Merge remote branch 'origin/master' into lp-binning

Conflicts:
	src/gallium/auxiliary/util/u_surface.c
	src/gallium/drivers/llvmpipe/Makefile
	src/gallium/drivers/llvmpipe/SConscript
	src/gallium/drivers/llvmpipe/lp_bld_arit.c
	src/gallium/drivers/llvmpipe/lp_bld_flow.c
	src/gallium/drivers/llvmpipe/lp_bld_interp.c
	src/gallium/drivers/llvmpipe/lp_clear.c
	src/gallium/drivers/llvmpipe/lp_context.c
	src/gallium/drivers/llvmpipe/lp_context.h
	src/gallium/drivers/llvmpipe/lp_draw_arrays.c
	src/gallium/drivers/llvmpipe/lp_jit.c
	src/gallium/drivers/llvmpipe/lp_jit.h
	src/gallium/drivers/llvmpipe/lp_prim_vbuf.c
	src/gallium/drivers/llvmpipe/lp_setup.c
	src/gallium/drivers/llvmpipe/lp_setup_point.c
	src/gallium/drivers/llvmpipe/lp_state.h
	src/gallium/drivers/llvmpipe/lp_state_blend.c
	src/gallium/drivers/llvmpipe/lp_state_derived.c
	src/gallium/drivers/llvmpipe/lp_state_fs.c
	src/gallium/drivers/llvmpipe/lp_state_sampler.c
	src/gallium/drivers/llvmpipe/lp_state_surface.c
	src/gallium/drivers/llvmpipe/lp_tex_cache.c
	src/gallium/drivers/llvmpipe/lp_tex_cache.h
	src/gallium/drivers/llvmpipe/lp_tex_sample.h
	src/gallium/drivers/llvmpipe/lp_tile_cache.c

144 files changed, 8326 insertions, 6304 deletions

diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index 5446eb68a9..e3af41c6e0 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -1,12 +1,177 @@
-# src/gallium/auxiliary/Makefile
 TOP = ../../..
 include $(TOP)/configs/current
 
