Merge commit 'origin/gallium-master-merge'

This is the big merge of the gallium-0.2 branch into master. gallium-master-merge was just the staging area for it. Both gallium-0.2 and gallium-master-merge are considered closed now. Conflicts: progs/demos/Makefile src/mesa/main/state.c src/mesa/main/texenvprogram.c
author: Brian Paul <[email protected]> 2009-02-10 16:44:02 -0700
committer: Brian Paul <[email protected]> 2009-02-10 16:44:02 -0700
commit: 5340b6dff73a0a23531ce2a5f28fba8303adab6e (patch)
tree: b141fc3648568dd8b941c966059e6ed32a8bd0ad /src/mesa
parent: 9fd26daec24f21dbe17afcb2e2ab272667ee9a69 (diff)
parent: ee4c921b65fb76998711f3c40330505cbc49a0e0 (diff)
169 files changed, 21060 insertions, 768 deletions
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index 6b4057030ef..78b71d4676f 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -3,8 +3,7 @@
 TOP = ../..
 include $(TOP)/configs/current
 
-include sources
-
+include sources.mak
 
 .SUFFIXES : .cpp
 
@@ -19,9 +18,11 @@ include sources
 
 
 
+
 # Default: build dependencies, then asm_subdirs, then convenience
 # libs (.a) and finally the device drivers:
-default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs
+default: depend asm_subdirs libmesa.a libmesagallium.a libglapi.a \
+	driver_subdirs
 
 
 
@@ -32,10 +33,17 @@ default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs
 libmesa.a: $(MESA_OBJECTS)
 	@ $(MKLIB) -o mesa -static $(MESA_OBJECTS)
 
+# Make archive of subset of core mesa object files for gallium
+libmesagallium.a: $(MESA_GALLIUM_OBJECTS)
+	@ $(MKLIB) -o mesagallium -static $(MESA_GALLIUM_OBJECTS)
+
 # Make archive of gl* API dispatcher functions only
 libglapi.a: $(GLAPI_OBJECTS)
-	@ $(MKLIB) -o glapi -static $(GLAPI_OBJECTS)
-
+	@if [ "${WINDOW_SYSTEM}" = "dri" ] ; then \
+		touch libglapi.a ; \
+	else \
+		$(MKLIB) -o glapi -static $(GLAPI_OBJECTS) ; \
+	fi
 
 ######################################################################
 # Device drivers
diff --git a/src/mesa/Makefile.DJ b/src/mesa/Makefile.DJ
deleted file mode 100644
index 06a13fb1abd..00000000000
--- a/src/mesa/Makefile.DJ
+++ /dev/null
@@ -1,166 +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.
-
-# DOS/DJGPP core makefile v1.7 for Mesa
-#
-#  Copyright (C) 2002 - Daniel Borca
-#  Email : [email protected]
-#  Web   : http://www.geocities.com/dborca
-
-
-#
-#  Available options:
-#
-#    Environment variables:
-#	CFLAGS
-#
-#	GLIDE		path to Glide3 SDK; used with FX.
-#			default = $(TOP)/glide3
-#	FX=1		build for 3dfx Glide3. Note that this disables
-#			compilation of most DMesa code and requires fxMesa.
-#			As a consequence, you'll need the DJGPP Glide3
-#			library to build any application.
-#			default = no
-#	X86=1		optimize for x86 (if possible, use MMX, SSE, 3DNow).
-#			default = no
-#
-#    Targets:
-#	all:		build GL
-#	clean:		remove object files
-#
-
-
-
-.PHONY: all clean
-.INTERMEDIATE: x86/gen_matypes.exe
-
-TOP = ../..
-GLIDE ?= $(TOP)/glide3
-LIBDIR = $(TOP)/lib
-GL_LIB = libgl.a
-GL_DXE = gl.dxe
-GL_IMP = libigl.a
-
-export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib
-
-CC = gcc
-CFLAGS += $(INCLUDE_DIRS)
-CFLAGS += -DUSE_EXTERNAL_DXTN_LIB=1
-ifeq ($(FX),1)
-CFLAGS += -D__DOS__
-CFLAGS += -I$(GLIDE)/include -DFX
-LIBNAME = "Mesa/FX DJGPP"
-else
-LIBNAME = "Mesa DJGPP"
-endif
-
-AR = ar
-ARFLAGS = crus
-
-HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe)
-
-ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),)
-UNLINK = del $(subst /,\,$(1))
-else
-UNLINK = $(RM) $(1)
-endif
-
-include sources
-
-ifeq ($(X86),1)
-CFLAGS += -DUSE_X86_ASM
-CFLAGS += -DUSE_MMX_ASM
-CFLAGS += -DUSE_SSE_ASM
-CFLAGS += -DUSE_3DNOW_ASM
-X86_SOURCES += $(X86_API)
-else
-X86_SOURCES =
-endif
-
-DRIVER_SOURCES = \
-	drivers/dos/dmesa.c
-ifeq ($(FX),1)
-DRIVER_SOURCES += \
-	$(GLIDE_DRIVER_SOURCES)
-else
-DRIVER_SOURCES += \
-	$(OSMESA_DRIVER_SOURCES) \
-	drivers/dos/video.c \
-	drivers/dos/vesa.c \
-	drivers/dos/blit.S \
-	drivers/dos/vga.c \
-	drivers/dos/null.c \
-	drivers/dos/dpmi.c
-endif
-
-SOURCES = $(CORE_SOURCES) $(X86_SOURCES) $(COMMON_DRIVER_SOURCES) $(DRIVER_SOURCES)
-
-OBJECTS = $(addsuffix .o,$(basename $(SOURCES)))
-
-X86_OBJECTS = $(addsuffix .o,$(basename $(X86_SOURCES)))
-
-.c.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.S.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.s.o:
-	$(CC) -o $@ $(CFLAGS) -x assembler-with-cpp -c $<
-
-all: $(LIBDIR)/$(GL_LIB) $(LIBDIR)/$(GL_DXE) $(LIBDIR)/$(GL_IMP)
-
-$(LIBDIR)/$(GL_LIB): $(OBJECTS)
-	$(AR) $(ARFLAGS) $@ $^
-
-$(LIBDIR)/$(GL_DXE) $(LIBDIR)/$(GL_IMP): $(OBJECTS)
-ifeq ($(HAVEDXE3),)
-	$(warning Missing DXE3 package... Skipping $(GL_DXE))
-else
-ifeq ($(FX),1)
-	-dxe3gen -o $(LIBDIR)/$(GL_DXE) -Y $(LIBDIR)/$(GL_IMP) -D $(LIBNAME) -E _gl -E _DMesa -P glide3x.dxe -U $^
-else
-	-dxe3gen -o $(LIBDIR)/$(GL_DXE) -Y $(LIBDIR)/$(GL_IMP) -D $(LIBNAME) -E _gl -E _DMesa -U $^
-endif
-endif
-
-$(X86_OBJECTS): x86/matypes.h
-
-x86/matypes.h: x86/gen_matypes.exe
-	$< > $@
-
-x86/gen_matypes.exe: x86/gen_matypes.c
-	$(CC) -o $@ $(CFLAGS) -s $<
-
-clean:
-	-$(call UNLINK,array_cache/*.o)
-	-$(call UNLINK,glapi/*.o)
-	-$(call UNLINK,main/*.o)
-	-$(call UNLINK,math/*.o)
-	-$(call UNLINK,shader/*.o)
-	-$(call UNLINK,sparc/*.o)
-	-$(call UNLINK,ppc/*.o)
-	-$(call UNLINK,swrast/*.o)
-	-$(call UNLINK,swrast_setup/*.o)
-	-$(call UNLINK,tnl/*.o)
-	-$(call UNLINK,x86/*.o)
-	-$(call UNLINK,drivers/common/*.o)
-	-$(call UNLINK,drivers/dos/*.o)
-	-$(call UNLINK,drivers/glide/*.o)
diff --git a/src/mesa/Makefile.mgw b/src/mesa/Makefile.mgw
index 886b344ec9b..097c390a83e 100644
--- a/src/mesa/Makefile.mgw
+++ b/src/mesa/Makefile.mgw
@@ -124,7 +124,7 @@ ifneq ($(wildcard $(addsuffix /rm,$(subst :, ,$(PATH)))),)
 UNLINK = $(RM) $(1)
 endif
 
-include sources
+include sources.mak
 
 CFLAGS += $(INCLUDE_DIRS) 
 
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
new file mode 100644
index 00000000000..a878d31dbb5
--- /dev/null
+++ b/src/mesa/SConscript
@@ -0,0 +1,343 @@
+#######################################################################
+# SConscript for Mesa
+
+
+Import('*')
+
+if env['platform'] != 'winddk':
+
+	env = env.Clone()
+	
+	env.Append(CPPPATH = [
+		'#/src/mesa',
+	])
+	
+	if env['platform'] == 'windows':
+		env.Append(CPPDEFINES = [
+			'_GDI32_', # prevent gl* being declared __declspec(dllimport) in MS headers
+			'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers 
+		])	 
+
+	#
+	# Source files
+	#
+	
+	main_sources = [
+		'main/api_arrayelt.c',
+		'main/api_exec.c',
+		'main/api_loopback.c',
+		'main/api_noop.c',
+		'main/api_validate.c',
+		'main/accum.c',
+		'main/attrib.c',
+		'main/arrayobj.c',
+		'main/blend.c',
+		'main/bufferobj.c',
+		'main/buffers.c',
+		'main/clear.c',
+		'main/clip.c',
+		'main/colortab.c',
+		'main/context.c',
+		'main/convolve.c',
+		'main/debug.c',
+		'main/depth.c',
+		'main/depthstencil.c',
+		'main/dlist.c',
+		'main/dlopen.c',
+		'main/drawpix.c',
+		'main/enable.c',
+		'main/enums.c',
+		'main/eval.c',
+		'main/execmem.c',
+		'main/extensions.c',
+		'main/fbobject.c',
+		'main/feedback.c',
+		'main/ffvertex_prog.c',
+		'main/fog.c',
+		'main/framebuffer.c',
+		'main/get.c',
+		'main/getstring.c',
+		'main/hash.c',
+		'main/hint.c',
+		'main/histogram.c',
+		'main/image.c',
+		'main/imports.c',
+		'main/light.c',
+		'main/lines.c',
+		'main/matrix.c',
+		'main/mipmap.c',
+		'main/mm.c',
+		'main/multisample.c',
+		'main/pixel.c',
+		'main/pixelstore.c',
+		'main/points.c',
+		'main/polygon.c',
+		'main/queryobj.c',
+		'main/rastpos.c',
+		'main/rbadaptors.c',
+		'main/readpix.c',
+		'main/renderbuffer.c',
+		'main/scissor.c',
+		'main/shaders.c',
+		'main/state.c',
+		'main/stencil.c',
+		'main/texcompress.c',
+		'main/texcompress_s3tc.c',
+		'main/texcompress_fxt1.c',
+		'main/texenv.c',
+		'main/texenvprogram.c',
+		'main/texformat.c',
+		'main/texgen.c',
+		'main/teximage.c',
+		'main/texobj.c',
+		'main/texparam.c',
+		'main/texrender.c',
+		'main/texstate.c',
+		'main/texstore.c',
+		'main/varray.c',
+		'main/vtxfmt.c',
+	]
+	
+	math_sources = [
+		'math/m_debug_clip.c',
+		'math/m_debug_norm.c',
+		'math/m_debug_xform.c',
+		'math/m_eval.c',
+		'math/m_matrix.c',
+		'math/m_translate.c',
+		'math/m_vector.c',
+		'math/m_xform.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',
+	]
+	
+	vf_sources = [
+		'vf/vf.c',
+		'vf/vf_generic.c',
+		'vf/vf_sse.c',
+	]
+	
+	statetracker_sources = [
+		'state_tracker/st_atom.c',
+		'state_tracker/st_atom_blend.c',
+		'state_tracker/st_atom_clip.c',
+		'state_tracker/st_atom_constbuf.c',
+		'state_tracker/st_atom_depth.c',
+		'state_tracker/st_atom_framebuffer.c',
+		'state_tracker/st_atom_pixeltransfer.c',
+		'state_tracker/st_atom_sampler.c',
+		'state_tracker/st_atom_scissor.c',
+		'state_tracker/st_atom_shader.c',
+		'state_tracker/st_atom_rasterizer.c',
+		'state_tracker/st_atom_stipple.c',
+		'state_tracker/st_atom_texture.c',
+		'state_tracker/st_atom_viewport.c',
+		'state_tracker/st_cb_accum.c',
+		'state_tracker/st_cb_bitmap.c',
+		'state_tracker/st_cb_blit.c',
+		'state_tracker/st_cb_bufferobjects.c',
+		'state_tracker/st_cb_clear.c',
+		'state_tracker/st_cb_flush.c',
+		'state_tracker/st_cb_drawpixels.c',
+		'state_tracker/st_cb_fbo.c',
+		'state_tracker/st_cb_get.c',
+		'state_tracker/st_cb_feedback.c',
+		'state_tracker/st_cb_program.c',
+		'state_tracker/st_cb_queryobj.c',
+		'state_tracker/st_cb_rasterpos.c',
+		'state_tracker/st_cb_readpixels.c',
+		'state_tracker/st_cb_strings.c',
+		'state_tracker/st_cb_texture.c',
+		'state_tracker/st_context.c',
+		'state_tracker/st_debug.c',
+		'state_tracker/st_draw.c',
+		'state_tracker/st_draw_feedback.c',
+		'state_tracker/st_extensions.c',
+		'state_tracker/st_format.c',
+		'state_tracker/st_framebuffer.c',
+		'state_tracker/st_gen_mipmap.c',
+		'state_tracker/st_mesa_to_tgsi.c',
+		'state_tracker/st_program.c',
+		'state_tracker/st_texture.c',
+	]
+	
+	shader_sources = [
+		'shader/arbprogparse.c',
+		'shader/arbprogram.c',
+		'shader/atifragshader.c',
+		'shader/grammar/grammar_mesa.c',
+		'shader/nvfragparse.c',
+		'shader/nvprogram.c',
+		'shader/nvvertparse.c',
+		'shader/program.c',
+		'shader/prog_cache.c',
+		'shader/prog_debug.c',
+		'shader/prog_execute.c',
+		'shader/prog_instruction.c',
+		'shader/prog_noise.c',
+		'shader/prog_parameter.c',
+		'shader/prog_print.c',
+		'shader/prog_statevars.c',
+		'shader/prog_uniform.c',
+		'shader/programopt.c',
+		'shader/shader_api.c',
+	]
+	
+	slang_sources = [
+		'shader/slang/slang_builtin.c',
+		'shader/slang/slang_codegen.c',
+		'shader/slang/slang_compile.c',
+		'shader/slang/slang_compile_function.c',
+		'shader/slang/slang_compile_operation.c',
+		'shader/slang/slang_compile_struct.c',
+		'shader/slang/slang_compile_variable.c',
+		'shader/slang/slang_emit.c',
+		'shader/slang/slang_ir.c',
+		'shader/slang/slang_label.c',
+		'shader/slang/slang_link.c',
+		'shader/slang/slang_log.c',
+		'shader/slang/slang_mem.c',
+		'shader/slang/slang_preprocess.c',
+		'shader/slang/slang_print.c',
+		'shader/slang/slang_simplify.c',
+		'shader/slang/slang_storage.c',
+		'shader/slang/slang_typeinfo.c',
+		'shader/slang/slang_vartable.c',
+		'shader/slang/slang_utility.c',
+	]
+	
+	mesa_sources = (
+		main_sources +
+		math_sources +
+		vbo_sources +
+		vf_sources +
+		statetracker_sources +
+		shader_sources +
+		slang_sources
+	)
+
+	glapi_sources = [
+		'main/dispatch.c',
+		'glapi/glapi.c',
+		'glapi/glapi_getproc.c',
+		'glapi/glthread.c',
+	]
+	
+	#
+	# Assembly sources
+	#
+	if gcc and env['machine'] == 'x86':
+		env.Append(CPPDEFINES = [
+			'USE_X86_ASM', 
+			'USE_MMX_ASM',
+			'USE_3DNOW_ASM',
+			'USE_SSE_ASM',
+		])
+		mesa_sources += [
+			'x86/common_x86.c',
+			'x86/x86.c',
+			'x86/3dnow.c',
+			'x86/sse.c',
+			'x86/common_x86_asm.S',
+			'x86/x86_xform2.S',
+			'x86/x86_xform3.S',
+			'x86/x86_xform4.S',
+			'x86/x86_cliptest.S',
+			'x86/mmx_blend.S',
+			'x86/3dnow_xform1.S',
+			'x86/3dnow_xform2.S',
+			'x86/3dnow_xform3.S',
+			'x86/3dnow_xform4.S',
+			'x86/3dnow_normal.S',
+			'x86/sse_xform1.S',
+			'x86/sse_xform2.S',
+			'x86/sse_xform3.S',
+			'x86/sse_xform4.S',
+			'x86/sse_normal.S',
+			'x86/read_rgba_span_x86.S',
+		]
+		glapi_sources += [
+			'x86/glapi_x86.S',
+		]
+	elif gcc and env['machine'] == 'x86_64':
+		env.Append(CPPDEFINES = [
+			'USE_X86_64_ASM', 
+		])
+		mesa_sources += [
+			'x86-64/x86-64.c',
+			'x86-64/xform4.S',
+		]
+		glapi_sources += [
+			'x86-64/glapi_x86-64.S'
+		]
+	elif gcc and env['machine'] == 'ppc':
+		env.Append(CPPDEFINES = [
+			'USE_PPC_ASM', 
+			'USE_VMX_ASM', 
+		])
+		mesa_sources += [
+			'ppc/common_ppc.c',
+		]
+		glapi_sources += [
+		]
+	elif gcc and env['machine'] == 'sparc':
+		mesa_sources += [
+			'sparc/sparc.c',
+			'sparc/clip.S',
+			'sparc/norm.S',
+			'sparc/xform.S',
+		]
+		glapi_sources += [
+			'sparc/glapi_sparc.S'
+		]
+	else:
+		pass
+	
+	# Generate matypes.h
+	if gcc and env['machine'] in ('x86', 'x86_64'):
+		# See http://www.scons.org/wiki/UsingCodeGenerators
+		gen_matypes = env.Program(
+			target = 'gen_matypes',
+			source = 'x86/gen_matypes.c',
+		)
+		matypes = env.Command(
+			'matypes.h',
+			gen_matypes,
+			gen_matypes[0].abspath + ' > $TARGET',
+		)
+		# Add the dir containing the generated header (somewhere inside  the
+		# build dir) to the include path  
+		env.Append(CPPPATH = [matypes[0].dir])
+	
+	#
+	# Libraries
+	# 
+
+	mesa = env.ConvenienceLibrary(
+		target = 'mesa',
+		source = mesa_sources,
+	)
+	Export('mesa')
+	
+	if not dri:
+		glapi = env.ConvenienceLibrary(
+			target = 'glapi',
+			source = glapi_sources,
+		)
+		Export('glapi')
+
diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template
index 2dc3664cc6a..2fa36bab3f9 100644
--- a/src/mesa/drivers/dri/Makefile.template
+++ b/src/mesa/drivers/dri/Makefile.template
@@ -2,13 +2,16 @@
 
 MESA_MODULES = $(TOP)/src/mesa/libmesa.a
 
-COMMON_SOURCES = \
+COMMON_GALLIUM_SOURCES = \
         ../common/utils.c \
-        ../common/texmem.c \
         ../common/vblank.c \
         ../common/dri_util.c \
-        ../common/xmlconfig.c \
-        ../common/drirenderbuffer.c 
+        ../common/xmlconfig.c
+
+COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
+        ../../common/driverfuncs.c \
+        ../common/texmem.c \
+        ../common/drirenderbuffer.c
 
 ifeq ($(WINDOW_SYSTEM),dri)
 WINOBJ=
@@ -59,9 +62,9 @@ SHARED_INCLUDES = \
 default: symlinks depend $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME)
 
 
-$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
+$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
 	$(MKLIB) -o $@ -noprefix -linker '$(CC)' -ldflags '$(LDFLAGS)' \
-		$(OBJECTS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS)
+		$(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS)
 
 
 $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME)
@@ -69,6 +72,7 @@ $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME)
 
 
 depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \
 		$(ASM_SOURCES)
diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h
index c6781f1c7ab..c95a5c8299a 100644
--- a/src/mesa/drivers/dri/common/dri_util.h
+++ b/src/mesa/drivers/dri/common/dri_util.h
@@ -1,25 +1,3 @@
-/* $XFree86: xc/lib/GL/dri/dri_util.h,v 1.1 2002/02/22 21:32:52 dawes Exp $ */
-/**
- * \file dri_util.h
- * DRI utility functions definitions.
- *
- * This module acts as glue between GLX and the actual hardware driver.  A DRI
- * driver doesn't really \e have to use any of this - it's optional.  But, some
- * useful stuff is done here that otherwise would have to be duplicated in most
- * drivers.
- * 
- * Basically, these utility functions take care of some of the dirty details of
- * screen initialization, context creation, context binding, DRM setup, etc.
- *
- * These functions are compiled into each DRI driver so libGL.so knows nothing
- * about them.
- *
- * \sa dri_util.c.
- * 
- * \author Kevin E. Martin <[email protected]>
- * \author Brian Paul <[email protected]>
- */
-
 /*
  * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
  * All Rights Reserved.
@@ -45,6 +23,26 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+/**
+ * \file dri_util.h
+ * DRI utility functions definitions.
+ *
+ * This module acts as glue between GLX and the actual hardware driver.  A DRI
+ * driver doesn't really \e have to use any of this - it's optional.  But, some
+ * useful stuff is done here that otherwise would have to be duplicated in most
+ * drivers.
+ * 
+ * Basically, these utility functions take care of some of the dirty details of
+ * screen initialization, context creation, context binding, DRM setup, etc.
+ *
+ * These functions are compiled into each DRI driver so libGL.so knows nothing
+ * about them.
+ *
+ * \sa dri_util.c.
+ * 
+ * \author Kevin E. Martin <[email protected]>
+ * \author Brian Paul <[email protected]>
+ */
 
 #ifndef _DRI_UTIL_H_
 #define _DRI_UTIL_H_
diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c
index b99bf2033b2..15af99136cd 100644
--- a/src/mesa/drivers/dri/common/drirenderbuffer.c
+++ b/src/mesa/drivers/dri/common/drirenderbuffer.c
@@ -209,6 +209,8 @@ driUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv)
    struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate;
    if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) {
       ctx->Driver.ResizeBuffers(ctx, fb, dPriv->w, dPriv->h);
+      /* if the driver needs the hw lock for ResizeBuffers, the drawable
+         might have changed again by now */
       assert(fb->Width == dPriv->w);
       assert(fb->Height == dPriv->h);
    }
diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c
index d6de1c0e9bb..5175ab0abbf 100644
--- a/src/mesa/drivers/dri/common/utils.c
+++ b/src/mesa/drivers/dri/common/utils.c
@@ -320,8 +320,10 @@ void driInitSingleExtension( GLcontext * ctx,
 	     */
 	    offset = _glapi_add_dispatch( functions, parameter_signature );
 	    if (offset == -1) {
+#if 0 /* this causes noise with egl */
 		fprintf(stderr, "DISPATCH ERROR! _glapi_add_dispatch failed "
 			"to add %s!\n", functions[0]);
+#endif
 	    }
 	    else if (ext->functions[i].remap_index != -1) {
 		driDispatchRemapTable[ ext->functions[i].remap_index ] = 
@@ -514,6 +516,9 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer,
  *                      \c GLX_SWAP_UNDEFINED_OML.  See the
  *                      GLX_OML_swap_method extension spec for more details.
  * \param num_db_modes  Number of entries in \c db_modes.
+ * \param msaa_samples  Array of msaa sample count. 0 represents a visual
+ *                      without a multisample buffer.
+ * \param num_msaa_modes Number of entries in \c msaa_samples.
  * \param visType       GLX visual type.  Usually either \c GLX_TRUE_COLOR or
  *                      \c GLX_DIRECT_COLOR.
  * 
@@ -533,7 +538,8 @@ __DRIconfig **
 driCreateConfigs(GLenum fb_format, GLenum fb_type,
 		 const uint8_t * depth_bits, const uint8_t * stencil_bits,
 		 unsigned num_depth_stencil_bits,
-		 const GLenum * db_modes, unsigned num_db_modes)
+		 const GLenum * db_modes, unsigned num_db_modes,
+	 	 const u_int8_t * msaa_samples, unsigned num_msaa_modes)
 {
    static const uint8_t bits_table[4][4] = {
      /* R  G  B  A */
@@ -593,9 +599,7 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
    int index;
    __DRIconfig **configs, **c;
    __GLcontextModes *modes;
-   unsigned i;
-   unsigned j;
-   unsigned k;
+   unsigned i, j, k, h;
    unsigned num_modes;
    unsigned num_accum_bits = 2;
 
@@ -668,7 +672,7 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
 	 break;
    }
 
-   num_modes = num_depth_stencil_bits * num_db_modes * num_accum_bits;
+   num_modes = num_depth_stencil_bits * num_db_modes * num_accum_bits * num_msaa_modes;
    configs = _mesa_calloc((num_modes + 1) * sizeof *configs);
    if (configs == NULL)
        return NULL;
@@ -676,66 +680,72 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
     c = configs;
     for ( k = 0 ; k < num_depth_stencil_bits ; k++ ) {
 	for ( i = 0 ; i < num_db_modes ; i++ ) {
-	    for ( j = 0 ; j < num_accum_bits ; j++ ) {
-		*c = _mesa_malloc (sizeof **c);
-		modes = &(*c)->modes;
-		c++;
-
-		memset(modes, 0, sizeof *modes);
-		modes->redBits   = bits[0];
-		modes->greenBits = bits[1];
-		modes->blueBits  = bits[2];
-		modes->alphaBits = bits[3];
-		modes->redMask   = masks[0];
-		modes->greenMask = masks[1];
-		modes->blueMask  = masks[2];
-		modes->alphaMask = masks[3];
-		modes->rgbBits   = modes->redBits + modes->greenBits
-		    + modes->blueBits + modes->alphaBits;
-
-		modes->accumRedBits   = 16 * j;
-		modes->accumGreenBits = 16 * j;
-		modes->accumBlueBits  = 16 * j;
-		modes->accumAlphaBits = (masks[3] != 0) ? 16 * j : 0;
-		modes->visualRating = (j == 0) ? GLX_NONE : GLX_SLOW_CONFIG;
-
-		modes->stencilBits = stencil_bits[k];
-		modes->depthBits = depth_bits[k];
-
-		modes->transparentPixel = GLX_NONE;
-		modes->transparentRed = GLX_DONT_CARE;
-		modes->transparentGreen = GLX_DONT_CARE;
-		modes->transparentBlue = GLX_DONT_CARE;
-		modes->transparentAlpha = GLX_DONT_CARE;
-		modes->transparentIndex = GLX_DONT_CARE;
-		modes->visualType = GLX_DONT_CARE;
-		modes->renderType = GLX_RGBA_BIT;
-		modes->drawableType = GLX_WINDOW_BIT;
-		modes->rgbMode = GL_TRUE;
-
-		if ( db_modes[i] == GLX_NONE ) {
-		    modes->doubleBufferMode = GL_FALSE;
-		}
-		else {
-		    modes->doubleBufferMode = GL_TRUE;
-		    modes->swapMethod = db_modes[i];
-		}
-
-		modes->haveAccumBuffer = ((modes->accumRedBits +
+	    for ( h = 0 ; h < num_msaa_modes; h++ ) {
+	    	for ( j = 0 ; j < num_accum_bits ; j++ ) {
+		    *c = _mesa_malloc (sizeof **c);
+		    modes = &(*c)->modes;
+		    c++;
+
+		    memset(modes, 0, sizeof *modes);
+		    modes->redBits   = bits[0];
+		    modes->greenBits = bits[1];
+		    modes->blueBits  = bits[2];
+		    modes->alphaBits = bits[3];
+		    modes->redMask   = masks[0];
+		    modes->greenMask = masks[1];
+		    modes->blueMask  = masks[2];
+		    modes->alphaMask = masks[3];
+		    modes->rgbBits   = modes->redBits + modes->greenBits
+		    	+ modes->blueBits + modes->alphaBits;
+
+		    modes->accumRedBits   = 16 * j;
+		    modes->accumGreenBits = 16 * j;
+		    modes->accumBlueBits  = 16 * j;
+		    modes->accumAlphaBits = (masks[3] != 0) ? 16 * j : 0;
+		    modes->visualRating = (j == 0) ? GLX_NONE : GLX_SLOW_CONFIG;
+
+		    modes->stencilBits = stencil_bits[k];
+		    modes->depthBits = depth_bits[k];
+
+		    modes->transparentPixel = GLX_NONE;
+		    modes->transparentRed = GLX_DONT_CARE;
+		    modes->transparentGreen = GLX_DONT_CARE;
+		    modes->transparentBlue = GLX_DONT_CARE;
+		    modes->transparentAlpha = GLX_DONT_CARE;
+		    modes->transparentIndex = GLX_DONT_CARE;
+		    modes->visualType = GLX_DONT_CARE;
+		    modes->renderType = GLX_RGBA_BIT;
+		    modes->drawableType = GLX_WINDOW_BIT;
+		    modes->rgbMode = GL_TRUE;
+
+		    if ( db_modes[i] == GLX_NONE ) {
+		    	modes->doubleBufferMode = GL_FALSE;
+		    }
+		    else {
+		    	modes->doubleBufferMode = GL_TRUE;
+		    	modes->swapMethod = db_modes[i];
+		    }
+
+		    modes->samples = msaa_samples[h];
+		    modes->sampleBuffers = modes->samples ? 1 : 0;
+
+
+		    modes->haveAccumBuffer = ((modes->accumRedBits +
 					   modes->accumGreenBits +
 					   modes->accumBlueBits +
 					   modes->accumAlphaBits) > 0);
-		modes->haveDepthBuffer = (modes->depthBits > 0);
-		modes->haveStencilBuffer = (modes->stencilBits > 0);
-
-		modes->bindToTextureRgb = GL_TRUE;
-		modes->bindToTextureRgba = GL_TRUE;
-		modes->bindToMipmapTexture = GL_FALSE;
-		modes->bindToTextureTargets = modes->rgbMode ?
-		    __DRI_ATTRIB_TEXTURE_1D_BIT |
-		    __DRI_ATTRIB_TEXTURE_2D_BIT |
-		    __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT :
-		    0;
+		    modes->haveDepthBuffer = (modes->depthBits > 0);
+		    modes->haveStencilBuffer = (modes->stencilBits > 0);
+
+		    modes->bindToTextureRgb = GL_TRUE;
+		    modes->bindToTextureRgba = GL_TRUE;
+		    modes->bindToMipmapTexture = GL_FALSE;
+		    modes->bindToTextureTargets = modes->rgbMode ?
+		    	__DRI_ATTRIB_TEXTURE_1D_BIT |
+		    	__DRI_ATTRIB_TEXTURE_2D_BIT |
+		    	__DRI_ATTRIB_TEXTURE_RECTANGLE_BIT :
+		    	0;
+		}
 	    }
 	}
     }
diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h
index b0908e530a5..a4ef5092478 100644
--- a/src/mesa/drivers/dri/common/utils.h
+++ b/src/mesa/drivers/dri/common/utils.h
@@ -131,7 +131,8 @@ extern __DRIconfig **
 driCreateConfigs(GLenum fb_format, GLenum fb_type,
 		 const uint8_t * depth_bits, const uint8_t * stencil_bits,
 		 unsigned num_depth_stencil_bits,
-		 const GLenum * db_modes, unsigned num_db_modes);
+		 const GLenum * db_modes, unsigned num_db_modes,
+    		 const uint8_t * msaa_samples, unsigned num_msaa_modes);
 
 const __DRIconfig **driConcatConfigs(const __DRIconfig **a,
 				     const __DRIconfig **b);
diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
index 0cdbbd0ace5..00bdcec7a69 100644
--- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c
+++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
@@ -628,6 +628,7 @@ ffbFillInModes( __DRIscreenPrivate *psp,
 
    uint8_t depth_bits_array[3];
    uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    depth_bits_array[1] = depth_bits;
@@ -641,6 +642,8 @@ ffbFillInModes( __DRIscreenPrivate *psp,
    stencil_bits_array[1] = 0;
    stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
    back_buffer_factor  = (have_back_buffer) ? 3 : 1;
 
@@ -656,9 +659,10 @@ ffbFillInModes( __DRIscreenPrivate *psp,
    configs = driCreateConfigs(fb_format, fb_type,
 			      depth_bits_array, stencil_bits_array,
 			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
+			      back_buffer_factor,
+                               msaa_samples_array, 1);
    if (configs == NULL) {
-    fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+      fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
               __LINE__);
       return NULL;
    }
diff --git a/src/mesa/drivers/dri/glcore/Makefile b/src/mesa/drivers/dri/glcore/Makefile
new file mode 100644
index 00000000000..ac7e1de9285
--- /dev/null
+++ b/src/mesa/drivers/dri/glcore/Makefile
@@ -0,0 +1,84 @@
+# src/mesa/drivers/dri/glcore/Makefile
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = glcore_dri.so
+
+DRIVER_SOURCES = glcore_driver.c \
+		 $(TOP)/src/mesa/drivers/common/driverfuncs.c \
+		 ../common/dri_util.c
+
+C_SOURCES = \
+	$(DRIVER_SOURCES) \
+	$(DRI_SOURCES) 
+
+
+# Include directories
+INCLUDE_DIRS = \
+	-I. \
+	-I../common \
+	-I../dri_client \
+	-I../dri_client/imports \
+	-Iserver \
+	-I$(TOP)/include \
+	-I$(DRM_SOURCE_PATH)/shared-core \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/mesa/glapi \
+	-I$(TOP)/src/mesa/math \
+	-I$(TOP)/src/mesa/transform \
+	-I$(TOP)/src/mesa/shader \
+	-I$(TOP)/src/mesa/swrast \
+	-I$(TOP)/src/mesa/swrast_setup
+
+# Core Mesa objects
+MESA_MODULES = $(TOP)/src/mesa/libmesa.a
+
+# Libraries that the driver shared lib depends on
+LIB_DEPS = -lm -lpthread -lc
+# LIB_DEPS = -lGL -lm -lpthread -lc
+
+
+ASM_SOURCES = 
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend $(TOP)/$(LIB_DIR)/$(LIBNAME)
+
+
+$(TOP)/$(LIB_DIR)/$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile
+	CC="$(CC)" CXX="$(CXX)" $(TOP)/bin/mklib -o $(LIBNAME) -noprefix -install $(TOP)/$(LIB_DIR) \
+		$(OBJECTS) $(WINLIB) $(LIB_DEPS) $(WINOBJ) $(MESA_MODULES)
+
+
+depend: $(C_SOURCES) $(ASM_SOURCES)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(C_SOURCES) $(ASM_SOURCES) \
+		> /dev/null 
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+clean:
+	-rm -f *.o server/*.o
+
+
+include depend
diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c
index 48603f5d79f..9a5a39c1dbc 100644
--- a/src/mesa/drivers/dri/i810/i810screen.c
+++ b/src/mesa/drivers/dri/i810/i810screen.c
@@ -77,6 +77,7 @@ i810FillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -88,13 +89,16 @@ i810FillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
     configs = driCreateConfigs(GL_RGB, GL_UNSIGNED_SHORT_5_6_5,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index e17f18a98f2..4c06170cdf3 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -478,6 +478,7 @@ intelFillInModes(__DRIscreenPrivate *psp,
 
    uint8_t depth_bits_array[3];
    uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    depth_bits_array[1] = depth_bits;
@@ -494,6 +495,8 @@ intelFillInModes(__DRIscreenPrivate *psp,
 
    stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
    back_buffer_factor = (have_back_buffer) ? 3 : 1;
 
@@ -509,7 +512,9 @@ intelFillInModes(__DRIscreenPrivate *psp,
    configs = driCreateConfigs(fb_format, fb_type,
 			      depth_bits_array, stencil_bits_array,
 			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
+			      back_buffer_factor,
+                              msaa_samples_array, 1);
+
    if (configs == NULL) {
     fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
               __LINE__);
@@ -679,7 +684,7 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
    static const GLenum back_buffer_modes[] = {
       GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
    };
-   uint8_t depth_bits[4], stencil_bits[4];
+   uint8_t depth_bits[4], stencil_bits[4], msaa_samples_array[1];
    int color;
    const __DRIconfig **configs = NULL;
 
@@ -730,6 +735,8 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
    depth_bits[3] = 24;
    stencil_bits[3] = 8;
 
+   msaa_samples_array[0] = 0;
+
    fb_format[0] = GL_RGB;
    fb_type[0] = GL_UNSIGNED_SHORT_5_6_5;
 
@@ -748,7 +755,9 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
 			  stencil_bits,
 			  ARRAY_SIZE(depth_bits),
 			  back_buffer_modes,
-			  ARRAY_SIZE(back_buffer_modes));
+			  ARRAY_SIZE(back_buffer_modes),
+                          msaa_samples_array, ARRAY_SIZE(msaa_samples_array));
+
       if (configs == NULL)
 	 configs = new_configs;
       else
diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c
index 6bfb4c32b15..43e59594dd5 100644
--- a/src/mesa/drivers/dri/mach64/mach64_screen.c
+++ b/src/mesa/drivers/dri/mach64/mach64_screen.c
@@ -93,6 +93,7 @@ mach64FillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -104,6 +105,8 @@ mach64FillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -119,7 +122,8 @@ mach64FillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor, back_buffer_modes,
-			       back_buffer_factor);
+			       back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
        fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
 	       __func__, __LINE__);
diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c
index 2d0f387cdc6..926534d6dff 100644
--- a/src/mesa/drivers/dri/mga/mga_xmesa.c
+++ b/src/mesa/drivers/dri/mga/mga_xmesa.c
@@ -129,6 +129,7 @@ mgaFillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[3];
     uint8_t stencil_bits_array[3];
+    uint8_t msaa_samples_array[1];
 
 
     depth_bits_array[0] = 0;
@@ -143,6 +144,8 @@ mgaFillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[1] = 0;
     stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -158,7 +161,8 @@ mgaFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
diff --git a/src/mesa/drivers/dri/r128/r128_screen.c b/src/mesa/drivers/dri/r128/r128_screen.c
index cb3a147dbac..7cda4ca5d38 100644
--- a/src/mesa/drivers/dri/r128/r128_screen.c
+++ b/src/mesa/drivers/dri/r128/r128_screen.c
@@ -422,7 +422,7 @@ r128FillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
-
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -434,6 +434,8 @@ r128FillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -446,26 +448,27 @@ r128FillInModes( __DRIscreenPrivate *psp,
         fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
     }
 
-   configs = driCreateConfigs(fb_format, fb_type,
-			      depth_bits_array, stencil_bits_array,
-			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
-   if (configs == NULL) {
-    fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
-              __LINE__);
-      return NULL;
-   }
+    configs = driCreateConfigs(fb_format, fb_type,
+                               depth_bits_array, stencil_bits_array,
+                               depth_buffer_factor, back_buffer_modes,
+                               back_buffer_factor,
+                               msaa_samples_array, 1);
+    if (configs == NULL) {
+        fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+                __LINE__);
+        return NULL;
+    }
 
-   /* Mark the visual as slow if there are "fake" stencil bits.
-    */
-   for (i = 0; configs[i]; i++) {
-      m = &configs[i]->modes;
-      if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
-         m->visualRating = GLX_SLOW_CONFIG;
-      }
-   }
+    /* Mark the visual as slow if there are "fake" stencil bits.
+     */
+    for (i = 0; configs[i]; i++) {
+        m = &configs[i]->modes;
+        if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
+            m->visualRating = GLX_SLOW_CONFIG;
+        }
+    }
 
-   return (const __DRIconfig **) configs;
+    return (const __DRIconfig **) configs;
 }
 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index f5665c4af19..56c22faafa7 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -276,7 +276,7 @@ radeonFillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
-
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -288,6 +288,8 @@ radeonFillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -303,7 +305,8 @@ radeonFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+			       msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c
index c5decb093af..326d595352f 100644
--- a/src/mesa/drivers/dri/savage/savage_xmesa.c
+++ b/src/mesa/drivers/dri/savage/savage_xmesa.c
@@ -910,7 +910,7 @@ savageFillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
-
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -922,6 +922,8 @@ savageFillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -937,7 +939,8 @@ savageFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
diff --git a/src/mesa/drivers/dri/sis/sis_screen.c b/src/mesa/drivers/dri/sis/sis_screen.c
index b1a5d15236f..9eb27fef087 100644
--- a/src/mesa/drivers/dri/sis/sis_screen.c
+++ b/src/mesa/drivers/dri/sis/sis_screen.c
@@ -77,6 +77,7 @@ sisFillInModes(__DRIscreenPrivate *psp, int bpp)
    };
    uint8_t depth_bits_array[4];
    uint8_t stencil_bits_array[4];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    stencil_bits_array[0] = 0;
@@ -87,6 +88,8 @@ sisFillInModes(__DRIscreenPrivate *psp, int bpp)
    depth_bits_array[3] = 32;
    stencil_bits_array[3] = 0;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = 4;
    back_buffer_factor = 2;
 
@@ -100,7 +103,8 @@ sisFillInModes(__DRIscreenPrivate *psp, int bpp)
 
    configs = driCreateConfigs(fb_format, fb_type, depth_bits_array,
 			      stencil_bits_array, depth_buffer_factor,
-			      back_buffer_modes, back_buffer_factor);
+			      back_buffer_modes, back_buffer_factor,
+                              msaa_samples_array, 1);
    if (configs == NULL) {
       fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, __LINE__);
       return NULL;
diff --git a/src/mesa/drivers/dri/swrast/swrast.c b/src/mesa/drivers/dri/swrast/swrast.c
index a90b30b49d3..b00f4ff2fe2 100644
--- a/src/mesa/drivers/dri/swrast/swrast.c
+++ b/src/mesa/drivers/dri/swrast/swrast.c
@@ -147,6 +147,7 @@ swrastFillInModes(__DRIscreen *psp,
 
     uint8_t depth_bits_array[4];
     uint8_t stencil_bits_array[4];
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = 0;
     depth_bits_array[1] = 0;
@@ -161,6 +162,8 @@ swrastFillInModes(__DRIscreen *psp,
     stencil_bits_array[2] = 0;
     stencil_bits_array[3] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = 4;
     back_buffer_factor = 2;
 
@@ -190,7 +193,7 @@ swrastFillInModes(__DRIscreen *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor, back_buffer_modes,
-			       back_buffer_factor);
+			       back_buffer_factor, msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
 		__LINE__);
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.c b/src/mesa/drivers/dri/tdfx/tdfx_screen.c
index cd22b849511..6d509a4d883 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_screen.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.c
@@ -361,6 +361,7 @@ tdfxFillInModes(__DRIscreenPrivate *psp,
 	static const GLenum db_modes[2] = { GLX_NONE, GLX_SWAP_UNDEFINED_OML };
 	uint8_t depth_bits_array[4];
 	uint8_t stencil_bits_array[4];
+        uint8_t msaa_samples_array[1];
 	if(deep) {
 		depth_bits_array[0] = 0;
 		depth_bits_array[1] = 24;
@@ -377,13 +378,16 @@ tdfxFillInModes(__DRIscreenPrivate *psp,
 		stencil_bits_array[3] = 8;
 	}
 
+	msaa_samples_array[0] = 0;
+
 	return driCreateConfigs(
 		deep ? GL_RGBA : GL_RGB,
 		deep ? GL_UNSIGNED_INT_8_8_8_8 : GL_UNSIGNED_SHORT_5_6_5,
 		depth_bits_array,
 		stencil_bits_array,
 		deep ? 2 : 4,
-		db_modes, 2);
+		db_modes, 2,
+                msaa_samples_array, 1);
 }
 
 /**
diff --git a/src/mesa/drivers/dri/unichrome/via_screen.c b/src/mesa/drivers/dri/unichrome/via_screen.c
index 988f9935ac1..e9f566161c2 100644
--- a/src/mesa/drivers/dri/unichrome/via_screen.c
+++ b/src/mesa/drivers/dri/unichrome/via_screen.c
@@ -342,6 +342,7 @@ viaFillInModes( __DRIscreenPrivate *psp,
      */
     static const uint8_t depth_bits_array[4]   = { 0, 16, 24, 32 };
     static const uint8_t stencil_bits_array[4] = { 0,  0,  8,  0 };
+    uint8_t msaa_samples_array[1] = { 0 };
     const unsigned depth_buffer_factor = 3;
 
     if ( pixel_bits == 16 ) {
@@ -356,7 +357,8 @@ viaFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor, back_buffer_modes,
-			       back_buffer_factor);
+			       back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
 		__LINE__);
diff --git a/src/mesa/glapi/descrip.mms b/src/mesa/glapi/descrip.mms
deleted file mode 100644
index ee96baa6fcc..00000000000
--- a/src/mesa/glapi/descrip.mms
+++ /dev/null
@@ -1,39 +0,0 @@
-# Makefile for core library for VMS
-# contributed by Jouk Jansen  [email protected]
-# Last revision : 29 September 2008
-
-.first
-	define gl [---.include.gl]
-	define main [-.main]
-
-.include [---]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = [---.include],[-.main]
-LIBDIR = [---.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = glapi.c glthread.c glapi_getproc.c
-
-OBJECTS =  glapi.obj,glthread.obj,glapi_getproc.obj
-
-##### RULES #####
-
-VERSION=Mesa V3.4
-
-##### TARGETS #####
-# Make the library
-$(LIBDIR)$(GL_LIB) : $(OBJECTS)
-  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
-
-clean :
-	purge
-	delete *.obj;*
-
-glapi.obj : glapi.c
-
-glthread.obj : glthread.c
-glapi_getproc.obj : glapi_getproc.c
diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c
index c3ebf60719f..9b5144a88b6 100644
--- a/src/mesa/glapi/glapi.c
+++ b/src/mesa/glapi/glapi.c
@@ -98,6 +98,7 @@ _glapi_set_warning_func( _glapi_warning_func func )
 static GLboolean
 warn(void)
 {
+#if !defined(_WIN32_WCE)
    if ((WarnFlag || getenv("MESA_DEBUG") || getenv("LIBGL_DEBUG"))
        && warning_func) {
       return GL_TRUE;
@@ -105,6 +106,9 @@ warn(void)
    else {
       return GL_FALSE;
    }
+#else
+   return GL_FALSE;
+#endif
 }
 
 
diff --git a/src/mesa/glapi/glthread.c b/src/mesa/glapi/glthread.c
index b818f4e26eb..56ddf7c25ac 100644
--- a/src/mesa/glapi/glthread.c
+++ b/src/mesa/glapi/glthread.c
@@ -25,16 +25,14 @@
 
 /*
  * XXX There's probably some work to do in order to make this file
- * truly reusable outside of Mesa.
+ * truly reusable outside of Mesa.  First, the glheader.h include must go.
  */
 
-
 #ifdef HAVE_DIX_CONFIG_H
 #include <dix-config.h>
 #endif
 
-#include <stdlib.h>
-#include <stdio.h>
+#include "main/glheader.h"
 #include "glthread.h"
 
 
diff --git a/src/mesa/glapi/glthread.h b/src/mesa/glapi/glthread.h
index 27ccd2e2e7b..ae2f79f37dd 100644
--- a/src/mesa/glapi/glthread.h
+++ b/src/mesa/glapi/glthread.h
@@ -116,9 +116,49 @@ typedef pthread_mutex_t _glthread_Mutex;
 #define _glthread_UNLOCK_MUTEX(name) \
    (void) pthread_mutex_unlock(&(name))
 
-#endif /* PTHREADS */
+typedef pthread_cond_t _glthread_Cond;
 
+#define _glthread_DECLARE_STATIC_COND(name) \
+   static _glthread_Cond name = PTHREAD_COND_INITIALIZER
 
+#define _glthread_INIT_COND(cond)			\
+   pthread_cond_init(&(cond), NULL)
+
+#define _glthread_DESTROY_COND(name) \
+   pthread_cond_destroy(&(name))
+
+#define _glthread_COND_WAIT(cond, mutex) \
+  pthread_cond_wait(&(cond), &(mutex))
+
+#define _glthread_COND_SIGNAL(cond) \
+  pthread_cond_signal(&(cond))
+
+#define _glthread_COND_BROADCAST(cond) \
+  pthread_cond_broadcast(&(cond))
+
+
+#else /* PTHREADS */
+
+typedef unsigned int _glthread_Cond;
+#define _glthread_DECLARE_STATIC_COND(name) \
+//  #warning Condition variables not implemented.
+
+#define _glthread_INIT_COND(cond)	    \
+  ASSERT(0);
+
+#define _glthread_DESTROY_COND(name) \
+  ASSERT(0);
+
+#define _glthread_COND_WAIT(cond, mutex) \
+  ASSERT(0);
+
+#define _glthread_COND_SIGNAL(cond) \
+  ASSERT(0);
+
+#define _glthread_COND_BROADCAST(cond) \
+  ASSERT(0);
+
+#endif
 
 
 /*
@@ -259,11 +299,11 @@ typedef benaphore _glthread_Mutex;
  * THREADS not defined
  */
 
-typedef int _glthread_TSD;
+typedef GLuint _glthread_TSD;
 
-typedef int _glthread_Thread;
+typedef GLuint _glthread_Thread;
 
-typedef int _glthread_Mutex;
+typedef GLuint _glthread_Mutex;
 
 #define _glthread_DECLARE_STATIC_MUTEX(name)  static _glthread_Mutex name = 0
 
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index 59c2debb229..5c8955d7c8d 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -133,6 +133,12 @@ _mesa_validate_DrawElements(GLcontext *ctx,
       /* use indices in the buffer object */
       GLuint indexBytes;
 
+      if (!ctx->Array.ElementArrayBufferObj->Size) {
+         _mesa_warning(ctx,
+                       "glDrawElements called with empty array elements buffer");
+         return GL_FALSE;
+      }
+
       if (type == GL_UNSIGNED_INT) {
          indexBytes = count * sizeof(GLuint);
       }
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index 59fe8e25def..016543da01f 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -38,6 +38,13 @@
 #include "bufferobj.h"
 
 
+#ifdef FEATURE_OES_mapbuffer
+#define DEFAULT_ACCESS GL_WRITE_ONLY;
+#else
+#define DEFAULT_ACCESS GL_READ_WRITE;
+#endif
+
+
 /**
  * Get the buffer object bound to the specified target in a GL context.
  *
@@ -255,7 +262,7 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
    obj->RefCount = 1;
    obj->Name = name;
    obj->Usage = GL_STATIC_DRAW_ARB;
-   obj->Access = GL_READ_WRITE_ARB;
+   obj->Access = DEFAULT_ACCESS;
 }
 
 
@@ -664,8 +671,8 @@ _mesa_map_drawpix_pbo(GLcontext *ctx,
  * \sa _mesa_unmap_bitmap_pbo
  */
 void
-_mesa_unmap_drapix_pbo(GLcontext *ctx,
-                       const struct gl_pixelstore_attrib *unpack)
+_mesa_unmap_drawpix_pbo(GLcontext *ctx,
+                        const struct gl_pixelstore_attrib *unpack)
 {
    if (unpack->BufferObj->Name) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
@@ -1065,7 +1072,7 @@ _mesa_UnmapBufferARB(GLenum target)
       status = ctx->Driver.UnmapBuffer( ctx, target, bufObj );
    }
 
-   bufObj->Access = GL_READ_WRITE_ARB; /* initial value, OK? */
+   bufObj->Access = DEFAULT_ACCESS;
    bufObj->Pointer = NULL;
 
    return status;
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 163fe241aab..3c08f0083cf 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -103,9 +103,8 @@ _mesa_map_drawpix_pbo(GLcontext *ctx,
                       const GLvoid *pixels);
 
 extern void
-_mesa_unmap_drapix_pbo(GLcontext *ctx,
-                       const struct gl_pixelstore_attrib *unpack);
-
+_mesa_unmap_drawpix_pbo(GLcontext *ctx,
+                        const struct gl_pixelstore_attrib *unpack);
 
 extern void *
 _mesa_map_readpix_pbo(GLcontext *ctx,
diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c
index 7877ee00451..818d068a12b 100644
--- a/src/mesa/main/buffers.c
+++ b/src/mesa/main/buffers.c
@@ -416,7 +416,7 @@ _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers,
       }
    }
 
-   ctx->NewState |= _NEW_COLOR;
+   ctx->NewState |= _NEW_BUFFERS;
 }
 
 
@@ -491,6 +491,7 @@ _mesa_ReadBuffer(GLenum buffer)
    /* OK, all error checking has been completed now */
 
    _mesa_readbuffer(ctx, buffer, srcBuffer);
+   ctx->NewState |= _NEW_BUFFERS;
 
    /*
     * Call device driver function.
diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
index cf1198cd0d5..c3feffda986 100644
--- a/src/mesa/main/config.h
+++ b/src/mesa/main/config.h
@@ -177,13 +177,11 @@
 /** For GL_ARB_fragment_program */
 /*@{*/
 #define MAX_FRAGMENT_PROGRAM_ADDRESS_REGS 0
-#define MAX_FRAGMENT_PROGRAM_ALU_INSTRUCTIONS 48
-#define MAX_FRAGMENT_PROGRAM_TEX_INSTRUCTIONS 24
-#define MAX_FRAGMENT_PROGRAM_TEX_INDIRECTIONS  4
 /*@}*/
 
 /** For any program target/extension */
 /*@{*/
+#define MAX_PROGRAM_INSTRUCTIONS  (16 * 1024)
 #define MAX_PROGRAM_LOCAL_PARAMS 256 /**< per-program constants (power of two) */
 #define MAX_PROGRAM_ENV_PARAMS 128
 #define MAX_PROGRAM_MATRICES 8
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 62c44733564..39c7fa037e0 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -145,9 +145,7 @@
 #include "glapi/glthread.h"
 #include "glapi/glapioffsets.h"
 #include "glapi/glapitable.h"
-#if FEATURE_NV_vertex_program || FEATURE_NV_fragment_program
 #include "shader/program.h"
-#endif
 #include "shader/shader_api.h"
 #if FEATURE_ATI_fragment_shader
 #include "shader/atifragshader.h"
@@ -186,9 +184,11 @@ GLfloat _mesa_ubyte_to_float_color_tab[256];
  * We have to finish any pending rendering.
  */
 void
-_mesa_notifySwapBuffers(__GLcontext *gc)
+_mesa_notifySwapBuffers(__GLcontext *ctx)
 {
-   FLUSH_VERTICES( gc, 0 );
+   if (ctx->Driver.Flush) {
+      ctx->Driver.Flush(ctx);
+   }
 }
 
 
@@ -611,6 +611,7 @@ delete_program_cb(GLuint id, void *data, void *userData)
    ctx->Driver.DeleteProgram(ctx, prog);
 }
 
+#if FEATURE_ATI_fragment_shader
 /**
  * Callback for deleting an ATI fragment shader object.
  * Called by _mesa_HashDeleteAll().
@@ -622,6 +623,7 @@ delete_fragshader_cb(GLuint id, void *data, void *userData)
    GLcontext *ctx = (GLcontext *) userData;
    _mesa_delete_ati_fragment_shader(ctx, shader);
 }
+#endif
 
 /**
  * Callback for deleting a buffer object.  Called by _mesa_HashDeleteAll().
@@ -712,7 +714,6 @@ delete_renderbuffer_cb(GLuint id, void *data, void *userData)
 }
 
 
-
 /**
  * Deallocate a shared state object and all children structures.
  *
@@ -818,11 +819,33 @@ _mesa_init_current(GLcontext *ctx)
 
 
 /**
- * Init vertex/fragment program native limits from logical limits.
+ * Init vertex/fragment program limits.
+ * Important: drivers should override these with actual limits.
  */
 static void
-init_natives(struct gl_program_constants *prog)
+init_program_limits(GLenum type, struct gl_program_constants *prog)
 {
+   prog->MaxInstructions = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxAluInstructions = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxTexInstructions = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxTexIndirections = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxTemps = MAX_PROGRAM_TEMPS;
+   prog->MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
+   prog->MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
+   prog->MaxUniformComponents = 4 * MAX_UNIFORMS;
+
+   if (type == GL_VERTEX_PROGRAM_ARB) {
+      prog->MaxParameters = MAX_NV_VERTEX_PROGRAM_PARAMS;
+      prog->MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
+      prog->MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
+   }
+   else {
+      prog->MaxParameters = MAX_NV_FRAGMENT_PROGRAM_PARAMS;
+      prog->MaxAttribs = MAX_NV_FRAGMENT_PROGRAM_INPUTS;
+      prog->MaxAddressRegs = MAX_FRAGMENT_PROGRAM_ADDRESS_REGS;
+   }
+
+   /* copy the above limits to init native limits */
    prog->MaxNativeInstructions = prog->MaxInstructions;
    prog->MaxNativeAluInstructions = prog->MaxAluInstructions;
    prog->MaxNativeTexInstructions = prog->MaxTexInstructions;
@@ -881,33 +904,10 @@ _mesa_init_constants(GLcontext *ctx)
    ctx->Const.MaxViewportWidth = MAX_WIDTH;
    ctx->Const.MaxViewportHeight = MAX_HEIGHT;
 #if FEATURE_ARB_vertex_program
-   ctx->Const.VertexProgram.MaxInstructions = MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS;
-   ctx->Const.VertexProgram.MaxAluInstructions = 0;
-   ctx->Const.VertexProgram.MaxTexInstructions = 0;
-   ctx->Const.VertexProgram.MaxTexIndirections = 0;
-   ctx->Const.VertexProgram.MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
-   ctx->Const.VertexProgram.MaxTemps = MAX_PROGRAM_TEMPS;
-   ctx->Const.VertexProgram.MaxParameters = MAX_NV_VERTEX_PROGRAM_PARAMS;
-   ctx->Const.VertexProgram.MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
-   ctx->Const.VertexProgram.MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
-   ctx->Const.VertexProgram.MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
-   ctx->Const.VertexProgram.MaxUniformComponents = 4 * MAX_UNIFORMS;
-   init_natives(&ctx->Const.VertexProgram);
+   init_program_limits(GL_VERTEX_PROGRAM_ARB, &ctx->Const.VertexProgram);
 #endif
-
 #if FEATURE_ARB_fragment_program
-   ctx->Const.FragmentProgram.MaxInstructions = MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS;
-   ctx->Const.FragmentProgram.MaxAluInstructions = MAX_FRAGMENT_PROGRAM_ALU_INSTRUCTIONS;
-   ctx->Const.FragmentProgram.MaxTexInstructions = MAX_FRAGMENT_PROGRAM_TEX_INSTRUCTIONS;
-   ctx->Const.FragmentProgram.MaxTexIndirections = MAX_FRAGMENT_PROGRAM_TEX_INDIRECTIONS;
-   ctx->Const.FragmentProgram.MaxAttribs = MAX_NV_FRAGMENT_PROGRAM_INPUTS;
-   ctx->Const.FragmentProgram.MaxTemps = MAX_PROGRAM_TEMPS;
-   ctx->Const.FragmentProgram.MaxParameters = MAX_NV_FRAGMENT_PROGRAM_PARAMS;
-   ctx->Const.FragmentProgram.MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
-   ctx->Const.FragmentProgram.MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
-   ctx->Const.FragmentProgram.MaxAddressRegs = MAX_FRAGMENT_PROGRAM_ADDRESS_REGS;
-   ctx->Const.FragmentProgram.MaxUniformComponents = 4 * MAX_UNIFORMS;
-   init_natives(&ctx->Const.FragmentProgram);
+   init_program_limits(GL_FRAGMENT_PROGRAM_ARB, &ctx->Const.FragmentProgram);
 #endif
    ctx->Const.MaxProgramMatrices = MAX_PROGRAM_MATRICES;
    ctx->Const.MaxProgramMatrixStackDepth = MAX_PROGRAM_MATRIX_STACK_DEPTH;
@@ -1066,6 +1066,7 @@ init_attrib_groups(GLcontext *ctx)
    /* Miscellaneous */
    ctx->NewState = _NEW_ALL;
    ctx->ErrorValue = (GLenum) GL_NO_ERROR;
+   ctx->varying_vp_inputs = ~0;
 
    return GL_TRUE;
 }
@@ -1165,7 +1166,7 @@ _mesa_initialize_context(GLcontext *ctx,
                          const struct dd_function_table *driverFunctions,
                          void *driverContext)
 {
-   ASSERT(driverContext);
+   /*ASSERT(driverContext);*/
    assert(driverFunctions->NewTextureObject);
    assert(driverFunctions->FreeTexImageData);
 
@@ -1269,7 +1270,7 @@ _mesa_create_context(const GLvisual *visual,
    GLcontext *ctx;
 
    ASSERT(visual);
-   ASSERT(driverContext);
+   /*ASSERT(driverContext);*/
 
    ctx = (GLcontext *) _mesa_calloc(sizeof(GLcontext));
    if (!ctx)
@@ -1317,7 +1318,9 @@ _mesa_free_context_data( GLcontext *ctx )
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, NULL);
 
+#if FEATURE_attrib_stack
    _mesa_free_attrib_data(ctx);
+#endif
    _mesa_free_lighting_data( ctx );
 #if FEATURE_evaluators
    _mesa_free_eval_data( ctx );
@@ -1634,7 +1637,22 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer,
           * or not bound to a user-created FBO.
           */
          if (!newCtx->DrawBuffer || newCtx->DrawBuffer->Name == 0) {
+            /* KW: merge conflict here, revisit. 
+             */
+            /* fix up the fb fields - these will end up wrong otherwise
+             * if the DRIdrawable changes, and everything relies on them.
+             * This is a bit messy (same as needed in _mesa_BindFramebufferEXT)
+             */
+            unsigned int i;
+            GLenum buffers[MAX_DRAW_BUFFERS];
+
             _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer);
+
+            for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) {
+               buffers[i] = newCtx->Color.DrawBuffer[i];
+            }
+
+            _mesa_drawbuffers(newCtx, newCtx->Const.MaxDrawBuffers, buffers, NULL);
          }
          if (!newCtx->ReadBuffer || newCtx->ReadBuffer->Name == 0) {
             _mesa_reference_framebuffer(&newCtx->ReadBuffer, readBuffer);
diff --git a/src/mesa/main/dlopen.c b/src/mesa/main/dlopen.c
index becef8173ef..8bc83c094fd 100644
--- a/src/mesa/main/dlopen.c
+++ b/src/mesa/main/dlopen.c
@@ -48,7 +48,7 @@ _mesa_dlopen(const char *libname, int flags)
    flags = RTLD_LAZY | RTLD_GLOBAL; /* Overriding flags at this time */
    return dlopen(libname, flags);
 #elif defined(__MINGW32__)
-   return LoadLibrary(libname);
+   return LoadLibraryA(libname);
 #else
    return NULL;
 #endif
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index ac97bc1ff9e..13cfa0e756e 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -153,7 +153,7 @@ _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height,
       return;
    }
 
-   if (!ctx->Current.RasterPosValid) {
+   if (!ctx->Current.RasterPosValid || width ==0 || height == 0) {
       return;
    }
 
@@ -218,7 +218,7 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
 
    if (ctx->RenderMode == GL_RENDER) {
       /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
-      const GLfloat epsilon = (const GLfloat)0.0001;
+      const GLfloat epsilon = 0.0001F;
       GLint x = IFLOOR(ctx->Current.RasterPos[0] + epsilon - xorig);
       GLint y = IFLOOR(ctx->Current.RasterPos[1] + epsilon - yorig);
 
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index 1b034105356..2b54fac6949 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -951,6 +951,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
          break;
 
       /* GL_MESA_program_debug */
+#if FEATURE_MESA_program_debug
       case GL_FRAGMENT_PROGRAM_CALLBACK_MESA:
          CHECK_EXTENSION(MESA_program_debug, cap);
          ctx->FragmentProgram.CallbackEnabled = state;
@@ -959,6 +960,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
          CHECK_EXTENSION(MESA_program_debug, cap);
          ctx->VertexProgram.CallbackEnabled = state;
          break;
+#endif
 
 #if FEATURE_ATI_fragment_shader
       case GL_FRAGMENT_SHADER_ATI:
@@ -1403,12 +1405,15 @@ _mesa_IsEnabled( GLenum cap )
          return ctx->Depth.BoundsTest;
 
       /* GL_MESA_program_debug */
+#if FEATURE_MESA_program_debug
       case GL_FRAGMENT_PROGRAM_CALLBACK_MESA:
          CHECK_EXTENSION(MESA_program_debug);
          return ctx->FragmentProgram.CallbackEnabled;
       case GL_VERTEX_PROGRAM_CALLBACK_MESA:
          CHECK_EXTENSION(MESA_program_debug);
          return ctx->VertexProgram.CallbackEnabled;
+#endif
+
 #if FEATURE_ATI_fragment_shader
       case GL_FRAGMENT_SHADER_ATI:
 	 CHECK_EXTENSION(ATI_fragment_shader);
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index e38e72dbbe0..1b31e283cd3 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -87,7 +87,7 @@ static const struct {
    { OFF, "GL_EXT_blend_logic_op",             F(EXT_blend_logic_op) },
    { OFF, "GL_EXT_blend_minmax",               F(EXT_blend_minmax) },
    { OFF, "GL_EXT_blend_subtract",             F(EXT_blend_subtract) },
-   { ON,  "GL_EXT_clip_volume_hint",           F(EXT_clip_volume_hint) },
+   { OFF, "GL_EXT_clip_volume_hint",           F(EXT_clip_volume_hint) },
    { OFF, "GL_EXT_cull_vertex",                F(EXT_cull_vertex) },
    { ON,  "GL_EXT_compiled_vertex_array",      F(EXT_compiled_vertex_array) },
    { OFF, "GL_EXT_convolution",                F(EXT_convolution) },
diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index f5e17c76750..c3cdc110379 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -193,7 +193,7 @@ _mesa_remove_attachment(GLcontext *ctx, struct gl_renderbuffer_attachment *att)
    if (att->Type == GL_TEXTURE) {
       ASSERT(att->Texture);
       if (ctx->Driver.FinishRenderTexture) {
-         /* tell driver we're done rendering to this texobj */
+         /* tell driver that we're done rendering to this texture. */
          ctx->Driver.FinishRenderTexture(ctx, att);
       }
       _mesa_reference_texobj(&att->Texture, NULL); /* unbind */
diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c
index cc1b44de642..d70b78f2586 100644
--- a/src/mesa/main/ffvertex_prog.c
+++ b/src/mesa/main/ffvertex_prog.c
@@ -47,17 +47,17 @@
 
 
 struct state_key {
+   unsigned light_color_material_mask:12;
+   unsigned light_material_mask:12;
    unsigned light_global_enabled:1;
    unsigned light_local_viewer:1;
    unsigned light_twoside:1;
    unsigned light_color_material:1;
-   unsigned light_color_material_mask:12;
-   unsigned light_material_mask:12;
    unsigned material_shininess_is_zero:1;
-
    unsigned need_eye_coords:1;
    unsigned normalize:1;
    unsigned rescale_normals:1;
+
    unsigned fog_source_is_depth:1;
    unsigned tnl_do_vertex_fog:1;
    unsigned separate_specular:1;
@@ -67,6 +67,8 @@ struct state_key {
    unsigned texture_enabled_global:1;
    unsigned fragprog_inputs_read:12;
 
+   unsigned varying_vp_inputs;
+
    struct {
       unsigned light_enabled:1;
       unsigned light_eyepos3_is_zero:1;
@@ -194,6 +196,7 @@ static void make_state_key( GLcontext *ctx, struct state_key *key )
    key->need_eye_coords = ctx->_NeedEyeCoords;
 
    key->fragprog_inputs_read = fp->Base.InputsRead;
+   key->varying_vp_inputs = ctx->varying_vp_inputs;
 
    if (ctx->RenderMode == GL_FEEDBACK) {
       /* make sure the vertprog emits color and tex0 */
@@ -447,20 +450,52 @@ static void release_temp( struct tnl_program *p, struct ureg reg )
    }
 }
 
-
 static void release_temps( struct tnl_program *p )
 {
    p->temp_in_use = p->temp_reserved;
 }
 
 
+static struct ureg register_param5(struct tnl_program *p, 
+				   GLint s0,
+				   GLint s1,
+				   GLint s2,
+				   GLint s3,
+                                   GLint s4)
+{
+   gl_state_index tokens[STATE_LENGTH];
+   GLint idx;
+   tokens[0] = s0;
+   tokens[1] = s1;
+   tokens[2] = s2;
+   tokens[3] = s3;
+   tokens[4] = s4;
+   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
+   return make_ureg(PROGRAM_STATE_VAR, idx);
+}
+
+
+#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
+#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
+#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
+#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
+
+
+
 /**
  * \param input  one of VERT_ATTRIB_x tokens.
  */
 static struct ureg register_input( struct tnl_program *p, GLuint input )
 {
-   p->program->Base.InputsRead |= (1<<input);
-   return make_ureg(PROGRAM_INPUT, input);
+   /* Material attribs are passed here as inputs >= 32
+    */
+   if (input >= 32 || (p->state->varying_vp_inputs & (1<<input))) {
+      p->program->Base.InputsRead |= (1<<input);
+      return make_ureg(PROGRAM_INPUT, input);
+   }
+   else {
+      return register_param3( p, STATE_INTERNAL, STATE_CURRENT_ATTRIB, input );
+   }
 }
 
 
@@ -493,7 +528,6 @@ static struct ureg register_const4f( struct tnl_program *p,
    return make_ureg(PROGRAM_CONSTANT, idx);
 }
 
-
 #define register_const1f(p, s0)         register_const4f(p, s0, 0, 0, 1)
 #define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
 #define register_const2f(p, s0, s1)     register_const4f(p, s0, s1, 0, 1)
@@ -513,32 +547,6 @@ static struct ureg get_identity_param( struct tnl_program *p )
    return p->identity;
 }
 
-
-static struct ureg register_param5(struct tnl_program *p, 
-				   GLint s0,
-				   GLint s1,
-				   GLint s2,
-				   GLint s3,
-                                   GLint s4)
-{
-   gl_state_index tokens[STATE_LENGTH];
-   GLint idx;
-   tokens[0] = s0;
-   tokens[1] = s1;
-   tokens[2] = s2;
-   tokens[3] = s3;
-   tokens[4] = s4;
-   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
-   return make_ureg(PROGRAM_STATE_VAR, idx);
-}
-
-
-#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
-#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
-#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
-#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
-
-
 static void register_matrix_param5( struct tnl_program *p,
 				    GLint s0, /* modelview, projection, etc */
 				    GLint s1, /* texture matrix number */
diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h
index 626806d35f6..5657976711e 100644
--- a/src/mesa/main/glheader.h
+++ b/src/mesa/main/glheader.h
@@ -64,22 +64,33 @@
 #include <stdarg.h>
 
 
-/* Get typedefs for uintptr_t and friends */
-#if defined(__MINGW32__) || defined(__NetBSD__)
-#  include <stdint.h>
-#elif defined(_WIN32)
-#  include <BaseTsd.h>
-#  if _MSC_VER == 1200
-     typedef UINT_PTR uintptr_t;
-#  endif 
-#elif defined(__INTERIX)
-/* Interix 3.x has a gcc that shadows this. */
-#  ifndef _UINTPTR_T_DEFINED
-     typedef unsigned long uintptr_t;
-#  define _UINTPTR_T_DEFINED
+/* Get standard integer types */
+#if defined(_MSC_VER)
+
+   typedef __int8             int8_t;
+   typedef unsigned __int8    uint8_t;
+   typedef __int16            int16_t;
+   typedef unsigned __int16   uint16_t;
+#  ifndef __eglplatform_h_
+     typedef __int32            int32_t;
+#  endif
+   typedef unsigned __int32   uint32_t;
+   typedef __int64            int64_t;
+   typedef unsigned __int64   uint64_t;
+
+#  if defined(_WIN64)
+     typedef __int64            intptr_t;
+     typedef unsigned __int64   uintptr_t;
+#  else
+     typedef __int32            intptr_t;
+     typedef unsigned __int32   uintptr_t;
 #  endif
+
+#  define INT64_C(__val) __val##i64
+#  define UINT64_C(__val) __val##ui64
+
 #else
-#  include <inttypes.h>
+#  include <stdint.h>
 #endif
 
 
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 0c2c894eef0..cb04594c1fd 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -557,7 +557,7 @@ _mesa_pow(double x, double y)
  * Find the first bit set in a word.
  */
 int
-_mesa_ffs(int i)
+_mesa_ffs(int32_t i)
 {
 #if (defined(_WIN32) ) || defined(__IBMC__) || defined(__IBMCPP__)
    register int bit = 0;
@@ -594,11 +594,7 @@ _mesa_ffs(int i)
  *          if no bits set.
  */
 int
-#ifdef __MINGW32__
-_mesa_ffsll(long val)
-#else
-_mesa_ffsll(long long val)
-#endif
+_mesa_ffsll(int64_t val)
 {
 #ifdef ffsll
    return ffsll(val);
@@ -607,11 +603,11 @@ _mesa_ffsll(long long val)
 
    assert(sizeof(val) == 8);
 
-   bit = _mesa_ffs(val);
+   bit = _mesa_ffs((int32_t)val);
    if (bit != 0)
       return bit;
 
-   bit = _mesa_ffs(val >> 32);
+   bit = _mesa_ffs((int32_t)(val >> 32));
    if (bit != 0)
       return 32 + bit;
 
@@ -950,12 +946,10 @@ _mesa_snprintf( char *str, size_t size, const char *fmt, ... )
 void
 _mesa_printf( const char *fmtString, ... )
 {
-   char s[MAXSTRING];
    va_list args;
    va_start( args, fmtString );  
-   vsnprintf(s, MAXSTRING, fmtString, args);
+   vfprintf(stderr, fmtString, args);
    va_end( args );
-   fprintf(stderr, "%s", s);
 }
 
 /** Wrapper around fprintf(), using vsprintf() for the formatting. */
diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h
index ec270a0cb30..a421eb5e755 100644
--- a/src/mesa/main/imports.h
+++ b/src/mesa/main/imports.h
@@ -708,14 +708,10 @@ extern double
 _mesa_pow(double x, double y);
 
 extern int
-_mesa_ffs(int i);
+_mesa_ffs(int32_t i);
 
 extern int
-#ifdef __MINGW32__
-_mesa_ffsll(long i);
-#else
-_mesa_ffsll(long long i);
-#endif
+_mesa_ffsll(int64_t i);
 
 extern unsigned int
 _mesa_bitcount(unsigned int n);
diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c
index ce50224d30a..a15c1f0be05 100644
--- a/src/mesa/main/light.c
+++ b/src/mesa/main/light.c
@@ -1368,6 +1368,7 @@ _mesa_init_lighting( GLcontext *ctx )
    /* Miscellaneous */
    ctx->Light._NeedEyeCoords = GL_FALSE;
    ctx->_NeedEyeCoords = GL_FALSE;
+   ctx->_ForceEyeCoords = GL_FALSE;
    ctx->_ModelViewInvScale = 1.0;
 }
 
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 92bce76ad51..be982afe39c 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -125,6 +125,8 @@ struct gl_program_cache;
 struct gl_texture_format;
 struct gl_texture_image;
 struct gl_texture_object;
+struct st_context;
+struct pipe_surface;
 typedef struct __GLcontextRec GLcontext;
 typedef struct __GLcontextModesRec GLvisual;
 typedef struct gl_framebuffer GLframebuffer;
@@ -1579,7 +1581,6 @@ struct gl_texture_attrib
 
    struct gl_texture_unit Unit[MAX_TEXTURE_UNITS];
 
-   /** Proxy texture objects */
    struct gl_texture_object *ProxyTex[NUM_TEXTURE_TARGETS];
 
    /** GL_EXT_shared_texture_palette */
@@ -2003,6 +2004,8 @@ struct gl_vertex_program_state
    GLboolean CallbackEnabled;
    GLuint CurrentPosition;
 #endif
+
+   GLboolean _Overriden;
 };
 
 
@@ -2760,6 +2763,7 @@ struct gl_matrix_stack
 #define _NEW_MULTISAMPLE        0x2000000  /**< __GLcontextRec::Multisample */
 #define _NEW_TRACK_MATRIX       0x4000000  /**< __GLcontextRec::VertexProgram */
 #define _NEW_PROGRAM            0x8000000  /**< __GLcontextRec::VertexProgram */
+#define _NEW_CURRENT_ATTRIB     0x10000000  /**< __GLcontextRec::Current */
 #define _NEW_ALL ~0
 /*@}*/
 
@@ -3094,6 +3098,8 @@ struct __GLcontextRec
    GLenum RenderMode;        /**< either GL_RENDER, GL_SELECT, GL_FEEDBACK */
    GLbitfield NewState;      /**< bitwise-or of _NEW_* flags */
 
+   GLbitfield varying_vp_inputs;  /**< mask of VERT_BIT_* flags */
+
    /** \name Derived state */
    /*@{*/
    /** Bitwise-or of DD_* flags.  Note that this bitfield may be used before
@@ -3139,7 +3145,7 @@ struct __GLcontextRec
    void *swsetup_context;
    void *swtnl_context;
    void *swtnl_im;
-   void *acache_context;
+   struct st_context *st;
    void *aelt_context;
    /*@}*/
 };
diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c
index 6047ebac4af..38be8266e05 100644
--- a/src/mesa/main/renderbuffer.c
+++ b/src/mesa/main/renderbuffer.c
@@ -1199,7 +1199,8 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
 
    if (width > 0 && height > 0) {
       /* allocate new buffer storage */
-      rb->Data = _mesa_malloc(width * height * pixelSize);
+      rb->Data = malloc(width * height * pixelSize);
+
       if (rb->Data == NULL) {
          rb->Width = 0;
          rb->Height = 0;
diff --git a/src/mesa/main/sources b/src/mesa/main/sources
index 1b43e7e00f0..4e58702c746 100644
--- a/src/mesa/main/sources
+++ b/src/mesa/main/sources
@@ -79,7 +79,6 @@ vsnprintf.c
 MESA_MAIN_HEADERS = \
 accum.h \
 api_arrayelt.h \
-api_eval.h \
 api_exec.h \
 api_loopback.h \
 api_noop.h \
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index 3cf70d6a3ca..0a39279bff3 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -448,6 +448,9 @@ _mesa_update_state_locked( GLcontext *ctx )
    GLbitfield new_state = ctx->NewState;
    GLbitfield prog_flags = _NEW_PROGRAM;
 
+   if (new_state == _NEW_CURRENT_ATTRIB) 
+      goto out;
+
    if (MESA_VERBOSE & VERBOSE_STATE)
       _mesa_print_state("_mesa_update_state", new_state);
 
@@ -475,7 +478,7 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & (_NEW_PROGRAM|_NEW_TEXTURE|_NEW_TEXTURE_MATRIX))
       _mesa_update_texture( ctx, new_state );
 
-   if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL))
+   if (new_state & _NEW_BUFFERS)
       _mesa_update_framebuffer(ctx);
 
    if (new_state & (_NEW_SCISSOR | _NEW_BUFFERS | _NEW_VIEWPORT))
@@ -540,6 +543,7 @@ _mesa_update_state_locked( GLcontext *ctx )
     * Set ctx->NewState to zero to avoid recursion if
     * Driver.UpdateState() has to call FLUSH_VERTICES().  (fixed?)
     */
+ out:
    new_state = ctx->NewState;
    ctx->NewState = 0;
    ctx->Driver.UpdateState(ctx, new_state);
@@ -556,3 +560,59 @@ _mesa_update_state( GLcontext *ctx )
    _mesa_update_state_locked(ctx);
    _mesa_unlock_context_textures(ctx);
 }
+
+
+
+
+/**
+ * Want to figure out which fragment program inputs are actually
+ * constant/current values from ctx->Current.  These should be
+ * referenced as a tracked state variable rather than a fragment
+ * program input, to save the overhead of putting a constant value in
+ * every submitted vertex, transferring it to hardware, interpolating
+ * it across the triangle, etc...
+ *
+ * When there is a VP bound, just use vp->outputs.  But when we're
+ * generating vp from fixed function state, basically want to
+ * calculate:
+ *
+ * vp_out_2_fp_in( vp_in_2_vp_out( varying_inputs ) | 
+ *                 potential_vp_outputs )
+ *
+ * Where potential_vp_outputs is calculated by looking at enabled
+ * texgen, etc.
+ * 
+ * The generated fragment program should then only declare inputs that
+ * may vary or otherwise differ from the ctx->Current values.
+ * Otherwise, the fp should track them as state values instead.
+ */
+void
+_mesa_set_varying_vp_inputs( GLcontext *ctx,
+                             GLbitfield varying_inputs )
+{
+   if (ctx->varying_vp_inputs != varying_inputs) {
+      ctx->varying_vp_inputs = varying_inputs;
+      ctx->NewState |= _NEW_ARRAY;
+      /*_mesa_printf("%s %x\n", __FUNCTION__, varying_inputs);*/
+   }
+}
+
+
+/**
+ * Used by drivers to tell core Mesa that the driver is going to
+ * install/ use its own vertex program.  In particular, this will
+ * prevent generated fragment programs from using state vars instead
+ * of ordinary varyings/inputs.
+ */
+void
+_mesa_set_vp_override(GLcontext *ctx, GLboolean flag)
+{
+   if (ctx->VertexProgram._Overriden != flag) {
+      ctx->VertexProgram._Overriden = flag;
+
+      /* Set one of the bits which will trigger fragment program
+       * regeneration:
+       */
+      ctx->NewState |= _NEW_ARRAY; 
+   }
+}
diff --git a/src/mesa/main/state.h b/src/mesa/main/state.h
index bb7cb8f32a3..29db08a0b95 100644
--- a/src/mesa/main/state.h
+++ b/src/mesa/main/state.h
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -29,13 +29,21 @@
 #include "mtypes.h"
 
 extern void
-_mesa_update_state( GLcontext *ctx );
+_mesa_update_state(GLcontext *ctx);
 
