_BACK_COLORS
+#undef VERT_SET_IND
+#endif
+#endif
+
+#if !HAVE_SPEC
+#undef VERT_SET_SPEC
+#undef VERT_COPY_SPEC
+#undef VERT_SAVE_SPEC
+#undef VERT_RESTORE_SPEC
+#if HAVE_BACK_COLORS
+#undef VERT_COPY_SPEC1
+#endif
+#else
+#if HAVE_BACK_COLORS
+#undef VERT_SET_SPEC
+#endif
+#endif
+
+#if !HAVE_BACK_COLORS
+#undef VERT_COPY_SPEC1
+#undef VERT_COPY_IND1
+#undef VERT_COPY_RGBA1
+#endif
+
+#ifndef INSANE_VERTICES
+#undef VERT_SET_Z
+#undef VERT_Z_ADD
+#endif
diff --git a/src/mesa/tnl_dd/t_dd_unfilled.h b/src/mesa/tnl_dd/t_dd_unfilled.h
new file mode 100644
index 0000000000..46415ea5f6
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_unfilled.h
@@ -0,0 +1,212 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#if HAVE_RGBA
+#define VERT_SET_IND( v, c )
+#define VERT_COPY_IND( v0, v1 )
+#define VERT_SAVE_IND( idx )
+#define VERT_RESTORE_IND( idx )
+#endif
+
+#if !HAVE_SPEC
+#define VERT_SET_SPEC( v, c )
+#define VERT_COPY_SPEC( v0, v1 )
+#define VERT_SAVE_SPEC( idx )
+#define VERT_RESTORE_SPEC( idx )
+#endif
+
+static void TAG(unfilled_tri)( GLcontext *ctx,
+			       GLenum mode,
+			       GLuint e0, GLuint e1, GLuint e2 )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLubyte *ef = VB->EdgeFlag;
+   VERTEX *v[3];
+   LOCAL_VARS(3);
+
+   v[0] = (VERTEX *)GET_VERTEX(e0);
+   v[1] = (VERTEX *)GET_VERTEX(e1);
+   v[2] = (VERTEX *)GET_VERTEX(e2);
+
+   if ((ctx->_TriangleCaps & DD_FLATSHADE) && HAVE_HW_FLATSHADE) {
+      if (HAVE_RGBA) {
+	 VERT_SAVE_RGBA(0);
+	 VERT_SAVE_RGBA(1);
+	 VERT_COPY_RGBA(v[0], v[2]);
+	 VERT_COPY_RGBA(v[1], v[2]);
+
+	 if (HAVE_SPEC) {
+	    VERT_SAVE_SPEC(0);
+	    VERT_SAVE_SPEC(1);
+	    VERT_COPY_SPEC(v[0], v[2]);
+	    VERT_COPY_SPEC(v[1], v[2]);
+	 }
+      } else {
+	 VERT_SAVE_IND(0);
+	 VERT_SAVE_IND(1);
+	 VERT_COPY_IND(v[0], v[2]);
+	 VERT_COPY_IND(v[1], v[2]);
+      }
+   }
+
+/*     fprintf(stderr, "%s %s %d %d %d\n", __FUNCTION__, */
+/*  	   _mesa_lookup_enum_by_nr( mode ), */
+/*  	   ef[e0], ef[e1], ef[e2]); */
+
+   if (mode == GL_POINT) {
+      RASTERIZE(GL_POINTS);
+      if (ef[e0]) POINT( v[0] );
+      if (ef[e1]) POINT( v[1] );
+      if (ef[e2]) POINT( v[2] );
+   }
+   else {
+      RASTERIZE(GL_LINES);
+      if (RENDER_PRIMITIVE == GL_POLYGON) {
+	 if (ef[e2]) LINE( v[2], v[0] );
+	 if (ef[e0]) LINE( v[0], v[1] );
+	 if (ef[e1]) LINE( v[1], v[2] );
+      }
+      else {
+	 if (ef[e0]) LINE( v[0], v[1] );
+	 if (ef[e1]) LINE( v[1], v[2] );
+	 if (ef[e2]) LINE( v[2], v[0] );
+      }
+   }
+
+   if ((ctx->_TriangleCaps & DD_FLATSHADE) && HAVE_HW_FLATSHADE) {
+      if (HAVE_RGBA) {
+	 VERT_RESTORE_RGBA(0);
+	 VERT_RESTORE_RGBA(1);
+
+	 if (HAVE_SPEC) {
+	    VERT_RESTORE_SPEC(0);
+	    VERT_RESTORE_SPEC(1);
+	 }
+      } else {
+	 VERT_RESTORE_IND(0);
+	 VERT_RESTORE_IND(1);
+      }
+   }
+}
+
+
+static void TAG(unfilled_quad)( GLcontext *ctx,
+				GLenum mode,
+				GLuint e0, GLuint e1,
+				GLuint e2, GLuint e3 )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLubyte *ef = VB->EdgeFlag;
+   VERTEX *v[4];
+   LOCAL_VARS(4);
+
+   v[0] = (VERTEX *)GET_VERTEX(e0);
+   v[1] = (VERTEX *)GET_VERTEX(e1);
+   v[2] = (VERTEX *)GET_VERTEX(e2);
+   v[3] = (VERTEX *)GET_VERTEX(e3);
+
+   /* Hardware flatshading breaks down here.  If the hardware doesn't
+    * support flatshading, this will already have been done:
+    */
+   if ((ctx->_TriangleCaps & DD_FLATSHADE) && HAVE_HW_FLATSHADE) {
+      if (HAVE_RGBA) {
+	 VERT_SAVE_RGBA(0);
+	 VERT_SAVE_RGBA(1);
+	 VERT_SAVE_RGBA(2);
+	 VERT_COPY_RGBA(v[0], v[3]);
+	 VERT_COPY_RGBA(v[1], v[3]);
+	 VERT_COPY_RGBA(v[2], v[3]);
+
+	 if (HAVE_SPEC) {
+	    VERT_SAVE_SPEC(0);
+	    VERT_SAVE_SPEC(1);
+	    VERT_SAVE_SPEC(2);
+	    VERT_COPY_SPEC(v[0], v[3]);
+	    VERT_COPY_SPEC(v[1], v[3]);
+	    VERT_COPY_SPEC(v[2], v[3]);
+	 }
+      } else {
+	 VERT_SAVE_IND(0);
+	 VERT_SAVE_IND(1);
+	 VERT_SAVE_IND(2);
+	 VERT_COPY_IND(v[0], v[3]);
+	 VERT_COPY_IND(v[1], v[3]);
+	 VERT_COPY_IND(v[2], v[3]);
+      }
+   }
+
+   if (mode == GL_POINT) {
+      RASTERIZE(GL_POINTS);
+      if (ef[e0]) POINT( v[0] );
+      if (ef[e1]) POINT( v[1] );
+      if (ef[e2]) POINT( v[2] );
+      if (ef[e3]) POINT( v[3] );
+   }
+   else {
+      RASTERIZE(GL_LINES);
+      if (ef[e0]) LINE( v[0], v[1] );
+      if (ef[e1]) LINE( v[1], v[2] );
+      if (ef[e2]) LINE( v[2], v[3] );
+      if (ef[e3]) LINE( v[3], v[0] );
+   }
+
+   if ((ctx->_TriangleCaps & DD_FLATSHADE) && HAVE_HW_FLATSHADE) {
+      if (HAVE_RGBA) {
+	 VERT_RESTORE_RGBA(0);
+	 VERT_RESTORE_RGBA(1);
+	 VERT_RESTORE_RGBA(2);
+
+	 if (HAVE_SPEC) {
+	    VERT_RESTORE_SPEC(0);
+	    VERT_RESTORE_SPEC(1);
+	    VERT_RESTORE_SPEC(2);
+	 }
+      } else {
+	 VERT_RESTORE_IND(0);
+	 VERT_RESTORE_IND(1);
+	 VERT_RESTORE_IND(2);
+      }
+   }
+}
+
+
+#if HAVE_RGBA
+#undef VERT_SET_IND
+#undef VERT_COPY_IND
+#undef VERT_SAVE_IND
+#undef VERT_RESTORE_IND
+#endif
+
+#if !HAVE_SPEC
+#undef VERT_SET_SPEC
+#undef VERT_COPY_SPEC
+#undef VERT_SAVE_SPEC
+#undef VERT_RESTORE_SPEC
+#endif
+
+#undef TAG
diff --git a/src/mesa/tnl_dd/t_dd_vb.c b/src/mesa/tnl_dd/t_dd_vb.c
new file mode 100644
index 0000000000..4a742bcbbe
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_vb.c
@@ -0,0 +1,392 @@
+
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+#include "math/m_translate.h"
+
+#if (HAVE_HW_VIEWPORT)
+#define UNVIEWPORT_VARS
+#define UNVIEWPORT_X(x) x
+#define UNVIEWPORT_Y(x) x
+#define UNVIEWPORT_Z(x) x
+#endif
+
+#ifndef LOCALVARS
+#define LOCALVARS
+#endif
+
+#ifndef CHECK_HW_DIVIDE
+#define CHECK_HW_DIVIDE 1
+#endif
+
+/* These don't need to be duplicated, but there's currently nowhere
+ * really convenient to put them.  Need to build some actual .o files in
+ * this directory?
