Merge remote-tracking branch 'origin/released'

Conflicts:
	apps/xwininfo/configure.ac
	apps/xwininfo/xwininfo.c
	libX11/configure.ac
	libX11/specs/libX11/AppC.xml
	libX11/specs/libX11/AppD.xml
	libX11/specs/libX11/CH03.xml
	libX11/specs/libX11/CH04.xml
	libX11/specs/libX11/CH05.xml
	libX11/specs/libX11/CH06.xml
	libX11/specs/libX11/CH07.xml
	libX11/specs/libX11/CH08.xml
	libX11/specs/libX11/CH09.xml
	libX11/specs/libX11/CH11.xml
	libX11/specs/libX11/CH12.xml
	libX11/specs/libX11/CH13.xml
	libX11/specs/libX11/CH14.xml
	libX11/specs/libX11/CH15.xml
	libX11/specs/libX11/CH16.xml
	libxcb/configure.ac
	libxcb/src/c_client.py
	libxcb/src/xcb_auth.c
	libxcb/src/xcb_util.c
	mesalib/common.py
	mesalib/configs/linux-dri
	mesalib/docs/GL3.txt
	mesalib/docs/download.html
	mesalib/docs/install.html
	mesalib/include/GL/internal/dri_interface.h
	mesalib/scons/custom.py
	mesalib/scons/gallium.py
	mesalib/src/gallium/auxiliary/util/u_math.h
	mesalib/src/gallium/auxiliary/util/u_vbuf_mgr.c
	mesalib/src/glsl/ast_function.cpp
	mesalib/src/glsl/ast_to_hir.cpp
	mesalib/src/glsl/glcpp/glcpp-parse.y
	mesalib/src/glsl/glsl_parser_extras.cpp
	mesalib/src/glsl/glsl_parser_extras.h
	mesalib/src/glsl/ir.cpp
	mesalib/src/glsl/ir.h
	mesalib/src/glsl/ir_clone.cpp
	mesalib/src/glsl/ir_print_visitor.cpp
	mesalib/src/glsl/ir_validate.cpp
	mesalib/src/glsl/linker.cpp
	mesalib/src/glsl/main.cpp
	mesalib/src/glsl/opt_tree_grafting.cpp
	mesalib/src/mapi/glapi/gen/Makefile
	mesalib/src/mapi/glapi/gen/gl_XML.py
	mesalib/src/mapi/glapi/gen/gl_table.py
	mesalib/src/mapi/glapi/glapi_mapi_tmp.h
	mesalib/src/mapi/glapi/glapi_sparc.S
	mesalib/src/mapi/glapi/glapi_x86-64.S
	mesalib/src/mapi/glapi/glapi_x86.S
	mesalib/src/mapi/glapi/glapitemp.h
	mesalib/src/mapi/glapi/glprocs.h
	mesalib/src/mapi/mapi/u_thread.c
	mesalib/src/mapi/mapi/u_thread.h
	mesalib/src/mesa/SConscript
	mesalib/src/mesa/drivers/common/driverfuncs.c
	mesalib/src/mesa/drivers/windows/gldirect/dglcontext.c
	mesalib/src/mesa/main/api_arrayelt.c
	mesalib/src/mesa/main/bufferobj.c
	mesalib/src/mesa/main/compiler.h
	mesalib/src/mesa/main/dd.h
	mesalib/src/mesa/main/dlist.c
	mesalib/src/mesa/main/enums.c
	mesalib/src/mesa/main/es_generator.py
	mesalib/src/mesa/main/fbobject.c
	mesalib/src/mesa/main/imports.h
	mesalib/src/mesa/main/mtypes.h
	mesalib/src/mesa/main/pbo.c
	mesalib/src/mesa/main/remap_helper.h
	mesalib/src/mesa/main/shared.c
	mesalib/src/mesa/main/texgetimage.c
	mesalib/src/mesa/main/teximage.c
	mesalib/src/mesa/main/uniforms.c
	mesalib/src/mesa/program/ir_to_mesa.cpp
	mesalib/src/mesa/program/prog_optimize.c
	mesalib/src/mesa/program/register_allocate.c
	mesalib/src/mesa/program/register_allocate.h
	mesalib/src/mesa/sources.mak
	mesalib/src/mesa/state_tracker/st_cb_bufferobjects.c
	mesalib/src/mesa/state_tracker/st_cb_texture.c
	mesalib/src/mesa/state_tracker/st_texture.h
	mesalib/src/mesa/swrast/s_context.c
	mesalib/src/mesa/tnl/t_draw.c
	mesalib/src/mesa/vbo/vbo_exec_array.c
	mesalib/src/mesa/vbo/vbo_save_api.c
	pixman/pixman/pixman-arm-common.h
	pixman/pixman/pixman-fast-path.c
	pixman/pixman/pixman-inlines.h
	pixman/test/Makefile.am
	pixman/test/utils.c
	pixman/test/utils.h
	xorg-server/configure.ac
	xorg-server/fb/fbblt.c
	xorg-server/hw/xquartz/darwin.c
	xorg-server/xkeyboard-config/symbols/de

