Merge branch 'vbo-0.2'

Conflicts:

	src/mesa/main/texcompress_s3tc.c
	src/mesa/tnl/t_array_api.c

160 files changed, 4184 insertions, 17877 deletions

diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index 1cd9e5b2bb..0ab4fc5149 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -142,7 +142,7 @@ depend: $(ALL_SOURCES)
 	@ echo "running $(MKDEP)"
 	@ touch depend
 	@$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \
-		> /dev/null 
+		> /dev/null 2>/dev/null
 
 
 subdirs:
diff --git a/src/mesa/array_cache/ac_context.c b/src/mesa/array_cache/ac_context.c
deleted file mode 100644
index 5d5ad5e363..0000000000
--- a/src/mesa/array_cache/ac_context.c
+++ /dev/null
@@ -1,375 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * 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>
- */
-
-#include "glheader.h"
-#include "macros.h"
-#include "imports.h"
-#include "mtypes.h"
-
-#include "array_cache/ac_context.h"
-
-
-/*
- * Initialize the array fallbacks.  That is, by default the fallback arrays
- * point into the current vertex attribute values in ctx->Current.Attrib[]
- */
-static void _ac_fallbacks_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   struct gl_client_array *cl;
-   GLuint i;
-
-   cl = &ac->Fallback.Normal;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.Color;
-   cl->Size = 4;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.SecondaryColor;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.FogCoord;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) &ctx->Current.Attrib[VERT_ATTRIB_FOG];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.Index;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) &ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0 ; i < MAX_TEXTURE_COORD_UNITS ; i++) {
-      cl = &ac->Fallback.TexCoord[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 0;
-      cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_TEX0 + i];
-      cl->Enabled = 1;
-      cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-
-   cl = &ac->Fallback.EdgeFlag;
-   cl->Size = 1;
-   cl->Type = GL_UNSIGNED_BYTE;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) &ctx->Current.EdgeFlag;
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      cl = &ac->Fallback.Attrib[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 0;
-      cl->Ptr = (GLubyte *) ctx->Current.Attrib[i];
-      cl->Enabled = 1;
-      cl->Flags = CA_CLIENT_DATA; /* hack */
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-}
-
-
-/*
- * Initialize the array cache pointers, types, strides, etc.
- */
-static void _ac_cache_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   struct gl_client_array *cl;
-   GLuint size = ctx->Const.MaxArrayLockSize + MAX_CLIPPED_VERTICES;
-   GLuint i;
-
-   cl = &ac->Cache.Vertex;
-   cl->Size = 4;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 4 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.Normal;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 3 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.Color;
-   cl->Size = 4;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 4 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.SecondaryColor;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 4 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.FogCoord;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.Index;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
-      cl = &ac->Cache.TexCoord[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 4 * sizeof(GLfloat);
-      cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-      cl->Enabled = 1;
-      cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-
-   cl = &ac->Cache.EdgeFlag;
-   cl->Size = 1;
-   cl->Type = GL_UNSIGNED_BYTE;
-   cl->Stride = 0;
-   cl->StrideB = sizeof(GLubyte);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0 ; i < VERT_ATTRIB_MAX; i++) {
-      cl = &ac->Cache.Attrib[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 4 * sizeof(GLfloat);
-      cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-      cl->Enabled = 1;
-      cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-}
-
-
-/* This storage used to hold translated client data if type or stride
- * need to be fixed.
- */
-static void _ac_elts_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   GLuint size = 1000;
-
-   ac->Elts = (GLuint *)MALLOC( sizeof(GLuint) * size );
-   ac->elt_size = size;
-}
-
-static void _ac_raw_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   GLuint i;
-
-   ac->Raw.Color = ac->Fallback.Color;
-   ac->Raw.EdgeFlag = ac->Fallback.EdgeFlag;
-   ac->Raw.FogCoord = ac->Fallback.FogCoord;
-   ac->Raw.Index = ac->Fallback.Index;
-   ac->Raw.Normal = ac->Fallback.Normal;
-   ac->Raw.SecondaryColor = ac->Fallback.SecondaryColor;
-   ac->Raw.Vertex = ctx->Array.ArrayObj->Vertex;
-
-   ac->IsCached.Color = GL_FALSE;
-   ac->IsCached.EdgeFlag = GL_FALSE;
-   ac->IsCached.FogCoord = GL_FALSE;
-   ac->IsCached.Index = GL_FALSE;
-   ac->IsCached.Normal = GL_FALSE;
-   ac->IsCached.SecondaryColor = GL_FALSE;
-   ac->IsCached.Vertex = GL_FALSE;
-
-   for (i = 0 ; i < MAX_TEXTURE_COORD_UNITS ; i++) {
-      ac->Raw.TexCoord[i] = ac->Fallback.TexCoord[i];
-      ac->IsCached.TexCoord[i] = GL_FALSE;
-   }
-
-   for (i = 0 ; i < VERT_ATTRIB_MAX ; i++) {
-      ac->Raw.Attrib[i] = ac->Fallback.Attrib[i];
-      ac->IsCached.Attrib[i] = GL_FALSE;
-   }
-}
-
-GLboolean _ac_CreateContext( GLcontext *ctx )
-{
-   ctx->acache_context = CALLOC(sizeof(ACcontext));
-   if (ctx->acache_context) {
-      _ac_cache_init( ctx );
-      _ac_fallbacks_init( ctx );
-      _ac_raw_init( ctx );
-      _ac_elts_init( ctx );
-      return GL_TRUE;
-   }
-   return GL_FALSE;
-}
-
-void _ac_DestroyContext( GLcontext *ctx )
-{
-   struct gl_buffer_object *nullObj = ctx->Array.NullBufferObj;
-   ACcontext *ac = AC_CONTEXT(ctx);
-   GLint i;
-
-   /* only free vertex data if it's really a pointer to vertex data and
-    * not an offset into a buffer object.
-    */
-   if (ac->Cache.Vertex.Ptr && ac->Cache.Vertex.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Vertex.Ptr );
-   if (ac->Cache.Normal.Ptr && ac->Cache.Normal.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Normal.Ptr );
-   if (ac->Cache.Color.Ptr && ac->Cache.Color.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Color.Ptr );
-   if (ac->Cache.SecondaryColor.Ptr && ac->Cache.SecondaryColor.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.SecondaryColor.Ptr );
-   if (ac->Cache.EdgeFlag.Ptr && ac->Cache.EdgeFlag.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.EdgeFlag.Ptr );
-   if (ac->Cache.Index.Ptr && ac->Cache.Index.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Index.Ptr );
-   if (ac->Cache.FogCoord.Ptr && ac->Cache.FogCoord.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.FogCoord.Ptr );
-
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
-      if (ac->Cache.TexCoord[i].Ptr && ac->Cache.TexCoord[i].BufferObj == nullObj)
-	 FREE( (void *) ac->Cache.TexCoord[i].Ptr );
-   }
-
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      if (ac->Cache.Attrib[i].Ptr && ac->Cache.Attrib[i].BufferObj == nullObj)
-	 FREE( (void *) ac->Cache.Attrib[i].Ptr );
-   }
-
-   if (ac->Elts)
-      FREE( ac->Elts );
-
-   /* Free the context structure itself */
-   FREE(ac);
-   ctx->acache_context = NULL;
-}
-
-void _ac_InvalidateState( GLcontext *ctx, GLuint new_state )
-{
-   AC_CONTEXT(ctx)->NewState |= new_state;
-   AC_CONTEXT(ctx)->NewArrayState |= ctx->Array.NewState;
-}
diff --git a/src/mesa/array_cache/ac_context.h b/src/mesa/array_cache/ac_context.h
deleted file mode 100644
index 36aded3152..0000000000
--- a/src/mesa/array_cache/ac_context.h
+++ /dev/null
@@ -1,99 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * 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>
- */
-
-#ifndef _AC_CONTEXT_H
-#define _AC_CONTEXT_H
-
-#include "glheader.h"
-#include "mtypes.h"
-
-#include "array_cache/acache.h"
-
-/* These are used to make the ctx->Current values look like
- * arrays (with zero StrideB).
- */
-struct ac_arrays {
-   struct gl_client_array Vertex;
-   struct gl_client_array Normal;
-   struct gl_client_array Color;
-   struct gl_client_array SecondaryColor;
-   struct gl_client_array FogCoord;
-   struct gl_client_array Index;
-   struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS];
-   struct gl_client_array EdgeFlag;
-   struct gl_client_array Attrib[VERT_ATTRIB_MAX];  /* GL_NV_vertex_program */
-};
-
-struct ac_array_pointers {
-   struct gl_client_array *Vertex;
-   struct gl_client_array *Normal;
-   struct gl_client_array *Color;
-   struct gl_client_array *SecondaryColor;
-   struct gl_client_array *FogCoord;
-   struct gl_client_array *Index;
-   struct gl_client_array *TexCoord[MAX_TEXTURE_COORD_UNITS];
-   struct gl_client_array *EdgeFlag;
-   struct gl_client_array *Attrib[VERT_ATTRIB_MAX];  /* GL_NV_vertex_program */
-};
-
-struct ac_array_flags {
-   GLboolean Vertex;
-   GLboolean Normal;
-   GLboolean Color;
-   GLboolean SecondaryColor;
-   GLboolean FogCoord;
-   GLboolean Index;
-   GLboolean TexCoord[MAX_TEXTURE_COORD_UNITS];
-   GLboolean EdgeFlag;
-   GLboolean Attrib[VERT_ATTRIB_MAX];  /* GL_NV_vertex_program */
-};
-
-
-typedef struct {
-   GLuint NewState;		/* not needed? */
-   GLuint NewArrayState;
-
-   /* Facility for importing and caching array data:
-    */
-   struct ac_arrays Fallback;
-   struct ac_arrays Cache;
-   struct ac_arrays Raw;
-   struct ac_array_flags IsCached;
-   GLuint start;
-   GLuint count;
-
-   /* Facility for importing element lists:
-    */
-   GLuint *Elts;
-   GLuint elt_size;
-
-} ACcontext;
-
-#define AC_CONTEXT(ctx) ((ACcontext *)ctx->acache_context)
-
-#endif
diff --git a/src/mesa/array_cache/ac_import.c b/src/mesa/array_cache/ac_import.c
deleted file mode 100644
index 119e58bf24..0000000000
--- a/src/mesa/array_cache/ac_import.c
+++ /dev/null
@@ -1,922 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5
- *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "macros.h"
-#include "imports.h"
-#include "mtypes.h"
-
-#include "math/m_translate.h"
-#include "array_cache/ac_context.h"
-#include "math/m_translate.h"
-
-#define STRIDE_ARRAY( array, offset ) 					\
-do {									\
-   GLubyte *tmp = ADD_POINTERS( (array).BufferObj->Data, (array).Ptr )	\
-                + (offset) * (array).StrideB;				\
-   (array).Ptr = tmp;							\
-} while (0)
-
-
-/* Set the array pointer back to its source when the cached data is
- * invalidated:
- */
-static void
-reset_texcoord( GLcontext *ctx, GLuint unit )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->TexCoord[unit].Enabled) {
-      ac->Raw.TexCoord[unit] = ctx->Array.ArrayObj->TexCoord[unit];
-      STRIDE_ARRAY(ac->Raw.TexCoord[unit], ac->start);
-   }
-   else {
-      ac->Raw.TexCoord[unit] = ac->Fallback.TexCoord[unit];
-
-      if (ctx->Current.Attrib[VERT_ATTRIB_TEX0 + unit][3] != 1.0)
-	 ac->Raw.TexCoord[unit].Size = 4;
-      else if (ctx->Current.Attrib[VERT_ATTRIB_TEX0 + unit][2] != 0.0)
-	 ac->Raw.TexCoord[unit].Size = 3;
-      else
-	 ac->Raw.TexCoord[unit].Size = 2;
-   }
-
-   ac->IsCached.TexCoord[unit] = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_TEXCOORD(unit);
-}
-
-static void
-reset_vertex( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   ASSERT(ctx->Array.ArrayObj->Vertex.Enabled
-          || (ctx->VertexProgram._Enabled && ctx->Array.ArrayObj->VertexAttrib[0].Enabled));
-   ac->Raw.Vertex = ctx->Array.ArrayObj->Vertex;
-   STRIDE_ARRAY(ac->Raw.Vertex, ac->start);
-   ac->IsCached.Vertex = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_VERTEX;
-}
-
-
-static void
-reset_normal( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->Normal.Enabled) {
-      ac->Raw.Normal = ctx->Array.ArrayObj->Normal;
-      STRIDE_ARRAY(ac->Raw.Normal, ac->start);
-   }
-   else {
-      ac->Raw.Normal = ac->Fallback.Normal;
-   }
-
-   ac->IsCached.Normal = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_NORMAL;
-}
-
-
-static void
-reset_color( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->Color.Enabled) {
-      ac->Raw.Color = ctx->Array.ArrayObj->Color;
-      STRIDE_ARRAY(ac->Raw.Color, ac->start);
-   }
-   else
-      ac->Raw.Color = ac->Fallback.Color;
-
-   ac->IsCached.Color = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_COLOR0;
-}
-
-
-static void
-reset_secondarycolor( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      ac->Raw.SecondaryColor = ctx->Array.ArrayObj->SecondaryColor;
-      STRIDE_ARRAY(ac->Raw.SecondaryColor, ac->start);
-   }
-   else
-      ac->Raw.SecondaryColor = ac->Fallback.SecondaryColor;
-
-   ac->IsCached.SecondaryColor = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_COLOR1;
-}
-
-
-static void
-reset_index( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->Index.Enabled) {
-      ac->Raw.Index = ctx->Array.ArrayObj->Index;
-      STRIDE_ARRAY(ac->Raw.Index, ac->start);
-   }
-   else
-      ac->Raw.Index = ac->Fallback.Index;
-
-   ac->IsCached.Index = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_INDEX;
-}
-
-
-static void
-reset_fogcoord( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      ac->Raw.FogCoord = ctx->Array.ArrayObj->FogCoord;
-      STRIDE_ARRAY(ac->Raw.FogCoord, ac->start);
-   }
-   else
-      ac->Raw.FogCoord = ac->Fallback.FogCoord;
-
-   ac->IsCached.FogCoord = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_FOGCOORD;
-}
-
-
-static void
-reset_edgeflag( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      ac->Raw.EdgeFlag = ctx->Array.ArrayObj->EdgeFlag;
-      STRIDE_ARRAY(ac->Raw.EdgeFlag, ac->start);
-   }
-   else
-      ac->Raw.EdgeFlag = ac->Fallback.EdgeFlag;
-
-   ac->IsCached.EdgeFlag = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_EDGEFLAG;
-}
-
-
-/**
- * \param index  the generic vertex array number.
- */
-static void
-reset_attrib( GLcontext *ctx, GLuint index )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->VertexAttrib[index].Enabled) {
-      ac->Raw.Attrib[index] = ctx->Array.ArrayObj->VertexAttrib[index];
-      STRIDE_ARRAY(ac->Raw.Attrib[index], ac->start);
-   }
-   else
-      ac->Raw.Attrib[index] = ac->Fallback.Attrib[index];
-
-   ac->IsCached.Attrib[index] = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_ATTRIB(index);
-}
-
-
-/**
- * Generic import function for color data
- */
-static void
-import( const GLcontext *ctx,
-        GLenum destType,
-        struct gl_client_array *to,
-        const struct gl_client_array *from )
-{
-   const ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (destType == 0) 
-      destType = from->Type;
-
-   switch (destType) {
-   case GL_FLOAT:
-      _math_trans_4fn( (GLfloat (*)[4]) to->Ptr,
-                       from->Ptr,
-		       from->StrideB,
-		       from->Type,
-		       from->Size,
-		       0,
-		       ac->count - ac->start);
-
-      to->StrideB = 4 * sizeof(GLfloat);
-      to->Type = GL_FLOAT;
-      break;
-      
-   case GL_UNSIGNED_BYTE:
-      _math_trans_4ub( (GLubyte (*)[4]) to->Ptr,
-                       from->Ptr,
-		       from->StrideB,
-		       from->Type,
-		       from->Size,
-		       0,
-		       ac->count - ac->start);
-
-      to->StrideB = 4 * sizeof(GLubyte);
-      to->Type = GL_UNSIGNED_BYTE;
-      break;
-
-   case GL_UNSIGNED_SHORT:
-      _math_trans_4us( (GLushort (*)[4]) to->Ptr,
-                       from->Ptr,
-		       from->StrideB,
-		       from->Type,
-		       from->Size,
-		       0,
-		       ac->count - ac->start);
-
-      to->StrideB = 4 * sizeof(GLushort);
-      to->Type = GL_UNSIGNED_SHORT;
-      break;
-      
-   default:
-      _mesa_problem(ctx, "Unexpected dest format in import()");
-      break;
-   }
-}
-
-
-
-/*
- * Functions to import array ranges with specified types and strides.
- * For example, if the vertex data is GLshort[2] and we want GLfloat[3]
- * we'll use an import function to do the data conversion.
- */
-
-static void
-import_texcoord( GLcontext *ctx, GLuint unit, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.TexCoord[unit];
-   struct gl_client_array *to = &ac->Cache.TexCoord[unit];
-   (void) type; (void) stride;
-
-   ASSERT(unit < ctx->Const.MaxTextureCoordUnits);
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 4*sizeof(GLfloat) || stride == 0);
-   ASSERT(ac->count - ac->start < ctx->Const.MaxArrayLockSize);
-
-   _math_trans_4f( (GLfloat (*)[4]) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   from->Size,
-		   0,
-		   ac->count - ac->start);
-
-   to->Size = from->Size;
-   to->StrideB = 4 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.TexCoord[unit] = GL_TRUE;
-}
-
-static void
-import_vertex( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Vertex;
-   struct gl_client_array *to = &ac->Cache.Vertex;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 4*sizeof(GLfloat) || stride == 0);
-
-   _math_trans_4f( (GLfloat (*)[4]) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   from->Size,
-		   0,
-		   ac->count - ac->start);
-
-   to->Size = from->Size;
-   to->StrideB = 4 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.Vertex = GL_TRUE;
-}
-
-static void
-import_normal( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Normal;
-   struct gl_client_array *to = &ac->Cache.Normal;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 3*sizeof(GLfloat) || stride == 0);
-
-   _math_trans_3fn((GLfloat (*)[3]) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   0,
-		   ac->count - ac->start);
-
-   to->StrideB = 3 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.Normal = GL_TRUE;
-}
-
-static void
-import_color( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Color;
-   struct gl_client_array *to = &ac->Cache.Color;
-   (void) stride;
-
-   import( ctx, type, to, from );
-   
-   ac->IsCached.Color = GL_TRUE;
-}
-
-static void
-import_index( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Index;
-   struct gl_client_array *to = &ac->Cache.Index;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_UNSIGNED_INT);
-   ASSERT(stride == sizeof(GLuint) || stride == 0);
-
-   _math_trans_1ui( (GLuint *) to->Ptr,
-		    from->Ptr,
-		    from->StrideB,
-		    from->Type,
-		    0,
-		    ac->count - ac->start);
-
-   to->StrideB = sizeof(GLuint);
-   to->Type = GL_UNSIGNED_INT;
-   ac->IsCached.Index = GL_TRUE;
-}
-
-static void
-import_secondarycolor( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.SecondaryColor;
-   struct gl_client_array *to = &ac->Cache.SecondaryColor;
-   (void) stride;
-
-   import( ctx, type, to, from );
-
-   ac->IsCached.SecondaryColor = GL_TRUE;
-}
-
-static void
-import_fogcoord( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.FogCoord;
-   struct gl_client_array *to = &ac->Cache.FogCoord;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == sizeof(GLfloat) || stride == 0);
-
-   _math_trans_1f( (GLfloat *) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   0,
-		   ac->count - ac->start);
-
-   to->StrideB = sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.FogCoord = GL_TRUE;
-}
-
-static void
-import_edgeflag( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.EdgeFlag;
-   struct gl_client_array *to = &ac->Cache.EdgeFlag;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_UNSIGNED_BYTE);
-   ASSERT(stride == sizeof(GLubyte) || stride == 0);
-
-   _math_trans_1ub( (GLubyte *) to->Ptr,
-		    from->Ptr,
-		    from->StrideB,
-		    from->Type,
-		    0,
-		    ac->count - ac->start);
-
-   to->StrideB = sizeof(GLubyte);
-   to->Type = GL_UNSIGNED_BYTE;
-   ac->IsCached.EdgeFlag = GL_TRUE;
-}
-
-/**
- * \param index  the generic vertex array number
- */
-static void
-import_attrib( GLcontext *ctx, GLuint index, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Attrib[index];
-   struct gl_client_array *to = &ac->Cache.Attrib[index];
-   (void) type; (void) stride;
-
-   ASSERT(index < MAX_VERTEX_PROGRAM_ATTRIBS);
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 4*sizeof(GLfloat) || stride == 0);
-   ASSERT(ac->count - ac->start < ctx->Const.MaxArrayLockSize);
-
-   if (from->Normalized) {
-      _math_trans_4fn( (GLfloat (*)[4]) to->Ptr,
-                      from->Ptr,
-                      from->StrideB,
-                      from->Type,
-                      from->Size,
-                      0,
-                      ac->count - ac->start);
-   }
-   else {
-      _math_trans_4f( (GLfloat (*)[4]) to->Ptr,
-                      from->Ptr,
-                      from->StrideB,
-                      from->Type,
-                      from->Size,
-                      0,
-                      ac->count - ac->start);
-   }
-
-   to->Size = from->Size;
-   to->StrideB = 4 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.Attrib[index] = GL_TRUE;
-}
-
-
-
-/*
- * Externals to request arrays with specific properties:
- */
-
-
-struct gl_client_array *
-_ac_import_texcoord( GLcontext *ctx,
-                     GLuint unit,
-                     GLenum type,
-                     GLuint reqstride,
-                     GLuint reqsize,
-                     GLboolean reqwriteable,
-                     GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   ASSERT(unit < MAX_TEXTURE_COORD_UNITS);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_TEXCOORD(unit))
-      reset_texcoord( ctx, unit );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.TexCoord[unit].Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.TexCoord[unit].Type != type ||
-       (reqstride != 0 && ac->Raw.TexCoord[unit].StrideB != (GLint)reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.TexCoord[unit])
-	 import_texcoord(ctx, unit, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.TexCoord[unit];
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.TexCoord[unit];
-   }
-}
-
-struct gl_client_array *
-_ac_import_vertex( GLcontext *ctx,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLuint reqsize,
-                   GLboolean reqwriteable,
-                   GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_VERTEX)
-      reset_vertex( ctx );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.Vertex.Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Vertex.Type != type ||
-       (reqstride != 0 && ac->Raw.Vertex.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Vertex)
-	 import_vertex(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Vertex;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Vertex;
-   }
-}
-
-struct gl_client_array *
-_ac_import_normal( GLcontext *ctx,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLboolean reqwriteable,
-                   GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_NORMAL)
-      reset_normal( ctx );
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Normal.Type != type ||
-       (reqstride != 0 && ac->Raw.Normal.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Normal)
-	 import_normal(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Normal;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Normal;
-   }
-}
-
-struct gl_client_array *
-_ac_import_color( GLcontext *ctx,
-                  GLenum type,
-                  GLuint reqstride,
-                  GLuint reqsize,
-                  GLboolean reqwriteable,
-                  GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_COLOR0)
-      reset_color( ctx );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.Color.Size > (GLint) reqsize) {
-      return NULL;
-   }
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if ((type != 0 && ac->Raw.Color.Type != type) ||
-       (reqstride != 0 && ac->Raw.Color.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Color) {
-      	 import_color(ctx, type, reqstride );
-      }
-      *writeable = GL_TRUE;
-      return &ac->Cache.Color;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Color;
-   }
-}
-
-struct gl_client_array *
-_ac_import_index( GLcontext *ctx,
-                  GLenum type,
-                  GLuint reqstride,
-                  GLboolean reqwriteable,
-                  GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_INDEX)
-      reset_index( ctx );
-
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Index.Type != type ||
-       (reqstride != 0 && ac->Raw.Index.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Index)
-	 import_index(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Index;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Index;
-   }
-}
-
-struct gl_client_array *
-_ac_import_secondarycolor( GLcontext *ctx,
-                           GLenum type,
-                           GLuint reqstride,
-                           GLuint reqsize,
-                           GLboolean reqwriteable,
-                           GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_COLOR1)
-      reset_secondarycolor( ctx );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.SecondaryColor.Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if ((type != 0 && ac->Raw.SecondaryColor.Type != type) ||
-       (reqstride != 0 && ac->Raw.SecondaryColor.StrideB != (GLint)reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.SecondaryColor)
-	 import_secondarycolor(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.SecondaryColor;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.SecondaryColor;
-   }
-}
-
-struct gl_client_array *
-_ac_import_fogcoord( GLcontext *ctx,
-                     GLenum type,
-                     GLuint reqstride,
-                     GLboolean reqwriteable,
-                     GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_FOGCOORD)
-      reset_fogcoord( ctx );
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.FogCoord.Type != type ||
-       (reqstride != 0 && ac->Raw.FogCoord.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.FogCoord)
-	 import_fogcoord(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.FogCoord;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.FogCoord;
-   }
-}
-
-struct gl_client_array *
-_ac_import_edgeflag( GLcontext *ctx,
-                     GLenum type,
-                     GLuint reqstride,
-                     GLboolean reqwriteable,
-                     GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_EDGEFLAG)
-      reset_edgeflag( ctx );
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.EdgeFlag.Type != type ||
-       (reqstride != 0 && ac->Raw.EdgeFlag.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.EdgeFlag)
-	 import_edgeflag(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.EdgeFlag;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.EdgeFlag;
-   }
-}
-
-/**
- * For GL_ARB/NV_vertex_program
- * \param index  index of the vertex array, starting at zero.
- */
-struct gl_client_array *
-_ac_import_attrib( GLcontext *ctx,
-                   GLuint index,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLuint reqsize,
-                   GLboolean reqwriteable,
-                   GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   ASSERT(index < VERT_ATTRIB_MAX);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_ATTRIB(index))
-      reset_attrib( ctx, index );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.Attrib[index].Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Attrib[index].Type != type ||
-       (reqstride != 0 && ac->Raw.Attrib[index].StrideB != (GLint)reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Attrib[index])
-	 import_attrib(ctx, index, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Attrib[index];
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Attrib[index];
-   }
-}
-
-
-/* Clients must call this function to validate state and set bounds
- * before importing any data:
- */
-void
-_ac_import_range( GLcontext *ctx, GLuint start, GLuint count )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (!ctx->Array.LockCount) {
-      /* Not locked, discard cached data.  Changes to lock
-       * status are caught via. _ac_invalidate_state().
-       */
-      ac->NewArrayState = _NEW_ARRAY_ALL;
-      ac->start = start;
-      ac->count = count;
-   }
-   else {
-      /* Locked, discard data for any disabled arrays.  Require that
-       * the whole locked range always be dealt with, otherwise hard to
-       * maintain cached data in the face of clipping.
-       */
-      ac->NewArrayState |= ~ctx->Array.ArrayObj->_Enabled;
-      ac->start = ctx->Array.LockFirst;
-      ac->count = ctx->Array.LockCount;
-      ASSERT(ac->start == start); /* hmm? */
-      ASSERT(ac->count == count);
-   }
-}
-
-
-
-/* Additional convienence function for importing the element list
- * for glDrawElements() and glDrawRangeElements().
- */
-CONST void *
-_ac_import_elements( GLcontext *ctx,
-		     GLenum new_type,
-		     GLuint count,
-		     GLenum old_type,
-		     CONST void *indices )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (old_type == new_type)
-      return indices;
-
-   if (ac->elt_size < count * sizeof(GLuint)) {
-      if (ac->Elts) FREE(ac->Elts);
-      while (ac->elt_size < count * sizeof(GLuint))
-	 ac->elt_size *= 2;
-      ac->Elts = (GLuint *) MALLOC(ac->elt_size);
-   }
-
-   switch (new_type) {
-   case GL_UNSIGNED_BYTE:
-      ASSERT(0);
-      return NULL;
-   case GL_UNSIGNED_SHORT:
-      ASSERT(0);
-      return NULL;
-   case GL_UNSIGNED_INT: {
-      GLuint *out = (GLuint *)ac->Elts;
-      GLuint i;
-
-      switch (old_type) {
-      case GL_UNSIGNED_BYTE: {
-	 CONST GLubyte *in = (CONST GLubyte *)indices;
-	 for (i = 0 ; i < count ; i++)
-	    out[i] = in[i];
-	 break;
-      }
-      case GL_UNSIGNED_SHORT: {
-	 CONST GLushort *in = (CONST GLushort *)indices;
-	 for (i = 0 ; i < count ; i++)
-	       out[i] = in[i];
-	 break;
-      }
-      default:
-	 ASSERT(0);
-      }
-
-      return (CONST void *)out;
-   }
-   default:
-      ASSERT(0);
-      break;
-   }
-
-   return NULL;
-}
diff --git a/src/mesa/array_cache/acache.h b/src/mesa/array_cache/acache.h
deleted file mode 100644
index 2bb64f3656..0000000000
--- a/src/mesa/array_cache/acache.h
+++ /dev/null
@@ -1,132 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  4.1
- *
- * 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>
- */
-
-#ifndef _ARRAYCACHE_H
-#define _ARRAYCACHE_H
-
-#include "mtypes.h"
-
-
-extern GLboolean
-_ac_CreateContext( GLcontext *ctx );
-
-extern void
-_ac_DestroyContext( GLcontext *ctx );
-
-extern void
-_ac_InvalidateState( GLcontext *ctx, GLuint new_state );
-
-extern struct gl_client_array *
-_ac_import_vertex( GLcontext *ctx,
-		   GLenum type,
-		   GLuint reqstride,
-		   GLuint reqsize,
-		   GLboolean reqwritable,
-		   GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_normal( GLcontext *ctx,
-		   GLenum type,
-		   GLuint reqstride,
-		   GLboolean reqwritable,
-		   GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_color( GLcontext *ctx,
-		  GLenum type,
-		  GLuint reqstride,
-		  GLuint reqsize,
-		  GLboolean reqwritable,
-		  GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_index( GLcontext *ctx,
-		  GLenum type,
-		  GLuint reqstride,
-		  GLboolean reqwritable,
-		  GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_secondarycolor( GLcontext *ctx,
-			   GLenum type,
-			   GLuint reqstride,
-			   GLuint reqsize,
-			   GLboolean reqwritable,
-			   GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_fogcoord( GLcontext *ctx,
-		     GLenum type,
-		     GLuint reqstride,
-		     GLboolean reqwritable,
-		     GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_edgeflag( GLcontext *ctx,
-		     GLenum type,
-		     GLuint reqstride,
-		     GLboolean reqwritable,
-		     GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_texcoord( GLcontext *ctx,
-		     GLuint unit,
-		     GLenum type,
-		     GLuint reqstride,
-		     GLuint reqsize,
-		     GLboolean reqwritable,
-		     GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_attrib( GLcontext *ctx,
-                   GLuint index,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLuint reqsize,
-                   GLboolean reqwritable,
-                   GLboolean *writable );
-
-
-/* Clients must call this function to validate state and set bounds
- * before importing any data:
- */
-extern void
-_ac_import_range( GLcontext *ctx, GLuint start, GLuint count );
-
-
-/* Additional convenience function:
- */
-extern CONST void *
-_ac_import_elements( GLcontext *ctx,
-		     GLenum new_type,
-		     GLuint count,
-		     GLenum old_type,
-		     CONST void *indices );
-
-
-#endif
diff --git a/src/mesa/array_cache/descrip.mms b/src/mesa/array_cache/descrip.mms
deleted file mode 100644
index 0607a01d1f..0000000000
--- a/src/mesa/array_cache/descrip.mms
+++ /dev/null
@@ -1,37 +0,0 @@
-# Makefile for core library for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-# Last revision : 16 June 2003
-
-.first
-	define gl [---.include.gl]
-	define math [-.math]
-	define array_cache [-.array_cache]
-
-.include [---]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = [---.include],[-.main],[-.glapi]
-LIBDIR = [---.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = ac_context.c ac_import.c
-
-OBJECTS = ac_context.obj,ac_import.obj
-##### RULES #####
-
-VERSION=Mesa V3.4
-
-##### TARGETS #####
-# Make the library
-$(LIBDIR)$(GL_LIB) : $(OBJECTS)
-  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
-
-clean :
-	purge
-	delete *.obj;*
-
-ac_context.obj : ac_context.c
-ac_import.obj : ac_import.c
diff --git a/src/mesa/array_cache/sources b/src/mesa/array_cache/sources
deleted file mode 100644
index fb3328d10e..0000000000
--- a/src/mesa/array_cache/sources
+++ /dev/null
@@ -1,7 +0,0 @@
-MESA_ARRAY_CACHE_SOURCES = \
-ac_context.c \
-ac_import.c
-
-MESA_ARRAY_CACHE_HEADERS = \
-ac_context.h \
-acache.h
diff --git a/src/mesa/drivers/directfb/idirectfbgl_mesa.c b/src/mesa/drivers/directfb/idirectfbgl_mesa.c
index ebcbfbcc65..9e76e0f6e0 100644
--- a/src/mesa/drivers/directfb/idirectfbgl_mesa.c
+++ b/src/mesa/drivers/directfb/idirectfbgl_mesa.c
@@ -46,7 +46,7 @@
 #include "texformat.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
diff --git a/src/mesa/drivers/dri/fb/fb_dri.c b/src/mesa/drivers/dri/fb/fb_dri.c
index d6ba23bc6e..08b52b4d8f 100644
--- a/src/mesa/drivers/dri/fb/fb_dri.c
+++ b/src/mesa/drivers/dri/fb/fb_dri.c
@@ -50,7 +50,7 @@
 #include "extensions.h"
 #include "framebuffer.h"
 #include "renderbuffer.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -93,7 +93,7 @@ update_state( GLcontext *ctx, GLuint new_state )
    /* not much to do here - pass it on */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
 }
 
@@ -365,7 +365,7 @@ fbCreateContext( const __GLcontextModes *glVisual,
 
    /* Create module contexts */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _swsetup_Wakeup( ctx );
@@ -399,7 +399,7 @@ fbDestroyContext( __DRIcontextPrivate *driContextPriv )
    if ( fbmesa ) {
       _swsetup_DestroyContext( fbmesa->glCtx );
       _tnl_DestroyContext( fbmesa->glCtx );
-      _ac_DestroyContext( fbmesa->glCtx );
+      _vbo_DestroyContext( fbmesa->glCtx );
       _swrast_DestroyContext( fbmesa->glCtx );
 
       /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/fb/fb_egl.c b/src/mesa/drivers/dri/fb/fb_egl.c
index 5c74b0b0f4..517e71f888 100644
--- a/src/mesa/drivers/dri/fb/fb_egl.c
+++ b/src/mesa/drivers/dri/fb/fb_egl.c
@@ -17,7 +17,7 @@
 #include "extensions.h"
 #include "framebuffer.h"
 #include "renderbuffer.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -388,7 +388,7 @@ update_state( GLcontext *ctx, GLuint new_state )
    /* not much to do here - pass it on */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
 }
 
@@ -491,7 +491,7 @@ fbCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext sh
 
    /* Create module contexts */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _swsetup_Wakeup( ctx );
diff --git a/src/mesa/drivers/dri/ffb/ffb_state.c b/src/mesa/drivers/dri/ffb/ffb_state.c
index b81d94de25..eb13478166 100644
--- a/src/mesa/drivers/dri/ffb/ffb_state.c
+++ b/src/mesa/drivers/dri/ffb/ffb_state.c
@@ -40,7 +40,7 @@
 #include "enums.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1034,7 +1034,7 @@ static void ffbDDUpdateState(GLcontext *ctx, GLuint newstate)
 
 	_swrast_InvalidateState( ctx, newstate );
 	_swsetup_InvalidateState( ctx, newstate );
-	_ac_InvalidateState( ctx, newstate );
+	_vbo_InvalidateState( ctx, newstate );
 	_tnl_InvalidateState( ctx, newstate );
 
 	if (newstate & _NEW_TEXTURE)
diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
index 99256d76b6..215aaf8ffb 100644
--- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c
+++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
@@ -38,7 +38,7 @@
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "drivers/common/driverfuncs.h"
 
 #include "ffb_context.h"
@@ -277,7 +277,7 @@ ffbCreateContext(const __GLcontextModes *mesaVis,
 
 	/* Initialize the software rasterizer and helper modules. */
 	_swrast_CreateContext( ctx );
-	_ac_CreateContext( ctx );
+	_vbo_CreateContext( ctx );
 	_tnl_CreateContext( ctx );
 	_swsetup_CreateContext( ctx );
 
@@ -313,7 +313,7 @@ ffbDestroyContext(__DRIcontextPrivate *driContextPriv)
 
 		_swsetup_DestroyContext( fmesa->glCtx );
 		_tnl_DestroyContext( fmesa->glCtx );
-		_ac_DestroyContext( fmesa->glCtx );
+		_vbo_DestroyContext( fmesa->glCtx );
 		_swrast_DestroyContext( fmesa->glCtx );
 
                 /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/gamma/gamma_context.c b/src/mesa/drivers/dri/gamma/gamma_context.c
index ffaf45459b..b1dcbfcdcf 100644
--- a/src/mesa/drivers/dri/gamma/gamma_context.c
+++ b/src/mesa/drivers/dri/gamma/gamma_context.c
@@ -28,7 +28,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -147,7 +147,7 @@ GLboolean gammaCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
diff --git a/src/mesa/drivers/dri/gamma/gamma_state.c b/src/mesa/drivers/dri/gamma/gamma_state.c
index 08749bf5d3..8dbe0a97ca 100644
--- a/src/mesa/drivers/dri/gamma/gamma_state.c
+++ b/src/mesa/drivers/dri/gamma/gamma_state.c
@@ -33,7 +33,7 @@
 #include "colormac.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 
 #define ENABLELIGHTING 0
@@ -1663,7 +1663,7 @@ static void gammaDDUpdateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    GAMMA_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/gamma/gamma_xmesa.c b/src/mesa/drivers/dri/gamma/gamma_xmesa.c
index 00f6aa32ae..e8922b1503 100644
--- a/src/mesa/drivers/dri/gamma/gamma_xmesa.c
+++ b/src/mesa/drivers/dri/gamma/gamma_xmesa.c
@@ -34,7 +34,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 static GLboolean 
 gammaInitDriver(__DRIscreenPrivate *sPriv)
@@ -57,7 +57,7 @@ gammaDestroyContext(__DRIcontextPrivate *driContextPriv)
     if (gmesa) {
       _swsetup_DestroyContext( gmesa->glCtx );
       _tnl_DestroyContext( gmesa->glCtx );
-      _ac_DestroyContext( gmesa->glCtx );
+      _vbo_DestroyContext( gmesa->glCtx );
       _swrast_DestroyContext( gmesa->glCtx );
 
       gammaFreeVB( gmesa->glCtx );
diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c
index c0f4efd7a3..db8f7a19a2 100644
--- a/src/mesa/drivers/dri/i810/i810context.c
+++ b/src/mesa/drivers/dri/i810/i810context.c
@@ -44,7 +44,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -287,7 +287,7 @@ i810CreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -350,7 +350,7 @@ i810DestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (imesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( imesa->glCtx );
       _tnl_DestroyContext( imesa->glCtx );
-      _ac_DestroyContext( imesa->glCtx );
+      _vbo_DestroyContext( imesa->glCtx );
       _swrast_DestroyContext( imesa->glCtx );
 
       i810FreeVB( imesa->glCtx );
diff --git a/src/mesa/drivers/dri/i810/i810state.c b/src/mesa/drivers/dri/i810/i810state.c
index 6f9d4b5bd4..3ad25282d9 100644
--- a/src/mesa/drivers/dri/i810/i810state.c
+++ b/src/mesa/drivers/dri/i810/i810state.c
@@ -21,8 +21,8 @@
 #include "i810ioctl.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
 #include "tnl/tnl.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 
 #include "tnl/t_pipeline.h"
@@ -953,7 +953,7 @@ static void i810InvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    I810_CONTEXT(ctx)->new_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index 2f78fd60b2..6ec34e5bde 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -36,7 +36,8 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
+
 
 #include "utils.h"
 #include "i915_reg.h"
@@ -63,7 +64,7 @@ static void i915InvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
    INTEL_CONTEXT(ctx)->NewGLState |= new_state;
diff --git a/src/mesa/drivers/dri/i915/intel_context.c b/src/mesa/drivers/dri/i915/intel_context.c
index 0a2e33ffd5..7bb00d9833 100644
--- a/src/mesa/drivers/dri/i915/intel_context.c
+++ b/src/mesa/drivers/dri/i915/intel_context.c
@@ -37,7 +37,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
@@ -228,7 +228,7 @@ static void intelInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
    INTEL_CONTEXT(ctx)->NewGLState |= new_state;
@@ -304,7 +304,7 @@ GLboolean intelInitContext( intelContextPtr intel,
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -423,7 +423,7 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
       _swsetup_DestroyContext (&intel->ctx);
       _tnl_DestroyContext (&intel->ctx);
-      _ac_DestroyContext (&intel->ctx);
+      _vbo_DestroyContext (&intel->ctx);
 
       _swrast_DestroyContext (&intel->ctx);
       intel->Fallback = 0;	/* don't call _swrast_Flush later */
diff --git a/src/mesa/drivers/dri/i915tex/i915_context.c b/src/mesa/drivers/dri/i915tex/i915_context.c
index 4cbe29d79d..9b4d72eab3 100644
--- a/src/mesa/drivers/dri/i915tex/i915_context.c
+++ b/src/mesa/drivers/dri/i915tex/i915_context.c
@@ -36,7 +36,6 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
 
 #include "utils.h"
 #include "i915_reg.h"
@@ -67,7 +66,7 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state)
 {
    _swrast_InvalidateState(ctx, new_state);
    _swsetup_InvalidateState(ctx, new_state);
-   _ac_InvalidateState(ctx, new_state);
+   _vbo_InvalidateState(ctx, new_state);
    _tnl_InvalidateState(ctx, new_state);
    _tnl_invalidate_vertex_state(ctx, new_state);
    intel_context(ctx)->NewGLState |= new_state;
diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c
index c7ab621919..aa76875a4a 100644
--- a/src/mesa/drivers/dri/i915tex/intel_context.c
+++ b/src/mesa/drivers/dri/i915tex/intel_context.c
@@ -37,7 +37,6 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
 
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
@@ -241,7 +240,7 @@ intelInvalidateState(GLcontext * ctx, GLuint new_state)
 {
    _swrast_InvalidateState(ctx, new_state);
    _swsetup_InvalidateState(ctx, new_state);
-   _ac_InvalidateState(ctx, new_state);
+   _vbo_InvalidateState(ctx, new_state);
    _tnl_InvalidateState(ctx, new_state);
    _tnl_invalidate_vertex_state(ctx, new_state);
    intel_context(ctx)->NewGLState |= new_state;
@@ -394,7 +393,7 @@ intelInitContext(struct intel_context *intel,
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext(ctx);
-   _ac_CreateContext(ctx);
+   _vbo_CreateContext(ctx);
    _tnl_CreateContext(ctx);
    _swsetup_CreateContext(ctx);
 
@@ -504,7 +503,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
       _swsetup_DestroyContext(&intel->ctx);
       _tnl_DestroyContext(&intel->ctx);
-      _ac_DestroyContext(&intel->ctx);
+      _vbo_DestroyContext(&intel->ctx);
 
       _swrast_DestroyContext(&intel->ctx);
       intel->Fallback = 0;      /* don't call _swrast_Flush later */
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index f0a6fa5740..9e4ff112dc 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -35,17 +35,11 @@ DRIVER_SOURCES = \
 	brw_context.c \
 	brw_curbe.c \
 	brw_draw.c \
-	brw_draw_current.c \
 	brw_draw_upload.c \
 	brw_eu.c \
 	brw_eu_debug.c \
 	brw_eu_emit.c \
 	brw_eu_util.c \
-	brw_exec.c \
-	brw_exec_api.c \
-	brw_exec_array.c \
-	brw_exec_draw.c \
-	brw_exec_eval.c \
 	brw_fallback.c \
 	brw_gs.c \
 	brw_gs_emit.c \
@@ -54,9 +48,6 @@ DRIVER_SOURCES = \
 	brw_metaops.c \
 	brw_misc_state.c \
 	brw_program.c \
-	brw_save.c \
-	brw_save_api.c \
-	brw_save_draw.c \
 	brw_sf.c \
 	brw_sf_emit.c \
 	brw_sf_state.c \
diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h
index c3967c8c61..49b2770a51 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.h
+++ b/src/mesa/drivers/dri/i965/brw_clip.h
@@ -116,7 +116,7 @@ struct brw_clip_compile {
    GLuint last_mrf;
 
    GLuint header_position_offset;
-   GLuint offset[BRW_ATTRIB_MAX];
+   GLuint offset[VERT_ATTRIB_MAX];
 };
 
 #define ATTR_SIZE  (4*4)
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index bc422c1a50..6faee65542 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -34,8 +34,6 @@
 #include "brw_aub.h"
 #include "brw_defines.h"
 #include "brw_draw.h"
-#include "brw_exec.h"
-#include "brw_save.h"
 #include "brw_vs.h"
 #include "imports.h"
 #include "intel_tex.h"
@@ -158,12 +156,6 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
 
    brw_FrameBufferTexInit( brw );
 
-   /* Hook our functions into exec and compile dispatch tables.  Only
-    * fallback on out-of-memory situations.
-    */
-   brw_exec_init( ctx );
-   brw_save_init( ctx );
-
    {
       const char *filename = getenv("INTEL_REPLAY");
       if (filename) {
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 9ee81b8725..08fdc54520 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -36,7 +36,6 @@
 #include "intel_context.h"
 #include "brw_structs.h"
 #include "imports.h"
-#include "brw_attrib.h"
 
 
 /* Glossary:
@@ -215,7 +214,7 @@ struct brw_vs_prog_data {
    GLuint total_grf;
    GLuint outputs_written;
 
-   GLuint64EXT inputs_read;
+   GLuint inputs_read;
 
    /* Used for calculating urb partitions:
     */
@@ -382,10 +381,10 @@ struct brw_cached_batch_item {
    
 
 
-/* Protect against a future where BRW_ATTRIB_MAX > 32.  Wouldn't life
+/* Protect against a future where VERT_ATTRIB_MAX > 32.  Wouldn't life
  * be easier if C allowed arrays of packed elements?
  */
-#define ATTRIB_BIT_DWORDS  ((BRW_ATTRIB_MAX+31)/32)
+#define ATTRIB_BIT_DWORDS  ((VERT_ATTRIB_MAX+31)/32)
 
 struct brw_vertex_element {
    const struct gl_client_array *glarray;
@@ -401,8 +400,8 @@ struct brw_vertex_element {
 
 
 struct brw_vertex_info {
-   GLuint64EXT varying;  /* varying:1[BRW_ATTRIB_MAX] */
-   GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[BRW_ATTRIB_MAX] */
+   GLuint varying;  /* varying:1[VERT_ATTRIB_MAX] */
+   GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */
 };
 
 
@@ -449,14 +448,13 @@ struct brw_context
    struct brw_cached_batch_item *cached_batch_items;
 
    struct {
-      /* Fallback values for inputs not supplied: 
-       */
-      struct gl_client_array current_values[BRW_ATTRIB_MAX];
 
       /* Arrays with buffer objects to copy non-bufferobj arrays into
        * for upload:
        */
-      struct gl_client_array vbo_array[BRW_ATTRIB_MAX];
+      struct gl_client_array vbo_array[VERT_ATTRIB_MAX];
+
+      struct brw_vertex_element inputs[VERT_ATTRIB_MAX];
 
 #define BRW_NR_UPLOAD_BUFS 17
 #define BRW_UPLOAD_INIT_SIZE (128*1024)
@@ -469,11 +467,6 @@ struct brw_context
 	 GLuint wrap;
       } upload;
 
-      /* Currenly bound arrays, including fallbacks to current_values
-       * above:
-       */
-      struct brw_vertex_element inputs[BRW_ATTRIB_MAX];
-
       /* Summary of size and varying of active arrays, so we can check
        * for changes to this state:
        */
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 471fda9f7e..092c6bafc2 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -35,7 +35,6 @@
 
 #include "brw_draw.h"
 #include "brw_defines.h"
-#include "brw_attrib.h"
 #include "brw_context.h"
 #include "brw_aub.h"
 #include "brw_state.h"
@@ -45,7 +44,8 @@
 #include "intel_batchbuffer.h"
 #include "intel_buffer_objects.h"
 
-
+#include "tnl/tnl.h"
+#include "vbo/vbo_context.h"
 
 
 
@@ -143,7 +143,7 @@ static void brw_emit_cliprect( struct brw_context *brw,
 
 
 static void brw_emit_prim( struct brw_context *brw, 
-			   const struct brw_draw_prim *prim )
+			   const struct _mesa_prim *prim )
 
 {
    struct brw_3d_primitive prim_packet;
@@ -170,34 +170,9 @@ static void brw_emit_prim( struct brw_context *brw,
    }
 }
 
-
-
-static void update_current_size( struct gl_client_array *array)
-{
-   const GLfloat *ptr = (const GLfloat *)array->Ptr;
-
-   assert(array->StrideB == 0);
-   assert(array->Type == GL_FLOAT || array->Type == GL_UNSIGNED_BYTE);
-
-   if (ptr[3] != 1.0) 
-      array->Size = 4;
-   else if (ptr[2] != 0.0) 
-      array->Size = 3;
-   else if (ptr[1] != 0.0) 
-      array->Size = 2;
-   else
-      array->Size = 1;
-}
-
-
-
-/* Fill in any gaps in passed arrays with pointers to current
- * attributes:
- */
 static void brw_merge_inputs( struct brw_context *brw,
 		       const struct gl_client_array *arrays[])
 {
-   struct gl_client_array *current_values = brw->vb.current_values;
    struct brw_vertex_element *inputs = brw->vb.inputs;
    struct brw_vertex_info old = brw->vb.info;
    GLuint i;
@@ -205,19 +180,16 @@ static void brw_merge_inputs( struct brw_context *brw,
    memset(inputs, 0, sizeof(*inputs));
    memset(&brw->vb.info, 0, sizeof(brw->vb.info));
 
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) {
-      if (arrays[i] && arrays[i]->Enabled)
-      {
-	 brw->vb.inputs[i].glarray = arrays[i];
-	 brw->vb.info.varying |= (GLuint64EXT) 1 << i;
-      }
-      else 
-      {
-	 brw->vb.inputs[i].glarray = &current_values[i];
-	 update_current_size(&current_values[i]);
-      }
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      brw->vb.inputs[i].glarray = arrays[i];
+
+      /* XXX: metaops passes null arrays */
+      if (arrays[i]) {
+	 if (arrays[i]->StrideB != 0)
+	    brw->vb.info.varying |= 1 << i;
 
-      brw->vb.info.sizes[i/16] |= (inputs[i].glarray->Size - 1) << ((i%16) * 2);
+	 brw->vb.info.sizes[i/16] |= (inputs[i].glarray->Size - 1) << ((i%16) * 2);
+      }
    }
 
    /* Raise statechanges if input sizes and varying have changed: 
@@ -229,8 +201,11 @@ static void brw_merge_inputs( struct brw_context *brw,
       brw->state.dirty.brw |= BRW_NEW_INPUT_VARYING;
 }
 
+/* XXX: could split the primitive list to fallback only on the
+ * non-conformant primitives.
+ */
 static GLboolean check_fallbacks( struct brw_context *brw,
-				  const struct brw_draw_prim *prim,
+				  const struct _mesa_prim *prim,
 				  GLuint nr_prims )
 {
    GLuint i;
@@ -281,15 +256,16 @@ static GLboolean check_fallbacks( struct brw_context *brw,
    return GL_FALSE;
 }
 
-
+/* May fail if out of video memory for texture or vbo upload, or on
+ * fallback conditions.
+ */
 static GLboolean brw_try_draw_prims( GLcontext *ctx,
 				     const struct gl_client_array *arrays[],
-				     const struct brw_draw_prim *prim,
+				     const struct _mesa_prim *prim,
 				     GLuint nr_prims,
-				     const struct brw_draw_index_buffer *ib,
+				     const struct _mesa_index_buffer *ib,
 				     GLuint min_index,
-				     GLuint max_index,
-				     GLuint flags )
+				     GLuint max_index )
 {
    struct intel_context *intel = intel_context(ctx);
    struct brw_context *brw = brw_context(ctx);
@@ -298,11 +274,11 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
 
    if (ctx->NewState)
       _mesa_update_state( ctx );
-      
+
    /* Bind all inputs, derive varying and size information:
     */
    brw_merge_inputs( brw, arrays );
-
+      
    /* Have to validate state quite late.  Will rebuild tnl_program,
     * which depends on varying information.  
     * 
@@ -319,10 +295,6 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    }
 
    {
-      assert(intel->locked);
-      
-
-
       /* Set the first primitive early, ahead of validate_state:
        */
       brw_set_prim(brw, prim[0].mode);
@@ -411,44 +383,88 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    return retval;
 }
 
+static GLboolean brw_need_rebase( GLcontext *ctx,
+				  const struct gl_client_array *arrays[],
+				  const struct _mesa_index_buffer *ib,
+				  GLuint min_index )
+{
+   if (min_index == 0) 
+      return GL_FALSE;
 
-GLboolean brw_draw_prims( GLcontext *ctx,
-			  const struct gl_client_array *arrays[],
-			  const struct brw_draw_prim *prim,
-			  GLuint nr_prims,
-			  const struct brw_draw_index_buffer *ib,
-			  GLuint min_index,
-			  GLuint max_index,
-			  GLuint flags )
+   if (ib) {
+      if (!vbo_all_varyings_in_vbos(arrays))
+	 return GL_TRUE;
+      else
+	 return GL_FALSE;
+   }
+   else {
+      /* Hmm.  This isn't quite what I wanted.  BRW can actually
+       * handle the mixed case well enough that we shouldn't need to
+       * rebase.  However, it's probably not very common, nor hugely
+       * expensive to do it this way:
+       */
+      if (!vbo_all_varyings_in_vbos(arrays))
+	 return GL_TRUE;
+      else
+	 return GL_FALSE;
+   }
+}
+				  
+
+void brw_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLuint min_index,
+		     GLuint max_index )
 {
    struct intel_context *intel = intel_context(ctx);
    GLboolean retval;
 
-   retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index, flags);
+   /* Decide if we want to rebase.  If so we end up recursing once
+    * only into this function.
+    */
+   if (brw_need_rebase( ctx, arrays, ib, min_index )) {
+      vbo_rebase_prims( ctx, arrays, 
+			prim, nr_prims, 
+			ib, min_index, max_index, 
+			brw_draw_prims );
+      
+      return;
+   }
 
-   
-   if (!retval && bmError(intel)) {
 
-      DBG("retrying\n");
-      /* This looks like out-of-memory but potentially we have
-       * situation where there is enough memory but it has become
-       * fragmented.  Clear out all heaps and start from scratch by
-       * faking a contended lock event:  (done elsewhere)
-       */
+   /* Make a first attempt at drawing:
+    */
+   retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
 
+   
+   /* This looks like out-of-memory but potentially we have
+    * situation where there is enough memory but it has become
+    * fragmented.  Clear out all heaps and start from scratch by
+    * faking a contended lock event:  (done elsewhere)
+    */
+   if (!retval && !intel->Fallback && bmError(intel)) {
+      DBG("retrying\n");
       /* Then try a second time only to upload textures and draw the
        * primitives:
        */
-      retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index, flags);
+      retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+   }
+
+   /* Otherwise, we really are out of memory.  Pass the drawing
+    * command to the software tnl module and which will in turn call
+    * swrast to do the drawing.
+    */
+   if (!retval) {
+      _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
    }
 
    if (intel->aub_file && (INTEL_DEBUG & DEBUG_SYNC)) {
       intelFinish( &intel->ctx );
       intel->aub_wrap = 1;
    }
-
-
-   return retval;
 }
 
 
@@ -461,21 +477,25 @@ static void brw_invalidate_vbo_cb( struct intel_context *intel, void *ptr )
 void brw_draw_init( struct brw_context *brw )
 {
    GLcontext *ctx = &brw->intel.ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLuint i;
    
+   /* Register our drawing function: 
+    */
+   vbo->draw_prims = brw_draw_prims;
+
    brw->vb.upload.size = BRW_UPLOAD_INIT_SIZE;
 
    for (i = 0; i < BRW_NR_UPLOAD_BUFS; i++) {
       brw->vb.upload.vbo[i] = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB);
       
-      /* XXX: Set these to no-backing-store
+      /* NOTE:  These are set to no-backing-store.
        */
       bmBufferSetInvalidateCB(&brw->intel,
 			      intel_bufferobj_buffer(intel_buffer_object(brw->vb.upload.vbo[i])),
 			      brw_invalidate_vbo_cb,
 			      &brw->intel,
 			      GL_TRUE);
-
    }
 
    ctx->Driver.BufferData( ctx, 
@@ -484,9 +504,6 @@ void brw_draw_init( struct brw_context *brw )
 			   NULL,
 			   GL_DYNAMIC_DRAW_ARB,
 			   brw->vb.upload.vbo[0] );
-						     
-   
-   brw_init_current_values(ctx, brw->vb.current_values);
 }
 
 void brw_draw_destroy( struct brw_context *brw )
diff --git a/src/mesa/drivers/dri/i965/brw_draw.h b/src/mesa/drivers/dri/i965/brw_draw.h
index 92640bf725..0f7b738310 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.h
+++ b/src/mesa/drivers/dri/i965/brw_draw.h
@@ -29,44 +29,18 @@
 #define BRW_DRAW_H
 
 #include "mtypes.h"		/* for GLcontext... */
-#include "brw_attrib.h"
+#include "vbo/vbo.h"
 
 struct brw_context;
 
-struct brw_draw_prim {
-   GLuint mode:8;
-   GLuint indexed:1;
-   GLuint begin:1;
-   GLuint end:1;
-   GLuint weak:1;
-   GLuint pad:20;
 
-   GLuint start;
-   GLuint count;
-};
-
-struct brw_draw_index_buffer {
-   GLuint count;
-   GLenum type;
-   struct gl_buffer_object *obj;
-   const void *ptr;
-   GLuint rebase;
-};
-
-
-#define BRW_DRAW_SORTED           0x1
-#define BRW_DRAW_ALL_INTERLEAVED  0x2
-#define BRW_DRAW_NON_INTERLEAVED  0x4
-#define BRW_DRAW_LOCKED           0x8
-
-GLboolean brw_draw_prims( GLcontext *ctx,
-			  const struct gl_client_array *arrays[],
-			  const struct brw_draw_prim *prims,
-			  GLuint nr_prims,
-			  const struct brw_draw_index_buffer *ib,
-			  GLuint min_index,
-			  GLuint max_index,
-			  GLuint flags );
+void brw_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prims,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLuint min_index,
+		     GLuint max_index );
 
 void brw_draw_init( struct brw_context *brw );
 void brw_draw_destroy( struct brw_context *brw );
@@ -80,25 +54,12 @@ void brw_init_current_values(GLcontext *ctx,
 /* brw_draw_upload.c
  */
 void brw_upload_indices( struct brw_context *brw,
-			 const struct brw_draw_index_buffer *index_buffer);
+			 const struct _mesa_index_buffer *index_buffer);
 
 GLboolean brw_upload_vertices( struct brw_context *brw,
 			       GLuint min_index,
 			       GLuint max_index );
 
 
-/* Helpers for save, exec.  Should probably have their own file:
- */
-struct brw_exec_context;
-struct brw_save_context;
-
-struct brw_exec_save {
-   struct brw_exec_context *exec;
-   struct brw_save_context *save;
-};
-
-/* Doesn't really belong here:
- */
-#define IMM_CONTEXT(ctx) ((struct brw_exec_save *)((ctx)->swtnl_im))
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_draw_current.c b/src/mesa/drivers/dri/i965/brw_draw_current.c
deleted file mode 100644
index 98d930738e..0000000000
--- a/src/mesa/drivers/dri/i965/brw_draw_current.c
+++ /dev/null
@@ -1,103 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include <stdlib.h>
-
-#include "glheader.h"
-#include "context.h"
-#include "state.h"
-#include "api_validate.h"
-#include "enums.h"
-
-#include "brw_context.h"
-#include "brw_draw.h"
-
-#include "bufmgr.h"
-#include "intel_buffer_objects.h"
-
-
-void brw_init_current_values(GLcontext *ctx,
-			     struct gl_client_array *arrays)
-{
-   GLuint i;
-
-   memset(arrays, 0, sizeof(*arrays) * BRW_ATTRIB_MAX);
-
-   /* Set up a constant (StrideB == 0) array for each current
-    * attribute:
-    */
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) {
-      struct gl_client_array *cl = &arrays[i];
-
-      switch (i) {
-      case BRW_ATTRIB_MAT_FRONT_SHININESS:
-      case BRW_ATTRIB_MAT_BACK_SHININESS:
-      case BRW_ATTRIB_INDEX:
-      case BRW_ATTRIB_EDGEFLAG:
-	 cl->Size = 1;
-	 break;
-      case BRW_ATTRIB_MAT_FRONT_INDEXES:
-      case BRW_ATTRIB_MAT_BACK_INDEXES:
-	 cl->Size = 3;
-	 break;
-      default:
-	 /* This is fixed for the material attributes, for others will
-	  * be determined at runtime:
-	  */
-	 if (i >= BRW_ATTRIB_MAT_FRONT_AMBIENT)
-	    cl->Size = 4;
-	 else
-	    cl->Size = 1;
-	 break;
-      }
-
-      switch (i) {
-      case BRW_ATTRIB_EDGEFLAG:
-	 cl->Type = GL_UNSIGNED_BYTE;
-	 cl->Ptr = (const void *)&ctx->Current.EdgeFlag;
-	 break;
-      case BRW_ATTRIB_INDEX:
-	 cl->Type = GL_FLOAT;
-	 cl->Ptr = (const void *)&ctx->Current.Index;
-	 break;
-      default:
-	 cl->Type = GL_FLOAT;
-	 if (i < BRW_ATTRIB_FIRST_MATERIAL)
-	    cl->Ptr = (const void *)ctx->Current.Attrib[i];
-	 else 
-	    cl->Ptr = (const void *)ctx->Light.Material.Attrib[i - BRW_ATTRIB_FIRST_MATERIAL];
-	 break;
-      }
-
-      cl->Stride = 0;
-      cl->StrideB = 0;
-      cl->Enabled = 1;
-      cl->Flags = 0;
-      cl->BufferObj = ctx->Array.NullBufferObj;
-   }
-}
-
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index 08741ad90f..6150cac4aa 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -35,7 +35,6 @@
 
 #include "brw_draw.h"
 #include "brw_defines.h"
-#include "brw_attrib.h"
 #include "brw_context.h"
 #include "brw_aub.h"
 #include "brw_state.h"
@@ -310,7 +309,6 @@ copy_array_to_vbo_array( struct brw_context *brw,
 			 GLuint i,
 			 const struct gl_client_array *array,
 			 GLuint element_size,
-			 GLuint min_index,
 			 GLuint count)
 {
    GLcontext *ctx = &brw->intel.ctx;
@@ -337,7 +335,6 @@ copy_array_to_vbo_array( struct brw_context *brw,
    vbo_array->Enabled = 1;
    vbo_array->Normalized = array->Normalized;
    vbo_array->_MaxElement = array->_MaxElement;	/* ? */
-   vbo_array->Flags = array->Flags; /* ? */
    vbo_array->BufferObj = vbo;
 
    {
@@ -349,7 +346,7 @@ copy_array_to_vbo_array( struct brw_context *brw,
       map += offset;
 
       copy_strided_array( map, 
-			  array->Ptr + min_index * array->StrideB,
+			  array->Ptr,
 			  element_size,
 			  array->StrideB,
 			  count);
@@ -380,7 +377,6 @@ interleaved_vbo_array( struct brw_context *brw,
    vbo_array->Enabled = 1;
    vbo_array->Normalized = array->Normalized;
    vbo_array->_MaxElement = array->_MaxElement;	
-   vbo_array->Flags = array->Flags; /* ? */
    vbo_array->BufferObj = uploaded_array->BufferObj;
 
    return vbo_array;
@@ -393,17 +389,17 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 {
    GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = intel_context(ctx);
-   GLuint64EXT tmp = brw->vs.prog_data->inputs_read; 
+   GLuint tmp = brw->vs.prog_data->inputs_read; 
    struct brw_vertex_element_packet vep;
    struct brw_array_state vbp;
    GLuint i;
    const void *ptr = NULL;
    GLuint interleave = 0;
 
-   struct brw_vertex_element *enabled[BRW_ATTRIB_MAX];
+   struct brw_vertex_element *enabled[VERT_ATTRIB_MAX];
    GLuint nr_enabled = 0;
 
-   struct brw_vertex_element *upload[BRW_ATTRIB_MAX];
+   struct brw_vertex_element *upload[VERT_ATTRIB_MAX];
    GLuint nr_uploads = 0;
    
 
@@ -412,17 +408,19 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 
    /* First build an array of pointers to ve's in vb.inputs_read
     */
+   if (0)
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
    
    while (tmp) {
       GLuint i = _mesa_ffsll(tmp)-1;
       struct brw_vertex_element *input = &brw->vb.inputs[i];
 
-      tmp &= ~((GLuint64EXT)1<<i);
+      tmp &= ~(1<<i);
       enabled[nr_enabled++] = input;
 
       input->index = i;
       input->element_size = get_size(input->glarray->Type) * input->glarray->Size;
-      input->count = input->glarray->StrideB ? max_index - min_index : 1;
+      input->count = input->glarray->StrideB ? max_index + 1 - min_index : 1;
 
       if (!input->glarray->BufferObj->Name) {
 	 if (i == 0) {
@@ -441,10 +439,16 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 	 }
 
 	 upload[nr_uploads++] = input;
-	 input->vbo_rebase_offset = 0;
+	 
+	 /* We rebase drawing to start at element zero only when
+	  * varyings are not in vbos, which means we can end up
+	  * uploading non-varying arrays (stride != 0) when min_index
+	  * is zero.  This doesn't matter as the amount to upload is
+	  * the same for these arrays whether the draw call is rebased
+	  * or not - we just have to upload the one element.
+	  */
+	 assert(min_index == 0 || input->glarray->StrideB == 0);
       }
-      else 
-	 input->vbo_rebase_offset = min_index * input->glarray->StrideB;
    }
 
    /* Upload interleaved arrays if all uploads are interleaved
@@ -457,7 +461,6 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
       input0->glarray = copy_array_to_vbo_array(brw, 0,
 						input0->glarray, 
 						interleave,
-						min_index,
 						input0->count);
 
       for (i = 1; i < nr_uploads; i++) {
@@ -475,7 +478,6 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 	 input->glarray = copy_array_to_vbo_array(brw, i, 
 						  input->glarray,
 						  input->element_size,
-						  min_index,
 						  input->count);
 
       }
@@ -523,9 +525,9 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
       vbp.vb[i].vb0.bits.pad = 0;
       vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA;
       vbp.vb[i].vb0.bits.vb_index = i;
-      vbp.vb[i].offset = (GLuint)input->glarray->Ptr + input->vbo_rebase_offset;
+      vbp.vb[i].offset = (GLuint)input->glarray->Ptr;
       vbp.vb[i].buffer = array_buffer(input->glarray);
-      vbp.vb[i].max_index = max_index - min_index;
+      vbp.vb[i].max_index = max_index;
    }
 
 
@@ -566,94 +568,32 @@ static GLuint element_size( GLenum type )
 
 
 
-
-static void rebase_indices_to_vbo_indices( struct brw_context *brw, 
-					   const struct brw_draw_index_buffer *index_buffer,
-					   struct gl_buffer_object **vbo_return,
-					   GLuint *offset_return )
+void brw_upload_indices( struct brw_context *brw,
+			 const struct _mesa_index_buffer *index_buffer )
 {
    GLcontext *ctx = &brw->intel.ctx;
-   GLuint min_index = index_buffer->rebase;
-   const void *indices = index_buffer->ptr;
-   GLsizei count = index_buffer->count;
-   GLenum type = index_buffer->type;
-   GLuint size = element_size(type) * count;
-   struct gl_buffer_object *bufferobj;
-   GLuint offset;
-   GLuint i;
-
-   get_space(brw, size, &bufferobj, &offset);
+   struct intel_context *intel = &brw->intel;
+   GLuint ib_size = get_size(index_buffer->type) * index_buffer->count;
+   struct gl_buffer_object *bufferobj = index_buffer->obj;
+   GLuint offset = (GLuint)index_buffer->ptr;
 
-   *vbo_return = bufferobj;
-   *offset_return = offset;
+   /* Turn into a proper VBO:
+    */
+   if (!bufferobj->Name) {
+     
+      /* Get new bufferobj, offset:
+       */
+      get_space(brw, ib_size, &bufferobj, &offset);
 
-   if (min_index == 0) {
       /* Straight upload
        */
       ctx->Driver.BufferSubData( ctx,
 				 GL_ELEMENT_ARRAY_BUFFER_ARB,
 				 offset, 
-				 size,
-				 indices,
+				 ib_size,
+				 index_buffer->ptr,
 				 bufferobj);
    }
-   else {
-      void *map = ctx->Driver.MapBuffer(ctx,
-					GL_ELEMENT_ARRAY_BUFFER_ARB,
-					GL_DYNAMIC_DRAW_ARB,
-					bufferobj);
-
-      map += offset;
-
-      switch (type) {
-      case GL_UNSIGNED_INT: {
-	 GLuint *ui_map = (GLuint *)map;
-	 const GLuint *ui_indices = (const GLuint *)indices;
-
-	 for (i = 0; i < count; i++)
-	    ui_map[i] = ui_indices[i] - min_index;
-	 break;
-      }
-      case GL_UNSIGNED_SHORT:  {
-	 GLushort *us_map = (GLushort *)map;
-	 const GLushort *us_indices = (const GLushort *)indices;
-
-	 for (i = 0; i < count; i++)
-	    us_map[i] = us_indices[i] - min_index;
-	 break;
-      }
-      case GL_UNSIGNED_BYTE:  {
-	 GLubyte *ub_map = (GLubyte *)map;
-	 const GLubyte *ub_indices = (const GLubyte *)indices;
-
-	 for (i = 0; i < count; i++)
-	    ub_map[i] = ub_indices[i] - min_index;
-	 break;
-      }
-      }
-
-      ctx->Driver.UnmapBuffer(ctx, 
-			      GL_ELEMENT_ARRAY_BUFFER_ARB, 
-			      bufferobj);
-
-   }
-}
-
-
-
-void brw_upload_indices( struct brw_context *brw,
-			 const struct brw_draw_index_buffer *index_buffer)
-{
-   struct intel_context *intel = &brw->intel;
-   GLuint ib_size = get_size(index_buffer->type) * index_buffer->count;
-   struct gl_buffer_object *bufferobj = index_buffer->obj;
-   GLuint offset = (GLuint)index_buffer->ptr;
-
-   /* Already turned into a proper VBO:
-    */
-   if (!index_buffer->obj->Name) {
-      rebase_indices_to_vbo_indices(brw, index_buffer, &bufferobj, &offset );
-   }
 
    /* Emit the indexbuffer packet:
     */
diff --git a/src/mesa/drivers/dri/i965/brw_exec.h b/src/mesa/drivers/dri/i965/brw_exec.h
deleted file mode 100644
index f07b448587..0000000000
--- a/src/mesa/drivers/dri/i965/brw_exec.h
+++ /dev/null
@@ -1,150 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#ifndef __BRW_EXEC_H__
-#define __BRW_EXEC_H__
-
-#include "mtypes.h"
-#include "brw_attrib.h"
-#include "brw_draw.h"
-
-
-#define BRW_MAX_PRIM 64
-
-/* Wierd implementation stuff:
- */
-#define BRW_VERT_BUFFER_SIZE (1024*16)	/* dwords == 64k */
-#define BRW_MAX_ATTR_CODEGEN 16 
-#define ERROR_ATTRIB 16
-
-
-
-
-struct brw_exec_eval1_map {
-   struct gl_1d_map *map;
-   GLuint sz;
-};
-
-struct brw_exec_eval2_map {
-   struct gl_2d_map *map;
-   GLuint sz;
-};
-
-
-
-struct brw_exec_copied_vtx {
-   GLfloat buffer[BRW_ATTRIB_MAX * 4 * BRW_MAX_COPIED_VERTS];
-   GLuint nr;
-};
-
-
-typedef void (*brw_attrfv_func)( const GLfloat * );
-
-
-struct brw_exec_context
-{
-   GLcontext *ctx;   
-   GLvertexformat vtxfmt;
-
-   struct {
-      struct gl_buffer_object *bufferobj;
-      GLubyte *buffer_map;
-
-      GLuint vertex_size;
-
-      struct brw_draw_prim prim[BRW_MAX_PRIM];
-      GLuint prim_count;
-
-      GLfloat *vbptr;		     /* cursor, points into buffer */
-      GLfloat vertex[BRW_ATTRIB_MAX*4]; /* current vertex */
-
-      GLfloat *current[BRW_ATTRIB_MAX]; /* points into ctx->Current, ctx->Light.Material */
-      GLfloat CurrentFloatEdgeFlag;
-
-      GLuint vert_count;
-      GLuint max_vert;
-      struct brw_exec_copied_vtx copied;
-
-      GLubyte attrsz[BRW_ATTRIB_MAX];
-      GLubyte active_sz[BRW_ATTRIB_MAX];
-
-      GLfloat *attrptr[BRW_ATTRIB_MAX]; 
-      struct gl_client_array arrays[BRW_ATTRIB_MAX];
-      const struct gl_client_array *inputs[BRW_ATTRIB_MAX];
-   } vtx;
-
-   
-   struct {
-      GLboolean recalculate_maps;
-      struct brw_exec_eval1_map map1[BRW_ATTRIB_MAX];
-      struct brw_exec_eval2_map map2[BRW_ATTRIB_MAX];
-   } eval;
-
-   struct {
-      const struct gl_client_array *inputs[BRW_ATTRIB_MAX];
-
-      struct gl_buffer_object *index_obj;
-   } array;
-};
-
-
-
-/* External API:
- */
-void brw_exec_init( GLcontext *ctx );
-void brw_exec_destroy( GLcontext *ctx );
-void brw_exec_invalidate_state( GLcontext *ctx, GLuint new_state );
-void brw_exec_FlushVertices( GLcontext *ctx, GLuint flags );
-void brw_exec_wakeup( GLcontext *ctx );
-
-
-/* Internal functions:
- */
-void brw_exec_array_init( struct brw_exec_context *exec );
-void brw_exec_array_destroy( struct brw_exec_context *exec );
-
-
-void brw_exec_vtx_init( struct brw_exec_context *exec );
-void brw_exec_vtx_destroy( struct brw_exec_context *exec );
-void brw_exec_vtx_flush( struct brw_exec_context *exec );
-void brw_exec_vtx_wrap( struct brw_exec_context *exec );
-
-void brw_exec_eval_update( struct brw_exec_context *exec );
-
-void brw_exec_do_EvalCoord2f( struct brw_exec_context *exec, 
-				     GLfloat u, GLfloat v );
-
-void brw_exec_do_EvalCoord1f( struct brw_exec_context *exec,
-				     GLfloat u);
-
-#endif
diff --git a/src/mesa/drivers/dri/i965/brw_exec_array.c b/src/mesa/drivers/dri/i965/brw_exec_array.c
deleted file mode 100644
index ca19a19837..0000000000
--- a/src/mesa/drivers/dri/i965/brw_exec_array.c
+++ /dev/null
@@ -1,283 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include "glheader.h"
-#include "context.h"
-#include "state.h"
-#include "api_validate.h"
-#include "api_noop.h"
-#include "dispatch.h"
-
-#include "brw_attrib.h"
-#include "brw_draw.h"
-#include "brw_exec.h"
-#include "brw_fallback.h"
-
-static GLuint get_max_index( GLuint count, GLuint type, 
-			     const GLvoid *indices )
-{
-   GLint i;
-
-   /* Compute max element.  This is only needed for upload of non-VBO,
-    * non-constant data elements.
-    *
-    * XXX: Postpone this calculation until it is known that it is
-    * needed.  Otherwise could scan this pointlessly in the all-vbo
-    * case.
-    */
-   switch(type) {
-   case GL_UNSIGNED_INT: {
-      const GLuint *ui_indices = (const GLuint *)indices;
-      GLuint max_ui = 0;
-      for (i = 0; i < count; i++)
-	 if (ui_indices[i] > max_ui)
-	    max_ui = ui_indices[i];
-      return max_ui;
-   }
-   case GL_UNSIGNED_SHORT: {
-      const GLushort *us_indices = (const GLushort *)indices;
-      GLuint max_us = 0;
-      for (i = 0; i < count; i++)
-	 if (us_indices[i] > max_us)
-	    max_us = us_indices[i];
-      return max_us;
-   }
-   case GL_UNSIGNED_BYTE: {
-      const GLubyte *ub_indices = (const GLubyte *)indices;
-      GLuint max_ub = 0;
-      for (i = 0; i < count; i++)
-	 if (ub_indices[i] > max_ub)
-	    max_ub = ub_indices[i];
-      return max_ub;
-   }
-   default:
-      return 0;
-   }
-}
-
-
-
-
-/***********************************************************************
- * API functions.
- */
-
-static void GLAPIENTRY
-brw_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
-   struct brw_draw_prim prim[1];
-   GLboolean ok;
-
-   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-      return;
-
-   FLUSH_CURRENT( ctx, 0 );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-      
-   prim[0].begin = 1;
-   prim[0].end = 1;
-   prim[0].weak = 0;
-   prim[0].pad = 0;
-
-   if (exec->array.inputs[0]->BufferObj->Name) {
-      /* Use vertex attribute as a hint to tell us if we expect all
-       * arrays to be in VBO's and if so, don't worry about avoiding
-       * the upload of elements < start.
-       */
-      prim[0].mode = mode;
-      prim[0].start = start;
-      prim[0].count = count;
-      prim[0].indexed = 0;
-
-      ok = brw_draw_prims( ctx, exec->array.inputs, prim, 1, NULL, 0, start + count, 0 );
-   }
-   else {
-      /* If not using VBO's, we don't want to upload any more elements
-       * than necessary from the arrays as they will not be valid next
-       * time the application tries to draw with them.
-       */
-      prim[0].mode = mode;
-      prim[0].start = 0;
-      prim[0].count = count;
-      prim[0].indexed = 0;
-
-      ok = brw_draw_prims( ctx, exec->array.inputs, prim, 1, NULL, start, start + count, 0 );
-   }
-   
-   if (!ok) {
-      brw_fallback(ctx);
-      CALL_DrawArrays(ctx->Exec, ( mode, start, count ));
-      brw_unfallback(ctx);
-   }
-}
-
-
-
-static void GLAPIENTRY
-brw_exec_DrawRangeElements(GLenum mode,
-			   GLuint start, GLuint end,
-			   GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
-   struct brw_draw_index_buffer ib;
-   struct brw_draw_prim prim[1];
-
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
-      return;
-
-   FLUSH_CURRENT( ctx, 0 );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-      
-   ib.count = count;
-   ib.type = type; 
-   ib.obj = ctx->Array.ElementArrayBufferObj;
-   ib.ptr = indices;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* Use the fact that indices are in a VBO as a hint that the
-       * program has put all the arrays in VBO's and we don't have to
-       * worry about performance implications of start > 0.
-       *
-       * XXX: consider passing start as min_index to draw_prims instead.
-       */
-      ib.rebase = 0;
-   }
-   else {
-      ib.rebase = start;
-   }
-
-   prim[0].begin = 1;
-   prim[0].end = 1;
-   prim[0].weak = 0;
-   prim[0].pad = 0;
-   prim[0].mode = mode;
-   prim[0].start = 0;
-   prim[0].count = count;
-   prim[0].indexed = 1;
-
-   if (!brw_draw_prims( ctx, exec->array.inputs, prim, 1, &ib, ib.rebase, end+1, 0 )) {
-      brw_fallback(ctx);
-      CALL_DrawRangeElements(ctx->Exec, (mode, start, end, count, type, indices));
-      brw_unfallback(ctx);
-   }
-}
-
-
-static void GLAPIENTRY
-brw_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint max_index;
-
-   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-      return;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      const GLvoid *map = ctx->Driver.MapBuffer(ctx,
-						 GL_ELEMENT_ARRAY_BUFFER_ARB,
-						 GL_DYNAMIC_READ_ARB,
-						 ctx->Array.ElementArrayBufferObj);
-
-      max_index = get_max_index(count, type, ADD_POINTERS(map, indices));
-
-      ctx->Driver.UnmapBuffer(ctx,
-			      GL_ELEMENT_ARRAY_BUFFER_ARB,
-			      ctx->Array.ElementArrayBufferObj);
-   }
-   else {
-      max_index = get_max_index(count, type, indices);
-   }
-
-   brw_exec_DrawRangeElements(mode, 0, max_index, count, type, indices);
-}
-
-
-/***********************************************************************
- * Initialization
- */
-
-
-static void init_arrays( GLcontext *ctx, 
-			 const struct gl_client_array *arrays[] )
-{
-   struct gl_array_object *obj = ctx->Array.ArrayObj;
-   GLuint i;
-
-   memset(arrays, 0, sizeof(*arrays) * BRW_ATTRIB_MAX);
-
-   arrays[BRW_ATTRIB_POS] = &obj->Vertex;
-   arrays[BRW_ATTRIB_NORMAL] = &obj->Normal;
-   arrays[BRW_ATTRIB_COLOR0] = &obj->Color;
-   arrays[BRW_ATTRIB_COLOR1] = &obj->SecondaryColor;
-   arrays[BRW_ATTRIB_FOG] = &obj->FogCoord;
-
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
-      arrays[BRW_ATTRIB_TEX0 + i] = &obj->TexCoord[i];
-
-   arrays[BRW_ATTRIB_INDEX] = &obj->Index;
-   arrays[BRW_ATTRIB_EDGEFLAG] = &obj->EdgeFlag;
-
-   for (i = BRW_ATTRIB_GENERIC0; i <= BRW_ATTRIB_GENERIC15; i++)
-      arrays[i] = &obj->VertexAttrib[i - BRW_ATTRIB_GENERIC0];
-}
-
-
-
-
-void brw_exec_array_init( struct brw_exec_context *exec )
-{
-   GLcontext *ctx = exec->ctx;
-
-   init_arrays(ctx, exec->array.inputs);
-
-#if 1
-   exec->vtxfmt.DrawArrays = brw_exec_DrawArrays;
-   exec->vtxfmt.DrawElements = brw_exec_DrawElements;
-   exec->vtxfmt.DrawRangeElements = brw_exec_DrawRangeElements;
-#else
-   exec->vtxfmt.DrawArrays = _mesa_noop_DrawArrays;
-   exec->vtxfmt.DrawElements = _mesa_noop_DrawElements;
-   exec->vtxfmt.DrawRangeElements = _mesa_noop_DrawRangeElements;
-#endif
-
-   exec->array.index_obj = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB);
-}
-
-
-void brw_exec_array_destroy( struct brw_exec_context *exec )
-{
-   GLcontext *ctx = exec->ctx;
-
-   ctx->Driver.DeleteBuffer(ctx, exec->array.index_obj);
-}
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 56e21e339e..86464b2ec5 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -30,8 +30,6 @@
 #include "tnl/tnl.h"
 #include "context.h"
 #include "brw_context.h"
-#include "brw_exec.h"
-#include "brw_save.h"
 #include "brw_fallback.h"
 
 #include "glheader.h"
@@ -40,297 +38,6 @@
 #include "imports.h"
 #include "macros.h"
 #include "mtypes.h"
-#include "dispatch.h"
-
-
-typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
-
-
-/* Wrapper functions in case glVertexAttrib*fvNV doesn't exist */
-static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc[4] = {
-   VertexAttrib1fvNV,
-   VertexAttrib2fvNV,
-   VertexAttrib3fvNV,
-   VertexAttrib4fvNV
-};
-
-#if 0
-static void VertexAttrib1fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvARB(ctx->Exec, (target, v));
-}
-
-
-static attr_func vert_attrfunc_arb[4] = {
-   VertexAttrib1fvARB,
-   VertexAttrib2fvARB,
-   VertexAttrib3fvARB,
-   VertexAttrib4fvARB
-};
-#endif
-
-
-
-
-
-
-static void mat_attr1fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case BRW_ATTRIB_MAT_FRONT_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SHININESS, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SHININESS, v ));
-      break;
-   }
-}
-
-
-static void mat_attr3fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case BRW_ATTRIB_MAT_FRONT_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_COLOR_INDEXES, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_COLOR_INDEXES, v ));
-      break;
-   }
-}
-
-
-static void mat_attr4fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case BRW_ATTRIB_MAT_FRONT_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_EMISSION, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_EMISSION, v ));
-      break;
-   case BRW_ATTRIB_MAT_FRONT_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_AMBIENT, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_AMBIENT, v ));
-      break;
-   case BRW_ATTRIB_MAT_FRONT_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_DIFFUSE, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_DIFFUSE, v ));
-      break;
-   case BRW_ATTRIB_MAT_FRONT_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SPECULAR, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SPECULAR, v ));
-      break;
-   }
-}
-
-
-static attr_func mat_attrfunc[4] = {
-   mat_attr1fv,
-   NULL,
-   mat_attr3fv,
-   mat_attr4fv
-};
-
-
-static void index_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_Indexf(ctx->Exec, (v[0]));
-}
-
-static void edgeflag_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_EdgeFlag(ctx->Exec, ((GLboolean)(v[0] == 1.0)));
-}
-
-struct loopback_attr {
-   GLint target;
-   GLint sz;
-   attr_func func;
-};
-
-/* Don't emit ends and begins on wrapped primitives.  Don't replay
- * wrapped vertices.  If we get here, it's probably because the the
- * precalculated wrapping is wrong.
- */
-static void loopback_prim( GLcontext *ctx,
-			   const GLfloat *buffer,
-			   const struct brw_draw_prim *prim,
-			   GLuint wrap_count,
-			   GLuint vertex_size,
-			   const struct loopback_attr *la, GLuint nr )
-{
-   GLint start = prim->start;
-   GLint end = start + prim->count;
-   const GLfloat *data;
-   GLint j;
-   GLuint k;
-
-   if (0)
-      _mesa_printf("loopback prim %s(%s,%s) verts %d..%d\n",
-		   _mesa_lookup_enum_by_nr(prim->mode),
-		   prim->begin ? "begin" : "..",
-		   prim->end ? "end" : "..",
-		   start, 
-		   end);
-
-   if (prim->begin) {
-      CALL_Begin(GET_DISPATCH(), ( prim->mode ));
-   }
-   else {
-      assert(start == 0);
-      start += wrap_count;
-   }
-
-   data = buffer + start * vertex_size;
-
-   for (j = start ; j < end ; j++) {
-      const GLfloat *tmp = data + la[0].sz;
-
-      for (k = 1 ; k < nr ; k++) {
-	 la[k].func( ctx, la[k].target, tmp );
-	 tmp += la[k].sz;
-      }
-	 
-      /* Fire the vertex
-       */
-      la[0].func( ctx, BRW_ATTRIB_POS, data );
-      data = tmp;
-   }
-
-   if (prim->end) {
-      CALL_End(GET_DISPATCH(), ());
-   }
-}
-
-/* Primitives generated by DrawArrays/DrawElements/Rectf may be
- * caught here.  If there is no primitive in progress, execute them
- * normally, otherwise need to track and discard the generated
- * primitives.
- */
-static void loopback_weak_prim( GLcontext *ctx,
-				const struct brw_draw_prim *prim )
-{
-   /* Use the prim_weak flag to ensure that if this primitive
-    * wraps, we don't mistake future vertex_lists for part of the
-    * surrounding primitive.
-    *
-    * While this flag is set, we are simply disposing of data
-    * generated by an operation now known to be a noop.
-    */
-   if (prim->begin)
-      ctx->Driver.CurrentExecPrimitive |= BRW_SAVE_PRIM_WEAK;
-   if (prim->end)
-      ctx->Driver.CurrentExecPrimitive &= ~BRW_SAVE_PRIM_WEAK;
-}
-
-
-void brw_loopback_vertex_list( GLcontext *ctx,
-			       const GLfloat *buffer,
-			       const GLubyte *attrsz,
-			       const struct brw_draw_prim *prim,
-			       GLuint prim_count,
-			       GLuint wrap_count,
-			       GLuint vertex_size)
-{
-   struct loopback_attr la[BRW_ATTRIB_MAX];
-   GLuint i, nr = 0;
-
-   for (i = 0 ; i <= BRW_ATTRIB_TEX7 ; i++) {
-      if (i == BRW_ATTRIB_INDEX || i == BRW_ATTRIB_EDGEFLAG)
-                   continue;
-
-      if (attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = attrsz[i];
-	 la[nr].func = vert_attrfunc[attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   for (i = BRW_ATTRIB_MAT_FRONT_AMBIENT ; 
-	i <= BRW_ATTRIB_MAT_BACK_INDEXES ; 
-	i++) {
-      if (attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = attrsz[i];
-	 la[nr].func = mat_attrfunc[attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   if (attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      la[nr].target = BRW_ATTRIB_EDGEFLAG;
-      la[nr].sz = attrsz[BRW_ATTRIB_EDGEFLAG];
-      la[nr].func = edgeflag_attr1fv;
-      nr++;
-   }
-
-   if (attrsz[BRW_ATTRIB_INDEX]) {
-      la[nr].target = BRW_ATTRIB_INDEX;
-      la[nr].sz = attrsz[BRW_ATTRIB_INDEX];
-      la[nr].func = index_attr1fv;
-      nr++;
-   }
-
-   /* XXX ARB vertex attribs */
-
-   for (i = 0 ; i < prim_count ; i++) {
-      if ((prim[i].mode & BRW_SAVE_PRIM_WEAK) &&
-	  (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END))
-      {
-	 loopback_weak_prim( ctx, &prim[i] );
-      }
-      else
-      {
-	 loopback_prim( ctx, buffer, &prim[i], wrap_count, vertex_size, la, nr );
-      }
-   }
-}
-
 