+ */
+static void copy_pv_rgba4_spec5( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   LOCALVARS   
+   GLubyte *verts = GET_VERTEX_STORE();
+   GLuint shift = GET_VERTEX_STRIDE_SHIFT();
+   GLuint *dst = (GLuint *)(verts + (edst << shift));
+   GLuint *src = (GLuint *)(verts + (esrc << shift));
+   dst[4] = src[4];
+   dst[5] = src[5];
+}
+
+static void copy_pv_rgba4( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   LOCALVARS
+   GLubyte *verts = GET_VERTEX_STORE();
+   GLuint shift = GET_VERTEX_STRIDE_SHIFT();
+   GLuint *dst = (GLuint *)(verts + (edst << shift));
+   GLuint *src = (GLuint *)(verts + (esrc << shift));
+   dst[4] = src[4];
+}
+
+static void copy_pv_rgba3( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   LOCALVARS
+   GLubyte *verts = GET_VERTEX_STORE();
+   GLuint shift = GET_VERTEX_STRIDE_SHIFT();
+   GLuint *dst = (GLuint *)(verts + (edst << shift));
+   GLuint *src = (GLuint *)(verts + (esrc << shift));
+   dst[3] = src[3];
+}
+
+
+void TAG(translate_vertex)(GLcontext *ctx,
+			   const VERTEX *src,
+			   SWvertex *dst)
+{
+   LOCALVARS
+   GLuint format = GET_VERTEX_FORMAT();
+   GLfloat *s = ctx->Viewport._WindowMap.m;
+   UNVIEWPORT_VARS;
+
+   if (format == TINY_VERTEX_FORMAT) {
+      if (HAVE_HW_VIEWPORT) {
+	 dst->win[0] = s[0]  * src->v.x + s[12];
+	 dst->win[1] = s[5]  * src->v.y + s[13];
+	 dst->win[2] = s[10] * src->v.z + s[14];
+	 dst->win[3] = 1.0;
+      } else {
+	 dst->win[0] = UNVIEWPORT_X( src->v.x );
+	 dst->win[1] = UNVIEWPORT_Y( src->v.y );
+	 dst->win[2] = UNVIEWPORT_Z( src->v.z );
+	 dst->win[3] = 1.0;
+      }
+
+      dst->color[0] = src->tv.color.red;
+      dst->color[1] = src->tv.color.green;
+      dst->color[2] = src->tv.color.blue;
+      dst->color[3] = src->tv.color.alpha;
+   }
+   else {
+      if (HAVE_HW_VIEWPORT) {
+	 if (HAVE_HW_DIVIDE && CHECK_HW_DIVIDE) {
+	    GLfloat oow = 1.0 / src->v.w;
+	    dst->win[0] = s[0]  * src->v.x * oow + s[12];
+	    dst->win[1] = s[5]  * src->v.y * oow + s[13];
+	    dst->win[2] = s[10] * src->v.z * oow + s[14];
+	    dst->win[3] = oow;
+	 } else {
+	    dst->win[0] = s[0]  * src->v.x + s[12];
+	    dst->win[1] = s[5]  * src->v.y + s[13];
+	    dst->win[2] = s[10] * src->v.z + s[14];
+	    dst->win[3] = src->v.w;
+	 }
+      } else {
+	 dst->win[0] = UNVIEWPORT_X( src->v.x );
+	 dst->win[1] = UNVIEWPORT_Y( src->v.y );
+	 dst->win[2] = UNVIEWPORT_Z( src->v.z );
+	 dst->win[3] = src->v.w;
+      }
+
+      dst->color[0] = src->v.color.red;
+      dst->color[1] = src->v.color.green;
+      dst->color[2] = src->v.color.blue;
+      dst->color[3] = src->v.color.alpha;
+
+      dst->specular[0] = src->v.specular.red;
+      dst->specular[1] = src->v.specular.green;
+      dst->specular[2] = src->v.specular.blue;
+
+      dst->fog = src->v.specular.alpha/255.0;
+
+      if (HAVE_PTEX_VERTICES &&
+	  ((HAVE_TEX2_VERTICES && format == PROJ_TEX3_VERTEX_FORMAT) ||
+	   (format == PROJ_TEX1_VERTEX_FORMAT))) {
+
+	 dst->texcoord[0][0] = src->pv.u0;
+	 dst->texcoord[0][1] = src->pv.v0;
+	 dst->texcoord[0][3] = src->pv.q0;
+
+	 dst->texcoord[1][0] = src->pv.u1;
+	 dst->texcoord[1][1] = src->pv.v1;
+	 dst->texcoord[1][3] = src->pv.q1;
+
+	 if (HAVE_TEX2_VERTICES) {
+	    dst->texcoord[2][0] = src->pv.u2;
+	    dst->texcoord[2][1] = src->pv.v2;
+	    dst->texcoord[2][3] = src->pv.q2;
+	 }
+
+	 if (HAVE_TEX3_VERTICES) {
+	    dst->texcoord[3][0] = src->pv.u3;
+	    dst->texcoord[3][1] = src->pv.v3;
+	    dst->texcoord[3][3] = src->pv.q3;
+	 }
+      }
+      else {
+	 dst->texcoord[0][0] = src->v.u0;
+	 dst->texcoord[0][1] = src->v.v0;
+	 dst->texcoord[0][3] = 1.0;
+
+	 dst->texcoord[1][0] = src->v.u1;
+	 dst->texcoord[1][1] = src->v.v1;
+	 dst->texcoord[1][3] = 1.0;
+
+	 if (HAVE_TEX2_VERTICES) {
+	    dst->texcoord[2][0] = src->v.u2;