3 files changed, 346 insertions, 325 deletions

diff --git a/mesalib/src/mesa/tnl/NOTES b/mesalib/src/mesa/tnl/NOTES
index 9a35e948b..43c469201 100644
--- a/mesalib/src/mesa/tnl/NOTES
+++ b/mesalib/src/mesa/tnl/NOTES
@@ -1,102 +1,100 @@
-INTRODUCTION

-

-A generic, configurable software implementation of GL transformation &

-lighting.

-

-This module provides an implementation of the routines required by the

-'vtxfmt' mechanism of core mesa for tnl functionality in all

-combinations of compile and execute modes.

-

-Most current drivers use the tnl module exclusively to provide this

-functionality, though there is an experimental alternate

-implementation provided by the tnl_dd/t_dd_imm_* files which can

-handle a small subset of GL states in execute mode only.

-

-

-STATE

-

-To create and destroy the module:

-

-	GLboolean _tnl_CreateContext( struct gl_context *ctx );

-	void _tnl_DestroyContext( struct gl_context *ctx );

-

-The module is not active by default, and must be installed by calling

-_tnl_Wakeup().  This function installs internal tnl functions into all

-the vtxfmt dispatch hooks, thus taking over the task of transformation

-and lighting entirely:

-

-        void _tnl_wakeup_exec( struct gl_context *ctx );

-	void _tnl_wakeup_save_exec( struct gl_context *ctx );

-

-   

-This module tracks state changes internally and maintains derived

-values based on the current state.  For this to work, the driver

-ensure the following funciton is called whenever the state changes and

-the swsetup module is 'awake':

-

-	void _tnl_InvalidateState( struct gl_context *ctx, GLuint new_state );

-

-There is no explicit call to put the tnl module to sleep.  Simply

-install other function pointers into all the vtxfmt dispatch slots,

-and (optionally) cease calling _tnl_InvalidateState().

-

-CUSTOMIZATION

-

-The module provides customizability through several mechanisms.  The

-most important is by allowing drivers to specify the pipeline through

-which vertex data is passed, including its eventual transfer to

-rasterization hardware (or software).

-

-The default pipeline is specified in t_pipeline.c, and is usually a

-starting point for driver pipelines.  Some drivers will remove a stage

-where hardware provides support for the implemented operation (for

-instance fog where per-pixel hardware fog is available, as in the dri

-tdfx driver), or add stages to shortcircuit latter operations (for

-example taking advantage of hardware support for strips and other

-higher-level primitives (for example the radeon driver).

-

-In addition, the following functions provide further tweaks:

-

-extern void

-_tnl_need_projected_coords( struct gl_context *ctx, GLboolean flag );

-

-	- Direct the default vertex transformation stage to

-          produce/not produce projected clip coordinates.

-	  

-extern void

-_tnl_need_dlist_loopback( struct gl_context *ctx, GLboolean flag );

-      

-        - Direct the display list component of the tnl module to

-          replay display lists as 'glVertex' type calls, rather than

-          passing the display list data directly into the tnl pipeline

-          mechanism.  