 
 
@@ -405,62 +112,6 @@ const struct brw_tracked_state brw_check_fallback = {
 
 
 
-/* If there is a fallback, fallback to software rasterization and
- * transformation together.  There is never a requirement to have
- * software t&l but hardware rasterization.
- * 
- * Further, all fallbacks are based on GL state, not on eg. primitive
- * or vertex data.
- */
-
-static void do_fallback( struct brw_context *brw,
-			 GLboolean fallback )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   /* flush:
-    */
-   ctx->Driver.Flush( ctx );
-
-   if (fallback) {
-      _swsetup_Wakeup( ctx );
-      _tnl_wakeup_exec( ctx );	
-
-      /* Need this because tnl_wakeup_exec does too much: 
-       */
-      brw_save_wakeup(ctx);
-      brw_save_fallback(ctx, GL_TRUE);
-   }
-   else {
-      /* Flush vertices and copy-to-current:
-       */
-      FLUSH_CURRENT(ctx, 0); 
-
-      _swrast_flush( ctx );
-
-      brw_exec_wakeup(ctx);
-
-      /* Need this because tnl_wakeup_exec does too much: 
-       */
-      brw_save_wakeup(ctx);
-      brw_save_fallback(ctx, GL_FALSE);	 
-   }
-}
-
-
-void brw_fallback( GLcontext *ctx )
-{
-   struct brw_context *brw = brw_context(ctx);
-   do_fallback(brw, 1);
-}
-
-
-void brw_unfallback( GLcontext *ctx )
-{
-   struct brw_context *brw = brw_context(ctx);
-   do_fallback(brw, 0);
-}
-
 /* Not used:
  */
 void intelFallback( struct intel_context *intel, GLuint bit, GLboolean mode )
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.h b/src/mesa/drivers/dri/i965/brw_fallback.h
index 81a2d344b8..684a46cd17 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.h
+++ b/src/mesa/drivers/dri/i965/brw_fallback.h
@@ -31,7 +31,7 @@
 #include "mtypes.h"		/* for GLcontext... */
 
 struct brw_context;
-struct brw_draw_prim;
+struct vbo_prim;
 
 void brw_fallback( GLcontext *ctx );
 void brw_unfallback( GLcontext *ctx );
@@ -39,7 +39,7 @@ void brw_unfallback( GLcontext *ctx );
 void brw_loopback_vertex_list( GLcontext *ctx,
 			       const GLfloat *buffer,
 			       const GLubyte *attrsz,
-			       const struct brw_draw_prim *prim,
+			       const struct vbo_prim *prim,
 			       GLuint prim_count,
 			       GLuint wrap_count,
 			       GLuint vertex_size);
diff --git a/src/mesa/drivers/dri/i965/brw_metaops.c b/src/mesa/drivers/dri/i965/brw_metaops.c
index acd1d58191..1579762b6d 100644
--- a/src/mesa/drivers/dri/i965/brw_metaops.c
+++ b/src/mesa/drivers/dri/i965/brw_metaops.c
@@ -47,7 +47,6 @@
 #include "brw_context.h"
 #include "brw_defines.h"
 #include "brw_draw.h"
-#include "brw_attrib.h"
 #include "brw_fallback.h"
 
 #define INIT(brw, STRUCT, ATTRIB) 		\
@@ -388,8 +387,8 @@ static void meta_draw_quad(struct intel_context *intel,
    struct brw_context *brw = brw_context(&intel->ctx);
    struct gl_client_array pos_array;
    struct gl_client_array color_array;
-   struct gl_client_array *attribs[BRW_ATTRIB_MAX];
-   struct brw_draw_prim prim[1];
+   struct gl_client_array *attribs[VERT_ATTRIB_MAX];
+   struct _mesa_prim prim[1];
    GLfloat pos[4][3];
    GLubyte color[4];
 
@@ -439,29 +438,29 @@ static void meta_draw_quad(struct intel_context *intel,
    /* Ignoring texture coords. 
     */
 
-   memset(attribs, 0, BRW_ATTRIB_MAX * sizeof(*attribs));
-
-   attribs[BRW_ATTRIB_POS] = &pos_array;
-   attribs[BRW_ATTRIB_POS]->Ptr = 0;
-   attribs[BRW_ATTRIB_POS]->Type = GL_FLOAT;
-   attribs[BRW_ATTRIB_POS]->Enabled = 1;
-   attribs[BRW_ATTRIB_POS]->Size = 3;
-   attribs[BRW_ATTRIB_POS]->StrideB = 3 * sizeof(GLfloat);
-   attribs[BRW_ATTRIB_POS]->Stride = 3 * sizeof(GLfloat);
-   attribs[BRW_ATTRIB_POS]->_MaxElement = 4;
-   attribs[BRW_ATTRIB_POS]->Normalized = 0;
-   attribs[BRW_ATTRIB_POS]->BufferObj = brw->metaops.vbo;
-
-   attribs[BRW_ATTRIB_COLOR0] = &color_array;
-   attribs[BRW_ATTRIB_COLOR0]->Ptr = (const GLubyte *)sizeof(pos);
-   attribs[BRW_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE;
-   attribs[BRW_ATTRIB_COLOR0]->Enabled = 1;
-   attribs[BRW_ATTRIB_COLOR0]->Size = 4;
-   attribs[BRW_ATTRIB_COLOR0]->StrideB = 0;
-   attribs[BRW_ATTRIB_COLOR0]->Stride = 0;
-   attribs[BRW_ATTRIB_COLOR0]->_MaxElement = 1;
-   attribs[BRW_ATTRIB_COLOR0]->Normalized = 1;
-   attribs[BRW_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo;
+   memset(attribs, 0, VERT_ATTRIB_MAX * sizeof(*attribs));
+
+   attribs[VERT_ATTRIB_POS] = &pos_array;
+   attribs[VERT_ATTRIB_POS]->Ptr = 0;
+   attribs[VERT_ATTRIB_POS]->Type = GL_FLOAT;
+   attribs[VERT_ATTRIB_POS]->Enabled = 1;
+   attribs[VERT_ATTRIB_POS]->Size = 3;
+   attribs[VERT_ATTRIB_POS]->StrideB = 3 * sizeof(GLfloat);
+   attribs[VERT_ATTRIB_POS]->Stride = 3 * sizeof(GLfloat);
+   attribs[VERT_ATTRIB_POS]->_MaxElement = 4;
+   attribs[VERT_ATTRIB_POS]->Normalized = 0;
+   attribs[VERT_ATTRIB_POS]->BufferObj = brw->metaops.vbo;
+
+   attribs[VERT_ATTRIB_COLOR0] = &color_array;
+   attribs[VERT_ATTRIB_COLOR0]->Ptr = (const GLubyte *)sizeof(pos);
+   attribs[VERT_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE;
+   attribs[VERT_ATTRIB_COLOR0]->Enabled = 1;
+   attribs[VERT_ATTRIB_COLOR0]->Size = 4;
+   attribs[VERT_ATTRIB_COLOR0]->StrideB = 0;
+   attribs[VERT_ATTRIB_COLOR0]->Stride = 0;
+   attribs[VERT_ATTRIB_COLOR0]->_MaxElement = 1;
+   attribs[VERT_ATTRIB_COLOR0]->Normalized = 1;
+   attribs[VERT_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo;
    
    /* Just ignoring texture coordinates for now. 
     */
@@ -476,19 +475,12 @@ static void meta_draw_quad(struct intel_context *intel,
    prim[0].start = 0;
    prim[0].count = 4;
 
-   if (!brw_draw_prims(&brw->intel.ctx, 
-		       (const struct gl_client_array **)attribs,
-		       prim, 1,
-		       NULL,
-		       0,
-		       4,
-		       BRW_DRAW_LOCKED ))
-   {
-      /* This should not be possible:
-       */
-      _mesa_printf("brw_draw_prims failed in metaops!\n");
-      assert(0);
-   }
+   brw_draw_prims(&brw->intel.ctx, 
+		  (const struct gl_client_array **)attribs,
+		  prim, 1,
+		  NULL,
+		  0,
+		  3 );
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_save.c b/src/mesa/drivers/dri/i965/brw_save.c
deleted file mode 100644
index 1af7791c4d..0000000000
--- a/src/mesa/drivers/dri/i965/brw_save.c
+++ /dev/null
@@ -1,126 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.3
- *
- * Copyright (C) 1999-2005  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 "mtypes.h"
-#include "api_arrayelt.h"
-#include "dlist.h"
-#include "vtxfmt.h"
-#include "imports.h"
-
-#include "brw_save.h"
-
-
-
-void brw_save_init( GLcontext *ctx )
-{
-   struct brw_save_context *save = CALLOC_STRUCT(brw_save_context);
-
-   if (ctx->swtnl_im == NULL) {
-      ctx->swtnl_im = CALLOC_STRUCT(brw_exec_save);
-   }
-
-   save->ctx = ctx;
-   IMM_CONTEXT(ctx)->save = save;
-
-   /* Initialize the arrayelt helper
-    */
-   if (!ctx->aelt_context &&
-       !_ae_create_context( ctx )) 
-      return;
-
-   brw_save_api_init( save );
-   brw_save_wakeup(ctx);
-
-   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
-}
-
-
-void brw_save_destroy( GLcontext *ctx )
-{
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
-   if (save) {
-      FREE(save);
-      IMM_CONTEXT(ctx)->save = NULL;
-   }
-
-   if (ctx->aelt_context) {
-      _ae_destroy_context( ctx );
-      ctx->aelt_context = NULL;
-   }
-
-   if (IMM_CONTEXT(ctx)->exec == NULL &&
-       IMM_CONTEXT(ctx)->save == NULL) {
-      FREE(IMM_CONTEXT(ctx));
-      ctx->swtnl_im = NULL;
-   }
-}
-
-
-void brw_save_invalidate_state( GLcontext *ctx, GLuint new_state )
-{
-   _ae_invalidate_state(ctx, new_state);
-}
-
-
-/* Note that this can occur during the playback of a display list:
- */
-void brw_save_fallback( GLcontext *ctx, GLboolean fallback )
-{
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
-
-   if (fallback)
-      save->replay_flags |= BRW_SAVE_FALLBACK;
-   else
-      save->replay_flags &= ~BRW_SAVE_FALLBACK;
-}
-
-
-/* I don't see any reason to swap this code out on fallbacks.  It
- * wouldn't really mean anything to do so anyway as the old lists are
- * still around from pre-fallback.  Instead, the above code ensures
- * that vertices are routed back through immediate mode dispatch on
- * fallback.
- *
- * The below can be moved into init or removed:
- */
-void brw_save_wakeup( GLcontext *ctx )
-{
-   ctx->Driver.NewList = brw_save_NewList;
-   ctx->Driver.EndList = brw_save_EndList;
-   ctx->Driver.SaveFlushVertices = brw_save_SaveFlushVertices;
-   ctx->Driver.BeginCallList = brw_save_BeginCallList;
-   ctx->Driver.EndCallList = brw_save_EndCallList;
-   ctx->Driver.NotifySaveBegin = brw_save_NotifyBegin;
-
-   /* Assume we haven't been getting state updates either:
-    */
-   brw_save_invalidate_state( ctx, ~0 );
-}
-
-
-
diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c
index e5a28b96e3..74c9d88e46 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.c
+++ b/src/mesa/drivers/dri/i965/brw_vs.c
@@ -54,12 +54,11 @@ static void do_vs_prog( struct brw_context *brw,
    c.vp = vp;
 
    c.prog_data.outputs_written = vp->program.Base.OutputsWritten;
-   c.prog_data.inputs_read = brw_translate_inputs(brw->intel.ctx.VertexProgram._Enabled,
-						  vp->program.Base.InputsRead);
+   c.prog_data.inputs_read = vp->program.Base.InputsRead;
 
    if (c.key.copy_edgeflag) {
       c.prog_data.outputs_written |= 1<<VERT_RESULT_EDGE;
-      c.prog_data.inputs_read |= 1<<BRW_ATTRIB_EDGEFLAG;
+      c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG;
    }
 
    if (0)
diff --git a/src/mesa/drivers/dri/i965/brw_vs_constval.c b/src/mesa/drivers/dri/i965/brw_vs_constval.c
index 502d828397..528e164db8 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_constval.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_constval.c
@@ -162,6 +162,7 @@ static GLuint get_input_size(struct brw_context *brw,
    GLuint sizes_dword = brw->vb.info.sizes[attr/16];
    GLuint sizes_bits = (sizes_dword>>((attr%16)*2)) & 0x3;
    return sizes_bits + 1;
+/*    return brw->vb.inputs[attr].glarray->Size; */
 }
 
 /* Calculate sizes of vertex program outputs.  Size is the largest
@@ -176,8 +177,6 @@ static void calc_wm_input_sizes( struct brw_context *brw )
    struct tracker t;
    GLuint insn;
    GLuint i;
-   GLuint64EXT inputs = brw_translate_inputs(brw->intel.ctx.VertexProgram._Enabled,
-					     vp->program.Base.InputsRead);
 
    memset(&t, 0, sizeof(t));
 
@@ -185,8 +184,8 @@ static void calc_wm_input_sizes( struct brw_context *brw )
    if (brw->attribs.Light->Model.TwoSide)
       t.twoside = 1;
 
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) 
-      if (inputs & (1<<i))
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) 
+      if (vp->program.Base.InputsRead & (1<<i))
 	 set_active_component(&t, PROGRAM_INPUT, i, 
 			      szflag[get_input_size(brw, i)]);
       
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 8403e1bd7b..ffdb843e85 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -77,8 +77,8 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    /* Allocate input regs:  
     */
    c->nr_inputs = 0;
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) {
-      if (c->prog_data.inputs_read & ((GLuint64EXT)1<<i)) {
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      if (c->prog_data.inputs_read & (1<<i)) {
 	 c->nr_inputs++;
 	 c->regs[PROGRAM_INPUT][i] = brw_vec8_grf(reg, 0);
 	 reg++;
@@ -791,7 +791,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    if (c->key.copy_edgeflag) {
       brw_MOV(p, 
 	      get_reg(c, PROGRAM_OUTPUT, VERT_RESULT_EDGE),
-	      get_reg(c, PROGRAM_INPUT, BRW_ATTRIB_EDGEFLAG));
+	      get_reg(c, PROGRAM_INPUT, VERT_ATTRIB_EDGEFLAG));
    }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
index 82c1958887..0d61092247 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
@@ -146,9 +146,13 @@ static void make_state_key( GLcontext *ctx, struct state_key *key )
       }
 
       /* BRW_NEW_INPUT_VARYING */
-      for (i = BRW_ATTRIB_MAT_FRONT_AMBIENT ; i < BRW_ATTRIB_MAX ; i++) 
-	 if (brw->vb.info.varying & ((GLuint64EXT)1<<i)) 
-	    key->light_material_mask |= 1<<(i-BRW_ATTRIB_MAT_FRONT_AMBIENT);
+
+      /* For these programs, material values are stuffed into the
+       * generic slots:
+       */
+      for (i = 0 ; i < MAT_ATTRIB_MAX ; i++) 
+	 if (brw->vb.info.varying & (1<<(VERT_ATTRIB_GENERIC0 + i))) 
+	    key->light_material_mask |= 1<<i;
 
       for (i = 0; i < MAX_LIGHTS; i++) {
 	 struct gl_light *light = &brw->attribs.Light->Light[i];
@@ -374,17 +378,10 @@ static void release_temps( struct tnl_program *p )
 
 static struct ureg register_input( struct tnl_program *p, GLuint input )
 {
-   GLuint orig_input = input;
-   /* Cram the material flags into the generic range.  We'll translate
-    * them back later.
-    */
-   if (input >= BRW_ATTRIB_MAT_FRONT_AMBIENT)
-      input -= BRW_ATTRIB_MAT_FRONT_AMBIENT - BRW_ATTRIB_GENERIC0;
-
    assert(input < 32);
 
    p->program->Base.InputsRead |= (1<<input);
-   return make_ureg(PROGRAM_INPUT, orig_input);
+   return make_ureg(PROGRAM_INPUT, input);
 }
 
 static struct ureg register_output( struct tnl_program *p, GLuint output )
@@ -647,7 +644,7 @@ static void emit_passthrough( struct tnl_program *p,
 static struct ureg get_eye_position( struct tnl_program *p )
 {
    if (is_undef(p->eye_position)) {
-      struct ureg pos = register_input( p, BRW_ATTRIB_POS ); 
+      struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
       struct ureg modelview[4];
 
       p->eye_position = reserve_temp(p);
@@ -710,7 +707,7 @@ static struct ureg get_eye_position_normalized( struct tnl_program *p )
 static struct ureg get_eye_normal( struct tnl_program *p )
 {
    if (is_undef(p->eye_normal)) {
-      struct ureg normal = register_input(p, BRW_ATTRIB_NORMAL );
+      struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL );
       struct ureg mvinv[3];
 
       register_matrix_param6( p, STATE_MATRIX, STATE_MODELVIEW, 0, 0, 2,
@@ -743,7 +740,7 @@ static struct ureg get_eye_normal( struct tnl_program *p )
 
 static void build_hpos( struct tnl_program *p )
 {
-   struct ureg pos = register_input( p, BRW_ATTRIB_POS ); 
+   struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
    struct ureg hpos = register_output( p, VERT_RESULT_HPOS );
    struct ureg mvp[4];
 
@@ -787,9 +784,9 @@ static struct ureg get_material( struct tnl_program *p, GLuint side,
    GLuint attrib = material_attrib(side, property);
 
    if (p->color_materials & (1<<attrib))
-      return register_input(p, BRW_ATTRIB_COLOR0);
+      return register_input(p, VERT_ATTRIB_COLOR0);
    else if (p->materials & (1<<attrib)) 
-      return register_input( p, attrib + BRW_ATTRIB_MAT_FRONT_AMBIENT );
+      return register_input( p, attrib + _TNL_ATTRIB_MAT_FRONT_AMBIENT );
    else
       return register_param3( p, STATE_MATERIAL, side, property );
 }
@@ -1157,7 +1154,7 @@ static void build_fog( struct tnl_program *p )
       input = swizzle1(get_eye_position(p), Z);
    }
    else {
-      input = swizzle1(register_input(p, BRW_ATTRIB_FOG), X);
+      input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
    }
 
    if (p->state->fog_option &&
@@ -1299,7 +1296,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    for (j = 0; j < 4; j++) {
 	       switch (modes[j]) {
 	       case TXG_OBJ_LINEAR: {
-		  struct ureg obj = register_input(p, BRW_ATTRIB_POS);
+		  struct ureg obj = register_input(p, VERT_ATTRIB_POS);
 		  struct ureg plane = 
 		     register_param3(p, STATE_TEXGEN, i,
 				     STATE_TEXGEN_OBJECT_S + j);
@@ -1348,7 +1345,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    }
 
 	    if (copy_mask) {
-	       struct ureg in = register_input(p, BRW_ATTRIB_TEX0+i);
+	       struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i);
 	       emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in );
 	    }
 	 }
@@ -1357,7 +1354,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    struct ureg texmat[4];
 	    struct ureg in = (!is_undef(out_texgen) ? 
 			      out_texgen : 
-			      register_input(p, BRW_ATTRIB_TEX0+i));
+			      register_input(p, VERT_ATTRIB_TEX0+i));
 	    if (PREFER_DP4) {
 	       register_matrix_param6( p, STATE_MATRIX, STATE_TEXTURE, i, 
 				       0, 3, STATE_MATRIX, texmat );
@@ -1373,7 +1370,7 @@ static void build_texture_transform( struct tnl_program *p )
 	 release_temps(p);
       } 
       else {
-	 emit_passthrough(p, BRW_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
+	 emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
       }
    }
 }
@@ -1425,10 +1422,10 @@ static void build_tnl_program( struct tnl_program *p )
 	 build_lighting(p);
       else {
 	 if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-	    emit_passthrough(p, BRW_ATTRIB_COLOR0, VERT_RESULT_COL0);
+	    emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0);
 
 	 if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
-	    emit_passthrough(p, BRW_ATTRIB_COLOR1, VERT_RESULT_COL1);
+	    emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1);
       }
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c
index ac09754e3a..786f30e641 100644
--- a/src/mesa/drivers/dri/i965/brw_vtbl.c
+++ b/src/mesa/drivers/dri/i965/brw_vtbl.c
@@ -46,8 +46,6 @@
 #include "brw_state.h"
 
 #include "brw_draw.h"
-#include "brw_exec.h"
-#include "brw_save.h"
 #include "brw_state.h"
 #include "brw_aub.h"
 #include "brw_fallback.h"
@@ -68,9 +66,6 @@ static void brw_destroy_context( struct intel_context *intel )
    brw_destroy_state(brw);
    brw_draw_destroy( brw );
 
-   brw_exec_destroy( ctx );
-   brw_save_destroy( ctx );
-
    brw_ProgramCacheDestroy( ctx );
    brw_FrameBufferTexDestroy( brw );
 }
@@ -166,10 +161,7 @@ static GLuint brw_flush_cmd( void )
 
 static void brw_invalidate_state( struct intel_context *intel, GLuint new_state )
 {
-   GLcontext *ctx = &intel->ctx;
-
-   brw_exec_invalidate_state(ctx, new_state);
-   brw_save_invalidate_state(ctx, new_state);
+   /* nothing */
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/intel_context.c b/src/mesa/drivers/dri/i965/intel_context.c
index 4486a28adc..34560e4353 100644
--- a/src/mesa/drivers/dri/i965/intel_context.c
+++ b/src/mesa/drivers/dri/i965/intel_context.c
@@ -38,7 +38,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
@@ -223,7 +223,7 @@ static void intelInvalidateState( GLcontext *ctx, GLuint new_state )
 
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
    
@@ -372,7 +372,7 @@ GLboolean intelInitContext( struct intel_context *intel,
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -517,7 +517,7 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
       _swsetup_DestroyContext (&intel->ctx);
       _tnl_DestroyContext (&intel->ctx);
-      _ac_DestroyContext (&intel->ctx);
+      _vbo_DestroyContext (&intel->ctx);
 
       _swrast_DestroyContext (&intel->ctx);
       intel->Fallback = 0;	/* don't call _swrast_Flush later */
diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c
index 2ab1cf2617..5a6c301da2 100644
--- a/src/mesa/drivers/dri/mach64/mach64_context.c
+++ b/src/mesa/drivers/dri/mach64/mach64_context.c
@@ -38,7 +38,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -225,7 +225,7 @@ GLboolean mach64CreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -278,7 +278,7 @@ void mach64DestroyContext( __DRIcontextPrivate *driContextPriv  )
 
       _swsetup_DestroyContext( mmesa->glCtx );
       _tnl_DestroyContext( mmesa->glCtx );
-      _ac_DestroyContext( mmesa->glCtx );
+      _vbo_DestroyContext( mmesa->glCtx );
       _swrast_DestroyContext( mmesa->glCtx );
 
       if (release_texture_heaps) {
diff --git a/src/mesa/drivers/dri/mach64/mach64_state.c b/src/mesa/drivers/dri/mach64/mach64_state.c
index d4804a2c55..667a394520 100644
--- a/src/mesa/drivers/dri/mach64/mach64_state.c
+++ b/src/mesa/drivers/dri/mach64/mach64_state.c
@@ -40,7 +40,7 @@
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1023,7 +1023,7 @@ static void mach64DDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    MACH64_CONTEXT(ctx)->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c
index f02913c6af..f024f73eb6 100644
--- a/src/mesa/drivers/dri/mga/mga_xmesa.c
+++ b/src/mesa/drivers/dri/mga/mga_xmesa.c
@@ -45,7 +45,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -593,7 +593,7 @@ mgaCreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    
    _swsetup_CreateContext( ctx );
@@ -676,7 +676,7 @@ mgaDestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (mmesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( mmesa->glCtx );
       _tnl_DestroyContext( mmesa->glCtx );
-      _ac_DestroyContext( mmesa->glCtx );
+      _vbo_DestroyContext( mmesa->glCtx );
       _swrast_DestroyContext( mmesa->glCtx );
 
       mgaFreeVB( mmesa->glCtx );
diff --git a/src/mesa/drivers/dri/mga/mgastate.c b/src/mesa/drivers/dri/mga/mgastate.c
index 075fd348cf..c20a76f29e 100644
--- a/src/mesa/drivers/dri/mga/mgastate.c
+++ b/src/mesa/drivers/dri/mga/mgastate.c
@@ -41,7 +41,7 @@
 #include "mgaregs.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
@@ -1042,7 +1042,7 @@ static void mgaDDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    MGA_CONTEXT(ctx)->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.c b/src/mesa/drivers/dri/nouveau/nouveau_context.c
index e18cebcd29..f811dc1b72 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_context.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.c
@@ -31,7 +31,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "matrix.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
 #include "framebuffer.h"
 
 #include "tnl/tnl.h"
@@ -200,7 +199,7 @@ GLboolean nouveauCreateContext( const __GLcontextModes *glVisual,
 
 	/* Initialize the swrast */
 	_swrast_CreateContext( ctx );
-	_ac_CreateContext( ctx );
+	_vbo_CreateContext( ctx );
 	_tnl_CreateContext( ctx );
 	_swsetup_CreateContext( ctx );
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.c b/src/mesa/drivers/dri/nouveau/nouveau_state.c
index 18f6ffb2ad..aef1f63494 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_state.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.c
@@ -30,7 +30,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "nouveau_fifo.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -144,7 +143,7 @@ static void nouveauDDInvalidateState(GLcontext *ctx, GLuint new_state)
 {
     _swrast_InvalidateState( ctx, new_state );
     _swsetup_InvalidateState( ctx, new_state );
-    _ac_InvalidateState( ctx, new_state );
+    _vbo_InvalidateState( ctx, new_state );
     _tnl_InvalidateState( ctx, new_state );
     NOUVEAU_CONTEXT(ctx)->new_render_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/r128/r128_context.c b/src/mesa/drivers/dri/r128/r128_context.c
index 8ec027542a..89ddafa02a 100644
--- a/src/mesa/drivers/dri/r128/r128_context.c
+++ b/src/mesa/drivers/dri/r128/r128_context.c
@@ -42,7 +42,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -235,7 +235,7 @@ GLboolean r128CreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -293,7 +293,7 @@ void r128DestroyContext( __DRIcontextPrivate *driContextPriv  )
 
       _swsetup_DestroyContext( rmesa->glCtx );
       _tnl_DestroyContext( rmesa->glCtx );
-      _ac_DestroyContext( rmesa->glCtx );
+      _vbo_DestroyContext( rmesa->glCtx );
       _swrast_DestroyContext( rmesa->glCtx );
 
       if ( release_texture_heaps ) {
diff --git a/src/mesa/drivers/dri/r128/r128_state.c b/src/mesa/drivers/dri/r128/r128_state.c
index 1bfd370937..e476afa5d8 100644
--- a/src/mesa/drivers/dri/r128/r128_state.c
+++ b/src/mesa/drivers/dri/r128/r128_state.c
@@ -44,7 +44,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1250,7 +1250,7 @@ static void r128DDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    R128_CONTEXT(ctx)->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile
index 75c09ff867..e9144ac75c 100644
--- a/src/mesa/drivers/dri/r200/Makefile
+++ b/src/mesa/drivers/dri/r200/Makefile
@@ -26,10 +26,6 @@ DRIVER_SOURCES = r200_context.c \
 		 r200_span.c \
 		 r200_maos.c \
 		 r200_sanity.c \
-		 r200_vtxfmt.c \
-		 r200_vtxfmt_c.c \
-		 r200_vtxfmt_sse.c \
-		 r200_vtxfmt_x86.c \
 		 r200_fragshader.c \
 		 r200_vertprog.c \
 		 radeon_screen.c \
@@ -37,7 +33,7 @@ DRIVER_SOURCES = r200_context.c \
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
-X86_SOURCES = r200_vtxtmp_x86.S
+X86_SOURCES = 
 
 DRIVER_DEFINES = -DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R200
 
diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index bb7a16c932..fc6eb93daa 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -45,7 +45,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -60,7 +60,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tex.h"
 #include "r200_swtcl.h"
 #include "r200_tcl.h"
-#include "r200_vtxfmt.h"
 #include "r200_maos.h"
 #include "r200_vertprog.h"
 
@@ -415,7 +414,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _ae_create_context( ctx );
@@ -424,11 +423,10 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
     */
    _tnl_destroy_pipeline( ctx );
    _tnl_install_pipeline( ctx, r200_pipeline );
-   ctx->Driver.FlushVertices = r200FlushVertices;
 
    /* Try and keep materials and vertices separate:
     */
-   _tnl_isolate_materials( ctx, GL_TRUE );
+/*    _tnl_isolate_materials( ctx, GL_TRUE ); */
 
 
    /* Configure swrast and TNL to match hardware characteristics:
@@ -533,12 +531,6 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
       TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1);
    }
 
-   if (rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL) {
-      if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	 r200VtxfmtInit( ctx, tcl_mode >= DRI_CONF_TCL_CODEGEN );
-
-      _tnl_need_dlist_norm_lengths( ctx, GL_FALSE );
-   }
    return GL_TRUE;
 }
 
@@ -568,7 +560,7 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
       release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( rmesa->glCtx );
       _tnl_DestroyContext( rmesa->glCtx );
-      _ac_DestroyContext( rmesa->glCtx );
+      _vbo_DestroyContext( rmesa->glCtx );
       _swrast_DestroyContext( rmesa->glCtx );
 
       r200DestroySwtcl( rmesa->glCtx );
@@ -579,12 +571,6 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
 	 r200FlushCmdBuf( rmesa, __FUNCTION__ );
       }
 
-      if (!(rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)) {
-	 int tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-	 if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	    r200VtxfmtDestroy( rmesa->glCtx );
-      }
-
       if (rmesa->state.scissor.pClipRects) {
 	 FREE(rmesa->state.scissor.pClipRects);
 	 rmesa->state.scissor.pClipRects = NULL;
@@ -700,9 +686,6 @@ r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
 			  (GLframebuffer *) driDrawPriv->driverPrivate,
 			  (GLframebuffer *) driReadPriv->driverPrivate );
 
-      if (newCtx->vb.enabled)
-	 r200VtxfmtMakeCurrent( newCtx->glCtx );
-
       _mesa_update_state( newCtx->glCtx );
       r200ValidateState( newCtx->glCtx );
 
@@ -727,6 +710,5 @@ r200UnbindContext( __DRIcontextPrivate *driContextPriv )
    if (R200_DEBUG & DEBUG_DRI)
       fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->glCtx);
 
-   r200VtxfmtUnbindContext( rmesa->glCtx );
    return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index dc1fbef72e..bab767838d 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -43,7 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "framebuffer.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -54,7 +54,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
-#include "r200_vtxfmt.h"
 #include "r200_vertprog.h"
 
 #include "drirenderbuffer.h"
@@ -2536,11 +2535,10 @@ static void r200InvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
    R200_CONTEXT(ctx)->NewGLState |= new_state;
-   r200VtxfmtInvalidate( ctx );
 }
 
 /* A hack.  The r200 can actually cope just fine with materials
diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index d95a80c7bb..b40d0bdcb7 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -39,7 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_arrayelt.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -50,7 +50,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
-#include "r200_vtxfmt.h"
 
 #include "xmlpool.h"
 
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c
index c14a275f7a..25d229d8ed 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.c
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.c
@@ -48,7 +48,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "tnl/t_vtx_api.h"
 
 #include "r200_context.h"
 #include "r200_ioctl.h"
@@ -936,13 +935,6 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 }
 
 
-void r200FlushVertices( GLcontext *ctx, GLuint flags )
-{
-   _tnl_FlushVertices( ctx, flags );
-
-   if (flags & FLUSH_STORED_VERTICES)
-      R200_NEWPRIM( R200_CONTEXT( ctx ) );
-}
 
 /**********************************************************************/
 /*                            Initialization.                         */
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h
index ce2b6b5f06..ccf817988c 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.h
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.h
@@ -42,7 +42,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 extern void r200InitSwtcl( GLcontext *ctx );
 extern void r200DestroySwtcl( GLcontext *ctx );
 
-extern void r200FlushVertices( GLcontext *ctx, GLuint flags );
 extern void r200ChooseRenderState( GLcontext *ctx );
 extern void r200ChooseVertexState( GLcontext *ctx );
 
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index 86edf7132c..62c335a707 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -40,7 +40,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "colormac.h"
 #include "light.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt.c b/src/mesa/drivers/dri/r200/r200_vtxfmt.c
deleted file mode 100644
index d73fbbafd5..0000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt.c
+++ /dev/null
@@ -1,1234 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt.c,v 1.4 2003/05/06 23:52:08 daenzer Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "r200_context.h"
-#include "r200_state.h"
-#include "r200_ioctl.h"
-#include "r200_tex.h"
-#include "r200_tcl.h"
-#include "r200_swtcl.h"
-#include "r200_vtxfmt.h"
-
-#include "api_noop.h"
-#include "api_arrayelt.h"
-#include "context.h"
-#include "mtypes.h"
-#include "enums.h"
-#include "glapi.h"
-#include "colormac.h"
-#include "light.h"
-#include "state.h"
-#include "vtxfmt.h"
-
-#include "tnl/tnl.h"
-#include "tnl/t_context.h"
-#include "tnl/t_array_api.h"
-#include "tnl/t_save_api.h"
-
-#include "dispatch.h"
-
-static void r200VtxFmtFlushVertices( GLcontext *, GLuint );
-
-static void count_func( const char *name,  struct dynfn *l )
-{
-   int i = 0;
-   struct dynfn *f;
-   foreach (f, l) i++;
-   if (i) fprintf(stderr, "%s: %d\n", name, i );
-}
-
-static void count_funcs( r200ContextPtr rmesa )
-{
-   count_func( "Vertex2f", &rmesa->vb.dfn_cache.Vertex2f );
-   count_func( "Vertex2fv", &rmesa->vb.dfn_cache.Vertex2fv );
-   count_func( "Vertex3f", &rmesa->vb.dfn_cache.Vertex3f );
-   count_func( "Vertex3fv", &rmesa->vb.dfn_cache.Vertex3fv );
-   count_func( "Color4ub", &rmesa->vb.dfn_cache.Color4ub ); 
-   count_func( "Color4ubv", &rmesa->vb.dfn_cache.Color4ubv ); 
-   count_func( "Color3ub", &rmesa->vb.dfn_cache.Color3ub ); 
-   count_func( "Color3ubv", &rmesa->vb.dfn_cache.Color3ubv ); 
-   count_func( "Color4f", &rmesa->vb.dfn_cache.Color4f );
-   count_func( "Color4fv", &rmesa->vb.dfn_cache.Color4fv );
-   count_func( "Color3f", &rmesa->vb.dfn_cache.Color3f );
-   count_func( "Color3fv", &rmesa->vb.dfn_cache.Color3fv );
-   count_func( "SecondaryColor3f", &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   count_func( "SecondaryColor3fv", &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   count_func( "SecondaryColor3ub", &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   count_func( "SecondaryColor3ubv", &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   count_func( "Normal3f", &rmesa->vb.dfn_cache.Normal3f );
-   count_func( "Normal3fv", &rmesa->vb.dfn_cache.Normal3fv );
-   count_func( "TexCoord3f", &rmesa->vb.dfn_cache.TexCoord3f );
-   count_func( "TexCoord3fv", &rmesa->vb.dfn_cache.TexCoord3fv );
-   count_func( "TexCoord2f", &rmesa->vb.dfn_cache.TexCoord2f );
-   count_func( "TexCoord2fv", &rmesa->vb.dfn_cache.TexCoord2fv );
-   count_func( "TexCoord1f", &rmesa->vb.dfn_cache.TexCoord1f );
-   count_func( "TexCoord1fv", &rmesa->vb.dfn_cache.TexCoord1fv );
-   count_func( "MultiTexCoord3fARB", &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
-   count_func( "MultiTexCoord3fvARB", &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
-   count_func( "MultiTexCoord2fARB", &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   count_func( "MultiTexCoord2fvARB", &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   count_func( "MultiTexCoord1fARB", &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   count_func( "MultiTexCoord1fvARB", &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-/*   count_func( "FogCoordfEXT", &rmesa->vb.dfn_cache.FogCoordfEXT );
-   count_func( "FogCoordfvEXT", &rmesa->vb.dfn_cache.FogCoordfvEXT );*/
-}
-
-static void r200NewList( GLcontext *ctx, GLuint list, GLenum mode )
-{
-   VFMT_FALLBACK( __FUNCTION__ );
-   _tnl_NewList( ctx, list, mode );
-   return;
-}
-
-void r200_copy_to_current( GLcontext *ctx ) 
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   unsigned i;
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT);
-
-   if (rmesa->vb.vtxfmt_0 & R200_VTX_N0) {
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0] = rmesa->vb.normalptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1] = rmesa->vb.normalptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2] = rmesa->vb.normalptr[2];
-   }
-
-   if (rmesa->vb.vtxfmt_0 & R200_VTX_DISCRETE_FOG) {
-      ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = rmesa->vb.fogptr[0];
-   }
-
-   switch( VTX_COLOR(rmesa->vb.vtxfmt_0, 0) ) {
-   case R200_VTX_PK_RGBA:
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->blue );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->alpha );
-      break;
-
-   case R200_VTX_FP_RGB:
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = rmesa->vb.floatcolorptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = rmesa->vb.floatcolorptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = rmesa->vb.floatcolorptr[2];
-      break;
-
-   case R200_VTX_FP_RGBA:
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = rmesa->vb.floatcolorptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = rmesa->vb.floatcolorptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = rmesa->vb.floatcolorptr[2];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = rmesa->vb.floatcolorptr[3];
-      break;
-      
-   default:
-      break;
-   }
-      
-   if (VTX_COLOR(rmesa->vb.vtxfmt_0, 1) == R200_VTX_PK_RGBA) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] = UBYTE_TO_FLOAT( rmesa->vb.specptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] = UBYTE_TO_FLOAT( rmesa->vb.specptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] = UBYTE_TO_FLOAT( rmesa->vb.specptr->blue );
-   } 
-
-   for ( i = 0 ; i < ctx->Const.MaxTextureUnits ; i++ ) {
-      const unsigned count = VTX_TEXn_COUNT( rmesa->vb.vtxfmt_1, i );
-      GLfloat * const src = rmesa->vb.texcoordptr[i];
-
-      if ( count != 0 ) {
-	 switch( count ) {
-	 case 3:
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][1] = src[1];
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][2] = src[2];
-	    break;
-	 case 2:
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][1] = src[1];
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][2] = 0.0F;
-	    break;
-	 case 1:
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][1] = 0.0F;
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][2] = 0.0F;
-	    break;
-	 }
-
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][0] = src[0];
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][3] = 1.0F;
-      }
-   }
-
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-static GLboolean discreet_gl_prim[GL_POLYGON+1] = {
-   1,				/* 0 points */
-   1,				/* 1 lines */
-   0,				/* 2 line_strip */
-   0,				/* 3 line_loop */
-   1,				/* 4 tris */
-   0,				/* 5 tri_fan */
-   0,				/* 6 tri_strip */
-   1,				/* 7 quads */
-   0,				/* 8 quadstrip */
-   0,				/* 9 poly */
-};
-
-static void flush_prims( r200ContextPtr rmesa )
-{
-   int i,j;
-   struct r200_dma_region tmp = rmesa->dma.current;
-   
-   tmp.buf->refcount++;
-   tmp.aos_size = rmesa->vb.vertex_size;
-   tmp.aos_stride = rmesa->vb.vertex_size;
-   tmp.aos_start = GET_START(&tmp);
-
-   rmesa->dma.current.ptr = rmesa->dma.current.start += 
-      (rmesa->vb.initial_counter - rmesa->vb.counter) * 
-      rmesa->vb.vertex_size * 4; 
-
-   rmesa->tcl.vertex_format = rmesa->vb.vtxfmt_0;
-   rmesa->tcl.aos_components[0] = &tmp;
-   rmesa->tcl.nr_aos_components = 1;
-   rmesa->dma.flush = NULL;
-
-   /* Optimize the primitive list:
-    */
-   if (rmesa->vb.nrprims > 1) {
-      for (j = 0, i = 1 ; i < rmesa->vb.nrprims; i++) {
-	 int pj = rmesa->vb.primlist[j].prim & 0xf;
-	 int pi = rmesa->vb.primlist[i].prim & 0xf;
-      
-	 if (pj == pi && discreet_gl_prim[pj] &&
-	     rmesa->vb.primlist[i].start == rmesa->vb.primlist[j].end) {
-	    rmesa->vb.primlist[j].end = rmesa->vb.primlist[i].end;
-	 }
-	 else {
-	    j++;
-	    if (j != i) rmesa->vb.primlist[j] = rmesa->vb.primlist[i];
-	 }
-      }
-      rmesa->vb.nrprims = j+1;
-   }
-
-   if (rmesa->vb.vtxfmt_0 != rmesa->hw.vtx.cmd[VTX_VTXFMT_0] ||
-       rmesa->vb.vtxfmt_1 != rmesa->hw.vtx.cmd[VTX_VTXFMT_1]) { 
-      R200_STATECHANGE( rmesa, vtx ); 
-      rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = rmesa->vb.vtxfmt_0;
-      rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = rmesa->vb.vtxfmt_1;
-   } 
-
-
-   for (i = 0 ; i < rmesa->vb.nrprims; i++) {
-      if (R200_DEBUG & DEBUG_PRIMS)
-	 fprintf(stderr, "vtxfmt prim %d: %s %d..%d\n", i,
-		 _mesa_lookup_enum_by_nr( rmesa->vb.primlist[i].prim & 
-					  PRIM_MODE_MASK ),
-		 rmesa->vb.primlist[i].start,
-		 rmesa->vb.primlist[i].end);
-
-      if (rmesa->vb.primlist[i].start < rmesa->vb.primlist[i].end)
-	 r200EmitPrimitive( rmesa->glCtx,
-			    rmesa->vb.primlist[i].start,
-			    rmesa->vb.primlist[i].end,
-			    rmesa->vb.primlist[i].prim );
-   }
-
-   rmesa->vb.nrprims = 0;
-   r200ReleaseDmaRegion( rmesa, &tmp, __FUNCTION__ );
-}
-
-
-static void start_prim( r200ContextPtr rmesa, GLuint mode )
-{
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, 
-	      rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   rmesa->vb.primlist[rmesa->vb.nrprims].start = 
-      rmesa->vb.initial_counter - rmesa->vb.counter;
-   rmesa->vb.primlist[rmesa->vb.nrprims].prim = mode;
-}
-
-static void note_last_prim( r200ContextPtr rmesa, GLuint flags )
-{
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, 
-	      rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      rmesa->vb.primlist[rmesa->vb.nrprims].prim |= flags;
-      rmesa->vb.primlist[rmesa->vb.nrprims].end = 
-	 rmesa->vb.initial_counter - rmesa->vb.counter;
-
-      if (++(rmesa->vb.nrprims) == R200_MAX_PRIMS)
-	 flush_prims( rmesa );
-   }
-}
-
-
-static void copy_vertex( r200ContextPtr rmesa, GLuint n, GLfloat *dst )
-{
-   GLuint i;
-   GLfloat *src = (GLfloat *)(rmesa->dma.current.address + 
-			      rmesa->dma.current.ptr + 
-			      (rmesa->vb.primlist[rmesa->vb.nrprims].start + n) * 
-			      rmesa->vb.vertex_size * 4);
-
-   if (R200_DEBUG & DEBUG_VFMT) 
-      fprintf(stderr, "copy_vertex %d\n", rmesa->vb.primlist[rmesa->vb.nrprims].start + n);
-
-   for (i = 0 ; i < rmesa->vb.vertex_size; i++) {
-      dst[i] = src[i];
-   }
-}
-
-/* NOTE: This actually reads the copied vertices back from uncached
- * memory.  Could also use the counter/notify mechanism to populate
- * tmp on the fly as vertices are generated.  
- */
-static GLuint copy_dma_verts( r200ContextPtr rmesa, GLfloat (*tmp)[R200_MAX_VERTEX_SIZE] )
-{
-   GLuint ovf, i;
-   GLuint nr = (rmesa->vb.initial_counter - rmesa->vb.counter) - 
-      rmesa->vb.primlist[rmesa->vb.nrprims].start;
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d verts\n", __FUNCTION__, nr);
-
-   switch( rmesa->vb.prim[0] )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      copy_vertex( rmesa, nr-1, tmp[0] );
-      return 1;
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 return 1;
-      } else {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 copy_vertex( rmesa, nr-1, tmp[1] );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      ovf = MIN2( nr, 2 );
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-static void VFMT_FALLBACK_OUTSIDE_BEGIN_END( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (ctx->Driver.NeedFlush) 
-      r200VtxFmtFlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx ); /* clear state so fell_back sticks */
-
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = r200FlushVertices;
-   ctx->Driver.NewList = _tnl_NewList;
-
-   assert( rmesa->dma.flush == 0 );
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-}
-
-
-/**
- * \todo
- * An interesting optimization of this function would be to have 3 element
- * table with the dispatch offsets of the TexCoord?fv functions, use count
- * to look-up the table, and a specialized version of GL_CALL that used the
- * offset number instead of the name.
- */
-static void dispatch_multitexcoord( GLuint count, GLuint unit, GLfloat * f )
-{
-   switch( count ) {
-   case 3:
-      CALL_MultiTexCoord3fvARB(GET_DISPATCH(), (GL_TEXTURE0+unit, f));
-      break;
-   case 2:
-      CALL_MultiTexCoord2fvARB(GET_DISPATCH(), (GL_TEXTURE0+unit, f));
-      break;
-   case 1:
-      CALL_MultiTexCoord1fvARB(GET_DISPATCH(), (GL_TEXTURE0+unit, f));
-      break;
-   default:
-      assert( count == 0 );
-      break;
-   }
-}
-
-void VFMT_FALLBACK( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat tmp[3][R200_MAX_VERTEX_SIZE];
-   GLuint i, prim;
-   GLuint ind0 = rmesa->vb.vtxfmt_0;
-   GLuint ind1 = rmesa->vb.vtxfmt_1;
-   GLuint nrverts;
-   GLfloat alpha = 1.0;
-   GLuint count;
-   GLuint unit;
-
-   if (R200_DEBUG & (DEBUG_FALLBACKS|DEBUG_VFMT))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      VFMT_FALLBACK_OUTSIDE_BEGIN_END( __FUNCTION__ );
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   nrverts = copy_dma_verts( rmesa, tmp );
-
-   /* Finish the prim at this point:
-    */
-   note_last_prim( rmesa, 0 );
-   flush_prims( rmesa );
-
-   /* Update ctx->Driver.CurrentExecPrimitive and swap in swtnl. 
-    */
-   prim = rmesa->vb.prim[0];
-   ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1;
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = r200FlushVertices;
-
-   assert(rmesa->dma.flush == 0);
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-   CALL_Begin(GET_DISPATCH(), (prim));
-
-   if (rmesa->vb.installed_color_3f_sz == 4)
-      alpha = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-
-   /* Replay saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      GLuint offset = 3;
-
-      if (ind0 & R200_VTX_N0) {
-	 CALL_Normal3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset += 3;
-      }
-
-      if (ind0 & R200_VTX_DISCRETE_FOG) {
-	 CALL_FogCoordfvEXT(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset++;
-      }
-
-      if (VTX_COLOR(ind0, 0) == R200_VTX_PK_RGBA) {
-	 CALL_Color4ubv(GET_DISPATCH(), ((GLubyte *)&tmp[i][offset]));
-	 offset++;
-      }
-      else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGBA) {
-	 CALL_Color4fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=4;
-      } 
-      else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGB) {
-	 CALL_Color3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=3;
-      }
-
-      if (VTX_COLOR(ind0, 1) == R200_VTX_PK_RGBA) {
-	 CALL_SecondaryColor3ubvEXT(GET_DISPATCH(), ((GLubyte *)&tmp[i][offset]));
-	 offset++;
-      }
-
-      for ( unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++ ) {
-	 count = VTX_TEXn_COUNT( ind1, unit );
-	 dispatch_multitexcoord( count, unit, &tmp[i][offset] );
-	 offset += count;
-      }
-
-      CALL_Vertex3fv(GET_DISPATCH(), (&tmp[i][0]));
-   }
-
-   /* Replay current vertex
-    */
-   if (ind0 & R200_VTX_N0) 
-      CALL_Normal3fv(GET_DISPATCH(), (rmesa->vb.normalptr));
-   if (ind0 & R200_VTX_DISCRETE_FOG) {
-      CALL_FogCoordfvEXT(GET_DISPATCH(), (rmesa->vb.fogptr));
-   }
-
-   if (VTX_COLOR(ind0, 0) == R200_VTX_PK_RGBA) {
-      CALL_Color4ub(GET_DISPATCH(), (rmesa->vb.colorptr->red,
-				     rmesa->vb.colorptr->green,
-				     rmesa->vb.colorptr->blue,
-				     rmesa->vb.colorptr->alpha));
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGBA) {
-      CALL_Color4fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGB) {
-      if (rmesa->vb.installed_color_3f_sz == 4 && alpha != 1.0) {
-	 CALL_Color4f(GET_DISPATCH(), (rmesa->vb.floatcolorptr[0],
-				       rmesa->vb.floatcolorptr[1],
-				       rmesa->vb.floatcolorptr[2],
-				       alpha));
-      }
-      else {
-	 CALL_Color3fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-      }
-   }
-
-   if (VTX_COLOR(ind0, 1) == R200_VTX_PK_RGBA) 
-       CALL_SecondaryColor3ubEXT(GET_DISPATCH(), (rmesa->vb.specptr->red, 
-						  rmesa->vb.specptr->green,
-						  rmesa->vb.specptr->blue));
-
-   for ( unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++ ) {
-      count = VTX_TEXn_COUNT( ind1, unit );
-      dispatch_multitexcoord( count, unit, rmesa->vb.texcoordptr[unit] );
-   }
-}
-
-
-
-static void wrap_buffer( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat tmp[3][R200_MAX_VERTEX_SIZE];
-   GLuint i, nrverts;
-
-   if (R200_DEBUG & (DEBUG_VFMT|DEBUG_PRIMS))
-      fprintf(stderr, "%s %d\n", __FUNCTION__,
-	      rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   /* Don't deal with parity.
-    */
-   if ((((rmesa->vb.initial_counter - rmesa->vb.counter) -  
-	 rmesa->vb.primlist[rmesa->vb.nrprims].start) & 1)) {
-      rmesa->vb.counter++;
-      rmesa->vb.initial_counter++;
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) 
-      nrverts = 0;
-   else {
-      nrverts = copy_dma_verts( rmesa, tmp );
-
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%d vertices to copy\n", nrverts);
-   
-      /* Finish the prim at this point:
-       */
-      note_last_prim( rmesa, 0 );
-   }
-
-   /* Fire any buffered primitives
-    */
-   flush_prims( rmesa );
-
-   /* Get new buffer
-    */
-   r200RefillCurrentDmaRegion( rmesa );
-
-   /* Reset counter, dmaptr
-    */
-   rmesa->vb.dmaptr = (int *)(rmesa->dma.current.ptr + rmesa->dma.current.address);
-   rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-      (rmesa->vb.vertex_size * 4);
-   rmesa->vb.counter--;
-   rmesa->vb.initial_counter = rmesa->vb.counter;
-   rmesa->vb.notify = wrap_buffer;
-
-   rmesa->dma.flush = flush_prims;
-
-   /* Restart wrapped primitive:
-    */
-   if (rmesa->vb.prim[0] != GL_POLYGON+1)
-      start_prim( rmesa, rmesa->vb.prim[0] );
-
-
-   /* Reemit saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      if (R200_DEBUG & DEBUG_VERTS) {
-	 int j;
-	 fprintf(stderr, "re-emit vertex %d to %p\n", i, 
-		 (void *)rmesa->vb.dmaptr);
-	 if (R200_DEBUG & DEBUG_VERBOSE)
-	    for (j = 0 ; j < rmesa->vb.vertex_size; j++) 
-	       fprintf(stderr, "\t%08x/%f\n", *(int*)&tmp[i][j], tmp[i][j]);
-      }
-
-      memcpy( rmesa->vb.dmaptr, tmp[i], rmesa->vb.vertex_size * 4 );
-      rmesa->vb.dmaptr += rmesa->vb.vertex_size;
-      rmesa->vb.counter--;
-   }
-}
-
-
-/**
- * Determines the hardware vertex format based on the current state vector.
- * 
- * \returns
- * If the hardware TCL unit is capable of handling the current state vector,
- * \c GL_TRUE is returned.  Otherwise, \c GL_FALSE is returned.
- *
- * \todo
- * Make this color format selection data driven.  If we receive only ubytes,
- * send color as ubytes.  Also check if converting (with free checking for
- * overflow) is cheaper than sending floats directly.
- *
- * \todo
- * When intializing texture coordinates, it might be faster to just copy the
- * entire \c VERT_ATTRIB_TEX0 vector into the vertex buffer.  It may mean that
- * some of the data (i.e., the last texture coordinate components) get copied
- * over, but that still may be faster than the conditional branching.  If
- * nothing else, the code will be smaller and easier to follow.
- */
-static GLboolean check_vtx_fmt( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint ind0 = R200_VTX_Z0;
-   GLuint ind1 = 0;
-   GLuint i;
-   GLuint count[R200_MAX_TEXTURE_UNITS];
-
-   if (rmesa->TclFallback || rmesa->vb.fell_back || ctx->CompileFlag ||
-      (ctx->Fog.Enabled && (ctx->Fog.FogCoordinateSource == GL_FOG_COORD)) ||
-      /* TODO: set tcl out fmt/compsel and reenable vtxfmt code */
-      ctx->VertexProgram._Enabled)
-      return GL_FALSE;
-
-   if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) 
-      ctx->Driver.FlushVertices( ctx, FLUSH_UPDATE_CURRENT );
-   
-   /* Make all this event-driven:
-    */
-   if (ctx->Light.Enabled) {
-      ind0 |= R200_VTX_N0;
-
-      if (ctx->Light.ColorMaterialEnabled) 
-	 ind0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT;
-      else
-	 ind0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT;
-   }
-   else {
-      /* TODO: make this data driven?
-       */
-      ind0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT;
-	 
-      if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
-	 ind0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT;
-      }
-   }
-
-   if ( ctx->Fog.FogCoordinateSource == GL_FOG_COORD ) {
-      ind0 |= R200_VTX_DISCRETE_FOG;
-   }
-
-   for ( i = 0 ; i < ctx->Const.MaxTextureUnits ; i++ ) {
-      count[i] = 0;
-
-      if (ctx->Texture.Unit[i]._ReallyEnabled) {
-	 if (rmesa->TexGenNeedNormals[i]) {
-	    ind0 |= R200_VTX_N0;
-	 }
-	 else {
-	    switch( ctx->Texture.Unit[i]._ReallyEnabled ) {
-	    case TEXTURE_CUBE_BIT:
-	    case TEXTURE_3D_BIT:
-	       count[i] = 3;
-	       break;
-	    case TEXTURE_2D_BIT:
-	    case TEXTURE_RECT_BIT:
-	       count[i] = 2;
-	       break;
-	    case TEXTURE_1D_BIT:
-	       count[i] = 1;
-	       break;
-	    }
-
-	    ind1 |= count[i] << (3 * i);
-	 }
-      }
-   }
-
-   if (R200_DEBUG & (DEBUG_VFMT|DEBUG_STATE))
-      fprintf(stderr, "%s: format: 0x%x, 0x%x\n", __FUNCTION__, ind0, ind1 );
-
-   R200_NEWPRIM(rmesa);
-   rmesa->vb.vtxfmt_0 = ind0;
-   rmesa->vb.vtxfmt_1 = ind1;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-
-   rmesa->vb.vertex_size = 3;
-   rmesa->vb.normalptr = ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
-   rmesa->vb.colorptr = NULL;
-   rmesa->vb.floatcolorptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-   rmesa->vb.fogptr = ctx->Current.Attrib[VERT_ATTRIB_FOG];
-   rmesa->vb.specptr = NULL;
-   rmesa->vb.floatspecptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
-   rmesa->vb.texcoordptr[0] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];
-   rmesa->vb.texcoordptr[1] = ctx->Current.Attrib[VERT_ATTRIB_TEX1];
-   rmesa->vb.texcoordptr[2] = ctx->Current.Attrib[VERT_ATTRIB_TEX2];
-   rmesa->vb.texcoordptr[3] = ctx->Current.Attrib[VERT_ATTRIB_TEX3];
-   rmesa->vb.texcoordptr[4] = ctx->Current.Attrib[VERT_ATTRIB_TEX4];
-   rmesa->vb.texcoordptr[5] = ctx->Current.Attrib[VERT_ATTRIB_TEX5];
-   rmesa->vb.texcoordptr[6] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];	/* dummy */
-   rmesa->vb.texcoordptr[7] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];	/* dummy */
-
-   /* Run through and initialize the vertex components in the order
-    * the hardware understands:
-    */
-   if (ind0 & R200_VTX_N0) {
-      rmesa->vb.normalptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.normalptr[0] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0];
-      rmesa->vb.normalptr[1] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1];
-      rmesa->vb.normalptr[2] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2];
-   }
-
-   if (ind0 & R200_VTX_DISCRETE_FOG) {
-      rmesa->vb.fogptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 1;
-      rmesa->vb.fogptr[0] = ctx->Current.Attrib[VERT_ATTRIB_FOG][0];
-   }
-
-   if (VTX_COLOR(ind0, 0) == R200_VTX_PK_RGBA) {
-      rmesa->vb.colorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->alpha, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] );
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGBA) {
-      rmesa->vb.floatcolorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 4;
-      rmesa->vb.floatcolorptr[0] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0];
-      rmesa->vb.floatcolorptr[1] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1];
-      rmesa->vb.floatcolorptr[2] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2];
-      rmesa->vb.floatcolorptr[3] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGB) {
-      rmesa->vb.floatcolorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.floatcolorptr[0] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0];
-      rmesa->vb.floatcolorptr[1] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1];
-      rmesa->vb.floatcolorptr[2] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2];
-   }   
-   
-   if (VTX_COLOR(ind0, 1) == R200_VTX_PK_RGBA) {
-      rmesa->vb.specptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] );
-   }
-
-
-   for ( i = 0 ; i < ctx->Const.MaxTextureUnits ; i++ ) {
-      if ( count[i] != 0 ) {
-	 float * const attr = ctx->Current.Attrib[VERT_ATTRIB_TEX0+i];
-	 unsigned  j;
-
-	 rmesa->vb.texcoordptr[i] = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-
-	 for ( j = 0 ; j < count[i] ; j++ ) {
-	    rmesa->vb.texcoordptr[i][j] = attr[j];
-	 }
-
-	 rmesa->vb.vertex_size += count[i];
-      }
-   }
-
-   if (rmesa->vb.installed_vertex_format != rmesa->vb.vtxfmt_0) {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on vertex_format change\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      rmesa->vb.installed_vertex_format = rmesa->vb.vtxfmt_0;
-   }
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s -- success\n", __FUNCTION__);
-
-   return GL_TRUE;
-}
-
-
-void r200VtxfmtInvalidate( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   rmesa->vb.recheck = GL_TRUE;
-   rmesa->vb.fell_back = GL_FALSE;
-}
-
-
-static void r200VtxfmtValidate( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   rmesa->vb.recheck = GL_FALSE;
-
-   if (check_vtx_fmt( ctx )) {
-      if (!rmesa->vb.installed) {
-	 if (R200_DEBUG & DEBUG_VFMT)
-	    fprintf(stderr, "reinstall (new install)\n");
-
-	 _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-	 ctx->Driver.FlushVertices = r200VtxFmtFlushVertices;
-	 ctx->Driver.NewList = r200NewList;
-	 rmesa->vb.installed = GL_TRUE;
-      }
-      else if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: already installed", __FUNCTION__);
-   } 
-   else {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: failed\n", __FUNCTION__);
-
-      if (rmesa->vb.installed) {
-	 if (rmesa->dma.flush)
-	    rmesa->dma.flush( rmesa );
-	 _tnl_wakeup_exec( ctx );
-	 ctx->Driver.FlushVertices = r200FlushVertices;
-	 ctx->Driver.NewList =_tnl_NewList;
-	 rmesa->vb.installed = GL_FALSE;
-      }
-   }      
-}
-
-
-
-/* Materials:
- */
-static void r200_Materialfv( GLenum face, GLenum pname, 
-			       const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      VFMT_FALLBACK( __FUNCTION__ );
-      CALL_Materialfv(GET_DISPATCH(), (face, pname, params));
-      return;
-   }
-   _mesa_noop_Materialfv( face, pname, params );
-   r200UpdateMaterial( ctx );
-}
-
-
-/* Begin/End
- */
-static void r200_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s( %s )\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   if (mode > GL_POLYGON) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glBegin" );
-      return;
-   }
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-      return;
-   }
-   
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
-
-   if (rmesa->vb.recheck) 
-      r200VtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      CALL_Begin(GET_DISPATCH(), (mode));
-      return;
-   }
-
-
-   if (rmesa->dma.flush && rmesa->vb.counter < 12) {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: flush almost-empty buffers\n", __FUNCTION__);
-      flush_prims( rmesa );
-   }
-
-   /* Need to arrange to save vertices here?  Or always copy from dma (yuk)?
-    */
-   if (!rmesa->dma.flush) {
-      if (rmesa->dma.current.ptr + 12*rmesa->vb.vertex_size*4 > 
-	  rmesa->dma.current.end) {
-	 R200_NEWPRIM( rmesa );
-	 r200RefillCurrentDmaRegion( rmesa );
-      }
-
-      rmesa->vb.dmaptr = (int *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-	 (rmesa->vb.vertex_size * 4);
-      rmesa->vb.counter--;
-      rmesa->vb.initial_counter = rmesa->vb.counter;
-      rmesa->vb.notify = wrap_buffer;
-      rmesa->dma.flush = flush_prims;
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   }
-   
-   
-   rmesa->vb.prim[0] = mode;
-   start_prim( rmesa, mode | PRIM_BEGIN );
-}
-
-
-
-static void r200_End( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-      return;
-   }
-
-   note_last_prim( rmesa, PRIM_END );
-   rmesa->vb.prim[0] = GL_POLYGON+1;
-}
-
-
-/* Fallback on difficult entrypoints:
- */
-#define PRE_LOOPBACK( FUNC )			\
-do {						\
-   if (R200_DEBUG & DEBUG_VFMT) 		\
-      fprintf(stderr, "%s\n", __FUNCTION__);	\
-   VFMT_FALLBACK( __FUNCTION__ );		\
-} while (0)
-#define TAG(x) r200_fallback_##x
-#include "vtxfmt_tmp.h"
-
-
-
-static GLboolean r200NotifyBegin( GLcontext *ctx, GLenum p )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(!rmesa->vb.installed);
-
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (rmesa->vb.recheck) 
-      r200VtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s -- failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   r200_Begin( p );
-   return GL_TRUE;
-}
-
-static void r200VtxFmtFlushVertices( GLcontext *ctx, GLuint flags )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(rmesa->vb.installed);
-
-   if (flags & FLUSH_UPDATE_CURRENT) {
-      r200_copy_to_current( ctx );
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on update_current\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-   }
-
-   if (flags & FLUSH_STORED_VERTICES) {
-      assert (rmesa->dma.flush == 0 ||
-	      rmesa->dma.flush == flush_prims);
-      if (rmesa->dma.flush == flush_prims)
-	 flush_prims( rmesa );
-      ctx->Driver.NeedFlush &= ~FLUSH_STORED_VERTICES;
-   }
-}
-
-
-
-/* At this point, don't expect very many versions of each function to
- * be generated, so not concerned about freeing them?
- */
-
-
-/**
- * Called once during context creation.
- */
-void r200VtxfmtInit( GLcontext *ctx, GLboolean useCodegen )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   GLvertexformat *vfmt = &(rmesa->vb.vtxfmt);
-
-   /* start by initializing to no-op functions */
-   _mesa_noop_vtxfmt_init(vfmt);
-
-   /* Hook in chooser functions for codegen, etc:
-    */
-   r200VtxfmtInitChoosers( vfmt );
-
-   /* Handled fully in supported states, but no codegen:
-    */
-   vfmt->Materialfv = r200_Materialfv;
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = r200_Begin;
-   vfmt->End = r200_End;
-
-   /* Fallback for performance reasons:  (Fix with cva/elt path here and
-    * dmatmp2.h style primitive-merging)
-    *
-    * These should call NotifyBegin(), as should _tnl_EvalMesh, to allow
-    * a driver-hook.
-    */
-   vfmt->DrawArrays = r200_fallback_DrawArrays;
-   vfmt->DrawElements = r200_fallback_DrawElements;
-   vfmt->DrawRangeElements = r200_fallback_DrawRangeElements; 
-
-   /* Active but unsupported -- fallback if we receive these:
-    */
-   vfmt->CallList = r200_fallback_CallList;
-   vfmt->CallLists = r200_fallback_CallLists;
-   vfmt->EvalCoord1f = r200_fallback_EvalCoord1f;
-   vfmt->EvalCoord1fv = r200_fallback_EvalCoord1fv;
-   vfmt->EvalCoord2f = r200_fallback_EvalCoord2f;
-   vfmt->EvalCoord2fv = r200_fallback_EvalCoord2fv;
-   vfmt->EvalMesh1 = r200_fallback_EvalMesh1;
-   vfmt->EvalMesh2 = r200_fallback_EvalMesh2;
-   vfmt->EvalPoint1 = r200_fallback_EvalPoint1;
-   vfmt->EvalPoint2 = r200_fallback_EvalPoint2;
-   vfmt->TexCoord4f = r200_fallback_TexCoord4f;
-   vfmt->TexCoord4fv = r200_fallback_TexCoord4fv;
-   vfmt->MultiTexCoord4fARB = r200_fallback_MultiTexCoord4fARB;
-   vfmt->MultiTexCoord4fvARB = r200_fallback_MultiTexCoord4fvARB;
-   vfmt->Vertex4f = r200_fallback_Vertex4f;
-   vfmt->Vertex4fv = r200_fallback_Vertex4fv;
-   vfmt->VertexAttrib1fNV  = r200_fallback_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = r200_fallback_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV  = r200_fallback_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = r200_fallback_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV  = r200_fallback_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = r200_fallback_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV  = r200_fallback_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = r200_fallback_VertexAttrib4fvNV;
-   vfmt->FogCoordfEXT = r200_fallback_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = r200_fallback_FogCoordfvEXT;
-   
-   (void)r200_fallback_vtxfmt;
-
-   TNL_CONTEXT(ctx)->Driver.NotifyBegin = r200NotifyBegin;
-
-   rmesa->vb.enabled = 1;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-   rmesa->vb.primflags = 0;
-
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord3f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1fv );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-/*   make_empty_list( &rmesa->vb.dfn_cache.FogCoordfEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.FogCoordfvEXT );*/
-
-   r200InitCodegen( &rmesa->vb.codegen, useCodegen );
-}
-
-static void free_funcs( struct dynfn *l )
-{
-   struct dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      _mesa_exec_free( f->code );
-      _mesa_free( f );
-   }
-}
-
-void r200VtxfmtUnbindContext( GLcontext *ctx )
-{
-}
-
-
-void r200VtxfmtMakeCurrent( GLcontext *ctx )
-{
-}
-
-
-void r200VtxfmtDestroy( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   count_funcs( rmesa );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2fv );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4f );
-   free_funcs( &rmesa->vb.dfn_cache.Color4fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3f );
-   free_funcs( &rmesa->vb.dfn_cache.Color3fv );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3f );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord3f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord3fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1fv );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-/*   free_funcs( &rmesa->vb.dfn_cache.FogCoordfEXT );
-   free_funcs( &rmesa->vb.dfn_cache.FogCoordfvEXT );*/
-}
-
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt.h b/src/mesa/drivers/dri/r200/r200_vtxfmt.h
deleted file mode 100644
index 46999191e0..0000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt.h
+++ /dev/null
@@ -1,123 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt.h,v 1.1 2002/10/30 12:51:53 alanh Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_VTXFMT_H__
-#define __R200_VTXFMT_H__
-
-#include "r200_context.h"
-
-
-
-extern void r200VtxfmtUpdate( GLcontext *ctx );
-extern void r200VtxfmtInit( GLcontext *ctx, GLboolean useCodegen );
-extern void r200VtxfmtInvalidate( GLcontext *ctx );
-extern void r200VtxfmtDestroy( GLcontext *ctx );
-extern void r200VtxfmtInitChoosers( GLvertexformat *vfmt );
-
-extern void r200VtxfmtMakeCurrent( GLcontext *ctx );
-extern void r200VtxfmtUnbindContext( GLcontext *ctx );
-
-extern void r200_copy_to_current( GLcontext *ctx );
-extern void VFMT_FALLBACK( const char *caller );
-
-#define DFN( FUNC, CACHE)				\
-do {							\
-   char *start = (char *)&FUNC;				\
-   char *end = (char *)&FUNC##_end;			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key[0] = key[0];				\
-   dfn->key[1] = key[1];				\
-   dfn->code = _mesa_exec_malloc( end - start );	\
-   _mesa_memcpy(dfn->code, start, end - start);		\
-}							\
-while ( 0 )
-
-#define FIXUP( CODE, OFFSET, CHECKVAL, NEWVAL )	\
-do {						\
-   int *icode = (int *)(CODE+OFFSET);		\
-   assert (*icode == CHECKVAL);			\
-   *icode = (int)NEWVAL;			\
-} while (0)
-
-
-/* Useful for figuring out the offsets:
- */
-#define FIXUP2( CODE, OFFSET, CHECKVAL, NEWVAL )		\
-do {								\
-   while (*(int *)(CODE+OFFSET) != CHECKVAL) OFFSET++;		\
-   /*fprintf(stderr, "%s/%d CVAL %x OFFSET %d VAL %x\n", __FUNCTION__,*/ \
-   /*	   __LINE__, CHECKVAL, OFFSET, (int)(NEWVAL));*/		\
-   *(int *)(CODE+OFFSET) = (int)(NEWVAL);				\
-   OFFSET += 4;							\
-} while (0)
-
-/* 
- */
-void r200InitCodegen( struct dfn_generators *gen, GLboolean useCodegen );
-void r200InitX86Codegen( struct dfn_generators *gen );
-void r200InitSSECodegen( struct dfn_generators *gen );
-
-
-
-/* Defined in r200_vtxfmt_x86.c
- */
-struct dynfn *r200_makeX86Vertex2f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Vertex2fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Vertex3f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Vertex3fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4ub( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4ubv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3ub( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3ubv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3ubEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3ubvEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3fEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3fvEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86Normal3f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Normal3fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord2f( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord2fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord1f( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord1fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord2fARB( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord2fvARB( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord1fARB( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord1fvARB( GLcontext *, const int * );
-
-#endif
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt_c.c b/src/mesa/drivers/dri/r200/r200_vtxfmt_c.c
deleted file mode 100644
index 1db5950c8f..0000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt_c.c
+++ /dev/null
@@ -1,1002 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt_c.c,v 1.2 2002/12/16 16:18:56 dawes Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "mtypes.h"
-#include "colormac.h"
-#include "simple_list.h"
-#include "api_noop.h"
-#include "vtxfmt.h"
-
-#include "r200_vtxfmt.h"
-#include "r200_tcl.h"
-
-#include "dispatch.h"
-
-/* Fallback versions of all the entrypoints for situations where
- * codegen isn't available.  This is still a lot faster than the
- * vb/pipeline implementation in Mesa.
- */
-static void r200_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = *(int *)&z;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void r200_Vertex3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = *(int *)&v[2];
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void r200_Vertex2f( GLfloat x, GLfloat y )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-   
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void r200_Vertex2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-
-/* Color for ubyte (packed) color formats:
- */
-#if 0
-static void r200_Color3ub_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void r200_Color3ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-
-static void r200_Color4ub_ub( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= a;
-}
-
-static void r200_Color4ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   *(GLuint *)rmesa->vb.colorptr = LE32_TO_CPU(*(GLuint *)v);
-}
-#endif /* 0 */
-
-static void r200_Color3f_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void r200_Color3fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-static void r200_Color4f_ub( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, a );
-}
-
-static void r200_Color4fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, v[3] );
-}
-
-
-/* Color for float color+alpha formats:
- */
-#if 0
-static void r200_Color3ub_4f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void r200_Color3ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-
-static void r200_Color4ub_4f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = UBYTE_TO_FLOAT(a);
-}
-
-static void r200_Color4ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void r200_Color3f_4f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;		
-}
-
-static void r200_Color3fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-static void r200_Color4f_4f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = a;
-}
-
-static void r200_Color4fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = v[3];
-}
-
-
-/* Color for float color formats:
- */
-#if 0
-static void r200_Color3ub_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-}
-
-static void r200_Color3ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-}
-
-static void r200_Color4ub_3f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(a);
-}
-
-static void r200_Color4ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void r200_Color3f_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-}
-
-static void r200_Color3fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-static void r200_Color4f_3f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = a;
-}
-
-static void r200_Color4fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = v[3]; 
-}
-
-
-/* Secondary Color:
- */
-#if 0
-static void r200_SecondaryColor3ubEXT_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void r200_SecondaryColor3ubvEXT_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-#endif /* 0 */
-
-static void r200_SecondaryColor3fEXT_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void r200_SecondaryColor3fvEXT_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-#if 0
-static void r200_SecondaryColor3ubEXT_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void r200_SecondaryColor3ubvEXT_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-#endif /* 0 */
-
-static void r200_SecondaryColor3fEXT_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;
-}
-
-static void r200_SecondaryColor3fvEXT_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-
-
-/* Normal
- */
-static void r200_Normal3f( GLfloat n0, GLfloat n1, GLfloat n2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = n0;
-   dest[1] = n1;
-   dest[2] = n2;
-}
-
-static void r200_Normal3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-
-/* FogCoord
- */
-static void r200_FogCoordfEXT( GLfloat f )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.fogptr;
-   dest[0] = r200ComputeFogBlendFactor( ctx, f );
-/*   ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = f;*/
-}
-
-static void r200_FogCoordfvEXT( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.fogptr;
-   dest[0] = r200ComputeFogBlendFactor( ctx, v[0] );
-/*   ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = v[0];*/
-}
-
-
-/* TexCoord
- */
-
-/* \todo maybe (target & 4 ? target & 5 : target & 3) is more save than (target & 7) */
-static void r200_MultiTexCoord1fARB(GLenum target, GLfloat s)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLint unit = (target & 7);
-   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
-
-   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
-   case TEXTURE_CUBE_BIT:
-   case TEXTURE_3D_BIT:
-      dest[2] = 0.0;
-      /* FALLTHROUGH */
-   case TEXTURE_2D_BIT:
-   case TEXTURE_RECT_BIT:
-      dest[1] = 0.0;
-      /* FALLTHROUGH */
-   case TEXTURE_1D_BIT:
-      dest[0] = s;
-   }
-}
-
-static void r200_MultiTexCoord2fARB(GLenum target, GLfloat s, GLfloat t)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLint unit = (target & 7);
-   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
-
-   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
-   case TEXTURE_CUBE_BIT:
-   case TEXTURE_3D_BIT:
-      dest[2] = 0.0;
-      /* FALLTHROUGH */
-   case TEXTURE_2D_BIT:
-   case TEXTURE_RECT_BIT:
-      dest[1] = t;
-      dest[0] = s;
-      break;
-   default:
-      VFMT_FALLBACK(__FUNCTION__);
-      CALL_MultiTexCoord2fARB(GET_DISPATCH(), (target, s, t));
-      return;	
-   }
-}
-
-static void r200_MultiTexCoord3fARB(GLenum target, GLfloat s, GLfloat t, GLfloat r)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLint unit = (target & 7);
-   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
-
-   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
-   case TEXTURE_CUBE_BIT:
-   case TEXTURE_3D_BIT:
-      dest[2] = r;
-      dest[1] = t;
-      dest[0] = s;
-      break;
-   default:
-      VFMT_FALLBACK(__FUNCTION__);
-      CALL_MultiTexCoord3fARB(GET_DISPATCH(), (target, s, t, r));
-      return;	
-   }
-}
-
-static void r200_TexCoord1f(GLfloat s)
-{
-   r200_MultiTexCoord1fARB(GL_TEXTURE0, s);
-}
-
-static void r200_TexCoord2f(GLfloat s, GLfloat t)
-{
-   r200_MultiTexCoord2fARB(GL_TEXTURE0, s, t);
-}
-
-static void r200_TexCoord3f(GLfloat s, GLfloat t, GLfloat r)
-{
-   r200_MultiTexCoord3fARB(GL_TEXTURE0, s, t, r);
-}
-
-static void r200_TexCoord1fv(const GLfloat *v)
-{
-   r200_MultiTexCoord1fARB(GL_TEXTURE0, v[0]);
-}
-
-static void r200_TexCoord2fv(const GLfloat *v)
-{
-   r200_MultiTexCoord2fARB(GL_TEXTURE0, v[0], v[1]);
-}
-
-static void r200_TexCoord3fv(const GLfloat *v)
-{
-   r200_MultiTexCoord3fARB(GL_TEXTURE0, v[0], v[1], v[2]);
-}
-
-static void r200_MultiTexCoord1fvARB(GLenum target, const GLfloat *v)
-{
-   r200_MultiTexCoord1fARB(target, v[0]);
-}
-
-static void r200_MultiTexCoord2fvARB(GLenum target, const GLfloat *v)
-{
-   r200_MultiTexCoord2fARB(target, v[0], v[1]);
-}
-
-static void r200_MultiTexCoord3fvARB(GLenum target, const GLfloat *v)
-{
-   r200_MultiTexCoord3fARB(target, v[0], v[1], v[2]);
-}
-
-
-static struct dynfn *lookup( struct dynfn *l, const int *key )
-{
-   struct dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key[0] == key[0] && f->key[1] == key[1]) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-/* Can't use the loopback template for this:
- */
-
-#define CHOOSE(FN, FNTYPE, MASK0, MASK1, ARGS1, ARGS2 )			\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);				\
-   int key[2];								\
-   struct dynfn *dfn;							\
-									\
-   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
-   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );				\
-   else if (R200_DEBUG & DEBUG_CODEGEN)					\
-      fprintf(stderr, "%s -- cached codegen\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (R200_DEBUG & DEBUG_CODEGEN)					\
-	 fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, r200_##FN);				\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* For the _3f case, only allow one color function to be hooked in at
- * a time.  Eventually, use a similar mechanism to allow selecting the
- * color component of the vertex format based on client behaviour.  
- *
- * Note:  Perform these actions even if there is a codegen or cached 
- * codegen version of the chosen function.
- */
-#define CHOOSE_COLOR(FN, FNTYPE, NR, MASK0, MASK1, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
-   int key[2];								\
-   struct dynfn *dfn;							\
-									\
-   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
-   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
-									\
-   if (VTX_COLOR(rmesa->vb.vtxfmt_0,0) == R200_VTX_PK_RGBA) {		\
-      SET_ ## FN (ctx->Exec, r200_##FN##_ub);				\
-   }									\
-   else if (VTX_COLOR(rmesa->vb.vtxfmt_0,0) == R200_VTX_FP_RGB) {	\
-									\
-      if (rmesa->vb.installed_color_3f_sz != NR) {			\
-         rmesa->vb.installed_color_3f_sz = NR;				\
-         if (NR == 3) ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = 1.0;	\
-         if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) {		\
-            r200_copy_to_current( ctx );				\
-            _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );	\
-            CALL_ ## FN (ctx->Exec, ARGS2);				\
-            return;							\
-         }								\
-      }									\
-									\
-      SET_ ## FN (ctx->Exec, r200_##FN##_3f);				\
-   }									\
-   else {								\
-      SET_ ## FN (ctx->Exec, r200_##FN##_4f);				\
-   }									\
-									\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (!dfn) dfn = rmesa->vb.codegen.FN( ctx, key );			\
-									\
-   if (dfn) {								\
-      if (R200_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- codegen version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, (FNTYPE)dfn->code);			\
-   }									\
-   else if (R200_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- 'c' version\n", __FUNCTION__ );		\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* Right now there are both _ub and _3f versions of the secondary color
- * functions.  Currently, we only set-up the hardware to use the _ub versions.
- * The _3f versions are needed for the cases where secondary color isn't used
- * in the vertex format, but it still needs to be stored in the context
- * state vector.
- */
-#define CHOOSE_SECONDARY_COLOR(FN, FNTYPE, MASK0, MASK1, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
-   int key[2];								\
-   struct dynfn *dfn;							\
-									\
-   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
-   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );			\
-   else  if (R200_DEBUG & DEBUG_CODEGEN)				\
-      fprintf(stderr, "%s -- cached version\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (R200_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, (VTX_COLOR(rmesa->vb.vtxfmt_0,1) == R200_VTX_PK_RGBA) \
-	  ? r200_##FN##_ub : r200_##FN##_3f);				\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;		\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-
-
-
-
-/* VTXFMT_0
- */
-#define MASK_XYZW  (R200_VTX_W0|R200_VTX_Z0)
-#define MASK_NORM  (MASK_XYZW|R200_VTX_N0)
-#define MASK_FOG   (MASK_NORM |R200_VTX_DISCRETE_FOG)
-#define MASK_COLOR (MASK_FOG |(R200_VTX_COLOR_MASK<<R200_VTX_COLOR_0_SHIFT))
-#define MASK_SPEC  (MASK_COLOR|(R200_VTX_COLOR_MASK<<R200_VTX_COLOR_1_SHIFT))
-
-/* VTXFMT_1
- */
-#define MASK_ST0 (0x7 << R200_VTX_TEX0_COMP_CNT_SHIFT)
-/* FIXME: maybe something like in the radeon driver is needed here? */
-
-
-typedef void (*p4f)( GLfloat, GLfloat, GLfloat, GLfloat );
-typedef void (*p3f)( GLfloat, GLfloat, GLfloat );
-typedef void (*p2f)( GLfloat, GLfloat );
-typedef void (*p1f)( GLfloat );
-typedef void (*pe3f)( GLenum, GLfloat, GLfloat, GLfloat );
-typedef void (*pe2f)( GLenum, GLfloat, GLfloat );
-typedef void (*pe1f)( GLenum, GLfloat );
-typedef void (*p4ub)( GLubyte, GLubyte, GLubyte, GLubyte );
-typedef void (*p3ub)( GLubyte, GLubyte, GLubyte );
-typedef void (*pfv)( const GLfloat * );
-typedef void (*pefv)( GLenum, const GLfloat * );
-typedef void (*pubv)( const GLubyte * );
-
-
-CHOOSE(Normal3f, p3f, MASK_NORM, 0, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Normal3fv, pfv, MASK_NORM, 0, 
-       (const GLfloat *v), (v))
-
-#if 0
-CHOOSE_COLOR(Color4ub, p4ub, 4, MASK_COLOR, 0,
-	(GLubyte a,GLubyte b, GLubyte c, GLubyte d), (a,b,c,d))
-CHOOSE_COLOR(Color4ubv, pubv, 4, MASK_COLOR, 0, 
-	(const GLubyte *v), (v))
-CHOOSE_COLOR(Color3ub, p3ub, 3, MASK_COLOR, 0, 
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_COLOR(Color3ubv, pubv, 3, MASK_COLOR, 0, 
-	(const GLubyte *v), (v))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubEXT, p3ub, MASK_SPEC, 0, 
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubvEXT, pubv, MASK_SPEC, 0, 
-	(const GLubyte *v), (v))
-#endif
-
-CHOOSE_COLOR(Color4f, p4f, 4, MASK_COLOR, 0, 
-	(GLfloat a,GLfloat b, GLfloat c, GLfloat d), (a,b,c,d))
-CHOOSE_COLOR(Color4fv, pfv, 4, MASK_COLOR, 0, 
-	(const GLfloat *v), (v))
-CHOOSE_COLOR(Color3f, p3f, 3, MASK_COLOR, 0,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_COLOR(Color3fv, pfv, 3, MASK_COLOR, 0,
-	(const GLfloat *v), (v))
-
-
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fEXT, p3f, MASK_SPEC, 0,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fvEXT, pfv, MASK_SPEC, 0,
-	(const GLfloat *v), (v))
-
-CHOOSE(TexCoord3f, p3f, ~0, MASK_ST0, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(TexCoord3fv, pfv, ~0, MASK_ST0, 
-       (const GLfloat *v), (v))
-CHOOSE(TexCoord2f, p2f, ~0, MASK_ST0, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(TexCoord2fv, pfv, ~0, MASK_ST0, 
-       (const GLfloat *v), (v))
-CHOOSE(TexCoord1f, p1f, ~0, MASK_ST0, 
-       (GLfloat a), (a))
-CHOOSE(TexCoord1fv, pfv, ~0, MASK_ST0, 
-       (const GLfloat *v), (v))
-
-CHOOSE(MultiTexCoord3fARB, pe3f, ~0, ~0,
-	 (GLenum u,GLfloat a,GLfloat b,GLfloat c), (u,a,b,c))
-CHOOSE(MultiTexCoord3fvARB, pefv, ~0, ~0,
-	(GLenum u,const GLfloat *v), (u,v))
-CHOOSE(MultiTexCoord2fARB, pe2f, ~0, ~0,
-	 (GLenum u,GLfloat a,GLfloat b), (u,a,b))
-CHOOSE(MultiTexCoord2fvARB, pefv, ~0, ~0,
-	(GLenum u,const GLfloat *v), (u,v))
-CHOOSE(MultiTexCoord1fARB, pe1f, ~0, ~0,
-	 (GLenum u,GLfloat a), (u,a))
-CHOOSE(MultiTexCoord1fvARB, pefv, ~0, ~0,
-	(GLenum u,const GLfloat *v), (u,v))
-
-CHOOSE(Vertex3f, p3f, ~0, ~0, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Vertex3fv, pfv, ~0, ~0, 
-       (const GLfloat *v), (v))
-CHOOSE(Vertex2f, p2f, ~0, ~0, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(Vertex2fv, pfv, ~0, ~0, 
-       (const GLfloat *v), (v))
-
-CHOOSE(FogCoordfEXT, p1f, MASK_FOG, ~0, 
-       (GLfloat f), (f))
-CHOOSE(FogCoordfvEXT, pfv, MASK_FOG, ~0, 
-       (const GLfloat *f), (f))
-
-
-
-
-void r200VtxfmtInitChoosers( GLvertexformat *vfmt )
-{
-   vfmt->Color3f = choose_Color3f;
-   vfmt->Color3fv = choose_Color3fv;
-   vfmt->Color4f = choose_Color4f;
-   vfmt->Color4fv = choose_Color4fv;
-   vfmt->SecondaryColor3fEXT = choose_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = choose_SecondaryColor3fvEXT;
-   vfmt->MultiTexCoord1fARB = choose_MultiTexCoord1fARB;
-   vfmt->MultiTexCoord1fvARB = choose_MultiTexCoord1fvARB;
-   vfmt->MultiTexCoord2fARB = choose_MultiTexCoord2fARB;
-   vfmt->MultiTexCoord2fvARB = choose_MultiTexCoord2fvARB;
-   vfmt->MultiTexCoord3fARB = choose_MultiTexCoord3fARB;
-   vfmt->MultiTexCoord3fvARB = choose_MultiTexCoord3fvARB;
-   vfmt->Normal3f = choose_Normal3f;
-   vfmt->Normal3fv = choose_Normal3fv;
-   vfmt->TexCoord1f = choose_TexCoord1f;
-   vfmt->TexCoord1fv = choose_TexCoord1fv;
-   vfmt->TexCoord2f = choose_TexCoord2f;
-   vfmt->TexCoord2fv = choose_TexCoord2fv;
-   vfmt->TexCoord3f = choose_TexCoord3f;
-   vfmt->TexCoord3fv = choose_TexCoord3fv;
-   vfmt->Vertex2f = choose_Vertex2f;
-   vfmt->Vertex2fv = choose_Vertex2fv;
-   vfmt->Vertex3f = choose_Vertex3f;
-   vfmt->Vertex3fv = choose_Vertex3fv;
-/*   vfmt->FogCoordfEXT = choose_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = choose_FogCoordfvEXT;*/
-
-   /* TODO: restore ubyte colors to vtxfmt.
-    */
-#if 0
-   vfmt->Color3ub = choose_Color3ub;
-   vfmt->Color3ubv = choose_Color3ubv;
-   vfmt->Color4ub = choose_Color4ub;
-   vfmt->Color4ubv = choose_Color4ubv;
-   vfmt->SecondaryColor3ubEXT = choose_SecondaryColor3ubEXT;
-   vfmt->SecondaryColor3ubvEXT = choose_SecondaryColor3ubvEXT;
-#endif
-}
-
-
-static struct dynfn *codegen_noop( GLcontext *ctx, const int *key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void r200InitCodegen( struct dfn_generators *gen, GLboolean useCodegen )
-{
-   gen->Vertex3f = codegen_noop;
-   gen->Vertex3fv = codegen_noop;
-   gen->Color4ub = codegen_noop;
-   gen->Color4ubv = codegen_noop;
-   gen->Normal3f = codegen_noop;
-   gen->Normal3fv = codegen_noop;
-
-   gen->TexCoord3f = codegen_noop;
-   gen->TexCoord3fv = codegen_noop;
-   gen->TexCoord2f = codegen_noop;
-   gen->TexCoord2fv = codegen_noop;
-   gen->TexCoord1f = codegen_noop;
-   gen->TexCoord1fv = codegen_noop;
-
-   gen->MultiTexCoord3fARB = codegen_noop;
-   gen->MultiTexCoord3fvARB = codegen_noop;
-   gen->MultiTexCoord2fARB = codegen_noop;
-   gen->MultiTexCoord2fvARB = codegen_noop;
-   gen->MultiTexCoord1fARB = codegen_noop;
-   gen->MultiTexCoord1fvARB = codegen_noop;
-/*   gen->FogCoordfEXT = codegen_noop;
-   gen->FogCoordfvEXT = codegen_noop;*/
-
-   gen->Vertex2f = codegen_noop;
-   gen->Vertex2fv = codegen_noop;
-   gen->Color3ub = codegen_noop;
-   gen->Color3ubv = codegen_noop;
-   gen->Color4f = codegen_noop;
-   gen->Color4fv = codegen_noop;
-   gen->Color3f = codegen_noop;
-   gen->Color3fv = codegen_noop;
-   gen->SecondaryColor3fEXT = codegen_noop;
-   gen->SecondaryColor3fvEXT = codegen_noop;
-   gen->SecondaryColor3ubEXT = codegen_noop;
-   gen->SecondaryColor3ubvEXT = codegen_noop;
-
-   if (useCodegen) {
-#if defined(USE_X86_ASM)
-      r200InitX86Codegen( gen );
-#endif
-
-#if defined(USE_SSE_ASM)
-      r200InitSSECodegen( gen );
-#endif
-   }
-}
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c b/src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c
deleted file mode 100644
index 5901730494..0000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c
+++ /dev/null
@@ -1,234 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt_sse.c,v 1.1 2002/10/30 12:51:53 alanh Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "r200_vtxfmt.h"
-
-#if defined(USE_SSE_ASM)
-#include "x86/common_x86_asm.h"
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN( _sse_Attribute2fv );
-EXTERN( _sse_Attribute2f );
-EXTERN( _sse_Attribute3fv );
-EXTERN( _sse_Attribute3f );
-EXTERN( _sse_MultiTexCoord2fv );
-EXTERN( _sse_MultiTexCoord2f );
-EXTERN( _sse_MultiTexCoord2fv_2 );
-EXTERN( _sse_MultiTexCoord2f_2 );
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.
- */
-
-static struct dynfn *r200_makeSSEAttribute2fv( struct dynfn * cache, const int * key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 10, 0x0, (int)dest);
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEAttribute2f( struct dynfn * cache, const int * key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 8, 0x0, (int)dest); 
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEAttribute3fv( struct dynfn * cache, const int * key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 13, 0x0, (int)dest);
-   FIXUP(dfn->code, 18, 0x8, 8+(int)dest);
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEAttribute3f( struct dynfn * cache, const int * key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 12, 0x0, (int)dest); 
-   FIXUP(dfn->code, 17, 0x8, 8+(int)dest); 
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSENormal3fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				    __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *r200_makeSSENormal3f( GLcontext *ctx, const int * key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				   __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *r200_makeSSEColor3fv( GLcontext *ctx, const int * key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-				       __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-static struct dynfn *r200_makeSSEColor3f( GLcontext *ctx, const int * key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-				      __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-static struct dynfn *r200_makeSSETexCoord2fv( GLcontext *ctx, const int * key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				    __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-static struct dynfn *r200_makeSSETexCoord2f( GLcontext *ctx, const int * key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				   __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-static struct dynfn *r200_makeSSEMultiTexCoord2fv( GLcontext *ctx, const int * key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _sse_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 18, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEMultiTexCoord2f( GLcontext *ctx, const int * key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _sse_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 16, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 15, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-#endif
-
-void r200InitSSECodegen( struct dfn_generators *gen )
-{
-   if ( cpu_has_xmm ) {
-      gen->Normal3fv = (void *) r200_makeSSENormal3fv;
-      gen->Normal3f = (void *) r200_makeSSENormal3f;
-      gen->Color3fv = (void *) r200_makeSSEColor3fv;
-      gen->Color3f = (void *) r200_makeSSEColor3f;
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-      gen->TexCoord2fv = (void *) r200_makeSSETexCoord2fv;
-      gen->TexCoord2f = (void *) r200_makeSSETexCoord2f;
-      gen->MultiTexCoord2fvARB = (void *) r200_makeSSEMultiTexCoord2fv;
-      gen->MultiTexCoord2fARB = (void *) r200_makeSSEMultiTexCoord2f;
-#endif
-   }
-}
-
-#else 
-
-void r200InitSSECodegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c b/src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c
deleted file mode 100644
index b78a55c31f..0000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c
+++ /dev/null
@@ -1,440 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt_x86.c,v 1.2 2002/12/16 16:18:56 dawes Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "r200_vtxfmt.h"
-
-#if defined(USE_X86_ASM)
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN ( _x86_Attribute2fv );
-EXTERN ( _x86_Attribute2f );
-EXTERN ( _x86_Attribute3fv );
-EXTERN ( _x86_Attribute3f );
-EXTERN ( _x86_Vertex3fv_6 );
-EXTERN ( _x86_Vertex3fv_8 );
-EXTERN ( _x86_Vertex3fv );
-EXTERN ( _x86_Vertex3f_4 );
-EXTERN ( _x86_Vertex3f_6 );
-EXTERN ( _x86_Vertex3f );
-EXTERN ( _x86_Color4ubv_ub );
-EXTERN ( _x86_Color4ubv_4f );
-EXTERN ( _x86_Color4ub_ub );
-EXTERN ( _x86_MultiTexCoord2fv );
-EXTERN ( _x86_MultiTexCoord2fv_2 );
-EXTERN ( _x86_MultiTexCoord2f );
-EXTERN ( _x86_MultiTexCoord2f_2 );
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-struct dynfn *r200_makeX86Vertex3f( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x %d\n", __FUNCTION__, 
-	      key[0], key[1], rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 4: {
-
-      DFN ( _x86_Vertex3f_4, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 2, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 25, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 36, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 46, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 51, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 60, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   case 6: {
-
-      DFN ( _x86_Vertex3f_6, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 28, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 34, 0x0, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 40, 0x0, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 57, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 63, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 70, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 79, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   default: {
-
-      DFN ( _x86_Vertex3f, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 9, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 37, 0x0, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 44, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 50, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 56, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-
-
-struct dynfn *r200_makeX86Vertex3fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x %d\n", __FUNCTION__, 
-	      key[0], key[1], rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 6: {
-
-      DFN ( _x86_Vertex3fv_6, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x00000024, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 56, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 61, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 76, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-   case 8: {
-
-      DFN ( _x86_Vertex3fv_8, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x0000001c, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 51, 0x00000020, (int)&rmesa->vb.vertex[6]);
-      FIXUP(dfn->code, 63, 0x00000024, (int)&rmesa->vb.vertex[7]);
-      FIXUP(dfn->code, 74, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 79, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 85, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 94, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-
-   default: {
-
-      DFN ( _x86_Vertex3fv, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 8, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 32, 0x00000006, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 37, 0x00000058, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 45, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 50, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 58, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-static struct dynfn *
-r200_makeX86Attribute2fv( struct dynfn * cache, const int *key,
-			  const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 11, 0x0, (int)dest); 
-   FIXUP(dfn->code, 16, 0x4, 4+(int)dest); 
-
-   return dfn;
-}
-
-static struct dynfn *
-r200_makeX86Attribute2f( struct dynfn * cache, const int *key,
-			 const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 1, 0x0, (int)dest); 
-
-   return dfn;
-}
-
-
-static struct dynfn *
-r200_makeX86Attribute3fv( struct dynfn * cache, const int *key,
-			  const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-static struct dynfn *
-r200_makeX86Attribute3f( struct dynfn * cache, const int *key,
-			 const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-struct dynfn *r200_makeX86Normal3fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				    __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *r200_makeX86Normal3f( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				   __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *r200_makeX86Color4ubv( GLcontext *ctx, const int *key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (VTX_COLOR(key[0],0) == R200_VTX_PK_RGBA) {
-      DFN ( _x86_Color4ubv_ub, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 5, 0x12345678, (int)rmesa->vb.colorptr); 
-      return dfn;
-   } 
-   else {
-
-      DFN ( _x86_Color4ubv_4f, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 2, 0x00000000, (int)_mesa_ubyte_to_float_color_tab); 
-      FIXUP(dfn->code, 27, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr); 
-      FIXUP(dfn->code, 33, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+4); 
-      FIXUP(dfn->code, 55, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+8); 
-      FIXUP(dfn->code, 61, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+12); 
-      return dfn;
-   }
-}
-
-struct dynfn *r200_makeX86Color4ub( GLcontext *ctx, const int *key )
-{
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (VTX_COLOR(key[0],0) == R200_VTX_PK_RGBA) {
-      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      DFN ( _x86_Color4ub_ub, rmesa->vb.dfn_cache.Color4ub );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.colorptr); 
-      FIXUP(dfn->code, 24, 0x0, (int)rmesa->vb.colorptr+1); 
-      FIXUP(dfn->code, 30, 0x0, (int)rmesa->vb.colorptr+2); 
-      FIXUP(dfn->code, 36, 0x0, (int)rmesa->vb.colorptr+3); 
-      return dfn;
-   }
-   else
-      return NULL;
-}
-
-
-struct dynfn *r200_makeX86Color3fv( GLcontext *ctx, const int *key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-				       __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-struct dynfn *r200_makeX86Color3f( GLcontext *ctx, const int *key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeX86Attribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-				      __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-
-
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-struct dynfn *r200_makeX86TexCoord2fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				    __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-struct dynfn *r200_makeX86TexCoord2f( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				   __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-struct dynfn *r200_makeX86MultiTexCoord2fvARB( GLcontext *ctx, const int *key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _x86_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 21, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-      FIXUP(dfn->code, 27, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4);
-   } else {
-      DFN ( _x86_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-struct dynfn *r200_makeX86MultiTexCoord2fARB( GLcontext *ctx, 
-					      const int *key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _x86_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 20, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]); 
-      FIXUP(dfn->code, 26, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4); 
-   }
-   else {
-      /* Note: this might get generated multiple times, even though the
-       * actual emitted code is the same.
-       */
-      DFN ( _x86_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.texcoordptr); 
-   }      
-   return dfn;
-}
-#endif
-
-void r200InitX86Codegen( struct dfn_generators *gen )
-{
-   gen->Vertex3f = r200_makeX86Vertex3f;
-   gen->Vertex3fv = r200_makeX86Vertex3fv;
-   gen->Color4ub = r200_makeX86Color4ub; /* PKCOLOR only */
-   gen->Color4ubv = r200_makeX86Color4ubv; /* PKCOLOR only */
-   gen->Normal3f = r200_makeX86Normal3f;
-   gen->Normal3fv = r200_makeX86Normal3fv;
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-   gen->TexCoord2f = r200_makeX86TexCoord2f;
-   gen->TexCoord2fv = r200_makeX86TexCoord2fv;
-   gen->MultiTexCoord2fARB = r200_makeX86MultiTexCoord2fARB;
-   gen->MultiTexCoord2fvARB = r200_makeX86MultiTexCoord2fvARB;
-#endif
-   gen->Color3f = r200_makeX86Color3f;
-   gen->Color3fv = r200_makeX86Color3fv;
-
-   /* Not done:
-    */
-/*     gen->Vertex2f = r200_makeX86Vertex2f; */
-/*     gen->Vertex2fv = r200_makeX86Vertex2fv; */
-/*     gen->Color3ub = r200_makeX86Color3ub; */
-/*     gen->Color3ubv = r200_makeX86Color3ubv; */
-/*     gen->Color4f = r200_makeX86Color4f; */
-/*     gen->Color4fv = r200_makeX86Color4fv; */
-/*     gen->TexCoord1f = r200_makeX86TexCoord1f; */
-/*     gen->TexCoord1fv = r200_makeX86TexCoord1fv; */
-/*     gen->MultiTexCoord1fARB = r200_makeX86MultiTexCoord1fARB; */
-/*     gen->MultiTexCoord1fvARB = r200_makeX86MultiTexCoord1fvARB; */
-}
-
-
-#else 
-
-void r200InitX86Codegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S b/src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S
deleted file mode 100644
index 5e33c7bdee..0000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S
+++ /dev/null
@@ -1,499 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxtmp_x86.S,v 1.2 2002/11/07 18:31:59 tsi Exp $ */
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-#define GLOBL( x )	\
-.globl x;		\
-x:
-
-.data
-.align 4
-
-/*
-	vertex 3f vertex size 4
-*/
-	
-GLOBL ( _x86_Vertex3f_4 )
-	movl	(0), %ecx
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	%eax, (%ecx)
-	movl	%edx, 4(%ecx)
-	movl	12(%esp), %eax
-	movl	(0), %edx
-	movl	%eax, 8(%ecx)
-	movl	%edx, 12(%ecx)
-	movl	(0), %eax
-	addl	$16, %ecx
-	dec 	%eax
-	movl	%ecx, (0)
-	movl	%eax, (0)
-	je	.1 
-	ret
-.1:	jmp	*0
-	
-GLOBL ( _x86_Vertex3f_4_end )
-
-/*
-	vertex 3f vertex size 6
-*/
-GLOBL ( _x86_Vertex3f_6 )
-	push	%edi
-	movl	(0), %edi
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	movl	(0), %eax
-	movl	(0), %edx
-	movl	(0), %ecx
-	movl	%eax, 12(%edi)
-	movl	%edx, 16(%edi)
-	movl	%ecx, 20(%edi)
-	addl	$24, %edi
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	pop 	%edi
-	movl	%eax, (0)
-	je	.2
-	ret
-.2:	jmp	*0
-GLOBL ( _x86_Vertex3f_6_end )
-/*
-	vertex 3f generic size
-*/
-GLOBL ( _x86_Vertex3f )
-	push	%edi
-	push	%esi
-	movl	$0, %esi
-	movl	(0), %edi
-	movl	12(%esp), %eax
-	movl	16(%esp), %edx
-	movl	20(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	addl	$12, %edi
-	movl	$0, %ecx
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	movl	%eax, (0)
-	pop 	%esi
-	pop 	%edi
-	je  	.3
-	ret
-.3:	jmp	*0
-
-GLOBL ( _x86_Vertex3f_end )
-
-/*
-	Vertex 3fv vertex size 6
-*/
-GLOBL ( _x86_Vertex3fv_6 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx, (%eax)
-	movl	4(%ecx), %edx
-	movl	8(%ecx), %ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(36), %edx
-	movl	%edx, 20(%eax)
-	addl	$24, %eax
-	movl	%eax, 0
-	movl	4, %eax
-	dec 	%eax
-	movl	%eax, 4
-	je	.4
-	ret
-.4:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_6_end )
-
-/*
-	Vertex 3fv vertex size 8
-*/
-GLOBL ( _x86_Vertex3fv_8 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx ,(%eax)
-	movl	4(%ecx) ,%edx
-	movl	8(%ecx) ,%ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 20(%eax)
-	movl	%ecx, 24(%eax)
-	movl	(36), %edx
-	movl	%edx, 28(%eax)
-	addl	$32, %eax
-	movl	%eax, (0)
-	movl	4, %eax
-	dec	%eax
-	movl    %eax, (4)
-	je	.5
-	ret
-.5:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_8_end )
-
-/*
-	Vertex 3fv generic vertex size
-*/
-GLOBL ( _x86_Vertex3fv )
-	movl	4(%esp), %edx
-	push	%edi
-	push	%esi
-	movl	(0x1010101), %edi
-	movl	(%edx), %eax
-	movl	4(%edx), %ecx
-	movl	8(%edx), %esi
-	movl	%eax, (%edi)
-	movl	%ecx, 4(%edi)
-	movl	%esi, 8(%edi)
-	addl	$12, %edi
-	movl	$6, %ecx
-	movl	$0x58, %esi
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	%edi, (0x1010101)
-	movl	(0x2020202), %eax
-	pop	%esi
-	pop	%edi
-	dec	%eax
-	movl	%eax, (0x2020202)
-	je	.6
-	ret
-.6:	jmp    *0
-GLOBL ( _x86_Vertex3fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL ( _x86_Attribute2f )
-	movl	$0x0, %edx
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	movl	%eax, (%edx)
-	movl	%ecx, 4(%edx)
-	ret
-GLOBL ( _x86_Attribute2f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _x86_Attribute2fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %eax      /* load v[1] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %eax, 4      	/* store v[1] to current vertex */
-	ret
-GLOBL ( _x86_Attribute2fv_end )
-
-
-/**
- * Generic handler for 3 float format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL ( _x86_Attribute3f )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	movl	12(%esp), %eax
-	movl	%ecx, 0
-	movl	%edx, 4
-	movl	%eax, 8
-	ret
-GLOBL ( _x86_Attribute3f_end )
-
-/**
- * Generic handler for 3 float vector format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL( _x86_Attribute3fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %edx      /* load v[1] */
-	movl 8(%eax), %eax      /* load v[2] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %edx, 4      	/* store v[1] to current vertex */
-	movl %eax, 8      	/* store v[2] to current vertex */
-	ret
-GLOBL ( _x86_Attribute3fv_end )
-
-
-/*
-	Color 4ubv_ub
-*/
-GLOBL ( _x86_Color4ubv_ub )
-	movl 4(%esp), %eax
-	movl $0x12345678, %edx
-	movl (%eax), %eax
-	movl %eax, (%edx)
-	ret
-GLOBL ( _x86_Color4ubv_ub_end )
-
-/*
-	Color 4ubv 4f
-*/
-GLOBL ( _x86_Color4ubv_4f )
-	push	%ebx
-	movl	$0, %edx
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	movl	8(%esp), %ebx
-	movl	(%ebx), %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4),%eax
-	movl	(%edx,%ecx,4),%ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	shr	$16, %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4), %eax
-	movl	(%edx,%ecx,4), %ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	pop	%ebx
-	ret
-GLOBL ( _x86_Color4ubv_4f_end )
-
-/*
-
-	Color4ub_ub
-*/
-GLOBL( _x86_Color4ub_ub )
-	push	%ebx
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	20(%esp), %ebx
-	mov	%al, (0)
-	mov	%dl, (0)
-	mov	%cl, (0)
-	mov	%bl, (0)
-	pop	%ebx
-	ret
-GLOBL( _x86_Color4ub_ub_end )
-
-
-/* \todo: change the "and $7, %eax" to something like "target & 4 ? target & 5 : target & 3)" */
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$7, %eax
-	movl	(%ecx), %edx
-	shl	$3, %eax
-	movl	4(%ecx), %ecx
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-
-GLOBL( _x86_MultiTexCoord2fv_2 )
-	movl	4(%esp,1), %eax
-	movl	8(%esp,1), %ecx
-	and	$0x7, %eax
-	movl	0(,%eax,4), %edx
-	movl	(%ecx), %eax
-	movl	%eax, (%edx)
-	movl	4(%ecx), %eax
-	movl	%eax, 4(%edx)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp), %ecx
-	and	$7, %eax
-	shl	$3, %eax
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _x86_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp,1), %ecx
-	and	$7,%eax
-	movl	0(,%eax,4), %eax
-	movl	%edx, (%eax)
-	movl	%ecx, 4(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_2_end )
-
-#if defined(USE_SSE_ASM)
-/**
- * This can be used as a template for either Color3fv (when the color
- * target is also a 3f) or Normal3fv.
- */
-
-GLOBL( _sse_Attribute3fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movl	8(%eax), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3fv_end )
-
-/**
- * This can be used as a template for either Color3f (when the color
- * target is also a 3f) or Normal3f.
- */
-
-GLOBL( _sse_Attribute3f )
-	movlps	4(%esp), %xmm0
-	movl	12(%esp), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2f )
-	movlps	4(%esp), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2f_end )
-
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$7, %eax
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-GLOBL( _sse_MultiTexCoord2fv_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$0x7, %eax
-	movl	0(,%eax,4), %edx
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, (%edx)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	and	$7, %eax
-	movlps	8(%esp), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _sse_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movlps	8(%esp), %xmm0
-	and	$7,%eax
-	movl	0(,%eax,4), %eax
-	movlps	%xmm0, (%eax)
-	ret
-GLOBL( _sse_MultiTexCoord2f_2_end )
-#endif
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 71402761ae..00489ba14e 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -44,7 +44,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -292,7 +292,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext(ctx);
-	_ac_CreateContext(ctx);
+	_vbo_CreateContext(ctx);
 	_tnl_CreateContext(ctx);
 	_swsetup_CreateContext(ctx);
 	_swsetup_Wakeup(ctx);
@@ -305,7 +305,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 	/* Try and keep materials and vertices separate:
 	 */
-	_tnl_isolate_materials(ctx, GL_TRUE);
+/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
 