+	    dst->texcoord[2][1] = src->v.v2;
+	    dst->texcoord[2][3] = 1.0;
+	 }
+
+	 if (HAVE_TEX3_VERTICES) {
+	    dst->texcoord[3][0] = src->v.u3;
+	    dst->texcoord[3][1] = src->v.v3;
+	    dst->texcoord[3][3] = 1.0;
+	 }
+      }
+   }
+
+   dst->pointSize = ctx->Point._Size;
+}
+
+
+
+void TAG(print_vertex)( GLcontext *ctx, const VERTEX *v )
+{
+   LOCALVARS
+   GLuint format = GET_VERTEX_FORMAT();
+
+   fprintf(stderr, "(%x) ", format);
+
+   switch (format) {
+#if HAVE_TINY_VERTICES
+   case TINY_VERTEX_FORMAT:
+      fprintf(stderr, "xyz %.4f,%.4f,%.4f rgba %x:%x:%x:%x\n",
+	      v->v.x, v->v.y, v->v.z,
+	      v->tv.color.red,
+	      v->tv.color.green,
+	      v->tv.color.blue,
+	      v->tv.color.alpha);
+      break;
+#endif
+#if HAVE_NOTEX_VERTICES
+   case NOTEX_VERTEX_FORMAT:
+      fprintf(stderr, "xyzw %.4f,%.4f,%.4f,%.4f rgba %x:%x:%x:%x spec %x:%x:%x:%x\n",
+	      v->v.x, v->v.y, v->v.z, v->v.w,
+	      v->v.color.red,
+	      v->v.color.green,
+	      v->v.color.blue,
+	      v->v.color.alpha,
+	      v->v.specular.red,
+	      v->v.specular.green,
+	      v->v.specular.blue,
+	      v->v.specular.alpha);
+      break;
+#endif
+#if HAVE_TEX0_VERTICES
+   case TEX0_VERTEX_FORMAT:
+      fprintf(stderr, "xyzw %.4f,%.4f,%.4f,%.4f rgba %x:%x:%x:%x st %.4f,%.4f\n",
+	      v->v.x, v->v.y, v->v.z, v->v.w,
+	      v->v.color.red,
+	      v->v.color.green,
+	      v->v.color.blue,
+	      v->v.color.alpha,
+	      v->v.u0,
+	      v->v.v0);
+      break;
+#endif
+#if HAVE_TEX1_VERTICES
+   case TEX1_VERTEX_FORMAT:
+      fprintf(stderr, "xyzw %.4f,%.4f,%.4f,%.4f rgba %x:%x:%x:%x st %.4f,%.4f st %.4f,%.4f\n",
+	      v->v.x, v->v.y, v->v.z, v->v.w,
+	      v->v.color.red,
+	      v->v.color.green,
+	      v->v.color.blue,
+	      v->v.color.alpha,
+	      v->v.u0,
+	      v->v.v0,
+	      v->v.u1,
+	      v->v.u2);
+      break;
+#endif
+#if HAVE_PTEX_VERTICES
+   case PROJ_TEX1_VERTEX_FORMAT:
+      fprintf(stderr, "xyzw %.4f,%.4f,%.4f,%.4f rgba %x:%x:%x:%x stq %.4f,%.4f,%.4f stq %.4f,%.4f,%.4f\n",
+	      v->v.x, v->v.y, v->v.z, v->v.w,
+	      v->v.color.red,
+	      v->v.color.green,
+	      v->v.color.blue,
+	      v->v.color.alpha,
+	      v->pv.u0,
+	      v->pv.v0,
+	      v->pv.q0,
+	      v->pv.u1,
+	      v->pv.v1,
+	      v->pv.q1);
+      break;
+#endif      
+   default:
+      fprintf(stderr, "???\n");
+      break;
+   }
+
+   fprintf(stderr, "\n");
+}
+
+static void do_import( struct vertex_buffer *VB,
+		       struct gl_client_array *to,
+		       struct gl_client_array *from )
+{
+   GLuint count = VB->Count;
+
+   if (!to->Ptr) {
+      to->Ptr = ALIGN_MALLOC( VB->Size * 4 * sizeof(GLubyte), 32 );
+      to->Type = GL_UNSIGNED_BYTE;
+   }
+
+   /* No need to transform the same value 3000 times.
+    */
+   if (!from->StrideB) {
+      to->StrideB = 0;
+      count = 1;
+   }
+   else
+      to->StrideB = 4 * sizeof(GLubyte);
+   
+   _math_trans_4ub( (GLubyte (*)[4]) to->Ptr,
+		    from->Ptr,
+		    from->StrideB,
+		    from->Type,
+		    from->Size,
+		    0,
+		    count);
+}
+
+#ifndef IMPORT_QUALIFIER
+#define IMPORT_QUALIFIER static
+#endif
+
+IMPORT_QUALIFIER void TAG(import_float_colors)( GLcontext *ctx )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct gl_client_array *to = GET_UBYTE_COLOR_STORE();
+   do_import( VB, to, VB->ColorPtr[0] );
+   VB->ColorPtr[0] = to;
+}
+
+IMPORT_QUALIFIER void TAG(import_float_spec_colors)( GLcontext *ctx )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct gl_client_array *to = GET_UBYTE_SPEC_COLOR_STORE();
+   do_import( VB, to, VB->SecondaryColorPtr[0] );
+   VB->SecondaryColorPtr[0] = to;
+}
+
+/* Interpolate the elements of the VB not included in typical hardware
+ * vertices.  
+ *
+ * NOTE: All these arrays are guarenteed by tnl to be writeable and
+ * have good stride.