-

-	  This allows display lists to be replayed by the tnl module

-	  even when the module is not strictly active.

-

-

-extern void

-_tnl_need_dlist_norm_lengths( struct gl_context *ctx, GLboolean flag );

-

-	- Direct the display list component to enable/disable caching

-          1/length values for display list normals.  Doing so is

-          ususally helpful when lighting is performed in software, but

-          wasteful otherwise.

-

-

-DRIVER INTERFACE

-

-The module itself offers a minimal driver interface:

-

-	 void (*RunPipeline)( struct gl_context *ctx );

-

-Normally this is set to _tnl_RunPipeline(), however the driver can use

-this hook to wrap checks or other code around this call. 

-

-In addition, the driver interface for the default render pipeline

-stage is housed in the tnl context struct (this could be cleaner).  

-

-

-RENDER DRIVER INTERFACE

-

+INTRODUCTION
+
+A generic, configurable software implementation of GL transformation &
+lighting.
+
+This module provides an implementation of the routines required by the
+'vtxfmt' mechanism of core mesa for tnl functionality in all
+combinations of compile and execute modes.
+
+Most current drivers use the tnl module exclusively to provide this
+functionality.
+
+
+STATE
+
+To create and destroy the module:
+
+	GLboolean _tnl_CreateContext( struct gl_context *ctx );
+	void _tnl_DestroyContext( struct gl_context *ctx );
+
+The module is not active by default, and must be installed by calling
+_tnl_Wakeup().  This function installs internal tnl functions into all
+the vtxfmt dispatch hooks, thus taking over the task of transformation
+and lighting entirely:
+
+        void _tnl_wakeup_exec( struct gl_context *ctx );
+	void _tnl_wakeup_save_exec( struct gl_context *ctx );
+
+   
+This module tracks state changes internally and maintains derived
+values based on the current state.  For this to work, the driver
+ensure the following funciton is called whenever the state changes and
+the swsetup module is 'awake':
+
+	void _tnl_InvalidateState( struct gl_context *ctx, GLuint new_state );
+
+There is no explicit call to put the tnl module to sleep.  Simply
+install other function pointers into all the vtxfmt dispatch slots,
+and (optionally) cease calling _tnl_InvalidateState().
+
+CUSTOMIZATION
+
+The module provides customizability through several mechanisms.  The
+most important is by allowing drivers to specify the pipeline through
+which vertex data is passed, including its eventual transfer to
+rasterization hardware (or software).
+
+The default pipeline is specified in t_pipeline.c, and is usually a
+starting point for driver pipelines.  Some drivers will remove a stage
+where hardware provides support for the implemented operation (for
+instance fog where per-pixel hardware fog is available),
+or add stages to shortcircuit latter operations (for
+example taking advantage of hardware support for strips and other
+higher-level primitives (for example the radeon driver).
+
+In addition, the following functions provide further tweaks:
+
+extern void
+_tnl_need_projected_coords( struct gl_context *ctx, GLboolean flag );
+
+	- Direct the default vertex transformation stage to
+          produce/not produce projected clip coordinates.
+	  
+extern void
+_tnl_need_dlist_loopback( struct gl_context *ctx, GLboolean flag );
+      
+        - Direct the display list component of the tnl module to
+          replay display lists as 'glVertex' type calls, rather than
+          passing the display list data directly into the tnl pipeline
+          mechanism.  
+
+	  This allows display lists to be replayed by the tnl module
+	  even when the module is not strictly active.
+
+
+extern void
+_tnl_need_dlist_norm_lengths( struct gl_context *ctx, GLboolean flag );
+
+	- Direct the display list component to enable/disable caching
+          1/length values for display list normals.  Doing so is
+          ususally helpful when lighting is performed in software, but
+          wasteful otherwise.
+
+
+DRIVER INTERFACE
+
+The module itself offers a minimal driver interface:
+
+	 void (*RunPipeline)( struct gl_context *ctx );
+
+Normally this is set to _tnl_RunPipeline(), however the driver can use
+this hook to wrap checks or other code around this call. 
+
+In addition, the driver interface for the default render pipeline
+stage is housed in the tnl context struct (this could be cleaner).  
+
+
+RENDER DRIVER INTERFACE
+
 See t_context.h for the definition and explanation of this.
\ No newline at end of file
diff --git a/mesalib/src/mesa/tnl/t_draw.c b/mesalib/src/mesa/tnl/t_draw.c
index a2ca6225b..f3f324314 100644
--- a/mesalib/src/mesa/tnl/t_draw.c
+++ b/mesalib/src/mesa/tnl/t_draw.c
@@ -280,10 +280,9 @@ static void bind_inputs( struct gl_context *ctx,
 	 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);