 /* As above but can only be called between _mesa_lock_context_textures() and 
  * _mesa_unlock_context_textures().
  */
 extern void
-_mesa_update_state_locked( GLcontext *ctx );
+_mesa_update_state_locked(GLcontext *ctx);
+
+
+extern void
+_mesa_set_varying_vp_inputs(GLcontext *ctx, GLbitfield varying_inputs);
+
+
+extern void
+_mesa_set_vp_override(GLcontext *ctx, GLboolean flag);
 
 
 #endif
diff --git a/src/mesa/main/texcompress.c b/src/mesa/main/texcompress.c
index 09eba151c06..c1b8c7675a2 100644
--- a/src/mesa/main/texcompress.c
+++ b/src/mesa/main/texcompress.c
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -135,8 +136,10 @@ _mesa_compressed_texture_size( GLcontext *ctx,
 
    ASSERT(depth == 1);
    (void) depth;
+   (void) size;
 
    switch (mesaFormat) {
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
       /* round up width to next multiple of 8, height to next multiple of 4 */
@@ -148,8 +151,14 @@ _mesa_compressed_texture_size( GLcontext *ctx,
        * take 16 bytes.
        */
       return size;
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+#endif
       /* round up width, height to next multiple of 4 */
       width = (width + 3) & ~3;
       height = (height + 3) & ~3;
@@ -161,6 +170,10 @@ _mesa_compressed_texture_size( GLcontext *ctx,
       return size;
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       /* round up width, height to next multiple of 4 */
       width = (width + 3) & ~3;
       height = (height + 3) & ~3;
@@ -170,6 +183,7 @@ _mesa_compressed_texture_size( GLcontext *ctx,
        * take 16 bytes.
        */
       return size;
+#endif
    default:
       _mesa_problem(ctx, "bad mesaFormat in _mesa_compressed_texture_size");
       return 0;
@@ -194,12 +208,15 @@ _mesa_compressed_texture_size_glenum(GLcontext *ctx,
    GLuint mesaFormat;
 
    switch (glformat) {
+#if FEATURE_texture_fxt1
    case GL_COMPRESSED_RGB_FXT1_3DFX:
       mesaFormat = MESA_FORMAT_RGB_FXT1;
       break;
    case GL_COMPRESSED_RGBA_FXT1_3DFX:
       mesaFormat = MESA_FORMAT_RGBA_FXT1;
       break;
+#endif
+#if FEATURE_texture_s3tc
    case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
    case GL_RGB_S3TC:
       mesaFormat = MESA_FORMAT_RGB_DXT1;
@@ -216,6 +233,21 @@ _mesa_compressed_texture_size_glenum(GLcontext *ctx,
    case GL_RGBA4_S3TC:
       mesaFormat = MESA_FORMAT_RGBA_DXT5;
       break;
+#if FEATURE_EXT_texture_sRGB
+   case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+      mesaFormat = MESA_FORMAT_SRGB_DXT1;
+      break;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+      mesaFormat = MESA_FORMAT_SRGBA_DXT1;
+      break;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+      mesaFormat = MESA_FORMAT_SRGBA_DXT3;
+      break;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
+      mesaFormat = MESA_FORMAT_SRGBA_DXT5;
+      break;
+#endif
+#endif
    default:
       return 0;
    }
@@ -237,18 +269,30 @@ _mesa_compressed_row_stride(GLuint mesaFormat, GLsizei width)
    GLint stride;
 
    switch (mesaFormat) {
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
       stride = ((width + 7) / 8) * 16; /* 16 bytes per 8x4 tile */
       break;
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+#endif
       stride = ((width + 3) / 4) * 8; /* 8 bytes per 4x4 tile */
       break;
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       stride = ((width + 3) / 4) * 16; /* 16 bytes per 4x4 tile */
       break;
+#endif
    default:
       _mesa_problem(NULL, "bad mesaFormat in _mesa_compressed_row_stride");
       return 0;
@@ -285,18 +329,30 @@ _mesa_compressed_image_address(GLint col, GLint row, GLint img,
     */
 
    switch (mesaFormat) {
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
       addr = (GLubyte *) image + 16 * (((width + 7) / 8) * (row / 4) + col / 8);
       break;
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+#endif
       addr = (GLubyte *) image + 8 * (((width + 3) / 4) * (row / 4) + col / 4);
       break;
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       addr = (GLubyte *) image + 16 * (((width + 3) / 4) * (row / 4) + col / 4);
       break;
+#endif
    default:
       _mesa_problem(NULL, "bad mesaFormat in _mesa_compressed_image_address");
       addr = NULL;
diff --git a/src/mesa/main/texcompress_fxt1.c b/src/mesa/main/texcompress_fxt1.c
index 45f344b0c52..fc151605c9e 100644
--- a/src/mesa/main/texcompress_fxt1.c
+++ b/src/mesa/main/texcompress_fxt1.c
@@ -298,17 +298,17 @@ const struct gl_texture_format _mesa_texformat_rgba_fxt1 = {
 /*
  * Define a 64-bit unsigned integer type and macros
  */
-#ifdef GL_EXT_timer_query  /* this extensions defines the GLuint64EXT type */
+#if 1
 
 #define FX64_NATIVE 1
 
-typedef GLuint64EXT Fx64;
+typedef uint64_t Fx64;
 
 #define FX64_MOV32(a, b) a = b
 #define FX64_OR32(a, b)  a |= b
 #define FX64_SHL(a, c)   a <<= c
 
-#else  /* !GL_EXT_timer_query */
+#else
 
 #define FX64_NATIVE 0
 
@@ -330,7 +330,7 @@ typedef struct {
        }                                               \
    } while (0)
 
-#endif  /* !GL_EXT_timer_query */
+#endif
 
 
 #define F(i) (GLfloat)1 /* can be used to obtain an oblong metric: 0.30 / 0.59 / 0.11 */
diff --git a/src/mesa/main/texcompress_s3tc.c b/src/mesa/main/texcompress_s3tc.c
index a19c161b2c0..d17e18da6b5 100644
--- a/src/mesa/main/texcompress_s3tc.c
+++ b/src/mesa/main/texcompress_s3tc.c
@@ -3,6 +3,7 @@
  * Version:  6.5.3
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -53,6 +54,34 @@
 #define DXTN_LIBNAME "libtxc_dxtn.so"
 #endif
 
+#if FEATURE_EXT_texture_sRGB
+/**
+ * Convert an 8-bit sRGB value from non-linear space to a
+ * linear RGB value in [0, 1].
+ * Implemented with a 256-entry lookup table.
+ */
+static INLINE GLfloat
+nonlinear_to_linear(GLubyte cs8)
+{
+   static GLfloat table[256];
+   static GLboolean tableReady = GL_FALSE;
+   if (!tableReady) {
+      /* compute lookup table now */
+      GLuint i;
+      for (i = 0; i < 256; i++) {
+         const GLfloat cs = UBYTE_TO_FLOAT(i);
+         if (cs <= 0.04045) {
+            table[i] = cs / 12.92f;
+         }
+         else {
+            table[i] = (GLfloat) _mesa_pow((cs + 0.055) / 1.055, 2.4);
+         }
+      }
+      tableReady = GL_TRUE;
+   }
+   return table[cs8];
+}
+#endif /* FEATURE_EXT_texture_sRGB */
 
 typedef void (*dxtFetchTexelFuncExt)( GLint srcRowstride, GLubyte *pixdata, GLint col, GLint row, GLvoid *texelOut );
 
@@ -483,6 +512,59 @@ fetch_texel_2d_f_rgba_dxt5( const struct gl_texture_image *texImage,
    texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
 }
 
+#if FEATURE_EXT_texture_sRGB
+static void
+fetch_texel_2d_f_srgb_dxt1( const struct gl_texture_image *texImage,
+                            GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgb_dxt1(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+
+static void
+fetch_texel_2d_f_srgba_dxt1( const struct gl_texture_image *texImage,
+                             GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgba_dxt1(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+
+static void
+fetch_texel_2d_f_srgba_dxt3( const struct gl_texture_image *texImage,
+                             GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgba_dxt3(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+
+static void
+fetch_texel_2d_f_srgba_dxt5( const struct gl_texture_image *texImage,
+                             GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgba_dxt5(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+#endif
 
 const struct gl_texture_format _mesa_texformat_rgb_dxt1 = {
    MESA_FORMAT_RGB_DXT1,		/* MesaFormat */
@@ -508,6 +590,78 @@ const struct gl_texture_format _mesa_texformat_rgb_dxt1 = {
    NULL					/* StoreTexel */
 };
 
+const struct gl_texture_format _mesa_texformat_rgba_dxt1 = {
+   MESA_FORMAT_RGBA_DXT1,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4, /*approx*/			/* RedBits */
+   4, /*approx*/			/* GreenBits */
+   4, /*approx*/			/* BlueBits */
+   1, /*approx*/			/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   0,					/* TexelBytes */
+   texstore_rgba_dxt1,			/* StoreTexImageFunc */
+   NULL, /*impossible*/ 		/* FetchTexel1D */
+   fetch_texel_2d_rgba_dxt1, 		/* FetchTexel2D */
+   NULL, /*impossible*/ 		/* FetchTexel3D */
+   NULL, /*impossible*/ 		/* FetchTexel1Df */
+   fetch_texel_2d_f_rgba_dxt1, 		/* FetchTexel2Df */
+   NULL, /*impossible*/ 		/* FetchTexel3Df */
+   NULL					/* StoreTexel */
+};
+
+const struct gl_texture_format _mesa_texformat_rgba_dxt3 = {
+   MESA_FORMAT_RGBA_DXT3,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4, /*approx*/			/* RedBits */
+   4, /*approx*/			/* GreenBits */
+   4, /*approx*/			/* BlueBits */
+   4, /*approx*/			/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   0,					/* TexelBytes */
+   texstore_rgba_dxt3,			/* StoreTexImageFunc */
+   NULL, /*impossible*/ 		/* FetchTexel1D */
+   fetch_texel_2d_rgba_dxt3, 		/* FetchTexel2D */
+   NULL, /*impossible*/ 		/* FetchTexel3D */
+   NULL, /*impossible*/ 		/* FetchTexel1Df */
+   fetch_texel_2d_f_rgba_dxt3, 		/* FetchTexel2Df */
+   NULL, /*impossible*/ 		/* FetchTexel3Df */
+   NULL					/* StoreTexel */
+};
+
+const struct gl_texture_format _mesa_texformat_rgba_dxt5 = {
+   MESA_FORMAT_RGBA_DXT5,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4,/*approx*/				/* RedBits */
+   4,/*approx*/				/* GreenBits */
+   4,/*approx*/				/* BlueBits */
+   4,/*approx*/				/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   0,					/* TexelBytes */
+   texstore_rgba_dxt5,			/* StoreTexImageFunc */
+   NULL, /*impossible*/ 		/* FetchTexel1D */
+   fetch_texel_2d_rgba_dxt5, 		/* FetchTexel2D */
+   NULL, /*impossible*/ 		/* FetchTexel3D */
+   NULL, /*impossible*/ 		/* FetchTexel1Df */
+   fetch_texel_2d_f_rgba_dxt5, 		/* FetchTexel2Df */
+   NULL, /*impossible*/ 		/* FetchTexel3Df */
+   NULL					/* StoreTexel */
+};
+
 #if FEATURE_EXT_texture_sRGB
 const struct gl_texture_format _mesa_texformat_srgb_dxt1 = {
    MESA_FORMAT_SRGB_DXT1,		/* MesaFormat */
@@ -525,17 +679,16 @@ const struct gl_texture_format _mesa_texformat_srgb_dxt1 = {
    0,					/* TexelBytes */
    texstore_rgb_dxt1,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgb_dxt1, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgb_dxt1, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgb_dxt1, 		/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
-#endif
 
-const struct gl_texture_format _mesa_texformat_rgba_dxt1 = {
-   MESA_FORMAT_RGBA_DXT1,		/* MesaFormat */
+const struct gl_texture_format _mesa_texformat_srgba_dxt1 = {
+   MESA_FORMAT_SRGBA_DXT1,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
    GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
    4, /*approx*/			/* RedBits */
@@ -550,16 +703,16 @@ const struct gl_texture_format _mesa_texformat_rgba_dxt1 = {
    0,					/* TexelBytes */
    texstore_rgba_dxt1,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgba_dxt1, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgba_dxt1, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgba_dxt1, 	/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
 
-const struct gl_texture_format _mesa_texformat_rgba_dxt3 = {
-   MESA_FORMAT_RGBA_DXT3,		/* MesaFormat */
+const struct gl_texture_format _mesa_texformat_srgba_dxt3 = {
+   MESA_FORMAT_SRGBA_DXT3,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
    GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
    4, /*approx*/			/* RedBits */
@@ -574,16 +727,16 @@ const struct gl_texture_format _mesa_texformat_rgba_dxt3 = {
    0,					/* TexelBytes */
    texstore_rgba_dxt3,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgba_dxt3, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgba_dxt3, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgba_dxt3, 	/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
 
-const struct gl_texture_format _mesa_texformat_rgba_dxt5 = {
-   MESA_FORMAT_RGBA_DXT5,		/* MesaFormat */
+const struct gl_texture_format _mesa_texformat_srgba_dxt5 = {
+   MESA_FORMAT_SRGBA_DXT5,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
    GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
    4,/*approx*/				/* RedBits */
@@ -598,10 +751,11 @@ const struct gl_texture_format _mesa_texformat_rgba_dxt5 = {
    0,					/* TexelBytes */
    texstore_rgba_dxt5,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgba_dxt5, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgba_dxt5, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgba_dxt5, 	/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
+#endif
diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c
index d1f03405c1d..af51a206a56 100644
--- a/src/mesa/main/texenvprogram.c
+++ b/src/mesa/main/texenvprogram.c
@@ -67,11 +67,9 @@ struct texenvprog_cache_item
 
 
 /**
- * This MAX is probably a bit generous, but that's OK.  There can be
- * up to four instructions per texture unit (TEX + 3 for combine),
- * then there's fog and specular add.
+ * Up to nine instructions per tex unit, plus fog, specular color.
  */
-#define MAX_INSTRUCTIONS ((MAX_TEXTURE_COORD_UNITS * 4) + 12)
+#define MAX_INSTRUCTIONS ((MAX_TEXTURE_COORD_UNITS * 9) + 12)
 
 #define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM)
 
@@ -81,10 +79,12 @@ struct mode_opt {
 };
 
 struct state_key {
-   GLbitfield enabled_units;
+   GLuint nr_enabled_units:8;
+   GLuint enabled_units:8;
    GLuint separate_specular:1;
    GLuint fog_enabled:1;
    GLuint fog_mode:2;
+   GLuint inputs_available:12;
 
    struct {
       GLuint enabled:1;
@@ -232,6 +232,8 @@ static GLuint translate_mode( GLenum envMode, GLenum mode )
 #define TEXTURE_UNKNOWN_INDEX 7
 static GLuint translate_tex_src_bit( GLbitfield bit )
 {
+   /* make sure number of switch cases is correct */
+   assert(NUM_TEXTURE_TARGETS == 7);
    switch (bit) {
    case TEXTURE_1D_BIT:   return TEXTURE_1D_INDEX;
    case TEXTURE_2D_BIT:   return TEXTURE_2D_INDEX;
@@ -246,6 +248,102 @@ static GLuint translate_tex_src_bit( GLbitfield bit )
    }
 }
 
+#define VERT_BIT_TEX_ANY    (0xff << VERT_ATTRIB_TEX0)
+#define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
+
+/**
+ * Identify all possible varying inputs.  The fragment program will
+ * never reference non-varying inputs, but will track them via state
+ * constants instead.
+ *
+ * This function figures out all the inputs that the fragment program
+ * has access to.  The bitmask is later reduced to just those which
+ * are actually referenced.
+ */
+static GLbitfield get_fp_input_mask( GLcontext *ctx )
+{
+   const GLboolean vertexShader = (ctx->Shader.CurrentProgram &&
+                                   ctx->Shader.CurrentProgram->VertexProgram);
+   const GLboolean vertexProgram = ctx->VertexProgram._Enabled;
+   GLbitfield fp_inputs = 0x0;
+
+   if (ctx->VertexProgram._Overriden) {
+      /* Somebody's messing with the vertex program and we don't have
+       * a clue what's happening.  Assume that it could be producing
+       * all possible outputs.
+       */
+      fp_inputs = ~0;
+   }
+   else if (ctx->RenderMode == GL_FEEDBACK) {
+      fp_inputs = (FRAG_BIT_COL0 | FRAG_BIT_TEX0);
+   }
+   else if (!(vertexProgram || vertexShader) ||
+            !ctx->VertexProgram._Current) {
+      /* Fixed function vertex logic */
+      GLbitfield varying_inputs = ctx->varying_vp_inputs;
+
+      /* These get generated in the setup routine regardless of the
+       * vertex program:
+       */
+      if (ctx->Point.PointSprite)
+         varying_inputs |= FRAG_BITS_TEX_ANY;
+
+      /* First look at what values may be computed by the generated
+       * vertex program:
+       */
+      if (ctx->Light.Enabled) {
+         fp_inputs |= FRAG_BIT_COL0;
+
+         if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+            fp_inputs |= FRAG_BIT_COL1;
+      }
+
+      fp_inputs |= (ctx->Texture._TexGenEnabled |
+                    ctx->Texture._TexMatEnabled) << FRAG_ATTRIB_TEX0;
+
+      /* Then look at what might be varying as a result of enabled
+       * arrays, etc:
+       */
+      if (varying_inputs & VERT_BIT_COLOR0) fp_inputs |= FRAG_BIT_COL0;
+      if (varying_inputs & VERT_BIT_COLOR1) fp_inputs |= FRAG_BIT_COL1;
+
+      fp_inputs |= (((varying_inputs & VERT_BIT_TEX_ANY) >> VERT_ATTRIB_TEX0) 
+                    << FRAG_ATTRIB_TEX0);
+
+   }
+   else {
+      /* calculate from vp->outputs */
+      struct gl_vertex_program *vprog;
+      GLbitfield vp_outputs;
+
+      /* Choose GLSL vertex shader over ARB vertex program.  Need this
+       * since vertex shader state validation comes after fragment state
+       * validation (see additional comments in state.c).
+       */
+      if (vertexShader)
+         vprog = ctx->Shader.CurrentProgram->VertexProgram;
+      else
+         vprog = ctx->VertexProgram._Current;
+
+      vp_outputs = vprog->Base.OutputsWritten;
+
+      /* These get generated in the setup routine regardless of the
+       * vertex program:
+       */
+      if (ctx->Point.PointSprite)
+         vp_outputs |= FRAG_BITS_TEX_ANY;
+
+      if (vp_outputs & (1 << VERT_RESULT_COL0)) fp_inputs |= FRAG_BIT_COL0;
+      if (vp_outputs & (1 << VERT_RESULT_COL1)) fp_inputs |= FRAG_BIT_COL1;
+
+      fp_inputs |= (((vp_outputs & VERT_RESULT_TEX_ANY) >> VERT_RESULT_TEX0) 
+                    << FRAG_ATTRIB_TEX0);
+   }
+   
+   return fp_inputs;
+}
+
+
 /**
  * Examine current texture environment state and generate a unique
  * key to identify it.
@@ -253,7 +351,9 @@ static GLuint translate_tex_src_bit( GLbitfield bit )
 static void make_state_key( GLcontext *ctx,  struct state_key *key )
 {
    GLuint i, j;
-	
+   GLbitfield inputs_referenced = FRAG_BIT_COL0;
+   GLbitfield inputs_available = get_fp_input_mask( ctx );
+
    memset(key, 0, sizeof(*key));
 
    for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
@@ -267,6 +367,8 @@ static void make_state_key( GLcontext *ctx,  struct state_key *key )
 
       key->unit[i].enabled = 1;
       key->enabled_units |= (1<<i);
+      key->nr_enabled_units = i+1;
+      inputs_referenced |= FRAG_BIT_TEX(i);
 
       key->unit[i].source_index = 
 	 translate_tex_src_bit(texUnit->_ReallyEnabled);		
@@ -297,16 +399,22 @@ static void make_state_key( GLcontext *ctx,  struct state_key *key )
       }
    }
 	
-   if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+   if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
       key->separate_specular = 1;
+      inputs_referenced |= FRAG_BIT_COL1;
+   }
 
    if (ctx->Fog.Enabled) {
       key->fog_enabled = 1;
       key->fog_mode = translate_fog_mode(ctx->Fog.Mode);
+      inputs_referenced |= FRAG_BIT_FOGC; /* maybe */
    }
+
+   key->inputs_available = (inputs_available & inputs_referenced);
 }
 
-/* Use uregs to represent registers internally, translate to Mesa's
+/**
+ * Use uregs to represent registers internally, translate to Mesa's
  * expected formats on emit.  
  *
  * NOTE: These are passed by value extensively in this file rather
@@ -339,16 +447,16 @@ static const struct ureg undef = {
 };
 
 
-/* State used to build the fragment program:
+/** State used to build the fragment program:
  */
 struct texenv_fragment_program {
    struct gl_fragment_program *program;
    GLcontext *ctx;
    struct state_key *state;
 
-   GLbitfield alu_temps;	/* Track texture indirections, see spec. */
-   GLbitfield temps_output;	/* Track texture indirections, see spec. */
-   GLbitfield temp_in_use;	/* Tracks temporary regs which are in use. */
+   GLbitfield alu_temps;	/**< Track texture indirections, see spec. */
+   GLbitfield temps_output;	/**< Track texture indirections, see spec. */
+   GLbitfield temp_in_use;	/**< Tracks temporary regs which are in use. */
    GLboolean error;
 
    struct ureg src_texture[MAX_TEXTURE_COORD_UNITS];   
@@ -356,11 +464,11 @@ struct texenv_fragment_program {
     * else undef.
     */
 
-   struct ureg src_previous;	/* Reg containing color from previous 
+   struct ureg src_previous;	/**< Reg containing color from previous 
 				 * stage.  May need to be decl'd.
 				 */
 
-   GLuint last_tex_stage;	/* Number of last enabled texture unit */
+   GLuint last_tex_stage;	/**< Number of last enabled texture unit */
 
    struct ureg half;
    struct ureg one;
@@ -438,7 +546,7 @@ static struct ureg get_tex_temp( struct texenv_fragment_program *p )
 {
    int bit;
    
-   /* First try to find availble temp not previously used (to avoid
+   /* First try to find available temp not previously used (to avoid
     * starting a new texture indirection).  According to the spec, the
     * ~p->temps_output isn't necessary, but will keep it there for
     * now:
@@ -463,6 +571,14 @@ static struct ureg get_tex_temp( struct texenv_fragment_program *p )
 }
 
 
+/** Mark a temp reg as being no longer allocatable. */
+static void reserve_temp( struct texenv_fragment_program *p, struct ureg r )
+{
+   if (r.file == PROGRAM_TEMPORARY)
+      p->temps_output |= (1 << r.idx);
+}
+
+
 static void release_temps(GLcontext *ctx, struct texenv_fragment_program *p )
 {
    GLuint max_temp = ctx->Const.FragmentProgram.MaxTemps;
@@ -501,11 +617,29 @@ static struct ureg register_param5( struct texenv_fragment_program *p,
 #define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
 
+static GLuint frag_to_vert_attrib( GLuint attrib )
+{
+   switch (attrib) {
+   case FRAG_ATTRIB_COL0: return VERT_ATTRIB_COLOR0;
+   case FRAG_ATTRIB_COL1: return VERT_ATTRIB_COLOR1;
+   default:
+      assert(attrib >= FRAG_ATTRIB_TEX0);
+      assert(attrib <= FRAG_ATTRIB_TEX7);
+      return attrib - FRAG_ATTRIB_TEX0 + VERT_ATTRIB_TEX0;
+   }
+}
+
 
 static struct ureg register_input( struct texenv_fragment_program *p, GLuint input )
 {
-   p->program->Base.InputsRead |= (1 << input);
-   return make_ureg(PROGRAM_INPUT, input);
+   if (p->state->inputs_available & (1<<input)) {
+      p->program->Base.InputsRead |= (1 << input);
+      return make_ureg(PROGRAM_INPUT, input);
+   }
+   else {
+      GLuint idx = frag_to_vert_attrib( input );
+      return register_param3( p, STATE_INTERNAL, STATE_CURRENT_ATTRIB, idx );
+   }
 }
 
 
@@ -556,10 +690,12 @@ emit_op(struct texenv_fragment_program *p,
 
    emit_dst( &inst->DstReg, dest, mask );
 
+#if 0
    /* Accounting for indirection tracking:
     */
    if (dest.file == PROGRAM_TEMPORARY)
       p->temps_output |= 1 << dest.idx;
+#endif
 
    return inst;
 }
@@ -614,6 +750,10 @@ static struct ureg emit_texld( struct texenv_fragment_program *p,
 
    p->program->Base.NumTexInstructions++;
 
+   /* Accounting for indirection tracking:
+    */
+   reserve_temp(p, dest);
+
    /* Is this a texture indirection?
     */
    if ((coord.file == PROGRAM_TEMPORARY &&
@@ -638,14 +778,16 @@ static struct ureg register_const4f( struct texenv_fragment_program *p,
 {
    GLfloat values[4];
    GLuint idx, swizzle;
+   struct ureg r;
    values[0] = s0;
    values[1] = s1;
    values[2] = s2;
    values[3] = s3;
    idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
                                      &swizzle );
-   ASSERT(swizzle == SWIZZLE_NOOP);
-   return make_ureg(PROGRAM_CONSTANT, idx);
+   r = make_ureg(PROGRAM_CONSTANT, idx);
+   r.swz = swizzle;
+   return r;
 }
 
 #define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
@@ -1031,6 +1173,7 @@ static void load_texture( struct texenv_fragment_program *p, GLuint unit )
 	 p->src_texture[unit] = emit_texld( p, OPCODE_TXP,
 					    tmp, WRITEMASK_XYZW, 
 					    unit, dim, texcoord );
+
 	 if (p->state->unit[unit].shadow)
 	    p->program->Base.ShadowSamplers |= 1 << unit;
 
@@ -1157,6 +1300,7 @@ create_new_program(GLcontext *ctx, struct state_key *key,
       for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++)
 	 if (key->enabled_units & (1<<unit)) {
 	    p.src_previous = emit_texenv( &p, unit );
+            reserve_temp(&p, p.src_previous); /* don't re-use this temp reg */
 	    release_temps(ctx, &p);	/* release all temps */
 	 }
    }
diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c
index ffbb7fb19dc..16d05cc7d07 100644
--- a/src/mesa/main/texformat.c
+++ b/src/mesa/main/texformat.c
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -333,6 +334,30 @@ const struct gl_texture_format _mesa_texformat_srgba8 = {
    store_texel_srgba8			/* StoreTexel */
 };
 
+const struct gl_texture_format _mesa_texformat_sargb8 = {
+   MESA_FORMAT_SARGB8,			/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   8,					/* RedBits */
+   8,					/* GreenBits */
+   8,					/* BlueBits */
+   8,					/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   4,					/* TexelBytes */
+   _mesa_texstore_sargb8,		/* StoreTexImageFunc */
+   NULL,				/* FetchTexel1D */
+   NULL,				/* FetchTexel2D */
+   NULL,				/* FetchTexel3D */
+   fetch_texel_1d_sargb8,		/* FetchTexel1Df */
+   fetch_texel_2d_sargb8,		/* FetchTexel2Df */
+   fetch_texel_3d_sargb8,		/* FetchTexel3Df */
+   store_texel_sargb8			/* StoreTexel */
+};
+
 const struct gl_texture_format _mesa_texformat_sl8 = {
    MESA_FORMAT_SL8,			/* MesaFormat */
    GL_LUMINANCE,			/* BaseFormat */
@@ -1624,21 +1649,40 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat,
          case GL_SLUMINANCE_ALPHA_EXT:
          case GL_SLUMINANCE8_ALPHA8_EXT:
             return &_mesa_texformat_sla8;
-         /* NOTE: not supporting any compression of sRGB at this time */
-         case GL_COMPRESSED_SRGB_EXT:
-            return &_mesa_texformat_srgb8;
-         case GL_COMPRESSED_SRGB_ALPHA_EXT:
-            return &_mesa_texformat_srgba8;
          case GL_COMPRESSED_SLUMINANCE_EXT:
             return &_mesa_texformat_sl8;
          case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
             return &_mesa_texformat_sla8;
-         case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+         case GL_COMPRESSED_SRGB_EXT:
+#if FEATURE_texture_s3tc
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgb_dxt1;
+#endif
             return &_mesa_texformat_srgb8;
+         case GL_COMPRESSED_SRGB_ALPHA_EXT:
+#if FEATURE_texture_s3tc
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt3; /* Not srgba_dxt1, see spec */
+#endif
+            return &_mesa_texformat_srgba8;
+#if FEATURE_texture_s3tc
+         case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgb_dxt1;
+            break;
          case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt1;
+            break;
          case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt3;
+            break;
          case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
-            return &_mesa_texformat_srgba8;
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt5;
+            break;
+#endif
          default:
             ; /* fallthrough */
       }
@@ -1719,6 +1763,11 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
       *comps = 1; /* XXX OK? */
       return;
 
+   case MESA_FORMAT_S8_Z24:
+      *datatype = GL_UNSIGNED_INT;
+      *comps = 1; /* XXX OK? */
+      return;
+
    case MESA_FORMAT_Z16:
       *datatype = GL_UNSIGNED_SHORT;
       *comps = 1;
@@ -1729,11 +1778,13 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
       *comps = 1;
       return;
 
+#if FEATURE_EXT_texture_sRGB
    case MESA_FORMAT_SRGB8:
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 3;
       return;
    case MESA_FORMAT_SRGBA8:
+   case MESA_FORMAT_SARGB8:
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 4;
       return;
@@ -1745,17 +1796,28 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 2;
       return;
+#endif
 
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       /* XXX generate error instead? */
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 0;
       return;
+#endif
 
    case MESA_FORMAT_RGBA:
       *datatype = CHAN_TYPE;
diff --git a/src/mesa/main/texformat.h b/src/mesa/main/texformat.h
index b1604d4a494..31364c36b1a 100644
--- a/src/mesa/main/texformat.h
+++ b/src/mesa/main/texformat.h
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -98,9 +99,15 @@ enum _format {
    /*@{*/
    MESA_FORMAT_SRGB8,
    MESA_FORMAT_SRGBA8,
+   MESA_FORMAT_SARGB8,
    MESA_FORMAT_SL8,
    MESA_FORMAT_SLA8,
+#if FEATURE_texture_s3tc
    MESA_FORMAT_SRGB_DXT1,
+   MESA_FORMAT_SRGBA_DXT1,
+   MESA_FORMAT_SRGBA_DXT3,
+   MESA_FORMAT_SRGBA_DXT5,
+#endif
    /*@}*/
 #endif
 
@@ -108,12 +115,16 @@ enum _format {
     * \name Compressed texture formats.
     */
    /*@{*/
+#if FEATURE_texture_fxt1
    MESA_FORMAT_RGB_FXT1,
    MESA_FORMAT_RGBA_FXT1,
+#endif
+#if FEATURE_texture_s3tc
    MESA_FORMAT_RGB_DXT1,
    MESA_FORMAT_RGBA_DXT1,
    MESA_FORMAT_RGBA_DXT3,
    MESA_FORMAT_RGBA_DXT5,
+#endif
    /*@}*/
 
    /**
@@ -170,9 +181,15 @@ extern const struct gl_texture_format _mesa_texformat_intensity;
 /*@{*/
 extern const struct gl_texture_format _mesa_texformat_srgb8;
 extern const struct gl_texture_format _mesa_texformat_srgba8;
+extern const struct gl_texture_format _mesa_texformat_sargb8;
 extern const struct gl_texture_format _mesa_texformat_sl8;
 extern const struct gl_texture_format _mesa_texformat_sla8;
+#if FEATURE_texture_s3tc
 extern const struct gl_texture_format _mesa_texformat_srgb_dxt1;
+extern const struct gl_texture_format _mesa_texformat_srgba_dxt1;
+extern const struct gl_texture_format _mesa_texformat_srgba_dxt3;
+extern const struct gl_texture_format _mesa_texformat_srgba_dxt5;
+#endif
 /*@}*/
 #endif
 
@@ -229,12 +246,16 @@ extern const struct gl_texture_format _mesa_texformat_ycbcr_rev;
 
 /** \name Compressed formats */
 /*@{*/
+#if FEATURE_texture_fxt1
 extern const struct gl_texture_format _mesa_texformat_rgb_fxt1;
 extern const struct gl_texture_format _mesa_texformat_rgba_fxt1;
+#endif
+#if FEATURE_texture_s3tc
 extern const struct gl_texture_format _mesa_texformat_rgb_dxt1;
 extern const struct gl_texture_format _mesa_texformat_rgba_dxt1;
 extern const struct gl_texture_format _mesa_texformat_rgba_dxt3;
 extern const struct gl_texture_format _mesa_texformat_rgba_dxt5;
+#endif
 /*@}*/
 
 /** \name The null format */
diff --git a/src/mesa/main/texformat_tmp.h b/src/mesa/main/texformat_tmp.h
index 5f0c674d4fb..275340cabd3 100644
--- a/src/mesa/main/texformat_tmp.h
+++ b/src/mesa/main/texformat_tmp.h
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -1222,11 +1223,11 @@ static void store_texel_srgb8(struct gl_texture_image *texImage,
 static void FETCH(srgba8)(const struct gl_texture_image *texImage,
                           GLint i, GLint j, GLint k, GLfloat *texel )
 {
-   const GLubyte *src = TEXEL_ADDR(GLubyte, texImage, i, j, k, 4);
-   texel[RCOMP] = nonlinear_to_linear(src[0]);
-   texel[GCOMP] = nonlinear_to_linear(src[1]);
-   texel[BCOMP] = nonlinear_to_linear(src[2]);
-   texel[ACOMP] = UBYTE_TO_FLOAT(src[3]); /* linear! */
+   const GLuint s = *TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   texel[RCOMP] = nonlinear_to_linear( (s >> 24) );
+   texel[GCOMP] = nonlinear_to_linear( (s >> 16) & 0xff );
+   texel[BCOMP] = nonlinear_to_linear( (s >>  8) & 0xff );
+   texel[ACOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff ); /* linear! */
 }
 
 #if DIM == 3
@@ -1234,10 +1235,29 @@ static void store_texel_srgba8(struct gl_texture_image *texImage,
                                GLint i, GLint j, GLint k, const void *texel)
 {
    const GLubyte *rgba = (const GLubyte *) texel;
-   GLubyte *dst = TEXEL_ADDR(GLubyte, texImage, i, j, k, 4);
-   dst[0] = rgba[RCOMP];
-   dst[1] = rgba[GCOMP];
-   dst[2] = rgba[BCOMP];
+   GLuint *dst = TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   *dst = PACK_COLOR_8888(rgba[RCOMP], rgba[GCOMP], rgba[BCOMP], rgba[ACOMP]);
+}
+#endif
+
+/* Fetch texel from 1D, 2D or 3D sargb8 texture, return 4 GLfloats */
+static void FETCH(sargb8)(const struct gl_texture_image *texImage,
+                          GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   const GLuint s = *TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   texel[RCOMP] = nonlinear_to_linear( (s >> 16) & 0xff );
+   texel[GCOMP] = nonlinear_to_linear( (s >>  8) & 0xff );
+   texel[BCOMP] = nonlinear_to_linear( (s      ) & 0xff );
+   texel[ACOMP] = UBYTE_TO_FLOAT( (s >> 24) ); /* linear! */
+}
+
+#if DIM == 3
+static void store_texel_sargb8(struct gl_texture_image *texImage,
+                               GLint i, GLint j, GLint k, const void *texel)
+{
+   const GLubyte *rgba = (const GLubyte *) texel;
+   GLuint *dst = TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   *dst = PACK_COLOR_8888(rgba[ACOMP], rgba[RCOMP], rgba[GCOMP], rgba[BCOMP]);
 }
 #endif
 
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index 18ca8450d36..8fd69c4f6ac 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -1161,19 +1161,30 @@ _mesa_init_teximage_fields(GLcontext *ctx, GLenum target,
    img->Width = width;
    img->Height = height;
    img->Depth = depth;
+
    img->Width2 = width - 2 * border;   /* == 1 << img->WidthLog2; */
-   img->Height2 = height - 2 * border; /* == 1 << img->HeightLog2; */
-   img->Depth2 = depth - 2 * border;   /* == 1 << img->DepthLog2; */
    img->WidthLog2 = logbase2(img->Width2);
-   if (height == 1)  /* 1-D texture */
+
+   if (height == 1) { /* 1-D texture */
+      img->Height2 = 1;
       img->HeightLog2 = 0;
-   else
+   }
+   else {
+      img->Height2 = height - 2 * border; /* == 1 << img->HeightLog2; */
       img->HeightLog2 = logbase2(img->Height2);
-   if (depth == 1)   /* 2-D texture */
+   }
+
+   if (depth == 1) {  /* 2-D texture */
+      img->Depth2 = 1;
       img->DepthLog2 = 0;
-   else
+   }
+   else {
+      img->Depth2 = depth - 2 * border;   /* == 1 << img->DepthLog2; */
       img->DepthLog2 = logbase2(img->Depth2);
+   }
+
    img->MaxLog2 = MAX2(img->WidthLog2, img->HeightLog2);
+
    img->IsCompressed = GL_FALSE;
    img->CompressedSize = 0;
 
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 67cfca25651..6360ca15f81 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -3,6 +3,7 @@
  * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -885,8 +886,8 @@ _mesa_swizzle_ubyte_image(GLcontext *ctx,
 
 /*    _mesa_printf("map %d %d %d %d\n", map[0], map[1], map[2], map[3]);  */
 
-   if (srcRowStride == dstRowStride &&
-       srcComponents == dstComponents &&
+   if (srcComponents == dstComponents &&
+       srcRowStride == dstRowStride &&
        srcRowStride == srcWidth * srcComponents &&
        dimensions < 3) {
       /* 1 and 2D images only */
@@ -2769,7 +2770,6 @@ _mesa_texstore_rgba_float16(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_srgb8(TEXSTORE_PARAMS)
 {
-   const GLboolean littleEndian = _mesa_little_endian();
    const struct gl_texture_format *newDstFormat;
    StoreTexImageFunc store;
    GLboolean k;
@@ -2777,14 +2777,8 @@ _mesa_texstore_srgb8(TEXSTORE_PARAMS)
    ASSERT(dstFormat == &_mesa_texformat_srgb8);
 
    /* reuse normal rgb texstore code */
-   if (littleEndian) {
-      newDstFormat = &_mesa_texformat_bgr888;
-      store = _mesa_texstore_bgr888;
-   }
-   else {
-      newDstFormat = &_mesa_texformat_rgb888;
-      store = _mesa_texstore_rgb888;
-   }
+   newDstFormat = &_mesa_texformat_rgb888;
+   store = _mesa_texstore_rgb888;
 
    k = store(ctx, dims, baseInternalFormat,
              newDstFormat, dstAddr,
@@ -2800,17 +2794,13 @@ _mesa_texstore_srgb8(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_srgba8(TEXSTORE_PARAMS)
 {
-   const GLboolean littleEndian = _mesa_little_endian();
    const struct gl_texture_format *newDstFormat;
    GLboolean k;
 
    ASSERT(dstFormat == &_mesa_texformat_srgba8);
 
    /* reuse normal rgba texstore code */
-   if (littleEndian)
-      newDstFormat = &_mesa_texformat_rgba8888_rev;
-   else
-      newDstFormat = &_mesa_texformat_rgba8888;
+   newDstFormat = &_mesa_texformat_rgba8888;
 
    k = _mesa_texstore_rgba8888(ctx, dims, baseInternalFormat,
                                newDstFormat, dstAddr,
@@ -2824,6 +2814,28 @@ _mesa_texstore_srgba8(TEXSTORE_PARAMS)
 
 
 GLboolean
+_mesa_texstore_sargb8(TEXSTORE_PARAMS)
+{
+   const struct gl_texture_format *newDstFormat;
+   GLboolean k;
+
+   ASSERT(dstFormat == &_mesa_texformat_sargb8);
+
+   /* reuse normal rgba texstore code */
+   newDstFormat = &_mesa_texformat_argb8888;
+
+   k = _mesa_texstore_argb8888(ctx, dims, baseInternalFormat,
+                               newDstFormat, dstAddr,
+                               dstXoffset, dstYoffset, dstZoffset,
+                               dstRowStride, dstImageOffsets,
+                               srcWidth, srcHeight, srcDepth,
+                               srcFormat, srcType,
+                               srcAddr, srcPacking);
+   return k;
+}
+
+
+GLboolean
 _mesa_texstore_sl8(TEXSTORE_PARAMS)
 {
    const struct gl_texture_format *newDstFormat;
@@ -2848,17 +2860,13 @@ _mesa_texstore_sl8(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_sla8(TEXSTORE_PARAMS)
 {
-   const GLboolean littleEndian = _mesa_little_endian();
    const struct gl_texture_format *newDstFormat;
    GLboolean k;
 
    ASSERT(dstFormat == &_mesa_texformat_sla8);
 
    /* reuse normal luminance/alpha texstore code */
-   if (littleEndian)
-      newDstFormat = &_mesa_texformat_al88;
-   else
-      newDstFormat = &_mesa_texformat_al88_rev;
+   newDstFormat = &_mesa_texformat_al88;
 
    k = _mesa_texstore_al88(ctx, dims, baseInternalFormat,
                            newDstFormat, dstAddr,
@@ -3688,16 +3696,40 @@ is_srgb_teximage(const struct gl_texture_image *texImage)
    switch (texImage->TexFormat->MesaFormat) {
    case MESA_FORMAT_SRGB8:
    case MESA_FORMAT_SRGBA8:
+   case MESA_FORMAT_SARGB8:
    case MESA_FORMAT_SL8:
    case MESA_FORMAT_SLA8:
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
       return GL_TRUE;
    default:
       return GL_FALSE;
    }
 }
 
-#endif /* FEATURE_EXT_texture_sRGB */
 
+/**
+ * Convert a float value from linear space to a
+ * non-linear sRGB value in [0, 255].
+ * Not terribly efficient.
+ */
+static INLINE GLfloat
+linear_to_nonlinear(GLfloat cl)
+{
+   /* can't have values outside [0, 1] */
+   GLfloat cs;
+   if (cl < 0.0031308) {
+      cs = 12.92 * cl;
+   }
+   else {
+      cs = 1.055 * _mesa_pow(cl, 0.41666) - 0.055;
+   }
+   return cs;
+}
+
+#endif /* FEATURE_EXT_texture_sRGB */
 
 /**
  * This is the software fallback for Driver.GetTexImage().
@@ -3814,12 +3846,34 @@ _mesa_get_teximage(GLcontext *ctx, GLenum target, GLint level,
             }
 #if FEATURE_EXT_texture_sRGB
             else if (is_srgb_teximage(texImage)) {
-               /* no pixel transfer and no non-linear to linear conversion */
-               const GLint comps = texImage->TexFormat->TexelBytes;
-               const GLint rowstride = comps * texImage->RowStride;
-               MEMCPY(dest,
-                      (const GLubyte *) texImage->Data + row * rowstride,
-                      comps * width * sizeof(GLubyte));
+               /* special case this since need to backconvert values */
+               /* convert row to RGBA format */
+               GLfloat rgba[MAX_WIDTH][4];
+               GLint col;
+               GLbitfield transferOps = 0x0;
+
+               for (col = 0; col < width; col++) {
+                  (*texImage->FetchTexelf)(texImage, col, row, img, rgba[col]);
+                  if (texImage->TexFormat->BaseFormat == GL_LUMINANCE) {
+                     rgba[col][RCOMP] = linear_to_nonlinear(rgba[col][RCOMP]);
+                     rgba[col][GCOMP] = 0.0;
+                     rgba[col][BCOMP] = 0.0;
+                  }
+                  else if (texImage->TexFormat->BaseFormat == GL_LUMINANCE_ALPHA) {
+                     rgba[col][RCOMP] = linear_to_nonlinear(rgba[col][RCOMP]);
+                     rgba[col][GCOMP] = 0.0;
+                     rgba[col][BCOMP] = 0.0;
+                  }
+                  else if (texImage->TexFormat->BaseFormat == GL_RGB ||
+                     texImage->TexFormat->BaseFormat == GL_RGBA) {
+                     rgba[col][RCOMP] = linear_to_nonlinear(rgba[col][RCOMP]);
+                     rgba[col][GCOMP] = linear_to_nonlinear(rgba[col][GCOMP]);
+                     rgba[col][BCOMP] = linear_to_nonlinear(rgba[col][BCOMP]);
+                  }
+               }
+               _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba,
+                                          format, type, dest,
+                                          &ctx->Pack, transferOps /*image xfer ops*/);
             }
 #endif /* FEATURE_EXT_texture_sRGB */
             else {
diff --git a/src/mesa/main/texstore.h b/src/mesa/main/texstore.h
index 951d0a112bb..b03386b2acc 100644
--- a/src/mesa/main/texstore.h
+++ b/src/mesa/main/texstore.h
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -73,6 +74,7 @@ extern GLboolean _mesa_texstore_rgba_dxt5(TEXSTORE_PARAMS);
 #if FEATURE_EXT_texture_sRGB
 extern GLboolean _mesa_texstore_srgb8(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_srgba8(TEXSTORE_PARAMS);
+extern GLboolean _mesa_texstore_sargb8(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_sl8(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_sla8(TEXSTORE_PARAMS);
 #endif
diff --git a/src/mesa/main/varray.h b/src/mesa/main/varray.h
index 6385689fcc9..97d5c8219d0 100644
--- a/src/mesa/main/varray.h
+++ b/src/mesa/main/varray.h
@@ -156,6 +156,20 @@ _mesa_LockArraysEXT(GLint first, GLsizei count);
 extern void GLAPIENTRY
 _mesa_UnlockArraysEXT( void );
 
+
+extern void GLAPIENTRY
+_mesa_DrawArrays(GLenum mode, GLint first, GLsizei count);
+
+extern void GLAPIENTRY
+_mesa_DrawElements(GLenum mode, GLsizei count, GLenum type,
+                   const GLvoid *indices);
+
+extern void GLAPIENTRY
+_mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
+                        GLenum type, const GLvoid *indices);
+
+
+
 extern void
 _mesa_init_varray( GLcontext * ctx );
 
diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h
index f5bf6e2c855..9287e8515cc 100644
--- a/src/mesa/main/version.h
+++ b/src/mesa/main/version.h
@@ -1,6 +1,10 @@
 /*
  * Mesa 3-D graphics library
+<<<<<<< HEAD:src/mesa/main/version.h
  * Version:  7.5
+=======
+ * Version:  7.4
+>>>>>>> origin/gallium-0.2:src/mesa/main/version.h
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
diff --git a/src/mesa/math/m_debug_xform.c b/src/mesa/math/m_debug_xform.c
index ec22c7052db..df8cc066b6e 100644
--- a/src/mesa/math/m_debug_xform.c
+++ b/src/mesa/math/m_debug_xform.c
@@ -206,7 +206,8 @@ static int test_transform_function( transform_func func, int psize,
          case VAR:
             break;
          default:
-            abort();
+            ASSERT(0);
+            return 0;
          }
       }
    }
diff --git a/src/mesa/ppc/common_ppc.c b/src/mesa/ppc/common_ppc.c
index 786e1817acf..c7cec96e5d5 100644
--- a/src/mesa/ppc/common_ppc.c
+++ b/src/mesa/ppc/common_ppc.c
@@ -85,7 +85,7 @@ void _mesa_init_all_ppc_transform_asm( void )
    }
    
 # ifndef USE_VMX_ASM
-   _mesa_ppc_cpu_features &= ~PPC_FEATURES_HAS_ALTIVEC;
+   _mesa_ppc_cpu_features &= ~PPC_FEATURE_HAS_ALTIVEC;
 # endif
 #endif
 }
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index c65d886a91d..2e0fc3694a4 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -30,6 +30,27 @@
  * \author Karl Rasche
  */
 
+/**
+Notes on program parameters, etc.
+
+The instructions we emit will use six kinds of source registers:
+
+  PROGRAM_INPUT      - input registers
+  PROGRAM_TEMPORARY  - temp registers
+  PROGRAM_ADDRESS    - address/indirect register
+  PROGRAM_SAMPLER    - texture sampler
+  PROGRAM_CONSTANT   - indexes into program->Parameters, a known constant/literal
+  PROGRAM_STATE_VAR  - indexes into program->Parameters, and may actually be:
+                       + a state variable, like "state.fog.color", or
+                       + a pointer to a "program.local[k]" parameter, or
+                       + a pointer to a "program.env[k]" parameter
+
+Basically, all the program.local[] and program.env[] values will get mapped
+into the unified gl_program->Parameters array.  This solves the problem of
+having three separate program parameter arrays.
+*/
+
+
 #include "main/glheader.h"
 #include "main/imports.h"
 #include "main/context.h"
@@ -38,16 +59,11 @@
 #include "shader/grammar/grammar_mesa.h"
 #include "arbprogparse.h"
 #include "program.h"
+#include "programopt.h"
 #include "prog_parameter.h"
 #include "prog_statevars.h"
 #include "prog_instruction.h"
 
-
-/* For ARB programs, use the NV instruction limits */
-#define MAX_INSTRUCTIONS MAX2(MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS, \
-                              MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS)
-
-
 /**
  * This is basically a union of the vertex_program and fragment_program
  * structs that we can use to parse the program into
@@ -1895,7 +1911,11 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
                                         const_values, 4);
          if (param_var->param_binding_begin == ~0U)
             param_var->param_binding_begin = idx;
-         param_var->param_binding_type = PROGRAM_CONSTANT;
+         param_var->param_binding_type = PROGRAM_STATE_VAR;
+         /* Note: when we reference this parameter in an instruction later,
+          * we'll check if it's really a constant/immediate and set the
+          * instruction register type appropriately.
+          */
          param_var->param_binding_length++;
          Program->Base.NumParameters++;
          break;
@@ -2602,6 +2622,18 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
          return 1;
    }
 
+   if (*File == PROGRAM_STATE_VAR) {
+      enum register_file file;
+
+      /* If we're referencing the Program->Parameters[] array, check if the
+       * parameter is really a constant/literal.  If so, set File to CONSTANT.
+       */
+      assert(*Index < (GLint) Program->Base.Parameters->NumParameters);
+      file = Program->Base.Parameters->Parameters[*Index].Type;
+      if (file == PROGRAM_CONSTANT)
+         *File = PROGRAM_CONSTANT;
+   }
+
    /* Add attributes to InputsRead only if they are used the program.
     * This avoids the handling of unused ATTRIB declarations in the drivers. */
    if (*File == PROGRAM_INPUT)
@@ -3378,11 +3410,11 @@ debug_variables (GLcontext * ctx, struct var_cache *vc_head,
                fprintf (stderr, "%s\n",
                         Program->Base.Parameters->Parameters[a + b].Name);
                if (Program->Base.Parameters->Parameters[a + b].Type == PROGRAM_STATE_VAR) {
-                  const char *s;
+                  char *s;
                   s = _mesa_program_state_string(Program->Base.Parameters->Parameters
                                                  [a + b].StateIndexes);
                   fprintf(stderr, "%s\n", s);
-                  _mesa_free((char *) s);
+                  _mesa_free(s);
                }
                else
                   fprintf (stderr, "%f %f %f %f\n",
@@ -3430,7 +3462,7 @@ parse_instructions(GLcontext * ctx, const GLubyte * inst,
       : ctx->Const.VertexProgram.MaxInstructions;
    GLint err = 0;
 
-   ASSERT(MAX_INSTRUCTIONS >= maxInst);
+   ASSERT(MAX_PROGRAM_INSTRUCTIONS >= maxInst);
 
    Program->MajorVersion = (GLuint) * inst++;
    Program->MinorVersion = (GLuint) * inst++;
@@ -3784,7 +3816,7 @@ _mesa_parse_arb_program(GLcontext *ctx, GLenum target,
 
    /* Initialize the arb_program struct */
    program->Base.String = strz;
-   program->Base.Instructions = _mesa_alloc_instructions(MAX_INSTRUCTIONS);
+   program->Base.Instructions = _mesa_alloc_instructions(MAX_PROGRAM_INSTRUCTIONS);
    program->Base.NumInstructions =
    program->Base.NumTemporaries =
    program->Base.NumParameters =
@@ -3829,12 +3861,12 @@ _mesa_parse_arb_program(GLcontext *ctx, GLenum target,
 
    _mesa_free (parsed);
 
-   /* Reallocate the instruction array from size [MAX_INSTRUCTIONS]
+   /* Reallocate the instruction array from size [MAX_PROGRAM_INSTRUCTIONS]
     * to size [ap.Base.NumInstructions].
     */
    program->Base.Instructions
       = _mesa_realloc_instructions(program->Base.Instructions,
-                                   MAX_INSTRUCTIONS,
+                                   MAX_PROGRAM_INSTRUCTIONS,
                                    program->Base.NumInstructions);
 
    return !err;
@@ -3898,6 +3930,16 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target,
       _mesa_free_parameter_list(program->Base.Parameters);
    program->Base.Parameters    = ap.Base.Parameters;
 
+   /* Append fog instructions now if the program has "OPTION ARB_fog_exp"
+    * or similar.  We used to leave this up to drivers, but it appears
+    * there's no hardware that wants to do fog in a discrete stage separate
+    * from the fragment shader.
+    */
+   if (program->FogOption != GL_NONE) {
+      _mesa_append_fog_code(ctx, program);
+      program->FogOption = GL_NONE;
+   }
+
 #if DEBUG_FP
    _mesa_printf("____________Fragment program %u ________\n", program->Base.Id);
    _mesa_print_program(&program->Base);
diff --git a/src/mesa/shader/atifragshader.h b/src/mesa/shader/atifragshader.h
index 32fb3a80190..e1dc20e6065 100644
--- a/src/mesa/shader/atifragshader.h
+++ b/src/mesa/shader/atifragshader.h
@@ -8,6 +8,8 @@
 #ifndef ATIFRAGSHADER_H
 #define ATIFRAGSHADER_H
 
+#include "main/mtypes.h"
+
 #define MAX_NUM_INSTRUCTIONS_PER_PASS_ATI 8
 #define MAX_NUM_PASSES_ATI                2
 #define MAX_NUM_FRAGMENT_REGISTERS_ATI    6
diff --git a/src/mesa/shader/grammar/descrip.mms b/src/mesa/shader/grammar/descrip.mms
deleted file mode 100644
index 6976b70d6ac..00000000000
--- a/src/mesa/shader/grammar/descrip.mms
+++ /dev/null
@@ -1,42 +0,0 @@
-# Makefile for core library for VMS
-# contributed by Jouk Jansen  [email protected]
-# Last revision : 29 September 2008
-
-.first
-	define gl [----.include.gl]
-	define math [--.math]
-	define swrast [--.swrast]
-	define array_cache [--.array_cache]
-	define main [--.main]
-
-.include [----]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = [----.include],[],[--.main],[--.glapi],[-.slang]
-LIBDIR = [----.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = grammar_mesa.c
-
-OBJECTS = grammar_mesa.obj
-
-##### RULES #####
-
-VERSION=Mesa V3.4
-
-##### TARGETS #####
-all : 
-	$(MMS)$(MMSQUALIFIERS) $(LIBDIR)$(GL_LIB)
-
-# Make the library
-$(LIBDIR)$(GL_LIB) : $(OBJECTS)
-  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
-
-clean :
-	purge
-	delete *.obj;*
-
-grammar_mesa.obj : grammar_mesa.c grammar.c
diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c
index 36a25377c55..9437e596138 100644
--- a/src/mesa/shader/prog_cache.c
+++ b/src/mesa/shader/prog_cache.c
@@ -44,6 +44,7 @@ struct cache_item
 struct gl_program_cache
 {
    struct cache_item **items;
+   struct cache_item *last;
    GLuint size, n_items;
 };
 
@@ -83,6 +84,8 @@ rehash(struct gl_program_cache *cache)
    struct cache_item *c, *next;
    GLuint size, i;
 
+   cache->last = NULL;
+
    size = cache->size * 3;
    items = (struct cache_item**) _mesa_malloc(size * sizeof(*items));
    _mesa_memset(items, 0, size * sizeof(*items));
@@ -105,6 +108,8 @@ clear_cache(GLcontext *ctx, struct gl_program_cache *cache)
 {
    struct cache_item *c, *next;
    GLuint i;
+   
+   cache->last = NULL;
 
    for (i = 0; i < cache->size; i++) {
       for (c = cache->items[i]; c; c = next) {
@@ -149,18 +154,26 @@ _mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *cache)
 
 
 struct gl_program *
-_mesa_search_program_cache(const struct gl_program_cache *cache,
+_mesa_search_program_cache(struct gl_program_cache *cache,
                            const void *key, GLuint keysize)
 {
-   const GLuint hash = hash_key(key, keysize);
-   struct cache_item *c;
-
-   for (c = cache->items[hash % cache->size]; c; c = c->next) {
-      if (c->hash == hash && memcmp(c->key, key, keysize) == 0)
-	 return c->program;
+   if (cache->last && 
+       memcmp(cache->last->key, key, keysize) == 0) {
+      return cache->last->program;
    }
+   else {
+      const GLuint hash = hash_key(key, keysize);
+      struct cache_item *c;
+
+      for (c = cache->items[hash % cache->size]; c; c = c->next) {
+         if (c->hash == hash && memcmp(c->key, key, keysize) == 0) {
+            cache->last = c;
+            return c->program;
+         }
+      }
 
-   return NULL;
+      return NULL;
+   }
 }
 
 
diff --git a/src/mesa/shader/prog_cache.h b/src/mesa/shader/prog_cache.h
index a8c91fba011..4e1ccac03ff 100644
--- a/src/mesa/shader/prog_cache.h
+++ b/src/mesa/shader/prog_cache.h
@@ -42,7 +42,7 @@ _mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *pc);
 
 
 extern struct gl_program *
-_mesa_search_program_cache(const struct gl_program_cache *cache,
+_mesa_search_program_cache(struct gl_program_cache *cache,
                            const void *key, GLuint keysize);
 
 extern void
diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h
index 268afc528c4..c649b3db5ee 100644
--- a/src/mesa/shader/prog_instruction.h
+++ b/src/mesa/shader/prog_instruction.h
@@ -418,11 +418,13 @@ struct prog_instruction
     */
    GLint BranchTarget;
 
+#if 01 /* XXX just use this for i965 driver for now! */
    /**
     * For TEX instructions in shaders, the sampler to use for the
     * texture lookup.
     */
    GLint Sampler;
+#endif
 
    const char *Comment;
 };
diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c
index dc48e84ab29..8ae961241f6 100644
--- a/src/mesa/shader/prog_parameter.c
+++ b/src/mesa/shader/prog_parameter.c
@@ -424,7 +424,7 @@ _mesa_add_state_reference(struct gl_program_parameter_list *paramList,
                           const gl_state_index stateTokens[STATE_LENGTH])
 {
    const GLuint size = 4; /* XXX fix */
-   const char *name;
+   char *name;
    GLint index;
 
    /* Check if the state reference is already in the list */
@@ -451,7 +451,7 @@ _mesa_add_state_reference(struct gl_program_parameter_list *paramList,
    paramList->StateFlags |= _mesa_program_state_flags(stateTokens);
 
    /* free name string here since we duplicated it in add_parameter() */
-   _mesa_free((void *) name);
+   _mesa_free(name);
 
    return index;
 }
diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c
index 8b0f16e688f..d0ed22e98e1 100644
--- a/src/mesa/shader/prog_print.c
+++ b/src/mesa/shader/prog_print.c
@@ -253,7 +253,9 @@ reg_string(enum register_file f, GLint index, gl_prog_print_mode mode,
          {
             struct gl_program_parameter *param
                = prog->Parameters->Parameters + index;
-            sprintf(str, _mesa_program_state_string(param->StateIndexes));
+            char *state = _mesa_program_state_string(param->StateIndexes);
+            sprintf(str, state);
+            _mesa_free(state);
          }
          break;
       case PROGRAM_ADDRESS:
@@ -359,6 +361,19 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended)
 }
 
 
+void
+_mesa_print_swizzle(GLuint swizzle)
+{
+   if (swizzle == SWIZZLE_XYZW) {
+      _mesa_printf(".xyzw\n");
+   }
+   else {
+      const char *s = _mesa_swizzle_string(swizzle, 0, 0);
+      _mesa_printf("%s\n", s);
+   }
+}
+
+
 const char *
 _mesa_writemask_string(GLuint writeMask)
 {
diff --git a/src/mesa/shader/prog_print.h b/src/mesa/shader/prog_print.h
index aa7f902feeb..d55661cebb8 100644
--- a/src/mesa/shader/prog_print.h
+++ b/src/mesa/shader/prog_print.h
@@ -47,6 +47,9 @@ const char *
 _mesa_writemask_string(GLuint writeMask);
 
 extern void
+_mesa_print_swizzle(GLuint swizzle);
+
+extern void
 _mesa_print_alu_instruction(const struct prog_instruction *inst,
                             const char *opcode_string, GLuint numRegs);
 
diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c
index 3ce60427bf4..e1db30b78f4 100644
--- a/src/mesa/shader/prog_statevars.c
+++ b/src/mesa/shader/prog_statevars.c
@@ -395,6 +395,12 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
 
    case STATE_INTERNAL:
       switch (state[1]) {
+      case STATE_CURRENT_ATTRIB: {
+         const GLuint idx = (GLuint) state[2];
+         COPY_4V(value, ctx->Current.Attrib[idx]);
+         return;
+      }						  
+
       case STATE_NORMAL_SCALE:
          ASSIGN_4V(value, 
                    ctx->_ModelViewInvScale, 
@@ -501,6 +507,9 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
          }
          return;
 
+         /* XXX: make sure new tokens added here are also handled in the 
+          * _mesa_program_state_flags() switch, below.
+          */
       default:
          /* unknown state indexes are silently ignored
           *  should be handled by the driver.
@@ -574,11 +583,29 @@ _mesa_program_state_flags(const gl_state_index state[STATE_LENGTH])
 
    case STATE_INTERNAL:
       switch (state[1]) {
+      case STATE_CURRENT_ATTRIB:
+         return _NEW_CURRENT_ATTRIB;
+
+      case STATE_NORMAL_SCALE:
+         return _NEW_MODELVIEW;
+
       case STATE_TEXRECT_SCALE:
       case STATE_SHADOW_AMBIENT:
 	 return _NEW_TEXTURE;
       case STATE_FOG_PARAMS_OPTIMIZED:
 	 return _NEW_FOG;
+      case STATE_LIGHT_SPOT_DIR_NORMALIZED:
+      case STATE_LIGHT_POSITION:
+      case STATE_LIGHT_POSITION_NORMALIZED:
+      case STATE_LIGHT_HALF_VECTOR:
+         return _NEW_LIGHT;
+
+      case STATE_PT_SCALE:
+      case STATE_PT_BIAS:
+      case STATE_PCM_SCALE:
+      case STATE_PCM_BIAS:
+         return _NEW_PIXEL;
+
       default:
          /* unknown state indexes are silently ignored and
          *  no flag set, since it is handled by the driver.
@@ -807,7 +834,7 @@ append_index(char *dst, GLint index)
  * For example, return "state.matrix.texture[2].inverse".
  * Use _mesa_free() to deallocate the string.
  */
-const char *
+char *
 _mesa_program_state_string(const gl_state_index state[STATE_LENGTH])
 {
    char str[1000] = "";
diff --git a/src/mesa/shader/prog_statevars.h b/src/mesa/shader/prog_statevars.h
index 909bce0a4f3..d5358a1d042 100644
--- a/src/mesa/shader/prog_statevars.h
+++ b/src/mesa/shader/prog_statevars.h
@@ -104,6 +104,7 @@ typedef enum gl_state_index_ {
    STATE_LOCAL,
 
    STATE_INTERNAL,		/* Mesa additions */
+   STATE_CURRENT_ATTRIB,        /* ctx->Current vertex attrib value */
    STATE_NORMAL_SCALE,
    STATE_TEXRECT_SCALE,
    STATE_FOG_PARAMS_OPTIMIZED,  /* for faster fog calc */
@@ -130,7 +131,7 @@ extern GLbitfield
 _mesa_program_state_flags(const gl_state_index state[STATE_LENGTH]);
 
 
-extern const char *
+extern char *
 _mesa_program_state_string(const gl_state_index state[STATE_LENGTH]);
 
 
diff --git a/src/mesa/shader/prog_uniform.c b/src/mesa/shader/prog_uniform.c
index dc76be8e46c..0642713148c 100644
--- a/src/mesa/shader/prog_uniform.c
+++ b/src/mesa/shader/prog_uniform.c
@@ -142,8 +142,8 @@ _mesa_longest_uniform_name(const struct gl_uniform_list *list)
    GLint max = 0;
    GLuint i;
    for (i = 0; list && i < list->NumUniforms; i++) {
-      GLuint len = _mesa_strlen(list->Uniforms[i].Name);
-      if (len > (GLuint)max)
+      GLint len = (GLint)_mesa_strlen(list->Uniforms[i].Name);
+      if (len > max)
          max = len;
    }
    return max;
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index d114828ec6a..7a3b827352a 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -554,7 +554,6 @@ _mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count)
    return GL_TRUE;
 }
 
-
 /**
  * Delete 'count' instructions at 'start' in the given program.
  * Adjust branch targets accordingly.
@@ -691,17 +690,47 @@ _mesa_combine_programs(GLcontext *ctx,
 
    if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) {
       struct gl_fragment_program *fprogA, *fprogB, *newFprog;
+      GLbitfield progB_inputsRead = progB->InputsRead;
+      GLint progB_colorFile, progB_colorIndex;
+
       fprogA = (struct gl_fragment_program *) progA;
       fprogB = (struct gl_fragment_program *) progB;
       newFprog = (struct gl_fragment_program *) newProg;
 
       newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill;
 
+      /* We'll do a search and replace for instances
+       * of progB_colorFile/progB_colorIndex below...
+       */
+      progB_colorFile = PROGRAM_INPUT;
+      progB_colorIndex = FRAG_ATTRIB_COL0;
+
+      /*
+       * The fragment program may get color from a state var rather than
+       * a fragment input (vertex output) if it's constant.
+       * See the texenvprogram.c code.
+       * So, search the program's parameter list now to see if the program
+       * gets color from a state var instead of a conventional fragment
+       * input register.
+       */
+      for (i = 0; i < progB->Parameters->NumParameters; i++) {
+         struct gl_program_parameter *p = &progB->Parameters->Parameters[i];
+         if (p->Type == PROGRAM_STATE_VAR &&
+             p->StateIndexes[0] == STATE_INTERNAL &&
+             p->StateIndexes[1] == STATE_CURRENT_ATTRIB &&
+             p->StateIndexes[2] == VERT_ATTRIB_COLOR0) {
+            progB_inputsRead |= FRAG_BIT_COL0;
+            progB_colorFile = PROGRAM_STATE_VAR;
+            progB_colorIndex = i;
+            break;
+         }
+      }
+
       /* Connect color outputs of fprogA to color inputs of fprogB, via a
        * new temporary register.
        */
       if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) &&
-          (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) {
+          (progB_inputsRead & FRAG_BIT_COL0)) {
          GLint tempReg = _mesa_find_free_register(newProg, PROGRAM_TEMPORARY);
          if (tempReg < 0) {
             _mesa_problem(ctx, "No free temp regs found in "
@@ -712,13 +741,14 @@ _mesa_combine_programs(GLcontext *ctx,
          replace_registers(newInst, lenA,
                            PROGRAM_OUTPUT, FRAG_RESULT_COLR,
                            PROGRAM_TEMPORARY, tempReg);
-         /* replace reads from input.color[0] with tempReg */
+         /* replace reads from the input color with tempReg */
          replace_registers(newInst + lenA, lenB,
-                           PROGRAM_INPUT, FRAG_ATTRIB_COL0,
-                           PROGRAM_TEMPORARY, tempReg);
+                           progB_colorFile, progB_colorIndex, /* search for */
+                           PROGRAM_TEMPORARY, tempReg  /* replace with */ );
       }
 
-      inputsB = progB->InputsRead;
+      /* compute combined program's InputsRead */
+      inputsB = progB_inputsRead;
       if (progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) {
          inputsB &= ~(1 << FRAG_ATTRIB_COL0);
       }
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index fb2a7e54fe6..a5cf1ca1d44 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -406,7 +406,7 @@ _mesa_init_shader_state(GLcontext * ctx)
     * are generated by the GLSL compiler.
     */
    ctx->Shader.EmitHighLevelInstructions = GL_TRUE;
-   ctx->Shader.EmitCondCodes = GL_TRUE; /* XXX probably want GL_FALSE... */
+   ctx->Shader.EmitCondCodes = GL_FALSE;/*GL_TRUE;*/ /* XXX probably want GL_FALSE... */
    ctx->Shader.EmitComments = GL_FALSE;
    ctx->Shader.Flags = get_shader_flags();
 }
@@ -1143,7 +1143,8 @@ get_matrix_dims(GLenum type, GLint *rows, GLint *cols)
 
 /**
  * Determine the number of rows and columns occupied by a uniform
- * according to its datatype.
+ * according to its datatype.  For non-matrix types (such as GL_FLOAT_VEC4),
+ * the number of rows = 1 and cols = number of elements in the vector.
  */
 static void
 get_uniform_rows_cols(const struct gl_program_parameter *p,
@@ -1152,11 +1153,17 @@ get_uniform_rows_cols(const struct gl_program_parameter *p,
    get_matrix_dims(p->DataType, rows, cols);
    if (*rows == 0 && *cols == 0) {
       /* not a matrix type, probably a float or vector */
-      *rows = p->Size / 4 + 1;
-      if (p->Size % 4 == 0)
-         *cols = 4;
-      else
-         *cols = p->Size % 4;
+      if (p->Size <= 4) {
+         *rows = 1;
+         *cols = p->Size;
+      }
+      else {
+         *rows = p->Size / 4 + 1;
+         if (p->Size % 4 == 0)
+            *cols = 4;
+         else
+            *cols = p->Size % 4;
+      }
    }
 }
 
@@ -1556,7 +1563,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
       return;
    }
 
-   if (index + offset > program->Parameters->Size) {
+   if (index + offset > (GLint) program->Parameters->Size) {
       /* out of bounds! */
       return;
    }
@@ -1621,7 +1628,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
          /* if the uniform is bool-valued, convert to 1.0 or 0.0 */
          if (is_boolean_type(param->DataType)) {
             for (i = 0; i < elems; i++) {
-               uniformVal[i] = uniformVal[i] ? 1.0 : 0.0;
+               uniformVal[i] = uniformVal[i] ? 1.0f : 0.0f;
             }
          }
       }
diff --git a/src/mesa/shader/slang/library/slang_version_syn.h b/src/mesa/shader/slang/library/slang_version_syn.h
deleted file mode 100644
index 3b94d85927a..00000000000
--- a/src/mesa/shader/slang/library/slang_version_syn.h
+++ /dev/null
@@ -1,64 +0,0 @@
-".syntax version_directive;\n"
-"version_directive\n"
-"	version_directive_1 .and .loop version_directive_2;\n"
-"version_directive_1\n"
-"	prior_optional_spaces .and optional_version_directive .and .true .emit $;\n"
-"version_directive_2\n"
-"	prior_optional_spaces .and version_directive_body .and .true .emit $;\n"
-"optional_version_directive\n"
-"	version_directive_body .or .true .emit 10 .emit 1;\n"
-"version_directive_body\n"
-"	'#' .and optional_space .and \"version\" .and space .and version_number .and optional_space .and\n"
-"	new_line;\n"
-"version_number\n"
-"	version_number_110;\n"
-"version_number_110\n"
-"	leading_zeroes .and \"110\" .emit 10 .emit 1;\n"
-"leading_zeroes\n"
-"	.loop zero;\n"
-"zero\n"
-"	'0';\n"
-"space\n"
-" single_space .and .loop single_space;\n"
-"optional_space\n"
-" .loop single_space;\n"
-"single_space\n"
-" ' ' .or '\\t';\n"
-"prior_optional_spaces\n"
-"	.loop prior_space;\n"
-"prior_space\n"
-"	c_style_comment_block .or cpp_style_comment_block .or space .or new_line;\n"
-"c_style_comment_block\n"
-" '/' .and '*' .and c_style_comment_rest;\n"
-"c_style_comment_rest\n"
-" .loop c_style_comment_char_no_star .and c_style_comment_rest_1;\n"
-"c_style_comment_rest_1\n"
-" c_style_comment_end .or c_style_comment_rest_2;\n"
-"c_style_comment_rest_2\n"
-" '*' .and c_style_comment_rest;\n"
-"c_style_comment_char_no_star\n"
-" '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n"
-"c_style_comment_end\n"
-" '*' .and '/';\n"
-"cpp_style_comment_block\n"
-" '/' .and '/' .and cpp_style_comment_block_1;\n"
-"cpp_style_comment_block_1\n"
-" cpp_style_comment_block_2 .or cpp_style_comment_block_3;\n"
-"cpp_style_comment_block_2\n"
-" .loop cpp_style_comment_char .and new_line;\n"
-"cpp_style_comment_block_3\n"
-" .loop cpp_style_comment_char;\n"
-"cpp_style_comment_char\n"
-" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n"
-"new_line\n"
-" cr_lf .or lf_cr .or '\\n' .or '\\r';\n"
-"cr_lf\n"
-" '\\r' .and '\\n';\n"
-"lf_cr\n"
-"	'\\n' .and '\\r';\n"
-".string __string_filter;\n"
-"__string_filter\n"
-" .loop __identifier_char;\n"
-"__identifier_char\n"
-" 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9';\n"
-""
diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c
index 848a64f932a..99f2cbdcc05 100644
--- a/src/mesa/shader/slang/slang_link.c
+++ b/src/mesa/shader/slang/slang_link.c
@@ -502,7 +502,6 @@ _slang_update_inputs_outputs(struct gl_program *prog)
          maxAddrReg = MAX2(maxAddrReg, inst->DstReg.Index + 1);
       }
    }
-
    prog->NumAddressRegs = maxAddrReg;
 }
 
diff --git a/src/mesa/sources b/src/mesa/sources.mak
index 01a3fb89e61..dfafbc982bf 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources.mak
@@ -162,6 +162,54 @@ VBO_SOURCES = \
 	vbo/vbo_save_draw.c \
 	vbo/vbo_save_loopback.c 
 
+VF_SOURCES = \
+	vf/vf.c \
+	vf/vf_generic.c \
+	vf/vf_sse.c
+
+STATETRACKER_SOURCES = \
+	state_tracker/st_atom.c \
+	state_tracker/st_atom_blend.c \
+	state_tracker/st_atom_clip.c \
+	state_tracker/st_atom_constbuf.c \
+	state_tracker/st_atom_depth.c \
+	state_tracker/st_atom_framebuffer.c \
+	state_tracker/st_atom_pixeltransfer.c \
+	state_tracker/st_atom_sampler.c \
+	state_tracker/st_atom_scissor.c \
+	state_tracker/st_atom_shader.c \
+	state_tracker/st_atom_rasterizer.c \
+	state_tracker/st_atom_stipple.c \
+	state_tracker/st_atom_texture.c \
+	state_tracker/st_atom_viewport.c \
+	state_tracker/st_cb_accum.c \
+	state_tracker/st_cb_bitmap.c \
+	state_tracker/st_cb_blit.c \
+	state_tracker/st_cb_bufferobjects.c \
+	state_tracker/st_cb_clear.c \
+	state_tracker/st_cb_flush.c \
+	state_tracker/st_cb_get.c \
+	state_tracker/st_cb_drawpixels.c \
+	state_tracker/st_cb_fbo.c \
+	state_tracker/st_cb_feedback.c \
+	state_tracker/st_cb_program.c \
+	state_tracker/st_cb_queryobj.c \
+	state_tracker/st_cb_rasterpos.c \
+	state_tracker/st_cb_readpixels.c \
+	state_tracker/st_cb_strings.c \
+	state_tracker/st_cb_texture.c \
+	state_tracker/st_api.c \
+	state_tracker/st_context.c \
+	state_tracker/st_debug.c \
+	state_tracker/st_draw.c \
+	state_tracker/st_draw_feedback.c \
+	state_tracker/st_extensions.c \
+	state_tracker/st_format.c \
+	state_tracker/st_framebuffer.c \
+	state_tracker/st_gen_mipmap.c \
+	state_tracker/st_mesa_to_tgsi.c \
+	state_tracker/st_program.c \
+	state_tracker/st_texture.c
 
 SHADER_SOURCES = \
 	shader/arbprogparse.c \
@@ -267,27 +315,38 @@ MESA_SOURCES = \
 	$(SHADER_SOURCES)	\
 	$(SWRAST_SOURCES)	\
 	$(SWRAST_SETUP_SOURCES)	\
-	$(COMMON_DRIVER_SOURCES) \
+	$(COMMON_DRIVER_SOURCES)\
 	$(ASM_C_SOURCES)	\
 	$(SLANG_SOURCES)
 
 ALL_SOURCES = \
 	$(MESA_SOURCES)		\
 	$(GLAPI_SOURCES)	\
-	$(ASM_SOURCES)		\
-	$(COMMON_DRIVER_SOURCES)
+	$(MESA_ASM_SOURCES)
 
+MESA_GALLIUM_SOURCES = \
+	$(MAIN_SOURCES)		\
+	$(MATH_SOURCES)		\
+	$(VBO_SOURCES)		\
+	$(STATETRACKER_SOURCES)	\
+	$(SHADER_SOURCES)	\
+	$(ASM_C_SOURCES)	\
+	$(SLANG_SOURCES)
 
 
 ### Object files
 
 MESA_OBJECTS = \
 	$(MESA_SOURCES:.c=.o) \
-	$(ASM_SOURCES:.S=.o)
+	$(MESA_ASM_SOURCES:.S=.o)
+
+MESA_GALLIUM_OBJECTS = \
+	$(MESA_GALLIUM_SOURCES:.c=.o) \
+	$(MESA_ASM_SOURCES:.S=.o)
 
 GLAPI_OBJECTS = \
 	$(GLAPI_SOURCES:.c=.o) \
-	$(ASM_API:.S=.o)
+	$(GLAPI_ASM_SOURCES:.S=.o)
 
 
 COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o)
@@ -295,11 +354,10 @@ COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o)
 
 
 ### Include directories
-### XXX we should be able to trim this down to just -Iinclude/ and -Isrc/mesa/
-### since most #includes have been changed from #include "mtypes.h" to
-### #include "main/mtypes.h", etc.
 