-SUBDIRS = $(GALLIUM_AUXILIARY_DIRS)
+LIBNAME = gallium
 
-default install clean:
-	@for dir in $(SUBDIRS) ; do \
-		if [ -d $$dir ] ; then \
-			(cd $$dir && $(MAKE) $@) || exit 1; \
-		fi \
-	done
+C_SOURCES = \
+	cso_cache/cso_context.c \
+	cso_cache/cso_cache.c \
+	cso_cache/cso_hash.c \
+	draw/draw_context.c \
+	draw/draw_gs.c \
+	draw/draw_pipe.c \
+	draw/draw_pipe_aaline.c \
+	draw/draw_pipe_aapoint.c \
+	draw/draw_pipe_clip.c \
+	draw/draw_pipe_cull.c \
+	draw/draw_pipe_flatshade.c \
+	draw/draw_pipe_offset.c \
+	draw/draw_pipe_pstipple.c \
+	draw/draw_pipe_stipple.c \
+	draw/draw_pipe_twoside.c \
+	draw/draw_pipe_unfilled.c \
+	draw/draw_pipe_util.c \
+	draw/draw_pipe_validate.c \
+	draw/draw_pipe_vbuf.c \
+	draw/draw_pipe_wide_line.c \
+	draw/draw_pipe_wide_point.c \
+	draw/draw_pt.c \
+	draw/draw_pt_elts.c \
+	draw/draw_pt_emit.c \
+	draw/draw_pt_fetch.c \
+	draw/draw_pt_fetch_emit.c \
+	draw/draw_pt_fetch_shade_emit.c \
+	draw/draw_pt_fetch_shade_pipeline.c \
+	draw/draw_pt_post_vs.c \
+	draw/draw_pt_util.c \
+	draw/draw_pt_varray.c \
+	draw/draw_pt_vcache.c \
+	draw/draw_vertex.c \
+	draw/draw_vs.c \
+	draw/draw_vs_varient.c \
+	draw/draw_vs_aos.c \
+	draw/draw_vs_aos_io.c \
+	draw/draw_vs_aos_machine.c \
+	draw/draw_vs_exec.c \
+	draw/draw_vs_llvm.c \
+	draw/draw_vs_ppc.c \
+	draw/draw_vs_sse.c \
+	indices/u_indices_gen.c \
+	indices/u_unfilled_gen.c \
+	pipebuffer/pb_buffer_fenced.c \
+	pipebuffer/pb_buffer_malloc.c \
+	pipebuffer/pb_bufmgr_alt.c \
+	pipebuffer/pb_bufmgr_cache.c \
+	pipebuffer/pb_bufmgr_debug.c \
+	pipebuffer/pb_bufmgr_fenced.c \
+	pipebuffer/pb_bufmgr_mm.c \
+	pipebuffer/pb_bufmgr_ondemand.c \
+	pipebuffer/pb_bufmgr_pool.c \
+	pipebuffer/pb_bufmgr_slab.c \
+	pipebuffer/pb_validate.c \
+	rbug/rbug_connection.c \
+	rbug/rbug_core.c \
+	rbug/rbug_texture.c \
+	rbug/rbug_context.c \
+	rbug/rbug_shader.c \
+	rbug/rbug_demarshal.c \
+	rtasm/rtasm_cpu.c \
+	rtasm/rtasm_execmem.c \
+	rtasm/rtasm_x86sse.c \
+	rtasm/rtasm_ppc.c \
+	rtasm/rtasm_ppc_spe.c \
+	tgsi/tgsi_sanity.c \
+	tgsi/tgsi_build.c \
+	tgsi/tgsi_dump.c \
+	tgsi/tgsi_exec.c \
+	tgsi/tgsi_info.c \
+	tgsi/tgsi_iterate.c \
+	tgsi/tgsi_parse.c \
+	tgsi/tgsi_ppc.c \
+	tgsi/tgsi_scan.c \
+	tgsi/tgsi_sse2.c \
+	tgsi/tgsi_text.c \
+	tgsi/tgsi_transform.c \
+	tgsi/tgsi_ureg.c \
+	tgsi/tgsi_util.c \
+	translate/translate_generic.c \
+	translate/translate_sse.c \
+	translate/translate.c \
+	translate/translate_cache.c \
+	util/u_debug.c \
+	util/u_debug_dump.c \
+	util/u_debug_symbol.c \
+	util/u_debug_stack.c \
+	util/u_blit.c \
+	util/u_blitter.c \
+	util/u_cache.c \
+	util/u_cpu_detect.c \
+	util/u_dl.c \
+	util/u_draw_quad.c \
+	util/u_format.c \
+	util/u_format_access.c \
+	util/u_format_table.c \
+	util/u_gen_mipmap.c \
+	util/u_handle_table.c \
+	util/u_hash_table.c \
+	util/u_hash.c \
+	util/u_keymap.c \
+	util/u_linear.c \
+	util/u_network.c \
+	util/u_math.c \
+	util/u_mm.c \
+	util/u_rect.c \
+	util/u_simple_shaders.c \
+	util/u_snprintf.c \
+	util/u_stream_stdc.c \
+	util/u_stream_wd.c \
+	util/u_surface.c \
+	util/u_texture.c \
+	util/u_tile.c \
+	util/u_time.c \
+	util/u_timed_winsys.c \
+	util/u_upload_mgr.c \
+	util/u_simple_screen.c \
+	vl/vl_bitstream_parser.c \
+	vl/vl_mpeg12_mc_renderer.c \
+	vl/vl_compositor.c \
+	vl/vl_csc.c \
+	vl/vl_shader_build.c
+
+GALLIVM_SOURCES = \
+	gallivm/gallivm.cpp  \
+	gallivm/gallivm_cpu.cpp \
+	gallivm/instructions.cpp  \
+	gallivm/loweringpass.cpp \
+	gallivm/tgsitollvm.cpp \
+	gallivm/storage.cpp \
+	gallivm/storagesoa.cpp \
+	gallivm/instructionssoa.cpp
+
+INC_SOURCES = \
+	gallivm/gallivm_builtins.cpp \
+	gallivm/gallivmsoabuiltins.cpp
+
+# XXX: gallivm doesn't build correctly so disable for now
+#ifeq ($(MESA_LLVM),1)
+#DEFINES += -D__STDC_CONSTANT_MACROS -D__STDC_LIMIT_MACROS
+#CPP_SOURCES += \
+#	$(GALLIVM_SOURCES)
+#endif
+
+
+include ../Makefile.template
+
+
+gallivm/gallivm_builtins.cpp: gallivm/llvm_builtins.c
+	clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp1.bin
+	(echo "static const unsigned char llvm_builtins_data[] = {"; od -txC temp1.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/,0x00};/") >$@
+	rm temp1.bin
+
+gallivm/gallivmsoabuiltins.cpp: gallivm/soabuiltins.c
+	clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp2.bin
+	(echo "static const unsigned char soabuiltins_data[] = {"; od -txC temp2.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/,0x00};/") >$@
+	rm temp2.bin
+
+
+indices/u_indices_gen.c: indices/u_indices_gen.py
+	python $< > $@
+
+indices/u_unfilled_gen.c: indices/u_unfilled_gen.py
+	python $< > $@
+
+util/u_format_table.c: util/u_format_table.py util/u_format_parse.py util/u_format.csv
+	python util/u_format_table.py util/u_format.csv > $@
+
+util/u_format_access.c: util/u_format_access.py util/u_format_parse.py util/u_format.csv
+	python util/u_format_access.py util/u_format.csv > $@
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
new file mode 100644
index 0000000000..782eb53386
--- /dev/null
+++ b/src/gallium/auxiliary/SConscript
@@ -0,0 +1,185 @@
+Import('*')
+
+from sys import executable as python_cmd
+
+env.Append(CPPPATH = [
+    'indices',
+    'util',
+])
+
+env.CodeGenerate(
+    target = 'indices/u_indices_gen.c', 
+    script = 'indices/u_indices_gen.py', 
+    source = [],
+    command = python_cmd + ' $SCRIPT > $TARGET'
+)
+
+env.CodeGenerate(
+    target = 'indices/u_unfilled_gen.c', 
+    script = 'indices/u_unfilled_gen.py', 
+    source = [],
+    command = python_cmd + ' $SCRIPT > $TARGET'
+)
+
+env.CodeGenerate(
+    target = 'util/u_format_table.c',
+    script = 'util/u_format_table.py',
+    source = ['util/u_format.csv'],
+    command = 'python $SCRIPT $SOURCE > $TARGET'
+)
+
+env.CodeGenerate(
+    target = 'util/u_format_access.c',
+    script = 'util/u_format_access.py',
+    source = ['util/u_format.csv'],
+    command = 'python $SCRIPT $SOURCE > $TARGET'
+)
+
+source = [
+    'cso_cache/cso_context.c',
+    'cso_cache/cso_cache.c',
+    'cso_cache/cso_hash.c',
+    'draw/draw_context.c',
+    'draw/draw_pipe.c',
+    'draw/draw_pipe_aaline.c',
+    'draw/draw_pipe_aapoint.c',
+    'draw/draw_pipe_clip.c',
+    'draw/draw_pipe_cull.c',
+    'draw/draw_pipe_flatshade.c',
+    'draw/draw_pipe_offset.c',
+    'draw/draw_pipe_pstipple.c',
+    'draw/draw_pipe_stipple.c',
+    'draw/draw_pipe_twoside.c',
+    'draw/draw_pipe_unfilled.c',
+    'draw/draw_pipe_util.c',
+    'draw/draw_pipe_validate.c',
+    'draw/draw_pipe_vbuf.c',
+    'draw/draw_pipe_wide_line.c',
+    'draw/draw_pipe_wide_point.c',
+    'draw/draw_pt.c',
+    'draw/draw_pt_elts.c',
+    'draw/draw_pt_emit.c',
+    'draw/draw_pt_fetch.c',
+    'draw/draw_pt_fetch_emit.c',
+    'draw/draw_pt_fetch_shade_emit.c',
+    'draw/draw_pt_fetch_shade_pipeline.c',
+    'draw/draw_pt_post_vs.c',
+    'draw/draw_pt_util.c',
+    'draw/draw_pt_varray.c',
+    'draw/draw_pt_vcache.c',
+    'draw/draw_vertex.c',
+    'draw/draw_vs.c',
+    'draw/draw_vs_aos.c',
+    'draw/draw_vs_aos_io.c',
+    'draw/draw_vs_aos_machine.c',
+    'draw/draw_vs_exec.c',
+    'draw/draw_vs_llvm.c',
+    'draw/draw_vs_ppc.c',
+    'draw/draw_vs_sse.c',
+    'draw/draw_vs_varient.c',
+    'draw/draw_gs.c',
+    #'indices/u_indices.c',
+    #'indices/u_unfilled_indices.c',
+    'indices/u_indices_gen.c',
+    'indices/u_unfilled_gen.c',
+    'pipebuffer/pb_buffer_fenced.c',
+    'pipebuffer/pb_buffer_malloc.c',
+    'pipebuffer/pb_bufmgr_alt.c',
+    'pipebuffer/pb_bufmgr_cache.c',
+    'pipebuffer/pb_bufmgr_debug.c',
+    'pipebuffer/pb_bufmgr_fenced.c',
+    'pipebuffer/pb_bufmgr_mm.c',
+    'pipebuffer/pb_bufmgr_ondemand.c',
+    'pipebuffer/pb_bufmgr_pool.c',
+    'pipebuffer/pb_bufmgr_slab.c',
+    'pipebuffer/pb_validate.c',
+    'rbug/rbug_core.c',
+    'rbug/rbug_shader.c',
+    'rbug/rbug_context.c',
+    'rbug/rbug_texture.c',
+    'rbug/rbug_demarshal.c',
+    'rbug/rbug_connection.c',
+    'rtasm/rtasm_cpu.c',
+    'rtasm/rtasm_execmem.c',
+    'rtasm/rtasm_x86sse.c',
+    'rtasm/rtasm_ppc.c',
+    'rtasm/rtasm_ppc_spe.c',
+    'tgsi/tgsi_build.c',
+    'tgsi/tgsi_dump.c',
+    'tgsi/tgsi_dump_c.c',
+    'tgsi/tgsi_exec.c',
+    'tgsi/tgsi_info.c',
+    'tgsi/tgsi_iterate.c',
+    'tgsi/tgsi_parse.c',
+    'tgsi/tgsi_sanity.c',
+    'tgsi/tgsi_scan.c',
+    'tgsi/tgsi_ppc.c',
+    'tgsi/tgsi_sse2.c',
+    'tgsi/tgsi_text.c',
+    'tgsi/tgsi_transform.c',
+    'tgsi/tgsi_ureg.c',
+    'tgsi/tgsi_util.c',
+    'translate/translate_generic.c',
+    'translate/translate_sse.c',
+    'translate/translate.c',
+    'translate/translate_cache.c',
+    'util/u_bitmask.c',
+    'util/u_blit.c',
+    'util/u_blitter.c',
+    'util/u_cache.c',
+    'util/u_cpu_detect.c',
+    'util/u_debug.c',
+    'util/u_debug_dump.c',
+    'util/u_debug_memory.c',
+    'util/u_debug_stack.c',
+    'util/u_debug_symbol.c',
+    'util/u_dl.c',
+    'util/u_draw_quad.c',
+    'util/u_format.c',
+    'util/u_format_access.c',
+    'util/u_format_table.c',
+    'util/u_gen_mipmap.c',
+    'util/u_handle_table.c',
+    'util/u_hash.c',
+    'util/u_hash_table.c',
+    'util/u_keymap.c',
+    'util/u_network.c',
+    'util/u_math.c',
+    'util/u_mm.c',
+    'util/u_rect.c',
+    'util/u_simple_shaders.c',
+    'util/u_snprintf.c',
+    'util/u_stream_stdc.c',
+    'util/u_stream_wd.c',
+    'util/u_surface.c',
+    'util/u_texture.c',
+    'util/u_tile.c',
+    'util/u_time.c',
+    'util/u_timed_winsys.c',
+    'util/u_upload_mgr.c',
+    'util/u_simple_screen.c',
+    'vl/vl_bitstream_parser.c',
+    'vl/vl_mpeg12_mc_renderer.c',
+    'vl/vl_compositor.c',
+    'vl/vl_csc.c',
+    'vl/vl_shader_build.c',
+]
+
+if env['llvm']:
+    source += [
+        'gallivm/gallivm.cpp',
+        'gallivm/gallivm_cpu.cpp',
+        'gallivm/instructions.cpp',
+        'gallivm/loweringpass.cpp',
+        'gallivm/tgsitollvm.cpp',
+        'gallivm/storage.cpp',
+        'gallivm/storagesoa.cpp',
+        'gallivm/instructionssoa.cpp',
+    ]
+
+gallium = env.ConvenienceLibrary(
+    target = 'gallium',
+    source = source,
+)
+
+Export('gallium')
diff --git a/src/gallium/auxiliary/cso_cache/Makefile b/src/gallium/auxiliary/cso_cache/Makefile
deleted file mode 100644
index 8726afcd94..0000000000
--- a/src/gallium/auxiliary/cso_cache/Makefile
+++ /dev/null
@@ -1,11 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = cso_cache
-
-C_SOURCES = \
-	cso_context.c \
-	cso_cache.c \
-	cso_hash.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/cso_cache/SConscript b/src/gallium/auxiliary/cso_cache/SConscript
deleted file mode 100644
index 651e68a191..0000000000
--- a/src/gallium/auxiliary/cso_cache/SConscript
+++ /dev/null
@@ -1,11 +0,0 @@
-Import('*')
-
-cso_cache = env.ConvenienceLibrary(
-	target = 'cso_cache',
-	source = [
-		'cso_context.c',
-		'cso_cache.c',
-		'cso_hash.c',
-	])
-
-auxiliaries.insert(0, cso_cache)
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index 4f13b3e2ba..2b16332e14 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -42,6 +42,7 @@
 #include "cso_cache/cso_context.h"
 #include "cso_cache/cso_cache.h"
 #include "cso_cache/cso_hash.h"
+#include "cso_context.h"
 