+	    ctx->Driver.MapBufferRange(ctx, 0, inputs[i]->BufferObj->Size,

+				       GL_MAP_READ_BIT,

+				       inputs[i]->BufferObj);

 	    

 	    assert(inputs[i]->BufferObj->Pointer);

 	 }

@@ -348,18 +347,32 @@ static void bind_indices( struct gl_context *ctx,
    }

 

    if (ib->obj->Name && !ib->obj->Pointer) {

+      unsigned map_size;

+

+      switch (ib->type) {

+      case GL_UNSIGNED_BYTE:

+	 map_size = ib->count * sizeof(GLubyte);

+	 break;

+      case GL_UNSIGNED_SHORT:

+	 map_size = ib->count * sizeof(GLushort);

+	 break;

+      case GL_UNSIGNED_INT:

+	 map_size = ib->count * sizeof(GLuint);

+	 break;

+      default:

+	 assert(0);

+	 map_size = 0;

+      }

+

       bo[*nr_bo] = ib->obj;

       (*nr_bo)++;

-      ctx->Driver.MapBuffer(ctx, 

-			    GL_ELEMENT_ARRAY_BUFFER,

-			    GL_READ_ONLY_ARB,

-			    ib->obj);

-

+      ptr = ctx->Driver.MapBufferRange(ctx, (GLsizeiptr) ib->ptr, map_size,

+				       GL_MAP_READ_BIT, ib->obj);

       assert(ib->obj->Pointer);

+   } else {

+      ptr = ib->ptr;

    }

 

-   ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);

-

    if (ib->type == GL_UNSIGNED_INT && VB->Primitive[0].basevertex == 0) {

       VB->Elts = (GLuint *) ptr;

    }

@@ -402,9 +415,7 @@ static void unmap_vbos( struct gl_context *ctx,
 {

    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]);

+      ctx->Driver.UnmapBuffer(ctx, bo[i]);

    }

 }

 

diff --git a/mesalib/src/mesa/tnl/t_pipeline.c b/mesalib/src/mesa/tnl/t_pipeline.c
index 27d7ecff7..881d5d5f5 100644
--- a/mesalib/src/mesa/tnl/t_pipeline.c
+++ b/mesalib/src/mesa/tnl/t_pipeline.c
@@ -1,210 +1,222 @@
-/*

- * Mesa 3-D graphics library

- * Version:  6.5.3

- *

- * Copyright (C) 1999-2007  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 "main/glheader.h"

-#include "main/context.h"

-#include "main/imports.h"

-#include "main/mtypes.h"

-

-#include "t_context.h"

-#include "t_pipeline.h"

-#include "t_vp_build.h"

-#include "t_vertex.h"

-

-void _tnl_install_pipeline( struct gl_context *ctx,

-			    const struct tnl_pipeline_stage **stages )

-{

-   TNLcontext *tnl = TNL_CONTEXT(ctx);

-   GLuint i;

-

-   tnl->pipeline.new_state = ~0;

-

-   /* Create a writeable copy of each stage.