 	/* Configure swrast and TNL to match hardware characteristics:
 	 */
@@ -486,7 +486,7 @@ void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
 		_swsetup_DestroyContext(r300->radeon.glCtx);
 		_tnl_ProgramCacheDestroy(r300->radeon.glCtx);
 		_tnl_DestroyContext(r300->radeon.glCtx);
-		_ac_DestroyContext(r300->radeon.glCtx);
+		_vbo_DestroyContext(r300->radeon.glCtx);
 		_swrast_DestroyContext(r300->radeon.glCtx);
 		
 		if (r300->dma.current.buf) {
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index dd3ecbb235..02f8e9107d 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -48,7 +48,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_context.h"
 
 #define USER_BUFFERS
-#define RADEON_VTXFMT_A
+/* KW: Disable this code.  Driver should hook into vbo module
+ * directly, see i965 driver for example.
+ */
+/* #define RADEON_VTXFMT_A */
 #define HW_VBOS
 
 /* We don't handle 16 bits elts swapping yet */
@@ -754,7 +757,7 @@ struct radeon_vertex_buffer {
 	
 	struct dt AttribPtr[VERT_ATTRIB_MAX];
 	
-	struct tnl_prim  *Primitive;	              
+	const struct _mesa_prim  *Primitive;	              
 	GLuint      PrimitiveCount;
 	GLint LockFirst;
 	GLsizei LockCount;
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 91305cb5a2..3d094b9db5 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -42,7 +42,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_arrayelt.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_vp_build.h"
 
@@ -352,7 +352,7 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
 	r300EmitState(rmesa);
 	
 	for(i=0; i < VB->PrimitiveCount; i++){
-		GLuint prim = VB->Primitive[i].mode;
+		GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 		GLuint start = VB->Primitive[i].start;
 		GLuint length = VB->Primitive[i].count;
 		
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 6a22ccad2f..e06999aa26 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -46,7 +46,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_arrayelt.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "texformat.h"
 
@@ -1885,7 +1885,7 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
 	
 	_swrast_InvalidateState(ctx, new_state);
 	_swsetup_InvalidateState(ctx, new_state);
-	_ac_InvalidateState(ctx, new_state);
+	_vbo_InvalidateState(ctx, new_state);
 	_tnl_InvalidateState(ctx, new_state);
 	_ae_invalidate_state(ctx, new_state);
 
diff --git a/src/mesa/drivers/dri/r300/radeon_state.c b/src/mesa/drivers/dri/r300/radeon_state.c
index 0ceaf7ebb7..ddadf83a00 100644
--- a/src/mesa/drivers/dri/r300/radeon_state.c
+++ b/src/mesa/drivers/dri/r300/radeon_state.c
@@ -41,7 +41,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "light.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
diff --git a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
index 72c03c53ad..0625e5bc57 100644
--- a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
+++ b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
@@ -46,6 +46,8 @@
 #include "state.h"
 #include "image.h"
 
+#include "vbo/vbo_context.h"
+
 #define CONV_VB(a, b) rvb->AttribPtr[(a)].size = vb->b->size, \
 			rvb->AttribPtr[(a)].type = GL_FLOAT, \
 			rvb->AttribPtr[(a)].stride = vb->b->stride, \
@@ -129,15 +131,7 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 			CONV(i, VertexAttrib[i]);
 	
 	for (i=0; i < VERT_ATTRIB_MAX; i++) {
-		if (enabled & (1 << i)) {
-			rmesa->state.VB.AttribPtr[i].data += rmesa->state.VB.AttribPtr[i].stride * start;
-		} else {
-			def.data = ctx->Current.Attrib[i];
-			memcpy(&rmesa->state.VB.AttribPtr[i], &def, sizeof(struct dt));
-		}
-		
-		/*if(rmesa->state.VB.AttribPtr[i].data == ctx->Current.Attrib[i])
-			fprintf(stderr, "%d is default coord\n", i);*/
+	   rmesa->state.VB.AttribPtr[i].data += rmesa->state.VB.AttribPtr[i].stride * start;
 	}
 	
 	for(i=0; i < VERT_ATTRIB_MAX; i++){
@@ -177,177 +171,18 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 
 void radeon_init_vtxfmt_a(r300ContextPtr rmesa);
 
-static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const GLvoid *c_indices )
-{
-	GET_CURRENT_CONTEXT(ctx);
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int elt_size;
-	int i;
-	unsigned int min = ~0, max = 0;
-	struct tnl_prim prim;
-	static void *ptr = NULL;
-	struct r300_dma_region rvb;
-	const GLvoid *indices = c_indices;
-	
-	if (count > 65535) {
-		WARN_ONCE("Too many verts!\n");
-		goto fallback;
-	}
-	
-	if (ctx->Array.ElementArrayBufferObj->Name) {
-		/* use indices in the buffer object */
-		if (!ctx->Array.ElementArrayBufferObj->Data) {
-			_mesa_warning(ctx, "DrawRangeElements with empty vertex elements buffer!");
-			return;
-		}
-		/* actual address is the sum of pointers */
-		indices = (GLvoid *)
-		ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data, (const GLubyte *) c_indices);
-	}
-	
-	if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-		return;
-	
-	FLUSH_CURRENT( ctx, 0 );
-	
-	memset(&rvb, 0, sizeof(rvb));
-	switch (type) {
-	case GL_UNSIGNED_BYTE:
-		for (i=0; i < count; i++) {
-			if(((unsigned char *)indices)[i] < min)
-				min = ((unsigned char *)indices)[i];
-			if(((unsigned char *)indices)[i] > max)
-				max = ((unsigned char *)indices)[i];
-		}
-		
-#ifdef FORCE_32BITS_ELTS
-		elt_size = 4;
-#else
-		elt_size = 2;
-#endif		
-		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
-		rvb.aos_offset = GET_START(&rvb);
-		ptr = rvb.address + rvb.start;
-			
-#ifdef FORCE_32BITS_ELTS
-		for (i=0; i < count; i++)
-			((unsigned int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
-#else
-		for (i=0; i < count; i++)
-			((unsigned short int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
-#endif
-	break;
-		
-	case GL_UNSIGNED_SHORT:
-		for (i=0; i < count; i++) {
-			if(((unsigned short int *)indices)[i] < min)
-				min = ((unsigned short int *)indices)[i];
-			if(((unsigned short int *)indices)[i] > max)
-				max = ((unsigned short int *)indices)[i];
-		}
-		
-#ifdef FORCE_32BITS_ELTS
-		elt_size = 4;
-#else
-		elt_size = 2;
-#endif
-		
-		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
-		rvb.aos_offset = GET_START(&rvb);
-		ptr = rvb.address + rvb.start;
-		
-#ifdef FORCE_32BITS_ELTS
-		for (i=0; i < count; i++)
-			((unsigned int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
-#else
-		for (i=0; i < count; i++)
-			((unsigned short int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
-#endif
-	break;
-	
-	case GL_UNSIGNED_INT:
-		for (i=0; i < count; i++) {
-			if(((unsigned int *)indices)[i] < min)
-				min = ((unsigned int *)indices)[i];
-			if(((unsigned int *)indices)[i] > max)
-				max = ((unsigned int *)indices)[i];
-		}
-		
-#ifdef FORCE_32BITS_ELTS
-		elt_size = 4;
-#else
-		if (max - min <= 65535)
-			elt_size = 2;
-		else 
-			elt_size = 4;
-#endif
-		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
-		rvb.aos_offset = GET_START(&rvb);
-		ptr = rvb.address + rvb.start;
-		
-		
-		if (elt_size == 2)
-			for (i=0; i < count; i++)
-				((unsigned short int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
-		else
-			for (i=0; i < count; i++)
-				((unsigned int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
-	break;
-	
-	default:
-		WARN_ONCE("Unknown elt type!\n");
-	goto fallback;
-	}
-	
-	if (ctx->NewState) 
-		_mesa_update_state( ctx );
-	
-	r300UpdateShaders(rmesa);
-	
-	if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL) {
-		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
-	}
-	
-	rmesa->state.VB.Count = max - min + 1;
-	
-	r300UpdateShaderStates(rmesa);
-	
-	rmesa->state.VB.Primitive = &prim;
-	rmesa->state.VB.PrimitiveCount = 1;
-	
-	prim.mode = mode | PRIM_BEGIN | PRIM_END;
-	if (rmesa->state.VB.LockCount)
-		prim.start = min - rmesa->state.VB.LockFirst;
-	else
-		prim.start = 0;
-	prim.count = count;
-	
-	rmesa->state.VB.Elts = ptr;
-	rmesa->state.VB.elt_size = elt_size;
-	
-	if (r300_run_vb_render(ctx, NULL)) {
-		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
-	}
-	
-	if(rvb.buf)
-		radeon_mm_use(rmesa, rvb.buf->id);
-	
-	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-	return;
-	
-	fallback:
-	_tnl_array_init(ctx);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-	CALL_DrawElements(GET_DISPATCH(), (mode, count, type, c_indices));
-	radeon_init_vtxfmt_a(rmesa);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-}
 
-static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei count, GLenum type, const GLvoid *c_indices)
+static void radeonDrawRangeElements(GLcontext *ctx,
+				    GLenum mode,
+				    GLuint min,
+				    GLuint max,
+				    GLsizei count, 
+				    GLenum type,
+				    const GLvoid *c_indices)
 {
-	GET_CURRENT_CONTEXT(ctx);
+#if 1
+	return GL_FALSE;
+#else
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct tnl_prim prim;
 	int elt_size;
@@ -371,26 +206,23 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 				indices += i * _mesa_sizeof_type(type);
 				count -= i;
 			}
-			return ;
+			return GL_TRUE;
 		}
 		WARN_ONCE("Too many verts!\n");
-		goto fallback;
+		return GL_FALSE;
 	}
 	
 	if (ctx->Array.ElementArrayBufferObj->Name) {
 		/* use indices in the buffer object */
 		if (!ctx->Array.ElementArrayBufferObj->Data) {
 			_mesa_warning(ctx, "DrawRangeElements with empty vertex elements buffer!");
-			return;
+			return GL_TRUE;
 		}
 		/* actual address is the sum of pointers */
 		indices = (GLvoid *)
 		ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data, (const GLubyte *) c_indices);
 	}
 	
-	if (!_mesa_validate_DrawRangeElements( ctx, mode, min, max, count, type, indices ))
-		return;
-	
 	FLUSH_CURRENT( ctx, 0 );
 #ifdef OPTIMIZE_ELTS
 	min = 0;
@@ -465,7 +297,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	
 	default:
 		WARN_ONCE("Unknown elt type!\n");
-	goto fallback;
+		return GL_FALSE;
 	}
 	
 	/* XXX: setup_arrays before state update? */
@@ -477,7 +309,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 
 	if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL) {
 		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
+		return GL_FALSE;
 	}
 
 	rmesa->state.VB.Count = max - min + 1;
@@ -501,37 +333,34 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	
 	if (r300_run_vb_render(ctx, NULL)) {
 		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
+		return GL_FALSE;
 	}
 	
 	if(rvb.buf)
 		radeon_mm_use(rmesa, rvb.buf->id);
 	
 	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-	return ;
-	
-	fallback:
-	_tnl_array_init(ctx);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-	CALL_DrawRangeElements(GET_DISPATCH(), (mode, min, max, count, type, c_indices));
-	radeon_init_vtxfmt_a(rmesa);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
+	return GL_TRUE;
+#endif
 }
 
-static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
+static GLboolean radeonDrawArrays( GLcontext *ctx,
+				   GLenum mode, GLint start, GLsizei count )
 {
+#if 1
+	return GL_FALSE;
+#else
 	GET_CURRENT_CONTEXT(ctx);
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct tnl_prim prim;
 	
 	if (count > 65535) {
+	   /* TODO: split into multiple draws.
+	    */
 		WARN_ONCE("Too many verts!\n");
-		goto fallback;
+		return GL_FALSE;
 	}
 	
-	if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-		return;
-	
 	FLUSH_CURRENT( ctx, 0 );
 	
 	if (ctx->NewState) 
@@ -542,7 +371,7 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 	r300UpdateShaders(rmesa);
 
 	if (setup_arrays(rmesa, start) >= R300_FALLBACK_TCL)
-		goto fallback;
+		return GL_FALSE;
 
 	rmesa->state.VB.Count = count;
 
@@ -564,31 +393,70 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 	rmesa->state.VB.elt_max = 0;
 	
 	if (r300_run_vb_render(ctx, NULL))
-		goto fallback;
+	   return GL_FALSE;
 
-	return ;
-	
-	fallback:
-	_tnl_array_init(ctx);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-	CALL_DrawArrays(GET_DISPATCH(), (mode, start, count));
-	radeon_init_vtxfmt_a(rmesa);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
+	return GL_TRUE;
+#endif
 }
 
+static void radeon_draw_prims( GLcontext *ctx,
+			       const struct gl_client_array *arrays[],
+			       const struct _mesa_prim *prim,
+			       GLuint nr_prims,
+			       const struct _mesa_index_buffer *ib,
+			       GLuint min_index,
+			       GLuint max_index)
+{
+	if (ib == NULL) {
+		for (i = 0; i < nr_prims; i++) {
+			if (!radeonDrawArrays(ctx,
+					      prim->mode,
+					      prim->start,
+					      prim->count)) {
+				/* Fallback
+				 */
+				_tnl_draw_prims(ctx,
+						arrays,
+						prim + i,
+						nr_prims - i,
+						ib, 
+						min_index,
+						max_index);
+				return;
+			}
+		}
+	} else {
+		for (i = 0; i < nr_prims; i++) {
+			if (!radeonDrawRangeElements(ctx,
+						     prim->mode,
+						     min_index,
+						     max_index,
+						     prim->count,
+						     ib->types,
+						     ib->ptr)) {
+				/* Fallback
+				 */
+				_tnl_draw_prims(ctx,
+						arrays,
+						prim + i,
+						nr_prims - i,
+						ib, 
+						min_index,
+						max_index);
+				return;
+			}
+		}
+	}
+}
+	   
 void radeon_init_vtxfmt_a(r300ContextPtr rmesa)
 {
 	GLcontext *ctx;
-	GLvertexformat *vfmt;
-	
-	ctx = rmesa->radeon.glCtx; 
-	vfmt = (GLvertexformat *)ctx->TnlModule.Current;
-   
-	vfmt->DrawElements = radeonDrawElements;
-	vfmt->DrawArrays = radeonDrawArrays;
-	vfmt->DrawRangeElements = radeonDrawRangeElements;
+	struct vbo_context *vbo = vbo_context(ctx);
 	
+	vbo->draw_prims = radeon_draw_prims;
 }
+
 #endif
 
 #ifdef HW_VBOS
diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile
index 6e46620b39..f223b2d922 100644
--- a/src/mesa/drivers/dri/radeon/Makefile
+++ b/src/mesa/drivers/dri/radeon/Makefile
@@ -22,11 +22,7 @@ DRIVER_SOURCES = \
 	radeon_swtcl.c \
 	radeon_span.c \
 	radeon_maos.c \
-	radeon_sanity.c \
-	radeon_vtxfmt.c \
-	radeon_vtxfmt_c.c \
-	radeon_vtxfmt_sse.c \
-	radeon_vtxfmt_x86.c 
+	radeon_sanity.c 
 
 C_SOURCES = \
 	$(COMMON_SOURCES) \
@@ -34,8 +30,7 @@ C_SOURCES = \
 
 DRIVER_DEFINES = -DRADEON_COMMON=0
 
-X86_SOURCES = \
-	radeon_vtxtmp_x86.S 
+X86_SOURCES = 
 
 include ../Makefile.template
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index 8b92e3fe65..6bc2c4aa5c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -46,7 +46,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -60,7 +60,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
 #include "radeon_tcl.h"
-#include "radeon_vtxfmt.h"
 #include "radeon_maos.h"
 
 #define need_GL_ARB_multisample
@@ -361,7 +360,7 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _ae_create_context( ctx );
@@ -370,13 +369,10 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     */
    _tnl_destroy_pipeline( ctx );
    _tnl_install_pipeline( ctx, radeon_pipeline );
-   ctx->Driver.FlushVertices = radeonFlushVertices;
 
    /* Try and keep materials and vertices separate:
     */
-   _tnl_isolate_materials( ctx, GL_TRUE );
-
-/*     _mesa_allow_light_in_model( ctx, GL_FALSE ); */
+/*    _tnl_isolate_materials( ctx, GL_TRUE ); */
 
    /* Configure swrast and T&L to match hardware characteristics:
     */
@@ -450,10 +446,7 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    }
 
    if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-      if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	 radeonVtxfmtInit( ctx, tcl_mode >= DRI_CONF_TCL_CODEGEN );
-
-      _tnl_need_dlist_norm_lengths( ctx, GL_FALSE );
+/*       _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
    }
    return GL_TRUE;
 }
@@ -484,7 +477,7 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv )
       release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( rmesa->glCtx );
       _tnl_DestroyContext( rmesa->glCtx );
-      _ac_DestroyContext( rmesa->glCtx );
+      _vbo_DestroyContext( rmesa->glCtx );
       _swrast_DestroyContext( rmesa->glCtx );
 
       radeonDestroySwtcl( rmesa->glCtx );
@@ -494,12 +487,6 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv )
 	 radeonFlushCmdBuf( rmesa, __FUNCTION__ );
       }
 
-      if (!(rmesa->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)) {
-	 int tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-	 if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	    radeonVtxfmtDestroy( rmesa->glCtx );
-      }
-
       _mesa_vector4f_free( &rmesa->tcl.ObjClean );
 
       if (rmesa->state.scissor.pClipRects) {
@@ -621,9 +608,6 @@ radeonMakeCurrent( __DRIcontextPrivate *driContextPriv,
 			  (GLframebuffer *) driDrawPriv->driverPrivate,
 			  (GLframebuffer *) driReadPriv->driverPrivate );
 
-      if (newCtx->vb.enabled)
-	 radeonVtxfmtMakeCurrent( newCtx->glCtx );
-
    } else {
       if (RADEON_DEBUG & DEBUG_DRI)
 	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
index 91a60bb9f1..65dbecf7a6 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
@@ -37,7 +37,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "imports.h"
 #include "mtypes.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "math/m_translate.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 216d7dca17..e19202fa44 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -42,7 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "context.h"
 #include "framebuffer.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -53,7 +53,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tcl.h"
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
-#include "radeon_vtxfmt.h"
 #include "drirenderbuffer.h"
 
 static void radeonUpdateSpecular( GLcontext *ctx );
@@ -2303,11 +2302,10 @@ static void radeonInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
    RADEON_CONTEXT(ctx)->NewGLState |= new_state;
-   radeonVtxfmtInvalidate( ctx );
 }
 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index 5682d95ae1..5fc34f0933 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -33,7 +33,7 @@
 #include "api_arrayelt.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -44,7 +44,6 @@
 #include "radeon_tcl.h"
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
-#include "radeon_vtxfmt.h"
 
 #include "xmlpool.h"
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
index e36a710d33..7ce1fa67cf 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
@@ -45,7 +45,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "tnl/t_vtx_api.h"	/* for _tnl_FlushVertices */
 
 #include "radeon_context.h"
 #include "radeon_ioctl.h"
@@ -848,14 +847,6 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 }
 
 
-void radeonFlushVertices( GLcontext *ctx, GLuint flags )
-{
-   _tnl_FlushVertices( ctx, flags );
-
-   if (flags & FLUSH_STORED_VERTICES)
-      RADEON_NEWPRIM( RADEON_CONTEXT( ctx ) );
-}
-
 /**********************************************************************/
 /*                            Initialization.                         */
 /**********************************************************************/
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
index f95a52c0b8..64f9019513 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
@@ -43,7 +43,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 extern void radeonInitSwtcl( GLcontext *ctx );
 extern void radeonDestroySwtcl( GLcontext *ctx );
 
-extern void radeonFlushVertices( GLcontext *ctx, GLuint flags );
 extern void radeonChooseRenderState( GLcontext *ctx );
 extern void radeonChooseVertexState( GLcontext *ctx );
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c
index ffd49b525b..5ad044c262 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c
@@ -39,7 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "mtypes.h"
 #include "enums.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c
deleted file mode 100644
index a5a9eb144b..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c
+++ /dev/null
@@ -1,1086 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt.c,v 1.6 2003/05/06 23:52:08 daenzer Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-#include "glheader.h"
-#include "imports.h"
-#include "api_noop.h"
-#include "api_arrayelt.h"
-#include "context.h"
-#include "mtypes.h"
-#include "enums.h"
-#include "glapi.h"
-#include "colormac.h"
-#include "light.h"
-#include "state.h"
-#include "vtxfmt.h"
-
-#include "tnl/tnl.h"
-#include "tnl/t_context.h"
-#include "tnl/t_array_api.h"
-#include "tnl/t_save_api.h"
-
-#include "radeon_context.h"
-#include "radeon_state.h"
-#include "radeon_ioctl.h"
-#include "radeon_tex.h"
-#include "radeon_tcl.h"
-#include "radeon_swtcl.h"
-#include "radeon_vtxfmt.h"
-
-#define VERT_ATTRIB_TEX(u)	(VERT_ATTRIB_TEX0 + (u))
-
-#include "dispatch.h"
-
-static void radeonVtxfmtFlushVertices( GLcontext *, GLuint );
-
-static void count_func( const char *name,  struct dynfn *l )
-{
-   int i = 0;
-   struct dynfn *f;
-   foreach (f, l) i++;
-   if (i) fprintf(stderr, "%s: %d\n", name, i );
-}
-
-static void count_funcs( radeonContextPtr rmesa )
-{
-   count_func( "Vertex2f", &rmesa->vb.dfn_cache.Vertex2f );
-   count_func( "Vertex2fv", &rmesa->vb.dfn_cache.Vertex2fv );
-   count_func( "Vertex3f", &rmesa->vb.dfn_cache.Vertex3f );
-   count_func( "Vertex3fv", &rmesa->vb.dfn_cache.Vertex3fv );
-   count_func( "Color4ub", &rmesa->vb.dfn_cache.Color4ub );
-   count_func( "Color4ubv", &rmesa->vb.dfn_cache.Color4ubv );
-   count_func( "Color3ub", &rmesa->vb.dfn_cache.Color3ub );
-   count_func( "Color3ubv", &rmesa->vb.dfn_cache.Color3ubv );
-   count_func( "Color4f", &rmesa->vb.dfn_cache.Color4f );
-   count_func( "Color4fv", &rmesa->vb.dfn_cache.Color4fv );
-   count_func( "Color3f", &rmesa->vb.dfn_cache.Color3f );
-   count_func( "Color3fv", &rmesa->vb.dfn_cache.Color3fv );
-   count_func( "SecondaryColor3f", &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   count_func( "SecondaryColor3fv", &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   count_func( "SecondaryColor3ub", &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   count_func( "SecondaryColor3ubv", &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   count_func( "Normal3f", &rmesa->vb.dfn_cache.Normal3f );
-   count_func( "Normal3fv", &rmesa->vb.dfn_cache.Normal3fv );
-   count_func( "TexCoord2f", &rmesa->vb.dfn_cache.TexCoord2f );
-   count_func( "TexCoord2fv", &rmesa->vb.dfn_cache.TexCoord2fv );
-   count_func( "TexCoord1f", &rmesa->vb.dfn_cache.TexCoord1f );
-   count_func( "TexCoord1fv", &rmesa->vb.dfn_cache.TexCoord1fv );
-   count_func( "MultiTexCoord2fARB", &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   count_func( "MultiTexCoord2fvARB", &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   count_func( "MultiTexCoord1fARB", &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   count_func( "MultiTexCoord1fvARB", &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-}
-
-
-void radeon_copy_to_current( GLcontext *ctx ) 
-{
-   GLuint unit;
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   assert(ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT);
-
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_N0) {
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0] = rmesa->vb.normalptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1] = rmesa->vb.normalptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2] = rmesa->vb.normalptr[2];
-   }
-
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_PKCOLOR) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->blue );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->alpha );
-   } 
-   
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_FPCOLOR) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = rmesa->vb.floatcolorptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = rmesa->vb.floatcolorptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = rmesa->vb.floatcolorptr[2];
-   }
-
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_FPALPHA)
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = rmesa->vb.floatcolorptr[3];
-      
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_PKSPEC) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] = UBYTE_TO_FLOAT( rmesa->vb.specptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] = UBYTE_TO_FLOAT( rmesa->vb.specptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] = UBYTE_TO_FLOAT( rmesa->vb.specptr->blue );
-   } 
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (rmesa->vb.vertex_format & RADEON_ST_BIT(unit)) {
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][0] = rmesa->vb.texcoordptr[unit][0];
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][1] = rmesa->vb.texcoordptr[unit][1];
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][2] = 0.0F;
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][3] = 1.0F;
-      }
-   }
-
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-static GLboolean discreet_gl_prim[GL_POLYGON+1] = {
-   1,				/* 0 points */
-   1,				/* 1 lines */
-   0,				/* 2 line_strip */
-   0,				/* 3 line_loop */
-   1,				/* 4 tris */
-   0,				/* 5 tri_fan */
-   0,				/* 6 tri_strip */
-   1,				/* 7 quads */
-   0,				/* 8 quadstrip */
-   0,				/* 9 poly */
-};
-
-static void flush_prims( radeonContextPtr rmesa )
-{
-   int i,j;
-   struct radeon_dma_region tmp = rmesa->dma.current;
-   
-   tmp.buf->refcount++;
-   tmp.aos_size = rmesa->vb.vertex_size;
-   tmp.aos_stride = rmesa->vb.vertex_size;
-   tmp.aos_start = GET_START(&tmp);
-
-   rmesa->dma.current.ptr = rmesa->dma.current.start += 
-      (rmesa->vb.initial_counter - rmesa->vb.counter) * rmesa->vb.vertex_size * 4; 
-
-   rmesa->tcl.vertex_format = rmesa->vb.vertex_format;
-   rmesa->tcl.aos_components[0] = &tmp;
-   rmesa->tcl.nr_aos_components = 1;
-   rmesa->dma.flush = NULL;
-
-   /* Optimize the primitive list:
-    */
-   if (rmesa->vb.nrprims > 1) {
-      for (j = 0, i = 1 ; i < rmesa->vb.nrprims; i++) {
-	 int pj = rmesa->vb.primlist[j].prim & 0xf;
-	 int pi = rmesa->vb.primlist[i].prim & 0xf;
-      
-	 if (pj == pi && discreet_gl_prim[pj] &&
-	     rmesa->vb.primlist[i].start == rmesa->vb.primlist[j].end) {
-	    rmesa->vb.primlist[j].end = rmesa->vb.primlist[i].end;
-	 }
-	 else {
-	    j++;
-	    if (j != i) rmesa->vb.primlist[j] = rmesa->vb.primlist[i];
-	 }
-      }
-      rmesa->vb.nrprims = j+1;
-   }
-
-   for (i = 0 ; i < rmesa->vb.nrprims; i++) {
-      if (RADEON_DEBUG & DEBUG_PRIMS)
-	 fprintf(stderr, "vtxfmt prim %d: %s %d..%d\n", i,
-		 _mesa_lookup_enum_by_nr( rmesa->vb.primlist[i].prim & 
-					  PRIM_MODE_MASK ),
-		 rmesa->vb.primlist[i].start,
-		 rmesa->vb.primlist[i].end);
-
-      radeonEmitPrimitive( rmesa->glCtx,
-			   rmesa->vb.primlist[i].start,
-			   rmesa->vb.primlist[i].end,
-			   rmesa->vb.primlist[i].prim );
-   }
-
-   rmesa->vb.nrprims = 0;
-   radeonReleaseDmaRegion( rmesa, &tmp, __FUNCTION__ );
-}
-
-
-static void start_prim( radeonContextPtr rmesa, GLuint mode )
-{
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   rmesa->vb.primlist[rmesa->vb.nrprims].start = rmesa->vb.initial_counter - rmesa->vb.counter;
-   rmesa->vb.primlist[rmesa->vb.nrprims].prim = mode;
-}
-
-static void note_last_prim( radeonContextPtr rmesa, GLuint flags )
-{
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      rmesa->vb.primlist[rmesa->vb.nrprims].prim |= flags;
-      rmesa->vb.primlist[rmesa->vb.nrprims].end = rmesa->vb.initial_counter - rmesa->vb.counter;
-
-      if (++(rmesa->vb.nrprims) == RADEON_MAX_PRIMS)
-	 flush_prims( rmesa );
-   }
-}
-
-
-static void copy_vertex( radeonContextPtr rmesa, GLuint n, GLfloat *dst )
-{
-   GLuint i;
-   GLfloat *src = (GLfloat *)(rmesa->dma.current.address + 
-			      rmesa->dma.current.ptr + 
-			      (rmesa->vb.primlist[rmesa->vb.nrprims].start + n) * 
-			      rmesa->vb.vertex_size * 4);
-
-   if (RADEON_DEBUG & DEBUG_VFMT) 
-      fprintf(stderr, "copy_vertex %d\n", rmesa->vb.primlist[rmesa->vb.nrprims].start + n);
-
-   for (i = 0 ; i < rmesa->vb.vertex_size; i++) {
-      dst[i] = src[i];
-   }
-}
-
-/* NOTE: This actually reads the copied vertices back from uncached
- * memory.  Could also use the counter/notify mechanism to populate
- * tmp on the fly as vertices are generated.  
- */
-static GLuint copy_dma_verts( radeonContextPtr rmesa, GLfloat (*tmp)[RADEON_MAX_VERTEX_SIZE] )
-{
-   GLuint ovf, i;
-   GLuint nr = (rmesa->vb.initial_counter - rmesa->vb.counter) - rmesa->vb.primlist[rmesa->vb.nrprims].start;
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d verts\n", __FUNCTION__, nr);
-
-   switch( rmesa->vb.prim[0] )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      copy_vertex( rmesa, nr-1, tmp[0] );
-      return 1;
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 return 1;
-      } else {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 copy_vertex( rmesa, nr-1, tmp[1] );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      ovf = MIN2(nr, 2);
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-static void VFMT_FALLBACK_OUTSIDE_BEGIN_END( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (ctx->Driver.NeedFlush) 
-      radeonVtxfmtFlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx ); /* clear state so fell_back sticks */
-
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = radeonFlushVertices;
-   ctx->Driver.NewList =_tnl_NewList;
-
-   assert( rmesa->dma.flush == 0 );
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-}
-
-
-static void VFMT_FALLBACK( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat tmp[3][RADEON_MAX_VERTEX_SIZE];
-   GLuint i, prim;
-   GLuint ind = rmesa->vb.vertex_format;
-   GLuint nrverts;
-   GLfloat alpha = 1.0;
-   GLuint unit;
-
-   if (RADEON_DEBUG & (DEBUG_FALLBACKS|DEBUG_VFMT))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      VFMT_FALLBACK_OUTSIDE_BEGIN_END( __FUNCTION__ );
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   nrverts = copy_dma_verts( rmesa, tmp );
-
-   /* Finish the prim at this point:
-    */
-   note_last_prim( rmesa, 0 );
-   flush_prims( rmesa );
-
-   /* Update ctx->Driver.CurrentExecPrimitive and swap in swtnl. 
-    */
-   prim = rmesa->vb.prim[0];
-   ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1;
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = radeonFlushVertices;
-
-   assert(rmesa->dma.flush == 0);
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-   CALL_Begin(GET_DISPATCH(), (prim));
-   
-   if (rmesa->vb.installed_color_3f_sz == 4)
-      alpha = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-
-   /* Replay saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      GLuint offset = 3;
-      if (ind & RADEON_CP_VC_FRMT_N0) {
-	 CALL_Normal3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset += 3;
-      }
-
-      if (ind & RADEON_CP_VC_FRMT_PKCOLOR) {
-	 radeon_color_t *col = (radeon_color_t *)&tmp[i][offset];
-	 CALL_Color4ub(GET_DISPATCH(), (col->red, col->green, col->blue, col->alpha));
-	 offset++;
-      }
-      else if (ind & RADEON_CP_VC_FRMT_FPALPHA) {
-	 CALL_Color4fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=4;
-      } 
-      else if (ind & RADEON_CP_VC_FRMT_FPCOLOR) {
-	 CALL_Color3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=3;
-      }
-
-      if (ind & RADEON_CP_VC_FRMT_PKSPEC) {
-	 radeon_color_t *spec = (radeon_color_t *)&tmp[i][offset];
-	 CALL_SecondaryColor3ubEXT(GET_DISPATCH(), (spec->red, spec->green, spec->blue));
-	 offset++;
-      }
-
-      for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-	 if (ind & RADEON_ST_BIT(unit)) {
-	    CALL_MultiTexCoord2fvARB(GET_DISPATCH(), ((GL_TEXTURE0 + unit), &tmp[i][offset]));
-	    offset += 2;
-	 }
-      }
-      CALL_Vertex3fv(GET_DISPATCH(), (&tmp[i][0]));
-   }
-
-   /* Replay current vertex
-    */
-   if (ind & RADEON_CP_VC_FRMT_N0) 
-      CALL_Normal3fv(GET_DISPATCH(), (rmesa->vb.normalptr));
-
-   if (ind & RADEON_CP_VC_FRMT_PKCOLOR)
-      CALL_Color4ub(GET_DISPATCH(), (rmesa->vb.colorptr->red, rmesa->vb.colorptr->green,
-      				     rmesa->vb.colorptr->blue, rmesa->vb.colorptr->alpha));
-   else if (ind & RADEON_CP_VC_FRMT_FPALPHA)
-      CALL_Color4fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-   else if (ind & RADEON_CP_VC_FRMT_FPCOLOR) {
-      if (rmesa->vb.installed_color_3f_sz == 4 && alpha != 1.0)
-	 CALL_Color4f(GET_DISPATCH(), (rmesa->vb.floatcolorptr[0],
-				       rmesa->vb.floatcolorptr[1],
-				       rmesa->vb.floatcolorptr[2],
-				       alpha));
-      else
-	 CALL_Color3fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-   }
-
-   if (ind & RADEON_CP_VC_FRMT_PKSPEC) 
-       CALL_SecondaryColor3ubEXT(GET_DISPATCH(), (rmesa->vb.specptr->red,
-       						  rmesa->vb.specptr->green,
-						  rmesa->vb.specptr->blue));
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ind & RADEON_ST_BIT(unit)) {
-	 CALL_MultiTexCoord2fvARB(GET_DISPATCH(), ((GL_TEXTURE0 + unit),
-	 			  rmesa->vb.texcoordptr[unit]));
-      }
-   }
-}
-
-static void radeonNewList( GLcontext *ctx, GLuint list, GLenum mode )
-{
-   VFMT_FALLBACK( __FUNCTION__ );
-   _tnl_NewList( ctx, list, mode );
-   return;
-}
-
-
-static void wrap_buffer( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat tmp[3][RADEON_MAX_VERTEX_SIZE];
-   GLuint i, nrverts;
-
-   if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_PRIMS))
-      fprintf(stderr, "%s %d\n", __FUNCTION__, rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   /* Don't deal with parity.
-    */
-   if ((((rmesa->vb.initial_counter - rmesa->vb.counter) -  
-	 rmesa->vb.primlist[rmesa->vb.nrprims].start) & 1)) {
-      rmesa->vb.counter++;
-      rmesa->vb.initial_counter++;
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) 
-      nrverts = 0;
-   else {
-      nrverts = copy_dma_verts( rmesa, tmp );
-
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%d vertices to copy\n", nrverts);
-   
-      /* Finish the prim at this point:
-       */
-      note_last_prim( rmesa, 0 );
-   }
-
-   /* Fire any buffered primitives
-    */
-   flush_prims( rmesa );
-
-   /* Get new buffer
-    */
-   radeonRefillCurrentDmaRegion( rmesa );
-
-   /* Reset counter, dmaptr
-    */
-   rmesa->vb.dmaptr = (int *)(rmesa->dma.current.ptr + rmesa->dma.current.address);
-   rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-      (rmesa->vb.vertex_size * 4);
-   rmesa->vb.counter--;
-   rmesa->vb.initial_counter = rmesa->vb.counter;
-   rmesa->vb.notify = wrap_buffer;
-
-   rmesa->dma.flush = flush_prims;
-
-   /* Restart wrapped primitive:
-    */
-   if (rmesa->vb.prim[0] != GL_POLYGON+1)
-      start_prim( rmesa, rmesa->vb.prim[0] );
-
-   /* Reemit saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      if (RADEON_DEBUG & DEBUG_VERTS) {
-	 int j;
-	 fprintf(stderr, "re-emit vertex %d to %p\n", i, (void *)rmesa->vb.dmaptr);
-	 if (RADEON_DEBUG & DEBUG_VERBOSE)
-	    for (j = 0 ; j < rmesa->vb.vertex_size; j++) 
-	       fprintf(stderr, "\t%08x/%f\n", *(int*)&tmp[i][j], tmp[i][j]);
-      }
-
-      memcpy( rmesa->vb.dmaptr, tmp[i], rmesa->vb.vertex_size * 4 );
-      rmesa->vb.dmaptr += rmesa->vb.vertex_size;
-      rmesa->vb.counter--;
-   }
-}
-
-
-
-static GLboolean check_vtx_fmt( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint ind = RADEON_CP_VC_FRMT_Z;
-   GLuint unit;
-
-   if (rmesa->TclFallback || rmesa->vb.fell_back || ctx->CompileFlag ||
-      (ctx->Fog.Enabled && (ctx->Fog.FogCoordinateSource == GL_FOG_COORD)))
-      return GL_FALSE;
-
-   if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) 
-      ctx->Driver.FlushVertices( ctx, FLUSH_UPDATE_CURRENT );
-   
-   /* Make all this event-driven:
-    */
-   if (ctx->Light.Enabled) {
-      ind |= RADEON_CP_VC_FRMT_N0;
-
-      /* TODO: make this data driven: If we receive only ubytes, send
-       * color as ubytes.  Also check if converting (with free
-       * checking for overflow) is cheaper than sending floats
-       * directly.
-       */
-      if (ctx->Light.ColorMaterialEnabled) {
-	 ind |= (RADEON_CP_VC_FRMT_FPCOLOR |
-		 RADEON_CP_VC_FRMT_FPALPHA);
-      }
-      else
-	 ind |= RADEON_CP_VC_FRMT_PKCOLOR; /* for alpha? */
-   }
-   else {
-      /* TODO: make this data driven?
-       */
-      ind |= RADEON_CP_VC_FRMT_PKCOLOR;
-	 
-      if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
-	 ind |= RADEON_CP_VC_FRMT_PKSPEC;
-      }
-   }
-
-   if ( ctx->Fog.FogCoordinateSource == GL_FOG_COORD ) {
-      ind |= RADEON_CP_VC_FRMT_PKSPEC;
-   }
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ctx->Texture.Unit[unit]._ReallyEnabled) {
-	 if (ctx->Texture.Unit[unit].TexGenEnabled) {
-	    if (rmesa->TexGenNeedNormals[unit]) {
-	       ind |= RADEON_CP_VC_FRMT_N0;
-	    }
-	 } else {
-	    if (ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][3] != 1.0) {
-	       if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-		  fprintf(stderr, "%s: q%u\n", __FUNCTION__, unit);
-	       return GL_FALSE;
-	    }
-	    ind |= RADEON_ST_BIT(unit);
-	 }
-      }
-   }
-
-   if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_STATE))
-      fprintf(stderr, "%s: format: 0x%x\n", __FUNCTION__, ind );
-
-   RADEON_NEWPRIM(rmesa);
-   rmesa->vb.vertex_format = ind;
-   rmesa->vb.vertex_size = 3;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-
-   rmesa->vb.normalptr = ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
-   rmesa->vb.colorptr = NULL;
-   rmesa->vb.floatcolorptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-   rmesa->vb.specptr = NULL;
-   rmesa->vb.floatspecptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
-   rmesa->vb.texcoordptr[0] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];
-   rmesa->vb.texcoordptr[1] = ctx->Current.Attrib[VERT_ATTRIB_TEX1];
-   rmesa->vb.texcoordptr[2] = ctx->Current.Attrib[VERT_ATTRIB_TEX2];
-   rmesa->vb.texcoordptr[3] = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; /* dummy */
-
-   /* Run through and initialize the vertex components in the order
-    * the hardware understands:
-    */
-   if (ind & RADEON_CP_VC_FRMT_N0) {
-      rmesa->vb.normalptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.normalptr[0] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0];
-      rmesa->vb.normalptr[1] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1];
-      rmesa->vb.normalptr[2] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2];
-   }
-
-   if (ind & RADEON_CP_VC_FRMT_PKCOLOR) {
-      rmesa->vb.colorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->alpha, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] );
-   }
-
-   if (ind & RADEON_CP_VC_FRMT_FPCOLOR) {
-      assert(!(ind & RADEON_CP_VC_FRMT_PKCOLOR));
-      rmesa->vb.floatcolorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.floatcolorptr[0] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0];
-      rmesa->vb.floatcolorptr[1] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1];
-      rmesa->vb.floatcolorptr[2] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2];
-
-      if (ind & RADEON_CP_VC_FRMT_FPALPHA) {
-	 rmesa->vb.vertex_size += 1;
-	 rmesa->vb.floatcolorptr[3] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-      }
-   }
-   
-   if (ind & RADEON_CP_VC_FRMT_PKSPEC) {
-      rmesa->vb.specptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] );
-      /* fog ??? */
-/*      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->alpha,
-			       radeonComputeFogFactor(ctx->Current.Attrib[VERT_ATTRIB_FOG][0]) ); */
-   }
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ind & RADEON_ST_BIT(unit)) {
-	 rmesa->vb.texcoordptr[unit] = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-	 rmesa->vb.vertex_size += 2;
-	 rmesa->vb.texcoordptr[unit][0] = ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][0];
-	 rmesa->vb.texcoordptr[unit][1] = ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][1];
-      }
-   }
-
-   if (rmesa->vb.installed_vertex_format != rmesa->vb.vertex_format) {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on vertex_format change\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      rmesa->vb.installed_vertex_format = rmesa->vb.vertex_format;
-   }
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s -- success\n", __FUNCTION__);
-   
-   return GL_TRUE;
-}
-
-void radeonVtxfmtInvalidate( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   rmesa->vb.recheck = GL_TRUE;
-   rmesa->vb.fell_back = GL_FALSE;
-}
-
-
-static void radeonVtxfmtValidate( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   rmesa->vb.recheck = GL_FALSE;
-
-   if (check_vtx_fmt( ctx )) {
-      if (!rmesa->vb.installed) {
-	 if (RADEON_DEBUG & DEBUG_VFMT)
-	    fprintf(stderr, "reinstall (new install)\n");
-
-	 _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-	 ctx->Driver.FlushVertices = radeonVtxfmtFlushVertices;
-	 ctx->Driver.NewList = radeonNewList;
-	 rmesa->vb.installed = GL_TRUE;
-      }
-      else if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: already installed", __FUNCTION__);
-   } 
-   else {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: failed\n", __FUNCTION__);
-
-      if (rmesa->vb.installed) {
-	 if (rmesa->dma.flush)
-	    rmesa->dma.flush( rmesa );
-	 _tnl_wakeup_exec( ctx );
-	 ctx->Driver.FlushVertices = radeonFlushVertices;
-	 ctx->Driver.NewList =_tnl_NewList;
-	 rmesa->vb.installed = GL_FALSE;
-      }
-   }      
-}
-
-
-
-/* Materials:
- */
-static void radeon_Materialfv( GLenum face, GLenum pname, 
-			       const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      VFMT_FALLBACK( __FUNCTION__ );
-      CALL_Materialfv(GET_DISPATCH(), (face, pname, params));
-      return;
-   }
-   _mesa_noop_Materialfv( face, pname, params );
-   radeonUpdateMaterial( ctx );
-}
-
-
-/* Begin/End
- */
-static void radeon_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s( %s )\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   if (mode > GL_POLYGON) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glBegin" );
-      return;
-   }
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-      return;
-   }
-   
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      radeonValidateState( ctx );
-
-   if (rmesa->vb.recheck) 
-      radeonVtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      CALL_Begin(GET_DISPATCH(), (mode));
-      return;
-   }
-
-
-   if (rmesa->dma.flush && rmesa->vb.counter < 12) {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: flush almost-empty buffers\n", __FUNCTION__);
-      flush_prims( rmesa );
-   }
-
-   /* Need to arrange to save vertices here?  Or always copy from dma (yuk)?
-    */
-   if (!rmesa->dma.flush) {
-/* FIXME: what are these constants? */
-      if (rmesa->dma.current.ptr + 12*rmesa->vb.vertex_size*4 > 
-	  rmesa->dma.current.end) {
-	 RADEON_NEWPRIM( rmesa );
-	 radeonRefillCurrentDmaRegion( rmesa );
-      }
-
-      rmesa->vb.dmaptr = (int *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-	 (rmesa->vb.vertex_size * 4);
-      rmesa->vb.counter--;
-      rmesa->vb.initial_counter = rmesa->vb.counter;
-      rmesa->vb.notify = wrap_buffer;
-      rmesa->dma.flush = flush_prims;
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   }
-   
-   
-   rmesa->vb.prim[0] = mode;
-   start_prim( rmesa, mode | PRIM_BEGIN );
-}
-
-
-
-static void radeon_End( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-      return;
-   }
-	  
-   note_last_prim( rmesa, PRIM_END );
-   rmesa->vb.prim[0] = GL_POLYGON+1;
-}
-
-
-/* Fallback on difficult entrypoints:
- */
-#define PRE_LOOPBACK( FUNC )			\
-do {						\
-   if (RADEON_DEBUG & DEBUG_VFMT) 		\
-      fprintf(stderr, "%s\n", __FUNCTION__);	\
-   VFMT_FALLBACK( __FUNCTION__ );		\
-} while (0)
-#define TAG(x) radeon_fallback_##x
-#include "vtxfmt_tmp.h"
-
-
-
-static GLboolean radeonNotifyBegin( GLcontext *ctx, GLenum p )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-   
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(!rmesa->vb.installed);
-
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      radeonValidateState( ctx );
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (rmesa->vb.recheck) 
-      radeonVtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s -- failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   radeon_Begin( p );
-   return GL_TRUE;
-}
-
-static void radeonVtxfmtFlushVertices( GLcontext *ctx, GLuint flags )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(rmesa->vb.installed);
-
-   if (flags & FLUSH_UPDATE_CURRENT) {
-      radeon_copy_to_current( ctx );
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on update_current\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-   }
-
-   if (flags & FLUSH_STORED_VERTICES) {
-      radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-      assert (rmesa->dma.flush == 0 ||
-	      rmesa->dma.flush == flush_prims);
-      if (rmesa->dma.flush == flush_prims)
-	 flush_prims( RADEON_CONTEXT( ctx ) );
-      ctx->Driver.NeedFlush &= ~FLUSH_STORED_VERTICES;
-   }
-}
-
-
-
-/* At this point, don't expect very many versions of each function to
- * be generated, so not concerned about freeing them?
- */
-
-
-void radeonVtxfmtInit( GLcontext *ctx, GLboolean useCodegen )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-   GLvertexformat *vfmt = &(rmesa->vb.vtxfmt);
-
-   /* start by initializing to no-op functions */
-   _mesa_noop_vtxfmt_init(vfmt);
-
-   /* Hook in chooser functions for codegen, etc:
-    */
-   radeonVtxfmtInitChoosers( vfmt );
-
-   /* Handled fully in supported states, but no codegen:
-    */
-   vfmt->Materialfv = radeon_Materialfv;
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = radeon_Begin;
-   vfmt->End = radeon_End;
-
-   /* Fallback for performance reasons:  (Fix with cva/elt path here and
-    * dmatmp2.h style primitive-merging)
-    *
-    * These should call NotifyBegin(), as should _tnl_EvalMesh, to allow
-    * a driver-hook.
-    */
-   vfmt->DrawArrays = radeon_fallback_DrawArrays;
-   vfmt->DrawElements = radeon_fallback_DrawElements;
-   vfmt->DrawRangeElements = radeon_fallback_DrawRangeElements; 
-
-   /* Active but unsupported -- fallback if we receive these:
-    */
-   vfmt->CallList = radeon_fallback_CallList;
-   vfmt->CallLists = radeon_fallback_CallLists;
-   vfmt->EvalCoord1f = radeon_fallback_EvalCoord1f;
-   vfmt->EvalCoord1fv = radeon_fallback_EvalCoord1fv;
-   vfmt->EvalCoord2f = radeon_fallback_EvalCoord2f;
-   vfmt->EvalCoord2fv = radeon_fallback_EvalCoord2fv;
-   vfmt->EvalMesh1 = radeon_fallback_EvalMesh1;
-   vfmt->EvalMesh2 = radeon_fallback_EvalMesh2;
-   vfmt->EvalPoint1 = radeon_fallback_EvalPoint1;
-   vfmt->EvalPoint2 = radeon_fallback_EvalPoint2;
-   vfmt->TexCoord3f = radeon_fallback_TexCoord3f;
-   vfmt->TexCoord3fv = radeon_fallback_TexCoord3fv;
-   vfmt->TexCoord4f = radeon_fallback_TexCoord4f;
-   vfmt->TexCoord4fv = radeon_fallback_TexCoord4fv;
-   vfmt->MultiTexCoord3fARB = radeon_fallback_MultiTexCoord3fARB;
-   vfmt->MultiTexCoord3fvARB = radeon_fallback_MultiTexCoord3fvARB;
-   vfmt->MultiTexCoord4fARB = radeon_fallback_MultiTexCoord4fARB;
-   vfmt->MultiTexCoord4fvARB = radeon_fallback_MultiTexCoord4fvARB;
-   vfmt->Vertex4f = radeon_fallback_Vertex4f;
-   vfmt->Vertex4fv = radeon_fallback_Vertex4fv;
-   vfmt->VertexAttrib1fNV  = radeon_fallback_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = radeon_fallback_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV  = radeon_fallback_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = radeon_fallback_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV  = radeon_fallback_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = radeon_fallback_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV  = radeon_fallback_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = radeon_fallback_VertexAttrib4fvNV;
-   vfmt->FogCoordfEXT = radeon_fallback_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = radeon_fallback_FogCoordfvEXT;
-
-   (void)radeon_fallback_vtxfmt;
-
-   TNL_CONTEXT(ctx)->Driver.NotifyBegin = radeonNotifyBegin;
-
-   rmesa->vb.enabled = 1;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-   rmesa->vb.primflags = 0;
-
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1fv );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-
-   radeonInitCodegen( &rmesa->vb.codegen, useCodegen );
-}
-
-static void free_funcs( struct dynfn *l )
-{
-   struct dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      _mesa_exec_free( f->code );
-      _mesa_free( f );
-   }
-}
-
-
-
-void radeonVtxfmtMakeCurrent( GLcontext *ctx )
-{
-}
-
-
-void radeonVtxfmtDestroy( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   count_funcs( rmesa );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2fv );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4f );
-   free_funcs( &rmesa->vb.dfn_cache.Color4fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3f );
-   free_funcs( &rmesa->vb.dfn_cache.Color3fv );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3f );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1fv );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-}
-
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h b/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h
deleted file mode 100644
index a656e49e44..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h
+++ /dev/null
@@ -1,120 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt.h,v 1.3 2002/12/21 17:02:16 dawes Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __RADEON_VTXFMT_H__
-#define __RADEON_VTXFMT_H__
-
-#include "radeon_context.h"
-
-
-extern void radeonVtxfmtUpdate( GLcontext *ctx );
-extern void radeonVtxfmtInit( GLcontext *ctx, GLboolean useCodegen );
-extern void radeonVtxfmtInvalidate( GLcontext *ctx );
-extern void radeonVtxfmtDestroy( GLcontext *ctx );
-extern void radeonVtxfmtInitChoosers( GLvertexformat *vfmt );
-
-extern void radeonVtxfmtMakeCurrent( GLcontext *ctx );
-extern void radeonVtxfmtUnbindContext( GLcontext *ctx );
-
-extern void radeon_copy_to_current( GLcontext *ctx );
-
-#define DFN( FUNC, CACHE)				\
-do {							\
-   char *start = (char *)&FUNC;				\
-   char *end = (char *)&FUNC##_end;			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key = key;					\
-   dfn->code = _mesa_exec_malloc( end - start );	\
-   _mesa_memcpy(dfn->code, start, end - start);		\
-}							\
-while ( 0 )
-
-#define FIXUP( CODE, OFFSET, CHECKVAL, NEWVAL )	\
-do {						\
-   int *icode = (int *)(CODE+OFFSET);		\
-   assert (*icode == CHECKVAL);			\
-   *icode = (int)NEWVAL;			\
-} while (0)
-
-
-/* Useful for figuring out the offsets:
- */
-#define FIXUP2( CODE, OFFSET, CHECKVAL, NEWVAL )		\
-do {								\
-   while (*(int *)(CODE+OFFSET) != CHECKVAL) OFFSET++;		\
-   fprintf(stderr, "%s/%d CVAL %x OFFSET %d VAL %x\n", __FUNCTION__,	\
-	   __LINE__, CHECKVAL, OFFSET, (int)(NEWVAL));			\
-   *(int *)(CODE+OFFSET) = (int)(NEWVAL);				\
-   OFFSET += 4;							\
-} while (0)
-
-/* 
- */
-void radeonInitCodegen( struct dfn_generators *gen, GLboolean useCodegen );
-void radeonInitX86Codegen( struct dfn_generators *gen );
-void radeonInitSSECodegen( struct dfn_generators *gen );
-
-
-
-/* Defined in radeon_vtxfmt_x86.c
- */
-struct dynfn *radeon_makeX86Vertex2f( GLcontext *, int );
-struct dynfn *radeon_makeX86Vertex2fv( GLcontext *, int );
-struct dynfn *radeon_makeX86Vertex3f( GLcontext *, int );
-struct dynfn *radeon_makeX86Vertex3fv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4ub( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4ubv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3ub( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3ubv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4f( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4fv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3f( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3fv( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3ubEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3ubvEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3fEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3fvEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86Normal3f( GLcontext *, int );
-struct dynfn *radeon_makeX86Normal3fv( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord2f( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord2fv( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord1f( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord1fv( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord2fARB( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord2fvARB( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord1fARB( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord1fvARB( GLcontext *, int );
-
-#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c
deleted file mode 100644
index aac029aa0e..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c
+++ /dev/null
@@ -1,924 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt_c.c,v 1.2 2002/12/16 16:18:59 dawes Exp $ */
-/**************************************************************************
-
-Copyright 2002 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-#include "glheader.h"
-#include "mtypes.h"
-#include "colormac.h"
-#include "simple_list.h"
-#include "api_noop.h"
-#include "vtxfmt.h"
-
-#include "radeon_vtxfmt.h"
-
-#include "dispatch.h"
-
-/* Fallback versions of all the entrypoints for situations where
- * codegen isn't available.  This is still a lot faster than the
- * vb/pipeline implementation in Mesa.
- */
-static void radeon_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = *(int *)&z;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void radeon_Vertex3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = *(int *)&v[2];
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void radeon_Vertex2f( GLfloat x, GLfloat y )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = *(int *)&rmesa->vb.vertex[i];
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void radeon_Vertex2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-#if 0
-/* Color for ubyte (packed) color formats:
- */
-static void radeon_Color3ub_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void radeon_Color3ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-
-static void radeon_Color4ub_ub( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= a;
-}
-
-static void radeon_Color4ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   *(GLuint *)rmesa->vb.colorptr = LE32_TO_CPU(*(GLuint *)v);
-}
-#endif /* 0 */
-
-static void radeon_Color3f_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void radeon_Color3fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-static void radeon_Color4f_ub( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, a );
-}
-
-static void radeon_Color4fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, v[3] );
-}
-
-
-/* Color for float color+alpha formats:
- */
-#if 0
-static void radeon_Color3ub_4f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void radeon_Color3ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-
-static void radeon_Color4ub_4f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = UBYTE_TO_FLOAT(a);
-}
-
-static void radeon_Color4ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void radeon_Color3f_4f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;		
-}
-
-static void radeon_Color3fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-static void radeon_Color4f_4f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = a;
-}
-
-static void radeon_Color4fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = v[3];
-}
-
-
-/* Color for float color formats:
- */
-#if 0
-static void radeon_Color3ub_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-}
-
-static void radeon_Color3ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-}
-
-static void radeon_Color4ub_3f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(a);
-}
-
-static void radeon_Color4ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void radeon_Color3f_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-}
-
-static void radeon_Color3fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-static void radeon_Color4f_3f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = a;
-}
-
-static void radeon_Color4fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = v[3]; 
-}
-
-
-/* Secondary Color:
- */
-#if 0
-static void radeon_SecondaryColor3ubEXT_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void radeon_SecondaryColor3ubvEXT_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-#endif /* 0 */
-
-static void radeon_SecondaryColor3fEXT_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void radeon_SecondaryColor3fvEXT_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-#if 0
-static void radeon_SecondaryColor3ubEXT_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void radeon_SecondaryColor3ubvEXT_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-#endif /* 0 */
-
-static void radeon_SecondaryColor3fEXT_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;
-}
-
-static void radeon_SecondaryColor3fvEXT_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-
-/* Normal
- */
-static void radeon_Normal3f( GLfloat n0, GLfloat n1, GLfloat n2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = n0;
-   dest[1] = n1;
-   dest[2] = n2;
-}
-
-static void radeon_Normal3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-
-/* TexCoord
- */
-static void radeon_TexCoord1f( GLfloat s )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = s;
-   dest[1] = 0;
-}
-
-static void radeon_TexCoord1fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = v[0];
-   dest[1] = 0;
-}
-
-static void radeon_TexCoord2f( GLfloat s, GLfloat t )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = s;
-   dest[1] = t;
-}
-
-static void radeon_TexCoord2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = v[0];
-   dest[1] = v[1];
-}
-
-
-/* MultiTexcoord
- * 
- * Technically speaking, these functions should subtract GL_TEXTURE0 from
- * \c target before masking and using it.  The value of GL_TEXTURE0 is 0x84C0,
- * which has the low-order 5 bits 0.  For all possible valid values of 
- * \c target.  Subtracting GL_TEXTURE0 has the net effect of masking \c target
- * with 0x1F.  Masking with 0x1F and then masking with 0x01 is redundant, so
- * the subtraction has been omitted.
- */
-
-static void radeon_MultiTexCoord1fARB( GLenum target, GLfloat s  )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = s;
-   dest[1] = 0;
-}
-
-static void radeon_MultiTexCoord1fvARB( GLenum target, const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = v[0];
-   dest[1] = 0;
-}
-
-static void radeon_MultiTexCoord2fARB( GLenum target, GLfloat s, GLfloat t )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = s;
-   dest[1] = t;
-}
-
-static void radeon_MultiTexCoord2fvARB( GLenum target, const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = v[0];
-   dest[1] = v[1];
-}
-
-static struct dynfn *lookup( struct dynfn *l, int key )
-{
-   struct dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key == key) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-/* Can't use the loopback template for this:
- */
-
-#define CHOOSE(FN, FNTYPE, MASK, ACTIVE, ARGS1, ARGS2 )			\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
-   struct dynfn *dfn;							\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );				\
-   else if (RADEON_DEBUG & DEBUG_CODEGEN)				\
-      fprintf(stderr, "%s -- cached codegen\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (RADEON_DEBUG & DEBUG_CODEGEN)					\
-	 fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, radeon_##FN);				\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* For the _3f case, only allow one color function to be hooked in at
- * a time.  Eventually, use a similar mechanism to allow selecting the
- * color component of the vertex format based on client behaviour.  
- *
- * Note:  Perform these actions even if there is a codegen or cached 
- * codegen version of the chosen function.
- */
-#define CHOOSE_COLOR(FN, FNTYPE, NR, MASK, ACTIVE, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx); \
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
-   struct dynfn *dfn;							\
-									\
-   if (rmesa->vb.vertex_format & ACTIVE_PKCOLOR) {			\
-      SET_ ## FN (ctx->Exec, radeon_##FN##_ub);				\
-   }									\
-   else if ((rmesa->vb.vertex_format &					\
-            (ACTIVE_FPCOLOR|ACTIVE_FPALPHA)) == ACTIVE_FPCOLOR) {	\
-									\
-      if (rmesa->vb.installed_color_3f_sz != NR) {			\
-         rmesa->vb.installed_color_3f_sz = NR;				\
-         if (NR == 3) ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = 1.0;	\
-         if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) {		\
-            radeon_copy_to_current( ctx );				\
-            _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );	\
-            CALL_ ## FN (ctx->Exec, ARGS2);				\
-            return;							\
-         }								\
-      }									\
-									\
-      SET_ ## FN (ctx->Exec, radeon_##FN##_3f);				\
-   }									\
-   else {								\
-      SET_ ## FN (ctx->Exec, radeon_##FN##_4f);				\
-   }									\
-									\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (!dfn) dfn = rmesa->vb.codegen.FN( ctx, key );			\
-									\
-   if (dfn) {								\
-      if (RADEON_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- codegen version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, (FNTYPE)dfn->code);			\
-   }									\
-   else if (RADEON_DEBUG & DEBUG_CODEGEN)				\
-         fprintf(stderr, "%s -- 'c' version\n", __FUNCTION__ );		\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* Right now there are both _ub and _3f versions of the secondary color
- * functions.  Currently, we only set-up the hardware to use the _ub versions.
- * The _3f versions are needed for the cases where secondary color isn't used
- * in the vertex format, but it still needs to be stored in the context
- * state vector.
- */
-#define CHOOSE_SECONDARY_COLOR(FN, FNTYPE, MASK, ACTIVE, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
-   struct dynfn *dfn = lookup( &rmesa->vb.dfn_cache.FN, key );		\
-									\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );				\
-   else  if (RADEON_DEBUG & DEBUG_CODEGEN)				\
-      fprintf(stderr, "%s -- cached version\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (RADEON_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, ((rmesa->vb.vertex_format & ACTIVE_PKSPEC) != 0)	\
-	  ? radeon_##FN##_ub : radeon_##FN##_3f);			\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-
-
-/* Shorthands
- */
-#define ACTIVE_XYZW (RADEON_CP_VC_FRMT_W0|RADEON_CP_VC_FRMT_Z)
-#define ACTIVE_NORM RADEON_CP_VC_FRMT_N0
-
-#define ACTIVE_PKCOLOR RADEON_CP_VC_FRMT_PKCOLOR
-#define ACTIVE_FPCOLOR RADEON_CP_VC_FRMT_FPCOLOR
-#define ACTIVE_FPALPHA RADEON_CP_VC_FRMT_FPALPHA
-#define ACTIVE_COLOR (ACTIVE_FPCOLOR|ACTIVE_PKCOLOR)
-
-#define ACTIVE_PKSPEC RADEON_CP_VC_FRMT_PKSPEC
-#define ACTIVE_FPSPEC RADEON_CP_VC_FRMT_FPSPEC
-#define ACTIVE_SPEC   (ACTIVE_FPSPEC|ACTIVE_PKSPEC)
-
-#define ACTIVE_ST0 RADEON_CP_VC_FRMT_ST0
-#define ACTIVE_ST1 RADEON_CP_VC_FRMT_ST1
-#define ACTIVE_ST2 RADEON_CP_VC_FRMT_ST2
-#define ACTIVE_ST_ALL (RADEON_CP_VC_FRMT_ST1|RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST2)
-
-/* Each codegen function should be able to be fully specified by a
- * subsetted version of rmesa->vb.vertex_format.
- */
-#define MASK_NORM    (ACTIVE_XYZW)
-#define MASK_COLOR   (MASK_NORM|ACTIVE_NORM)
-#define MASK_SPEC    (MASK_COLOR|ACTIVE_COLOR)
-#define MASK_ST0     (MASK_SPEC|ACTIVE_SPEC)
-#define MASK_ST1     (MASK_ST0|ACTIVE_ST0)
-#define MASK_ST2     (MASK_ST1|ACTIVE_ST1)
-#define MASK_ST_ALL  (MASK_ST2|ACTIVE_ST2)
-#define MASK_VERTEX  (MASK_ST_ALL|ACTIVE_FPALPHA) 
-
-
-typedef void (*p4f)( GLfloat, GLfloat, GLfloat, GLfloat );
-typedef void (*p3f)( GLfloat, GLfloat, GLfloat );
-typedef void (*p2f)( GLfloat, GLfloat );
-typedef void (*p1f)( GLfloat );
-typedef void (*pe2f)( GLenum, GLfloat, GLfloat );
-typedef void (*pe1f)( GLenum, GLfloat );
-typedef void (*p4ub)( GLubyte, GLubyte, GLubyte, GLubyte );
-typedef void (*p3ub)( GLubyte, GLubyte, GLubyte );
-typedef void (*pfv)( const GLfloat * );
-typedef void (*pefv)( GLenum, const GLfloat * );
-typedef void (*pubv)( const GLubyte * );
-
-
-CHOOSE(Normal3f, p3f, MASK_NORM, ACTIVE_NORM, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Normal3fv, pfv, MASK_NORM, ACTIVE_NORM, 
-       (const GLfloat *v), (v))
-
-#if 0
-CHOOSE_COLOR(Color4ub, p4ub, 4, MASK_COLOR, ACTIVE_COLOR,
-	(GLubyte a,GLubyte b, GLubyte c, GLubyte d), (a,b,c,d))
-CHOOSE_COLOR(Color4ubv, pubv, 4, MASK_COLOR, ACTIVE_COLOR, 
-	(const GLubyte *v), (v))
-CHOOSE_COLOR(Color3ub, p3ub, 3, MASK_COLOR, ACTIVE_COLOR, 
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_COLOR(Color3ubv, pubv, 3, MASK_COLOR, ACTIVE_COLOR, 
-	(const GLubyte *v), (v))
-#endif
-
-CHOOSE_COLOR(Color4f, p4f, 4, MASK_COLOR, ACTIVE_COLOR, 
-	(GLfloat a,GLfloat b, GLfloat c, GLfloat d), (a,b,c,d))
-CHOOSE_COLOR(Color4fv, pfv, 4, MASK_COLOR, ACTIVE_COLOR, 
-	(const GLfloat *v), (v))
-CHOOSE_COLOR(Color3f, p3f, 3, MASK_COLOR, ACTIVE_COLOR,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_COLOR(Color3fv, pfv, 3, MASK_COLOR, ACTIVE_COLOR,
-	(const GLfloat *v), (v))
-
-
-#if 0
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubEXT, p3ub, MASK_SPEC, ACTIVE_SPEC,
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubvEXT, pubv, MASK_SPEC, ACTIVE_SPEC,
-	(const GLubyte *v), (v))
-#endif
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fEXT, p3f, MASK_SPEC, ACTIVE_SPEC,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fvEXT, pfv, MASK_SPEC, ACTIVE_SPEC,
-	(const GLfloat *v), (v))
-
-CHOOSE(TexCoord2f, p2f, MASK_ST0, ACTIVE_ST0, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(TexCoord2fv, pfv, MASK_ST0, ACTIVE_ST0, 
-       (const GLfloat *v), (v))
-CHOOSE(TexCoord1f, p1f, MASK_ST0, ACTIVE_ST0, 
-       (GLfloat a), (a))
-CHOOSE(TexCoord1fv, pfv, MASK_ST0, ACTIVE_ST0, 
-       (const GLfloat *v), (v))
-
-CHOOSE(MultiTexCoord2fARB, pe2f, MASK_ST_ALL, ACTIVE_ST_ALL,
-	 (GLenum u,GLfloat a,GLfloat b), (u,a,b))
-CHOOSE(MultiTexCoord2fvARB, pefv, MASK_ST_ALL, ACTIVE_ST_ALL,
-	(GLenum u,const GLfloat *v), (u,v))
-CHOOSE(MultiTexCoord1fARB, pe1f, MASK_ST_ALL, ACTIVE_ST_ALL,
-	 (GLenum u,GLfloat a), (u,a))
-CHOOSE(MultiTexCoord1fvARB, pefv, MASK_ST_ALL, ACTIVE_ST_ALL,
-	(GLenum u,const GLfloat *v), (u,v))
-
-CHOOSE(Vertex3f, p3f, MASK_VERTEX, MASK_VERTEX, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Vertex3fv, pfv, MASK_VERTEX, MASK_VERTEX, 
-       (const GLfloat *v), (v))
-CHOOSE(Vertex2f, p2f, MASK_VERTEX, MASK_VERTEX, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(Vertex2fv, pfv, MASK_VERTEX, MASK_VERTEX, 
-       (const GLfloat *v), (v))
-
-
-
-
-
-void radeonVtxfmtInitChoosers( GLvertexformat *vfmt )
-{
-   vfmt->Color3f = choose_Color3f;
-   vfmt->Color3fv = choose_Color3fv;
-   vfmt->Color4f = choose_Color4f;
-   vfmt->Color4fv = choose_Color4fv;
-   vfmt->SecondaryColor3fEXT = choose_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = choose_SecondaryColor3fvEXT;
-   vfmt->MultiTexCoord1fARB = choose_MultiTexCoord1fARB;
-   vfmt->MultiTexCoord1fvARB = choose_MultiTexCoord1fvARB;
-   vfmt->MultiTexCoord2fARB = choose_MultiTexCoord2fARB;
-   vfmt->MultiTexCoord2fvARB = choose_MultiTexCoord2fvARB;
-   vfmt->Normal3f = choose_Normal3f;
-   vfmt->Normal3fv = choose_Normal3fv;
-   vfmt->TexCoord1f = choose_TexCoord1f;
-   vfmt->TexCoord1fv = choose_TexCoord1fv;
-   vfmt->TexCoord2f = choose_TexCoord2f;
-   vfmt->TexCoord2fv = choose_TexCoord2fv;
-   vfmt->Vertex2f = choose_Vertex2f;
-   vfmt->Vertex2fv = choose_Vertex2fv;
-   vfmt->Vertex3f = choose_Vertex3f;
-   vfmt->Vertex3fv = choose_Vertex3fv;
-
-#if 0
-   vfmt->Color3ub = choose_Color3ub;
-   vfmt->Color3ubv = choose_Color3ubv;
-   vfmt->Color4ub = choose_Color4ub;
-   vfmt->Color4ubv = choose_Color4ubv;
-   vfmt->SecondaryColor3ubEXT = choose_SecondaryColor3ubEXT;
-   vfmt->SecondaryColor3ubvEXT = choose_SecondaryColor3ubvEXT;
-#endif
-}
-
-
-static struct dynfn *codegen_noop( GLcontext *ctx, int key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void radeonInitCodegen( struct dfn_generators *gen, GLboolean useCodegen )
-{
-   gen->Vertex3f = codegen_noop;
-   gen->Vertex3fv = codegen_noop;
-   gen->Color4ub = codegen_noop;
-   gen->Color4ubv = codegen_noop;
-   gen->Normal3f = codegen_noop;
-   gen->Normal3fv = codegen_noop;
-   gen->TexCoord2f = codegen_noop;
-   gen->TexCoord2fv = codegen_noop;
-   gen->MultiTexCoord2fARB = codegen_noop;
-   gen->MultiTexCoord2fvARB = codegen_noop;
-   gen->Vertex2f = codegen_noop;
-   gen->Vertex2fv = codegen_noop;
-   gen->Color3ub = codegen_noop;
-   gen->Color3ubv = codegen_noop;
-   gen->Color4f = codegen_noop;
-   gen->Color4fv = codegen_noop;
-   gen->Color3f = codegen_noop;
-   gen->Color3fv = codegen_noop;
-   gen->SecondaryColor3fEXT = codegen_noop;
-   gen->SecondaryColor3fvEXT = codegen_noop;
-   gen->SecondaryColor3ubEXT = codegen_noop;
-   gen->SecondaryColor3ubvEXT = codegen_noop;
-   gen->TexCoord1f = codegen_noop;
-   gen->TexCoord1fv = codegen_noop;
-   gen->MultiTexCoord1fARB = codegen_noop;
-   gen->MultiTexCoord1fvARB = codegen_noop;
-
-   if (useCodegen) {
-#if defined(USE_X86_ASM)
-      radeonInitX86Codegen( gen );
-#endif
-
-#if defined(USE_SSE_ASM)
-      radeonInitSSECodegen( gen );
-#endif
-   }
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c
deleted file mode 100644
index 0f0fc9e065..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c
+++ /dev/null
@@ -1,236 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt_sse.c,v 1.1 2002/10/30 12:51:58 alanh Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "radeon_vtxfmt.h"
-
-#if defined(USE_SSE_ASM)
-#include "x86/common_x86_asm.h"
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN( _sse_Attribute2fv );
-EXTERN( _sse_Attribute2f );
-EXTERN( _sse_Attribute3fv );
-EXTERN( _sse_Attribute3f );
-EXTERN( _sse_MultiTexCoord2fv );
-EXTERN( _sse_MultiTexCoord2f );
-EXTERN( _sse_MultiTexCoord2fv_2 );
-EXTERN( _sse_MultiTexCoord2f_2 );
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.
- */
-
-static struct dynfn *radeon_makeSSEAttribute2fv( struct dynfn * cache, int key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 10, 0x0, (int)dest);
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEAttribute2f( struct dynfn * cache, int key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 8, 0x0, (int)dest); 
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEAttribute3fv( struct dynfn * cache, int key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 13, 0x0, (int)dest);
-   FIXUP(dfn->code, 18, 0x8, 8+(int)dest);
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEAttribute3f( struct dynfn * cache, int key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 12, 0x0, (int)dest); 
-   FIXUP(dfn->code, 17, 0x8, 8+(int)dest); 
-   return dfn;
-}
-
-static struct dynfn * radeon_makeSSENormal3fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				      __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *radeon_makeSSENormal3f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				     __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *radeon_makeSSEColor3fv( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-					 __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-static struct dynfn *radeon_makeSSEColor3f( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-					__FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-static struct dynfn *radeon_makeSSETexCoord2fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				       __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-static struct dynfn *radeon_makeSSETexCoord2f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				     __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-static struct dynfn *radeon_makeSSEMultiTexCoord2fv( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-      (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _sse_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 18, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEMultiTexCoord2f( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-      (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _sse_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 16, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 15, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-#endif
-
-void radeonInitSSECodegen( struct dfn_generators *gen )
-{
-   if ( cpu_has_xmm ) {
-      gen->Normal3fv = (void *) radeon_makeSSENormal3fv;
-      gen->Normal3f = (void *) radeon_makeSSENormal3f;
-      gen->Color3fv = (void *) radeon_makeSSEColor3fv;
-      gen->Color3f = (void *) radeon_makeSSEColor3f;
-      gen->TexCoord2fv = (void *) radeon_makeSSETexCoord2fv;
-      gen->TexCoord2f = (void *) radeon_makeSSETexCoord2f;
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-      gen->MultiTexCoord2fvARB = (void *) radeon_makeSSEMultiTexCoord2fv;
-      gen->MultiTexCoord2fARB = (void *) radeon_makeSSEMultiTexCoord2f;
-#endif
-   }
-}
-
-#else 
-
-void radeonInitSSECodegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c
deleted file mode 100644
index 529e79065e..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c
+++ /dev/null
@@ -1,440 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt_x86.c,v 1.2 2002/12/21 17:02:16 dawes Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
-
-The above copyright notice and this permission notice (including the
-next paragraph) shall be included in all copies or substantial
-portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
-IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "radeon_vtxfmt.h"
-
-#if defined(USE_X86_ASM)
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN ( _x86_Attribute2fv );
-EXTERN ( _x86_Attribute2f );
-EXTERN ( _x86_Attribute3fv );
-EXTERN ( _x86_Attribute3f );
-EXTERN ( _x86_Vertex3fv_6 );
-EXTERN ( _x86_Vertex3fv_8 );
-EXTERN ( _x86_Vertex3fv );
-EXTERN ( _x86_Vertex3f_4 );
-EXTERN ( _x86_Vertex3f_6 );
-EXTERN ( _x86_Vertex3f );
-EXTERN ( _x86_Color4ubv_ub );
-EXTERN ( _x86_Color4ubv_4f );
-EXTERN ( _x86_Color4ub_ub );
-EXTERN ( _x86_MultiTexCoord2fv );
-EXTERN ( _x86_MultiTexCoord2fv_2 );
-EXTERN ( _x86_MultiTexCoord2f );
-EXTERN ( _x86_MultiTexCoord2f_2 );
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-struct dynfn *radeon_makeX86Vertex3f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x %d\n", __FUNCTION__, key, rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 4: {
-
-      DFN ( _x86_Vertex3f_4, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 2, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 25, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 36, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 46, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 51, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 60, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   case 6: {
-
-      DFN ( _x86_Vertex3f_6, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 28, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 34, 0x0, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 40, 0x0, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 57, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 63, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 70, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 79, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   default: {
-
-      DFN ( _x86_Vertex3f, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 9, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 37, 0x0, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 44, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 50, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 56, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-
-
-struct dynfn *radeon_makeX86Vertex3fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x %d\n", __FUNCTION__, key, rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 6: {
-
-      DFN ( _x86_Vertex3fv_6, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x00000024, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 56, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 61, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 76, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-   case 8: {
-
-      DFN ( _x86_Vertex3fv_8, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x0000001c, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 51, 0x00000020, (int)&rmesa->vb.vertex[6]);
-      FIXUP(dfn->code, 63, 0x00000024, (int)&rmesa->vb.vertex[7]);
-      FIXUP(dfn->code, 74, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 79, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 85, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 94, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-
-   default: {
-
-      DFN ( _x86_Vertex3fv, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 8, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 32, 0x00000006, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 37, 0x00000058, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 45, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 50, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 58, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-static struct dynfn *
-radeon_makeX86Attribute2fv( struct dynfn * cache, int key,
-			    const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 11, 0x0, (int)dest); 
-   FIXUP(dfn->code, 16, 0x4, 4+(int)dest); 
-
-   return dfn;
-}
-
-static struct dynfn *
-radeon_makeX86Attribute2f( struct dynfn * cache, int key,
-			   const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 1, 0x0, (int)dest); 
-
-   return dfn;
-}
-
-
-static struct dynfn *
-radeon_makeX86Attribute3fv( struct dynfn * cache, int key,
-			    const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-static struct dynfn *
-radeon_makeX86Attribute3f( struct dynfn * cache, int key,
-			 const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-struct dynfn *radeon_makeX86Normal3fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				      __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *radeon_makeX86Normal3f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				     __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *radeon_makeX86Color4ubv( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if (key & RADEON_CP_VC_FRMT_PKCOLOR) {
-      DFN ( _x86_Color4ubv_ub, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 5, 0x12345678, (int)rmesa->vb.colorptr); 
-      return dfn;
-   } 
-   else {
-
-      DFN ( _x86_Color4ubv_4f, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 2, 0x00000000, (int)_mesa_ubyte_to_float_color_tab); 
-      FIXUP(dfn->code, 27, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr); 
-      FIXUP(dfn->code, 33, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+4); 
-      FIXUP(dfn->code, 55, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+8); 
-      FIXUP(dfn->code, 61, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+12); 
-      return dfn;
-   }
-}
-
-struct dynfn *radeon_makeX86Color4ub( GLcontext *ctx, int key )
-{
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if (key & RADEON_CP_VC_FRMT_PKCOLOR) {
-      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      DFN ( _x86_Color4ub_ub, rmesa->vb.dfn_cache.Color4ub );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.colorptr); 
-      FIXUP(dfn->code, 24, 0x0, (int)rmesa->vb.colorptr+1); 
-      FIXUP(dfn->code, 30, 0x0, (int)rmesa->vb.colorptr+2); 
-      FIXUP(dfn->code, 36, 0x0, (int)rmesa->vb.colorptr+3); 
-      return dfn;
-   }
-   else
-      return NULL;
-}
-
-
-struct dynfn *radeon_makeX86Color3fv( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-					 __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-struct dynfn *radeon_makeX86Color3f( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeX86Attribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-					__FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-
-
-struct dynfn *radeon_makeX86TexCoord2fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				      __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-struct dynfn *radeon_makeX86TexCoord2f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				     __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-struct dynfn *radeon_makeX86MultiTexCoord2fvARB( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-      (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _x86_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 21, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);
-      FIXUP(dfn->code, 27, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4);
-   } else {
-      DFN ( _x86_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-struct dynfn *radeon_makeX86MultiTexCoord2fARB( GLcontext *ctx, 
-						int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-       (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _x86_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 20, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);
-      FIXUP(dfn->code, 26, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4); 
-   }
-   else {
-      /* Note: this might get generated multiple times, even though the
-       * actual emitted code is the same.
-       */
-      DFN ( _x86_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.texcoordptr); 
-   }      
-   return dfn;
-}
-#endif
-
-void radeonInitX86Codegen( struct dfn_generators *gen )
-{
-   gen->Vertex3f = radeon_makeX86Vertex3f;
-   gen->Vertex3fv = radeon_makeX86Vertex3fv;
-   gen->Color4ub = radeon_makeX86Color4ub; /* PKCOLOR only */
-   gen->Color4ubv = radeon_makeX86Color4ubv; /* PKCOLOR only */
-   gen->Normal3f = radeon_makeX86Normal3f;
-   gen->Normal3fv = radeon_makeX86Normal3fv;
-   gen->TexCoord2f = radeon_makeX86TexCoord2f;
-   gen->TexCoord2fv = radeon_makeX86TexCoord2fv;
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-   gen->MultiTexCoord2fARB = radeon_makeX86MultiTexCoord2fARB;
-   gen->MultiTexCoord2fvARB = radeon_makeX86MultiTexCoord2fvARB;
-#endif
-   gen->Color3f = radeon_makeX86Color3f;
-   gen->Color3fv = radeon_makeX86Color3fv;
-
-   /* Not done:
-    */
-/*     gen->Vertex2f = radeon_makeX86Vertex2f; */
-/*     gen->Vertex2fv = radeon_makeX86Vertex2fv; */
-/*     gen->Color3ub = radeon_makeX86Color3ub; */
-/*     gen->Color3ubv = radeon_makeX86Color3ubv; */
-/*     gen->Color4f = radeon_makeX86Color4f; */
-/*     gen->Color4fv = radeon_makeX86Color4fv; */
-/*     gen->TexCoord1f = radeon_makeX86TexCoord1f; */
-/*     gen->TexCoord1fv = radeon_makeX86TexCoord1fv; */
-/*     gen->MultiTexCoord1fARB = radeon_makeX86MultiTexCoord1fARB; */
-/*     gen->MultiTexCoord1fvARB = radeon_makeX86MultiTexCoord1fvARB; */
-}
-
-
-#else 
-
-void radeonInitX86Codegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxtmp_x86.S b/src/mesa/drivers/dri/radeon/radeon_vtxtmp_x86.S
deleted file mode 100644
index 1b433491aa..0000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxtmp_x86.S
+++ /dev/null
@@ -1,498 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxtmp_x86.S,v 1.1 2002/10/30 12:51:58 alanh Exp $ */
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-#define GLOBL( x )	\
-.globl x;		\
-x:
-
-.data
-.align 4
-
-/*
-	vertex 3f vertex size 4
-*/
-	
-GLOBL ( _x86_Vertex3f_4 )
-	movl	(0), %ecx
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	%eax, (%ecx)
-	movl	%edx, 4(%ecx)
-	movl	12(%esp), %eax
-	movl	(0), %edx
-	movl	%eax, 8(%ecx)
-	movl	%edx, 12(%ecx)
-	movl	(0), %eax
-	addl	$16, %ecx
-	dec 	%eax
-	movl	%ecx, (0)
-	movl	%eax, (0)
-	je	.1 
-	ret
-.1:	jmp	*0
-	
-GLOBL ( _x86_Vertex3f_4_end )
-
-/*
-	vertex 3f vertex size 6
-*/
-GLOBL ( _x86_Vertex3f_6 )
-	push	%edi
-	movl	(0), %edi
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	movl	(0), %eax
-	movl	(0), %edx
-	movl	(0), %ecx
-	movl	%eax, 12(%edi)
-	movl	%edx, 16(%edi)
-	movl	%ecx, 20(%edi)
-	addl	$24, %edi
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	pop 	%edi
-	movl	%eax, (0)
-	je	.2
-	ret
-.2:	jmp	*0
-GLOBL ( _x86_Vertex3f_6_end )
-/*
-	vertex 3f generic size
-*/
-GLOBL ( _x86_Vertex3f )
-	push	%edi
-	push	%esi
-	movl	$0, %esi
-	movl	(0), %edi
-	movl	12(%esp), %eax
-	movl	16(%esp), %edx
-	movl	20(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	addl	$12, %edi
-	movl	$0, %ecx
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	movl	%eax, (0)
-	pop 	%esi
-	pop 	%edi
-	je  	.3
-	ret
-.3:	jmp	*0
-
-GLOBL ( _x86_Vertex3f_end )
-
-/*
-	Vertex 3fv vertex size 6
-*/
-GLOBL ( _x86_Vertex3fv_6 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx, (%eax)
-	movl	4(%ecx), %edx
-	movl	8(%ecx), %ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(36), %edx
-	movl	%edx, 20(%eax)
-	addl	$24, %eax
-	movl	%eax, 0
-	movl	4, %eax
-	dec 	%eax
-	movl	%eax, 4
-	je	.4
-	ret
-.4:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_6_end )
-
-/*
-	Vertex 3fv vertex size 8
-*/
-GLOBL ( _x86_Vertex3fv_8 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx ,(%eax)
-	movl	4(%ecx) ,%edx
-	movl	8(%ecx) ,%ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 20(%eax)
-	movl	%ecx, 24(%eax)
-	movl	(36), %edx
-	movl	%edx, 28(%eax)
-	addl	$32, %eax
-	movl	%eax, (0)
-	movl	4, %eax
-	dec	%eax
-	movl    %eax, (4)
-	je	.5
-	ret
-.5:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_8_end )
-
-/*
-	Vertex 3fv generic vertex size
-*/
-GLOBL ( _x86_Vertex3fv )
-	movl	4(%esp), %edx
-	push	%edi
-	push	%esi
-	movl	(0x1010101), %edi
-	movl	(%edx), %eax
-	movl	4(%edx), %ecx
-	movl	8(%edx), %esi
-	movl	%eax, (%edi)
-	movl	%ecx, 4(%edi)
-	movl	%esi, 8(%edi)
-	addl	$12, %edi
-	movl	$6, %ecx
-	movl	$0x58, %esi
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	%edi, (0x1010101)
-	movl	(0x2020202), %eax
-	pop	%esi
-	pop	%edi
-	dec	%eax
-	movl	%eax, (0x2020202)
-	je	.6
-	ret
-.6:	jmp    *0
-GLOBL ( _x86_Vertex3fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL ( _x86_Attribute2f )
-	movl	$0x0, %edx
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	movl	%eax, (%edx)
-	movl	%ecx, 4(%edx)
-	ret
-GLOBL ( _x86_Attribute2f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _x86_Attribute2fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %eax      /* load v[1] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %eax, 4      	/* store v[1] to current vertex */
-	ret
-GLOBL ( _x86_Attribute2fv_end )
-
-
-/**
- * Generic handler for 3 float format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL ( _x86_Attribute3f )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	movl	12(%esp), %eax
-	movl	%ecx, 0
-	movl	%edx, 4
-	movl	%eax, 8
-	ret
-GLOBL ( _x86_Attribute3f_end )
-
-/**
- * Generic handler for 3 float vector format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL( _x86_Attribute3fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %edx      /* load v[1] */
-	movl 8(%eax), %eax      /* load v[2] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %edx, 4      	/* store v[1] to current vertex */
-	movl %eax, 8      	/* store v[2] to current vertex */
-	ret
-GLOBL ( _x86_Attribute3fv_end )
-
-
-/*
-	Color 4ubv_ub
-*/
-GLOBL ( _x86_Color4ubv_ub )
-	movl 4(%esp), %eax
-	movl $0x12345678, %edx
-	movl (%eax), %eax
-	movl %eax, (%edx)
-	ret
-GLOBL ( _x86_Color4ubv_ub_end )
-
-/*
-	Color 4ubv 4f
-*/
-GLOBL ( _x86_Color4ubv_4f )
-	push	%ebx
-	movl	$0, %edx
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	movl	8(%esp), %ebx
-	movl	(%ebx), %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4),%eax
-	movl	(%edx,%ecx,4),%ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	shr	$16, %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4), %eax
-	movl	(%edx,%ecx,4), %ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	pop	%ebx
-	ret
-GLOBL ( _x86_Color4ubv_4f_end )
-
-/*
-
-	Color4ub_ub
-*/
-GLOBL( _x86_Color4ub_ub )
-	push	%ebx
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	20(%esp), %ebx
-	mov	%al, (0)
-	mov	%dl, (0)
-	mov	%cl, (0)
-	mov	%bl, (0)
-	pop	%ebx
-	ret
-GLOBL( _x86_Color4ub_ub_end )
-
-
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$3, %eax
-	movl	(%ecx), %edx
-	shl	$3, %eax
-	movl	4(%ecx), %ecx
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-
-GLOBL( _x86_MultiTexCoord2fv_2 )
-	movl	4(%esp,1), %eax
-	movl	8(%esp,1), %ecx
-	and	$3, %eax
-	movl	0(,%eax,4), %edx
-	movl	(%ecx), %eax
-	movl	%eax, (%edx)
-	movl	4(%ecx), %eax
-	movl	%eax, 4(%edx)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp), %ecx
-	and	$3, %eax
-	shl	$3, %eax
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _x86_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp,1), %ecx
-	and	$3,%eax
-	movl	0(,%eax,4), %eax
-	movl	%edx, (%eax)
-	movl	%ecx, 4(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_2_end )
-
-#if defined(USE_SSE_ASM)
-/**
- * This can be used as a template for either Color3fv (when the color
- * target is also a 3f) or Normal3fv.
- */
-
-GLOBL( _sse_Attribute3fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movl	8(%eax), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3fv_end )
-
-/**
- * This can be used as a template for either Color3f (when the color
- * target is also a 3f) or Normal3f.
- */
-
-GLOBL( _sse_Attribute3f )
-	movlps	4(%esp), %xmm0
-	movl	12(%esp), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2f )
-	movlps	4(%esp), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2f_end )
-
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$3, %eax
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-GLOBL( _sse_MultiTexCoord2fv_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$3, %eax
-	movl	0(,%eax,4), %edx
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, (%edx)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	and	$3, %eax
-	movlps	8(%esp), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _sse_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movlps	8(%esp), %xmm0
-	and	$3,%eax
-	movl	0(,%eax,4), %eax
-	movlps	%xmm0, (%eax)
-	ret
-GLOBL( _sse_MultiTexCoord2f_2_end )
-#endif
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/drivers/dri/s3v/s3v_context.c b/src/mesa/drivers/dri/s3v/s3v_context.c
index 7b0aa0daee..2d2f704ad7 100644
--- a/src/mesa/drivers/dri/s3v/s3v_context.c
+++ b/src/mesa/drivers/dri/s3v/s3v_context.c
@@ -6,7 +6,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -139,7 +139,7 @@ GLboolean s3vCreateContext(const __GLcontextModes *glVisual,
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext( ctx );
-	_ac_CreateContext( ctx );
+	_vbo_CreateContext( ctx );
 	_tnl_CreateContext( ctx );
 	_swsetup_CreateContext( ctx );
 