 INCLUDE_DIRS = \
 	-I$(TOP)/include \
 	-I$(TOP)/src/mesa \
-	-I$(TOP)/src/mesa/main
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary
diff --git a/src/mesa/state_tracker/Makefile b/src/mesa/state_tracker/Makefile
new file mode 100644
index 00000000000..0ab1dc6e6bd
--- /dev/null
+++ b/src/mesa/state_tracker/Makefile
@@ -0,0 +1,2 @@
+default:
+	cd ../.. ; make
+\ No newline at end of file
diff --git a/src/mesa/state_tracker/st_api.c b/src/mesa/state_tracker/st_api.c
new file mode 100644
index 00000000000..fc0e9a23164
--- /dev/null
+++ b/src/mesa/state_tracker/st_api.c
@@ -0,0 +1,33 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 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.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Just a global symbol for EGL to look for to identify the supported
+ * graphics API.
+ */
+int st_api_OpenGL = 1;
diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c
new file mode 100644
index 00000000000..fc8587f4598
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom.c
@@ -0,0 +1,218 @@
+/**************************************************************************
+ * 
+ * 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 "main/glheader.h"
+#include "main/context.h"
+
+#include "pipe/p_defines.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_bitmap.h"
+#include "st_program.h"
+
+       
+
+/* This is used to initialize st->atoms[].  We could use this list
+ * directly except for a single atom, st_update_constants, which has a
+ * .dirty value which changes according to the parameters of the
+ * current fragment and vertex programs, and so cannot be a static
+ * value.
+ */
+static const struct st_tracked_state *atoms[] =
+{
+   &st_update_depth_stencil_alpha,
+   &st_update_clip,
+
+   &st_finalize_textures,
+   &st_update_shader,
+
+   &st_update_rasterizer,
+   &st_update_polygon_stipple,
+   &st_update_viewport,
+   &st_update_scissor,
+   &st_update_blend,
+   &st_update_sampler,
+   &st_update_texture,
+   &st_update_framebuffer,
+   &st_update_vs_constants,
+   &st_update_fs_constants,
+   &st_update_pixel_transfer
+};
+
+
+void st_init_atoms( struct st_context *st )
+{
+   GLuint i;
+
+   st->atoms = malloc(sizeof(atoms));
+   st->nr_atoms = sizeof(atoms)/sizeof(*atoms);
+   memcpy(st->atoms, atoms, sizeof(atoms));
+
+   /* Patch in a pointer to the dynamic state atom:
+    */
+   for (i = 0; i < st->nr_atoms; i++) {
+      if (st->atoms[i] == &st_update_vs_constants) {
+	 st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_VERTEX];
+	 st->atoms[i][0] = st_update_vs_constants;
+      }
+
+      if (st->atoms[i] == &st_update_fs_constants) {
+	 st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_FRAGMENT];
+	 st->atoms[i][0] = st_update_fs_constants;
+      }
+   }
+}
+
+
+void st_destroy_atoms( struct st_context *st )
+{
+   if (st->atoms) {
+      free(st->atoms);
+      st->atoms = NULL;
+   }
+}
+
+
+/***********************************************************************
+ */
+
+static GLboolean check_state( const struct st_state_flags *a,
+			      const struct st_state_flags *b )
+{
+   return ((a->mesa & b->mesa) ||
+	   (a->st & b->st));
+}
+
+static void accumulate_state( struct st_state_flags *a,
+			      const struct st_state_flags *b )
+{
+   a->mesa |= b->mesa;
+   a->st |= b->st;
+}
+
+
+static void xor_states( struct st_state_flags *result,
+			     const struct st_state_flags *a,
+			      const struct st_state_flags *b )
+{
+   result->mesa = a->mesa ^ b->mesa;
+   result->st = a->st ^ b->st;
+}
+
+
+/* Too complex to figure out, just check every time:
+ */
+static void check_program_state( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+
+   if (ctx->VertexProgram._Current != &st->vp->Base)
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+
+   if (ctx->FragmentProgram._Current != &st->fp->Base)
+      st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+
+}
+
+
+/***********************************************************************
+ * Update all derived state:
+ */
+
+void st_validate_state( struct st_context *st )
+{
+   struct st_state_flags *state = &st->dirty;
+   GLuint i;
+
+   /* The bitmap cache is immune to pixel unpack changes.
+    * Note that GLUT makes several calls to glPixelStore for each
+    * bitmap char it draws so this is an important check.
+    */
+   if (state->mesa & ~_NEW_PACKUNPACK)
+      st_flush_bitmap_cache(st);
+
+   check_program_state( st );
+
+   if (state->st == 0)
+      return;
+
+//   _mesa_printf("%s %x/%x\n", __FUNCTION__, state->mesa, state->st);
+
+   if (1) {
+      /* Debug version which enforces various sanity checks on the
+       * state flags which are generated and checked to help ensure
+       * state atoms are ordered correctly in the list.
+       */
+      struct st_state_flags examined, prev;      
+      memset(&examined, 0, sizeof(examined));
+      prev = *state;
+
+      for (i = 0; i < st->nr_atoms; i++) {	 
+	 const struct st_tracked_state *atom = st->atoms[i];
+	 struct st_state_flags generated;
+	 
+//	 _mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st);
+
+	 if (!(atom->dirty.mesa || atom->dirty.st) ||
+	     !atom->update) {
+	    _mesa_printf("malformed atom %s\n", atom->name);
+	    assert(0);
+	 }
+
+	 if (check_state(state, &atom->dirty)) {
+	    st->atoms[i]->update( st );
+//	    _mesa_printf("after: %x\n", atom->dirty.mesa);
+	 }
+
+	 accumulate_state(&examined, &atom->dirty);
+
+	 /* generated = (prev ^ state)
+	  * if (examined & generated)
+	  *     fail;
+	  */
+	 xor_states(&generated, &prev, state);
+	 assert(!check_state(&examined, &generated));
+	 prev = *state;
+      }
+//      _mesa_printf("\n");
+
+   }
+   else {
+      const GLuint nr = st->nr_atoms;
+
+      for (i = 0; i < nr; i++) {	 
+	 if (check_state(state, &st->atoms[i]->dirty))
+	    st->atoms[i]->update( st );
+      }
+   }
+
+   memset(state, 0, sizeof(*state));
+}
+
+
+
diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h
new file mode 100644
index 00000000000..c7cffd85c8a
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom.h
@@ -0,0 +1,66 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+    
+
+#ifndef ST_ATOM_H
+#define ST_ATOM_H
+
+struct st_context;
+struct st_tracked_state;
+
+void st_init_atoms( struct st_context *st );
+void st_destroy_atoms( struct st_context *st );
+
+
+void st_validate_state( struct st_context *st );
+
+
+extern const struct st_tracked_state st_update_framebuffer;
+extern const struct st_tracked_state st_update_clip;
+extern const struct st_tracked_state st_update_depth_stencil_alpha;
+extern const struct st_tracked_state st_update_shader;
+extern const struct st_tracked_state st_update_rasterizer;
+extern const struct st_tracked_state st_update_polygon_stipple;
+extern const struct st_tracked_state st_update_viewport;
+extern const struct st_tracked_state st_update_scissor;
+extern const struct st_tracked_state st_update_blend;
+extern const struct st_tracked_state st_update_sampler;
+extern const struct st_tracked_state st_update_texture;
+extern const struct st_tracked_state st_finalize_textures;
+extern const struct st_tracked_state st_update_fs_constants;
+extern const struct st_tracked_state st_update_vs_constants;
+extern const struct st_tracked_state st_update_pixel_transfer;
+
+
+uint st_compare_func_to_pipe(GLenum func);
+
+#endif
diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c
new file mode 100644
index 00000000000..35c09c3e08f
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_blend.c
@@ -0,0 +1,232 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_context.h"
+
+#include "main/macros.h"
+
+/**
+ * Convert GLenum blend tokens to pipe tokens.
+ * Both blend factors and blend funcs are accepted.
+ */
+static GLuint
+translate_blend(GLenum blend)
+{
+   switch (blend) {
+   /* blend functions */
+   case GL_FUNC_ADD:
+      return PIPE_BLEND_ADD;
+   case GL_FUNC_SUBTRACT:
+      return PIPE_BLEND_SUBTRACT;
+   case GL_FUNC_REVERSE_SUBTRACT:
+      return PIPE_BLEND_REVERSE_SUBTRACT;
+   case GL_MIN:
+      return PIPE_BLEND_MIN;
+   case GL_MAX:
+      return PIPE_BLEND_MAX;
+
+   /* blend factors */
+   case GL_ONE:
+      return PIPE_BLENDFACTOR_ONE;
+   case GL_SRC_COLOR:
+      return PIPE_BLENDFACTOR_SRC_COLOR;
+   case GL_SRC_ALPHA:
+      return PIPE_BLENDFACTOR_SRC_ALPHA;
+   case GL_DST_ALPHA:
+      return PIPE_BLENDFACTOR_DST_ALPHA;
+   case GL_DST_COLOR:
+      return PIPE_BLENDFACTOR_DST_COLOR;
+   case GL_SRC_ALPHA_SATURATE:
+      return PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE;
+   case GL_CONSTANT_COLOR:
+      return PIPE_BLENDFACTOR_CONST_COLOR;
+   case GL_CONSTANT_ALPHA:
+      return PIPE_BLENDFACTOR_CONST_ALPHA;
+      /*
+      return PIPE_BLENDFACTOR_SRC1_COLOR;
+      return PIPE_BLENDFACTOR_SRC1_ALPHA;
+      */
+   case GL_ZERO:
+      return PIPE_BLENDFACTOR_ZERO;
+   case GL_ONE_MINUS_SRC_COLOR:
+      return PIPE_BLENDFACTOR_INV_SRC_COLOR;
+   case GL_ONE_MINUS_SRC_ALPHA:
+      return PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+   case GL_ONE_MINUS_DST_COLOR:
+      return PIPE_BLENDFACTOR_INV_DST_COLOR;
+   case GL_ONE_MINUS_DST_ALPHA:
+      return PIPE_BLENDFACTOR_INV_DST_ALPHA;
+   case GL_ONE_MINUS_CONSTANT_COLOR:
+      return PIPE_BLENDFACTOR_INV_CONST_COLOR;
+   case GL_ONE_MINUS_CONSTANT_ALPHA:
+      return PIPE_BLENDFACTOR_INV_CONST_ALPHA;
+      /*
+      return PIPE_BLENDFACTOR_INV_SRC1_COLOR;
+      return PIPE_BLENDFACTOR_INV_SRC1_ALPHA;
+      */
+   default:
+      assert("invalid GL token in translate_blend()" == NULL);
+      return 0;
+   }
+}
+
+
+/**
+ * Convert GLenum logicop tokens to pipe tokens.
+ */
+static GLuint
+translate_logicop(GLenum logicop)
+{
+   switch (logicop) {
+   case GL_CLEAR:
+      return PIPE_LOGICOP_CLEAR;
+   case GL_NOR:
+      return PIPE_LOGICOP_NOR;
+   case GL_AND_INVERTED:
+      return PIPE_LOGICOP_AND_INVERTED;
+   case GL_COPY_INVERTED:
+      return PIPE_LOGICOP_COPY_INVERTED;
+   case GL_AND_REVERSE:
+      return PIPE_LOGICOP_AND_REVERSE;
+   case GL_INVERT:
+      return PIPE_LOGICOP_INVERT;
+   case GL_XOR:
+      return PIPE_LOGICOP_XOR;
+   case GL_NAND:
+      return PIPE_LOGICOP_NAND;
+   case GL_AND:
+      return PIPE_LOGICOP_AND;
+   case GL_EQUIV:
+      return PIPE_LOGICOP_EQUIV;
+   case GL_NOOP:
+      return PIPE_LOGICOP_NOOP;
+   case GL_OR_INVERTED:
+      return PIPE_LOGICOP_OR_INVERTED;
+   case GL_COPY:
+      return PIPE_LOGICOP_COPY;
+   case GL_OR_REVERSE:
+      return PIPE_LOGICOP_OR_REVERSE;
+   case GL_OR:
+      return PIPE_LOGICOP_OR;
+   case GL_SET:
+      return PIPE_LOGICOP_SET;
+   default:
+      assert("invalid GL token in translate_logicop()" == NULL);
+      return 0;
+   }
+}
+
+
+static void 
+update_blend( struct st_context *st )
+{
+   struct pipe_blend_state *blend = &st->state.blend;
+
+   memset(blend, 0, sizeof(*blend));
+
+   if (st->ctx->Color.ColorLogicOpEnabled ||
+       (st->ctx->Color.BlendEnabled &&
+        st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) {
+      /* logicop enabled */
+      blend->logicop_enable = 1;
+      blend->logicop_func = translate_logicop(st->ctx->Color.LogicOp);
+   }
+   else if (st->ctx->Color.BlendEnabled) {
+      /* blending enabled */
+      blend->blend_enable = 1;
+
+      blend->rgb_func = translate_blend(st->ctx->Color.BlendEquationRGB);
+      if (st->ctx->Color.BlendEquationRGB == GL_MIN ||
+          st->ctx->Color.BlendEquationRGB == GL_MAX) {
+         /* Min/max are special */
+         blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor = PIPE_BLENDFACTOR_ONE;
+      }
+      else {
+         blend->rgb_src_factor = translate_blend(st->ctx->Color.BlendSrcRGB);
+         blend->rgb_dst_factor = translate_blend(st->ctx->Color.BlendDstRGB);
+      }
+
+      blend->alpha_func = translate_blend(st->ctx->Color.BlendEquationA);
+      if (st->ctx->Color.BlendEquationA == GL_MIN ||
+          st->ctx->Color.BlendEquationA == GL_MAX) {
+         /* Min/max are special */
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ONE;
+      }
+      else {
+         blend->alpha_src_factor = translate_blend(st->ctx->Color.BlendSrcA);
+         blend->alpha_dst_factor = translate_blend(st->ctx->Color.BlendDstA);
+      }
+   }
+   else {
+      /* no blending / logicop */
+   }
+
+   /* Colormask - maybe reverse these bits? */
+   if (st->ctx->Color.ColorMask[0])
+      blend->colormask |= PIPE_MASK_R;
+   if (st->ctx->Color.ColorMask[1])
+      blend->colormask |= PIPE_MASK_G;
+   if (st->ctx->Color.ColorMask[2])
+      blend->colormask |= PIPE_MASK_B;
+   if (st->ctx->Color.ColorMask[3])
+      blend->colormask |= PIPE_MASK_A;
+
+   if (st->ctx->Color.DitherFlag)
+      blend->dither = 1;
+
+   cso_set_blend(st->cso_context, blend);
+
+   {
+      struct pipe_blend_color bc;
+      COPY_4FV(bc.color, st->ctx->Color.BlendColor);
+      cso_set_blend_color(st->cso_context, &bc);
+   }
+}
+
+
+const struct st_tracked_state st_update_blend = {
+   "st_update_blend",					/* name */
+   {							/* dirty */
+      (_NEW_COLOR),  /* XXX _NEW_BLEND someday? */	/* mesa */
+      0,						/* st */
+   },
+   update_blend,					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c
new file mode 100644
index 00000000000..23d709b814c
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_clip.c
@@ -0,0 +1,71 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+ 
+
+#include "st_context.h"
+#include "pipe/p_context.h"
+#include "st_atom.h"
+
+
+/* Second state atom for user clip planes:
+ */
+static void update_clip( struct st_context *st )
+{
+   struct pipe_clip_state clip;
+   GLuint i;
+
+   memset(&clip, 0, sizeof(clip));
+
+   for (i = 0; i < PIPE_MAX_CLIP_PLANES; i++) {
+      if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) {
+	 memcpy(clip.ucp[clip.nr], 
+		st->ctx->Transform._ClipUserPlane[i], 
+		sizeof(clip.ucp[0]));
+	 clip.nr++;
+      }
+   }
+      
+   if (memcmp(&clip, &st->state.clip, sizeof(clip)) != 0) {
+      st->state.clip = clip;
+      st->pipe->set_clip_state(st->pipe, &clip);
+   }
+}
+
+
+const struct st_tracked_state st_update_clip = {
+   "st_update_clip",					/* name */
+   {							/* dirty */
+      (_NEW_TRANSFORM),					/* mesa */
+      0,						/* st */
+   },
+   update_clip						/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
new file mode 100644
index 00000000000..514b10cd022
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -0,0 +1,140 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_program.h"
+
+
+/**
+ * Pass the given program parameters to the graphics pipe as a
+ * constant buffer.
+ * \param id  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
+ */
+void st_upload_constants( struct st_context *st,
+                          struct gl_program_parameter_list *params,
+                          unsigned id)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_constant_buffer *cbuf = &st->state.constants[id];
+
+   assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT);
+
+   /* update constants */
+   if (params && params->NumParameters) {
+      const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4;
+
+      /* Update our own dependency flags.  This works because this
+       * function will also be called whenever the program changes.
+       */
+      st->constants.tracked_state[id].dirty.mesa =
+         (params->StateFlags | _NEW_PROGRAM);
+
+      _mesa_load_state_parameters(st->ctx, params);
+
+      /* We always need to get a new buffer, to keep the drivers simple and
+       * avoid gratuitous rendering synchronization.
+       */
+      pipe_buffer_reference(pipe->screen, &cbuf->buffer, NULL );
+      cbuf->buffer = pipe_buffer_create(pipe->screen, 16, PIPE_BUFFER_USAGE_CONSTANT,
+					paramBytes );
+
+      if (0)
+      {
+	 printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
+                __FUNCTION__, id, params->NumParameters, params->StateFlags);
+         _mesa_print_parameter_list(params);
+      }
+
+      /* load Mesa constants into the constant buffer */
+      if (cbuf->buffer) {
+         void *map = pipe_buffer_map(pipe->screen, cbuf->buffer,
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+         memcpy(map, params->ParameterValues, paramBytes);
+         pipe_buffer_unmap(pipe->screen, cbuf->buffer);
+      }
+
+      st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf);
+   }
+   else {
+      st->constants.tracked_state[id].dirty.mesa = 0;
+      //  st->pipe->set_constant_buffer(st->pipe, id, 0, NULL);
+   }
+}
+
+/* Vertex shader:
+ */
+static void update_vs_constants(struct st_context *st )
+{
+   struct st_vertex_program *vp = st->vp;
+   struct gl_program_parameter_list *params = vp->Base.Base.Parameters;
+
+   st_upload_constants( st, params, PIPE_SHADER_VERTEX );
+}
+
+const struct st_tracked_state st_update_vs_constants = {
+   "st_update_vs_constants",				/* name */
+   {							/* dirty */
+      0,  /* set dynamically above */			/* mesa */
+      ST_NEW_VERTEX_PROGRAM,				/* st */
+   },
+   update_vs_constants					/* update */
+};
+
+/* Fragment shader:
+ */
+static void update_fs_constants(struct st_context *st )
+{
+   struct st_fragment_program *fp = st->fp;
+   struct gl_program_parameter_list *params = fp->Base.Base.Parameters;
+
+   st_upload_constants( st, params, PIPE_SHADER_FRAGMENT );
+}
+
+const struct st_tracked_state st_update_fs_constants = {
+   "st_update_fs_constants",				/* name */
+   {							/* dirty */
+      0,  /* set dynamically above */			/* mesa */
+      ST_NEW_FRAGMENT_PROGRAM,				/* st */
+   },
+   update_fs_constants					/* update */
+};
+
diff --git a/src/mesa/state_tracker/st_atom_constbuf.h b/src/mesa/state_tracker/st_atom_constbuf.h
new file mode 100644
index 00000000000..f707534e2cf
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_constbuf.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_ATOM_CONSTBUF_H
+#define ST_ATOM_CONSTBUF_H
+
+
+void st_upload_constants( struct st_context *st,
+                          struct gl_program_parameter_list *params,
+                          unsigned id);
+
+
+#endif /* ST_ATOM_CONSTBUF_H */
diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c
new file mode 100644
index 00000000000..2d617bd95dc
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_depth.c
@@ -0,0 +1,151 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  *   Zack  Rusin
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Convert an OpenGL compare mode to a pipe tokens.
+ */
+GLuint
+st_compare_func_to_pipe(GLenum func)
+{
+   /* Same values, just biased */
+   assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER);
+   assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER);
+   assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER);
+   assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER);
+   assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER);
+   assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER);
+   assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER);
+   assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER);
+   assert(func >= GL_NEVER);
+   assert(func <= GL_ALWAYS);
+   return func - GL_NEVER;
+}
+
+
+/**
+ * Convert GLenum stencil op tokens to pipe tokens.
+ */
+static GLuint
+gl_stencil_op_to_pipe(GLenum func)
+{
+   switch (func) {
+   case GL_KEEP:
+      return PIPE_STENCIL_OP_KEEP;
+   case GL_ZERO:
+      return PIPE_STENCIL_OP_ZERO;
+   case GL_REPLACE:
+      return PIPE_STENCIL_OP_REPLACE;
+   case GL_INCR:
+      return PIPE_STENCIL_OP_INCR;
+   case GL_DECR:
+      return PIPE_STENCIL_OP_DECR;
+   case GL_INCR_WRAP:
+      return PIPE_STENCIL_OP_INCR_WRAP;
+   case GL_DECR_WRAP:
+      return PIPE_STENCIL_OP_DECR_WRAP;
+   case GL_INVERT:
+      return PIPE_STENCIL_OP_INVERT;
+   default:
+      assert("invalid GL token in gl_stencil_op_to_pipe()" == NULL);
+      return 0;
+   }
+}
+
+static void
+update_depth_stencil_alpha(struct st_context *st)
+{
+   struct pipe_depth_stencil_alpha_state *dsa = &st->state.depth_stencil;
+
+   memset(dsa, 0, sizeof(*dsa));
+
+   dsa->depth.enabled = st->ctx->Depth.Test;
+   dsa->depth.writemask = st->ctx->Depth.Mask;
+   dsa->depth.func = st_compare_func_to_pipe(st->ctx->Depth.Func);
+
+   if (st->ctx->Query.CurrentOcclusionObject &&
+       st->ctx->Query.CurrentOcclusionObject->Active)
+      dsa->depth.occlusion_count = 1;
+
+   if (st->ctx->Stencil.Enabled && st->ctx->Visual.stencilBits > 0) {
+      dsa->stencil[0].enabled = 1;
+      dsa->stencil[0].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]);
+      dsa->stencil[0].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]);
+      dsa->stencil[0].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]);
+      dsa->stencil[0].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]);
+      dsa->stencil[0].ref_value = st->ctx->Stencil.Ref[0] & 0xff;
+      dsa->stencil[0].valuemask = st->ctx->Stencil.ValueMask[0] & 0xff;
+      dsa->stencil[0].writemask = st->ctx->Stencil.WriteMask[0] & 0xff;
+
+      if (st->ctx->Stencil._TestTwoSide) {
+         dsa->stencil[1].enabled = 1;
+         dsa->stencil[1].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]);
+         dsa->stencil[1].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]);
+         dsa->stencil[1].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]);
+         dsa->stencil[1].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]);
+         dsa->stencil[1].ref_value = st->ctx->Stencil.Ref[1] & 0xff;
+         dsa->stencil[1].valuemask = st->ctx->Stencil.ValueMask[1] & 0xff;
+         dsa->stencil[1].writemask = st->ctx->Stencil.WriteMask[1] & 0xff;
+      }
+      else {
+         dsa->stencil[1] = dsa->stencil[0];
+         dsa->stencil[1].enabled = 0;
+      }
+   }
+
+   if (st->ctx->Color.AlphaEnabled) {
+      dsa->alpha.enabled = 1;
+      dsa->alpha.func = st_compare_func_to_pipe(st->ctx->Color.AlphaFunc);
+      dsa->alpha.ref_value = st->ctx->Color.AlphaRef;
+   }
+
+   cso_set_depth_stencil_alpha(st->cso_context, dsa);
+}
+
+
+const struct st_tracked_state st_update_depth_stencil_alpha = {
+   "st_update_depth_stencil",				/* name */
+   {							/* dirty */
+      (_NEW_DEPTH|_NEW_STENCIL|_NEW_COLOR),		/* mesa */
+      0,						/* st */
+   },
+   update_depth_stencil_alpha				/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
new file mode 100644
index 00000000000..625efdd66b5
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -0,0 +1,176 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+ 
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_fbo.h"
+#include "st_public.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "cso_cache/cso_context.h"
+
+
+
+/**
+ * When doing GL render to texture, we have to be sure that finalize_texture()
+ * didn't yank out the pipe_texture that we earlier created a surface for.
+ * Check for that here and create a new surface if needed.
+ */
+static void
+update_renderbuffer_surface(struct st_context *st,
+                            struct st_renderbuffer *strb)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_texture *texture = strb->rtt->pt;
+   int rtt_width = strb->Base.Width;
+   int rtt_height = strb->Base.Height;
+
+   if (!strb->surface ||
+       strb->surface->texture != texture ||
+       strb->surface->width != rtt_width ||
+       strb->surface->height != rtt_height) {
+      GLuint level;
+      /* find matching mipmap level size */
+      for (level = 0; level <= texture->last_level; level++) {
+         if (texture->width[level] == rtt_width &&
+             texture->height[level] == rtt_height) {
+
+            pipe_surface_reference(&strb->surface, NULL);
+
+            strb->surface = screen->get_tex_surface(screen,
+                                              texture,
+                                              strb->rtt_face,
+                                              level,
+                                              strb->rtt_slice,
+                                              PIPE_BUFFER_USAGE_GPU_READ |
+                                              PIPE_BUFFER_USAGE_GPU_WRITE);
+#if 0
+            printf("-- alloc new surface %d x %d into tex %p\n",
+                   strb->surface->width, strb->surface->height,
+                   texture);
+#endif
+            break;
+         }
+      }
+   }
+}
+
+
+/**
+ * Update framebuffer state (color, depth, stencil, etc. buffers)
+ */
+static void
+update_framebuffer_state( struct st_context *st )
+{
+   struct pipe_framebuffer_state *framebuffer = &st->state.framebuffer;
+   struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   struct st_renderbuffer *strb;
+   GLuint i;
+
+   memset(framebuffer, 0, sizeof(*framebuffer));
+
+   framebuffer->width = fb->Width;
+   framebuffer->height = fb->Height;
+
+   /*printf("------ fb size %d x %d\n", fb->Width, fb->Height);*/
+
+   /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state
+    * to determine which surfaces to draw to
+    */
+   framebuffer->nr_cbufs = 0;
+   for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
+      strb = st_renderbuffer(fb->_ColorDrawBuffers[i]);
+
+      /*printf("--------- framebuffer surface rtt %p\n", strb->rtt);*/
+      if (strb->rtt) {
+         /* rendering to a GL texture, may have to update surface */
+         update_renderbuffer_surface(st, strb);
+      }
+
+      if (strb->surface) {
+         framebuffer->cbufs[framebuffer->nr_cbufs] = strb->surface;
+         framebuffer->nr_cbufs++;
+      }
+   }
+
+   strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer);
+   if (strb) {
+      strb = st_renderbuffer(strb->Base.Wrapped);
+      if (strb->rtt) {
+         /* rendering to a GL texture, may have to update surface */
+         update_renderbuffer_surface(st, strb);
+      }
+
+      framebuffer->zsbuf = strb->surface;
+   }
+   else {
+      strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer);
+      if (strb) {
+         strb = st_renderbuffer(strb->Base.Wrapped);
+         assert(strb->surface);
+         framebuffer->zsbuf = strb->surface;
+      }
+   }
+
+   cso_set_framebuffer(st->cso_context, framebuffer);
+
+   if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
+      if (st->frontbuffer_status == FRONT_STATUS_COPY_OF_BACK) {
+         /* copy back color buffer to front color buffer */
+         struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+	 struct pipe_surface *surf_front, *surf_back;
+         (void) st_get_framebuffer_surface(stfb, ST_SURFACE_FRONT_LEFT, &surf_front);
+         (void) st_get_framebuffer_surface(stfb, ST_SURFACE_BACK_LEFT, &surf_back);
+
+         st->pipe->surface_copy(st->pipe,
+                                FALSE,
+                                surf_front, 0, 0,  /* dest */
+                                surf_back, 0, 0,   /* src */
+                                fb->Width, fb->Height);
+      }
+      /* we're assuming we'll really draw to the front buffer */
+      st->frontbuffer_status = FRONT_STATUS_DIRTY;
+   }
+}
+
+
+const struct st_tracked_state st_update_framebuffer = {
+   "st_update_framebuffer",				/* name */
+   {							/* dirty */
+      _NEW_BUFFERS,					/* mesa */
+      ST_NEW_FRAMEBUFFER,				/* st */
+   },
+   update_framebuffer_state				/* update */
+};
+
diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c
new file mode 100644
index 00000000000..a357b716771
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c
@@ -0,0 +1,474 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+/*
+ * Generate fragment programs to implement pixel transfer ops, such as
+ * scale/bias, colormatrix, colortable, convolution...
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "shader/program.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_format.h"
+#include "st_program.h"
+#include "st_texture.h"
+
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_pack_color.h"
+
+
+struct state_key
+{
+   GLuint scaleAndBias:1;
+   GLuint colorMatrix:1;
+   GLuint colorMatrixPostScaleBias:1;
+   GLuint pixelMaps:1;
+
+#if 0
+   GLfloat Maps[3][256][4];
+   int NumMaps;
+   GLint NumStages;
+   pipeline_stage Stages[STAGE_MAX];
+   GLboolean StagesUsed[STAGE_MAX];
+   GLfloat Scale1[4], Bias1[4];
+   GLfloat Scale2[4], Bias2[4];
+#endif
+};
+
+
+static GLboolean
+is_identity(const GLfloat m[16])
+{
+   GLuint i;
+   for (i = 0; i < 16; i++) {
+      const int row = i % 4, col = i / 4;
+      const float val = (GLfloat)(row == col);
+      if (m[i] != val)
+         return GL_FALSE;
+   }
+   return GL_TRUE;
+}
+
+
+static void
+make_state_key(GLcontext *ctx,  struct state_key *key)
+{
+   static const GLfloat zero[4] = { 0.0f, 0.0f, 0.0f, 0.0f };
+   static const GLfloat one[4] = { 1.0f, 1.0f, 1.0f, 1.0f };
+
+   memset(key, 0, sizeof(*key));
+
+   if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 ||
+       ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 ||
+       ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 ||
+       ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) {
+      key->scaleAndBias = 1;
+   }
+
+   if (!is_identity(ctx->ColorMatrixStack.Top->m)) {
+      key->colorMatrix = 1;
+   }
+
+   if (!TEST_EQ_4V(ctx->Pixel.PostColorMatrixScale, one) ||
+       !TEST_EQ_4V(ctx->Pixel.PostColorMatrixBias, zero)) {
+      key->colorMatrixPostScaleBias = 1;
+   }
+
+   key->pixelMaps = ctx->Pixel.MapColorFlag;
+}
+
+
+static struct pipe_texture *
+create_color_map_texture(GLcontext *ctx)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_texture *pt;
+   enum pipe_format format;
+   const uint texSize = 256; /* simple, and usually perfect */
+
+   /* find an RGBA texture format */
+   format = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER);
+
+   /* create texture for color map/table */
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0,
+                          texSize, texSize, 1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   return pt;
+}
+
+
+/**
+ * Update the pixelmap texture with the contents of the R/G/B/A pixel maps.
+ */
+static void
+load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surface;
+   const GLuint rSize = ctx->PixelMaps.RtoR.Size;
+   const GLuint gSize = ctx->PixelMaps.GtoG.Size;
+   const GLuint bSize = ctx->PixelMaps.BtoB.Size;
+   const GLuint aSize = ctx->PixelMaps.AtoA.Size;
+   const uint texSize = pt->width[0];
+   uint *dest;
+   uint i, j;
+
+   surface = screen->get_tex_surface(screen, pt, 0, 0, 0, 
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+   dest = (uint *) screen->surface_map(screen, surface,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* Pack four 1D maps into a 2D texture:
+    * R map is placed horizontally, indexed by S, in channel 0
+    * G map is placed vertically, indexed by T, in channel 1
+    * B map is placed horizontally, indexed by S, in channel 2
+    * A map is placed vertically, indexed by T, in channel 3
+    */
+   for (i = 0; i < texSize; i++) {
+      for (j = 0; j < texSize; j++) {
+         int k = (i * texSize + j);
+         ubyte r = ctx->PixelMaps.RtoR.Map8[j * rSize / texSize];
+         ubyte g = ctx->PixelMaps.GtoG.Map8[i * gSize / texSize];
+         ubyte b = ctx->PixelMaps.BtoB.Map8[j * bSize / texSize];
+         ubyte a = ctx->PixelMaps.AtoA.Map8[i * aSize / texSize];
+         util_pack_color_ub(r, g, b, a, pt->format, dest + k);
+      }
+   }
+
+   screen->surface_unmap(screen, surface);
+   pipe_surface_reference(&surface, NULL);
+}
+
+
+
+#define MAX_INST 100
+
+/**
+ * Returns a fragment program which implements the current pixel transfer ops.
+ */
+static struct gl_fragment_program *
+get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key)
+{
+   struct st_context *st = ctx->st;
+   struct prog_instruction inst[MAX_INST];
+   struct gl_program_parameter_list *params;
+   struct gl_fragment_program *fp;
+   GLuint ic = 0;
+   const GLuint colorTemp = 0;
+
+   fp = (struct gl_fragment_program *)
+      ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!fp)
+      return NULL;
+
+   params = _mesa_new_parameter_list();
+
+   /*
+    * Get initial pixel color from the texture.
+    * TEX colorTemp, fragment.texcoord[0], texture[0], 2D;
+    */
+   _mesa_init_instructions(inst + ic, 1);
+   inst[ic].Opcode = OPCODE_TEX;
+   inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+   inst[ic].DstReg.Index = colorTemp;
+   inst[ic].SrcReg[0].File = PROGRAM_INPUT;
+   inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   inst[ic].TexSrcUnit = 0;
+   inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+   fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0);
+   fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR);
+   fp->Base.SamplersUsed = 0x1;  /* sampler 0 (bit 0) is used */
+
+   if (key->scaleAndBias) {
+      static const gl_state_index scale_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 };
+      static const gl_state_index bias_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 };
+      GLfloat scale[4], bias[4];
+      GLint scale_p, bias_p;
+
+      scale[0] = ctx->Pixel.RedScale;
+      scale[1] = ctx->Pixel.GreenScale;
+      scale[2] = ctx->Pixel.BlueScale;
+      scale[3] = ctx->Pixel.AlphaScale;
+      bias[0] = ctx->Pixel.RedBias;
+      bias[1] = ctx->Pixel.GreenBias;
+      bias[2] = ctx->Pixel.BlueBias;
+      bias[3] = ctx->Pixel.AlphaBias;
+
+      scale_p = _mesa_add_state_reference(params, scale_state);
+      bias_p = _mesa_add_state_reference(params, bias_state);
+
+      /* MAD colorTemp, colorTemp, scale, bias; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MAD;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = scale_p;
+      inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[2].Index = bias_p;
+      ic++;
+   }
+
+   if (key->pixelMaps) {
+      const GLuint temp = 1;
+
+      /* create the colormap/texture now if not already done */
+      if (!st->pixel_xfer.pixelmap_texture) {
+         st->pixel_xfer.pixelmap_texture = create_color_map_texture(ctx);
+      }
+
+      /* with a little effort, we can do four pixel map look-ups with
+       * two TEX instructions:
+       */
+
+      /* TEX temp.rg, colorTemp.rgba, texture[1], 2D; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_TEX;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_XY; /* write R,G */
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].TexSrcUnit = 1;
+      inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+      ic++;
+
+      /* TEX temp.ba, colorTemp.baba, texture[1], 2D; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_TEX;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_ZW; /* write B,A */
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W,
+                                                 SWIZZLE_Z, SWIZZLE_W);
+      inst[ic].TexSrcUnit = 1;
+      inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+      ic++;
+
+      /* MOV colorTemp, temp; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MOV;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = temp;
+      ic++;
+
+      fp->Base.SamplersUsed |= (1 << 1);  /* sampler 1 is used */
+   }
+
+   if (key->colorMatrix) {
+      static const gl_state_index row0_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 0, 0, 0 };
+      static const gl_state_index row1_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 1, 1, 0 };
+      static const gl_state_index row2_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 2, 2, 0 };
+      static const gl_state_index row3_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 3, 3, 0 };
+
+      GLint row0_p = _mesa_add_state_reference(params, row0_state);
+      GLint row1_p = _mesa_add_state_reference(params, row1_state);
+      GLint row2_p = _mesa_add_state_reference(params, row2_state);
+      GLint row3_p = _mesa_add_state_reference(params, row3_state);
+      const GLuint temp = 1;
+
+      /* DP4 temp.x, colorTemp, matrow0; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_X;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row0_p;
+      ic++;
+
+      /* DP4 temp.y, colorTemp, matrow1; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_Y;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row1_p;
+      ic++;
+
+      /* DP4 temp.z, colorTemp, matrow2; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_Z;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row2_p;
+      ic++;
+
+      /* DP4 temp.w, colorTemp, matrow3; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_W;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row3_p;
+      ic++;
+
+      /* MOV colorTemp, temp; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MOV;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = temp;
+      ic++;
+   }
+
+   if (key->colorMatrixPostScaleBias) {
+      static const gl_state_index scale_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 };
+      static const gl_state_index bias_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 };
+      GLint scale_param, bias_param;
+
+      scale_param = _mesa_add_state_reference(params, scale_state);
+      bias_param = _mesa_add_state_reference(params, bias_state);
+
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MAD;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = scale_param;
+      inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[2].Index = bias_param;
+      ic++;
+   }
+
+   /* Modify last instruction's dst reg to write to result.color */
+   {
+      struct prog_instruction *last = &inst[ic - 1];
+      last->DstReg.File = PROGRAM_OUTPUT;
+      last->DstReg.Index = FRAG_RESULT_COLR;
+   }
+
+   /* END; */
+   _mesa_init_instructions(inst + ic, 1);
+   inst[ic].Opcode = OPCODE_END;
+   ic++;
+
+   assert(ic <= MAX_INST);
+
+
+   fp->Base.Instructions = _mesa_alloc_instructions(ic);
+   if (!fp->Base.Instructions) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY,
+                  "generating pixel transfer program");
+      return NULL;
+   }
+
+   _mesa_copy_instructions(fp->Base.Instructions, inst, ic);
+   fp->Base.NumInstructions = ic;
+   fp->Base.Parameters = params;
+
+#if 0
+   printf("========= pixel transfer prog\n");
+   _mesa_print_program(&fp->Base);
+   _mesa_print_parameter_list(fp->Base.Parameters);
+#endif
+
+   return fp;
+}
+
+
+
+/**
+ * Update st->pixel_xfer.program in response to new pixel-transfer state.
+ */
+static void
+update_pixel_transfer(struct st_context *st)
+{
+   GLcontext *ctx = st->ctx;
+   struct state_key key;
+   struct gl_fragment_program *fp;
+
+   make_state_key(st->ctx, &key);
+
+   fp = (struct gl_fragment_program *)
+      _mesa_search_program_cache(st->pixel_xfer.cache, &key, sizeof(key));
+   if (!fp) {
+      fp = get_pixel_transfer_program(st->ctx, &key);
+      _mesa_program_cache_insert(st->ctx, st->pixel_xfer.cache,
+                                 &key, sizeof(key), &fp->Base);
+   }
+
+   if (ctx->Pixel.MapColorFlag) {
+      load_color_map_texture(ctx, st->pixel_xfer.pixelmap_texture);
+   }
+   st->pixel_xfer.pixelmap_enabled = ctx->Pixel.MapColorFlag;
+
+   st->pixel_xfer.program = (struct st_fragment_program *) fp;
+}
+
+
+
+const struct st_tracked_state st_update_pixel_transfer = {
+   "st_update_pixel_transfer",				/* name */
+   {							/* dirty */
+      _NEW_PIXEL | _NEW_COLOR_MATRIX,			/* mesa */
+      0,						/* st */
+   },
+   update_pixel_transfer				/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c
new file mode 100644
index 00000000000..ea76487bcfc
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_rasterizer.c
@@ -0,0 +1,289 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+ 
+#include "main/macros.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_context.h"
+
+
+static GLuint translate_fill( GLenum mode )
+{
+   switch (mode) {
+   case GL_POINT:
+      return PIPE_POLYGON_MODE_POINT;
+   case GL_LINE:
+      return PIPE_POLYGON_MODE_LINE;
+   case GL_FILL:
+      return PIPE_POLYGON_MODE_FILL;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static GLboolean get_offset_flag( GLuint fill_mode, 
+				  const struct gl_polygon_attrib *p )
+{
+   switch (fill_mode) {
+   case PIPE_POLYGON_MODE_POINT:
+      return p->OffsetPoint;
+   case PIPE_POLYGON_MODE_LINE:
+      return p->OffsetLine;
+   case PIPE_POLYGON_MODE_FILL:
+      return p->OffsetFill;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+static void update_raster_state( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+   struct pipe_rasterizer_state *raster = &st->state.rasterizer;
+   const struct gl_vertex_program *vertProg = ctx->VertexProgram._Current;
+   uint i;
+
+   memset(raster, 0, sizeof(*raster));
+
+   raster->origin_lower_left = 1; /* Always true for OpenGL */
+   
+   /* _NEW_POLYGON, _NEW_BUFFERS
+    */
+   {
+      if (ctx->Polygon.FrontFace == GL_CCW)
+         raster->front_winding = PIPE_WINDING_CCW;
+      else
+         raster->front_winding = PIPE_WINDING_CW;
+
+      /* XXX
+       * I think the intention here is that user-created framebuffer objects
+       * use Y=0=TOP layout instead of OpenGL's normal Y=0=bottom layout.
+       * Flipping Y changes CW to CCW and vice-versa.
+       * But this is an implementation/driver-specific artifact - remove...
+       */
+      if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0)
+         raster->front_winding ^= PIPE_WINDING_BOTH;
+   }
+
+   /* _NEW_LIGHT
+    */
+   if (ctx->Light.ShadeModel == GL_FLAT)
+      raster->flatshade = 1;
+
+   /* _NEW_LIGHT | _NEW_PROGRAM
+    *
+    * Back-face colors can come from traditional lighting (when
+    * GL_LIGHT_MODEL_TWO_SIDE is set) or from vertex programs/shaders (when
+    * GL_VERTEX_PROGRAM_TWO_SIDE is set).  Note the logic here.
+    */
+   if (ctx->VertexProgram._Current) {
+      if (ctx->VertexProgram._Enabled ||
+          (ctx->Shader.CurrentProgram &&
+           ctx->Shader.CurrentProgram->VertexProgram &&
+           ctx->Shader.CurrentProgram->LinkStatus)) {
+         /* user-defined vertex program or shader */
+         raster->light_twoside = ctx->VertexProgram.TwoSideEnabled;
+      }
+      else {
+         /* TNL-generated program */
+         raster->light_twoside = ctx->Light.Enabled && ctx->Light.Model.TwoSide;
+      }
+   }
+   else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
+      raster->light_twoside = 1;
+   }
+
+   /* _NEW_POLYGON
+    */
+   if (ctx->Polygon.CullFlag) {
+      if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) {
+	 raster->cull_mode = PIPE_WINDING_BOTH;
+      }
+      else if (ctx->Polygon.CullFaceMode == GL_FRONT) {
+	 raster->cull_mode = raster->front_winding;
+      }
+      else {
+	 raster->cull_mode = raster->front_winding ^ PIPE_WINDING_BOTH;
+      }
+   }
+
+   /* _NEW_POLYGON
+    */
+   {
+      GLuint fill_front = translate_fill( ctx->Polygon.FrontMode );
+      GLuint fill_back = translate_fill( ctx->Polygon.BackMode );
+      
+      if (raster->front_winding == PIPE_WINDING_CW) {
+	 raster->fill_cw = fill_front;
+	 raster->fill_ccw = fill_back;
+      }
+      else {
+	 raster->fill_cw = fill_back;
+	 raster->fill_ccw = fill_front;
+      }
+
+      /* Simplify when culling is active:
+       */
+      if (raster->cull_mode & PIPE_WINDING_CW) {
+	 raster->fill_cw = raster->fill_ccw;
+      }
+      
+      if (raster->cull_mode & PIPE_WINDING_CCW) {
+	 raster->fill_ccw = raster->fill_cw;
+      }
+   }
+
+   /* _NEW_POLYGON 
+    */
+   if (ctx->Polygon.OffsetUnits != 0.0 ||
+       ctx->Polygon.OffsetFactor != 0.0) {
+      raster->offset_cw = get_offset_flag( raster->fill_cw, &ctx->Polygon );
+      raster->offset_ccw = get_offset_flag( raster->fill_ccw, &ctx->Polygon );
+      raster->offset_units = ctx->Polygon.OffsetUnits;
+      raster->offset_scale = ctx->Polygon.OffsetFactor;
+   }
+
+   if (ctx->Polygon.SmoothFlag)
+      raster->poly_smooth = 1;
+
+   if (ctx->Polygon.StippleFlag)
+      raster->poly_stipple_enable = 1;
+
+
+   /* _NEW_BUFFERS, _NEW_POLYGON
+    */
+   if (raster->fill_cw != PIPE_POLYGON_MODE_FILL ||
+       raster->fill_ccw != PIPE_POLYGON_MODE_FILL)
+   {
+      GLfloat mrd = (ctx->DrawBuffer ? 
+		     ctx->DrawBuffer->_MRD : 
+		     1.0f);
+
+      raster->offset_units = ctx->Polygon.OffsetFactor * mrd;
+      raster->offset_scale = (ctx->Polygon.OffsetUnits * mrd *
+			    st->polygon_offset_scale);
+   }
+      
+   /* _NEW_POINT
+    */
+   raster->point_size = ctx->Point.Size;
+
+   raster->point_size_min = 0;    /* temporary, will go away */
+   raster->point_size_max = 1000; /* temporary, will go away */
+
+   raster->point_smooth = ctx->Point.SmoothFlag;
+   raster->point_sprite = ctx->Point.PointSprite;
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+      if (ctx->Point.CoordReplace[i]) {
+         if (ctx->Point.SpriteOrigin == GL_UPPER_LEFT)
+            raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_UPPER_LEFT;
+         else 
+            raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_LOWER_LEFT;
+      }
+      else {
+         raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE;
+      }
+   }
+
+   /* ST_NEW_VERTEX_PROGRAM
+    */
+   if (vertProg) {
+      if (vertProg->Base.Id == 0) {
+         if (vertProg->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
+            /* generated program which emits point size */
+            raster->point_size_per_vertex = TRUE;
+         }
+      }
+      else if (ctx->VertexProgram.PointSizeEnabled) {
+         /* user-defined program and GL_VERTEX_PROGRAM_POINT_SIZE set */
+         raster->point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled;
+      }
+   }
+   if (!raster->point_size_per_vertex) {
+      /* clamp size now */
+      raster->point_size = CLAMP(ctx->Point.Size,
+                                 ctx->Point.MinSize,
+                                 ctx->Point.MaxSize);
+   }
+
+   /* _NEW_LINE
+    */
+   raster->line_smooth = ctx->Line.SmoothFlag;
+   if (ctx->Line.SmoothFlag) {
+      raster->line_width = CLAMP(ctx->Line.Width,
+                                ctx->Const.MinLineWidthAA,
+                                ctx->Const.MaxLineWidthAA);
+   }
+   else {
+      raster->line_width = CLAMP(ctx->Line.Width,
+                                ctx->Const.MinLineWidth,
+                                ctx->Const.MaxLineWidth);
+   }
+
+   raster->line_stipple_enable = ctx->Line.StippleFlag;
+   raster->line_stipple_pattern = ctx->Line.StipplePattern;
+   /* GL stipple factor is in [1,256], remap to [0, 255] here */
+   raster->line_stipple_factor = ctx->Line.StippleFactor - 1;
+
+   /* _NEW_MULTISAMPLE */
+   if (ctx->Multisample._Enabled || st->force_msaa)
+      raster->multisample = 1;
+
+   /* _NEW_SCISSOR */
+   if (ctx->Scissor.Enabled)
+      raster->scissor = 1;
+
+   raster->gl_rasterization_rules = 1;
+
+   cso_set_rasterizer(st->cso_context, raster);
+}
+
+const struct st_tracked_state st_update_rasterizer = {
+   "st_update_rasterizer",    /* name */
+   {
+      (_NEW_BUFFERS |
+       _NEW_LIGHT |
+       _NEW_LINE |
+       _NEW_MULTISAMPLE |
+       _NEW_POINT |
+       _NEW_POLYGON |
+       _NEW_PROGRAM |
+       _NEW_SCISSOR),      /* mesa state dependencies*/
+      ST_NEW_VERTEX_PROGRAM,  /* state tracker dependencies */
+   },
+   update_raster_state     /* update function */
+};
diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c
new file mode 100644
index 00000000000..d7b904354f2
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_sampler.c
@@ -0,0 +1,217 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+ 
+
+#include "main/macros.h"
+
+#include "st_context.h"
+#include "st_cb_texture.h"
+#include "st_atom.h"
+#include "st_program.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Convert GLenum texcoord wrap tokens to pipe tokens.
+ */
+static GLuint
+gl_wrap_to_sp(GLenum wrap)
+{
+   switch (wrap) {
+   case GL_REPEAT:
+      return PIPE_TEX_WRAP_REPEAT;
+   case GL_CLAMP:
+      return PIPE_TEX_WRAP_CLAMP;
+   case GL_CLAMP_TO_EDGE:
+      return PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   case GL_CLAMP_TO_BORDER:
+      return PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+   case GL_MIRRORED_REPEAT:
+      return PIPE_TEX_WRAP_MIRROR_REPEAT;
+   case GL_MIRROR_CLAMP_EXT:
+      return PIPE_TEX_WRAP_MIRROR_CLAMP;
+   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
+      return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE;
+   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
+      return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+static GLuint
+gl_filter_to_mip_filter(GLenum filter)
+{
+   switch (filter) {
+   case GL_NEAREST:
+   case GL_LINEAR:
+      return PIPE_TEX_MIPFILTER_NONE;
+
+   case GL_NEAREST_MIPMAP_NEAREST:
+   case GL_LINEAR_MIPMAP_NEAREST:
+      return PIPE_TEX_MIPFILTER_NEAREST;
+
+   case GL_NEAREST_MIPMAP_LINEAR:
+   case GL_LINEAR_MIPMAP_LINEAR:
+      return PIPE_TEX_MIPFILTER_LINEAR;
+
+   default:
+      assert(0);
+      return PIPE_TEX_MIPFILTER_NONE;
+   }
+}
+
+
+static GLuint
+gl_filter_to_img_filter(GLenum filter)
+{
+   switch (filter) {
+   case GL_NEAREST:
+   case GL_NEAREST_MIPMAP_NEAREST:
+   case GL_NEAREST_MIPMAP_LINEAR:
+      return PIPE_TEX_FILTER_NEAREST;
+
+   case GL_LINEAR:
+   case GL_LINEAR_MIPMAP_NEAREST:
+   case GL_LINEAR_MIPMAP_LINEAR:
+      return PIPE_TEX_FILTER_LINEAR;
+
+   default:
+      assert(0);
+      return PIPE_TEX_FILTER_NEAREST;
+   }
+}
+
+
+static void 
+update_samplers(struct st_context *st)
+{
+   struct gl_vertex_program *vprog = st->ctx->VertexProgram._Current;
+   struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current;
+   const GLbitfield samplersUsed = (vprog->Base.SamplersUsed |
+                                    fprog->Base.SamplersUsed);
+   GLuint su;
+
+   st->state.num_samplers = 0;
+
+   /* loop over sampler units (aka tex image units) */
+   for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) {
+      struct pipe_sampler_state *sampler = st->state.samplers + su;
+
+      memset(sampler, 0, sizeof(*sampler));
+
+      if (samplersUsed & (1 << su)) {
+         struct gl_texture_object *texobj;
+         GLuint texUnit;
+
+         if (fprog->Base.SamplersUsed & (1 << su))
+            texUnit = fprog->Base.SamplerUnits[su];
+         else
+            texUnit = vprog->Base.SamplerUnits[su];
+
+         texobj = st->ctx->Texture.Unit[texUnit]._Current;
+         if (!texobj) {
+            texobj = st_get_default_texture(st);
+         }
+
+         sampler->wrap_s = gl_wrap_to_sp(texobj->WrapS);
+         sampler->wrap_t = gl_wrap_to_sp(texobj->WrapT);
+         sampler->wrap_r = gl_wrap_to_sp(texobj->WrapR);
+
+         sampler->min_img_filter = gl_filter_to_img_filter(texobj->MinFilter);
+         sampler->min_mip_filter = gl_filter_to_mip_filter(texobj->MinFilter);
+         sampler->mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter);
+
+         if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB)
+            sampler->normalized_coords = 1;
+
+         sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias;
+         sampler->min_lod = MAX2(0.0f, texobj->MinLod);
+         sampler->max_lod = MIN2(texobj->MaxLevel - texobj->BaseLevel,
+                                 texobj->MaxLod);
+         if (sampler->max_lod < sampler->min_lod) {
+            /* The GL spec doesn't seem to specify what to do in this case.
+             * Swap the values.
+             */
+            float tmp = sampler->max_lod;
+            sampler->max_lod = sampler->min_lod;
+            sampler->min_lod = tmp;
+            assert(sampler->min_lod <= sampler->max_lod);
+         }
+
+         sampler->border_color[0] = texobj->BorderColor[RCOMP];
+         sampler->border_color[1] = texobj->BorderColor[GCOMP];
+         sampler->border_color[2] = texobj->BorderColor[BCOMP];
+         sampler->border_color[3] = texobj->BorderColor[ACOMP];
+
+	 sampler->max_anisotropy = texobj->MaxAnisotropy;
+         if (sampler->max_anisotropy > 1.0) {
+            sampler->min_img_filter = PIPE_TEX_FILTER_ANISO;
+            sampler->mag_img_filter = PIPE_TEX_FILTER_ANISO;
+         }
+
+         /* only care about ARB_shadow, not SGI shadow */
+         if (texobj->CompareMode == GL_COMPARE_R_TO_TEXTURE) {
+            sampler->compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE;
+            sampler->compare_func
+               = st_compare_func_to_pipe(texobj->CompareFunc);
+         }
+
+         st->state.num_samplers = su + 1;
+
+         /*printf("%s su=%u non-null\n", __FUNCTION__, su);*/
+         cso_single_sampler(st->cso_context, su, sampler);
+      }
+      else {
+         /*printf("%s su=%u null\n", __FUNCTION__, su);*/
+         cso_single_sampler(st->cso_context, su, NULL);
+      }
+   }
+
+   cso_single_sampler_done(st->cso_context);
+}
+
+
+const struct st_tracked_state st_update_sampler = {
+   "st_update_sampler",					/* name */
+   {							/* dirty */
+      _NEW_TEXTURE,					/* mesa */
+      0,						/* st */
+   },
+   update_samplers					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c
new file mode 100644
index 00000000000..3fd59e1945c
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_scissor.c
@@ -0,0 +1,92 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+ 
+
+#include "st_context.h"
+#include "pipe/p_context.h"
+#include "st_atom.h"
+
+
+/**
+ * Scissor depends on the scissor box, and the framebuffer dimensions.
+ */
+static void
+update_scissor( struct st_context *st )
+{
+   struct pipe_scissor_state scissor;
+   const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   GLint miny, maxy;
+
+   scissor.minx = 0;
+   scissor.miny = 0;
+   scissor.maxx = fb->Width;
+   scissor.maxy = fb->Height;
+
+   if (st->ctx->Scissor.Enabled) {
+      if ((GLuint)st->ctx->Scissor.X > scissor.minx)
+         scissor.minx = st->ctx->Scissor.X;
+      if ((GLuint)st->ctx->Scissor.Y > scissor.miny)
+         scissor.miny = st->ctx->Scissor.Y;
+
+      if ((GLuint)st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx)
+         scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width;
+      if ((GLuint)st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy)
+         scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height;
+
+      /* check for null space */
+      if (scissor.minx >= scissor.maxx || scissor.miny >= scissor.maxy)
+         scissor.minx = scissor.miny = scissor.maxx = scissor.maxy = 0;
+   }
+
+   /* Now invert Y.  Pipe drivers use the convention Y=0=top for surfaces
+    */
+   miny = fb->Height - scissor.maxy;
+   maxy = fb->Height - scissor.miny;
+   scissor.miny = miny;
+   scissor.maxy = maxy;
+
+   if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) {
+      /* state has changed */
+      st->state.scissor = scissor;  /* struct copy */
+      st->pipe->set_scissor_state(st->pipe, &scissor); /* activate */
+   }
+}
+
+
+const struct st_tracked_state st_update_scissor = {
+   "st_update_scissor",					/* name */
+   {							/* dirty */
+      (_NEW_SCISSOR | _NEW_BUFFERS),			/* mesa */
+      0,						/* st */
+   },
+   update_scissor					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
new file mode 100644
index 00000000000..cbd414e2d3b
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -0,0 +1,369 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+/**
+ * State validation for vertex/fragment shaders.
+ * Note that we have to delay most vertex/fragment shader translation
+ * until rendering time since the linkage between the vertex outputs and
+ * fragment inputs can vary depending on the pairing of shaders.
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "shader/program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_simple_shaders.h"
+
+#include "cso_cache/cso_context.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_program.h"
+#include "st_atom_shader.h"
+#include "st_mesa_to_tgsi.h"
+
+
+/**
+ * This represents a vertex program, especially translated to match
+ * the inputs of a particular fragment shader.
+ */
+struct translated_vertex_program
+{
+   struct st_vertex_program *master;
+
+   /** The fragment shader "signature" this vertex shader is meant for: */
+   GLbitfield frag_inputs;
+
+   /** Compared against master vertex program's serialNo: */
+   GLuint serialNo;
+
+   /** Maps VERT_RESULT_x to slot */
+   GLuint output_to_slot[VERT_RESULT_MAX];
+   ubyte output_to_semantic_name[VERT_RESULT_MAX];
+   ubyte output_to_semantic_index[VERT_RESULT_MAX];
+
+   /** Pointer to the translated vertex program */
+   struct st_vertex_program *vp;
+
+   struct translated_vertex_program *next;  /**< next in linked list */
+};
+
+
+
+/**
+ * Given a vertex program output attribute, return the corresponding
+ * fragment program input attribute.
+ * \return -1 for vertex outputs that have no corresponding fragment input
+ */
+static GLint
+vp_out_to_fp_in(GLuint vertResult)
+{
+   if (vertResult >= VERT_RESULT_TEX0 &&
+       vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS)
+      return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0);
+
+   if (vertResult >= VERT_RESULT_VAR0 &&
+       vertResult < VERT_RESULT_VAR0 + MAX_VARYING)
+      return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0);
+
+   switch (vertResult) {
+   case VERT_RESULT_HPOS:
+      return FRAG_ATTRIB_WPOS;
+   case VERT_RESULT_COL0:
+      return FRAG_ATTRIB_COL0;
+   case VERT_RESULT_COL1:
+      return FRAG_ATTRIB_COL1;
+   case VERT_RESULT_FOGC:
+      return FRAG_ATTRIB_FOGC;
+   default:
+      /* Back-face colors, edge flags, etc */
+      return -1;
+   }
+}
+
+
+/**
+ * Find a translated vertex program that corresponds to stvp and
+ * has outputs matched to stfp's inputs.
+ * This performs vertex and fragment translation (to TGSI) when needed.
+ */
+static struct translated_vertex_program *
+find_translated_vp(struct st_context *st,
+                   struct st_vertex_program *stvp,
+                   struct st_fragment_program *stfp)
+{
+   static const GLuint UNUSED = ~0;
+   struct translated_vertex_program *xvp;
+   const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead;
+
+   /*
+    * Translate fragment program if needed.
+    */
+   if (!stfp->state.tokens) {
+      GLuint inAttr, numIn = 0;
+
+      for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) {
+         if (fragInputsRead & (1 << inAttr)) {
+            stfp->input_to_slot[inAttr] = numIn;
+            numIn++;
+         }
+         else {
+            stfp->input_to_slot[inAttr] = UNUSED;
+         }
+      }
+
+      stfp->num_input_slots = numIn;
+
+      assert(stfp->Base.Base.NumInstructions > 1);
+
+      st_translate_fragment_program(st, stfp, stfp->input_to_slot);
+   }
+
+
+   /* See if we've got a translated vertex program whose outputs match
+    * the fragment program's inputs.
+    * XXX This could be a hash lookup, using InputsRead as the key.
+    */
+   for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) {
+      if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) {
+         break;
+      }
+   }
+
+   /* No?  Allocate translated vp object now */
+   if (!xvp) {
+      xvp = CALLOC_STRUCT(translated_vertex_program);
+      xvp->frag_inputs = fragInputsRead;
+      xvp->master = stvp;
+
+      xvp->next = stfp->vertex_programs;
+      stfp->vertex_programs = xvp;
+   }
+
+   /* See if we need to translate vertex program to TGSI form */
+   if (xvp->serialNo != stvp->serialNo) {
+      GLuint outAttr, dummySlot;
+      const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
+      GLuint numVpOuts = 0;
+      GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE;
+      GLint maxGeneric;
+
+      /* Compute mapping of vertex program outputs to slots, which depends
+       * on the fragment program's input->slot mapping.
+       */
+      for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
+         /* set defaults: */
+         xvp->output_to_slot[outAttr] = UNUSED;
+         xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_COUNT;
+         xvp->output_to_semantic_index[outAttr] = 99;
+
+         if (outAttr == VERT_RESULT_HPOS) {
+            /* always put xformed position into slot zero */
+            xvp->output_to_slot[VERT_RESULT_HPOS] = 0;
+            xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION;
+            xvp->output_to_semantic_index[outAttr] = 0;
+            numVpOuts++;
+         }
+         else if (outputsWritten & (1 << outAttr)) {
+            /* see if the frag prog wants this vert output */
+            GLint fpInAttrib = vp_out_to_fp_in(outAttr);
+            if (fpInAttrib >= 0) {
+               GLuint fpInSlot = stfp->input_to_slot[fpInAttrib];
+               if (fpInSlot != ~0) {
+                  /* match this vp output to the fp input */
+                  GLuint vpOutSlot = stfp->input_map[fpInSlot];
+                  xvp->output_to_slot[outAttr] = vpOutSlot;
+                  xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot];
+                  xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot];
+                  numVpOuts++;
+               }
+            }
+            else if (outAttr == VERT_RESULT_PSIZ)
+               emitPntSize = GL_TRUE;
+            else if (outAttr == VERT_RESULT_BFC0)
+               emitBFC0 = GL_TRUE;
+            else if (outAttr == VERT_RESULT_BFC1)
+               emitBFC1 = GL_TRUE;
+         }
+#if 0 /*debug*/
+         printf("assign vp output_to_slot[%d] = %d\n", outAttr, 
+                xvp->output_to_slot[outAttr]);
+#endif
+      }
+
+      /* must do these last */
+      if (emitPntSize) {
+         xvp->output_to_slot[VERT_RESULT_PSIZ] = numVpOuts++;
+         xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE;
+         xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0;
+      }
+      if (emitBFC0) {
+         xvp->output_to_slot[VERT_RESULT_BFC0] = numVpOuts++;
+         xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
+         xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0;
+      }
+      if (emitBFC1) {
+         xvp->output_to_slot[VERT_RESULT_BFC1] = numVpOuts++;
+         xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
+         xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 1;
+      }
+
+      /* Unneeded vertex program outputs will go to this slot.
+       * We could use this info to do dead code elimination in the
+       * vertex program.
+       */
+      dummySlot = numVpOuts;
+
+      /* find max GENERIC slot index */
+      maxGeneric = -1;
+      for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
+         if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) {
+            maxGeneric = MAX2(maxGeneric,
+                              xvp->output_to_semantic_index[outAttr]);
+         }
+      }
+
+      /* Map vert program outputs that aren't used to the dummy slot
+       * (and an unused generic attribute slot).
+       */
+      for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
+         if (outputsWritten & (1 << outAttr)) {
+            if (xvp->output_to_slot[outAttr] == UNUSED) {
+               xvp->output_to_slot[outAttr] = dummySlot;
+               xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC;
+               xvp->output_to_semantic_index[outAttr] = maxGeneric + 1;
+            }
+         }
+
+#if 0 /*debug*/
+         printf("vp output_to_slot[%d] = %d\n", outAttr, 
+                xvp->output_to_slot[outAttr]);
+#endif
+      }
+
+      assert(stvp->Base.Base.NumInstructions > 1);
+
+      st_translate_vertex_program(st, stvp, xvp->output_to_slot,
+                                  xvp->output_to_semantic_name,
+                                  xvp->output_to_semantic_index);
+
+      xvp->vp = stvp;
+
+      /* translated VP is up to date now */
+      xvp->serialNo = stvp->serialNo;
+   }
+
+   return xvp;
+}
+
+
+void
+st_free_translated_vertex_programs(struct st_context *st,
+                                   struct translated_vertex_program *xvp)
+{
+   struct translated_vertex_program *next;
+
+   while (xvp) {
+      next = xvp->next;
+      free(xvp);
+      xvp = next;
+   }
+}
+
+
+static void *
+get_passthrough_fs(struct st_context *st)
+{
+   struct pipe_shader_state shader;
+
+   if (!st->passthrough_fs) {
+      st->passthrough_fs =
+         util_make_fragment_passthrough_shader(st->pipe, &shader);
+#if 0      /* We actually need to keep the tokens around at this time */
+      free((void *) shader.tokens);
+#endif
+   }
+
+   return st->passthrough_fs;
+}
+
+
+static void
+update_linkage( struct st_context *st )
+{
+   struct st_vertex_program *stvp;
+   struct st_fragment_program *stfp;
+   struct translated_vertex_program *xvp;
+
+   /* find active shader and params -- Should be covered by
+    * ST_NEW_VERTEX_PROGRAM
+    */
+   assert(st->ctx->VertexProgram._Current);
+   stvp = st_vertex_program(st->ctx->VertexProgram._Current);
+   assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB);
+
+   assert(st->ctx->FragmentProgram._Current);
+   stfp = st_fragment_program(st->ctx->FragmentProgram._Current);
+   assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB);
+
+   xvp = find_translated_vp(st, stvp, stfp);
+
+   st_reference_vertprog(st, &st->vp, stvp);
+   st_reference_fragprog(st, &st->fp, stfp);
+
+   cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader);
+
+   if (st->missing_textures) {
+      /* use a pass-through frag shader that uses no textures */
+      void *fs = get_passthrough_fs(st);
+      cso_set_fragment_shader_handle(st->cso_context, fs);
+   }
+   else {
+      cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader);
+   }
+
+   st->vertex_result_to_slot = xvp->output_to_slot;
+}
+
+
+const struct st_tracked_state st_update_shader = {
+   "st_update_shader",					/* name */
+   {							/* dirty */
+      0,						/* mesa */
+      ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM	/* st */
+   },
+   update_linkage					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h
new file mode 100644
index 00000000000..8403bc66c92
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_shader.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_ATOM_SHADER_H
+#define ST_ATOM_SHADER_H
+
+
+extern void
+st_free_translated_vertex_programs(struct st_context *st,
+                                   struct translated_vertex_program *xvp);
+
+
+#endif /* ST_ATOM_SHADER_H */
diff --git a/src/mesa/state_tracker/st_atom_stipple.c b/src/mesa/state_tracker/st_atom_stipple.c
new file mode 100644
index 00000000000..f395930ab40
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_stipple.c
@@ -0,0 +1,63 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * \brief polygon stipple state
+  *
+  * Authors:
+  *   Brian Paul
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+
+
+static void 
+update_stipple( struct st_context *st )
+{
+   const GLuint sz = sizeof(st->state.poly_stipple.stipple);
+   assert(sz == sizeof(st->ctx->PolygonStipple));
+
+   if (memcmp(&st->state.poly_stipple.stipple, st->ctx->PolygonStipple, sz)) {
+      /* state has changed */
+      memcpy(st->state.poly_stipple.stipple, st->ctx->PolygonStipple, sz);
+      st->pipe->set_polygon_stipple(st->pipe, &st->state.poly_stipple);
+   }
+}
+
+
+const struct st_tracked_state st_update_polygon_stipple = {
+   "st_update_polygon_stipple",				/* name */
+   {							/* dirty */
+      (_NEW_POLYGONSTIPPLE),				/* mesa */
+      0,						/* st */
+   },
+   update_stipple					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c
new file mode 100644
index 00000000000..21f7321f97a
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_texture.c
@@ -0,0 +1,162 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_texture.h"
+#include "st_cb_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "cso_cache/cso_context.h"
+
+
+static void 
+update_textures(struct st_context *st)
+{
+   struct gl_vertex_program *vprog = st->ctx->VertexProgram._Current;
+   struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current;
+   const GLbitfield samplersUsed = (vprog->Base.SamplersUsed |
+                                    fprog->Base.SamplersUsed);
+   GLuint su;
+
+   st->state.num_textures = 0;
+
+   /* loop over sampler units (aka tex image units) */
+   for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) {
+      struct pipe_texture *pt = NULL;
+
+      if (samplersUsed & (1 << su)) {
+         struct gl_texture_object *texObj;
+         struct st_texture_object *stObj;
+         GLboolean flush, retval;
+         GLuint texUnit;
+
+         if (fprog->Base.SamplersUsed & (1 << su))
+            texUnit = fprog->Base.SamplerUnits[su];
+         else
+            texUnit = vprog->Base.SamplerUnits[su];
+
+         texObj = st->ctx->Texture.Unit[texUnit]._Current;
+
+         if (!texObj) {
+            texObj = st_get_default_texture(st);
+         }
+         stObj = st_texture_object(texObj);
+
+         retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush);
+         if (!retval) {
+            /* out of mem */
+            continue;
+         }
+
+         st->state.num_textures = su + 1;
+
+         pt = st_get_stobj_texture(stObj);
+      }
+
+      /*
+      if (pt) {
+         printf("%s su=%u non-null\n", __FUNCTION__, su);
+      }
+      else {
+         printf("%s su=%u null\n", __FUNCTION__, su);
+      }
+      */
+
+      pipe_texture_reference(&st->state.sampler_texture[su], pt);
+   }
+
+   cso_set_sampler_textures(st->cso_context,
+                            st->state.num_textures,
+                            st->state.sampler_texture);
+}
+
+
+const struct st_tracked_state st_update_texture = {
+   "st_update_texture",					/* name */
+   {							/* dirty */
+      _NEW_TEXTURE,					/* mesa */
+      ST_NEW_FRAGMENT_PROGRAM,				/* st */
+   },
+   update_textures					/* update */
+};
+
+
+
+
+static void 
+finalize_textures(struct st_context *st)
+{
+   struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current;
+   const GLboolean prev_missing_textures = st->missing_textures;
+   GLuint su;
+
+   st->missing_textures = GL_FALSE;
+
+   for (su = 0; su < st->ctx->Const.MaxTextureCoordUnits; su++) {
+      if (fprog->Base.SamplersUsed & (1 << su)) {
+         const GLuint texUnit = fprog->Base.SamplerUnits[su];
+         struct gl_texture_object *texObj
+            = st->ctx->Texture.Unit[texUnit]._Current;
+         struct st_texture_object *stObj = st_texture_object(texObj);
+
+         if (texObj) {
+            GLboolean flush, retval;
+
+            retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush);
+            if (!retval) {
+               /* out of mem */
+               st->missing_textures = GL_TRUE;
+               continue;
+            }
+
+            stObj->teximage_realloc = TRUE;
+         }
+      }
+   }
+
+   if (prev_missing_textures != st->missing_textures)
+      st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+}
+
+
+
+const struct st_tracked_state st_finalize_textures = {
+   "st_finalize_textures",		/* name */
+   {					/* dirty */
+      _NEW_TEXTURE,			/* mesa */
+      0,				/* st */
+   },
+   finalize_textures			/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c
new file mode 100644
index 00000000000..27ec2eb0331
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_viewport.c
@@ -0,0 +1,91 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "main/context.h"
+#include "main/colormac.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "cso_cache/cso_context.h"
+
+/**
+ * Update the viewport transformation matrix.  Depends on:
+ *  - viewport pos/size
+ *  - depthrange
+ *  - window pos/size or FBO size
+ */
+static void
+update_viewport( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+   GLfloat yScale, yBias;
+
+   /* _NEW_BUFFERS
+    */
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      yScale = -1;
+      yBias = (GLfloat)ctx->DrawBuffer->Height;
+   }
+   else {
+      yScale = 1.0;
+      yBias = 0.0;
+   }
+
+   /* _NEW_VIEWPORT 
+    */
+   {
+      GLfloat x = (GLfloat)ctx->Viewport.X;
+      GLfloat y = (GLfloat)ctx->Viewport.Y;
+      GLfloat z = ctx->Viewport.Near;
+      GLfloat half_width = (GLfloat)ctx->Viewport.Width / 2.0f;
+      GLfloat half_height = (GLfloat)ctx->Viewport.Height / 2.0f;
+      GLfloat half_depth = (GLfloat)(ctx->Viewport.Far - ctx->Viewport.Near) / 2.0f;
+      
+      st->state.viewport.scale[0] = half_width;
+      st->state.viewport.scale[1] = half_height * yScale;
+      st->state.viewport.scale[2] = half_depth;
+      st->state.viewport.scale[3] = 1.0;
+
+      st->state.viewport.translate[0] = half_width + x;
+      st->state.viewport.translate[1] = (half_height + y) * yScale + yBias;
+      st->state.viewport.translate[2] = half_depth + z;
+      st->state.viewport.translate[3] = 0.0;
+
+      cso_set_viewport(st->cso_context, &st->state.viewport);
+   }
+}
+
+
+const struct st_tracked_state st_update_viewport = {
+   "st_update_viewport",				/* name */
+   {							/* dirty */
+      _NEW_BUFFERS | _NEW_VIEWPORT,			/* mesa */
+      0,						/* st */
+   },
+   update_viewport					/* update */
+};
diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h
new file mode 100644
index 00000000000..b81de316ec9
--- /dev/null
+++ b/src/mesa/state_tracker/st_cache.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ *
+ * Copyright 2007 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.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin <[email protected]>
+  */
+
+#ifndef ST_CACHE_H
+#define ST_CACHE_H
+
+#include "cso_cache/cso_cache.h"
+
+struct pipe_blend_state;
+struct pipe_sampler_state;
+struct st_context;
+
+
+const struct cso_blend *
+st_cached_blend_state(struct st_context *st,
+                      const struct pipe_blend_state *blend);
+
+const struct cso_sampler *
+st_cached_sampler_state(struct st_context *st,
+                        const struct pipe_sampler_state *sampler);
+
+const struct cso_depth_stencil_alpha *
+st_cached_depth_stencil_alpha_state(struct st_context *st,
+                              const struct pipe_depth_stencil_alpha_state *depth_stencil);
+
+const struct cso_rasterizer *
+st_cached_rasterizer_state(struct st_context *st,
+                           const struct pipe_rasterizer_state *raster);
+
+const struct cso_fragment_shader *
+st_cached_fs_state(struct st_context *st,
+                   const struct pipe_shader_state *templ);
+
+
+const struct cso_vertex_shader *
+st_cached_vs_state(struct st_context *st,
+                   const struct pipe_shader_state *templ);
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
new file mode 100644
index 00000000000..a4e72b48ed4
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -0,0 +1,362 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+
+#include "st_context.h"
+#include "st_cb_accum.h"
+#include "st_cb_fbo.h"
+#include "st_draw.h"
+#include "st_public.h"
+#include "st_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+
+#define UNCLAMPED_FLOAT_TO_SHORT(us, f)  \
+   us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) )
+
+
+/**
+ * For hardware that supports deep color buffers, we could accelerate
+ * most/all the accum operations with blending/texturing.
+ * For now, just use the get/put_tile() functions and do things in software.
+ */
+
+
+/**
+ * Wrapper for pipe_get_tile_rgba().  Do format/cpp override to make the
+ * tile util function think the surface is 16bit/channel, even if it's not.
+ * See also: st_renderbuffer_alloc_storage()
+ */
+static void
+acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps,
+                  uint x, uint y, uint w, uint h, float *p)
+{
+   const enum pipe_format f = acc_ps->format;
+   const struct pipe_format_block b = acc_ps->block;
+
+   acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT;
+   acc_ps->block.size = 8;
+   acc_ps->block.width = 1;
+   acc_ps->block.height = 1;
+
+   pipe_get_tile_rgba(acc_ps, x, y, w, h, p);
+
+   acc_ps->format = f;
+   acc_ps->block = b;
+}
+
+
+/**
+ * Wrapper for pipe_put_tile_rgba().  Do format/cpp override to make the
+ * tile util function think the surface is 16bit/channel, even if it's not.
+ * See also: st_renderbuffer_alloc_storage()
+ */
+static void
+acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps,
+                  uint x, uint y, uint w, uint h, const float *p)
+{
+   enum pipe_format f = acc_ps->format;
+   const struct pipe_format_block b = acc_ps->block;
+
+   acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT;
+   acc_ps->block.size = 8;
+   acc_ps->block.width = 1;
+   acc_ps->block.height = 1;
+
+   pipe_put_tile_rgba(acc_ps, x, y, w, h, p);
+
+   acc_ps->format = f;
+   acc_ps->block = b;
+}
+
+
+
+void
+st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *acc_strb = st_renderbuffer(rb);
+   struct pipe_surface *acc_ps;
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   const GLint xpos = ctx->DrawBuffer->_Xmin;
+   const GLint ypos = ctx->DrawBuffer->_Ymin;
+   const GLint width = ctx->DrawBuffer->_Xmax - xpos;
+   const GLint height = ctx->DrawBuffer->_Ymax - ypos;
+   GLubyte *map;
+
+   acc_ps = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                    PIPE_BUFFER_USAGE_CPU_WRITE);
+   map = screen->surface_map(screen, acc_ps,
+                             PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* note acc_strb->format might not equal acc_ps->format */
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         GLshort r = FLOAT_TO_SHORT(ctx->Accum.ClearColor[0]);
+         GLshort g = FLOAT_TO_SHORT(ctx->Accum.ClearColor[1]);
+         GLshort b = FLOAT_TO_SHORT(ctx->Accum.ClearColor[2]);
+         GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]);
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *dst = (GLshort *) (map + (ypos + i) * acc_ps->stride + xpos * 8);
+            for (j = 0; j < width; j++) {
+               dst[0] = r;
+               dst[1] = g;
+               dst[2] = b;
+               dst[3] = a;
+               dst += 4;
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()");
+   }
+
+   screen->surface_unmap(screen, acc_ps);
+   pipe_surface_reference(&acc_ps, NULL);
+}
+
+
+/** For ADD/MULT */
+static void
+accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias,
+          GLint xpos, GLint ypos, GLint width, GLint height,
+          struct st_renderbuffer *acc_strb)
+{
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct pipe_surface *acc_ps = acc_strb->surface;
+   GLubyte *map;
+
+   map = screen->surface_map(screen, acc_ps, 
+                             PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* note acc_strb->format might not equal acc_ps->format */
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *acc = (GLshort *) (map + (ypos + i) * acc_ps->stride + xpos * 8);
+            for (j = 0; j < width * 4; j++) {
+               float val = SHORT_TO_FLOAT(acc[j]) * scale + bias;
+               acc[j] = FLOAT_TO_SHORT(val);
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
+   }
+
+   screen->surface_unmap(screen, acc_ps);
+}
+
+
+static void
+accum_accum(struct pipe_context *pipe, GLfloat value,
+            GLint xpos, GLint ypos, GLint width, GLint height,
+            struct st_renderbuffer *acc_strb,
+            struct st_renderbuffer *color_strb)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *acc_surf, *color_surf;
+   GLfloat *colorBuf, *accBuf;
+   GLint i;
+
+   acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                      (PIPE_BUFFER_USAGE_CPU_WRITE |
+                                       PIPE_BUFFER_USAGE_CPU_READ));
+
+   color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+
+   colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+   accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+   pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, colorBuf);
+   acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf);
+
+   for (i = 0; i < 4 * width * height; i++) {
+      accBuf[i] = accBuf[i] + colorBuf[i] * value;
+   }
+
+   acc_put_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf);
+
+   free(colorBuf);
+   free(accBuf);
+   pipe_surface_reference(&acc_surf, NULL);
+   pipe_surface_reference(&color_surf, NULL);
+}
+
+
+static void
+accum_load(struct pipe_context *pipe, GLfloat value,
+           GLint xpos, GLint ypos, GLint width, GLint height,
+           struct st_renderbuffer *acc_strb,
+           struct st_renderbuffer *color_strb)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *acc_surf, *color_surf;
+   GLfloat *buf;
+   GLint i;
+
+   acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                      PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+
+   buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+   pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, buf);
+
+   for (i = 0; i < 4 * width * height; i++) {
+      buf[i] = buf[i] * value;
+   }
+
+   acc_put_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, buf);
+
+   free(buf);
+   pipe_surface_reference(&acc_surf, NULL);
+   pipe_surface_reference(&color_surf, NULL);
+}
+
+
+static void
+accum_return(GLcontext *ctx, GLfloat value,
+             GLint xpos, GLint ypos, GLint width, GLint height,
+             struct st_renderbuffer *acc_strb,
+             struct st_renderbuffer *color_strb)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const GLubyte *colormask = ctx->Color.ColorMask;
+   struct pipe_surface *acc_surf, *color_surf;
+   GLfloat *abuf, *cbuf = NULL;
+   GLint i, ch;
+
+   abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+   acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                      PIPE_BUFFER_USAGE_CPU_READ);
+
+   color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0,
+                                        (PIPE_BUFFER_USAGE_CPU_READ |
+                                         PIPE_BUFFER_USAGE_CPU_WRITE));
+
+   acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, abuf);
+
+   if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) {
+      cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+      pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, cbuf);
+   }
+
+   for (i = 0; i < width * height; i++) {
+      for (ch = 0; ch < 4; ch++) {
+         if (colormask[ch]) {
+            GLfloat val = abuf[i * 4 + ch] * value;
+            abuf[i * 4 + ch] = CLAMP(val, 0.0f, 1.0f);
+         }
+         else {
+            abuf[i * 4 + ch] = cbuf[i * 4 + ch];
+         }
+      }
+   }
+
+   pipe_put_tile_rgba(color_surf, xpos, ypos, width, height, abuf);
+
+   free(abuf);
+   if (cbuf)
+      free(cbuf);
+   pipe_surface_reference(&acc_surf, NULL);
+   pipe_surface_reference(&color_surf, NULL);
+}
+
+
+static void
+st_Accum(GLcontext *ctx, GLenum op, GLfloat value)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct st_renderbuffer *acc_strb
+     = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
+   struct st_renderbuffer *color_strb
+      = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer);
+
+   const GLint xpos = ctx->DrawBuffer->_Xmin;
+   const GLint ypos = ctx->DrawBuffer->_Ymin;
+   const GLint width = ctx->DrawBuffer->_Xmax - xpos;
+   const GLint height = ctx->DrawBuffer->_Ymax - ypos;
+
+   /* make sure color bufs aren't cached */
+   st_flush( st, PIPE_FLUSH_RENDER_CACHE, NULL );
+
+   switch (op) {
+   case GL_ADD:
+      if (value != 0.0F) {
+         accum_mad(ctx, 1.0, value, xpos, ypos, width, height, acc_strb);
+      }
+      break;
+   case GL_MULT:
+      if (value != 1.0F) {
+         accum_mad(ctx, value, 0.0, xpos, ypos, width, height, acc_strb);
+      }
+      break;
+   case GL_ACCUM:
+      if (value != 0.0F) {
+         accum_accum(pipe, value, xpos, ypos, width, height, acc_strb, color_strb);
+      }
+      break;
+   case GL_LOAD:
+      accum_load(pipe, value, xpos, ypos, width, height, acc_strb, color_strb);
+      break;
+   case GL_RETURN:
+      accum_return(ctx, value, xpos, ypos, width, height, acc_strb, color_strb);
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+
+void st_init_accum_functions(struct dd_function_table *functions)
+{
+   functions->Accum = st_Accum;
+}
diff --git a/src/mesa/state_tracker/st_cb_accum.h b/src/mesa/state_tracker/st_cb_accum.h
new file mode 100644
index 00000000000..ed9b7dab943
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_accum.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_ACCUM_H
+#define ST_CB_ACCUM_H
+
+
+extern void
+st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb);
+
+extern void st_init_accum_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_ACCUM_H */
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
new file mode 100644
index 00000000000..f14e562400b
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -0,0 +1,833 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_program.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_program.h"
+#include "st_mesa_to_tgsi.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_draw_quad.h"
+#include "util/u_simple_shaders.h"
+#include "shader/prog_instruction.h"
+#include "cso_cache/cso_context.h"
+
+
+
+/**
+ * glBitmaps are drawn as textured quads.  The user's bitmap pattern
+ * is stored in a texture image.  An alpha8 texture format is used.
+ * The fragment shader samples a bit (texel) from the texture, then
+ * discards the fragment if the bit is off.
+ *
+ * Note that we actually store the inverse image of the bitmap to
+ * simplify the fragment program.  An "on" bit gets stored as texel=0x0
+ * and an "off" bit is stored as texel=0xff.  Then we kill the
+ * fragment if the negated texel value is less than zero.
+ */
+
+
+/**
+ * The bitmap cache attempts to accumulate multiple glBitmap calls in a
+ * buffer which is then rendered en mass upon a flush, state change, etc.
+ * A wide, short buffer is used to target the common case of a series
+ * of glBitmap calls being used to draw text.
+ */
+static GLboolean UseBitmapCache = GL_TRUE;
+
+
+#define BITMAP_CACHE_WIDTH  512
+#define BITMAP_CACHE_HEIGHT 32
+
+struct bitmap_cache
+{
+   /** Window pos to render the cached image */
+   GLint xpos, ypos;
+   /** Bounds of region used in window coords */
+   GLint xmin, ymin, xmax, ymax;
+
+   GLfloat color[4];
+
+   struct pipe_texture *texture;
+   struct pipe_surface *surf;
+
+   GLboolean empty;
+
+   /** An I8 texture image: */
+   ubyte *buffer;
+};
+
+
+
+
+/**
+ * Make fragment program for glBitmap:
+ *   Sample the texture and kill the fragment if the bit is 0.
+ * This program will be combined with the user's fragment program.
+ */
+static struct st_fragment_program *
+make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex)
+{
+   struct st_fragment_program *stfp;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   p->NumInstructions = 3;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+