 struct cso_context {
    struct pipe_context *pipe;
@@ -50,28 +51,43 @@ struct cso_context {
    struct {
       void *samplers[PIPE_MAX_SAMPLERS];
       unsigned nr_samplers;
+
+      void *vertex_samplers[PIPE_MAX_VERTEX_SAMPLERS];
+      unsigned nr_vertex_samplers;
    } hw;
 
    void *samplers[PIPE_MAX_SAMPLERS];
    unsigned nr_samplers;
 
+   void *vertex_samplers[PIPE_MAX_VERTEX_SAMPLERS];
+   unsigned nr_vertex_samplers;
+
    unsigned nr_samplers_saved;
    void *samplers_saved[PIPE_MAX_SAMPLERS];
 
+   unsigned nr_vertex_samplers_saved;
+   void *vertex_samplers_saved[PIPE_MAX_VERTEX_SAMPLERS];
+
    struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
    uint nr_textures;
 
+   struct pipe_texture *vertex_textures[PIPE_MAX_VERTEX_SAMPLERS];
+   uint nr_vertex_textures;
+
    uint nr_textures_saved;
    struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
 
+   uint nr_vertex_textures_saved;
+   struct pipe_texture *vertex_textures_saved[PIPE_MAX_SAMPLERS];
+
    /** Current and saved state.
     * The saved state is used as a 1-deep stack.
     */
    void *blend, *blend_saved;
    void *depth_stencil, *depth_stencil_saved;
    void *rasterizer, *rasterizer_saved;
-   void *fragment_shader, *fragment_shader_saved;
-   void *vertex_shader, *vertex_shader_saved;
+   void *fragment_shader, *fragment_shader_saved, *geometry_shader;
+   void *vertex_shader, *vertex_shader_saved, *geometry_shader_saved;
 
    struct pipe_framebuffer_state fb, fb_saved;
    struct pipe_viewport_state vp, vp_saved;
@@ -244,7 +260,9 @@ void cso_release_all( struct cso_context *ctx )
    if (ctx->pipe) {
       ctx->pipe->bind_blend_state( ctx->pipe, NULL );
       ctx->pipe->bind_rasterizer_state( ctx->pipe, NULL );
-      ctx->pipe->bind_sampler_states( ctx->pipe, 0, NULL );
+      ctx->pipe->bind_fragment_sampler_states( ctx->pipe, 0, NULL );
+      if (ctx->pipe->bind_vertex_sampler_states)
+         ctx->pipe->bind_vertex_sampler_states(ctx->pipe, 0, NULL);
       ctx->pipe->bind_depth_stencil_alpha_state( ctx->pipe, NULL );
       ctx->pipe->bind_fs_state( ctx->pipe, NULL );
       ctx->pipe->bind_vs_state( ctx->pipe, NULL );
@@ -255,6 +273,11 @@ void cso_release_all( struct cso_context *ctx )
       pipe_texture_reference(&ctx->textures_saved[i], NULL);
    }
 
+   for (i = 0; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
+      pipe_texture_reference(&ctx->vertex_textures[i], NULL);
+      pipe_texture_reference(&ctx->vertex_textures_saved[i], NULL);
+   }
+
    free_framebuffer_state(&ctx->fb);
    free_framebuffer_state(&ctx->fb_saved);
 
@@ -378,6 +401,46 @@ enum pipe_error cso_single_sampler(struct cso_context *ctx,
    return PIPE_OK;
 }
 
+enum pipe_error
+cso_single_vertex_sampler(struct cso_context *ctx,
+                          unsigned idx,
+                          const struct pipe_sampler_state *templ)
+{
+   void *handle = NULL;
+
+   if (templ != NULL) {
+      unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state));
+      struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                          hash_key, CSO_SAMPLER,
+                                                          (void*)templ);
+
+      if (cso_hash_iter_is_null(iter)) {
+         struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler));
+         if (!cso)
+            return PIPE_ERROR_OUT_OF_MEMORY;
+
+         memcpy(&cso->state, templ, sizeof(*templ));
+         cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state);
+         cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state;
+         cso->context = ctx->pipe;
+
+         iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso);
+         if (cso_hash_iter_is_null(iter)) {
+            FREE(cso);
+            return PIPE_ERROR_OUT_OF_MEMORY;
+         }
+
+         handle = cso->data;
+      }
+      else {
+         handle = ((struct cso_sampler *)cso_hash_iter_data(iter))->data;
+      }
+   }
+
+   ctx->vertex_samplers[idx] = handle;
+   return PIPE_OK;
+}
+
 void cso_single_sampler_done( struct cso_context *ctx )
 {
    unsigned i;
@@ -398,7 +461,36 @@ void cso_single_sampler_done( struct cso_context *ctx )
       memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *));
       ctx->hw.nr_samplers = ctx->nr_samplers;
 
-      ctx->pipe->bind_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers);
+      ctx->pipe->bind_fragment_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers);
+   }
+}
+
+void
+cso_single_vertex_sampler_done(struct cso_context *ctx)
+{
+   unsigned i;
+
+   /* find highest non-null sampler */
+   for (i = PIPE_MAX_VERTEX_SAMPLERS; i > 0; i--) {
+      if (ctx->vertex_samplers[i - 1] != NULL)
+         break;
+   }
+
+   ctx->nr_vertex_samplers = i;
+
+   if (ctx->hw.nr_vertex_samplers != ctx->nr_vertex_samplers ||
+       memcmp(ctx->hw.vertex_samplers,
+              ctx->vertex_samplers,
+              ctx->nr_vertex_samplers * sizeof(void *)) != 0) 
+   {
+      memcpy(ctx->hw.vertex_samplers,
+             ctx->vertex_samplers,
+             ctx->nr_vertex_samplers * sizeof(void *));
+      ctx->hw.nr_vertex_samplers = ctx->nr_vertex_samplers;
+
+      ctx->pipe->bind_vertex_sampler_states(ctx->pipe,
+                                            ctx->nr_vertex_samplers,
+                                            ctx->vertex_samplers);
    }
 }
 