diff --git a/src/mesa/drivers/dri/s3v/s3v_state.c b/src/mesa/drivers/dri/s3v/s3v_state.c
index fcea901c1c..b86b618c11 100644
--- a/src/mesa/drivers/dri/s3v/s3v_state.c
+++ b/src/mesa/drivers/dri/s3v/s3v_state.c
@@ -10,7 +10,7 @@
 #include "colormac.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 
 /* #define DEBUG(str) printf str */
@@ -825,7 +825,7 @@ static void s3vDDUpdateState( GLcontext *ctx, GLuint new_state )
 {
 	_swrast_InvalidateState( ctx, new_state );
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 	S3V_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/s3v/s3v_xmesa.c b/src/mesa/drivers/dri/s3v/s3v_xmesa.c
index 533424cdea..c451f7452b 100644
--- a/src/mesa/drivers/dri/s3v/s3v_xmesa.c
+++ b/src/mesa/drivers/dri/s3v/s3v_xmesa.c
@@ -13,7 +13,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 /* #define DEBUG(str) printf str */
 
@@ -38,7 +38,7 @@ s3vDestroyContext(__DRIcontextPrivate *driContextPriv)
     if (vmesa) {
       _swsetup_DestroyContext( vmesa->glCtx );
       _tnl_DestroyContext( vmesa->glCtx );
-      _ac_DestroyContext( vmesa->glCtx );
+      _vbo_DestroyContext( vmesa->glCtx );
       _swrast_DestroyContext( vmesa->glCtx );
 
       s3vFreeVB( vmesa->glCtx );
diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c
index f51e321a4a..ad79b9235c 100644
--- a/src/mesa/drivers/dri/savage/savage_xmesa.c
+++ b/src/mesa/drivers/dri/savage/savage_xmesa.c
@@ -40,7 +40,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -509,7 +509,7 @@ savageCreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    
    _swsetup_CreateContext( ctx );
@@ -599,7 +599,7 @@ savageDestroyContext(__DRIcontextPrivate *driContextPriv)
 
       _swsetup_DestroyContext(imesa->glCtx );
       _tnl_DestroyContext( imesa->glCtx );
-      _ac_DestroyContext( imesa->glCtx );
+      _vbo_DestroyContext( imesa->glCtx );
       _swrast_DestroyContext( imesa->glCtx );
 
       /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/savage/savagestate.c b/src/mesa/drivers/dri/savage/savagestate.c
index 2f20dc23b6..741a9dda4f 100644
--- a/src/mesa/drivers/dri/savage/savagestate.c
+++ b/src/mesa/drivers/dri/savage/savagestate.c
@@ -41,7 +41,7 @@
 #include "savage_bci.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1673,7 +1673,7 @@ static void savageDDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    SAVAGE_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/sis/sis6326_state.c b/src/mesa/drivers/dri/sis/sis6326_state.c
index 6bc2c6de3b..08402fb3e2 100644
--- a/src/mesa/drivers/dri/sis/sis6326_state.c
+++ b/src/mesa/drivers/dri/sis/sis6326_state.c
@@ -37,7 +37,7 @@
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -645,7 +645,7 @@ sis6326DDInvalidateState( GLcontext *ctx, GLuint new_state )
 
 	_swrast_InvalidateState( ctx, new_state );
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 	smesa->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/sis/sis_context.c b/src/mesa/drivers/dri/sis/sis_context.c
index a300a080ec..89b81da347 100644
--- a/src/mesa/drivers/dri/sis/sis_context.c
+++ b/src/mesa/drivers/dri/sis/sis_context.c
@@ -53,7 +53,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -306,7 +306,7 @@ sisCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -353,7 +353,7 @@ sisDestroyContext ( __DRIcontextPrivate *driContextPriv )
    if ( smesa != NULL ) {
       _swsetup_DestroyContext( smesa->glCtx );
       _tnl_DestroyContext( smesa->glCtx );
-      _ac_DestroyContext( smesa->glCtx );
+      _vbo_DestroyContext( smesa->glCtx );
       _swrast_DestroyContext( smesa->glCtx );
 
       if (smesa->using_agp)
diff --git a/src/mesa/drivers/dri/sis/sis_state.c b/src/mesa/drivers/dri/sis/sis_state.c
index 2a10a5fbf6..33a2f089b8 100644
--- a/src/mesa/drivers/dri/sis/sis_state.c
+++ b/src/mesa/drivers/dri/sis/sis_state.c
@@ -42,7 +42,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -707,7 +707,7 @@ sisDDInvalidateState( GLcontext *ctx, GLuint new_state )
 
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    smesa->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_context.c b/src/mesa/drivers/dri/tdfx/tdfx_context.c
index bd9dade89b..a9163f49a8 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_context.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_context.c
@@ -49,7 +49,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -328,7 +328,7 @@ GLboolean tdfxCreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -609,7 +609,7 @@ tdfxDestroyContext( __DRIcontextPrivate *driContextPriv )
 
       _swsetup_DestroyContext( fxMesa->glCtx );
       _tnl_DestroyContext( fxMesa->glCtx );
-      _ac_DestroyContext( fxMesa->glCtx );
+      _vbo_DestroyContext( fxMesa->glCtx );
       _swrast_DestroyContext( fxMesa->glCtx );
 
       tdfxFreeVB( fxMesa->glCtx );
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.c b/src/mesa/drivers/dri/tdfx/tdfx_state.c
index 59e6549e5e..42cb5dfaa3 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_state.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_state.c
@@ -46,7 +46,7 @@
 #include "teximage.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -1234,7 +1234,7 @@ static void tdfxDDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    TDFX_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/trident/trident_context.c b/src/mesa/drivers/dri/trident/trident_context.c
index d380658647..dbbd1ac0c6 100644
--- a/src/mesa/drivers/dri/trident/trident_context.c
+++ b/src/mesa/drivers/dri/trident/trident_context.c
@@ -30,7 +30,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -143,7 +143,7 @@ tridentCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -183,7 +183,7 @@ tridentDestroyContext(__DRIcontextPrivate *driContextPriv)
     if (tmesa) {
       _swsetup_DestroyContext( tmesa->glCtx );
       _tnl_DestroyContext( tmesa->glCtx );
-      _ac_DestroyContext( tmesa->glCtx );
+      _vbo_DestroyContext( tmesa->glCtx );
       _swrast_DestroyContext( tmesa->glCtx );
 
       /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/trident/trident_state.c b/src/mesa/drivers/dri/trident/trident_state.c
index 037cce5f86..5303bd422e 100644
--- a/src/mesa/drivers/dri/trident/trident_state.c
+++ b/src/mesa/drivers/dri/trident/trident_state.c
@@ -26,7 +26,7 @@
  */
 #include "trident_context.h"
 #include "trident_lock.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -468,7 +468,7 @@ tridentDDUpdateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    TRIDENT_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/unichrome/via_context.c b/src/mesa/drivers/dri/unichrome/via_context.c
index 49c2465655..bc5a414df6 100644
--- a/src/mesa/drivers/dri/unichrome/via_context.c
+++ b/src/mesa/drivers/dri/unichrome/via_context.c
@@ -42,7 +42,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -583,7 +583,7 @@ viaCreateContext(const __GLcontextModes *visual,
     /* Initialize the software rasterizer and helper modules.
      */
     _swrast_CreateContext(ctx);
-    _ac_CreateContext(ctx);
+    _vbo_CreateContext(ctx);
     _tnl_CreateContext(ctx);
     _swsetup_CreateContext(ctx);
 
@@ -708,7 +708,7 @@ viaDestroyContext(__DRIcontextPrivate *driContextPriv)
 	
 	_swsetup_DestroyContext(vmesa->glCtx);
         _tnl_DestroyContext(vmesa->glCtx);
-        _ac_DestroyContext(vmesa->glCtx);
+        _vbo_DestroyContext(vmesa->glCtx);
         _swrast_DestroyContext(vmesa->glCtx);
         /* free the Mesa context */
 	_mesa_destroy_context(vmesa->glCtx);
diff --git a/src/mesa/drivers/dri/unichrome/via_state.c b/src/mesa/drivers/dri/unichrome/via_state.c
index 2c9cfca522..30b9dc289a 100644
--- a/src/mesa/drivers/dri/unichrome/via_state.c
+++ b/src/mesa/drivers/dri/unichrome/via_state.c
@@ -40,7 +40,7 @@
 #include "via_3d_reg.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1507,7 +1507,7 @@ static void viaInvalidateState(GLcontext *ctx, GLuint newState)
 
     _swrast_InvalidateState(ctx, newState);
     _swsetup_InvalidateState(ctx, newState);
-    _ac_InvalidateState(ctx, newState);
+    _vbo_InvalidateState(ctx, newState);
     _tnl_InvalidateState(ctx, newState);
 }
 
diff --git a/src/mesa/drivers/glide/fxdd.c b/src/mesa/drivers/glide/fxdd.c
index ef60602c83..f49c3889f6 100644
--- a/src/mesa/drivers/glide/fxdd.c
+++ b/src/mesa/drivers/glide/fxdd.c
@@ -55,7 +55,7 @@
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 
 
diff --git a/src/mesa/drivers/osmesa/osmesa.c b/src/mesa/drivers/osmesa/osmesa.c
index 0b027eddec..c4fc8821e5 100644
--- a/src/mesa/drivers/osmesa/osmesa.c
+++ b/src/mesa/drivers/osmesa/osmesa.c
@@ -41,7 +41,6 @@
 #include "imports.h"
 #include "mtypes.h"
 #include "renderbuffer.h"
-#include "array_cache/acache.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast/s_context.h"
@@ -51,6 +50,7 @@
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
 #include "drivers/common/driverfuncs.h"
+#include "vbo/vbo.h"
 
 
 
@@ -110,8 +110,8 @@ osmesa_update_state( GLcontext *ctx, GLuint new_state )
    /* easy - just propogate */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
 }
 
 
@@ -1261,7 +1261,7 @@ OSMesaCreateContextExt( GLenum format, GLint depthBits, GLint stencilBits,
          TNLcontext *tnl;
 
 	 if (!_swrast_CreateContext( ctx ) ||
-             !_ac_CreateContext( ctx ) ||
+             !_vbo_CreateContext( ctx ) ||
              !_tnl_CreateContext( ctx ) ||
              !_swsetup_CreateContext( ctx )) {
             _mesa_destroy_visual(osmesa->gl_visual);
@@ -1299,7 +1299,7 @@ OSMesaDestroyContext( OSMesaContext osmesa )
    if (osmesa) {
       _swsetup_DestroyContext( &osmesa->mesa );
       _tnl_DestroyContext( &osmesa->mesa );
-      _ac_DestroyContext( &osmesa->mesa );
+      _vbo_DestroyContext( &osmesa->mesa );
       _swrast_DestroyContext( &osmesa->mesa );
 
       _mesa_destroy_visual( osmesa->gl_visual );
diff --git a/src/mesa/drivers/svga/svgamesa15.c b/src/mesa/drivers/svga/svgamesa15.c
index a4c0901e38..ae5104d0c0 100644
--- a/src/mesa/drivers/svga/svgamesa15.c
+++ b/src/mesa/drivers/svga/svgamesa15.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa15.c,v 1.12 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa15.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa16.c b/src/mesa/drivers/svga/svgamesa16.c
index 999c6331c9..a59937bfb4 100644
--- a/src/mesa/drivers/svga/svgamesa16.c
+++ b/src/mesa/drivers/svga/svgamesa16.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa16.c,v 1.12 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa16.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa24.c b/src/mesa/drivers/svga/svgamesa24.c
index 93c84167eb..dd15bf38db 100644
--- a/src/mesa/drivers/svga/svgamesa24.c
+++ b/src/mesa/drivers/svga/svgamesa24.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa24.c,v 1.13 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa24.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa32.c b/src/mesa/drivers/svga/svgamesa32.c
index 5a3a6776ce..4da18795d8 100644
--- a/src/mesa/drivers/svga/svgamesa32.c
+++ b/src/mesa/drivers/svga/svgamesa32.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa32.c,v 1.13 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa32.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa8.c b/src/mesa/drivers/svga/svgamesa8.c
index e3f5eaebf6..4264fcd959 100644
--- a/src/mesa/drivers/svga/svgamesa8.c
+++ b/src/mesa/drivers/svga/svgamesa8.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa8.c,v 1.10 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa8.c,v 1.9.10.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c
index 510fbd5849..b0ef422de2 100644
--- a/src/mesa/drivers/x11/xm_api.c
+++ b/src/mesa/drivers/x11/xm_api.c
@@ -74,7 +74,7 @@
 #include "renderbuffer.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
@@ -1562,7 +1562,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
    /* Initialize the software rasterizer and helper modules.
     */
    if (!_swrast_CreateContext( mesaCtx ) ||
-       !_ac_CreateContext( mesaCtx ) ||
+       !_vbo_CreateContext( mesaCtx ) ||
        !_tnl_CreateContext( mesaCtx ) ||
        !_swsetup_CreateContext( mesaCtx )) {
       _mesa_free_context_data(&c->mesa);
@@ -1594,7 +1594,7 @@ void XMesaDestroyContext( XMesaContext c )
    _swsetup_DestroyContext( mesaCtx );
    _swrast_DestroyContext( mesaCtx );
    _tnl_DestroyContext( mesaCtx );
-   _ac_DestroyContext( mesaCtx );
+   _vbo_DestroyContext( mesaCtx );
    _mesa_free_context_data( mesaCtx );
    _mesa_free( c );
 }
diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c
index d272630d18..dbac3b8cdb 100644
--- a/src/mesa/drivers/x11/xm_dd.c
+++ b/src/mesa/drivers/x11/xm_dd.c
@@ -47,7 +47,6 @@
 #include "texstore.h"
 #include "texformat.h"
 #include "xmesaP.h"
-#include "array_cache/acache.h"
 #include "swrast/swrast.h"
 #include "swrast/s_context.h"
 #include "swrast_setup/swrast_setup.h"
@@ -829,8 +828,8 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state )
     * modules.  The X11 driver has no internal GL-dependent state.
     */
    _swrast_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
 
    if (ctx->DrawBuffer->Name != 0)
diff --git a/src/mesa/main/api_noop.c b/src/mesa/main/api_noop.c
index 33d44e4329..0c1a35361f 100644
--- a/src/mesa/main/api_noop.c
+++ b/src/mesa/main/api_noop.c
@@ -45,7 +45,7 @@
 static void GLAPIENTRY _mesa_noop_EdgeFlag( GLboolean b )
 {
    GET_CURRENT_CONTEXT(ctx);
-   ctx->Current.EdgeFlag = b;
+   ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] = (GLfloat)b;
 }
 
 static void GLAPIENTRY _mesa_noop_Indexf( GLfloat f )
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index 852b9aaee9..d601ee461e 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -114,40 +114,34 @@ _mesa_initialize_array_object( GLcontext *ctx,
    obj->Vertex.StrideB = 0;
    obj->Vertex.Ptr = NULL;
    obj->Vertex.Enabled = GL_FALSE;
-   obj->Vertex.Flags = CA_CLIENT_DATA;
    obj->Normal.Type = GL_FLOAT;
    obj->Normal.Stride = 0;
    obj->Normal.StrideB = 0;
    obj->Normal.Ptr = NULL;
    obj->Normal.Enabled = GL_FALSE;
-   obj->Normal.Flags = CA_CLIENT_DATA;
    obj->Color.Size = 4;
    obj->Color.Type = GL_FLOAT;
    obj->Color.Stride = 0;
    obj->Color.StrideB = 0;
    obj->Color.Ptr = NULL;
    obj->Color.Enabled = GL_FALSE;
-   obj->Color.Flags = CA_CLIENT_DATA;
    obj->SecondaryColor.Size = 4;
    obj->SecondaryColor.Type = GL_FLOAT;
    obj->SecondaryColor.Stride = 0;
    obj->SecondaryColor.StrideB = 0;
    obj->SecondaryColor.Ptr = NULL;
    obj->SecondaryColor.Enabled = GL_FALSE;
-   obj->SecondaryColor.Flags = CA_CLIENT_DATA;
    obj->FogCoord.Size = 1;
    obj->FogCoord.Type = GL_FLOAT;
    obj->FogCoord.Stride = 0;
    obj->FogCoord.StrideB = 0;
    obj->FogCoord.Ptr = NULL;
    obj->FogCoord.Enabled = GL_FALSE;
-   obj->FogCoord.Flags = CA_CLIENT_DATA;
    obj->Index.Type = GL_FLOAT;
    obj->Index.Stride = 0;
    obj->Index.StrideB = 0;
    obj->Index.Ptr = NULL;
    obj->Index.Enabled = GL_FALSE;
-   obj->Index.Flags = CA_CLIENT_DATA;
    for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
       obj->TexCoord[i].Size = 4;
       obj->TexCoord[i].Type = GL_FLOAT;
@@ -155,13 +149,11 @@ _mesa_initialize_array_object( GLcontext *ctx,
       obj->TexCoord[i].StrideB = 0;
       obj->TexCoord[i].Ptr = NULL;
       obj->TexCoord[i].Enabled = GL_FALSE;
-      obj->TexCoord[i].Flags = CA_CLIENT_DATA;
    }
    obj->EdgeFlag.Stride = 0;
    obj->EdgeFlag.StrideB = 0;
    obj->EdgeFlag.Ptr = NULL;
    obj->EdgeFlag.Enabled = GL_FALSE;
-   obj->EdgeFlag.Flags = CA_CLIENT_DATA;
    for (i = 0; i < VERT_ATTRIB_MAX; i++) {
       obj->VertexAttrib[i].Size = 4;
       obj->VertexAttrib[i].Type = GL_FLOAT;
@@ -170,7 +162,6 @@ _mesa_initialize_array_object( GLcontext *ctx,
       obj->VertexAttrib[i].Ptr = NULL;
       obj->VertexAttrib[i].Enabled = GL_FALSE;
       obj->VertexAttrib[i].Normalized = GL_FALSE;
-      obj->VertexAttrib[i].Flags = CA_CLIENT_DATA;
    }
 
 #if FEATURE_ARB_vertex_buffer_object
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 99f4dc9dfd..9b3759b6c8 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -976,9 +976,8 @@ _mesa_init_current( GLcontext *ctx )
    ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_NORMAL], 0.0, 0.0, 1.0, 1.0 );
    ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_COLOR0], 1.0, 1.0, 1.0, 1.0 );
    ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_COLOR1], 0.0, 0.0, 0.0, 1.0 );
-   ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_FOG], 0.0, 0.0, 0.0, 0.0 );
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX][0] = 1.0;
-   ctx->Current.EdgeFlag = GL_TRUE;
+   ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX], 1.0, 0.0, 0.0, 1.0 );
+   ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG], 1.0, 0.0, 0.0, 1.0 );
 }
 
 
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index ca48403482..49bc2933bf 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -343,8 +343,6 @@ typedef enum
    OPCODE_ATTR_3F_ARB,
    OPCODE_ATTR_4F_ARB,
    OPCODE_MATERIAL,
-   OPCODE_INDEX,
-   OPCODE_EDGEFLAG,
    OPCODE_BEGIN,
    OPCODE_END,
    OPCODE_RECTF,
@@ -5110,45 +5108,19 @@ save_EvalPoint2(GLint x, GLint y)
 static void GLAPIENTRY
 save_Indexf(GLfloat x)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   Node *n;
-   SAVE_FLUSH_VERTICES(ctx);
-   n = ALLOC_INSTRUCTION(ctx, OPCODE_INDEX, 1);
-   if (n) {
-      n[1].f = x;
-   }
-
-   ctx->ListState.ActiveIndex = 1;
-   ctx->ListState.CurrentIndex = x;
-
-   if (ctx->ExecuteFlag) {
-      CALL_Indexf(ctx->Exec, (x));
-   }
+   save_Attr1fNV(VERT_ATTRIB_COLOR_INDEX, x);
 }
 
 static void GLAPIENTRY
 save_Indexfv(const GLfloat * v)
 {
-   save_Indexf(v[0]);
+   save_Attr1fNV(VERT_ATTRIB_COLOR_INDEX, v[0]);
 }
 
 static void GLAPIENTRY
 save_EdgeFlag(GLboolean x)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   Node *n;
-   SAVE_FLUSH_VERTICES(ctx);
-   n = ALLOC_INSTRUCTION(ctx, OPCODE_EDGEFLAG, 1);
-   if (n) {
-      n[1].b = x;
-   }
-
-   ctx->ListState.ActiveEdgeFlag = 1;
-   ctx->ListState.CurrentEdgeFlag = x;
-
-   if (ctx->ExecuteFlag) {
-      CALL_EdgeFlag(ctx->Exec, (x));
-   }
+   save_Attr1fNV(VERT_ATTRIB_EDGEFLAG, x ? 1.0 : 0.0);
 }
 
 static void GLAPIENTRY
@@ -6602,12 +6574,6 @@ execute_list(GLcontext *ctx, GLuint list)
                CALL_Materialfv(ctx->Exec, (n[1].e, n[2].e, f));
             }
             break;
-         case OPCODE_INDEX:
-            CALL_Indexf(ctx->Exec, (n[1].f));
-            break;
-         case OPCODE_EDGEFLAG:
-            CALL_EdgeFlag(ctx->Exec, (n[1].b));
-            break;
          case OPCODE_BEGIN:
             CALL_Begin(ctx->Exec, (n[1].e));
             break;
@@ -6793,9 +6759,6 @@ _mesa_NewList(GLuint list, GLenum mode)
    for (i = 0; i < MAT_ATTRIB_MAX; i++)
       ctx->ListState.ActiveMaterialSize[i] = 0;
 
-   ctx->ListState.ActiveIndex = 0;
-   ctx->ListState.ActiveEdgeFlag = 0;
-
    ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
    ctx->Driver.NewList(ctx, list, mode);
 
@@ -8421,12 +8384,6 @@ print_list(GLcontext *ctx, GLuint list)
             _mesa_printf("MATERIAL %x %x: %f %f %f %f\n",
                          n[1].i, n[2].i, n[3].f, n[4].f, n[5].f, n[6].f);
             break;
-         case OPCODE_INDEX:
-            _mesa_printf("INDEX: %f\n", n[1].f);
-            break;
-         case OPCODE_EDGEFLAG:
-            _mesa_printf("EDGEFLAG: %d\n", n[1].i);
-            break;
          case OPCODE_BEGIN:
             _mesa_printf("BEGIN %x\n", n[1].i);
             break;
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index 076d8731f8..0d54c29949 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -56,7 +56,7 @@ static void
 client_state(GLcontext *ctx, GLenum cap, GLboolean state)
 {
    GLuint flag;
-   GLuint *var;
+   GLboolean *var;
 
    switch (cap) {
       case GL_VERTEX_ARRAY:
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index 7845ea018e..b61cb821f7 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -136,6 +136,7 @@ static const struct {
    { OFF, "GL_ATI_texture_env_combine3",       F(ATI_texture_env_combine3)},
    { OFF, "GL_ATI_texture_mirror_once",        F(ATI_texture_mirror_once)},
    { OFF, "GL_ATI_fragment_shader",            F(ATI_fragment_shader)},
+   { OFF, "GL_ATI_separate_stencil",           F(ATI_separate_stencil)},
    { OFF, "GL_IBM_multimode_draw_arrays",      F(IBM_multimode_draw_arrays) },
    { ON,  "GL_IBM_rasterpos_clip",             F(IBM_rasterpos_clip) },
    { OFF, "GL_IBM_texture_mirrored_repeat",    F(ARB_texture_mirrored_repeat)},
@@ -223,6 +224,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #endif
    ctx->Extensions.ATI_texture_env_combine3 = GL_TRUE;
    ctx->Extensions.ATI_texture_mirror_once = GL_TRUE;
+   ctx->Extensions.ATI_separate_stencil = GL_TRUE;
    ctx->Extensions.EXT_blend_color = GL_TRUE;
    ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
    ctx->Extensions.EXT_blend_func_separate = GL_TRUE;
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index 858c822659..9b2a42f7c1 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -323,7 +323,7 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
       case GL_EDGE_FLAG:
          {
          FLUSH_CURRENT(ctx, 0);
-         params[0] = ctx->Current.EdgeFlag;
+         params[0] = (ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0);
          }
          break;
       case GL_FEEDBACK_BUFFER_SIZE:
@@ -2150,7 +2150,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
       case GL_EDGE_FLAG:
          {
          FLUSH_CURRENT(ctx, 0);
-         params[0] = BOOLEAN_TO_FLOAT(ctx->Current.EdgeFlag);
+         params[0] = BOOLEAN_TO_FLOAT((ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0));
          }
          break;
       case GL_FEEDBACK_BUFFER_SIZE:
@@ -3977,7 +3977,7 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
       case GL_EDGE_FLAG:
          {
          FLUSH_CURRENT(ctx, 0);
-         params[0] = BOOLEAN_TO_INT(ctx->Current.EdgeFlag);
+         params[0] = BOOLEAN_TO_INT((ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0));
          }
          break;
       case GL_FEEDBACK_BUFFER_SIZE:
diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py
index 0f2ca00b83..c18216d4a8 100644
--- a/src/mesa/main/get_gen.py
+++ b/src/mesa/main/get_gen.py
@@ -190,7 +190,7 @@ StateVars = [
 	( "GL_DOUBLEBUFFER", GLboolean,
 	  ["ctx->DrawBuffer->Visual.doubleBufferMode"], "", None ),
 	( "GL_DRAW_BUFFER", GLenum, ["ctx->DrawBuffer->ColorDrawBuffer[0]"], "", None ),
-	( "GL_EDGE_FLAG", GLboolean, ["ctx->Current.EdgeFlag"],
+	( "GL_EDGE_FLAG", GLboolean, ["(ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0)"],
 	  "FLUSH_CURRENT(ctx, 0);", None ),
 	( "GL_FEEDBACK_BUFFER_SIZE", GLint, ["ctx->Feedback.BufferSize"], "", None ),
 	( "GL_FEEDBACK_BUFFER_TYPE", GLenum, ["ctx->Feedback.Type"], "", None ),
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 0ed73beb3c..2ad37d3a8d 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -143,7 +143,7 @@ enum
    VERT_ATTRIB_COLOR1 = 4,
    VERT_ATTRIB_FOG = 5,
    VERT_ATTRIB_COLOR_INDEX = 6,
-   VERT_ATTRIB_SEVEN = 7,
+   VERT_ATTRIB_EDGEFLAG = 7,
    VERT_ATTRIB_TEX0 = 8,
    VERT_ATTRIB_TEX1 = 9,
    VERT_ATTRIB_TEX2 = 10,
@@ -183,7 +183,7 @@ enum
 #define VERT_BIT_COLOR1      (1 << VERT_ATTRIB_COLOR1)
 #define VERT_BIT_FOG         (1 << VERT_ATTRIB_FOG)
 #define VERT_BIT_COLOR_INDEX (1 << VERT_ATTRIB_COLOR_INDEX)
-#define VERT_BIT_SEVEN       (1 << VERT_ATTRIB_SEVEN)
+#define VERT_BIT_EDGEFLAG    (1 << VERT_ATTRIB_EDGEFLAG)
 #define VERT_BIT_TEX0        (1 << VERT_ATTRIB_TEX0)
 #define VERT_BIT_TEX1        (1 << VERT_ATTRIB_TEX1)
 #define VERT_BIT_TEX2        (1 << VERT_ATTRIB_TEX2)
@@ -616,11 +616,11 @@ struct gl_current_attrib
    /**
     * \name Current vertex attributes.
     * \note Values are valid only after FLUSH_VERTICES has been called.
+    * \note Index and Edgeflag current values are stored as floats in the 
+    * SIX and SEVEN attribute slots.
     */
    /*@{*/
    GLfloat Attrib[VERT_ATTRIB_MAX][4];	/**< Position, color, texcoords, etc */
-   GLfloat Index;			/**< Current color index */
-   GLboolean EdgeFlag;			/**< Current edge flag */
    /*@}*/
 
    /**
@@ -1640,8 +1640,6 @@ struct gl_pixelstore_attrib
 };
 
 
-#define CA_CLIENT_DATA     0x1	/**< Data not allocated by mesa */
-
 
 /**
  * Client vertex array attributes
@@ -1653,14 +1651,12 @@ struct gl_client_array
    GLsizei Stride;		/**< user-specified stride */
    GLsizei StrideB;		/**< actual stride in bytes */
    const GLubyte *Ptr;          /**< Points to array data */
-   GLbitfield Enabled;		/**< one of the _NEW_ARRAY_ bits */
+   GLboolean Enabled;		/**< Enabled flag is a boolean */
    GLboolean Normalized;        /**< GL_ARB_vertex_program */
 
    /**< GL_ARB_vertex_buffer_object */
    struct gl_buffer_object *BufferObj;
    GLuint _MaxElement;
-
-   GLbitfield Flags;
 };
 
 
@@ -1681,8 +1677,8 @@ struct gl_array_object
    struct gl_client_array SecondaryColor;
    struct gl_client_array FogCoord;
    struct gl_client_array Index;
-   struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS];
    struct gl_client_array EdgeFlag;
+   struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS];
    /*@}*/
 
    /** Generic arrays for vertex programs/shaders */
@@ -2478,6 +2474,7 @@ struct gl_extensions
    GLboolean ATI_texture_mirror_once;
    GLboolean ATI_texture_env_combine3;
    GLboolean ATI_fragment_shader;
+   GLboolean ATI_separate_stencil;
    GLboolean IBM_rasterpos_clip;
    GLboolean IBM_multimode_draw_arrays;
    GLboolean MESA_pack_invert;
@@ -2611,7 +2608,7 @@ struct matrix_stack
 #define _NEW_ARRAY_COLOR1           VERT_BIT_COLOR1
 #define _NEW_ARRAY_FOGCOORD         VERT_BIT_FOG
 #define _NEW_ARRAY_INDEX            VERT_BIT_COLOR_INDEX
-#define _NEW_ARRAY_EDGEFLAG         VERT_BIT_SEVEN
+#define _NEW_ARRAY_EDGEFLAG         VERT_BIT_EDGEFLAG
 #define _NEW_ARRAY_TEXCOORD_0       VERT_BIT_TEX0
 #define _NEW_ARRAY_TEXCOORD_1       VERT_BIT_TEX1
 #define _NEW_ARRAY_TEXCOORD_2       VERT_BIT_TEX2
@@ -2620,7 +2617,7 @@ struct matrix_stack
 #define _NEW_ARRAY_TEXCOORD_5       VERT_BIT_TEX5
 #define _NEW_ARRAY_TEXCOORD_6       VERT_BIT_TEX6
 #define _NEW_ARRAY_TEXCOORD_7       VERT_BIT_TEX7
-#define _NEW_ARRAY_ATTRIB_0         0x10000  /* start at bit 16 */
+#define _NEW_ARRAY_ATTRIB_0         VERT_BIT_GENERIC0  /* start at bit 16 */
 #define _NEW_ARRAY_ALL              0xffffffff
 
 
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index 4184aeb9d9..55fb4552d2 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -903,8 +903,8 @@ update_arrays( GLcontext *ctx )
 
    /* 7 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_SEVEN].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_SEVEN]._MaxElement);
+       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
+      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]._MaxElement);
    }
 
    /* 8..15 */
diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c
index ad71a81f6e..e61eb0030c 100644
--- a/src/mesa/main/stencil.c
+++ b/src/mesa/main/stencil.c
@@ -119,23 +119,7 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
 
    ref = CLAMP( ref, 0, stencilMax );
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
-      if (ctx->Stencil.Function[face] == func &&
-          ctx->Stencil.ValueMask[face] == mask &&
-          ctx->Stencil.Ref[face] == ref)
-         return;
-      FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.Function[face] = func;
-      ctx->Stencil.Ref[face] = ref;
-      ctx->Stencil.ValueMask[face] = mask;
-      if (ctx->Driver.StencilFuncSeparate) {
-         ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT,
-                                         func, ref, mask);
-      }
-   }
-   else {
+   if (ctx->Extensions.ATI_separate_stencil) {
       /* set both front and back state */
       if (ctx->Stencil.Function[0] == func &&
           ctx->Stencil.Function[1] == func &&
@@ -153,6 +137,22 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
                                          func, ref, mask);
       }
    }
+   else {
+      /* only set active face state */
+      const GLint face = ctx->Stencil.ActiveFace;
+      if (ctx->Stencil.Function[face] == func &&
+          ctx->Stencil.ValueMask[face] == mask &&
+          ctx->Stencil.Ref[face] == ref)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_STENCIL);
+      ctx->Stencil.Function[face] = func;
+      ctx->Stencil.Ref[face] = ref;
+      ctx->Stencil.ValueMask[face] = mask;
+      if (ctx->Driver.StencilFuncSeparate) {
+         ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT,
+                                         func, ref, mask);
+      }
+   }
 }
 
 
@@ -173,26 +173,26 @@ _mesa_StencilMask( GLuint mask )
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
-      if (ctx->Stencil.WriteMask[face] == mask)
+   if (ctx->Extensions.ATI_separate_stencil) {
+      /* set both front and back state */
+      if (ctx->Stencil.WriteMask[0] == mask &&
+          ctx->Stencil.WriteMask[1] == mask)
          return;
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.WriteMask[face] = mask;
+      ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask;
       if (ctx->Driver.StencilMaskSeparate) {
-         ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask);
+         ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask);
       }
    }
    else {
-      /* set both front and back state */
-      if (ctx->Stencil.WriteMask[0] == mask &&
-          ctx->Stencil.WriteMask[1] == mask)
+      /* only set active face state */
+      const GLint face = ctx->Stencil.ActiveFace;
+      if (ctx->Stencil.WriteMask[face] == mask)
          return;
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask;
+      ctx->Stencil.WriteMask[face] = mask;
       if (ctx->Driver.StencilMaskSeparate) {
-         ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask);
+         ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask);
       }
    }
 }
@@ -273,23 +273,7 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
          return;
    }
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
-      if (ctx->Stencil.ZFailFunc[face] == zfail &&
-          ctx->Stencil.ZPassFunc[face] == zpass &&
-          ctx->Stencil.FailFunc[face] == fail)
-         return;
-      FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.ZFailFunc[face] = zfail;
-      ctx->Stencil.ZPassFunc[face] = zpass;
-      ctx->Stencil.FailFunc[face] = fail;
-      if (ctx->Driver.StencilOpSeparate) {
-         ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT,
-                                       fail, zfail, zpass);
-      }
-   }
-   else {
+   if (ctx->Extensions.ATI_separate_stencil) {
       /* set both front and back state */
       if (ctx->Stencil.ZFailFunc[0] == zfail &&
           ctx->Stencil.ZFailFunc[1] == zfail &&
@@ -307,6 +291,22 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
                                        fail, zfail, zpass);
       }
    }
+   else {
+      /* only set active face state */
+      const GLint face = ctx->Stencil.ActiveFace;
+      if (ctx->Stencil.ZFailFunc[face] == zfail &&
+          ctx->Stencil.ZPassFunc[face] == zpass &&
+          ctx->Stencil.FailFunc[face] == fail)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_STENCIL);
+      ctx->Stencil.ZFailFunc[face] = zfail;
+      ctx->Stencil.ZPassFunc[face] = zpass;
+      ctx->Stencil.FailFunc[face] = fail;
+      if (ctx->Driver.StencilOpSeparate) {
+         ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT,
+                                       fail, zfail, zpass);
+      }
+   }
 }
 
 
diff --git a/src/mesa/sources b/src/mesa/sources
index 09692c6f15..6d6d22861c 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources
@@ -78,10 +78,6 @@ MATH_SOURCES = \
 	math/m_vector.c \
 	math/m_xform.c
 
-ARRAY_CACHE_SOURCES = \
-	array_cache/ac_context.c \
-	array_cache/ac_import.c
-
 SWRAST_SOURCES = \
 	swrast/s_aaline.c \
 	swrast/s_aatriangle.c \
@@ -119,13 +115,9 @@ SWRAST_SETUP_SOURCES = \
 	swrast_setup/ss_triangle.c 
 
 TNL_SOURCES = \
-	tnl/t_array_api.c \
-	tnl/t_array_import.c \
 	tnl/t_context.c \
 	tnl/t_pipeline.c \
-	tnl/t_save_api.c \
-	tnl/t_save_loopback.c \
-	tnl/t_save_playback.c \
+	tnl/t_draw.c \
 	tnl/t_vb_arbprogram.c \
 	tnl/t_vb_arbprogram_sse.c \
 	tnl/t_vb_arbshader.c\
@@ -142,13 +134,23 @@ TNL_SOURCES = \
 	tnl/t_vp_build.c \
 	tnl/t_vertex.c \
 	tnl/t_vertex_sse.c \
-	tnl/t_vertex_generic.c \
-	tnl/t_vtx_api.c \
-	tnl/t_vtx_generic.c \
-	tnl/t_vtx_x86.c \
-	tnl/t_vtx_eval.c \
-	tnl/t_vtx_exec.c 
-
+	tnl/t_vertex_generic.c 
+
+VBO_SOURCES = \
+	vbo/vbo_context.c \
+	vbo/vbo_exec.c \
+	vbo/vbo_exec_api.c \
+	vbo/vbo_exec_array.c \
+	vbo/vbo_exec_draw.c \
+	vbo/vbo_exec_eval.c \
+	vbo/vbo_rebase.c \
+	vbo/vbo_split.c \
+	vbo/vbo_split_copy.c \
+	vbo/vbo_split_inplace.c \
+	vbo/vbo_save.c \
+	vbo/vbo_save_api.c \
+	vbo/vbo_save_draw.c \
+	vbo/vbo_save_loopback.c 
 
 
 SHADER_SOURCES = \
@@ -242,8 +244,7 @@ X86_SOURCES =			\
 	x86/sse_xform3.S	\
 	x86/sse_xform4.S	\
 	x86/sse_normal.S	\
-	x86/read_rgba_span_x86.S \
-	tnl/t_vtx_x86_gcc.S
+	x86/read_rgba_span_x86.S
 
 X86_API =			\
 	x86/glapi_x86.S
@@ -317,7 +318,7 @@ ALL_SOURCES = \
 SOLO_SOURCES = \
 	$(MAIN_SOURCES)		\
 	$(MATH_SOURCES)		\
-	$(ARRAY_CACHE_SOURCES)	\
+	$(VBO_SOURCES)		\
 	$(TNL_SOURCES)		\
 	$(SHADER_SOURCES)	\
 	$(SWRAST_SOURCES)	\
diff --git a/src/mesa/tnl/sources b/src/mesa/tnl/sources
index e01f55dbaf..a0888be11d 100644
--- a/src/mesa/tnl/sources
+++ b/src/mesa/tnl/sources
@@ -1,12 +1,7 @@
 # List of source files in this directory used for X.org xserver build
 MESA_TNL_SOURCES = \
-t_array_api.c \
-t_array_import.c \
 t_context.c \
 t_pipeline.c \
-t_save_api.c \
-t_save_loopback.c \
-t_save_playback.c \
 t_vb_arbprogram.c \
 t_vb_arbprogram_sse.c \
 t_vb_arbshader.c \
@@ -23,24 +18,17 @@ t_vb_vertex.c \
 t_vertex.c \
 t_vertex_generic.c \
 t_vertex_sse.c \
-t_vp_build.c \
-t_vtx_api.c \
-t_vtx_eval.c \
-t_vtx_exec.c \
-t_vtx_generic.c \
-t_vtx_x86.c
+t_vp_build.c 
 
 MESA_TNL_HEADERS = \
 t_array_api.h \
 t_array_import.h \
 t_context.h \
 t_pipeline.h \
-t_save_api.h \
 t_vb_arbprogram.h \
 t_vb_cliptmp.h \
 t_vb_lighttmp.h \
 t_vb_rendertmp.h \
 t_vertex.h \
 t_vp_build.h \
-t_vtx_api.h \
 tnl.h
diff --git a/src/mesa/tnl/t_array_api.c b/src/mesa/tnl/t_array_api.c
deleted file mode 100644
index 087f6af08c..0000000000
--- a/src/mesa/tnl/t_array_api.c
+++ /dev/null
@@ -1,432 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.2
- *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/**
- * \file t_array_api.c
- * \brief Vertex array API functions (glDrawArrays, etc)
- * \author Keith Whitwell
- */
-
-#include "glheader.h"
-#include "api_validate.h"
-#include "context.h"
-#include "imports.h"
-#include "macros.h"
-#include "mtypes.h"
-#include "state.h"
-
-#include "array_cache/acache.h"
-
-#include "t_array_api.h"
-#include "t_array_import.h"
-#include "t_save_api.h"
-#include "t_context.h"
-#include "t_pipeline.h"
-#include "dispatch.h"
-
-static void fallback_drawarrays( GLcontext *ctx, GLenum mode, GLint start,
-				 GLsizei count )
-{
-   GLint i;
-
-   assert(!ctx->CompileFlag);
-   assert(ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END);
-
-   CALL_Begin(GET_DISPATCH(), (mode));
-   for (i = 0; i < count; i++) 
-       CALL_ArrayElement(GET_DISPATCH(), ( start + i ));
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-static void fallback_drawelements( GLcontext *ctx, GLenum mode, GLsizei count,
-				   const GLuint *indices)
-{
-   GLint i;
-
-   assert(!ctx->CompileFlag);
-   assert(ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END);
-
-   /* Here, indices will already reflect the buffer object if active */
-
-   CALL_Begin(GET_DISPATCH(), (mode));
-   for (i = 0 ; i < count ; i++) {
-      CALL_ArrayElement(GET_DISPATCH(), ( indices[i] ));
-   }
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-/* Note this function no longer takes a 'start' value, the range is
- * assumed to start at zero.  The old trick of subtracting 'start'
- * from each index won't work if the indices are not in writeable
- * memory.
- */
-static void _tnl_draw_range_elements( GLcontext *ctx, GLenum mode,
-				      GLuint max_index,
-				      GLsizei index_count, GLuint *indices )
-
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_prim prim;
-   FLUSH_CURRENT( ctx, 0 );
-   
-   _tnl_vb_bind_arrays( ctx, 0, max_index );
-
-   tnl->vb.Primitive = &prim;
-   tnl->vb.Primitive[0].mode = mode | PRIM_BEGIN | PRIM_END;
-   tnl->vb.Primitive[0].start = 0;
-   tnl->vb.Primitive[0].count = index_count;
-   tnl->vb.PrimitiveCount = 1;
-
-   tnl->vb.Elts = (GLuint *)indices;
-
-   tnl->Driver.RunPipeline( ctx );
-}
-
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   /* It's tempting to get rid of this threshold value because we take
-    * very different paths if 'count' is less than or greater than 'thresh'.
-    * I've found/fixed at least one bug which only occured for particular
-    * array sizes.  Also, several conformance tests use very short arrays
-    * which means the long-array path doesn't get tested.  -Brian
-    */
-   GLuint thresh = (ctx->Driver.NeedFlush & FLUSH_STORED_VERTICES) ? 30 : 10;
-   
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawArrays %d %d\n", start, count); 
-   
-   /* Check arguments, etc.
-    */
-   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-      return;
-
-   assert(!ctx->CompileFlag);
-
-   if (!ctx->Array.LockCount && (GLuint) count < thresh) {
-      /* Small primitives: attempt to share a vb (at the expense of
-       * using the immediate interface).
-      */
-      fallback_drawarrays( ctx, mode, start, count );
-   } 
-   else if (start >= (GLint) ctx->Array.LockFirst &&
-	    start + count <= (GLint)(ctx->Array.LockFirst + ctx->Array.LockCount)) {
-      
-      struct tnl_prim prim;
-
-      /* Locked primitives which can fit in a single vertex buffer:
-       */
-      FLUSH_CURRENT( ctx, 0 );
-
-      /* Locked drawarrays.  Reuse any previously transformed data.
-       */
-      _tnl_vb_bind_arrays( ctx, ctx->Array.LockFirst, 
-			   ctx->Array.LockFirst + ctx->Array.LockCount );
-
-      tnl->vb.Primitive = &prim;
-      tnl->vb.Primitive[0].mode = mode | PRIM_BEGIN | PRIM_END;
-      tnl->vb.Primitive[0].start = start;
-      tnl->vb.Primitive[0].count = count;
-      tnl->vb.PrimitiveCount = 1;
-
-      tnl->Driver.RunPipeline( ctx );
-   } 
-   else {
-      int bufsz = 256;		/* Use a small buffer for cache goodness */
-      int j, nr;
-      int minimum, modulo, skip;
-
-      /* Large primitives requiring decomposition to multiple vertex
-       * buffers:
-       */
-      switch (mode) {
-      case GL_POINTS:
-	 minimum = 0;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_LINES:
-	 minimum = 1;
-	 modulo = 2;
-	 skip = 1;
-	 break;
-      case GL_LINE_STRIP:
-	 minimum = 1;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_TRIANGLES:
-	 minimum = 2;
-	 modulo = 3;
-	 skip = 2;
-	 break;
-      case GL_TRIANGLE_STRIP:
-	 minimum = 2;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_QUADS:
-	 minimum = 3;
-	 modulo = 4;
-	 skip = 3;
-	 break;
-      case GL_QUAD_STRIP:
-	 minimum = 3;
-	 modulo = 2;
-	 skip = 0;
-	 break;
-      case GL_LINE_LOOP:
-      case GL_TRIANGLE_FAN:
-      case GL_POLYGON:
-      default:
-	 /* Primitives requiring a copied vertex (fan-like primitives)
-	  * must use the slow path if they cannot fit in a single
-	  * vertex buffer.  
-	  */
-	 if (count <= (GLint) ctx->Const.MaxArrayLockSize) {
-	    bufsz = ctx->Const.MaxArrayLockSize;
-	    minimum = 0;
-	    modulo = 1;
-	    skip = 0;
-	 }
-	 else {
-	    fallback_drawarrays( ctx, mode, start, count );
-	    return;
-	 }
-      }
-
-      FLUSH_CURRENT( ctx, 0 );
-
-      bufsz -= bufsz % modulo;
-      bufsz -= minimum;
-      count += start;
-
-      for (j = start + minimum ; j < count ; j += nr + skip ) {
-
-	 struct tnl_prim prim;
-
-	 nr = MIN2( bufsz, count - j );
-
-         /* XXX is the last parameter a count or index into the array??? */
-	 _tnl_vb_bind_arrays( ctx, j - minimum, j + nr );
-
-	 tnl->vb.Primitive = &prim;
-	 tnl->vb.Primitive[0].mode = mode;
-
-	 if (j == start + minimum)
-	    tnl->vb.Primitive[0].mode |= PRIM_BEGIN;
-
-	 if (j + nr + skip >= count)
-	    tnl->vb.Primitive[0].mode |= PRIM_END;
-
-	 tnl->vb.Primitive[0].start = 0;
-	 tnl->vb.Primitive[0].count = nr + minimum;
-	 tnl->vb.PrimitiveCount = 1;
-
-	 tnl->Driver.RunPipeline( ctx );
-      }
-   }
-}
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawRangeElements(GLenum mode,
-		       GLuint start, GLuint end,
-		       GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint *ui_indices;
-
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawRangeElements %d %d %d\n", start, end, count); 
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* use indices in the buffer object */
-      if (!ctx->Array.ElementArrayBufferObj->Data) {
-         _mesa_warning(ctx,
-                       "DrawRangeElements with empty vertex elements buffer!");
-         return;
-      }
-      /* actual address is the sum of pointers */
-      indices = (const GLvoid *)
-         ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data,
-                      (const GLubyte *) indices);
-   }
-
-   /* Check arguments, etc.
-    */
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count,
-                                          type, indices ))
-      return;
-
-   ui_indices = (GLuint *)_ac_import_elements( ctx, GL_UNSIGNED_INT,
-					       count, type, indices );
-
-#ifdef DEBUG
-   /* check that array indices really fall inside [start, end] range */
-   {
-      GLint i;
-      for (i = 0; i < count; i++) {
-         if (ui_indices[i] < start || ui_indices[i] > end) {
-            _mesa_warning(ctx, "Invalid array index in "
-                          "glDrawRangeElements(index=%u)", ui_indices[i]);
-         }
-      }
-   }
-#endif
-
-   assert(!ctx->CompileFlag);
-
-   if (ctx->Array.LockCount) {
-      /* Are the arrays already locked?  If so we currently have to look
-       * at the whole locked range.
-       */
-
-      if (start == 0 && ctx->Array.LockFirst == 0 && 
-	  end < (ctx->Array.LockFirst + ctx->Array.LockCount))
-	 _tnl_draw_range_elements( ctx, mode,
-				   ctx->Array.LockCount,
-				   count, ui_indices );
-      else {
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-      }
-   }
-   else if (start == 0 && end < ctx->Const.MaxArrayLockSize) {
-      /* The arrays aren't locked but we can still fit them inside a
-       * single vertexbuffer.
-       */
-      _tnl_draw_range_elements( ctx, mode, end + 1, count, ui_indices );
-   }
-   else {
-      /* Range is too big to optimize:
-       */
-      fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-}
-
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawElements(GLenum mode, GLsizei count, GLenum type,
-		  const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint *ui_indices;
-
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawElements %d\n", count); 
-
-   /* Check arguments, etc. */
-   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-      return;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* actual address is the sum of pointers */
-      indices = (const GLvoid *)
-         ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data,
-                      (const GLubyte *) indices);
-   }
-
-   ui_indices = (GLuint *)_ac_import_elements( ctx, GL_UNSIGNED_INT,
-					       count, type, indices );
-
-   assert(!ctx->CompileFlag);
-
-   if (ctx->Array.LockCount) {
-      if (ctx->Array.LockFirst == 0)
-	 _tnl_draw_range_elements( ctx, mode,
-				   ctx->Array.LockCount,
-				   count, ui_indices );
-      else
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-   else {
-      /* Scan the index list and see if we can use the locked path anyway.
-       */
-      GLuint max_elt = 0;
-      GLint i;
-
-      for (i = 0 ; i < count ; i++)
-	 if (ui_indices[i] > max_elt)
-            max_elt = ui_indices[i];
-
-      if (max_elt < ctx->Const.MaxArrayLockSize && /* can we use it? */
-	  max_elt < (GLuint) count) 	           /* do we want to use it? */
-	 _tnl_draw_range_elements( ctx, mode, max_elt+1, count, ui_indices );
-      else
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-}
-
-
-/**
- * Initialize context's vertex array fields.  Called during T 'n L context
- * creation.
- */
-void _tnl_array_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_vertex_arrays *tmp = &tnl->array_inputs;
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-   GLuint i;
-
-   vfmt->DrawArrays = _tnl_DrawArrays;
-   vfmt->DrawElements = _tnl_DrawElements;
-   vfmt->DrawRangeElements = _tnl_DrawRangeElements;
-
-   /* Setup vector pointers that will be used to bind arrays to VB's.
-    */
-   _mesa_vector4f_init( &tmp->Obj, 0, NULL);
-   _mesa_vector4f_init( &tmp->Normal, 0, NULL);   
-   _mesa_vector4f_init( &tmp->FogCoord, 0, NULL);
-   _mesa_vector4f_init( &tmp->Index, 0, NULL);
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++)
-      _mesa_vector4f_init( &tmp->TexCoord[i], 0, NULL);
-}
-
-
-/**
- * Destroy the context's vertex array stuff.
- * Called during T 'n L context destruction.
- */
-void _tnl_array_destroy( GLcontext *ctx )
-{
-   (void) ctx;
-}
diff --git a/src/mesa/tnl/t_array_api.h b/src/mesa/tnl/t_array_api.h
deleted file mode 100644
index 61d1f696b0..0000000000
--- a/src/mesa/tnl/t_array_api.h
+++ /dev/null
@@ -1,46 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#ifndef _T_VARRAY_H
-#define _T_VARRAY_H
-
-#include "mtypes.h"
-#include "t_context.h"
-
-
-extern void GLAPIENTRY _tnl_DrawArrays(GLenum mode, GLint first, GLsizei count);
-
-extern void GLAPIENTRY _tnl_DrawElements(GLenum mode, GLsizei count, GLenum type,
-			      const GLvoid *indices);
-
-extern void GLAPIENTRY _tnl_DrawRangeElements(GLenum mode, GLuint start,
-				   GLuint end, GLsizei count, GLenum type,
-				   const GLvoid *indices);
-
-
-extern void _tnl_array_init( GLcontext *ctx );
-extern void _tnl_array_destroy( GLcontext *ctx );
-
-#endif
diff --git a/src/mesa/tnl/t_array_import.c b/src/mesa/tnl/t_array_import.c
deleted file mode 100644
index 13c5689ceb..0000000000
--- a/src/mesa/tnl/t_array_import.c
+++ /dev/null
@@ -1,376 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.1
- *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "imports.h"
-#include "state.h"
-#include "mtypes.h"
-
-#include "array_cache/acache.h"
-
-#include "t_array_import.h"
-#include "t_context.h"
-
-
-/**
- * XXX writable and stride are always false in these functions...
- */
-static void _tnl_import_vertex( GLcontext *ctx,
-				GLboolean writable,
-				GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_vertex(ctx,
-			   GL_FLOAT,
-			   stride ? 4*sizeof(GLfloat) : 0,
-			   0,
-			   writable,
-			   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Obj.data = (GLfloat (*)[4]) data;
-   inputs->Obj.start = (GLfloat *) data;
-   inputs->Obj.stride = tmp->StrideB;
-   inputs->Obj.size = tmp->Size;
-}
-
-static void _tnl_import_normal( GLcontext *ctx,
-				GLboolean writable,
-				GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_normal(ctx, GL_FLOAT,
-			   stride ? 3*sizeof(GLfloat) : 0, writable,
-			   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Normal.data = (GLfloat (*)[4]) data;
-   inputs->Normal.start = (GLfloat *) data;
-   inputs->Normal.stride = tmp->StrideB;
-   inputs->Normal.size = 3;
-}
-
-
-static void _tnl_import_color( GLcontext *ctx,
-			       GLboolean writable,
-			       GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_color(ctx,
-			  GL_FLOAT,
-			  stride ? 4*sizeof(GLfloat) : 0,
-			  4,
-			  writable,
-			  &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Color.data = (GLfloat (*)[4]) data;
-   inputs->Color.start = (GLfloat *) data;
-   inputs->Color.stride = tmp->StrideB;
-   inputs->Color.size = tmp->Size;
-}
-
-
-static void _tnl_import_secondarycolor( GLcontext *ctx,
-					GLboolean writable,
-					GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_secondarycolor(ctx, 
-				   GL_FLOAT,
-				   stride ? 4*sizeof(GLfloat) : 0,
-				   4,
-				   writable,
-				   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->SecondaryColor.data = (GLfloat (*)[4]) data;
-   inputs->SecondaryColor.start = (GLfloat *) data;
-   inputs->SecondaryColor.stride = tmp->StrideB;
-   inputs->SecondaryColor.size = tmp->Size;
-}
-
-static void _tnl_import_fogcoord( GLcontext *ctx,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_fogcoord(ctx, GL_FLOAT,
-			     stride ? sizeof(GLfloat) : 0, writable,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->FogCoord.data = (GLfloat (*)[4]) data;
-   inputs->FogCoord.start = (GLfloat *) data;
-   inputs->FogCoord.stride = tmp->StrideB;
-}
-
-static void _tnl_import_index( GLcontext *ctx,
-			       GLboolean writable,
-			       GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_index(ctx, GL_FLOAT,
-			  stride ? sizeof(GLfloat) : 0, writable,
-			  &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Index.data = (GLfloat (*)[4]) data;
-   inputs->Index.start = (GLfloat *) data;
-   inputs->Index.stride = tmp->StrideB;
-}
-
-
-static void _tnl_import_texcoord( GLcontext *ctx,
-				  GLuint unit,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_texcoord(ctx, unit, GL_FLOAT,
-			     stride ? 4 * sizeof(GLfloat) : 0,
-			     0,
-			     writable,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->TexCoord[unit].data = (GLfloat (*)[4]) data;
-   inputs->TexCoord[unit].start = (GLfloat *) data;
-   inputs->TexCoord[unit].stride = tmp->StrideB;
-   inputs->TexCoord[unit].size = tmp->Size;
-}
-
-
-static void _tnl_import_edgeflag( GLcontext *ctx,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-   (void) writable; (void) stride;
-
-   tmp = _ac_import_edgeflag(ctx, GL_UNSIGNED_BYTE,
-			     sizeof(GLubyte),
-			     0,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->EdgeFlag = (GLubyte *) data;
-}
-
-
-
-static void _tnl_import_attrib( GLcontext *ctx,
-                                GLuint index,
-                                GLboolean writable,
-                                GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   ASSERT(index < MAX_VERTEX_PROGRAM_ATTRIBS);
-
-   tmp = _ac_import_attrib(ctx, index, GL_FLOAT,
-                           stride ? 4 * sizeof(GLfloat) : 0,
-                           4,  /* want GLfloat[4] */
-                           writable,
-                           &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Attribs[index].data = (GLfloat (*)[4]) data;
-   inputs->Attribs[index].start = (GLfloat *) data;
-   inputs->Attribs[index].stride = tmp->StrideB;
-   inputs->Attribs[index].size = tmp->Size;
-}
-
-
-static void _tnl_constant_attrib( TNLcontext *tnl,
-				  struct tnl_vertex_arrays *tmp,
-				  GLuint i )
-{
-   tmp->Attribs[i].count = 1;
-   tmp->Attribs[i].data = (GLfloat (*)[4]) tnl->vtx.current[i];
-   tmp->Attribs[i].start = tnl->vtx.current[i];
-   tmp->Attribs[i].size = 4; 
-   tmp->Attribs[i].stride = 0;
-   tnl->vb.AttribPtr[i] = &tmp->Attribs[i];
-}
-
-
-
-void _tnl_vb_bind_arrays( GLcontext *ctx, GLint start, GLint end)
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->array_inputs;
-   const struct gl_vertex_program *program
-      = ctx->VertexProgram._Enabled ? ctx->VertexProgram.Current : NULL;
-   GLuint i, index;
-
-   VB->Count = end - start;
-   VB->Elts = NULL;
-
-   _ac_import_range( ctx, start, end );
-
-   /* Note that the generic attribute arrays are treated differently
-    * depending on whether an NV or ARB vertex program is enabled
-    * (corresponding to aliasing vs. non-aliasing behaviour).
-    * Generic array 0 always aliases vertex position.
-    */
-   for (index = 0; index < VERT_ATTRIB_MAX; index++) {
-      if (ctx->VertexProgram._Enabled
-          && (program->IsNVProgram || index == 0)
-          && ctx->Array.ArrayObj->VertexAttrib[index].Enabled) {
-         /* Use generic attribute array.  If an NV vertex program is active,
-          * the generic arrays override the conventional attributes.
-          * Otherwise, if an ARB vertex program is active, we'll import the
-          * generic attributes without aliasing over conventional attribs
-          * (see below).
-          */
-         _tnl_import_attrib( ctx, index, GL_FALSE, GL_TRUE );
-         VB->AttribPtr[index] = &tmp->Attribs[index];
-      }
-      /* use conventional arrays... */
-      else if (index == VERT_ATTRIB_POS) {
-	 _tnl_import_vertex( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Obj.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_POS] = &tmp->Obj;
-      }
-      else if (index == VERT_ATTRIB_NORMAL) {
-	 _tnl_import_normal( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Normal.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_NORMAL] = &tmp->Normal;
-      }
-      else if (index == VERT_ATTRIB_COLOR0) {
-	 _tnl_import_color( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Color.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &tmp->Color;
-      }
-      else if (index == VERT_ATTRIB_COLOR1) {
-	 _tnl_import_secondarycolor( ctx, GL_FALSE, GL_FALSE );
-	 tmp->SecondaryColor.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_COLOR1] = &tmp->SecondaryColor;
-      }
-      else if (index == VERT_ATTRIB_FOG) {
-	 _tnl_import_fogcoord( ctx, GL_FALSE, GL_FALSE );
-	 tmp->FogCoord.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_FOG] = &tmp->FogCoord;
-      }
-      else if (index == VERT_ATTRIB_COLOR_INDEX) {
-         _tnl_import_index( ctx, GL_FALSE, GL_FALSE );
-         tmp->Index.count = VB->Count;
-         VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX] = &tmp->Index;
-      }
-      else if (index >= VERT_ATTRIB_TEX0 && index <= VERT_ATTRIB_TEX7) {
-	 i = index - VERT_ATTRIB_TEX0;
-	 _tnl_import_texcoord( ctx, i, GL_FALSE, GL_FALSE );
-	 tmp->TexCoord[i].count = VB->Count;
-	 VB->AttribPtr[index] = &tmp->TexCoord[i];
-      }
-      else if (index >= VERT_ATTRIB_GENERIC1 &&
-               index <= VERT_ATTRIB_GENERIC15) {
-         const GLuint arrayIndex = index - VERT_ATTRIB_GENERIC0;
-         if (program && !program->IsNVProgram &&
-	     ctx->Array.ArrayObj->VertexAttrib[arrayIndex].Enabled) {
-            /* GL_ARB_vertex_program: bind a generic attribute array */
-            _tnl_import_attrib(ctx, arrayIndex, GL_FALSE, GL_TRUE);
-            VB->AttribPtr[index] = &tmp->Attribs[arrayIndex];
-         }
-	 else {
-            _tnl_constant_attrib(tnl, tmp, index);
-         }
-      }
-      else {
-	 _tnl_constant_attrib(tnl, tmp, index);
-      }
-      assert(VB->AttribPtr[index]);
-      assert(VB->AttribPtr[index]->size);
-   }
-
-   /* odd-ball vertex attributes */
-   {
-      _tnl_import_edgeflag( ctx, GL_TRUE, sizeof(GLboolean) );
-      VB->EdgeFlag = (GLboolean *) tmp->EdgeFlag;
-   }
-
-   /* These are constant & could be precalculated:
-    */
-   for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-      _tnl_constant_attrib(tnl, tmp, i);
-   }
-
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
diff --git a/src/mesa/tnl/t_array_import.h b/src/mesa/tnl/t_array_import.h
deleted file mode 100644
index 39b77641d5..0000000000
--- a/src/mesa/tnl/t_array_import.h
+++ /dev/null
@@ -1,36 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-#ifndef _T_ARRAY_IMPORT_H
-#define _T_ARRAY_IMPORT_H
-
-#include "mtypes.h"
-#include "t_context.h"
-
-extern void _tnl_vb_bind_arrays( GLcontext *ctx, GLint start, GLint end );
-
-extern void _tnl_array_import_init( GLcontext *ctx );
-
-#endif
diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index 154780cc97..d9458b74ec 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -26,38 +26,19 @@
  */
 
 
-#include "api_arrayelt.h"
 #include "glheader.h"
 #include "imports.h"
 #include "context.h"
 #include "macros.h"
 #include "mtypes.h"
-#include "dlist.h"
 #include "light.h"
-#include "vtxfmt.h"
 