+   /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */
+   p->Instructions[ic].Opcode = OPCODE_TEX;
+   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].DstReg.Index = 0;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   p->Instructions[ic].TexSrcUnit = samplerIndex;
+   p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+
+   /* KIL if -tmp0 < 0 # texel=0 -> keep / texel=0 -> discard */
+   p->Instructions[ic].Opcode = OPCODE_KIL;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].SrcReg[0].Index = 0;
+   p->Instructions[ic].SrcReg[0].NegateBase = NEGATE_XYZW;
+   ic++;
+
+   /* END; */
+   p->Instructions[ic++].Opcode = OPCODE_END;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = FRAG_BIT_TEX0;
+   p->OutputsWritten = 0x0;
+   p->SamplersUsed = (1 << samplerIndex);
+
+   stfp = (struct st_fragment_program *) p;
+   stfp->Base.UsesKill = GL_TRUE;
+   st_translate_fragment_program(ctx->st, stfp, NULL);
+
+   return stfp;
+}
+
+
+static int
+find_free_bit(uint bitfield)
+{
+   int i;
+   for (i = 0; i < 32; i++) {
+      if ((bitfield & (1 << i)) == 0) {
+         return i;
+      }
+   }
+   return -1;
+}
+
+
+/**
+ * Combine basic bitmap fragment program with the user-defined program.
+ */
+static struct st_fragment_program *
+combined_bitmap_fragment_program(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+   struct st_fragment_program *stfp = st->fp;
+
+   if (!stfp->bitmap_program) {
+      /*
+       * Generate new program which is the user-defined program prefixed
+       * with the bitmap sampler/kill instructions.
+       */
+      struct st_fragment_program *bitmap_prog;
+      uint sampler;
+
+      sampler = find_free_bit(st->fp->Base.Base.SamplersUsed);
+      bitmap_prog = make_bitmap_fragment_program(ctx, sampler);
+
+      stfp->bitmap_program = (struct st_fragment_program *)
+         _mesa_combine_programs(ctx,
+                                &bitmap_prog->Base.Base, &stfp->Base.Base);
+      stfp->bitmap_program->bitmap_sampler = sampler;
+
+      /* done with this after combining */
+      st_reference_fragprog(st, &bitmap_prog, NULL);
+
+#if 0
+      {
+         struct gl_program *p = &stfp->bitmap_program->Base.Base;
+         printf("Combined bitmap program:\n");
+         _mesa_print_program(p);
+         printf("InputsRead: 0x%x\n", p->InputsRead);
+         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
+         _mesa_print_parameter_list(p->Parameters);
+      }
+#endif
+
+      /* translate to TGSI tokens */
+      st_translate_fragment_program(st, stfp->bitmap_program, NULL);
+   }
+
+   return stfp->bitmap_program;
+}
+
+
+/**
+ * Copy user-provide bitmap bits into texture buffer, expanding
+ * bits into texels.
+ * "On" bits will set texels to 0xff.
+ * "Off" bits will not modify texels.
+ * Note that the image is actually going to be upside down in
+ * the texture.  We deal with that with texcoords.
+ */
+static void
+unpack_bitmap(struct st_context *st,
+              GLint px, GLint py, GLsizei width, GLsizei height,
+              const struct gl_pixelstore_attrib *unpack,
+              const GLubyte *bitmap,
+              ubyte *destBuffer, uint destStride)
+{
+   GLint row, col;
+
+#define SET_PIXEL(COL, ROW) \
+   destBuffer[(py + (ROW)) * destStride + px + (COL)] = 0x0;
+
+   for (row = 0; row < height; row++) {
+      const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack,
+                 bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0);
+
+      if (unpack->LsbFirst) {
+         /* Lsb first */
+         GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            if (*src & mask) {
+               SET_PIXEL(col, row);
+            }
+
+            if (mask == 128U) {
+               src++;
+               mask = 1U;
+            }
+            else {
+               mask = mask << 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 1)
+            src++;
+      }
+      else {
+         /* Msb first */
+         GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            if (*src & mask) {
+               SET_PIXEL(col, row);
+            }
+
+            if (mask == 1U) {
+               src++;
+               mask = 128U;
+            }
+            else {
+               mask = mask >> 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 128)
+            src++;
+      }
+
+   } /* row */
+
+#undef SET_PIXEL
+}
+
+
+/**
+ * Create a texture which represents a bitmap image.
+ */
+static struct pipe_texture *
+make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLubyte *bitmap)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surface;
+   ubyte *dest;
+   struct pipe_texture *pt;
+
+   /* PBO source... */
+   bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap);
+   if (!bitmap) {
+      return NULL;
+   }
+
+   /**
+    * Create texture to hold bitmap pattern.
+    */
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format,
+                          0, width, height, 1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   if (!pt) {
+      _mesa_unmap_bitmap_pbo(ctx, unpack);
+      return NULL;
+   }
+
+   surface = screen->get_tex_surface(screen, pt, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* map texture surface */
+   dest = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* Put image into texture surface */
+   memset(dest, 0xff, height * surface->stride);
+   unpack_bitmap(ctx->st, 0, 0, width, height, unpack, bitmap,
+                 dest, surface->stride);
+
+   _mesa_unmap_bitmap_pbo(ctx, unpack);
+
+   /* Release surface */
+   screen->surface_unmap(screen, surface);
+   pipe_surface_reference(&surface, NULL);
+
+   return pt;
+}
+
+static GLuint
+setup_bitmap_vertex_data(struct st_context *st,
+                         int x, int y, int width, int height,
+                         float z, const float color[4])
+{
+   struct pipe_context *pipe = st->pipe;
+   const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   const GLfloat fb_width = (GLfloat)fb->Width;
+   const GLfloat fb_height = (GLfloat)fb->Height;
+   const GLfloat x0 = (GLfloat)x;
+   const GLfloat x1 = (GLfloat)(x + width);
+   const GLfloat y0 = (GLfloat)y;
+   const GLfloat y1 = (GLfloat)(y + height);
+   const GLfloat sLeft = (GLfloat)0.0, sRight = (GLfloat)1.0;
+   const GLfloat tTop = (GLfloat)0.0, tBot = (GLfloat)1.0 - tTop;
+   const GLfloat clip_x0 = (GLfloat)(x0 / fb_width * 2.0 - 1.0);
+   const GLfloat clip_y0 = (GLfloat)(y0 / fb_height * 2.0 - 1.0);
+   const GLfloat clip_x1 = (GLfloat)(x1 / fb_width * 2.0 - 1.0);
+   const GLfloat clip_y1 = (GLfloat)(y1 / fb_height * 2.0 - 1.0);
+   const GLuint max_slots = 4096 / sizeof(st->bitmap.vertices);
+   GLuint i;
+
+   if (st->bitmap.vbuf_slot >= max_slots) {
+      pipe_buffer_reference(pipe->screen, &st->bitmap.vbuf, NULL);
+      st->bitmap.vbuf_slot = 0;
+   }
+
+   if (!st->bitmap.vbuf) {
+      st->bitmap.vbuf = pipe_buffer_create(pipe->screen, 32, 
+                                           PIPE_BUFFER_USAGE_VERTEX,
+                                           max_slots * sizeof(st->bitmap.vertices));
+   }
+
+   /* Positions are in clip coords since we need to do clipping in case
+    * the bitmap quad goes beyond the window bounds.
+    */
+   st->bitmap.vertices[0][0][0] = clip_x0;
+   st->bitmap.vertices[0][0][1] = clip_y0;
+   st->bitmap.vertices[0][2][0] = sLeft;
+   st->bitmap.vertices[0][2][1] = tTop;
+
+   st->bitmap.vertices[1][0][0] = clip_x1;
+   st->bitmap.vertices[1][0][1] = clip_y0;
+   st->bitmap.vertices[1][2][0] = sRight;
+   st->bitmap.vertices[1][2][1] = tTop;
+   
+   st->bitmap.vertices[2][0][0] = clip_x1;
+   st->bitmap.vertices[2][0][1] = clip_y1;
+   st->bitmap.vertices[2][2][0] = sRight;
+   st->bitmap.vertices[2][2][1] = tBot;
+   
+   st->bitmap.vertices[3][0][0] = clip_x0;
+   st->bitmap.vertices[3][0][1] = clip_y1;
+   st->bitmap.vertices[3][2][0] = sLeft;
+   st->bitmap.vertices[3][2][1] = tBot;
+   
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->bitmap.vertices[i][0][2] = z;
+      st->bitmap.vertices[i][0][3] = 1.0;
+      st->bitmap.vertices[i][1][0] = color[0];
+      st->bitmap.vertices[i][1][1] = color[1];
+      st->bitmap.vertices[i][1][2] = color[2];
+      st->bitmap.vertices[i][1][3] = color[3];
+      st->bitmap.vertices[i][2][2] = 0.0; /*R*/
+      st->bitmap.vertices[i][2][3] = 1.0; /*Q*/
+   }
+
+   /* put vertex data into vbuf */
+   {
+      char *buf = pipe_buffer_map(pipe->screen, 
+                                  st->bitmap.vbuf, 
+                                  PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      memcpy(buf + st->bitmap.vbuf_slot * sizeof st->bitmap.vertices, 
+             st->bitmap.vertices, 
+             sizeof st->bitmap.vertices);
+
+      pipe_buffer_unmap(pipe->screen, st->bitmap.vbuf);
+   }
+
+   return st->bitmap.vbuf_slot++ * sizeof st->bitmap.vertices;
+}
+
+
+
+/**
+ * Render a glBitmap by drawing a textured quad
+ */
+static void
+draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
+                 GLsizei width, GLsizei height,
+                 struct pipe_texture *pt,
+                 const GLfloat *color)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct cso_context *cso = ctx->st->cso_context;
+   struct st_fragment_program *stfp;
+   GLuint maxSize;
+   GLuint offset;
+
+   stfp = combined_bitmap_fragment_program(ctx);
+
+   /* As an optimization, Mesa's fragment programs will sometimes get the
+    * primary color from a statevar/constant rather than a varying variable.
+    * when that's the case, we need to ensure that we use the 'color'
+    * parameter and not the current attribute color (which may have changed
+    * through glRasterPos and state validation.
+    * So, we force the proper color here.  Not elegant, but it works.
+    */
+   {
+      GLfloat colorSave[4];
+      COPY_4V(colorSave, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
+      COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], color);
+      st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
+      COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], colorSave);
+   }
+
+
+   /* limit checks */
+   /* XXX if the bitmap is larger than the max texture size, break
+    * it up into chunks.
+    */
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+   assert(width <= (GLsizei)maxSize);
+   assert(height <= (GLsizei)maxSize);
+
+   cso_save_rasterizer(cso);
+   cso_save_samplers(cso);
+   cso_save_sampler_textures(cso);
+   cso_save_viewport(cso);
+   cso_save_fragment_shader(cso);
+   cso_save_vertex_shader(cso);
+
+   /* rasterizer state: just scissor */
+   st->bitmap.rasterizer.scissor = ctx->Scissor.Enabled;
+   cso_set_rasterizer(cso, &st->bitmap.rasterizer);
+
+   /* fragment shader state: TEX lookup program */
+   cso_set_fragment_shader_handle(cso, stfp->driver_shader);
+
+   /* vertex shader state: position + texcoord pass-through */
+   cso_set_vertex_shader_handle(cso, st->bitmap.vs);
+
+   /* user samplers, plus our bitmap sampler */
+   {
+      struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+      uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_samplers);
+      uint i;
+      for (i = 0; i < st->state.num_samplers; i++) {
+         samplers[i] = &st->state.samplers[i];
+      }
+      samplers[stfp->bitmap_sampler] = &st->bitmap.sampler;
+      cso_set_samplers(cso, num, (const struct pipe_sampler_state **) samplers);
+   }
+
+   /* user textures, plus the bitmap texture */
+   {
+      struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+      uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_textures);
+      memcpy(textures, st->state.sampler_texture, sizeof(textures));
+      textures[stfp->bitmap_sampler] = pt;
+      cso_set_sampler_textures(cso, num, textures);
+   }
+
+   /* viewport state: viewport matching window dims */
+   {
+      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+      const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP);
+      const GLfloat width = (GLfloat)fb->Width;
+      const GLfloat height = (GLfloat)fb->Height;
+      struct pipe_viewport_state vp;
+      vp.scale[0] =  0.5f * width;
+      vp.scale[1] = height * (invert ? -0.5f : 0.5f);
+      vp.scale[2] = 1.0f;
+      vp.scale[3] = 1.0f;
+      vp.translate[0] = 0.5f * width;
+      vp.translate[1] = 0.5f * height;
+      vp.translate[2] = 0.0f;
+      vp.translate[3] = 0.0f;
+      cso_set_viewport(cso, &vp);
+   }
+
+   /* draw textured quad */
+   offset = setup_bitmap_vertex_data(st, x, y, width, height,
+                                     ctx->Current.RasterPos[2],
+                                     color);
+
+   util_draw_vertex_buffer(pipe, st->bitmap.vbuf, offset,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           3); /* attribs/vert */
+
+
+   /* restore state */
+   cso_restore_rasterizer(cso);
+   cso_restore_samplers(cso);
+   cso_restore_sampler_textures(cso);
+   cso_restore_viewport(cso);
+   cso_restore_fragment_shader(cso);
+   cso_restore_vertex_shader(cso);
+}
+
+
+static void
+reset_cache(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct bitmap_cache *cache = st->bitmap.cache;
+
+   //memset(cache->buffer, 0xff, sizeof(cache->buffer));
+   cache->empty = GL_TRUE;
+
+   cache->xmin = 1000000;
+   cache->xmax = -1000000;
+   cache->ymin = 1000000;
+   cache->ymax = -1000000;
+
+   if (cache->surf)
+      screen->tex_surface_release(screen, &cache->surf);
+
+   assert(!cache->texture);
+
+   /* allocate a new texture */
+   cache->texture = st_texture_create(st, PIPE_TEXTURE_2D,
+                                      st->bitmap.tex_format, 0,
+                                      BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT,
+                                      1, 0,
+                                      PIPE_TEXTURE_USAGE_SAMPLER);
+
+   /* Map the texture surface.
+    * Subsequent glBitmap calls will write into the texture image.
+    */
+   cache->surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0,
+                                         PIPE_BUFFER_USAGE_CPU_WRITE);
+   cache->buffer = screen->surface_map(screen, cache->surf,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* init image to all 0xff */
+   memset(cache->buffer, 0xff, BITMAP_CACHE_WIDTH * BITMAP_CACHE_HEIGHT);
+}
+
+
+/**
+ * If there's anything in the bitmap cache, draw/flush it now.
+ */
+void
+st_flush_bitmap_cache(struct st_context *st)
+{
+   if (!st->bitmap.cache->empty) {
+      struct bitmap_cache *cache = st->bitmap.cache;
+
+      if (st->ctx->DrawBuffer) {
+         struct pipe_context *pipe = st->pipe;
+         struct pipe_screen *screen = pipe->screen;
+
+         assert(cache->xmin <= cache->xmax);
+ 
+/*         printf("flush size %d x %d  at %d, %d\n",
+                cache->xmax - cache->xmin,
+                cache->ymax - cache->ymin,
+                cache->xpos, cache->ypos);
+*/
+
+         /* The texture surface has been mapped until now.
+          * So unmap and release the texture surface before drawing.
+          */
+         screen->surface_unmap(screen, cache->surf);
+         cache->buffer = NULL;
+
+         screen->tex_surface_release(screen, &cache->surf);
+
+         draw_bitmap_quad(st->ctx,
+                          cache->xpos,
+                          cache->ypos,
+                          st->ctx->Current.RasterPos[2],
+                          BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT,
+                          cache->texture,
+                          cache->color);
+      }
+
+      /* release/free the texture */
+      pipe_texture_reference(&cache->texture, NULL);
+
+      reset_cache(st);
+   }
+}
+
+/* Flush bitmap cache and release vertex buffer.
+ */
+void
+st_flush_bitmap( struct st_context *st )
+{
+   st_flush_bitmap_cache(st);
+
+   /* Release vertex buffer to avoid synchronous rendering if we were
+    * to map it in the next frame.
+    */
+   pipe_buffer_reference(st->pipe->screen, &st->bitmap.vbuf, NULL);
+   st->bitmap.vbuf_slot = 0;
+}
+
+
+/**
+ * Try to accumulate this glBitmap call in the bitmap cache.
+ * \return  GL_TRUE for success, GL_FALSE if bitmap is too large, etc.
+ */
+static GLboolean
+accum_bitmap(struct st_context *st,
+             GLint x, GLint y, GLsizei width, GLsizei height,
+             const struct gl_pixelstore_attrib *unpack,
+             const GLubyte *bitmap )
+{
+   struct bitmap_cache *cache = st->bitmap.cache;
+   int px = -999, py;
+
+   if (width > BITMAP_CACHE_WIDTH ||
+       height > BITMAP_CACHE_HEIGHT)
+      return GL_FALSE; /* too big to cache */
+
+   if (!cache->empty) {
+      px = x - cache->xpos;  /* pos in buffer */
+      py = y - cache->ypos;
+      if (px < 0 || px + width > BITMAP_CACHE_WIDTH ||
+          py < 0 || py + height > BITMAP_CACHE_HEIGHT ||
+          !TEST_EQ_4V(st->ctx->Current.RasterColor, cache->color)) {
+         /* This bitmap would extend beyond cache bounds, or the bitmap
+          * color is changing
+          * so flush and continue.
+          */
+         st_flush_bitmap_cache(st);
+      }
+   }
+
+   if (cache->empty) {
+      /* Initialize.  Center bitmap vertically in the buffer. */
+      px = 0;
+      py = (BITMAP_CACHE_HEIGHT - height) / 2;
+      cache->xpos = x;
+      cache->ypos = y - py;
+      cache->empty = GL_FALSE;
+      COPY_4FV(cache->color, st->ctx->Current.RasterColor);
+   }
+
+   assert(px != -999);
+
+   if (x < cache->xmin)
+      cache->xmin = x;
+   if (y < cache->ymin)
+      cache->ymin = y;
+   if (x + width > cache->xmax)
+      cache->xmax = x + width;
+   if (y + height > cache->ymax)
+      cache->ymax = y + height;
+
+   unpack_bitmap(st, px, py, width, height, unpack, bitmap,
+                 cache->buffer, BITMAP_CACHE_WIDTH);
+
+   return GL_TRUE; /* accumulated */
+}
+
+
+
+/**
+ * Called via ctx->Driver.Bitmap()
+ */
+static void
+st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+          const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap )
+{
+   struct st_context *st = ctx->st;
+   struct pipe_texture *pt;
+
+   if (width == 0 || height == 0)
+      return;
+
+   st_validate_state(st);
+
+   if (!st->bitmap.vs) {
+      /* create pass-through vertex shader now */
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indexes[] = { 0, 0, 0 };
+      st->bitmap.vs = util_make_vertex_passthrough_shader(st->pipe, 3,
+                                                          semantic_names,
+                                                          semantic_indexes,
+                                                          &st->bitmap.vert_shader);
+   }
+
+   if (UseBitmapCache && accum_bitmap(st, x, y, width, height, unpack, bitmap))
+      return;
+
+   pt = make_bitmap_texture(ctx, width, height, unpack, bitmap);
+   if (pt) {
+      assert(pt->target == PIPE_TEXTURE_2D);
+      draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2],
+                       width, height, pt,
+                       st->ctx->Current.RasterColor);
+      /* release/free the texture */
+      pipe_texture_reference(&pt, NULL);
+   }
+}
+
+
+/** Per-context init */
+void
+st_init_bitmap_functions(struct dd_function_table *functions)
+{
+   functions->Bitmap = st_Bitmap;
+}
+
+
+/** Per-context init */
+void
+st_init_bitmap(struct st_context *st)
+{
+   struct pipe_sampler_state *sampler = &st->bitmap.sampler;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   /* init sampler state once */
+   memset(sampler, 0, sizeof(*sampler));
+   sampler->wrap_s = PIPE_TEX_WRAP_CLAMP;
+   sampler->wrap_t = PIPE_TEX_WRAP_CLAMP;
+   sampler->wrap_r = PIPE_TEX_WRAP_CLAMP;
+   sampler->min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler->min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   sampler->mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler->normalized_coords = 1;
+
+   /* init baseline rasterizer state once */
+   memset(&st->bitmap.rasterizer, 0, sizeof(st->bitmap.rasterizer));
+   st->bitmap.rasterizer.gl_rasterization_rules = 1;
+   st->bitmap.rasterizer.bypass_vs = 1;
+
+   /* find a usable texture format */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM, PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      st->bitmap.tex_format = PIPE_FORMAT_I8_UNORM;
+   }
+   else {
+      /* XXX support more formats */
+      assert(0);
+   }
+
+   /* alloc bitmap cache object */
+   st->bitmap.cache = CALLOC_STRUCT(bitmap_cache);
+
+   reset_cache(st);
+}
+
+
+/** Per-context tear-down */
+void
+st_destroy_bitmap(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct bitmap_cache *cache = st->bitmap.cache;
+
+   screen->surface_unmap(screen, cache->surf);
+   screen->tex_surface_release(screen, &cache->surf);
+
+   if (st->bitmap.vs) {
+      cso_delete_vertex_shader(st->cso_context, st->bitmap.vs);
+      st->bitmap.vs = NULL;
+   }
+
+   if (st->bitmap.vbuf) {
+      pipe_buffer_reference(pipe->screen, &st->bitmap.vbuf, NULL);
+      st->bitmap.vbuf = NULL;
+   }
+
+   if (st->bitmap.cache) {
+      pipe_texture_release(&st->bitmap.cache->texture);
+      FREE(st->bitmap.cache);
+      st->bitmap.cache = NULL;
+   }
+}
diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h
new file mode 100644
index 00000000000..81cf61981dc
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bitmap.h
@@ -0,0 +1,52 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_BITMAP_H
+#define ST_CB_BITMAP_H
+
+
+extern void
+st_init_bitmap_functions(struct dd_function_table *functions);
+
+extern void
+st_init_bitmap(struct st_context *st);
+
+extern void
+st_destroy_bitmap(struct st_context *st);
+
+extern void
+st_flush_bitmap_cache(struct st_context *st);
+
+/* Flush bitmap cache and release vertex buffer.  Needed at end of
+ * frame to avoid synchronous rendering.
+ */
+extern void
+st_flush_bitmap(struct st_context *st);
+
+
+#endif /* ST_CB_BITMAP_H */
diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c
new file mode 100644
index 00000000000..28526234720
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_blit.c
@@ -0,0 +1,111 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_program.h"
+#include "st_cb_blit.h"
+#include "st_cb_fbo.h"
+
+#include "util/u_blit.h"
+
+#include "cso_cache/cso_context.h"
+
+
+void
+st_init_blit(struct st_context *st)
+{
+   st->blit = util_create_blit(st->pipe, st->cso_context);
+}
+
+
+void
+st_destroy_blit(struct st_context *st)
+{
+   util_destroy_blit(st->blit);
+   st->blit = NULL;
+}
+
+
+static void
+st_BlitFramebuffer(GLcontext *ctx,
+                   GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                   GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                   GLbitfield mask, GLenum filter)
+{
+   struct st_context *st = ctx->st;
+
+   const uint pFilter = ((filter == GL_NEAREST)
+                         ? PIPE_TEX_MIPFILTER_NEAREST
+                         : PIPE_TEX_MIPFILTER_LINEAR);
+
+   if (mask & GL_COLOR_BUFFER_BIT) {
+      struct st_renderbuffer *srcRb = 
+         st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer);
+      struct st_renderbuffer *dstRb = 
+         st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
+      struct pipe_surface *srcSurf = srcRb->surface;
+      struct pipe_surface *dstSurf = dstRb->surface;
+
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         /* invert Y */
+         srcY0 = srcRb->Base.Height - srcY0;
+         srcY1 = srcRb->Base.Height - srcY1;
+
+         dstY0 = dstRb->Base.Height - dstY0;
+         dstY1 = dstRb->Base.Height - dstY1;
+      }
+
+      util_blit_pixels(st->blit,
+                       srcSurf, srcX0, srcY0, srcX1, srcY1,
+                       dstSurf, dstX0, dstY0, dstX1, dstY1,
+                       0.0, pFilter);
+
+   }
+}
+
+
+
+void
+st_init_blit_functions(struct dd_function_table *functions)
+{
+#if FEATURE_EXT_framebuffer_blit
+   functions->BlitFramebuffer = st_BlitFramebuffer;
+#endif
+}
diff --git a/src/mesa/state_tracker/st_cb_blit.h b/src/mesa/state_tracker/st_cb_blit.h
new file mode 100644
index 00000000000..ed22986b531
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_blit.h
@@ -0,0 +1,46 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_BLIT_H
+#define ST_CB_BLIT_H
+
+
+#include "st_context.h"
+
+
+
+extern void
+st_init_blit(struct st_context *st);
+
+extern void
+st_destroy_blit(struct st_context *st);
+
+extern void
+st_init_blit_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_BLIT_H */
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
new file mode 100644
index 00000000000..07fa2afce05
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -0,0 +1,240 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "main/imports.h"
+#include "main/mtypes.h"
+#include "main/bufferobj.h"
+
+#include "st_context.h"
+#include "st_cb_bufferobjects.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+
+
+/* Pixel buffers and Vertex/index buffers are handled through these
+ * mesa callbacks.  Framebuffer/Renderbuffer objects are
+ * created/managed elsewhere.
+ */
+
+
+
+/**
+ * There is some duplication between mesa's bufferobjects and our
+ * bufmgr buffers.  Both have an integer handle and a hashtable to
+ * lookup an opaque structure.  It would be nice if the handles and
+ * internal structure where somehow shared.
+ */
+static struct gl_buffer_object *
+st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target)
+{
+   struct st_buffer_object *st_obj = CALLOC_STRUCT(st_buffer_object);
+
+   if (!st_obj)
+      return NULL;
+
+   _mesa_initialize_buffer_object(&st_obj->Base, name, target);
+
+   return &st_obj->Base;
+}
+
+
+
+/**
+ * Deallocate/free a vertex/pixel buffer object.
+ * Called via glDeleteBuffersARB().
+ */
+static void
+st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+
+   if (st_obj->buffer) 
+      pipe_buffer_reference(pipe->screen, &st_obj->buffer, NULL);
+
+   free(st_obj);
+}
+
+
+
+/**
+ * Replace data in a subrange of buffer object.  If the data range
+ * specified by size + offset extends beyond the end of the buffer or
+ * if data is NULL, no copy is performed.
+ * Called via glBufferSubDataARB().
+ */
+static void
+st_bufferobj_subdata(GLcontext *ctx,
+		     GLenum target,
+		     GLintptrARB offset,
+		     GLsizeiptrARB size,
+		     const GLvoid * data, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   char *map;
+
+   if (offset >= st_obj->size || size > (st_obj->size - offset))
+      return;
+
+   map = pipe_buffer_map(pipe->screen, st_obj->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+   memcpy(map + offset, data, size);
+   pipe_buffer_unmap(pipe->screen, st_obj->buffer);
+}
+
+
+/**
+ * Called via glGetBufferSubDataARB().
+ */
+static void
+st_bufferobj_get_subdata(GLcontext *ctx,
+                         GLenum target,
+                         GLintptrARB offset,
+                         GLsizeiptrARB size,
+                         GLvoid * data, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   char *map;
+
+   if (offset >= st_obj->size || size > (st_obj->size - offset))
+      return;
+
+   map = pipe_buffer_map(pipe->screen, st_obj->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+   memcpy(data, map + offset, size);
+   pipe_buffer_unmap(pipe->screen, st_obj->buffer);
+}
+
+
+/**
+ * Allocate space for and store data in a buffer object.  Any data that was
+ * previously stored in the buffer object is lost.  If data is NULL,
+ * memory will be allocated, but no copy will occur.
+ * Called via glBufferDataARB().
+ */
+static void
+st_bufferobj_data(GLcontext *ctx,
+		  GLenum target,
+		  GLsizeiptrARB size,
+		  const GLvoid * data,
+		  GLenum usage, 
+		  struct gl_buffer_object *obj)
+{
+   struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   unsigned buffer_usage;
+
+   st_obj->Base.Size = size;
+   st_obj->Base.Usage = usage;
+   
+   switch(target) {
+   case GL_PIXEL_PACK_BUFFER_ARB:
+   case GL_PIXEL_UNPACK_BUFFER_ARB:
+      buffer_usage = PIPE_BUFFER_USAGE_PIXEL;
+      break;
+   case GL_ARRAY_BUFFER_ARB:
+      buffer_usage = PIPE_BUFFER_USAGE_VERTEX;
+      break;
+   case GL_ELEMENT_ARRAY_BUFFER_ARB:
+      buffer_usage = PIPE_BUFFER_USAGE_INDEX;
+      break;
+   default:
+      buffer_usage = 0;
+   }
+
+   pipe_buffer_reference( pipe->screen, &st_obj->buffer, NULL );
+
+   st_obj->buffer = pipe_buffer_create( pipe->screen, 32, buffer_usage, size );
+
+   st_obj->size = size;
+
+   if (data)
+      st_bufferobj_subdata(ctx, target, 0, size, data, obj);
+}
+
+
+/**
+ * Called via glMapBufferARB().
+ */
+static void *
+st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
+                 struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   GLuint flags;
+
+   switch (access) {
+   case GL_WRITE_ONLY:
+      flags = PIPE_BUFFER_USAGE_CPU_WRITE;
+      break;
+   case GL_READ_ONLY:
+      flags = PIPE_BUFFER_USAGE_CPU_READ;
+      break;
+   case GL_READ_WRITE:
+      /* fall-through */
+   default:
+      flags = PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE;
+      break;      
+   }
+
+   obj->Pointer = pipe_buffer_map(pipe->screen, st_obj->buffer, flags);
+   return obj->Pointer;
+}
+
+
+/**
+ * Called via glMapBufferARB().
+ */
+static GLboolean
+st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+
+   pipe_buffer_unmap(pipe->screen, st_obj->buffer);
+   obj->Pointer = NULL;
+   return GL_TRUE;
+}
+
+
+void
+st_init_bufferobject_functions(struct dd_function_table *functions)
+{
+   functions->NewBufferObject = st_bufferobj_alloc;
+   functions->DeleteBuffer = st_bufferobj_free;
+   functions->BufferData = st_bufferobj_data;
+   functions->BufferSubData = st_bufferobj_subdata;
+   functions->GetBufferSubData = st_bufferobj_get_subdata;
+   functions->MapBuffer = st_bufferobj_map;
+   functions->UnmapBuffer = st_bufferobj_unmap;
+}
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h
new file mode 100644
index 00000000000..dcbb5a52336
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.h
@@ -0,0 +1,66 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_BUFFEROBJECTS_H
+#define ST_CB_BUFFEROBJECTS_H
+
+struct st_context;
+struct gl_buffer_object;
+struct pipe_buffer;
+
+/**
+ * State_tracker vertex/pixel buffer object, derived from Mesa's
+ * gl_buffer_object.
+ */
+struct st_buffer_object
+{
+   struct gl_buffer_object Base;
+   struct pipe_buffer *buffer;  
+   GLsizeiptrARB size;
+};
+
+
+/* Are the obj->Name tests necessary?  Unfortunately yes, mesa
+ * allocates a couple of gl_buffer_object structs statically, and the
+ * Name == 0 test is the only way to identify them and avoid casting
+ * them erroneously to our structs.
+ */
+static INLINE struct st_buffer_object *
+st_buffer_object(struct gl_buffer_object *obj)
+{
+   if (obj->Name)
+      return (struct st_buffer_object *) obj;
+   else
+      return NULL;
+}
+
+
+extern void
+st_init_bufferobject_functions(struct dd_function_table *functions);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
new file mode 100644
index 00000000000..668c3f9ebf4
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -0,0 +1,602 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "shader/prog_instruction.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_accum.h"
+#include "st_cb_clear.h"
+#include "st_cb_fbo.h"
+#include "st_draw.h"
+#include "st_program.h"
+#include "st_public.h"
+#include "st_mesa_to_tgsi.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_pack_color.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_draw_quad.h"
+
+#include "cso_cache/cso_context.h"
+
+
+void
+st_init_clear(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   /* rasterizer state: bypass clipping */
+   memset(&st->clear.raster, 0, sizeof(st->clear.raster));
+   st->clear.raster.gl_rasterization_rules = 1;
+   st->clear.raster.bypass_clipping = 1;
+
+   /* viewport state: identity since we're drawing in window coords */
+   st->clear.viewport.scale[0] = 1.0;
+   st->clear.viewport.scale[1] = 1.0;
+   st->clear.viewport.scale[2] = 1.0;
+   st->clear.viewport.scale[3] = 1.0;
+   st->clear.viewport.translate[0] = 0.0;
+   st->clear.viewport.translate[1] = 0.0;
+   st->clear.viewport.translate[2] = 0.0;
+   st->clear.viewport.translate[3] = 0.0;
+
+   /* fragment shader state: color pass-through program */
+   st->clear.fs =
+      util_make_fragment_passthrough_shader(pipe, &st->clear.frag_shader);
+
+   /* vertex shader state: color/position pass-through */
+   {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR };
+      const uint semantic_indexes[] = { 0, 0 };
+      st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2,
+                                                         semantic_names,
+                                                         semantic_indexes,
+                                                         &st->clear.vert_shader);
+   }
+}
+
+
+void
+st_destroy_clear(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (st->clear.vert_shader.tokens) {
+      FREE((void *) st->clear.vert_shader.tokens);
+      st->clear.vert_shader.tokens = NULL;
+   }
+
+   if (st->clear.frag_shader.tokens) {
+      FREE((void *) st->clear.frag_shader.tokens);
+      st->clear.frag_shader.tokens = NULL;
+   }
+
+   if (st->clear.fs) {
+      cso_delete_fragment_shader(st->cso_context, st->clear.fs);
+      st->clear.fs = NULL;
+   }
+   if (st->clear.vs) {
+      cso_delete_vertex_shader(st->cso_context, st->clear.vs);
+      st->clear.vs = NULL;
+   }
+   if (st->clear.vbuf) {
+      pipe_buffer_reference(pipe->screen, &st->clear.vbuf, NULL);
+      st->clear.vbuf = NULL;
+   }
+}
+
+
+static GLboolean
+is_depth_stencil_format(enum pipe_format pipeFormat)
+{
+   switch (pipeFormat) {
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_Z24S8_UNORM:
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+
+
+/**
+ * Draw a screen-aligned quadrilateral.
+ * Coords are window coords with y=0=bottom.  These coords will be transformed
+ * by the vertex shader and viewport transform (which will flip Y if needed).
+ */
+static void
+draw_quad(GLcontext *ctx,
+          float x0, float y0, float x1, float y1, GLfloat z,
+          const GLfloat color[4])
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   const GLuint max_slots = 1024 / sizeof(st->clear.vertices);
+   GLuint i;
+   void *buf;
+
+   if (st->clear.vbuf_slot >= max_slots) {
+      pipe_buffer_reference(pipe->screen, &st->clear.vbuf, NULL);
+      st->clear.vbuf_slot = 0;
+   }
+
+   if (!st->clear.vbuf) {
+      st->clear.vbuf = pipe_buffer_create(pipe->screen, 32, PIPE_BUFFER_USAGE_VERTEX,
+                                          max_slots * sizeof(st->clear.vertices));
+   }
+
+   /* positions */
+   st->clear.vertices[0][0][0] = x0;
+   st->clear.vertices[0][0][1] = y0;
+
+   st->clear.vertices[1][0][0] = x1;
+   st->clear.vertices[1][0][1] = y0;
+
+   st->clear.vertices[2][0][0] = x1;
+   st->clear.vertices[2][0][1] = y1;
+
+   st->clear.vertices[3][0][0] = x0;
+   st->clear.vertices[3][0][1] = y1;
+
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->clear.vertices[i][0][2] = z;
+      st->clear.vertices[i][0][3] = 1.0;
+      st->clear.vertices[i][1][0] = color[0];
+      st->clear.vertices[i][1][1] = color[1];
+      st->clear.vertices[i][1][2] = color[2];
+      st->clear.vertices[i][1][3] = color[3];
+   }
+
+   /* put vertex data into vbuf */
+   buf = pipe_buffer_map(pipe->screen, st->clear.vbuf, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy((char *)buf + st->clear.vbuf_slot * sizeof(st->clear.vertices), 
+          st->clear.vertices, 
+          sizeof(st->clear.vertices));
+
+   pipe_buffer_unmap(pipe->screen, st->clear.vbuf);
+
+   /* draw */
+   util_draw_vertex_buffer(pipe, 
+                           st->clear.vbuf, 
+                           st->clear.vbuf_slot * sizeof(st->clear.vertices),
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+   /* Increment slot */
+   st->clear.vbuf_slot++;
+}
+
+
+
+/**
+ * Do glClear by drawing a quadrilateral.
+ * The vertices of the quad will be computed from the
+ * ctx->DrawBuffer->_X/Ymin/max fields.
+ */
+static void
+clear_with_quad(GLcontext *ctx,
+                GLboolean color, GLboolean depth, GLboolean stencil)
+{
+   struct st_context *st = ctx->st;
+   const GLfloat x0 = (GLfloat) ctx->DrawBuffer->_Xmin;
+   const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax;
+   GLfloat y0, y1;
+
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      y0 = (GLfloat) (ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax);
+      y1 = (GLfloat) (ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin);
+   }
+   else {
+      y0 = (GLfloat) ctx->DrawBuffer->_Ymin;
+      y1 = (GLfloat) ctx->DrawBuffer->_Ymax;
+   }
+
+   /*
+   printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__, 
+	  color ? "color, " : "",
+	  depth ? "depth, " : "",
+	  stencil ? "stencil" : "",
+	  x0, y0,
+	  x1, y1);
+   */
+
+   cso_save_blend(st->cso_context);
+   cso_save_depth_stencil_alpha(st->cso_context);
+   cso_save_rasterizer(st->cso_context);
+   cso_save_viewport(st->cso_context);
+   cso_save_fragment_shader(st->cso_context);
+   cso_save_vertex_shader(st->cso_context);
+
+   /* blend state: RGBA masking */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      if (color) {
+         if (ctx->Color.ColorMask[0])
+            blend.colormask |= PIPE_MASK_R;
+         if (ctx->Color.ColorMask[1])
+            blend.colormask |= PIPE_MASK_G;
+         if (ctx->Color.ColorMask[2])
+            blend.colormask |= PIPE_MASK_B;
+         if (ctx->Color.ColorMask[3])
+            blend.colormask |= PIPE_MASK_A;
+         if (st->ctx->Color.DitherFlag)
+            blend.dither = 1;
+      }
+      cso_set_blend(st->cso_context, &blend);
+   }
+
+   /* depth_stencil state: always pass/set to ref value */
+   {
+      struct pipe_depth_stencil_alpha_state depth_stencil;
+      memset(&depth_stencil, 0, sizeof(depth_stencil));
+      if (depth) {
+         depth_stencil.depth.enabled = 1;
+         depth_stencil.depth.writemask = 1;
+         depth_stencil.depth.func = PIPE_FUNC_ALWAYS;
+      }
+
+      if (stencil) {
+         depth_stencil.stencil[0].enabled = 1;
+         depth_stencil.stencil[0].func = PIPE_FUNC_ALWAYS;
+         depth_stencil.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+         depth_stencil.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+         depth_stencil.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+         depth_stencil.stencil[0].ref_value = ctx->Stencil.Clear;
+         depth_stencil.stencil[0].valuemask = 0xff;
+         depth_stencil.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
+      }
+
+      cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil);
+   }
+
+   cso_set_rasterizer(st->cso_context, &st->clear.raster);
+   cso_set_viewport(st->cso_context, &st->clear.viewport);
+
+   cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
+   cso_set_vertex_shader_handle(st->cso_context, st->clear.vs);
+
+   /* draw quad matching scissor rect (XXX verify coord round-off) */
+   draw_quad(ctx, x0, y0, x1, y1, (GLfloat) ctx->Depth.Clear, ctx->Color.ClearColor);
+
+   /* Restore pipe state */
+   cso_restore_blend(st->cso_context);
+   cso_restore_depth_stencil_alpha(st->cso_context);
+   cso_restore_rasterizer(st->cso_context);
+   cso_restore_viewport(st->cso_context);
+   cso_restore_fragment_shader(st->cso_context);
+   cso_restore_vertex_shader(st->cso_context);
+}
+
+
+/**
+ * Determine if we need to clear the depth buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_color_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (!ctx->Color.ColorMask[0] ||
+       !ctx->Color.ColorMask[1] ||
+       !ctx->Color.ColorMask[2] ||
+       !ctx->Color.ColorMask[3])
+      return TRUE;
+
+   return FALSE;
+}
+
+
+static INLINE GLboolean
+check_clear_depth_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
+   GLboolean maskStencil
+      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (maskStencil)
+      return TRUE;
+
+   return FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the depth buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (isDS && 
+       strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
+       ctx->DrawBuffer->Visual.stencilBits > 0)
+      return TRUE;
+
+   return FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the stencil buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
+   const GLboolean maskStencil
+      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (maskStencil) 
+      return TRUE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   /* This is correct, but it is necessary to look at the depth clear
+    * value held in the surface when it comes time to issue the clear,
+    * rather than taking depth and stencil clear values from the
+    * current state.
+    */
+   if (isDS && 
+       strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
+       ctx->DrawBuffer->Visual.depthBits > 0)
+      return TRUE;
+
+   return FALSE;
+}
+
+
+
+static void
+clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_color_with_quad( ctx, rb )) {
+      /* masking or scissoring */
+      clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE);
+   }
+   else {
+      /* clear whole buffer w/out masking */
+      uint clearValue;
+      /* NOTE: we always pass the clear color as PIPE_FORMAT_A8R8G8B8_UNORM
+       * at this time!
+       */
+      util_pack_color(ctx->Color.ClearColor, PIPE_FORMAT_A8R8G8B8_UNORM, &clearValue);
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+static void
+clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_depth_with_quad(ctx, rb)) {
+      /* scissoring or we have a combined depth/stencil buffer */
+      clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE);
+   }
+   else {
+      /* simple clear of whole buffer */
+      uint clearValue = util_pack_z(strb->surface->format, ctx->Depth.Clear);
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+static void
+clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_stencil_with_quad(ctx, rb)) {
+      /* masking or scissoring or combined depth/stencil buffer */
+      clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE);
+   }
+   else {
+      /* simple clear of whole buffer */
+      GLuint clearValue = ctx->Stencil.Clear;
+
+      switch (strb->surface->format) {
+      case PIPE_FORMAT_S8Z24_UNORM:
+         clearValue <<= 24;
+         break;
+      default:
+         ; /* no-op, stencil value is in least significant bits */
+      }  
+
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+static void
+clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_depth_stencil_with_quad(ctx, rb)) {
+      /* masking or scissoring */
+      clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_TRUE);
+   }
+   else {
+      /* clear whole buffer w/out masking */
+      GLuint clearValue = util_pack_z(strb->surface->format, ctx->Depth.Clear);
+
+      switch (strb->surface->format) {
+      case PIPE_FORMAT_S8Z24_UNORM:
+         clearValue |= ctx->Stencil.Clear << 24;
+         break;
+      case PIPE_FORMAT_Z24S8_UNORM:
+         clearValue |= ctx->Stencil.Clear;
+         break;
+      default:
+         assert(0);
+      }  
+
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+void st_flush_clear( struct st_context *st )
+{
+   /* Release vertex buffer to avoid synchronous rendering if we were
+    * to map it in the next frame.
+    */
+   pipe_buffer_reference(st->pipe->screen, &st->clear.vbuf, NULL);
+   st->clear.vbuf_slot = 0;
+}
+ 
+
+
+/**
+ * Called via ctx->Driver.Clear()
+ * XXX: doesn't pick up the differences between front/back/left/right
+ * clears.  Need to sort that out...
+ */
+static void st_clear(GLcontext *ctx, GLbitfield mask)
+{
+   static const GLbitfield BUFFER_BITS_DS
+      = (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
+   struct st_context *st = ctx->st;
+   struct gl_renderbuffer *depthRb
+      = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
+   struct gl_renderbuffer *stencilRb
+      = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
+   GLbitfield cmask = mask & BUFFER_BITS_COLOR;
+
+   /* This makes sure the softpipe has the latest scissor, etc values */
+   st_validate_state( st );
+
+   /*
+    * XXX TO-DO:
+    * If we're going to use clear_with_quad() for any reason, use it to
+    * clear as many other buffers as possible.
+    * As it is now, we sometimes call clear_with_quad() three times to clear
+    * color/depth/stencil individually...
+    */
+
+   if (cmask) {
+      GLuint b;
+      for (b = 0; cmask; b++) {
+         if (cmask & (1 << b)) {
+            struct gl_renderbuffer *rb
+               = ctx->DrawBuffer->Attachment[b].Renderbuffer;
+            assert(rb);
+            clear_color_buffer(ctx, rb);
+            cmask &= ~(1 << b); /* turn off bit */
+         }
+         assert(b < BUFFER_COUNT);
+      }
+   }
+
+   if (mask & BUFFER_BIT_ACCUM) {
+      st_clear_accum_buffer(ctx,
+                       ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
+   }
+
+   if ((mask & BUFFER_BITS_DS) == BUFFER_BITS_DS && depthRb == stencilRb) {
+      /* clearing combined depth + stencil */
+      clear_depth_stencil_buffer(ctx, depthRb);
+   }
+   else {
+      /* separate depth/stencil clears */
+      if (mask & BUFFER_BIT_DEPTH) {
+         clear_depth_buffer(ctx, depthRb);
+      }
+      if (mask & BUFFER_BIT_STENCIL) {
+         clear_stencil_buffer(ctx, stencilRb);
+      }
+   }
+}
+
+
+void st_init_clear_functions(struct dd_function_table *functions)
+{
+   functions->Clear = st_clear;
+}
diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h
new file mode 100644
index 00000000000..bc035ac25ca
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_clear.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_CLEAR_H
+#define ST_CB_CLEAR_H
+
+
+extern void
+st_init_clear(struct st_context *st);
+
+
+extern void
+st_destroy_clear(struct st_context *st);
+
+extern void
+st_flush_clear(struct st_context *st);
+
+
+extern void
+st_init_clear_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_CLEAR_H */
+
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
new file mode 100644
index 00000000000..32bf21411da
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -0,0 +1,1104 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "main/state.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_draw.h"
+#include "st_program.h"
+#include "st_cb_drawpixels.h"
+#include "st_cb_readpixels.h"
+#include "st_cb_fbo.h"
+#include "st_cb_texture.h"
+#include "st_draw.h"
+#include "st_format.h"
+#include "st_mesa_to_tgsi.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_draw_quad.h"
+#include "shader/prog_instruction.h"
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Check if the given program is:
+ * 0: MOVE result.color, fragment.color;
+ * 1: END;
+ */
+static GLboolean
+is_passthrough_program(const struct gl_fragment_program *prog)
+{
+   if (prog->Base.NumInstructions == 2) {
+      const struct prog_instruction *inst = prog->Base.Instructions;
+      if (inst[0].Opcode == OPCODE_MOV &&
+          inst[1].Opcode == OPCODE_END &&
+          inst[0].DstReg.File == PROGRAM_OUTPUT &&
+          inst[0].DstReg.Index == FRAG_RESULT_COLR &&
+          inst[0].DstReg.WriteMask == WRITEMASK_XYZW &&
+          inst[0].SrcReg[0].File == PROGRAM_INPUT &&
+          inst[0].SrcReg[0].Index == FRAG_ATTRIB_COL0 &&
+          inst[0].SrcReg[0].Swizzle == SWIZZLE_XYZW) {
+         return GL_TRUE;
+      }
+   }
+   return GL_FALSE;
+}
+
+
+
+/**
+ * Make fragment shader for glDraw/CopyPixels.  This shader is made
+ * by combining the pixel transfer shader with the user-defined shader.
+ */
+static struct st_fragment_program *
+combined_drawpix_fragment_program(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+   struct st_fragment_program *stfp;
+
+   if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn
+       && st->fp->serialNo == st->pixel_xfer.user_prog_sn) {
+      /* the pixel tranfer program has not changed and the user-defined
+       * program has not changed, so re-use the combined program.
+       */
+      stfp = st->pixel_xfer.combined_prog;
+   }
+   else {
+      /* Concatenate the pixel transfer program with the current user-
+       * defined program.
+       */
+      if (is_passthrough_program(&st->fp->Base)) {
+         stfp = (struct st_fragment_program *)
+            _mesa_clone_program(ctx, &st->pixel_xfer.program->Base.Base);
+      }
+      else {
+#if 0
+         printf("Base program:\n");
+         _mesa_print_program(&st->fp->Base.Base);
+         printf("DrawPix program:\n");
+         _mesa_print_program(&st->pixel_xfer.program->Base.Base);
+#endif
+         stfp = (struct st_fragment_program *)
+            _mesa_combine_programs(ctx,
+                                   &st->pixel_xfer.program->Base.Base,
+                                   &st->fp->Base.Base);
+      }
+
+#if 0
+      {
+         struct gl_program *p = &stfp->Base.Base;
+         printf("Combined DrawPixels program:\n");
+         _mesa_print_program(p);
+         printf("InputsRead: 0x%x\n", p->InputsRead);
+         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
+         _mesa_print_parameter_list(p->Parameters);
+      }
+#endif
+
+      /* translate to TGSI tokens */
+      st_translate_fragment_program(st, stfp, NULL);
+
+      /* save new program, update serial numbers */
+      st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo;
+      st->pixel_xfer.user_prog_sn = st->fp->serialNo;
+      st->pixel_xfer.combined_prog_sn = stfp->serialNo;
+      st->pixel_xfer.combined_prog = stfp;
+   }
+
+   /* Ideally we'd have updated the pipe constants during the normal
+    * st/atom mechanism.  But we can't since this is specific to glDrawPixels.
+    */
+   st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
+
+   return stfp;
+}
+
+
+/**
+ * Create fragment shader that does a TEX() instruction to get a Z
+ * value, then writes to FRAG_RESULT_DEPR.
+ * Pass fragment color through as-is.
+ */
+static struct st_fragment_program *
+make_fragment_shader_z(struct st_context *st)
+{
+   GLcontext *ctx = st->ctx;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   if (st->drawpix.z_shader) {
+      return st->drawpix.z_shader;
+   }
+
+   /*
+    * Create shader now
+    */
+   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   p->NumInstructions = 3;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+
+   /* TEX result.depth, fragment.texcoord[0], texture[0], 2D; */
+   p->Instructions[ic].Opcode = OPCODE_TEX;
+   p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPR;
+   p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Z;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   p->Instructions[ic].TexSrcUnit = 0;
+   p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+
+   /* MOV result.color, fragment.color */
+   p->Instructions[ic].Opcode = OPCODE_MOV;
+   p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0;
+   ic++;
+
+   /* END; */
+   p->Instructions[ic++].Opcode = OPCODE_END;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0;
+   p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR);
+   p->SamplersUsed = 0x1;  /* sampler 0 (bit 0) is used */
+
+   st->drawpix.z_shader = (struct st_fragment_program *) p;
+   st_translate_fragment_program(st, st->drawpix.z_shader, NULL);
+
+   return st->drawpix.z_shader;
+}
+
+
+
+/**
+ * Create a simple vertex shader that just passes through the
+ * vertex position and texcoord (and optionally, color).
+ */
+static struct st_vertex_program *
+st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor)
+{
+   GLcontext *ctx = st->ctx;
+   struct st_vertex_program *stvp;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   if (st->drawpix.vert_shaders[passColor])
+      return st->drawpix.vert_shaders[passColor];
+
+   /*
+    * Create shader now
+    */
+   p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   if (passColor)
+      p->NumInstructions = 4;
+   else
+      p->NumInstructions = 3;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+   /* MOV result.pos, vertex.pos; */
+   p->Instructions[0].Opcode = OPCODE_MOV;
+   p->Instructions[0].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS;
+   p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS;
+   /* MOV result.texcoord0, vertex.texcoord0; */
+   p->Instructions[1].Opcode = OPCODE_MOV;
+   p->Instructions[1].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[1].DstReg.Index = VERT_RESULT_TEX0;
+   p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_TEX0;
+   ic = 2;
+   if (passColor) {
+      /* MOV result.color0, vertex.color0; */
+      p->Instructions[ic].Opcode = OPCODE_MOV;
+      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
+      p->Instructions[ic].DstReg.Index = VERT_RESULT_COL0;
+      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+      p->Instructions[ic].SrcReg[0].Index = VERT_ATTRIB_COLOR0;
+      ic++;
+   }
+
+   /* END; */
+   p->Instructions[ic].Opcode = OPCODE_END;
+   ic++;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = VERT_BIT_POS | VERT_BIT_TEX0;
+   p->OutputsWritten = ((1 << VERT_RESULT_TEX0) |
+                        (1 << VERT_RESULT_HPOS));
+   if (passColor) {
+      p->InputsRead |= VERT_BIT_COLOR0;
+      p->OutputsWritten |= (1 << VERT_RESULT_COL0);
+   }
+
+   stvp = (struct st_vertex_program *) p;
+   st_translate_vertex_program(st, stvp, NULL, NULL, NULL);
+
+   st->drawpix.vert_shaders[passColor] = stvp;
+
+   return stvp;
+}
+
+
+static GLenum
+_mesa_base_format(GLenum format)
+{
+   switch (format) {
+   case GL_DEPTH_COMPONENT:
+      return GL_DEPTH_COMPONENT;
+   case GL_STENCIL_INDEX:
+      return GL_STENCIL_INDEX;
+   default:
+      return GL_RGBA;
+   }
+}
+
+
+/**
+ * Make texture containing an image for glDrawPixels image.
+ * If 'pixels' is NULL, leave the texture image data undefined.
+ */
+static struct pipe_texture *
+make_texture(struct st_context *st,
+	     GLsizei width, GLsizei height, GLenum format, GLenum type,
+	     const struct gl_pixelstore_attrib *unpack,
+	     const GLvoid *pixels)
+{
+   GLcontext *ctx = st->ctx;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const struct gl_texture_format *mformat;
+   struct pipe_texture *pt;
+   enum pipe_format pipeFormat;
+   GLuint cpp;
+   GLenum baseFormat;
+
+   baseFormat = _mesa_base_format(format);
+
+   mformat = st_ChooseTextureFormat(ctx, baseFormat, format, type);
+   assert(mformat);
+
+   pipeFormat = st_mesa_format_to_pipe_format(mformat->MesaFormat);
+   assert(pipeFormat);
+   cpp = st_sizeof_format(pipeFormat);
+
+   pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+   if (!pixels)
+      return NULL;
+
+   pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height,
+			  1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   if (!pt) {
+      _mesa_unmap_drawpix_pbo(ctx, unpack);
+      return NULL;
+   }
+
+   {
+      struct pipe_surface *surface;
+      static const GLuint dstImageOffsets = 0;
+      GLboolean success;
+      GLubyte *dest;
+      const GLbitfield imageTransferStateSave = ctx->_ImageTransferState;
+
+      /* we'll do pixel transfer in a fragment shader */
+      ctx->_ImageTransferState = 0x0;
+
+      surface = screen->get_tex_surface(screen, pt, 0, 0, 0,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      /* map texture surface */
+      dest = screen->surface_map(screen, surface,
+                                 PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      /* Put image into texture surface.
+       * Note that the image is actually going to be upside down in
+       * the texture.  We deal with that with texcoords.
+       */
+      success = mformat->StoreImage(ctx, 2,           /* dims */
+                                    baseFormat,       /* baseInternalFormat */
+                                    mformat,          /* gl_texture_format */
+                                    dest,             /* dest */
+                                    0, 0, 0,          /* dstX/Y/Zoffset */
+                                    surface->stride,  /* dstRowStride, bytes */
+                                    &dstImageOffsets, /* dstImageOffsets */
+                                    width, height, 1, /* size */
+                                    format, type,     /* src format/type */
+                                    pixels,           /* data source */
+                                    unpack);
+
+      /* unmap */
+      screen->surface_unmap(screen, surface);
+      pipe_surface_reference(&surface, NULL);
+
+      assert(success);
+
+      /* restore */
+      ctx->_ImageTransferState = imageTransferStateSave;
+   }
+
+   _mesa_unmap_drawpix_pbo(ctx, unpack);
+
+   return pt;
+}
+
+
+/**
+ * Draw quad with texcoords and optional color.
+ * Coords are window coords with y=0=bottom.
+ * \param color  may be null
+ * \param invertTex  if true, flip texcoords vertically
+ */
+static void
+draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
+          GLfloat x1, GLfloat y1, const GLfloat *color,
+          GLboolean invertTex)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */
+
+   /* setup vertex data */
+   {
+      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+      const GLfloat fb_width = (GLfloat) fb->Width;
+      const GLfloat fb_height = (GLfloat) fb->Height;
+      const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f;
+      const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f;
+      const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f;
+      const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f;
+      const GLfloat sLeft = 0.0f, sRight = 1.0f;
+      const GLfloat tTop = invertTex, tBot = 1.0f - tTop;
+      GLuint tex, i;
+
+      /* upper-left */
+      verts[0][0][0] = clip_x0;    /* v[0].attr[0].x */
+      verts[0][0][1] = clip_y0;    /* v[0].attr[0].y */
+
+      /* upper-right */
+      verts[1][0][0] = clip_x1;
+      verts[1][0][1] = clip_y0;
+
+      /* lower-right */
+      verts[2][0][0] = clip_x1;
+      verts[2][0][1] = clip_y1;
+
+      /* lower-left */
+      verts[3][0][0] = clip_x0;
+      verts[3][0][1] = clip_y1;
+
+      tex = color ? 2 : 1;
+      verts[0][tex][0] = sLeft; /* v[0].attr[tex].s */
+      verts[0][tex][1] = tTop;  /* v[0].attr[tex].t */
+      verts[1][tex][0] = sRight;
+      verts[1][tex][1] = tTop;
+      verts[2][tex][0] = sRight;
+      verts[2][tex][1] = tBot;
+      verts[3][tex][0] = sLeft;
+      verts[3][tex][1] = tBot;
+
+      /* same for all verts: */
+      if (color) {
+         for (i = 0; i < 4; i++) {
+            verts[i][0][2] = z;   /*Z*/
+            verts[i][0][3] = 1.0f; /*W*/
+            verts[i][1][0] = color[0];
+            verts[i][1][1] = color[1];
+            verts[i][1][2] = color[2];
+            verts[i][1][3] = color[3];
+            verts[i][2][2] = 0.0f; /*R*/
+            verts[i][2][3] = 1.0f; /*Q*/
+         }
+      }
+      else {
+         for (i = 0; i < 4; i++) {
+            verts[i][0][2] = z;   /*Z*/
+            verts[i][0][3] = 1.0f; /*W*/
+            verts[i][1][2] = 0.0f; /*R*/
+            verts[i][1][3] = 1.0f; /*Q*/
+         }
+      }
+   }
+
+   {
+      struct pipe_buffer *buf;
+      ubyte *map;
+
+      /* allocate/load buffer object with vertex data */
+      buf = pipe_buffer_create(pipe->screen, 32, PIPE_BUFFER_USAGE_VERTEX,
+                               sizeof(verts));
+      map = pipe_buffer_map(pipe->screen, buf, PIPE_BUFFER_USAGE_CPU_WRITE);
+      memcpy(map, verts, sizeof(verts));
+      pipe_buffer_unmap(pipe->screen, buf);
+
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_QUADS,
+                              4,  /* verts */
+                              3); /* attribs/vert */
+      pipe_buffer_reference(pipe->screen, &buf, NULL);
+   }
+}
+
+
+
+static void
+draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
+                   GLsizei width, GLsizei height,
+                   GLfloat zoomX, GLfloat zoomY,
+                   struct pipe_texture *pt,
+                   struct st_vertex_program *stvp,
+                   struct st_fragment_program *stfp,
+                   const GLfloat *color,
+                   GLboolean invertTex)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct cso_context *cso = ctx->st->cso_context;
+   GLfloat x0, y0, x1, y1;
+   GLsizei maxSize;
+
+   /* limit checks */
+   /* XXX if DrawPixels image is larger than max texture size, break
+    * it up into chunks.
+    */
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+   assert(width <= maxSize);
+   assert(height <= maxSize);
+
+   cso_save_rasterizer(cso);
+   cso_save_viewport(cso);
+   cso_save_samplers(cso);
+   cso_save_sampler_textures(cso);
+   cso_save_fragment_shader(cso);
+   cso_save_vertex_shader(cso);
+
+   /* rasterizer state: just scissor */
+   {
+      struct pipe_rasterizer_state rasterizer;
+      memset(&rasterizer, 0, sizeof(rasterizer));
+      rasterizer.gl_rasterization_rules = 1;
+      rasterizer.scissor = ctx->Scissor.Enabled;
+      cso_set_rasterizer(cso, &rasterizer);
+   }
+
+   /* fragment shader state: TEX lookup program */
+   cso_set_fragment_shader_handle(cso, stfp->driver_shader);
+
+   /* vertex shader state: position + texcoord pass-through */
+   cso_set_vertex_shader_handle(cso, stvp->driver_shader);
+
+
+   /* texture sampling state: */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+
+      cso_single_sampler(cso, 0, &sampler);
+      if (st->pixel_xfer.pixelmap_enabled) {
+         cso_single_sampler(cso, 1, &sampler);
+      }
+      cso_single_sampler_done(cso);
+   }
+
+   /* viewport state: viewport matching window dims */
+   {
+      const float width = (float) ctx->DrawBuffer->Width;
+      const float height = (float) ctx->DrawBuffer->Height;
+      struct pipe_viewport_state vp;
+      vp.scale[0] =  0.5f * width;
+      vp.scale[1] = -0.5f * height;
+      vp.scale[2] = 1.0f;
+      vp.scale[3] = 1.0f;
+      vp.translate[0] = 0.5f * width;
+      vp.translate[1] = 0.5f * height;
+      vp.translate[2] = 0.0f;
+      vp.translate[3] = 0.0f;
+      cso_set_viewport(cso, &vp);
+   }
+
+   /* texture state: */
+   if (st->pixel_xfer.pixelmap_enabled) {
+      struct pipe_texture *textures[2];
+      textures[0] = pt;
+      textures[1] = st->pixel_xfer.pixelmap_texture;
+      pipe->set_sampler_textures(pipe, 2, textures);
+   }
+   else {
+      pipe->set_sampler_textures(pipe, 1, &pt);
+   }
+
+   /* Compute window coords (y=0=bottom) with pixel zoom.
+    * Recall that these coords are transformed by the current
+    * vertex shader and viewport transformation.
+    */
+   x0 = (GLfloat) x;
+   x1 = x + width * ctx->Pixel.ZoomX;
+   y0 = (GLfloat) y;
+   y1 = y + height * ctx->Pixel.ZoomY;
+   //if(!color)
+   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex);
+   //else
+   //printf("skip draw quad\n");
+   /* restore state */
+   cso_restore_rasterizer(cso);
+   cso_restore_viewport(cso);
+   cso_restore_samplers(cso);
+   cso_restore_sampler_textures(cso);
+   cso_restore_fragment_shader(cso);
+   cso_restore_vertex_shader(cso);
+}
+
+
+/**
+ * Check if a GL format/type combination is a match to the given pipe format.
+ * XXX probably move this to a re-usable place.
+ */
+static GLboolean
+compatible_formats(GLenum format, GLenum type, enum pipe_format pipeFormat)
+{
+   static const GLuint one = 1;
+   GLubyte littleEndian = *((GLubyte *) &one);
+
+   if (pipeFormat == PIPE_FORMAT_R8G8B8A8_UNORM &&
+       format == GL_RGBA &&
+       type == GL_UNSIGNED_BYTE &&
+       !littleEndian) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_R8G8B8A8_UNORM &&
+            format == GL_ABGR_EXT &&
+            type == GL_UNSIGNED_BYTE &&
+            littleEndian) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_A8R8G8B8_UNORM &&
+            format == GL_BGRA &&
+            type == GL_UNSIGNED_BYTE &&
+            littleEndian) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_R5G6B5_UNORM &&
+            format == GL_RGB &&
+            type == GL_UNSIGNED_SHORT_5_6_5) {
+      /* endian don't care */
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_R5G6B5_UNORM &&
+            format == GL_BGR &&
+            type == GL_UNSIGNED_SHORT_5_6_5_REV) {
+      /* endian don't care */
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_S8_UNORM &&
+            format == GL_STENCIL_INDEX &&
+            type == GL_UNSIGNED_BYTE) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_Z32_UNORM &&
+            format == GL_DEPTH_COMPONENT &&
+            type == GL_UNSIGNED_INT) {
+      return GL_TRUE;
+   }
+   /* XXX add more cases */
+   else {
+      return GL_FALSE;
+   }
+}
+
+
+/**
+ * Check if any per-fragment ops are enabled.
+ * XXX probably move this to a re-usable place.
+ */
+static GLboolean
+any_fragment_ops(const struct st_context *st)
+{
+   if (st->state.depth_stencil.alpha.enabled ||
+       st->state.depth_stencil.depth.enabled ||
+       st->state.blend.blend_enable ||
+       st->state.blend.logicop_enable)
+      /* XXX more checks */
+      return GL_TRUE;
+   else
+      return GL_FALSE;
+}
+
+
+/**
+ * Check if any pixel transfer ops are enabled.
+ * XXX probably move this to a re-usable place.
+ */
+static GLboolean
+any_pixel_transfer_ops(const struct st_context *st)
+{
+   if (st->ctx->Pixel.RedScale != 1.0 ||
+       st->ctx->Pixel.RedBias != 0.0 ||
+       st->ctx->Pixel.GreenScale != 1.0 ||
+       st->ctx->Pixel.GreenBias != 0.0 ||
+       st->ctx->Pixel.BlueScale != 1.0 ||
+       st->ctx->Pixel.BlueBias != 0.0 ||
+       st->ctx->Pixel.AlphaScale != 1.0 ||
+       st->ctx->Pixel.AlphaBias != 0.0 ||
+       st->ctx->Pixel.MapColorFlag)
+      /* XXX more checks */
+      return GL_TRUE;
+   else
+      return GL_FALSE;
+}
+
+
+/**
+ * Draw image with a blit, or other non-textured quad method.
+ */
+static void
+draw_blit(struct st_context *st,
+          GLsizei width, GLsizei height,
+          GLenum format, GLenum type, const GLvoid *pixels)
+{
+
+
+}
+
+
+static void
+draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+                    GLsizei width, GLsizei height, GLenum type,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLvoid *pixels)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct st_renderbuffer *strb;
+   struct pipe_surface *ps;
+   const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
+   GLint skipPixels;
+   ubyte *stmap;
+
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   strb = st_renderbuffer(ctx->DrawBuffer->
+                          Attachment[BUFFER_STENCIL].Renderbuffer);
+   ps = screen->get_tex_surface(screen, strb->texture, 0, 0, 0,
+                                PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* map the stencil buffer */
+   stmap = screen->surface_map(screen, ps, 
+                               PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* if width > MAX_WIDTH, have to process image in chunks */
+   skipPixels = 0;
+   while (skipPixels < width) {
+      const GLint spanX = x + skipPixels;
+      const GLint spanWidth = MIN2(width - skipPixels, MAX_WIDTH);
+      GLint row;
+      for (row = 0; row < height; row++) {
+         GLint spanY = y + row;
+         GLubyte values[MAX_WIDTH];
+         GLenum destType = GL_UNSIGNED_BYTE;
+         const GLvoid *source = _mesa_image_address2d(unpack, pixels,
+                                                      width, height,
+                                                      GL_COLOR_INDEX, type,
+                                                      row, skipPixels);
+         _mesa_unpack_stencil_span(ctx, spanWidth, destType, values,
+                                   type, source, unpack,
+                                   ctx->_ImageTransferState);
+         if (zoom) {
+            /*
+            _swrast_write_zoomed_stencil_span(ctx, x, y, spanWidth,
+                                              spanX, spanY, values);
+            */
+         }
+         else {
+            if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+               spanY = ctx->DrawBuffer->Height - spanY - 1;
+            }
+
+            switch (ps->format) {
+            case PIPE_FORMAT_S8_UNORM:
+               {
+                  ubyte *dest = stmap + spanY * ps->stride + spanX;
+                  memcpy(dest, values, spanWidth);
+               }
+               break;
+            case PIPE_FORMAT_S8Z24_UNORM:
+               {
+                  uint *dest = (uint *) (stmap + spanY * ps->stride + spanX*4);
+                  GLint k;
+                  for (k = 0; k < spanWidth; k++) {
+                     uint p = dest[k];
+                     p = (p & 0xffffff) | (values[k] << 24);
+                     dest[k] = p;
+                  }
+               }
+               break;
+            default:
+               assert(0);
+            }
+         }
+      }
+      skipPixels += spanWidth;
+   }
+
+   /* unmap the stencil buffer */
+   screen->surface_unmap(screen, ps);
+   pipe_surface_reference(&ps, NULL);
+}
+
+
+/**
+ * Called via ctx->Driver.DrawPixels()
+ */
+static void
+st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+              GLenum format, GLenum type,
+              const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels)
+{
+   struct st_fragment_program *stfp;
+   struct st_vertex_program *stvp;
+   struct st_context *st = ctx->st;
+   struct pipe_surface *ps;
+   GLuint bufferFormat;
+   const GLfloat *color;
+
+   if (format == GL_STENCIL_INDEX) {
+      draw_stencil_pixels(ctx, x, y, width, height, type, unpack, pixels);
+      return;
+   }
+
+   _mesa_set_vp_override( ctx, TRUE );
+   _mesa_update_state( ctx );
+
+   st_validate_state(st);
+
+   if (format == GL_DEPTH_COMPONENT) {
+      ps = st->state.framebuffer.zsbuf;
+      stfp = make_fragment_shader_z(ctx->st);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE);
+      color = ctx->Current.RasterColor;
+   }
+   else if (format == GL_STENCIL_INDEX) {
+      ps = st->state.framebuffer.zsbuf;
+      /* XXX special case - can't use texture map */
+      color = NULL;
+   }
+   else {
+      ps = st->state.framebuffer.cbufs[0];
+      stfp = combined_drawpix_fragment_program(ctx);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE);
+      color = NULL;
+   }
+
+   bufferFormat = ps->format;
+
+   if (1/*any_fragment_ops(st) ||
+       any_pixel_transfer_ops(st) ||
+       !compatible_formats(format, type, ps->format)*/) {
+      /* textured quad */
+      struct pipe_texture *pt
+         = make_texture(ctx->st, width, height, format, type, unpack, pixels);
+      if (pt) {
+         draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
+                            width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
+                            pt, stvp, stfp, color, GL_FALSE);
+         pipe_texture_reference(&pt, NULL);
+      }
+   }
+   else {
+      /* blit */
+      draw_blit(st, width, height, format, type, pixels);
+   }
+
+   _mesa_set_vp_override( ctx, FALSE );
+}
+
+
+
+static void
+copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
+                    GLsizei width, GLsizei height,
+                    GLint dstx, GLint dsty)
+{
+   struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer);
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct pipe_surface *psDraw;
+   ubyte *drawMap;
+   ubyte *buffer;
+   int i;
+
+   buffer = malloc(width * height * sizeof(ubyte));
+   if (!buffer) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)");
+      return;
+   }
+
+   /* this will do stencil pixel transfer ops */
+   st_read_stencil_pixels(ctx, srcx, srcy, width, height, GL_UNSIGNED_BYTE,
+                          &ctx->DefaultPacking, buffer);
+
+   psDraw = screen->get_tex_surface(screen, rbDraw->texture, 0, 0, 0,
+                                    PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   assert(psDraw->block.width == 1);
+   assert(psDraw->block.height == 1);
+   
+   /* map the stencil buffer */
+   drawMap = screen->surface_map(screen, psDraw, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* draw */
+   /* XXX PixelZoom not handled yet */
+   for (i = 0; i < height; i++) {
+      ubyte *dst;
+      const ubyte *src;
+      int y;
+
+      y = dsty + i;
+
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         y = ctx->DrawBuffer->Height - y - 1;
+      }
+
+      dst = drawMap + y * psDraw->stride + dstx * psDraw->block.size;
+      src = buffer + i * width;
+
+      switch (psDraw->format) {
+      case PIPE_FORMAT_S8Z24_UNORM:
+         {
+            uint *dst4 = (uint *) dst;
+            int j;
+            for (j = 0; j < width; j++) {
+               *dst4 = (*dst4 & 0xffffff) | (src[j] << 24);
+               dst4++;
+            }
+         }
+         break;
+      case PIPE_FORMAT_S8_UNORM:
+         memcpy(dst, src, width);
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   free(buffer);
+
+   /* unmap the stencil buffer */
+   screen->surface_unmap(screen, psDraw);
+   pipe_surface_reference(&psDraw, NULL);
+}
+
+
+static void
+st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
+              GLsizei width, GLsizei height,
+              GLint dstx, GLint dsty, GLenum type)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct st_renderbuffer *rbRead;
+   struct st_vertex_program *stvp;
+   struct st_fragment_program *stfp;
+   struct pipe_surface *psTex;
+   struct pipe_texture *pt;
+   GLfloat *color;
+   enum pipe_format srcFormat, texFormat;
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   st_validate_state(st);
+
+   if (type == GL_STENCIL) {
+      /* can't use texturing to do stencil */
+      copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
+      return;
+   }
+
+   if (type == GL_COLOR) {
+      rbRead = st_get_color_read_renderbuffer(ctx);
+      color = NULL;
+      stfp = combined_drawpix_fragment_program(ctx);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE);
+   }
+   else {
+      assert(type == GL_DEPTH);
+      rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
+      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
+      stfp = make_fragment_shader_z(ctx->st);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE);
+   }
+
+   srcFormat = rbRead->texture->format;
+
+   if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      texFormat = srcFormat;
+   }
+   else {
+      /* srcFormat can't be used as a texture format */
+      if (type == GL_DEPTH) {
+         texFormat = st_choose_format(pipe, GL_DEPTH_COMPONENT, PIPE_TEXTURE_2D, 
+                                      PIPE_TEXTURE_USAGE_DEPTH_STENCIL);
+         assert(texFormat != PIPE_FORMAT_NONE); /* XXX no depth texture formats??? */
+      }
+      else {
+         /* default color format */
+         texFormat = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE_2D, 
+                                      PIPE_TEXTURE_USAGE_SAMPLER);
+         assert(texFormat != PIPE_FORMAT_NONE);
+      }
+   }
+
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0,
+                          width, height, 1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   if (!pt)
+      return;
+
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      srcy = ctx->DrawBuffer->Height - srcy - height;
+   }
+
+   if (srcFormat == texFormat) {
+      /* copy source framebuffer surface into mipmap/texture */
+      struct pipe_surface *psRead = screen->get_tex_surface(screen,
+                                       rbRead->texture, 0, 0, 0,
+                                       PIPE_BUFFER_USAGE_GPU_READ);
+      psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, 
+                                      PIPE_BUFFER_USAGE_GPU_WRITE );
+      pipe->surface_copy(pipe,
+                         FALSE,
+			 psTex, /* dest */
+			 0, 0, /* destx/y */
+			 psRead,
+			 srcx, srcy, width, height);
+      pipe_surface_reference(&psRead, NULL);
+   }
+   else {
+      /* CPU-based fallback/conversion */
+      struct pipe_surface *psRead = screen->get_tex_surface(screen,
+                                       rbRead->texture, 0, 0, 0,
+                                       PIPE_BUFFER_USAGE_CPU_READ);
+
+      psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, 
+                                      PIPE_BUFFER_USAGE_CPU_WRITE );
+
+      if (type == GL_COLOR) {
+         /* alternate path using get/put_tile() */
+         GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+         pipe_get_tile_rgba(psRead, srcx, srcy, width, height, buf);
+         pipe_put_tile_rgba(psTex, 0, 0, width, height, buf);
+
+         free(buf);
+      }
+      else {
+         /* GL_DEPTH */
+         GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint));
+         pipe_get_tile_z(psRead, srcx, srcy, width, height, buf);
+         pipe_put_tile_z(psTex, 0, 0, width, height, buf);
+         free(buf);
+      }
+      pipe_surface_reference(&psRead, NULL);
+   }
+
+   pipe_surface_reference(&psTex, NULL);
+
+   /* draw textured quad */
+   draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
+                      width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
+                      pt, stvp, stfp, color, GL_TRUE);
+
+   pipe_texture_reference(&pt, NULL);
+}
+
+
+
+void st_init_drawpixels_functions(struct dd_function_table *functions)
+{
+   functions->DrawPixels = st_DrawPixels;
+   functions->CopyPixels = st_CopyPixels;
+}
+
+
+void
+st_destroy_drawpix(struct st_context *st)
+{
+   st_reference_fragprog(st, &st->drawpix.z_shader, NULL);
+   st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
+   st_reference_vertprog(st, &st->drawpix.vert_shaders[0], NULL);
+   st_reference_vertprog(st, &st->drawpix.vert_shaders[1], NULL);
+}
+
+
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h
new file mode 100644
index 00000000000..26fe864d185
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_drawpixels.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_DRAWPIXELS_H
+#define ST_CB_DRAWPIXELS_H
+
+
+extern void st_init_drawpixels_functions(struct dd_function_table *functions);
+
+extern void
+st_destroy_drawpix(struct st_context *st);
+
+
+#endif /* ST_CB_DRAWPIXELS_H */
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
new file mode 100644
index 00000000000..9af648b8d0f
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -0,0 +1,458 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Framebuffer/renderbuffer functions.
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "st_context.h"
+#include "st_cb_fbo.h"
+#include "st_cb_texture.h"
+#include "st_format.h"
+#include "st_public.h"
+#include "st_texture.h"
+
+
+
+/**
+ * Compute the renderbuffer's Red/Green/EtcBit fields from the pipe format.
+ */
+static int
+init_renderbuffer_bits(struct st_renderbuffer *strb,
+                       enum pipe_format pipeFormat)
+{
+   struct pipe_format_info info;
+
+   if (!st_get_format_info( pipeFormat, &info )) {
+      assert( 0 );
+   }
+
+   strb->Base._ActualFormat = info.base_format;
+   strb->Base.RedBits = info.red_bits;
+   strb->Base.GreenBits = info.green_bits;
+   strb->Base.BlueBits = info.blue_bits;
+   strb->Base.AlphaBits = info.alpha_bits;
+   strb->Base.DepthBits = info.depth_bits;
+   strb->Base.StencilBits = info.stencil_bits;
+   strb->Base.DataType = st_format_datatype(pipeFormat);
+
+   return info.size;
+}
+
+/**
+ * gl_renderbuffer::AllocStorage()
+ * This is called to allocate the original drawing surface, and
+ * during window resize.
+ */
+static GLboolean
+st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                              GLenum internalFormat,
+                              GLuint width, GLuint height)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+   struct pipe_texture template;
+   unsigned surface_usage;
+
+   /* Free the old surface and texture
+    */
+   pipe_surface_reference( &strb->surface, NULL );
+   pipe_texture_reference( &strb->texture, NULL );
+
+
+   memset(&template, 0, sizeof(template));
+
+   if (strb->format != PIPE_FORMAT_NONE) {
+      template.format = strb->format;
+   }
+   else {
+      template.format = st_choose_renderbuffer_format(pipe, internalFormat);
+   }
+
+   strb->Base.Width  = width;
+   strb->Base.Height = height;
+   init_renderbuffer_bits(strb, template.format);
+
+   template.target = PIPE_TEXTURE_2D;
+   template.compressed = 0;
+   pf_get_block(template.format, &template.block);
+   template.width[0] = width;
+   template.height[0] = height;
+   template.depth[0] = 1;
+   template.last_level = 0;
+   template.nr_samples = rb->NumSamples;
+
+   if (pf_is_depth_stencil(template.format)) {
+      template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   }
+   else {
+      template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                            PIPE_TEXTURE_USAGE_RENDER_TARGET);
+   }
+
+
+   /* Probably need dedicated flags for surface usage too: 
+    */
+   surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
+                    PIPE_BUFFER_USAGE_GPU_WRITE);
+#if 0
+                    PIPE_BUFFER_USAGE_CPU_READ |
+                    PIPE_BUFFER_USAGE_CPU_WRITE);
+#endif
+
+   strb->texture = pipe->screen->texture_create( pipe->screen,
+                                                 &template );
+
+   /* Special path for accum buffers.  
+    *
+    * Try a different surface format.  Since accum buffers are s/w
+    * only for now, the surface pixel format doesn't really matter,
+    * only that the buffer is large enough.
+    */
+   if (!strb->texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) 
+   {
+      /* Actually, just setting this usage value should be sufficient
+       * to tell the driver to go ahead and allocate the buffer, even
+       * if HW doesn't support the format.
+       */
+      template.tex_usage = 0;
+      surface_usage = (PIPE_BUFFER_USAGE_CPU_READ |
+                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      strb->texture = pipe->screen->texture_create( pipe->screen,
+                                                    &template );
+
+   }
+
+   if (!strb->texture) 
+      return FALSE;
+
+   strb->surface = pipe->screen->get_tex_surface( pipe->screen,
+                                                  strb->texture,
+                                                  0, 0, 0,
+                                                  surface_usage );
+
+   assert(strb->surface->texture);
+   assert(strb->surface->format);
+   assert(strb->surface->block.size);
+   assert(strb->surface->block.width);
+   assert(strb->surface->block.height);
+   assert(strb->surface->width == width);
+   assert(strb->surface->height == height);
+   assert(strb->surface->stride);
+
+
+   return strb->surface != NULL;
+}
+
+
+/**
+ * gl_renderbuffer::Delete()
+ */
+static void
+st_renderbuffer_delete(struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+   ASSERT(strb);
+   pipe_surface_reference(&strb->surface, NULL);
+   pipe_texture_reference(&strb->texture, NULL);
+   free(strb);
+}
+
+
+/**
+ * gl_renderbuffer::GetPointer()
+ */
+static void *
+null_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb,
+                 GLint x, GLint y)
+{
+   /* By returning NULL we force all software rendering to go through
+    * the span routines.
+    */
+#if 0
+   assert(0);  /* Should never get called with softpipe */
+#endif
+   return NULL;
+}
+
+
+/**
+ * Called via ctx->Driver.NewFramebuffer()
+ */
+static struct gl_framebuffer *
+st_new_framebuffer(GLcontext *ctx, GLuint name)
+{
+   /* XXX not sure we need to subclass gl_framebuffer for pipe */
+   return _mesa_new_framebuffer(ctx, name);
+}
+
+
+/**
+ * Called via ctx->Driver.NewRenderbuffer()
+ */
+static struct gl_renderbuffer *
+st_new_renderbuffer(GLcontext *ctx, GLuint name)
+{
+   struct st_renderbuffer *strb = CALLOC_STRUCT(st_renderbuffer);
+   if (strb) {
+      _mesa_init_renderbuffer(&strb->Base, name);
+      strb->Base.Delete = st_renderbuffer_delete;
+      strb->Base.AllocStorage = st_renderbuffer_alloc_storage;
+      strb->Base.GetPointer = null_get_pointer;
+      strb->format = PIPE_FORMAT_NONE;
+      return &strb->Base;
+   }
+   return NULL;
+}
+
+
+/**
+ * Allocate a renderbuffer for a an on-screen window (not a user-created
+ * renderbuffer).  The window system code determines the format.
+ */
+struct gl_renderbuffer *
+st_new_renderbuffer_fb(enum pipe_format format, int samples)
+{
+   struct st_renderbuffer *strb;
+
+   strb = CALLOC_STRUCT(st_renderbuffer);
+   if (!strb) {
+      _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer");
+      return NULL;
+   }
+
+   _mesa_init_renderbuffer(&strb->Base, 0);
+   strb->Base.ClassID = 0x4242; /* just a unique value */
+   strb->Base.NumSamples = samples;
+   strb->format = format;
+
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      strb->Base.InternalFormat = GL_RGBA;
+      strb->Base._BaseFormat = GL_RGBA;
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      strb->Base.InternalFormat = GL_DEPTH_COMPONENT16;
+      strb->Base._BaseFormat = GL_DEPTH_COMPONENT;
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      strb->Base.InternalFormat = GL_DEPTH_COMPONENT32;
+      strb->Base._BaseFormat = GL_DEPTH_COMPONENT;
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      strb->Base.InternalFormat = GL_DEPTH24_STENCIL8_EXT;
+      strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+      break;
+   case PIPE_FORMAT_S8_UNORM:
+      strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT;
+      strb->Base._BaseFormat = GL_STENCIL_INDEX;
+      break;
+   case DEFAULT_ACCUM_PIPE_FORMAT: /*PIPE_FORMAT_R16G16B16A16_SNORM*/
+      strb->Base.InternalFormat = GL_RGBA16;
+      strb->Base._BaseFormat = GL_RGBA;
+      break;
+   default:
+      _mesa_problem(NULL,
+		    "Unexpected format in st_new_renderbuffer_fb");
+      return NULL;
+   }
+
+   /* st-specific methods */
+   strb->Base.Delete = st_renderbuffer_delete;
+   strb->Base.AllocStorage = st_renderbuffer_alloc_storage;
+   strb->Base.GetPointer = null_get_pointer;
+
+   /* surface is allocated in st_renderbuffer_alloc_storage() */
+   strb->surface = NULL;
+
+   return &strb->Base;
+}
+
+
+
+
+/**
+ * Called via ctx->Driver.BindFramebufferEXT().
+ */
+static void
+st_bind_framebuffer(GLcontext *ctx, GLenum target,
+                    struct gl_framebuffer *fb, struct gl_framebuffer *fbread)
+{
+
+}
+
+/**
+ * Called by ctx->Driver.FramebufferRenderbuffer
+ */
+static void
+st_framebuffer_renderbuffer(GLcontext *ctx, 