@@ -447,6 +539,21 @@ void cso_restore_samplers(struct cso_context *ctx)
    cso_single_sampler_done( ctx );
 }
 
+void
+cso_save_vertex_samplers(struct cso_context *ctx)
+{
+   ctx->nr_vertex_samplers_saved = ctx->nr_vertex_samplers;
+   memcpy(ctx->vertex_samplers_saved, ctx->vertex_samplers, sizeof(ctx->vertex_samplers));
+}
+
+void
+cso_restore_vertex_samplers(struct cso_context *ctx)
+{
+   ctx->nr_vertex_samplers = ctx->nr_vertex_samplers_saved;
+   memcpy(ctx->vertex_samplers, ctx->vertex_samplers_saved, sizeof(ctx->vertex_samplers));
+   cso_single_vertex_sampler_done(ctx);
+}
+
 
 enum pipe_error cso_set_sampler_textures( struct cso_context *ctx,
                                           uint count,
@@ -461,7 +568,7 @@ enum pipe_error cso_set_sampler_textures( struct cso_context *ctx,
    for ( ; i < PIPE_MAX_SAMPLERS; i++)
       pipe_texture_reference(&ctx->textures[i], NULL);
 
-   ctx->pipe->set_sampler_textures(ctx->pipe, count, textures);
+   ctx->pipe->set_fragment_sampler_textures(ctx->pipe, count, textures);
 
    return PIPE_OK;
 }
@@ -491,13 +598,71 @@ void cso_restore_sampler_textures( struct cso_context *ctx )
    for ( ; i < PIPE_MAX_SAMPLERS; i++)
       pipe_texture_reference(&ctx->textures[i], NULL);
 
-   ctx->pipe->set_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures);
+   ctx->pipe->set_fragment_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures);
 
    ctx->nr_textures_saved = 0;
 }
 
 
 
+enum pipe_error
+cso_set_vertex_sampler_textures(struct cso_context *ctx,
+                                uint count,
+                                struct pipe_texture **textures)
+{
+   uint i;
+
+   ctx->nr_vertex_textures = count;
+
+   for (i = 0; i < count; i++) {
+      pipe_texture_reference(&ctx->vertex_textures[i], textures[i]);
+   }
+   for ( ; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
+      pipe_texture_reference(&ctx->vertex_textures[i], NULL);
+   }
+
+   ctx->pipe->set_vertex_sampler_textures(ctx->pipe, count, textures);
+
+   return PIPE_OK;
+}
+
+void
+cso_save_vertex_sampler_textures(struct cso_context *ctx)
+{
+   uint i;
+
+   ctx->nr_vertex_textures_saved = ctx->nr_vertex_textures;
+   for (i = 0; i < ctx->nr_vertex_textures; i++) {
+      assert(!ctx->vertex_textures_saved[i]);
+      pipe_texture_reference(&ctx->vertex_textures_saved[i], ctx->vertex_textures[i]);
+   }
+}
+
+void
+cso_restore_vertex_sampler_textures(struct cso_context *ctx)
+{
+   uint i;
+
+   ctx->nr_vertex_textures = ctx->nr_vertex_textures_saved;
+
+   for (i = 0; i < ctx->nr_vertex_textures; i++) {
+      pipe_texture_reference(&ctx->vertex_textures[i], NULL);
+      ctx->vertex_textures[i] = ctx->vertex_textures_saved[i];
+      ctx->vertex_textures_saved[i] = NULL;
+   }
+   for ( ; i < PIPE_MAX_VERTEX_SAMPLERS; i++) {
+      pipe_texture_reference(&ctx->vertex_textures[i], NULL);
+   }
+
+   ctx->pipe->set_vertex_sampler_textures(ctx->pipe,
+                                          ctx->nr_vertex_textures,
+                                          ctx->vertex_textures);
+
+   ctx->nr_vertex_textures_saved = 0;
+}
+
+
+
 enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx,
                                             const struct pipe_depth_stencil_alpha_state *templ)
 {
@@ -863,3 +1028,38 @@ enum pipe_error cso_set_blend_color(struct cso_context *ctx,
    }
    return PIPE_OK;
 }
+
+enum pipe_error cso_set_geometry_shader_handle(struct cso_context *ctx,
+                                               void *handle)
+{
+   if (ctx->geometry_shader != handle) {
+      ctx->geometry_shader = handle;
+      ctx->pipe->bind_gs_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_delete_geometry_shader(struct cso_context *ctx, void *handle)
+{
+    if (handle == ctx->geometry_shader) {
+      /* unbind before deleting */
+      ctx->pipe->bind_gs_state(ctx->pipe, NULL);
+      ctx->geometry_shader = NULL;
+   }
+   ctx->pipe->delete_gs_state(ctx->pipe, handle);
+}
+
+void cso_save_geometry_shader(struct cso_context *ctx)
+{
+   assert(!ctx->geometry_shader_saved);
+   ctx->geometry_shader_saved = ctx->geometry_shader;
+}
+
+void cso_restore_geometry_shader(struct cso_context *ctx)
+{
+   if (ctx->geometry_shader_saved != ctx->geometry_shader) {
+      ctx->pipe->bind_gs_state(ctx->pipe, ctx->geometry_shader_saved);
+      ctx->geometry_shader = ctx->geometry_shader_saved;
+   }
+   ctx->geometry_shader_saved = NULL;
+}
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h
index b04e98bfa1..b9e313e32d 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.h
+++ b/src/gallium/auxiliary/cso_cache/cso_context.h
@@ -31,7 +31,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 
 
 #ifdef	__cplusplus
@@ -84,6 +84,20 @@ enum pipe_error cso_single_sampler( struct cso_context *cso,
 
 void cso_single_sampler_done( struct cso_context *cso );
 
+void
+cso_save_vertex_samplers(struct cso_context *cso);
+
+void
+cso_restore_vertex_samplers(struct cso_context *cso);
+
+enum pipe_error
+cso_single_vertex_sampler(struct cso_context *cso,
+                          unsigned nr,
+                          const struct pipe_sampler_state *states);
+
+void
+cso_single_vertex_sampler_done(struct cso_context *cso);
+
 
 
 enum pipe_error cso_set_sampler_textures( struct cso_context *cso,
@@ -94,6 +108,17 @@ void cso_restore_sampler_textures( struct cso_context *cso );
 
 
 
+enum pipe_error
+cso_set_vertex_sampler_textures(struct cso_context *cso,
+                                uint count,
+                                struct pipe_texture **textures);
+void
+cso_save_vertex_sampler_textures(struct cso_context *cso);
+void
+cso_restore_vertex_sampler_textures(struct cso_context *cso);
+
+
+
 /* These aren't really sensible -- most of the time the api provides
  * object semantics for shaders anyway, and the cases where it doesn't
  * (eg mesa's internall-generated texenv programs), it will be up to
@@ -121,6 +146,13 @@ void cso_save_vertex_shader(struct cso_context *cso);
 void cso_restore_vertex_shader(struct cso_context *cso);
 