 #include "tnl.h"
-#include "t_array_api.h"
 #include "t_context.h"
 #include "t_pipeline.h"
-#include "t_save_api.h"
 #include "t_vp_build.h"
-#include "t_vtx_api.h"
-
-
-
-static void
-install_driver_callbacks( GLcontext *ctx )
-{
-   ctx->Driver.NewList = _tnl_NewList;
-   ctx->Driver.EndList = _tnl_EndList;
-   ctx->Driver.FlushVertices = _tnl_FlushVertices;
-   ctx->Driver.SaveFlushVertices = _tnl_SaveFlushVertices;
-   ctx->Driver.BeginCallList = _tnl_BeginCallList;
-   ctx->Driver.EndCallList = _tnl_EndCallList;
-}
-
 
+#include "vbo/vbo.h"
 
 GLboolean
 _tnl_CreateContext( GLcontext *ctx )
@@ -72,20 +53,13 @@ _tnl_CreateContext( GLcontext *ctx )
       return GL_FALSE;
    }
 
-   if (_mesa_getenv("MESA_CODEGEN"))
-      tnl->AllowCodegen = GL_TRUE;
-
    /* Initialize the VB.
     */
    tnl->vb.Size = ctx->Const.MaxArrayLockSize + MAX_CLIPPED_VERTICES;
 
 
-   /* Initialize tnl state and tnl->vtxfmt.
+   /* Initialize tnl state.
     */
-   _tnl_save_init( ctx );
-   _tnl_array_init( ctx );
-   _tnl_vtx_init( ctx );
-
    if (ctx->_MaintainTnlProgram) {
       _tnl_ProgramCacheInit( ctx );
       _tnl_install_pipeline( ctx, _tnl_vp_pipeline );
@@ -93,34 +67,18 @@ _tnl_CreateContext( GLcontext *ctx )
       _tnl_install_pipeline( ctx, _tnl_default_pipeline );
    }
 
-   /* Initialize the arrayelt helper
-    */
-   if (!_ae_create_context( ctx ))
-      return GL_FALSE;
-
-
    tnl->NeedNdcCoords = GL_TRUE;
-   tnl->LoopbackDListCassettes = GL_FALSE;
-   tnl->CalcDListNormalLengths = GL_TRUE;
    tnl->AllowVertexFog = GL_TRUE;
    tnl->AllowPixelFog = GL_TRUE;
 
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
-
    /* Set a few default values in the driver struct.
     */
-   install_driver_callbacks(ctx);
-   ctx->Driver.NeedFlush = 0;
-   ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
-
    tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
    tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
    tnl->Driver.NotifyMaterialChange = _mesa_validate_all_lighting_tables;
 
+   tnl->nr_blocks = 0;
+
    return GL_TRUE;
 }
 
@@ -130,11 +88,7 @@ _tnl_DestroyContext( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
 
-   _tnl_array_destroy( ctx );
-   _tnl_vtx_destroy( ctx );
-   _tnl_save_destroy( ctx );
    _tnl_destroy_pipeline( ctx );
-   _ae_destroy_context( ctx );
 
    if (ctx->_MaintainTnlProgram)
       _tnl_ProgramCacheDestroy( ctx );
@@ -155,10 +109,7 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
          || !tnl->AllowPixelFog;
    }
 
-   _ae_invalidate_state(ctx, new_state);
-
    tnl->pipeline.new_state |= new_state;
-   tnl->vtx.eval.new_state |= new_state;
 
    /* Calculate tnl->render_inputs:
     */
@@ -205,36 +156,21 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
 
 
 void
-_tnl_wakeup_exec( GLcontext *ctx )
+_tnl_wakeup( GLcontext *ctx )
 {
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   install_driver_callbacks(ctx);
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
    /* Assume we haven't been getting state updates either:
     */
    _tnl_InvalidateState( ctx, ~0 );
 
+#if 0
    if (ctx->Light.ColorMaterialEnabled) {
       _mesa_update_color_material( ctx, 
 				   ctx->Current.Attrib[VERT_ATTRIB_COLOR0] );
    }
+#endif
 }
 
 
-void
-_tnl_wakeup_save_exec( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   _tnl_wakeup_exec( ctx );
-   _mesa_install_save_vtxfmt( ctx, &tnl->save_vtxfmt );
-}
 
 
 /**
@@ -250,27 +186,6 @@ _tnl_need_projected_coords( GLcontext *ctx, GLboolean mode )
 }
 
 void
-_tnl_need_dlist_loopback( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->LoopbackDListCassettes = mode;
-}
-
-void
-_tnl_need_dlist_norm_lengths( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->CalcDListNormalLengths = mode;
-}
-
-void
-_tnl_isolate_materials( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->IsolateMaterials = mode;
-}
-
-void
 _tnl_allow_vertex_fog( GLcontext *ctx, GLboolean value )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h
index d5414bd730..a872f26177 100644
--- a/src/mesa/tnl/t_context.h
+++ b/src/mesa/tnl/t_context.h
@@ -43,29 +43,6 @@
  * stages to the vertex_buffer TNLcontext::vb, where the vertex data
  * is stored. The last stage in the pipeline is the rasterizer.
  *
- * The initial vertex_buffer data may either come from an ::immediate
- * structure or client vertex_arrays or display lists:
- *
- *
- * - The ::immediate structure records all the GL commands issued between
- * glBegin and glEnd.  \n
- * The structure accumulates data, until it is either full or it is
- * flushed (usually by a state change). Before starting then the pipeline,
- * the collected vertex data in ::immediate has to be pushed into
- * TNLcontext::vb.
- * This happens in ::_tnl_vb_bind_immediate. The pipeline is then run by
- * calling tnl_device_driver::RunPipeline = ::_tnl_run_pipeline, which
- * is stored in TNLcontext::Driver.   \n
- * An ::immediate does (for performance reasons) usually not finish with a
- * glEnd, and hence it also does not need to start with a glBegin.
- * This means that the last vertices of one ::immediate may need to be
- * saved for the next one.
- *
- *
- * - NOT SURE ABOUT THIS: The vertex_arrays structure is used to handle
- * glDrawArrays etc.  \n
- * Here, the data of the vertex_arrays is copied by ::_tnl_vb_bind_arrays
- * into TNLcontext::vb, so that the pipeline can be started.
  */
 
 
@@ -79,6 +56,7 @@
 #include "math/m_vector.h"
 #include "math/m_xform.h"
 
+#include "vbo/vbo.h"
 
 #define MAX_PIPELINE_STAGES     30
 
@@ -106,7 +84,7 @@ enum {
 	_TNL_ATTRIB_COLOR1 = 4,
 	_TNL_ATTRIB_FOG = 5,
 	_TNL_ATTRIB_COLOR_INDEX = 6,
-	_TNL_ATTRIB_SEVEN = 7,
+	_TNL_ATTRIB_EDGEFLAG = 7,
 	_TNL_ATTRIB_TEX0 = 8,
 	_TNL_ATTRIB_TEX1 = 9,
 	_TNL_ATTRIB_TEX2 = 10,
@@ -115,7 +93,8 @@ enum {
 	_TNL_ATTRIB_TEX5 = 13,
 	_TNL_ATTRIB_TEX6 = 14,
 	_TNL_ATTRIB_TEX7 = 15,
-	_TNL_ATTRIB_GENERIC0 = 16,
+
+	_TNL_ATTRIB_GENERIC0 = 16, /* doesn't really exist! */
 	_TNL_ATTRIB_GENERIC1 = 17,
 	_TNL_ATTRIB_GENERIC2 = 18,
 	_TNL_ATTRIB_GENERIC3 = 19,
@@ -131,21 +110,36 @@ enum {
 	_TNL_ATTRIB_GENERIC13 = 29,
 	_TNL_ATTRIB_GENERIC14 = 30,
 	_TNL_ATTRIB_GENERIC15 = 31,
-	_TNL_ATTRIB_MAT_FRONT_AMBIENT = 32,
-	_TNL_ATTRIB_MAT_BACK_AMBIENT = 33,
-	_TNL_ATTRIB_MAT_FRONT_DIFFUSE = 34,
-	_TNL_ATTRIB_MAT_BACK_DIFFUSE = 35,
-	_TNL_ATTRIB_MAT_FRONT_SPECULAR = 36,
-	_TNL_ATTRIB_MAT_BACK_SPECULAR = 37,
-	_TNL_ATTRIB_MAT_FRONT_EMISSION = 38,
-	_TNL_ATTRIB_MAT_BACK_EMISSION = 39,
-	_TNL_ATTRIB_MAT_FRONT_SHININESS = 40,
-	_TNL_ATTRIB_MAT_BACK_SHININESS = 41,
-	_TNL_ATTRIB_MAT_FRONT_INDEXES = 42,
-	_TNL_ATTRIB_MAT_BACK_INDEXES = 43,
-	_TNL_ATTRIB_EDGEFLAG = 44,
-	_TNL_ATTRIB_POINTSIZE = 45,
-	_TNL_ATTRIB_MAX = 46
+
+	/* These alias with the generics, but they are not active
+	 * concurrently, so it's not a problem.  The TNL module
+	 * doesn't have to do anything about this as this is how they
+	 * are passed into the _draw_prims callback.
+	 *
+	 * When we generate fixed-function replacement programs (in
+	 * t_vp_build.c currently), they refer to the appropriate
+	 * generic attribute in order to pick up per-vertex material
+	 * data.
+	 */
+	_TNL_ATTRIB_MAT_FRONT_AMBIENT = 16,
+	_TNL_ATTRIB_MAT_BACK_AMBIENT = 17,
+	_TNL_ATTRIB_MAT_FRONT_DIFFUSE = 18,
+	_TNL_ATTRIB_MAT_BACK_DIFFUSE = 19,
+	_TNL_ATTRIB_MAT_FRONT_SPECULAR = 20,
+	_TNL_ATTRIB_MAT_BACK_SPECULAR = 21,
+	_TNL_ATTRIB_MAT_FRONT_EMISSION = 22,
+	_TNL_ATTRIB_MAT_BACK_EMISSION = 23,
+	_TNL_ATTRIB_MAT_FRONT_SHININESS = 24,
+	_TNL_ATTRIB_MAT_BACK_SHININESS = 25,
+	_TNL_ATTRIB_MAT_FRONT_INDEXES = 26,
+	_TNL_ATTRIB_MAT_BACK_INDEXES = 27,
+
+	/* This is really a VERT_RESULT, not an attrib.  Need to fix
+	 * tnl to understand the difference.
+	 */
+	_TNL_ATTRIB_POINTSIZE = 16,
+
+	_TNL_ATTRIB_MAX = 32
 } ;
 
 #define _TNL_ATTRIB_TEX(u)       (_TNL_ATTRIB_TEX0 + (u))
@@ -166,8 +160,8 @@ enum {
 #define _TNL_FIRST_GENERIC _TNL_ATTRIB_GENERIC0
 #define _TNL_LAST_GENERIC  _TNL_ATTRIB_GENERIC15
 
-#define _TNL_FIRST_MAT       _TNL_ATTRIB_MAT_FRONT_AMBIENT
-#define _TNL_LAST_MAT        _TNL_ATTRIB_MAT_BACK_INDEXES
+#define _TNL_FIRST_MAT       _TNL_ATTRIB_MAT_FRONT_AMBIENT /* GENERIC0 */
+#define _TNL_LAST_MAT        _TNL_ATTRIB_MAT_BACK_INDEXES  /* GENERIC11 */
 
 /* Number of available generic attributes */
 #define _TNL_NUM_GENERIC 16
@@ -175,221 +169,22 @@ enum {
 /* Number of attributes used for evaluators */
 #define _TNL_NUM_EVAL 16
 
+
 #define PRIM_BEGIN     0x10
 #define PRIM_END       0x20
-#define PRIM_WEAK      0x40
 #define PRIM_MODE_MASK 0x0f
 
-/* 
- */
-struct tnl_prim {
-   GLuint mode;
-   GLuint start;
-   GLuint count;
-};
-
-
-
-struct tnl_eval1_map {
-   struct gl_1d_map *map;
-   GLuint sz;
-};
-
-struct tnl_eval2_map {
-   struct gl_2d_map *map;
-   GLuint sz;
-};
-
-struct tnl_eval {
-   GLuint new_state;
-   struct tnl_eval1_map map1[_TNL_NUM_EVAL];
-   struct tnl_eval2_map map2[_TNL_NUM_EVAL];
-};
-
-
-#define TNL_MAX_PRIM 16
-#define TNL_MAX_COPIED_VERTS 3
-
-struct tnl_copied_vtx {
-   GLfloat buffer[_TNL_ATTRIB_MAX * 4 * TNL_MAX_COPIED_VERTS];
-   GLuint nr;
-};
-
-#define VERT_BUFFER_SIZE 2048	/* 8kbytes */
-
-
-typedef void (*tnl_attrfv_func)( const GLfloat * );
-
-struct _tnl_dynfn {
-   struct _tnl_dynfn *next, *prev;
-   GLuint key;
-   char *code;
-};
-
-struct _tnl_dynfn_lists {
-   struct _tnl_dynfn Vertex[4];
-   struct _tnl_dynfn Attribute[4];
-};
-
-struct _tnl_dynfn_generators {
-   struct _tnl_dynfn *(*Vertex[4])( GLcontext *ctx, int key );
-   struct _tnl_dynfn *(*Attribute[4])( GLcontext *ctx, int key );
-};
-
-#define _TNL_MAX_ATTR_CODEGEN 32
-
-
-/**
- * The assembly of vertices in immediate mode is separated from
- * display list compilation.  This allows a simpler immediate mode
- * treatment and a display list compiler better suited to
- * hardware-acceleration.
- */
-struct tnl_vtx {
-   GLfloat buffer[VERT_BUFFER_SIZE];
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLubyte active_sz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-   struct tnl_prim prim[TNL_MAX_PRIM];
-   GLuint prim_count;
-   GLfloat *vbptr;		      /* cursor, points into buffer */
-   GLfloat vertex[_TNL_ATTRIB_MAX*4]; /* current vertex */
-   GLfloat *attrptr[_TNL_ATTRIB_MAX]; /* points into vertex */
-   GLfloat *current[_TNL_ATTRIB_MAX]; /* points into ctx->Current, etc */
-   GLfloat CurrentFloatEdgeFlag;
-   GLuint counter, initial_counter;
-   struct tnl_copied_vtx copied;
-
-   /** Note extra space for error handler: */
-   tnl_attrfv_func tabfv[_TNL_ATTRIB_ERROR+1][4];
-
-   struct _tnl_dynfn_lists cache;
-   struct _tnl_dynfn_generators gen;
-
-   struct tnl_eval eval;
-   GLboolean *edgeflag_tmp;
-   GLboolean have_materials;
-};
-
-
-
-
-/* For display lists, this structure holds a run of vertices of the
- * same format, and a strictly well-formed set of begin/end pairs,
- * starting on the first vertex and ending at the last.  Vertex
- * copying on buffer breaks is precomputed according to these
- * primitives, though there are situations where the copying will need
- * correction at execute-time, perhaps by replaying the list as
- * immediate mode commands.
- *
- * On executing this list, the 'current' values may be updated with
- * the values of the final vertex, and often no fixup of the start of
- * the vertex list is required.
- *
- * Eval and other commands that don't fit into these vertex lists are
- * compiled using the fallback opcode mechanism provided by dlist.c.
- */
-struct tnl_vertex_list {
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-
-   GLfloat *buffer;
-   GLuint count;
-   GLuint wrap_count;		/* number of copied vertices at start */
-   GLboolean have_materials;	/* bit of a hack - quick check for materials */
-   GLboolean dangling_attr_ref;	/* current attr implicitly referenced 
-				   outside the list */
-
-   GLfloat *normal_lengths;
-   struct tnl_prim *prim;
-   GLuint prim_count;
-
-   struct tnl_vertex_store *vertex_store;
-   struct tnl_primitive_store *prim_store;
-};
-
-/* These buffers should be a reasonable size to support upload to
- * hardware?  Maybe drivers should stitch them back together, or
- * specify a desired size?
- */
-#define SAVE_BUFFER_SIZE (16*1024)
-#define SAVE_PRIM_SIZE   128
-
-/* Storage to be shared among several vertex_lists.
- */
-struct tnl_vertex_store {
-   GLfloat buffer[SAVE_BUFFER_SIZE];
-   GLuint used;
-   GLuint refcount;
-};
-
-struct tnl_primitive_store {
-   struct tnl_prim buffer[SAVE_PRIM_SIZE];
-   GLuint used;
-   GLuint refcount;
-};
-
-
-struct tnl_save {
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-
-   GLfloat *buffer;
-   GLuint count;
-   GLuint wrap_count;
-   GLuint replay_flags;
-
-   struct tnl_prim *prim;
-   GLuint prim_count, prim_max;
-
-   struct tnl_vertex_store *vertex_store;
-   struct tnl_primitive_store *prim_store;
-
-   GLfloat *vbptr;		   /* cursor, points into buffer */
-   GLfloat vertex[_TNL_ATTRIB_MAX*4];	   /* current values */
-   GLfloat *attrptr[_TNL_ATTRIB_MAX];
-   GLuint counter, initial_counter;
-   GLboolean dangling_attr_ref;
-   GLboolean have_materials;
-
-   GLuint opcode_vertex_list;
-
-   struct tnl_copied_vtx copied;
-   
-   GLfloat CurrentFloatEdgeFlag;
-
-   GLfloat *current[_TNL_ATTRIB_MAX]; /* points into ctx->ListState */
-   GLubyte *currentsz[_TNL_ATTRIB_MAX];
+static INLINE GLuint _tnl_translate_prim( const struct _mesa_prim *prim )
+{
+   GLuint flag;
+   flag = prim->mode;
+   if (prim->begin) flag |= PRIM_BEGIN;
+   if (prim->end) flag |= PRIM_END;
+   return flag;
+}
 
-   void (*tabfv[_TNL_ATTRIB_MAX][4])( const GLfloat * );
-};
 
 
-/**
- * A collection of vertex arrays.
- */
-struct tnl_vertex_arrays
-{
-   /* Conventional vertex attribute arrays */
-   GLvector4f  Obj;
-   GLvector4f  Normal;
-   GLvector4f  Color;
-   GLvector4f  SecondaryColor;
-   GLvector4f  FogCoord;
-   GLvector4f  TexCoord[MAX_TEXTURE_COORD_UNITS];
-   GLvector4f  Index;
-
-   GLubyte     *EdgeFlag;
-   GLuint      *Elt;
-
-   /* These attributes don't alias with the conventional attributes.
-    * The GL_NV_vertex_program extension defines 16 extra sets of vertex
-    * arrays which have precedent over the conventional arrays when enabled.
-    */
-   /* XXX I think the array size is wronge (47 vs. 16) */
-   GLvector4f  Attribs[_TNL_ATTRIB_MAX];
-};
-
 
 /**
  * Contains the current state of a running pipeline.
@@ -424,11 +219,10 @@ struct vertex_buffer
    GLvector4f  *FogCoordPtr;	                /* _TNL_BIT_FOG */
    GLvector4f  *VaryingPtr[MAX_VARYING_VECTORS];
 
-   struct tnl_prim  *Primitive;	              
+   const struct _mesa_prim  *Primitive;	              
    GLuint      PrimitiveCount;	      
 
    /* Inputs to the vertex program stage */
-   /* XXX This array may be too large (47 vs. 16) */
    GLvector4f *AttribPtr[_TNL_ATTRIB_MAX];      /* GL_NV_vertex_program */
 };
 
@@ -626,12 +420,6 @@ struct tnl_device_driver
     * arrays.
     */
 
-   GLboolean (*NotifyBegin)(GLcontext *ctx, GLenum p);
-   /* Allow drivers to hook in optimized begin/end engines.
-    * Return value:  GL_TRUE - driver handled the begin
-    *                GL_FALSE - driver didn't handle the begin
-    */
-
    /***
     *** Rendering -- These functions called only from t_vb_render.c
     ***/
@@ -737,26 +525,11 @@ typedef struct
     */
    struct tnl_device_driver Driver;
 
-   /* Execute:
-    */
-   struct tnl_vtx vtx;
-   
-   /* Compile:
-    */
-   struct tnl_save save;
-
    /* Pipeline
     */
    struct tnl_pipeline pipeline;
    struct vertex_buffer vb;
 
-   /* GLvectors for binding to vb:
-    */
-   struct tnl_vertex_arrays vtx_inputs;
-   struct tnl_vertex_arrays save_inputs;
-   struct tnl_vertex_arrays current;
-   struct tnl_vertex_arrays array_inputs;
-
    /* Clipspace/ndc/window vertex managment:
     */
    struct tnl_clipspace clipspace;
@@ -764,26 +537,21 @@ typedef struct
    /* Probably need a better configuration mechanism:
     */
    GLboolean NeedNdcCoords;
-   GLboolean LoopbackDListCassettes;
-   GLboolean CalcDListNormalLengths;
-   GLboolean IsolateMaterials;
    GLboolean AllowVertexFog;
    GLboolean AllowPixelFog;
-   GLboolean AllowCodegen;
-
    GLboolean _DoVertexFog;  /* eval fog function at each vertex? */
 
-   /* If True, it means we started a glBegin/End primtive with an invalid
-    * vertex/fragment program or incomplete framebuffer.  In that case,
-    * discard any buffered vertex data.
-    */
-   GLboolean DiscardPrimitive;
-
    DECLARE_RENDERINPUTS(render_inputs_bitset);
 
-   GLvertexformat exec_vtxfmt;
-   GLvertexformat save_vtxfmt;
+   GLvector4f tmp_inputs[VERT_ATTRIB_MAX];
 
+   /* Temp storage for t_draw.c: 
+    */
+   GLubyte *block[VERT_ATTRIB_MAX];
+   GLuint nr_blocks;
+
+   /* Cache of fixed-function-replacing vertex programs:
+    */
    struct tnl_cache *vp_cache;
 
 } TNLcontext;
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
new file mode 100644
index 0000000000..c97cf5f7b2
--- /dev/null
+++ b/src/mesa/tnl/t_draw.c
@@ -0,0 +1,406 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "context.h"
+#include "imports.h"
+#include "state.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "enums.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+#include "t_vp_build.h"
+#include "t_vertex.h"
+#include "tnl.h"
+
+
+
+static GLubyte *get_space(GLcontext *ctx, GLuint bytes)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLubyte *space = _mesa_malloc(bytes);
+   
+   tnl->block[tnl->nr_blocks++] = space;
+   return space;
+}
+
+
+static void free_space(GLcontext *ctx)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+   for (i = 0; i < tnl->nr_blocks; i++)
+      _mesa_free(tnl->block[i]);
+   tnl->nr_blocks = 0;
+}
+
+
+/* Convert the incoming array to GLfloats.  Understands the
+ * array->Normalized flag and selects the correct conversion method.
+ */
+#define CONVERT( TYPE, MACRO ) do {		\
+   GLuint i, j;					\
+   if (input->Normalized) {			\
+      for (i = 0; i < count; i++) {		\
+	 const TYPE *in = (TYPE *)ptr;		\
+	 for (j = 0; j < sz; j++) {		\
+	    *fptr++ = MACRO(*in);		\
+	    in++;				\
+	 }					\
+	 ptr += input->StrideB;			\
+      }						\
+   } else {					\
+      for (i = 0; i < count; i++) {		\
+	 const TYPE *in = (TYPE *)ptr;		\
+	 for (j = 0; j < sz; j++) {		\
+	    *fptr++ = (GLfloat)(*in);		\
+	    in++;				\
+	 }					\
+	 ptr += input->StrideB;			\
+      }						\
+   }						\
+} while (0)
+
+
+
+/* Adjust pointer to point at first requested element, convert to
+ * floating point, populate VB->AttribPtr[].
+ */
+static void _tnl_import_array( GLcontext *ctx,
+			       GLuint attrib,
+			       GLuint count,
+			       const struct gl_client_array *input,
+			       const GLubyte *ptr )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint stride = input->StrideB;
+
+   if (input->Type != GL_FLOAT) {
+      const GLuint sz = input->Size;
+      GLubyte *buf = get_space(ctx, count * sz * sizeof(GLfloat));
+      GLfloat *fptr = (GLfloat *)buf;
+
+      switch (input->Type) {
+      case GL_BYTE: 
+	 CONVERT(GLbyte, BYTE_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_BYTE: 
+	 CONVERT(GLubyte, UBYTE_TO_FLOAT); 
+	 break;
+      case GL_SHORT: 
+	 CONVERT(GLshort, SHORT_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_SHORT: 
+	 CONVERT(GLushort, USHORT_TO_FLOAT); 
+	 break;
+      case GL_INT: 
+	 CONVERT(GLint, INT_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_INT: 
+	 CONVERT(GLuint, UINT_TO_FLOAT); 
+	 break;
+      case GL_DOUBLE: 
+	 CONVERT(GLdouble, (GLfloat)); 
+	 break;
+      default:
+	 assert(0);
+	 break;
+      }
+
+      ptr = buf;
+      stride = sz * sizeof(GLfloat);
+   }
+
+   VB->AttribPtr[attrib] = &tnl->tmp_inputs[attrib];
+   VB->AttribPtr[attrib]->data = (GLfloat (*)[4])ptr;
+   VB->AttribPtr[attrib]->start = (GLfloat *)ptr;
+   VB->AttribPtr[attrib]->count = count;
+   VB->AttribPtr[attrib]->stride = stride;
+   VB->AttribPtr[attrib]->size = input->Size;
+
+   /* This should die, but so should the whole GLvector4f concept: 
+    */
+   VB->AttribPtr[attrib]->flags = (((1<<input->Size)-1) | 
+				   VEC_NOT_WRITEABLE |
+				   (stride == 4*sizeof(GLfloat) ? 0 : VEC_BAD_STRIDE));
+   
+   VB->AttribPtr[attrib]->storage = NULL;
+}
+
+#define CLIPVERTS  ((6 + MAX_CLIP_PLANES) * 2)
+
+
+static GLboolean *_tnl_import_edgeflag( GLcontext *ctx,
+					const GLvector4f *input,
+					GLuint count)
+{
+   const GLubyte *ptr = (const GLubyte *)input->data;
+   const GLuint stride = input->stride;
+   GLboolean *space = (GLboolean *)get_space(ctx, count + CLIPVERTS);
+   GLboolean *bptr = space;
+   GLuint i;
+
+   for (i = 0; i < count; i++) {
+      *bptr++ = ((GLfloat *)ptr)[0] == 1.0;
+      ptr += stride;
+   }
+
+   return space;
+}
+
+
+static void bind_inputs( GLcontext *ctx, 
+			 const struct gl_client_array *inputs[],
+			 GLint count,
+			 struct gl_buffer_object **bo,
+			 GLuint *nr_bo )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+
+   /* Map all the VBOs
+    */
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      const void *ptr;
+
+      if (inputs[i]->BufferObj->Name) { 
+	 if (!inputs[i]->BufferObj->Pointer) {
+	    bo[*nr_bo] = inputs[i]->BufferObj;
+	    (*nr_bo)++;
+	    ctx->Driver.MapBuffer(ctx, 
+				  GL_ARRAY_BUFFER,
+				  GL_READ_ONLY_ARB,
+				  inputs[i]->BufferObj);
+	    
+	    assert(inputs[i]->BufferObj->Pointer);
+	 }
+	 
+	 ptr = ADD_POINTERS(inputs[i]->BufferObj->Pointer,
+			    inputs[i]->Ptr);
+      }
+      else
+	 ptr = inputs[i]->Ptr;
+
+      /* Just make sure the array is floating point, otherwise convert to
+       * temporary storage.  
+       *
+       * XXX: remove the GLvector4f type at some stage and just use
+       * client arrays.
+       */
+      _tnl_import_array(ctx, i, count, inputs[i], ptr);
+   }
+
+   /* We process only the vertices between min & max index:
+    */
+   VB->Count = count;
+
+
+   /* Legacy pointers -- remove one day.
+    */
+   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
+   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
+   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
+   VB->ColorPtr[1] = NULL;
+   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
+   VB->IndexPtr[1] = NULL;
+   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
+   VB->SecondaryColorPtr[1] = NULL;
+   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
+   }
+
+   /* Clipping and drawing code still requires this to be a packed
+    * array of ubytes which can be written into.  TODO: Fix and
+    * remove.
+    */
+   if (ctx->Polygon.FrontMode != GL_FILL ||
+       ctx->Polygon.BackMode != GL_FILL)
+   {
+      VB->EdgeFlag = _tnl_import_edgeflag( ctx, 
+					   VB->AttribPtr[_TNL_ATTRIB_EDGEFLAG],
+					   VB->Count );
+   }
+
+}
+
+
+/* Translate indices to GLuints and store in VB->Elts.
+ */
+static void bind_indices( GLcontext *ctx,
+			  const struct _mesa_index_buffer *ib,
+			  struct gl_buffer_object **bo,
+			  GLuint *nr_bo)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+   void *ptr;
+
+   if (!ib) {
+      VB->Elts = NULL;
+      return;
+   }
+
+   if (ib->obj->Name && !ib->obj->Pointer) {
+      bo[*nr_bo] = ib->obj;
+      (*nr_bo)++;
+      ctx->Driver.MapBuffer(ctx, 
+			    GL_ELEMENT_ARRAY_BUFFER,
+			    GL_READ_ONLY_ARB,
+			    ib->obj);
+
+      assert(ib->obj->Pointer);
+   }
+
+   ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);
+
+   if (ib->type == GL_UNSIGNED_INT) {
+      VB->Elts = (GLuint *) ptr;
+   }
+   else {
+      GLuint *elts = (GLuint *)get_space(ctx, ib->count * sizeof(GLuint));
+      VB->Elts = elts;
+
+      if (ib->type == GL_UNSIGNED_SHORT) {
+	 const GLushort *in = (GLushort *)ptr;
+	 for (i = 0; i < ib->count; i++) 
+	    *elts++ = (GLuint)(*in++);
+      }
+      else {
+	 const GLubyte *in = (GLubyte *)ptr;
+	 for (i = 0; i < ib->count; i++) 
+	    *elts++ = (GLuint)(*in++);
+      }
+   }
+}
+
+static void bind_prims( GLcontext *ctx,
+			const struct _mesa_prim *prim,
+			GLuint nr_prims )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+
+   VB->Primitive = prim;
+   VB->PrimitiveCount = nr_prims;
+}
+
+static void unmap_vbos( GLcontext *ctx,
+			struct gl_buffer_object **bo,
+			GLuint nr_bo )
+{
+   GLuint i;
+   for (i = 0; i < nr_bo; i++) { 
+      ctx->Driver.UnmapBuffer(ctx, 
+			      0, /* target -- I don't see why this would be needed */
+			      bo[i]);
+   }
+}
+
+
+
+/* This is the main entrypoint into the slimmed-down software tnl
+ * module.  In a regular swtnl driver, this can be plugged straight
+ * into the vbo->Driver.DrawPrims() callback.
+ */
+void _tnl_draw_prims( GLcontext *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   const GLuint TEST_SPLIT = 0;
+   const GLint max = TEST_SPLIT ? 8 : tnl->vb.Size - MAX_CLIPPED_VERTICES;
+
+   if (0)
+   {
+      GLuint i;
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
+      for (i = 0; i < nr_prims; i++)
+	 _mesa_printf("prim %d: %s start %d count %d\n", i, 
+		      _mesa_lookup_enum_by_nr(prim[i].mode),
+		      prim[i].start,
+		      prim[i].count);
+   }
+
+   if (min_index) {
+      /* We always translate away calls with min_index != 0. 
+       */
+      vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, 
+			min_index, max_index,
+			_tnl_draw_prims );
+      return;
+   }
+   else if (max_index >= max) {
+      /* The software TNL pipeline has a fixed amount of storage for
+       * vertices and it is necessary to split incoming drawing commands
+       * if they exceed that limit.
+       */
+      struct split_limits limits;
+      limits.max_verts = max;
+      limits.max_vb_size = ~0;
+      limits.max_indices = ~0;
+
+      /* This will split the buffers one way or another and
+       * recursively call back into this function.
+       */
+      vbo_split_prims( ctx, arrays, prim, nr_prims, ib, 
+		       0, max_index,
+		       _tnl_draw_prims,
+		       &limits );
+   }
+   else {
+      /* May need to map a vertex buffer object for every attribute plus
+       * one for the index buffer.
+       */
+      struct gl_buffer_object *bo[VERT_ATTRIB_MAX + 1];
+      GLuint nr_bo = 0;
+
+      /* Binding inputs may imply mapping some vertex buffer objects.
+       * They will need to be unmapped below.
+       */
+      bind_inputs(ctx, arrays, max_index+1, bo, &nr_bo);
+      bind_indices(ctx, ib, bo, &nr_bo);
+      bind_prims(ctx, prim, nr_prims );
+
+      TNL_CONTEXT(ctx)->Driver.RunPipeline(ctx);
+
+      unmap_vbos(ctx, bo, nr_bo);
+      free_space(ctx);
+   }
+}
+
diff --git a/src/mesa/tnl/t_save_api.c b/src/mesa/tnl/t_save_api.c
deleted file mode 100644
index dbbd095fe4..0000000000
--- a/src/mesa/tnl/t_save_api.c
+++ /dev/null
@@ -1,1741 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-
-
-/**
- * The display list compiler attempts to store lists of vertices with the
- * same vertex layout.  Additionally it attempts to minimize the need
- * for execute-time fixup of these vertex lists, allowing them to be
- * cached on hardware.
- *
- * There are still some circumstances where this can be thwarted, for
- * example by building a list that consists of one very long primitive
- * (eg Begin(Triangles), 1000 vertices, End), and calling that list
- * from inside a different begin/end object (Begin(Lines), CallList,
- * End).  
- *
- * In that case the code will have to replay the list as individual
- * commands through the Exec dispatch table, or fix up the copied
- * vertices at execute-time.
- *
- * The other case where fixup is required is when a vertex attribute
- * is introduced in the middle of a primitive.  Eg:
- *  Begin(Lines)
- *  TexCoord1f()           Vertex2f()
- *  TexCoord1f() Color3f() Vertex2f()
- *  End()
- *
- *  If the current value of Color isn't known at compile-time, this
- *  primitive will require fixup.
- *
- *
- * The list compiler currently doesn't attempt to compile lists
- * containing EvalCoord or EvalPoint commands.  On encountering one of
- * these, compilation falls back to opcodes.  
- *
- * This could be improved to fallback only when a mix of EvalCoord and
- * Vertex commands are issued within a single primitive.
- */
-
-
-#include "glheader.h"
-#include "context.h"
-#include "dlist.h"
-#include "enums.h"
-#include "macros.h"
-#include "api_validate.h"
-#include "api_arrayelt.h"
-#include "vtxfmt.h"
-#include "t_save_api.h"
-#include "dispatch.h"
-
-/*
- * NOTE: Old 'parity' issue is gone, but copying can still be
- * wrong-footed on replay.
- */
-static GLuint _save_copy_vertices( GLcontext *ctx, 
-				   const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT( ctx );
-   const struct tnl_prim *prim = &node->prim[node->prim_count-1];
-   GLuint nr = prim->count;
-   GLuint sz = tnl->save.vertex_size;
-   const GLfloat *src = node->buffer + prim->start * sz;
-   GLfloat *dst = tnl->save.copied.buffer;
-   GLuint ovf, i;
-
-   if (prim->mode & PRIM_END)
-      return 0;
-	 
-   switch( prim->mode & PRIM_MODE_MASK )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz*sizeof(GLfloat) );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz*sizeof(GLfloat) );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz*sizeof(GLfloat) );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      else {
-	 _mesa_memcpy( dst, src+(nr-1)*sz, sz*sizeof(GLfloat) );
-	 return 1;
-      }
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 _mesa_memcpy( dst, src+0, sz*sizeof(GLfloat) );
-	 return 1;
-      } else {
-	 _mesa_memcpy( dst, src+0, sz*sizeof(GLfloat) );
-	 _mesa_memcpy( dst+sz, src+(nr-1)*sz, sz*sizeof(GLfloat) );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz*sizeof(GLfloat) );
-      return i;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-
-static void
-build_normal_lengths( struct tnl_vertex_list *node )
-{
-   GLuint i;
-   GLfloat *len;
-   GLfloat *n = node->buffer;
-   GLuint stride = node->vertex_size;
-   GLuint count = node->count;
-
-   len = node->normal_lengths = (GLfloat *) MALLOC( count * sizeof(GLfloat) );
-   if (!len)
-      return;
-
-   /* Find the normal of the first vertex:
-    */
-   for (i = 0 ; i < _TNL_ATTRIB_NORMAL ; i++) 
-      n += node->attrsz[i];
-
-   for (i = 0 ; i < count ; i++, n += stride) {
-      len[i] = LEN_3FV( n );
-      if (len[i] > 0.0F) len[i] = 1.0F / len[i];
-   } 
-}
-
-static struct tnl_vertex_store *alloc_vertex_store( GLcontext *ctx )
-{
-   struct tnl_vertex_store *store = MALLOC_STRUCT(tnl_vertex_store);
-   (void) ctx;
-   store->used = 0;
-   store->refcount = 1;
-   return store;
-}
-
-static struct tnl_primitive_store *alloc_prim_store( GLcontext *ctx )
-{
-   struct tnl_primitive_store *store = MALLOC_STRUCT(tnl_primitive_store);
-   (void) ctx;
-   store->used = 0;
-   store->refcount = 1;
-   return store;
-}
-
-static void _save_reset_counters( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   tnl->save.prim = tnl->save.prim_store->buffer + tnl->save.prim_store->used;
-   tnl->save.buffer = (tnl->save.vertex_store->buffer + 
-		       tnl->save.vertex_store->used);
-
-   if (tnl->save.vertex_size)
-      tnl->save.initial_counter = ((SAVE_BUFFER_SIZE - 
-				    tnl->save.vertex_store->used) / 
-				   tnl->save.vertex_size);
-   else
-      tnl->save.initial_counter = 0;
-
-   if (tnl->save.initial_counter > ctx->Const.MaxArrayLockSize )
-      tnl->save.initial_counter = ctx->Const.MaxArrayLockSize;
-
-   tnl->save.counter = tnl->save.initial_counter;
-   tnl->save.prim_count = 0;
-   tnl->save.prim_max = SAVE_PRIM_SIZE - tnl->save.prim_store->used;
-   tnl->save.copied.nr = 0;
-   tnl->save.dangling_attr_ref = 0;
-}
-
-
-/* Insert the active immediate struct onto the display list currently
- * being built.
- */
-static void _save_compile_vertex_list( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_vertex_list *node;
-
-   /* Allocate space for this structure in the display list currently
-    * being compiled.
-    */
-   node = (struct tnl_vertex_list *)
-      _mesa_alloc_instruction(ctx, tnl->save.opcode_vertex_list, sizeof(*node));
-
-   if (!node)
-      return;
-
-   /* Duplicate our template, increment refcounts to the storage structs:
-    */
-   _mesa_memcpy(node->attrsz, tnl->save.attrsz, sizeof(node->attrsz));
-   node->vertex_size = tnl->save.vertex_size;
-   node->buffer = tnl->save.buffer;
-   node->count = tnl->save.initial_counter - tnl->save.counter;
-   node->wrap_count = tnl->save.copied.nr;
-   node->have_materials = tnl->save.have_materials;
-   node->dangling_attr_ref = tnl->save.dangling_attr_ref;
-   node->normal_lengths = NULL;
-   node->prim = tnl->save.prim;
-   node->prim_count = tnl->save.prim_count;
-   node->vertex_store = tnl->save.vertex_store;
-   node->prim_store = tnl->save.prim_store;
-
-   node->vertex_store->refcount++;
-   node->prim_store->refcount++;
-
-   assert(node->attrsz[_TNL_ATTRIB_POS] != 0 ||
-	  node->count == 0);
-
-   if (tnl->save.dangling_attr_ref)
-      ctx->ListState.CurrentList->flags |= MESA_DLIST_DANGLING_REFS;
-
-   /* Maybe calculate normal lengths:
-    */
-   if (tnl->CalcDListNormalLengths && 
-       node->attrsz[_TNL_ATTRIB_NORMAL] == 3 &&
-       !(ctx->ListState.CurrentList->flags & MESA_DLIST_DANGLING_REFS))
-      build_normal_lengths( node );
-
-
-   tnl->save.vertex_store->used += tnl->save.vertex_size * node->count;
-   tnl->save.prim_store->used += node->prim_count;
-
-   /* Decide whether the storage structs are full, or can be used for
-    * the next vertex lists as well.
-    */
-   if (tnl->save.vertex_store->used > 
-       SAVE_BUFFER_SIZE - 16 * (tnl->save.vertex_size + 4)) {
-
-      tnl->save.vertex_store->refcount--; 
-      assert(tnl->save.vertex_store->refcount != 0);
-      tnl->save.vertex_store = alloc_vertex_store( ctx );
-      tnl->save.vbptr = tnl->save.vertex_store->buffer;
-   } 
-
-   if (tnl->save.prim_store->used > SAVE_PRIM_SIZE - 6) {
-      tnl->save.prim_store->refcount--; 
-      assert(tnl->save.prim_store->refcount != 0);
-      tnl->save.prim_store = alloc_prim_store( ctx );
-   } 
-
-   /* Reset our structures for the next run of vertices:
-    */
-   _save_reset_counters( ctx );
-
-   /* Copy duplicated vertices 
-    */
-   tnl->save.copied.nr = _save_copy_vertices( ctx, node );
-
-
-   /* Deal with GL_COMPILE_AND_EXECUTE:
-    */
-   if (ctx->ExecuteFlag) {
-      _tnl_playback_vertex_list( ctx, (void *) node );
-   }
-}
-
-
-/* TODO -- If no new vertices have been stored, don't bother saving
- * it.
- */
-static void _save_wrap_buffers( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLint i = tnl->save.prim_count - 1;
-   GLenum mode;
-
-   assert(i < (GLint) tnl->save.prim_max);
-   assert(i >= 0);
-
-   /* Close off in-progress primitive.
-    */
-   tnl->save.prim[i].count = ((tnl->save.initial_counter - tnl->save.counter) - 
-			     tnl->save.prim[i].start);
-   mode = tnl->save.prim[i].mode & ~(PRIM_BEGIN|PRIM_END);
-   
-   /* store the copied vertices, and allocate a new list.
-    */
-   _save_compile_vertex_list( ctx );
-
-   /* Restart interrupted primitive
-    */
-   tnl->save.prim[0].mode = mode;
-   tnl->save.prim[0].start = 0;
-   tnl->save.prim[0].count = 0;
-   tnl->save.prim_count = 1;
-}
-
-
-
-/* Called only when buffers are wrapped as the result of filling the
- * vertex_store struct.  
- */
-static void _save_wrap_filled_vertex( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat *data = tnl->save.copied.buffer;
-   GLuint i;
-
-   /* Emit a glEnd to close off the last vertex list.
-    */
-   _save_wrap_buffers( ctx );
-   
-    /* Copy stored stored vertices to start of new list.
-    */
-   assert(tnl->save.counter > tnl->save.copied.nr);
-
-   for (i = 0 ; i < tnl->save.copied.nr ; i++) {
-      _mesa_memcpy( tnl->save.vbptr, data, tnl->save.vertex_size * sizeof(GLfloat));
-      data += tnl->save.vertex_size;
-      tnl->save.vbptr += tnl->save.vertex_size;
-      tnl->save.counter--;
-   }
-}
-
-
-static void _save_copy_to_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   /* XXX Use _TNL_FIRST_* and _TNL_LAST_* values instead? */
-   for (i = _TNL_ATTRIB_POS+1 ; i <= _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (tnl->save.attrsz[i]) {
-	 tnl->save.currentsz[i][0] = tnl->save.attrsz[i];
-	 COPY_CLEAN_4V(tnl->save.current[i], 
-		    tnl->save.attrsz[i], 
-		    tnl->save.attrptr[i]);
-      }
-   }
-
-   /* Edgeflag requires special treatment: 
-    *
-    * TODO: change edgeflag to GLfloat in Mesa.
-    */
-   if (tnl->save.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->ListState.ActiveEdgeFlag = 1;
-      tnl->save.CurrentFloatEdgeFlag = 
-	 tnl->save.attrptr[_TNL_ATTRIB_EDGEFLAG][0];
-      ctx->ListState.CurrentEdgeFlag = 
-	 (tnl->save.CurrentFloatEdgeFlag == 1.0);
-   }
-}
-
-
-static void _save_copy_from_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLint i;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i <= _TNL_ATTRIB_EDGEFLAG ; i++) 
-      switch (tnl->save.attrsz[i]) {
-      case 4: tnl->save.attrptr[i][3] = tnl->save.current[i][3];
-      case 3: tnl->save.attrptr[i][2] = tnl->save.current[i][2];
-      case 2: tnl->save.attrptr[i][1] = tnl->save.current[i][1];
-      case 1: tnl->save.attrptr[i][0] = tnl->save.current[i][0];
-      case 0: break;
-      }
-
-   /* Edgeflag requires special treatment:
-    */
-   if (tnl->save.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      tnl->save.CurrentFloatEdgeFlag = (GLfloat)ctx->ListState.CurrentEdgeFlag;
-      tnl->save.attrptr[_TNL_ATTRIB_EDGEFLAG][0] = tnl->save.CurrentFloatEdgeFlag;
-   }
-}
-
-
-
-
-/* Flush existing data, set new attrib size, replay copied vertices.
- */ 
-static void _save_upgrade_vertex( GLcontext *ctx, 
-				 GLuint attr,
-				 GLuint newsz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint oldsz;
-   GLuint i;
-   GLfloat *tmp;
-
-   /* Store the current run of vertices, and emit a GL_END.  Emit a
-    * BEGIN in the new buffer.
-    */
-   if (tnl->save.initial_counter != tnl->save.counter) 
-      _save_wrap_buffers( ctx );
-   else
-      assert( tnl->save.copied.nr == 0 );
-
-   /* Do a COPY_TO_CURRENT to ensure back-copying works for the case
-    * when the attribute already exists in the vertex and is having
-    * its size increased.  
-    */
-   _save_copy_to_current( ctx );
-
-   /* Fix up sizes:
-    */
-   oldsz = tnl->save.attrsz[attr];
-   tnl->save.attrsz[attr] = newsz;
-
-   tnl->save.vertex_size += newsz - oldsz;
-   tnl->save.counter = ((SAVE_BUFFER_SIZE - tnl->save.vertex_store->used) / 
-			tnl->save.vertex_size);
-   if (tnl->save.counter > ctx->Const.MaxArrayLockSize )
-      tnl->save.counter = ctx->Const.MaxArrayLockSize;
-   tnl->save.initial_counter = tnl->save.counter;
-
-   /* Recalculate all the attrptr[] values:
-    */
-   for (i = 0, tmp = tnl->save.vertex ; i < _TNL_ATTRIB_MAX ; i++) {
-      if (tnl->save.attrsz[i]) {
-	 tnl->save.attrptr[i] = tmp;
-	 tmp += tnl->save.attrsz[i];
-      }
-      else 
-	 tnl->save.attrptr[i] = NULL; /* will not be dereferenced. */
-   }
-
-   /* Copy from current to repopulate the vertex with correct values.
-    */
-   _save_copy_from_current( ctx );
-
-   /* Replay stored vertices to translate them to new format here.
-    *
-    * If there are copied vertices and the new (upgraded) attribute
-    * has not been defined before, this list is somewhat degenerate,
-    * and will need fixup at runtime.
-    */
-   if (tnl->save.copied.nr)
-   {
-      GLfloat *data = tnl->save.copied.buffer;
-      GLfloat *dest = tnl->save.buffer;
-      GLuint j;
-
-      /* Need to note this and fix up at runtime (or loopback):
-       */
-      if (tnl->save.currentsz[attr][0] == 0) {
-	 assert(oldsz == 0);
-	 tnl->save.dangling_attr_ref = GL_TRUE;
-
-/* 	 _mesa_debug(NULL, "_save_upgrade_vertex: dangling reference attr %d\n",  */
-/* 		     attr);  */
-
-#if 0
-	 /* The current strategy is to punt these degenerate cases
-	  * through _tnl_loopback_vertex_list(), a lower-performance
-	  * option.  To minimize the impact of this, artificially
-	  * reduce the size of this vertex_list.
-	  */
-	 if (t->save.counter > 10) {
-	    t->save.initial_counter = 10;
-	    t->save.counter = 10;
-	 }
-#endif
-      }
-
-      for (i = 0 ; i < tnl->save.copied.nr ; i++) {
-	 for (j = 0 ; j < _TNL_ATTRIB_MAX ; j++) {
-	    if (tnl->save.attrsz[j]) {
-	       if (j == attr) {
-		  if (oldsz) {
-		     COPY_CLEAN_4V( dest, oldsz, data );
-		     data += oldsz;
-		     dest += newsz;
-		  }
-		  else {
-		     COPY_SZ_4V( dest, newsz, tnl->save.current[attr] );
-		     dest += newsz;
-		  }
-	       }
-	       else {
-		  GLint sz = tnl->save.attrsz[j];
-		  COPY_SZ_4V( dest, sz, data );
-		  data += sz;
-		  dest += sz;
-	       }
-	    }
-	 }
-      }
-
-      tnl->save.vbptr = dest;
-      tnl->save.counter -= tnl->save.copied.nr;
-   }
-}
-
-
-
-
-/* Helper function for 'CHOOSE' macro.  Do what's necessary when an
- * entrypoint is called for the first time.
- */
-static void do_choose( GLuint attr, GLuint sz, 
-		       void (*attr_func)( const GLfloat *),
-		       void (*choose1)( const GLfloat *),
-		       void (*choose2)( const GLfloat *),
-		       void (*choose3)( const GLfloat *),
-		       void (*choose4)( const GLfloat *),
-		       const GLfloat *v )
-{ 
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   static GLfloat id[4] = { 0, 0, 0, 1 };
-   int i;
-
-   if (tnl->save.attrsz[attr] < sz) {
-      /* New size is larger.  Need to flush existing vertices and get
-       * an enlarged vertex format.
-       */
-      _save_upgrade_vertex( ctx, attr, sz );
-   }
-   else {
-      /* New size is equal or smaller - just need to fill in some
-       * zeros.
-       */
-      for (i = sz ; i <= tnl->save.attrsz[attr] ; i++)
-	 tnl->save.attrptr[attr][i-1] = id[i-1];
-   }
-
-   /* Reset any active pointers for this attribute 
-    */
-   tnl->save.tabfv[attr][0] = choose1;
-   tnl->save.tabfv[attr][1] = choose2;
-   tnl->save.tabfv[attr][2] = choose3;
-   tnl->save.tabfv[attr][3] = choose4;
-
-   /* Update the secondary dispatch table with the new function
-    */
-   tnl->save.tabfv[attr][sz-1] = attr_func;
-
-   (*attr_func)(v);
-}
-
-
-
-/* Only one size for each attribute may be active at once.  Eg. if
- * Color3f is installed/active, then Color4f may not be, even if the
- * vertex actually contains 4 color coordinates.  This is because the
- * 3f version won't otherwise set color[3] to 1.0 -- this is the job
- * of the chooser function when switching between Color4f and Color3f.
- */
-#define ATTRFV( ATTR, N )					\
-static void save_choose_##ATTR##_##N( const GLfloat *v );	\
-								\
-static void save_attrib_##ATTR##_##N( const GLfloat *v )	\
-{								\
-   GET_CURRENT_CONTEXT( ctx );					\
-   TNLcontext *tnl = TNL_CONTEXT(ctx);				\
-								\
-   if ((ATTR) == 0) {						\
-      GLuint i;							\
-								\
-      if (N>0) tnl->save.vbptr[0] = v[0];			\
-      if (N>1) tnl->save.vbptr[1] = v[1];			\
-      if (N>2) tnl->save.vbptr[2] = v[2];			\
-      if (N>3) tnl->save.vbptr[3] = v[3];			\
-								\
-      for (i = N; i < tnl->save.vertex_size; i++)		\
-	 tnl->save.vbptr[i] = tnl->save.vertex[i];		\
-								\
-      tnl->save.vbptr += tnl->save.vertex_size;			\
-								\
-      if (--tnl->save.counter == 0)				\
-	 _save_wrap_filled_vertex( ctx );			\
-   }								\
-   else {							\
-      GLfloat *dest = tnl->save.attrptr[ATTR];			\
-      if (N>0) dest[0] = v[0];					\
-      if (N>1) dest[1] = v[1];					\
-      if (N>2) dest[2] = v[2];					\
-      if (N>3) dest[3] = v[3];					\
-   }								\
-}
-
-#define CHOOSE( ATTR, N )					\
-static void save_choose_##ATTR##_##N( const GLfloat *v )	\
-{								\
-   do_choose(ATTR, N,						\
-	     save_attrib_##ATTR##_##N,				\
-	     save_choose_##ATTR##_1,				\
-	     save_choose_##ATTR##_2,				\
-	     save_choose_##ATTR##_3,				\
-	     save_choose_##ATTR##_4,				\
-	     v );						\
-}
-
-#define INIT(ATTR)					\
-static void save_init_##ATTR( TNLcontext *tnl )		\
-{							\
-   tnl->save.tabfv[ATTR][0] = save_choose_##ATTR##_1;	\
-   tnl->save.tabfv[ATTR][1] = save_choose_##ATTR##_2;	\
-   tnl->save.tabfv[ATTR][2] = save_choose_##ATTR##_3;	\
-   tnl->save.tabfv[ATTR][3] = save_choose_##ATTR##_4;	\
-}
-   
-#define ATTRS( ATTRIB )				\
-   ATTRFV( ATTRIB, 1 )				\
-   ATTRFV( ATTRIB, 2 )				\
-   ATTRFV( ATTRIB, 3 )				\
-   ATTRFV( ATTRIB, 4 )				\
-   CHOOSE( ATTRIB, 1 )				\
-   CHOOSE( ATTRIB, 2 )				\
-   CHOOSE( ATTRIB, 3 )				\
-   CHOOSE( ATTRIB, 4 )				\
-   INIT( ATTRIB )				\
-
-
-/* Generate a lot of functions.  These are the actual worker
- * functions, which are equivalent to those generated via codegen
- * elsewhere.
- */
-ATTRS( 0 )
-ATTRS( 1 )
-ATTRS( 2 )
-ATTRS( 3 )
-ATTRS( 4 )
-ATTRS( 5 )
-ATTRS( 6 )
-ATTRS( 7 )
-ATTRS( 8 )
-ATTRS( 9 )
-ATTRS( 10 )
-ATTRS( 11 )
-ATTRS( 12 )
-ATTRS( 13 )
-ATTRS( 14 )
-ATTRS( 15 )
-
-
-static void _save_reset_vertex( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint i;
-
-   save_init_0( tnl );
-   save_init_1( tnl );
-   save_init_2( tnl );
-   save_init_3( tnl );
-   save_init_4( tnl );
-   save_init_5( tnl );
-   save_init_6( tnl );
-   save_init_7( tnl );
-   save_init_8( tnl );
-   save_init_9( tnl );
-   save_init_10( tnl );
-   save_init_11( tnl );
-   save_init_12( tnl );
-   save_init_13( tnl );
-   save_init_14( tnl );
-   save_init_15( tnl );
-      
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++)
-      tnl->save.attrsz[i] = 0;
-      
-   tnl->save.vertex_size = 0;
-   tnl->save.have_materials = 0;
-
-   _save_reset_counters( ctx ); 
-}
-
-
-
-/* Cope with aliasing of classic Vertex, Normal, etc. and the fan-out
- * of glMultTexCoord and glProgramParamterNV by routing all these
- * through a second level dispatch table.
- */
-#define DISPATCH_ATTRFV( ATTR, COUNT, P )	\
-do {						\
-   GET_CURRENT_CONTEXT( ctx ); 			\
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 		\
-   tnl->save.tabfv[ATTR][COUNT-1]( P );		\
-} while (0)
-
-#define DISPATCH_ATTR1FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 1, V )
-#define DISPATCH_ATTR2FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 2, V )
-#define DISPATCH_ATTR3FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 3, V )
-#define DISPATCH_ATTR4FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 4, V )
-
-#define DISPATCH_ATTR1F( ATTR, S ) DISPATCH_ATTRFV( ATTR, 1, &(S) )
-
-#if defined(USE_X86_ASM) && 0 /* will break register calling convention */
-/* Naughty cheat:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) DISPATCH_ATTRFV( ATTR, 2, &(S) )
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) DISPATCH_ATTRFV( ATTR, 3, &(S) )
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q ) DISPATCH_ATTRFV( ATTR, 4, &(S) )
-#else
-/* Safe:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) 		\
-do { 						\
-   GLfloat v[2]; 				\
-   v[0] = S; v[1] = T;				\
-   DISPATCH_ATTR2FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) 		\
-do { 						\
-   GLfloat v[3]; 				\
-   v[0] = S; v[1] = T; v[2] = R;		\
-   DISPATCH_ATTR3FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q )	\
-do { 						\
-   GLfloat v[4]; 				\
-   v[0] = S; v[1] = T; v[2] = R; v[3] = Q;	\
-   DISPATCH_ATTR4FV( ATTR, v );			\
-} while (0)
-#endif
-
-
-static void enum_error( void )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   _mesa_compile_error( ctx, GL_INVALID_ENUM, "glVertexAttrib" );
-}
-
-static void GLAPIENTRY _save_Vertex2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_POS, x, y );
-}
-
-static void GLAPIENTRY _save_Vertex2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _save_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_POS, x, y, z );
-}
-
-static void GLAPIENTRY _save_Vertex3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _save_Vertex4f( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_POS, x, y, z, w );
-}
-
-static void GLAPIENTRY _save_Vertex4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _save_TexCoord1f( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_TEX0, x );
-}
-
-static void GLAPIENTRY _save_TexCoord1fv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _save_TexCoord2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_TEX0, x, y );
-}
-
-static void GLAPIENTRY _save_TexCoord2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _save_TexCoord3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_TEX0, x, y, z );
-}
-
-static void GLAPIENTRY _save_TexCoord3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _save_TexCoord4f( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_TEX0, x, y, z, w );
-}
-
-static void GLAPIENTRY _save_TexCoord4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _save_Normal3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_NORMAL, x, y, z );
-}
-
-static void GLAPIENTRY _save_Normal3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_NORMAL, v );
-}
-
-static void GLAPIENTRY _save_FogCoordfEXT( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_FOG, x );
-}
-
-static void GLAPIENTRY _save_FogCoordfvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_FOG, v );
-}
-
-static void GLAPIENTRY _save_Color3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR0, x, y, z );
-}
-
-static void GLAPIENTRY _save_Color3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _save_Color4f( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_COLOR0, x, y, z, w );
-}
-
-static void GLAPIENTRY _save_Color4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _save_SecondaryColor3fEXT( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR1, x, y, z );
-}
-
-static void GLAPIENTRY _save_SecondaryColor3fvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR1, v );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord1f( GLenum target, GLfloat x  )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1F( attr, x );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord1fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1FV( attr, v );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord2f( GLenum target, GLfloat x, GLfloat y )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2F( attr, x, y );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord2fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2FV( attr, v );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord3f( GLenum target, GLfloat x, GLfloat y,
-				    GLfloat z)
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3F( attr, x, y, z );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord3fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3FV( attr, v );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord4f( GLenum target, GLfloat x, GLfloat y,
-				    GLfloat z, GLfloat w )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4F( attr, x, y, z, w );
-}
-
-static void GLAPIENTRY _save_MultiTexCoord4fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4FV( attr, v );
-}
-
-static void GLAPIENTRY _save_VertexAttrib1fNV( GLuint index, GLfloat x )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR1F( index, x );
-   else
-      enum_error(); 
-}
-
-static void GLAPIENTRY _save_VertexAttrib1fvNV( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR1FV( index, v );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY _save_VertexAttrib2fNV( GLuint index, GLfloat x, GLfloat y )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR2F( index, x, y );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY _save_VertexAttrib2fvNV( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR2FV( index, v );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY _save_VertexAttrib3fNV( GLuint index, GLfloat x, GLfloat y, 
-				  GLfloat z )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR3F( index, x, y, z );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY _save_VertexAttrib3fvNV( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR3FV( index, v );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY _save_VertexAttrib4fNV( GLuint index, GLfloat x, GLfloat y,
-				  GLfloat z, GLfloat w )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR4F( index, x, y, z, w );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY _save_VertexAttrib4fvNV( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_PROGRAM_ATTRIBS)
-      DISPATCH_ATTR4FV( index, v );
-   else
-      enum_error();
-}
-
-
-static void GLAPIENTRY
-_save_VertexAttrib1fARB( GLuint index, GLfloat x )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR1F( index, x );
-   else
-      enum_error(); 
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib1fvARB( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR1FV( index, v );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib2fARB( GLuint index, GLfloat x, GLfloat y )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR2F( index, x, y );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib2fvARB( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR2FV( index, v );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib3fARB( GLuint index, GLfloat x, GLfloat y, GLfloat z )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR3F( index, x, y, z );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib3fvARB( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR3FV( index, v );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib4fARB( GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR4F( index, x, y, z, w );
-   else
-      enum_error();
-}
-
-static void GLAPIENTRY
-_save_VertexAttrib4fvARB( GLuint index, const GLfloat *v )
-{
-   if (index < MAX_VERTEX_ATTRIBS)
-      DISPATCH_ATTR4FV( index, v );
-   else
-      enum_error();
-}
-
-
-/* Materials:  
- * 
- * These are treated as per-vertex attributes, at indices above where
- * the NV_vertex_program leaves off.  There are a lot of good things
- * about treating materials this way.  
- *
- * However: I don't want to double the number of generated functions
- * just to cope with this, so I unroll the 'C' varients of CHOOSE and
- * ATTRF into this function, and dispense with codegen and
- * second-level dispatch.
- *
- * There is no aliasing of material attributes with other entrypoints.
- */
-#define MAT_ATTR( A, N, params )			\
-do {							\
-   if (tnl->save.attrsz[A] < N) {			\
-      _save_upgrade_vertex( ctx, A, N );		\
-      tnl->save.have_materials = GL_TRUE;               \
-   }							\
-							\
-   {							\
-      GLfloat *dest = tnl->save.attrptr[A];	      	\
-      if (N>0) dest[0] = params[0];			\
-      if (N>1) dest[1] = params[1];			\
-      if (N>2) dest[2] = params[2];			\
-      if (N>3) dest[3] = params[3];			\
-   }							\
-} while (0)
-
-
-#define MAT( ATTR, N, face, params )			\
-do {							\
-   if (face != GL_BACK)					\
-      MAT_ATTR( ATTR, N, params ); /* front */		\
-   if (face != GL_FRONT)				\
-      MAT_ATTR( ATTR + 1, N, params ); /* back */	\
-} while (0)
-
-
-/* NOTE: Have to remove/deal-with colormaterial crossovers, probably
- * later on - in the meantime just store everything.  
- */
-static void GLAPIENTRY _save_Materialfv( GLenum face, GLenum pname, 
-			       const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (pname) {
-   case GL_EMISSION:
-      MAT( _TNL_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
-      break;
-   case GL_AMBIENT:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      break;
-   case GL_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   case GL_SPECULAR:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
-      break;
-   case GL_SHININESS:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
-      break;
-   case GL_COLOR_INDEXES:
-      MAT( _TNL_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
-      break;
-   case GL_AMBIENT_AND_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   default:
-      _mesa_compile_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-}
-
-
-#define IDX_ATTR( A, IDX )				\
-do {							\
-   GET_CURRENT_CONTEXT( ctx );				\
-   TNLcontext *tnl = TNL_CONTEXT(ctx);			\
-							\
-   if (tnl->save.attrsz[A] < 1) {			\
-      _save_upgrade_vertex( ctx, A, 1 );		\
-   }							\
-							\
-   {							\
-      GLfloat *dest = tnl->save.attrptr[A];		\
-      dest[0] = IDX;				\
-   }							\
-} while (0)
-
-
-static void GLAPIENTRY _save_EdgeFlag( GLboolean b )
-{
-   IDX_ATTR( _TNL_ATTRIB_EDGEFLAG, (GLfloat)b );
-}
-
-
-static void GLAPIENTRY _save_Indexf( GLfloat f )
-{
-   IDX_ATTR( _TNL_ATTRIB_COLOR_INDEX, f );
-}
-
-static void GLAPIENTRY _save_Indexfv( const GLfloat *f )
-{
-   IDX_ATTR( _TNL_ATTRIB_COLOR_INDEX, f[0] );
-}
-
-
-
-
-/* Cope with EvalCoord/CallList called within a begin/end object:
- *     -- Flush current buffer
- *     -- Fallback to opcodes for the rest of the begin/end object.
- */
-#define FALLBACK(ctx) 							\
-do {									\
-   TNLcontext *tnl = TNL_CONTEXT(ctx);					\
-									\
-   if (tnl->save.initial_counter != tnl->save.counter ||		\
-       tnl->save.prim_count) 						\
-      _save_compile_vertex_list( ctx );					\
-									\
-   _save_copy_to_current( ctx );					\
-   _save_reset_vertex( ctx );						\
-   _mesa_install_save_vtxfmt( ctx, &ctx->ListState.ListVtxfmt );	\
-   ctx->Driver.SaveNeedFlush = 0;					\
-} while (0)
-
-static void GLAPIENTRY _save_EvalCoord1f( GLfloat u )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_EvalCoord1f(ctx->Save, ( u ));
-}
-
-static void GLAPIENTRY _save_EvalCoord1fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_EvalCoord1fv(ctx->Save, ( v ));
-}
-
-static void GLAPIENTRY _save_EvalCoord2f( GLfloat u, GLfloat v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_EvalCoord2f(ctx->Save, ( u, v ));
-}
-
-static void GLAPIENTRY _save_EvalCoord2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_EvalCoord2fv(ctx->Save, ( v ));
-}
-
-static void GLAPIENTRY _save_EvalPoint1( GLint i )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_EvalPoint1(ctx->Save, ( i ));
-}
-
-static void GLAPIENTRY _save_EvalPoint2( GLint i, GLint j )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_EvalPoint2(ctx->Save, ( i, j ));
-}
-
-static void GLAPIENTRY _save_CallList( GLuint l )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_CallList(ctx->Save, ( l ));
-}
-
-static void GLAPIENTRY _save_CallLists( GLsizei n, GLenum type, const GLvoid *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   FALLBACK(ctx);
-   CALL_CallLists(ctx->Save, ( n, type, v ));
-}
-
-
-
-
-/**
- * Called via ctx->Driver.NotifySaveBegin(ctx, mode) when we get a
- * glBegin() call while compiling a display list.
- * See save_Begin() in dlist.c
- *
- * This plugs in our special TNL-related display list functions.
- * All subsequent glBegin/glVertex/glEnd()s found while compiling a
- * display list will get routed to the functions in this file.
- *
- * Updating of ctx->Driver.CurrentSavePrimitive is already taken care of.
- */
-static GLboolean _save_NotifyBegin( GLcontext *ctx, GLenum mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-
-   if (1) {
-      GLuint i = tnl->save.prim_count++;
-
-      assert(i < tnl->save.prim_max);
-      tnl->save.prim[i].mode = mode | PRIM_BEGIN;
-      tnl->save.prim[i].start = tnl->save.initial_counter - tnl->save.counter;
-      tnl->save.prim[i].count = 0;   
-
-      _mesa_install_save_vtxfmt( ctx, &tnl->save_vtxfmt );      
-      ctx->Driver.SaveNeedFlush = 1;
-      return GL_TRUE;
-   }
-   else 
-      return GL_FALSE;
-}
-
-
-
-static void GLAPIENTRY _save_End( void )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLint i = tnl->save.prim_count - 1;
-
-   ctx->Driver.CurrentSavePrimitive = PRIM_OUTSIDE_BEGIN_END;
-   if (ctx->ExecuteFlag)
-      ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-
-   tnl->save.prim[i].mode |= PRIM_END;
-   tnl->save.prim[i].count = ((tnl->save.initial_counter - tnl->save.counter) - 
-			      tnl->save.prim[i].start);
-
-   if (i == (GLint) tnl->save.prim_max - 1) {
-      _save_compile_vertex_list( ctx );
-      assert(tnl->save.copied.nr == 0);
-   }
-
-   /* Swap out this vertex format while outside begin/end.  Any color,
-    * etc. received between here and the next begin will be compiled
-    * as opcodes.
-    */   
-   _mesa_install_save_vtxfmt( ctx, &ctx->ListState.ListVtxfmt );
-}
-
-
-/* These are all errors as this vtxfmt is only installed inside
- * begin/end pairs.
- */
-static void GLAPIENTRY _save_DrawElements(GLenum mode, GLsizei count, GLenum type,
-			       const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   (void) mode; (void) count; (void) type; (void) indices;
-   _mesa_compile_error( ctx, GL_INVALID_OPERATION, "glDrawElements" );
-}
-
-
-static void GLAPIENTRY _save_DrawRangeElements(GLenum mode,
-				    GLuint start, GLuint end,
-				    GLsizei count, GLenum type,
-				    const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   (void) mode; (void) start; (void) end; (void) count; (void) type; (void) indices;
-   _mesa_compile_error( ctx, GL_INVALID_OPERATION, "glDrawRangeElements" );
-}
-
-static void GLAPIENTRY _save_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   (void) mode; (void) start; (void) count;
-   _mesa_compile_error( ctx, GL_INVALID_OPERATION, "glDrawArrays" );
-}
-
-static void GLAPIENTRY _save_Rectf( GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   (void) x1; (void) y1; (void) x2; (void) y2;
-   _mesa_compile_error( ctx, GL_INVALID_OPERATION, "glRectf" );
-}
-
-static void GLAPIENTRY _save_EvalMesh1( GLenum mode, GLint i1, GLint i2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   (void) mode; (void) i1; (void) i2;
-   _mesa_compile_error( ctx, GL_INVALID_OPERATION, "glEvalMesh1" );
-}
-
-static void GLAPIENTRY _save_EvalMesh2( GLenum mode, GLint i1, GLint i2,
-				  GLint j1, GLint j2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   (void) mode; (void) i1; (void) i2; (void) j1; (void) j2;
-   _mesa_compile_error( ctx, GL_INVALID_OPERATION, "glEvalMesh2" );
-}
-
-/**
- * This is only called if someone tries to compile nested glBegin()s 
- * in their display list.
- */
-static void GLAPIENTRY _save_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   (void) mode;
-   _mesa_compile_error(ctx, GL_INVALID_OPERATION,
-                       "glBegin(called inside glBegin/End)");
-}
-
-
-/* Unlike the functions above, these are to be hooked into the vtxfmt
- * maintained in ctx->ListState, active when the list is known or
- * suspected to be outside any begin/end primitive.
- */
-static void GLAPIENTRY _save_OBE_Rectf( GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   _save_NotifyBegin( ctx, GL_QUADS | PRIM_WEAK );
-   CALL_Vertex2f(GET_DISPATCH(), ( x1, y1 ));
-   CALL_Vertex2f(GET_DISPATCH(), ( x2, y1 ));
-   CALL_Vertex2f(GET_DISPATCH(), ( x2, y2 ));
-   CALL_Vertex2f(GET_DISPATCH(), ( x1, y2 ));
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-static void GLAPIENTRY _save_OBE_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLint i;
-
-   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-      return;
-
-   _ae_map_vbos( ctx );
-
-   _save_NotifyBegin( ctx, mode | PRIM_WEAK );
-   for (i = 0; i < count; i++)
-       CALL_ArrayElement(GET_DISPATCH(), (start + i));
-   CALL_End(GET_DISPATCH(), ());
-
-   _ae_unmap_vbos( ctx );
-}
-
-
-static void GLAPIENTRY _save_OBE_DrawElements(GLenum mode, GLsizei count, GLenum type,
-				   const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLint i;
-
-   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-      return;
-
-   _ae_map_vbos( ctx );
-
-   _save_NotifyBegin( ctx, mode | PRIM_WEAK );
-
-   switch (type) {
-   case GL_UNSIGNED_BYTE:
-      for (i = 0 ; i < count ; i++)
-	  CALL_ArrayElement(GET_DISPATCH(), ( ((GLubyte *)indices)[i] ));
-      break;
-   case GL_UNSIGNED_SHORT:
-      for (i = 0 ; i < count ; i++)
-	  CALL_ArrayElement(GET_DISPATCH(), ( ((GLushort *)indices)[i] ));
-      break;
-   case GL_UNSIGNED_INT:
-      for (i = 0 ; i < count ; i++)
-	  CALL_ArrayElement(GET_DISPATCH(), ( ((GLuint *)indices)[i] ));
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glDrawElements(type)" );
-      break;
-   }
-
-   CALL_End(GET_DISPATCH(), ());
-
-   _ae_unmap_vbos( ctx );
-}
-
-static void GLAPIENTRY _save_OBE_DrawRangeElements(GLenum mode,
-					GLuint start, GLuint end,
-					GLsizei count, GLenum type,
-					const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   if (_mesa_validate_DrawRangeElements( ctx, mode,
-					 start, end,
-					 count, type, indices ))
-      _save_OBE_DrawElements( mode, count, type, indices );
-}
-
-
-
-
-
-static void _save_vtxfmt_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLvertexformat *vfmt = &tnl->save_vtxfmt;
-
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = _save_Begin;
-   vfmt->Color3f = _save_Color3f;
-   vfmt->Color3fv = _save_Color3fv;
-   vfmt->Color4f = _save_Color4f;
-   vfmt->Color4fv = _save_Color4fv;
-   vfmt->EdgeFlag = _save_EdgeFlag;
-   vfmt->End = _save_End;
-   vfmt->FogCoordfEXT = _save_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = _save_FogCoordfvEXT;
-   vfmt->Indexf = _save_Indexf;
-   vfmt->Indexfv = _save_Indexfv;
-   vfmt->Materialfv = _save_Materialfv;
-   vfmt->MultiTexCoord1fARB = _save_MultiTexCoord1f;
-   vfmt->MultiTexCoord1fvARB = _save_MultiTexCoord1fv;
-   vfmt->MultiTexCoord2fARB = _save_MultiTexCoord2f;
-   vfmt->MultiTexCoord2fvARB = _save_MultiTexCoord2fv;
-   vfmt->MultiTexCoord3fARB = _save_MultiTexCoord3f;
-   vfmt->MultiTexCoord3fvARB = _save_MultiTexCoord3fv;
-   vfmt->MultiTexCoord4fARB = _save_MultiTexCoord4f;
-   vfmt->MultiTexCoord4fvARB = _save_MultiTexCoord4fv;
-   vfmt->Normal3f = _save_Normal3f;
-   vfmt->Normal3fv = _save_Normal3fv;
-   vfmt->SecondaryColor3fEXT = _save_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = _save_SecondaryColor3fvEXT;
-   vfmt->TexCoord1f = _save_TexCoord1f;
-   vfmt->TexCoord1fv = _save_TexCoord1fv;
-   vfmt->TexCoord2f = _save_TexCoord2f;
-   vfmt->TexCoord2fv = _save_TexCoord2fv;
-   vfmt->TexCoord3f = _save_TexCoord3f;
-   vfmt->TexCoord3fv = _save_TexCoord3fv;
-   vfmt->TexCoord4f = _save_TexCoord4f;
-   vfmt->TexCoord4fv = _save_TexCoord4fv;
-   vfmt->Vertex2f = _save_Vertex2f;
-   vfmt->Vertex2fv = _save_Vertex2fv;
-   vfmt->Vertex3f = _save_Vertex3f;
-   vfmt->Vertex3fv = _save_Vertex3fv;
-   vfmt->Vertex4f = _save_Vertex4f;
-   vfmt->Vertex4fv = _save_Vertex4fv;
-   vfmt->VertexAttrib1fNV = _save_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = _save_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV = _save_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = _save_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV = _save_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = _save_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV = _save_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = _save_VertexAttrib4fvNV;
-   vfmt->VertexAttrib1fARB = _save_VertexAttrib1fARB;
-   vfmt->VertexAttrib1fvARB = _save_VertexAttrib1fvARB;
-   vfmt->VertexAttrib2fARB = _save_VertexAttrib2fARB;
-   vfmt->VertexAttrib2fvARB = _save_VertexAttrib2fvARB;
-   vfmt->VertexAttrib3fARB = _save_VertexAttrib3fARB;
-   vfmt->VertexAttrib3fvARB = _save_VertexAttrib3fvARB;
-   vfmt->VertexAttrib4fARB = _save_VertexAttrib4fARB;
-   vfmt->VertexAttrib4fvARB = _save_VertexAttrib4fvARB;
-   
-   /* This will all require us to fallback to saving the list as opcodes:
-    */ 
-   vfmt->CallList = _save_CallList; /* inside begin/end */
-   vfmt->CallLists = _save_CallLists; /* inside begin/end */
-   vfmt->EvalCoord1f = _save_EvalCoord1f;
-   vfmt->EvalCoord1fv = _save_EvalCoord1fv;
-   vfmt->EvalCoord2f = _save_EvalCoord2f;
-   vfmt->EvalCoord2fv = _save_EvalCoord2fv;
-   vfmt->EvalPoint1 = _save_EvalPoint1;
-   vfmt->EvalPoint2 = _save_EvalPoint2;
-
-   /* These are all errors as we at least know we are in some sort of
-    * begin/end pair:
-    */
-   vfmt->EvalMesh1 = _save_EvalMesh1;	
-   vfmt->EvalMesh2 = _save_EvalMesh2;
-   vfmt->Begin = _save_Begin;
-   vfmt->Rectf = _save_Rectf;
-   vfmt->DrawArrays = _save_DrawArrays;
-   vfmt->DrawElements = _save_DrawElements;
-   vfmt->DrawRangeElements = _save_DrawRangeElements;
-
-}
-
-
-void _tnl_SaveFlushVertices( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* Noop when we are actually active:
-    */
-   if (ctx->Driver.CurrentSavePrimitive == PRIM_INSIDE_UNKNOWN_PRIM ||
-       ctx->Driver.CurrentSavePrimitive <= GL_POLYGON)
-      return;
-
-   if (tnl->save.initial_counter != tnl->save.counter ||
-       tnl->save.prim_count) 
-      _save_compile_vertex_list( ctx );
-   
-   _save_copy_to_current( ctx );
-   _save_reset_vertex( ctx );
-   ctx->Driver.SaveNeedFlush = 0;
-}
-
-void _tnl_NewList( GLcontext *ctx, GLuint list, GLenum mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   (void) list; (void) mode;
-
-   if (!tnl->save.prim_store)
-      tnl->save.prim_store = alloc_prim_store( ctx );
-
-   if (!tnl->save.vertex_store) {
-      tnl->save.vertex_store = alloc_vertex_store( ctx );
-      tnl->save.vbptr = tnl->save.vertex_store->buffer;
-   }
-   
-   _save_reset_vertex( ctx );
-   ctx->Driver.SaveNeedFlush = 0;
-}
-
-void _tnl_EndList( GLcontext *ctx )
-{
-   (void) ctx;
-   assert(TNL_CONTEXT(ctx)->save.vertex_size == 0);
-}
- 
-void _tnl_BeginCallList( GLcontext *ctx, struct mesa_display_list *dlist )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->save.replay_flags |= dlist->flags;
-   tnl->save.replay_flags |= tnl->LoopbackDListCassettes;
-}
-
-void _tnl_EndCallList( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   
-   if (ctx->ListState.CallDepth == 1)
-      tnl->save.replay_flags = 0;
-}
-
-
-static void _tnl_destroy_vertex_list( GLcontext *ctx, void *data )
-{
-   struct tnl_vertex_list *node = (struct tnl_vertex_list *)data;
-   (void) ctx;
-
-   if ( --node->vertex_store->refcount == 0 )
-      FREE( node->vertex_store );
-
-   if ( --node->prim_store->refcount == 0 )
-      FREE( node->prim_store );
-
-   if ( node->normal_lengths )
-      FREE( node->normal_lengths );
-}
-
-
-static void _tnl_print_vertex_list( GLcontext *ctx, void *data )
-{
-   struct tnl_vertex_list *node = (struct tnl_vertex_list *)data;
-   GLuint i;
-   (void) ctx;
-
-   _mesa_debug(NULL, "TNL-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
-               node->count,
-	       node->prim_count,
-	       node->vertex_size);
-
-   for (i = 0 ; i < node->prim_count ; i++) {
-      struct tnl_prim *prim = &node->prim[i];
-      _mesa_debug(NULL, "   prim %d: %s %d..%d %s %s\n",
-		  i, 
-		  _mesa_lookup_enum_by_nr(prim->mode & PRIM_MODE_MASK),
-		  prim->start, 
-		  prim->start + prim->count,
-		  (prim->mode & PRIM_BEGIN) ? "BEGIN" : "(wrap)",
-		  (prim->mode & PRIM_END) ? "END" : "(wrap)");
-   }
-}
-
-
-static void _save_current_init( GLcontext *ctx ) 
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLint i;
-
-   for (i = 0; i < _TNL_ATTRIB_MAT_FRONT_AMBIENT; i++) {
-      ASSERT(i < VERT_ATTRIB_MAX);
-      tnl->save.currentsz[i] = &ctx->ListState.ActiveAttribSize[i];
-      tnl->save.current[i] = ctx->ListState.CurrentAttrib[i];
-   }
-
-   for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-      const GLuint j = i - _TNL_FIRST_MAT;
-      ASSERT(j < MAT_ATTRIB_MAX);
-      tnl->save.currentsz[i] = &ctx->ListState.ActiveMaterialSize[j];
-      tnl->save.current[i] = ctx->ListState.CurrentMaterial[j];
-   }
-
-   tnl->save.currentsz[_TNL_ATTRIB_EDGEFLAG] = &ctx->ListState.ActiveEdgeFlag;
-   tnl->save.current[_TNL_ATTRIB_EDGEFLAG] = &tnl->save.CurrentFloatEdgeFlag;
-}
-
-/**
- * Initialize the display list compiler
- */
-void _tnl_save_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_vertex_arrays *tmp = &tnl->save_inputs;
-   GLuint i;
-
-
-   for (i = 0; i < _TNL_ATTRIB_MAX; i++)
-      _mesa_vector4f_init( &tmp->Attribs[i], 0, NULL);
-
-   tnl->save.opcode_vertex_list =
-      _mesa_alloc_opcode( ctx,
-			  sizeof(struct tnl_vertex_list),
-			  _tnl_playback_vertex_list,
-			  _tnl_destroy_vertex_list,
-			  _tnl_print_vertex_list );
-
-   ctx->Driver.NotifySaveBegin = _save_NotifyBegin;
-
-   _save_vtxfmt_init( ctx );
-   _save_current_init( ctx );
-
-   /* Hook our array functions into the outside-begin-end vtxfmt in 
-    * ctx->ListState.
-    */
-   ctx->ListState.ListVtxfmt.Rectf = _save_OBE_Rectf;
-   ctx->ListState.ListVtxfmt.DrawArrays = _save_OBE_DrawArrays;
-   ctx->ListState.ListVtxfmt.DrawElements = _save_OBE_DrawElements;
-   ctx->ListState.ListVtxfmt.DrawRangeElements = _save_OBE_DrawRangeElements;
-   _mesa_install_save_vtxfmt( ctx, &ctx->ListState.ListVtxfmt );
-}
-
-
-/**
- * Deallocate the immediate-mode buffer for the given context, if
- * its reference count goes to zero.
- */
-void _tnl_save_destroy( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* Decrement the refcounts.  References may still be held by
-    * display lists yet to be destroyed, so it may not yet be time to
-    * free these items.
-    */
-   if (tnl->save.prim_store &&
-       --tnl->save.prim_store->refcount == 0 )
-      FREE( tnl->save.prim_store );
-
-   if (tnl->save.vertex_store &&
-       --tnl->save.vertex_store->refcount == 0 )
-      FREE( tnl->save.vertex_store );
-}
diff --git a/src/mesa/tnl/t_save_api.h b/src/mesa/tnl/t_save_api.h
deleted file mode 100644
index c30e0e8ecd..0000000000
--- a/src/mesa/tnl/t_save_api.h
+++ /dev/null
@@ -1,57 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#ifndef __T_SAVE_API_H__
-#define __T_SAVE_API_H__
-
-#include "t_context.h"
-
-extern GLboolean _tnl_weak_begin( GLcontext *ctx, GLenum mode );
-
-extern void _tnl_EndList( GLcontext *ctx );
-extern void _tnl_NewList( GLcontext *ctx, GLuint list, GLenum mode );
-
-extern void _tnl_EndCallList( GLcontext *ctx );
-extern void _tnl_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
-
-extern void _tnl_SaveFlushVertices( GLcontext *ctx );
-
-extern void _tnl_save_init( GLcontext *ctx );
-extern void _tnl_save_destroy( GLcontext *ctx );
-
-extern void _tnl_loopback_vertex_list( GLcontext *ctx,
-				       const struct tnl_vertex_list *list );
-
-extern void _tnl_playback_vertex_list( GLcontext *ctx, void *data );
-
-#endif
diff --git a/src/mesa/tnl/t_save_loopback.c b/src/mesa/tnl/t_save_loopback.c
deleted file mode 100644
index b96bb4ec55..0000000000
--- a/src/mesa/tnl/t_save_loopback.c
+++ /dev/null
@@ -1,330 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.3
- *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/* Author:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "enums.h"
-#include "glapi.h"
-#include "imports.h"
-#include "macros.h"
-#include "mtypes.h"
-#include "t_context.h"
-#include "t_save_api.h"
-#include "dispatch.h"
-
-/* If someone compiles a display list like:
- *      glBegin(Triangles)
- *      glVertex() 
- *      ... lots of vertices ...
- *      glEnd()
- *
- * or: 
- *      glDrawArrays(...)
- * 
- * and then tries to execute it like this:
- *
- *      glBegin(Lines)
- *      glCallList()
- *      glEnd()
- *
- * it will wind up in here, as the vertex copying used when wrapping
- * buffers in list compilation (Triangles) won't be right for how the
- * list is being executed (as Lines). 
- *
- * This could be avoided by not compiling as vertex_lists until after
- * the first glEnd() has been seen.  However, that would miss an
- * important category of display lists, for the sake of a degenerate
- * usage. 
- *
- * Further, replaying degenerately-called lists in this fashion is
- * probably still faster than the replay using opcodes.
- */
-
-typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
-
-
-/* Wrapper functions in case glVertexAttrib*fvNV doesn't exist */
-static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc[4] = {
-   VertexAttrib1fvNV,
-   VertexAttrib2fvNV,
-   VertexAttrib3fvNV,
-   VertexAttrib4fvNV
-};
-
-
-static void VertexAttrib1fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvARB(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc_arb[4] = {
-   VertexAttrib1fvARB,
-   VertexAttrib2fvARB,
-   VertexAttrib3fvARB,
-   VertexAttrib4fvARB
-};
-
-
-
-
-
-
-
-static void mat_attr1fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SHININESS, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SHININESS, v ));
-      break;
-   }
-}
-
-
-static void mat_attr3fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_COLOR_INDEXES, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_COLOR_INDEXES, v ));
-      break;
-   }
-}
-
-
-static void mat_attr4fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_EMISSION, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_EMISSION, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_AMBIENT, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_AMBIENT, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_DIFFUSE, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_DIFFUSE, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SPECULAR, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SPECULAR, v ));
-      break;
-   }
-}
-
-
-static attr_func mat_attrfunc[4] = {
-   mat_attr1fv,
-   NULL,
-   mat_attr3fv,
-   mat_attr4fv
-};
-
-
-static void edgeflag_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_EdgeFlag(ctx->Exec, ((GLboolean)(v[0] == 1.0)));
-}
-
-struct loopback_attr {
-   GLint target;
-   GLint sz;
-   attr_func func;
-};
-
-/* Don't emit ends and begins on wrapped primitives.  Don't replay
- * wrapped vertices.  If we get here, it's probably because the the
- * precalculated wrapping is wrong.
- */
-static void loopback_prim( GLcontext *ctx,
-			   const struct tnl_vertex_list *list, GLuint i,
-			   const struct loopback_attr *la, GLuint nr )
-{
-   struct tnl_prim *prim = &list->prim[i];
-   GLint begin = prim->start;
-   GLint end = begin + prim->count;
-   GLfloat *data;
-   GLint j;
-   GLuint k;
-
-   if (prim->mode & PRIM_BEGIN) {
-      CALL_Begin(GET_DISPATCH(), ( prim->mode & PRIM_MODE_MASK ));
-   }
-   else {
-      assert(i == 0);
-      assert(begin == 0);
-      begin += list->wrap_count;
-   }
-
-   data = list->buffer + begin * list->vertex_size;
-
-   for (j = begin ; j < end ; j++) {
-      GLfloat *tmp = data + la[0].sz;
-
-      for (k = 1 ; k < nr ; k++) {
-	 la[k].func( ctx, la[k].target, tmp );
-	 tmp += la[k].sz;
-      }
-	 
-      /* Fire the vertex
-       */
-      la[0].func( ctx, VERT_ATTRIB_POS, data );
-      data = tmp;
-   }
-
-   if (prim->mode & PRIM_END) {
-      CALL_End(GET_DISPATCH(), ());
-   }
-   else {
-      assert (i == list->prim_count-1);
-   }
-}
-
-/* Primitives generated by DrawArrays/DrawElements/Rectf may be
- * caught here.  If there is no primitive in progress, execute them
- * normally, otherwise need to track and discard the generated
- * primitives.
- */
-static void loopback_weak_prim( GLcontext *ctx,
-				const struct tnl_vertex_list *list, GLuint i,
-				const struct loopback_attr *la, GLuint nr )
-{
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) 
-      loopback_prim( ctx, list, i, la, nr );
-   else {
-      struct tnl_prim *prim = &list->prim[i];
-
-      /* Use the prim_weak flag to ensure that if this primitive
-       * wraps, we don't mistake future vertex_lists for part of the
-       * surrounding primitive.
-       *
-       * While this flag is set, we are simply disposing of data
-       * generated by an operation now known to be a noop.
-       */
-      if (prim->mode & PRIM_BEGIN)
-	 ctx->Driver.CurrentExecPrimitive |= PRIM_WEAK;
-      if (prim->mode & PRIM_END)
-	 ctx->Driver.CurrentExecPrimitive &= ~PRIM_WEAK;
-   }
-}
-
-
-
-void _tnl_loopback_vertex_list( GLcontext *ctx,
-                                const struct tnl_vertex_list *list )
-{
-   struct loopback_attr la[_TNL_ATTRIB_MAX];
-   GLuint i, nr = 0;
-
-   /* conventional + generic attributes */
-   for (i = 0 ; i <= _TNL_ATTRIB_GENERIC15 ; i++) {
-      if (list->attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = list->attrsz[i];
-	 la[nr].func = vert_attrfunc[list->attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   /* material attributes */
-   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT ; 
-	i <= _TNL_ATTRIB_MAT_BACK_INDEXES ; 
-	i++) {
-      if (list->attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = list->attrsz[i];
-	 la[nr].func = mat_attrfunc[list->attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   /* special-case: edgeflag */
-   if (list->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      la[nr].target = _TNL_ATTRIB_EDGEFLAG;
-      la[nr].sz = list->attrsz[_TNL_ATTRIB_EDGEFLAG];
-      la[nr].func = edgeflag_attr1fv;
-      nr++;
-   }
-
-   for (i = 0 ; i < list->prim_count ; i++) {
-      if (list->prim[i].mode & PRIM_WEAK)
-	 loopback_weak_prim( ctx, list, i, la, nr );
-      else
-	 loopback_prim( ctx, list, i, la, nr );
-   }
-}
diff --git a/src/mesa/tnl/t_save_playback.c b/src/mesa/tnl/t_save_playback.c
deleted file mode 100644
index 426e94372e..0000000000
--- a/src/mesa/tnl/t_save_playback.c
+++ /dev/null
@@ -1,216 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.1
- *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/* Author:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "imports.h"
-#include "mtypes.h"
-#include "macros.h"
-#include "light.h"
-#include "state.h"
-#include "t_pipeline.h"
-#include "t_save_api.h"
-#include "t_vtx_api.h"
-
-static INLINE GLint get_size( const GLfloat *f )
-{
-   if (f[3] != 1.0) return 4;
-   if (f[2] != 0.0) return 3;
-   return 2;
-}
-
-
-/* Some nasty stuff still hanging on here.  
- *
- * TODO - remove VB->ColorPtr, etc and just use the AttrPtr's.
- */
-static void _tnl_bind_vertex_list( GLcontext *ctx,
-                                   const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->save_inputs;
-   GLfloat *data = node->buffer;
-   GLuint attr, i;
-
-   /* Setup constant data in the VB.
-    */
-   VB->Count = node->count;
-   VB->Primitive = node->prim;
-   VB->PrimitiveCount = node->prim_count;
-   VB->Elts = NULL;
-   VB->NormalLengthPtr = node->normal_lengths;
-
-   for (attr = 0; attr <= _TNL_ATTRIB_EDGEFLAG; attr++) {
-      if (node->attrsz[attr]) {
-	 tmp->Attribs[attr].count = node->count;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) data;
-	 tmp->Attribs[attr].start = data;
-	 tmp->Attribs[attr].size = node->attrsz[attr];
-	 tmp->Attribs[attr].stride = node->vertex_size * sizeof(GLfloat);
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-	 data += node->attrsz[attr];
-      }
-      else {
-	 tmp->Attribs[attr].count = 1;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) tnl->vtx.current[attr];
-	 tmp->Attribs[attr].start = tnl->vtx.current[attr];
-	 tmp->Attribs[attr].size = get_size( tnl->vtx.current[attr] );
-	 tmp->Attribs[attr].stride = 0;
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-      }
-   }
-
-   
-   /* Copy edgeflag to a contiguous array
-    */
-   if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL) {
-      if (node->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-	 VB->EdgeFlag = _tnl_translate_edgeflag( ctx, data, 
-						 node->count,
-						 node->vertex_size );
-	 data++;
-      }
-      else 
-	 VB->EdgeFlag = _tnl_import_current_edgeflag( ctx, node->count );
-   }
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
-
-static void _playback_copy_to_current( GLcontext *ctx,
-				       const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   const GLfloat *data;
-   GLuint i;
-
-   if (node->count)
-      data = node->buffer + (node->count-1) * node->vertex_size;
-   else
-      data = node->buffer;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i <= _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (node->attrsz[i]) {
-	 COPY_CLEAN_4V(tnl->vtx.current[i], node->attrsz[i], data);
-	 data += node->attrsz[i];
-      }
-   }
-
-   /* Edgeflag requires special treatment:
-    */
-   if (node->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (data[0] == 1.0);
-   }
-
-   /* Colormaterial -- this kindof sucks.
-    */
-   if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
-   }
-
-   if (node->have_materials) {
-      tnl->Driver.NotifyMaterialChange( ctx );
-      ctx->NewState |= _NEW_LIGHT;
-   }
-
-   /* CurrentExecPrimitive
-    */
-   if (node->prim_count) {
-      GLenum mode = node->prim[node->prim_count - 1].mode;
-      if (mode & PRIM_END)
-	 ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-      else
-	 ctx->Driver.CurrentExecPrimitive = (mode & PRIM_MODE_MASK);
-   }
-}
-
-
-/**
- * Execute the buffer and save copied verts.
- */
-void _tnl_playback_vertex_list( GLcontext *ctx, void *data )
-{
-   const struct tnl_vertex_list *node = (const struct tnl_vertex_list *) data;
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   FLUSH_CURRENT(ctx, 0);
-
-   if (node->prim_count > 0 && node->count > 0) {
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END &&
-	  (node->prim[0].mode & PRIM_BEGIN)) {
-
-	 /* Degenerate case: list is called inside begin/end pair and
-	  * includes operations such as glBegin or glDrawArrays.
-	  */
-	 _mesa_error( ctx, GL_INVALID_OPERATION, "displaylist recursive begin");
-	 _tnl_loopback_vertex_list( ctx, node );
-	 return;
-      }
-      else if (tnl->save.replay_flags) {
-	 /* Various degnerate cases: translate into immediate mode
-	  * calls rather than trying to execute in place.
-	  */
-	 _tnl_loopback_vertex_list( ctx, node );
-	 return;
-      }
-      
-      if (ctx->NewState)
-	 _mesa_update_state( ctx );
-
-      if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-          (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glBegin (invalid vertex/fragment program)");
-         return;
-      }
-
-      _tnl_bind_vertex_list( ctx, node );
-
-      tnl->Driver.RunPipeline( ctx );
-   }
-
-   /* Copy to current?
-    */
-   _playback_copy_to_current( ctx, node );
-}
diff --git a/src/mesa/tnl/t_vb_render.c b/src/mesa/tnl/t_vb_render.c
index 163fa31257..c38f0745e1 100644
--- a/src/mesa/tnl/t_vb_render.c
+++ b/src/mesa/tnl/t_vb_render.c
@@ -305,7 +305,7 @@ static GLboolean run_render( GLcontext *ctx,
 
       for (i = 0 ; i < VB->PrimitiveCount ; i++)
       {
-	 GLuint prim = VB->Primitive[i].mode;
+	 GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 	 GLuint start = VB->Primitive[i].start;
 	 GLuint length = VB->Primitive[i].count;
 
diff --git a/src/mesa/tnl/t_vb_rendertmp.h b/src/mesa/tnl/t_vb_rendertmp.h
index 90319eddfc..2b5f4e93b2 100644
--- a/src/mesa/tnl/t_vb_rendertmp.h
+++ b/src/mesa/tnl/t_vb_rendertmp.h
@@ -45,8 +45,8 @@
 #endif
 