+                            struct gl_framebuffer *fb,
+                            GLenum attachment,
+                            struct gl_renderbuffer *rb)
+{
+   /* XXX no need for derivation? */
+   _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb);
+}
+
+
+/**
+ * Called by ctx->Driver.RenderTexture
+ */
+static void
+st_render_texture(GLcontext *ctx,
+                  struct gl_framebuffer *fb,
+                  struct gl_renderbuffer_attachment *att)
+{
+   struct st_renderbuffer *strb;
+   struct gl_renderbuffer *rb;
+   struct pipe_texture *pt = st_get_texobj_texture(att->Texture);
+   struct st_texture_object *stObj;
+   const struct gl_texture_image *texImage =
+      att->Texture->Image[att->CubeMapFace][att->TextureLevel];
+
+   if (!pt) 
+      return;
+
+   assert(!att->Renderbuffer);
+
+   /* create new renderbuffer which wraps the texture image */
+   rb = st_new_renderbuffer(ctx, 0);
+   if (!rb) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture()");
+      return;
+   }
+
+   _mesa_reference_renderbuffer(&att->Renderbuffer, rb);
+   assert(rb->RefCount == 1);
+   rb->AllocStorage = NULL; /* should not get called */
+   strb = st_renderbuffer(rb);
+
+   /* get the texture for the texture object */
+   stObj = st_texture_object(att->Texture);
+
+   /* point renderbuffer at texobject */
+   strb->rtt = stObj;
+   strb->rtt_level = att->TextureLevel;
+   strb->rtt_face = att->CubeMapFace;
+   strb->rtt_slice = att->Zoffset;
+
+   rb->Width = texImage->Width2;
+   rb->Height = texImage->Height2;
+   /*printf("***** render to texture level %d: %d x %d\n", att->TextureLevel, rb->Width, rb->Height);*/
+
+   /*printf("***** pipe texture %d x %d\n", pt->width[0], pt->height[0]);*/
+
+   pipe_texture_reference( &strb->texture, pt );
+
+   pipe_surface_reference(&strb->surface, NULL);
+
+   /* the new surface will be created during framebuffer validation */
+
+   init_renderbuffer_bits(strb, pt->format);
+
+   /*
+   printf("RENDER TO TEXTURE obj=%p pt=%p surf=%p  %d x %d\n",
+          att->Texture, pt, strb->surface, rb->Width, rb->Height);
+   */
+
+   /* Invalidate buffer state so that the pipe's framebuffer state
+    * gets updated.
+    * That's where the new renderbuffer (which we just created) gets
+    * passed to the pipe as a (color/depth) render target.
+    */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+/**
+ * Called via ctx->Driver.FinishRenderTexture.
+ */
+static void
+st_finish_render_texture(GLcontext *ctx,
+                         struct gl_renderbuffer_attachment *att)
+{
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer);
+
+   if (!strb)
+      return;
+
+   st_flush( ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL );
+
+   if (strb->surface)
+      screen->tex_surface_release( screen, &strb->surface );
+
+   strb->rtt = NULL;
+
+   /*
+   printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface);
+   */
+
+   _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+
+   /* restore previous framebuffer state */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+
+void st_init_fbo_functions(struct dd_function_table *functions)
+{
+   functions->NewFramebuffer = st_new_framebuffer;
+   functions->NewRenderbuffer = st_new_renderbuffer;
+   functions->BindFramebuffer = st_bind_framebuffer;
+   functions->FramebufferRenderbuffer = st_framebuffer_renderbuffer;
+   functions->RenderTexture = st_render_texture;
+   functions->FinishRenderTexture = st_finish_render_texture;
+   /* no longer needed by core Mesa, drivers handle resizes...
+   functions->ResizeBuffers = st_resize_buffers;
+   */
+}
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
new file mode 100644
index 00000000000..44fa9fe9a4f
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -0,0 +1,71 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_FBO_H
+#define ST_CB_FBO_H
+
+
+#define DEFAULT_ACCUM_PIPE_FORMAT PIPE_FORMAT_R16G16B16A16_SNORM
+
+
+
+/**
+ * Derived renderbuffer class.  Just need to add a pointer to the
+ * pipe surface.
+ */
+struct st_renderbuffer
+{
+   struct gl_renderbuffer Base;
+   struct pipe_texture *texture;
+   struct pipe_surface *surface; /* temporary view into texture */
+   enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
+
+   struct st_texture_object *rtt;  /**< GL render to texture's texture */
+   int rtt_level, rtt_face, rtt_slice;
+
+   /** Render to texture state */
+   struct pipe_texture *texture_save;
+   struct pipe_surface *surface_save;
+};
+
+
+static INLINE struct st_renderbuffer *
+st_renderbuffer(struct gl_renderbuffer *rb)
+{
+   return (struct st_renderbuffer *) rb;
+}
+
+
+extern struct gl_renderbuffer *
+st_new_renderbuffer_fb(enum pipe_format format, int samples);
+
+extern void
+st_init_fbo_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_FBO_H */
diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c
new file mode 100644
index 00000000000..c7e8aa7cc56
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_feedback.c
@@ -0,0 +1,309 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+/**
+ * GL_SELECT and GL_FEEDBACK render modes.
+ * Basically, we use a private instance of the 'draw' module for doing
+ * selection/feedback.  It would be nice to use the transform_feedback
+ * hardware feature, but it's defined as happening pre-clip and we want
+ * post-clipped primitives.  Also, there's concerns about the efficiency
+ * of using the hardware for this anyway.
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/feedback.h"
+#include "main/macros.h"
+
+#include "vbo/vbo.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_draw.h"
+#include "st_cb_feedback.h"
+#include "st_cb_bufferobjects.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_cache.h"
+
+#include "draw/draw_context.h"
+#include "draw/draw_pipe.h"
+
+
+/**
+ * This is actually used for both feedback and selection.
+ */
+struct feedback_stage
+{
+   struct draw_stage stage;   /**< Base class */
+   GLcontext *ctx;            /**< Rendering context */
+   GLboolean reset_stipple_counter;
+};
+
+
+/**********************************************************************
+ * GL Feedback functions
+ **********************************************************************/
+
+static INLINE struct feedback_stage *
+feedback_stage( struct draw_stage *stage )
+{
+   return (struct feedback_stage *)stage;
+}
+
+
+static void
+feedback_vertex(GLcontext *ctx, const struct draw_context *draw,
+                const struct vertex_header *v)
+{
+   const struct st_context *st = ctx->st;
+   GLfloat win[4];
+   const GLfloat *color, *texcoord;
+   const GLfloat ci = 0;
+   GLuint slot;
+
+   /* Recall that Y=0=Top of window for Gallium wincoords */
+   win[0] = v->data[0][0];
+   win[1] = ctx->DrawBuffer->Height - v->data[0][1];
+   win[2] = v->data[0][2];
+   win[3] = 1.0F / v->data[0][3];
+
+   /* XXX
+    * When we compute vertex layout, save info about position of the
+    * color and texcoord attribs to use here.
+    */
+
+   slot = st->vertex_result_to_slot[VERT_RESULT_COL0];
+   if (slot != ~0U)
+      color = v->data[slot];
+   else
+      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
+
+   slot = st->vertex_result_to_slot[VERT_RESULT_TEX0];
+   if (slot != ~0U)
+      texcoord = v->data[slot];
+   else
+      texcoord = ctx->Current.Attrib[VERT_ATTRIB_TEX0];
+
+   _mesa_feedback_vertex(ctx, win, color, ci, texcoord);
+}
+
+
+static void
+feedback_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   struct draw_context *draw = stage->draw;
+   FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_POLYGON_TOKEN);
+   FEEDBACK_TOKEN(fs->ctx, (GLfloat) 3); /* three vertices */
+   feedback_vertex(fs->ctx, draw, prim->v[0]);
+   feedback_vertex(fs->ctx, draw, prim->v[1]);
+   feedback_vertex(fs->ctx, draw, prim->v[2]);
+}
+
+
+static void
+feedback_line( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   struct draw_context *draw = stage->draw;
+   if (fs->reset_stipple_counter) {
+      FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_LINE_RESET_TOKEN);
+      fs->reset_stipple_counter = GL_FALSE;
+   }
+   else {
+      FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_LINE_TOKEN);
+   }
+   feedback_vertex(fs->ctx, draw, prim->v[0]);
+   feedback_vertex(fs->ctx, draw, prim->v[1]);
+}
+
+
+static void
+feedback_point( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   struct draw_context *draw = stage->draw;
+   FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_POINT_TOKEN);
+   feedback_vertex(fs->ctx, draw, prim->v[0]);
+}
+
+
+static void
+feedback_flush( struct draw_stage *stage, unsigned flags )
+{
+   /* no-op */
+}
+
+
+static void
+feedback_reset_stipple_counter( struct draw_stage *stage )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   fs->reset_stipple_counter = GL_TRUE;
+}
+
+
+static void
+feedback_destroy( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+/**
+ * Create GL feedback drawing stage.
+ */
+static struct draw_stage *
+draw_glfeedback_stage(GLcontext *ctx, struct draw_context *draw)
+{
+   struct feedback_stage *fs = CALLOC_STRUCT(feedback_stage);
+
+   fs->stage.draw = draw;
+   fs->stage.next = NULL;
+   fs->stage.point = feedback_point;
+   fs->stage.line = feedback_line;
+   fs->stage.tri = feedback_tri;
+   fs->stage.flush = feedback_flush;
+   fs->stage.reset_stipple_counter = feedback_reset_stipple_counter;
+   fs->stage.destroy = feedback_destroy;
+   fs->ctx = ctx;
+
+   return &fs->stage;
+}
+
+
+
+/**********************************************************************
+ * GL Selection functions
+ **********************************************************************/
+
+static void
+select_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] );
+   _mesa_update_hitflag( fs->ctx, prim->v[1]->data[0][2] );
+   _mesa_update_hitflag( fs->ctx, prim->v[2]->data[0][2] );
+}
+
+static void
+select_line( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] );
+   _mesa_update_hitflag( fs->ctx, prim->v[1]->data[0][2] );
+}
+
+
+static void
+select_point( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] );
+}
+
+
+static void
+select_flush( struct draw_stage *stage, unsigned flags )
+{
+   /* no-op */
+}
+
+
+static void
+select_reset_stipple_counter( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+static void
+select_destroy( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+
+/**
+ * Create GL selection mode drawing stage.
+ */
+static struct draw_stage *
+draw_glselect_stage(GLcontext *ctx, struct draw_context *draw)
+{
+   struct feedback_stage *fs = CALLOC_STRUCT(feedback_stage);
+
+   fs->stage.draw = draw;
+   fs->stage.next = NULL;
+   fs->stage.point = select_point;
+   fs->stage.line = select_line;
+   fs->stage.tri = select_tri;
+   fs->stage.flush = select_flush;
+   fs->stage.reset_stipple_counter = select_reset_stipple_counter;
+   fs->stage.destroy = select_destroy;
+   fs->ctx = ctx;
+
+   return &fs->stage;
+}
+
+
+static void
+st_RenderMode(GLcontext *ctx, GLenum newMode )
+{
+   struct st_context *st = ctx->st;
+   struct draw_context *draw = st->draw;
+
+   if (newMode == GL_RENDER) {
+      /* restore normal VBO draw function */
+      vbo_set_draw_func(ctx, st_draw_vbo);
+   }
+   else if (newMode == GL_SELECT) {
+      if (!st->selection_stage)
+         st->selection_stage = draw_glselect_stage(ctx, draw);
+      draw_set_rasterize_stage(draw, st->selection_stage);
+      /* Plug in new vbo draw function */
+      vbo_set_draw_func(ctx, st_feedback_draw_vbo);
+   }
+   else {
+      if (!st->feedback_stage)
+         st->feedback_stage = draw_glfeedback_stage(ctx, draw);
+      draw_set_rasterize_stage(draw, st->feedback_stage);
+      /* Plug in new vbo draw function */
+      vbo_set_draw_func(ctx, st_feedback_draw_vbo);
+      /* need to generate/use a vertex program that emits pos/color/tex */
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+   }
+}
+
+
+
+void st_init_feedback_functions(struct dd_function_table *functions)
+{
+   functions->RenderMode = st_RenderMode;
+}
diff --git a/src/mesa/state_tracker/st_cb_feedback.h b/src/mesa/state_tracker/st_cb_feedback.h
new file mode 100644
index 00000000000..2559ba3817b
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_feedback.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_FEEDBACK_H
+#define ST_CB_FEEDBACK_H
+
+
+extern void
+st_init_feedback_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_FEEDBACK_H */
diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c
new file mode 100644
index 00000000000..f8621ab125f
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_flush.c
@@ -0,0 +1,149 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "st_context.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_flush.h"
+#include "st_cb_clear.h"
+#include "st_cb_fbo.h"
+#include "st_public.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "util/u_gen_mipmap.h"
+#include "util/u_blit.h"
+
+
+static INLINE GLboolean
+is_front_buffer_dirty(struct st_context *st)
+{
+   return st->frontbuffer_status == FRONT_STATUS_DIRTY;
+}
+
+
+/**
+ * Tell the screen to display the front color buffer on-screen.
+ */
+static void
+display_front_buffer(struct st_context *st)
+{
+   GLframebuffer *fb = st->ctx->DrawBuffer;
+   struct st_renderbuffer *strb
+      = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+   struct pipe_surface *front_surf = strb->surface;
+
+   /* Hook for copying "fake" frontbuffer if necessary:
+    */
+   st->pipe->screen->flush_frontbuffer( st->pipe->screen, front_surf,
+                                        st->pipe->priv );
+
+   /*
+   st->frontbuffer_status = FRONT_STATUS_UNDEFINED;
+   */
+}
+
+
+void st_flush( struct st_context *st, uint pipeFlushFlags,
+               struct pipe_fence_handle **fence )
+{
+   FLUSH_VERTICES(st->ctx, 0);
+
+   /* Release any vertex buffers that might potentially be accessed in
+    * successive frames:
+    */
+   st_flush_bitmap(st);
+   st_flush_clear(st);
+   util_blit_flush(st->blit);
+   util_gen_mipmap_flush(st->gen_mipmap);
+
+   st->pipe->flush( st->pipe, pipeFlushFlags, fence );
+}
+
+
+/**
+ * Flush, and wait for completion.
+ */
+void st_finish( struct st_context *st )
+{
+   struct pipe_fence_handle *fence = NULL;
+
+   st_flush(st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, &fence);
+
+   if(fence) {
+      st->pipe->screen->fence_finish(st->pipe->screen, fence, 0);
+      st->pipe->screen->fence_reference(st->pipe->screen, &fence, NULL);
+   }
+}
+
+
+
+/**
+ * Called via ctx->Driver.Flush()
+ */
+static void st_glFlush(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+
+   if (is_front_buffer_dirty(st)) {
+      st_finish(st);
+      display_front_buffer(st);
+   }
+   else {
+      st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL);
+   }
+}
+
+
+/**
+ * Called via ctx->Driver.Finish()
+ */
+static void st_glFinish(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+
+   st_finish(st);
+
+   if (is_front_buffer_dirty(st)) {
+      display_front_buffer(st);
+   }
+}
+
+
+void st_init_flush_functions(struct dd_function_table *functions)
+{
+   functions->Flush = st_glFlush;
+   functions->Finish = st_glFinish;
+}
diff --git a/src/mesa/state_tracker/st_cb_flush.h b/src/mesa/state_tracker/st_cb_flush.h
new file mode 100644
index 00000000000..c26f7792259
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_flush.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_FLUSH_H
+#define ST_CB_FLUSH_H
+
+
+extern void
+st_init_flush_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_FLUSH_H */
+
diff --git a/src/mesa/state_tracker/st_cb_get.c b/src/mesa/state_tracker/st_cb_get.c
new file mode 100644
index 00000000000..e7d7f03bc9b
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_get.c
@@ -0,0 +1,97 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 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.
+ * 
+ **************************************************************************/
+
+
+/**
+ * glGet functions
+ *
+ * \author Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+
+#include "pipe/p_defines.h"
+
+#include "st_cb_fbo.h"
+#include "st_cb_get.h"
+
+
+
+/**
+ * Examine the current color read buffer format to determine
+ * which GL pixel format/type combo is the best match.
+ */
+static void
+get_preferred_read_format_type(GLcontext *ctx, GLint *format, GLint *type)
+{
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct st_renderbuffer *strb = st_renderbuffer(fb->_ColorReadBuffer);
+
+   /* defaults */
+   *format = ctx->Const.ColorReadFormat;
+   *type = ctx->Const.ColorReadType;
+
+   if (strb) {
+      /* XXX could add more cases here... */
+      if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM) {
+         *format = GL_BGRA;
+         if (_mesa_little_endian())
+            *type = GL_UNSIGNED_INT_8_8_8_8_REV;
+         else
+            *type = GL_UNSIGNED_INT_8_8_8_8;
+      }
+   }
+}
+
+
+/**
+ * We only intercept the OES preferred ReadPixels format/type.
+ * Everything else goes to the default _mesa_GetIntegerv.
+ */
+static GLboolean 
+st_GetIntegerv(GLcontext *ctx, GLenum pname, GLint *params)
+{
+   GLint dummy;
+
+   switch (pname) {
+   case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES:
+      get_preferred_read_format_type(ctx, &dummy, params);
+      return GL_TRUE;
+   case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES:
+      get_preferred_read_format_type(ctx, params, &dummy);
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+
+void st_init_get_functions(struct dd_function_table *functions)
+{
+   functions->GetIntegerv = st_GetIntegerv;
+}
diff --git a/src/mesa/state_tracker/st_cb_get.h b/src/mesa/state_tracker/st_cb_get.h
new file mode 100644
index 00000000000..8e9f3e93060
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_get.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_GET_H
+#define ST_CB_GET_H
+
+
+extern void
+st_init_get_functions(struct dd_function_table *functions);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
new file mode 100644
index 00000000000..ea0fa20012e
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -0,0 +1,266 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/enums.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/program.h"
+#include "shader/programopt.h"
+#include "shader/shader_api.h"
+
+#include "cso_cache/cso_context.h"
+#include "draw/draw_context.h"
+
+#include "st_context.h"
+#include "st_program.h"
+#include "st_atom_shader.h"
+#include "st_cb_program.h"
+
+
+static GLuint SerialNo = 1;
+
+
+/**
+ * Called via ctx->Driver.BindProgram() to bind an ARB vertex or
+ * fragment program.
+ */
+static void st_bind_program( GLcontext *ctx,
+			     GLenum target, 
+			     struct gl_program *prog )
+{
+   struct st_context *st = st_context(ctx);
+
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB: 
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+      break;
+   }
+}
+
+
+/**
+ * Called via ctx->Driver.UseProgram() to bind a linked GLSL program
+ * (vertex shader + fragment shader).
+ */
+static void st_use_program( GLcontext *ctx,
+			    GLuint program )
+{
+   struct st_context *st = st_context(ctx);
+
+   st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+   st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+
+   _mesa_use_program(ctx, program);
+}
+
+
+
+/**
+ * Called via ctx->Driver.NewProgram() to allocate a new vertex or
+ * fragment program.
+ */
+static struct gl_program *st_new_program( GLcontext *ctx,
+					  GLenum target,
+					  GLuint id )
+{
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB: {
+      struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program);
+
+      prog->serialNo = SerialNo++;
+
+      return _mesa_init_vertex_program( ctx, 
+					&prog->Base,
+					target, 
+					id );
+   }
+
+   case GL_FRAGMENT_PROGRAM_ARB:
+   case GL_FRAGMENT_PROGRAM_NV: {
+      struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program);
+
+      prog->serialNo = SerialNo++;
+
+      return _mesa_init_fragment_program( ctx, 
+					  &prog->Base,
+					  target, 
+					  id );
+   }
+
+   default:
+      assert(0);
+      return NULL;
+   }
+}
+
+
+void
+st_delete_program(GLcontext *ctx, struct gl_program *prog)
+{
+   struct st_context *st = st_context(ctx);
+
+   switch( prog->Target ) {
+   case GL_VERTEX_PROGRAM_ARB:
+      {
+         struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
+
+         if (stvp->driver_shader) {
+            cso_delete_vertex_shader(st->cso_context, stvp->driver_shader);
+            stvp->driver_shader = NULL;
+         }
+
+         if (stvp->draw_shader) {
+#if FEATURE_feedback || FEATURE_drawpix
+            /* this would only have been allocated for the RasterPos path */
+            draw_delete_vertex_shader(st->draw, stvp->draw_shader);
+            stvp->draw_shader = NULL;
+#endif
+         }
+
+         if (stvp->state.tokens) {
+            FREE((void *) stvp->state.tokens);
+            stvp->state.tokens = NULL;
+         }
+      }
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      {
+         struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
+
+         if (stfp->driver_shader) {
+            cso_delete_fragment_shader(st->cso_context, stfp->driver_shader);
+            stfp->driver_shader = NULL;
+         }
+         
+         if (stfp->state.tokens) {
+            FREE((void *) stfp->state.tokens);
+            stfp->state.tokens = NULL;
+         }
+
+         if (stfp->bitmap_program) {
+            struct gl_program *prg = &stfp->bitmap_program->Base.Base;
+            _mesa_reference_program(ctx, &prg, NULL);
+            stfp->bitmap_program = NULL;
+         }
+
+         st_free_translated_vertex_programs(st, stfp->vertex_programs);
+      }
+      break;
+   default:
+      assert(0); /* problem */
+   }
+
+   /* delete base class */
+   _mesa_delete_program( ctx, prog );
+}
+
+
+static GLboolean st_is_program_native( GLcontext *ctx,
+				       GLenum target, 
+				       struct gl_program *prog )
+{
+   return GL_TRUE;
+}
+
+
+static void st_program_string_notify( GLcontext *ctx,
+				      GLenum target,
+				      struct gl_program *prog )
+{
+   struct st_context *st = st_context(ctx);
+
+   if (target == GL_FRAGMENT_PROGRAM_ARB) {
+      struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
+
+      stfp->serialNo++;
+
+      if (stfp->driver_shader) {
+         cso_delete_fragment_shader(st->cso_context, stfp->driver_shader);
+         stfp->driver_shader = NULL;
+      }
+
+      if (stfp->state.tokens) {
+         FREE((void *) stfp->state.tokens);
+         stfp->state.tokens = NULL;
+      }
+
+      stfp->param_state = stfp->Base.Base.Parameters->StateFlags;
+
+      if (st->fp == stfp)
+	 st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+   }
+   else if (target == GL_VERTEX_PROGRAM_ARB) {
+      struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
+
+      stvp->serialNo++;
+
+      if (stvp->driver_shader) {
+         cso_delete_vertex_shader(st->cso_context, stvp->driver_shader);
+         stvp->driver_shader = NULL;
+      }
+
+      if (stvp->draw_shader) {
+#if FEATURE_feedback || FEATURE_drawpix
+         /* this would only have been allocated for the RasterPos path */
+         draw_delete_vertex_shader(st->draw, stvp->draw_shader);
+         stvp->draw_shader = NULL;
+#endif
+      }
+
+      if (stvp->state.tokens) {
+         FREE((void *) stvp->state.tokens);
+         stvp->state.tokens = NULL;
+      }
+
+      stvp->param_state = stvp->Base.Base.Parameters->StateFlags;
+
+      if (st->vp == stvp)
+	 st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+   }
+}
+
+
+
+void st_init_program_functions(struct dd_function_table *functions)
+{
+   functions->BindProgram = st_bind_program;
+   functions->UseProgram = st_use_program;
+   functions->NewProgram = st_new_program;
+   functions->DeleteProgram = st_delete_program;
+   functions->IsProgramNative = st_is_program_native;
+   functions->ProgramStringNotify = st_program_string_notify;
+}
diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h
new file mode 100644
index 00000000000..0de96f2fd22
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_program.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ *
+ * Copyright 2008 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.
+ *
+ **************************************************************************/
+
+#ifndef ST_CB_PROGRAM_H
+#define ST_CB_PROGRAM_H
+
+
+extern void
+st_init_program_functions(struct dd_function_table *functions);
+
+extern void
+st_delete_program(GLcontext *ctx, struct gl_program *prog);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c
new file mode 100644
index 00000000000..21c2c7dd9ff
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_queryobj.c
@@ -0,0 +1,171 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+/**
+ * glBegin/EndQuery interface to pipe
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "st_context.h"
+#include "st_cb_queryobj.h"
+#include "st_public.h"
+
+
+struct st_query_object
+{
+   struct gl_query_object base;
+   struct pipe_query *pq;
+};
+
+
+/**
+ * Cast wrapper
+ */
+static struct st_query_object *
+st_query_object(struct gl_query_object *q)
+{
+   return (struct st_query_object *) q;
+}
+
+
+static struct gl_query_object *
+st_NewQueryObject(GLcontext *ctx, GLuint id)
+{
+   struct st_query_object *stq = CALLOC_STRUCT(st_query_object);
+   if (stq) {
+      stq->base.Id = id;
+      stq->base.Ready = GL_TRUE;
+      stq->pq = NULL;
+      return &stq->base;
+   }
+   return NULL;
+}
+
+
+
+static void
+st_DeleteQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   if (stq->pq) {
+      pipe->destroy_query(pipe, stq->pq);
+      stq->pq = NULL;
+   }
+
+   FREE(stq);
+}
+
+
+static void
+st_BeginQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   switch (q->Target) {
+   case GL_SAMPLES_PASSED_ARB:
+      if (!stq->pq)
+	 stq->pq = pipe->create_query( pipe, PIPE_QUERY_OCCLUSION_COUNTER );
+      break;
+   default:
+      assert(0);
+      return;
+   }
+
+   pipe->begin_query(pipe, stq->pq);
+}
+
+
+static void
+st_EndQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   pipe->end_query(pipe, stq->pq);
+}
+
+
+static void
+st_WaitQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   /* this function should only be called if we don't have a ready result */
+   assert(!stq->base.Ready);
+
+   while (!stq->base.Ready &&
+	  !pipe->get_query_result(pipe, 
+				  stq->pq,
+				  TRUE,
+				  &q->Result))
+   {
+      /* nothing */
+   }
+			    
+   q->Ready = GL_TRUE;
+}
+
+
+static void
+st_CheckQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   if (!q->Ready) {
+      q->Ready = pipe->get_query_result(pipe, 
+					stq->pq,
+					FALSE,
+					&q->Result);
+   }
+}
+
+
+
+
+void st_init_query_functions(struct dd_function_table *functions)
+{
+   functions->NewQueryObject = st_NewQueryObject;
+   functions->DeleteQuery = st_DeleteQuery;
+   functions->BeginQuery = st_BeginQuery;
+   functions->EndQuery = st_EndQuery;
+   functions->WaitQuery = st_WaitQuery;
+   functions->CheckQuery = st_CheckQuery;
+}
diff --git a/src/mesa/state_tracker/st_cb_queryobj.h b/src/mesa/state_tracker/st_cb_queryobj.h
new file mode 100644
index 00000000000..9220a212b60
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_queryobj.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_QUERYOBJ_H
+#define ST_CB_QUERYOBJ_H
+
+
+extern void
+st_init_query_functions(struct dd_function_table *functions);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
new file mode 100644
index 00000000000..8867ca5652e
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -0,0 +1,261 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+/**
+ * glRasterPos implementation.  Basically render a GL_POINT with our
+ * private draw module.  Plug in a special "rasterpos" stage at the end
+ * of the 'draw' pipeline to capture the results and update the current
+ * raster pos attributes.
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/feedback.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_draw.h"
+#include "st_cb_rasterpos.h"
+#include "st_draw.h"
+#include "draw/draw_context.h"
+#include "draw/draw_pipe.h"
+#include "shader/prog_instruction.h"
+#include "vbo/vbo.h"
+
+
+
+/**
+ * Our special drawing pipeline stage (replaces rasterization).
+ */
+struct rastpos_stage
+{
+   struct draw_stage stage;   /**< Base class */
+   GLcontext *ctx;            /**< Rendering context */
+
+   /* vertex attrib info we can setup once and re-use */
+   struct gl_client_array array[VERT_ATTRIB_MAX];
+   const struct gl_client_array *arrays[VERT_ATTRIB_MAX];
+   struct _mesa_prim prim;
+};
+
+
+static INLINE struct rastpos_stage *
+rastpos_stage( struct draw_stage *stage )
+{
+   return (struct rastpos_stage *) stage;
+}
+
+static void
+rastpos_flush( struct draw_stage *stage, unsigned flags )
+{
+   /* no-op */
+}
+
+static void
+rastpos_reset_stipple_counter( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+static void
+rastpos_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   /* should never get here */
+   assert(0);
+}
+
+static void
+rastpos_line( struct draw_stage *stage, struct prim_header *prim )
+{
+   /* should never get here */
+   assert(0);
+}
+
+static void
+rastpos_destroy(struct draw_stage *stage)
+{
+   free(stage);
+}
+
+
+/**
+ * Update a raster pos attribute from the vertex result if it's present,
+ * else copy the current attrib.
+ */
+static void
+update_attrib(GLcontext *ctx, const GLuint *outputMapping,
+              const struct vertex_header *vert,
+              GLfloat *dest,
+              GLuint result, GLuint defaultAttrib)
+{
+   const GLfloat *src;
+   const GLuint k = outputMapping[result];
+   if (k != ~0U)
+      src = vert->data[k];
+   else
+      src = ctx->Current.Attrib[defaultAttrib];
+   COPY_4V(dest, src);
+}
+
+
+/**
+ * Normally, this function would render a GL_POINT.
+ */
+static void
+rastpos_point(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct rastpos_stage *rs = rastpos_stage(stage);
+   GLcontext *ctx = rs->ctx;
+   struct st_context *st = ctx->st;
+   const GLfloat height = (GLfloat) ctx->DrawBuffer->Height;
+   const GLuint *outputMapping = st->vertex_result_to_slot;
+   const GLfloat *pos;
+   GLuint i;
+
+   /* if we get here, we didn't get clipped */
+   ctx->Current.RasterPosValid = GL_TRUE;
+
+   /* update raster pos */
+   pos = prim->v[0]->data[0];
+   ctx->Current.RasterPos[0] = pos[0];
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP)
+      ctx->Current.RasterPos[1] = height - pos[1]; /* invert Y */
+   else
+      ctx->Current.RasterPos[1] = pos[1];
+   ctx->Current.RasterPos[2] = pos[2];
+   ctx->Current.RasterPos[3] = pos[3];
+
+   /* update other raster attribs */
+   update_attrib(ctx, outputMapping, prim->v[0],
+                 ctx->Current.RasterColor,
+                 VERT_RESULT_COL0, VERT_ATTRIB_COLOR0);
+
+   update_attrib(ctx, outputMapping, prim->v[0],
+                 ctx->Current.RasterSecondaryColor,
+                 VERT_RESULT_COL1, VERT_ATTRIB_COLOR1);
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      update_attrib(ctx, outputMapping, prim->v[0],
+                    ctx->Current.RasterTexCoords[i],
+                    VERT_RESULT_TEX0 + i, VERT_ATTRIB_TEX0 + i);
+   }
+
+   if (ctx->RenderMode == GL_SELECT) {
+      _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] );
+   }
+}
+
+
+/**
+ * Create rasterpos "drawing" stage.
+ */
+static struct rastpos_stage *
+new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw)
+{
+   struct rastpos_stage *rs = CALLOC_STRUCT(rastpos_stage);
+   GLuint i;
+
+   rs->stage.draw = draw;
+   rs->stage.next = NULL;
+   rs->stage.point = rastpos_point;
+   rs->stage.line = rastpos_line;
+   rs->stage.tri = rastpos_tri;
+   rs->stage.flush = rastpos_flush;
+   rs->stage.destroy = rastpos_destroy;
+   rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter;
+   rs->stage.destroy = rastpos_destroy;
+   rs->ctx = ctx;
+
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      rs->array[i].Size = 4;
+      rs->array[i].Type = GL_FLOAT;
+      rs->array[i].Stride = 0;
+      rs->array[i].StrideB = 0;
+      rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i];
+      rs->array[i].Enabled = GL_TRUE;
+      rs->array[i].Normalized = GL_TRUE;
+      rs->array[i].BufferObj = NULL;
+      rs->arrays[i] = &rs->array[i];
+   }
+
+   rs->prim.mode = GL_POINTS;
+   rs->prim.indexed = 0;
+   rs->prim.begin = 1;
+   rs->prim.end = 1;
+   rs->prim.weak = 0;
+   rs->prim.start = 0;
+   rs->prim.count = 1;
+
+   return rs;
+}
+
+
+static void
+st_RasterPos(GLcontext *ctx, const GLfloat v[4])
+{
+   struct st_context *st = ctx->st;
+   struct draw_context *draw = st->draw;
+   struct rastpos_stage *rs;
+
+   if (st->rastpos_stage) {
+      /* get rastpos stage info */
+      rs = rastpos_stage(st->rastpos_stage);
+   }
+   else {
+      /* create rastpos draw stage */
+      rs = new_draw_rastpos_stage(ctx, draw);
+      st->rastpos_stage = &rs->stage;
+   }
+
+   /* plug our rastpos stage into the draw module */
+   draw_set_rasterize_stage(st->draw, st->rastpos_stage);
+
+   /* make sure everything's up to date */
+   st_validate_state(ctx->st);
+
+   /* This will get set only if rastpos_point(), above, gets called */
+   ctx->Current.RasterPosValid = GL_FALSE;
+
+   /* All vertex attribs but position were previously initialized above.
+    * Just plug in position pointer now.
+    */
+   rs->array[0].Ptr = (GLubyte *) v;
+
+   /* draw the point */
+   st_feedback_draw_vbo(ctx, rs->arrays, &rs->prim, 1, NULL, 0, 1);
+}
+
+
+
+void st_init_rasterpos_functions(struct dd_function_table *functions)
+{
+   functions->RasterPos = st_RasterPos;
+}
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.h b/src/mesa/state_tracker/st_cb_rasterpos.h
new file mode 100644
index 00000000000..2b992e1405e
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_rasterpos.h
@@ -0,0 +1,33 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_RASTERPOS_H
+#define ST_CB_RASTERPOS_H
+
+extern void st_init_rasterpos_functions(struct dd_function_table *functions);
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
new file mode 100644
index 00000000000..646eaff1903
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -0,0 +1,464 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+/**
+ * glReadPixels interface to pipe
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/bufferobj.h"
+#include "main/context.h"
+#include "main/image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "st_context.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_readpixels.h"
+#include "st_cb_fbo.h"
+#include "st_format.h"
+#include "st_public.h"
+
+
+/**
+ * Special case for reading stencil buffer.
+ * For color/depth we use get_tile().  For stencil, map the stencil buffer.
+ */
+void
+st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+                       GLsizei width, GLsizei height, GLenum type,
+                       const struct gl_pixelstore_attrib *packing,
+                       GLvoid *pixels)
+{
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct st_renderbuffer *strb = st_renderbuffer(fb->_StencilBuffer);
+   struct pipe_surface *ps;
+   ubyte *stmap;
+   GLint j;
+
+   /* Create a CPU-READ surface/view into the renderbuffer's texture */
+   ps = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                PIPE_BUFFER_USAGE_CPU_READ);
+
+   /* map the stencil buffer */
+   stmap = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ);
+
+   /* width should never be > MAX_WIDTH since we did clipping earlier */
+   ASSERT(width <= MAX_WIDTH);
+
+   /* process image row by row */
+   for (j = 0; j < height; j++, y++) {
+      GLvoid *dest;
+      GLstencil values[MAX_WIDTH];
+      GLint srcY;
+
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         srcY = ctx->DrawBuffer->Height - y - 1;
+      }
+      else {
+         srcY = y;
+      }
+
+      /* get stencil values */
+      switch (ps->format) {
+      case PIPE_FORMAT_S8_UNORM:
+         {
+            const ubyte *src = stmap + srcY * ps->stride + x;
+            memcpy(values, src, width);
+         }
+         break;
+      case PIPE_FORMAT_S8Z24_UNORM:
+         {
+            const uint *src = (uint *) (stmap + srcY * ps->stride + x*4);
+            GLint k;
+            for (k = 0; k < width; k++) {
+               values[k] = src[k] >> 24;
+            }
+         }
+         break;
+      case PIPE_FORMAT_Z24S8_UNORM:
+         {
+            const uint *src = (uint *) (stmap + srcY * ps->stride + x*4);
+            GLint k;
+            for (k = 0; k < width; k++) {
+               values[k] = src[k] & 0xff;
+            }
+         }
+         break;
+      default:
+         assert(0);
+      }
+
+      /* store */
+      dest = _mesa_image_address2d(packing, pixels, width, height,
+                                   GL_STENCIL_INDEX, type, j, 0);
+
+      _mesa_pack_stencil_span(ctx, width, type, dest, values, packing);
+   }
+
+
+   /* unmap the stencil buffer */
+   screen->surface_unmap(screen, ps);
+   pipe_surface_reference(&ps, NULL);
+}
+
+
+/**
+ * Return renderbuffer to use for reading color pixels for glRead/CopyPixel
+ * commands.
+ * Special care is needed for the front buffer.
+ */
+struct st_renderbuffer *
+st_get_color_read_renderbuffer(GLcontext *ctx)
+{
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct st_renderbuffer *strb =
+      st_renderbuffer(fb->_ColorReadBuffer);
+   struct st_renderbuffer *front = 
+      st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+
+   if (strb == front
+       && ctx->st->frontbuffer_status == FRONT_STATUS_COPY_OF_BACK) {
+      /* reading from front color buffer, which is a logical copy of the
+       * back color buffer.
+       */
+      struct st_renderbuffer *back = 
+         st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+      strb = back;
+   }
+
+   return strb;
+}
+
+
+/**
+ * Try to do glReadPixels in a fast manner for common cases.
+ * \return GL_TRUE for success, GL_FALSE for failure
+ */
+static GLboolean
+st_fast_readpixels(GLcontext *ctx, struct st_renderbuffer *strb,
+                   GLint x, GLint y, GLsizei width, GLsizei height,
+                   GLenum format, GLenum type,
+                   const struct gl_pixelstore_attrib *pack,
+                   GLvoid *dest)
+{
+   enum combination {
+      A8R8G8B8_UNORM_TO_RGBA_UBYTE,
+      A8R8G8B8_UNORM_TO_RGB_UBYTE,
+      A8R8G8B8_UNORM_TO_BGRA_UINT
+   } combo;
+
+   if (ctx->_ImageTransferState)
+      return GL_FALSE;
+
+   if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM &&
+       format == GL_RGBA && type == GL_UNSIGNED_BYTE) {
+      combo = A8R8G8B8_UNORM_TO_RGBA_UBYTE;
+   }
+   else if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM &&
+            format == GL_RGB && type == GL_UNSIGNED_BYTE) {
+      combo = A8R8G8B8_UNORM_TO_RGB_UBYTE;
+   }
+   else if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM &&
+            format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV) {
+      combo = A8R8G8B8_UNORM_TO_BGRA_UINT;
+   }
+   else {
+      return GL_FALSE;
+   }
+
+   /*printf("st_fast_readpixels combo %d\n", (GLint) combo);*/
+
+   {
+      struct pipe_context *pipe = ctx->st->pipe;
+      struct pipe_screen *screen = pipe->screen;
+      struct pipe_surface *surf;
+      const GLubyte *map;
+      GLubyte *dst;
+      GLint row, col, dy, dstStride;
+
+      surf = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_READ);
+      if (!surf) {
+         return GL_FALSE;
+      }
+
+      map = screen->surface_map(screen, surf, PIPE_BUFFER_USAGE_CPU_READ);
+      if (!map) {
+         pipe_surface_reference(&surf, NULL);
+         return GL_FALSE;
+      }
+
+      if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+         y = surf->height - y - 1;
+         dy = -1;
+      }
+      else {
+         dy = 1;
+      }
+
+      dst = _mesa_image_address2d(pack, dest, width, height,
+                                  format, type, 0, 0);
+      dstStride = _mesa_image_row_stride(pack, width, format, type);
+
+      switch (combo) {
+      case A8R8G8B8_UNORM_TO_RGBA_UBYTE:
+         for (row = 0; row < height; row++) {
+            const GLubyte *src = map + y * surf->stride + x * 4;
+            for (col = 0; col < width; col++) {
+               GLuint pixel = ((GLuint *) src)[col];
+               dst[col*4+0] = (pixel >> 16) & 0xff;
+               dst[col*4+1] = (pixel >>  8) & 0xff;
+               dst[col*4+2] = (pixel >>  0) & 0xff;
+               dst[col*4+3] = (pixel >> 24) & 0xff;
+            }
+            dst += dstStride;
+            y += dy;
+         }
+         break;
+      case A8R8G8B8_UNORM_TO_RGB_UBYTE:
+         for (row = 0; row < height; row++) {
+            const GLubyte *src = map + y * surf->stride + x * 4;
+            for (col = 0; col < width; col++) {
+               GLuint pixel = ((GLuint *) src)[col];
+               dst[col*3+0] = (pixel >> 16) & 0xff;
+               dst[col*3+1] = (pixel >>  8) & 0xff;
+               dst[col*3+2] = (pixel >>  0) & 0xff;
+            }
+            dst += dstStride;
+            y += dy;
+         }
+         break;
+      case A8R8G8B8_UNORM_TO_BGRA_UINT:
+         for (row = 0; row < height; row++) {
+            const GLubyte *src = map + y * surf->stride + x * 4;
+            memcpy(dst, src, 4 * width);
+            dst += dstStride;
+            y += dy;
+         }
+         break;
+      default:
+         ; /* nothing */
+      }
+
+      screen->surface_unmap(screen, surf);
+      pipe_surface_reference(&surf, NULL);
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Do glReadPixels by getting rows from the framebuffer surface with
+ * get_tile().  Convert to requested format/type with Mesa image routines.
+ * Image transfer ops are done in software too.
+ */
+static void
+st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+              GLenum format, GLenum type,
+              const struct gl_pixelstore_attrib *pack,
+              GLvoid *dest)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   GLfloat temp[MAX_WIDTH][4];
+   const GLbitfield transferOps = ctx->_ImageTransferState;
+   GLsizei i, j;
+   GLint yStep, dfStride;
+   GLfloat *df;
+   struct st_renderbuffer *strb;
+   struct gl_pixelstore_attrib clippedPacking = *pack;
+   struct pipe_surface *surf;
+
+   assert(ctx->ReadBuffer->Width > 0);
+
+   /* XXX convolution not done yet */
+   assert((transferOps & IMAGE_CONVOLUTION_BIT) == 0);
+
+   /* Do all needed clipping here, so that we can forget about it later */
+   if (!_mesa_clip_readpixels(ctx, &x, &y, &width, &height, &clippedPacking)) {
+      /* The ReadPixels surface is totally outside the window bounds */
+      return;
+   }
+
+   dest = _mesa_map_readpix_pbo(ctx, &clippedPacking, dest);
+   if (!dest)
+      return;
+
+   /* make sure rendering has completed */
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   if (format == GL_STENCIL_INDEX) {
+      st_read_stencil_pixels(ctx, x, y, width, height, type, pack, dest);
+      return;
+   }
+   else if (format == GL_DEPTH_COMPONENT) {
+      strb = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
+   }
+   else {
+      /* Read color buffer */
+      strb = st_get_color_read_renderbuffer(ctx);
+   }
+
+   if (!strb)
+      return;
+
+   /* try a fast-path readpixels before anything else */
+   if (st_fast_readpixels(ctx, strb, x, y, width, height,
+                          format, type, pack, dest)) {
+      /* success! */
+      _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+      return;
+   }
+
+   if (format == GL_RGBA && type == GL_FLOAT) {
+      /* write tile(row) directly into user's buffer */
+      df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width,
+                                             height, format, type, 0, 0);
+      dfStride = width * 4;
+   }
+   else {
+      /* write tile(row) into temp row buffer */
+      df = (GLfloat *) temp;
+      dfStride = 0;
+   }
+
+   /* determine bottom-to-top vs. top-to-bottom order */
+   if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+      y = strb->Base.Height - 1 - y;
+      yStep = -1;
+   }
+   else {
+      yStep = 1;
+   }
+
+   /* Create a CPU-READ surface/view into the renderbuffer's texture */
+   surf = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_CPU_READ);
+
+   /*
+    * Copy pixels from pipe_surface to user memory
+    */
+   {
+      /* dest of first pixel in client memory */
+      GLubyte *dst = _mesa_image_address2d(&clippedPacking, dest, width,
+                                           height, format, type, 0, 0);
+      /* dest row stride */
+      const GLint dstStride = _mesa_image_row_stride(&clippedPacking, width,
+                                                     format, type);
+
+      if (surf->format == PIPE_FORMAT_S8Z24_UNORM ||
+          surf->format == PIPE_FORMAT_X8Z24_UNORM) {
+         if (format == GL_DEPTH_COMPONENT) {
+            for (i = 0; i < height; i++) {
+               GLuint ztemp[MAX_WIDTH];
+               GLfloat zfloat[MAX_WIDTH];
+               const double scale = 1.0 / ((1 << 24) - 1);
+               pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0);
+               y += yStep;
+               for (j = 0; j < width; j++) {
+                  zfloat[j] = (float) (scale * (ztemp[j] & 0xffffff));
+               }
+               _mesa_pack_depth_span(ctx, width, dst, type,
+                                     zfloat, &clippedPacking);
+               dst += dstStride;
+            }
+         }
+         else {
+            /* untested, but simple: */
+            assert(format == GL_DEPTH_STENCIL_EXT);
+            for (i = 0; i < height; i++) {
+               pipe_get_tile_raw(surf, x, y, width, 1, dst, 0);
+               y += yStep;
+               dst += dstStride;
+            }
+         }
+      }
+      else if (surf->format == PIPE_FORMAT_Z16_UNORM) {
+         for (i = 0; i < height; i++) {
+            GLushort ztemp[MAX_WIDTH];
+            GLfloat zfloat[MAX_WIDTH];
+            const double scale = 1.0 / 0xffff;
+            pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0);
+            y += yStep;
+            for (j = 0; j < width; j++) {
+               zfloat[j] = (float) (scale * ztemp[j]);
+            }
+            _mesa_pack_depth_span(ctx, width, dst, type,
+                                  zfloat, &clippedPacking);
+            dst += dstStride;
+         }
+      }
+      else if (surf->format == PIPE_FORMAT_Z32_UNORM) {
+         for (i = 0; i < height; i++) {
+            GLuint ztemp[MAX_WIDTH];
+            GLfloat zfloat[MAX_WIDTH];
+            const double scale = 1.0 / 0xffffffff;
+            pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0);
+            y += yStep;
+            for (j = 0; j < width; j++) {
+               zfloat[j] = (float) (scale * ztemp[j]);
+            }
+            _mesa_pack_depth_span(ctx, width, dst, type,
+                                  zfloat, &clippedPacking);
+            dst += dstStride;
+         }
+      }
+      else {
+         /* RGBA format */
+         /* Do a row at a time to flip image data vertically */
+         for (i = 0; i < height; i++) {
+            pipe_get_tile_rgba(surf, x, y, width, 1, df);
+            y += yStep;
+            df += dfStride;
+            if (!dfStride) {
+               _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst,
+                                          &clippedPacking, transferOps);
+               dst += dstStride;
+            }
+         }
+      }
+   }
+
+   pipe_surface_reference(&surf, NULL);
+
+   _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+}
+
+
+void st_init_readpixels_functions(struct dd_function_table *functions)
+{
+   functions->ReadPixels = st_readpixels;
+}
diff --git a/src/mesa/state_tracker/st_cb_readpixels.h b/src/mesa/state_tracker/st_cb_readpixels.h
new file mode 100644
index 00000000000..9e151be51fb
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_readpixels.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_READPIXELS_H
+#define ST_CB_READPIXELS_H
+
+extern struct st_renderbuffer *
+st_get_color_read_renderbuffer(GLcontext *ctx);
+
+extern void
+st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+                       GLsizei width, GLsizei height, GLenum type,
+                       const struct gl_pixelstore_attrib *packing,
+                       GLvoid *pixels);
+
+extern void
+st_init_readpixels_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_READPIXELS_H */
diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c
new file mode 100644
index 00000000000..2036ccafbfb
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_strings.c
@@ -0,0 +1,84 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/version.h"
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "st_context.h"
+#include "st_cb_strings.h"
+
+#define ST_VERSION_STRING "0.2"
+
+static const GLubyte *
+st_get_string(GLcontext * ctx, GLenum name)
+{
+   struct st_context *st = st_context(ctx);
+   struct pipe_screen *screen = st->pipe->screen;
+
+   switch (name) {
+   case GL_VENDOR: {
+      const char *vendor = screen->get_vendor( screen );
+      const char *tungsten = "Tungsten Graphics, Inc.";
+
+      /* Tungsten Graphics, Inc. developed the state_tracker module
+       * (and much of Mesa), but the driver itself may come from elsewhere.
+       * The additional string allows "and XyzCorp" to reflect this.
+       */
+      if (vendor && strcmp(vendor, tungsten) != 0)
+	 util_snprintf(st->vendor, sizeof(st->vendor),
+                  "%s and %s", tungsten, vendor);
+      else
+	 util_snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten);
+
+      return (GLubyte *) st->vendor;
+   }
+
+   case GL_RENDERER:
+      util_snprintf(st->renderer, sizeof(st->renderer), "Gallium %s on %s", 
+               ST_VERSION_STRING,
+	       screen->get_name( screen ));
+
+      return (GLubyte *) st->renderer;
+
+   default:
+      return NULL;
+   }
+}
+
+
+void st_init_string_functions(struct dd_function_table *functions)
+{
+   functions->GetString = st_get_string;
+}
diff --git a/src/mesa/state_tracker/st_cb_strings.h b/src/mesa/state_tracker/st_cb_strings.h
new file mode 100644
index 00000000000..3b765aaa592
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_strings.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_STRINGS_H
+#define ST_CB_STRINGS_H
+
+
+extern void
+st_init_string_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_CLEAR_H */
+
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
new file mode 100644
index 00000000000..9279a768c82
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -0,0 +1,1549 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "main/imports.h"
+#if FEATURE_convolve
+#include "main/convolve.h"
+#endif
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/mipmap.h"
+#include "main/pixel.h"
+#include "main/texcompress.h"
+#include "main/texformat.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+#include "main/texstore.h"
+
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_cb_fbo.h"
+#include "state_tracker/st_cb_texture.h"
+#include "state_tracker/st_format.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_texture.h"
+#include "state_tracker/st_gen_mipmap.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_blit.h"
+
+
+#define DBG if (0) printf
+
+
+static enum pipe_texture_target
+gl_target_to_pipe(GLenum target)
+{
+   switch (target) {
+   case GL_TEXTURE_1D:
+      return PIPE_TEXTURE_1D;
+
+   case GL_TEXTURE_2D:
+   case GL_TEXTURE_RECTANGLE_NV:
+      return PIPE_TEXTURE_2D;
+
+   case GL_TEXTURE_3D:
+      return PIPE_TEXTURE_3D;
+
+   case GL_TEXTURE_CUBE_MAP_ARB:
+      return PIPE_TEXTURE_CUBE;
+
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+/**
+ * Return nominal bytes per texel for a compressed format, 0 for non-compressed
+ * format.
+ */
+static int
+compressed_num_bytes(GLuint mesaFormat)
+{
+   switch(mesaFormat) {
+#if FEATURE_texture_fxt1
+   case MESA_FORMAT_RGB_FXT1:
+   case MESA_FORMAT_RGBA_FXT1:
+#endif
+#if FEATURE_texture_s3tc
+   case MESA_FORMAT_RGB_DXT1:
+   case MESA_FORMAT_RGBA_DXT1:
+      return 2;
+   case MESA_FORMAT_RGBA_DXT3:
+   case MESA_FORMAT_RGBA_DXT5:
+      return 4;
+#endif
+   default:
+      return 0;
+   }
+}
+
+
+/** called via ctx->Driver.NewTextureImage() */
+static struct gl_texture_image *
+st_NewTextureImage(GLcontext * ctx)
+{
+   DBG("%s\n", __FUNCTION__);
+   (void) ctx;
+   return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image);
+}
+
+
+/** called via ctx->Driver.NewTextureObject() */
+static struct gl_texture_object *
+st_NewTextureObject(GLcontext * ctx, GLuint name, GLenum target)
+{
+   struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object);
+
+   DBG("%s\n", __FUNCTION__);
+   _mesa_initialize_texture_object(&obj->base, name, target);
+
+   return &obj->base;
+}
+
+/** called via ctx->Driver.DeleteTextureImage() */
+static void 
+st_DeleteTextureObject(GLcontext *ctx,
+                       struct gl_texture_object *texObj)
+{
+   struct st_texture_object *stObj = st_texture_object(texObj);
+   if (stObj->pt)
+      pipe_texture_reference(&stObj->pt, NULL);
+
+   _mesa_delete_texture_object(ctx, texObj);
+}
+
+
+/** called via ctx->Driver.FreeTexImageData() */
+static void
+st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage)
+{
+   struct st_texture_image *stImage = st_texture_image(texImage);
+
+   DBG("%s\n", __FUNCTION__);
+
+   if (stImage->pt) {
+      pipe_texture_reference(&stImage->pt, NULL);
+   }
+
+   if (texImage->Data) {
+      _mesa_align_free(texImage->Data);
+      texImage->Data = NULL;
+   }
+}
+
+
+/**
+ * From linux kernel i386 header files, copes with odd sizes better
+ * than COPY_DWORDS would:
+ * XXX Put this in src/mesa/main/imports.h ???
+ */
+#if defined(i386) || defined(__i386__)
+static INLINE void *
+__memcpy(void *to, const void *from, size_t n)
+{
+   int d0, d1, d2;
+   __asm__ __volatile__("rep ; movsl\n\t"
+                        "testb $2,%b4\n\t"
+                        "je 1f\n\t"
+                        "movsw\n"
+                        "1:\ttestb $1,%b4\n\t"
+                        "je 2f\n\t"
+                        "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2)
+                        :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from)
+                        :"memory");
+   return (to);
+}
+#else
+#define __memcpy(a,b,c) memcpy(a,b,c)
+#endif
+
+
+/**
+ * The system memcpy (at least on ubuntu 5.10) has problems copying
+ * to agp (writecombined) memory from a source which isn't 64-byte
+ * aligned - there is a 4x performance falloff.
+ *
+ * The x86 __memcpy is immune to this but is slightly slower
+ * (10%-ish) than the system memcpy.
+ *
+ * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but
+ * isn't much faster than x86_memcpy for agp copies.
+ * 
+ * TODO: switch dynamically.
+ */
+static void *
+do_memcpy(void *dest, const void *src, size_t n)
+{
+   if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) {
+      return __memcpy(dest, src, n);
+   }
+   else
+      return memcpy(dest, src, n);
+}
+
+
+static int
+logbase2(int n)
+{
+   GLint i = 1, log2 = 0;
+   while (n > i) {
+      i *= 2;
+      log2++;
+   }
+   return log2;
+}
+
+
+/**
+ * Allocate a pipe_texture object for the given st_texture_object using
+ * the given st_texture_image to guess the mipmap size/levels.
+ *
+ * [comments...]
+ * Otherwise, store it in memory if (Border != 0) or (any dimension ==
+ * 1).
+ *    
+ * Otherwise, if max_level >= level >= min_level, create texture with
+ * space for images from min_level down to max_level.
+ *
+ * Otherwise, create texture with space for images from (level 0)..(1x1).
+ * Consider pruning this texture at a validation if the saving is worth it.
+ */
+static void
+guess_and_alloc_texture(struct st_context *st,
+			struct st_texture_object *stObj,
+			const struct st_texture_image *stImage)
+{
+   GLuint firstLevel;
+   GLuint lastLevel;
+   GLuint width = stImage->base.Width2;  /* size w/out border */
+   GLuint height = stImage->base.Height2;
+   GLuint depth = stImage->base.Depth2;
+   GLuint i, comp_byte = 0;
+   enum pipe_format fmt;
+
+   DBG("%s\n", __FUNCTION__);
+
+   assert(!stObj->pt);
+
+   if (stObj->pt &&
+       (GLint) stImage->level > stObj->base.BaseLevel &&
+       (stImage->base.Width == 1 ||
+        (stObj->base.Target != GL_TEXTURE_1D &&
+         stImage->base.Height == 1) ||
+        (stObj->base.Target == GL_TEXTURE_3D &&
+         stImage->base.Depth == 1)))
+      return;
+
+   /* If this image disrespects BaseLevel, allocate from level zero.
+    * Usually BaseLevel == 0, so it's unlikely to happen.
+    */
+   if ((GLint) stImage->level < stObj->base.BaseLevel)
+      firstLevel = 0;
+   else
+      firstLevel = stObj->base.BaseLevel;
+
+
+   /* Figure out image dimensions at start level. 
+    */
+   for (i = stImage->level; i > firstLevel; i--) {
+      if (width != 1)
+         width <<= 1;
+      if (height != 1)
+         height <<= 1;
+      if (depth != 1)
+         depth <<= 1;
+   }
+
+   if (width == 0 || height == 0 || depth == 0) {
+      /* no texture needed */
+      return;
+   }
+
+   /* Guess a reasonable value for lastLevel.  This is probably going
+    * to be wrong fairly often and might mean that we have to look at
+    * resizable buffers, or require that buffers implement lazy
+    * pagetable arrangements.
+    */
+   if ((stObj->base.MinFilter == GL_NEAREST ||
+        stObj->base.MinFilter == GL_LINEAR) &&
+       stImage->level == firstLevel) {
+      lastLevel = firstLevel;
+   }
+   else {
+      GLuint l2width = logbase2(width);
+      GLuint l2height = logbase2(height);
+      GLuint l2depth = logbase2(depth);
+      lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth);
+   }
+
+   if (stImage->base.IsCompressed)
+      comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat);
+
+   fmt = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat);
+   stObj->pt = st_texture_create(st,
+                                 gl_target_to_pipe(stObj->base.Target),
+                                 fmt,
+                                 lastLevel,
+                                 width,
+                                 height,
+                                 depth,
+                                 comp_byte,
+                                 ( (pf_is_depth_stencil(fmt) ?
+                                   PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
+                                   PIPE_TEXTURE_USAGE_RENDER_TARGET) |
+                                   PIPE_TEXTURE_USAGE_SAMPLER ));
+
+   DBG("%s - success\n", __FUNCTION__);
+}
+
+
+/**
+ * Adjust pixel unpack params and image dimensions to strip off the
+ * texture border.
+ * Gallium doesn't support texture borders.  They've seldem been used
+ * and seldom been implemented correctly anyway.
+ * \param unpackNew  returns the new pixel unpack parameters
+ */
+static void
+strip_texture_border(GLint border,
+                     GLint *width, GLint *height, GLint *depth,
+                     const struct gl_pixelstore_attrib *unpack,
+                     struct gl_pixelstore_attrib *unpackNew)
+{
+   assert(border > 0);  /* sanity check */
+
+   *unpackNew = *unpack;
+
+   if (unpackNew->RowLength == 0)
+      unpackNew->RowLength = *width;
+
+   if (depth && unpackNew->ImageHeight == 0)
+      unpackNew->ImageHeight = *height;
+
+   unpackNew->SkipPixels += border;
+   if (height)
+      unpackNew->SkipRows += border;
+   if (depth)
+      unpackNew->SkipImages += border;
+
+   assert(*width >= 3);
+   *width = *width - 2 * border;
+   if (height && *height >= 3)
+      *height = *height - 2 * border;
+   if (depth && *depth >= 3)
+      *depth = *depth - 2 * border;
+}
+
+
+/**
+ * Do glTexImage1/2/3D().
+ */
+static void
+st_TexImage(GLcontext * ctx,
+            GLint dims,
+            GLenum target, GLint level,
+            GLint internalFormat,
+            GLint width, GLint height, GLint depth,
+            GLint border,
+            GLenum format, GLenum type, const void *pixels,
+            const struct gl_pixelstore_attrib *unpack,
+            struct gl_texture_object *texObj,
+            struct gl_texture_image *texImage,
+            GLsizei imageSize, int compressed)
+{
+   struct st_texture_object *stObj = st_texture_object(texObj);
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   GLint postConvWidth, postConvHeight;
+   GLint texelBytes, sizeInBytes;
+   GLuint dstRowStride;
+   struct gl_pixelstore_attrib unpackNB;
+
+   DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
+       _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
+
+   /* gallium does not support texture borders, strip it off */
+   if (border) {
+      strip_texture_border(border, &width, &height, &depth,
+                           unpack, &unpackNB);
+      unpack = &unpackNB;
+      texImage->Width = width;
+      texImage->Height = height;
+      texImage->Depth = depth;
+      texImage->Border = 0;
+      border = 0;
+   }
+
+   postConvWidth = width;
+   postConvHeight = height;
+
+   stImage->face = _mesa_tex_target_to_face(target);
+   stImage->level = level;
+
+#if FEATURE_convolve
+   if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) {
+      _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth,
+                                         &postConvHeight);
+   }
+#endif
+
+   /* choose the texture format */
+   texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat,
+                                                format, type);
+
+   _mesa_set_fetch_functions(texImage, dims);
+
+   if (texImage->TexFormat->TexelBytes == 0) {
+      /* must be a compressed format */
+      texelBytes = 0;
+      texImage->IsCompressed = GL_TRUE;
+      texImage->CompressedSize =
+	 ctx->Driver.CompressedTextureSize(ctx, texImage->Width,
+					   texImage->Height, texImage->Depth,
+					   texImage->TexFormat->MesaFormat);
+   }
+   else {
+      texelBytes = texImage->TexFormat->TexelBytes;
+      
+      /* Minimum pitch of 32 bytes */
+      if (postConvWidth * texelBytes < 32) {
+	 postConvWidth = 32 / texelBytes;
+	 texImage->RowStride = postConvWidth;
+      }
+      
+      /* we'll set RowStride elsewhere when the texture is a "mapped" state */
+      /*assert(texImage->RowStride == postConvWidth);*/
+   }
+
+   /* Release the reference to a potentially orphaned buffer.   
+    * Release any old malloced memory.
+    */
+   if (stImage->pt) {
+      pipe_texture_reference(&stImage->pt, NULL);
+      assert(!texImage->Data);
+   }
+   else if (texImage->Data) {
+      _mesa_align_free(texImage->Data);
+   }
+
+   if (width == 0 || height == 0 || depth == 0) {
+      /* stop after freeing old image */
+      return;
+   }
+
+   /* If this is the only mipmap level in the texture, could call
+    * bmBufferData with NULL data to free the old block and avoid
+    * waiting on any outstanding fences.
+    */
+   if (stObj->pt &&
+       (stObj->teximage_realloc ||
+        (/*stObj->pt->first_level == level &&*/
+         stObj->pt->last_level == level &&
+         stObj->pt->target != PIPE_TEXTURE_CUBE &&
+         !st_texture_match_image(stObj->pt, &stImage->base,
+                                 stImage->face, stImage->level)))) {
+
+      DBG("release it\n");
+      pipe_texture_reference(&stObj->pt, NULL);
+      assert(!stObj->pt);
+      stObj->teximage_realloc = FALSE;
+   }
+
+   if (!stObj->pt) {
+      guess_and_alloc_texture(ctx->st, stObj, stImage);
+      if (!stObj->pt) {
+         /* Probably out of memory.
+          * Try flushing any pending rendering, then retry.
+          */
+         st_finish(ctx->st);
+         guess_and_alloc_texture(ctx->st, stObj, stImage);
+         if (!stObj->pt) {
+            _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+            return;
+         }
+      }
+   }
+
+   assert(!stImage->pt);
+
+   if (stObj->pt &&
+       st_texture_match_image(stObj->pt, &stImage->base,
+                                 stImage->face, stImage->level)) {
+
+      pipe_texture_reference(&stImage->pt, stObj->pt);
+      assert(stImage->pt);
+   }
+
+   if (!stImage->pt)
+      DBG("XXX: Image did not fit into texture - storing in local memory!\n");
+
+   /* st_CopyTexImage calls this function with pixels == NULL, with
+    * the expectation that the texture will be set up but nothing
+    * more will be done.  This is where those calls return:
+    */
+   if (compressed) {
+      pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels,
+						      unpack,
+						      "glCompressedTexImage");
+   } else {
+      pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1,
+					   format, type,
+					   pixels, unpack, "glTexImage");
+   }
+   if (!pixels)
+      return;
+
+   if (stImage->pt) {
+      texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (stImage->surface)
+         dstRowStride = stImage->surface->stride;
+   }
+   else {
+      /* Allocate regular memory and store the image there temporarily.   */
+      if (texImage->IsCompressed) {
+         sizeInBytes = texImage->CompressedSize;
+         dstRowStride =
+            _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width);
+         assert(dims != 3);
+      }
+      else {
+         dstRowStride = postConvWidth * texelBytes;
+         sizeInBytes = depth * dstRowStride * postConvHeight;
+      }
+
+      texImage->Data = _mesa_align_malloc(sizeInBytes, 16);
+   }
+
+   if (!texImage->Data) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+      return;
+   }
+
+   DBG("Upload image %dx%dx%d row_len %x pitch %x\n",
+       width, height, depth, width * texelBytes, dstRowStride);
+
+   /* Copy data.  Would like to know when it's ok for us to eg. use
+    * the blitter to copy.  Or, use the hardware to do the format
+    * conversion and copy:
+    */
+   if (compressed) {
+      memcpy(texImage->Data, pixels, imageSize);
+   }
+   else {
+      GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height,
+						       format, type);
+      int i;
+      const GLubyte *src = (const GLubyte *) pixels;
+
+      for (i = 0; i++ < depth;) {
+	 if (!texImage->TexFormat->StoreImage(ctx, dims, 
+					      texImage->_BaseFormat, 
+					      texImage->TexFormat, 
+					      texImage->Data,
+					      0, 0, 0, /* dstX/Y/Zoffset */
+					      dstRowStride,
+					      texImage->ImageOffsets,
+					      width, height, 1,
+					      format, type, src, unpack)) {
+	    _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+	 }
+
+	 if (stImage->pt && i < depth) {
+	    st_texture_image_unmap(ctx->st, stImage);
+	    texImage->Data = st_texture_image_map(ctx->st, stImage, i,
+                                                  PIPE_BUFFER_USAGE_CPU_WRITE);
+	    src += srcImageStride;
+	 }
+      }
+   }
+
+   _mesa_unmap_teximage_pbo(ctx, unpack);
+
+   if (stImage->pt) {
+      st_texture_image_unmap(ctx->st, stImage);
+      texImage->Data = NULL;
+   }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
+}
+
+
+static void
+st_TexImage3D(GLcontext * ctx,
+              GLenum target, GLint level,
+              GLint internalFormat,
+              GLint width, GLint height, GLint depth,
+              GLint border,
+              GLenum format, GLenum type, const void *pixels,
+              const struct gl_pixelstore_attrib *unpack,
+              struct gl_texture_object *texObj,
+              struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 3, target, level,
+                 internalFormat, width, height, depth, border,
+                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+}
+
+
+static void
+st_TexImage2D(GLcontext * ctx,
+              GLenum target, GLint level,
+              GLint internalFormat,
+              GLint width, GLint height, GLint border,
+              GLenum format, GLenum type, const void *pixels,
+              const struct gl_pixelstore_attrib *unpack,
+              struct gl_texture_object *texObj,
+              struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 2, target, level,
+                 internalFormat, width, height, 1, border,
+                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+}
+
+
+static void
+st_TexImage1D(GLcontext * ctx,
+              GLenum target, GLint level,
+              GLint internalFormat,
+              GLint width, GLint border,
+              GLenum format, GLenum type, const void *pixels,
+              const struct gl_pixelstore_attrib *unpack,
+              struct gl_texture_object *texObj,
+              struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 1, target, level,
+                 internalFormat, width, 1, 1, border,
+                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+}
+
+
+static void
+st_CompressedTexImage2D(GLcontext *ctx, GLenum target, GLint level,
+                        GLint internalFormat,
+                        GLint width, GLint height, GLint border,
+                        GLsizei imageSize, const GLvoid *data,
+                        struct gl_texture_object *texObj,
+                        struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 2, target, level,
+		 internalFormat, width, height, 1, border,
+		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1);
+}
+
+
+/**
+ * Need to map texture image into memory before copying image data,
+ * then unmap it.
+ */
+static void
+st_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
+                 GLenum format, GLenum type, GLvoid * pixels,
+                 struct gl_texture_object *texObj,
+                 struct gl_texture_image *texImage, int compressed)
+{
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack,
+                                                    texImage->Width,
+						    texImage->Height,
+                                                    format, type);
+   GLuint depth;
+   GLuint i;
+   GLubyte *dest;
+
+   /* Map */
+   if (stImage->pt) {
+      /* Image is stored in hardware format in a buffer managed by the
+       * kernel.  Need to explicitly map and unmap it.
+       */
+      texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
+                                            PIPE_BUFFER_USAGE_CPU_READ);
+      texImage->RowStride = stImage->surface->stride / stImage->pt->block.size;
+   }
+   else {
+      /* Otherwise, the image should actually be stored in
+       * texImage->Data.  This is pretty confusing for
+       * everybody, I'd much prefer to separate the two functions of
+       * texImage->Data - storage for texture images in main memory
+       * and access (ie mappings) of images.  In other words, we'd
+       * create a new texImage->Map field and leave Data simply for
+       * storage.
+       */
+      assert(texImage->Data);
+   }
+
+   depth = texImage->Depth;
+   texImage->Depth = 1;
+
+   dest = (GLubyte *) pixels;
+
+   for (i = 0; i++ < depth;) {
+      if (compressed) {
+	 _mesa_get_compressed_teximage(ctx, target, level, dest,
+				       texObj, texImage);
+      } else {
+	 _mesa_get_teximage(ctx, target, level, format, type, dest,
+			    texObj, texImage);
+      }
+
+      if (stImage->pt && i < depth) {
+	 st_texture_image_unmap(ctx->st, stImage);
+	 texImage->Data = st_texture_image_map(ctx->st, stImage, i,
+                                               PIPE_BUFFER_USAGE_CPU_READ);
+	 dest += dstImageStride;
+      }
+   }
+
+   texImage->Depth = depth;
+
+   /* Unmap */
+   if (stImage->pt) {
+      st_texture_image_unmap(ctx->st, stImage);
+      texImage->Data = NULL;
+   }
+}
+
+
+static void
+st_GetTexImage(GLcontext * ctx, GLenum target, GLint level,
+               GLenum format, GLenum type, GLvoid * pixels,
+               struct gl_texture_object *texObj,
+               struct gl_texture_image *texImage)
+{
+   st_get_tex_image(ctx, target, level, format, type, pixels,
+                    texObj, texImage, 0);
+}
+
+
+static void
+st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
+                         GLvoid *pixels,
+                         struct gl_texture_object *texObj,
+                         struct gl_texture_image *texImage)
+{
+   st_get_tex_image(ctx, target, level, 0, 0, pixels,
+                    (struct gl_texture_object *) texObj,
+                    (struct gl_texture_image *) texImage, 1);
+}
+
+
+
+static void
+st_TexSubimage(GLcontext * ctx,
+               GLint dims,
+               GLenum target, GLint level,
+               GLint xoffset, GLint yoffset, GLint zoffset,
+               GLint width, GLint height, GLint depth,
+               GLenum format, GLenum type, const void *pixels,
+               const struct gl_pixelstore_attrib *packing,
+               struct gl_texture_object *texObj,
+               struct gl_texture_image *texImage)
+{
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   GLuint dstRowStride;
+   GLuint srcImageStride = _mesa_image_image_stride(packing, width, height,
+						    format, type);
+   int i;
+   const GLubyte *src;
+
+   DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__,
+       _mesa_lookup_enum_by_nr(target),
+       level, xoffset, yoffset, width, height);
+
+   pixels =
+      _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format,
+                                  type, pixels, packing, "glTexSubImage2D");
+   if (!pixels)
+      return;
+
+   /* Map buffer if necessary.  Need to lock to prevent other contexts
+    * from uploading the buffer under us.
+    */
+   if (stImage->pt) {
+      texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, 
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (stImage->surface)
+         dstRowStride = stImage->surface->stride;
+   }
+
+   if (!texImage->Data) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+      return;
+   }
+
+   src = (const GLubyte *) pixels;
+
+   for (i = 0; i++ < depth;) {
+      if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
+					   texImage->TexFormat,
+					   texImage->Data,
+					   xoffset, yoffset, 0,
+					   dstRowStride,
+					   texImage->ImageOffsets,
+					   width, height, 1,
+					   format, type, src, packing)) {
+	 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+      }
+
+      if (stImage->pt && i < depth) {
+         /* map next slice of 3D texture */
+	 st_texture_image_unmap(ctx->st, stImage);
+	 texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i,
+                                               PIPE_BUFFER_USAGE_CPU_WRITE);
+	 src += srcImageStride;
+      }
+   }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
+
+   _mesa_unmap_teximage_pbo(ctx, packing);
+
+   if (stImage->pt) {
+      st_texture_image_unmap(ctx->st, stImage);
+      texImage->Data = NULL;
+   }
+}
+
+
+
+static void
+st_TexSubImage3D(GLcontext * ctx,
+                   GLenum target,
+                   GLint level,
+                   GLint xoffset, GLint yoffset, GLint zoffset,
+                   GLsizei width, GLsizei height, GLsizei depth,
+                   GLenum format, GLenum type,
+                   const GLvoid * pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage)
+{
+   st_TexSubimage(ctx, 3, target, level,
+                  xoffset, yoffset, zoffset,
+                  width, height, depth,
+                  format, type, pixels, packing, texObj, texImage);
+}
+
+
+static void
+st_TexSubImage2D(GLcontext * ctx,
+                   GLenum target,
+                   GLint level,
+                   GLint xoffset, GLint yoffset,
+                   GLsizei width, GLsizei height,
+                   GLenum format, GLenum type,
+                   const GLvoid * pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage)
+{
+   st_TexSubimage(ctx, 2, target, level,
+                  xoffset, yoffset, 0,
+                  width, height, 1,
+                  format, type, pixels, packing, texObj, texImage);
+}
+
+
+static void
+st_TexSubImage1D(GLcontext * ctx,
+                   GLenum target,
+                   GLint level,
+                   GLint xoffset,
+                   GLsizei width,
+                   GLenum format, GLenum type,
+                   const GLvoid * pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage)
+{
+   st_TexSubimage(ctx, 1, target, level,
+                  xoffset, 0, 0,
+                  width, 1, 1,
+                  format, type, pixels, packing, texObj, texImage);
+}
+
+
+
+/**
+ * Return 0 for GL_TEXTURE_CUBE_MAP_POSITIVE_X,
+ *        1 for GL_TEXTURE_CUBE_MAP_NEGATIVE_X,
+ *        etc.
+ * XXX duplicated from main/teximage.c
+ */
+static uint
+texture_face(GLenum target)
+{
+   if (target >= GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB &&
+       target <= GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB)
+      return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
+   else
+      return 0;
+}
+
+
+
+/**
+ * Do a CopyTexSubImage operation by mapping the source surface and
+ * dest surface and using get_tile()/put_tile() to access the pixels/texels.
+ *
+ * Note: srcY=0=TOP of renderbuffer
+ */
+static void
+fallback_copy_texsubimage(GLcontext *ctx,
+                          GLenum target,
+                          GLint level,
+                          struct st_renderbuffer *strb,
+                          struct st_texture_image *stImage,
+                          GLenum baseFormat,
+                          GLint destX, GLint destY, GLint destZ,
+                          GLint srcX, GLint srcY,
+                          GLsizei width, GLsizei height)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const uint face = texture_face(target);
+   struct pipe_texture *pt = stImage->pt;
+   struct pipe_surface *src_surf, *dest_surf;
+
+   /* We'd use strb->surface, here but it's created for GPU read/write only */
+   src_surf = pipe->screen->get_tex_surface( pipe->screen,
+                                             strb->texture,
+                                             0, 0, 0,
+                                             PIPE_BUFFER_USAGE_CPU_READ);
+
+   dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   assert(width <= MAX_WIDTH);
+
+   if (baseFormat == GL_DEPTH_COMPONENT) {
+      const GLboolean scaleOrBias = (ctx->Pixel.DepthScale != 1.0F ||
+                                     ctx->Pixel.DepthBias != 0.0F);
+      GLint row, yStep;
+
+      /* determine bottom-to-top vs. top-to-bottom order for src buffer */
+      if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+         srcY = strb->Base.Height - 1 - srcY;
+         yStep = -1;
+      }
+      else {
+         yStep = 1;
+      }
+
+      /* To avoid a large temp memory allocation, do copy row by row */
+      for (row = 0; row < height; row++, srcY += yStep, destY++) {
+         uint data[MAX_WIDTH];
+         pipe_get_tile_z(src_surf, srcX, srcY, width, 1, data);
+         if (scaleOrBias) {
+            _mesa_scale_and_bias_depth_uint(ctx, width, data);
+         }
+         pipe_put_tile_z(dest_surf, destX, destY, width, 1, data);
+      }
+   }
+   else {
+      /* RGBA format */
+      GLfloat *tempSrc =
+         (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
+      GLvoid *texDest =
+         st_texture_image_map(ctx->st, stImage, 0,PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      if (tempSrc && texDest) {
+         const GLint dims = 2;
+         struct gl_texture_image *texImage = &stImage->base;
+         GLint dstRowStride = stImage->surface->stride;
+         struct gl_pixelstore_attrib unpack = ctx->DefaultPacking;
+
+         if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+            /* need to invert src */
+            srcY = strb->Base.Height - srcY - height;
+            unpack.Invert = GL_TRUE;
+         }
+
+         /* get float/RGBA image from framebuffer */
+         /* XXX this usually involves a lot of int/float conversion.
+          * try to avoid that someday.
+          */
+         pipe_get_tile_rgba(src_surf, srcX, srcY, width, height, tempSrc);
+
+         /* Store into texture memory.
+          * Note that this does some special things such as pixel transfer
+          * ops and format conversion.  In particular, if the dest tex format
+          * is actually RGBA but the user created the texture as GL_RGB we
+          * need to fill-in/override the alpha channel with 1.0.
+          */
+         texImage->TexFormat->StoreImage(ctx, dims,
+                                         texImage->_BaseFormat, 
+                                         texImage->TexFormat, 
+                                         texDest,
+                                         destX, destY, destZ,
+                                         dstRowStride,
+                                         texImage->ImageOffsets,
+                                         width, height, 1,
+                                         GL_RGBA, GL_FLOAT, tempSrc, /* src */
+                                         &unpack);
+      }
+      else {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+      }
+
+      if (tempSrc)
+         _mesa_free(tempSrc);
+      if (texDest)
+         st_texture_image_unmap(ctx->st, stImage);
+   }
+
+   screen->tex_surface_release(screen, &dest_surf);
+   screen->tex_surface_release(screen, &src_surf);
+}
+
+
+/**
+ * Do a CopyTex[Sub]Image1/2/3D() using a hardware (blit) path if possible.
+ * Note that the region to copy has already been clipped so we know we
+ * won't read from outside the source renderbuffer's bounds.
+ *
+ * Note: srcY=0=Bottom of renderbuffer (GL convention)
+ */
+static void
+st_copy_texsubimage(GLcontext *ctx,
+                    GLenum target, GLint level,
+                    GLint destX, GLint destY, GLint destZ,
+                    GLint srcX, GLint srcY,
+                    GLsizei width, GLsizei height)
+{
+   struct gl_texture_unit *texUnit =
+      &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   struct gl_texture_object *texObj =
+      _mesa_select_tex_object(ctx, texUnit, target);
+   struct gl_texture_image *texImage =
+      _mesa_select_tex_image(ctx, texObj, target, level);
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   const GLenum texBaseFormat = texImage->InternalFormat;
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct st_renderbuffer *strb;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   enum pipe_format dest_format, src_format;
+   GLboolean use_fallback = GL_TRUE;
+   GLboolean matching_base_formats;
+
+   /* any rendering in progress must complete before we grab the fb image */
+   st_finish(ctx->st);
+
+   /* determine if copying depth or color data */
+   if (texBaseFormat == GL_DEPTH_COMPONENT) {
+      strb = st_renderbuffer(fb->_DepthBuffer);
+   }
+   else if (texBaseFormat == GL_DEPTH_STENCIL_EXT) {
+      strb = st_renderbuffer(fb->_StencilBuffer);
+   }
+   else {
+      /* texBaseFormat == GL_RGB, GL_RGBA, GL_ALPHA, etc */
+      strb = st_renderbuffer(fb->_ColorReadBuffer);
+   }
+
+   assert(strb);
+   assert(strb->surface);
+   assert(stImage->pt);
+
+   src_format = strb->surface->format;
+   dest_format = stImage->pt->format;
+
+   /*
+    * Determine if the src framebuffer and dest texture have the same
+    * base format.  We need this to detect a case such as the framebuffer
+    * being GL_RGBA but the texture being GL_RGB.  If the actual hardware
+    * texture format stores RGBA we need to set A=1 (overriding the
+    * framebuffer's alpha values).  We can't do that with the blit or
+    * textured-quad paths.
+    */
+   matching_base_formats = (strb->Base._BaseFormat == texImage->_BaseFormat);
+
+   if (matching_base_formats && ctx->_ImageTransferState == 0x0) {
+      /* try potential hardware path */
+      struct pipe_surface *dest_surface = NULL;
+
+      if (src_format == dest_format) {
+         /* use surface_copy() / blit */
+         boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP);
+
+         dest_surface = screen->get_tex_surface(screen, stImage->pt,
+                                                stImage->face, stImage->level,
+                                                destZ,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+         if (do_flip)
+            srcY = strb->surface->height - srcY - height;
+
+         /* for surface_copy(), y=0=top, always */
+         pipe->surface_copy(pipe,
+                            do_flip,
+                            /* dest */
+                            dest_surface,
+                            destX, destY,
+                            /* src */
+                            strb->surface,
+                            srcX, srcY,
+                            /* size */
+                            width, height);
+         use_fallback = GL_FALSE;
+      }
+      else if (screen->is_format_supported(screen, src_format,
+                                           PIPE_TEXTURE_2D, 
+                                           PIPE_TEXTURE_USAGE_SAMPLER,
+                                           0) &&
+               screen->is_format_supported(screen, dest_format,
+                                           PIPE_TEXTURE_2D, 
+                                           PIPE_TEXTURE_USAGE_RENDER_TARGET,
+                                           0)) {
+         /* draw textured quad to do the copy */
+         boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP);
+         int srcY0, srcY1;
+
+         dest_surface = screen->get_tex_surface(screen, stImage->pt,
+                                                stImage->face, stImage->level,
+                                                destZ,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+
+         if (do_flip) {
+            srcY1 = strb->Base.Height - srcY - height;
+            srcY0 = srcY1 + height;
+         }
+         else {
+            srcY0 = srcY;
+            srcY1 = srcY0 + height;
+         }
+         util_blit_pixels(ctx->st->blit,
+                          strb->surface,
+                          srcX, srcY0,
+                          srcX + width, srcY1,
+                          dest_surface,
+                          destX, destY,
+                          destX + width, destY + height,
+                          0.0, PIPE_TEX_MIPFILTER_NEAREST);
+         use_fallback = GL_FALSE;
+      }
+
+      if (dest_surface)
+         pipe_surface_reference(&dest_surface, NULL);
+   }
+
+   if (use_fallback) {
+      /* software fallback */