 
+enum pipe_error cso_set_geometry_shader_handle(struct cso_context *ctx,
+                                               void *handle);
+void cso_delete_geometry_shader(struct cso_context *ctx, void *handle);
+void cso_save_geometry_shader(struct cso_context *cso);
+void cso_restore_geometry_shader(struct cso_context *cso);
+
+
 
 enum pipe_error cso_set_framebuffer(struct cso_context *cso,
                                     const struct pipe_framebuffer_state *fb);
diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile
deleted file mode 100644
index 5041dcc072..0000000000
--- a/src/gallium/auxiliary/draw/Makefile
+++ /dev/null
@@ -1,46 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = draw
-
-C_SOURCES = \
-	draw_context.c \
-	draw_pipe.c \
-	draw_pipe_aaline.c \
-	draw_pipe_aapoint.c \
-	draw_pipe_clip.c \
-	draw_pipe_cull.c \
-	draw_pipe_flatshade.c \
-	draw_pipe_offset.c \
-	draw_pipe_pstipple.c \
-	draw_pipe_stipple.c \
-	draw_pipe_twoside.c \
-	draw_pipe_unfilled.c \
-	draw_pipe_util.c \
-	draw_pipe_validate.c \
-	draw_pipe_vbuf.c \
-	draw_pipe_wide_line.c \
-	draw_pipe_wide_point.c \
-	draw_pt.c \
-	draw_pt_elts.c \
-	draw_pt_emit.c \
-	draw_pt_fetch.c \
-	draw_pt_fetch_emit.c \
-	draw_pt_fetch_shade_emit.c \
-	draw_pt_fetch_shade_pipeline.c \
-	draw_pt_post_vs.c \
-        draw_pt_util.c \
-        draw_pt_varray.c \
-	draw_pt_vcache.c \
-	draw_vertex.c \
-	draw_vs.c \
-	draw_vs_varient.c \
-	draw_vs_aos.c \
-	draw_vs_aos_io.c \
-	draw_vs_aos_machine.c \
-	draw_vs_exec.c \
-	draw_vs_llvm.c \
-	draw_vs_ppc.c  \
-	draw_vs_sse.c 
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript
deleted file mode 100644
index 5f05aa324a..0000000000
--- a/src/gallium/auxiliary/draw/SConscript
+++ /dev/null
@@ -1,46 +0,0 @@
-Import('*')
-
-draw = env.ConvenienceLibrary(
-	target = 'draw',
-	source = [
-		'draw_context.c',
-		'draw_pipe.c',
-		'draw_pipe_aaline.c',
-		'draw_pipe_aapoint.c',
-		'draw_pipe_clip.c',
-		'draw_pipe_cull.c',
-		'draw_pipe_flatshade.c',
-		'draw_pipe_offset.c',
-		'draw_pipe_pstipple.c',
-		'draw_pipe_stipple.c',
-		'draw_pipe_twoside.c',
-		'draw_pipe_unfilled.c',
-		'draw_pipe_util.c',
-		'draw_pipe_validate.c',
-		'draw_pipe_vbuf.c',
-		'draw_pipe_wide_line.c',
-		'draw_pipe_wide_point.c',
-		'draw_pt.c',
-		'draw_pt_elts.c',
-		'draw_pt_emit.c',
-		'draw_pt_fetch.c',
-		'draw_pt_fetch_emit.c',
-		'draw_pt_fetch_shade_emit.c',
-		'draw_pt_fetch_shade_pipeline.c',
-		'draw_pt_post_vs.c',
-		'draw_pt_util.c',
-		'draw_pt_varray.c',
-		'draw_pt_vcache.c',
-		'draw_vertex.c',
-		'draw_vs.c',
-		'draw_vs_aos.c',
-		'draw_vs_aos_io.c',
-		'draw_vs_aos_machine.c',
-		'draw_vs_exec.c',
-		'draw_vs_llvm.c',
-		'draw_vs_ppc.c',
-		'draw_vs_sse.c',
-		'draw_vs_varient.c'
-	])
-
-auxiliaries.insert(0, draw)
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index a4f1fcddc1..667aa46b20 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -36,6 +36,7 @@
 #include "draw_context.h"
 #include "draw_vbuf.h"
 #include "draw_vs.h"
+#include "draw_gs.h"
 #include "draw_pt.h"
 #include "draw_pipe.h"
 
@@ -67,6 +68,9 @@ struct draw_context *draw_create( void )
    if (!draw_vs_init( draw ))
       goto fail;
 
+   if (!draw_gs_init( draw ))
+      goto fail;
+
    return draw;
 
 fail:
@@ -231,11 +235,19 @@ draw_set_mapped_vertex_buffer(struct draw_context *draw,
 
 void
 draw_set_mapped_constant_buffer(struct draw_context *draw,
-                                const void *buffer, 
+                                unsigned shader_type,
+                                const void *buffer,
                                 unsigned size )
 {
-   draw->pt.user.constants = buffer;
-   draw_vs_set_constants( draw, (const float (*)[4])buffer, size );
+   debug_assert(shader_type == PIPE_SHADER_VERTEX ||
+                shader_type == PIPE_SHADER_GEOMETRY);
+   if (shader_type == PIPE_SHADER_VERTEX) {
+      draw->pt.user.vs_constants = buffer;
+      draw_vs_set_constants( draw, (const float (*)[4])buffer, size );
+   } else if (shader_type == PIPE_SHADER_GEOMETRY) {
+      draw->pt.user.gs_constants = buffer;
+      draw_gs_set_constants( draw, (const float (*)[4])buffer, size );
+   }
 }
 
 
@@ -298,7 +310,7 @@ draw_set_force_passthrough( struct draw_context *draw, boolean enable )
  * a post-transformed vertex.
  *
  * With this function, drivers that use the draw module should have no reason
- * to track the current vertex shader.
+ * to track the current vertex/geometry shader.
  *
  * Note that the draw module may sometimes generate vertices with extra
  * attributes (such as texcoords for AA lines).  The driver can call this
@@ -309,43 +321,59 @@ draw_set_force_passthrough( struct draw_context *draw, boolean enable )
  * work for the drivers.
  */
 int
-draw_find_vs_output(const struct draw_context *draw,
-                    uint semantic_name, uint semantic_index)
+draw_find_shader_output(const struct draw_context *draw,
+                        uint semantic_name, uint semantic_index)
 {
    const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+   const struct draw_geometry_shader *gs = draw->gs.geometry_shader;
    uint i;
-   for (i = 0; i < vs->info.num_outputs; i++) {
-      if (vs->info.output_semantic_name[i] == semantic_name &&
-          vs->info.output_semantic_index[i] == semantic_index)
+   const struct tgsi_shader_info *info = &vs->info;
+
+   if (gs)
+      info = &gs->info;
+
+   for (i = 0; i < info->num_outputs; i++) {
+      if (info->output_semantic_name[i] == semantic_name &&
+          info->output_semantic_index[i] == semantic_index)
          return i;
    }
 
    /* XXX there may be more than one extra vertex attrib.
     * For example, simulated gl_FragCoord and gl_PointCoord.
     */
-   if (draw->extra_vp_outputs.semantic_name == semantic_name &&
-       draw->extra_vp_outputs.semantic_index == semantic_index) {
-      return draw->extra_vp_outputs.slot;
+   if (draw->extra_shader_outputs.semantic_name == semantic_name &&
+       draw->extra_shader_outputs.semantic_index == semantic_index) {
+      return draw->extra_shader_outputs.slot;
    }
+
    return 0;
 }
 
 
 /**
- * Return number of vertex shader outputs.
+ * Return number of the shader outputs.
+ *
+ * If geometry shader is present, its output will be returned,
+ * if not vertex shader is used.
  */
 uint
-draw_num_vs_outputs(const struct draw_context *draw)
+draw_num_shader_outputs(const struct draw_context *draw)
 {
    uint count = draw->vs.vertex_shader->info.num_outputs;
-   if (draw->extra_vp_outputs.slot > 0)
+
+   /* if geometry shader is present, its outputs go to te
+    * driver, not the vertex shaders */
+   if (draw->gs.geometry_shader)
+      count = draw->gs.geometry_shader->info.num_outputs;
+
+   if (draw->extra_shader_outputs.slot > 0)
       count++;
    return count;
 }
 