-    */

-   for (i = 0 ; i < MAX_PIPELINE_STAGES && stages[i] ; i++) {

-      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];

-      memcpy(s, stages[i], sizeof(*s));

-      if (s->create)

-	 s->create(ctx, s);

-   }

-

-   tnl->pipeline.nr_stages = i;

-}

-

-void _tnl_destroy_pipeline( struct gl_context *ctx )

-{

-   TNLcontext *tnl = TNL_CONTEXT(ctx);

-   GLuint i;

-

-   for (i = 0 ; i < tnl->pipeline.nr_stages ; i++) {

-      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];

-      if (s->destroy)

-	 s->destroy(s);

-   }

-

-   tnl->pipeline.nr_stages = 0;

-}

-

-

-

-static GLuint check_input_changes( struct gl_context *ctx )

-{

-   TNLcontext *tnl = TNL_CONTEXT(ctx);

-   GLuint i;

-   

-   for (i = 0; i <= _TNL_LAST_MAT; i++) {

-      if (tnl->vb.AttribPtr[i]->size != tnl->pipeline.last_attrib_size[i] ||

-	  tnl->vb.AttribPtr[i]->stride != tnl->pipeline.last_attrib_stride[i]) {

-	 tnl->pipeline.last_attrib_size[i] = tnl->vb.AttribPtr[i]->size;

-	 tnl->pipeline.last_attrib_stride[i] = tnl->vb.AttribPtr[i]->stride;

-	 tnl->pipeline.input_changes |= 1<<i;

-      }

-   }

-

-   return tnl->pipeline.input_changes;

-}

-

-

-static GLuint check_output_changes( struct gl_context *ctx )

-{

-#if 0

-   TNLcontext *tnl = TNL_CONTEXT(ctx);

-   

-   for (i = 0; i < VERT_RESULT_MAX; i++) {

-      if (tnl->vb.ResultPtr[i]->size != tnl->last_result_size[i] ||

-	  tnl->vb.ResultPtr[i]->stride != tnl->last_result_stride[i]) {

-	 tnl->last_result_size[i] = tnl->vb.ResultPtr[i]->size;

-	 tnl->last_result_stride[i] = tnl->vb.ResultPtr[i]->stride;

-	 tnl->pipeline.output_changes |= 1<<i;

-      }

-   }

-

-   if (tnl->pipeline.output_changes) 

-      tnl->Driver.NotifyOutputChanges( ctx, tnl->pipeline.output_changes );

-   

-   return tnl->pipeline.output_changes;

-#else

-   return ~0;

-#endif

-}

-

-

-void _tnl_run_pipeline( struct gl_context *ctx )

-{

-   TNLcontext *tnl = TNL_CONTEXT(ctx);

-   unsigned short __tmp;

-   GLuint i;

-

-   if (!tnl->vb.Count)

-      return;

-

-   /* Check for changed input sizes or change in stride to/from zero

-    * (ie const or non-const).

-    */

-   if (check_input_changes( ctx ) || tnl->pipeline.new_state) {

-      if (ctx->VertexProgram._MaintainTnlProgram)

-	 _tnl_UpdateFixedFunctionProgram( ctx );

-

-      for (i = 0; i < tnl->pipeline.nr_stages ; i++) {

-	 struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];

-	 if (s->validate)

-	    s->validate( ctx, s );

-      }

-      

-      tnl->pipeline.new_state = 0;

-      tnl->pipeline.input_changes = 0;

-      

-      /* Pipeline can only change its output in response to either a

-       * statechange or an input size/stride change.  No other changes

-       * are allowed.

-       */

-      if (check_output_changes( ctx ))

-	 _tnl_notify_pipeline_output_change( ctx );

-   }

-

-   START_FAST_MATH(__tmp);

-

-   for (i = 0; i < tnl->pipeline.nr_stages ; i++) {

-      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];

-      if (!s->run( ctx, s ))

-	 break;

-   }

-

-   END_FAST_MATH(__tmp);

-}

-

-

-

-/* The default pipeline.  This is useful for software rasterizers, and

- * simple hardware rasterizers.  For customization, I don't recommend

- * tampering with the internals of these stages in the way that

- * drivers did in Mesa 3.4.  These stages are basically black boxes,

- * and should be left intact.

- *

- * To customize the pipeline, consider:

- *

- * - removing redundant stages (making sure that the software rasterizer

- *   can cope with this on fallback paths).  An example is fog

- *   coordinate generation, which is not required in the FX driver.