+ */
+#ifndef INTERP_QUALIFIER 
+#define INTERP_QUALIFIER static
+#endif
+
+#define GET_COLOR(ptr, idx) (((GLchan (*)[4])((ptr)->Ptr))[idx])
+
+
+INTERP_QUALIFIER void TAG(interp_extras)( GLcontext *ctx,
+					  GLfloat t,
+					  GLuint dst, GLuint out, GLuint in,
+					  GLboolean force_boundary )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   if (VB->ColorPtr[1]) {
+      INTERP_4CHAN( t,
+		    GET_COLOR(VB->ColorPtr[1], dst),
+		    GET_COLOR(VB->ColorPtr[1], out),
+		    GET_COLOR(VB->ColorPtr[1], in) );
+
+      if (VB->SecondaryColorPtr[1]) {
+	 INTERP_3CHAN( t,
+		       GET_COLOR(VB->SecondaryColorPtr[1], dst),
+		       GET_COLOR(VB->SecondaryColorPtr[1], out),
+		       GET_COLOR(VB->SecondaryColorPtr[1], in) );
+      }
+   }
+
+   if (VB->EdgeFlag) {
+      VB->EdgeFlag[dst] = VB->EdgeFlag[out] || force_boundary;
+   }
+
+   INTERP_VERTEX(ctx, t, dst, out, in, force_boundary);
+}
+
+INTERP_QUALIFIER void TAG(copy_pv_extras)( GLcontext *ctx, 
+					   GLuint dst, GLuint src )
+{
+   LOCALVARS
+      struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   if (VB->ColorPtr[1]) {
+      COPY_CHAN4( GET_COLOR(VB->ColorPtr[1], dst), 
+		  GET_COLOR(VB->ColorPtr[1], src) );
+
+      if (VB->SecondaryColorPtr[1]) {
+	 COPY_CHAN4( GET_COLOR(VB->SecondaryColorPtr[1], dst), 
+		     GET_COLOR(VB->SecondaryColorPtr[1], src) );
+      }
+   }
+
+   COPY_PV_VERTEX(ctx, dst, src);
+}
+
+
+#undef INTERP_QUALIFIER
+#undef IMPORT_QUALIFIER
+#undef GET_COLOR
+
+#undef IND
+#undef TAG
diff --git a/src/mesa/tnl_dd/t_dd_vbtmp.h b/src/mesa/tnl_dd/t_dd_vbtmp.h
new file mode 100644
index 0000000000..5aad0b5622
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_vbtmp.h
@@ -0,0 +1,834 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  5.0.1
+ *
+ * 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>
+ */
+
+
+/* Unlike the other templates here, this assumes quite a bit about the
+ * underlying hardware.  Specifically it assumes a d3d-like vertex
+ * format, with a layout more or less constrained to look like the
+ * following:
+ *
+ * union {
+ *    struct {
+ *        float x, y, z, w;
+ *        struct { char r, g, b, a; } color;
+ *        struct { char r, g, b, fog; } spec;
+ *        float u0, v0;
+ *        float u1, v1;
+ *        float u2, v2;
+ *        float u3, v3;
+ *    } v;
+ *    struct {
+ *        float x, y, z, w;
+ *        struct { char r, g, b, a; } color;
+ *        struct { char r, g, b, fog; } spec;
+ *        float u0, v0, q0;
+ *        float u1, v1, q1;
+ *        float u2, v2, q2;
+ *        float u3, v3, q3;
+ *    } pv;
+ *    struct {
+ *        float x, y, z;
+ *        struct { char r, g, b, a; } color;
+ *    } tv;
+ *    float f[16];
+ *    unsigned int ui[16];
+ *    unsigned char ub4[4][16];
+ * }
+ *
+ 
+ * VERTEX:   hw vertex type as above
+ * VERTEX_COLOR: hw color struct type in VERTEX
+ *
+ * DO_XYZW:  Emit xyz and maybe w coordinates.
+ * DO_RGBA:  Emit color.
+ * DO_SPEC:  Emit specular color.
+ * DO_FOG:   Emit fog coordinate in specular alpha.
+ * DO_TEX0:  Emit tex0 u,v coordinates.
+ * DO_TEX1:  Emit tex1 u,v coordinates.
+ * DO_TEX2:  Emit tex2 u,v coordinates.
+ * DO_TEX3:  Emit tex3 u,v coordinates.
+ * DO_PTEX:  Emit tex0,1,2,3 q coordinates where possible.
+ *
+ * HAVE_RGBA_COLOR: Hardware takes color in rgba order (else bgra).
+ *
+ * HAVE_HW_VIEWPORT:  Hardware performs viewport transform.
+ * HAVE_HW_DIVIDE:  Hardware performs perspective divide.
+ *
+ * HAVE_TINY_VERTICES:  Hardware understands v.tv format.
+ * HAVE_PTEX_VERTICES:  Hardware understands v.pv format.
+ * HAVE_NOTEX_VERTICES:  Hardware understands v.v format with texcount 0.
+ *
+ * Additionally, this template assumes it is emitting *transformed*
+ * vertices; the modifications to emit untransformed vertices (ie. to
+ * t&l hardware) are probably too great to cooexist with the code
+ * already in this file.
+ *
+ * NOTE: The PTEX vertex format always includes TEX0 and TEX1, even if
+ * only TEX0 is enabled, in order to maintain a vertex size which is
+ * an exact number of quadwords.
+ */
+
+#if (HAVE_HW_VIEWPORT)
+#define VIEWPORT_X(dst,x) dst = x
+#define VIEWPORT_Y(dst,y) dst = y
+#define VIEWPORT_Z(dst,z) dst = z
+#else
+#define VIEWPORT_X(dst,x) dst = s[0]  * x + s[12]
+#define VIEWPORT_Y(dst,y) dst = s[5]  * y + s[13]
+#define VIEWPORT_Z(dst,z) dst = s[10] * z + s[14]
+#endif
+
+#if (HAVE_HW_DIVIDE && !HAVE_PTEX_VERTICES)
+#error "can't cope with this combination" 
+#endif 
+
+#ifndef LOCALVARS
+#define LOCALVARS
+#endif
+
+#ifndef CHECK_HW_DIVIDE
+#define CHECK_HW_DIVIDE 1
+#endif
+
+#if (HAVE_HW_DIVIDE || DO_SPEC || DO_TEX0 || DO_FOG || !HAVE_TINY_VERTICES)
+
+static void TAG(emit)( GLcontext *ctx,
+		       GLuint start, GLuint end,
+		       void *dest,
+		       GLuint stride )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLfloat (*tc0)[4], (*tc1)[4], (*fog)[4];
+   GLfloat (*tc2)[4], (*tc3)[4];
+   GLubyte (*col)[4], (*spec)[4];
+   GLuint tc0_stride, tc1_stride, col_stride, spec_stride, fog_stride;
+   GLuint tc2_stride, tc3_stride;
+   GLuint tc0_size, tc1_size;
+   GLuint tc2_size, tc3_size;
+   GLfloat (*coord)[4];
+   GLuint coord_stride;
+   VERTEX *v = (VERTEX *)dest;
+   const GLfloat *s = GET_VIEWPORT_MAT();
+   const GLubyte *mask = VB->ClipMask;
+   int i;
+
+/*     fprintf(stderr, "%s(big) importable %d %d..%d\n",  */
+/*  	   __FUNCTION__, VB->importable_data, start, end); */
+
+   if (HAVE_HW_VIEWPORT && HAVE_HW_DIVIDE && CHECK_HW_DIVIDE) {