 
 /**
- * Provide TGSI sampler objects for vertex shaders that use texture fetches.
+ * Provide TGSI sampler objects for vertex/geometry shaders that use texture fetches.
  * This might only be used by software drivers for the time being.
  */
 void
@@ -355,6 +383,8 @@ draw_texture_samplers(struct draw_context *draw,
 {
    draw->vs.num_samplers = num_samplers;
    draw->vs.samplers = samplers;
+   draw->gs.num_samplers = num_samplers;
+   draw->gs.samplers = samplers;
 }
 
 
@@ -366,13 +396,6 @@ void draw_set_render( struct draw_context *draw,
    draw->render = render;
 }
 
-void draw_set_edgeflags( struct draw_context *draw,
-                         const unsigned *edgeflag )
-{
-   draw->pt.user.edgeflag = edgeflag;
-}
-
-
 
 
 /**
@@ -428,3 +451,18 @@ void draw_do_flush( struct draw_context *draw, unsigned flags )
       draw->flushing = FALSE;
    }
 }
+
+
+int draw_current_shader_outputs(struct draw_context *draw)
+{
+   if (draw->gs.geometry_shader)
+      return draw->gs.num_gs_outputs;
+   return draw->vs.num_vs_outputs;
+}
+
+int draw_current_shader_position_output(struct draw_context *draw)
+{
+   if (draw->gs.geometry_shader)
+      return draw->gs.position_output;
+   return draw->vs.position_output;
+}
diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index d529e4e9a2..b716209df2 100644
--- a/src/gallium/auxiliary/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -45,6 +45,7 @@ struct pipe_context;
 struct draw_context;
 struct draw_stage;
 struct draw_vertex_shader;
+struct draw_geometry_shader;
 struct tgsi_sampler;
 
 
@@ -85,11 +86,11 @@ draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe
 
 
 int
-draw_find_vs_output(const struct draw_context *draw,
-                    uint semantic_name, uint semantic_index);
+draw_find_shader_output(const struct draw_context *draw,
+                        uint semantic_name, uint semantic_index);
 
 uint
-draw_num_vs_outputs(const struct draw_context *draw);
+draw_num_shader_outputs(const struct draw_context *draw);
 
 
 void
@@ -112,6 +113,17 @@ void draw_delete_vertex_shader(struct draw_context *draw,
                                struct draw_vertex_shader *dvs);
 
 
+/*
+ * Geometry shader functions
+ */
+struct draw_geometry_shader *
+draw_create_geometry_shader(struct draw_context *draw,
+                            const struct pipe_shader_state *shader);
+void draw_bind_geometry_shader(struct draw_context *draw,
+                               struct draw_geometry_shader *dvs);
+void draw_delete_geometry_shader(struct draw_context *draw,
+                                 struct draw_geometry_shader *dvs);
+
 
 /*
  * Vertex data functions
@@ -140,12 +152,10 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw,
                                    unsigned attr, const void *buffer);
 
 void draw_set_mapped_constant_buffer(struct draw_context *draw,
+                                     unsigned shader_type,
                                      const void *buffer,
                                      unsigned size );
 
-void draw_set_edgeflags( struct draw_context *draw,
-                         const unsigned *edgeflag );
-
 