 #ifndef TEST_PRIM_END
-#define TEST_PRIM_END(flags) (flags & PRIM_END)
-#define TEST_PRIM_BEGIN(flags) (flags & PRIM_BEGIN)
+#define TEST_PRIM_END(prim) (flags & PRIM_END)
+#define TEST_PRIM_BEGIN(prim) (flags & PRIM_BEGIN)
 #endif
 
 #ifndef ELT
diff --git a/src/mesa/tnl/t_vtx_api.c b/src/mesa/tnl/t_vtx_api.c
deleted file mode 100644
index 77eec8b3b4..0000000000
--- a/src/mesa/tnl/t_vtx_api.c
+++ /dev/null
@@ -1,1046 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-#include "simple_list.h"
-
-#include "dispatch.h"
-
-static void reset_attrfv( TNLcontext *tnl );
-
-/** Note extra space for error index: */
-static tnl_attrfv_func choose[_TNL_ATTRIB_ERROR+1][4];
-static tnl_attrfv_func generic_attr_func[_TNL_MAX_ATTR_CODEGEN][4];
-
-
-/* Close off the last primitive, execute the buffer, restart the
- * primitive.  
- */
-static void _tnl_wrap_buffers( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   
-
-   if (tnl->vtx.prim_count == 0) {
-      tnl->vtx.copied.nr = 0;
-      tnl->vtx.counter = tnl->vtx.initial_counter;
-      tnl->vtx.vbptr = tnl->vtx.buffer;
-   }
-   else {
-      GLuint last_prim = tnl->vtx.prim[tnl->vtx.prim_count-1].mode;
-      GLuint last_count;
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-	 GLint i = tnl->vtx.prim_count - 1;
-	 assert(i >= 0);
-	 tnl->vtx.prim[i].count = ((tnl->vtx.initial_counter - 
-				    tnl->vtx.counter) - 
-				   tnl->vtx.prim[i].start);
-      }
-
-      last_count = tnl->vtx.prim[tnl->vtx.prim_count-1].count;
-
-      /* Execute the buffer and save copied vertices.
-       */
-      if (tnl->vtx.counter != tnl->vtx.initial_counter)
-	 _tnl_flush_vtx( ctx );
-      else {
-	 tnl->vtx.prim_count = 0;
-	 tnl->vtx.copied.nr = 0;
-      }
-
-      /* Emit a glBegin to start the new list.
-       */
-      assert(tnl->vtx.prim_count == 0);
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-	 tnl->vtx.prim[0].mode = ctx->Driver.CurrentExecPrimitive;
-	 tnl->vtx.prim[0].start = 0;
-	 tnl->vtx.prim[0].count = 0;
-	 tnl->vtx.prim_count++;
-      
-	 if (tnl->vtx.copied.nr == last_count)
-	    tnl->vtx.prim[0].mode |= last_prim & PRIM_BEGIN;
-      }
-   }
-}
-
-
-/* Deal with buffer wrapping where provoked by the vertex buffer
- * filling up, as opposed to upgrade_vertex().
- *
- * Make it GLAPIENTRY, so we can tail from the codegen'ed Vertex*fv
- */
-void GLAPIENTRY _tnl_wrap_filled_vertex( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat *data = tnl->vtx.copied.buffer;
-   GLuint i;
-
-   /* Run pipeline on current vertices, copy wrapped vertices
-    * to tnl->copied.
-    */
-   _tnl_wrap_buffers( ctx );
-   
-   /* Copy stored stored vertices to start of new list. 
-    */
-   assert(tnl->vtx.counter > tnl->vtx.copied.nr);
-
-   for (i = 0 ; i < tnl->vtx.copied.nr ; i++) {
-      _mesa_memcpy( tnl->vtx.vbptr, data, 
-		    tnl->vtx.vertex_size * sizeof(GLfloat));
-      tnl->vtx.vbptr += tnl->vtx.vertex_size;
-      data += tnl->vtx.vertex_size;
-      tnl->vtx.counter--;
-   }
-
-   tnl->vtx.copied.nr = 0;
-}
-
-
-/*
- * Copy the active vertex's values to the ctx->Current fields.
- */
-static void _tnl_copy_to_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i < _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (tnl->vtx.attrsz[i]) {
-         /* Note: the tnl->vtx.current[i] pointers points to
-          * the ctx->Current fields.  The first 16 or so, anyway.
-          */
-	 COPY_CLEAN_4V(tnl->vtx.current[i], 
-                       tnl->vtx.attrsz[i], 
-                       tnl->vtx.attrptr[i]);
-      }
-   }
-
-   /* Edgeflag requires additional treatment:
-    */
-   if (tnl->vtx.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = 
-	 (tnl->vtx.CurrentFloatEdgeFlag == 1.0);
-   }
-
-   /* Colormaterial -- this kindof sucks.
-    */
-   if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, 
-				  ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
-   }
-
-   if (tnl->vtx.have_materials) {
-      tnl->Driver.NotifyMaterialChange( ctx );
-      ctx->NewState |= _NEW_LIGHT;
-   }
-         
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-
-static void _tnl_copy_from_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLint i;
-
-   /* Edgeflag requires additional treatment:
-    */
-   tnl->vtx.CurrentFloatEdgeFlag = (GLfloat) ctx->Current.EdgeFlag;
-   
-   for (i = _TNL_ATTRIB_POS+1 ; i < _TNL_ATTRIB_MAX ; i++) 
-      switch (tnl->vtx.attrsz[i]) {
-      case 4: tnl->vtx.attrptr[i][3] = tnl->vtx.current[i][3];
-      case 3: tnl->vtx.attrptr[i][2] = tnl->vtx.current[i][2];
-      case 2: tnl->vtx.attrptr[i][1] = tnl->vtx.current[i][1];
-      case 1: tnl->vtx.attrptr[i][0] = tnl->vtx.current[i][0];
-	 break;
-      }
-
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-}
-
-
-/* Flush existing data, set new attrib size, replay copied vertices.
- */ 
-static void _tnl_wrap_upgrade_vertex( GLcontext *ctx, 
-				      GLuint attr,
-				      GLuint newsz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint oldsz;
-   GLuint i;
-   GLfloat *tmp;
-   GLint lastcount = tnl->vtx.initial_counter - tnl->vtx.counter;
-
-   /* Run pipeline on current vertices, copy wrapped vertices
-    * to tnl->vtx.copied.
-    */
-   _tnl_wrap_buffers( ctx );
-
-
-   /* Do a COPY_TO_CURRENT to ensure back-copying works for the case
-    * when the attribute already exists in the vertex and is having
-    * its size increased.  
-    */
-   _tnl_copy_to_current( ctx );
-
-
-   /* Heuristic: Attempt to isolate attributes received outside
-    * begin/end so that they don't bloat the vertices.
-    */
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END &&
-       tnl->vtx.attrsz[attr] == 0 && 
-       lastcount > 8 &&
-       tnl->vtx.vertex_size) {
-      reset_attrfv( tnl );
-   }
-
-   /* Fix up sizes:
-    */
-   oldsz = tnl->vtx.attrsz[attr];
-   tnl->vtx.attrsz[attr] = newsz;
-
-   tnl->vtx.vertex_size += newsz - oldsz;
-   tnl->vtx.counter = MIN2( VERT_BUFFER_SIZE / tnl->vtx.vertex_size,
-			    ctx->Const.MaxArrayLockSize );
-   tnl->vtx.initial_counter = tnl->vtx.counter;
-   tnl->vtx.vbptr = tnl->vtx.buffer;
-
-
-   /* Recalculate all the attrptr[] values
-    */
-   for (i = 0, tmp = tnl->vtx.vertex ; i < _TNL_ATTRIB_MAX ; i++) {
-      if (tnl->vtx.attrsz[i]) {
-	 tnl->vtx.attrptr[i] = tmp;
-	 tmp += tnl->vtx.attrsz[i];
-      }
-      else 
-	 tnl->vtx.attrptr[i] = NULL; /* will not be dereferenced */
-   }
-
-   /* Copy from current to repopulate the vertex with correct values.
-    */
-   _tnl_copy_from_current( ctx );
-
-   /* Replay stored vertices to translate them
-    * to new format here.
-    *
-    * -- No need to replay - just copy piecewise
-    */
-   if (tnl->vtx.copied.nr)
-   {
-      const GLfloat *data = tnl->vtx.copied.buffer;
-      GLfloat *dest = tnl->vtx.buffer;
-      GLuint j;
-
-      for (i = 0 ; i < tnl->vtx.copied.nr ; i++) {
-	 for (j = 0 ; j < _TNL_ATTRIB_MAX ; j++) {
-	    if (tnl->vtx.attrsz[j]) {
-	       if (j == attr) {
-		  if (oldsz) {
-		     COPY_CLEAN_4V( dest, oldsz, data );
-		     data += oldsz;
-		     dest += newsz;
-		  } else {
-		     COPY_SZ_4V( dest, newsz, tnl->vtx.current[j] );
-		     dest += newsz;
-		  }
-	       }
-	       else {
-		  GLuint sz = tnl->vtx.attrsz[j];
-		  COPY_SZ_4V( dest, sz, data );
-		  dest += sz;
-		  data += sz;
-	       }
-	    }
-	 }
-      }
-
-      tnl->vtx.vbptr = dest;
-      tnl->vtx.counter -= tnl->vtx.copied.nr;
-      tnl->vtx.copied.nr = 0;
-   }
-
-   /* For codegen - attrptr's may have changed, so need to redo
-    * codegen.  Might be a reasonable place to try & detect attributes
-    * in the vertex which aren't being submitted any more.
-    */
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) 
-      if (tnl->vtx.attrsz[i]) {
-	 GLuint j = tnl->vtx.attrsz[i] - 1;
-
-	 if (i < _TNL_MAX_ATTR_CODEGEN)
-	    tnl->vtx.tabfv[i][j] = choose[i][j];
-      }
-
-}
-
-
-static void _tnl_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   static const GLfloat id[4] = { 0, 0, 0, 1 };
-   int i;
-
-   if (0) 
-      _mesa_printf("%s attr %d sz %d -> %d\n",
-		   __FUNCTION__, attr, tnl->vtx.attrsz[attr], sz);
-
-   if (tnl->vtx.attrsz[attr] < sz) {
-      /* New size is larger.  Need to flush existing vertices and get
-       * an enlarged vertex format.
-       */
-      _tnl_wrap_upgrade_vertex( ctx, attr, sz );
-   }
-   else if (sz < tnl->vtx.active_sz[attr]) {
-      /* New size is smaller - just need to fill in some
-       * zeros.  Don't need to flush or wrap.
-       */
-      for (i = sz ; i <= tnl->vtx.attrsz[attr] ; i++)
-	 tnl->vtx.attrptr[attr][i-1] = id[i-1];
-   }
-
-   tnl->vtx.active_sz[attr] = sz;
-
-   /* Does setting NeedFlush belong here?  Necessitates resetting
-    * vtxfmt on each flush (otherwise flags won't get reset
-    * afterwards).
-    */
-   if (attr == 0) 
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   else 
-      ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-}
-
-#ifdef USE_X86_ASM
-
-static struct _tnl_dynfn *lookup( struct _tnl_dynfn *l, GLuint key )
-{
-   struct _tnl_dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key == key) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-
-static tnl_attrfv_func do_codegen( GLcontext *ctx, GLuint attr, GLuint sz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   struct _tnl_dynfn *dfn = NULL;
-
-   if (attr == 0) {
-      GLuint key = tnl->vtx.vertex_size;
-
-      dfn = lookup( &tnl->vtx.cache.Vertex[sz-1], key );
-
-      if (!dfn)
-	 dfn = tnl->vtx.gen.Vertex[sz-1]( ctx, key );
-   }
-   else {
-      GLuint key = (GLuint) tnl->vtx.attrptr[attr];
-
-      dfn = lookup( &tnl->vtx.cache.Attribute[sz-1], key );
-
-      if (!dfn)
-	 dfn = tnl->vtx.gen.Attribute[sz-1]( ctx, key );
-   }
-
-   if (dfn) 
-      return *(tnl_attrfv_func *) &dfn->code;
-   else
-      return NULL;
-}
-
-#endif /* USE_X86_ASM */
-
-/* Helper function for 'CHOOSE' macro.  Do what's necessary when an
- * entrypoint is called for the first time.
- */
-
-static tnl_attrfv_func do_choose( GLuint attr, GLuint sz )
-{ 
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint oldsz = tnl->vtx.active_sz[attr];
-
-   assert(attr < _TNL_MAX_ATTR_CODEGEN);
-
-   if (oldsz != sz) {
-      /* Reset any active pointers for this attribute 
-       */
-      if (oldsz)
-	 tnl->vtx.tabfv[attr][oldsz-1] = choose[attr][oldsz-1];
-   
-      _tnl_fixup_vertex( ctx, attr, sz );
- 
-   }
-
-
-   /* Try to use codegen:
-    */
-#ifdef USE_X86_ASM
-   if (tnl->AllowCodegen)
-      tnl->vtx.tabfv[attr][sz-1] = do_codegen( ctx, attr, sz );
-   else
-#endif
-      tnl->vtx.tabfv[attr][sz-1] = NULL;
-
-   /* Else use generic version:
-    */
-   if (!tnl->vtx.tabfv[attr][sz-1])
-      tnl->vtx.tabfv[attr][sz-1] = generic_attr_func[attr][sz-1];
-
-   ASSERT(tnl->vtx.tabfv[attr][sz-1]);
-   return tnl->vtx.tabfv[attr][sz-1];
-}
-
-
-
-#define CHOOSE( ATTR, N )				\
-static void choose_##ATTR##_##N( const GLfloat *v )	\
-{							\
-   tnl_attrfv_func f = do_choose(ATTR, N);		\
-   ASSERT(f); \
-   f( v );						\
-}
-
-#define CHOOSERS( ATTRIB )	\
-   CHOOSE( ATTRIB, 1 )		\
-   CHOOSE( ATTRIB, 2 )		\
-   CHOOSE( ATTRIB, 3 )		\
-   CHOOSE( ATTRIB, 4 )		\
-
-
-#define INIT_CHOOSERS(ATTR)			\
-   ASSERT(ATTR <= _TNL_ATTRIB_ERROR);\
-   choose[ATTR][0] = choose_##ATTR##_1;		\
-   choose[ATTR][1] = choose_##ATTR##_2;		\
-   choose[ATTR][2] = choose_##ATTR##_3;		\
-   choose[ATTR][3] = choose_##ATTR##_4;
-
-/* conventional attributes */
-CHOOSERS( 0 )
-CHOOSERS( 1 )
-CHOOSERS( 2 )
-CHOOSERS( 3 )
-CHOOSERS( 4 )
-CHOOSERS( 5 )
-CHOOSERS( 6 )
-CHOOSERS( 7 )
-CHOOSERS( 8 )
-CHOOSERS( 9 )
-CHOOSERS( 10 )
-CHOOSERS( 11 )
-CHOOSERS( 12 )
-CHOOSERS( 13 )
-CHOOSERS( 14 )
-CHOOSERS( 15 )
-
-/* generic attributes */
-CHOOSERS( 16 )
-CHOOSERS( 17 )
-CHOOSERS( 18 )
-CHOOSERS( 19 )
-CHOOSERS( 20 )
-CHOOSERS( 21 )
-CHOOSERS( 22 )
-CHOOSERS( 23 )
-CHOOSERS( 24 )
-CHOOSERS( 25 )
-CHOOSERS( 26 )
-CHOOSERS( 27 )
-CHOOSERS( 28 )
-CHOOSERS( 29 )
-CHOOSERS( 30 )
-CHOOSERS( 31 )
-
-
-/**
- * This function will get called when glVertexAttribNV/ARB() is called
- * with an invalid index parameter.
- */
-static void
-error_attrib(const GLfloat *unused)
-{
-   GET_CURRENT_CONTEXT( ctx );
-   (void) unused;
-   _mesa_error( ctx, GL_INVALID_VALUE, "glVertexAttrib(index)" );
-}   
-
-
-
-/**
- * Reset all the per-vertex functions pointers to point to the default
- * "chooser" functions.
- */
-static void
-reset_attrfv(TNLcontext *tnl)
-{   
-   GLuint i;
-
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) 
-      if (tnl->vtx.attrsz[i]) {
-	 GLint j = tnl->vtx.attrsz[i] - 1;
-	 tnl->vtx.attrsz[i] = 0;
-	 tnl->vtx.active_sz[i] = 0;
-
-	 if (i < _TNL_MAX_ATTR_CODEGEN) {
-            while (j >= 0) {
-	       tnl->vtx.tabfv[i][j] = choose[i][j];
-               j--;
-            }
-         }
-      }
-
-   tnl->vtx.vertex_size = 0;
-   tnl->vtx.have_materials = 0;
-}
-      
-
-
-/**
- * Materials:  
- * 
- * These are treated as per-vertex attributes, at indices above where
- * the NV_vertex_program leaves off.  There are a lot of good things
- * about treating materials this way.  
- *
- * However: I don't want to double the number of generated functions
- * just to cope with this, so I unroll the 'C' varients of CHOOSE and
- * ATTRF into this function, and dispense with codegen and
- * second-level dispatch.
- *
- * There is no aliasing of material attributes with other entrypoints.
- */
-#define OTHER_ATTR( A, N, params )			\
-do {							\
-   if (tnl->vtx.active_sz[A] != N) {			\
-      _tnl_fixup_vertex( ctx, A, N );			\
-   }							\
-							\
-   {							\
-      GLfloat *dest = tnl->vtx.attrptr[A];		\
-      if (N>0) dest[0] = (params)[0];			\
-      if (N>1) dest[1] = (params)[1];			\
-      if (N>2) dest[2] = (params)[2];			\
-      if (N>3) dest[3] = (params)[3];			\
-   }							\
-} while (0)
-
-
-#define MAT( ATTR, N, face, params )			\
-do {							\
-   if (face != GL_BACK)					\
-      OTHER_ATTR( ATTR, N, params ); /* front */	\
-   if (face != GL_FRONT)				\
-      OTHER_ATTR( ATTR + 1, N, params ); /* back */	\
-} while (0)
-
-
-/**
- * Called by glMaterialfv().
- * Colormaterial is dealt with later on.
- */
-static void GLAPIENTRY
-_tnl_Materialfv( GLenum face, GLenum pname, const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (face) {
-   case GL_FRONT:
-   case GL_BACK:
-   case GL_FRONT_AND_BACK:
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-
-   switch (pname) {
-   case GL_EMISSION:
-      MAT( _TNL_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
-      break;
-   case GL_AMBIENT:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      break;
-   case GL_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   case GL_SPECULAR:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
-      break;
-   case GL_SHININESS:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
-      break;
-   case GL_COLOR_INDEXES:
-      MAT( _TNL_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
-      break;
-   case GL_AMBIENT_AND_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-
-   tnl->vtx.have_materials = GL_TRUE;
-}
-
-
-static void GLAPIENTRY _tnl_EdgeFlag( GLboolean b )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat f = (GLfloat)b;
-
-   OTHER_ATTR( _TNL_ATTRIB_EDGEFLAG, 1, &f );
-}
-
-
-/* Eval
- */
-static void GLAPIENTRY _tnl_EvalCoord1f( GLfloat u )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* TODO: use a CHOOSE() function for this: */
-   {
-      GLint i;
-      if (tnl->vtx.eval.new_state) 
-	 _tnl_update_eval( ctx );
-
-      for (i = 0; i < _TNL_NUM_EVAL; i++) {
-	 if (tnl->vtx.eval.map1[i].map) 
-	    if (tnl->vtx.active_sz[i] != tnl->vtx.eval.map1[i].sz)
-	       _tnl_fixup_vertex( ctx, i, tnl->vtx.eval.map1[i].sz );
-      }
-   }
-
-
-   _mesa_memcpy( tnl->vtx.copied.buffer, tnl->vtx.vertex, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-
-   _tnl_do_EvalCoord1f( ctx, u );
-
-   _mesa_memcpy( tnl->vtx.vertex, tnl->vtx.copied.buffer,
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-}
-
-static void GLAPIENTRY _tnl_EvalCoord2f( GLfloat u, GLfloat v )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* TODO: use a CHOOSE() function for this: */
-   {
-      GLint i;
-      if (tnl->vtx.eval.new_state) 
-	 _tnl_update_eval( ctx );
-
-      for (i = 0; i < _TNL_NUM_EVAL; i++) {
-	 if (tnl->vtx.eval.map2[i].map) 
-	    if (tnl->vtx.active_sz[i] != tnl->vtx.eval.map2[i].sz)
-	       _tnl_fixup_vertex( ctx, i, tnl->vtx.eval.map2[i].sz );
-      }
-
-      if (ctx->Eval.AutoNormal) 
-	 if (tnl->vtx.active_sz[_TNL_ATTRIB_NORMAL] != 3)
-	    _tnl_fixup_vertex( ctx, _TNL_ATTRIB_NORMAL, 3 );
-   }
-
-   _mesa_memcpy( tnl->vtx.copied.buffer, tnl->vtx.vertex, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-
-   _tnl_do_EvalCoord2f( ctx, u, v );
-
-   _mesa_memcpy( tnl->vtx.vertex, tnl->vtx.copied.buffer, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-}
-
-static void GLAPIENTRY _tnl_EvalCoord1fv( const GLfloat *u )
-{
-   _tnl_EvalCoord1f( u[0] );
-}
-
-static void GLAPIENTRY _tnl_EvalCoord2fv( const GLfloat *u )
-{
-   _tnl_EvalCoord2f( u[0], u[1] );
-}
-
-static void GLAPIENTRY _tnl_EvalPoint1( GLint i )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   GLfloat du = ((ctx->Eval.MapGrid1u2 - ctx->Eval.MapGrid1u1) /
-		 (GLfloat) ctx->Eval.MapGrid1un);
-   GLfloat u = i * du + ctx->Eval.MapGrid1u1;
-
-   _tnl_EvalCoord1f( u );
-}
-
-
-static void GLAPIENTRY _tnl_EvalPoint2( GLint i, GLint j )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   GLfloat du = ((ctx->Eval.MapGrid2u2 - ctx->Eval.MapGrid2u1) / 
-		 (GLfloat) ctx->Eval.MapGrid2un);
-   GLfloat dv = ((ctx->Eval.MapGrid2v2 - ctx->Eval.MapGrid2v1) / 
-		 (GLfloat) ctx->Eval.MapGrid2vn);
-   GLfloat u = i * du + ctx->Eval.MapGrid2u1;
-   GLfloat v = j * dv + ctx->Eval.MapGrid2v1;
-
-   _tnl_EvalCoord2f( u, v );
-}
-
-
-/**
- * Called from glBegin.
- * ctx->Driver.CurrentExecPrimitive will be set to <mode>.
- */
-static void GLAPIENTRY _tnl_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-
-   if (mode > GL_POLYGON) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glBegin(mode)");
-      return;
-   }
-
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) {
-      /* we're not inside a glBegin/End pair */
-      TNLcontext *tnl = TNL_CONTEXT(ctx); 
-      GLuint i;
-
-      if (ctx->NewState) {
-	 _mesa_update_state( ctx );
-
-         if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-            (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-            _mesa_error(ctx, GL_INVALID_OPERATION,
-                        "glBegin (invalid vertex/fragment program)");
-            tnl->DiscardPrimitive = GL_TRUE;
-            return;
-         }
-
-         if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
-            _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
-                        "glBegin(incomplete framebuffer)");
-            tnl->DiscardPrimitive = GL_TRUE;
-            return;
-         }
-
-         tnl->DiscardPrimitive = GL_FALSE;
-
-	 if (!(tnl->Driver.NotifyBegin && 
-	       tnl->Driver.NotifyBegin( ctx, mode )))
-	     CALL_Begin(ctx->Exec, (mode));
-	 return;
-      }
-
-      /* Heuristic: attempt to isolate attributes occuring outside
-       * begin/end pairs.
-       */
-      if (tnl->vtx.vertex_size && !tnl->vtx.attrsz[0]) 
-	 _tnl_FlushVertices( ctx, ~0 );
-
-      i = tnl->vtx.prim_count++;
-      tnl->vtx.prim[i].mode = mode | PRIM_BEGIN;
-      tnl->vtx.prim[i].start = tnl->vtx.initial_counter - tnl->vtx.counter;
-      tnl->vtx.prim[i].count = 0;
-
-      ctx->Driver.CurrentExecPrimitive = mode;
-   }
-   else {
-      /* already inside glBegin/End */
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-   }      
-}
-
-
-/**
- * Called from glEnd.
- */
-static void GLAPIENTRY _tnl_End( void )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-
-   if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-      /* closing an open glBegin primitive */
-      TNLcontext *tnl = TNL_CONTEXT(ctx); 
-      int idx = tnl->vtx.initial_counter - tnl->vtx.counter;
-      int i = tnl->vtx.prim_count - 1;
-
-      tnl->vtx.prim[i].mode |= PRIM_END; 
-      tnl->vtx.prim[i].count = idx - tnl->vtx.prim[i].start;
-
-      ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-
-      /* Two choices which effect the way vertex attributes are
-       * carried over (or not) between adjacent primitives.
-       */
-#if 0
-      if (tnl->vtx.prim_count == TNL_MAX_PRIM) 
-	 _tnl_FlushVertices( ctx, ~0 );
-#else
-      if (tnl->vtx.prim_count == TNL_MAX_PRIM)
-	 _tnl_flush_vtx( ctx );	
-#endif
-
-   }
-   else {
-      /* glBegin hasn't been called! */
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-   }
-}
-
-
-/**
- * XXX why aren't all members initialized here??
- */
-static void _tnl_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = _tnl_Begin;
-   vfmt->CallList = _mesa_CallList;
-   vfmt->CallLists = _mesa_CallLists;
-   vfmt->EdgeFlag = _tnl_EdgeFlag;
-   vfmt->End = _tnl_End;
-   vfmt->EvalCoord1f = _tnl_EvalCoord1f;
-   vfmt->EvalCoord1fv = _tnl_EvalCoord1fv;
-   vfmt->EvalCoord2f = _tnl_EvalCoord2f;
-   vfmt->EvalCoord2fv = _tnl_EvalCoord2fv;
-   vfmt->EvalPoint1 = _tnl_EvalPoint1;
-   vfmt->EvalPoint2 = _tnl_EvalPoint2;
-   vfmt->Materialfv = _tnl_Materialfv;
-
-   vfmt->Rectf = _mesa_noop_Rectf;
-   vfmt->EvalMesh1 = _mesa_noop_EvalMesh1;
-   vfmt->EvalMesh2 = _mesa_noop_EvalMesh2;
-}
-
-
-
-void _tnl_FlushVertices( GLcontext *ctx, GLuint flags )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   (void) flags;
-
-   if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-      /* still inside a glBegin/End pair.  How'd we get here??? */
-      return;
-   }
-
-   if (tnl->DiscardPrimitive) {
-      /* discard any primitives */
-      tnl->vtx.prim_count = 0;
-      tnl->vtx.counter = tnl->vtx.initial_counter;
-      tnl->vtx.vbptr = tnl->vtx.buffer;
-   }
-
-   if (tnl->vtx.counter != tnl->vtx.initial_counter) {
-      _tnl_flush_vtx( ctx );
-   }
-
-   if (tnl->vtx.vertex_size) {
-      _tnl_copy_to_current( ctx );
-      reset_attrfv( tnl );
-   }
-
-   ctx->Driver.NeedFlush = 0;
-}
-
-
-/**
- * Init the tnl->vtx->current[] pointers to point to the corresponding
- * fields in ctx->Current attribute group.
- */
-static void _tnl_current_init( GLcontext *ctx ) 
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLint i;
-
-   /* setup the pointers for the typical (32) vertex attributes */
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) 
-      tnl->vtx.current[i] = ctx->Current.Attrib[i];
-
-   /* setup pointers for the 12 material attributes */
-   for (i = 0; i < MAT_ATTRIB_MAX; i++)
-      tnl->vtx.current[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = 
-	 ctx->Light.Material.Attrib[i];
-
-   /* special case */
-   tnl->vtx.current[_TNL_ATTRIB_EDGEFLAG] = &tnl->vtx.CurrentFloatEdgeFlag;
-}
-
-static struct _tnl_dynfn *no_codegen( GLcontext *ctx, int key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void _tnl_vtx_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   struct tnl_vertex_arrays *tmp = &tnl->vtx_inputs;
-   GLuint i;
-   static int firsttime = 1;
-   
-   if (firsttime) {
-      firsttime = 0;
-
-      /* conventional attributes */
-      INIT_CHOOSERS( 0 );
-      INIT_CHOOSERS( 1 );
-      INIT_CHOOSERS( 2 );
-      INIT_CHOOSERS( 3 );
-      INIT_CHOOSERS( 4 );
-      INIT_CHOOSERS( 5 );
-      INIT_CHOOSERS( 6 );
-      INIT_CHOOSERS( 7 );
-      INIT_CHOOSERS( 8 );
-      INIT_CHOOSERS( 9 );
-      INIT_CHOOSERS( 10 );
-      INIT_CHOOSERS( 11 );
-      INIT_CHOOSERS( 12 );
-      INIT_CHOOSERS( 13 );
-      INIT_CHOOSERS( 14 );
-      INIT_CHOOSERS( 15 );
-
-      /* generic attributes */
-      INIT_CHOOSERS( 16 );
-      INIT_CHOOSERS( 17 );
-      INIT_CHOOSERS( 18 );
-      INIT_CHOOSERS( 19 );
-      INIT_CHOOSERS( 20 );
-      INIT_CHOOSERS( 21 );
-      INIT_CHOOSERS( 22 );
-      INIT_CHOOSERS( 23 );
-      INIT_CHOOSERS( 24 );
-      INIT_CHOOSERS( 25 );
-      INIT_CHOOSERS( 26 );
-      INIT_CHOOSERS( 27 );
-      INIT_CHOOSERS( 28 );
-      INIT_CHOOSERS( 29 );
-      INIT_CHOOSERS( 30 );
-      INIT_CHOOSERS( 31 );
-
-      choose[_TNL_ATTRIB_ERROR][0] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][1] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][2] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][3] = error_attrib;
-
-#ifdef USE_X86_ASM
-      if (tnl->AllowCodegen) {
-         _tnl_x86choosers(choose, do_choose); /* x86 INIT_CHOOSERS */
-      }
-#endif
-
-      _tnl_generic_attr_table_init( generic_attr_func );
-   }
-
-   for (i = 0; i < _TNL_ATTRIB_EDGEFLAG; i++)
-      _mesa_vector4f_init( &tmp->Attribs[i], 0, NULL);
-
-   for (i = 0; i < 4; i++) {
-      make_empty_list( &tnl->vtx.cache.Vertex[i] );
-      make_empty_list( &tnl->vtx.cache.Attribute[i] );
-      tnl->vtx.gen.Vertex[i] = no_codegen;
-      tnl->vtx.gen.Attribute[i] = no_codegen;
-   }
-
-#ifdef USE_X86_ASM
-   _tnl_InitX86Codegen( &tnl->vtx.gen );
-#endif
-
-   _tnl_current_init( ctx );
-   _tnl_exec_vtxfmt_init( ctx );
-   _tnl_generic_exec_vtxfmt_init( ctx );
-#ifdef USE_X86_ASM
-   if (tnl->AllowCodegen) {
-      _tnl_x86_exec_vtxfmt_init( ctx ); /* x86 DISPATCH_ATTRFV */
-   }
-#endif
-
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
-   _mesa_memcpy( tnl->vtx.tabfv, choose, sizeof(choose) );
-
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) {
-      tnl->vtx.attrsz[i] = 0;
-      tnl->vtx.active_sz[i] = 0;
-   }
-
-   tnl->vtx.vertex_size = 0;
-   tnl->vtx.have_materials = 0;
-}
-
-static void free_funcs( struct _tnl_dynfn *l )
-{
-   struct _tnl_dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      ALIGN_FREE( f->code );
-      FREE( f );
-   }
-}
-
-
-void _tnl_vtx_destroy( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   for (i = 0; i < 4; i++) {
-      free_funcs( &tnl->vtx.cache.Vertex[i] );
-      free_funcs( &tnl->vtx.cache.Attribute[i] ); 
-   }
-}
-
diff --git a/src/mesa/tnl/t_vtx_api.h b/src/mesa/tnl/t_vtx_api.h
deleted file mode 100644
index 5311f40a85..0000000000
--- a/src/mesa/tnl/t_vtx_api.h
+++ /dev/null
@@ -1,86 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#ifndef __T_VTX_API_H__
-#define __T_VTX_API_H__
-
-#include "t_context.h"
-
-
-/* t_vtx_api.c:
- */
-extern void _tnl_vtx_init( GLcontext *ctx );
-extern void _tnl_vtx_destroy( GLcontext *ctx );
-
-extern void _tnl_FlushVertices( GLcontext *ctx, GLuint flags );
-extern void _tnl_flush_vtx( GLcontext *ctx );
-
-extern void GLAPIENTRY _tnl_wrap_filled_vertex( GLcontext *ctx );
-
-/* t_vtx_exec.c:
- */
-
-extern void _tnl_do_EvalCoord2f( GLcontext* ctx, GLfloat u, GLfloat v );
-extern void _tnl_do_EvalCoord1f(GLcontext* ctx, GLfloat u);
-extern void _tnl_update_eval( GLcontext *ctx );
-
-extern GLboolean *_tnl_translate_edgeflag( GLcontext *ctx, 
-					   const GLfloat *data, 
-					   GLuint count,
-					   GLuint stride );
-
-extern GLboolean *_tnl_import_current_edgeflag( GLcontext *ctx,
-						GLuint count );
-
-
-
-/* t_vtx_generic.c:
- */
-extern void _tnl_generic_exec_vtxfmt_init( GLcontext *ctx );
-
-extern void _tnl_generic_attr_table_init( tnl_attrfv_func (*tab)[4] );
-
-/* t_vtx_x86.c:
- */
-extern void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen );
-
-extern void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx );
-
-extern void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-			      tnl_attrfv_func (*do_choose)( GLuint attr,
-							GLuint sz ));
-
-
-
-
-#endif
diff --git a/src/mesa/tnl/t_vtx_eval.c b/src/mesa/tnl/t_vtx_eval.c
deleted file mode 100644
index 71564076e3..0000000000
--- a/src/mesa/tnl/t_vtx_eval.c
+++ /dev/null
@@ -1,265 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.1
- *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "api_eval.h"
-#include "context.h"
-#include "macros.h"
-#include "math/m_eval.h"
-#include "t_vtx_api.h"
-#include "dispatch.h"
-
-
-static void clear_active_eval1( TNLcontext *tnl, GLuint attr ) 
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   tnl->vtx.eval.map1[attr].map = NULL;
-}
-
-static void clear_active_eval2( TNLcontext *tnl, GLuint attr ) 
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   tnl->vtx.eval.map2[attr].map = NULL;
-}
-
-static void set_active_eval1( TNLcontext *tnl, GLuint attr, GLuint dim, 
-			      struct gl_1d_map *map )
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   if (!tnl->vtx.eval.map1[attr].map) {
-      tnl->vtx.eval.map1[attr].map = map;
-      tnl->vtx.eval.map1[attr].sz = dim;
-   }
-} 
-
-static void set_active_eval2( TNLcontext *tnl, GLuint attr, GLuint dim, 
-			      struct gl_2d_map *map )
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   if (!tnl->vtx.eval.map2[attr].map) {
-      tnl->vtx.eval.map2[attr].map = map;
-      tnl->vtx.eval.map2[attr].sz = dim;
-   }
-} 
-
-void _tnl_update_eval( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   /* Vertex program maps have priority over conventional attribs */
-
-   for (attr = 0; attr < _TNL_NUM_EVAL; attr++) {
-      clear_active_eval1( tnl, attr );
-      clear_active_eval2( tnl, attr );
-   }
-
-   if (ctx->Eval.Map1Color4) 
-      set_active_eval1( tnl, VERT_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
-      
-   if (ctx->Eval.Map2Color4) 
-      set_active_eval2( tnl, VERT_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
-
-   if (ctx->Eval.Map1TextureCoord4) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
-   else if (ctx->Eval.Map1TextureCoord3) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
-   else if (ctx->Eval.Map1TextureCoord2) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
-   else if (ctx->Eval.Map1TextureCoord1) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
-
-   if (ctx->Eval.Map2TextureCoord4) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
-   else if (ctx->Eval.Map2TextureCoord3) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
-   else if (ctx->Eval.Map2TextureCoord2) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
-   else if (ctx->Eval.Map2TextureCoord1) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
-
-   if (ctx->Eval.Map1Normal) 
-      set_active_eval1( tnl, VERT_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
-
-   if (ctx->Eval.Map2Normal) 
-      set_active_eval2( tnl, VERT_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
-
-   if (ctx->Eval.Map1Vertex4) 
-      set_active_eval1( tnl, VERT_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
-   else if (ctx->Eval.Map1Vertex3) 
-      set_active_eval1( tnl, VERT_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
-
-   if (ctx->Eval.Map2Vertex4) 
-      set_active_eval2( tnl, VERT_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
-   else if (ctx->Eval.Map2Vertex3) 
-      set_active_eval2( tnl, VERT_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
-
-   /* Evaluators with generic attributes is only supported for NV vertex
-    * programs, not ARB vertex programs.  16 evaluator maps are supported.
-    * We do this after the conventional attributes since the spec says that
-    * these generic maps have higher priority.
-    */
-   if (ctx->VertexProgram._Enabled &&
-       ctx->VertexProgram._Current &&
-       ctx->VertexProgram._Current->IsNVProgram) {
-      for (attr = 0; attr < 16; attr++) {
-	 if (ctx->Eval.Map1Attrib[attr]) 
-	    set_active_eval1( tnl, attr, 4, &ctx->EvalMap.Map1Attrib[attr] );
-
-	 if (ctx->Eval.Map2Attrib[attr]) 
-	    set_active_eval2( tnl, attr, 4, &ctx->EvalMap.Map2Attrib[attr] );
-      }
-   }
-
-   tnl->vtx.eval.new_state = 0;
-}
-
-
-
-void _tnl_do_EvalCoord1f(GLcontext* ctx, GLfloat u)
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   for (attr = 1; attr < _TNL_NUM_EVAL; attr++) {
-      struct gl_1d_map *map = tnl->vtx.eval.map1[attr].map;
-      if (map) {
-	 GLfloat uu = (u - map->u1) * map->du;
-	 GLfloat data[4];
-
-	 ASSIGN_4V(data, 0, 0, 0, 1);
-
-	 _math_horner_bezier_curve(map->Points, data, uu, 
-				   tnl->vtx.eval.map1[attr].sz, 
-				   map->Order);
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[attr],
-		     tnl->vtx.attrsz[attr],
-		     data );
-      }
-   }
-
-   /** Vertex -- EvalCoord1f is a noop if this map not enabled:
-    **/
-   if (tnl->vtx.eval.map1[0].map) {
-      struct gl_1d_map *map = tnl->vtx.eval.map1[0].map;
-      GLfloat uu = (u - map->u1) * map->du;
-      GLfloat vertex[4];
-
-      ASSIGN_4V(vertex, 0, 0, 0, 1);
-
-      _math_horner_bezier_curve(map->Points, vertex, uu, 
-				tnl->vtx.eval.map1[0].sz, 
-				map->Order);
-
-      if (tnl->vtx.eval.map1[0].sz == 4) 
-	 CALL_Vertex4fv(GET_DISPATCH(), ( vertex ));
-      else
-	 CALL_Vertex3fv(GET_DISPATCH(), ( vertex )); 
-   }
-}
-
-
-
-void _tnl_do_EvalCoord2f( GLcontext* ctx, GLfloat u, GLfloat v )
-{   
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   for (attr = 1; attr < _TNL_NUM_EVAL; attr++) {
-      struct gl_2d_map *map = tnl->vtx.eval.map2[attr].map;
-      if (map) {
-	 GLfloat uu = (u - map->u1) * map->du;
-	 GLfloat vv = (v - map->v1) * map->dv;
-	 GLfloat data[4];
-
-	 ASSIGN_4V(data, 0, 0, 0, 1);
-
-	 _math_horner_bezier_surf(map->Points, 
-				  data, 
-				  uu, vv, 
-				  tnl->vtx.eval.map2[attr].sz, 
-				  map->Uorder, map->Vorder);
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[attr],
-		     tnl->vtx.attrsz[attr],
-		     data );
-      }
-   }
-
-   /** Vertex -- EvalCoord2f is a noop if this map not enabled:
-    **/
-   if (tnl->vtx.eval.map2[0].map) {
-      struct gl_2d_map *map = tnl->vtx.eval.map2[0].map;
-      GLfloat uu = (u - map->u1) * map->du;
-      GLfloat vv = (v - map->v1) * map->dv;
-      GLfloat vertex[4];
-
-      ASSIGN_4V(vertex, 0, 0, 0, 1);
-
-      if (ctx->Eval.AutoNormal) {
-	 GLfloat normal[4];
-         GLfloat du[4], dv[4];
-
-         _math_de_casteljau_surf(map->Points, vertex, du, dv, uu, vv, 
-				 tnl->vtx.eval.map2[0].sz,
-				 map->Uorder, map->Vorder);
-
-	 if (tnl->vtx.eval.map2[0].sz == 4) {
-	    du[0] = du[0]*vertex[3] - du[3]*vertex[0];
-	    du[1] = du[1]*vertex[3] - du[3]*vertex[1];
-	    du[2] = du[2]*vertex[3] - du[3]*vertex[2];
-	 
-	    dv[0] = dv[0]*vertex[3] - dv[3]*vertex[0];
-	    dv[1] = dv[1]*vertex[3] - dv[3]*vertex[1];
-	    dv[2] = dv[2]*vertex[3] - dv[3]*vertex[2];
-	 }
-
-
-         CROSS3(normal, du, dv);
-         NORMALIZE_3FV(normal);
-	 normal[3] = 1.0;
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[_TNL_ATTRIB_NORMAL],
-		     tnl->vtx.attrsz[_TNL_ATTRIB_NORMAL],
-		     normal );
-
-      }
-      else {
-         _math_horner_bezier_surf(map->Points, vertex, uu, vv, 
-				  tnl->vtx.eval.map2[0].sz,
-				  map->Uorder, map->Vorder);
-      }
-
-      if (tnl->vtx.attrsz[0] == 4) 
-	 CALL_Vertex4fv(GET_DISPATCH(), ( vertex ));
-      else
-	 CALL_Vertex3fv(GET_DISPATCH(), ( vertex )); 
-   }
-}
-
-
diff --git a/src/mesa/tnl/t_vtx_exec.c b/src/mesa/tnl/t_vtx_exec.c
deleted file mode 100644
index 900c4ab6cc..0000000000
--- a/src/mesa/tnl/t_vtx_exec.c
+++ /dev/null
@@ -1,293 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "api_eval.h"
-#include "context.h"
-#include "enums.h"
-#include "state.h"
-#include "macros.h"
-#include "math/m_eval.h"
-#include "t_vtx_api.h"
-#include "t_pipeline.h"
-
-
-static void _tnl_print_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint count = tnl->vtx.initial_counter - tnl->vtx.counter;
-   GLuint i;
-
-   _mesa_debug(ctx, "_tnl_print_vtx: %u vertices %d primitives, %d vertsize\n",
-               count,
-	       tnl->vtx.prim_count,
-	       tnl->vtx.vertex_size);
-
-   for (i = 0 ; i < tnl->vtx.prim_count ; i++) {
-      struct tnl_prim *prim = &tnl->vtx.prim[i];
-      _mesa_debug(NULL, "   prim %d: %s %d..%d %s %s\n",
-		  i, 
-		  _mesa_lookup_enum_by_nr(prim->mode & PRIM_MODE_MASK),
-		  prim->start, 
-		  prim->start + prim->count,
-		  (prim->mode & PRIM_BEGIN) ? "BEGIN" : "(wrap)",
-		  (prim->mode & PRIM_END) ? "END" : "(wrap)");
-   }
-}
-
-GLboolean *_tnl_translate_edgeflag( GLcontext *ctx, const GLfloat *data,
-				    GLuint count, GLuint stride )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLboolean *ef = tnl->vtx.edgeflag_tmp;
-   GLuint i;
-
-   if (!ef) 
-      ef = tnl->vtx.edgeflag_tmp = (GLboolean *) MALLOC( tnl->vb.Size );
-   
-   for (i = 0 ; i < count ; i++, data += stride)
-      ef[i] = (data[0] == 1.0);
-
-   return ef;
-}
-
-
-GLboolean *_tnl_import_current_edgeflag( GLcontext *ctx,
-					 GLuint count )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLboolean *ef = tnl->vtx.edgeflag_tmp;
-   GLboolean tmp = ctx->Current.EdgeFlag;
-   GLuint i;
-
-   if (!ef) 
-      ef = tnl->vtx.edgeflag_tmp = (GLboolean *) MALLOC( tnl->vb.Size );
-
-   for (i = 0 ; i < count ; i++)
-      ef[i] = tmp;
-
-   return ef;
-}
-
-static INLINE GLint get_size( const GLfloat *f )
-{
-   if (f[3] != 1.0) return 4;
-   if (f[2] != 0.0) return 3;
-   return 2;
-}
-
-/* Some nasty stuff still hanging on here.  
- *
- * TODO - remove VB->NormalPtr, etc and just use the AttrPtr's.
- */
-static void _tnl_vb_bind_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->vtx_inputs;
-   GLfloat *data = tnl->vtx.buffer;
-   GLuint count = tnl->vtx.initial_counter - tnl->vtx.counter;
-   GLuint attr, i;
-
-#undef DEBUG_VTX
-
-#ifdef DEBUG_VTX
-   fprintf(stderr, "_tnl_vb_bind_vtx(): %d verts %d vertsize\n",
-		  count, tnl->vtx.vertex_size);
-#endif
-
-
-   /* Setup constant data in the VB.
-    */
-   VB->Count = count;
-   VB->Primitive = tnl->vtx.prim;
-   VB->PrimitiveCount = tnl->vtx.prim_count;
-   VB->Elts = NULL;
-   VB->NormalLengthPtr = NULL;
-
-   for (attr = 0; attr <= _TNL_ATTRIB_EDGEFLAG ; attr++) {
-      if (tnl->vtx.attrsz[attr]) {
-	 tmp->Attribs[attr].count = count;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) data;
-	 tmp->Attribs[attr].start = data;
-	 tmp->Attribs[attr].size = tnl->vtx.attrsz[attr];
-	 tmp->Attribs[attr].stride = tnl->vtx.vertex_size * sizeof(GLfloat);
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-	 data += tnl->vtx.attrsz[attr];
-      }
-      else {
-/* 	 VB->AttribPtr[attr] = &tnl->current.Attribs[attr]; */
-
-
-	 tmp->Attribs[attr].count = 1;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) tnl->vtx.current[attr];
-	 tmp->Attribs[attr].start = tnl->vtx.current[attr];
-	 tmp->Attribs[attr].size = get_size( tnl->vtx.current[attr] );
-	 tmp->Attribs[attr].stride = 0;
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-      }
-   }
-
-   
-   /* Copy and translate EdgeFlag to a contiguous array of GLbooleans
-    */
-   if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL) {
-      if (tnl->vtx.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-	 VB->EdgeFlag = _tnl_translate_edgeflag( ctx, data, count,
-						 tnl->vtx.vertex_size );
-	 data++;
-      }
-      else 
-	 VB->EdgeFlag = _tnl_import_current_edgeflag( ctx, count );
-   }
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
-
-
-/*
- * NOTE: Need to have calculated primitives by this point -- do it on the fly.
- * NOTE: Old 'parity' issue is gone.
- */
-static GLuint _tnl_copy_vertices( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT( ctx );
-   GLuint nr = tnl->vtx.prim[tnl->vtx.prim_count-1].count;
-   GLuint ovf, i;
-   GLuint sz = tnl->vtx.vertex_size;
-   GLfloat *dst = tnl->vtx.copied.buffer;
-   GLfloat *src = (tnl->vtx.buffer + 
-		   tnl->vtx.prim[tnl->vtx.prim_count-1].start * 
-		   tnl->vtx.vertex_size);
-
-
-   switch( ctx->Driver.CurrentExecPrimitive )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      else {
-	 _mesa_memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 1;
-      }
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 return 1;
-      } else {
-	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 _mesa_memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      /* no parity issue, but need to make sure the tri is not drawn twice */
-      if (nr & 1) {
-	 tnl->vtx.prim[tnl->vtx.prim_count-1].count--;
-      }
-      /* fallthrough */
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case PRIM_OUTSIDE_BEGIN_END:
-      return 0;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-
-/**
- * Execute the buffer and save copied verts.
- */
-void _tnl_flush_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint vertex_count = tnl->vtx.initial_counter - tnl->vtx.counter;
-
-   if (0)
-      _tnl_print_vtx( ctx );
-
-   if (tnl->vtx.prim_count && vertex_count) {
-
-      tnl->vtx.copied.nr = _tnl_copy_vertices( ctx ); 
-
-      if (tnl->vtx.copied.nr != vertex_count) {
-	 if (ctx->NewState)
-	    _mesa_update_state( ctx );
-      
-	 _tnl_vb_bind_vtx( ctx );
-
-	 tnl->Driver.RunPipeline( ctx );
-      }
-   }
-
-   tnl->vtx.prim_count = 0;
-   tnl->vtx.counter = tnl->vtx.initial_counter;
-   tnl->vtx.vbptr = tnl->vtx.buffer;
-}
diff --git a/src/mesa/tnl/t_vtx_generic.c b/src/mesa/tnl/t_vtx_generic.c
deleted file mode 100644
index 28de445ed4..0000000000
--- a/src/mesa/tnl/t_vtx_generic.c
+++ /dev/null
@@ -1,615 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-
-
-/* Versions of all the entrypoints for situations where codegen isn't
- * available.  
- *
- * Note: Only one size for each attribute may be active at once.
- * Eg. if Color3f is installed/active, then Color4f may not be, even
- * if the vertex actually contains 4 color coordinates.  This is
- * because the 3f version won't otherwise set color[3] to 1.0 -- this
- * is the job of the chooser function when switching between Color4f
- * and Color3f.
- */
-#define ATTRFV( ATTR, N )				\
-static void attrib_##ATTR##_##N( const GLfloat *v )	\
-{							\
-   GET_CURRENT_CONTEXT( ctx );				\
-   TNLcontext *tnl = TNL_CONTEXT(ctx);			\
-							\
-   if ((ATTR) == 0) {					\
-      GLuint i;						\
-							\
-      if (N>0) tnl->vtx.vbptr[0] = v[0];		\
-      if (N>1) tnl->vtx.vbptr[1] = v[1];		\
-      if (N>2) tnl->vtx.vbptr[2] = v[2];		\
-      if (N>3) tnl->vtx.vbptr[3] = v[3];		\
-							\
-      for (i = N; i < tnl->vtx.vertex_size; i++)	\
-	 tnl->vtx.vbptr[i] = tnl->vtx.vertex[i];	\
-							\
-      tnl->vtx.vbptr += tnl->vtx.vertex_size;		\
-							\
-      if (--tnl->vtx.counter == 0)			\
-	 _tnl_wrap_filled_vertex( ctx );		\
-   }							\
-   else {						\
-      GLfloat *dest = tnl->vtx.attrptr[ATTR];		\
-      if (N>0) dest[0] = v[0];				\
-      if (N>1) dest[1] = v[1];				\
-      if (N>2) dest[2] = v[2];				\
-      if (N>3) dest[3] = v[3];				\
-   }							\
-}
-
-#define INIT(TAB, ATTR)						\
-   TAB[ATTR][0] = attrib_##ATTR##_1;				\
-   TAB[ATTR][1] = attrib_##ATTR##_2;				\
-   TAB[ATTR][2] = attrib_##ATTR##_3;				\
-   TAB[ATTR][3] = attrib_##ATTR##_4;
-
-
-#define ATTRS( ATTRIB )				\
-   ATTRFV( ATTRIB, 1 )				\
-   ATTRFV( ATTRIB, 2 )				\
-   ATTRFV( ATTRIB, 3 )				\
-   ATTRFV( ATTRIB, 4 )			
-
-/* conventional attribs */
-ATTRS( 0 )
-ATTRS( 1 )
-ATTRS( 2 )
-ATTRS( 3 )
-ATTRS( 4 )
-ATTRS( 5 )
-ATTRS( 6 )
-ATTRS( 7 )
-ATTRS( 8 )
-ATTRS( 9 )
-ATTRS( 10 )
-ATTRS( 11 )
-ATTRS( 12 )
-ATTRS( 13 )
-ATTRS( 14 )
-ATTRS( 15 )
-
-/* generic attribs */
-ATTRS( 16 )
-ATTRS( 17 )
-ATTRS( 18 )
-ATTRS( 19 )
-ATTRS( 20 )
-ATTRS( 21 )
-ATTRS( 22 )
-ATTRS( 23 )
-ATTRS( 24 )
-ATTRS( 25 )
-ATTRS( 26 )
-ATTRS( 27 )
-ATTRS( 28 )
-ATTRS( 29 )
-ATTRS( 30 )
-ATTRS( 31 )
-
-
-void _tnl_generic_attr_table_init( tnl_attrfv_func (*tab)[4] )
-{
-   /* conventional attribs */
-   INIT( tab, 0 );
-   INIT( tab, 1 );
-   INIT( tab, 2 );
-   INIT( tab, 3 );
-   INIT( tab, 4 );
-   INIT( tab, 5 );
-   INIT( tab, 6 );
-   INIT( tab, 7 );
-   INIT( tab, 8 );
-   INIT( tab, 9 );
-   INIT( tab, 10 );
-   INIT( tab, 11 );
-   INIT( tab, 12 );
-   INIT( tab, 13 );
-   INIT( tab, 14 );
-   INIT( tab, 15 );
-
-   /* generic attribs */
-   INIT( tab, 16 );
-   INIT( tab, 17 );
-   INIT( tab, 18 );
-   INIT( tab, 19 );
-   INIT( tab, 20 );
-   INIT( tab, 21 );
-   INIT( tab, 22 );
-   INIT( tab, 23 );
-   INIT( tab, 24 );
-   INIT( tab, 25 );
-   INIT( tab, 26 );
-   INIT( tab, 27 );
-   INIT( tab, 28 );
-   INIT( tab, 29 );
-   INIT( tab, 30 );
-   INIT( tab, 31 );
-}
-
-/* These can be made efficient with codegen.  Further, by adding more
- * logic to do_choose(), the double-dispatch for legacy entrypoints
- * like glVertex3f() can be removed.
- */
-#define DISPATCH_ATTRFV( ATTR, COUNT, P )	\
-do {						\
-   GET_CURRENT_CONTEXT( ctx ); 			\
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 		\
-   tnl->vtx.tabfv[ATTR][COUNT-1]( P );		\
-} while (0)
-
-#define DISPATCH_ATTR1FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 1, V )
-#define DISPATCH_ATTR2FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 2, V )
-#define DISPATCH_ATTR3FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 3, V )
-#define DISPATCH_ATTR4FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 4, V )
-
-#define DISPATCH_ATTR1F( ATTR, S ) DISPATCH_ATTRFV( ATTR, 1, &(S) )
-
-#if defined(USE_X86_ASM) && 0 /* will break register calling convention */
-/* Naughty cheat:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) DISPATCH_ATTRFV( ATTR, 2, &(S) )
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) DISPATCH_ATTRFV( ATTR, 3, &(S) )
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q ) DISPATCH_ATTRFV( ATTR, 4, &(S) )
-#else
-/* Safe:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) 		\
-do { 						\
-   GLfloat v[2]; 				\
-   v[0] = S; v[1] = T;				\
-   DISPATCH_ATTR2FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) 		\
-do { 						\
-   GLfloat v[3]; 				\
-   v[0] = S; v[1] = T; v[2] = R;		\
-   DISPATCH_ATTR3FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q )	\
-do { 						\
-   GLfloat v[4]; 				\
-   v[0] = S; v[1] = T; v[2] = R; v[3] = Q;	\
-   DISPATCH_ATTR4FV( ATTR, v );			\
-} while (0)
-#endif
-
-
-static void GLAPIENTRY _tnl_Vertex2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_POS, x, y );
-}
-
-static void GLAPIENTRY _tnl_Vertex2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_POS, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Vertex3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_Vertex4f( GLfloat x, GLfloat y, GLfloat z, 
-				      GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_POS, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_Vertex4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord1f( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_TEX0, x );
-}
-
-static void GLAPIENTRY _tnl_TexCoord1fv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_TEX0, x, y );
-}
-
-static void GLAPIENTRY _tnl_TexCoord2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_TEX0, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_TexCoord3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord4f( GLfloat x, GLfloat y, GLfloat z,
-					GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_TEX0, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_TexCoord4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_Normal3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_NORMAL, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Normal3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_NORMAL, v );
-}
-
-static void GLAPIENTRY _tnl_FogCoordfEXT( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_FOG, x );
-}
-
-static void GLAPIENTRY _tnl_FogCoordfvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_FOG, v );
-}
-
-static void GLAPIENTRY _tnl_Color3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR0, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Color3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _tnl_Color4f( GLfloat x, GLfloat y, GLfloat z, 
-				     GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_COLOR0, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_Color4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _tnl_Indexf( GLfloat v )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_COLOR_INDEX, v );
-}
-
-static void GLAPIENTRY _tnl_Indexfv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_COLOR_INDEX, v );
-}
-
-static void GLAPIENTRY _tnl_SecondaryColor3fEXT( GLfloat x, GLfloat y, 
-						 GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR1, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_SecondaryColor3fvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR1, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord1f( GLenum target, GLfloat x  )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1F( attr, x );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord1fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord2f( GLenum target, GLfloat x, GLfloat y )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2F( attr, x, y );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord2fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord3f( GLenum target, GLfloat x, 
-					     GLfloat y, GLfloat z)
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3F( attr, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord3fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord4f( GLenum target, GLfloat x, 
-					     GLfloat y, GLfloat z,
-					     GLfloat w )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4F( attr, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord4fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4FV( attr, v );
-}
-
-
-/**
- * GL_NV_vertex_program Vertex Attributes
- * Note that these attributes DO alias the conventional attributes.
- * Also, calling glVertexAttribNV(0, xxx) is equivalent to glVertex(xxx).
- */
-
-static void GLAPIENTRY _tnl_VertexAttrib1fNV( GLuint index, GLfloat x )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR1F( index, x );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib1fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR1FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fNV(GLuint index, GLfloat x, GLfloat y)
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR2F( index, x, y );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR2FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fNV( GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR3F( index, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR3FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fNV( GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z,
-					      GLfloat w )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR4F( index, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR4FV( index, v );
-}
-
-
-
-/**
- * GL_ARB_vertex_program Vertex Attributes
- * Note that these attributes do NOT alias the conventional attributes.
- * Also, calling glVertexAttribARB(0, xxx) is equivalent to glVertex(xxx).
- */
-
-static void GLAPIENTRY _tnl_VertexAttrib1fARB( GLuint index, GLfloat x )
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR1F( index, x );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib1fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR1FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fARB( GLuint index, GLfloat x, 
-					      GLfloat y )
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR2F( index, x, y );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR2FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fARB(GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR3F( index, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR3FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fARB(GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z, GLfloat w)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR4F( index, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR4FV( index, v );
-}
-
-
-/* Install the generic versions of the 2nd level dispatch
- * functions.  Some of these have a codegen alternative.
- */
-void _tnl_generic_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   vfmt->Color3f = _tnl_Color3f;
-   vfmt->Color3fv = _tnl_Color3fv;
-   vfmt->Color4f = _tnl_Color4f;
-   vfmt->Color4fv = _tnl_Color4fv;
-   vfmt->Indexf = _tnl_Indexf;
-   vfmt->Indexfv = _tnl_Indexfv;
-   vfmt->FogCoordfEXT = _tnl_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = _tnl_FogCoordfvEXT;
-   vfmt->MultiTexCoord1fARB = _tnl_MultiTexCoord1f;
-   vfmt->MultiTexCoord1fvARB = _tnl_MultiTexCoord1fv;
-   vfmt->MultiTexCoord2fARB = _tnl_MultiTexCoord2f;
-   vfmt->MultiTexCoord2fvARB = _tnl_MultiTexCoord2fv;
-   vfmt->MultiTexCoord3fARB = _tnl_MultiTexCoord3f;
-   vfmt->MultiTexCoord3fvARB = _tnl_MultiTexCoord3fv;
-   vfmt->MultiTexCoord4fARB = _tnl_MultiTexCoord4f;
-   vfmt->MultiTexCoord4fvARB = _tnl_MultiTexCoord4fv;
-   vfmt->Normal3f = _tnl_Normal3f;
-   vfmt->Normal3fv = _tnl_Normal3fv;
-   vfmt->SecondaryColor3fEXT = _tnl_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = _tnl_SecondaryColor3fvEXT;
-   vfmt->TexCoord1f = _tnl_TexCoord1f;
-   vfmt->TexCoord1fv = _tnl_TexCoord1fv;
-   vfmt->TexCoord2f = _tnl_TexCoord2f;
-   vfmt->TexCoord2fv = _tnl_TexCoord2fv;
-   vfmt->TexCoord3f = _tnl_TexCoord3f;
-   vfmt->TexCoord3fv = _tnl_TexCoord3fv;
-   vfmt->TexCoord4f = _tnl_TexCoord4f;
-   vfmt->TexCoord4fv = _tnl_TexCoord4fv;
-   vfmt->Vertex2f = _tnl_Vertex2f;
-   vfmt->Vertex2fv = _tnl_Vertex2fv;
-   vfmt->Vertex3f = _tnl_Vertex3f;
-   vfmt->Vertex3fv = _tnl_Vertex3fv;
-   vfmt->Vertex4f = _tnl_Vertex4f;
-   vfmt->Vertex4fv = _tnl_Vertex4fv;
-   vfmt->VertexAttrib1fNV = _tnl_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = _tnl_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV = _tnl_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = _tnl_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV = _tnl_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = _tnl_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV = _tnl_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = _tnl_VertexAttrib4fvNV;
-   vfmt->VertexAttrib1fARB = _tnl_VertexAttrib1fARB;
-   vfmt->VertexAttrib1fvARB = _tnl_VertexAttrib1fvARB;
-   vfmt->VertexAttrib2fARB = _tnl_VertexAttrib2fARB;
-   vfmt->VertexAttrib2fvARB = _tnl_VertexAttrib2fvARB;
-   vfmt->VertexAttrib3fARB = _tnl_VertexAttrib3fARB;
-   vfmt->VertexAttrib3fvARB = _tnl_VertexAttrib3fvARB;
-   vfmt->VertexAttrib4fARB = _tnl_VertexAttrib4fARB;
-   vfmt->VertexAttrib4fvARB = _tnl_VertexAttrib4fvARB;
-}
diff --git a/src/mesa/tnl/t_vtx_x86.c b/src/mesa/tnl/t_vtx_x86.c
deleted file mode 100644
index 99a0550932..0000000000
--- a/src/mesa/tnl/t_vtx_x86.c
+++ /dev/null
@@ -1,396 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Daniel Borca <dborca@yahoo.com>
- */
-
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-#include "simple_list.h"
-
-#if defined(USE_X86_ASM) && !defined(HAVE_NONSTANDARD_GLAPIENTRY)
-
-#define EXTERN( FUNC )		\
-extern const char FUNC[];	\
-extern const char FUNC##_end[]
-
-EXTERN( _tnl_x86_Attribute1fv );
-EXTERN( _tnl_x86_Attribute2fv );
-EXTERN( _tnl_x86_Attribute3fv );
-EXTERN( _tnl_x86_Attribute4fv );
-EXTERN( _tnl_x86_Vertex1fv );
-EXTERN( _tnl_x86_Vertex2fv );
-EXTERN( _tnl_x86_Vertex3fv );
-EXTERN( _tnl_x86_Vertex4fv );
-
-EXTERN( _tnl_x86_dispatch_attrf1 );
-EXTERN( _tnl_x86_dispatch_attrf2 );
-EXTERN( _tnl_x86_dispatch_attrf3 );
-EXTERN( _tnl_x86_dispatch_attrf4 );
-EXTERN( _tnl_x86_dispatch_attrfv );
-EXTERN( _tnl_x86_dispatch_multitexcoordf1 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf2 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf3 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf4 );
-EXTERN( _tnl_x86_dispatch_multitexcoordfv );
-EXTERN( _tnl_x86_dispatch_vertexattribf1 );
-EXTERN( _tnl_x86_dispatch_vertexattribf2 );
-EXTERN( _tnl_x86_dispatch_vertexattribf3 );
-EXTERN( _tnl_x86_dispatch_vertexattribf4 );
-EXTERN( _tnl_x86_dispatch_vertexattribfv );
-
-EXTERN( _tnl_x86_choose_fv );
-
-
-#define DONT_KNOW_OFFSETS 1
-
-
-#define DFN( FUNC, CACHE, KEY )				\
-   struct _tnl_dynfn *dfn = MALLOC_STRUCT( _tnl_dynfn );\
-   const char *start = FUNC;				\
-   const char *end = FUNC##_end;			\
-   int offset = 0;               			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key = KEY;					\
-   dfn->code = ALIGN_MALLOC( end - start, 16 );		\
-   _mesa_memcpy (dfn->code, start, end - start)
-
-
-
-#undef DEBUG_VTX
-
-#ifdef DEBUG_VTX
-#define FIXUP_PRINTF( offset, NEWVAL ) \
-      fprintf(stderr, "%s/%d: offset %d, new value: 0x%x\n", __FILE__, __LINE__, offset, (int)(NEWVAL))
-#define FIXUPREL_PRINTF( offset, NEWVAL, CODE ) \
-      fprintf(stderr, "%s/%d: offset %d, new value: 0x%x\n", __FILE__, __LINE__, offset, (int)(NEWVAL) - ((int)(CODE)+offset) - 4)
-#else
-#define FIXUP_PRINTF( offset, NEWVAL )
-#define FIXUPREL_PRINTF( offset, NEWVAL, CODE )
-#endif
-
-#define FIXUP( CODE, KNOWN_OFFSET, CHECKVAL, NEWVAL )	\
-do {							\
-   GLint subst = 0x10101010 + CHECKVAL;			\
-							\
-   if (DONT_KNOW_OFFSETS) {				\
-      while (*(int *)(CODE+offset) != subst) offset++;	\
-      *(int *)(CODE+offset) = (int)(NEWVAL);		\
-      FIXUP_PRINTF(offset, NEWVAL);			\
-      offset += 4;					\
-   }							\
-   else {						\
-      int *icode = (int *)(CODE+KNOWN_OFFSET);		\
-      assert (*icode == subst);				\
-      *icode = (int)NEWVAL;				\
-   }							\
-} while (0)
-
-
-
-#define FIXUPREL( CODE, KNOWN_OFFSET, CHECKVAL, NEWVAL )\
-do {							\
-   GLint subst = 0x10101010 + CHECKVAL;			\
-							\
-   if (DONT_KNOW_OFFSETS) {				\
-      while (*(int *)(CODE+offset) != subst) offset++;	\
-      *(int *)(CODE+offset) = (int)(NEWVAL) - ((int)(CODE)+offset) - 4; \
-      FIXUPREL_PRINTF(offset, NEWVAL, CODE);		\
-      offset += 4;					\
-   }							\
-   else {						\
-      int *icode = (int *)(CODE+KNOWN_OFFSET);		\
-      assert (*icode == subst);				\
-      *icode = (int)(NEWVAL) - (int)(icode) - 4;	\
-   }							\
-} while (0)
-
-
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-static struct _tnl_dynfn *makeX86Vertex1fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex1fv, tnl->vtx.cache.Vertex[1-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 1);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[1]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Vertex2fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex2fv, tnl->vtx.cache.Vertex[2-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 2);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[2]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Vertex3fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (vertex_size) {
-      default: {
-         DFN ( _tnl_x86_Vertex3fv, tnl->vtx.cache.Vertex[3-1], vertex_size );
-
-         FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-         FIXUP(dfn->code, 0, 1, vertex_size - 3);
-         FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[3]);
-         FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-         FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-         FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-         FIXUP(dfn->code, 0, 4, (int)ctx);
-         FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-         return dfn;
-      }
-   }
-}
-
-static struct _tnl_dynfn *makeX86Vertex4fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex4fv, tnl->vtx.cache.Vertex[4-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 4);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[4]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-
-static struct _tnl_dynfn *makeX86Attribute1fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute1fv, tnl->vtx.cache.Attribute[1-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute2fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute2fv, tnl->vtx.cache.Attribute[2-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute3fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute3fv, tnl->vtx.cache.Attribute[3-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-   FIXUP(dfn->code, 0, 2, 8+dest);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute4fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute4fv, tnl->vtx.cache.Attribute[4-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-   FIXUP(dfn->code, 0, 2, 8+dest);
-   FIXUP(dfn->code, 0, 3, 12+dest);
-
-   return dfn;
-}
-
-
-void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen )
-{
-   gen->Vertex[0] = makeX86Vertex1fv;
-   gen->Vertex[1] = makeX86Vertex2fv;
-   gen->Vertex[2] = makeX86Vertex3fv;
-   gen->Vertex[3] = makeX86Vertex4fv;
-   gen->Attribute[0] = makeX86Attribute1fv;
-   gen->Attribute[1] = makeX86Attribute2fv;
-   gen->Attribute[2] = makeX86Attribute3fv;
-   gen->Attribute[3] = makeX86Attribute4fv;
-}
-
-
-#define MKDISP(FUNC, SIZE, ATTR, WARP)					\
-do {									\
-   char *code;								\
-   const char *start = WARP;						\
-   const char *end = WARP##_end;					\
-   int offset = 0;							\
-   code = ALIGN_MALLOC( end - start, 16 );				\
-   _mesa_memcpy (code, start, end - start);					\
-   FIXUP(code, 0, 0, (int)&(TNL_CONTEXT(ctx)->vtx.tabfv[ATTR][SIZE-1]));\
-   *(void **)&vfmt->FUNC = code;					\
-} while (0)
-
-
-/* Install the codegen'ed versions of the 2nd level dispatch
- * functions.  We should keep a list and free them in the end...
- */
-void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   MKDISP(Color3f,             3, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrf3);
-   MKDISP(Color3fv,            3, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrfv);
-   MKDISP(Color4f,             4, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrf4);
-   MKDISP(Color4fv,            4, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrfv);
-   MKDISP(FogCoordfEXT,        1, _TNL_ATTRIB_FOG,    _tnl_x86_dispatch_attrf1);
-   MKDISP(FogCoordfvEXT,       1, _TNL_ATTRIB_FOG,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Normal3f,            3, _TNL_ATTRIB_NORMAL, _tnl_x86_dispatch_attrf3);
-   MKDISP(Normal3fv,           3, _TNL_ATTRIB_NORMAL, _tnl_x86_dispatch_attrfv);
-   MKDISP(SecondaryColor3fEXT, 3, _TNL_ATTRIB_COLOR1, _tnl_x86_dispatch_attrf3);
-   MKDISP(SecondaryColor3fvEXT,3, _TNL_ATTRIB_COLOR1, _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord1f,          1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf1);
-   MKDISP(TexCoord1fv,         1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord2f,          2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf2);
-   MKDISP(TexCoord2fv,         2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord3f,          3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf3);
-   MKDISP(TexCoord3fv,         3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord4f,          4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf4);
-   MKDISP(TexCoord4fv,         4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex2f,            2, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf2);
-   MKDISP(Vertex2fv,           2, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex3f,            3, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf3);
-   MKDISP(Vertex3fv,           3, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex4f,            4, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf4);
-   MKDISP(Vertex4fv,           4, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-
-   MKDISP(MultiTexCoord1fARB,  1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf1);
-   MKDISP(MultiTexCoord1fvARB, 1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord2fARB,  2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf2);
-   MKDISP(MultiTexCoord2fvARB, 2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord3fARB,  3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf3);
-   MKDISP(MultiTexCoord3fvARB, 3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord4fARB,  4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf4);
-   MKDISP(MultiTexCoord4fvARB, 4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-
-   MKDISP(VertexAttrib1fNV,    1, 0,                  _tnl_x86_dispatch_vertexattribf1);
-   MKDISP(VertexAttrib1fvNV,   1, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib2fNV,    2, 0,                  _tnl_x86_dispatch_vertexattribf2);
-   MKDISP(VertexAttrib2fvNV,   2, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib3fNV,    3, 0,                  _tnl_x86_dispatch_vertexattribf3);
-   MKDISP(VertexAttrib3fvNV,   3, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib4fNV,    4, 0,                  _tnl_x86_dispatch_vertexattribf4);
-   MKDISP(VertexAttrib4fvNV,   4, 0,                  _tnl_x86_dispatch_vertexattribfv);
-}
-
-
-/* Install the codegen'ed choosers.
- * We should keep a list and free them in the end...
- */
-void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-		       tnl_attrfv_func (*do_choose)( GLuint attr,
-						 GLuint sz ))
-{
-   int attr, size;
-
-   for (attr = 0; attr < _TNL_MAX_ATTR_CODEGEN; attr++) {
-      for (size = 0; size < 4; size++) {
-         char *code;
-         const char *start = _tnl_x86_choose_fv;
-         const char *end = _tnl_x86_choose_fv_end;
-         int offset = 0;
-         code = ALIGN_MALLOC( end - start, 16 );
-         _mesa_memcpy (code, start, end - start);
-         FIXUP(code, 0, 0, attr);
-         FIXUP(code, 0, 1, size + 1);
-         FIXUPREL(code, 0, 2, do_choose);
-         choose[attr][size] = (tnl_attrfv_func)code;
-      }
-   }
-}
-
-#else 
-
-void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen )
-{
-   (void) gen;
-}
-
-
-void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx )
-{
-   (void) ctx;
-}
-
-
-void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-		       tnl_attrfv_func (*do_choose)( GLuint attr,
-						 GLuint sz ))
-{
-   (void) choose;
-   (void) do_choose;
-}
-
-#endif
diff --git a/src/mesa/tnl/t_vtx_x86_gcc.S b/src/mesa/tnl/t_vtx_x86_gcc.S
deleted file mode 100644
index 93753be759..0000000000
--- a/src/mesa/tnl/t_vtx_x86_gcc.S
+++ /dev/null
@@ -1,561 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 Tungsten Graphics Inc., Cedar Park, Texas.
-
-All Rights Reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a
-copy of this software and associated documentation files (the "Software"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, and/or sell copies of the Software, and to permit persons to whom
-the Software is furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice (including the next
-paragraph) shall be included in all copies or substantial portions of the
-Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
-DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
-OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
-USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-**************************************************************************/
-
-/*
- * Authors:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Daniel Borca <dborca@yahoo.com>
- */
-
-#if defined (__DJGPP__) || defined (__MINGW32__) || defined (__CYGWIN__)
-#define GLOBL( x )	\
-.globl _##x;		\
-_##x:
-#else  /* !defined (__DJGPP__) && !defined (__MINGW32__) && !defined (__CYGWIN__) */
-#define GLOBL( x )	\
-.globl x;		\
-x:
-#endif /* !defined (__DJGPP__) && !defined (__MINGW32__) && !defined (__CYGWIN__) */
-
-
-#if !defined (STDCALL_API)
-#define RETCLEAN( x )	ret
-#else
-#define RETCLEAN( x )	ret $x
-#endif
-
-
-#define _JMP(x)		\
-.byte 0xe9;		\
-.long x
-
-#define _CALL(x)	\
-.byte 0xe8;		\
-.long x
-
-
-/* Someone who knew a lot about this sort of thing would use this
- * macro to note current offsets, etc in a special region of the
- * object file & just make everything work out neat.  I don't know
- * enough to do that...
- */
-
-#define SUBST( x ) (0x10101010 + x)
-
-
-.data
-
-
-/* [dBorca] TODO
- * Unfold functions for each vertex size?
- * Build super-specialized SSE versions?
- *
- * There is a trick in Vertex*fv: under certain conditions,
- * we tail to _tnl_wrap_filled_vertex(ctx). This means that
- * if Vertex*fv is STDCALL, then _tnl_wrap_filled_vertex must
- * be STDCALL as well, because (GLcontext *) and (GLfloat *)
- * have the same size.
- */
-.align 4
-GLOBL ( _tnl_x86_Vertex1fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* 0x0 --> tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	addl	$4, %edi	/* tnl->vtx.vbptr += 1 */
-	movl	$SUBST(1), %ecx	/* 0x1 --> (tnl->vtx.vertex_size - 1) */
-	movl	$SUBST(2), %esi	/* 0x2 --> (tnl->vtx.vertex + 1) */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* 0x0 --> tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* 0x3 --> counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* 0x3 --> counter */
-	je	.0		/* if (counter == 0) goto .0 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.0:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex1fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex2fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	addl	$8, %edi	/* tnl->vtx.vbptr += 2 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 2 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 2 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.1		/* if (counter == 0) goto .1 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.1:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex2fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex3fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	8(%ecx), %esi	/* load v[2] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	movl	%esi, 8(%edi)	/* tnl->vtx.vbptr[2] = v[2] */
-	addl	$12, %edi	/* tnl->vtx.vbptr += 3 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 3 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 3 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.2		/* if (counter == 0) goto .2 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.2:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex3fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex4fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	8(%ecx), %esi	/* load v[2] */
-	movl	12(%ecx), %ecx	/* load v[3] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	movl	%esi, 8(%edi)	/* tnl->vtx.vbptr[2] = v[2] */
-	movl	%ecx, 12(%edi)	/* tnl->vtx.vbptr[3] = v[3] */
-	addl	$16, %edi	/* tnl->vtx.vbptr += 4 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 4 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 4 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.3		/* if (counter == 0) goto .3 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.3:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex4fv_end )
-
-
-/**
- * Generic handlers for vector format data. 
- */
-GLOBL( _tnl_x86_Attribute1fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute1fv_end )
-
-GLOBL( _tnl_x86_Attribute2fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute2fv_end )
-
-GLOBL( _tnl_x86_Attribute3fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	8(%ecx), %ecx	/* load v[2] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	movl	%ecx, SUBST(2)	/* store v[2] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute3fv_end )
-
-GLOBL( _tnl_x86_Attribute4fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	movl	8(%ecx), %eax	/* load v[2] */
-	movl	12(%ecx), %edx	/* load v[3] */
-	movl	%eax, SUBST(2)	/* store v[2] to current vertex */
-	movl	%edx, SUBST(3)	/* store v[3] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute4fv_end )
-
-
-/* Choosers:
- *
- * Must generate all of these ahead of first usage.  Generate at
- * compile-time?
- */
-GLOBL( _tnl_x86_choose_fv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	$SUBST(0), (%esp)	/* arg 0 - attrib */
-	movl	$SUBST(1), 4(%esp)	/* arg 1 - N */
-	_CALL	(SUBST(2))		/* call do_choose */
-	add	$12, %esp		/* tear down stack frame */
-	jmp	*%eax			/* jump to new func */
-GLOBL ( _tnl_x86_choose_fv_end )
-
-
-/* FIRST LEVEL FUNCTIONS -- these are plugged directly into GL dispatch.
- *
- * In the 1st level dispatch functions, switch to a different
- * calling convention -- (const GLfloat *v) in %ecx.
- *
- * As with regular (x86) dispatch, don't create a new stack frame -
- * just let the 'ret' in the dispatched function return straight
- * back to the original caller.
- *
- * Vertex/Normal/Color, etc: the address of the function pointer
- * is known at codegen time.
- */
-
-/* Unfortunately, have to play with the stack in the non-fv case:
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_attrf1 )
-GLOBL( _tnl_x86_dispatch_attrf2 )
-GLOBL( _tnl_x86_dispatch_attrf3 )
-GLOBL( _tnl_x86_dispatch_attrf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$12, %esp	/* tear down frame */
-	ret			/* return */
-GLOBL( _tnl_x86_dispatch_attrf4_end )
-GLOBL( _tnl_x86_dispatch_attrf3_end )
-GLOBL( _tnl_x86_dispatch_attrf2_end )
-GLOBL( _tnl_x86_dispatch_attrf1_end )
-
-#else  /* defined(STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_attrf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$4		/* return */
-GLOBL( _tnl_x86_dispatch_attrf1_end )
-
-GLOBL( _tnl_x86_dispatch_attrf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_attrf2_end )
-
-GLOBL( _tnl_x86_dispatch_attrf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_attrf3_end )
-
-GLOBL( _tnl_x86_dispatch_attrf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_attrf4_end )
-#endif /* defined(STDCALL_API) */
-
-/* The fv case is simpler:
- */
-GLOBL( _tnl_x86_dispatch_attrfv )
-	jmp	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-GLOBL( _tnl_x86_dispatch_attrfv_end )
-
-
-/* MultiTexcoord: the address of the function pointer must be
- * calculated, but can use the index argument slot to hold 'v', and
- * avoid setting up a new stack frame.
- *
- * [dBorca]
- * right, this would be the preferred approach, but gcc does not
- * clean up the stack after each function call when optimizing (-fdefer-pop);
- * can it make assumptions about what's already on the stack?  I dunno,
- * but in this case, we can't mess with the caller's stack frame, and
- * we must use a model like `_x86_dispatch_attrfv' above.  Caveat emptor!
- */
-
-/* Also, will only need a maximum of four of each of these per context:
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_multitexcoordf1 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf2 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf3 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf4 )
-	movl	4(%esp), %ecx
-	leal	8(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, 4(%esp)
-	sall	$4, %ecx
-	jmp	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-GLOBL( _tnl_x86_dispatch_multitexcoordf4_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf3_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf2_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf1_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordfv )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, 4(%esp)
-	sall	$4, %ecx
-	jmp	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
-
-#else  /* defined (STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf1_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf2_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf3_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$20		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf4_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordfv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	movl	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
-#endif /* defined (STDCALL_API) */
-
-
-/* VertexAttrib: the address of the function pointer must be
- * calculated.
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_vertexattribf1 )
-GLOBL( _tnl_x86_dispatch_vertexattribf2 )
-GLOBL( _tnl_x86_dispatch_vertexattribf3 )
-GLOBL( _tnl_x86_dispatch_vertexattribf4 )
-	movl	4(%esp), %eax
-	cmpl	$16, %eax
-	jb	.8		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.8:
-	leal	8(%esp), %ecx	/* calculate 'v' */
-	movl	%ecx, 4(%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	jmp	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-GLOBL( _tnl_x86_dispatch_vertexattribf4_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf3_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf2_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf1_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribfv )
-	movl	4(%esp), %eax
-	cmpl	$16, %eax
-	jb	.9		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.9:
-	movl	8(%esp), %ecx	/* load 'v' */
-	movl	%ecx, 4(%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	jmp	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
-
-#else  /* defined (STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_vertexattribf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.81		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.81:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf1_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.82		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.82:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf2_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.83		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.83:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf3_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.84		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.84:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$20		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf4_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribfv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.9		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.9:
-	movl	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
-#endif /* defined (STDCALL_API) */
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index dab422429a..17cb30adc7 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -50,30 +50,14 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state );
  * dispatch and/or driver callbacks.
  */
 