+      (void) s;
+      coord = VB->ClipPtr->data;
+      coord_stride = VB->ClipPtr->stride;
+   }
+   else {
+      coord = VB->NdcPtr->data;
+      coord_stride = VB->NdcPtr->stride;
+   }
+
+   if (DO_TEX3) {
+      const GLuint t3 = GET_TEXSOURCE(3);
+      tc3 = VB->TexCoordPtr[t3]->data;
+      tc3_stride = VB->TexCoordPtr[t3]->stride;
+      if (DO_PTEX)
+	 tc3_size = VB->TexCoordPtr[t3]->size;
+   }
+
+   if (DO_TEX2) {
+      const GLuint t2 = GET_TEXSOURCE(2);
+      tc2 = VB->TexCoordPtr[t2]->data;
+      tc2_stride = VB->TexCoordPtr[t2]->stride;
+      if (DO_PTEX)
+	 tc2_size = VB->TexCoordPtr[t2]->size;
+   }
+
+   if (DO_TEX1) {
+      const GLuint t1 = GET_TEXSOURCE(1);
+      tc1 = VB->TexCoordPtr[t1]->data;
+      tc1_stride = VB->TexCoordPtr[t1]->stride;
+      if (DO_PTEX)
+	 tc1_size = VB->TexCoordPtr[t1]->size;
+   }
+
+   if (DO_TEX0) {
+      const GLuint t0 = GET_TEXSOURCE(0);
+      tc0_stride = VB->TexCoordPtr[t0]->stride;
+      tc0 = VB->TexCoordPtr[t0]->data;
+      if (DO_PTEX) 
+	 tc0_size = VB->TexCoordPtr[t0]->size;
+   }
+
+   if (DO_RGBA) {
+      if (VB->ColorPtr[0]->Type != GL_UNSIGNED_BYTE)
+	 IMPORT_FLOAT_COLORS( ctx );
+      col = (GLubyte (*)[4])VB->ColorPtr[0]->Ptr;
+      col_stride = VB->ColorPtr[0]->StrideB;
+   }
+
+   if (DO_SPEC) {
+      if (VB->SecondaryColorPtr[0]) {
+	 if (VB->SecondaryColorPtr[0]->Type != GL_UNSIGNED_BYTE)
+	    IMPORT_FLOAT_SPEC_COLORS( ctx );
+	 spec = (GLubyte (*)[4])VB->SecondaryColorPtr[0]->Ptr;
+	 spec_stride = VB->SecondaryColorPtr[0]->StrideB;
+      } else {
+	 GLubyte tmp[4];
+	 spec = &tmp;
+	 spec_stride = 0;
+      }
+   }
+
+   if (DO_FOG) {
+      if (VB->FogCoordPtr) {
+	 fog = VB->FogCoordPtr->data;
+	 fog_stride = VB->FogCoordPtr->stride;
+      }
+      else {
+	 static GLfloat tmp[4] = {0, 0, 0, 0};
+	 fog = &tmp;
+	 fog_stride = 0;
+      }
+   }
+
+   if (VB->importable_data || (DO_SPEC && !spec_stride) || (DO_FOG && !fog_stride)) {
+      /* May have nonstandard strides:
+       */
+      if (start) {
+	 coord =  (GLfloat (*)[4])((GLubyte *)coord + start * coord_stride);
+	 if (DO_TEX0)
+	    tc0 =  (GLfloat (*)[4])((GLubyte *)tc0 + start * tc0_stride);
+	 if (DO_TEX1) 
+	    tc1 =  (GLfloat (*)[4])((GLubyte *)tc1 + start * tc1_stride);
+	 if (DO_TEX2) 
+	    tc2 =  (GLfloat (*)[4])((GLubyte *)tc2 + start * tc2_stride);
+	 if (DO_TEX3) 
+	    tc3 =  (GLfloat (*)[4])((GLubyte *)tc3 + start * tc3_stride);
+	 if (DO_RGBA) 
+	    STRIDE_4UB(col, start * col_stride);
+	 if (DO_SPEC)
+	    STRIDE_4UB(spec, start * spec_stride);
+	 if (DO_FOG)
+	    /*STRIDE_F(fog, start * fog_stride);*/
+	    fog =  (GLfloat (*)[4])((GLubyte *)fog + start * fog_stride);
+      }
+
+      for (i=start; i < end; i++, v = (VERTEX *)((GLubyte *)v + stride)) {
+	 if (DO_XYZW) {
+	    if (HAVE_HW_VIEWPORT || mask[i] == 0) {
+	       VIEWPORT_X(v->v.x, coord[0][0]);
+	       VIEWPORT_Y(v->v.y, coord[0][1]);
+	       VIEWPORT_Z(v->v.z, coord[0][2]);
+	       v->v.w = coord[0][3];
+	    }
+/*  	    fprintf(stderr, "vert %d: %.2f %.2f %.2f %.2f\n",  */
+/*  		    i, v->v.x, v->v.y, v->v.z, v->v.w); */
+	    coord =  (GLfloat (*)[4])((GLubyte *)coord +  coord_stride);
+	 }
+	 if (DO_RGBA) {
+	    if (HAVE_RGBA_COLOR) {
+	       *(GLuint *)&v->v.color = LE32_TO_CPU(*(GLuint *)&col[0]);
+	       STRIDE_4UB(col, col_stride);
+	    } else {
+	       v->v.color.blue  = col[0][2];
+	       v->v.color.green = col[0][1];
+	       v->v.color.red   = col[0][0];
+	       v->v.color.alpha = col[0][3];
+	       STRIDE_4UB(col, col_stride);
+	    }
+	 }
+	 if (DO_SPEC) {
+	    v->v.specular.red = spec[0][0];
+	    v->v.specular.green = spec[0][1];
+	    v->v.specular.blue = spec[0][2];
+	    STRIDE_4UB(spec, spec_stride);
+	 }
+	 if (DO_FOG) {
+	    v->v.specular.alpha = fog[0][0] * 255.0;
+	    /*STRIDE_F(fog, fog_stride);*/
+	    fog =  (GLfloat (*)[4])((GLubyte *)fog + fog_stride);
+	 }
+	 if (DO_TEX0) {
+	    v->v.u0 = tc0[0][0];
+	    v->v.v0 = tc0[0][1];
+	    if (DO_PTEX) {
+	       if (HAVE_PTEX_VERTICES) {
+		  if (tc0_size == 4) 
+		     v->pv.q0 = tc0[0][3];
+		  else
+		     v->pv.q0 = 1.0;
+	       } 
+	       else if (tc0_size == 4) {
+		  float rhw = 1.0 / tc0[0][3];
+		  v->v.w *= tc0[0][3];
+		  v->v.u0 *= rhw;
+		  v->v.v0 *= rhw;
+	       } 
+	    } 
+	    tc0 =  (GLfloat (*)[4])((GLubyte *)tc0 +  tc0_stride);
+	 }
+	 if (DO_TEX1) {
+	    if (DO_PTEX) {
+	       v->pv.u1 = tc1[0][0];
+	       v->pv.v1 = tc1[0][1];
+	       if (tc1_size == 4) 
+		  v->pv.q1 = tc1[0][3];
+	       else
+		  v->pv.q1 = 1.0;
+	    } 
+	    else {
+	       v->v.u1 = tc1[0][0];
+	       v->v.v1 = tc1[0][1];
+	    }
+	    tc1 =  (GLfloat (*)[4])((GLubyte *)tc1 +  tc1_stride);
+	 } 
+	 else if (DO_PTEX) {
+	    *(GLuint *)&v->pv.q1 = 0;	/* avoid culling on radeon */
+	 }
+	 if (DO_TEX2) {
+	    if (DO_PTEX) {
+	       v->pv.u2 = tc2[0][0];
+	       v->pv.v2 = tc2[0][1];
+	       if (tc2_size == 4) 
+		  v->pv.q2 = tc2[0][3];
+	       else
+		  v->pv.q2 = 1.0;
+	    } 
+	    else {
+	       v->v.u2 = tc2[0][0];
+	       v->v.v2 = tc2[0][1];
+	    }
+	    tc2 =  (GLfloat (*)[4])((GLubyte *)tc2 +  tc2_stride);
+	 } 
+	 if (DO_TEX3) {
+	    if (DO_PTEX) {
+	       v->pv.u3 = tc3[0][0];
+	       v->pv.v3 = tc3[0][1];
+	       if (tc3_size == 4) 
+		  v->pv.q3 = tc3[0][3];
+	       else
+		  v->pv.q3 = 1.0;
+	    } 
+	    else {
+	       v->v.u3 = tc3[0][0];
+	       v->v.v3 = tc3[0][1];
+	    }