 /***********************************************************************
  * draw_prim.c 
diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
new file mode 100644
index 0000000000..5db2e75542
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -0,0 +1,338 @@
+/**************************************************************************
+ *
+ * Copyright 2009 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 "draw_gs.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
+
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#define MAX_PRIM_VERTICES 6
+/* fixme: move it from here */
+#define MAX_PRIMITIVES 64
+
+boolean
+draw_gs_init( struct draw_context *draw )
+{
+   draw->gs.machine = tgsi_exec_machine_create();
+   if (!draw->gs.machine)
+      return FALSE;
+
+   draw->gs.machine->Primitives = align_malloc(
+      MAX_PRIMITIVES * sizeof(struct tgsi_exec_vector), 16);
+   if (!draw->gs.machine->Primitives)
+      return FALSE;
+   memset(draw->gs.machine->Primitives, 0,
+          MAX_PRIMITIVES * sizeof(struct tgsi_exec_vector));
+
+   return TRUE;
+}
+
+
+void draw_gs_set_constants( struct draw_context *draw,
+                            const float (*constants)[4],
+                            unsigned size )
+{
+}
+
+
+struct draw_geometry_shader *
+draw_create_geometry_shader(struct draw_context *draw,
+                            const struct pipe_shader_state *state)
+{
+   struct draw_geometry_shader *gs;
+   int i;
+
+   gs = CALLOC_STRUCT(draw_geometry_shader);
+
+   if (!gs)
+      return NULL;
+
+   gs->state = *state;
+   gs->state.tokens = tgsi_dup_tokens(state->tokens);
+   if (!gs->state.tokens) {
+      FREE(gs);
+      return NULL;
+   }
+
+   tgsi_scan_shader(state->tokens, &gs->info);
+
+   /* setup the defaults */
+   gs->input_primitive = PIPE_PRIM_TRIANGLES;
+   gs->output_primitive = PIPE_PRIM_TRIANGLE_STRIP;
+   gs->max_output_vertices = 32;
+
+   for (i = 0; i < gs->info.num_properties; ++i) {
+      if (gs->info.properties[i].name ==
+          TGSI_PROPERTY_GS_INPUT_PRIM)
+         gs->input_primitive = gs->info.properties[i].data[0];
+      else if (gs->info.properties[i].name ==
+               TGSI_PROPERTY_GS_OUTPUT_PRIM)
+         gs->output_primitive = gs->info.properties[i].data[0];
+      else if (gs->info.properties[i].name ==
+               TGSI_PROPERTY_GS_MAX_VERTICES)
+         gs->max_output_vertices = gs->info.properties[i].data[0];
+   }
+
+   gs->machine = draw->gs.machine;
+
+   if (gs)
+   {
+      uint i;
+      for (i = 0; i < gs->info.num_outputs; i++) {
+         if (gs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION &&
+             gs->info.output_semantic_index[i] == 0)
+            gs->position_output = i;
+      }
+   }
+
+   return gs;
+}
+
+void draw_bind_geometry_shader(struct draw_context *draw,
+                               struct draw_geometry_shader *dgs)
+{
+   draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE);
+
+   if (dgs) {
+      draw->gs.geometry_shader = dgs;
+      draw->gs.num_gs_outputs = dgs->info.num_outputs;
+      draw->gs.position_output = dgs->position_output;
+      draw_geometry_shader_prepare(dgs, draw);
+   }
+   else {
+      draw->gs.geometry_shader = NULL;
+      draw->gs.num_gs_outputs = 0;
+   }
+}
+
+void draw_delete_geometry_shader(struct draw_context *draw,
+                                 struct draw_geometry_shader *dgs)
+{
+   FREE(dgs);
+}
+
+static INLINE int num_vertices_for_prim(int prim)
+{
+   switch(prim) {
+   case PIPE_PRIM_POINTS:
+      return 1;
+   case PIPE_PRIM_LINES:
+      return 2;
+   case PIPE_PRIM_LINE_LOOP:
+      return 2;
+   case PIPE_PRIM_LINE_STRIP:
+      return 2;
+   case PIPE_PRIM_TRIANGLES:
+      return 3;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      return 3;
+   case PIPE_PRIM_TRIANGLE_FAN:
+      return 3;
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      return 4;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      return 6;
+   default:
+      assert(!"Bad geometry shader input");
+      return 0;
+   }
+}
+
+static void draw_fetch_geometry_input(struct draw_geometry_shader *shader,
+                                      int start_primitive,
+                                      int num_primitives,
+                                      const float (*input_ptr)[4],
+                                      unsigned input_vertex_stride,
+                                      unsigned inputs_from_vs)
+{
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned slot, vs_slot, k, j;
+   unsigned num_vertices = num_vertices_for_prim(shader->input_primitive);
+   int idx = 0;
+
+   for (slot = 0, vs_slot = 0; slot < shader->info.num_inputs; slot++) {
+      /*debug_printf("Slot = %d (semantic = %d)\n", slot,
+        shader->info.input_semantic_name[slot]);*/
+      if (shader->info.input_semantic_name[slot] ==
+          TGSI_SEMANTIC_PRIMID) {
+         for (j = 0; j < num_primitives; ++j) {
+            machine->Inputs[idx].xyzw[0].f[j] = (float)start_primitive + j;
+            machine->Inputs[idx].xyzw[1].f[j] = (float)start_primitive + j;
+            machine->Inputs[idx].xyzw[2].f[j] = (float)start_primitive + j;
+            machine->Inputs[idx].xyzw[3].f[j] = (float)start_primitive + j;
+         }
+         ++idx;
+      } else {
+         for (j = 0; j < num_primitives; ++j) {
+            int vidx = idx;
+            const float (*prim_ptr)[4];
+            /*debug_printf("    %d) Prim (num_verts = %d)\n", start_primitive + j,
+              num_vertices);*/
+            prim_ptr = (const float (*)[4])(
+               (const char *)input_ptr +
+               (j * num_vertices * input_vertex_stride));
+
+            for (k = 0; k < num_vertices; ++k, ++vidx) {
+               const float (*input)[4];
+               input = (const float (*)[4])(
+                  (const char *)prim_ptr + (k * input_vertex_stride));
+               vidx = k * TGSI_EXEC_MAX_INPUT_ATTRIBS + slot;
+               /*debug_printf("\t%d)(%d) Input vert:\n", vidx, k);*/
+#if 1
+               assert(!util_is_inf_or_nan(input[vs_slot][0]));
+               assert(!util_is_inf_or_nan(input[vs_slot][1]));
+               assert(!util_is_inf_or_nan(input[vs_slot][2]));
+               assert(!util_is_inf_or_nan(input[vs_slot][3]));
+#endif
+               machine->Inputs[vidx].xyzw[0].f[j] = input[vs_slot][0];
+               machine->Inputs[vidx].xyzw[1].f[j] = input[vs_slot][1];
+               machine->Inputs[vidx].xyzw[2].f[j] = input[vs_slot][2];
+               machine->Inputs[vidx].xyzw[3].f[j] = input[vs_slot][3];
+#if 0
+               debug_printf("\t\t%d %f %f %f %f\n", slot,
+                            machine->Inputs[vidx].xyzw[0].f[j],
+                            machine->Inputs[vidx].xyzw[1].f[j],
+                            machine->Inputs[vidx].xyzw[2].f[j],
+                            machine->Inputs[vidx].xyzw[3].f[j]);
+#endif
+            }
+         }
+         ++vs_slot;
+         idx += num_vertices;
+      }
+   }
+}
+
+static INLINE void
+draw_geometry_fetch_outputs(struct draw_geometry_shader *shader,
+                            int num_primitives,
+                            float (*output)[4],
+                            unsigned vertex_size)
+{
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned prim_idx, j, slot;
+
+   /* Unswizzle all output results.
+    */
+   /* FIXME: handle all the primitives produced by the gs, not just
+    * the first one
+    unsigned prim_count =
+    mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0];*/
+   for (prim_idx = 0; prim_idx < num_primitives; ++prim_idx) {
+      unsigned num_verts_per_prim = machine->Primitives[0];
+      for (j = 0; j < num_verts_per_prim; j++) {
+         int idx = (prim_idx * num_verts_per_prim + j) *
+                   shader->info.num_outputs;
+#ifdef DEBUG_OUTPUTS
+         debug_printf("%d) Output vert:\n", idx);
+#endif
+         for (slot = 0; slot < shader->info.num_outputs; slot++) {
+            output[slot][0] = machine->Outputs[idx + slot].xyzw[0].f[prim_idx];
+            output[slot][1] = machine->Outputs[idx + slot].xyzw[1].f[prim_idx];
+            output[slot][2] = machine->Outputs[idx + slot].xyzw[2].f[prim_idx];
+            output[slot][3] = machine->Outputs[idx + slot].xyzw[3].f[prim_idx];
+#ifdef DEBUG_OUTPUTS
+            debug_printf("\t%d: %f %f %f %f\n", slot,
+                         output[slot][0],
+                         output[slot][1],
+                         output[slot][2],
+                         output[slot][3]);
+#endif
+            debug_assert(!util_is_inf_or_nan(output[slot][0]));
+         }
+         output = (float (*)[4])((char *)output + vertex_size);
+      }
+   }
+}
+
+void draw_geometry_shader_run(struct draw_geometry_shader *shader,
+                              const float (*input)[4],
+                              float (*output)[4],
+                              const float (*constants)[4],
+                              unsigned count,
+                              unsigned input_stride,
+                              unsigned vertex_size)
+{
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned int i;
+   unsigned num_vertices = num_vertices_for_prim(shader->input_primitive);
+   unsigned num_primitives = count/num_vertices;
+   unsigned inputs_from_vs = 0;
+
+   machine->Consts = constants;
+
+   for (i = 0; i < shader->info.num_inputs; ++i) {
+      if (shader->info.input_semantic_name[i] != TGSI_SEMANTIC_PRIMID)
+         ++inputs_from_vs;
+   }
+
+   for (i = 0; i < num_primitives; ++i) {
+      unsigned int max_primitives = 1;
+
+      draw_fetch_geometry_input(shader, i, max_primitives, input,
+                                input_stride, inputs_from_vs);
+
+      tgsi_set_exec_mask(machine,
+                         1,
+                         max_primitives > 1,
+                         max_primitives > 2,
+                         max_primitives > 3);
+
+      /* run interpreter */
+      tgsi_exec_machine_run(machine);
+
+      draw_geometry_fetch_outputs(shader, max_primitives,
+                                  output, vertex_size);
+   }
+}
+
+void draw_geometry_shader_delete(struct draw_geometry_shader *shader)
+{
+   FREE((void*) shader->state.tokens);
+   FREE(shader);
+}
+
+void draw_geometry_shader_prepare(struct draw_geometry_shader *shader,
+                                  struct draw_context *draw)
+{
+    if (shader->machine->Tokens != shader->state.tokens) {
+       tgsi_exec_machine_bind_shader(shader->machine,
+                                     shader->state.tokens,
+                                     draw->gs.num_samplers,
+                                     draw->gs.samplers);
+    }
+}
diff --git a/src/gallium/auxiliary/draw/draw_gs.h b/src/gallium/auxiliary/draw/draw_gs.h
new file mode 100644
index 0000000000..d6a97d9c4e
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_gs.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 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.
+ * 
+ **************************************************************************/
+
+#ifndef DRAW_GS_H
+#define DRAW_GS_H
+
+#include "draw_context.h"
+#include "draw_private.h"
+
+
+#define MAX_TGSI_PRIMITIVES 4
+
+struct draw_context;
+
+/**
+ * Private version of the compiled geometry shader
+ */
+struct draw_geometry_shader {
+   struct draw_context *draw;
+
+   struct tgsi_exec_machine *machine;
+
+   /* This member will disappear shortly:*/
+   struct pipe_shader_state state;
+
+   struct tgsi_shader_info info;
+   unsigned position_output;
+
+   unsigned max_output_vertices;
+   unsigned input_primitive;
+   unsigned output_primitive;
+
+   /* Extracted from shader:
+    */
+   const float (*immediates)[4];
+};
+
+void draw_geometry_shader_run(struct draw_geometry_shader *shader,
+                              const float (*input)[4],
+                              float (*output)[4],
+                              const float (*constants)[4],
+                              unsigned count,
+                              unsigned input_stride,
+                              unsigned output_stride);
+
+void draw_geometry_shader_prepare(struct draw_geometry_shader *shader,
+                                  struct draw_context *draw);
+
+void draw_geometry_shader_delete(struct draw_geometry_shader *shader);
+
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
index 9f956715a2..4585dcdb48 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -35,6 +35,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
@@ -139,29 +141,29 @@ aa_transform_decl(struct tgsi_transform_context *ctx,
    struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
 