-/* Restore just the ctx->Exec table:
- */
-extern void
-_tnl_wakeup_exec( GLcontext *ctx );
-
-/* Restore both ctx->Exec and ctx->Save:
- */
 extern void
-_tnl_wakeup_save_exec( GLcontext *ctx );
+_tnl_wakeup( GLcontext *ctx );
 
 /* Driver configuration options:
  */
 extern void
 _tnl_need_projected_coords( GLcontext *ctx, GLboolean flag );
 
-extern void
-_tnl_need_dlist_loopback( GLcontext *ctx, GLboolean flag );
-
-extern void
-_tnl_need_dlist_norm_lengths( GLcontext *ctx, GLboolean flag );
-
-extern void
-_tnl_isolate_materials( GLcontext *ctx, GLboolean flag );
-
 
 /* Control whether T&L does per-vertex fog
  */
@@ -86,4 +70,16 @@ _tnl_allow_pixel_fog( GLcontext *ctx, GLboolean value );
 extern void
 _tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program);
 
+struct _mesa_prim;
+struct _mesa_index_buffer;
+
+void
+_tnl_draw_prims( GLcontext *ctx,
+		 const struct gl_client_array *arrays[],
+		 const struct _mesa_prim *prim,
+		 GLuint nr_prims,
+		 const struct _mesa_index_buffer *ib,
+		 GLuint min_index,
+		 GLuint max_index);
+
 #endif
diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h
new file mode 100644
index 0000000000..874a5f9e0e
--- /dev/null
+++ b/src/mesa/vbo/vbo.h
@@ -0,0 +1,117 @@
+/*
+ * mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file vbo_context.h
+ * \brief VBO builder module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+#ifndef _VBO_H
+#define _VBO_H
+
+#include "mtypes.h"
+
+struct _mesa_prim {
+   GLuint mode:8;
+   GLuint indexed:1;
+   GLuint begin:1;
+   GLuint end:1;
+   GLuint weak:1;
+   GLuint pad:20;
+
+   GLuint start;
+   GLuint count;
+};
+
+/* Would like to call this a "vbo_index_buffer", but this would be
+ * confusing as the indices are not neccessarily yet in a non-null
+ * buffer object.
+ */
+struct _mesa_index_buffer {
+   GLuint count;
+   GLenum type;
+   struct gl_buffer_object *obj;
+   const void *ptr;
+};
+
+
+
+GLboolean _vbo_CreateContext( GLcontext *ctx );
+void _vbo_DestroyContext( GLcontext *ctx );
+void _vbo_InvalidateState( GLcontext *ctx, GLuint new_state );
+
+
+typedef void (*vbo_draw_func)( GLcontext *ctx,
+			       const struct gl_client_array **arrays,
+			       const struct _mesa_prim *prims,
+			       GLuint nr_prims,
+			       const struct _mesa_index_buffer *ib,
+			       GLuint min_index,
+			       GLuint max_index );
+
+
+
+
+/* Utility function to cope with various constraints on tnl modules or
+ * hardware.  This can be used to split an incoming set of arrays and
+ * primitives against the following constraints:
+ *    - Maximum number of indices in index buffer.
+ *    - Maximum number of vertices referenced by index buffer.
+ *    - Maximum hardware vertex buffer size.
+ */
+struct split_limits {
+   GLuint max_verts;
+   GLuint max_indices;
+   GLuint max_vb_size;		/* bytes */
+};
+
+
+void vbo_split_prims( GLcontext *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index,
+		      vbo_draw_func draw,
+		      const struct split_limits *limits );
+
+
+/* Helpers for dealing translating away non-zero min_index.
+ */
+GLboolean vbo_all_varyings_in_vbos( const struct gl_client_array *arrays[] );
+
+void vbo_rebase_prims( GLcontext *ctx,
+		       const struct gl_client_array *arrays[],
+		       const struct _mesa_prim *prim,
+		       GLuint nr_prims,
+		       const struct _mesa_index_buffer *ib,
+		       GLuint min_index,
+		       GLuint max_index,
+		       vbo_draw_func draw );
+
+
+#endif
diff --git a/src/mesa/drivers/dri/i965/brw_attrib.h b/src/mesa/vbo/vbo_attrib.h
index 12659bd1cf..0ae928f2af 100644
--- a/src/mesa/drivers/dri/i965/brw_attrib.h
+++ b/src/mesa/vbo/vbo_attrib.h
@@ -29,8 +29,8 @@
   *   Keith Whitwell <keith@tungstengraphics.com>
   */
 
-#ifndef BRW_ATTRIB_H
-#define BRW_ATTRIB_H
+#ifndef VBO_ATTRIB_H
+#define VBO_ATTRIB_H
 
 
 /*
@@ -44,71 +44,64 @@
  * program instruction formats and register layouts.
  */
 enum {
-	BRW_ATTRIB_POS = 0,
-	BRW_ATTRIB_WEIGHT = 1,
-	BRW_ATTRIB_NORMAL = 2,
-	BRW_ATTRIB_COLOR0 = 3,
-	BRW_ATTRIB_COLOR1 = 4,
-	BRW_ATTRIB_FOG = 5,
-	BRW_ATTRIB_INDEX = 6,        
-	BRW_ATTRIB_EDGEFLAG = 7,     
-	BRW_ATTRIB_TEX0 = 8,
-	BRW_ATTRIB_TEX1 = 9,
-	BRW_ATTRIB_TEX2 = 10,
-	BRW_ATTRIB_TEX3 = 11,
-	BRW_ATTRIB_TEX4 = 12,
-	BRW_ATTRIB_TEX5 = 13,
-	BRW_ATTRIB_TEX6 = 14,
-	BRW_ATTRIB_TEX7 = 15,
+	VBO_ATTRIB_POS = 0,
+	VBO_ATTRIB_WEIGHT = 1,
+	VBO_ATTRIB_NORMAL = 2,
+	VBO_ATTRIB_COLOR0 = 3,
+	VBO_ATTRIB_COLOR1 = 4,
+	VBO_ATTRIB_FOG = 5,
+	VBO_ATTRIB_INDEX = 6,        
+	VBO_ATTRIB_EDGEFLAG = 7,     
+	VBO_ATTRIB_TEX0 = 8,
+	VBO_ATTRIB_TEX1 = 9,
+	VBO_ATTRIB_TEX2 = 10,
+	VBO_ATTRIB_TEX3 = 11,
+	VBO_ATTRIB_TEX4 = 12,
+	VBO_ATTRIB_TEX5 = 13,
+	VBO_ATTRIB_TEX6 = 14,
+	VBO_ATTRIB_TEX7 = 15,
 
-	BRW_ATTRIB_GENERIC0 = 16, /* Not used? */
-	BRW_ATTRIB_GENERIC1 = 17,
-	BRW_ATTRIB_GENERIC2 = 18,
-	BRW_ATTRIB_GENERIC3 = 19,
-	BRW_ATTRIB_GENERIC4 = 20,
-	BRW_ATTRIB_GENERIC5 = 21,
-	BRW_ATTRIB_GENERIC6 = 22,
-	BRW_ATTRIB_GENERIC7 = 23,
-	BRW_ATTRIB_GENERIC8 = 24,
-	BRW_ATTRIB_GENERIC9 = 25,
-	BRW_ATTRIB_GENERIC10 = 26,
-	BRW_ATTRIB_GENERIC11 = 27,
-	BRW_ATTRIB_GENERIC12 = 28,
-	BRW_ATTRIB_GENERIC13 = 29,
-	BRW_ATTRIB_GENERIC14 = 30,
-	BRW_ATTRIB_GENERIC15 = 31,
+	VBO_ATTRIB_GENERIC0 = 16, /* Not used? */
+	VBO_ATTRIB_GENERIC1 = 17,
+	VBO_ATTRIB_GENERIC2 = 18,
+	VBO_ATTRIB_GENERIC3 = 19,
+	VBO_ATTRIB_GENERIC4 = 20,
+	VBO_ATTRIB_GENERIC5 = 21,
+	VBO_ATTRIB_GENERIC6 = 22,
+	VBO_ATTRIB_GENERIC7 = 23,
+	VBO_ATTRIB_GENERIC8 = 24,
+	VBO_ATTRIB_GENERIC9 = 25,
+	VBO_ATTRIB_GENERIC10 = 26,
+	VBO_ATTRIB_GENERIC11 = 27,
+	VBO_ATTRIB_GENERIC12 = 28,
+	VBO_ATTRIB_GENERIC13 = 29,
+	VBO_ATTRIB_GENERIC14 = 30,
+	VBO_ATTRIB_GENERIC15 = 31,
 
-	BRW_ATTRIB_MAT_FRONT_AMBIENT = 32, 
-	BRW_ATTRIB_MAT_BACK_AMBIENT = 33,
-	BRW_ATTRIB_MAT_FRONT_DIFFUSE = 34,
-	BRW_ATTRIB_MAT_BACK_DIFFUSE = 35,
-	BRW_ATTRIB_MAT_FRONT_SPECULAR = 36,
-	BRW_ATTRIB_MAT_BACK_SPECULAR = 37,
-	BRW_ATTRIB_MAT_FRONT_EMISSION = 38,
-	BRW_ATTRIB_MAT_BACK_EMISSION = 39,
-	BRW_ATTRIB_MAT_FRONT_SHININESS = 40,
-	BRW_ATTRIB_MAT_BACK_SHININESS = 41,
-	BRW_ATTRIB_MAT_FRONT_INDEXES = 42,
-	BRW_ATTRIB_MAT_BACK_INDEXES = 43, 
+	/* XXX: in the vertex program InputsRead flag, we alias
+	 * materials and generics and use knowledge about the program
+	 * (whether it is a fixed-function emulation) to
+	 * differentiate.  Here we must keep them apart instead.
+	 */
+	VBO_ATTRIB_MAT_FRONT_AMBIENT = 32, 
+	VBO_ATTRIB_MAT_BACK_AMBIENT = 33,
+	VBO_ATTRIB_MAT_FRONT_DIFFUSE = 34,
+	VBO_ATTRIB_MAT_BACK_DIFFUSE = 35,
+	VBO_ATTRIB_MAT_FRONT_SPECULAR = 36,
+	VBO_ATTRIB_MAT_BACK_SPECULAR = 37,
+	VBO_ATTRIB_MAT_FRONT_EMISSION = 38,
+	VBO_ATTRIB_MAT_BACK_EMISSION = 39,
+	VBO_ATTRIB_MAT_FRONT_SHININESS = 40,
+	VBO_ATTRIB_MAT_BACK_SHININESS = 41,
+	VBO_ATTRIB_MAT_FRONT_INDEXES = 42,
+	VBO_ATTRIB_MAT_BACK_INDEXES = 43, 
 
-	BRW_ATTRIB_MAX = 44
-} ;
+	VBO_ATTRIB_MAX = 44
+};
 
-#define BRW_ATTRIB_FIRST_MATERIAL BRW_ATTRIB_MAT_FRONT_AMBIENT
-#define BRW_ATTRIB_LAST_MATERIAL BRW_ATTRIB_MAT_BACK_INDEXES
-
-#define BRW_MAX_COPIED_VERTS 3
-
-
-static inline GLuint64EXT brw_translate_inputs( GLboolean vp_enabled,
-						GLuint mesa_inputs )
-{
-   GLuint64EXT inputs = mesa_inputs;
-   if (vp_enabled)
-      return inputs;
-   else 
-      return (inputs & 0xffff) | ((inputs & 0xffff0000) << 16);
-}
+#define VBO_ATTRIB_FIRST_MATERIAL VBO_ATTRIB_MAT_FRONT_AMBIENT
+#define VBO_ATTRIB_LAST_MATERIAL VBO_ATTRIB_MAT_BACK_INDEXES
 
+#define VBO_MAX_COPIED_VERTS 3
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_attrib_tmp.h b/src/mesa/vbo/vbo_attrib_tmp.h
index 3089bd6cac..ff11c7d59a 100644
--- a/src/mesa/drivers/dri/i965/brw_attrib_tmp.h
+++ b/src/mesa/vbo/vbo_attrib_tmp.h
@@ -40,192 +40,192 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 static void GLAPIENTRY TAG(Vertex2f)( GLfloat x, GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2F( BRW_ATTRIB_POS, x, y );
+   ATTR2F( VBO_ATTRIB_POS, x, y );
 }
 
 static void GLAPIENTRY TAG(Vertex2fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2FV( BRW_ATTRIB_POS, v );
+   ATTR2FV( VBO_ATTRIB_POS, v );
 }
 
 static void GLAPIENTRY TAG(Vertex3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_POS, x, y, z );
+   ATTR3F( VBO_ATTRIB_POS, x, y, z );
 }
 
 static void GLAPIENTRY TAG(Vertex3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_POS, v );
+   ATTR3FV( VBO_ATTRIB_POS, v );
 }
 
 static void GLAPIENTRY TAG(Vertex4f)( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4F( BRW_ATTRIB_POS, x, y, z, w );
+   ATTR4F( VBO_ATTRIB_POS, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(Vertex4fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4FV( BRW_ATTRIB_POS, v );
+   ATTR4FV( VBO_ATTRIB_POS, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord1f)( GLfloat x )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_TEX0, x );
+   ATTR1F( VBO_ATTRIB_TEX0, x );
 }
 
 static void GLAPIENTRY TAG(TexCoord1fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1FV( BRW_ATTRIB_TEX0, v );
+   ATTR1FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord2f)( GLfloat x, GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2F( BRW_ATTRIB_TEX0, x, y );
+   ATTR2F( VBO_ATTRIB_TEX0, x, y );
 }
 
 static void GLAPIENTRY TAG(TexCoord2fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2FV( BRW_ATTRIB_TEX0, v );
+   ATTR2FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_TEX0, x, y, z );
+   ATTR3F( VBO_ATTRIB_TEX0, x, y, z );
 }
 
 static void GLAPIENTRY TAG(TexCoord3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_TEX0, v );
+   ATTR3FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord4f)( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4F( BRW_ATTRIB_TEX0, x, y, z, w );
+   ATTR4F( VBO_ATTRIB_TEX0, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(TexCoord4fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4FV( BRW_ATTRIB_TEX0, v );
+   ATTR4FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(Normal3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_NORMAL, x, y, z );
+   ATTR3F( VBO_ATTRIB_NORMAL, x, y, z );
 }
 
 static void GLAPIENTRY TAG(Normal3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_NORMAL, v );
+   ATTR3FV( VBO_ATTRIB_NORMAL, v );
 }
 
 static void GLAPIENTRY TAG(FogCoordfEXT)( GLfloat x )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_FOG, x );
+   ATTR1F( VBO_ATTRIB_FOG, x );
 }
 
 static void GLAPIENTRY TAG(FogCoordfvEXT)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1FV( BRW_ATTRIB_FOG, v );
+   ATTR1FV( VBO_ATTRIB_FOG, v );
 }
 
 static void GLAPIENTRY TAG(Color3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_COLOR0, x, y, z );
+   ATTR3F( VBO_ATTRIB_COLOR0, x, y, z );
 }
 
 static void GLAPIENTRY TAG(Color3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_COLOR0, v );
+   ATTR3FV( VBO_ATTRIB_COLOR0, v );
 }
 
 static void GLAPIENTRY TAG(Color4f)( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4F( BRW_ATTRIB_COLOR0, x, y, z, w );
+   ATTR4F( VBO_ATTRIB_COLOR0, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(Color4fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4FV( BRW_ATTRIB_COLOR0, v );
+   ATTR4FV( VBO_ATTRIB_COLOR0, v );
 }
 
 static void GLAPIENTRY TAG(SecondaryColor3fEXT)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_COLOR1, x, y, z );
+   ATTR3F( VBO_ATTRIB_COLOR1, x, y, z );
 }
 
 static void GLAPIENTRY TAG(SecondaryColor3fvEXT)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_COLOR1, v );
+   ATTR3FV( VBO_ATTRIB_COLOR1, v );
 }
 
 
 static void GLAPIENTRY TAG(EdgeFlag)( GLboolean b )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_EDGEFLAG, (GLfloat)b );
+   ATTR1F( VBO_ATTRIB_EDGEFLAG, (GLfloat)b );
 }
 
 static void GLAPIENTRY TAG(Indexf)( GLfloat f )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_INDEX, f );
+   ATTR1F( VBO_ATTRIB_INDEX, f );
 }
 
 static void GLAPIENTRY TAG(Indexfv)( const GLfloat *f )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1FV( BRW_ATTRIB_INDEX, f );
+   ATTR1FV( VBO_ATTRIB_INDEX, f );
 }
 
 
 static void GLAPIENTRY TAG(MultiTexCoord1f)( GLenum target, GLfloat x  )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR1F( attr, x );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord1fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR1FV( attr, v );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord2f)( GLenum target, GLfloat x, GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR2F( attr, x, y );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord2fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR2FV( attr, v );
 }
 
@@ -233,14 +233,14 @@ static void GLAPIENTRY TAG(MultiTexCoord3f)( GLenum target, GLfloat x, GLfloat y
 				    GLfloat z)
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR3F( attr, x, y, z );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord3fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR3FV( attr, v );
 }
 
@@ -248,14 +248,14 @@ static void GLAPIENTRY TAG(MultiTexCoord4f)( GLenum target, GLfloat x, GLfloat y
 				    GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR4F( attr, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord4fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR4FV( attr, v );
 }
 
@@ -266,7 +266,7 @@ static void GLAPIENTRY TAG(VertexAttrib1fARB)( GLuint index, GLfloat x )
    if (index == 0) 
       ATTR1F(0, x);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR1F(BRW_ATTRIB_GENERIC0 + index, x);
+      ATTR1F(VBO_ATTRIB_GENERIC0 + index, x);
    else
       ERROR();
 }
@@ -278,7 +278,7 @@ static void GLAPIENTRY TAG(VertexAttrib1fvARB)( GLuint index,
    if (index == 0) 
       ATTR1FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR1FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR1FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
@@ -290,7 +290,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fARB)( GLuint index, GLfloat x,
    if (index == 0) 
       ATTR2F(0, x, y);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR2F(BRW_ATTRIB_GENERIC0 + index, x, y);
+      ATTR2F(VBO_ATTRIB_GENERIC0 + index, x, y);
    else
       ERROR();
 }
@@ -302,7 +302,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fvARB)( GLuint index,
    if (index == 0) 
       ATTR2FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR2FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR2FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
@@ -314,7 +314,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fARB)( GLuint index, GLfloat x,
    if (index == 0) 
       ATTR3F(0, x, y, z);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR3F(BRW_ATTRIB_GENERIC0 + index, x, y, z);
+      ATTR3F(VBO_ATTRIB_GENERIC0 + index, x, y, z);
    else
       ERROR();
 }
@@ -326,7 +326,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fvARB)( GLuint index,
    if (index == 0) 
       ATTR3FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR3FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR3FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
@@ -339,7 +339,7 @@ static void GLAPIENTRY TAG(VertexAttrib4fARB)( GLuint index, GLfloat x,
    if (index == 0) 
       ATTR4F(0, x, y, z, w);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR4F(BRW_ATTRIB_GENERIC0 + index, x, y, z, w);
+      ATTR4F(VBO_ATTRIB_GENERIC0 + index, x, y, z, w);
    else
       ERROR();
 }
@@ -351,20 +351,21 @@ static void GLAPIENTRY TAG(VertexAttrib4fvARB)( GLuint index,
    if (index == 0) 
       ATTR4FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR4FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR4FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
 
 
-/* Although we don't export NV_vertex_program, these entrypoints are
+/* In addition to supporting NV_vertex_program, these entrypoints are
  * used by the display list and other code specifically because of
- * their property of aliasing with other attributes.
+ * their property of aliasing with other attributes.  (See
+ * vbo_save_loopback.c)
  */
 static void GLAPIENTRY TAG(VertexAttrib1fNV)( GLuint index, GLfloat x )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX)
+   if (index < VBO_ATTRIB_MAX)
       ATTR1F(index, x);
 }
 
@@ -372,7 +373,7 @@ static void GLAPIENTRY TAG(VertexAttrib1fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR1FV(index, v);
 }
 
@@ -380,7 +381,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fNV)( GLuint index, GLfloat x,
 					      GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR2F(index, x, y);
 }
 
@@ -388,7 +389,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR2FV(index, v);
 }
 
@@ -396,7 +397,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fNV)( GLuint index, GLfloat x,
 					      GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR3F(index, x, y, z);
 }
 
@@ -404,7 +405,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR3FV(index, v);
 }
 
@@ -413,7 +414,7 @@ static void GLAPIENTRY TAG(VertexAttrib4fNV)( GLuint index, GLfloat x,
 					      GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR4F(index, x, y, z, w);
 }
 
@@ -421,7 +422,7 @@ static void GLAPIENTRY TAG(VertexAttrib4fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR4FV(index, v);
 }
 
@@ -443,26 +444,26 @@ static void GLAPIENTRY TAG(Materialfv)( GLenum face, GLenum pname,
    GET_CURRENT_CONTEXT( ctx );
    switch (pname) {
    case GL_EMISSION:
-      MAT( BRW_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
       break;
    case GL_AMBIENT:
-      MAT( BRW_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
       break;
    case GL_DIFFUSE:
-      MAT( BRW_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
       break;
    case GL_SPECULAR:
-      MAT( BRW_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
       break;
    case GL_SHININESS:
-      MAT( BRW_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
       break;
    case GL_COLOR_INDEXES:
-      MAT( BRW_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
       break;
    case GL_AMBIENT_AND_DIFFUSE:
-      MAT( BRW_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      MAT( BRW_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
       break;
    default:
       ERROR();
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
new file mode 100644
index 0000000000..65998e7ff7
--- /dev/null
+++ b/src/mesa/vbo/vbo_context.c
@@ -0,0 +1,246 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ *
+ * Copyright (C) 1999-2005  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 "mtypes.h"
+#include "vbo.h"
+#include "vbo_context.h"
+#include "imports.h"
+#include "api_arrayelt.h"
+
+/* Reach out and grab this to use as the default:
+ */
+extern void _tnl_draw_prims( GLcontext *ctx,
+			     const struct gl_client_array *arrays[],
+			     const struct _mesa_prim *prims,
+			     GLuint nr_prims,
+			     const struct _mesa_index_buffer *ib,
+			     GLuint min_index,
+			     GLuint max_index );
+
+
+
+#define NR_LEGACY_ATTRIBS 16
+#define NR_GENERIC_ATTRIBS 16
+#define NR_MAT_ATTRIBS 12
+
+static GLuint check_size( const GLfloat *attr )
+{
+   if (attr[3] != 1.0) return 4;
+   if (attr[2] != 0.0) return 3;
+   if (attr[1] != 0.0) return 2;
+   return 1;		
+}
+
+static void init_legacy_currval(GLcontext *ctx)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct gl_client_array *arrays = vbo->legacy_currval;
+   GLuint i;
+
+   memset(arrays, 0, sizeof(*arrays) * NR_LEGACY_ATTRIBS);
+
+   /* Set up a constant (StrideB == 0) array for each current
+    * attribute:
+    */
+   for (i = 0; i < NR_LEGACY_ATTRIBS; i++) {
+      struct gl_client_array *cl = &arrays[i];
+
+      /* Size will have to be determined at runtime:
+       */
+      cl->Size = check_size(ctx->Current.Attrib[i]);
+      cl->Stride = 0;
+      cl->StrideB = 0;
+      cl->Enabled = 1;
+      cl->Type = GL_FLOAT;
+      cl->Ptr = (const void *)ctx->Current.Attrib[i];
+      cl->BufferObj = ctx->Array.NullBufferObj;
+   }
+}
+
+
+static void init_generic_currval(GLcontext *ctx)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct gl_client_array *arrays = vbo->generic_currval;
+   GLuint i;
+
+   memset(arrays, 0, sizeof(*arrays) * NR_GENERIC_ATTRIBS);
+
+   for (i = 0; i < NR_GENERIC_ATTRIBS; i++) {
+      struct gl_client_array *cl = &arrays[i];
+
+      /* This will have to be determined at runtime:
+       */
+      cl->Size = 1;
+      cl->Type = GL_FLOAT;
+      cl->Ptr = (const void *)ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + i];
+      cl->Stride = 0;
+      cl->StrideB = 0;
+      cl->Enabled = 1;
+      cl->BufferObj = ctx->Array.NullBufferObj;
+   }
+}
+
+
+static void init_mat_currval(GLcontext *ctx)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct gl_client_array *arrays = vbo->mat_currval;
+   GLuint i;
+
+   memset(arrays, 0, sizeof(*arrays) * NR_MAT_ATTRIBS);
+
+   /* Set up a constant (StrideB == 0) array for each current
+    * attribute:
+    */
+   for (i = 0; i < NR_MAT_ATTRIBS; i++) {
+      struct gl_client_array *cl = &arrays[i];
+
+      /* Size is fixed for the material attributes, for others will
+       * be determined at runtime:
+       */
+      switch (i - VERT_ATTRIB_GENERIC0) {
+      case MAT_ATTRIB_FRONT_SHININESS:
+      case MAT_ATTRIB_BACK_SHININESS:
+	 cl->Size = 1;
+	 break;
+      case MAT_ATTRIB_FRONT_INDEXES:
+      case MAT_ATTRIB_BACK_INDEXES:
+	 cl->Size = 3;
+	 break;
+      default:
+	 cl->Size = 4;
+	 break;
+      }
+
+      if (i < MAT_ATTRIB_MAX)
+	 cl->Ptr = (const void *)ctx->Light.Material.Attrib[i];
+      else 
+	 cl->Ptr = (const void *)ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + i];
+
+      cl->Type = GL_FLOAT;
+      cl->Stride = 0;
+      cl->StrideB = 0;
+      cl->Enabled = 1;
+      cl->BufferObj = ctx->Array.NullBufferObj;
+   }
+}
+
+#if 0
+
+static void vbo_exec_current_init( struct vbo_exec_context *exec ) 
+{
+   GLcontext *ctx = exec->ctx;
+   GLint i;
+
+   /* setup the pointers for the typical 16 vertex attributes */
+   for (i = 0; i < VBO_ATTRIB_FIRST_MATERIAL; i++) 
+      exec->vtx.current[i] = ctx->Current.Attrib[i];
+
+   /* setup pointers for the 12 material attributes */
+   for (i = 0; i < MAT_ATTRIB_MAX; i++)
+      exec->vtx.current[VBO_ATTRIB_FIRST_MATERIAL + i] = 
+	 ctx->Light.Material.Attrib[i];
+}
+#endif
+
+GLboolean _vbo_CreateContext( GLcontext *ctx )
+{
+   struct vbo_context *vbo = CALLOC_STRUCT(vbo_context);
+
+   ctx->swtnl_im = (void *)vbo;
+
+   /* Initialize the arrayelt helper
+    */
+   if (!ctx->aelt_context &&
+       !_ae_create_context( ctx )) {
+      return GL_FALSE;
+   }
+
+   /* TODO: remove these pointers.
+    */
+   vbo->legacy_currval = &vbo->currval[VBO_ATTRIB_POS];
+   vbo->generic_currval = &vbo->currval[VBO_ATTRIB_GENERIC0];
+   vbo->mat_currval = &vbo->currval[VBO_ATTRIB_MAT_FRONT_AMBIENT];
+
+   init_legacy_currval( ctx );
+   init_generic_currval( ctx );
+   init_mat_currval( ctx );
+
+   /* Build mappings from VERT_ATTRIB -> VBO_ATTRIB depending on type
+    * of vertex program active.
+    */
+   {
+      GLuint i;
+
+      /* When no vertex program, pull in the material attributes in
+       * the 16..32 generic range.
+       */
+      for (i = 0; i < 16; i++) 
+	 vbo->map_vp_none[i] = i;
+      for (i = 0; i < 12; i++) 
+	 vbo->map_vp_none[16+i] = VBO_ATTRIB_MAT_FRONT_AMBIENT + i;
+      for (i = 0; i < 4; i++)
+	 vbo->map_vp_none[28+i] = i;	
+      
+      for (i = 0; i < VERT_ATTRIB_MAX; i++)
+	 vbo->map_vp_arb[i] = i;
+   }
+
+
+   /* By default: 
+    */
+   vbo->draw_prims = _tnl_draw_prims;
+
+   /* Hook our functions into exec and compile dispatch tables.  These
+    * will pretty much be permanently installed, which means that the
+    * vtxfmt mechanism can be removed now.
+    */
+   vbo_exec_init( ctx );
+   vbo_save_init( ctx );
+
+   
+   return GL_TRUE;
+}
+
+void _vbo_InvalidateState( GLcontext *ctx, GLuint new_state )
+{
+   _ae_invalidate_state(ctx, new_state);
+   vbo_exec_invalidate_state(ctx, new_state);
+}
+
+
+void _vbo_DestroyContext( GLcontext *ctx )
+{
+   if (ctx->aelt_context) {
+      _ae_destroy_context( ctx );
+      ctx->aelt_context = NULL;
+   }
+
+   FREE(vbo_context(ctx));
+   ctx->swtnl_im = NULL;
+}
diff --git a/src/mesa/vbo/vbo_context.h b/src/mesa/vbo/vbo_context.h
new file mode 100644
index 0000000000..0dc1019b39
--- /dev/null
+++ b/src/mesa/vbo/vbo_context.h
@@ -0,0 +1,108 @@
+/*
+ * mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file vbo_context.h
+ * \brief VBO builder module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+/**
+ * \mainpage The VBO builder module
+ *
+ * This module hooks into the GL dispatch table and catches all vertex
+ * building and drawing commands, such as glVertex3f, glBegin and
+ * glDrawArrays.  The module stores all incoming vertex data as arrays
+ * in GL vertex buffer objects (VBOs), and translates all drawing
+ * commands into calls to a driver supplied DrawPrimitives() callback.
+ *
+ * The module captures both immediate mode and display list drawing,
+ * and manages the allocation, reference counting and deallocation of
+ * vertex buffer objects itself.
+ * 
+ * The DrawPrimitives() callback can be either implemented by the
+ * driver itself or hooked to the tnl module's _tnl_draw_primitives()
+ * function for hardware without tnl capablilties or during fallbacks.
+ */
+
+
+#ifndef _VBO_CONTEXT_H
+#define _VBO_CONTEXT_H
+
+#include "vbo.h"
+#include "vbo_attrib.h"
+#include "vbo_save.h"
+#include "vbo_exec.h"
+
+
+struct vbo_context {
+   struct gl_client_array currval[VBO_ATTRIB_MAX];
+   
+   /* These point into the above.  TODO: remove. 
+    */
+   struct gl_client_array *legacy_currval;
+   struct gl_client_array *generic_currval;
+   struct gl_client_array *mat_currval;
+
+   GLuint map_vp_none[32];
+   GLuint map_vp_arb[32];
+
+   GLfloat *current[VBO_ATTRIB_MAX]; /* points into ctx->Current, ctx->Light.Material */
+   GLfloat CurrentFloatEdgeFlag;
+
+
+   struct vbo_exec_context exec;
+   struct vbo_save_context save;
+
+   /* Callback into the driver.  This must always succeed, the driver
+    * is responsible for initiating any fallback actions required:
+    */
+   vbo_draw_func draw_prims;
+};
+
+
+static INLINE struct vbo_context *vbo_context(GLcontext *ctx) 
+{
+   return (struct vbo_context *)(ctx->swtnl_im);
+}
+
+enum {
+   VP_NONE = 1,
+   VP_NV,
+   VP_ARB
+};
+
+static INLINE GLuint get_program_mode( GLcontext *ctx )
+{
+   if (!ctx->VertexProgram._Enabled)
+      return VP_NONE;
+   else if (ctx->VertexProgram.Current->IsNVProgram)
+      return VP_NV;
+   else
+      return VP_ARB;
+}
+
+
+#endif
diff --git a/src/mesa/drivers/dri/i965/brw_exec.c b/src/mesa/vbo/vbo_exec.c
index fc06c3c361..7d95873247 100644
--- a/src/mesa/drivers/dri/i965/brw_exec.c
+++ b/src/mesa/vbo/vbo_exec.c
@@ -35,19 +35,13 @@
 #include "dlist.h"
 #include "vtxfmt.h"
 
-#include "brw_exec.h"
+#include "vbo_context.h"
 
-
-void brw_exec_init( GLcontext *ctx )
+void vbo_exec_init( GLcontext *ctx )
 {
-   struct brw_exec_context *exec = CALLOC_STRUCT(brw_exec_context);
-
-   if (ctx->swtnl_im == NULL) {
-      ctx->swtnl_im = CALLOC_STRUCT(brw_exec_save);
-   }
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    exec->ctx = ctx;
-   IMM_CONTEXT(ctx)->exec = exec;
 
    /* Initialize the arrayelt helper
     */
@@ -55,48 +49,41 @@ void brw_exec_init( GLcontext *ctx )
        !_ae_create_context( ctx )) 
       return;
 
-   brw_exec_vtx_init( exec );
-   brw_exec_array_init( exec );
+   vbo_exec_vtx_init( exec );
+   vbo_exec_array_init( exec );
+
+   /* Hook our functions into exec and compile dispatch tables.
+    */
+   _mesa_install_exec_vtxfmt( ctx, &exec->vtxfmt );
 
    ctx->Driver.NeedFlush = 0;
    ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-   ctx->Driver.FlushVertices = brw_exec_FlushVertices;
+   ctx->Driver.FlushVertices = vbo_exec_FlushVertices;
 
-   exec->eval.recalculate_maps = 1;
+   vbo_exec_invalidate_state( ctx, ~0 );
 }
 
 
-void brw_exec_destroy( GLcontext *ctx )
+void vbo_exec_destroy( GLcontext *ctx )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    if (ctx->aelt_context) {
       _ae_destroy_context( ctx );
       ctx->aelt_context = NULL;
    }
 
-   brw_exec_vtx_destroy( exec );
-   brw_exec_array_destroy( exec );
-
-   if (exec) {
-      FREE(exec);
-      IMM_CONTEXT(ctx)->exec = NULL;
-   }
-
-   if (IMM_CONTEXT(ctx)->exec == NULL &&
-       IMM_CONTEXT(ctx)->save == NULL) {
-      FREE(IMM_CONTEXT(ctx));
-      ctx->swtnl_im = NULL;
-   }
+   vbo_exec_vtx_destroy( exec );
+   vbo_exec_array_destroy( exec );
 }
 
 /* Really want to install these callbacks to a central facility to be
  * invoked according to the state flags.  That will have to wait for a
  * mesa rework:
  */ 
-void brw_exec_invalidate_state( GLcontext *ctx, GLuint new_state )
+void vbo_exec_invalidate_state( GLcontext *ctx, GLuint new_state )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    if (new_state & (_NEW_PROGRAM|_NEW_EVAL))
       exec->eval.recalculate_maps = 1;
@@ -105,21 +92,6 @@ void brw_exec_invalidate_state( GLcontext *ctx, GLuint new_state )
 }
 
 
-void brw_exec_wakeup( GLcontext *ctx )
-{
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
-
-   ctx->Driver.FlushVertices = brw_exec_FlushVertices;
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &exec->vtxfmt );
-
-   /* Assume we haven't been getting state updates either:
-    */
-   brw_exec_invalidate_state( ctx, ~0 );
-}
 
 
 
diff --git a/src/mesa/vbo/vbo_exec.h b/src/mesa/vbo/vbo_exec.h
new file mode 100644
index 0000000000..a9b01e08e6
--- /dev/null
+++ b/src/mesa/vbo/vbo_exec.h
@@ -0,0 +1,165 @@
+/**************************************************************************
+
+Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ *
+ */
+
+#ifndef __VBO_EXEC_H__
+#define __VBO_EXEC_H__
+
+#include "mtypes.h"
+#include "vbo.h"
+#include "vbo_attrib.h"
+
+
+#define VBO_MAX_PRIM 64
+
+/* Wierd implementation stuff:
+ */
+#define VBO_VERT_BUFFER_SIZE (1024*16)	/* dwords == 64k */
+#define VBO_MAX_ATTR_CODEGEN 16 
+#define ERROR_ATTRIB 16
+
+
+
+
+struct vbo_exec_eval1_map {
+   struct gl_1d_map *map;
+   GLuint sz;
+};
+
+struct vbo_exec_eval2_map {
+   struct gl_2d_map *map;
+   GLuint sz;
+};
+
+
+
+struct vbo_exec_copied_vtx {
+   GLfloat buffer[VBO_ATTRIB_MAX * 4 * VBO_MAX_COPIED_VERTS];
+   GLuint nr;
+};
+
+
+typedef void (*vbo_attrfv_func)( const GLfloat * );
+
+
+struct vbo_exec_context
+{
+   GLcontext *ctx;   
+   GLvertexformat vtxfmt;
+
+   struct {
+      struct gl_buffer_object *bufferobj;
+      GLubyte *buffer_map;
+
+      GLuint vertex_size;
+
+      struct _mesa_prim prim[VBO_MAX_PRIM];
+      GLuint prim_count;
+
+      GLfloat *vbptr;		     /* cursor, points into buffer */
+      GLfloat vertex[VBO_ATTRIB_MAX*4]; /* current vertex */
+
+      GLuint vert_count;
+      GLuint max_vert;
+      struct vbo_exec_copied_vtx copied;
+
+      GLubyte attrsz[VBO_ATTRIB_MAX];
+      GLubyte active_sz[VBO_ATTRIB_MAX];
+
+      GLfloat *attrptr[VBO_ATTRIB_MAX]; 
+      struct gl_client_array arrays[VBO_ATTRIB_MAX];
+
+      /* According to program mode, the values above plus current
+       * values are squashed down to the 32 attributes passed to the
+       * vertex program below:
+       */
+      GLuint program_mode;
+      GLuint enabled_flags;
+      const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
+   } vtx;
+
+   
+   struct {
+      GLboolean recalculate_maps;
+      struct vbo_exec_eval1_map map1[VERT_ATTRIB_MAX];
+      struct vbo_exec_eval2_map map2[VERT_ATTRIB_MAX];
+   } eval;
+
+   struct {
+      GLuint program_mode;
+      GLuint enabled_flags;
+      GLuint array_obj;
+
+      /* These just mirror the current arrayobj (todo: make arrayobj
+       * look like this and remove the mirror):
+       */
+      const struct gl_client_array *legacy_array[16];
+      const struct gl_client_array *generic_array[16];
+
+      /* Arrays and current values manipulated according to program
+       * mode, etc.  These are the attributes as seen by vertex
+       * programs:
+       */
+      const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
+   } array;
+};
+
+
+
+/* External API:
+ */
+void vbo_exec_init( GLcontext *ctx );
+void vbo_exec_destroy( GLcontext *ctx );
+void vbo_exec_invalidate_state( GLcontext *ctx, GLuint new_state );
+void vbo_exec_FlushVertices( GLcontext *ctx, GLuint flags );
+
+
+/* Internal functions:
+ */
+void vbo_exec_array_init( struct vbo_exec_context *exec );
+void vbo_exec_array_destroy( struct vbo_exec_context *exec );
+
+
+void vbo_exec_vtx_init( struct vbo_exec_context *exec );
+void vbo_exec_vtx_destroy( struct vbo_exec_context *exec );
+void vbo_exec_vtx_flush( struct vbo_exec_context *exec );
+void vbo_exec_vtx_wrap( struct vbo_exec_context *exec );
+
+void vbo_exec_eval_update( struct vbo_exec_context *exec );
+
+void vbo_exec_do_EvalCoord2f( struct vbo_exec_context *exec, 
+				     GLfloat u, GLfloat v );
+
+void vbo_exec_do_EvalCoord1f( struct vbo_exec_context *exec,
+				     GLfloat u);
+
+#endif
diff --git a/src/mesa/drivers/dri/i965/brw_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index 8b243c6084..71fee8ca16 100644
--- a/src/mesa/drivers/dri/i965/brw_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -41,17 +41,15 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_noop.h"
 #include "dispatch.h"
 
-#include "brw_exec.h"
-#include "intel_context.h"
+#include "vbo_context.h"
 
-
-static void reset_attrfv( struct brw_exec_context *exec );
+static void reset_attrfv( struct vbo_exec_context *exec );
 
 
 /* Close off the last primitive, execute the buffer, restart the
  * primitive.  
  */
-static void brw_exec_wrap_buffers( struct brw_exec_context *exec )
+static void vbo_exec_wrap_buffers( struct vbo_exec_context *exec )
 {
    if (exec->vtx.prim_count == 0) {
       exec->vtx.copied.nr = 0;
@@ -74,7 +72,7 @@ static void brw_exec_wrap_buffers( struct brw_exec_context *exec )
       /* Execute the buffer and save copied vertices.
        */
       if (exec->vtx.vert_count)
-	 brw_exec_vtx_flush( exec );
+	 vbo_exec_vtx_flush( exec );
       else {
 	 exec->vtx.prim_count = 0;
 	 exec->vtx.copied.nr = 0;
@@ -100,7 +98,7 @@ static void brw_exec_wrap_buffers( struct brw_exec_context *exec )
 /* Deal with buffer wrapping where provoked by the vertex buffer
  * filling up, as opposed to upgrade_vertex().
  */
-void brw_exec_vtx_wrap( struct brw_exec_context *exec )
+void vbo_exec_vtx_wrap( struct vbo_exec_context *exec )
 {
    GLfloat *data = exec->vtx.copied.buffer;
    GLuint i;
@@ -108,7 +106,7 @@ void brw_exec_vtx_wrap( struct brw_exec_context *exec )
    /* Run pipeline on current vertices, copy wrapped vertices
     * to exec->vtx.copied.
     */
-   brw_exec_wrap_buffers( exec );
+   vbo_exec_wrap_buffers( exec );
    
    /* Copy stored stored vertices to start of new list. 
     */
@@ -129,72 +127,68 @@ void brw_exec_vtx_wrap( struct brw_exec_context *exec )
 /*
  * Copy the active vertex's values to the ctx->Current fields.
  */
-static void brw_exec_copy_to_current( struct brw_exec_context *exec )
+static void vbo_exec_copy_to_current( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLuint i;
 
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (exec->vtx.attrsz[i]) {
+	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+
          /* Note: the exec->vtx.current[i] pointers point into the
           * ctx->Current.Attrib and ctx->Light.Material.Attrib arrays.
           */
-	 COPY_CLEAN_4V(exec->vtx.current[i], 
+	 COPY_CLEAN_4V(current, 
 		       exec->vtx.attrsz[i], 
 		       exec->vtx.attrptr[i]);
 
+	 
+	 /* Given that we explicitly state size here, there is no need
+	  * for the COPY_CLEAN above, could just copy 16 bytes and be
+	  * done.  The only problem is when Mesa accesses ctx->Current
+	  * directly.
+	  */
+	 vbo->currval[i].Size = exec->vtx.attrsz[i];
+
 	 /* This triggers rather too much recalculation of Mesa state
 	  * that doesn't get used (eg light positions).
 	  */
-	 if (i >= BRW_ATTRIB_MAT_FRONT_AMBIENT &&
-	     i <= BRW_ATTRIB_MAT_BACK_INDEXES)
+	 if (i >= VBO_ATTRIB_MAT_FRONT_AMBIENT &&
+	     i <= VBO_ATTRIB_MAT_BACK_INDEXES)
 	    ctx->NewState |= _NEW_LIGHT;
       }
    }
 
-   /* color index is special (it's not a float[4] so COPY_CLEAN_4V above
-    * will trash adjacent memory!)
-    */
-   if (exec->vtx.attrsz[BRW_ATTRIB_INDEX]) {
-      ctx->Current.Index = exec->vtx.attrptr[BRW_ATTRIB_INDEX][0];
-   }
-
-   /* Edgeflag requires additional treatment:
-    */
-   if (exec->vtx.attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (exec->vtx.CurrentFloatEdgeFlag == 1.0);
-   }
-
    /* Colormaterial -- this kindof sucks.
     */
    if (ctx->Light.ColorMaterialEnabled &&
-       exec->vtx.attrsz[BRW_ATTRIB_COLOR0]) {
+       exec->vtx.attrsz[VBO_ATTRIB_COLOR0]) {
       _mesa_update_color_material(ctx, 
-				  ctx->Current.Attrib[BRW_ATTRIB_COLOR0]);
+				  ctx->Current.Attrib[VBO_ATTRIB_COLOR0]);
    }
 
    ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
 }
 
 
-static void brw_exec_copy_from_current( struct brw_exec_context *exec )
+static void vbo_exec_copy_from_current( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLint i;
 
-   /* Edgeflag requires additional treatment:
-    */
-   exec->vtx.CurrentFloatEdgeFlag = 
-      (GLfloat)ctx->Current.EdgeFlag;
-   
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) 
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
+      const GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
       switch (exec->vtx.attrsz[i]) {
-      case 4: exec->vtx.attrptr[i][3] = exec->vtx.current[i][3];
-      case 3: exec->vtx.attrptr[i][2] = exec->vtx.current[i][2];
-      case 2: exec->vtx.attrptr[i][1] = exec->vtx.current[i][1];
-      case 1: exec->vtx.attrptr[i][0] = exec->vtx.current[i][0];
+      case 4: exec->vtx.attrptr[i][3] = current[3];
+      case 3: exec->vtx.attrptr[i][2] = current[2];
+      case 2: exec->vtx.attrptr[i][1] = current[1];
+      case 1: exec->vtx.attrptr[i][0] = current[0];
 	 break;
       }
+   }
 
    ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
 }
@@ -202,11 +196,12 @@ static void brw_exec_copy_from_current( struct brw_exec_context *exec )
 
 /* Flush existing data, set new attrib size, replay copied vertices.
  */ 
-static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
+static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec,
 					  GLuint attr,
 					  GLuint newsz )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLint lastcount = exec->vtx.vert_count;
    GLfloat *tmp;
    GLuint oldsz;
@@ -215,14 +210,14 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
    /* Run pipeline on current vertices, copy wrapped vertices
     * to exec->vtx.copied.
     */
-   brw_exec_wrap_buffers( exec );
+   vbo_exec_wrap_buffers( exec );
 
 
    /* Do a COPY_TO_CURRENT to ensure back-copying works for the case
     * when the attribute already exists in the vertex and is having
     * its size increased.  
     */
-   brw_exec_copy_to_current( exec );
+   vbo_exec_copy_to_current( exec );
 
 
    /* Heuristic: Attempt to isolate attributes received outside
@@ -241,14 +236,14 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
    exec->vtx.attrsz[attr] = newsz;
 
    exec->vtx.vertex_size += newsz - oldsz;
-   exec->vtx.max_vert = BRW_VERT_BUFFER_SIZE / exec->vtx.vertex_size;
+   exec->vtx.max_vert = VBO_VERT_BUFFER_SIZE / exec->vtx.vertex_size;
    exec->vtx.vert_count = 0;
    exec->vtx.vbptr = (GLfloat *)exec->vtx.buffer_map;
    
 
    /* Recalculate all the attrptr[] values
     */
-   for (i = 0, tmp = exec->vtx.vertex ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0, tmp = exec->vtx.vertex ; i < VBO_ATTRIB_MAX ; i++) {
       if (exec->vtx.attrsz[i]) {
 	 exec->vtx.attrptr[i] = tmp;
 	 tmp += exec->vtx.attrsz[i];
@@ -259,7 +254,7 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
 
    /* Copy from current to repopulate the vertex with correct values.
     */
-   brw_exec_copy_from_current( exec );
+   vbo_exec_copy_from_current( exec );
 
    /* Replay stored vertices to translate them
     * to new format here.
@@ -275,7 +270,7 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
       assert(exec->vtx.vbptr == (GLfloat *)exec->vtx.buffer_map);
       
       for (i = 0 ; i < exec->vtx.copied.nr ; i++) {
-	 for (j = 0 ; j < BRW_ATTRIB_MAX ; j++) {
+	 for (j = 0 ; j < VBO_ATTRIB_MAX ; j++) {
 	    if (exec->vtx.attrsz[j]) {
 	       if (j == attr) {
 		  if (oldsz) {
@@ -283,7 +278,8 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
 		     data += oldsz;
 		     dest += newsz;
 		  } else {
-		     COPY_SZ_4V( dest, newsz, exec->vtx.current[j] );
+		     const GLfloat *current = (const GLfloat *)vbo->currval[j].Ptr;
+		     COPY_SZ_4V( dest, newsz, current );
 		     dest += newsz;
 		  }
 	       }
@@ -304,17 +300,17 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
 }
 
 
-static void brw_exec_fixup_vertex( GLcontext *ctx,
+static void vbo_exec_fixup_vertex( GLcontext *ctx,
 				   GLuint attr, GLuint sz )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
    int i;
 
    if (sz > exec->vtx.attrsz[attr]) {
       /* New size is larger.  Need to flush existing vertices and get
        * an enlarged vertex format.
        */
-      brw_exec_wrap_upgrade_vertex( exec, attr, sz );
+      vbo_exec_wrap_upgrade_vertex( exec, attr, sz );
    }
    else if (sz < exec->vtx.active_sz[attr]) {
       static const GLfloat id[4] = { 0, 0, 0, 1 };
@@ -345,10 +341,10 @@ static void brw_exec_fixup_vertex( GLcontext *ctx,
  */
 #define ATTR( A, N, V0, V1, V2, V3 )				\
 do {								\
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;	\
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;	\
 								\
    if (exec->vtx.active_sz[A] != N)				\
-      brw_exec_fixup_vertex(ctx, A, N);			\
+      vbo_exec_fixup_vertex(ctx, A, N);			\
 								\
    {								\
       GLfloat *dest = exec->vtx.attrptr[A];			\
@@ -368,15 +364,15 @@ do {								\
       exec->ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;	\
 								\
       if (++exec->vtx.vert_count >= exec->vtx.max_vert)		\
-	 brw_exec_vtx_wrap( exec );				\
+	 vbo_exec_vtx_wrap( exec );				\
    }								\
 } while (0)
 
 
 #define ERROR() _mesa_error( ctx, GL_INVALID_ENUM, __FUNCTION__ )
-#define TAG(x) brw_##x
+#define TAG(x) vbo_##x
 
-#include "brw_attrib_tmp.h"
+#include "vbo_attrib_tmp.h"
 
 
 
@@ -384,20 +380,20 @@ do {								\
 
 /* Eval
  */
-static void GLAPIENTRY brw_exec_EvalCoord1f( GLfloat u )
+static void GLAPIENTRY vbo_exec_EvalCoord1f( GLfloat u )
 {
    GET_CURRENT_CONTEXT( ctx );
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    {
       GLint i;
       if (exec->eval.recalculate_maps) 
-	 brw_exec_eval_update( exec );
+	 vbo_exec_eval_update( exec );
 
-      for (i = 0 ; i <= BRW_ATTRIB_INDEX ; i++) {
+      for (i = 0 ; i <= VBO_ATTRIB_INDEX ; i++) {
 	 if (exec->eval.map1[i].map) 
 	    if (exec->vtx.active_sz[i] != exec->eval.map1[i].sz)
-	       brw_exec_fixup_vertex( ctx, i, exec->eval.map1[i].sz );
+	       vbo_exec_fixup_vertex( ctx, i, exec->eval.map1[i].sz );
       }
    }
 
@@ -405,64 +401,64 @@ static void GLAPIENTRY brw_exec_EvalCoord1f( GLfloat u )
    _mesa_memcpy( exec->vtx.copied.buffer, exec->vtx.vertex, 
                  exec->vtx.vertex_size * sizeof(GLfloat));
 
-   brw_exec_do_EvalCoord1f( exec, u );
+   vbo_exec_do_EvalCoord1f( exec, u );
 
    _mesa_memcpy( exec->vtx.vertex, exec->vtx.copied.buffer,
                  exec->vtx.vertex_size * sizeof(GLfloat));
 }
 
-static void GLAPIENTRY brw_exec_EvalCoord2f( GLfloat u, GLfloat v )
+static void GLAPIENTRY vbo_exec_EvalCoord2f( GLfloat u, GLfloat v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    {
       GLint i;
       if (exec->eval.recalculate_maps) 
-	 brw_exec_eval_update( exec );
+	 vbo_exec_eval_update( exec );
 
-      for (i = 0 ; i <= BRW_ATTRIB_INDEX ; i++) {
+      for (i = 0 ; i <= VBO_ATTRIB_INDEX ; i++) {
 	 if (exec->eval.map2[i].map) 
 	    if (exec->vtx.active_sz[i] != exec->eval.map2[i].sz)
-	       brw_exec_fixup_vertex( ctx, i, exec->eval.map2[i].sz );
+	       vbo_exec_fixup_vertex( ctx, i, exec->eval.map2[i].sz );
       }
 
       if (ctx->Eval.AutoNormal) 
-	 if (exec->vtx.active_sz[BRW_ATTRIB_NORMAL] != 3)
-	    brw_exec_fixup_vertex( ctx, BRW_ATTRIB_NORMAL, 3 );
+	 if (exec->vtx.active_sz[VBO_ATTRIB_NORMAL] != 3)
+	    vbo_exec_fixup_vertex( ctx, VBO_ATTRIB_NORMAL, 3 );
    }
 
    _mesa_memcpy( exec->vtx.copied.buffer, exec->vtx.vertex, 
                  exec->vtx.vertex_size * sizeof(GLfloat));
 
-   brw_exec_do_EvalCoord2f( exec, u, v );
+   vbo_exec_do_EvalCoord2f( exec, u, v );
 
    _mesa_memcpy( exec->vtx.vertex, exec->vtx.copied.buffer, 
                  exec->vtx.vertex_size * sizeof(GLfloat));
 }
 
-static void GLAPIENTRY brw_exec_EvalCoord1fv( const GLfloat *u )
+static void GLAPIENTRY vbo_exec_EvalCoord1fv( const GLfloat *u )
 {
-   brw_exec_EvalCoord1f( u[0] );
+   vbo_exec_EvalCoord1f( u[0] );
 }
 
-static void GLAPIENTRY brw_exec_EvalCoord2fv( const GLfloat *u )
+static void GLAPIENTRY vbo_exec_EvalCoord2fv( const GLfloat *u )
 {
-   brw_exec_EvalCoord2f( u[0], u[1] );
+   vbo_exec_EvalCoord2f( u[0], u[1] );
 }
 
-static void GLAPIENTRY brw_exec_EvalPoint1( GLint i )
+static void GLAPIENTRY vbo_exec_EvalPoint1( GLint i )
 {
    GET_CURRENT_CONTEXT( ctx );
    GLfloat du = ((ctx->Eval.MapGrid1u2 - ctx->Eval.MapGrid1u1) /
 		 (GLfloat) ctx->Eval.MapGrid1un);
    GLfloat u = i * du + ctx->Eval.MapGrid1u1;
 
-   brw_exec_EvalCoord1f( u );
+   vbo_exec_EvalCoord1f( u );
 }
 
 
-static void GLAPIENTRY brw_exec_EvalPoint2( GLint i, GLint j )
+static void GLAPIENTRY vbo_exec_EvalPoint2( GLint i, GLint j )
 {
    GET_CURRENT_CONTEXT( ctx );
    GLfloat du = ((ctx->Eval.MapGrid2u2 - ctx->Eval.MapGrid2u1) / 
@@ -472,19 +468,19 @@ static void GLAPIENTRY brw_exec_EvalPoint2( GLint i, GLint j )
    GLfloat u = i * du + ctx->Eval.MapGrid2u1;
    GLfloat v = j * dv + ctx->Eval.MapGrid2v1;
 
-   brw_exec_EvalCoord2f( u, v );
+   vbo_exec_EvalCoord2f( u, v );
 }
 
 
 /* Build a list of primitives on the fly.  Keep
  * ctx->Driver.CurrentExecPrimitive uptodate as well.
  */
-static void GLAPIENTRY brw_exec_Begin( GLenum mode )
+static void GLAPIENTRY vbo_exec_Begin( GLenum mode )
 {
    GET_CURRENT_CONTEXT( ctx ); 
 
    if (ctx->Driver.CurrentExecPrimitive == GL_POLYGON+1) {
-      struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+      struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
       int i;
 
       if (ctx->NewState) {
@@ -505,7 +501,7 @@ static void GLAPIENTRY brw_exec_Begin( GLenum mode )
        * begin/end pairs.
        */
       if (exec->vtx.vertex_size && !exec->vtx.attrsz[0]) 
-	 brw_exec_FlushVertices( ctx, ~0 );
+	 vbo_exec_FlushVertices( ctx, ~0 );
 
       i = exec->vtx.prim_count++;
       exec->vtx.prim[i].mode = mode;
@@ -524,20 +520,12 @@ static void GLAPIENTRY brw_exec_Begin( GLenum mode )
       
 }
 
-static GLuint brw_max_prim( GLcontext *ctx )
-{
-        struct intel_context *intel = intel_context( ctx );
-        if (intel->numClipRects <= 1)
-                return BRW_MAX_PRIM;
-        return BRW_MAX_PRIM/intel->numClipRects;
-}
-
-static void GLAPIENTRY brw_exec_End( void )
+static void GLAPIENTRY vbo_exec_End( void )
 {
    GET_CURRENT_CONTEXT( ctx ); 
 
    if (ctx->Driver.CurrentExecPrimitive != GL_POLYGON+1) {
-      struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+      struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
       int idx = exec->vtx.vert_count;
       int i = exec->vtx.prim_count - 1;
 
@@ -546,29 +534,29 @@ static void GLAPIENTRY brw_exec_End( void )
 
       ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1;
 
-      if (exec->vtx.prim_count >= brw_max_prim(ctx))
-	 brw_exec_vtx_flush( exec );	
+      if (exec->vtx.prim_count == VBO_MAX_PRIM)
+	 vbo_exec_vtx_flush( exec );	
    }
    else 
       _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
 }
 
 
-static void brw_exec_vtxfmt_init( struct brw_exec_context *exec )
+static void vbo_exec_vtxfmt_init( struct vbo_exec_context *exec )
 {
    GLvertexformat *vfmt = &exec->vtxfmt;
 
    vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = brw_exec_Begin;
+   vfmt->Begin = vbo_exec_Begin;
    vfmt->CallList = _mesa_CallList;
    vfmt->CallLists = _mesa_CallLists;
-   vfmt->End = brw_exec_End;
-   vfmt->EvalCoord1f = brw_exec_EvalCoord1f;
-   vfmt->EvalCoord1fv = brw_exec_EvalCoord1fv;
-   vfmt->EvalCoord2f = brw_exec_EvalCoord2f;
-   vfmt->EvalCoord2fv = brw_exec_EvalCoord2fv;
-   vfmt->EvalPoint1 = brw_exec_EvalPoint1;
-   vfmt->EvalPoint2 = brw_exec_EvalPoint2;
+   vfmt->End = vbo_exec_End;
+   vfmt->EvalCoord1f = vbo_exec_EvalCoord1f;
+   vfmt->EvalCoord1fv = vbo_exec_EvalCoord1fv;
+   vfmt->EvalCoord2f = vbo_exec_EvalCoord2f;
+   vfmt->EvalCoord2fv = vbo_exec_EvalCoord2fv;
+   vfmt->EvalPoint1 = vbo_exec_EvalPoint1;
+   vfmt->EvalPoint2 = vbo_exec_EvalPoint2;
 
    vfmt->Rectf = _mesa_noop_Rectf;
    vfmt->EvalMesh1 = _mesa_noop_EvalMesh1;
@@ -577,113 +565,101 @@ static void brw_exec_vtxfmt_init( struct brw_exec_context *exec )
 
    /* from attrib_tmp.h:
     */
-   vfmt->Color3f = brw_Color3f;
-   vfmt->Color3fv = brw_Color3fv;
-   vfmt->Color4f = brw_Color4f;
-   vfmt->Color4fv = brw_Color4fv;
-   vfmt->FogCoordfEXT = brw_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = brw_FogCoordfvEXT;
-   vfmt->MultiTexCoord1fARB = brw_MultiTexCoord1f;
-   vfmt->MultiTexCoord1fvARB = brw_MultiTexCoord1fv;
-   vfmt->MultiTexCoord2fARB = brw_MultiTexCoord2f;
-   vfmt->MultiTexCoord2fvARB = brw_MultiTexCoord2fv;
-   vfmt->MultiTexCoord3fARB = brw_MultiTexCoord3f;
-   vfmt->MultiTexCoord3fvARB = brw_MultiTexCoord3fv;
-   vfmt->MultiTexCoord4fARB = brw_MultiTexCoord4f;
-   vfmt->MultiTexCoord4fvARB = brw_MultiTexCoord4fv;
-   vfmt->Normal3f = brw_Normal3f;
-   vfmt->Normal3fv = brw_Normal3fv;
-   vfmt->SecondaryColor3fEXT = brw_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = brw_SecondaryColor3fvEXT;
-   vfmt->TexCoord1f = brw_TexCoord1f;
-   vfmt->TexCoord1fv = brw_TexCoord1fv;
-   vfmt->TexCoord2f = brw_TexCoord2f;
-   vfmt->TexCoord2fv = brw_TexCoord2fv;
-   vfmt->TexCoord3f = brw_TexCoord3f;
-   vfmt->TexCoord3fv = brw_TexCoord3fv;
-   vfmt->TexCoord4f = brw_TexCoord4f;
-   vfmt->TexCoord4fv = brw_TexCoord4fv;
-   vfmt->Vertex2f = brw_Vertex2f;
-   vfmt->Vertex2fv = brw_Vertex2fv;
-   vfmt->Vertex3f = brw_Vertex3f;
-   vfmt->Vertex3fv = brw_Vertex3fv;
-   vfmt->Vertex4f = brw_Vertex4f;
-   vfmt->Vertex4fv = brw_Vertex4fv;
+   vfmt->Color3f = vbo_Color3f;
+   vfmt->Color3fv = vbo_Color3fv;
+   vfmt->Color4f = vbo_Color4f;
+   vfmt->Color4fv = vbo_Color4fv;
+   vfmt->FogCoordfEXT = vbo_FogCoordfEXT;
+   vfmt->FogCoordfvEXT = vbo_FogCoordfvEXT;
+   vfmt->MultiTexCoord1fARB = vbo_MultiTexCoord1f;
+   vfmt->MultiTexCoord1fvARB = vbo_MultiTexCoord1fv;
+   vfmt->MultiTexCoord2fARB = vbo_MultiTexCoord2f;
+   vfmt->MultiTexCoord2fvARB = vbo_MultiTexCoord2fv;
+   vfmt->MultiTexCoord3fARB = vbo_MultiTexCoord3f;
+   vfmt->MultiTexCoord3fvARB = vbo_MultiTexCoord3fv;
+   vfmt->MultiTexCoord4fARB = vbo_MultiTexCoord4f;
+   vfmt->MultiTexCoord4fvARB = vbo_MultiTexCoord4fv;
+   vfmt->Normal3f = vbo_Normal3f;
+   vfmt->Normal3fv = vbo_Normal3fv;
+   vfmt->SecondaryColor3fEXT = vbo_SecondaryColor3fEXT;
+   vfmt->SecondaryColor3fvEXT = vbo_SecondaryColor3fvEXT;
+   vfmt->TexCoord1f = vbo_TexCoord1f;
+   vfmt->TexCoord1fv = vbo_TexCoord1fv;
+   vfmt->TexCoord2f = vbo_TexCoord2f;
+   vfmt->TexCoord2fv = vbo_TexCoord2fv;
+   vfmt->TexCoord3f = vbo_TexCoord3f;
+   vfmt->TexCoord3fv = vbo_TexCoord3fv;
+   vfmt->TexCoord4f = vbo_TexCoord4f;
+   vfmt->TexCoord4fv = vbo_TexCoord4fv;
+   vfmt->Vertex2f = vbo_Vertex2f;
+   vfmt->Vertex2fv = vbo_Vertex2fv;
+   vfmt->Vertex3f = vbo_Vertex3f;
+   vfmt->Vertex3fv = vbo_Vertex3fv;
+   vfmt->Vertex4f = vbo_Vertex4f;
+   vfmt->Vertex4fv = vbo_Vertex4fv;
    
-   vfmt->VertexAttrib1fARB = brw_VertexAttrib1fARB;
-   vfmt->VertexAttrib1fvARB = brw_VertexAttrib1fvARB;
-   vfmt->VertexAttrib2fARB = brw_VertexAttrib2fARB;
-   vfmt->VertexAttrib2fvARB = brw_VertexAttrib2fvARB;
-   vfmt->VertexAttrib3fARB = brw_VertexAttrib3fARB;
-   vfmt->VertexAttrib3fvARB = brw_VertexAttrib3fvARB;
-   vfmt->VertexAttrib4fARB = brw_VertexAttrib4fARB;
-   vfmt->VertexAttrib4fvARB = brw_VertexAttrib4fvARB;
-
-   vfmt->VertexAttrib1fNV = brw_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = brw_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV = brw_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = brw_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV = brw_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = brw_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV = brw_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = brw_VertexAttrib4fvNV;
-
-   vfmt->Materialfv = brw_Materialfv;
-
-   vfmt->EdgeFlag = brw_EdgeFlag;
-   vfmt->Indexf = brw_Indexf;
-   vfmt->Indexfv = brw_Indexfv;
+   vfmt->VertexAttrib1fARB = vbo_VertexAttrib1fARB;
+   vfmt->VertexAttrib1fvARB = vbo_VertexAttrib1fvARB;
+   vfmt->VertexAttrib2fARB = vbo_VertexAttrib2fARB;
+   vfmt->VertexAttrib2fvARB = vbo_VertexAttrib2fvARB;
+   vfmt->VertexAttrib3fARB = vbo_VertexAttrib3fARB;
+   vfmt->VertexAttrib3fvARB = vbo_VertexAttrib3fvARB;
+   vfmt->VertexAttrib4fARB = vbo_VertexAttrib4fARB;
+   vfmt->VertexAttrib4fvARB = vbo_VertexAttrib4fvARB;
+
+   vfmt->VertexAttrib1fNV = vbo_VertexAttrib1fNV;
+   vfmt->VertexAttrib1fvNV = vbo_VertexAttrib1fvNV;
+   vfmt->VertexAttrib2fNV = vbo_VertexAttrib2fNV;
+   vfmt->VertexAttrib2fvNV = vbo_VertexAttrib2fvNV;
+   vfmt->VertexAttrib3fNV = vbo_VertexAttrib3fNV;
+   vfmt->VertexAttrib3fvNV = vbo_VertexAttrib3fvNV;
+   vfmt->VertexAttrib4fNV = vbo_VertexAttrib4fNV;
+   vfmt->VertexAttrib4fvNV = vbo_VertexAttrib4fvNV;
+
+   vfmt->Materialfv = vbo_Materialfv;
+
+   vfmt->EdgeFlag = vbo_EdgeFlag;
+   vfmt->Indexf = vbo_Indexf;
+   vfmt->Indexfv = vbo_Indexfv;
 
 }
 
 
-static void brw_exec_current_init( struct brw_exec_context *exec ) 
-{
-   GLcontext *ctx = exec->ctx;
-   GLint i;
-
-   /* setup the pointers for the typical 16 vertex attributes */
-   for (i = 0; i < BRW_ATTRIB_FIRST_MATERIAL; i++) 
-      exec->vtx.current[i] = ctx->Current.Attrib[i];
-
-   /* setup pointers for the 12 material attributes */
-   for (i = 0; i < MAT_ATTRIB_MAX; i++)
-      exec->vtx.current[BRW_ATTRIB_FIRST_MATERIAL + i] = 
-	 ctx->Light.Material.Attrib[i];
-
-   exec->vtx.current[BRW_ATTRIB_INDEX] = &ctx->Current.Index;
-   exec->vtx.current[BRW_ATTRIB_EDGEFLAG] = &exec->vtx.CurrentFloatEdgeFlag;
-}
-
-void brw_exec_vtx_init( struct brw_exec_context *exec )
+void vbo_exec_vtx_init( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLuint i;
 
    /* Allocate a buffer object.  Will just reuse this object
     * continuously.
     */
    exec->vtx.bufferobj = ctx->Array.NullBufferObj;
-   exec->vtx.buffer_map = ALIGN_MALLOC(BRW_VERT_BUFFER_SIZE * sizeof(GLfloat), 64);
+   exec->vtx.buffer_map = ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE * sizeof(GLfloat), 64);
 
-   brw_exec_current_init( exec );
-   brw_exec_vtxfmt_init( exec );
+   vbo_exec_vtxfmt_init( exec );
 
    /* Hook our functions into the dispatch table.
     */
    _mesa_install_exec_vtxfmt( exec->ctx, &exec->vtxfmt );
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
       exec->vtx.attrsz[i] = 0;
       exec->vtx.active_sz[i] = 0;
       exec->vtx.inputs[i] = &exec->vtx.arrays[i];
    }
- 
+   
+   {
+      struct gl_client_array *arrays = exec->vtx.arrays;
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+   }
+
    exec->vtx.vertex_size = 0;
 }
 
 
-void brw_exec_vtx_destroy( struct brw_exec_context *exec )
+void vbo_exec_vtx_destroy( struct vbo_exec_context *exec )
 {
    if (exec->vtx.buffer_map) {
       ALIGN_FREE(exec->vtx.buffer_map);
@@ -692,19 +668,19 @@ void brw_exec_vtx_destroy( struct brw_exec_context *exec )
 }
 
 
-void brw_exec_FlushVertices( GLcontext *ctx, GLuint flags )
+void vbo_exec_FlushVertices( GLcontext *ctx, GLuint flags )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    if (exec->ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END)
       return;
 
    if (exec->vtx.vert_count) {
-      brw_exec_vtx_flush( exec );
+      vbo_exec_vtx_flush( exec );
    }
 
    if (exec->vtx.vertex_size) {
-      brw_exec_copy_to_current( exec );
+      vbo_exec_copy_to_current( exec );
       reset_attrfv( exec );
    }
 
@@ -712,11 +688,11 @@ void brw_exec_FlushVertices( GLcontext *ctx, GLuint flags )
 }
 
 
-static void reset_attrfv( struct brw_exec_context *exec )
+static void reset_attrfv( struct vbo_exec_context *exec )
 {   
    GLuint i;
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
       exec->vtx.attrsz[i] = 0;
       exec->vtx.active_sz[i] = 0;
    }
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
new file mode 100644
index 0000000000..f96df5f164
--- /dev/null
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -0,0 +1,387 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "glheader.h"
+#include "context.h"
+#include "state.h"
+#include "api_validate.h"
+#include "api_noop.h"
+#include "dispatch.h"
+
+#include "vbo_context.h"
+
+/* Compute min and max elements for drawelements calls.
+ */
+static void get_minmax_index( GLuint count, GLuint type, 
+			      const GLvoid *indices,
+			      GLuint *min_index,
+			      GLuint *max_index)
+{
+   GLint i;
+
+   switch(type) {
+   case GL_UNSIGNED_INT: {
+      const GLuint *ui_indices = (const GLuint *)indices;
+      GLuint max_ui = ui_indices[0];
+      GLuint min_ui = ui_indices[0];
+      for (i = 1; i < count; i++) {
+	 if (ui_indices[i] > max_ui) max_ui = ui_indices[i];
+	 if (ui_indices[i] < min_ui) min_ui = ui_indices[i];
+      }
+      *min_index = min_ui;
+      *max_index = max_ui;
+      break;
+   }
+   case GL_UNSIGNED_SHORT: {
+      const GLushort *us_indices = (const GLushort *)indices;
+      GLuint max_us = us_indices[0];
+      GLuint min_us = us_indices[0];
+      for (i = 1; i < count; i++) {
+	 if (us_indices[i] > max_us) max_us = us_indices[i];
+	 if (us_indices[i] < min_us) min_us = us_indices[i];
+      }
+      *min_index = min_us;
+      *max_index = max_us;
+      break;
+   }
+   case GL_UNSIGNED_BYTE: {
+      const GLubyte *ub_indices = (const GLubyte *)indices;
+      GLuint max_ub = ub_indices[0];
+      GLuint min_ub = ub_indices[0];
+      for (i = 1; i < count; i++) {
+	 if (ub_indices[i] > max_ub) max_ub = ub_indices[i];
+	 if (ub_indices[i] < min_ub) min_ub = ub_indices[i];
+      }
+      *min_index = min_ub;
+      *max_index = max_ub;
+      break;
+   }
+   default:
+      assert(0);
+      break;
+   }
+}
+
+
+/* Just translate the arrayobj into a sane layout.
+ */
+static void bind_array_obj( GLcontext *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   GLuint i;
+
+   /* TODO: Fix the ArrayObj struct to keep legacy arrays in an array
+    * rather than as individual named arrays.  Then this function can
+    * go away.
+    */
+   exec->array.legacy_array[VERT_ATTRIB_POS] = &ctx->Array.ArrayObj->Vertex;
+   exec->array.legacy_array[VERT_ATTRIB_WEIGHT] = &vbo->legacy_currval[VERT_ATTRIB_WEIGHT];
+   exec->array.legacy_array[VERT_ATTRIB_NORMAL] = &ctx->Array.ArrayObj->Normal;
+   exec->array.legacy_array[VERT_ATTRIB_COLOR0] = &ctx->Array.ArrayObj->Color;
+   exec->array.legacy_array[VERT_ATTRIB_COLOR1] = &ctx->Array.ArrayObj->SecondaryColor;
+   exec->array.legacy_array[VERT_ATTRIB_FOG] = &ctx->Array.ArrayObj->FogCoord;
+   exec->array.legacy_array[VERT_ATTRIB_COLOR_INDEX] = &ctx->Array.ArrayObj->Index;
+   exec->array.legacy_array[VERT_ATTRIB_EDGEFLAG] = &ctx->Array.ArrayObj->EdgeFlag;
+
+   for (i = 0; i < 8; i++)
+      exec->array.legacy_array[VERT_ATTRIB_TEX0 + i] = &ctx->Array.ArrayObj->TexCoord[i];
+
+   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      exec->array.generic_array[i] = &ctx->Array.ArrayObj->VertexAttrib[i];
+   
+   exec->array.array_obj = ctx->Array.ArrayObj->Name;
+}
+
+static void recalculate_input_bindings( GLcontext *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   const struct gl_client_array **inputs = &exec->array.inputs[0];
+   GLuint i;
+
+   exec->array.program_mode = get_program_mode(ctx);
+   exec->array.enabled_flags = ctx->Array.ArrayObj->_Enabled;
+
+   switch (exec->array.program_mode) {
+   case VP_NONE:
+      /* When no vertex program is active, we put the material values
+       * into the generic slots.  This is the only situation where
+       * material values are available as per-vertex attributes.
+       */
+      for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
+	 if (exec->array.legacy_array[i]->Enabled)
+	    inputs[i] = exec->array.legacy_array[i];
+	 else
+	    inputs[i] = &vbo->legacy_currval[i];
+      }
+
+      for (i = 0; i < MAT_ATTRIB_MAX; i++) {
+	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->mat_currval[i];
+      }
+
+      /* Could use just about anything, just to fill in the empty
+       * slots:
+       */
+      for (i = MAT_ATTRIB_MAX; i < VERT_ATTRIB_MAX; i++)
+	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
+
+      break;
+   case VP_NV:
+      /* NV_vertex_program - attribute arrays alias and override
+       * conventional, legacy arrays.  No materials, and the generic
+       * slots are vacant.
+       */
+      for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
+	 if (exec->array.generic_array[i]->Enabled)
+	    inputs[i] = exec->array.generic_array[i];
+	 else if (exec->array.legacy_array[i]->Enabled)
+	    inputs[i] = exec->array.legacy_array[i];
+	 else
+	    inputs[i] = &vbo->legacy_currval[i];
+      }
+
+      /* Could use just about anything, just to fill in the empty
+       * slots:
+       */
+      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++)
+	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
+
+      break;
+   case VP_ARB:
+      /* ARB_vertex_program - Only the attribute zero (position) array
+       * aliases and overrides the legacy position array.  
+       *
+       * Otherwise, legacy attributes available in the legacy slots,
+       * generic attributes in the generic slots and materials are not
+       * available as per-vertex attributes.
+       */
+      if (exec->array.generic_array[0]->Enabled)
+	 inputs[0] = exec->array.generic_array[0];
+      else if (exec->array.legacy_array[0]->Enabled)
+	 inputs[0] = exec->array.legacy_array[0];
+      else
+	 inputs[0] = &vbo->legacy_currval[0];
+
+
+      for (i = 1; i <= VERT_ATTRIB_TEX7; i++) {
+	 if (exec->array.legacy_array[i]->Enabled)
+	    inputs[i] = exec->array.legacy_array[i];
+	 else
+	    inputs[i] = &vbo->legacy_currval[i];
+      }
+
+      for (i = 0; i < 16; i++) {
+	 if (exec->array.generic_array[0]->Enabled)
+	    inputs[VERT_ATTRIB_GENERIC0 + i] = exec->array.generic_array[i];
+	 else
+	    inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+      }
+      break;
+   }
+}
+
+static void bind_arrays( GLcontext *ctx )
+{
+#if 0
+   if (ctx->Array.ArrayObj.Name != exec->array.array_obj) {
+      bind_array_obj(ctx);
+      recalculate_input_bindings(ctx);
+   }
+   else if (exec->array.program_mode != get_program_mode(ctx) ||
+	    exec->array.enabled_flags != ctx->Array.ArrayObj->_Enabled) {
+      
+      recalculate_input_bindings(ctx);
+   }
+#else
+   bind_array_obj(ctx);
+   recalculate_input_bindings(ctx);
+#endif
+}
+
+
+
+/***********************************************************************
+ * API functions.
+ */
+
+static void GLAPIENTRY
+vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   struct _mesa_prim prim[1];
+
+   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
+      return;
+
+   FLUSH_CURRENT( ctx, 0 );
+
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+      
+   bind_arrays( ctx );
+
+   prim[0].begin = 1;
+   prim[0].end = 1;
+   prim[0].weak = 0;
+   prim[0].pad = 0;
+   prim[0].mode = mode;
+   prim[0].start = start;
+   prim[0].count = count;
+   prim[0].indexed = 0;
+
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, NULL, start, start + count - 1 );
+}
+
+
+
+static void GLAPIENTRY
+vbo_exec_DrawRangeElements(GLenum mode,
+			   GLuint start, GLuint end,
+			   GLsizei count, GLenum type, const GLvoid *indices)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   struct _mesa_index_buffer ib;
+   struct _mesa_prim prim[1];
+
+   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
+      return;
+
+   FLUSH_CURRENT( ctx, 0 );
+
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
+   bind_arrays( ctx );
+
+   ib.count = count;
+   ib.type = type; 
+   ib.obj = ctx->Array.ElementArrayBufferObj;
+   ib.ptr = indices;
+
+   prim[0].begin = 1;
+   prim[0].end = 1;
+   prim[0].weak = 0;
+   prim[0].pad = 0;
+   prim[0].mode = mode;
+   prim[0].start = 0;
+   prim[0].count = count;
+   prim[0].indexed = 1;
+
+   /* Need to give special consideration to rendering a range of
+    * indices starting somewhere above zero.  Typically the
+    * application is issuing multiple DrawRangeElements() to draw
+    * successive primitives layed out linearly in the vertex arrays.
+    * Unless the vertex arrays are all in a VBO (or locked as with
+    * CVA), the OpenGL semantics imply that we need to re-read or
+    * re-upload the vertex data on each draw call.  
+    *
+    * In the case of hardware tnl, we want to avoid starting the
+    * upload at zero, as it will mean every draw call uploads an
+    * increasing amount of not-used vertex data.  Worse - in the
+    * software tnl module, all those vertices might be transformed and
+    * lit but never rendered.
+    *
+    * If we just upload or transform the vertices in start..end,
+    * however, the indices will be incorrect.
+    *
+    * At this level, we don't know exactly what the requirements of
+    * the backend are going to be, though it will likely boil down to
+    * either:
+    *
+    * 1) Do nothing, everything is in a VBO and is processed once
+    *       only.
+    *
+    * 2) Adjust the indices and vertex arrays so that start becomes
+    *    zero.
+    *
+    * Rather than doing anything here, I'll provide a helper function
+    * for the latter case elsewhere.
+    */
+
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, &ib, start, end );
+}
+
+static void GLAPIENTRY
+vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLuint min_index = 0;
+   GLuint max_index = 0;
+
+   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
+      return;
+
+   if (ctx->Array.ElementArrayBufferObj->Name) {
+      const GLvoid *map = ctx->Driver.MapBuffer(ctx,
+						 GL_ELEMENT_ARRAY_BUFFER_ARB,
+						 GL_READ_ONLY,
+						 ctx->Array.ElementArrayBufferObj);
+
+      get_minmax_index(count, type, ADD_POINTERS(map, indices), &min_index, &max_index);
+
+      ctx->Driver.UnmapBuffer(ctx,
+			      GL_ELEMENT_ARRAY_BUFFER_ARB,
+			      ctx->Array.ElementArrayBufferObj);
+   }
+   else {
+      get_minmax_index(count, type, indices, &min_index, &max_index);
+   }
+
+   vbo_exec_DrawRangeElements(mode, min_index, max_index, count, type, indices);
+}
+
+
+/***********************************************************************
+ * Initialization
+ */
+
+
+
+
+void vbo_exec_array_init( struct vbo_exec_context *exec )
+{
+#if 1
+   exec->vtxfmt.DrawArrays = vbo_exec_DrawArrays;
+   exec->vtxfmt.DrawElements = vbo_exec_DrawElements;
+   exec->vtxfmt.DrawRangeElements = vbo_exec_DrawRangeElements;
+#else
+   exec->vtxfmt.DrawArrays = _mesa_noop_DrawArrays;
+   exec->vtxfmt.DrawElements = _mesa_noop_DrawElements;
+   exec->vtxfmt.DrawRangeElements = _mesa_noop_DrawRangeElements;
+#endif
+}
+
+
+void vbo_exec_array_destroy( struct vbo_exec_context *exec )
+{
+   /* nothing to do */
+}
diff --git a/src/mesa/drivers/dri/i965/brw_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 62bda9845b..2b52d7a1ff 100644
--- a/src/mesa/drivers/dri/i965/brw_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -31,12 +31,10 @@
 #include "state.h"
 #include "macros.h"
 
-#include "brw_exec.h"
-#include "brw_draw.h"
-#include "brw_fallback.h"
+#include "vbo_context.h"
 
 
-static void brw_exec_debug_verts( struct brw_exec_context *exec )
+static void vbo_exec_debug_verts( struct vbo_exec_context *exec )
 {
    GLuint count = exec->vtx.vert_count;
    GLuint i;
@@ -48,7 +46,7 @@ static void brw_exec_debug_verts( struct brw_exec_context *exec )
 		exec->vtx.vertex_size);
 
    for (i = 0 ; i < exec->vtx.prim_count ; i++) {
-      struct brw_draw_prim *prim = &exec->vtx.prim[i];
+      struct _mesa_prim *prim = &exec->vtx.prim[i];
       _mesa_printf("   prim %d: %s%s %d..%d %s %s\n",
 		   i, 
 		   _mesa_lookup_enum_by_nr(prim->mode),
@@ -65,7 +63,7 @@ static void brw_exec_debug_verts( struct brw_exec_context *exec )
  * NOTE: Need to have calculated primitives by this point -- do it on the fly.
  * NOTE: Old 'parity' issue is gone.
  */
-static GLuint brw_copy_vertices( struct brw_exec_context *exec )
+static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
 {
    GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
    GLuint ovf, i;
@@ -116,6 +114,11 @@ static GLuint brw_copy_vertices( struct brw_exec_context *exec )
 	 return 2;
       }
    case GL_TRIANGLE_STRIP:
+      /* no parity issue, but need to make sure the tri is not drawn twice */
+      if (nr & 1) {
+	 exec->vtx.prim[exec->vtx.prim_count-1].count--;
+      }
+      /* fallthrough */
    case GL_QUAD_STRIP:
       switch (nr) {
       case 0: ovf = 0; break;
@@ -134,24 +137,49 @@ static GLuint brw_copy_vertices( struct brw_exec_context *exec )
 }
 
 
+
 /* TODO: populate these as the vertex is defined:
  */
-static void brw_exec_bind_arrays( struct brw_exec_context *exec )
+static void vbo_exec_bind_arrays( GLcontext *ctx )
 {
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
    struct gl_client_array *arrays = exec->vtx.arrays;
    GLuint count = exec->vtx.vert_count;
    GLubyte *data = exec->vtx.buffer_map;
+   const GLuint *map;
    GLuint attr;
 
-   memset(arrays, 0, BRW_ATTRIB_MAX * sizeof(arrays[0]));
+   /* Install the default (ie Current) attributes first, then overlay
+    * all active ones.
+    */
+   switch (get_program_mode(exec->ctx)) {
+   case VP_NONE:
+      memcpy(arrays,      vbo->legacy_currval, 16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->mat_currval,    MAT_ATTRIB_MAX * sizeof(arrays[0]));
+      map = vbo->map_vp_none;
+      break;
+   case VP_NV:
+   case VP_ARB:
+      /* The aliasing of attributes for NV vertex programs has already
+       * occurred.  NV vertex programs cannot access material values,
+       * nor attributes greater than VERT_ATTRIB_TEX7.  
+       */
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+      map = vbo->map_vp_arb;
+      break;
+   }
 