+	    tc3 =  (GLfloat (*)[4])((GLubyte *)tc3 +  tc3_stride);
+	 } 
+      }
+   }
+   else {
+      for (i=start; i < end; i++, v = (VERTEX *)((GLubyte *)v + stride)) {
+	 if (DO_XYZW) {
+	    if (HAVE_HW_VIEWPORT || mask[i] == 0) {
+	       VIEWPORT_X(v->v.x, coord[i][0]);
+	       VIEWPORT_Y(v->v.y, coord[i][1]);
+	       VIEWPORT_Z(v->v.z, coord[i][2]);
+	       v->v.w = coord[i][3];
+	    }
+	 }
+	 if (DO_RGBA) {
+	    if (HAVE_RGBA_COLOR) {
+	       *(GLuint *)&v->v.color = LE32_TO_CPU(*(GLuint *)&col[i]);
+	    }
+	    else {
+	       v->v.color.blue  = col[i][2];
+	       v->v.color.green = col[i][1];
+	       v->v.color.red   = col[i][0];
+	       v->v.color.alpha = col[i][3];
+	    }
+	 }
+	 if (DO_SPEC) {
+	    v->v.specular.red   = spec[i][0];
+	    v->v.specular.green = spec[i][1];
+	    v->v.specular.blue  = spec[i][2];
+	 }
+	 if (DO_FOG) {
+	    v->v.specular.alpha = fog[i][0] * 255.0;
+	 }
+	 if (DO_TEX0) {
+	    v->v.u0 = tc0[i][0];
+	    v->v.v0 = tc0[i][1];
+	    if (DO_PTEX) {
+	       if (HAVE_PTEX_VERTICES) {
+		  if (tc0_size == 4) 
+		     v->pv.q0 = tc0[i][3];
+		  else
+		     v->pv.q0 = 1.0;
+
+		  v->pv.q1 = 0;	/* radeon */
+	       } 
+	       else if (tc0_size == 4) {
+		  float rhw = 1.0 / tc0[i][3];
+		  v->v.w *= tc0[i][3];
+		  v->v.u0 *= rhw;
+		  v->v.v0 *= rhw;
+	       } 
+	    } 
+	 }
+	 if (DO_TEX1) {
+	    if (DO_PTEX) {
+	       v->pv.u1 = tc1[i][0];
+	       v->pv.v1 = tc1[i][1];
+	       if (tc1_size == 4) 
+		  v->pv.q1 = tc1[i][3];
+	       else
+		  v->pv.q1 = 1.0;
+	    } 
+	    else {
+	       v->v.u1 = tc1[i][0];
+	       v->v.v1 = tc1[i][1];
+	    }
+	 }
+      }
+   }
+}
+#else
+#if DO_XYZW
+
+#if HAVE_HW_DIVIDE
+#error "cannot use tiny vertices with hw perspective divide"
+#endif
+
+static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end,
+		       void *dest, GLuint stride )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLubyte (*col)[4];
+   GLuint col_stride;
+   GLfloat (*coord)[4] = VB->NdcPtr->data;
+   GLuint coord_stride = VB->NdcPtr->stride;
+   GLfloat *v = (GLfloat *)dest;
+   const GLubyte *mask = VB->ClipMask;
+   const GLfloat *s = GET_VIEWPORT_MAT();
+   int i;
+
+   (void) s;
+
+   ASSERT(stride == 4);
+
+   if (VB->ColorPtr[0]->Type != GL_UNSIGNED_BYTE)
+      IMPORT_FLOAT_COLORS( ctx );
+
+   col = (GLubyte (*)[4])VB->ColorPtr[0]->Ptr;
+   col_stride = VB->ColorPtr[0]->StrideB;
+   ASSERT(VB->ColorPtr[0]->Type == GL_UNSIGNED_BYTE);
+
+/*     fprintf(stderr, "%s(small) importable %x\n",  */
+/*  	   __FUNCTION__, VB->importable_data); */
+
+   /* Pack what's left into a 4-dword vertex.  Color is in a different
+    * place, and there is no 'w' coordinate.
+    */
+   if (VB->importable_data) {
+      if (start) {
+	 coord =  (GLfloat (*)[4])((GLubyte *)coord + start * coord_stride);
+	 STRIDE_4UB(col, start * col_stride);
+      }
+
+      for (i=start; i < end; i++, v+=4) {
+	 if (HAVE_HW_VIEWPORT || mask[i] == 0) {
+	    VIEWPORT_X(v[0], coord[0][0]);
+	    VIEWPORT_Y(v[1], coord[0][1]);
+	    VIEWPORT_Z(v[2], coord[0][2]);
+	 }
+	 coord =  (GLfloat (*)[4])((GLubyte *)coord +  coord_stride);
+	 if (DO_RGBA) {
+	    if (HAVE_RGBA_COLOR) {
+	       *(GLuint *)&v[3] = LE32_TO_CPU(*(GLuint *)col);
+	    }
+	    else {
+	       VERTEX_COLOR *c = (VERTEX_COLOR *)&v[3];
+	       c->blue  = col[0][2];
+	       c->green = col[0][1];
+	       c->red   = col[0][0];
+	       c->alpha = col[0][3];
+	    }
+	    STRIDE_4UB( col, col_stride );
+	 }
+/*  	 fprintf(stderr, "vert %d: %.2f %.2f %.2f %x\n",  */
+/*  		 i, v[0], v[1], v[2], *(int *)&v[3]); */
+      }
+   }
+   else {
+      for (i=start; i < end; i++, v+=4) {
+	 if (HAVE_HW_VIEWPORT || mask[i] == 0) {
+	    VIEWPORT_X(v[0], coord[i][0]);
+	    VIEWPORT_Y(v[1], coord[i][1]);
+	    VIEWPORT_Z(v[2], coord[i][2]);
+	 }
+	 if (DO_RGBA) {
+	    if (HAVE_RGBA_COLOR) {
+	       *(GLuint *)&v[3] = LE32_TO_CPU(*(GLuint *)&col[i]);
+	    }
+	    else {
+	       VERTEX_COLOR *c = (VERTEX_COLOR *)&v[3];
+	       c->blue  = col[i][2];
+	       c->green = col[i][1];
+	       c->red   = col[i][0];
+	       c->alpha = col[i][3];
+	    }
+	 }
+/*  	 fprintf(stderr, "vert %d: %.2f %.2f %.2f %x\n",  */
+/*  		 i, v[0], v[1], v[2], *(int *)&v[3]); */
+
+      }
+   }
+}
+#else
+static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end,
+		       void *dest, GLuint stride )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLubyte (*col)[4];
+   GLuint col_stride;
+   GLfloat *v = (GLfloat *)dest;
+   int i;
+
+   if (VB->ColorPtr[0]->Type != GL_UNSIGNED_BYTE)
+      IMPORT_FLOAT_COLORS( ctx );
+
+   col = VB->ColorPtr[0]->Ptr;
+   col_stride = VB->ColorPtr[0]->StrideB;
+
+   if (start)
+      STRIDE_4UB(col, col_stride * start);
+
+   /* Need to figure out where color is:
+    */
+   if (GET_VERTEX_FORMAT() == TINY_VERTEX_FORMAT)
+      v += 3;
+   else
+      v += 4;
+
+   for (i=start; i < end; i++, STRIDE_F(v, stride)) {
+      if (HAVE_RGBA_COLOR) {
+	 *(GLuint *)v = LE32_TO_CPU(*(GLuint *)col[0]);
+      }
+      else {
+	 VERTEX_COLOR *c = (VERTEX_COLOR *)v;
+	 c->blue  = col[0][2];
+	 c->green = col[0][1];
+	 c->red   = col[0][0];
+	 c->alpha = col[0][3];
+      }
+      STRIDE_4UB( col, col_stride );
+   }
+}
+#endif /* emit */
+#endif /* emit */
+
+#if (DO_XYZW) && (DO_RGBA)
+
+
+#if (HAVE_PTEX_VERTICES)
+static GLboolean TAG(check_tex_sizes)( GLcontext *ctx )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   /* Force 'missing' texcoords to something valid.