- *

- * - replacing general-purpose machine-independent stages with

- *   general-purpose machine-specific stages.  There is no example of

- *   this to date, though it must be borne in mind that all subsequent

- *   stages that reference the output of the new stage must cope with

- *   any machine-specific data introduced.  This may not be easy

- *   unless there are no such stages (ie the new stage is the last in

- *   the pipe).

- *

- * - inserting optimized (but specialized) stages ahead of the

- *   general-purpose fallback implementation.  For example, the old

- *   fastpath mechanism, which only works when the VB->Elts input is

- *   available, can be duplicated by placing the fastpath stage at the

- *   head of this pipeline.  Such specialized stages are currently

- *   constrained to have no outputs (ie. they must either finish the *

- *   pipeline by returning GL_FALSE from run(), or do nothing).

- *

- * Some work can be done to lift some of the restrictions in the final

- * case, if it becomes necessary to do so.

- */

-const struct tnl_pipeline_stage *_tnl_default_pipeline[] = {

-   &_tnl_vertex_transform_stage,

-   &_tnl_normal_transform_stage,

-   &_tnl_lighting_stage,

-   &_tnl_texgen_stage,

-   &_tnl_texture_transform_stage,

-   &_tnl_point_attenuation_stage,

-   &_tnl_vertex_program_stage, 

-   &_tnl_fog_coordinate_stage,

-   &_tnl_render_stage,

-   NULL 

-};

-

-const struct tnl_pipeline_stage *_tnl_vp_pipeline[] = {

-   &_tnl_vertex_program_stage,

-   &_tnl_render_stage,

-   NULL

-};