+      fallback_copy_texsubimage(ctx, target, level,
+                                strb, stImage, texBaseFormat,
+                                destX, destY, destZ,
+                                srcX, srcY, width, height);
+   }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
+}
+
+
+
+static void
+st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
+                  GLenum internalFormat,
+                  GLint x, GLint y, GLsizei width, GLint border)
+{
+   struct gl_texture_unit *texUnit =
+      &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   struct gl_texture_object *texObj =
+      _mesa_select_tex_object(ctx, texUnit, target);
+   struct gl_texture_image *texImage =
+      _mesa_select_tex_image(ctx, texObj, target, level);
+
+#if 0
+   if (border)
+      goto fail;
+#endif
+
+   /* Setup or redefine the texture object, texture and texture
+    * image.  Don't populate yet.  
+    */
+   ctx->Driver.TexImage1D(ctx, target, level, internalFormat,
+                          width, border,
+                          GL_RGBA, CHAN_TYPE, NULL,
+                          &ctx->DefaultPacking, texObj, texImage);
+
+   st_copy_texsubimage(ctx, target, level,
+                       0, 0, 0,  /* destX,Y,Z */
+                       x, y, width, 1);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
+                  GLenum internalFormat,
+                  GLint x, GLint y, GLsizei width, GLsizei height,
+                  GLint border)
+{
+   struct gl_texture_unit *texUnit =
+      &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   struct gl_texture_object *texObj =
+      _mesa_select_tex_object(ctx, texUnit, target);
+   struct gl_texture_image *texImage =
+      _mesa_select_tex_image(ctx, texObj, target, level);
+
+   /* Setup or redefine the texture object, texture and texture
+    * image.  Don't populate yet.  
+    */
+   ctx->Driver.TexImage2D(ctx, target, level, internalFormat,
+                          width, height, border,
+                          GL_RGBA, CHAN_TYPE, NULL,
+                          &ctx->DefaultPacking, texObj, texImage);
+
+   st_copy_texsubimage(ctx, target, level,
+                       0, 0, 0,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
+                     GLint xoffset, GLint x, GLint y, GLsizei width)
+{
+   const GLint yoffset = 0, zoffset = 0;
+   const GLsizei height = 1;
+   st_copy_texsubimage(ctx, target, level,
+                       xoffset, yoffset, zoffset,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
+                     GLint xoffset, GLint yoffset,
+                     GLint x, GLint y, GLsizei width, GLsizei height)
+{
+   const GLint zoffset = 0;
+   st_copy_texsubimage(ctx, target, level,
+                       xoffset, yoffset, zoffset,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
+                     GLint xoffset, GLint yoffset, GLint zoffset,
+                     GLint x, GLint y, GLsizei width, GLsizei height)
+{
+   st_copy_texsubimage(ctx, target, level,
+                       xoffset, yoffset, zoffset,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+/**
+ * Compute which mipmap levels that really need to be sent to the hardware.
+ * This depends on the base image size, GL_TEXTURE_MIN_LOD,
+ * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL.
+ */
+static void
+calculate_first_last_level(struct st_texture_object *stObj)
+{
+   struct gl_texture_object *tObj = &stObj->base;
+
+   /* These must be signed values.  MinLod and MaxLod can be negative numbers,
+    * and having firstLevel and lastLevel as signed prevents the need for
+    * extra sign checks.
+    */
+   int firstLevel;
+   int lastLevel;
+
+   /* Yes, this looks overly complicated, but it's all needed.
+    */
+   switch (tObj->Target) {
+   case GL_TEXTURE_1D:
+   case GL_TEXTURE_2D:
+   case GL_TEXTURE_3D:
+   case GL_TEXTURE_CUBE_MAP:
+      if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) {
+         /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL.
+          */
+         firstLevel = lastLevel = tObj->BaseLevel;
+      }
+      else {
+         firstLevel = 0;
+         lastLevel = MIN2(tObj->MaxLevel,
+                          (int) tObj->Image[0][tObj->BaseLevel]->WidthLog2);
+      }
+      break;
+   case GL_TEXTURE_RECTANGLE_NV:
+   case GL_TEXTURE_4D_SGIS:
+      firstLevel = lastLevel = 0;
+      break;
+   default:
+      return;
+   }
+
+   stObj->lastLevel = lastLevel;
+}
+
+
+static void
+copy_image_data_to_texture(struct st_context *st,
+			   struct st_texture_object *stObj,
+                           GLuint dstLevel,
+			   struct st_texture_image *stImage)
+{
+   if (stImage->pt) {
+      /* Copy potentially with the blitter:
+       */
+      st_texture_image_copy(st->pipe,
+                            stObj->pt, dstLevel,  /* dest texture, level */
+                            stImage->pt, /* src texture */
+                            stImage->face
+                            );
+
+      pipe_texture_reference(&stImage->pt, NULL);
+   }
+   else if (stImage->base.Data) {
+      assert(stImage->base.Data != NULL);
+
+      /* More straightforward upload.  
+       */
+      st_texture_image_data(st->pipe,
+                               stObj->pt,
+                               stImage->face,
+                               dstLevel,
+                               stImage->base.Data,
+                               stImage->base.RowStride * 
+                               stObj->pt->block.size,
+                               stImage->base.RowStride *
+                               stImage->base.Height *
+                               stObj->pt->block.size);
+      _mesa_align_free(stImage->base.Data);
+      stImage->base.Data = NULL;
+   }
+
+   pipe_texture_reference(&stImage->pt, stObj->pt);
+}
+
+
+/**
+ * Called during state validation.  When this function is finished,
+ * the texture object should be ready for rendering.
+ * \return GL_TRUE for success, GL_FALSE for failure (out of mem)
+ */
+GLboolean
+st_finalize_texture(GLcontext *ctx,
+		    struct pipe_context *pipe,
+		    struct gl_texture_object *tObj,
+		    GLboolean *needFlush)
+{
+   struct st_texture_object *stObj = st_texture_object(tObj);
+   const GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+   int comp_byte = 0;
+   int cpp;
+   GLuint face;
+   struct st_texture_image *firstImage;
+
+   *needFlush = GL_FALSE;
+
+   /* We know/require this is true by now: 
+    */
+   assert(stObj->base._Complete);
+
+   /* What levels must the texture include at a minimum?
+    */
+   calculate_first_last_level(stObj);
+   firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]);
+
+   /* If both firstImage and stObj point to a texture which can contain
+    * all active images, favour firstImage.  Note that because of the
+    * completeness requirement, we know that the image dimensions
+    * will match.
+    */
+   if (firstImage->pt &&
+       firstImage->pt != stObj->pt &&
+       firstImage->pt->last_level >= stObj->lastLevel) {
+
+      pipe_texture_reference(&stObj->pt, firstImage->pt);
+   }
+
+   /* FIXME: determine format block instead of cpp */
+   if (firstImage->base.IsCompressed) {
+      comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat);
+      cpp = comp_byte;
+   }
+   else {
+      cpp = firstImage->base.TexFormat->TexelBytes;
+   }
+
+   /* If we already have a gallium texture, check that it matches the texture
+    * object's format, target, size, num_levels, etc.
+    */
+   if (stObj->pt) {
+      const enum pipe_format fmt =
+         st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat);
+      if (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) ||
+          stObj->pt->format != fmt ||
+          stObj->pt->last_level < stObj->lastLevel ||
+          stObj->pt->width[0] != firstImage->base.Width2 ||
+          stObj->pt->height[0] != firstImage->base.Height2 ||
+          stObj->pt->depth[0] != firstImage->base.Depth2 ||
+          stObj->pt->block.size/stObj->pt->block.width != cpp || /* Nominal bytes per pixel */
+          stObj->pt->compressed != firstImage->base.IsCompressed) {
+         pipe_texture_release(&stObj->pt);
+         ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER;
+      }
+   }
+
+   /* May need to create a new gallium texture:
+    */
+   if (!stObj->pt) {
+      const enum pipe_format fmt =
+         st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat);
+      stObj->pt = st_texture_create(ctx->st,
+                                    gl_target_to_pipe(stObj->base.Target),
+                                    fmt,
+                                    stObj->lastLevel,
+                                    firstImage->base.Width2,
+                                    firstImage->base.Height2,
+                                    firstImage->base.Depth2,
+                                    comp_byte,
+                                    ( (pf_is_depth_stencil(fmt) ?
+                                      PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET) |
+                                      PIPE_TEXTURE_USAGE_SAMPLER ));
+
+      if (!stObj->pt) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+         return GL_FALSE;
+      }
+   }
+
+   /* Pull in any images not in the object's texture:
+    */
+   for (face = 0; face < nr_faces; face++) {
+      GLuint level;
+      for (level = 0; level <= stObj->lastLevel; level++) {
+         struct st_texture_image *stImage =
+            st_texture_image(stObj->base.Image[face][stObj->base.BaseLevel + level]);
+
+         /* Need to import images in main memory or held in other textures.
+          */
+         if (stImage && stObj->pt != stImage->pt) {
+            copy_image_data_to_texture(ctx->st, stObj, level, stImage);
+	    *needFlush = GL_TRUE;
+         }
+      }
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Returns pointer to a default/dummy texture.
+ * This is typically used when the current shader has tex/sample instructions
+ * but the user has not provided a (any) texture(s).
+ */
+struct gl_texture_object *
+st_get_default_texture(struct st_context *st)
+{
+   if (!st->default_texture) {
+      static const GLenum target = GL_TEXTURE_2D;
+      GLubyte pixels[16][16][4];
+      struct gl_texture_object *texObj;
+      struct gl_texture_image *texImg;
+
+      /* init image to gray */
+      memset(pixels, 127, sizeof(pixels));
+
+      texObj = st->ctx->Driver.NewTextureObject(st->ctx, 0, target);
+
+      texImg = _mesa_get_tex_image(st->ctx, texObj, target, 0);
+
+      _mesa_init_teximage_fields(st->ctx, target, texImg,
+                                 16, 16, 1, 0,  /* w, h, d, border */
+                                 GL_RGBA);
+
+      st_TexImage(st->ctx, 2, target,
+                  0, GL_RGBA,    /* level, intformat */
+                  16, 16, 1, 0,  /* w, h, d, border */
+                  GL_RGBA, GL_UNSIGNED_BYTE, pixels,
+                  &st->ctx->DefaultPacking,
+                  texObj, texImg,
+                  0, 0);
+
+      texObj->MinFilter = GL_NEAREST;
+      texObj->MagFilter = GL_NEAREST;
+      texObj->_Complete = GL_TRUE;
+
+      st->default_texture = texObj;
+   }
+   return st->default_texture;
+}
+
+
+void
+st_init_texture_functions(struct dd_function_table *functions)
+{
+   functions->ChooseTextureFormat = st_ChooseTextureFormat;
+   functions->TexImage1D = st_TexImage1D;
+   functions->TexImage2D = st_TexImage2D;
+   functions->TexImage3D = st_TexImage3D;
+   functions->TexSubImage1D = st_TexSubImage1D;
+   functions->TexSubImage2D = st_TexSubImage2D;
+   functions->TexSubImage3D = st_TexSubImage3D;
+   functions->CopyTexImage1D = st_CopyTexImage1D;
+   functions->CopyTexImage2D = st_CopyTexImage2D;
+   functions->CopyTexSubImage1D = st_CopyTexSubImage1D;
+   functions->CopyTexSubImage2D = st_CopyTexSubImage2D;
+   functions->CopyTexSubImage3D = st_CopyTexSubImage3D;
+   functions->GenerateMipmap = st_generate_mipmap;
+
+   functions->GetTexImage = st_GetTexImage;
+
+   /* compressed texture functions */
+   functions->CompressedTexImage2D = st_CompressedTexImage2D;
+   functions->GetCompressedTexImage = st_GetCompressedTexImage;
+   functions->CompressedTextureSize = _mesa_compressed_texture_size;
+
+   functions->NewTextureObject = st_NewTextureObject;
+   functions->NewTextureImage = st_NewTextureImage;
+   functions->DeleteTexture = st_DeleteTextureObject;
+   functions->FreeTexImageData = st_FreeTextureImageData;
+   functions->UpdateTexturePalette = 0;
+
+   functions->TextureMemCpy = do_memcpy;
+
+   /* XXX Temporary until we can query pipe's texture sizes */
+   functions->TestProxyTexImage = _mesa_test_proxy_teximage;
+}
diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h
new file mode 100644
index 00000000000..f1fe0339cdd
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_texture.h
@@ -0,0 +1,48 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_TEXTURE_H
+#define ST_CB_TEXTURE_H
+
+
+extern GLboolean
+st_finalize_texture(GLcontext *ctx,
+		    struct pipe_context *pipe, 
+		    struct gl_texture_object *tObj,
+		    GLboolean *needFlush);
+
+
+extern struct gl_texture_object *
+st_get_default_texture(struct st_context *st);
+
+
+extern void
+st_init_texture_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_TEXTURE_H */
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
new file mode 100644
index 00000000000..e584a6ceeac
--- /dev/null
+++ b/src/mesa/state_tracker/st_context.c
@@ -0,0 +1,325 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "main/imports.h"
+#include "main/context.h"
+#include "main/extensions.h"
+#include "main/matrix.h"
+#include "main/buffers.h"
+#include "main/scissor.h"
+#include "vbo/vbo.h"
+#include "shader/shader_api.h"
+#include "glapi/glapi.h"
+#include "st_public.h"
+#include "st_context.h"
+#include "st_cb_accum.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_blit.h"
+#include "st_cb_bufferobjects.h"
+#include "st_cb_clear.h"
+#if FEATURE_drawpix
+#include "st_cb_drawpixels.h"
+#include "st_cb_rasterpos.h"
+#endif
+#ifdef FEATURE_OES_draw_texture
+#include "st_cb_drawtex.h"
+#endif
+#include "st_cb_fbo.h"
+#include "st_cb_get.h"
+#if FEATURE_feedback
+#include "st_cb_feedback.h"
+#endif
+#include "st_cb_program.h"
+#include "st_cb_queryobj.h"
+#include "st_cb_readpixels.h"
+#include "st_cb_texture.h"
+#include "st_cb_flush.h"
+#include "st_cb_strings.h"
+#include "st_atom.h"
+#include "st_draw.h"
+#include "st_extensions.h"
+#include "st_gen_mipmap.h"
+#include "st_program.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "draw/draw_context.h"
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Called via ctx->Driver.UpdateState()
+ */
+void st_invalidate_state(GLcontext * ctx, GLuint new_state)
+{
+   struct st_context *st = st_context(ctx);
+
+   st->dirty.mesa |= new_state;
+   st->dirty.st |= ST_NEW_MESA;
+
+   /* This is the only core Mesa module we depend upon.
+    * No longer use swrast, swsetup, tnl.
+    */
+   _vbo_InvalidateState(ctx, new_state);
+}
+
+
+/**
+ * Check for multisample env var override.
+ */
+int
+st_get_msaa(void)
+{
+   const char *msaa = _mesa_getenv("__GL_FSAA_MODE");
+   if (msaa)
+      return atoi(msaa);
+   return 0;
+}
+
+
+static struct st_context *
+st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe )
+{
+   uint i;
+   struct st_context *st = CALLOC_STRUCT( st_context );
+   
+   ctx->st = st;
+
+   st->ctx = ctx;
+   st->pipe = pipe;
+
+   /* state tracker needs the VBO module */
+   _vbo_CreateContext(ctx);
+
+#if FEATURE_feedback || FEATURE_drawpix
+   st->draw = draw_create(); /* for selection/feedback */
+
+   /* Disable draw options that might convert points/lines to tris, etc.
+    * as that would foul-up feedback/selection mode.
+    */
+   draw_wide_line_threshold(st->draw, 1000.0f);
+   draw_wide_point_threshold(st->draw, 1000.0f);
+   draw_enable_line_stipple(st->draw, FALSE);
+   draw_enable_point_sprites(st->draw, FALSE);
+#endif
+
+   st->dirty.mesa = ~0;
+   st->dirty.st = ~0;
+
+   st->cso_context = cso_create_context(pipe);
+
+   st_init_atoms( st );
+   st_init_bitmap(st);
+   st_init_clear(st);
+   st_init_draw( st );
+   st_init_generate_mipmap(st);
+   st_init_blit(st);
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+      st->state.sampler_list[i] = &st->state.samplers[i];
+
+   /* we want all vertex data to be placed in buffer objects */
+   vbo_use_buffer_objects(ctx);
+
+   /* Need these flags:
+    */
+   st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
+
+   st->ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
+
+   st->pixel_xfer.cache = _mesa_new_program_cache();
+
+   st->force_msaa = st_get_msaa();
+
+   /* GL limits and extensions */
+   st_init_limits(st);
+   st_init_extensions(st);
+
+   return st;
+}
+
+
+struct st_context *st_create_context(struct pipe_context *pipe,
+                                     const __GLcontextModes *visual,
+                                     struct st_context *share)
+{
+   GLcontext *ctx;
+   GLcontext *shareCtx = share ? share->ctx : NULL;
+   struct dd_function_table funcs;
+
+   memset(&funcs, 0, sizeof(funcs));
+   st_init_driver_functions(&funcs);
+
+   ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL);
+
+   return st_create_context_priv(ctx, pipe);
+}
+
+
+static void st_destroy_context_priv( struct st_context *st )
+{
+   uint i;
+
+#if FEATURE_feedback || FEATURE_drawpix
+   draw_destroy(st->draw);
+#endif
+   st_destroy_atoms( st );
+   st_destroy_draw( st );
+   st_destroy_generate_mipmap(st);
+#if FEATURE_EXT_framebuffer_blit
+   st_destroy_blit(st);
+#endif
+   st_destroy_clear(st);
+#if FEATURE_drawpix
+   st_destroy_bitmap(st);
+   st_destroy_drawpix(st);
+#endif
+#ifdef FEATURE_OES_draw_texture
+   st_destroy_drawtex(st);
+#endif
+
+   for (i = 0; i < Elements(st->state.sampler_texture); i++) {
+      pipe_texture_reference(&st->state.sampler_texture[i], NULL);
+   }
+
+   for (i = 0; i < Elements(st->state.constants); i++) {
+      if (st->state.constants[i].buffer) {
+         pipe_buffer_reference(st->pipe->screen, &st->state.constants[i].buffer, NULL);
+      }
+   }
+
+   if (st->default_texture) {
+      st->ctx->Driver.DeleteTexture(st->ctx, st->default_texture);
+      st->default_texture = NULL;
+   }
+
+   free( st );
+}
+
+ 
+void st_destroy_context( struct st_context *st )
+{
+   struct pipe_context *pipe = st->pipe;
+   struct cso_context *cso = st->cso_context;
+   GLcontext *ctx = st->ctx;
+
+   /* need to unbind and destroy CSO objects before anything else */
+   cso_release_all(st->cso_context);
+
+   st_reference_fragprog(st, &st->fp, NULL);
+   st_reference_vertprog(st, &st->vp, NULL);
+
+   _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache);
+
+   _vbo_DestroyContext(st->ctx);
+
+   _mesa_free_context_data(ctx);
+
+   st_destroy_context_priv(st);
+
+   cso_destroy_context(cso);
+
+   pipe->destroy( pipe );
+
+   free(ctx);
+}
+
+
+void st_make_current(struct st_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read)
+{
+   if (st) {
+      GLboolean firstTime = st->ctx->FirstTimeCurrent;
+      _mesa_make_current(st->ctx, &draw->Base, &read->Base);
+      /* Need to initialize viewport here since draw->Base->Width/Height
+       * will still be zero at this point.
+       * This could be improved, but would require rather extensive work
+       * elsewhere (allocate rb surface storage sooner)
+       */
+      if (firstTime) {
+         GLuint w = draw->InitWidth, h = draw->InitHeight;
+         _mesa_set_viewport(st->ctx, 0, 0, w, h);
+         _mesa_set_scissor(st->ctx, 0, 0, w, h);
+
+      }
+   }
+   else {
+      _mesa_make_current(NULL, NULL, NULL);
+   }
+}
+
+
+void st_copy_context_state(struct st_context *dst,
+                           struct st_context *src,
+                           uint mask)
+{
+   _mesa_copy_context(dst->ctx, src->ctx, mask);
+}
+
+
+
+st_proc st_get_proc_address(const char *procname)
+{
+   return (st_proc) _glapi_get_proc_address(procname);
+}
+
+
+
+void st_init_driver_functions(struct dd_function_table *functions)
+{
+   _mesa_init_glsl_driver_functions(functions);
+
+#if FEATURE_accum
+   st_init_accum_functions(functions);
+#endif
+#if FEATURE_EXT_framebuffer_blit
+   st_init_blit_functions(functions);
+#endif
+   st_init_bufferobject_functions(functions);
+   st_init_clear_functions(functions);
+#if FEATURE_drawpix
+   st_init_bitmap_functions(functions);
+   st_init_drawpixels_functions(functions);
+   st_init_rasterpos_functions(functions);
+#endif
+   st_init_fbo_functions(functions);
+   st_init_get_functions(functions);
+#if FEATURE_feedback
+   st_init_feedback_functions(functions);
+#endif
+   st_init_program_functions(functions);
+#if FEATURE_ARB_occlusion_query
+   st_init_query_functions(functions);
+#endif
+   st_init_readpixels_functions(functions);
+   st_init_texture_functions(functions);
+   st_init_flush_functions(functions);
+   st_init_string_functions(functions);
+
+   functions->UpdateState = st_invalidate_state;
+}
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
new file mode 100644
index 00000000000..695ac4a96f4
--- /dev/null
+++ b/src/mesa/state_tracker/st_context.h
@@ -0,0 +1,249 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CONTEXT_H
+#define ST_CONTEXT_H
+
+#include "main/mtypes.h"
+#include "shader/prog_cache.h"
+#include "pipe/p_state.h"
+
+
+struct st_context;
+struct st_texture_object;
+struct st_fragment_program;
+struct draw_context;
+struct draw_stage;
+struct cso_cache;
+struct cso_blend;
+struct gen_mipmap_state;
+struct blit_state;
+struct bitmap_cache;
+
+
+#define FRONT_STATUS_UNDEFINED    0
+#define FRONT_STATUS_DIRTY        1
+#define FRONT_STATUS_COPY_OF_BACK 2
+
+
+#define ST_NEW_MESA                    0x1 /* Mesa state has changed */
+#define ST_NEW_FRAGMENT_PROGRAM        0x2
+#define ST_NEW_VERTEX_PROGRAM          0x4
+#define ST_NEW_FRAMEBUFFER             0x8
+
+
+struct st_state_flags {
+   GLuint mesa;
+   GLuint st;
+};
+
+struct st_tracked_state {
+   const char *name;
+   struct st_state_flags dirty;
+   void (*update)( struct st_context *st );
+};
+
+
+
+struct st_context
+{
+   GLcontext *ctx;
+
+   struct pipe_context *pipe;
+
+   struct draw_context *draw;  /**< For selection/feedback/rastpos only */
+   struct draw_stage *feedback_stage;  /**< For GL_FEEDBACK rendermode */
+   struct draw_stage *selection_stage;  /**< For GL_SELECT rendermode */
+   struct draw_stage *rastpos_stage;  /**< For glRasterPos */
+
+   /* Some state is contained in constant objects.
+    * Other state is just parameter values.
+    */
+   struct {
+      struct pipe_blend_state               blend;
+      struct pipe_depth_stencil_alpha_state depth_stencil;
+      struct pipe_rasterizer_state          rasterizer;
+      struct pipe_sampler_state             samplers[PIPE_MAX_SAMPLERS];
+      struct pipe_sampler_state             *sampler_list[PIPE_MAX_SAMPLERS];
+      struct pipe_clip_state clip;
+      struct pipe_constant_buffer constants[2];
+      struct pipe_framebuffer_state framebuffer;
+      struct pipe_texture *sampler_texture[PIPE_MAX_SAMPLERS];
+      struct pipe_poly_stipple poly_stipple;
+      struct pipe_scissor_state scissor;
+      struct pipe_viewport_state viewport;
+
+      GLuint num_samplers;
+      GLuint num_textures;
+   } state;
+
+   struct {
+      struct st_tracked_state tracked_state[PIPE_SHADER_TYPES];
+   } constants;
+
+   /* XXX unused: */
+   struct {
+      struct gl_fragment_program *fragment_program;
+   } cb;
+
+   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ */
+
+   char vendor[100];
+   char renderer[100];
+
+   /* State to be validated:
+    */
+   struct st_tracked_state **atoms;
+   GLuint nr_atoms;
+
+   struct st_state_flags dirty;
+
+   GLboolean missing_textures;
+
+   GLfloat polygon_offset_scale; /* ?? */
+
+   /** Mapping from VERT_RESULT_x to post-transformed vertex slot */
+   const GLuint *vertex_result_to_slot;
+
+   struct st_vertex_program *vp;    /**< Currently bound vertex program */
+   struct st_fragment_program *fp;  /**< Currently bound fragment program */
+
+   struct gl_texture_object *default_texture;
+
+   struct {
+      struct gl_program_cache *cache;
+      struct st_fragment_program *program;  /**< cur pixel transfer prog */
+      GLuint xfer_prog_sn;  /**< pixel xfer program serial no. */
+      GLuint user_prog_sn;  /**< user fragment program serial no. */
+      struct st_fragment_program *combined_prog;
+      GLuint combined_prog_sn;
+      struct pipe_texture *pixelmap_texture;
+      boolean pixelmap_enabled;  /**< use the pixelmap texture? */
+   } pixel_xfer;
+
+   /** for glBitmap */
+   struct {
+      struct pipe_rasterizer_state rasterizer;
+      struct pipe_sampler_state sampler;
+      struct pipe_shader_state vert_shader;
+      enum pipe_format tex_format;
+      void *vs;
+      float vertices[4][3][4];  /**< vertex pos + color + texcoord */
+      struct pipe_buffer *vbuf;
+      unsigned vbuf_slot;       /* next free slot in vbuf */
+      struct bitmap_cache *cache;
+   } bitmap;
+
+   /** for glDraw/CopyPixels */
+   struct {
+      struct st_fragment_program *z_shader;
+      struct st_vertex_program *vert_shaders[2];
+   } drawpix;
+
+   /** for glClear */
+   struct {
+      struct pipe_shader_state vert_shader;
+      struct pipe_shader_state frag_shader;
+      struct pipe_rasterizer_state raster;
+      struct pipe_viewport_state viewport;
+      void *vs;
+      void *fs;
+      float vertices[4][2][4];  /**< vertex pos + color */
+      struct pipe_buffer *vbuf;
+      unsigned vbuf_slot;
+   } clear;
+
+   void *passthrough_fs;  /**< simple pass-through frag shader */
+
+   struct gen_mipmap_state *gen_mipmap;
+   struct blit_state *blit;
+
+   struct cso_context *cso_context;
+
+   int force_msaa;
+};
+
+
+/* Need this so that we can implement Mesa callbacks in this module.
+ */
+static INLINE struct st_context *st_context(GLcontext *ctx)
+{
+   return ctx->st;
+}
+
+
+/**
+ * Wrapper for GLframebuffer.
+ * This is an opaque type to the outside world.
+ */
+struct st_framebuffer
+{
+   GLframebuffer Base;
+   void *Private;
+   GLuint InitWidth, InitHeight;
+};
+
+
+extern void st_init_driver_functions(struct dd_function_table *functions);
+
+void st_invalidate_state(GLcontext * ctx, GLuint new_state);
+
+
+
+#define Y_0_TOP 1
+#define Y_0_BOTTOM 2
+
+static INLINE GLuint
+st_fb_orientation(const struct gl_framebuffer *fb)
+{
+   if (fb && fb->Name == 0) {
+      /* Drawing into a window (on-screen buffer).
+       *
+       * Negate Y scale to flip image vertically.
+       * The NDC Y coords prior to viewport transformation are in the range
+       * [y=-1=bottom, y=1=top]
+       * Hardware window coords are in the range [y=0=top, y=H-1=bottom] where
+       * H is the window height.
+       * Use the viewport transformation to invert Y.
+       */
+      return Y_0_TOP;
+   }
+   else {
+      /* Drawing into user-created FBO (very likely a texture).
+       *
+       * For textures, T=0=Bottom, so by extension Y=0=Bottom for rendering.
+       */
+      return Y_0_BOTTOM;
+   }
+}
+
+
+extern int
+st_get_msaa(void);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c
new file mode 100644
index 00000000000..c7d26ce33cf
--- /dev/null
+++ b/src/mesa/state_tracker/st_debug.c
@@ -0,0 +1,70 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "main/context.h"
+#include "shader/prog_print.h"
+
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "cso_cache/cso_cache.h"
+
+#include "st_context.h"
+#include "st_debug.h"
+#include "st_program.h"
+
+
+
+/**
+ * Print current state.  May be called from inside gdb to see currently
+ * bound vertex/fragment shaders and associated constants.
+ */
+void
+st_print_current(void)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct st_context *st = ctx->st;
+
+#if 0
+   int i;
+
+   printf("Vertex Transform Inputs:\n");
+   for (i = 0; i < st->vp->state.num_inputs; i++) {
+      printf("  Slot %d:  VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]);
+   }
+#endif
+
+   tgsi_dump( st->vp->state.tokens, 0 );
+   if (st->vp->Base.Base.Parameters)
+      _mesa_print_parameter_list(st->vp->Base.Base.Parameters);
+
+   tgsi_dump( st->fp->state.tokens, 0 );
+   if (st->fp->Base.Base.Parameters)
+      _mesa_print_parameter_list(st->fp->Base.Base.Parameters);
+}
diff --git a/src/mesa/state_tracker/st_debug.h b/src/mesa/state_tracker/st_debug.h
new file mode 100644
index 00000000000..49d752e1b23
--- /dev/null
+++ b/src/mesa/state_tracker/st_debug.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_DEBUG_H
+#define ST_DEBUG_H
+
+extern void
+st_print_current(void);
+
+
+#endif /* ST_DEBUG_H */
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
new file mode 100644
index 00000000000..ae71d586c2f
--- /dev/null
+++ b/src/mesa/state_tracker/st_draw.c
@@ -0,0 +1,699 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "shader/prog_uniform.h"
+
+#include "vbo/vbo.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_bufferobjects.h"
+#include "st_draw.h"
+#include "st_program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+
+static GLuint double_types[4] = {
+   PIPE_FORMAT_R64_FLOAT,
+   PIPE_FORMAT_R64G64_FLOAT,
+   PIPE_FORMAT_R64G64B64_FLOAT,
+   PIPE_FORMAT_R64G64B64A64_FLOAT
+};
+
+static GLuint float_types[4] = {
+   PIPE_FORMAT_R32_FLOAT,
+   PIPE_FORMAT_R32G32_FLOAT,
+   PIPE_FORMAT_R32G32B32_FLOAT,
+   PIPE_FORMAT_R32G32B32A32_FLOAT
+};
+
+static GLuint uint_types_norm[4] = {
+   PIPE_FORMAT_R32_UNORM,
+   PIPE_FORMAT_R32G32_UNORM,
+   PIPE_FORMAT_R32G32B32_UNORM,
+   PIPE_FORMAT_R32G32B32A32_UNORM
+};
+
+static GLuint uint_types_scale[4] = {
+   PIPE_FORMAT_R32_USCALED,
+   PIPE_FORMAT_R32G32_USCALED,
+   PIPE_FORMAT_R32G32B32_USCALED,
+   PIPE_FORMAT_R32G32B32A32_USCALED
+};
+
+static GLuint int_types_norm[4] = {
+   PIPE_FORMAT_R32_SNORM,
+   PIPE_FORMAT_R32G32_SNORM,
+   PIPE_FORMAT_R32G32B32_SNORM,
+   PIPE_FORMAT_R32G32B32A32_SNORM
+};
+
+static GLuint int_types_scale[4] = {
+   PIPE_FORMAT_R32_SSCALED,
+   PIPE_FORMAT_R32G32_SSCALED,
+   PIPE_FORMAT_R32G32B32_SSCALED,
+   PIPE_FORMAT_R32G32B32A32_SSCALED
+};
+
+static GLuint ushort_types_norm[4] = {
+   PIPE_FORMAT_R16_UNORM,
+   PIPE_FORMAT_R16G16_UNORM,
+   PIPE_FORMAT_R16G16B16_UNORM,
+   PIPE_FORMAT_R16G16B16A16_UNORM
+};
+
+static GLuint ushort_types_scale[4] = {
+   PIPE_FORMAT_R16_USCALED,
+   PIPE_FORMAT_R16G16_USCALED,
+   PIPE_FORMAT_R16G16B16_USCALED,
+   PIPE_FORMAT_R16G16B16A16_USCALED
+};
+
+static GLuint short_types_norm[4] = {
+   PIPE_FORMAT_R16_SNORM,
+   PIPE_FORMAT_R16G16_SNORM,
+   PIPE_FORMAT_R16G16B16_SNORM,
+   PIPE_FORMAT_R16G16B16A16_SNORM
+};
+
+static GLuint short_types_scale[4] = {
+   PIPE_FORMAT_R16_SSCALED,
+   PIPE_FORMAT_R16G16_SSCALED,
+   PIPE_FORMAT_R16G16B16_SSCALED,
+   PIPE_FORMAT_R16G16B16A16_SSCALED
+};
+
+static GLuint ubyte_types_norm[4] = {
+   PIPE_FORMAT_R8_UNORM,
+   PIPE_FORMAT_R8G8_UNORM,
+   PIPE_FORMAT_R8G8B8_UNORM,
+   PIPE_FORMAT_R8G8B8A8_UNORM
+};
+
+static GLuint ubyte_types_scale[4] = {
+   PIPE_FORMAT_R8_USCALED,
+   PIPE_FORMAT_R8G8_USCALED,
+   PIPE_FORMAT_R8G8B8_USCALED,
+   PIPE_FORMAT_R8G8B8A8_USCALED
+};
+
+static GLuint byte_types_norm[4] = {
+   PIPE_FORMAT_R8_SNORM,
+   PIPE_FORMAT_R8G8_SNORM,
+   PIPE_FORMAT_R8G8B8_SNORM,
+   PIPE_FORMAT_R8G8B8A8_SNORM
+};
+
+static GLuint byte_types_scale[4] = {
+   PIPE_FORMAT_R8_SSCALED,
+   PIPE_FORMAT_R8G8_SSCALED,
+   PIPE_FORMAT_R8G8B8_SSCALED,
+   PIPE_FORMAT_R8G8B8A8_SSCALED
+};
+
+static GLuint fixed_types[4] = {
+   PIPE_FORMAT_R32_FIXED,
+   PIPE_FORMAT_R32G32_FIXED,
+   PIPE_FORMAT_R32G32B32_FIXED,
+   PIPE_FORMAT_R32G32B32A32_FIXED
+};
+
+
+
+/**
+ * Return a PIPE_FORMAT_x for the given GL datatype and size.
+ */
+GLuint
+st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized)
+{
+   assert((type >= GL_BYTE && type <= GL_DOUBLE) ||
+          type == GL_FIXED);
+   assert(size >= 1);
+   assert(size <= 4);
+
+   if (normalized) {
+      switch (type) {
+      case GL_DOUBLE: return double_types[size-1];
+      case GL_FLOAT: return float_types[size-1];
+      case GL_INT: return int_types_norm[size-1];
+      case GL_SHORT: return short_types_norm[size-1];
+      case GL_BYTE: return byte_types_norm[size-1];
+      case GL_UNSIGNED_INT: return uint_types_norm[size-1];
+      case GL_UNSIGNED_SHORT: return ushort_types_norm[size-1];
+      case GL_UNSIGNED_BYTE: return ubyte_types_norm[size-1];
+      case GL_FIXED: return fixed_types[size-1];
+      default: assert(0); return 0;
+      }      
+   }
+   else {
+      switch (type) {
+      case GL_DOUBLE: return double_types[size-1];
+      case GL_FLOAT: return float_types[size-1];
+      case GL_INT: return int_types_scale[size-1];
+      case GL_SHORT: return short_types_scale[size-1];
+      case GL_BYTE: return byte_types_scale[size-1];
+      case GL_UNSIGNED_INT: return uint_types_scale[size-1];
+      case GL_UNSIGNED_SHORT: return ushort_types_scale[size-1];
+      case GL_UNSIGNED_BYTE: return ubyte_types_scale[size-1];
+      case GL_FIXED: return fixed_types[size-1];
+      default: assert(0); return 0;
+      }      
+   }
+   return 0; /* silence compiler warning */
+}
+
+
+/*
+ * If edge flags are needed, setup an bitvector of flags and call
+ * pipe->set_edgeflags().
+ * XXX memleak: need to free the returned pointer at some point
+ */
+static void *
+setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count,
+                const struct gl_client_array *array)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+
+   if ((primMode == GL_TRIANGLES ||
+        primMode == GL_QUADS ||
+        primMode == GL_POLYGON) &&
+       (ctx->Polygon.FrontMode != GL_FILL ||
+        ctx->Polygon.BackMode != GL_FILL)) {
+      /* need edge flags */
+      GLint i;
+      unsigned *vec;
+      struct st_buffer_object *stobj = st_buffer_object(array->BufferObj);
+      ubyte *map;
+
+      if (!stobj)
+         return NULL;
+
+      vec = (unsigned *) calloc(sizeof(unsigned), (count + 31) / 32);
+      if (!vec)
+         return NULL;
+
+      map = pipe_buffer_map(pipe->screen, stobj->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+      map = ADD_POINTERS(map, array->Ptr);
+
+      for (i = 0; i < count; i++) {
+         if (*((float *) map))
+            vec[i/32] |= 1 << (i % 32);
+
+         map += array->StrideB;
+      }
+
+      pipe_buffer_unmap(pipe->screen, stobj->buffer);
+
+      pipe->set_edgeflags(pipe, vec);
+
+      return vec;
+   }
+   else {
+      /* edge flags not needed */
+      pipe->set_edgeflags(pipe, NULL);
+      return NULL;
+   }
+}
+
+
+/**
+ * Examine the active arrays to determine if we have interleaved
+ * vertex arrays all living in one VBO, or all living in user space.
+ * \param userSpace  returns whether the arrays are in user space.
+ */
+static GLboolean
+is_interleaved_arrays(const struct st_vertex_program *vp,
+                      const struct gl_client_array **arrays,
+                      GLboolean *userSpace)
+{
+   GLuint attr;
+   const struct gl_buffer_object *firstBufObj = NULL;
+   GLint firstStride = -1;
+   GLuint num_client_arrays = 0;
+   const GLubyte *client_addr = NULL;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      const struct gl_buffer_object *bufObj = arrays[mesaAttr]->BufferObj;
+      const GLsizei stride = arrays[mesaAttr]->StrideB; /* in bytes */
+
+      if (firstStride < 0) {
+         firstStride = stride;
+      }
+      else if (firstStride != stride) {
+         return GL_FALSE;
+      }
+         
+      if (!bufObj || !bufObj->Name) {
+         num_client_arrays++;
+         /* Try to detect if the client-space arrays are
+          * "close" to each other.
+          */
+         if (!client_addr) {
+            client_addr = arrays[mesaAttr]->Ptr;
+         }
+         else if (abs(arrays[mesaAttr]->Ptr - client_addr) > firstStride) {
+            /* arrays start too far apart */
+            return GL_FALSE;
+         }
+      }
+      else if (!firstBufObj) {
+         firstBufObj = bufObj;
+      }
+      else if (bufObj != firstBufObj) {
+         return GL_FALSE;
+      }
+   }
+
+   *userSpace = (num_client_arrays == vp->num_inputs);
+   /* printf("user space: %d (%d %d)\n", (int) *userSpace,num_client_arrays,vp->num_inputs); */
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Compute the memory range occupied by the arrays.
+ */
+static void
+get_arrays_bounds(const struct st_vertex_program *vp,
+                       const struct gl_client_array **arrays,
+                       GLuint max_index,
+                       const GLubyte **low, const GLubyte **high)
+{
+   const GLubyte *low_addr = NULL;
+   GLuint attr;
+   GLint stride;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      const GLubyte *start = arrays[mesaAttr]->Ptr;
+      stride = arrays[mesaAttr]->StrideB;
+      if (attr == 0) {
+         low_addr = start;
+      }
+      else {
+         low_addr = MIN2(low_addr, start);
+      }
+   }
+
+   *low = low_addr;
+   *high = low_addr + (max_index + 1) * stride;
+}
+
+
+/**
+ * Set up for drawing interleaved arrays that all live in one VBO
+ * or all live in user space.
+ * \param vbuffer  returns vertex buffer info
+ * \param velements  returns vertex element info
+ */
+static void
+setup_interleaved_attribs(GLcontext *ctx,
+                          const struct st_vertex_program *vp,
+                          const struct gl_client_array **arrays,
+                          GLuint max_index,
+                          GLboolean userSpace,
+                          struct pipe_vertex_buffer *vbuffer,
+                          struct pipe_vertex_element velements[])
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   GLuint attr;
+   const GLubyte *offset0;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
+      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+      GLsizei stride = arrays[mesaAttr]->StrideB;
+
+      /*printf("stobj %u = %p\n", attr, (void*)stobj);*/
+
+      if (attr == 0) {
+         const GLubyte *low, *high;
+
+         get_arrays_bounds(vp, arrays, max_index, &low, &high);
+         /*printf("buffer range: %p %p  %d\n", low, high, high-low);*/
+
+         offset0 = low;
+         if (userSpace) {
+            vbuffer->buffer =
+               pipe_user_buffer_create(pipe->screen, (void *) low, high - low);
+            vbuffer->buffer_offset = 0;
+         }
+         else {
+            vbuffer->buffer = NULL;
+            pipe_buffer_reference(pipe->screen, &vbuffer->buffer, stobj->buffer);
+            vbuffer->buffer_offset = (unsigned) low;
+         }
+         vbuffer->stride = stride; /* in bytes */
+         vbuffer->max_index = max_index;
+      }
+
+      velements[attr].src_offset =
+         (unsigned) (arrays[mesaAttr]->Ptr - offset0);
+      velements[attr].vertex_buffer_index = 0;
+      velements[attr].nr_components = arrays[mesaAttr]->Size;
+      velements[attr].src_format =
+         st_pipe_vertex_format(arrays[mesaAttr]->Type,
+                               arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Normalized);
+      assert(velements[attr].src_format);
+   }
+}
+
+
+/**
+ * Set up a separate pipe_vertex_buffer and pipe_vertex_element for each
+ * vertex attribute.
+ * \param vbuffer  returns vertex buffer info
+ * \param velements  returns vertex element info
+ */
+static void
+setup_non_interleaved_attribs(GLcontext *ctx,
+                              const struct st_vertex_program *vp,
+                              const struct gl_client_array **arrays,
+                              GLuint max_index,
+                              GLboolean *userSpace,
+                              struct pipe_vertex_buffer vbuffer[],
+                              struct pipe_vertex_element velements[])
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   GLuint attr;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
+      GLsizei stride = arrays[mesaAttr]->StrideB;
+
+      *userSpace = GL_FALSE;
+
+      if (bufobj && bufobj->Name) {
+         /* Attribute data is in a VBO.
+          * Recall that for VBOs, the gl_client_array->Ptr field is
+          * really an offset from the start of the VBO, not a pointer.
+          */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         assert(stobj->buffer);
+         /*printf("stobj %u = %p\n", attr, (void*) stobj);*/
+
+         vbuffer[attr].buffer = NULL;
+         pipe_buffer_reference(pipe->screen, &vbuffer[attr].buffer, stobj->buffer);
+         vbuffer[attr].buffer_offset = (unsigned) arrays[mesaAttr]->Ptr;
+         velements[attr].src_offset = 0;
+      }
+      else {
+         /* attribute data is in user-space memory, not a VBO */
+         uint bytes;
+         /*printf("user-space array %d stride %d\n", attr, stride);*/
+	
+         *userSpace = GL_TRUE;
+
+         /* wrap user data */
+         if (arrays[mesaAttr]->Ptr) {
+            /* user's vertex array */
+            if (arrays[mesaAttr]->StrideB) {
+               bytes = arrays[mesaAttr]->StrideB * (max_index + 1);
+            }
+            else {
+               bytes = arrays[mesaAttr]->Size
+                  * _mesa_sizeof_type(arrays[mesaAttr]->Type);
+            }
+            vbuffer[attr].buffer = pipe_user_buffer_create(pipe->screen,
+                           (void *) arrays[mesaAttr]->Ptr, bytes);
+         }
+         else {
+            /* no array, use ctx->Current.Attrib[] value */
+            bytes = sizeof(ctx->Current.Attrib[0]);
+            vbuffer[attr].buffer = pipe_user_buffer_create(pipe->screen,
+                           (void *) ctx->Current.Attrib[mesaAttr], bytes);
+            stride = 0;
+         }
+
+         vbuffer[attr].buffer_offset = 0;
+         velements[attr].src_offset = 0;
+      }
+
+      assert(velements[attr].src_offset <= 2048); /* 11-bit field */
+
+      /* common-case setup */
+      vbuffer[attr].stride = stride; /* in bytes */
+      vbuffer[attr].max_index = max_index;
+      velements[attr].vertex_buffer_index = attr;
+      velements[attr].nr_components = arrays[mesaAttr]->Size;
+      velements[attr].src_format
+         = st_pipe_vertex_format(arrays[mesaAttr]->Type,
+                                 arrays[mesaAttr]->Size,
+                                 arrays[mesaAttr]->Normalized);
+      assert(velements[attr].src_format);
+   }
+}
+
+
+
+/**
+ * Prior to drawing, check that any uniforms referenced by the
+ * current shader have been set.  If a uniform has not been set,
+ * issue a warning.
+ */
+static void
+check_uniforms(GLcontext *ctx)
+{
+   const struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+   if (shProg && shProg->LinkStatus) {
+      GLuint i;
+      for (i = 0; i < shProg->Uniforms->NumUniforms; i++) {
+         const struct gl_uniform *u = &shProg->Uniforms->Uniforms[i];
+         if (!u->Initialized) {
+            _mesa_warning(ctx,
+                          "Using shader with uninitialized uniform: %s",
+                          u->Name);
+         }
+      }
+   }
+}
+
+
+/**
+ * This function gets plugged into the VBO module and is called when
+ * we have something to render.
+ * Basically, translate the information into the format expected by gallium.
+ */
+void
+st_draw_vbo(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)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   const struct st_vertex_program *vp;
+   const struct pipe_shader_state *vs;
+   struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS];
+   GLuint attr;
+   struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
+   unsigned num_vbuffers, num_velements;
+   GLboolean userSpace;
+
+   /* sanity check for pointer arithmetic below */
+   assert(sizeof(arrays[0]->Ptr[0]) == 1);
+
+   st_validate_state(ctx->st);
+
+   /* must get these after state validation! */
+   vp = ctx->st->vp;
+   vs = &ctx->st->vp->state;
+
+#if 0
+   if (MESA_VERBOSE & VERBOSE_GLSL) {
+      check_uniforms(ctx);
+   }
+#endif
+
+   /*
+    * Setup the vbuffer[] and velements[] arrays.
+    */
+   if (is_interleaved_arrays(vp, arrays, &userSpace)) {
+      /*printf("Draw interleaved\n");*/
+      setup_interleaved_attribs(ctx, vp, arrays, max_index, userSpace,
+                                vbuffer, velements);
+      num_vbuffers = 1;
+      num_velements = vp->num_inputs;
+      if (num_velements == 0)
+         num_vbuffers = 0;
+   }
+   else {
+      /*printf("Draw non-interleaved\n");*/
+      setup_non_interleaved_attribs(ctx, vp, arrays, max_index,
+                                    &userSpace, vbuffer, velements);
+      num_vbuffers = vp->num_inputs;
+      num_velements = vp->num_inputs;
+   }
+
+#if 0
+   {
+      GLuint i;
+      for (i = 0; i < num_vbuffers; i++) {
+         printf("buffers[%d].stride = %u\n", i, vbuffer[i].stride);
+         printf("buffers[%d].max_index = %u\n", i, vbuffer[i].max_index);
+         printf("buffers[%d].buffer_offset = %u\n", i, vbuffer[i].buffer_offset);
+         printf("buffers[%d].buffer = %p\n", i, (void*) vbuffer[i].buffer);
+      }
+      for (i = 0; i < num_velements; i++) {
+         printf("vlements[%d].vbuffer_index = %u\n", i, velements[i].vertex_buffer_index);
+         printf("vlements[%d].src_offset = %u\n", i, velements[i].src_offset);
+         printf("vlements[%d].nr_comps = %u\n", i, velements[i].nr_components);
+         printf("vlements[%d].format = %s\n", i, pf_name(velements[i].src_format));
+      }
+   }
+#endif
+
+   pipe->set_vertex_buffers(pipe, num_vbuffers, vbuffer);
+   pipe->set_vertex_elements(pipe, num_velements, velements);
+
+   if (num_vbuffers == 0 || num_velements == 0)
+      return;
+
+   /* do actual drawing */
+   if (ib) {
+      /* indexed primitive */
+      struct gl_buffer_object *bufobj = ib->obj;
+      struct pipe_buffer *indexBuf = NULL;
+      unsigned indexSize, indexOffset, i;
+
+      switch (ib->type) {
+      case GL_UNSIGNED_INT:
+         indexSize = 4;
+         break;
+      case GL_UNSIGNED_SHORT:
+         indexSize = 2;
+         break;
+      case GL_UNSIGNED_BYTE:
+         indexSize = 1;
+         break;
+      default:
+         assert(0);
+	 return;
+      }
+
+      /* get/create the index buffer object */
+      if (bufobj && bufobj->Name) {
+         /* elements/indexes are in a real VBO */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         pipe_buffer_reference(pipe->screen, &indexBuf, stobj->buffer);
+         indexOffset = (unsigned) ib->ptr / indexSize;
+      }
+      else {
+         /* element/indicies are in user space memory */
+         indexBuf = pipe_user_buffer_create(pipe->screen, (void *) ib->ptr,
+                                            ib->count * indexSize);
+         indexOffset = 0;
+      }
+
+      /* draw */
+      if (nr_prims == 1 && pipe->draw_range_elements != NULL) {
+         i = 0;
+
+         /* XXX: exercise temporary path to pass min/max directly
+          * through to driver & draw module.  These interfaces still
+          * need a bit of work...
+          */
+         setup_edgeflags(ctx, prims[i].mode,
+                         prims[i].start + indexOffset, prims[i].count,
+                         arrays[VERT_ATTRIB_EDGEFLAG]);
+
+         pipe->draw_range_elements(pipe, indexBuf, indexSize,
+                                   min_index,
+                                   max_index,
+                                   prims[i].mode,
+                                   prims[i].start + indexOffset, prims[i].count);
+      }
+      else {
+         for (i = 0; i < nr_prims; i++) {
+            setup_edgeflags(ctx, prims[i].mode,
+                            prims[i].start + indexOffset, prims[i].count,
+                            arrays[VERT_ATTRIB_EDGEFLAG]);
+            
+            pipe->draw_elements(pipe, indexBuf, indexSize,
+                                prims[i].mode,
+                                prims[i].start + indexOffset, prims[i].count);
+         }
+      }
+
+      pipe_buffer_reference(pipe->screen, &indexBuf, NULL);
+   }
+   else {
+      /* non-indexed */
+      GLuint i;
+      for (i = 0; i < nr_prims; i++) {
+         setup_edgeflags(ctx, prims[i].mode,
+                         prims[i].start, prims[i].count,
+                         arrays[VERT_ATTRIB_EDGEFLAG]);
+
+         pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count);
+      }
+   }
+
+   /* unreference buffers (frees wrapped user-space buffer objects) */
+   for (attr = 0; attr < num_vbuffers; attr++) {
+      pipe_buffer_reference(pipe->screen, &vbuffer[attr].buffer, NULL);
+      assert(!vbuffer[attr].buffer);
+   }
+
+   if (userSpace) 
+   {
+      pipe->set_vertex_buffers(pipe, 0, NULL);
+   }
+}
+
+
+void st_init_draw( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+
+   vbo_set_draw_func(ctx, st_draw_vbo);
+}
+
+
+void st_destroy_draw( struct st_context *st )
+{
+}
+
+
diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h
new file mode 100644
index 00000000000..659c3ac834f
--- /dev/null
+++ b/src/mesa/state_tracker/st_draw.h
@@ -0,0 +1,67 @@
+/**************************************************************************
+ * 
+ * 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 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+    
+
+#ifndef ST_DRAW_H
+#define ST_DRAW_H
+
+struct _mesa_prim;
+struct _mesa_index_buffer;
+
+void st_init_draw( struct st_context *st );
+
+void st_destroy_draw( struct st_context *st );
+
+extern void
+st_draw_vbo(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);
+
+extern void
+st_feedback_draw_vbo(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);
+
+/* Internal function:
+ */
+extern GLuint
+st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized);
+
+#endif
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
new file mode 100644
index 00000000000..5c9c4506c22
--- /dev/null
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -0,0 +1,262 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "shader/prog_uniform.h"
+
+#include "vbo/vbo.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_bufferobjects.h"
+#include "st_draw.h"
+#include "st_program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "draw/draw_private.h"
+#include "draw/draw_context.h"
+
+
+#if FEATURE_feedback || FEATURE_drawpix
+
+/**
+ * Set the (private) draw module's post-transformed vertex format when in
+ * GL_SELECT or GL_FEEDBACK mode or for glRasterPos.
+ */
+static void
+set_feedback_vertex_format(GLcontext *ctx)
+{
+#if 0
+   struct st_context *st = ctx->st;
+   struct vertex_info vinfo;
+   GLuint i;
+
+   memset(&vinfo, 0, sizeof(vinfo));
+
+   if (ctx->RenderMode == GL_SELECT) {
+      assert(ctx->RenderMode == GL_SELECT);
+      vinfo.num_attribs = 1;
+      vinfo.format[0] = FORMAT_4F;
+      vinfo.interp_mode[0] = INTERP_LINEAR;
+   }
+   else {
+      /* GL_FEEDBACK, or glRasterPos */
+      /* emit all attribs (pos, color, texcoord) as GLfloat[4] */
+      vinfo.num_attribs = st->state.vs->cso->state.num_outputs;
+      for (i = 0; i < vinfo.num_attribs; i++) {
+         vinfo.format[i] = FORMAT_4F;
+         vinfo.interp_mode[i] = INTERP_LINEAR;
+      }
+   }
+
+   draw_set_vertex_info(st->draw, &vinfo);
+#endif
+}
+
+
+/**
+ * Called by VBO to draw arrays when in selection or feedback mode and
+ * to implement glRasterPos.
+ * This is very much like the normal draw_vbo() function above.
+ * Look at code refactoring some day.
+ * Might move this into the failover module some day.
+ */
+void
+st_feedback_draw_vbo(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)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct draw_context *draw = st->draw;
+   const struct st_vertex_program *vp;
+   const struct pipe_shader_state *vs;
+   struct pipe_buffer *index_buffer_handle = 0;
+   struct pipe_vertex_buffer vbuffers[PIPE_MAX_SHADER_INPUTS];
+   struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
+   GLuint attr, i;
+   ubyte *mapped_constants;
+
+   assert(draw);
+
+   st_validate_state(ctx->st);
+
+   /* must get these after state validation! */
+   vp = ctx->st->vp;
+   vs = &st->vp->state;
+
+   if (!st->vp->draw_shader) {
+      st->vp->draw_shader = draw_create_vertex_shader(draw, vs);
+   }
+
+   /*
+    * Set up the draw module's state.
+    *
+    * We'd like to do this less frequently, but the normal state-update
+    * code sends state updates to the pipe, not to our private draw module.
+    */
+   assert(draw);
+   draw_set_viewport_state(draw, &st->state.viewport);
+   draw_set_clip_state(draw, &st->state.clip);
+   draw_set_rasterizer_state(draw, &st->state.rasterizer);
+   draw_bind_vertex_shader(draw, st->vp->draw_shader);
+   set_feedback_vertex_format(ctx);
+
+   /* loop over TGSI shader inputs to determine vertex buffer
+    * and attribute info
+    */
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
+      void *map;
+
+      if (bufobj && bufobj->Name) {
+         /* Attribute data is in a VBO.
+          * Recall that for VBOs, the gl_client_array->Ptr field is
+          * really an offset from the start of the VBO, not a pointer.
+          */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         assert(stobj->buffer);
+
+         vbuffers[attr].buffer = NULL;
+         pipe_buffer_reference(pipe->screen, &vbuffers[attr].buffer, stobj->buffer);
+         vbuffers[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */
+         velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr;
+      }
+      else {
+         /* attribute data is in user-space memory, not a VBO */
+         uint bytes = (arrays[mesaAttr]->Size
+                       * _mesa_sizeof_type(arrays[mesaAttr]->Type)
+                       * (max_index + 1));
+
+         /* wrap user data */
+         vbuffers[attr].buffer
+            = pipe_user_buffer_create(pipe->screen, (void *) arrays[mesaAttr]->Ptr,
+                                      bytes);
+         vbuffers[attr].buffer_offset = 0;
+         velements[attr].src_offset = 0;
+      }
+
+      /* common-case setup */
+      vbuffers[attr].stride = arrays[mesaAttr]->StrideB; /* in bytes */
+      vbuffers[attr].max_index = max_index;
+      velements[attr].vertex_buffer_index = attr;
+      velements[attr].nr_components = arrays[mesaAttr]->Size;
+      velements[attr].src_format = 
+         st_pipe_vertex_format(arrays[mesaAttr]->Type,
+                               arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Normalized);
+      assert(velements[attr].src_format);
+
+      /* tell draw about this attribute */
+#if 0
+      draw_set_vertex_buffer(draw, attr, &vbuffer[attr]);
+#endif
+
+      /* map the attrib buffer */
+      map = pipe_buffer_map(pipe->screen, vbuffers[attr].buffer,
+                            PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_vertex_buffer(draw, attr, map);
+   }
+
+   draw_set_vertex_buffers(draw, vp->num_inputs, vbuffers);
+   draw_set_vertex_elements(draw, vp->num_inputs, velements);
+
+   if (ib) {
+      unsigned indexSize;
+      struct gl_buffer_object *bufobj = ib->obj;
+      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+      void *map;
+
+      index_buffer_handle = stobj->buffer;
+
+      switch (ib->type) {
+      case GL_UNSIGNED_INT:
+         indexSize = 4;
+         break;
+      case GL_UNSIGNED_SHORT:
+         indexSize = 2;
+         break;
+      default:
+         assert(0);
+	 return;
+      }
+
+      map = pipe_buffer_map(pipe->screen, index_buffer_handle,
+                            PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_element_buffer(draw, indexSize, map);
+   }
+   else {
+      /* no index/element buffer */
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+
+
+   /* map constant buffers */
+   mapped_constants = pipe_buffer_map(pipe->screen,
+                                      st->state.constants[PIPE_SHADER_VERTEX].buffer,
+                                      PIPE_BUFFER_USAGE_CPU_READ);
+   draw_set_mapped_constant_buffer(st->draw, mapped_constants,
+                                   st->state.constants[PIPE_SHADER_VERTEX].buffer->size);
+
+
+   /* draw here */
+   for (i = 0; i < nr_prims; i++) {
+      draw_arrays(draw, prims[i].mode, prims[i].start, prims[i].count);
+   }
+
+
+   /* unmap constant buffers */
+   pipe_buffer_unmap(pipe->screen, st->state.constants[PIPE_SHADER_VERTEX].buffer);
+
+   /*
+    * unmap vertex/index buffers
+    */
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+      if (draw->pt.vertex_buffer[i].buffer) {
+         pipe_buffer_unmap(pipe->screen, draw->pt.vertex_buffer[i].buffer);
+         pipe_buffer_reference(pipe->screen, &draw->pt.vertex_buffer[i].buffer, NULL);
+         draw_set_mapped_vertex_buffer(draw, i, NULL);
+      }
+   }
+   if (ib) {
+      pipe_buffer_unmap(pipe->screen, index_buffer_handle);
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+}
+
+#endif /* FEATURE_feedback || FEATURE_drawpix */
+
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
new file mode 100644
index 00000000000..cd34529d385
--- /dev/null
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -0,0 +1,279 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * 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 "main/imports.h"
+#include "main/context.h"
+#include "main/extensions.h"
+#include "main/macros.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+
+#include "st_context.h"
+#include "st_extensions.h"
+
+
+static int _min(int a, int b)
+{
+   return (a < b) ? a : b;
+}
+
+static float _maxf(float a, float b)
+{
+   return (a > b) ? a : b;
+}
+
+static int _clamp(int a, int min, int max)
+{
+   if (a < min)
+      return min;
+   else if (a > max)
+      return max;
+   else
+      return a;
+}
+
+
+/**
+ * Query driver to get implementation limits.
+ * Note that we have to limit/clamp against Mesa's internal limits too.
+ */
+void st_init_limits(struct st_context *st)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct gl_constants *c = &st->ctx->Const;
+
+   c->MaxTextureLevels
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS),
+            MAX_TEXTURE_LEVELS);
+
+   c->Max3DTextureLevels
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_3D_LEVELS),
+            MAX_3D_TEXTURE_LEVELS);
+
+   c->MaxCubeTextureLevels
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS),
+            MAX_CUBE_TEXTURE_LEVELS);
+
+   c->MaxTextureRectSize
+      = _min(1 << (c->MaxTextureLevels - 1), MAX_TEXTURE_RECT_SIZE);
+
+   c->MaxTextureImageUnits
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS),
+            MAX_TEXTURE_IMAGE_UNITS);
+
+   c->MaxVertexTextureImageUnits
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS),
+             MAX_VERTEX_TEXTURE_IMAGE_UNITS);
+
+   c->MaxTextureCoordUnits
+      = _min(c->MaxTextureImageUnits, MAX_TEXTURE_COORD_UNITS);
+
+   c->MaxTextureUnits = _min(c->MaxTextureImageUnits, c->MaxTextureCoordUnits);
+
+   c->MaxDrawBuffers
+      = _clamp(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS),
+              1, MAX_DRAW_BUFFERS);
+
+   c->MaxLineWidth
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH));
+   c->MaxLineWidthAA
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA));
+
+   c->MaxPointSize
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH));
+   c->MaxPointSizeAA
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA));
+
+   c->MaxTextureMaxAnisotropy
+      = _maxf(2.0f, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY));
+
+   c->MaxTextureLodBias
+      = screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_LOD_BIAS);
+
+   c->MaxDrawBuffers
+      = CLAMP(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS),
+              1, MAX_DRAW_BUFFERS);
+}
+
+
+/**
+ * Use pipe_screen::get_param() to query PIPE_CAP_ values to determine
+ * which GL extensions are supported.
+ * Quite a few extensions are always supported because they are standard
+ * features or can be built on top of other gallium features.
+ * Some fine tuning may still be needed.
+ */
+void st_init_extensions(struct st_context *st)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   GLcontext *ctx = st->ctx;
+
+   /*
+    * Extensions that are supported by all Gallium drivers:
+    */
+   ctx->Extensions.ARB_multisample = GL_TRUE;
+   ctx->Extensions.ARB_fragment_program = GL_TRUE;
+   ctx->Extensions.ARB_texture_border_clamp = GL_TRUE; /* XXX temp */
+   ctx->Extensions.ARB_texture_compression = GL_TRUE;
+   ctx->Extensions.ARB_texture_cube_map = GL_TRUE;
+   ctx->Extensions.ARB_texture_env_combine = GL_TRUE;
+   ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE;
+   ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE;
+   ctx->Extensions.ARB_vertex_program = GL_TRUE;
+   ctx->Extensions.ARB_vertex_buffer_object = 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;
+   ctx->Extensions.EXT_blend_logic_op = GL_TRUE;
+   ctx->Extensions.EXT_blend_minmax = GL_TRUE;
+   ctx->Extensions.EXT_blend_subtract = GL_TRUE;
+   ctx->Extensions.EXT_framebuffer_blit = GL_TRUE;
+   ctx->Extensions.EXT_framebuffer_object = GL_TRUE;
+   ctx->Extensions.EXT_fog_coord = GL_TRUE;
+   ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;
+   ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE;
+   ctx->Extensions.EXT_point_parameters = GL_TRUE;
+   ctx->Extensions.EXT_secondary_color = GL_TRUE;
+   ctx->Extensions.EXT_stencil_wrap = GL_TRUE;
+   ctx->Extensions.EXT_texture_env_add = GL_TRUE;
+   ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
+   ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
+   ctx->Extensions.EXT_texture_lod_bias = GL_TRUE;
+
+   ctx->Extensions.NV_blend_square = GL_TRUE;
+   ctx->Extensions.NV_texgen_reflection = GL_TRUE;
+
+   ctx->Extensions.SGI_color_matrix = GL_TRUE;
+   ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
+
+   /*
+    * Extensions that depend on the driver/hardware:
+    */
+   if (screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS) > 0) {
+      ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_GLSL)) {
+      ctx->Extensions.ARB_fragment_shader = GL_TRUE;
+      ctx->Extensions.ARB_vertex_shader = GL_TRUE;
+      ctx->Extensions.ARB_shader_objects = GL_TRUE;
+      ctx->Extensions.ARB_shading_language_100 = GL_TRUE;
+      ctx->Extensions.ARB_shading_language_120 = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_REPEAT) > 0) {
+      ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_CLAMP) > 0) {
+      ctx->Extensions.EXT_texture_mirror_clamp = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) {
+      ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE;
+      ctx->Extensions.NV_texture_rectangle = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS) > 1) {
+      ctx->Extensions.ARB_multitexture = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) {
+      ctx->Extensions.ATI_separate_stencil = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_ANISOTROPIC_FILTER)) {
+      ctx->Extensions.EXT_texture_filter_anisotropic = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_POINT_SPRITE)) {
+      ctx->Extensions.ARB_point_sprite = GL_TRUE;
+      ctx->Extensions.NV_point_sprite = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_OCCLUSION_QUERY)) {
+      ctx->Extensions.ARB_occlusion_query = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TEXTURE_SHADOW_MAP)) {
+      ctx->Extensions.ARB_depth_texture = GL_TRUE;
+      ctx->Extensions.ARB_shadow = GL_TRUE;
+      ctx->Extensions.EXT_shadow_funcs = GL_TRUE;
+      /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/
+   }
+
+   /* GL_EXT_packed_depth_stencil requires both the ability to render to
+    * a depth/stencil buffer and texture from depth/stencil source.
+    */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_Z24S8_UNORM,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0) &&
+       screen->is_format_supported(screen, PIPE_FORMAT_Z24S8_UNORM,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE;
+   }
+   else if (screen->is_format_supported(screen, PIPE_FORMAT_S8Z24_UNORM,
+                                        PIPE_TEXTURE_2D, 
+                                        PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0) &&
+            screen->is_format_supported(screen, PIPE_FORMAT_S8Z24_UNORM,
+                                        PIPE_TEXTURE_2D, 
+                                        PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE;
+   }
+
+   /* sRGB support */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SRGB,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0) ||
+      screen->is_format_supported(screen, PIPE_FORMAT_A8R8G8B8_SRGB,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_texture_sRGB = GL_TRUE;
+   }
+
+   /* s3tc support */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE;
+   }
+
+   /* ycbcr support */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, 
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0) ||
+       screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, 
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.MESA_ycbcr_texture = GL_TRUE;
+   }
+}
diff --git a/src/mesa/state_tracker/st_extensions.h b/src/mesa/state_tracker/st_extensions.h
new file mode 100644
index 00000000000..2994f16dd33
--- /dev/null
+++ b/src/mesa/state_tracker/st_extensions.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_EXTENSIONS_H
+#define ST_EXTENSIONS_H
+
+
+extern void st_init_limits(struct st_context *st);
+
+extern void st_init_extensions(struct st_context *st);
+
+
+#endif /* ST_EXTENSIONS_H */
diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c
new file mode 100644
index 00000000000..9e2d60c9269
--- /dev/null
+++ b/src/mesa/state_tracker/st_format.c
@@ -0,0 +1,718 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * 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.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Texture Image-related functions.
+ * \author Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/texstore.h"
+#include "main/texformat.h"
+#include "main/enums.h"
+#include "main/macros.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "st_context.h"
+#include "st_format.h"
+
+static GLuint
+format_bits(
+   pipe_format_rgbazs_t  info,
+   GLuint comp )
+{
+   return pf_get_component_bits( (enum pipe_format) info, comp );
+}
+
+static GLuint
+format_max_bits(
+   pipe_format_rgbazs_t  info )
+{
+   GLuint   size = format_bits( info, PIPE_FORMAT_COMP_R );
+
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_G ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_B ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_A ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_Z ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_S ) );
+   return size;
+}
+
+static GLuint
+format_size(
+   pipe_format_rgbazs_t  info )
+{
+   return
+      format_bits( info, PIPE_FORMAT_COMP_R ) +
+      format_bits( info, PIPE_FORMAT_COMP_G ) +
+      format_bits( info, PIPE_FORMAT_COMP_B ) +
+      format_bits( info, PIPE_FORMAT_COMP_A ) +
+      format_bits( info, PIPE_FORMAT_COMP_Z ) +
+      format_bits( info, PIPE_FORMAT_COMP_S );
+}
+
+/*
+ * XXX temporary here
+ */
+GLboolean
+st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo)
+{
+   if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) {
+      pipe_format_rgbazs_t info;
+
+      info = format;
+
+#if 0
+      printf("%s\n", pf_name( format ) );
+#endif
+
+      /* Data type */
+      if (format == PIPE_FORMAT_A1R5G5B5_UNORM || format == PIPE_FORMAT_R5G6B5_UNORM) {
+         pinfo->datatype = GL_UNSIGNED_SHORT;
+      }
+      else {
+         GLuint size;
+
+         size = format_max_bits( info );
+         if (size == 8) {
+            if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM)
+               pinfo->datatype = GL_UNSIGNED_BYTE;
+            else
+               pinfo->datatype = GL_BYTE;
+         }
+         else if (size == 16) {
+            if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM)
+               pinfo->datatype = GL_UNSIGNED_SHORT;
+            else
+               pinfo->datatype = GL_SHORT;
+         }
+         else {
+            assert( size <= 32 );
+            if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM)
+               pinfo->datatype = GL_UNSIGNED_INT;
+            else
+               pinfo->datatype = GL_INT;
+         }
+      }
+
+      /* Component bits */
+      pinfo->red_bits = format_bits( info, PIPE_FORMAT_COMP_R );
+      pinfo->green_bits = format_bits( info, PIPE_FORMAT_COMP_G );
+      pinfo->blue_bits = format_bits( info, PIPE_FORMAT_COMP_B );
+      pinfo->alpha_bits = format_bits( info, PIPE_FORMAT_COMP_A );
+      pinfo->depth_bits = format_bits( info, PIPE_FORMAT_COMP_Z );
+      pinfo->stencil_bits = format_bits( info, PIPE_FORMAT_COMP_S );
+      pinfo->luminance_bits = 0;
+      pinfo->intensity_bits = 0;
+
+      /* Format size */
+      pinfo->size = format_size( info ) / 8;
+
+      /* Luminance & Intensity bits */
+      if( pf_swizzle_x(info) == PIPE_FORMAT_COMP_R &&
+          pf_swizzle_y(info) == PIPE_FORMAT_COMP_R &&
+          pf_swizzle_z(info) == PIPE_FORMAT_COMP_R ) {
+         if( pf_swizzle_w(info) == PIPE_FORMAT_COMP_R ) {
+            pinfo->intensity_bits = pinfo->red_bits;
+         }
+         else {
+            pinfo->luminance_bits = pinfo->red_bits;
+         }
+         pinfo->red_bits = 0;
+      }
+
+      /* Base format */
+      if (pinfo->depth_bits) {
+         if (pinfo->stencil_bits) {
+            pinfo->base_format = GL_DEPTH_STENCIL_EXT;
+         }
+         else {
+            pinfo->base_format = GL_DEPTH_COMPONENT;
+         }
+      }
+      else if (pinfo->stencil_bits) {
+         pinfo->base_format = GL_STENCIL_INDEX;
+      }
+      else {
+         pinfo->base_format = GL_RGBA;
+      }
+   }
+   else if (pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR) {
+      pinfo->base_format = GL_YCBCR_MESA;
+      pinfo->datatype = GL_UNSIGNED_SHORT;
+      pinfo->size = 2; /* two bytes per "texel" */
+   }
+   else {
+      /* compressed format? */
+      assert(0);
+   }
+
+#if 0
+   printf(
+      "ST_FORMAT: R(%u), G(%u), B(%u), A(%u), Z(%u), S(%u)\n",
+      pinfo->red_bits,
+      pinfo->green_bits,
+      pinfo->blue_bits,
+      pinfo->alpha_bits,
+      pinfo->depth_bits,
+      pinfo->stencil_bits );
+#endif
+
+   pinfo->format = format;
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Return bytes per pixel for the given format.
+ */
+GLuint
+st_sizeof_format(enum pipe_format format)
+{
+   struct pipe_format_info info;
+   if (!st_get_format_info( format, &info )) {
+      assert( 0 );
+      return 0;
+   }
+   return info.size;
+}
+
+
+/**
+ * Return bytes per pixel for the given format.
+ */
+GLenum
+st_format_datatype(enum pipe_format format)
+{
+   struct pipe_format_info info;
+   if (!st_get_format_info( format, &info )) {
+      assert( 0 );
+      return 0;
+   }
+   return info.datatype;
+}
+
+
+enum pipe_format
+st_mesa_format_to_pipe_format(GLuint mesaFormat)
+{
+   switch (mesaFormat) {
+      /* fix this */
+   case MESA_FORMAT_ARGB8888_REV:
+   case MESA_FORMAT_ARGB8888:
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   case MESA_FORMAT_ARGB1555:
+      return PIPE_FORMAT_A1R5G5B5_UNORM;
+   case MESA_FORMAT_ARGB4444:
+      return PIPE_FORMAT_A4R4G4B4_UNORM;
+   case MESA_FORMAT_RGB565:
+      return PIPE_FORMAT_R5G6B5_UNORM;
+   case MESA_FORMAT_AL88:
+      return PIPE_FORMAT_A8L8_UNORM;
+   case MESA_FORMAT_A8:
+      return PIPE_FORMAT_A8_UNORM;
+   case MESA_FORMAT_L8:
+      return PIPE_FORMAT_L8_UNORM;
+   case MESA_FORMAT_I8:
+      return PIPE_FORMAT_I8_UNORM;
+   case MESA_FORMAT_Z16:
+      return PIPE_FORMAT_Z16_UNORM;
+   case MESA_FORMAT_Z32:
+      return PIPE_FORMAT_Z32_UNORM;
+   case MESA_FORMAT_Z24_S8:
+      return PIPE_FORMAT_Z24S8_UNORM;
+   case MESA_FORMAT_S8_Z24:
+      return PIPE_FORMAT_S8Z24_UNORM;
+   case MESA_FORMAT_YCBCR:
+      return PIPE_FORMAT_YCBCR;
+#if FEATURE_texture_s3tc
+   case MESA_FORMAT_RGB_DXT1:
+      return PIPE_FORMAT_DXT1_RGB;
+   case MESA_FORMAT_RGBA_DXT1:
+      return PIPE_FORMAT_DXT1_RGBA;
+   case MESA_FORMAT_RGBA_DXT3:
+      return PIPE_FORMAT_DXT3_RGBA;
+   case MESA_FORMAT_RGBA_DXT5:
+      return PIPE_FORMAT_DXT5_RGBA;
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+      return PIPE_FORMAT_DXT1_SRGB;
+   case MESA_FORMAT_SRGBA_DXT1:
+      return PIPE_FORMAT_DXT1_SRGBA;
+   case MESA_FORMAT_SRGBA_DXT3:
+      return PIPE_FORMAT_DXT3_SRGBA;
+   case MESA_FORMAT_SRGBA_DXT5:
+      return PIPE_FORMAT_DXT5_SRGBA;
+#endif
+#endif
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SLA8:
+      return PIPE_FORMAT_A8L8_SRGB;
+   case MESA_FORMAT_SL8:
+      return PIPE_FORMAT_L8_SRGB;
+   case MESA_FORMAT_SRGB8:
+      return PIPE_FORMAT_R8G8B8_SRGB;
+   case MESA_FORMAT_SRGBA8:
+      return PIPE_FORMAT_R8G8B8A8_SRGB;
+   case MESA_FORMAT_SARGB8:
+      return PIPE_FORMAT_A8R8G8B8_SRGB;
+#endif
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+/**
+ * Find an RGBA format supported by the context/winsys.
+ */
+static enum pipe_format
+default_rgba_format(struct pipe_screen *screen, 
+                    enum pipe_texture_target target,
+                    unsigned tex_usage, 
+                    unsigned geom_flags)
+{
+   static const enum pipe_format colorFormats[] = {
+      PIPE_FORMAT_A8R8G8B8_UNORM,
+      PIPE_FORMAT_B8G8R8A8_UNORM,
+      PIPE_FORMAT_R8G8B8A8_UNORM,
+      PIPE_FORMAT_R5G6B5_UNORM
+   };
+   uint i;
+   for (i = 0; i < Elements(colorFormats); i++) {
+      if (screen->is_format_supported( screen, colorFormats[i], target, tex_usage, geom_flags )) {
+         return colorFormats[i];
+      }
+   }
+   return PIPE_FORMAT_NONE;
+}
+
+/**
+ * Find an sRGBA format supported by the context/winsys.
+ */
+static enum pipe_format
+default_srgba_format(struct pipe_screen *screen, 
+                    enum pipe_texture_target target,
+                    unsigned tex_usage, 
+                    unsigned geom_flags)
+{
+   static const enum pipe_format colorFormats[] = {
+      PIPE_FORMAT_A8R8G8B8_SRGB,
+      PIPE_FORMAT_B8G8R8A8_SRGB,
+      PIPE_FORMAT_R8G8B8A8_SRGB,
+   };
+   uint i;
+   for (i = 0; i < Elements(colorFormats); i++) {
+      if (screen->is_format_supported( screen, colorFormats[i], target, tex_usage, geom_flags )) {
+         return colorFormats[i];
+      }
+   }
+   return PIPE_FORMAT_NONE;
+}
+
+/**
+ * Search list of formats for first RGBA format with >8 bits/channel.
+ */
+static enum pipe_format
+default_deep_rgba_format(struct pipe_screen *screen, 
+                         enum pipe_texture_target target,
+                         unsigned tex_usage, 
+                         unsigned geom_flags)
+{
+   if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, target, tex_usage, geom_flags)) {
+      return PIPE_FORMAT_R16G16B16A16_SNORM;
+   }
+   if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET)
+      return default_rgba_format(screen, target, tex_usage, geom_flags);
+   else
+      return PIPE_FORMAT_NONE;
+}
+
+
+/**
+ * Find an Z format supported by the context/winsys.
+ */
+static enum pipe_format
+default_depth_format(struct pipe_screen *screen, 
+                     enum pipe_texture_target target,
+                     unsigned tex_usage, 
+                     unsigned geom_flags)
+{
+   static const enum pipe_format zFormats[] = {
+      PIPE_FORMAT_Z16_UNORM,
+      PIPE_FORMAT_Z32_UNORM,
+      PIPE_FORMAT_S8Z24_UNORM,
+      PIPE_FORMAT_Z24S8_UNORM
+   };
+   uint i;
+   for (i = 0; i < Elements(zFormats); i++) {
+      if (screen->is_format_supported( screen, zFormats[i], target, tex_usage, geom_flags )) {
+         return zFormats[i];
+      }
+   }
+   return PIPE_FORMAT_NONE;
+}
+
+
+/**
+ * Given an OpenGL internalFormat value for a texture or surface, return
+ * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match.
+ * \param target  one of PIPE_TEXTURE_x
+ * \param tex_usage  either PIPE_TEXTURE_USAGE_RENDER_TARGET
+ *                   or PIPE_TEXTURE_USAGE_SAMPLER
+ */
+enum pipe_format
+st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+                 enum pipe_texture_target target, unsigned tex_usage)
+{
+   struct pipe_screen *screen = pipe->screen;
+   unsigned geom_flags = 0;
+
+   switch (internalFormat) {
+   case 4:
+   case GL_RGBA:
+   case GL_COMPRESSED_RGBA:
+   case 3:
+   case GL_RGB:
+   case GL_COMPRESSED_RGB:
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+   case GL_RGBA16:
+      if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET)
+         return default_deep_rgba_format( screen, target, tex_usage, geom_flags );
+      else
+         return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGBA4:
+   case GL_RGBA2:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A4R4G4B4_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A4R4G4B4_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGB5_A1:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A1R5G5B5_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGB5:
+   case GL_RGB4:
+   case GL_R3_G3_B2:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A1R5G5B5_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_R5G6B5_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_ALPHA:
+   case GL_ALPHA4:
+   case GL_ALPHA8:
+   case GL_ALPHA12:
+   case GL_ALPHA16:
+   case GL_COMPRESSED_ALPHA:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case 1:
+   case GL_LUMINANCE:
+   case GL_LUMINANCE4:
+   case GL_LUMINANCE8:
+   case GL_LUMINANCE12:
+   case GL_LUMINANCE16:
+   case GL_COMPRESSED_LUMINANCE:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_L8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case 2:
+   case GL_LUMINANCE_ALPHA:
+   case GL_LUMINANCE4_ALPHA4:
+   case GL_LUMINANCE6_ALPHA2:
+   case GL_LUMINANCE8_ALPHA8:
+   case GL_LUMINANCE12_ALPHA4:
+   case GL_LUMINANCE12_ALPHA12:
+   case GL_LUMINANCE16_ALPHA16:
+   case GL_COMPRESSED_LUMINANCE_ALPHA:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A8L8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_INTENSITY:
+   case GL_INTENSITY4:
+   case GL_INTENSITY8:
+   case GL_INTENSITY12:
+   case GL_INTENSITY16:
+   case GL_COMPRESSED_INTENSITY:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_I8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_YCBCR_MESA:
+      if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR,
+                                      target, tex_usage, geom_flags)) {
+         return PIPE_FORMAT_YCBCR;
+      }
+      if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV,
+                                      target, tex_usage, geom_flags)) {
+         return PIPE_FORMAT_YCBCR_REV;
+      }
+      return PIPE_FORMAT_NONE;
+
+   case GL_RGB_S3TC:
+   case GL_RGB4_S3TC:
+   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_RGB;
+
+   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_RGBA;
+
+   case GL_RGBA_S3TC:
+   case GL_RGBA4_S3TC:
+   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+      return PIPE_FORMAT_DXT3_RGBA;
+
+   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+      return PIPE_FORMAT_DXT5_RGBA;
+
+#if 0
+   case GL_COMPRESSED_RGB_FXT1_3DFX:
+      return PIPE_FORMAT_RGB_FXT1;
+   case GL_COMPRESSED_RGBA_FXT1_3DFX:
+      return PIPE_FORMAT_RGB_FXT1;
+#endif
+
+   case GL_DEPTH_COMPONENT16:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z16_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z16_UNORM;
+      /* fall-through */
+   case GL_DEPTH_COMPONENT24:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8Z24_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z24S8_UNORM;
+      /* fall-through */
+   case GL_DEPTH_COMPONENT32:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z32_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z32_UNORM;
+      /* fall-through */
+   case GL_DEPTH_COMPONENT:
+      return default_depth_format( screen, target, tex_usage, geom_flags );
+
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8Z24_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z24S8_UNORM;
+      return PIPE_FORMAT_NONE;
+
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8Z24_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z24S8_UNORM;
+      return PIPE_FORMAT_NONE;
+
+   case GL_SRGB_EXT:
+   case GL_SRGB8_EXT:
+   case GL_COMPRESSED_SRGB_EXT:
+   case GL_COMPRESSED_SRGB_ALPHA_EXT:
+   case GL_SRGB_ALPHA_EXT:
+   case GL_SRGB8_ALPHA8_EXT:
+      return default_srgba_format( screen, target, tex_usage, geom_flags );
+   case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_SRGB;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_SRGBA;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+      return PIPE_FORMAT_DXT3_SRGBA;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
+      return PIPE_FORMAT_DXT5_SRGBA;
+
+   case GL_SLUMINANCE_ALPHA_EXT:
+   case GL_SLUMINANCE8_ALPHA8_EXT:
+   case GL_COMPRESSED_SLUMINANCE_EXT:
+   case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_SRGB, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A8L8_SRGB;
+      return default_srgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_SLUMINANCE_EXT:
+   case GL_SLUMINANCE8_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_L8_SRGB, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_L8_SRGB;
+      return default_srgba_format( screen, target, tex_usage, geom_flags );
+
+   default:
+      return PIPE_FORMAT_NONE;
+   }
+}
+
+
+static GLboolean
+is_stencil_format(GLenum format)
+{
+   switch (format) {
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+/**
+ * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces.
+ */
+enum pipe_format
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat)
+{
+   uint usage;
+   if (is_stencil_format(internalFormat))
+      usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   else
+      usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
+   return st_choose_format(pipe, internalFormat, PIPE_TEXTURE_2D, usage);
+}
+
+
+static const struct gl_texture_format *
+translate_gallium_format_to_mesa_format(enum pipe_format format)
+{
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return &_mesa_texformat_argb8888;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      return &_mesa_texformat_argb1555;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      return &_mesa_texformat_argb4444;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      return &_mesa_texformat_rgb565;
+   case PIPE_FORMAT_A8L8_UNORM:
+      return &_mesa_texformat_al88;
+   case PIPE_FORMAT_A8_UNORM:
+      return &_mesa_texformat_a8;
+   case PIPE_FORMAT_L8_UNORM:
+      return &_mesa_texformat_l8;
+   case PIPE_FORMAT_I8_UNORM:
+      return &_mesa_texformat_i8;
+   case PIPE_FORMAT_Z16_UNORM:
+      return &_mesa_texformat_z16;
+   case PIPE_FORMAT_Z32_UNORM:
+      return &_mesa_texformat_z32;
+   case PIPE_FORMAT_Z24S8_UNORM:
+      return &_mesa_texformat_z24_s8;
+   case PIPE_FORMAT_S8Z24_UNORM:
+      return &_mesa_texformat_s8_z24;
+   case PIPE_FORMAT_YCBCR:
+      return &_mesa_texformat_ycbcr;
+   case PIPE_FORMAT_YCBCR_REV:
+      return &_mesa_texformat_ycbcr_rev;
+#if FEATURE_texture_s3tc
+   case PIPE_FORMAT_DXT1_RGB:
+      return &_mesa_texformat_rgb_dxt1;
+   case PIPE_FORMAT_DXT1_RGBA:
+      return &_mesa_texformat_rgba_dxt1;
+   case PIPE_FORMAT_DXT3_RGBA:
+      return &_mesa_texformat_rgba_dxt3;
+   case PIPE_FORMAT_DXT5_RGBA:
+      return &_mesa_texformat_rgba_dxt5;
+#if FEATURE_EXT_texture_sRGB
+   case PIPE_FORMAT_DXT1_SRGB:
+      return &_mesa_texformat_srgb_dxt1;
+   case PIPE_FORMAT_DXT1_SRGBA:
+      return &_mesa_texformat_srgba_dxt1;
+   case PIPE_FORMAT_DXT3_SRGBA:
+      return &_mesa_texformat_srgba_dxt3;
+   case PIPE_FORMAT_DXT5_SRGBA:
+      return &_mesa_texformat_srgba_dxt5;
+#endif
+#endif
+#if FEATURE_EXT_texture_sRGB
+   case PIPE_FORMAT_A8L8_SRGB:
+      return &_mesa_texformat_sla8;
+   case PIPE_FORMAT_L8_SRGB:
+      return &_mesa_texformat_sl8;
+   case PIPE_FORMAT_R8G8B8_SRGB:
+      return &_mesa_texformat_srgb8;
+   case PIPE_FORMAT_R8G8B8A8_SRGB:
+      return &_mesa_texformat_srgba8;
+   case PIPE_FORMAT_A8R8G8B8_SRGB:
+      return &_mesa_texformat_sargb8;
+#endif
+   /* XXX add additional cases */
+   default:
+      assert(0);
+      return NULL;
+   }
+}
+
+
+/**
+ * Called via ctx->Driver.chooseTextureFormat().
+ */
+const struct gl_texture_format *
+st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat,
+                       GLenum format, GLenum type)
+{
+   enum pipe_format pFormat;
+
+   (void) format;
+   (void) type;
+
+   pFormat = st_choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE_2D, 
+                              PIPE_TEXTURE_USAGE_SAMPLER);
+   if (pFormat == PIPE_FORMAT_NONE)
+      return NULL;
+
+   return translate_gallium_format_to_mesa_format(pFormat);
+}
diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h
new file mode 100644
index 00000000000..3f5ac3201b1
--- /dev/null
+++ b/src/mesa/state_tracker/st_format.h
@@ -0,0 +1,79 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_FORMAT_H
+#define ST_FORMAT_H
+
+
+struct pipe_format_info
+{
+   enum pipe_format format;
+   GLenum base_format;
+   GLenum datatype;
+   GLubyte red_bits;
+   GLubyte green_bits;
+   GLubyte blue_bits;
+   GLubyte alpha_bits;
+   GLubyte luminance_bits;
+   GLubyte intensity_bits;
+   GLubyte depth_bits;
+   GLubyte stencil_bits;
+   GLubyte size;           /**< in bytes */
+};
+
+
+GLboolean
+st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo);
+
+
+extern GLuint
+st_sizeof_format(enum pipe_format format);
+
+
+extern GLenum
+st_format_datatype(enum pipe_format format);
+
+
+extern enum pipe_format
+st_mesa_format_to_pipe_format(GLuint mesaFormat);
+
+
+extern enum pipe_format
+st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+                 enum pipe_texture_target target, unsigned tex_usage);
+
+extern enum pipe_format
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat);
+
+
+extern const struct gl_texture_format *
+st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
+                       GLenum format, GLenum type);
+
+
+#endif /* ST_CB_TEXIMAGE_H */
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
new file mode 100644
index 00000000000..0d9c7b97e3d
--- /dev/null
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -0,0 +1,333 @@
+/**************************************************************************
+ * 
+ * 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 "main/imports.h"
+#include "main/buffers.h"
+#include "main/context.h"
+#include "main/framebuffer.h"
+#include "main/matrix.h"
+#include "main/renderbuffer.h"