+    */
+   if (DO_TEX3 && VB->TexCoordPtr[2] == 0)
+      VB->TexCoordPtr[2] = VB->TexCoordPtr[3];
+
+   if (DO_TEX2 && VB->TexCoordPtr[1] == 0)
+      VB->TexCoordPtr[1] = VB->TexCoordPtr[2];
+
+   if (DO_TEX1 && VB->TexCoordPtr[0] == 0)
+      VB->TexCoordPtr[0] = VB->TexCoordPtr[1];
+
+   if (DO_PTEX)
+      return GL_TRUE;
+   
+   if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) ||
+       (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) ||
+       (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4) ||
+       (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4))
+      return GL_FALSE;
+
+   return GL_TRUE;
+}
+#else
+static GLboolean TAG(check_tex_sizes)( GLcontext *ctx )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   /* Force 'missing' texcoords to something valid.
+    */
+   if (DO_TEX3 && VB->TexCoordPtr[2] == 0)
+      VB->TexCoordPtr[2] = VB->TexCoordPtr[3];
+
+   if (DO_TEX2 && VB->TexCoordPtr[1] == 0)
+      VB->TexCoordPtr[1] = VB->TexCoordPtr[2];
+
+   if (DO_TEX1 && VB->TexCoordPtr[0] == 0)
+      VB->TexCoordPtr[0] = VB->TexCoordPtr[1];
+
+   if (DO_PTEX)
+      return GL_TRUE;
+
+   /* No hardware support for projective texture.  Can fake it for
+    * TEX0 only.
+    */
+   if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) ||
+       (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) ||
+       (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) {
+      PTEX_FALLBACK();
+      return GL_FALSE;
+   }
+
+   if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) {
+      if (DO_TEX1 || DO_TEX2 || DO_TEX3) {
+	 PTEX_FALLBACK();
+      }
+      return GL_FALSE;
+   }
+
+   return GL_TRUE;
+}
+#endif /* ptex */
+
+
+static void TAG(interp)( GLcontext *ctx,
+			 GLfloat t,
+			 GLuint edst, GLuint eout, GLuint ein,
+			 GLboolean force_boundary )
+{
+   LOCALVARS
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLubyte *ddverts = GET_VERTEX_STORE();
+   GLuint shift = GET_VERTEX_STRIDE_SHIFT();
+   const GLfloat *dstclip = VB->ClipPtr->data[edst];
+   GLfloat w;
+   const GLfloat *s = GET_VIEWPORT_MAT();
+
+   VERTEX *dst = (VERTEX *)(ddverts + (edst << shift));
+   VERTEX *in  = (VERTEX *)(ddverts + (ein << shift));
+   VERTEX *out = (VERTEX *)(ddverts + (eout << shift));
+
+   (void)s;
+
+   if (HAVE_HW_DIVIDE && CHECK_HW_DIVIDE) {
+      VIEWPORT_X( dst->v.x, dstclip[0] );
+      VIEWPORT_Y( dst->v.y, dstclip[1] );
+      VIEWPORT_Z( dst->v.z, dstclip[2] );
+      w = dstclip[3];
+   }
+   else {
+      w = 1.0 / dstclip[3];
+      VIEWPORT_X( dst->v.x, dstclip[0] * w );
+      VIEWPORT_Y( dst->v.y, dstclip[1] * w );
+      VIEWPORT_Z( dst->v.z, dstclip[2] * w );
+   }
+
+   if ((HAVE_HW_DIVIDE && CHECK_HW_DIVIDE) || 
+       DO_FOG || DO_SPEC || DO_TEX0 || DO_TEX1 ||
+       DO_TEX2 || DO_TEX3 || !HAVE_TINY_VERTICES) {
+
+      dst->v.w = w;
+
+      INTERP_UB( t, dst->ub4[4][0], out->ub4[4][0], in->ub4[4][0] );
+      INTERP_UB( t, dst->ub4[4][1], out->ub4[4][1], in->ub4[4][1] );
+      INTERP_UB( t, dst->ub4[4][2], out->ub4[4][2], in->ub4[4][2] );
+      INTERP_UB( t, dst->ub4[4][3], out->ub4[4][3], in->ub4[4][3] );
+
+      if (DO_SPEC) {
+	 INTERP_UB( t, dst->v.specular.red,   out->v.specular.red,   in->v.specular.red );
+	 INTERP_UB( t, dst->v.specular.green, out->v.specular.green, in->v.specular.green );
+	 INTERP_UB( t, dst->v.specular.blue,  out->v.specular.blue,  in->v.specular.blue );
+      }
+      if (DO_FOG) {
+	 INTERP_UB( t, dst->v.specular.alpha, out->v.specular.alpha, in->v.specular.alpha );
+      }
+      if (DO_TEX0) {
+	 if (DO_PTEX) {
+	    if (HAVE_PTEX_VERTICES) {
+	       INTERP_F( t, dst->pv.u0, out->pv.u0, in->pv.u0 );
+	       INTERP_F( t, dst->pv.v0, out->pv.v0, in->pv.v0 );
+	       INTERP_F( t, dst->pv.q0, out->pv.q0, in->pv.q0 );
+	    } else {
+	       GLfloat wout = VB->NdcPtr->data[eout][3];
+	       GLfloat win = VB->NdcPtr->data[ein][3];
+	       GLfloat qout = out->pv.w / wout;
+	       GLfloat qin = in->pv.w / win;
+	       GLfloat qdst, rqdst;
+
+	       ASSERT( !HAVE_HW_DIVIDE );
+
+	       INTERP_F( t, dst->v.u0, out->v.u0 * qout, in->v.u0 * qin );
+	       INTERP_F( t, dst->v.v0, out->v.v0 * qout, in->v.v0 * qin );
+	       INTERP_F( t, qdst, qout, qin );
+
+	       rqdst = 1.0 / qdst;
+	       dst->v.u0 *= rqdst;
+	       dst->v.v0 *= rqdst;
+	       dst->v.w *= rqdst;
+	    }
+	 }
+	 else {
+	    INTERP_F( t, dst->v.u0, out->v.u0, in->v.u0 );
+	    INTERP_F( t, dst->v.v0, out->v.v0, in->v.v0 );
+	 }
+      }
+      if (DO_TEX1) {
+	 if (DO_PTEX) {
+	    INTERP_F( t, dst->pv.u1, out->pv.u1, in->pv.u1 );
+	    INTERP_F( t, dst->pv.v1, out->pv.v1, in->pv.v1 );
+	    INTERP_F( t, dst->pv.q1, out->pv.q1, in->pv.q1 );
+	 } else {
+	    INTERP_F( t, dst->v.u1, out->v.u1, in->v.u1 );
+	    INTERP_F( t, dst->v.v1, out->v.v1, in->v.v1 );
+	 }
+      }
+      else if (DO_PTEX) {
+	 dst->pv.q1 = 0.0;	/* must be a valid float on radeon */
+      }
+      if (DO_TEX2) {
+	 if (DO_PTEX) {
+	    INTERP_F( t, dst->pv.u2, out->pv.u2, in->pv.u2 );
+	    INTERP_F( t, dst->pv.v2, out->pv.v2, in->pv.v2 );
+	    INTERP_F( t, dst->pv.q2, out->pv.q2, in->pv.q2 );
+	 } else {
+	    INTERP_F( t, dst->v.u2, out->v.u2, in->v.u2 );
+	    INTERP_F( t, dst->v.v2, out->v.v2, in->v.v2 );
+	 }
+      }
+      if (DO_TEX3) {
+	 if (DO_PTEX) {
+	    INTERP_F( t, dst->pv.u3, out->pv.u3, in->pv.u3 );
+	    INTERP_F( t, dst->pv.v3, out->pv.v3, in->pv.v3 );
+	    INTERP_F( t, dst->pv.q3, out->pv.q3, in->pv.q3 );
+	 } else {
+	    INTERP_F( t, dst->v.u3, out->v.u3, in->v.u3 );
+	    INTERP_F( t, dst->v.v3, out->v.v3, in->v.v3 );
+	 }
+      }
+   } else {
+      /* 4-dword vertex.  Color is in v[3] and there is no oow coordinate.