+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5.3
+ *
+ * Copyright (C) 1999-2007  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 "main/glheader.h"
+#include "main/context.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+#include "t_vp_build.h"
+#include "t_vertex.h"
+
+void _tnl_install_pipeline( struct gl_context *ctx,
+			    const struct tnl_pipeline_stage **stages )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+
+   tnl->pipeline.new_state = ~0;
+
+   /* Create a writeable copy of each stage.
+    */
+   for (i = 0 ; i < MAX_PIPELINE_STAGES && stages[i] ; i++) {
+      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+      memcpy(s, stages[i], sizeof(*s));
+      if (s->create)
+	 s->create(ctx, s);
+   }
+
+   tnl->pipeline.nr_stages = i;
+}
+
+void _tnl_destroy_pipeline( struct gl_context *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+
+   for (i = 0 ; i < tnl->pipeline.nr_stages ; i++) {
+      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+      if (s->destroy)
+	 s->destroy(s);
+   }
+
+   tnl->pipeline.nr_stages = 0;
+}
+
+
+
+static GLuint check_input_changes( struct gl_context *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+   
+   for (i = 0; i <= _TNL_LAST_MAT; i++) {
+      if (tnl->vb.AttribPtr[i]->size != tnl->pipeline.last_attrib_size[i] ||
+	  tnl->vb.AttribPtr[i]->stride != tnl->pipeline.last_attrib_stride[i]) {
+	 tnl->pipeline.last_attrib_size[i] = tnl->vb.AttribPtr[i]->size;
+	 tnl->pipeline.last_attrib_stride[i] = tnl->vb.AttribPtr[i]->stride;
+	 tnl->pipeline.input_changes |= 1<<i;
+      }
+   }
+
+   return tnl->pipeline.input_changes;
+}
+
+
+static GLuint check_output_changes( struct gl_context *ctx )
+{
+#if 0
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   
+   for (i = 0; i < VERT_RESULT_MAX; i++) {
+      if (tnl->vb.ResultPtr[i]->size != tnl->last_result_size[i] ||
+	  tnl->vb.ResultPtr[i]->stride != tnl->last_result_stride[i]) {
+	 tnl->last_result_size[i] = tnl->vb.ResultPtr[i]->size;
+	 tnl->last_result_stride[i] = tnl->vb.ResultPtr[i]->stride;
+	 tnl->pipeline.output_changes |= 1<<i;
+      }
+   }
+
+   if (tnl->pipeline.output_changes) 
+      tnl->Driver.NotifyOutputChanges( ctx, tnl->pipeline.output_changes );
+   
+   return tnl->pipeline.output_changes;
+#else
+   return ~0;
+#endif
+}
+
+
+void _tnl_run_pipeline( struct gl_context *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   unsigned short __tmp;
+   GLuint i;
+
+   if (!tnl->vb.Count)
+      return;
+
+   /* Check for changed input sizes or change in stride to/from zero
+    * (ie const or non-const).
+    */
+   if (check_input_changes( ctx ) || tnl->pipeline.new_state) {
+      if (ctx->VertexProgram._MaintainTnlProgram)
+	 _tnl_UpdateFixedFunctionProgram( ctx );
+
+      for (i = 0; i < tnl->pipeline.nr_stages ; i++) {
+	 struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+	 if (s->validate)
+	    s->validate( ctx, s );
+      }
+      
+      tnl->pipeline.new_state = 0;
+      tnl->pipeline.input_changes = 0;
+      
+      /* Pipeline can only change its output in response to either a
+       * statechange or an input size/stride change.  No other changes
+       * are allowed.
+       */
+      if (check_output_changes( ctx ))
+	 _tnl_notify_pipeline_output_change( ctx );
+   }
+
+#ifndef _OPENMP
+   /* Don't adjust FPU precision mode in case multiple threads are to be used.
+    * This would require that the additional threads also changed the FPU mode
+    * which is quite a mess as this had to be done in all parallelized sections;
+    * otherwise the master thread and all other threads are running in different
+    * modes, producing inconsistent results.
+    * Note that all x64 implementations don't define/use START_FAST_MATH, so
+    * this is "hack" is only used in i386 mode
+    */
+   START_FAST_MATH(__tmp);
+#endif
+
+   for (i = 0; i < tnl->pipeline.nr_stages ; i++) {
+      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+      if (!s->run( ctx, s ))
+	 break;
+   }
+
+#ifndef _OPENMP
+   END_FAST_MATH(__tmp);
+#endif
+}
+
+
+
+/* The default pipeline.  This is useful for software rasterizers, and
+ * simple hardware rasterizers.  For customization, I don't recommend
+ * tampering with the internals of these stages in the way that
+ * drivers did in Mesa 3.4.  These stages are basically black boxes,
+ * and should be left intact.
+ *
+ * To customize the pipeline, consider:
+ *
+ * - removing redundant stages (making sure that the software rasterizer
+ *   can cope with this on fallback paths).  An example is fog
+ *   coordinate generation, which is not required in the FX driver.
+ *
+ * - replacing general-purpose machine-independent stages with
+ *   general-purpose machine-specific stages.  There is no example of
+ *   this to date, though it must be borne in mind that all subsequent
+ *   stages that reference the output of the new stage must cope with
+ *   any machine-specific data introduced.  This may not be easy
+ *   unless there are no such stages (ie the new stage is the last in
+ *   the pipe).
+ *
+ * - inserting optimized (but specialized) stages ahead of the
+ *   general-purpose fallback implementation.  For example, the old
+ *   fastpath mechanism, which only works when the VB->Elts input is
+ *   available, can be duplicated by placing the fastpath stage at the
+ *   head of this pipeline.  Such specialized stages are currently
+ *   constrained to have no outputs (ie. they must either finish the *
+ *   pipeline by returning GL_FALSE from run(), or do nothing).
+ *
+ * Some work can be done to lift some of the restrictions in the final
+ * case, if it becomes necessary to do so.
+ */
+const struct tnl_pipeline_stage *_tnl_default_pipeline[] = {
+   &_tnl_vertex_transform_stage,
+   &_tnl_normal_transform_stage,
+   &_tnl_lighting_stage,
+   &_tnl_texgen_stage,
+   &_tnl_texture_transform_stage,
+   &_tnl_point_attenuation_stage,
+   &_tnl_vertex_program_stage, 
+   &_tnl_fog_coordinate_stage,
+   &_tnl_render_stage,
+   NULL 
+};
+
+const struct tnl_pipeline_stage *_tnl_vp_pipeline[] = {
+   &_tnl_vertex_program_stage,
+   &_tnl_render_stage,
+   NULL
+};