+#include "main/scissor.h"
+#include "st_context.h"
+#include "st_cb_fbo.h"
+#include "st_public.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+
+
+struct st_framebuffer *
+st_create_framebuffer( const __GLcontextModes *visual,
+                       enum pipe_format colorFormat,
+                       enum pipe_format depthFormat,
+                       enum pipe_format stencilFormat,
+                       uint width, uint height,
+                       void *private)
+{
+   struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer);
+   if (stfb) {
+      int samples = st_get_msaa();
+
+      if (visual->sampleBuffers)
+         samples = visual->samples;
+
+      _mesa_initialize_framebuffer(&stfb->Base, visual);
+
+      {
+         /* fake frontbuffer */
+         /* XXX allocation should only happen in the unusual case
+            it's actually needed */
+         struct gl_renderbuffer *rb
+            = st_new_renderbuffer_fb(colorFormat, samples);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
+      }
+
+      if (visual->doubleBufferMode) {
+         struct gl_renderbuffer *rb
+            = st_new_renderbuffer_fb(colorFormat, samples);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
+      }
+
+      if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
+         /* combined depth/stencil buffer */
+         struct gl_renderbuffer *depthStencilRb
+            = st_new_renderbuffer_fb(depthFormat, samples);
+         /* note: bind RB to two attachment points */
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb);
+      }
+      else {
+         /* separate depth and/or stencil */
+
+         if (visual->depthBits == 32) {
+            /* 32-bit depth buffer */
+            struct gl_renderbuffer *depthRb
+               = st_new_renderbuffer_fb(depthFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
+         }
+         else if (visual->depthBits == 24) {
+            /* 24-bit depth buffer, ignore stencil bits */
+            struct gl_renderbuffer *depthRb
+               = st_new_renderbuffer_fb(depthFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
+         }
+         else if (visual->depthBits > 0) {
+            /* 16-bit depth buffer */
+            struct gl_renderbuffer *depthRb
+               = st_new_renderbuffer_fb(depthFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
+         }
+
+         if (visual->stencilBits > 0) {
+            /* 8-bit stencil */
+            struct gl_renderbuffer *stencilRb
+               = st_new_renderbuffer_fb(stencilFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb);
+         }
+      }
+
+      if (visual->accumRedBits > 0) {
+         /* 16-bit/channel accum */
+         struct gl_renderbuffer *accumRb
+            = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT, 0); /* XXX accum isn't multisampled right? */
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb);
+      }
+
+      stfb->Base.Initialized = GL_TRUE;
+      stfb->InitWidth = width;
+      stfb->InitHeight = height;
+      stfb->Private = private;
+   }
+   return stfb;
+}
+
+
+void st_resize_framebuffer( struct st_framebuffer *stfb,
+                            uint width, uint height )
+{
+   if (stfb->Base.Width != width || stfb->Base.Height != height) {
+      GET_CURRENT_CONTEXT(ctx);
+      if (ctx) {
+         if (stfb->InitWidth == 0 && stfb->InitHeight == 0) {
+            /* didn't have a valid size until now */
+            stfb->InitWidth = width;
+            stfb->InitHeight = height;
+            if (ctx->Viewport.Width <= 1) {
+               /* set context's initial viewport/scissor size */
+               _mesa_set_viewport(ctx, 0, 0, width, height);
+               _mesa_set_scissor(ctx, 0, 0, width, height);
+            }
+         }
+
+         _mesa_resize_framebuffer(ctx, &stfb->Base, width, height);
+
+         assert(stfb->Base.Width == width);
+         assert(stfb->Base.Height == height);
+      }
+   }
+}
+
+
+void st_unreference_framebuffer( struct st_framebuffer *stfb )
+{
+   _mesa_unreference_framebuffer((struct gl_framebuffer **) &stfb);
+}
+
+
+
+/**
+ * Set/replace a framebuffer surface.
+ * The user of the state tracker can use this instead of
+ * st_resize_framebuffer() to provide new surfaces when a window is resized.
+ */
+void
+st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                           uint surfIndex, struct pipe_surface *surf)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   static const GLuint invalid_size = 9999999;
+   struct st_renderbuffer *strb;
+   GLuint width, height, i;
+
+   assert(surfIndex < BUFFER_COUNT);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+
+   /* fail */
+   if (!strb) return;
+
+   /* replace the renderbuffer's surface/texture pointers */
+   pipe_surface_reference( &strb->surface, surf );
+   pipe_texture_reference( &strb->texture, surf->texture );
+
+   if (ctx) {
+      /* If ctx isn't set, we've likely not made current yet.
+       * But when we do, we need to start setting this dirty bit
+       * to ensure the renderbuffer attachements are up-to-date
+       * via update_framebuffer.
+       */
+      ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER;
+   }
+
+   /* update renderbuffer's width/height */
+   strb->Base.Width = surf->width;
+   strb->Base.Height = surf->height;
+
+   /* Try to update the framebuffer's width/height from the renderbuffer
+    * sizes.  Before we start drawing, all the rbs _should_ be the same size.
+    */
+   width = height = invalid_size;
+   for (i = 0; i < BUFFER_COUNT; i++) {
+      if (stfb->Base.Attachment[i].Renderbuffer) {
+         if (width == invalid_size) {
+            width = stfb->Base.Attachment[i].Renderbuffer->Width;
+            height = stfb->Base.Attachment[i].Renderbuffer->Height;
+         }
+         else if (width != stfb->Base.Attachment[i].Renderbuffer->Width ||
+                  height != stfb->Base.Attachment[i].Renderbuffer->Height) {
+            /* inconsistant renderbuffer sizes, bail out */
+            return;
+         }
+      }
+   }
+
+   if (width != invalid_size) {
+      /* OK, the renderbuffers are of a consistant size, so update the
+       * parent framebuffer's size.
+       */
+      stfb->Base.Width = width;
+      stfb->Base.Height = height;
+   }
+}
+
+
+
+/**
+ * Return the pipe_surface for the given renderbuffer.
+ */
+int
+st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex, struct pipe_surface **surface)
+{
+   struct st_renderbuffer *strb;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   /* sanity checks, ST tokens should match Mesa tokens */
+   assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT);
+   assert(ST_SURFACE_BACK_RIGHT == BUFFER_BACK_RIGHT);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+   if (strb) {
+      *surface = strb->surface;
+      return GL_TRUE;
+   }
+
+   *surface = NULL;
+   return GL_FALSE;
+}
+
+int
+st_get_framebuffer_texture(struct st_framebuffer *stfb, uint surfIndex, struct pipe_texture **texture)
+{
+   struct st_renderbuffer *strb;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   /* sanity checks, ST tokens should match Mesa tokens */
+   assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT);
+   assert(ST_SURFACE_BACK_RIGHT == BUFFER_BACK_RIGHT);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+   if (strb) {
+      *texture = strb->texture;
+      return GL_TRUE;
+   }
+
+   *texture = NULL;
+   return GL_FALSE;
+}
+
+/**
+ * This function is to be called prior to SwapBuffers on the given
+ * framebuffer.  It checks if the current context is bound to the framebuffer
+ * and flushes rendering if needed.
+ */
+void
+st_notify_swapbuffers(struct st_framebuffer *stfb)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      st_flush( ctx->st, 
+		PIPE_FLUSH_RENDER_CACHE | 
+		PIPE_FLUSH_SWAPBUFFERS |
+		PIPE_FLUSH_FRAME,
+                NULL );
+      ctx->st->frontbuffer_status = FRONT_STATUS_COPY_OF_BACK;
+   }
+}
+
+
+/** 
+ * Quick hack - allows the winsys to inform the driver that surface
+ * states are now undefined after a glXSwapBuffers or similar.
+ */
+void
+st_notify_swapbuffers_complete(struct st_framebuffer *stfb)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      struct st_renderbuffer *strb;
+
+      /* Mark back color buffers as undefined */
+      strb = st_renderbuffer(stfb->Base.Attachment[BUFFER_BACK_LEFT].
+                             Renderbuffer);
+      if (strb && strb->surface)
+         strb->surface->status = PIPE_SURFACE_STATUS_UNDEFINED;
+
+      strb = st_renderbuffer(stfb->Base.Attachment[BUFFER_BACK_RIGHT].
+                             Renderbuffer);
+      if (strb && strb->surface)
+         strb->surface->status = PIPE_SURFACE_STATUS_UNDEFINED;
+   }
+}
+
+
+void *st_framebuffer_private( struct st_framebuffer *stfb )
+{
+   return stfb->Private;
+}
+
+void st_get_framebuffer_dimensions( struct st_framebuffer *stfb,
+				    uint *width,
+				    uint *height)
+{
+   *width = stfb->Base.Width;
+   *height = stfb->Base.Height;
+}
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
new file mode 100644
index 00000000000..6a3455aaba2
--- /dev/null
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -0,0 +1,209 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 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 "main/imports.h"
+#include "main/mipmap.h"
+#include "main/teximage.h"
+#include "main/texformat.h"
+
+#include "shader/prog_instruction.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_gen_mipmap.h"
+
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_context.h"
+
+#include "st_context.h"
+#include "st_draw.h"
+#include "st_gen_mipmap.h"
+#include "st_program.h"
+#include "st_texture.h"
+#include "st_cb_texture.h"
+
+
+/**
+ * one-time init for generate mipmap
+ * XXX Note: there may be other times we need no-op/simple state like this.
+ * In that case, some code refactoring would be good.
+ */
+void
+st_init_generate_mipmap(struct st_context *st)
+{
+   st->gen_mipmap = util_create_gen_mipmap(st->pipe, st->cso_context);
+}
+
+
+void
+st_destroy_generate_mipmap(struct st_context *st)
+{
+   util_destroy_gen_mipmap(st->gen_mipmap);
+   st->gen_mipmap = NULL;
+}
+
+
+/**
+ * Generate mipmap levels using hardware rendering.
+ * \return TRUE if successful, FALSE if not possible
+ */
+static boolean
+st_render_mipmap(struct st_context *st,
+                 GLenum target,
+                 struct pipe_texture *pt,
+                 uint baseLevel, uint lastLevel)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const uint face = _mesa_tex_target_to_face(target);
+
+   assert(target != GL_TEXTURE_3D); /* not done yet */
+
+   /* check if we can render in the texture's format */
+   if (!screen->is_format_supported(screen, pt->format, target,
+                                    PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
+      return FALSE;
+   }
+
+   util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel,
+                   PIPE_TEX_FILTER_LINEAR);
+
+   return TRUE;
+}
+
+
+static void
+fallback_generate_mipmap(GLcontext *ctx, GLenum target,
+                         struct gl_texture_object *texObj)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *pt = st_get_texobj_texture(texObj);
+   const uint baseLevel = texObj->BaseLevel;
+   const uint lastLevel = pt->last_level;
+   const uint face = _mesa_tex_target_to_face(target), zslice = 0;
+   uint dstLevel;
+   GLenum datatype;
+   GLuint comps;
+
+   assert(target != GL_TEXTURE_3D); /* not done yet */
+
+   _mesa_format_to_type_and_comps(texObj->Image[face][baseLevel]->TexFormat,
+                                  &datatype, &comps);
+
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      struct pipe_surface *srcSurf, *dstSurf;
+      const ubyte *srcData;
+      ubyte *dstData;
+
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      srcData = (ubyte *) pipe_surface_map(srcSurf,
+                                           PIPE_BUFFER_USAGE_CPU_READ)
+              + srcSurf->offset;
+      dstData = (ubyte *) pipe_surface_map(dstSurf,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE)
+              + dstSurf->offset;
+
+      _mesa_generate_mipmap_level(target, datatype, comps,
+                   0 /*border*/,
+                   pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel],
+                   srcData,
+                   srcSurf->stride, /* stride in bytes */
+                   pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel],
+                   dstData,
+                   dstSurf->stride); /* stride in bytes */
+
+      pipe_surface_unmap(srcSurf);
+      pipe_surface_unmap(dstSurf);
+
+      pipe_surface_reference(&srcSurf, NULL);
+      pipe_surface_reference(&dstSurf, NULL);
+   }
+}
+
+
+void
+st_generate_mipmap(GLcontext *ctx, GLenum target,
+                   struct gl_texture_object *texObj)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_texture *pt = st_get_texobj_texture(texObj);
+   const uint baseLevel = texObj->BaseLevel;
+   uint lastLevel;
+   uint dstLevel;
+
+   if (!pt)
+      return;
+
+   lastLevel = pt->last_level;
+
+   if (!st_render_mipmap(st, target, pt, baseLevel, lastLevel)) {
+      fallback_generate_mipmap(ctx, target, texObj);
+   }
+
+   /* Fill in the Mesa gl_texture_image fields */
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      const struct gl_texture_image *srcImage
+         = _mesa_get_tex_image(ctx, texObj, target, srcLevel);
+      struct gl_texture_image *dstImage;
+      struct st_texture_image *stImage;
+      uint dstWidth = pt->width[dstLevel];
+      uint dstHeight = pt->height[dstLevel];
+      uint dstDepth = pt->depth[dstLevel];
+      uint border = srcImage->Border;
+
+      dstImage = _mesa_get_tex_image(ctx, texObj, target, dstLevel);
+      if (!dstImage) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "generating mipmaps");
+         return;
+      }
+
+      if (dstImage->ImageOffsets)
+         _mesa_free(dstImage->ImageOffsets);
+
+      /* Free old image data */
+      if (dstImage->Data)
+         ctx->Driver.FreeTexImageData(ctx, dstImage);
+
+      /* initialize new image */
+      _mesa_init_teximage_fields(ctx, target, dstImage, dstWidth, dstHeight,
+                                 dstDepth, border, srcImage->InternalFormat);
+
+      dstImage->TexFormat = srcImage->TexFormat;
+
+      stImage = (struct st_texture_image *) dstImage;
+      pipe_texture_reference(&stImage->pt, pt);
+   }
+}
diff --git a/src/mesa/state_tracker/st_gen_mipmap.h b/src/mesa/state_tracker/st_gen_mipmap.h
new file mode 100644
index 00000000000..00fbae93026
--- /dev/null
+++ b/src/mesa/state_tracker/st_gen_mipmap.h
@@ -0,0 +1,46 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_GEN_MIPMAP_H
+#define ST_GEN_MIPMAP_H
+
+
+extern void
+st_init_generate_mipmap(struct st_context *st);
+
+
+extern void
+st_destroy_generate_mipmap(struct st_context *st);
+
+
+extern void
+st_generate_mipmap(GLcontext *ctx, GLenum target,
+                   struct gl_texture_object *texObj);
+
+
+#endif /* ST_GEN_MIPMAP_H */
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
new file mode 100644
index 00000000000..8b9b18f94df
--- /dev/null
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -0,0 +1,1041 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 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.
+ * 
+ **************************************************************************/
+
+/*
+ * \author
+ * Michal Krol
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_sanity.h"
+#include "st_mesa_to_tgsi.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "pipe/p_debug.h"
+
+/*
+ * Map mesa register file to TGSI register file.
+ */
+static GLuint
+map_register_file(
+   enum register_file file,
+   GLuint index,
+   const GLuint immediateMapping[],
+   GLboolean indirectAccess )
+{
+   switch( file ) {
+   case PROGRAM_UNDEFINED:
+      return TGSI_FILE_NULL;
+   case PROGRAM_TEMPORARY:
+      return TGSI_FILE_TEMPORARY;
+   /*case PROGRAM_LOCAL_PARAM:*/
+   /*case PROGRAM_ENV_PARAM:*/
+
+      /* Because of the longstanding problem with mesa arb shaders
+       * where constants, immediates and state variables are all
+       * bundled together as PROGRAM_STATE_VAR, we can't tell from the
+       * mesa register file whether this is a CONSTANT or an
+       * IMMEDIATE, hence we need all the other information.
+       */
+   case PROGRAM_STATE_VAR:
+   case PROGRAM_NAMED_PARAM:
+   case PROGRAM_UNIFORM:
+      if (!indirectAccess && immediateMapping && immediateMapping[index] != ~0)
+         return TGSI_FILE_IMMEDIATE;
+      else
+	 return TGSI_FILE_CONSTANT;
+   case PROGRAM_CONSTANT:
+      if (indirectAccess)
+         return TGSI_FILE_CONSTANT;
+      assert(immediateMapping[index] != ~0);
+      return TGSI_FILE_IMMEDIATE;
+   case PROGRAM_INPUT:
+      return TGSI_FILE_INPUT;
+   case PROGRAM_OUTPUT:
+      return TGSI_FILE_OUTPUT;
+   case PROGRAM_ADDRESS:
+      return TGSI_FILE_ADDRESS;
+   default:
+      assert( 0 );
+      return TGSI_FILE_NULL;
+   }
+}
+
+/**
+ * Map mesa register file index to TGSI index.
+ * Take special care when processing input and output indices.
+ * \param file  one of TGSI_FILE_x
+ * \param index  the mesa register file index
+ * \param inputMapping  maps Mesa input indexes to TGSI input indexes
+ * \param outputMapping  maps Mesa output indexes to TGSI output indexes
+ */
+static GLuint
+map_register_file_index(
+   GLuint file,
+   GLuint index,
+   const GLuint inputMapping[],
+   const GLuint outputMapping[],
+   const GLuint immediateMapping[],
+   GLboolean indirectAccess )
+{
+   switch( file ) {
+   case TGSI_FILE_INPUT:
+      /* inputs are mapped according to the user-defined map */
+      return inputMapping[index];
+
+   case TGSI_FILE_OUTPUT:
+      return outputMapping[index];
+
+   case TGSI_FILE_IMMEDIATE:
+      if (indirectAccess)
+         return index;
+      assert(immediateMapping[index] != ~0);
+      return immediateMapping[index];
+
+   default:
+      return index;
+   }
+}
+
+/*
+ * Map mesa texture target to TGSI texture target.
+ */
+static GLuint
+map_texture_target(
+   GLuint textarget )
+{
+   switch( textarget ) {
+   case TEXTURE_1D_INDEX:
+      return TGSI_TEXTURE_1D;
+   case TEXTURE_2D_INDEX:
+      return TGSI_TEXTURE_2D;
+   case TEXTURE_3D_INDEX:
+      return TGSI_TEXTURE_3D;
+   case TEXTURE_CUBE_INDEX:
+      return TGSI_TEXTURE_CUBE;
+   case TEXTURE_RECT_INDEX:
+      return TGSI_TEXTURE_RECT;
+   default:
+      assert( 0 );
+   }
+
+   return TGSI_TEXTURE_1D;
+}
+
+static GLuint
+convert_sat(
+   GLuint sat )
+{
+   switch( sat ) {
+   case SATURATE_OFF:
+      return TGSI_SAT_NONE;
+   case SATURATE_ZERO_ONE:
+      return TGSI_SAT_ZERO_ONE;
+   case SATURATE_PLUS_MINUS_ONE:
+      return TGSI_SAT_MINUS_PLUS_ONE;
+   default:
+      assert( 0 );
+      return TGSI_SAT_NONE;
+   }
+}
+
+static GLuint
+convert_writemask(
+   GLuint writemask )
+{
+   assert( WRITEMASK_X == TGSI_WRITEMASK_X );
+   assert( WRITEMASK_Y == TGSI_WRITEMASK_Y );
+   assert( WRITEMASK_Z == TGSI_WRITEMASK_Z );
+   assert( WRITEMASK_W == TGSI_WRITEMASK_W );
+   assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 );
+
+   return writemask;
+}
+
+static struct tgsi_full_immediate
+make_immediate(const float *value, uint size)
+{
+   struct tgsi_full_immediate imm;
+
+   imm = tgsi_default_full_immediate();
+   imm.Immediate.NrTokens += size;
+   imm.Immediate.DataType = TGSI_IMM_FLOAT32;
+   imm.u.Pointer = value;
+   return imm;
+}
+
+static void
+compile_instruction(
+   const struct prog_instruction *inst,
+   struct tgsi_full_instruction *fullinst,
+   const GLuint inputMapping[],
+   const GLuint outputMapping[],
+   const GLuint immediateMapping[],
+   GLboolean indirectAccess,
+   GLuint preamble_size,
+   GLuint processor,
+   GLboolean *insideSubroutine)
+{
+   GLuint i;
+   struct tgsi_full_dst_register *fulldst;
+   struct tgsi_full_src_register *fullsrc;
+
+   *fullinst = tgsi_default_full_instruction();
+
+   fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode );
+   fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode );
+   fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode );
+
+   fulldst = &fullinst->FullDstRegisters[0];
+   fulldst->DstRegister.File = map_register_file( inst->DstReg.File, 0, NULL, GL_FALSE );
+   fulldst->DstRegister.Index = map_register_file_index(
+      fulldst->DstRegister.File,
+      inst->DstReg.Index,
+      inputMapping,
+      outputMapping,
+      NULL,
+      GL_FALSE );
+   fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask );
+
+   for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) {
+      GLuint j;
+
+      fullsrc = &fullinst->FullSrcRegisters[i];
+      fullsrc->SrcRegister.File = map_register_file(
+         inst->SrcReg[i].File,
+         inst->SrcReg[i].Index,
+         immediateMapping,
+         indirectAccess );
+      fullsrc->SrcRegister.Index = map_register_file_index(
+         fullsrc->SrcRegister.File,
+         inst->SrcReg[i].Index,
+         inputMapping,
+         outputMapping,
+         immediateMapping,
+         indirectAccess );
+
+
+      /* swizzle (ext swizzle also depends on negation) */
+      {
+         GLuint swz[4];
+         GLboolean extended = (inst->SrcReg[i].NegateBase != NEGATE_NONE &&
+                               inst->SrcReg[i].NegateBase != NEGATE_XYZW);
+         for( j = 0; j < 4; j++ ) {
+            swz[j] = GET_SWZ( inst->SrcReg[i].Swizzle, j );
+            if (swz[j] > SWIZZLE_W)
+               extended = GL_TRUE;
+         }
+         if (extended) {
+            for (j = 0; j < 4; j++) {
+               tgsi_util_set_src_register_extswizzle(&fullsrc->SrcRegisterExtSwz,
+                                                     swz[j], j);
+            }
+         }
+         else {
+            for (j = 0; j < 4; j++) {
+               tgsi_util_set_src_register_swizzle(&fullsrc->SrcRegister,
+                                                  swz[j], j);
+            }
+         }
+      }
+
+      if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) {
+         fullsrc->SrcRegister.Negate = 1;
+      }
+      else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) {
+         if( inst->SrcReg[i].NegateBase & NEGATE_X ) {
+            fullsrc->SrcRegisterExtSwz.NegateX = 1;
+         }
+         if( inst->SrcReg[i].NegateBase & NEGATE_Y ) {
+            fullsrc->SrcRegisterExtSwz.NegateY = 1;
+         }
+         if( inst->SrcReg[i].NegateBase & NEGATE_Z ) {
+            fullsrc->SrcRegisterExtSwz.NegateZ = 1;
+         }
+         if( inst->SrcReg[i].NegateBase & NEGATE_W ) {
+            fullsrc->SrcRegisterExtSwz.NegateW = 1;
+         }
+      }
+
+      if( inst->SrcReg[i].Abs ) {
+         fullsrc->SrcRegisterExtMod.Absolute = 1;
+      }
+
+      if( inst->SrcReg[i].NegateAbs ) {
+         fullsrc->SrcRegisterExtMod.Negate = 1;
+      }
+
+      if( inst->SrcReg[i].RelAddr ) {
+         fullsrc->SrcRegister.Indirect = 1;
+
+         fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS;
+         fullsrc->SrcRegisterInd.Index = 0;
+      }
+   }
+
+   switch( inst->Opcode ) {
+   case OPCODE_ARL:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ARL;
+      break;
+   case OPCODE_ABS:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ABS;
+      break;
+   case OPCODE_ADD:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ADD;
+      break;
+   case OPCODE_BGNLOOP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_BGNSUB:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB;
+      *insideSubroutine = GL_TRUE;
+      break;
+   case OPCODE_BRA:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BRA;
+      break;
+   case OPCODE_BRK:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BRK;
+      break;
+   case OPCODE_CAL:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_CAL;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_CMP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_CMP;
+      break;
+   case OPCODE_CONT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_CONT;
+      break;
+   case OPCODE_COS:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_COS;
+      break;
+   case OPCODE_DDX:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DDX;
+      break;
+   case OPCODE_DDY:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DDY;
+      break;
+   case OPCODE_DP2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP2;
+      break;
+   case OPCODE_DP2A:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP2A;
+      break;
+   case OPCODE_DP3:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP3;
+      break;
+   case OPCODE_DP4:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP4;
+      break;
+   case OPCODE_DPH:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DPH;
+      break;
+   case OPCODE_DST:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DST;
+      break;
+   case OPCODE_ELSE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_ENDIF:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF;
+      break;
+   case OPCODE_ENDLOOP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_ENDSUB:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB;
+      *insideSubroutine = GL_FALSE;
+      break;
+   case OPCODE_EX2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_EX2;
+      break;
+   case OPCODE_EXP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_EXP;
+      break;
+   case OPCODE_FLR:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_FLR;
+      break;
+   case OPCODE_FRC:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_FRC;
+      break;
+   case OPCODE_IF:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_IF;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_TRUNC:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TRUNC;
+      break;
+   case OPCODE_KIL:
+      /* conditional */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_KIL;
+      break;
+   case OPCODE_KIL_NV:
+      /* predicated */
+      assert(inst->DstReg.CondMask == COND_TR);
+      fullinst->Instruction.Opcode = TGSI_OPCODE_KILP;
+      break;
+   case OPCODE_LG2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LG2;
+      break;
+   case OPCODE_LOG:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LOG;
+      break;
+   case OPCODE_LIT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LIT;
+      break;
+   case OPCODE_LRP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LRP;
+      break;
+   case OPCODE_MAD:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MAD;
+      break;
+   case OPCODE_MAX:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MAX;
+      break;
+   case OPCODE_MIN:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MIN;
+      break;
+   case OPCODE_MOV:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MOV;
+      break;
+   case OPCODE_MUL:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MUL;
+      break;
+   case OPCODE_NOISE1:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1;
+      break;
+   case OPCODE_NOISE2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2;
+      break;
+   case OPCODE_NOISE3:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3;
+      break;
+   case OPCODE_NOISE4:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4;
+      break;
+   case OPCODE_NOP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOP;
+      break;
+   case OPCODE_NRM3:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NRM;
+      break;
+   case OPCODE_NRM4:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NRM4;
+      break;
+   case OPCODE_POW:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_POW;
+      break;
+   case OPCODE_RCP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_RCP;
+      break;
+   case OPCODE_RET:
+      /* If RET is used inside main (not a real subroutine) we may want
+       * to execute END instead of RET.  TBD...
+       */
+      if (1 /*  *insideSubroutine */) {
+         fullinst->Instruction.Opcode = TGSI_OPCODE_RET;
+      }
+      else {
+         /* inside main() pseudo-function */
+         fullinst->Instruction.Opcode = TGSI_OPCODE_END;
+      }
+      break;
+   case OPCODE_RSQ:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ;
+
+      /* KW: Don't do this here.  If particular hardware needs to do
+       * this, can do so in the driver..
+       */
+#if 0
+       tgsi_util_set_full_src_register_sign_mode(
+          &fullinst->FullSrcRegisters[0],
+          TGSI_UTIL_SIGN_CLEAR ); 
+#endif
+      break;
+   case OPCODE_SCS:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SCS;
+      fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY;
+      break;
+   case OPCODE_SEQ:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ;
+      break;
+   case OPCODE_SGE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SGE;
+      break;
+   case OPCODE_SGT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SGT;
+      break;
+   case OPCODE_SIN:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SIN;
+      break;
+   case OPCODE_SLE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SLE;
+      break;
+   case OPCODE_SLT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SLT;
+      break;
+   case OPCODE_SNE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SNE;
+      break;
+   case OPCODE_SSG:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SSG;
+      break;
+   case OPCODE_SUB:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SUB;
+      break;
+   case OPCODE_SWZ:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ;
+      break;
+   case OPCODE_TEX:
+      /* ordinary texture lookup */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TEX;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXB:
+      /* texture lookup with LOD bias */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXB;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXD:
+      /* texture lookup with explicit partial derivatives */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXD;
+      fullinst->Instruction.NumSrcRegs = 4;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      /* src[0] = coord, src[1] = d[strq]/dx, src[2] = d[strq]/dy */
+      fullinst->FullSrcRegisters[3].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[3].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXL:
+      /* texture lookup with explicit LOD */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXL;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXP:
+      /* texture lookup with divide by Q component */
+      /* convert to TEX w/ special flag for division */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXP;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_XPD:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_XPD;
+      fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ;
+      break;
+   case OPCODE_END:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_END;
+      break;
+   default:
+      assert( 0 );
+   }
+}
+
+/**
+ * \param usage_mask  bitfield of TGSI_WRITEMASK_{XYZW} tokens
+ */
+static struct tgsi_full_declaration
+make_input_decl(
+   GLuint index,
+   GLboolean interpolate_info,
+   GLuint interpolate,
+   GLuint usage_mask,
+   GLboolean semantic_info,
+   GLuint semantic_name,
+   GLbitfield semantic_index,
+   GLbitfield input_flags)
+{
+   struct tgsi_full_declaration decl;
+
+   assert(semantic_name < TGSI_SEMANTIC_COUNT);
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_INPUT;
+   decl.Declaration.UsageMask = usage_mask;
+   decl.Declaration.Semantic = semantic_info;
+   decl.DeclarationRange.First = index;
+   decl.DeclarationRange.Last = index;
+   if (semantic_info) {
+      decl.Semantic.SemanticName = semantic_name;
+      decl.Semantic.SemanticIndex = semantic_index;
+   }
+   if (interpolate_info) {
+      decl.Declaration.Interpolate = interpolate;
+   }
+   if (input_flags & PROG_PARAM_BIT_CENTROID)
+      decl.Declaration.Centroid = 1;
+   if (input_flags & PROG_PARAM_BIT_INVARIANT)
+      decl.Declaration.Invariant = 1;
+
+   return decl;
+}
+
+/**
+ * \param usage_mask  bitfield of TGSI_WRITEMASK_{XYZW} tokens
+ */
+static struct tgsi_full_declaration
+make_output_decl(
+   GLuint index,
+   GLuint semantic_name,
+   GLuint semantic_index,
+   GLuint usage_mask,
+   GLbitfield output_flags)
+{
+   struct tgsi_full_declaration decl;
+
+   assert(semantic_name < TGSI_SEMANTIC_COUNT);
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_OUTPUT;
+   decl.Declaration.UsageMask = usage_mask;
+   decl.Declaration.Semantic = 1;
+   decl.DeclarationRange.First = index;
+   decl.DeclarationRange.Last = index;
+   decl.Semantic.SemanticName = semantic_name;
+   decl.Semantic.SemanticIndex = semantic_index;
+   if (output_flags & PROG_PARAM_BIT_CENTROID)
+      decl.Declaration.Centroid = 1;
+   if (output_flags & PROG_PARAM_BIT_INVARIANT)
+      decl.Declaration.Invariant = 1;
+
+   return decl;
+}
+
+
+static struct tgsi_full_declaration
+make_temp_decl(
+   GLuint start_index,
+   GLuint end_index )
+{
+   struct tgsi_full_declaration decl;
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_TEMPORARY;
+   decl.DeclarationRange.First = start_index;
+   decl.DeclarationRange.Last = end_index;
+   return decl;
+}
+
+static struct tgsi_full_declaration
+make_addr_decl(
+   GLuint start_index,
+   GLuint end_index )
+{
+   struct tgsi_full_declaration decl;
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_ADDRESS;
+   decl.DeclarationRange.First = start_index;
+   decl.DeclarationRange.Last = end_index;
+   return decl;
+}
+
+static struct tgsi_full_declaration
+make_sampler_decl(GLuint index)
+{
+   struct tgsi_full_declaration decl;
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_SAMPLER;
+   decl.DeclarationRange.First = index;
+   decl.DeclarationRange.Last = index;
+   return decl;
+}
+
+/** Reference into a constant buffer */
+static struct tgsi_full_declaration
+make_constant_decl(GLuint first, GLuint last)
+{
+   struct tgsi_full_declaration decl;
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_CONSTANT;
+   decl.DeclarationRange.First = first;
+   decl.DeclarationRange.Last = last;
+   return decl;
+}
+
+
+
+/**
+ * Find the temporaries which are used in the given program.
+ */
+static void
+find_temporaries(const struct gl_program *program,
+                 GLboolean tempsUsed[MAX_PROGRAM_TEMPS])
+{
+   GLuint i, j;
+
+   for (i = 0; i < MAX_PROGRAM_TEMPS; i++)
+      tempsUsed[i] = GL_FALSE;
+
+   for (i = 0; i < program->NumInstructions; i++) {
+      const struct prog_instruction *inst = program->Instructions + i;
+      const GLuint n = _mesa_num_inst_src_regs( inst->Opcode );
+      for (j = 0; j < n; j++) {
+         if (inst->SrcReg[j].File == PROGRAM_TEMPORARY)
+            tempsUsed[inst->SrcReg[j].Index] = GL_TRUE;
+         if (inst->DstReg.File == PROGRAM_TEMPORARY)
+            tempsUsed[inst->DstReg.Index] = GL_TRUE;
+      }
+   }
+}
+
+
+
+
+/**
+ * Translate Mesa program to TGSI format.
+ * \param program  the program to translate
+ * \param numInputs  number of input registers used
+ * \param inputMapping  maps Mesa fragment program inputs to TGSI generic
+ *                      input indexes
+ * \param inputSemanticName  the TGSI_SEMANTIC flag for each input
+ * \param inputSemanticIndex  the semantic index (ex: which texcoord) for each input
+ * \param interpMode  the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input
+
+ * \param numOutputs  number of output registers used
+ * \param outputMapping  maps Mesa fragment program outputs to TGSI
+ *                       generic outputs
+ * \param outputSemanticName  the TGSI_SEMANTIC flag for each output
+ * \param outputSemanticIndex  the semantic index (ex: which texcoord) for each output
+ * \param tokens  array to store translated tokens in
+ * \param maxTokens  size of the tokens array
+ *
+ * \return number of tokens placed in 'tokens' buffer, or zero if error
+ */
+GLuint
+st_translate_mesa_program(
+   GLcontext *ctx,
+   uint procType,
+   const struct gl_program *program,
+   GLuint numInputs,
+   const GLuint inputMapping[],
+   const ubyte inputSemanticName[],
+   const ubyte inputSemanticIndex[],
+   const GLuint interpMode[],
+   const GLbitfield inputFlags[],
+   GLuint numOutputs,
+   const GLuint outputMapping[],
+   const ubyte outputSemanticName[],
+   const ubyte outputSemanticIndex[],
+   const GLbitfield outputFlags[],
+   struct tgsi_token *tokens,
+   GLuint maxTokens )
+{
+   GLuint i;
+   GLuint ti;  /* token index */
+   struct tgsi_header *header;
+   struct tgsi_processor *processor;
+   struct tgsi_full_instruction fullinst;
+   GLuint preamble_size = 0;
+   GLuint immediates[1000];
+   GLuint numImmediates = 0;
+   GLboolean insideSubroutine = GL_FALSE;
+   GLboolean indirectAccess = GL_FALSE;
+
+   assert(procType == TGSI_PROCESSOR_FRAGMENT ||
+          procType == TGSI_PROCESSOR_VERTEX);
+
+   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+
+   header = (struct tgsi_header *) &tokens[1];
+   *header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) &tokens[2];
+   *processor = tgsi_build_processor( procType, header );
+
+   ti = 3;
+
+   /*
+    * Declare input attributes.
+    */
+   if (procType == TGSI_PROCESSOR_FRAGMENT) {
+      for (i = 0; i < numInputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         fulldecl = make_input_decl(i,
+                                    GL_TRUE, interpMode[i],
+                                    TGSI_WRITEMASK_XYZW,
+                                    GL_TRUE, inputSemanticName[i],
+                                    inputSemanticIndex[i],
+                                    inputFlags[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+   else {
+      /* vertex prog */
+      /* XXX: this could probaby be merged with the clause above.
+       * the only difference is the semantic tags.
+       */
+      for (i = 0; i < numInputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         fulldecl = make_input_decl(i,
+                                    GL_FALSE, 0,
+                                    TGSI_WRITEMASK_XYZW,
+                                    GL_FALSE, 0, 0,
+                                    inputFlags[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+
+   /*
+    * Declare output attributes.
+    */
+   if (procType == TGSI_PROCESSOR_FRAGMENT) {
+      for (i = 0; i < numOutputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         switch (outputSemanticName[i]) {
+         case TGSI_SEMANTIC_POSITION:
+            fulldecl = make_output_decl(i,
+                                        TGSI_SEMANTIC_POSITION, /* Z / Depth */
+                                        outputSemanticIndex[i],
+                                        TGSI_WRITEMASK_Z,
+                                        outputFlags[i]);
+            break;
+         case TGSI_SEMANTIC_COLOR:
+            fulldecl = make_output_decl(i,
+                                        TGSI_SEMANTIC_COLOR,
+                                        outputSemanticIndex[i],
+                                        TGSI_WRITEMASK_XYZW,
+                                        outputFlags[i]);
+            break;
+         default:
+            assert(0);
+            return 0;
+         }
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+   else {
+      /* vertex prog */
+      for (i = 0; i < numOutputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         fulldecl = make_output_decl(i,
+                                     outputSemanticName[i],
+                                     outputSemanticIndex[i],
+                                     TGSI_WRITEMASK_XYZW,
+                                     outputFlags[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+
+   /* temporary decls */
+   {
+      GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1];
+      GLboolean inside_range = GL_FALSE;
+      GLuint start_range = 0;
+
+      find_temporaries(program, tempsUsed);
+      tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE;
+      for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) {
+         if (tempsUsed[i] && !inside_range) {
+            inside_range = GL_TRUE;
+            start_range = i;
+         }
+         else if (!tempsUsed[i] && inside_range) {
+            struct tgsi_full_declaration fulldecl;
+
+            inside_range = GL_FALSE;
+            fulldecl = make_temp_decl( start_range, i - 1 );
+            ti += tgsi_build_full_declaration(
+               &fulldecl,
+               &tokens[ti],
+               header,
+               maxTokens - ti );
+         }
+      }
+   }
+
+   /* Declare address register.
+   */
+   if (program->NumAddressRegs > 0) {
+      struct tgsi_full_declaration fulldecl;
+
+      assert( program->NumAddressRegs == 1 );
+
+      fulldecl = make_addr_decl( 0, 0 );
+      ti += tgsi_build_full_declaration(
+         &fulldecl,
+         &tokens[ti],
+         header,
+         maxTokens - ti );
+
+      indirectAccess = GL_TRUE;
+   }
+
+   /* immediates/literals */
+   memset(immediates, ~0, sizeof(immediates));
+
+   /* Emit immediates only when there is no address register in use.
+    * FIXME: Be smarter and recognize param arrays -- indirect addressing is
+    *        only valid within the referenced array.
+    */
+   if (program->Parameters && !indirectAccess) {
+      for (i = 0; i < program->Parameters->NumParameters; i++) {
+         if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) {
+            struct tgsi_full_immediate fullimm;
+
+            fullimm = make_immediate( program->Parameters->ParameterValues[i], 4 );
+            ti += tgsi_build_full_immediate(
+               &fullimm,
+               &tokens[ti],
+               header,
+               maxTokens - ti );
+            immediates[i] = numImmediates;
+            numImmediates++;
+         }
+      }
+   }
+
+   /* constant buffer refs */
+   if (program->Parameters) {
+      GLint start = -1, end = -1;
+
+      for (i = 0; i < program->Parameters->NumParameters; i++) {
+         GLboolean emit = (i == program->Parameters->NumParameters - 1);
+         GLboolean matches;
+
+         switch (program->Parameters->Parameters[i].Type) {
+         case PROGRAM_ENV_PARAM:
+         case PROGRAM_STATE_VAR:
+         case PROGRAM_NAMED_PARAM:
+         case PROGRAM_UNIFORM:
+            matches = GL_TRUE;
+            break;
+         case PROGRAM_CONSTANT:
+            matches = indirectAccess;
+            break;
+         default:
+            matches = GL_FALSE;
+         }
+
+         if (matches) {
+            if (start == -1) {
+               /* begin a sequence */
+               start = i;
+               end = i;
+            }
+            else {
+               /* continue sequence */
+               end = i;
+            }
+         }
+         else {
+            if (start != -1) {
+               /* end of sequence */
+               emit = GL_TRUE;
+            }
+         }
+
+         if (emit && start >= 0) {
+            struct tgsi_full_declaration fulldecl;
+
+            fulldecl = make_constant_decl( start, end );
+            ti += tgsi_build_full_declaration(
+               &fulldecl,
+               &tokens[ti],
+               header,
+               maxTokens - ti );
+            start = end = -1;
+         }
+      }
+   }
+
+   /* texture samplers */
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
+      if (program->SamplersUsed & (1 << i)) {
+         struct tgsi_full_declaration fulldecl;
+
+         fulldecl = make_sampler_decl( i );
+         ti += tgsi_build_full_declaration(
+            &fulldecl,
+            &tokens[ti],
+            header,
+            maxTokens - ti );
+      }
+   }
+
+   for (i = 0; i < program->NumInstructions; i++) {
+      compile_instruction(
+         &program->Instructions[i],
+         &fullinst,
+         inputMapping,
+         outputMapping,
+         immediates,
+         indirectAccess,
+         preamble_size,
+         procType,
+         &insideSubroutine );
+
+      ti += tgsi_build_full_instruction(
+         &fullinst,
+         &tokens[ti],
+         header,
+         maxTokens - ti );
+   }
+
+#if DEBUG
+   if(!tgsi_sanity_check(tokens)) {
+      debug_printf("Due to sanity check failure(s) above the following shader program is invalid:\n");
+      debug_printf("\nOriginal program:\n%s", program->String);
+      debug_printf("\nMesa program:\n");
+      _mesa_print_program(program);
+      debug_printf("\nTGSI program:\n");
+      tgsi_dump(tokens, 0);
+      assert(0);
+   }
+#endif
+
+   return ti;
+}
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h
new file mode 100644
index 00000000000..b465b3bddcf
--- /dev/null
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h
@@ -0,0 +1,67 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_MESA_TO_TGSI_H
+#define ST_MESA_TO_TGSI_H
+
+#include "main/mtypes.h"
+
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_token;
+struct gl_program;
+
+GLuint
+st_translate_mesa_program(
+   GLcontext *ctx,
+   uint procType,
+   const struct gl_program *program,
+   GLuint numInputs,
+   const GLuint inputMapping[],
+   const ubyte inputSemanticName[],
+   const ubyte inputSemanticIndex[],
+   const GLuint interpMode[],
+   const GLbitfield inputFlags[],
+   GLuint numOutputs,
+   const GLuint outputMapping[],
+   const ubyte outputSemanticName[],
+   const ubyte outputSemanticIndex[],
+   const GLbitfield outputFlags[],
+   struct tgsi_token *tokens,
+   GLuint maxTokens );
+
+
+#if defined __cplusplus
+} /* extern "C" */
+#endif
+
+#endif /* ST_MESA_TO_TGSI_H */
+
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
new file mode 100644
index 00000000000..442eeed1470
--- /dev/null
+++ b/src/mesa/state_tracker/st_program.c
@@ -0,0 +1,556 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  *   Brian Paul
+  */
+
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "shader/prog_print.h"
+#include "shader/programopt.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_program.h"
+#include "st_mesa_to_tgsi.h"
+#include "cso_cache/cso_context.h"
+
+
+#define ST_MAX_SHADER_TOKENS 4096
+
+
+#define TGSI_DEBUG 0
+
+
+/** XXX we should use the version of this from u_memory.h but including
+ * that header causes symbol collisions.
+ */
+static INLINE void *
+mem_dup(const void *src, uint size)
+{
+   void *dup = MALLOC(size);
+   if (dup)
+      memcpy(dup, src, size);
+   return dup;
+}
+
+
+
+/**
+ * Translate a Mesa vertex shader into a TGSI shader.
+ * \param outputMapping  to map vertex program output registers (VERT_RESULT_x)
+ *       to TGSI output slots
+ * \param tokensOut  destination for TGSI tokens
+ * \return  pointer to cached pipe_shader object.
+ */
+void
+st_translate_vertex_program(struct st_context *st,
+                            struct st_vertex_program *stvp,
+                            const GLuint outputMapping[],
+                            const ubyte *outputSemanticName,
+                            const ubyte *outputSemanticIndex)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   GLuint defaultOutputMapping[VERT_RESULT_MAX];
+   struct pipe_shader_state vs;
+   GLuint attr, i;
+   GLuint num_generic = 0;
+   GLuint num_tokens;
+
+   ubyte vs_input_semantic_name[PIPE_MAX_SHADER_INPUTS];
+   ubyte vs_input_semantic_index[PIPE_MAX_SHADER_INPUTS];
+   uint vs_num_inputs = 0;
+
+   ubyte vs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS];
+   ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
+   uint vs_num_outputs = 0;
+
+   GLbitfield input_flags[MAX_PROGRAM_INPUTS];
+   GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
+
+   memset(&vs, 0, sizeof(vs));
+   memset(input_flags, 0, sizeof(input_flags));
+   memset(output_flags, 0, sizeof(output_flags));
+
+   if (stvp->Base.IsPositionInvariant)
+      _mesa_insert_mvp_code(st->ctx, &stvp->Base);
+
+   /*
+    * Determine number of inputs, the mappings between VERT_ATTRIB_x
+    * and TGSI generic input indexes, plus input attrib semantic info.
+    */
+   for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
+      if (stvp->Base.Base.InputsRead & (1 << attr)) {
+         const GLuint slot = vs_num_inputs;
+
+         vs_num_inputs++;
+
+         stvp->input_to_index[attr] = slot;
+         stvp->index_to_input[slot] = attr;
+
+         switch (attr) {
+         case VERT_ATTRIB_POS:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_WEIGHT:
+            /* fall-through */
+         case VERT_ATTRIB_NORMAL:
+            /* just label as a generic */
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_COLOR0:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_COLOR1:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_input_semantic_index[slot] = 1;
+            break;
+         case VERT_ATTRIB_FOG:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_POINT_SIZE:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_PSIZE;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_TEX0:
+         case VERT_ATTRIB_TEX1:
+         case VERT_ATTRIB_TEX2:
+         case VERT_ATTRIB_TEX3:
+         case VERT_ATTRIB_TEX4:
+         case VERT_ATTRIB_TEX5:
+         case VERT_ATTRIB_TEX6:
+         case VERT_ATTRIB_TEX7:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            vs_input_semantic_index[slot] = num_generic++;
+            break;
+         case VERT_ATTRIB_GENERIC0:
+         case VERT_ATTRIB_GENERIC1:
+         case VERT_ATTRIB_GENERIC2:
+         case VERT_ATTRIB_GENERIC3:
+         case VERT_ATTRIB_GENERIC4:
+         case VERT_ATTRIB_GENERIC5:
+         case VERT_ATTRIB_GENERIC6:
+         case VERT_ATTRIB_GENERIC7:
+            assert(attr < VERT_ATTRIB_MAX);
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            vs_input_semantic_index[slot] = num_generic++;
+            break;
+         default:
+            assert(0);
+         }
+
+         input_flags[slot] = stvp->Base.Base.InputFlags[attr];
+      }
+   }
+
+#if 0
+   if (outputMapping && outputSemanticName) {
+      printf("VERT_RESULT  written  out_slot  semantic_name  semantic_index\n");
+      for (attr = 0; attr < VERT_RESULT_MAX; attr++) {
+         printf("    %-2d          %c       %3d          %2d              %2d\n",
+                attr, 
+                ((stvp->Base.Base.OutputsWritten & (1 << attr)) ? 'Y' : ' '),
+                outputMapping[attr],
+                outputSemanticName[attr],
+                outputSemanticIndex[attr]);
+      }
+   }
+#endif
+
+   /* initialize output semantics to defaults */
+   for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) {
+      vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC;
+      vs_output_semantic_index[i] = 0;
+      output_flags[i] = 0x0;
+   }
+
+   num_generic = 0;
+   /*
+    * Determine number of outputs, the (default) output register
+    * mapping and the semantic information for each output.
+    */
+   for (attr = 0; attr < VERT_RESULT_MAX; attr++) {
+      if (stvp->Base.Base.OutputsWritten & (1 << attr)) {
+         GLuint slot;
+
+         /* XXX
+          * Pass in the fragment program's input's semantic info.
+          * Use the generic semantic indexes from there, instead of
+          * guessing below.
+          */
+
+         if (outputMapping) {
+            slot = outputMapping[attr];
+            assert(slot != ~0);
+         }
+         else {
+            slot = vs_num_outputs;
+            vs_num_outputs++;
+            defaultOutputMapping[attr] = slot;
+         }
+
+         switch (attr) {
+         case VERT_RESULT_HPOS:
+            assert(slot == 0);
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_COL0:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_COL1:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_output_semantic_index[slot] = 1;
+            break;
+         case VERT_RESULT_BFC0:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_BFC1:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR;
+            vs_output_semantic_index[slot] = 1;
+            break;
+         case VERT_RESULT_FOGC:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_PSIZ:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_EDGE:
+            assert(0);
+            break;
+         case VERT_RESULT_TEX0:
+         case VERT_RESULT_TEX1:
+         case VERT_RESULT_TEX2:
+         case VERT_RESULT_TEX3:
+         case VERT_RESULT_TEX4:
+         case VERT_RESULT_TEX5:
+         case VERT_RESULT_TEX6:
+         case VERT_RESULT_TEX7:
+            /* fall-through */
+         case VERT_RESULT_VAR0:
+            /* fall-through */
+         default:
+            if (outputSemanticName) {
+               /* use provided semantic into */
+               assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT);
+               vs_output_semantic_name[slot] = outputSemanticName[attr];
+               vs_output_semantic_index[slot] = outputSemanticIndex[attr];
+            }
+            else {
+               /* use default semantic info */
+               vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+               vs_output_semantic_index[slot] = num_generic++;
+            }
+         }
+
+         output_flags[slot] = stvp->Base.Base.OutputFlags[attr];
+      }
+   }
+
+   assert(vs_output_semantic_name[0] == TGSI_SEMANTIC_POSITION);
+
+
+   if (outputMapping) {
+      /* find max output slot referenced to compute vs_num_outputs */
+      GLuint maxSlot = 0;
+      for (attr = 0; attr < VERT_RESULT_MAX; attr++) {
+         if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot)
+            maxSlot = outputMapping[attr];
+      }
+      vs_num_outputs = maxSlot + 1;
+   }
+   else {
+      outputMapping = defaultOutputMapping;
+   }
+
+   /* free old shader state, if any */
+   if (stvp->state.tokens) {
+      FREE((void *) stvp->state.tokens);
+      stvp->state.tokens = NULL;
+   }
+   if (stvp->driver_shader) {
+      cso_delete_vertex_shader(st->cso_context, stvp->driver_shader);
+      stvp->driver_shader = NULL;
+   }
+
+   /* XXX: fix static allocation of tokens:
+    */
+   num_tokens = st_translate_mesa_program(st->ctx,
+                                          TGSI_PROCESSOR_VERTEX,
+                                          &stvp->Base.Base,
+                                          /* inputs */
+                                          vs_num_inputs,
+                                          stvp->input_to_index,
+                                          vs_input_semantic_name,
+                                          vs_input_semantic_index,
+                                          NULL,
+                                          input_flags,
+                                          /* outputs */
+                                          vs_num_outputs,
+                                          outputMapping,
+                                          vs_output_semantic_name,
+                                          vs_output_semantic_index,
+                                          output_flags,
+                                          /* tokenized result */
+                                          tokens, ST_MAX_SHADER_TOKENS);
+
+   assert(num_tokens < ST_MAX_SHADER_TOKENS);
+
+   vs.tokens = (struct tgsi_token *)
+      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+
+   stvp->num_inputs = vs_num_inputs;
+   stvp->state = vs; /* struct copy */
+   stvp->driver_shader = pipe->create_vs_state(pipe, &vs);
+
+   if (0)
+      _mesa_print_program(&stvp->Base.Base);
+
+   if (TGSI_DEBUG)
+      tgsi_dump( vs.tokens, 0 );
+}
+
+
+
+/**
+ * Translate a Mesa fragment shader into a TGSI shader.
+ * \param inputMapping  to map fragment program input registers to TGSI
+ *                      input slots
+ * \param tokensOut  destination for TGSI tokens
+ * \return  pointer to cached pipe_shader object.
+ */
+void
+st_translate_fragment_program(struct st_context *st,
+                              struct st_fragment_program *stfp,
+                              const GLuint inputMapping[])
+{
+   struct pipe_context *pipe = st->pipe;
+   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   GLuint outputMapping[FRAG_RESULT_MAX];
+   GLuint defaultInputMapping[FRAG_ATTRIB_MAX];
+   struct pipe_shader_state fs;
+   GLuint interpMode[16];  /* XXX size? */
+   GLuint attr;
+   const GLbitfield inputsRead = stfp->Base.Base.InputsRead;
+   GLuint vslot = 0;
+   GLuint num_generic = 0;
+   GLuint num_tokens;
+
+   uint fs_num_inputs = 0;
+
+   ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS];
+   ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
+   uint fs_num_outputs = 0;
+
+   GLbitfield input_flags[MAX_PROGRAM_INPUTS];
+   GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
+
+   memset(&fs, 0, sizeof(fs));
+   memset(input_flags, 0, sizeof(input_flags));
+   memset(output_flags, 0, sizeof(output_flags));
+
+   /* which vertex output goes to the first fragment input: */
+   if (inputsRead & FRAG_BIT_WPOS)
+      vslot = 0;
+   else
+      vslot = 1;
+
+   /*
+    * Convert Mesa program inputs to TGSI input register semantics.
+    */
+   for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) {
+      if (inputsRead & (1 << attr)) {
+         const GLuint slot = fs_num_inputs;
+
+         defaultInputMapping[attr] = slot;
+
+         stfp->input_map[slot] = vslot++;
+
+         fs_num_inputs++;
+
+         switch (attr) {
+         case FRAG_ATTRIB_WPOS:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+            break;
+         case FRAG_ATTRIB_COL0:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+            break;
+         case FRAG_ATTRIB_COL1:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            stfp->input_semantic_index[slot] = 1;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+            break;
+         case FRAG_ATTRIB_FOGC:
+            if (stfp->Base.UsesPointCoord)
+               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            else
+               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+            break;
+         case FRAG_ATTRIB_TEX0:
+         case FRAG_ATTRIB_TEX1:
+         case FRAG_ATTRIB_TEX2:
+         case FRAG_ATTRIB_TEX3:
+         case FRAG_ATTRIB_TEX4:
+         case FRAG_ATTRIB_TEX5:
+         case FRAG_ATTRIB_TEX6:
+         case FRAG_ATTRIB_TEX7:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            stfp->input_semantic_index[slot] = num_generic++;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+            break;
+         case FRAG_ATTRIB_VAR0:
+            /* fall-through */
+         default:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            stfp->input_semantic_index[slot] = num_generic++;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+         }
+
+         input_flags[slot] = stfp->Base.Base.InputFlags[attr];
+      }
+   }
+
+   /*
+    * Semantics and mapping for outputs
+    */
+   {
+      uint numColors = 0;
+      GLbitfield outputsWritten = stfp->Base.Base.OutputsWritten;
+
+      /* if z is written, emit that first */
+      if (outputsWritten & (1 << FRAG_RESULT_DEPR)) {
+         fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION;
+         fs_output_semantic_index[fs_num_outputs] = 0;
+         outputMapping[FRAG_RESULT_DEPR] = fs_num_outputs;
+         fs_num_outputs++;
+         outputsWritten &= ~(1 << FRAG_RESULT_DEPR);
+      }
+
+      /* handle remaning outputs (color) */
+      for (attr = 0; attr < FRAG_RESULT_MAX; attr++) {
+         if (outputsWritten & (1 << attr)) {
+            switch (attr) {
+            case FRAG_RESULT_DEPR:
+               /* handled above */
+               assert(0);
+               break;
+            case FRAG_RESULT_COLR:
+               fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR;
+               fs_output_semantic_index[fs_num_outputs] = numColors;
+               outputMapping[attr] = fs_num_outputs;
+               numColors++;
+               break;
+            default:
+               assert(0);
+            }
+
+            output_flags[fs_num_outputs] = stfp->Base.Base.OutputFlags[attr];
+
+            fs_num_outputs++;
+         }
+      }
+   }
+
+   if (!inputMapping)
+      inputMapping = defaultInputMapping;
+
+   /* XXX: fix static allocation of tokens:
+    */
+   num_tokens = st_translate_mesa_program(st->ctx,
+                                          TGSI_PROCESSOR_FRAGMENT,
+                                          &stfp->Base.Base,
+                                          /* inputs */
+                                          fs_num_inputs,
+                                          inputMapping,
+                                          stfp->input_semantic_name,
+                                          stfp->input_semantic_index,
+                                          interpMode,
+                                          input_flags,
+                                          /* outputs */
+                                          fs_num_outputs,
+                                          outputMapping,
+                                          fs_output_semantic_name,
+                                          fs_output_semantic_index,
+                                          output_flags,
+                                          /* tokenized result */
+                                          tokens, ST_MAX_SHADER_TOKENS);
+
+   assert(num_tokens < ST_MAX_SHADER_TOKENS);
+
+   fs.tokens = (struct tgsi_token *)
+      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+
+   stfp->state = fs; /* struct copy */
+   stfp->driver_shader = pipe->create_fs_state(pipe, &fs);
+
+   if (0)
+      _mesa_print_program(&stfp->Base.Base);
+
+   if (TGSI_DEBUG)
+      tgsi_dump( fs.tokens, 0/*TGSI_DUMP_VERBOSE*/ );
+}
+
+
+/**
+ * Debug- print current shader text
+ */
+void
+st_print_shaders(GLcontext *ctx)
+{
+   struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+   if (shProg) {
+      GLuint i;
+      for (i = 0; i < shProg->NumShaders; i++) {
+         printf("GLSL shader %u of %u:\n", i, shProg->NumShaders);
+         printf("%s\n", shProg->Shaders[i]->Source);
+      }
+   }
+}
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
new file mode 100644
index 00000000000..e2e5eddef22
--- /dev/null
+++ b/src/mesa/state_tracker/st_program.h
@@ -0,0 +1,158 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+    
+
+#ifndef ST_PROGRAM_H
+#define ST_PROGRAM_H
+
+#include "main/mtypes.h"
+#include "shader/program.h"
+#include "pipe/p_shader_tokens.h"
+
+
+struct cso_fragment_shader;
+struct cso_vertex_shader;
+struct translated_vertex_program;
+
+
+/**
+ * Derived from Mesa gl_fragment_program:
+ */
+struct st_fragment_program
+{
+   struct gl_fragment_program Base;
+   GLuint serialNo;
+
+   GLuint input_to_slot[FRAG_ATTRIB_MAX];  /**< Maps FRAG_ATTRIB_x to slot */
+   GLuint num_input_slots;
+
+   /** map FP input back to VP output */
+   GLuint input_map[PIPE_MAX_SHADER_INPUTS];
+
+   ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS];
+   ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
+
+   struct pipe_shader_state state;
+   void *driver_shader;
+
+   GLuint param_state;
+
+   /** List of vertex programs which have been translated such that their
+    * outputs match this fragment program's inputs.
+    */
+   struct translated_vertex_program *vertex_programs;
+
+   /** Program prefixed with glBitmap prologue */
+   struct st_fragment_program *bitmap_program;
+   uint bitmap_sampler;
+};
+
+
+/**
+ * Derived from Mesa gl_fragment_program:
+ */
+struct st_vertex_program
+{
+   struct gl_vertex_program Base;  /**< The Mesa vertex program */
+   GLuint serialNo;
+
+   /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */
+   GLuint input_to_index[VERT_ATTRIB_MAX];
+   /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */
+   GLuint index_to_input[PIPE_MAX_SHADER_INPUTS];
+
+   GLuint num_inputs;
+
+   struct pipe_shader_state state;
+   void *driver_shader;
+
+   /** For using our private draw module (glRasterPos) */
+   struct draw_vertex_shader *draw_shader;
+
+   GLuint param_state;
+};
+
+
+static INLINE struct st_fragment_program *
+st_fragment_program( struct gl_fragment_program *fp )
+{
+   return (struct st_fragment_program *)fp;
+}
+
+
+static INLINE struct st_vertex_program *
+st_vertex_program( struct gl_vertex_program *vp )
+{
+   return (struct st_vertex_program *)vp;
+}
+
+
+static INLINE void
+st_reference_vertprog(struct st_context *st,
+                      struct st_vertex_program **ptr,
+                      struct st_vertex_program *prog)
+{
+   _mesa_reference_program(st->ctx,
+                           (struct gl_program **) ptr,
+                           (struct gl_program *) prog);
+}
+
+static INLINE void
+st_reference_fragprog(struct st_context *st,
+                      struct st_fragment_program **ptr,
+                      struct st_fragment_program *prog)
+{
+   _mesa_reference_program(st->ctx,
+                           (struct gl_program **) ptr,
+                           (struct gl_program *) prog);
+}
+
+
+extern void
+st_translate_fragment_program(struct st_context *st,
+                              struct st_fragment_program *fp,
+                              const GLuint inputMapping[]);
+
+
+extern void
+st_translate_vertex_program(struct st_context *st,
+                            struct st_vertex_program *vp,
+                            const GLuint vert_output_to_slot[],
+                            const ubyte *fs_input_semantic_name,
+                            const ubyte *fs_input_semantic_index);
+
+
+extern void
+st_print_shaders(GLcontext *ctx);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h
new file mode 100644
index 00000000000..414218bb589
--- /dev/null
+++ b/src/mesa/state_tracker/st_public.h
@@ -0,0 +1,122 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_PUBLIC_H
+#define ST_PUBLIC_H
+
+#include "GL/gl.h"
+#include "GL/internal/glcore.h"  /* for __GLcontextModes */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+
+#define ST_SURFACE_FRONT_LEFT   0
+#define ST_SURFACE_BACK_LEFT    1
+#define ST_SURFACE_FRONT_RIGHT  2
+#define ST_SURFACE_BACK_RIGHT   3
+#define ST_SURFACE_DEPTH        8
+
+#define ST_TEXTURE_2D    0x2
+#define ST_TEXTURE_RECT  0x4
+
+#define ST_TEXTURE_RGB   0x1
+#define ST_TEXTURE_RGBA  0x2
+
+
+struct st_context;
+struct st_framebuffer;
+struct pipe_context;
+struct pipe_fence_handle;
+struct pipe_surface;
+struct pipe_texture;
+
+
+struct st_context *st_create_context(struct pipe_context *pipe,
+                                     const __GLcontextModes *visual,
+                                     struct st_context *share);
+
+void st_destroy_context( struct st_context *st );
+
+void st_copy_context_state(struct st_context *dst, struct st_context *src,
+                           uint mask);
+
+struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual,
+                                              enum pipe_format colorFormat,
+                                              enum pipe_format depthFormat,
+                                              enum pipe_format stencilFormat,
+                                              uint width, uint height,
+                                              void *privateData);
+
+void st_resize_framebuffer( struct st_framebuffer *stfb,
+                            uint width, uint height );
+
+void st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                                uint surfIndex, struct pipe_surface *surf);
+
+void st_get_framebuffer_dimensions( struct st_framebuffer *stfb,
+				    uint *width, uint *height);
+
+int st_get_framebuffer_surface(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_surface **surface);
+
+int st_get_framebuffer_texture(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_texture **texture);
+
+void *st_framebuffer_private( struct st_framebuffer *stfb );
+
+void st_unreference_framebuffer( struct st_framebuffer *stfb );
+
+void st_make_current(struct st_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read);
+
+void st_flush( struct st_context *st, uint pipeFlushFlags,
+               struct pipe_fence_handle **fence );
+void st_finish( struct st_context *st );
+
+void st_notify_swapbuffers(struct st_framebuffer *stfb);
+void st_notify_swapbuffers_complete(struct st_framebuffer *stfb);
+
+int st_set_teximage(struct pipe_texture *pt, int target);
+
+/** Redirect rendering into stfb's surface to a texture image */
+int st_bind_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                     int target, int format, int level);
+
+/** Undo surface-to-texture binding */
+int st_release_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                        int target, int format, int level);
+
+
+/** Generic function type */
+typedef void (*st_proc)();
+
+st_proc st_get_proc_address(const char *procname);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
new file mode 100644
index 00000000000..63cfe5fc163
--- /dev/null
+++ b/src/mesa/state_tracker/st_texture.c
@@ -0,0 +1,488 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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 "st_context.h"
+#include "st_format.h"
+#include "st_public.h"
+#include "st_texture.h"
+#include "st_cb_fbo.h"
+#include "main/enums.h"
+#include "main/teximage.h"
+#include "main/texstore.h"
+
+#undef Elements  /* fix re-defined macro warning */
+
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_rect.h"
+
+
+#define DBG if(0) printf
+
+#if 0
+static GLenum
+target_to_target(GLenum target)
+{
+   switch (target) {
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB:
+      return GL_TEXTURE_CUBE_MAP_ARB;
+   default:
+      return target;
+   }
+}
+#endif
+
+
+/**
+ * Allocate a new pipe_texture object
+ * width0, height0, depth0 are the dimensions of the level 0 image
+ * (the highest resolution).  last_level indicates how many mipmap levels
+ * to allocate storage for.  For non-mipmapped textures, this will be zero.
+ */
+struct pipe_texture *
+st_texture_create(struct st_context *st,
+                  enum pipe_texture_target target,
+		  enum pipe_format format,
+		  GLuint last_level,
+		  GLuint width0,
+		  GLuint height0,
+		  GLuint depth0,
+		  GLuint compress_byte,
+                  GLuint usage )
+{
+   struct pipe_texture pt, *newtex;
+   struct pipe_screen *screen = st->pipe->screen;
+
+   assert(target <= PIPE_TEXTURE_CUBE);
+
+   DBG("%s target %s format %s last_level %d\n", __FUNCTION__,
+       _mesa_lookup_enum_by_nr(target),
+       _mesa_lookup_enum_by_nr(format), last_level);
+
+   assert(format);
+   assert(screen->is_format_supported(screen, format, target, 
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   memset(&pt, 0, sizeof(pt));
+   pt.target = target;
+   pt.format = format;
+   pt.last_level = last_level;
+   pt.width[0] = width0;
+   pt.height[0] = height0;
+   pt.depth[0] = depth0;
+   pt.compressed = compress_byte ? 1 : 0;
+   pf_get_block(format, &pt.block);
+   pt.tex_usage = usage;
+
+   newtex = screen->texture_create(screen, &pt);
+
+   assert(!newtex || newtex->refcount == 1);
+
+   return newtex;
+}
+
+
+/**
+ * Check if a texture image be pulled into a unified mipmap texture.
+ * This mirrors the completeness test in a lot of ways.
+ *
+ * Not sure whether I want to pass gl_texture_image here.
+ */
+GLboolean
+st_texture_match_image(const struct pipe_texture *pt,
+                       const struct gl_texture_image *image,
+                       GLuint face, GLuint level)
+{
+   /* Images with borders are never pulled into mipmap textures. 
+    */
+   if (image->Border) 
+      return GL_FALSE;
+
+   if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format ||
+       image->IsCompressed != pt->compressed)
+      return GL_FALSE;
+
+   /* Test image dimensions against the base level image adjusted for
+    * minification.  This will also catch images not present in the
+    * texture, changed targets, etc.
+    */
+   if (image->Width != pt->width[level] ||
+       image->Height != pt->height[level] ||
+       image->Depth != pt->depth[level])
+      return GL_FALSE;
+
+   return GL_TRUE;
+}
+
+
+#if 000
+/* Although we use the image_offset[] array to store relative offsets
+ * to cube faces, Mesa doesn't know anything about this and expects
+ * each cube face to be treated as a separate image.
+ *
+ * These functions present that view to mesa:
+ */
+const GLuint *
+st_texture_depth_offsets(struct pipe_texture *pt, GLuint level)
+{
+   static const GLuint zero = 0;
+
+   if (pt->target != PIPE_TEXTURE_3D || pt->level[level].nr_images == 1)
+      return &zero;
+   else
+      return pt->level[level].image_offset;
+}
+
+
+/**
+ * Return the offset to the given mipmap texture image within the
+ * texture memory buffer, in bytes.
+ */
+GLuint
+st_texture_image_offset(const struct pipe_texture * pt,
+                        GLuint face, GLuint level)
+{
+   if (pt->target == PIPE_TEXTURE_CUBE)
+      return (pt->level[level].level_offset +
+              pt->level[level].image_offset[face] * pt->cpp);
+   else
+      return pt->level[level].level_offset;
+}
+#endif
+
+
+/**
+ * Map a teximage in a mipmap texture.
+ * \param row_stride  returns row stride in bytes
+ * \param image_stride  returns image stride in bytes (for 3D textures).
+ * \return address of mapping
+ */
+GLubyte *
+st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
+		     GLuint zoffset,
+                     GLuint flags )
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_texture *pt = stImage->pt;
+   DBG("%s \n", __FUNCTION__);
+
+   stImage->surface = screen->get_tex_surface(screen, pt, stImage->face,
+                                              stImage->level, zoffset, 
+                                              flags);
+
+   if (stImage->surface)
+      return screen->surface_map(screen, stImage->surface, flags);
+   else
+      return NULL;
+}
+
+
+void
+st_texture_image_unmap(struct st_context *st,
+                       struct st_texture_image *stImage)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+
+   DBG("%s\n", __FUNCTION__);
+
+   screen->surface_unmap(screen, stImage->surface);
+
+   pipe_surface_reference(&stImage->surface, NULL);
+}
+
+
+
+/**
+ * Upload data to a rectangular sub-region.  Lots of choices how to do this:
+ *
+ * - memcpy by span to current destination
+ * - upload data as new buffer and blit
+ *
+ * Currently always memcpy.
+ */
+static void
+st_surface_data(struct pipe_context *pipe,
+		struct pipe_surface *dst,
+		unsigned dstx, unsigned dsty,
+		const void *src, unsigned src_stride,
+		unsigned srcx, unsigned srcy, unsigned width, unsigned height)
+{
+   struct pipe_screen *screen = pipe->screen;
+   void *map = screen->surface_map(screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   pipe_copy_rect(map,
+                  &dst->block,
+                  dst->stride,
+                  dstx, dsty, 
+                  width, height, 
+                  src, src_stride, 
+                  srcx, srcy);
+
+   screen->surface_unmap(screen, dst);
+}
+
+
+/* Upload data for a particular image.
+ */
+void
+st_texture_image_data(struct pipe_context *pipe,
+                      struct pipe_texture *dst,
+                      GLuint face,
+                      GLuint level,
+                      void *src,
+                      GLuint src_row_stride, GLuint src_image_stride)
+{
+   struct pipe_screen *screen = pipe->screen;
+   GLuint depth = dst->depth[level];
+   GLuint i;
+   const GLubyte *srcUB = src;
+   struct pipe_surface *dst_surface;
+
+   DBG("%s\n", __FUNCTION__);
+   for (i = 0; i < depth; i++) {
+      dst_surface = screen->get_tex_surface(screen, dst, face, level, i,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      st_surface_data(pipe, dst_surface,
+		      0, 0,                             /* dstx, dsty */
+		      srcUB,
+		      src_row_stride,
+		      0, 0,                             /* source x, y */
+		      dst->width[level], dst->height[level]);       /* width, height */
+
+      screen->tex_surface_release(screen, &dst_surface);
+
+      srcUB += src_image_stride;
+   }
+}
+
+
+/* Copy mipmap image between textures
+ */
+void
+st_texture_image_copy(struct pipe_context *pipe,
+                      struct pipe_texture *dst, GLuint dstLevel,
+                      struct pipe_texture *src,
+                      GLuint face)
+{
+   struct pipe_screen *screen = pipe->screen;
+   GLuint width = dst->width[dstLevel];
+   GLuint height = dst->height[dstLevel];
+   GLuint depth = dst->depth[dstLevel];
+   struct pipe_surface *src_surface;
+   struct pipe_surface *dst_surface;
+   GLuint i;
+
+   for (i = 0; i < depth; i++) {
+      GLuint srcLevel;
+
+      /* find src texture level of needed size */
+      for (srcLevel = 0; srcLevel <= src->last_level; srcLevel++) {
+         if (src->width[srcLevel] == width &&
+             src->height[srcLevel] == height) {
+            break;
+         }
+      }
+      assert(src->width[srcLevel] == width);
+      assert(src->height[srcLevel] == height);
+
+#if 0
+      {
+         src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i,
+                                               PIPE_BUFFER_USAGE_CPU_READ);
+         ubyte *map = screen->surface_map(screen, src_surface, PIPE_BUFFER_USAGE_CPU_READ);
+         map += src_surface->width * src_surface->height * 4 / 2;
+         printf("%s center pixel: %d %d %d %d (pt %p[%d] -> %p[%d])\n",
+                __FUNCTION__,
+                map[0], map[1], map[2], map[3],
+                src, srcLevel, dst, dstLevel);
+
+         screen->surface_unmap(screen, src_surface);
+         pipe_surface_reference(&src_surface, NULL);
+      }
+#endif
+
+      dst_surface = screen->get_tex_surface(screen, dst, face, dstLevel, i,
+                                            PIPE_BUFFER_USAGE_GPU_WRITE);
+
+      src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i,
+                                            PIPE_BUFFER_USAGE_GPU_READ);
+
+      pipe->surface_copy(pipe,
+                         FALSE,
+			 dst_surface,
+			 0, 0, /* destX, Y */
+			 src_surface,
+			 0, 0, /* srcX, Y */
+			 width, height);
+
+      screen->tex_surface_release(screen, &src_surface);
+      screen->tex_surface_release(screen, &dst_surface);
+   }
+}
+
+/** Bind a pipe surface for use as a texture image */
+int
+st_set_teximage(struct pipe_texture *pt, int target)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint unit = ctx->Texture.CurrentUnit;
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   struct st_texture_image *stImage;
+   int internalFormat;
+
+   switch (pt->format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      internalFormat = GL_RGBA8;
+      break;
+   default:
+      return 0;
+   };
+
+   switch (target) {
+   case ST_TEXTURE_2D:
+      target = GL_TEXTURE_2D;
+      break;
+   case ST_TEXTURE_RECT:
+      target = GL_TEXTURE_RECTANGLE_ARB;
+      break;
+   default:
+      return 0;
+   }
+
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+   texImage = _mesa_get_tex_image(ctx, texObj, target, 0);
+   stImage = st_texture_image(texImage);
+   
+   _mesa_init_teximage_fields(ctx, GL_TEXTURE_2D, texImage, pt->width[0],
+                              pt->height[0], 1, 0, internalFormat);
+
+   texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat, GL_RGBA,
+                                                GL_UNSIGNED_BYTE);
+   _mesa_set_fetch_functions(texImage, 2);
+
+   pipe_texture_reference(&stImage->pt, pt);
+
+   return 1;
+}
+
+/** Redirect rendering into stfb's surface to a texture image */
+int
+st_bind_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                 int target, int format, int level)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const GLuint unit = ctx->Texture.CurrentUnit;
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   struct st_texture_image *stImage;
+   struct st_renderbuffer *strb;
+   GLint face = 0, slice = 0;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+
+   if (strb->texture_save || strb->surface_save) {
+      /* Error! */
+      return 0;
+   }
+
+   if (target == ST_TEXTURE_2D) {
+      texObj = texUnit->Current2D;
+      texImage = _mesa_get_tex_image(ctx, texObj, GL_TEXTURE_2D, level);
+      stImage = st_texture_image(texImage);
+   }
+   else {
+      /* unsupported target */
+      return 0;
+   }
+
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* save the renderbuffer's surface/texture info */
+   pipe_texture_reference(&strb->texture_save, strb->texture);
+   pipe_surface_reference(&strb->surface_save, strb->surface);
+
+   /* plug in new surface/texture info */
+   pipe_texture_reference(&strb->texture, stImage->pt);
+   strb->surface = screen->get_tex_surface(screen, strb->texture,
+                                           face, level, slice,
+                                           (PIPE_BUFFER_USAGE_GPU_READ |
+                                            PIPE_BUFFER_USAGE_GPU_WRITE));
+
+   st->dirty.st |= ST_NEW_FRAMEBUFFER;
+
+   return 1;
+}
+
+
+/** Undo surface-to-texture binding */
+int
+st_release_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                    int target, int format, int level)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct st_context *st = ctx->st;
+   struct st_renderbuffer *strb;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+
+   if (!strb->texture_save || !strb->surface_save) {
+      /* Error! */
+      return 0;
+   }
+
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* free tex surface, restore original */
+   pipe_surface_reference(&strb->surface, strb->surface_save);
+   pipe_texture_reference(&strb->texture, strb->texture_save);
+
+   pipe_surface_reference(&strb->surface_save, NULL);
+   pipe_texture_reference(&strb->texture_save, NULL);
+
+   st->dirty.st |= ST_NEW_FRAMEBUFFER;
+
+   return 1;
+}
diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h
new file mode 100644
index 00000000000..31f66ad52cf
--- /dev/null
+++ b/src/mesa/state_tracker/st_texture.h
@@ -0,0 +1,179 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 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.
+ * 
+ **************************************************************************/
+
+#ifndef ST_TEXTURE_H
+#define ST_TEXTURE_H
+
+
+#include "main/mtypes.h"
+
+struct pipe_context;
+struct pipe_texture;
+
+
+struct st_texture_image
+{
+   struct gl_texture_image base;
+
+   /* These aren't stored in gl_texture_image 
+    */
+   GLuint level;
+   GLuint face;
+
+   /* If stImage->pt != NULL, image data is stored here.
+    * Else if stImage->base.Data != NULL, image is stored there.
+    * Else there is no image data.
+    */
+   struct pipe_texture *pt;
+
+   struct pipe_surface *surface;
+};
+
+
+
+struct st_texture_object
+{
+   struct gl_texture_object base;       /* The "parent" object */
+
+   /* The texture must include at levels [0..lastLevel] once validated:
+    */
+   GLuint lastLevel;
+
+   /* On validation any active images held in main memory or in other
+    * textures will be copied to this texture and the old storage freed.
+    */
+   struct pipe_texture *pt;
+
+   GLboolean teximage_realloc;
+};
+
+
+static INLINE struct st_texture_image *
+st_texture_image(struct gl_texture_image *img)
+{
+   return (struct st_texture_image *) img;
+}
+
+static INLINE struct st_texture_object *
+st_texture_object(struct gl_texture_object *obj)
+{
+   return (struct st_texture_object *) obj;
+}
+
+
+static INLINE struct pipe_texture *
+st_get_texobj_texture(struct gl_texture_object *texObj)
+{
+   struct st_texture_object *stObj = st_texture_object(texObj);
+   return stObj ? stObj->pt : NULL;
+}
+
+
+static INLINE struct pipe_texture *
+st_get_stobj_texture(struct st_texture_object *stObj)
+{
+   return stObj ? stObj->pt : NULL;
+}
+
+static INLINE GLboolean pf_is_depth_stencil( enum pipe_format format )
+{
+   return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) +
+           pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0;
+}
+
+
+extern struct pipe_texture *
+st_texture_create(struct st_context *st,
+                  enum pipe_texture_target target,
+		  enum pipe_format format,
+                  GLuint last_level,
+                  GLuint width0,
+                  GLuint height0,
+                  GLuint depth0,
+                  GLuint compress_byte,
+                  GLuint tex_usage );
+
+
+/* Check if an image fits into an existing texture object.
+ */
+extern GLboolean
+st_texture_match_image(const struct pipe_texture *pt,
+                       const struct gl_texture_image *image,
+                       GLuint face, GLuint level);
+
+/* Return a pointer to an image within a texture.  Return image stride as
+ * well.
+ */
+extern GLubyte *
+st_texture_image_map(struct st_context *st,
+                     struct st_texture_image *stImage,
+		     GLuint zoffset,
+                     GLuint flags);
+
+extern void
+st_texture_image_unmap(struct st_context *st,
+                       struct st_texture_image *stImage);
+
+
+/* Return pointers to each 2d slice within an image.  Indexed by depth
+ * value.
+ */
+extern const GLuint *
+st_texture_depth_offsets(struct pipe_texture *pt, GLuint level);
+
+
+/* Return the linear offset of an image relative to the start of its region.
+ */
+extern GLuint
+st_texture_image_offset(const struct pipe_texture *pt,
+                        GLuint face, GLuint level);
+
+extern GLuint
+st_texture_texel_offset(const struct pipe_texture * pt,
+                        GLuint face, GLuint level,
+                        GLuint col, GLuint row, GLuint img);
+
+
+/* Upload an image into a texture
+ */
+extern void
+st_texture_image_data(struct pipe_context *pipe,
+                      struct pipe_texture *dst,
+                      GLuint face, GLuint level, void *src,
+                      GLuint src_row_pitch, GLuint src_image_pitch);
+
+
+/* Copy an image between two textures
+ */
+extern void
+st_texture_image_copy(struct pipe_context *pipe,
+                      struct pipe_texture *dst, GLuint dstLevel,
+                      struct pipe_texture *src,
+                      GLuint face);
+
+
+#endif
diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c
index 62857ddeb0d..297940adbd6 100644
--- a/src/mesa/swrast/s_context.c
+++ b/src/mesa/swrast/s_context.c
@@ -533,6 +533,9 @@ _swrast_update_texture_samplers(GLcontext *ctx)
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLuint u;
 