+       */
+      INTERP_UB( t, dst->ub4[3][0], out->ub4[3][0], in->ub4[3][0] );
+      INTERP_UB( t, dst->ub4[3][1], out->ub4[3][1], in->ub4[3][1] );
+      INTERP_UB( t, dst->ub4[3][2], out->ub4[3][2], in->ub4[3][2] );
+      INTERP_UB( t, dst->ub4[3][3], out->ub4[3][3], in->ub4[3][3] );
+   }
+}
+
+#endif /* rgba && xyzw */
+
+
+static void TAG(init)( void )
+{
+   setup_tab[IND].emit = TAG(emit);
+
+#if (DO_XYZW && DO_RGBA)
+   setup_tab[IND].check_tex_sizes = TAG(check_tex_sizes);
+   setup_tab[IND].interp = TAG(interp);
+#endif
+
+   if (DO_SPEC)
+      setup_tab[IND].copy_pv = copy_pv_rgba4_spec5;
+   else if (HAVE_HW_DIVIDE || DO_SPEC || DO_FOG || DO_TEX0 || DO_TEX1 ||
+	    DO_TEX2 || DO_TEX3 || !HAVE_TINY_VERTICES)
+      setup_tab[IND].copy_pv = copy_pv_rgba4;
+   else
+      setup_tab[IND].copy_pv = copy_pv_rgba3;
+
+   if (DO_TEX3) {
+      if (DO_PTEX) {
+	 ASSERT(HAVE_PTEX_VERTICES);
+	 setup_tab[IND].vertex_format = PROJ_TEX3_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 18;
+	 setup_tab[IND].vertex_stride_shift = 7;
+      }
+      else {
+	 setup_tab[IND].vertex_format = TEX3_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 14;
+	 setup_tab[IND].vertex_stride_shift = 6;
+      }
+   }
+   else if (DO_TEX2) {
+      if (DO_PTEX) {
+	 ASSERT(HAVE_PTEX_VERTICES);
+	 setup_tab[IND].vertex_format = PROJ_TEX3_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 18;
+	 setup_tab[IND].vertex_stride_shift = 7;
+      }
+      else {
+	 setup_tab[IND].vertex_format = TEX2_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 12;
+	 setup_tab[IND].vertex_stride_shift = 6;
+      }
+   }
+   else if (DO_TEX1) {
+      if (DO_PTEX) {
+	 ASSERT(HAVE_PTEX_VERTICES);
+	 setup_tab[IND].vertex_format = PROJ_TEX1_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 12;
+	 setup_tab[IND].vertex_stride_shift = 6;
+      }
+      else {
+	 setup_tab[IND].vertex_format = TEX1_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 10;
+	 setup_tab[IND].vertex_stride_shift = 6;
+      }
+   }
+   else if (DO_TEX0) {
+      if (DO_PTEX && HAVE_PTEX_VERTICES) {
+	 setup_tab[IND].vertex_format = PROJ_TEX1_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 12;
+	 setup_tab[IND].vertex_stride_shift = 6;
+      } else {
+	 setup_tab[IND].vertex_format = TEX0_VERTEX_FORMAT;
+	 setup_tab[IND].vertex_size = 8;
+	 setup_tab[IND].vertex_stride_shift = 5;
+      }
+   }
+   else if (!HAVE_HW_DIVIDE && !DO_SPEC && !DO_FOG && HAVE_TINY_VERTICES) {
+      setup_tab[IND].vertex_format = TINY_VERTEX_FORMAT;
+      setup_tab[IND].vertex_size = 4;
+      setup_tab[IND].vertex_stride_shift = 4;
+   } else if (HAVE_NOTEX_VERTICES) {
+      setup_tab[IND].vertex_format = NOTEX_VERTEX_FORMAT;
+      setup_tab[IND].vertex_size = 6;
+      setup_tab[IND].vertex_stride_shift = 5;
+   } else {
+      setup_tab[IND].vertex_format = TEX0_VERTEX_FORMAT;
+      setup_tab[IND].vertex_size = 8;
+      setup_tab[IND].vertex_stride_shift = 5;
+   }
+
+   assert(setup_tab[IND].vertex_size * 4 <=
+          1 << setup_tab[IND].vertex_stride_shift);
+}
+
+
+#undef IND
+#undef TAG
diff --git a/src/mesa/tnl_dd/t_dd_vertex.h b/src/mesa/tnl_dd/t_dd_vertex.h
new file mode 100644
index 0000000000..d45dd09fbd
--- /dev/null
+++ b/src/mesa/tnl_dd/t_dd_vertex.h
@@ -0,0 +1,78 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  4.0.3
+ *
+ * Copyright (C) 1999-2002  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>
+ */
+
+typedef struct {
+   GLfloat x, y, z, w;
+} TAG(_coord_t);
+
+#ifdef COLOR_IS_RGBA
+typedef struct {
+#if defined(BYTE_ORDER) && defined(BIG_ENDIAN) && BYTE_ORDER == BIG_ENDIAN
+   GLubyte alpha, blue, green, red;
+#else
+   GLubyte red, green, blue, alpha;
+#endif
+} TAG(_color_t);
+#else
+typedef struct {
+#if defined(BYTE_ORDER) && defined(BIG_ENDIAN) && BYTE_ORDER == BIG_ENDIAN
+   GLubyte alpha, red, green, blue;
+#else
+   GLubyte blue, green, red, alpha;
+#endif
+} TAG(_color_t);
+#endif
+
+typedef union {
+   struct {
+      GLfloat x, y, z, w;
+      TAG(_color_t) color;
+      TAG(_color_t) specular;
+      GLfloat u0, v0;
+      GLfloat u1, v1;
+      GLfloat u2, v2;
+      GLfloat u3, v3;
+   } v;
+   struct {
+      GLfloat x, y, z, w;
+      TAG(_color_t) color;
+      TAG(_color_t) specular;
+      GLfloat u0, v0, q0;
+      GLfloat u1, v1, q1;
+      GLfloat u2, v2, q2;
+      GLfloat u3, v3, q3;
+   } pv;
+   struct {
+      GLfloat x, y, z;
+      TAG(_color_t) color;
+   } tv;
+   GLfloat f[24];
+   GLuint  ui[24];
+   GLubyte ub4[24][4];
+} TAG(Vertex), *TAG(VertexPtr);
+