aboutsummaryrefslogtreecommitdiff
path: root/mesalib/src/mesa/tnl
diff options
context:
space:
mode:
authormarha <marha@users.sourceforge.net>2011-08-29 08:51:20 +0200
committermarha <marha@users.sourceforge.net>2011-08-29 08:51:20 +0200
commit01df5d59e56a1b060568f8cad2e89f7eea22fc70 (patch)
tree9db83037fd85d0974b60fc1a05e0665083f26000 /mesalib/src/mesa/tnl
parentfd1f4d9fe3ea67fa6def8ee4927a8f71e0440f12 (diff)
downloadvcxsrv-01df5d59e56a1b060568f8cad2e89f7eea22fc70.tar.gz
vcxsrv-01df5d59e56a1b060568f8cad2e89f7eea22fc70.tar.bz2
vcxsrv-01df5d59e56a1b060568f8cad2e89f7eea22fc70.zip
xwininfo libX11 libXmu libxcb mesa xserver xkeyboard-config git update 29
aug 2011
Diffstat (limited to 'mesalib/src/mesa/tnl')
-rw-r--r--mesalib/src/mesa/tnl/NOTES200
-rw-r--r--mesalib/src/mesa/tnl/t_draw.c39
-rw-r--r--mesalib/src/mesa/tnl/t_pipeline.c432
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 b1967e654..86af4b7cf 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
+};