    /* Make all active attributes (including edgeflag) available as
     * arrays of floats.
     */
-   for (attr = 0; attr < BRW_ATTRIB_MAX ; attr++) {
-      if (exec->vtx.attrsz[attr]) {
+   for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
+      GLuint src = map[attr];
+
+      if (exec->vtx.attrsz[src]) {
 	 arrays[attr].Ptr = (void *)data;
-	 arrays[attr].Size = exec->vtx.attrsz[attr];
+	 arrays[attr].Size = exec->vtx.attrsz[src];
 	 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Type = GL_FLOAT;
@@ -165,59 +193,32 @@ static void brw_exec_bind_arrays( struct brw_exec_context *exec )
 }
 
 
-static
-void brw_exec_loopback_vertex_list( GLcontext *ctx,
-				    struct brw_exec_context *exec )
-{
-   const GLfloat *buffer = (const GLfloat *)exec->vtx.buffer_map;
-
-   brw_fallback(ctx);
-
-   brw_loopback_vertex_list( ctx,
-			     buffer,
-			     exec->vtx.attrsz,
-			     exec->vtx.prim,
-			     exec->vtx.prim_count,
-			     0,	/* XXX - copied.nr */
-			     exec->vtx.vertex_size);
-
-
-   brw_unfallback(ctx);
-}
-
-
 /**
  * Execute the buffer and save copied verts.
  */
-void brw_exec_vtx_flush( struct brw_exec_context *exec )
+void vbo_exec_vtx_flush( struct vbo_exec_context *exec )
 {
    if (0)
-      brw_exec_debug_verts( exec );
+      vbo_exec_debug_verts( exec );
 
 
    if (exec->vtx.prim_count && 
        exec->vtx.vert_count) {
 
-      exec->vtx.copied.nr = brw_copy_vertices( exec ); 
+      exec->vtx.copied.nr = vbo_copy_vertices( exec ); 
 
       if (exec->vtx.copied.nr != exec->vtx.vert_count) {
 	 GLcontext *ctx = exec->ctx;
 
-	 brw_exec_bind_arrays( exec );
-
-
-	 if (!brw_draw_prims( ctx, 
-			      exec->vtx.inputs, 
-			      exec->vtx.prim, 
-			      exec->vtx.prim_count,
-			      NULL,
-			      0,
-			      exec->vtx.vert_count,
-			      0 )) {
-	    /* Fallback path: 
-	     */
-	    brw_exec_loopback_vertex_list(ctx, exec);
-	 }
+	 vbo_exec_bind_arrays( ctx );
+
+	 vbo_context(ctx)->draw_prims( ctx, 
+				       exec->vtx.inputs, 
+				       exec->vtx.prim, 
+				       exec->vtx.prim_count,
+				       NULL,
+				       0,
+				       exec->vtx.vert_count );
       }
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_exec_eval.c b/src/mesa/vbo/vbo_exec_eval.c
index 9dbeb1b58e..99a8044576 100644
--- a/src/mesa/drivers/dri/i965/brw_exec_eval.c
+++ b/src/mesa/vbo/vbo_exec_eval.c
@@ -30,21 +30,21 @@
 #include "context.h"
 #include "macros.h"
 #include "math/m_eval.h"
-#include "brw_exec.h"
+#include "vbo_exec.h"
 #include "dispatch.h"
 
 
-static void clear_active_eval1( struct brw_exec_context *exec, GLuint attr ) 
+static void clear_active_eval1( struct vbo_exec_context *exec, GLuint attr ) 
 {
    exec->eval.map1[attr].map = NULL;
 }
 
-static void clear_active_eval2( struct brw_exec_context *exec, GLuint attr ) 
+static void clear_active_eval2( struct vbo_exec_context *exec, GLuint attr ) 
 {
    exec->eval.map2[attr].map = NULL;
 }
 
-static void set_active_eval1( struct brw_exec_context *exec, GLuint attr, GLuint dim, 
+static void set_active_eval1( struct vbo_exec_context *exec, GLuint attr, GLuint dim, 
 			      struct gl_1d_map *map )
 {
    if (!exec->eval.map1[attr].map) {
@@ -53,7 +53,7 @@ static void set_active_eval1( struct brw_exec_context *exec, GLuint attr, GLuint
    }
 } 
 
-static void set_active_eval2( struct brw_exec_context *exec, GLuint attr, GLuint dim, 
+static void set_active_eval2( struct vbo_exec_context *exec, GLuint attr, GLuint dim, 
 			      struct gl_2d_map *map )
 {
    if (!exec->eval.map2[attr].map) {
@@ -62,21 +62,21 @@ static void set_active_eval2( struct brw_exec_context *exec, GLuint attr, GLuint
    }
 } 
 
-void brw_exec_eval_update( struct brw_exec_context *exec )
+void vbo_exec_eval_update( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
    GLuint attr;
 
    /* Vertex program maps have priority over conventional attribs */
 
-   for (attr = 0; attr < BRW_ATTRIB_FIRST_MATERIAL; attr++) {
+   for (attr = 0; attr < VBO_ATTRIB_FIRST_MATERIAL; attr++) {
       clear_active_eval1( exec, attr );
       clear_active_eval2( exec, attr );
    }
 
    /* _NEW_PROGRAM */
    if (ctx->VertexProgram._Enabled) {
-      for (attr = 0; attr < BRW_ATTRIB_FIRST_MATERIAL; attr++) {
+      for (attr = 0; attr < VBO_ATTRIB_FIRST_MATERIAL; attr++) {
 	 /* _NEW_EVAL */
 	 if (ctx->Eval.Map1Attrib[attr]) 
 	    set_active_eval1( exec, attr, 4, &ctx->EvalMap.Map1Attrib[attr] );
@@ -87,55 +87,55 @@ void brw_exec_eval_update( struct brw_exec_context *exec )
    }
 
    if (ctx->Eval.Map1Color4) 
-      set_active_eval1( exec, BRW_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
+      set_active_eval1( exec, VBO_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
       
    if (ctx->Eval.Map2Color4) 
-      set_active_eval2( exec, BRW_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
+      set_active_eval2( exec, VBO_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
 
    if (ctx->Eval.Map1TextureCoord4) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
    else if (ctx->Eval.Map1TextureCoord3) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
    else if (ctx->Eval.Map1TextureCoord2) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
    else if (ctx->Eval.Map1TextureCoord1) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
 
    if (ctx->Eval.Map2TextureCoord4) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
    else if (ctx->Eval.Map2TextureCoord3) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
    else if (ctx->Eval.Map2TextureCoord2) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
    else if (ctx->Eval.Map2TextureCoord1) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
 
    if (ctx->Eval.Map1Normal) 
-      set_active_eval1( exec, BRW_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
+      set_active_eval1( exec, VBO_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
 
    if (ctx->Eval.Map2Normal) 
-      set_active_eval2( exec, BRW_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
+      set_active_eval2( exec, VBO_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
 
    if (ctx->Eval.Map1Vertex4) 
-      set_active_eval1( exec, BRW_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
+      set_active_eval1( exec, VBO_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
    else if (ctx->Eval.Map1Vertex3) 
-      set_active_eval1( exec, BRW_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
+      set_active_eval1( exec, VBO_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
 
    if (ctx->Eval.Map2Vertex4) 
-      set_active_eval2( exec, BRW_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
+      set_active_eval2( exec, VBO_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
    else if (ctx->Eval.Map2Vertex3) 
-      set_active_eval2( exec, BRW_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
+      set_active_eval2( exec, VBO_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
 
    exec->eval.recalculate_maps = 0;
 }
 
 
 
-void brw_exec_do_EvalCoord1f(struct brw_exec_context *exec, GLfloat u)
+void vbo_exec_do_EvalCoord1f(struct vbo_exec_context *exec, GLfloat u)
 {
    GLuint attr;
 
-   for (attr = 1; attr <= BRW_ATTRIB_INDEX; attr++) {
+   for (attr = 1; attr <= VBO_ATTRIB_INDEX; attr++) {
       struct gl_1d_map *map = exec->eval.map1[attr].map;
       if (map) {
 	 GLfloat uu = (u - map->u1) * map->du;
@@ -175,12 +175,12 @@ void brw_exec_do_EvalCoord1f(struct brw_exec_context *exec, GLfloat u)
 
 
 
-void brw_exec_do_EvalCoord2f( struct brw_exec_context *exec, 
+void vbo_exec_do_EvalCoord2f( struct vbo_exec_context *exec, 
 			      GLfloat u, GLfloat v )
 {   
    GLuint attr;
 
-   for (attr = 1; attr <= BRW_ATTRIB_INDEX; attr++) {
+   for (attr = 1; attr <= VBO_ATTRIB_INDEX; attr++) {
       struct gl_2d_map *map = exec->eval.map2[attr].map;
       if (map) {
 	 GLfloat uu = (u - map->u1) * map->du;
@@ -234,8 +234,8 @@ void brw_exec_do_EvalCoord2f( struct brw_exec_context *exec,
          NORMALIZE_3FV(normal);
 	 normal[3] = 1.0;
 
-	 COPY_SZ_4V( exec->vtx.attrptr[BRW_ATTRIB_NORMAL],
-		     exec->vtx.attrsz[BRW_ATTRIB_NORMAL],
+	 COPY_SZ_4V( exec->vtx.attrptr[VBO_ATTRIB_NORMAL],
+		     exec->vtx.attrsz[VBO_ATTRIB_NORMAL],
 		     normal );
 
       }
diff --git a/src/mesa/vbo/vbo_rebase.c b/src/mesa/vbo/vbo_rebase.c
new file mode 100644
index 0000000000..bc4211d852
--- /dev/null
+++ b/src/mesa/vbo/vbo_rebase.c
@@ -0,0 +1,222 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Helper for drivers which find themselves rendering a range of
+ * indices starting somewhere above zero.  Typically the application
+ * is issuing multiple DrawArrays() or DrawElements() to draw
+ * successive primitives layed out linearly in the vertex arrays.
+ * Unless the vertex arrays are all in a VBO, the OpenGL semantics
+ * imply that we need to re-upload the vertex data on each draw call.
+ * In that case, we want to avoid starting the upload at zero, as it
+ * will mean every draw call uploads an increasing amount of not-used
+ * vertex data.  Worse - in the software tnl module, all those
+ * vertices will be transformed and lit.
+ *
+ * If we just upload the new data, however, the indices will be
+ * incorrect as we tend to upload each set of vertex data to a new
+ * region.  
+ *
+ * This file provides a helper to adjust the arrays, primitives and
+ * indices of a draw call so that it can be re-issued with a min_index
+ * of zero.
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mtypes.h"
+
+#include "vbo.h"
+
+
+#define REBASE(TYPE) 						\
+static void *rebase_##TYPE( const void *ptr,			\
+			  GLuint count, 			\
+			  TYPE min_index )			\
+{								\
+   const TYPE *in = (TYPE *)ptr;				\
+   TYPE *tmp_indices = malloc(count * sizeof(TYPE));	\
+   GLuint i;							\
+								\
+   for (i = 0; i < count; i++)  				\
+      tmp_indices[i] = in[i] - min_index;			\
+								\
+   return (void *)tmp_indices;					\
+}
+
+
+REBASE(GLuint)
+REBASE(GLushort)
+REBASE(GLubyte)
+
+GLboolean vbo_all_varyings_in_vbos( const struct gl_client_array *arrays[] )
+{
+   GLuint i;
+   
+   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      if (arrays[i]->StrideB &&
+	  arrays[i]->BufferObj->Name == 0)
+	 return GL_FALSE;
+
+   return GL_TRUE;
+}
+
+/* Adjust primitives, indices and vertex definitions so that min_index
+ * becomes zero. There are lots of reasons for wanting to do this, eg:
+ *
+ * Software tnl:
+ *    - any time min_index != 0, otherwise unused vertices lower than
+ *      min_index will be transformed.
+ *
+ * Hardware tnl:
+ *    - if ib != NULL and min_index != 0, otherwise vertices lower than 
+ *      min_index will be uploaded.  Requires adjusting index values.
+ *
+ *    - if ib == NULL and min_index != 0, just for convenience so this doesn't
+ *      have to be handled within the driver.
+ *
+ * Hardware tnl with VBO support:
+ *    - as above, but only when vertices are not (all?) in VBO's.
+ *    - can't save time by trying to upload half a vbo - typically it is
+ *      all or nothing.
+ */
+void vbo_rebase_prims( GLcontext *ctx,
+		       const struct gl_client_array *arrays[],
+		       const struct _mesa_prim *prim,
+		       GLuint nr_prims,
+		       const struct _mesa_index_buffer *ib,
+		       GLuint min_index,
+		       GLuint max_index,
+		       vbo_draw_func draw )
+{
+   struct gl_client_array tmp_arrays[VERT_ATTRIB_MAX];
+   const struct gl_client_array *tmp_array_pointers[VERT_ATTRIB_MAX];
+
+   struct _mesa_index_buffer tmp_ib;
+   struct _mesa_prim *tmp_prims = NULL;
+   void *tmp_indices = NULL;
+   GLuint i;
+
+   assert(min_index != 0);
+
+   if (0)
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
+
+   if (ib) {
+      /* Unfortunately need to adjust each index individually.
+       */
+      GLboolean map_ib = ib->obj->Name && !ib->obj->Pointer;
+      void *ptr;
+
+      if (map_ib) 
+	 ctx->Driver.MapBuffer(ctx, 
+			       GL_ELEMENT_ARRAY_BUFFER,
+			       GL_READ_ONLY_ARB,
+			       ib->obj);
+
+
+      ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);
+
+      /* Some users might prefer it if we translated elements to
+       * GLuints here.  Others wouldn't...
+       */
+      switch (ib->type) {
+      case GL_UNSIGNED_INT: 
+	 tmp_indices = rebase_GLuint( ptr, ib->count, min_index );
+	 break;
+      case GL_UNSIGNED_SHORT: 
+	 tmp_indices = rebase_GLushort( ptr, ib->count, min_index );
+	 break;
+      case GL_UNSIGNED_BYTE: 
+	 tmp_indices = rebase_GLubyte( ptr, ib->count, min_index );
+	 break;
+      }      
+
+      if (map_ib) 
+	 ctx->Driver.UnmapBuffer(ctx, 
+				 GL_ELEMENT_ARRAY_BUFFER,
+				 ib->obj);
+
+      tmp_ib.obj = ctx->Array.NullBufferObj;
+      tmp_ib.ptr = tmp_indices;
+      tmp_ib.count = ib->count;
+      tmp_ib.type = ib->type;
+
+      ib = &tmp_ib;
+   }
+   else {
+      /* Otherwise the primitives need adjustment.
+       */
+      tmp_prims = (struct _mesa_prim *)_mesa_malloc(sizeof(*prim) * nr_prims);
+
+      for (i = 0; i < nr_prims; i++) {
+	 /* If this fails, it could indicate an application error:
+	  */
+	 assert(prim[i].start >= min_index);
+
+	 tmp_prims[i] = prim[i];
+	 tmp_prims[i].start -= min_index;
+      }
+
+      prim = tmp_prims;
+   }
+
+   /* Just need to adjust the pointer values on each incoming array.
+    * This works for VBO and non-vbo rendering and shouldn't pesimize
+    * VBO-based upload schemes.  However this may still not be a fast
+    * path for hardware tnl for VBO based rendering as most machines
+    * will be happier if you just specify a starting vertex value in
+    * each primitive.
+    *
+    * For drivers with hardware tnl, you only want to do this if you
+    * are forced to, eg non-VBO indexed rendering with start != 0.
+    */
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      tmp_arrays[i] = *arrays[i];
+      tmp_arrays[i].Ptr += min_index * tmp_arrays[i].StrideB;
+      tmp_array_pointers[i] = &tmp_arrays[i];
+   }
+   
+   /* Re-issue the draw call.
+    */
+   draw( ctx, 
+	 tmp_array_pointers, 
+	 prim, 
+	 nr_prims, 
+	 ib, 
+	 0, 
+	 max_index - min_index );
+   
+   if (tmp_indices)
+      _mesa_free(tmp_indices);
+   
+   if (tmp_prims)
+      _mesa_free(tmp_prims);
+}
+
+
+
diff --git a/src/mesa/vbo/vbo_save.c b/src/mesa/vbo/vbo_save.c
new file mode 100644
index 0000000000..e7f4687963
--- /dev/null
+++ b/src/mesa/vbo/vbo_save.c
@@ -0,0 +1,89 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ *
+ * Copyright (C) 1999-2005  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 "mtypes.h"
+#include "dlist.h"
+#include "vtxfmt.h"
+#include "imports.h"
+
+#include "vbo_context.h"
+
+
+
+static void vbo_save_callback_init( GLcontext *ctx )
+{
+   ctx->Driver.NewList = vbo_save_NewList;
+   ctx->Driver.EndList = vbo_save_EndList;
+   ctx->Driver.SaveFlushVertices = vbo_save_SaveFlushVertices;
+   ctx->Driver.BeginCallList = vbo_save_BeginCallList;
+   ctx->Driver.EndCallList = vbo_save_EndCallList;
+   ctx->Driver.NotifySaveBegin = vbo_save_NotifyBegin;
+}
+
+
+
+void vbo_save_init( GLcontext *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_save_context *save = &vbo->save;
+
+   save->ctx = ctx;
+
+   vbo_save_api_init( save );
+   vbo_save_callback_init(ctx);
+
+   {
+      struct gl_client_array *arrays = save->arrays;
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+   }
+
+   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
+}
+
+
+void vbo_save_destroy( GLcontext *ctx )
+{
+}
+
+
+
+
+/* Note that this can occur during the playback of a display list:
+ */
+void vbo_save_fallback( GLcontext *ctx, GLboolean fallback )
+{
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
+
+   if (fallback)
+      save->replay_flags |= VBO_SAVE_FALLBACK;
+   else
+      save->replay_flags &= ~VBO_SAVE_FALLBACK;
+}
+
+
diff --git a/src/mesa/drivers/dri/i965/brw_save.h b/src/mesa/vbo/vbo_save.h
index 41cabe6508..b81f275a60 100644
--- a/src/mesa/drivers/dri/i965/brw_save.h
+++ b/src/mesa/vbo/vbo_save.h
@@ -31,16 +31,16 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  */
 
-#ifndef BRW_SAVE_H
-#define BRW_SAVE_H
+#ifndef VBO_SAVE_H
+#define VBO_SAVE_H
 
 #include "mtypes.h"
-#include "brw_attrib.h"
-#include "brw_draw.h"
+#include "vbo.h"
+#include "vbo_attrib.h"
 
 
-struct brw_save_copied_vtx {
-   GLfloat buffer[BRW_ATTRIB_MAX * 4 * BRW_MAX_COPIED_VERTS];
+struct vbo_save_copied_vtx {
+   GLfloat buffer[VBO_ATTRIB_MAX * 4 * VBO_MAX_COPIED_VERTS];
    GLuint nr;
 };
 
@@ -60,8 +60,8 @@ struct brw_save_copied_vtx {
  * Eval and other commands that don't fit into these vertex lists are
  * compiled using the fallback opcode mechanism provided by dlist.c.
  */
-struct brw_save_vertex_list {
-   GLubyte attrsz[BRW_ATTRIB_MAX];
+struct vbo_save_vertex_list {
+   GLubyte attrsz[VBO_ATTRIB_MAX];
    GLuint vertex_size;
 
    GLuint buffer_offset;
@@ -70,15 +70,15 @@ struct brw_save_vertex_list {
    GLboolean dangling_attr_ref;	/* current attr implicitly referenced 
 				   outside the list */
 
-   struct brw_draw_prim *prim;
+   struct _mesa_prim *prim;
    GLuint prim_count;
 
-   struct brw_save_vertex_store *vertex_store;
-   struct brw_save_primitive_store *prim_store;
+   struct vbo_save_vertex_store *vertex_store;
+   struct vbo_save_primitive_store *prim_store;
 };
 
 /* These buffers should be a reasonable size to support upload to
- * hardware.  Current brw implementation will re-upload on any
+ * hardware.  Current vbo implementation will re-upload on any
  * changes, so don't make too big or apps which dynamically create
  * dlists and use only a few times will suffer.
  *
@@ -87,36 +87,36 @@ struct brw_save_vertex_list {
  * likelyhood as it occurs.  No reason we couldn't change usage
  * internally even though this probably isn't allowed for client VBOs?
  */
-#define BRW_SAVE_BUFFER_SIZE (8*1024) /* dwords */
-#define BRW_SAVE_PRIM_SIZE   128
-#define BRW_SAVE_PRIM_WEAK 0x40
+#define VBO_SAVE_BUFFER_SIZE (8*1024) /* dwords */
+#define VBO_SAVE_PRIM_SIZE   128
+#define VBO_SAVE_PRIM_WEAK 0x40
 
-#define BRW_SAVE_FALLBACK    0x10000000
+#define VBO_SAVE_FALLBACK    0x10000000
 
 /* Storage to be shared among several vertex_lists.
  */
-struct brw_save_vertex_store {
+struct vbo_save_vertex_store {
    struct gl_buffer_object *bufferobj;
    GLfloat *buffer;
    GLuint used;
    GLuint refcount;
 };
 
-struct brw_save_primitive_store {
-   struct brw_draw_prim buffer[BRW_SAVE_PRIM_SIZE];
+struct vbo_save_primitive_store {
+   struct _mesa_prim buffer[VBO_SAVE_PRIM_SIZE];
    GLuint used;
    GLuint refcount;
 };
 
 
-struct brw_save_context {
+struct vbo_save_context {
    GLcontext *ctx;
    GLvertexformat vtxfmt;
-   struct gl_client_array arrays[BRW_ATTRIB_MAX];
-   const struct gl_client_array *inputs[BRW_ATTRIB_MAX];
+   struct gl_client_array arrays[VBO_ATTRIB_MAX];
+   const struct gl_client_array *inputs[VBO_ATTRIB_MAX];
 
-   GLubyte attrsz[BRW_ATTRIB_MAX];
-   GLubyte active_sz[BRW_ATTRIB_MAX];
+   GLubyte attrsz[VBO_ATTRIB_MAX];
+   GLubyte active_sz[VBO_ATTRIB_MAX];
    GLuint vertex_size;
 
    GLfloat *buffer;
@@ -124,15 +124,15 @@ struct brw_save_context {
    GLuint wrap_count;
    GLuint replay_flags;
 
-   struct brw_draw_prim *prim;
+   struct _mesa_prim *prim;
    GLuint prim_count, prim_max;
 
-   struct brw_save_vertex_store *vertex_store;
-   struct brw_save_primitive_store *prim_store;
+   struct vbo_save_vertex_store *vertex_store;
+   struct vbo_save_primitive_store *prim_store;
 
    GLfloat *vbptr;		   /* cursor, points into buffer */
-   GLfloat vertex[BRW_ATTRIB_MAX*4];	   /* current values */
-   GLfloat *attrptr[BRW_ATTRIB_MAX];
+   GLfloat vertex[VBO_ATTRIB_MAX*4];	   /* current values */
+   GLfloat *attrptr[VBO_ATTRIB_MAX];
    GLuint vert_count;
    GLuint max_vert;
    GLboolean dangling_attr_ref;
@@ -140,32 +140,38 @@ struct brw_save_context {
 
    GLuint opcode_vertex_list;
 
-   struct brw_save_copied_vtx copied;
+   struct vbo_save_copied_vtx copied;
    
-   GLfloat CurrentFloatEdgeFlag;
-
-   GLfloat *current[BRW_ATTRIB_MAX]; /* points into ctx->ListState */
-   GLubyte *currentsz[BRW_ATTRIB_MAX];
+   GLfloat *current[VBO_ATTRIB_MAX]; /* points into ctx->ListState */
+   GLubyte *currentsz[VBO_ATTRIB_MAX];
 };
 
 
-void brw_save_init( GLcontext *ctx );
-void brw_save_destroy( GLcontext *ctx );
-void brw_save_wakeup( GLcontext *ctx );
-void brw_save_invalidate_state( GLcontext *ctx, GLuint new_state );
-void brw_save_fallback( GLcontext *ctx, GLboolean fallback );
+void vbo_save_init( GLcontext *ctx );
+void vbo_save_destroy( GLcontext *ctx );
+void vbo_save_fallback( GLcontext *ctx, GLboolean fallback );
+
+/* save_loopback.c:
+ */
+void vbo_loopback_vertex_list( GLcontext *ctx,
+			       const GLfloat *buffer,
+			       const GLubyte *attrsz,
+			       const struct _mesa_prim *prim,
+			       GLuint prim_count,
+			       GLuint wrap_count,
+			       GLuint vertex_size);
 
 /* Callbacks:
  */
-void brw_save_EndList( GLcontext *ctx );
-void brw_save_NewList( GLcontext *ctx, GLuint list, GLenum mode );
-void brw_save_EndCallList( GLcontext *ctx );
-void brw_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
-void brw_save_SaveFlushVertices( GLcontext *ctx );
-GLboolean brw_save_NotifyBegin( GLcontext *ctx, GLenum mode );
+void vbo_save_EndList( GLcontext *ctx );
+void vbo_save_NewList( GLcontext *ctx, GLuint list, GLenum mode );
+void vbo_save_EndCallList( GLcontext *ctx );
+void vbo_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
+void vbo_save_SaveFlushVertices( GLcontext *ctx );
+GLboolean vbo_save_NotifyBegin( GLcontext *ctx, GLenum mode );
 
-void brw_save_playback_vertex_list( GLcontext *ctx, void *data );
+void vbo_save_playback_vertex_list( GLcontext *ctx, void *data );
 
-void brw_save_api_init( struct brw_save_context *save );
+void vbo_save_api_init( struct vbo_save_context *save );
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_save_api.c b/src/mesa/vbo/vbo_save_api.c
index 06ed1d23aa..c08fd1fe55 100644
--- a/src/mesa/drivers/dri/i965/brw_save_api.c
+++ b/src/mesa/vbo/vbo_save_api.c
@@ -77,9 +77,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "vtxfmt.h"
 #include "dispatch.h"
 
-#include "brw_save.h"
-#include "brw_fallback.h"
-
+#include "vbo_context.h"
 
 
 
@@ -88,11 +86,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * wrong-footed on replay.
  */
 static GLuint _save_copy_vertices( GLcontext *ctx, 
-				   const struct brw_save_vertex_list *node,
+				   const struct vbo_save_vertex_list *node,
 				   const GLfloat *src_buffer)
 {
-   struct brw_save_context *save = IMM_CONTEXT( ctx )->save;
-   const struct brw_draw_prim *prim = &node->prim[node->prim_count-1];
+   struct vbo_save_context *save = &vbo_context( ctx )->save;
+   const struct _mesa_prim *prim = &node->prim[node->prim_count-1];
    GLuint nr = prim->count;
    GLuint sz = save->vertex_size;
    const GLfloat *src = src_buffer + prim->start * sz;
@@ -158,9 +156,9 @@ static GLuint _save_copy_vertices( GLcontext *ctx,
 }
 
 
-static struct brw_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
+static struct vbo_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
 {
-   struct brw_save_vertex_store *vertex_store = CALLOC_STRUCT(brw_save_vertex_store);
+   struct vbo_save_vertex_store *vertex_store = CALLOC_STRUCT(vbo_save_vertex_store);
 
    /* obj->Name needs to be non-zero, but won't ever be examined more
     * closely than that.  In particular these buffers won't be entered
@@ -172,7 +170,7 @@ static struct brw_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
 
    ctx->Driver.BufferData( ctx, 
 			   GL_ARRAY_BUFFER_ARB, 
-			   BRW_SAVE_BUFFER_SIZE * sizeof(GLfloat),
+			   VBO_SAVE_BUFFER_SIZE * sizeof(GLfloat),
 			   NULL,
 			   GL_STATIC_DRAW_ARB,
 			   vertex_store->bufferobj);
@@ -184,7 +182,7 @@ static struct brw_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
    return vertex_store;
 }
 
-static void free_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *vertex_store )
+static void free_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *vertex_store )
 {
    assert(!vertex_store->buffer);
 
@@ -194,29 +192,29 @@ static void free_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *ver
    FREE( vertex_store );
 }
 
-static GLfloat *map_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *vertex_store )
+static GLfloat *map_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *vertex_store )
 {
    assert(vertex_store->bufferobj);
    assert(!vertex_store->buffer);
    vertex_store->buffer = (GLfloat *)ctx->Driver.MapBuffer(ctx, 
 							   GL_ARRAY_BUFFER_ARB,	/* not used */
-							   GL_STATIC_DRAW_ARB, /* not used */
+							   GL_WRITE_ONLY, /* not used */
 							   vertex_store->bufferobj); 
 
    assert(vertex_store->buffer);
    return vertex_store->buffer + vertex_store->used;
 }
 
-static void unmap_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *vertex_store )
+static void unmap_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *vertex_store )
 {
    ctx->Driver.UnmapBuffer( ctx, GL_ARRAY_BUFFER_ARB, vertex_store->bufferobj );
    vertex_store->buffer = NULL;
 }
 
 
-static struct brw_save_primitive_store *alloc_prim_store( GLcontext *ctx )
+static struct vbo_save_primitive_store *alloc_prim_store( GLcontext *ctx )
 {
-   struct brw_save_primitive_store *store = CALLOC_STRUCT(brw_save_primitive_store);
+   struct vbo_save_primitive_store *store = CALLOC_STRUCT(vbo_save_primitive_store);
    (void) ctx;
    store->used = 0;
    store->refcount = 1;
@@ -225,7 +223,7 @@ static struct brw_save_primitive_store *alloc_prim_store( GLcontext *ctx )
 
 static void _save_reset_counters( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    save->prim = save->prim_store->buffer + save->prim_store->used;
    save->buffer = (save->vertex_store->buffer + 
@@ -234,14 +232,14 @@ static void _save_reset_counters( GLcontext *ctx )
    assert(save->buffer == save->vbptr);
 
    if (save->vertex_size)
-      save->max_vert = ((BRW_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
+      save->max_vert = ((VBO_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
 			 save->vertex_size);
    else
       save->max_vert = 0;
 
    save->vert_count = 0;
    save->prim_count = 0;
-   save->prim_max = BRW_SAVE_PRIM_SIZE - save->prim_store->used;
+   save->prim_max = VBO_SAVE_PRIM_SIZE - save->prim_store->used;
    save->dangling_attr_ref = 0;
 }
 
@@ -251,13 +249,13 @@ static void _save_reset_counters( GLcontext *ctx )
  */
 static void _save_compile_vertex_list( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
-   struct brw_save_vertex_list *node;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
+   struct vbo_save_vertex_list *node;
 
    /* Allocate space for this structure in the display list currently
     * being compiled.
     */
-   node = (struct brw_save_vertex_list *)
+   node = (struct vbo_save_vertex_list *)
       _mesa_alloc_instruction(ctx, save->opcode_vertex_list, sizeof(*node));
 
    if (!node)
@@ -279,7 +277,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
    node->vertex_store->refcount++;
    node->prim_store->refcount++;
 
-   assert(node->attrsz[BRW_ATTRIB_POS] != 0 ||
+   assert(node->attrsz[VBO_ATTRIB_POS] != 0 ||
 	  node->count == 0);
 
    if (save->dangling_attr_ref)
@@ -301,7 +299,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
 
       _glapi_set_dispatch(ctx->Exec);
 
-      brw_loopback_vertex_list( ctx,
+      vbo_loopback_vertex_list( ctx,
 				(const GLfloat *)((const char *)save->vertex_store->buffer + 
 						  node->buffer_offset),
 				node->attrsz,
@@ -318,7 +316,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
     * the next vertex lists as well.
     */
    if (save->vertex_store->used > 
-       BRW_SAVE_BUFFER_SIZE - 16 * (save->vertex_size + 4)) {
+       VBO_SAVE_BUFFER_SIZE - 16 * (save->vertex_size + 4)) {
 
       /* Unmap old store:
        */
@@ -336,7 +334,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
       save->vbptr = map_vertex_store( ctx, save->vertex_store );
    } 
 
-   if (save->prim_store->used > BRW_SAVE_PRIM_SIZE - 6) {
+   if (save->prim_store->used > VBO_SAVE_PRIM_SIZE - 6) {
       save->prim_store->refcount--; 
       assert(save->prim_store->refcount != 0);
       save->prim_store = alloc_prim_store( ctx );
@@ -353,7 +351,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
  */
 static void _save_wrap_buffers( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLint i = save->prim_count - 1;
    GLenum mode;
    GLboolean weak;
@@ -391,7 +389,7 @@ static void _save_wrap_buffers( GLcontext *ctx )
  */
 static void _save_wrap_filled_vertex( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLfloat *data = save->copied.buffer;
    GLuint i;
 
@@ -414,38 +412,26 @@ static void _save_wrap_filled_vertex( GLcontext *ctx )
 
 static void _save_copy_to_current( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
    GLuint i;
 
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (save->attrsz[i]) {
 	 save->currentsz[i][0] = save->attrsz[i];
 	 COPY_CLEAN_4V(save->current[i], 
-		    save->attrsz[i], 
-		    save->attrptr[i]);
+		       save->attrsz[i], 
+		       save->attrptr[i]);
       }
    }
-
-   /* Edgeflag requires special treatment: 
-    *
-    * TODO: change edgeflag to GLfloat in Mesa.
-    */
-   if (save->attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      ctx->ListState.ActiveEdgeFlag = 1;
-      save->CurrentFloatEdgeFlag = 
-	 save->attrptr[BRW_ATTRIB_EDGEFLAG][0];
-      ctx->ListState.CurrentEdgeFlag = 
-	 (save->CurrentFloatEdgeFlag == 1.0);
-   }
 }
 
 
 static void _save_copy_from_current( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
    GLint i;
 
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) 
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       switch (save->attrsz[i]) {
       case 4: save->attrptr[i][3] = save->current[i][3];
       case 3: save->attrptr[i][2] = save->current[i][2];
@@ -453,12 +439,6 @@ static void _save_copy_from_current( GLcontext *ctx )
       case 1: save->attrptr[i][0] = save->current[i][0];
       case 0: break;
       }
-
-   /* Edgeflag requires special treatment:
-    */
-   if (save->attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      save->CurrentFloatEdgeFlag = (GLfloat)ctx->ListState.CurrentEdgeFlag;
-      save->attrptr[BRW_ATTRIB_EDGEFLAG][0] = save->CurrentFloatEdgeFlag;
    }
 }
 
@@ -471,7 +451,7 @@ static void _save_upgrade_vertex( GLcontext *ctx,
 				 GLuint attr,
 				 GLuint newsz )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLuint oldsz;
    GLuint i;
    GLfloat *tmp;
@@ -496,13 +476,13 @@ static void _save_upgrade_vertex( GLcontext *ctx,
    save->attrsz[attr] = newsz;
 
    save->vertex_size += newsz - oldsz;
-   save->max_vert = ((BRW_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
+   save->max_vert = ((VBO_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
 		      save->vertex_size);
    save->vert_count = 0;
 
    /* Recalculate all the attrptr[] values:
     */
-   for (i = 0, tmp = save->vertex ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0, tmp = save->vertex ; i < VBO_ATTRIB_MAX ; i++) {
       if (save->attrsz[i]) {
 	 save->attrptr[i] = tmp;
 	 tmp += save->attrsz[i];
@@ -529,13 +509,13 @@ static void _save_upgrade_vertex( GLcontext *ctx,
 
       /* Need to note this and fix up at runtime (or loopback):
        */
-      if (attr != BRW_ATTRIB_POS && save->currentsz[attr][0] == 0) {
+      if (attr != VBO_ATTRIB_POS && save->currentsz[attr][0] == 0) {
 	 assert(oldsz == 0);
 	 save->dangling_attr_ref = GL_TRUE;
       }
 
       for (i = 0 ; i < save->copied.nr ; i++) {
-	 for (j = 0 ; j < BRW_ATTRIB_MAX ; j++) {
+	 for (j = 0 ; j < VBO_ATTRIB_MAX ; j++) {
 	    if (save->attrsz[j]) {
 	       if (j == attr) {
 		  if (oldsz) {
@@ -565,7 +545,7 @@ static void _save_upgrade_vertex( GLcontext *ctx,
 
 static void save_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
 
    if (sz > save->attrsz[attr]) {
       /* New size is larger.  Need to flush existing vertices and get
@@ -589,10 +569,10 @@ static void save_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
 
 static void _save_reset_vertex( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLuint i;
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
       save->attrsz[i] = 0;
       save->active_sz[i] = 0;
    }
@@ -613,7 +593,7 @@ static void _save_reset_vertex( GLcontext *ctx )
  */
 #define ATTR( A, N, V0, V1, V2, V3 )				\
 do {								\
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;	\
+   struct vbo_save_context *save = &vbo_context(ctx)->save;	\
 								\
    if (save->active_sz[A] != N)				\
       save_fixup_vertex(ctx, A, N);				\
@@ -641,7 +621,7 @@ do {								\
 
 #define TAG(x) _save_##x
 
-#include "brw_attrib_tmp.h"
+#include "vbo_attrib_tmp.h"
 
 
 
@@ -652,7 +632,7 @@ do {								\
  */
 #define DO_FALLBACK(ctx) 							\
 do {									\
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;					\
+   struct vbo_save_context *save = &vbo_context(ctx)->save;					\
 									\
    if (save->vert_count || save->prim_count) 						\
       _save_compile_vertex_list( ctx );					\
@@ -726,17 +706,17 @@ static void GLAPIENTRY _save_CallLists( GLsizei n, GLenum type, const GLvoid *v
 /* This begin is hooked into ...  Updating of
  * ctx->Driver.CurrentSavePrimitive is already taken care of.
  */
-GLboolean brw_save_NotifyBegin( GLcontext *ctx, GLenum mode )
+GLboolean vbo_save_NotifyBegin( GLcontext *ctx, GLenum mode )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
 
    GLuint i = save->prim_count++;
 
    assert(i < save->prim_max);
-   save->prim[i].mode = mode & ~BRW_SAVE_PRIM_WEAK;
+   save->prim[i].mode = mode & ~VBO_SAVE_PRIM_WEAK;
    save->prim[i].begin = 1;
    save->prim[i].end = 0;
-   save->prim[i].weak = (mode & BRW_SAVE_PRIM_WEAK) ? 1 : 0;
+   save->prim[i].weak = (mode & VBO_SAVE_PRIM_WEAK) ? 1 : 0;
    save->prim[i].pad = 0;
    save->prim[i].start = save->vert_count;
    save->prim[i].count = 0;   
@@ -751,7 +731,7 @@ GLboolean brw_save_NotifyBegin( GLcontext *ctx, GLenum mode )
 static void GLAPIENTRY _save_End( void )
 {
    GET_CURRENT_CONTEXT( ctx ); 
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
    GLint i = save->prim_count - 1;
 
    ctx->Driver.CurrentSavePrimitive = PRIM_OUTSIDE_BEGIN_END;
@@ -838,7 +818,7 @@ static void GLAPIENTRY _save_Begin( GLenum mode )
 static void GLAPIENTRY _save_OBE_Rectf( GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2 )
 {
    GET_CURRENT_CONTEXT(ctx);
-   brw_save_NotifyBegin( ctx, GL_QUADS | BRW_SAVE_PRIM_WEAK );
+   vbo_save_NotifyBegin( ctx, GL_QUADS | VBO_SAVE_PRIM_WEAK );
    CALL_Vertex2f(GET_DISPATCH(), ( x1, y1 ));
    CALL_Vertex2f(GET_DISPATCH(), ( x2, y1 ));
    CALL_Vertex2f(GET_DISPATCH(), ( x2, y2 ));
@@ -857,7 +837,8 @@ static void GLAPIENTRY _save_OBE_DrawArrays(GLenum mode, GLint start, GLsizei co
 
    _ae_map_vbos( ctx );
 
-   brw_save_NotifyBegin( ctx, mode | BRW_SAVE_PRIM_WEAK );
+   vbo_save_NotifyBegin( ctx, mode | VBO_SAVE_PRIM_WEAK );
+
    for (i = 0; i < count; i++)
        CALL_ArrayElement(GET_DISPATCH(), (start + i));
    CALL_End(GET_DISPATCH(), ());
@@ -879,7 +860,7 @@ static void GLAPIENTRY _save_OBE_DrawElements(GLenum mode, GLsizei count, GLenum
 
    _ae_map_vbos( ctx );
 
-   brw_save_NotifyBegin( ctx, mode | BRW_SAVE_PRIM_WEAK );
+   vbo_save_NotifyBegin( ctx, mode | VBO_SAVE_PRIM_WEAK );
 
    switch (type) {
    case GL_UNSIGNED_BYTE:
@@ -922,7 +903,7 @@ static void GLAPIENTRY _save_OBE_DrawRangeElements(GLenum mode,
 
 static void _save_vtxfmt_init( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLvertexformat *vfmt = &save->vtxfmt;
 
    vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
@@ -1007,9 +988,9 @@ static void _save_vtxfmt_init( GLcontext *ctx )
 }
 
 
-void brw_save_SaveFlushVertices( GLcontext *ctx )
+void vbo_save_SaveFlushVertices( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    /* Noop when we are actually active:
     */
@@ -1027,9 +1008,9 @@ void brw_save_SaveFlushVertices( GLcontext *ctx )
    ctx->Driver.SaveNeedFlush = 0;
 }
 
-void brw_save_NewList( GLcontext *ctx, GLuint list, GLenum mode )
+void vbo_save_NewList( GLcontext *ctx, GLuint list, GLenum mode )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    (void) list; (void) mode;
 
@@ -1046,36 +1027,36 @@ void brw_save_NewList( GLcontext *ctx, GLuint list, GLenum mode )
    ctx->Driver.SaveNeedFlush = 0;
 }
 
-void brw_save_EndList( GLcontext *ctx )
+void vbo_save_EndList( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    unmap_vertex_store( ctx, save->vertex_store );
 
    assert(save->vertex_size == 0);
 }
  
-void brw_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *dlist )
+void vbo_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *dlist )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    save->replay_flags |= dlist->flags;
 }
 
-void brw_save_EndCallList( GLcontext *ctx )
+void vbo_save_EndCallList( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    
    if (ctx->ListState.CallDepth == 1) {
-      /* This is correct: want to keep only the BRW_SAVE_FALLBACK
+      /* This is correct: want to keep only the VBO_SAVE_FALLBACK
        * flag, if it is set:
        */
-      save->replay_flags &= BRW_SAVE_FALLBACK;
+      save->replay_flags &= VBO_SAVE_FALLBACK;
    }
 }
 
 
-static void brw_destroy_vertex_list( GLcontext *ctx, void *data )
+static void vbo_destroy_vertex_list( GLcontext *ctx, void *data )
 {
-   struct brw_save_vertex_list *node = (struct brw_save_vertex_list *)data;
+   struct vbo_save_vertex_list *node = (struct vbo_save_vertex_list *)data;
    (void) ctx;
 
    if ( --node->vertex_store->refcount == 0 ) 
@@ -1086,19 +1067,19 @@ static void brw_destroy_vertex_list( GLcontext *ctx, void *data )
 }
 
 
-static void brw_print_vertex_list( GLcontext *ctx, void *data )
+static void vbo_print_vertex_list( GLcontext *ctx, void *data )
 {
-   struct brw_save_vertex_list *node = (struct brw_save_vertex_list *)data;
+   struct vbo_save_vertex_list *node = (struct vbo_save_vertex_list *)data;
    GLuint i;
    (void) ctx;
 
-   _mesa_debug(NULL, "BRW-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
+   _mesa_debug(NULL, "VBO-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
                node->count,
 	       node->prim_count,
 	       node->vertex_size);
 
    for (i = 0 ; i < node->prim_count ; i++) {
-      struct brw_draw_prim *prim = &node->prim[i];
+      struct _mesa_prim *prim = &node->prim[i];
       _mesa_debug(NULL, "   prim %d: %s%s %d..%d %s %s\n",
 		  i, 
 		  _mesa_lookup_enum_by_nr(prim->mode),
@@ -1113,49 +1094,45 @@ static void brw_print_vertex_list( GLcontext *ctx, void *data )
 
 static void _save_current_init( GLcontext *ctx ) 
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLint i;
 
-   for (i = 0; i < BRW_ATTRIB_FIRST_MATERIAL; i++) {
-      save->currentsz[i] = &ctx->ListState.ActiveAttribSize[i];
-      save->current[i] = ctx->ListState.CurrentAttrib[i];
+   for (i = VBO_ATTRIB_POS; i <= VBO_ATTRIB_GENERIC15; i++) {
+      const GLuint j = i - VBO_ATTRIB_POS;
+      ASSERT(j < VERT_ATTRIB_MAX);
+      save->currentsz[i] = &ctx->ListState.ActiveAttribSize[j];
+      save->current[i] = ctx->ListState.CurrentAttrib[j];
    }
 
-   for (i = BRW_ATTRIB_FIRST_MATERIAL; i <= BRW_ATTRIB_LAST_MATERIAL; i++) {
-      const GLuint j = i - BRW_ATTRIB_FIRST_MATERIAL;
+   for (i = VBO_ATTRIB_FIRST_MATERIAL; i <= VBO_ATTRIB_LAST_MATERIAL; i++) {
+      const GLuint j = i - VBO_ATTRIB_FIRST_MATERIAL;
       ASSERT(j < MAT_ATTRIB_MAX);
       save->currentsz[i] = &ctx->ListState.ActiveMaterialSize[j];
       save->current[i] = ctx->ListState.CurrentMaterial[j];
    }
-
-   save->currentsz[BRW_ATTRIB_INDEX] = &ctx->ListState.ActiveIndex;
-   save->current[BRW_ATTRIB_INDEX] = &ctx->ListState.CurrentIndex;
-
-   save->currentsz[BRW_ATTRIB_EDGEFLAG] = &ctx->ListState.ActiveEdgeFlag;
-   save->current[BRW_ATTRIB_EDGEFLAG] = &save->CurrentFloatEdgeFlag;
 }
 
 /**
  * Initialize the display list compiler
  */
-void brw_save_api_init( struct brw_save_context *save )
+void vbo_save_api_init( struct vbo_save_context *save )
 {
    GLcontext *ctx = save->ctx;
    GLuint i;
 
    save->opcode_vertex_list =
       _mesa_alloc_opcode( ctx,
-			  sizeof(struct brw_save_vertex_list),
-			  brw_save_playback_vertex_list,
-			  brw_destroy_vertex_list,
-			  brw_print_vertex_list );
+			  sizeof(struct vbo_save_vertex_list),
+			  vbo_save_playback_vertex_list,
+			  vbo_destroy_vertex_list,
+			  vbo_print_vertex_list );
 
-   ctx->Driver.NotifySaveBegin = brw_save_NotifyBegin;
+   ctx->Driver.NotifySaveBegin = vbo_save_NotifyBegin;
 
    _save_vtxfmt_init( ctx );
    _save_current_init( ctx );
 
-   for (i = 0; i < BRW_ATTRIB_MAX; i++)
+   for (i = 0; i < VBO_ATTRIB_MAX; i++)
       save->inputs[i] = &save->arrays[i];
 
    /* Hook our array functions into the outside-begin-end vtxfmt in 
diff --git a/src/mesa/drivers/dri/i965/brw_save_draw.c b/src/mesa/vbo/vbo_save_draw.c
index cebdd8d060..a50d1cd904 100644
--- a/src/mesa/drivers/dri/i965/brw_save_draw.c
+++ b/src/mesa/vbo/vbo_save_draw.c
@@ -34,19 +34,18 @@
 #include "light.h"
 #include "state.h"
 
-#include "brw_save.h"
-#include "brw_draw.h"
-#include "brw_fallback.h"
+#include "vbo_context.h"
+
 
 /*
  * After playback, copy everything but the position from the
  * last vertex to the saved state
  */
 static void _playback_copy_to_current( GLcontext *ctx,
-				       const struct brw_save_vertex_list *node )
+				       const struct vbo_save_vertex_list *node )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
-   GLfloat vertex[BRW_ATTRIB_MAX * 4], *data = vertex;
+   struct vbo_context *vbo = vbo_context(ctx);
+   GLfloat vertex[VBO_ATTRIB_MAX * 4], *data = vertex;
    GLuint i, offset;
 
    if (node->count)
@@ -59,43 +58,34 @@ static void _playback_copy_to_current( GLcontext *ctx,
 				 node->vertex_size * sizeof(GLfloat), 
 				 data, node->vertex_store->bufferobj );
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (node->attrsz[i]) {
-	 if (i != BRW_ATTRIB_POS)
-	    COPY_CLEAN_4V(save->current[i], node->attrsz[i], data);
+	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
 
-	 if (i >= BRW_ATTRIB_MAT_FRONT_AMBIENT &&
-	     i <= BRW_ATTRIB_MAT_BACK_INDEXES)
-	    ctx->NewState |= _NEW_LIGHT;
+	 COPY_CLEAN_4V(current, 
+		       node->attrsz[i], 
+		       data);
 
-	 /* Edgeflag requires special treatment:
-	  */
-	 if (i == BRW_ATTRIB_EDGEFLAG)
-	    ctx->Current.EdgeFlag = (data[0] == 1.0);
+	 vbo->currval[i].Size = node->attrsz[i];
 
-	 data += node->attrsz[i] * sizeof(GLfloat);
-      }
-   }
+	 data += node->attrsz[i];
 
-   /* Edgeflag requires special treatment:
-    */
-   if (node->attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (data[0] == 1.0);
+	 if (i >= VBO_ATTRIB_FIRST_MATERIAL &&
+	     i <= VBO_ATTRIB_LAST_MATERIAL)
+	    ctx->NewState |= _NEW_LIGHT;
+      }
    }
 
-
-#if 1
    /* Colormaterial -- this kindof sucks.
     */
    if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, ctx->Current.Attrib[BRW_ATTRIB_COLOR0]);
+      _mesa_update_color_material(ctx, ctx->Current.Attrib[VBO_ATTRIB_COLOR0]);
    }
-#endif
 
    /* CurrentExecPrimitive
     */
    if (node->prim_count) {
-      const struct brw_draw_prim *prim = &node->prim[node->prim_count - 1];
+      const struct _mesa_prim *prim = &node->prim[node->prim_count - 1];
       if (prim->end)
 	 ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
       else
@@ -108,16 +98,38 @@ static void _playback_copy_to_current( GLcontext *ctx,
 /* Treat the vertex storage as a VBO, define vertex arrays pointing
  * into it:
  */
-static void brw_bind_vertex_list( struct brw_save_context *save,
-                                   const struct brw_save_vertex_list *node )
+static void vbo_bind_vertex_list( GLcontext *ctx,
+                                   const struct vbo_save_vertex_list *node )
 {
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_save_context *save = &vbo->save;
    struct gl_client_array *arrays = save->arrays;
    GLuint data = node->buffer_offset;
+   const GLuint *map;
    GLuint attr;
 
-   memset(arrays, 0, BRW_ATTRIB_MAX * sizeof(arrays[0]));
+   /* Install the default (ie Current) attributes first, then overlay
+    * all active ones.
+    */
+   switch (get_program_mode(ctx)) {
+   case VP_NONE:
+      memcpy(arrays,      vbo->legacy_currval, 16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->mat_currval,    MAT_ATTRIB_MAX * sizeof(arrays[0]));
+      map = vbo->map_vp_none;
+      break;
+   case VP_NV:
+   case VP_ARB:
+      /* The aliasing of attributes for NV vertex programs has already
+       * occurred.  NV vertex programs cannot access material values,
+       * nor attributes greater than VERT_ATTRIB_TEX7.  
+       */
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+      map = vbo->map_vp_arb;
+      break;
+   }
 
-   for (attr = 0; attr < BRW_ATTRIB_MAX; attr++) {
+   for (attr = 0; attr < VBO_ATTRIB_MAX; attr++) {
       if (node->attrsz[attr]) {
 	 arrays[attr].Ptr = (const GLubyte *)data;
 	 arrays[attr].Size = node->attrsz[attr];
@@ -135,15 +147,15 @@ static void brw_bind_vertex_list( struct brw_save_context *save,
    }
 }
 
-static void brw_save_loopback_vertex_list( GLcontext *ctx,
-					   const struct brw_save_vertex_list *list )
+static void vbo_save_loopback_vertex_list( GLcontext *ctx,
+					   const struct vbo_save_vertex_list *list )
 {
    const char *buffer = ctx->Driver.MapBuffer(ctx, 
 					      GL_ARRAY_BUFFER_ARB, 
-					      GL_DYNAMIC_READ_ARB, /* ? */
+					      GL_READ_ONLY, /* ? */
 					       list->vertex_store->bufferobj);
 
-   brw_loopback_vertex_list( ctx,
+   vbo_loopback_vertex_list( ctx,
 			     (const GLfloat *)(buffer + list->buffer_offset),
 			     list->attrsz,
 			     list->prim,
@@ -159,10 +171,10 @@ static void brw_save_loopback_vertex_list( GLcontext *ctx,
 /**
  * Execute the buffer and save copied verts.
  */
-void brw_save_playback_vertex_list( GLcontext *ctx, void *data )
+void vbo_save_playback_vertex_list( GLcontext *ctx, void *data )
 {
-   const struct brw_save_vertex_list *node = (const struct brw_save_vertex_list *) data;
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   const struct vbo_save_vertex_list *node = (const struct vbo_save_vertex_list *) data;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    FLUSH_CURRENT(ctx, 0);
 
@@ -177,14 +189,14 @@ void brw_save_playback_vertex_list( GLcontext *ctx, void *data )
 	 if (0)
 	    _mesa_printf("displaylist recursive begin");
 
-	 brw_save_loopback_vertex_list( ctx, node );
+	 vbo_save_loopback_vertex_list( ctx, node );
 	 return;
       }
       else if (save->replay_flags) {
 	 /* Various degnerate cases: translate into immediate mode
 	  * calls rather than trying to execute in place.
 	  */
-	 brw_save_loopback_vertex_list( ctx, node );
+	 vbo_save_loopback_vertex_list( ctx, node );
 	 return;
       }
       
@@ -198,21 +210,15 @@ void brw_save_playback_vertex_list( GLcontext *ctx, void *data )
          return;
       }
 
-      brw_bind_vertex_list( save, node );
-
-      if (!brw_draw_prims( save->ctx, 
-			   save->inputs, 
-			   node->prim, 
-			   node->prim_count,
-			   NULL,
-			   0,	/* Node is a VBO, so this is ok */
-			   node->count,
-			   0 )) {
-	 brw_fallback(ctx);
-	 brw_save_loopback_vertex_list( ctx, node );
-	 brw_unfallback(ctx);
-	 return;
-      }
+      vbo_bind_vertex_list( ctx, node );
+
+      vbo_context(ctx)->draw_prims( ctx, 
+				    save->inputs, 
+				    node->prim, 
+				    node->prim_count,
+				    NULL,
+				    0,	/* Node is a VBO, so this is ok */
+				    node->count );
    }
 
    /* Copy to current?
diff --git a/src/mesa/vbo/vbo_save_loopback.c b/src/mesa/vbo/vbo_save_loopback.c
new file mode 100644
index 0000000000..430333b84d
--- /dev/null
+++ b/src/mesa/vbo/vbo_save_loopback.c
@@ -0,0 +1,194 @@
+/**************************************************************************
+ * 
+ * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "swrast_setup/swrast_setup.h"
+#include "swrast/swrast.h"
+#include "tnl/tnl.h"
+#include "context.h"
+
+#include "vbo_context.h"
+
+#include "glheader.h"
+#include "enums.h"
+#include "glapi.h"
+#include "imports.h"
+#include "macros.h"
+#include "mtypes.h"
+#include "dispatch.h"
+
+
+typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
+
+
+/* This file makes heavy use of the aliasing of NV vertex attributes
+ * with the legacy attributes, and also with ARB and Material
+ * attributes as currently implemented.
+ */
+static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
+}
+
+static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
+}
+
+static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
+}
+
+static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
+}
+
+static attr_func vert_attrfunc[4] = {
+   VertexAttrib1fvNV,
+   VertexAttrib2fvNV,
+   VertexAttrib3fvNV,
+   VertexAttrib4fvNV
+};
+
+struct loopback_attr {
+   GLint target;
+   GLint sz;
+   attr_func func;
+};
+
+/* Don't emit ends and begins on wrapped primitives.  Don't replay
+ * wrapped vertices.  If we get here, it's probably because the the
+ * precalculated wrapping is wrong.
+ */
+static void loopback_prim( GLcontext *ctx,
+			   const GLfloat *buffer,
+			   const struct _mesa_prim *prim,
+			   GLuint wrap_count,
+			   GLuint vertex_size,
+			   const struct loopback_attr *la, GLuint nr )
+{
+   GLint start = prim->start;
+   GLint end = start + prim->count;
+   const GLfloat *data;
+   GLint j;
+   GLuint k;
+
+   if (0)
+      _mesa_printf("loopback prim %s(%s,%s) verts %d..%d\n",
+		   _mesa_lookup_enum_by_nr(prim->mode),
+		   prim->begin ? "begin" : "..",
+		   prim->end ? "end" : "..",
+		   start, 
+		   end);
+
+   if (prim->begin) {
+      CALL_Begin(GET_DISPATCH(), ( prim->mode ));
+   }
+   else {
+      assert(start == 0);
+      start += wrap_count;
+   }
+
+   data = buffer + start * vertex_size;
+
+   for (j = start ; j < end ; j++) {
+      const GLfloat *tmp = data + la[0].sz;
+
+      for (k = 1 ; k < nr ; k++) {
+	 la[k].func( ctx, la[k].target, tmp );
+	 tmp += la[k].sz;
+      }
+	 
+      /* Fire the vertex
+       */
+      la[0].func( ctx, VBO_ATTRIB_POS, data );
+      data = tmp;
+   }
+
+   if (prim->end) {
+      CALL_End(GET_DISPATCH(), ());
+   }
+}
+
+/* Primitives generated by DrawArrays/DrawElements/Rectf may be
+ * caught here.  If there is no primitive in progress, execute them
+ * normally, otherwise need to track and discard the generated
+ * primitives.
+ */
+static void loopback_weak_prim( GLcontext *ctx,
+				const struct _mesa_prim *prim )
+{
+   /* Use the prim_weak flag to ensure that if this primitive
+    * wraps, we don't mistake future vertex_lists for part of the
+    * surrounding primitive.
+    *
+    * While this flag is set, we are simply disposing of data
+    * generated by an operation now known to be a noop.
+    */
+   if (prim->begin)
+      ctx->Driver.CurrentExecPrimitive |= VBO_SAVE_PRIM_WEAK;
+   if (prim->end)
+      ctx->Driver.CurrentExecPrimitive &= ~VBO_SAVE_PRIM_WEAK;
+}
+
+
+void vbo_loopback_vertex_list( GLcontext *ctx,
+			       const GLfloat *buffer,
+			       const GLubyte *attrsz,
+			       const struct _mesa_prim *prim,
+			       GLuint prim_count,
+			       GLuint wrap_count,
+			       GLuint vertex_size)
+{
+   struct loopback_attr la[VBO_ATTRIB_MAX];
+   GLuint i, nr = 0;
+
+   /* All Legacy, NV, ARB and Material attributes are routed through
+    * the NV attributes entrypoints:
+    */
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
+      if (attrsz[i]) {
+	 la[nr].target = i;
+	 la[nr].sz = attrsz[i];
+	 la[nr].func = vert_attrfunc[attrsz[i]-1];
+	 nr++;
+      }
+   }
+
+   for (i = 0 ; i < prim_count ; i++) {
+      if ((prim[i].mode & VBO_SAVE_PRIM_WEAK) &&
+	  (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END))
+      {
+	 loopback_weak_prim( ctx, &prim[i] );
+      }
+      else
+      {
+	 loopback_prim( ctx, buffer, &prim[i], wrap_count, vertex_size, la, nr );
+      }
+   }
+}
diff --git a/src/mesa/vbo/vbo_split.c b/src/mesa/vbo/vbo_split.c
new file mode 100644
index 0000000000..ef205a3bb1
--- /dev/null
+++ b/src/mesa/vbo/vbo_split.c
@@ -0,0 +1,161 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Deal with hardware and/or swtnl maximums:
+ * - maximum number of vertices in buffer
+ * - maximum number of elements (maybe zero)
+ *
+ * The maximums may vary with opengl state (eg if a larger hardware
+ * vertex is required in this state, the maximum number of vertices
+ * may be smaller than in another state).
+ *
+ * We want buffer splitting to be a convenience function for the code
+ * actually drawing the primitives rather than a system-wide maximum,
+ * otherwise it is hard to avoid pessimism.  
+ *
+ * For instance, if a driver has no hardware limits on vertex buffer
+ * dimensions, it would not ordinarily want to split vbos.  But if
+ * there is an unexpected fallback, eg memory manager fails to upload
+ * textures, it will want to pass the drawing commands onto swtnl,
+ * which does have limitations.  A convenience function allows swtnl
+ * to split the drawing and vbos internally without imposing its
+ * limitations on drivers which want to use it as a fallback path.
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mtypes.h"
+
+#include "vbo_split.h"
+#include "vbo.h"
+
+/* True if a primitive can be split without copying of vertices, false
+ * otherwise.
+ */
+GLboolean split_prim_inplace(GLenum mode, GLuint *first, GLuint *incr)
+{
+   switch (mode) {
+   case GL_POINTS:
+      *first = 1;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_LINES:
+      *first = 2;
+      *incr = 2;
+      return GL_TRUE;
+   case GL_LINE_STRIP:
+      *first = 2;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_TRIANGLES:
+      *first = 3;
+      *incr = 3;
+      return GL_TRUE;
+   case GL_TRIANGLE_STRIP:
+      *first = 3;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_QUADS:
+      *first = 4;
+      *incr = 4;
+      return GL_TRUE;
+   case GL_QUAD_STRIP:
+      *first = 4;
+      *incr = 2;
+      return GL_TRUE;
+   default:
+      *first = 0;
+      *incr = 1;		/* so that count % incr works */
+      return GL_FALSE;
+   }
+}
+
+
+
+void vbo_split_prims( GLcontext *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index,
+		      vbo_draw_func draw,
+		      const struct split_limits *limits )
+{
+  
+   if (ib) {
+      if (limits->max_indices == 0) {
+	 /* Could traverse the indices, re-emitting vertices in turn.
+	  * But it's hard to see why this case would be needed - for
+	  * software tnl, it is better to convert to non-indexed
+	  * rendering after transformation is complete, as is done in
+	  * the t_dd_rendertmp.h templates.  Are there any devices
+	  * with hardware tnl that cannot do indexed rendering?
+	  *
+	  * For now, this path is disabled.
+	  */
+	 assert(0);
+      }
+      else if (max_index - min_index >= limits->max_verts) {
+	 /* The vertex buffers are too large for hardware (or the
+	  * swtnl module).  Traverse the indices, re-emitting vertices
+	  * in turn.  Use a vertex cache to preserve some of the
+	  * sharing from the original index list.
+	  */
+	 vbo_split_copy(ctx, arrays, prim, nr_prims, ib,
+			draw, limits );
+      }
+      else if (ib->count > limits->max_indices) {
+	 /* The index buffer is too large for hardware.  Try to split
+	  * on whole-primitive boundaries, otherwise try to split the
+	  * individual primitives.
+	  */
+	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
+			   min_index, max_index, draw, limits );
+      }
+      else {
+	 /* Why were we called? */
+	 assert(0);
+      }
+   }
+   else {
+      if (max_index - min_index >= limits->max_verts) {
+	 /* The vertex buffer is too large for hardware (or the swtnl
+	  * module).  Try to split on whole-primitive boundaries,
+	  * otherwise try to split the individual primitives.
+	  */
+	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
+			   min_index, max_index, draw, limits );
+      }
+      else {
+	 /* Why were we called? */
+	 assert(0);
+      }
+   }
+}
+
diff --git a/src/mesa/vbo/vbo_split.h b/src/mesa/vbo/vbo_split.h
new file mode 100644
index 0000000000..05888d048c
--- /dev/null
+++ b/src/mesa/vbo/vbo_split.h
@@ -0,0 +1,72 @@
+/*
+ * mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file vbo_context.h
+ * \brief VBO builder module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+/**
+ * \mainpage The VBO splitter
+ *
+ * This is the private data used internally to the vbo_split_prims()
+ * helper function.  Nobody outside the vbo_split* files needs to
+ * include or know about this structure.
+ */
+
+
+#ifndef _VBO_SPLIT_H
+#define _VBO_SPLIT_H
+
+#include "vbo.h"
+
+
+/* True if a primitive can be split without copying of vertices, false
+ * otherwise.
+ */
+GLboolean split_prim_inplace(GLenum mode, GLuint *first, GLuint *incr);
+
+void vbo_split_inplace( GLcontext *ctx,
+			const struct gl_client_array *arrays[],
+			const struct _mesa_prim *prim,
+			GLuint nr_prims,
+			const struct _mesa_index_buffer *ib,
+			GLuint min_index,
+			GLuint max_index,
+			vbo_draw_func draw,
+			const struct split_limits *limits );
+
+/* Requires ib != NULL:
+ */
+void vbo_split_copy( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     vbo_draw_func draw,
+		     const struct split_limits *limits );
+
+#endif
diff --git a/src/mesa/vbo/vbo_split_copy.c b/src/mesa/vbo/vbo_split_copy.c
new file mode 100644
index 0000000000..e142dde680
--- /dev/null
+++ b/src/mesa/vbo/vbo_split_copy.c
@@ -0,0 +1,551 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Split indexed primitives with per-vertex copying.
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "macros.h"
+#include "enums.h"
+#include "mtypes.h"
+
+#include "vbo_split.h"
+#include "vbo.h"
+
+
+#define ELT_TABLE_SIZE 16
+
+/* Used for vertex-level splitting of indexed buffers.  Note that
+ * non-indexed primitives may be converted to indexed in some cases
+ * (eg loops, fans) in order to use this splitting path.
+ */
+struct copy_context {
+
+   GLcontext *ctx;
+   const struct gl_client_array **array;
+   const struct _mesa_prim *prim;
+   GLuint nr_prims;
+   const struct _mesa_index_buffer *ib;
+   vbo_draw_func draw;
+
+   const struct split_limits *limits;
+
+   struct {
+      GLuint attr;
+      GLuint size;
+      const struct gl_client_array *array;
+      const GLubyte *src_ptr;
+
+      struct gl_client_array dstarray;
+
+   } varying[VERT_ATTRIB_MAX];
+   GLuint nr_varying;
+
+   const struct gl_client_array *dstarray_ptr[VERT_ATTRIB_MAX];
+   struct _mesa_index_buffer dstib;
+
+   GLuint *translated_elt_buf;
+   const GLuint *srcelt;
+
+   /* A baby hash table to avoid re-emitting (some) duplicate
+    * vertices when splitting indexed primitives.
+    */
+   struct { 
+      GLuint in;
+      GLuint out;
+   } vert_cache[ELT_TABLE_SIZE];
+      
+
+   GLuint vertex_size;
+   GLubyte *dstbuf;
+   GLubyte *dstptr;		/* dstptr == dstbuf + dstelt_max * vertsize */
+   GLuint dstbuf_size;	/* in vertices */
+   GLuint dstbuf_nr;		/* count of emitted vertices, also the
+				 * largest value in dstelt.  Our
+				 * MaxIndex.
+				 */
+
+   GLuint *dstelt;
+   GLuint dstelt_nr;
+   GLuint dstelt_size;
+
+#define MAX_PRIM 32
+   struct _mesa_prim dstprim[MAX_PRIM];
+   GLuint dstprim_nr;
+
+};
+
+
+static GLuint type_size( GLenum type )
+{
+   switch(type) {
+   case GL_BYTE: return sizeof(GLbyte);
+   case GL_UNSIGNED_BYTE: return sizeof(GLubyte);
+   case GL_SHORT: return sizeof(GLshort);
+   case GL_UNSIGNED_SHORT: return sizeof(GLushort);
+   case GL_INT: return sizeof(GLint);
+   case GL_UNSIGNED_INT: return sizeof(GLuint);
+   case GL_FLOAT: return sizeof(GLfloat);
+   case GL_DOUBLE: return sizeof(GLdouble);
+   default: return 0;
+   }
+}
+
+static GLuint attr_size( const struct gl_client_array *array )
+{
+   return array->Size * type_size(array->Type);
+}
+
+
+/* Starts returning true slightly before the buffer fills, to ensure
+ * that there is sufficient room for any remaining vertices to finish
+ * off the prim:
+ */
+static GLboolean check_flush( struct copy_context *copy )
+{
+   if (copy->dstbuf_nr + 4 > copy->dstbuf_size)
+      return GL_TRUE;
+
+   if (copy->dstelt_nr + 4 > copy->dstelt_size)
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
+
+static void flush( struct copy_context *copy )
+{
+   GLuint i;
+
+   /* Set some counters: 
+    */
+   copy->dstib.count = copy->dstelt_nr;
+
+   copy->draw( copy->ctx,
+	       copy->dstarray_ptr,
+	       copy->dstprim,
+	       copy->dstprim_nr,
+	       &copy->dstib,
+	       0,
+	       copy->dstbuf_nr );
+
+   /* Reset all pointers: 
+    */
+   copy->dstprim_nr = 0;
+   copy->dstelt_nr = 0;
+   copy->dstbuf_nr = 0;
+   copy->dstptr = copy->dstbuf;
+
+   /* Clear the vertex cache:
+    */
+   for (i = 0; i < ELT_TABLE_SIZE; i++)
+      copy->vert_cache[i].in = ~0;
+}
+
+
+
+static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
+{
+   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
+
+/*    _mesa_printf("begin %s (%d)\n", _mesa_lookup_enum_by_nr(mode), begin_flag); */
+		
+   prim->mode = mode;
+   prim->begin = begin_flag;
+}
+
+
+/* Use a hashtable to attempt to identify recently-emitted vertices
+ * and avoid re-emitting them.
+ */
+static GLuint elt(struct copy_context *copy, GLuint elt_idx)
+{
+   GLuint elt = copy->srcelt[elt_idx];
+   GLuint slot = elt & (ELT_TABLE_SIZE-1);
+
+/*    _mesa_printf("elt %d\n", elt); */
+
+   /* Look up the incoming element in the vertex cache.  Re-emit if
+    * necessary.   
+    */
+   if (copy->vert_cache[slot].in != elt) {
+      GLubyte *csr = copy->dstptr;
+      GLuint i;
+
+/*       _mesa_printf("  --> emit to dstelt %d\n", copy->dstbuf_nr); */
+
+      for (i = 0; i < copy->nr_varying; i++) {
+	 const struct gl_client_array *srcarray = copy->varying[i].array;
+	 const GLubyte *srcptr = copy->varying[i].src_ptr + elt * srcarray->StrideB;
+
+	 memcpy(csr, srcptr, copy->varying[i].size);
+	 csr += copy->varying[i].size;
+
+	 if (0) 
+	 {
+	    const GLuint *f = (const GLuint *)srcptr;
+	    GLuint j;
+	    _mesa_printf("  varying %d: ", i);
+	    for(j = 0; j < copy->varying[i].size / 4; j++)
+	       _mesa_printf("%x ", f[j]);
+	    _mesa_printf("\n");
+	 }
+	       
+      }
+
+      copy->vert_cache[slot].in = elt;
+      copy->vert_cache[slot].out = copy->dstbuf_nr++;
+      copy->dstptr += copy->vertex_size;
+
+      assert(csr == copy->dstptr);
+      assert(copy->dstptr == (copy->dstbuf + 
+				    copy->dstbuf_nr * 
+				    copy->vertex_size));
+   }
+/*    else */
+/*       _mesa_printf("  --> reuse vertex\n"); */
+   
+/*    _mesa_printf("  --> emit %d\n", copy->vert_cache[slot].out); */
+   copy->dstelt[copy->dstelt_nr++] = copy->vert_cache[slot].out;
+   return check_flush(copy);
+}
+
+static void end( struct copy_context *copy, GLboolean end_flag )
+{
+   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
+
+/*    _mesa_printf("end (%d)\n", end_flag); */
+
+   prim->end = end_flag;
+   prim->count = copy->dstelt_nr - prim->start;
+
+   if (++copy->dstprim_nr == MAX_PRIM ||
+       check_flush(copy)) 
+      flush(copy);
+}
+
+
+
+static void replay_elts( struct copy_context *copy )
+{
+   GLuint i, j, k;
+   GLboolean split;
+
+   for (i = 0; i < copy->nr_prims; i++) {
+      const struct _mesa_prim *prim = &copy->prim[i];
+      const GLuint start = prim->start;
+      GLuint first, incr;
+
+      switch (prim->mode) {
+	 
+      case GL_LINE_LOOP:
+	 /* Convert to linestrip and emit the final vertex explicitly,
+	  * but only in the resultant strip that requires it.
+	  */
+	 j = 0;
+	 while (j != prim->count) {
+	    begin(copy, GL_LINE_STRIP, prim->begin && j == 0);
+
+	    for (split = GL_FALSE; j != prim->count && !split; j++)
+	       split = elt(copy, start + j);
+
+	    if (j == prim->count) {
+	       /* Done, emit final line.  Split doesn't matter as
+		* it is always raised a bit early so we can emit
+		* the last verts if necessary!
+		*/
+	       if (prim->end) 
+		  (void)elt(copy, start + 0);
+
+	       end(copy, prim->end);
+	    }
+	    else {
+	       /* Wrap
+		*/
+	       assert(split);
+	       end(copy, 0);
+	       j--;
+	    }
+	 }
+	 break;
+
+      case GL_TRIANGLE_FAN:
+      case GL_POLYGON:
+	 j = 2;
+	 while (j != prim->count) {
+	    begin(copy, prim->mode, prim->begin && j == 0);
+
+	    split = elt(copy, start+0); 
+	    assert(!split);
+
+	    split = elt(copy, start+j-1); 
+	    assert(!split);
+
+	    for (; j != prim->count && !split; j++)
+	       split = elt(copy, start+j);
+
+	    end(copy, prim->end && j == prim->count);
+
+	    if (j != prim->count) {
+	       /* Wrapped the primitive, need to repeat some vertices:
+		*/
+	       j -= 1;
+	    }
+	 }
+	 break;
+
+      default:
+	 (void)split_prim_inplace(prim->mode, &first, &incr);
+	 
+	 j = 0;
+	 while (j != prim->count) {
+
+	    begin(copy, prim->mode, prim->begin && j == 0);
+
+	    split = 0;
+	    for (k = 0; k < first; k++, j++)
+	       split |= elt(copy, start+j);
+
+	    assert(!split);
+
+	    for (; j != prim->count && !split; )
+	       for (k = 0; k < incr; k++, j++)
+		  split |= elt(copy, start+j);
+
+	    end(copy, prim->end && j == prim->count);
+
+	    if (j != prim->count) {
+	       /* Wrapped the primitive, need to repeat some vertices:
+		*/
+	       assert(j > first - incr);
+	       j -= (first - incr);
+	    }
+	 }
+	 break;
+      }
+   }
+
+   if (copy->dstprim_nr)
+      flush(copy);
+}
+
+
+static void replay_init( struct copy_context *copy )
+{
+   GLcontext *ctx = copy->ctx;
+   GLuint i;
+   GLuint offset;
+   const GLvoid *srcptr;
+
+   /* Make a list of varying attributes and their vbo's.  Also
+    * calculate vertex size.
+    */
+   copy->vertex_size = 0;
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      struct gl_buffer_object *vbo = copy->array[i]->BufferObj;
+
+      if (copy->array[i]->StrideB == 0) {
+	 copy->dstarray_ptr[i] = copy->array[i];
+      }
+      else {
+	 GLuint j = copy->nr_varying++;
+	 
+	 copy->varying[j].attr = i;
+	 copy->varying[j].array = copy->array[i];
+	 copy->varying[j].size = attr_size(copy->array[i]);
+	 copy->vertex_size += attr_size(copy->array[i]);
+      
+	 if (vbo->Name && !vbo->Pointer) 
+	    ctx->Driver.MapBuffer(ctx,
+				  GL_ARRAY_BUFFER_ARB, 
+				  GL_WRITE_ONLY, /* XXX */
+				  vbo);
+
+	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
+						 copy->array[i]->Ptr);
+
+	 copy->dstarray_ptr[i] = &copy->varying[j].dstarray;
+      }
+   }
+
+   /* There must always be an index buffer.  Currently require the
+    * caller convert non-indexed prims to indexed.  Could alternately
+    * do it internally.
+    */
+   if (copy->ib->obj->Name && !copy->ib->obj->Pointer) 
+      ctx->Driver.MapBuffer(ctx, 
+			    GL_ARRAY_BUFFER_ARB, /* XXX */
+			    GL_WRITE_ONLY, /* XXX */
+			    copy->ib->obj);
+
+   srcptr = (const GLubyte *)ADD_POINTERS(copy->ib->obj->Pointer, copy->ib->ptr);
+
+   switch (copy->ib->type) {
+   case GL_UNSIGNED_BYTE:
+      copy->translated_elt_buf = _mesa_malloc(sizeof(GLuint) * copy->ib->count);
+      copy->srcelt = copy->translated_elt_buf;
+
+      for (i = 0; i < copy->ib->count; i++)
+	 copy->translated_elt_buf[i] = ((const GLubyte *)srcptr)[i];
+      break;
+
+   case GL_UNSIGNED_SHORT:
+      copy->translated_elt_buf = _mesa_malloc(sizeof(GLuint) * copy->ib->count);
+      copy->srcelt = copy->translated_elt_buf;
+
+      for (i = 0; i < copy->ib->count; i++)
+	 copy->translated_elt_buf[i] = ((const GLushort *)srcptr)[i];
+      break;
+
+   case GL_UNSIGNED_INT:
+      copy->translated_elt_buf = NULL;
+      copy->srcelt = (const GLuint *)srcptr;
+      break;
+   }
+   
+
+   /* Figure out the maximum allowed vertex buffer size:
+    */
+   if (copy->vertex_size * copy->limits->max_verts <= copy->limits->max_vb_size) {
+      copy->dstbuf_size = copy->limits->max_verts;
+   }
+   else {
+      copy->dstbuf_size = copy->limits->max_vb_size / copy->vertex_size;
+   }
+
+   /* Allocate an output vertex buffer:
+    *
+    * XXX:  This should be a VBO!
+    */
+   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * 
+			       copy->vertex_size);   
+   copy->dstptr = copy->dstbuf;
+
+   /* Setup new vertex arrays to point into the output buffer: 
+    */
+   for (offset = 0, i = 0; i < copy->nr_varying; i++) {
+      const struct gl_client_array *src = copy->varying[i].array;
+      struct gl_client_array *dst = &copy->varying[i].dstarray;
+
+      dst->Size = src->Size;
+      dst->Type = src->Type;
+      dst->Stride = copy->vertex_size;
+      dst->StrideB = copy->vertex_size;
+      dst->Ptr = copy->dstbuf + offset;
+      dst->Enabled = GL_TRUE;
+      dst->Normalized = GL_TRUE;
+      dst->BufferObj = ctx->Array.NullBufferObj;
+      dst->_MaxElement = copy->dstbuf_size; /* may be less! */
+
+      offset += copy->varying[i].size;
+   }
+
+   /* Allocate an output element list:
+    */
+   copy->dstelt_size = MIN2(65536,
+			    copy->ib->count * 2 + 3);
+   copy->dstelt_size = MIN2(copy->dstelt_size,
+			    copy->limits->max_indices);
+   copy->dstelt = _mesa_malloc(sizeof(GLuint) * copy->dstelt_size);
+   copy->dstelt_nr = 0;
+
+   /* Setup the new index buffer to point to the allocated element
+    * list:
+    */
+   copy->dstib.count = 0;	/* duplicates dstelt_nr */
+   copy->dstib.type = GL_UNSIGNED_INT;
+   copy->dstib.obj = ctx->Array.NullBufferObj;
+   copy->dstib.ptr = copy->dstelt;
+}
+
+
+static void replay_finish( struct copy_context *copy )
+{
+   GLcontext *ctx = copy->ctx;
+   GLuint i;
+
+   /* Free our vertex and index buffers: 
+    */
+   _mesa_free(copy->translated_elt_buf);
+   _mesa_free(copy->dstbuf);
+   _mesa_free(copy->dstelt);
+   
+   /* Unmap VBO's 
+    */
+   for (i = 0; i < copy->nr_varying; i++) {
+      struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
+
+      if (vbo->Name && vbo->Pointer) 
+	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, vbo);
+   }
+
+   /* Unmap index buffer:
+    */
+   if (copy->ib->obj->Name && copy->ib->obj->Pointer) {
+      ctx->Driver.UnmapBuffer(ctx, 
+			      GL_ARRAY_BUFFER_ARB, /* XXX */
+			      copy->ib->obj);
+   }
+}
+
+void vbo_split_copy( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     vbo_draw_func draw,
+		     const struct split_limits *limits )
+{
+   struct copy_context copy;
+   GLuint i;
+
+   memset(&copy, 0, sizeof(copy));
+
+   /* Require indexed primitives:
+    */
+   assert(ib);
+   
+   copy.ctx = ctx;
+   copy.array = arrays;
+   copy.prim = prim;
+   copy.nr_prims = nr_prims;
+   copy.ib = ib;
+   copy.draw = draw;
+   copy.limits = limits;
+
+
+   /* Clear the vertex cache:
+    */
+   for (i = 0; i < ELT_TABLE_SIZE; i++)
+      copy.vert_cache[i].in = ~0;
+
+
+   replay_init(&copy);
+   replay_elts(&copy);
+   replay_finish(&copy);
+}
diff --git a/src/mesa/vbo/vbo_split_inplace.c b/src/mesa/vbo/vbo_split_inplace.c
new file mode 100644
index 0000000000..ea62866e7c
--- /dev/null
+++ b/src/mesa/vbo/vbo_split_inplace.c
@@ -0,0 +1,286 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "mtypes.h"
+#include "macros.h"
+#include "enums.h"
+#include "vbo_split.h"
+
+
+#define MAX_PRIM 32
+
+/* Used for splitting without copying.
+ */
+struct split_context {
+   GLcontext *ctx;
+   const struct gl_client_array **array;
+   const struct _mesa_prim *prim;
+   GLuint nr_prims;
+   const struct _mesa_index_buffer *ib;
+   GLuint min_index;
+   GLuint max_index;
+   vbo_draw_func draw;
+
+   const struct split_limits *limits;
+
+   struct _mesa_prim dstprim[MAX_PRIM];
+   GLuint dstprim_nr;
+};
+
+
+
+
+static void flush_vertex( struct split_context *split )
+{
+   GLint min_index, max_index;
+
+   if (!split->dstprim_nr) 
+      return;
+
+   if (split->ib) {
+      /* This should basically be multipass rendering over the same
+       * unchanging set of VBO's.  Would like the driver not to
+       * re-upload the data, or swtnl not to re-transform the
+       * vertices.
+       */
+      assert(split->max_index - split->min_index < split->limits->max_verts);
+      min_index = split->min_index;
+      max_index = split->max_index;
+   }
+   else {
+      /* Non-indexed rendering.  Cannot assume that the primitives are
+       * ordered by increasing vertex, because of entrypoints like
+       * MultiDrawArrays.
+       */
+      GLuint i;
+      min_index = split->dstprim[0].start;
+      max_index = min_index + split->dstprim[0].count - 1;
+
+      for (i = 1; i < split->dstprim_nr; i++) {
+	 GLuint tmp_min = split->dstprim[i].start;
+	 GLuint tmp_max = tmp_min + split->dstprim[i].count - 1;
+
+	 if (tmp_min < min_index) 
+	    min_index = tmp_min;
+
+	 if (tmp_max > max_index) 
+	    max_index = tmp_max;
+      }
+   }
+
+   assert(max_index >= min_index);
+
+   split->draw( split->ctx, 
+		split->array, 
+		split->dstprim,
+		split->dstprim_nr,
+		NULL,
+		min_index,
+		max_index);
+
+   split->dstprim_nr = 0;
+}
+
+
+static struct _mesa_prim *next_outprim( struct split_context *split )
+{
+   if (split->dstprim_nr == MAX_PRIM-1) {
+      flush_vertex(split);
+   }
+
+   {
+      struct _mesa_prim *prim = &split->dstprim[split->dstprim_nr++];
+      memset(prim, 0, sizeof(*prim));
+      return prim;
+   }
+}
+
+static int align(int value, int alignment)
+{
+   return (value + alignment - 1) & ~(alignment - 1);
+}
+
+
+
+/* Break large primitives into smaller ones.  If not possible, convert
+ * the primitive to indexed and pass to split_elts().
+ */
+static void split_prims( struct split_context *split) 
+{
+   GLuint csr = 0;
+   GLuint i;
+
+   for (i = 0; i < split->nr_prims; i++) {
+      const struct _mesa_prim *prim = &split->prim[i];
+      GLuint first, incr;
+      GLboolean split_inplace = split_prim_inplace(prim->mode, &first, &incr);
+      GLuint count;
+
+      /* Always wrap on an even numbered vertex to avoid problems with
+       * triangle strips.  
+       */
+      GLuint available = align(split->limits->max_verts - csr - 1, 2); 
+      assert(split->limits->max_verts >= csr);
+
+      if (prim->count < first)
+	 continue;
+      
+      count = prim->count - (prim->count - first) % incr; 
+
+
+      if ((available < count && !split_inplace) || 
+	  (available < first && split_inplace)) {
+	 flush_vertex(split);
+	 csr = 0;
+	 available = align(split->limits->max_verts - csr - 1, 2);
+      }
+      
+      if (available >= count) {
+	 struct _mesa_prim *outprim = next_outprim(split);
+	 *outprim = *prim;
+	 csr += prim->count;
+	 available = align(split->limits->max_verts - csr - 1, 2); 
+      } 
+      else if (split_inplace) {
+	 GLuint j, nr;
+
+
+	 for (j = 0 ; j < count ; ) {
+	    GLuint remaining = count - j;
+	    struct _mesa_prim *outprim = next_outprim(split);
+
+	    nr = MIN2( available, remaining );
+	    
+	    nr -= (nr - first) % incr;
+	    
+	    outprim->mode = prim->mode;
+	    outprim->begin = (j == 0 && prim->begin);
+	    outprim->end = (nr == remaining && prim->end);
+	    outprim->start = prim->start + j;
+	    outprim->count = nr;
+	    
+	    if (nr == remaining) {
+	       /* Finished. 
+		*/
+	       j += nr;		
+	       csr += nr;
+	       available = align(split->limits->max_verts - csr - 1, 2); 
+	    }
+	    else {
+	       /* Wrapped the primitive: 
+		*/
+	       j += nr - (first - incr);
+	       flush_vertex(split);
+	       csr = 0;
+	       available = align(split->limits->max_verts - csr - 1, 2); 
+	    }
+	 }
+      }
+      else if (split->ib == NULL) {
+	 /* XXX: could at least send the first max_verts off from the
+	  * inplace buffers.
+	  */
+
+	 /* else convert to indexed primitive and pass to split_elts,
+	  * which will do the necessary copying and turn it back into a
+	  * vertex primitive for rendering...
+	  */
+	 struct _mesa_index_buffer ib;
+	 struct _mesa_prim tmpprim;
+	 GLuint *elts = malloc(count * sizeof(GLuint));
+	 GLuint j;
+	 
+	 for (j = 0; j < count; j++)
+	    elts[j] = prim->start + j;
+
+	 ib.count = count;
+	 ib.type = GL_UNSIGNED_INT;
+	 ib.obj = split->ctx->Array.NullBufferObj;
+	 ib.ptr = elts;
+	    
+	 tmpprim = *prim;
+	 tmpprim.indexed = 1;
+	 tmpprim.start = 0;
+	 tmpprim.count = count;
+
+	 flush_vertex(split);
+
+	 vbo_split_copy(split->ctx,
+			split->array,
+			&tmpprim, 1, 
+			&ib,
+			split->draw,
+			split->limits);
+	    
+	 free(elts);
+      }
+      else {
+	 flush_vertex(split);
+
+	 vbo_split_copy(split->ctx,
+			split->array,
+			prim, 1, 
+			split->ib,
+			split->draw,
+			split->limits);
+      }
+   }
+
+   flush_vertex(split);
+}
+
+
+void vbo_split_inplace( GLcontext *ctx,
+			const struct gl_client_array *arrays[],
+			const struct _mesa_prim *prim,
+			GLuint nr_prims,
+			const struct _mesa_index_buffer *ib,
+			GLuint min_index,
+			GLuint max_index,
+			vbo_draw_func draw,
+			const struct split_limits *limits )
+{
+   struct split_context split;
+
+   memset(&split, 0, sizeof(split));
+
+   split.ctx = ctx;
+   split.array = arrays;
+   split.prim = prim;
+   split.nr_prims = nr_prims;
+   split.ib = ib;
+   split.min_index = min_index;
+   split.max_index = max_index;
+   split.draw = draw;
+   split.limits = limits;
+
+   split_prims( &split );
+}
+
+