+   if (!swrast)
+      return; /* pipe hack */
+
    for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {
       const struct gl_texture_object *tObj = ctx->Texture.Unit[u]._Current;
       /* Note: If tObj is NULL, the sample function will be a simple
diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c
index 2fc9fd0d503..700f76d4bc4 100644
--- a/src/mesa/swrast/s_drawpix.c
+++ b/src/mesa/swrast/s_drawpix.c
@@ -881,7 +881,7 @@ _swrast_DrawPixels( GLcontext *ctx,
 
    swrast_render_finish(ctx);
 
-   _mesa_unmap_drapix_pbo(ctx, unpack);
+   _mesa_unmap_drawpix_pbo(ctx, unpack);
 }
 
 
diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index 19ec556dcd4..7bfc7b00610 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -78,6 +78,9 @@ _tnl_CreateContext( GLcontext *ctx )
 
    tnl->nr_blocks = 0;
 
+   /* plug in the VBO drawing function */
+   vbo_set_draw_func(ctx, _tnl_draw_prims);
+
    return GL_TRUE;
 }
 
diff --git a/src/mesa/tnl/t_rasterpos.c b/src/mesa/tnl/t_rasterpos.c
index dfbe0f18069..14d79eab541 100644
--- a/src/mesa/tnl/t_rasterpos.c
+++ b/src/mesa/tnl/t_rasterpos.c
@@ -26,11 +26,11 @@
 #include "main/glheader.h"
 #include "main/colormac.h"
 #include "main/context.h"
-#include "feedback.h"
-#include "light.h"
+#include "main/feedback.h"
+#include "main/light.h"
 #include "main/macros.h"
-#include "rastpos.h"
-#include "simple_list.h"
+#include "main/rastpos.h"
+#include "main/simple_list.h"
 #include "main/mtypes.h"
 
 #include "math/m_matrix.h"
diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h
index d1c147ff032..d6ebc66c045 100644
--- a/src/mesa/tnl/t_vp_build.h
+++ b/src/mesa/tnl/t_vp_build.h
@@ -29,6 +29,13 @@
 
 #include "main/mtypes.h"
 
+#define TNL_FIXED_FUNCTION_STATE_FLAGS (_NEW_PROGRAM |		\
+					_NEW_LIGHT |		\
+					_NEW_TEXTURE |		\
+					_NEW_TEXTURE_MATRIX |		\
+					_NEW_TRANSFORM |	\
+					_NEW_FOG |		\
+					_NEW_POINT)
 
 extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx );
 
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index 9a1717bf898..4d628aa9a60 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -81,6 +81,8 @@ _tnl_draw_prims( GLcontext *ctx,
 		 GLuint min_index,
 		 GLuint max_index);
 
+extern void
+_mesa_load_tracked_matrices(GLcontext *ctx);
 
 extern void
 _tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4]);
diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h
index 4c51b44cdad..5362226c2f9 100644
--- a/src/mesa/vbo/vbo.h
+++ b/src/mesa/vbo/vbo.h
@@ -117,4 +117,22 @@ void vbo_rebase_prims( GLcontext *ctx,
 void vbo_use_buffer_objects(GLcontext *ctx);
 
 
+void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func);
+
+
+void GLAPIENTRY
+_vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a);
+
+void GLAPIENTRY
+_vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z);
+
+void GLAPIENTRY
+_vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q);
+
+void GLAPIENTRY
+_vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params);
+
+void GLAPIENTRY
+_vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+
 #endif
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
index 274d23248f1..467ab4e3e55 100644
--- a/src/mesa/vbo/vbo_context.c
+++ b/src/mesa/vbo/vbo_context.c
@@ -31,6 +31,7 @@
 #include "vbo.h"
 #include "vbo_context.h"
 
+#if 0
 /* Reach out and grab this to use as the default:
  */
 extern void _tnl_draw_prims( GLcontext *ctx,
@@ -40,6 +41,7 @@ extern void _tnl_draw_prims( GLcontext *ctx,
 			     const struct _mesa_index_buffer *ib,
 			     GLuint min_index,
 			     GLuint max_index );
+#endif
 
 
 
@@ -215,7 +217,9 @@ GLboolean _vbo_CreateContext( GLcontext *ctx )
 
    /* By default: 
     */
+#if 0 /* dead - see vbo_set_draw_func() */
    vbo->draw_prims = _tnl_draw_prims;
+#endif
 
    /* Hook our functions into exec and compile dispatch tables.  These
     * will pretty much be permanently installed, which means that the
@@ -243,10 +247,20 @@ void _vbo_DestroyContext( GLcontext *ctx )
       ctx->aelt_context = NULL;
    }
 
-   vbo_exec_destroy(ctx);
+   if (vbo_context(ctx)) {
+      vbo_exec_destroy(ctx);
 #if FEATURE_dlist
-   vbo_save_destroy(ctx);
+      vbo_save_destroy(ctx);
 #endif
-   FREE(vbo_context(ctx));
-   ctx->swtnl_im = NULL;
+      FREE(vbo_context(ctx));
+      ctx->swtnl_im = NULL;
+   }
+}
+
+
+void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   vbo->draw_prims = func;
 }
+
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index f6daa253fdd..2743bf6b55c 100644
--- a/src/mesa/vbo/vbo_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -143,32 +143,36 @@ static void vbo_exec_copy_to_current( struct vbo_exec_context *exec )
 
    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.
           */
-         if (exec->vtx.attrptr[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 >= VBO_ATTRIB_MAT_FRONT_AMBIENT &&
-	     i <= VBO_ATTRIB_MAT_BACK_INDEXES)
-	    ctx->NewState |= _NEW_LIGHT;
+	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+         GLfloat tmp[4];
+
+         COPY_CLEAN_4V(tmp, 
+                       exec->vtx.attrsz[i], 
+                       exec->vtx.attrptr[i]);
+         
+         if (memcmp(current, tmp, sizeof(tmp)) != 0)
+         { 
+            memcpy(current, tmp, sizeof(tmp));
+	 
+            /* 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 >= VBO_ATTRIB_MAT_FRONT_AMBIENT &&
+                i <= VBO_ATTRIB_MAT_BACK_INDEXES)
+               ctx->NewState |= _NEW_LIGHT;
+            
+            ctx->NewState |= _NEW_CURRENT_ATTRIB;
+         }
       }
    }
 
@@ -696,7 +700,8 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec )
    GLuint i;
 
    /* Allocate a buffer object.  Will just reuse this object
-    * continuously.
+    * continuously, unless vbo_use_buffer_objects() is called to enable
+    * use of real VBOs.
     */
    _mesa_reference_buffer_object(ctx,
                                  &exec->vtx.bufferobj,
@@ -775,3 +780,36 @@ static void reset_attrfv( struct vbo_exec_context *exec )
    exec->vtx.vertex_size = 0;
 }
       
+
+void GLAPIENTRY
+_vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a)
+{
+   vbo_Color4f(r, g, b, a);
+}
+
+
+void GLAPIENTRY
+_vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z)
+{
+   vbo_Normal3f(x, y, z);
+}
+
+
+void GLAPIENTRY
+_vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q)
+{
+   vbo_MultiTexCoord4f(target, s, t, r, q);
+}
+
+void GLAPIENTRY
+_vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params)
+{
+   vbo_Materialfv(face, pname, params);
+}
+
+
+void GLAPIENTRY
+_vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w)
+{
+   vbo_VertexAttrib4fARB(index, x, y, z, w);
+}
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
index 82f4db17d19..8871e10cf60 100644
--- a/src/mesa/vbo/vbo_exec_array.c
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -30,6 +30,8 @@
 #include "main/state.h"
 #include "main/api_validate.h"
 #include "main/api_noop.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
 
 #include "vbo_context.h"
 
@@ -125,6 +127,7 @@ 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];
+   GLbitfield const_inputs = 0x0;
    GLuint i;
 
    exec->array.program_mode = get_program_mode(ctx);
@@ -139,19 +142,24 @@ static void recalculate_input_bindings( GLcontext *ctx )
       for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
-	 else
+	 else {
 	    inputs[i] = &vbo->legacy_currval[i];
+            const_inputs |= 1 << i;
+         }
       }
 
       for (i = 0; i < MAT_ATTRIB_MAX; i++) {
 	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->mat_currval[i];
+         const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
       }
 
       /* Could use just about anything, just to fill in the empty
        * slots:
        */
-      for (i = MAT_ATTRIB_MAX; i < VERT_ATTRIB_MAX - VERT_ATTRIB_GENERIC0; i++)
+      for (i = MAT_ATTRIB_MAX; i < VERT_ATTRIB_MAX - VERT_ATTRIB_GENERIC0; i++) {
 	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+         const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
+      }
 
       break;
    case VP_NV:
@@ -164,15 +172,19 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	    inputs[i] = exec->array.generic_array[i];
 	 else if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
-	 else
+	 else {
 	    inputs[i] = &vbo->legacy_currval[i];
+            const_inputs |= 1 << i;
+         }
       }
 
       /* Could use just about anything, just to fill in the empty
        * slots:
        */
-      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++)
+      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++) {
 	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
+         const_inputs |= 1 << i;
+      }
 
       break;
    case VP_ARB:
@@ -187,25 +199,34 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	 inputs[0] = exec->array.generic_array[0];
       else if (exec->array.legacy_array[0]->Enabled)
 	 inputs[0] = exec->array.legacy_array[0];
-      else
+      else {
 	 inputs[0] = &vbo->legacy_currval[0];
+         const_inputs |= 1 << 0;
+      }
 
 
       for (i = 1; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
-	 else
+	 else {
 	    inputs[i] = &vbo->legacy_currval[i];
+            const_inputs |= 1 << i;
+         }
       }
 
       for (i = 0; i < 16; i++) {
 	 if (exec->array.generic_array[i]->Enabled)
 	    inputs[VERT_ATTRIB_GENERIC0 + i] = exec->array.generic_array[i];
-	 else
+	 else {
 	    inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+            const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
+         }
+
       }
       break;
    }
+
+   _mesa_set_varying_vp_inputs( ctx, ~const_inputs );
 }
 
 static void bind_arrays( GLcontext *ctx )
@@ -255,6 +276,11 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
 
    bind_arrays( ctx );
 
+   /* Again...
+    */
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
    prim[0].begin = 1;
    prim[0].end = 1;
    prim[0].weak = 0;
@@ -295,6 +321,9 @@ vbo_exec_DrawRangeElements(GLenum mode,
 
    bind_arrays( ctx );
 
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
    ib.count = count;
    ib.type = type; 
    ib.obj = ctx->Array.ElementArrayBufferObj;
@@ -403,3 +432,29 @@ void vbo_exec_array_destroy( struct vbo_exec_context *exec )
 {
    /* nothing to do */
 }
+
+
+/* This API entrypoint is not ordinarily used */
+void GLAPIENTRY
+_mesa_DrawArrays(GLenum mode, GLint first, GLsizei count)
+{
+   vbo_exec_DrawArrays(mode, first, count);
+}
+
+
+/* This API entrypoint is not ordinarily used */
+void GLAPIENTRY
+_mesa_DrawElements(GLenum mode, GLsizei count, GLenum type,
+                   const GLvoid *indices)
+{
+   vbo_exec_DrawElements(mode, count, type, indices);
+}
+
+
+/* This API entrypoint is not ordinarily used */
+void GLAPIENTRY
+_mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
+                        GLenum type, const GLvoid *indices)
+{
+   vbo_exec_DrawRangeElements(mode, start, end, count, type, indices);
+}
diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 7204aec1dcb..45133fa4f32 100644
--- a/src/mesa/vbo/vbo_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -150,6 +150,7 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
    GLubyte *data = exec->vtx.buffer_map;
    const GLuint *map;
    GLuint attr;
+   GLbitfield varying_inputs = 0x0;
 
    /* Install the default (ie Current) attributes first, then overlay
     * all active ones.
@@ -225,8 +226,11 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 	 arrays[attr]._MaxElement = count; /* ??? */
 
 	 data += exec->vtx.attrsz[src] * sizeof(GLfloat);
+         varying_inputs |= 1<<attr;
       }
    }
+
+   _mesa_set_varying_vp_inputs( ctx, varying_inputs );
 }
 
 
@@ -256,6 +260,9 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec )
 	  */
 	 vbo_exec_bind_arrays( ctx );
 
+         if (ctx->NewState)
+            _mesa_update_state( ctx );
+
          /* if using a real VBO, unmap it before drawing */
          if (exec->vtx.bufferobj->Name) {
             ctx->Driver.UnmapBuffer(ctx, target, exec->vtx.bufferobj);
diff --git a/src/mesa/vbo/vbo_save_draw.c b/src/mesa/vbo/vbo_save_draw.c
index 7ee0a9a33f4..617c5a77baf 100644
--- a/src/mesa/vbo/vbo_save_draw.c
+++ b/src/mesa/vbo/vbo_save_draw.c
@@ -64,18 +64,26 @@ static void _playback_copy_to_current( GLcontext *ctx,
    for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (node->attrsz[i]) {
 	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+         GLfloat tmp[4];
 
-	 COPY_CLEAN_4V(current, 
-		       node->attrsz[i], 
-		       data);
+         COPY_CLEAN_4V(tmp, 
+                       node->attrsz[i], 
+                       data);
+         
+         if (memcmp(current, tmp, 4 * sizeof(GLfloat)) != 0)
+         {
+            memcpy(current, tmp, 4 * sizeof(GLfloat));
 
-	 vbo->currval[i].Size = node->attrsz[i];
+            vbo->currval[i].Size = node->attrsz[i];
 
-	 data += node->attrsz[i];
+            if (i >= VBO_ATTRIB_FIRST_MATERIAL &&
+                i <= VBO_ATTRIB_LAST_MATERIAL)
+               ctx->NewState |= _NEW_LIGHT;
+
+            ctx->NewState |= _NEW_CURRENT_ATTRIB;
+         }
 
-	 if (i >= VBO_ATTRIB_FIRST_MATERIAL &&
-	     i <= VBO_ATTRIB_LAST_MATERIAL)
-	    ctx->NewState |= _NEW_LIGHT;
+	 data += node->attrsz[i];
       }
    }
 
@@ -111,6 +119,7 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
    const GLuint *map;
    GLuint attr;
    GLubyte node_attrsz[VBO_ATTRIB_MAX];  /* copy of node->attrsz[] */
+   GLbitfield varying_inputs = 0x0;
 
    memcpy(node_attrsz, node->attrsz, sizeof(node->attrsz));
 
@@ -176,8 +185,11 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
 	 assert(arrays[attr].BufferObj->Name);
 
 	 data += node->attrsz[src] * sizeof(GLfloat);
+         varying_inputs |= 1<<attr;
       }
    }
+
+   _mesa_set_varying_vp_inputs( ctx, varying_inputs );
 }
 
 static void vbo_save_loopback_vertex_list( GLcontext *ctx,
@@ -246,6 +258,11 @@ void vbo_save_playback_vertex_list( GLcontext *ctx, void *data )
 
       vbo_bind_vertex_list( ctx, node );
 
+      /* Again...
+       */
+      if (ctx->NewState)
+	 _mesa_update_state( ctx );
+
       vbo_context(ctx)->draw_prims( ctx, 
 				    save->inputs, 
 				    node->prim, 
diff --git a/src/mesa/vf/vf.c b/src/mesa/vf/vf.c
new file mode 100644
index 00000000000..82f3d2b641d
--- /dev/null
+++ b/src/mesa/vf/vf.c
@@ -0,0 +1,372 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <[email protected]>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+
+#include "vf.h"
+
+#define DBG 0
+
+
+
+static GLboolean match_fastpath( struct vertex_fetch *vf,
+				 const struct vf_fastpath *fp)
+{
+   GLuint j;
+
+   if (vf->attr_count != fp->attr_count) 
+      return GL_FALSE;
+
+   for (j = 0; j < vf->attr_count; j++) 
+      if (vf->attr[j].format != fp->attr[j].format ||
+	  vf->attr[j].inputsize != fp->attr[j].size ||
+	  vf->attr[j].vertoffset != fp->attr[j].offset) 
+	 return GL_FALSE;
+      
+   if (fp->match_strides) {
+      if (vf->vertex_stride != fp->vertex_stride)
+	 return GL_FALSE;
+
+      for (j = 0; j < vf->attr_count; j++) 
+	 if (vf->attr[j].inputstride != fp->attr[j].stride) 
+	    return GL_FALSE;
+   }
+   
+   return GL_TRUE;
+}
+
+static GLboolean search_fastpath_emit( struct vertex_fetch *vf )
+{
+   struct vf_fastpath *fp = vf->fastpath;
+
+   for ( ; fp ; fp = fp->next) {
+      if (match_fastpath(vf, fp)) {
+         vf->emit = fp->func;
+	 return GL_TRUE;
+      }
+   }
+
+   return GL_FALSE;
+}
+
+void vf_register_fastpath( struct vertex_fetch *vf,
+			     GLboolean match_strides )
+{
+   struct vf_fastpath *fastpath = CALLOC_STRUCT(vf_fastpath);
+   GLuint i;
+
+   fastpath->vertex_stride = vf->vertex_stride;
+   fastpath->attr_count = vf->attr_count;
+   fastpath->match_strides = match_strides;
+   fastpath->func = vf->emit;
+   fastpath->attr = (struct vf_attr_type *)
+      _mesa_malloc(vf->attr_count * sizeof(fastpath->attr[0]));
+
+   for (i = 0; i < vf->attr_count; i++) {
+      fastpath->attr[i].format = vf->attr[i].format;
+      fastpath->attr[i].stride = vf->attr[i].inputstride;
+      fastpath->attr[i].size = vf->attr[i].inputsize;
+      fastpath->attr[i].offset = vf->attr[i].vertoffset;
+   }
+
+   fastpath->next = vf->fastpath;
+   vf->fastpath = fastpath;
+}
+
+
+
+
+/***********************************************************************
+ * Build codegen functions or return generic ones:
+ */
+static void choose_emit_func( struct vertex_fetch *vf, 
+			      GLuint count, 
+			      GLubyte *dest)
+{
+   vf->emit = NULL;
+   
+   /* Does this match an existing (hardwired, codegen or known-bad)
+    * fastpath?
+    */
+   if (search_fastpath_emit(vf)) {
+      /* Use this result.  If it is null, then it is already known
+       * that the current state will fail for codegen and there is no
+       * point trying again.
+       */
+   }
+   else if (vf->codegen_emit) {
+      vf->codegen_emit( vf );
+   }
+
+   if (!vf->emit) {
+      vf_generate_hardwired_emit(vf);
+   }
+
+   /* Otherwise use the generic version:
+    */
+   if (!vf->emit)
+      vf->emit = vf_generic_emit;
+
+   vf->emit( vf, count, dest );
+}
+
+
+
+
+
+/***********************************************************************
+ * Public entrypoints, mostly dispatch to the above:
+ */
+
+
+
+GLuint vf_set_vertex_attributes( struct vertex_fetch *vf, 
+				 const struct vf_attr_map *map,
+				 GLuint nr, 
+				 GLuint vertex_stride )
+{
+   GLuint offset = 0;
+   GLuint i, j;
+
+   assert(nr < VF_ATTRIB_MAX);
+
+   memset(vf->lookup, 0, sizeof(vf->lookup));
+
+   for (j = 0, i = 0; i < nr; i++) {
+      const GLuint format = map[i].format;
+      if (format == EMIT_PAD) {
+	 if (DBG)
+	    _mesa_printf("%d: pad %d, offset %d\n", i,  
+			 map[i].offset, offset);  
+
+	 offset += map[i].offset;
+
+      }
+      else {
+	 assert(vf->lookup[map[i].attrib] == 0);
+	 vf->lookup[map[i].attrib] = &vf->attr[j];
+
+	 vf->attr[j].attrib = map[i].attrib;
+	 vf->attr[j].format = format;
+	 vf->attr[j].insert = vf_format_info[format].insert;
+	 vf->attr[j].extract = vf_format_info[format].extract;
+	 vf->attr[j].vertattrsize = vf_format_info[format].attrsize;
+	 vf->attr[j].vertoffset = offset;
+	 
+	 if (DBG)
+	    _mesa_printf("%d: %s, offset %d\n", i,  
+			 vf_format_info[format].name,
+			 vf->attr[j].vertoffset);   
+
+	 offset += vf_format_info[format].attrsize;
+	 j++;
+      }
+   }
+
+   vf->attr_count = j;
+   vf->vertex_stride = vertex_stride ? vertex_stride : offset;
+   vf->emit = choose_emit_func;
+
+   assert(vf->vertex_stride >= offset);
+   return vf->vertex_stride;
+}
+
+
+
+void vf_set_vp_matrix( struct vertex_fetch *vf,
+		       const GLfloat *viewport )
+{
+   assert(vf->allow_viewport_emits);
+
+   /* scale */
+   vf->vp[0] = viewport[MAT_SX];
+   vf->vp[1] = viewport[MAT_SY];
+   vf->vp[2] = viewport[MAT_SZ];
+   vf->vp[3] = 1.0;
+
+   /* translate */
+   vf->vp[4] = viewport[MAT_TX];
+   vf->vp[5] = viewport[MAT_TY];
+   vf->vp[6] = viewport[MAT_TZ];
+   vf->vp[7] = 0.0;
+}
+
+void vf_set_vp_scale_translate( struct vertex_fetch *vf,
+				const GLfloat *scale,
+				const GLfloat *translate )
+{
+   assert(vf->allow_viewport_emits);
+
+   vf->vp[0] = scale[0];
+   vf->vp[1] = scale[1];
+   vf->vp[2] = scale[2];
+   vf->vp[3] = scale[3];
+
+   vf->vp[4] = translate[0];
+   vf->vp[5] = translate[1];
+   vf->vp[6] = translate[2];
+   vf->vp[7] = translate[3];
+}
+
+
+/* Set attribute pointers, adjusted for start position:
+ */
+void vf_set_sources( struct vertex_fetch *vf,
+		     GLvector4f * const sources[],
+		     GLuint start )
+{
+   struct vf_attr *a = vf->attr;
+   GLuint j;
+   
+   for (j = 0; j < vf->attr_count; j++) {
+      const GLvector4f *vptr = sources[a[j].attrib];
+      
+      if ((a[j].inputstride != vptr->stride) ||
+	  (a[j].inputsize != vptr->size))
+	 vf->emit = choose_emit_func;
+      
+      a[j].inputstride = vptr->stride;
+      a[j].inputsize = vptr->size;
+      a[j].do_insert = a[j].insert[vptr->size - 1]; 
+      a[j].inputptr = ((GLubyte *)vptr->data) + start * vptr->stride;
+   }
+}
+
+
+
+/* Emit count VB vertices to dest.  
+ */
+void vf_emit_vertices( struct vertex_fetch *vf,
+		       GLuint count,
+		       void *dest )
+{
+   vf->emit( vf, count, (GLubyte*) dest );	
+}
+
+
+/* Extract a named attribute from a hardware vertex.  Will have to
+ * reverse any viewport transformation, swizzling or other conversions
+ * which may have been applied.
+ *
+ * This is mainly required for on-the-fly vertex translations to
+ * swrast format.
+ */
+void vf_get_attr( struct vertex_fetch *vf,
+		  const void *vertex,
+		  GLenum attr, 
+		  const GLfloat *dflt,
+		  GLfloat *dest )
+{
+   const struct vf_attr *a = vf->attr;
+   const GLuint attr_count = vf->attr_count;
+   GLuint j;
+
+   for (j = 0; j < attr_count; j++) {
+      if (a[j].attrib == attr) {
+	 a[j].extract( &a[j], dest, (GLubyte *)vertex + a[j].vertoffset );
+	 return;
+      }
+   }
+
+   /* Else return the value from ctx->Current.
+    */
+   _mesa_memcpy( dest, dflt, 4*sizeof(GLfloat));
+}
+
+
+
+
+struct vertex_fetch *vf_create( GLboolean allow_viewport_emits )
+{
+   struct vertex_fetch *vf = CALLOC_STRUCT(vertex_fetch);
+   GLuint i;
+
+   for (i = 0; i < VF_ATTRIB_MAX; i++)
+      vf->attr[i].vf = vf;
+
+   vf->allow_viewport_emits = allow_viewport_emits;
+
+   switch(CHAN_TYPE) {
+   case GL_UNSIGNED_BYTE:
+      vf->chan_scale[0] = 255.0;
+      vf->chan_scale[1] = 255.0;
+      vf->chan_scale[2] = 255.0;
+      vf->chan_scale[3] = 255.0;
+      break;
+   case GL_UNSIGNED_SHORT:
+      vf->chan_scale[0] = 65535.0;
+      vf->chan_scale[1] = 65535.0;
+      vf->chan_scale[2] = 65535.0;
+      vf->chan_scale[3] = 65535.0;
+      break;
+   default:
+      vf->chan_scale[0] = 1.0;
+      vf->chan_scale[1] = 1.0;
+      vf->chan_scale[2] = 1.0;
+      vf->chan_scale[3] = 1.0;
+      break;
+   }
+
+   vf->identity[0] = 0.0;
+   vf->identity[1] = 0.0;
+   vf->identity[2] = 0.0;
+   vf->identity[3] = 1.0;
+
+   vf->codegen_emit = NULL;
+
+#ifdef USE_SSE_ASM
+   if (!_mesa_getenv("MESA_NO_CODEGEN"))
+      vf->codegen_emit = vf_generate_sse_emit;
+#endif
+
+   return vf;
+}
+
+
+void vf_destroy( struct vertex_fetch *vf )
+{
+   struct vf_fastpath *fp, *tmp;
+
+   for (fp = vf->fastpath ; fp ; fp = tmp) {
+      tmp = fp->next;
+      FREE(fp->attr);
+
+      /* KW: At the moment, fp->func is constrained to be allocated by
+       * _mesa_exec_alloc(), as the hardwired fastpaths in
+       * t_vertex_generic.c are handled specially.  It would be nice
+       * to unify them, but this probably won't change until this
+       * module gets another overhaul.
+       */
+      _mesa_exec_free((void *) fp->func);
+      FREE(fp);
+   }
+   
+   vf->fastpath = NULL;
+   FREE(vf);
+}
diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h
new file mode 100644
index 00000000000..83d7547619c
--- /dev/null
+++ b/src/mesa/vf/vf.h
@@ -0,0 +1,234 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <[email protected]>
+ */
+
+#ifndef VF_VERTEX_H
+#define VF_VERTEX_H
+
+#include "main/mtypes.h"
+#include "math/m_vector.h"
+
+enum {
+   VF_ATTRIB_POS = 0,
+   VF_ATTRIB_WEIGHT = 1,
+   VF_ATTRIB_NORMAL = 2,
+   VF_ATTRIB_COLOR0 = 3,
+   VF_ATTRIB_COLOR1 = 4,
+   VF_ATTRIB_FOG = 5,
+   VF_ATTRIB_COLOR_INDEX = 6,
+   VF_ATTRIB_EDGEFLAG = 7,
+   VF_ATTRIB_TEX0 = 8,
+   VF_ATTRIB_TEX1 = 9,
+   VF_ATTRIB_TEX2 = 10,
+   VF_ATTRIB_TEX3 = 11,
+   VF_ATTRIB_TEX4 = 12,
+   VF_ATTRIB_TEX5 = 13,
+   VF_ATTRIB_TEX6 = 14,
+   VF_ATTRIB_TEX7 = 15,
+   VF_ATTRIB_VAR0 = 16,
+   VF_ATTRIB_VAR1 = 17,
+   VF_ATTRIB_VAR2 = 18,
+   VF_ATTRIB_VAR3 = 19,
+   VF_ATTRIB_VAR4 = 20,
+   VF_ATTRIB_VAR5 = 21,
+   VF_ATTRIB_VAR6 = 22,
+   VF_ATTRIB_VAR7 = 23,
+   VF_ATTRIB_POINTSIZE = 24,
+   VF_ATTRIB_BFC0 = 25,
+   VF_ATTRIB_BFC1 = 26,
+   VF_ATTRIB_CLIP_POS = 27,
+   VF_ATTRIB_VERTEX_HEADER = 28,
+   VF_ATTRIB_MAX = 29
+};
+
+
+enum vf_attr_format {
+   EMIT_1F,
+   EMIT_2F,
+   EMIT_3F,
+   EMIT_4F,
+   EMIT_2F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_3F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_4F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_3F_XYW,			/* for projective texture */
+   EMIT_1UB_1F,			/* for fog coordinate */
+   EMIT_3UB_3F_RGB,		/* for specular color */
+   EMIT_3UB_3F_BGR,		/* for specular color */
+   EMIT_4UB_4F_RGBA,		/* for color */
+   EMIT_4UB_4F_BGRA,		/* for color */
+   EMIT_4UB_4F_ARGB,		/* for color */
+   EMIT_4UB_4F_ABGR,		/* for color */
+   EMIT_4CHAN_4F_RGBA,		/* for swrast color */
+   EMIT_PAD,			/* leave a hole of 'offset' bytes */
+   EMIT_MAX
+};
+
+struct vf_attr_map {
+   GLuint attrib;
+   enum vf_attr_format format;
+   GLuint offset;
+};
+
+struct vertex_fetch;
+
+void vf_set_vp_matrix( struct vertex_fetch *vf,
+		      const GLfloat *viewport );
+
+void vf_set_vp_scale_translate( struct vertex_fetch *vf,
+				const GLfloat *scale,
+				const GLfloat *translate );
+
+GLuint vf_set_vertex_attributes( struct vertex_fetch *vf,
+				 const struct vf_attr_map *map,
+				 GLuint nr, 
+				 GLuint vertex_stride );
+
+void vf_set_sources( struct vertex_fetch *vf,
+		     GLvector4f * const attrib[],
+		     GLuint start ); 
+
+void vf_emit_vertices( struct vertex_fetch *vf,
+		       GLuint count,
+		       void *dest );
+
+void vf_get_attr( struct vertex_fetch *vf,
+		  const void *vertex,
+		  GLenum attr, 
+		  const GLfloat *dflt,
+		  GLfloat *dest );
+
+struct vertex_fetch *vf_create( GLboolean allow_viewport_emits );
+
+void vf_destroy( struct vertex_fetch *vf );
+
+
+
+/***********************************************************************
+ * Internal functions and structs:
+ */
+
+struct vf_attr;
+
+typedef void (*vf_extract_func)( const struct vf_attr *a, 
+				 GLfloat *out, 
+				 const GLubyte *v );
+
+typedef void (*vf_insert_func)( const struct vf_attr *a, 
+				GLubyte *v, 
+				const GLfloat *in );
+
+typedef void (*vf_emit_func)( struct vertex_fetch *vf,
+			      GLuint count, 
+			      GLubyte *dest );
+
+
+
+/* Describes how to convert/move a vertex attribute from a vertex
+ * array to a vertex structure.
+ */
+struct vf_attr
+{
+   struct vertex_fetch *vf;
+
+   GLuint format;
+   GLuint inputsize;
+   GLuint inputstride;
+   GLuint vertoffset;      /* position of the attrib in the vertex struct */
+
+   GLuint attrib;          /* which vertex attrib (0=position, etc) */
+   GLuint vertattrsize;    /* size of the attribute in bytes */
+
+   GLubyte *inputptr;
+   const vf_insert_func *insert;
+   vf_insert_func do_insert;
+   vf_extract_func extract;
+};
+
+struct vertex_fetch
+{
+   struct vf_attr attr[VF_ATTRIB_MAX];
+   GLuint attr_count;
+   GLuint vertex_stride;
+
+   struct vf_attr *lookup[VF_ATTRIB_MAX];
+   
+   vf_emit_func emit;
+
+   /* Parameters and constants for codegen:
+    */
+   GLboolean allow_viewport_emits;
+   GLfloat vp[8];		
+   GLfloat chan_scale[4];
+   GLfloat identity[4];
+
+   struct vf_fastpath *fastpath;
+   
+   void (*codegen_emit)( struct vertex_fetch *vf );
+};
+
+
+struct vf_attr_type {
+   GLuint format;
+   GLuint size;
+   GLuint stride;
+   GLuint offset;
+};
+
+struct vf_fastpath {
+   GLuint vertex_stride;
+   GLuint attr_count;
+   GLboolean match_strides;
+
+   struct vf_attr_type *attr;
+
+   vf_emit_func func;
+   struct vf_fastpath *next;
+};
+
+
+void vf_register_fastpath( struct vertex_fetch *vtx,
+			     GLboolean match_strides );
+
+void vf_generic_emit( struct vertex_fetch *vf,
+			GLuint count,
+			GLubyte *v );
+
+void vf_generate_hardwired_emit( struct vertex_fetch *vf );
+
+void vf_generate_sse_emit( struct vertex_fetch *vf );
+
+
+struct vf_format_info {
+   const char *name;
+   vf_extract_func extract;
+   vf_insert_func insert[4];
+   const GLuint attrsize;
+};
+
+const struct vf_format_info vf_format_info[EMIT_MAX];
+
+
+#endif
diff --git a/src/mesa/vf/vf_generic.c b/src/mesa/vf/vf_generic.c
new file mode 100644
index 00000000000..baa00af29a6
--- /dev/null
+++ b/src/mesa/vf/vf_generic.c
@@ -0,0 +1,981 @@
+
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <[email protected]>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+
+#include "vf/vf.h"
+
+
+/*
+ * These functions take the NDC coordinates pointed to by 'in', apply the
+ * NDC->Viewport mapping and store the results at 'v'.
+ */
+
+static INLINE void insert_4f_viewport_4( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+   out[3] = in[3];
+}
+
+static INLINE void insert_4f_viewport_3( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_viewport_2( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] =                    trans[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_viewport_1( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] =                    trans[1];
+   out[2] =                    trans[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_3f_viewport_3( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+}
+
+static INLINE void insert_3f_viewport_2( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+}
+
+static INLINE void insert_3f_viewport_1( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] =                    trans[1];
+   out[2] =                    trans[2];
+}
+
+static INLINE void insert_2f_viewport_2( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+}
+
+static INLINE void insert_2f_viewport_1( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = trans[1];
+}
+
+
+/*
+ * These functions do the same as above, except for the viewport mapping.
+ */
+
+static INLINE void insert_4f_4( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = in[3];
+}
+
+static INLINE void insert_4f_3( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static INLINE void insert_3f_xyw_4( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[3];
+}
+
+static INLINE void insert_3f_xyw_err( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   (void) a; (void) v; (void) in;
+   _mesa_exit(1);
+}
+
+static INLINE void insert_3f_3( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+}
+
+static INLINE void insert_3f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+}
+
+static INLINE void insert_3f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+}
+
+
+static INLINE void insert_2f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+}
+
+static INLINE void insert_2f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+}
+
+static INLINE void insert_1f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+
+   out[0] = in[0];
+}
+
+static INLINE void insert_null( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   (void) a; (void) v; (void) in;
+}
+
+static INLINE void insert_4chan_4f_rgba_4( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[3], in[3]);
+}
+
+static INLINE void insert_4chan_4f_rgba_3( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); 
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4chan_4f_rgba_2( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   c[2] = 0;
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4chan_4f_rgba_1( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   c[1] = 0;
+   c[2] = 0;
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4ub_4f_rgba_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+}
+
+static INLINE void insert_4ub_4f_rgba_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_rgba_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[2] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_rgba_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   v[1] = 0;
+   v[2] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+}
+
+static INLINE void insert_4ub_4f_bgra_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[0] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   v[1] = 0;
+   v[0] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]);
+}
+
+static INLINE void insert_4ub_4f_argb_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]);
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   v[3] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   v[2] = 0x00;
+   v[3] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]);
+}
+
+static INLINE void insert_4ub_4f_abgr_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]);
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   v[1] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   v[2] = 0x00;
+   v[1] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_3ub_3f_rgb_3( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+}
+
+static INLINE void insert_3ub_3f_rgb_2( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[2] = 0;
+}
+
+static INLINE void insert_3ub_3f_rgb_1( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   v[1] = 0;
+   v[2] = 0;
+}
+
+static INLINE void insert_3ub_3f_bgr_3( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+}
+
+static INLINE void insert_3ub_3f_bgr_2( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[0] = 0;
+}
+
+static INLINE void insert_3ub_3f_bgr_1( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   v[1] = 0;
+   v[0] = 0;
+}
+
+
+static INLINE void insert_1ub_1f_1( const struct vf_attr *a, GLubyte *v, 
+				    const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+}
+
+
+/***********************************************************************
+ * Functions to perform the reverse operations to the above, for
+ * swrast translation and clip-interpolation.
+ * 
+ * Currently always extracts a full 4 floats.
+ */
+
+static void extract_4f_viewport( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   /* Although included for completeness, the position coordinate is
+    * usually handled differently during clipping.
+    */
+   out[0] = (in[0] - trans[0]) / scale[0];
+   out[1] = (in[1] - trans[1]) / scale[1];
+   out[2] = (in[2] - trans[2]) / scale[2];
+   out[3] = in[3];
+}
+
+static void extract_3f_viewport( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = (in[0] - trans[0]) / scale[0];
+   out[1] = (in[1] - trans[1]) / scale[1];
+   out[2] = (in[2] - trans[2]) / scale[2];
+   out[3] = 1;
+}
+
+
+static void extract_2f_viewport( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = (in[0] - trans[0]) / scale[0];
+   out[1] = (in[1] - trans[1]) / scale[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+
+static void extract_4f( const struct vf_attr *a, GLfloat *out, const GLubyte *v  )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = in[3];
+}
+
+static void extract_3f_xyw( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = in[2];
+}
+
+
+static void extract_3f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = 1;
+}
+
+
+static void extract_2f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static void extract_1f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static void extract_4chan_4f_rgba( const struct vf_attr *a, GLfloat *out, 
+				   const GLubyte *v )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+
+   out[0] = CHAN_TO_FLOAT(c[0]);
+   out[1] = CHAN_TO_FLOAT(c[1]);
+   out[2] = CHAN_TO_FLOAT(c[2]);
+   out[3] = CHAN_TO_FLOAT(c[3]);
+}
+
+static void extract_4ub_4f_rgba( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[2] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_bgra( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[2] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[0] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_argb( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[3] = UBYTE_TO_FLOAT(v[0]);
+   out[0] = UBYTE_TO_FLOAT(v[1]);
+   out[1] = UBYTE_TO_FLOAT(v[2]);
+   out[2] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_abgr( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[3] = UBYTE_TO_FLOAT(v[0]);
+   out[2] = UBYTE_TO_FLOAT(v[1]);
+   out[1] = UBYTE_TO_FLOAT(v[2]);
+   out[0] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_3ub_3f_rgb( const struct vf_attr *a, GLfloat *out, 
+				const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[2] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = 1;
+}
+
+static void extract_3ub_3f_bgr( const struct vf_attr *a, GLfloat *out, 
+				const GLubyte *v )
+{
+   (void) a;
+   out[2] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[0] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = 1;
+}
+
+static void extract_1ub_1f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+
+const struct vf_format_info vf_format_info[EMIT_MAX] = 
+{
+   { "1f",
+     extract_1f,
+     { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 },
+     sizeof(GLfloat) },
+
+   { "2f",
+     extract_2f,
+     { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 },
+     2 * sizeof(GLfloat) },
+
+   { "3f",
+     extract_3f,
+     { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 },
+     3 * sizeof(GLfloat) },
+
+   { "4f",
+     extract_4f,
+     { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 },
+     4 * sizeof(GLfloat) },
+
+   { "2f_viewport",
+     extract_2f_viewport,
+     { insert_2f_viewport_1, insert_2f_viewport_2, insert_2f_viewport_2,
+       insert_2f_viewport_2 },
+     2 * sizeof(GLfloat) },
+
+   { "3f_viewport",
+     extract_3f_viewport,
+     { insert_3f_viewport_1, insert_3f_viewport_2, insert_3f_viewport_3,
+       insert_3f_viewport_3 },
+     3 * sizeof(GLfloat) },
+
+   { "4f_viewport",
+     extract_4f_viewport,
+     { insert_4f_viewport_1, insert_4f_viewport_2, insert_4f_viewport_3,
+       insert_4f_viewport_4 }, 
+     4 * sizeof(GLfloat) },
+
+   { "3f_xyw",
+     extract_3f_xyw,
+     { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, 
+       insert_3f_xyw_4 },
+     3 * sizeof(GLfloat) },
+
+   { "1ub_1f",
+     extract_1ub_1f,
+     { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 },
+     sizeof(GLubyte) },
+
+   { "3ub_3f_rgb",
+     extract_3ub_3f_rgb,
+     { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3,
+       insert_3ub_3f_rgb_3 },
+     3 * sizeof(GLubyte) },
+
+   { "3ub_3f_bgr",
+     extract_3ub_3f_bgr,
+     { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3,
+       insert_3ub_3f_bgr_3 },
+     3 * sizeof(GLubyte) },
+
+   { "4ub_4f_rgba",
+     extract_4ub_4f_rgba,
+     { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, 
+       insert_4ub_4f_rgba_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_bgra",
+     extract_4ub_4f_bgra,
+     { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3,
+       insert_4ub_4f_bgra_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_argb",
+     extract_4ub_4f_argb,
+     { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3,
+       insert_4ub_4f_argb_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_abgr",
+     extract_4ub_4f_abgr,
+     { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3,
+       insert_4ub_4f_abgr_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4chan_4f_rgba",
+     extract_4chan_4f_rgba,
+     { insert_4chan_4f_rgba_1, insert_4chan_4f_rgba_2, insert_4chan_4f_rgba_3,
+       insert_4chan_4f_rgba_4 },
+     4 * sizeof(GLchan) },
+
+   { "pad",
+     NULL,
+     { NULL, NULL, NULL, NULL },
+     0 }
+
+};
+
+
+
+    
+/***********************************************************************
+ * Hardwired fastpaths for emitting whole vertices or groups of
+ * vertices
+ */
+#define EMIT5(NR, F0, F1, F2, F3, F4, NAME)				\
+static void NAME( struct vertex_fetch *vf,				\
+		  GLuint count,						\
+		  GLubyte *v )						\
+{									\
+   struct vf_attr *a = vf->attr;				\
+   GLuint i;								\
+									\
+   for (i = 0 ; i < count ; i++, v += vf->vertex_stride) {		\
+      if (NR > 0) {							\
+	 F0( &a[0], v + a[0].vertoffset, (GLfloat *)a[0].inputptr );	\
+	 a[0].inputptr += a[0].inputstride;				\
+      }									\
+      									\
+      if (NR > 1) {							\
+	 F1( &a[1], v + a[1].vertoffset, (GLfloat *)a[1].inputptr );	\
+	 a[1].inputptr += a[1].inputstride;				\
+      }									\
+      									\
+      if (NR > 2) {							\
+	 F2( &a[2], v + a[2].vertoffset, (GLfloat *)a[2].inputptr );	\
+	 a[2].inputptr += a[2].inputstride;				\
+      }									\
+      									\
+      if (NR > 3) {							\
+	 F3( &a[3], v + a[3].vertoffset, (GLfloat *)a[3].inputptr );	\
+	 a[3].inputptr += a[3].inputstride;				\
+      }									\
+									\
+      if (NR > 4) {							\
+	 F4( &a[4], v + a[4].vertoffset, (GLfloat *)a[4].inputptr );	\
+	 a[4].inputptr += a[4].inputstride;				\
+      }									\
+   }									\
+}
+
+   
+#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \
+				  insert_null, insert_null, NAME)
+
+#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \
+				      insert_null, NAME)
+   
+#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \
+				          insert_null, NAME)
+   
+
+EMIT2(insert_3f_viewport_3, insert_4ub_4f_rgba_4, emit_viewport3_rgba4)
+EMIT2(insert_3f_viewport_3, insert_4ub_4f_bgra_4, emit_viewport3_bgra4)
+EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4)
+
+EMIT3(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_viewport4_rgba4_st2)
+EMIT3(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2,  emit_viewport4_bgra4_st2)
+EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2)
+
+EMIT4(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_viewport4_rgba4_st2_st2)
+EMIT4(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, insert_2f_2,  emit_viewport4_bgra4_st2_st2)
+EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2)
+
+
+/* Use the codegen paths to select one of a number of hardwired
+ * fastpaths.
+ */
+void vf_generate_hardwired_emit( struct vertex_fetch *vf )
+{
+   vf_emit_func func = NULL;
+
+   /* Does it fit a hardwired fastpath?  Help! this is growing out of
+    * control!
+    */
+   switch (vf->attr_count) {
+   case 2:
+      if (vf->attr[0].do_insert == insert_3f_viewport_3) {
+	 if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4) 
+	    func = emit_viewport3_bgra4;
+	 else if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) 
+	    func = emit_viewport3_rgba4;
+      }
+      else if (vf->attr[0].do_insert == insert_3f_3 &&
+	       vf->attr[1].do_insert == insert_4ub_4f_rgba_4) {
+ 	 func = emit_xyz3_rgba4; 
+      }
+      break;
+   case 3:
+      if (vf->attr[2].do_insert == insert_2f_2) {
+	 if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) {
+	    if (vf->attr[0].do_insert == insert_4f_viewport_4)
+	       func = emit_viewport4_rgba4_st2;
+	    else if (vf->attr[0].do_insert == insert_4f_4) 
+	       func = emit_xyzw4_rgba4_st2;
+	 }
+	 else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 &&
+		  vf->attr[0].do_insert == insert_4f_viewport_4)
+	    func = emit_viewport4_bgra4_st2;
+      }
+      break;
+   case 4:
+      if (vf->attr[2].do_insert == insert_2f_2 &&
+	  vf->attr[3].do_insert == insert_2f_2) {
+	 if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) {
+	    if (vf->attr[0].do_insert == insert_4f_viewport_4)
+	       func = emit_viewport4_rgba4_st2_st2;
+	    else if (vf->attr[0].do_insert == insert_4f_4) 
+	       func = emit_xyzw4_rgba4_st2_st2;
+	 }
+	 else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 &&
+		  vf->attr[0].do_insert == insert_4f_viewport_4)
+	    func = emit_viewport4_bgra4_st2_st2;
+      }
+      break;
+   }
+
+   vf->emit = func;
+}
+
+/***********************************************************************
+ * Generic (non-codegen) functions for whole vertices or groups of
+ * vertices
+ */
+
+void vf_generic_emit( struct vertex_fetch *vf,
+		      GLuint count,
+		      GLubyte *v )
+{
+   struct vf_attr *a = vf->attr;
+   const GLuint attr_count = vf->attr_count;
+   const GLuint stride = vf->vertex_stride;
+   GLuint i, j;
+
+   for (i = 0 ; i < count ; i++, v += stride) {
+      for (j = 0; j < attr_count; j++) {
+	 GLfloat *in = (GLfloat *)a[j].inputptr;
+	 a[j].inputptr += a[j].inputstride;
+	 a[j].do_insert( &a[j], v + a[j].vertoffset, in );
+      }
+   }
+}
+
+
diff --git a/src/mesa/vf/vf_sse.c b/src/mesa/vf/vf_sse.c
new file mode 100644
index 00000000000..4d70196ffed
--- /dev/null
+++ b/src/mesa/vf/vf_sse.c
@@ -0,0 +1,662 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <[email protected]>
+ */
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+#include "main/enums.h"
+
+#include "vf/vf.h"
+
+#if defined(USE_SSE_ASM)
+
+#include "x86/rtasm/x86sse.h"
+#include "x86/common_x86_asm.h"
+
+
+#define X    0
+#define Y    1
+#define Z    2
+#define W    3
+
+
+struct x86_program {
+   struct x86_function func;
+
+   struct vertex_fetch *vf;
+   GLboolean inputs_safe;
+   GLboolean outputs_safe;
+   GLboolean have_sse2;
+   
+   struct x86_reg identity;
+   struct x86_reg chan0;
+};
+
+
+static struct x86_reg get_identity( struct x86_program *p )
+{
+   return p->identity;
+}
+
+static void emit_load4f_4( struct x86_program *p, 			   
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movups(&p->func, dest, arg0);
+}
+
+static void emit_load4f_3( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Have to jump through some hoops:
+    *
+    * c 0 0 0
+    * c 0 0 1
+    * 0 0 c 1
+    * a b c 1
+    */
+   sse_movss(&p->func, dest, x86_make_disp(arg0, 8));
+   sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) );
+   sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) );
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load4f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Initialize from identity, then pull in low two words:
+    */
+   sse_movups(&p->func, dest, get_identity(p));
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load4f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Pull in low word, then swizzle in identity */
+   sse_movss(&p->func, dest, arg0);
+   sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) );
+}
+
+
+
+static void emit_load3f_3( struct x86_program *p, 			   
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Over-reads by 1 dword - potential SEGV if input is a vertex
+    * array.
+    */
+   if (p->inputs_safe) {
+      sse_movups(&p->func, dest, arg0);
+   } 
+   else {
+      /* c 0 0 0
+       * c c c c
+       * a b c c 
+       */
+      sse_movss(&p->func, dest, x86_make_disp(arg0, 8));
+      sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X));
+      sse_movlps(&p->func, dest, arg0);
+   }
+}
+
+static void emit_load3f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_2(p, dest, arg0);
+}
+
+static void emit_load3f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_1(p, dest, arg0);
+}
+
+static void emit_load2f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load2f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_1(p, dest, arg0);
+}
+
+static void emit_load1f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movss(&p->func, dest, arg0);
+}
+
+static void (*load[4][4])( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 ) = {
+   { emit_load1f_1, 
+     emit_load1f_1, 
+     emit_load1f_1, 
+     emit_load1f_1 },
+
+   { emit_load2f_1, 
+     emit_load2f_2, 
+     emit_load2f_2, 
+     emit_load2f_2 },
+
+   { emit_load3f_1, 
+     emit_load3f_2, 
+     emit_load3f_3, 
+     emit_load3f_3 },
+
+   { emit_load4f_1, 
+     emit_load4f_2, 
+     emit_load4f_3, 
+     emit_load4f_4 } 
+};
+
+static void emit_load( struct x86_program *p,
+		       struct x86_reg dest,
+		       GLuint sz,
+		       struct x86_reg src,
+		       GLuint src_sz)
+{
+   load[sz-1][src_sz-1](p, dest, src);
+}
+
+static void emit_store4f( struct x86_program *p, 			   
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   sse_movups(&p->func, dest, arg0);
+}
+
+static void emit_store3f( struct x86_program *p, 
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   if (p->outputs_safe) {
+      /* Emit the extra dword anyway.  This may hurt writecombining,
+       * may cause other problems.
+       */
+      sse_movups(&p->func, dest, arg0);
+   }
+   else {
+      /* Alternate strategy - emit two, shuffle, emit one.
+       */
+      sse_movlps(&p->func, dest, arg0);
+      sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */
+      sse_movss(&p->func, x86_make_disp(dest,8), arg0);
+   }
+}
+
+static void emit_store2f( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_store1f( struct x86_program *p, 
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   sse_movss(&p->func, dest, arg0);
+}
+
+
+static void (*store[4])( struct x86_program *p, 
+			 struct x86_reg dest,
+			 struct x86_reg arg0 ) = 
+{
+   emit_store1f, 
+   emit_store2f, 
+   emit_store3f, 
+   emit_store4f 
+};
+
+static void emit_store( struct x86_program *p,
+			struct x86_reg dest,
+			GLuint sz,
+			struct x86_reg temp )
+
+{
+   store[sz-1](p, dest, temp);
+}
+
+static void emit_pack_store_4ub( struct x86_program *p,
+				 struct x86_reg dest,
+				 struct x86_reg temp )
+{
+   /* Scale by 255.0
+    */
+   sse_mulps(&p->func, temp, p->chan0);
+
+   if (p->have_sse2) {
+      sse2_cvtps2dq(&p->func, temp, temp);
+      sse2_packssdw(&p->func, temp, temp);
+      sse2_packuswb(&p->func, temp, temp);
+      sse_movss(&p->func, dest, temp);
+   }
+   else {
+      struct x86_reg mmx0 = x86_make_reg(file_MMX, 0);
+      struct x86_reg mmx1 = x86_make_reg(file_MMX, 1);
+      sse_cvtps2pi(&p->func, mmx0, temp);
+      sse_movhlps(&p->func, temp, temp);
+      sse_cvtps2pi(&p->func, mmx1, temp);
+      mmx_packssdw(&p->func, mmx0, mmx1);
+      mmx_packuswb(&p->func, mmx0, mmx0);
+      mmx_movd(&p->func, dest, mmx0);
+   }
+}
+
+static GLint get_offset( const void *a, const void *b )
+{
+   return (const char *)b - (const char *)a;
+}
+
+/* Not much happens here.  Eventually use this function to try and
+ * avoid saving/reloading the source pointers each vertex (if some of
+ * them can fit in registers).
+ */
+static void get_src_ptr( struct x86_program *p,
+			 struct x86_reg srcREG,
+			 struct x86_reg vfREG,
+			 struct vf_attr *a )
+{
+   struct vertex_fetch *vf = p->vf;
+   struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr));
+
+   /* Load current a[j].inputptr
+    */
+   x86_mov(&p->func, srcREG, ptr_to_src);
+}
+
+static void update_src_ptr( struct x86_program *p,
+			 struct x86_reg srcREG,
+			 struct x86_reg vfREG,
+			 struct vf_attr *a )
+{
+   if (a->inputstride) {
+      struct vertex_fetch *vf = p->vf;
+      struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr));
+
+      /* add a[j].inputstride (hardcoded value - could just as easily
+       * pull the stride value from memory each time).
+       */
+      x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride));
+      
+      /* save new value of a[j].inputptr 
+       */
+      x86_mov(&p->func, ptr_to_src, srcREG);
+   }
+}
+
+
+/* Lots of hardcoding
+ *
+ * EAX -- pointer to current output vertex
+ * ECX -- pointer to current attribute 
+ * 
+ */
+static GLboolean build_vertex_emit( struct x86_program *p )
+{
+   struct vertex_fetch *vf = p->vf;
+   GLuint j = 0;
+
+   struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX);
+   struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX);
+   struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP);
+   struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI);
+   struct x86_reg temp = x86_make_reg(file_XMM, 0);
+   struct x86_reg vp0 = x86_make_reg(file_XMM, 1);
+   struct x86_reg vp1 = x86_make_reg(file_XMM, 2);
+   GLubyte *fixup, *label;
+
+   /* Push a few regs?
+    */
+   x86_push(&p->func, countEBP);
+   x86_push(&p->func, vfESI);
+
+
+   /* Get vertex count, compare to zero
+    */
+   x86_xor(&p->func, srcECX, srcECX);
+   x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2));
+   x86_cmp(&p->func, countEBP, srcECX);
+   fixup = x86_jcc_forward(&p->func, cc_E);
+
+   /* Initialize destination register. 
+    */
+   x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3));
+
+   /* Move argument 1 (vf) into a reg:
+    */
+   x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1));
+
+   
+   /* Possibly load vp0, vp1 for viewport calcs:
+    */
+   if (vf->allow_viewport_emits) {
+      sse_movups(&p->func, vp0, x86_make_disp(vfESI, get_offset(vf, &vf->vp[0])));
+      sse_movups(&p->func, vp1, x86_make_disp(vfESI, get_offset(vf, &vf->vp[4])));
+   }
+
+   /* always load, needed or not:
+    */
+   sse_movups(&p->func, p->chan0, x86_make_disp(vfESI, get_offset(vf, &vf->chan_scale[0])));
+   sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0])));
+
+   /* Note address for loop jump */
+   label = x86_get_label(&p->func);
+
+   /* Emit code for each of the attributes.  Currently routes
+    * everything through SSE registers, even when it might be more
+    * efficient to stick with regular old x86.  No optimization or
+    * other tricks - enough new ground to cover here just getting
+    * things working.
+    */
+   while (j < vf->attr_count) {
+      struct vf_attr *a = &vf->attr[j];
+      struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset);
+
+      /* Now, load an XMM reg from src, perhaps transform, then save.
+       * Could be shortcircuited in specific cases:
+       */
+      switch (a->format) {
+      case EMIT_1F:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 1, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_2F:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 2, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_3F:
+	 /* Potentially the worst case - hardcode 2+1 copying:
+	  */
+	 if (0) {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 3, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 else {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 2, temp);
+	    if (a->inputsize > 2) {
+	       emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1);
+	       emit_store(p, x86_make_disp(dest,8), 1, temp);
+	    }
+	    else {
+	       sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p));
+	    }
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 break;
+      case EMIT_4F:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 4, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_2F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 2, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_3F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 3, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 4, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_3F_XYW:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z));
+	 emit_store(p, dest, 3, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+
+      case EMIT_1UB_1F:	 
+	 /* Test for PAD3 + 1UB:
+	  */
+	 if (j > 0 &&
+	     a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3)
+	 {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize);
+	    sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X));
+	    emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 else {
+	    _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize );
+	    return GL_FALSE;
+	 }
+	 break;
+      case EMIT_3UB_3F_RGB:
+      case EMIT_3UB_3F_BGR:
+	 /* Test for 3UB + PAD1:
+	  */
+	 if (j == vf->attr_count - 1 ||
+	     a[1].vertoffset >= a->vertoffset + 4) {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    if (a->format == EMIT_3UB_3F_BGR)
+	       sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W));
+	    emit_pack_store_4ub(p, dest, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 /* Test for 3UB + 1UB:
+	  */
+	 else if (j < vf->attr_count - 1 &&
+		  a[1].format == EMIT_1UB_1F &&
+		  a[1].vertoffset == a->vertoffset + 3) {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    update_src_ptr(p, srcECX, vfESI, a);
+
+	    /* Make room for incoming value:
+	     */
+	    sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
+
+	    get_src_ptr(p, srcECX, vfESI, &a[1]);
+	    emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize);
+	    update_src_ptr(p, srcECX, vfESI, &a[1]);
+
+	    /* Rearrange and possibly do BGR conversion:
+	     */
+	    if (a->format == EMIT_3UB_3F_BGR)
+	       sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X));
+	    else
+	       sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X));
+
+	    emit_pack_store_4ub(p, dest, temp);
+	    j++;		/* NOTE: two attrs consumed */
+	 }
+	 else {
+	    _mesa_printf("Can't emit 3ub\n");
+	 }
+	 return GL_FALSE;	/* add this later */
+	 break;
+
+      case EMIT_4UB_4F_RGBA:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4UB_4F_BGRA:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4UB_4F_ARGB:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4UB_4F_ABGR:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4CHAN_4F_RGBA:
+	 switch (CHAN_TYPE) {
+	 case GL_UNSIGNED_BYTE:
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	    emit_pack_store_4ub(p, dest, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	    break;
+	 case GL_FLOAT:
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 4, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	    break;
+	 case GL_UNSIGNED_SHORT:
+	 default:
+	    _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE));
+	    return GL_FALSE;
+	 }
+	 break;
+      default:
+	 _mesa_printf("unknown a[%d].format %d\n", j, a->format);
+	 return GL_FALSE;	/* catch any new opcodes */
+      }
+      
+      /* Increment j by at least 1 - may have been incremented above also:
+       */
+      j++;
+   }
+
+   /* Next vertex:
+    */
+   x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride));
+
+   /* decr count, loop if not zero
+    */
+   x86_dec(&p->func, countEBP);
+   x86_test(&p->func, countEBP, countEBP); 
+   x86_jcc(&p->func, cc_NZ, label);
+
+   /* Exit mmx state?
+    */
+   if (p->func.need_emms)
+      mmx_emms(&p->func);
+
+   /* Land forward jump here:
+    */
+   x86_fixup_fwd_jump(&p->func, fixup);
+
+   /* Pop regs and return
+    */
+   x86_pop(&p->func, x86_get_base_reg(vfESI));
+   x86_pop(&p->func, countEBP);
+   x86_ret(&p->func);
+
+   vf->emit = (vf_emit_func)x86_get_func(&p->func);
+   return GL_TRUE;
+}
+
+
+
+void vf_generate_sse_emit( struct vertex_fetch *vf )
+{
+   struct x86_program p;   
+
+   if (!cpu_has_xmm) {
+      vf->codegen_emit = NULL;
+      return;
+   }
+
+   _mesa_memset(&p, 0, sizeof(p));
+
+   p.vf = vf;
+   p.inputs_safe = 0;		/* for now */
+   p.outputs_safe = 0;		/* for now */
+   p.have_sse2 = cpu_has_xmm2;
+   p.identity = x86_make_reg(file_XMM, 6);
+   p.chan0 = x86_make_reg(file_XMM, 7);
+
+   x86_init_func(&p.func);
+
+   if (build_vertex_emit(&p)) {
+      vf_register_fastpath( vf, GL_TRUE );
+   }
+   else {
+      /* Note the failure so that we don't keep trying to codegen an
+       * impossible state:
+       */
+      vf_register_fastpath( vf, GL_FALSE );
+      x86_release_func(&p.func);
+   }
+}
+
+#else
+
+void vf_generate_sse_emit( struct vertex_fetch *vf )
+{
+   /* Dummy version for when USE_SSE_ASM not defined */
+}
+
+#endif
diff --git a/src/mesa/x86/Makefile b/src/mesa/x86/Makefile
index dc8c7f355eb..aa49a9134aa 100644
--- a/src/mesa/x86/Makefile
+++ b/src/mesa/x86/Makefile
@@ -21,7 +21,7 @@ clean:
 
 
 gen_matypes: gen_matypes.c
-	$(CC) $(INCLUDE_DIRS) $(CFLAGS) gen_matypes.c -o gen_matypes
+	$(HOST_CC) $(INCLUDE_DIRS) $(HOST_CFLAGS) gen_matypes.c -o gen_matypes
 
 # need some special rules here, unfortunately
 matypes.h: ../main/mtypes.h ../tnl/t_context.h gen_matypes
diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c
index e68f8dfec13..5aedf5b04be 100644
--- a/src/mesa/x86/rtasm/x86sse.c
+++ b/src/mesa/x86/rtasm/x86sse.c
@@ -1,3 +1,4 @@
+#ifdef USE_X86_ASM
 #if defined(__i386__) || defined(__386__)
 
 #include "main/imports.h"
@@ -290,7 +291,7 @@ void x86_call( struct x86_function *p, void (*label)())
 void x86_call( struct x86_function *p, struct x86_reg reg)
 {
    emit_1ub(p, 0xff);
-   emit_modrm(p, reg, reg);
+   emit_modrm_noreg(p, 2, reg);
 }
 #endif
 
@@ -1192,3 +1193,9 @@ void x86sse_dummy( void )
 }
 
 #endif
+
+#else  /* USE_X86_ASM */
+
+int x86sse_c_dummy_var; /* silence warning */
+
+#endif /* USE_X86_ASM */
author	Brian Paul <[email protected]>	2009-02-10 16:44:02 -0700
committer	Brian Paul <[email protected]>	2009-02-10 16:44:02 -0700
commit	5340b6dff73a0a23531ce2a5f28fba8303adab6e (patch)
tree	b141fc3648568dd8b941c966059e6ed32a8bd0ad /src/mesa
parent	9fd26daec24f21dbe17afcb2e2ab272667ee9a69 (diff)
parent	ee4c921b65fb76998711f3c40330505cbc49a0e0 (diff)