    if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
-       decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR &&
-       decl->Semantic.SemanticIndex == 0) {
-      aactx->colorOutput = decl->DeclarationRange.First;
+       decl->Semantic.Name == TGSI_SEMANTIC_COLOR &&
+       decl->Semantic.Index == 0) {
+      aactx->colorOutput = decl->Range.First;
    }
    else if (decl->Declaration.File == TGSI_FILE_SAMPLER) {
       uint i;
-      for (i = decl->DeclarationRange.First;
-           i <= decl->DeclarationRange.Last; i++) {
+      for (i = decl->Range.First;
+           i <= decl->Range.Last; i++) {
          aactx->samplersUsed |= 1 << i;
       }
    }
    else if (decl->Declaration.File == TGSI_FILE_INPUT) {
-      if ((int) decl->DeclarationRange.Last > aactx->maxInput)
-         aactx->maxInput = decl->DeclarationRange.Last;
-      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC &&
-           (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) {
-         aactx->maxGeneric = decl->Semantic.SemanticIndex;
+      if ((int) decl->Range.Last > aactx->maxInput)
+         aactx->maxInput = decl->Range.Last;
+      if (decl->Semantic.Name == TGSI_SEMANTIC_GENERIC &&
+           (int) decl->Semantic.Index > aactx->maxGeneric) {
+         aactx->maxGeneric = decl->Semantic.Index;
       }
    }
    else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
       uint i;
-      for (i = decl->DeclarationRange.First;
-           i <= decl->DeclarationRange.Last; i++) {
+      for (i = decl->Range.First;
+           i <= decl->Range.Last; i++) {
          aactx->tempsUsed |= (1 << i);
       }
    }
@@ -228,30 +230,30 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       /* XXX this could be linear... */
       decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
       decl.Declaration.Semantic = 1;
-      decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-      decl.Semantic.SemanticIndex = aactx->maxGeneric + 1;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = aactx->maxInput + 1;
+      decl.Semantic.Name = TGSI_SEMANTIC_GENERIC;
+      decl.Semantic.Index = aactx->maxGeneric + 1;
+      decl.Range.First = 
+      decl.Range.Last = aactx->maxInput + 1;
       ctx->emit_declaration(ctx, &decl);
 
       /* declare new sampler */
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_SAMPLER;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = aactx->freeSampler;
+      decl.Range.First = 
+      decl.Range.Last = aactx->freeSampler;
       ctx->emit_declaration(ctx, &decl);
 
       /* declare new temp regs */
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_TEMPORARY;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = aactx->texTemp;
+      decl.Range.First = 
+      decl.Range.Last = aactx->texTemp;
       ctx->emit_declaration(ctx, &decl);
 
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_TEMPORARY;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = aactx->colorTemp;
+      decl.Range.First = 
+      decl.Range.Last = aactx->colorTemp;
       ctx->emit_declaration(ctx, &decl);
 
       aactx->firstInstruction = FALSE;
@@ -265,14 +267,15 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_TEX;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = aactx->texTemp;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = aactx->texTemp;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->freeSampler;
+      newInst.Instruction.Texture = TRUE;
+      newInst.Texture.Texture = TGSI_TEXTURE_2D;
+      newInst.Src[0].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[0].Register.Index = aactx->maxInput + 1;
+      newInst.Src[1].Register.File = TGSI_FILE_SAMPLER;
+      newInst.Src[1].Register.Index = aactx->freeSampler;
 
       ctx->emit_instruction(ctx, &newInst);
 
@@ -280,26 +283,26 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ;
+      newInst.Dst[0].Register.File = TGSI_FILE_OUTPUT;
+      newInst.Dst[0].Register.Index = aactx->colorOutput;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_XYZ;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = aactx->colorTemp;
       ctx->emit_instruction(ctx, &newInst);
 
       /* MUL alpha */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Dst[0].Register.File = TGSI_FILE_OUTPUT;
+      newInst.Dst[0].Register.Index = aactx->colorOutput;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_W;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->texTemp;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = aactx->colorTemp;
+      newInst.Src[1].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[1].Register.Index = aactx->texTemp;
       ctx->emit_instruction(ctx, &newInst);
 
       /* END */
@@ -316,11 +319,11 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       uint i;
 
       for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
-         struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
-         if (dst->DstRegister.File == TGSI_FILE_OUTPUT &&
-             dst->DstRegister.Index == aactx->colorOutput) {
-            dst->DstRegister.File = TGSI_FILE_TEMPORARY;
-            dst->DstRegister.Index = aactx->colorTemp;
+         struct tgsi_full_dst_register *dst = &inst->Dst[i];
+         if (dst->Register.File == TGSI_FILE_OUTPUT &&
+             dst->Register.Index == aactx->colorOutput) {
+            dst->Register.File = TGSI_FILE_TEMPORARY;
+            dst->Register.Index = aactx->colorTemp;
          }
       }
 
@@ -398,10 +401,9 @@ aaline_create_texture(struct aaline_stage *aaline)
    texTemp.target = PIPE_TEXTURE_2D;
    texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */
    texTemp.last_level = MAX_TEXTURE_LEVEL;
-   texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL;
-   texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL;
-   texTemp.depth[0] = 1;
-   pf_get_block(texTemp.format, &texTemp.block);
+   texTemp.width0 = 1 << MAX_TEXTURE_LEVEL;
+   texTemp.height0 = 1 << MAX_TEXTURE_LEVEL;
+   texTemp.depth0 = 1;
 
    aaline->texture = screen->texture_create(screen, &texTemp);
    if (!aaline->texture)
@@ -413,11 +415,11 @@ aaline_create_texture(struct aaline_stage *aaline)
     */
    for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) {
       struct pipe_transfer *transfer;
-      const uint size = aaline->texture->width[level];
+      const uint size = u_minify(aaline->texture->width0, level);
       ubyte *data;
       uint i, j;
 
-      assert(aaline->texture->width[level] == aaline->texture->height[level]);
+      assert(aaline->texture->width0 == aaline->texture->height0);
 
       /* This texture is new, no need to flush. 
        */
@@ -658,13 +660,13 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header)
    }
 
    /* update vertex attrib info */
-   aaline->tex_slot = draw->vs.num_vs_outputs;
-   aaline->pos_slot = draw->vs.position_output;
+   aaline->tex_slot = draw_current_shader_outputs(draw);
+   aaline->pos_slot = draw_current_shader_position_output(draw);;
 
    /* advertise the extra post-transformed vertex attribute */
-   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-   draw->extra_vp_outputs.semantic_index = aaline->fs->generic_attrib;
-   draw->extra_vp_outputs.slot = aaline->tex_slot;
+   draw->extra_shader_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+   draw->extra_shader_outputs.semantic_index = aaline->fs->generic_attrib;
+   draw->extra_shader_outputs.slot = aaline->tex_slot;
 
    /* how many samplers? */
    /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
@@ -705,7 +707,7 @@ aaline_flush(struct draw_stage *stage, unsigned flags)
                                        aaline->state.texture);
    draw->suspend_flushing = FALSE;
 
-   draw->extra_vp_outputs.slot = 0;
+   draw->extra_shader_outputs.slot = 0;
 }
 
 
@@ -896,16 +898,16 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe)
    aaline->driver_bind_fs_state = pipe->bind_fs_state;
    aaline->driver_delete_fs_state = pipe->delete_fs_state;
 
-   aaline->driver_bind_sampler_states = pipe->bind_sampler_states;
-   aaline->driver_set_sampler_textures = pipe->set_sampler_textures;
+   aaline->driver_bind_sampler_states = pipe->bind_fragment_sampler_states;
+   aaline->driver_set_sampler_textures = pipe->set_fragment_sampler_textures;
 
    /* override the driver's functions */
    pipe->create_fs_state = aaline_create_fs_state;
    pipe->bind_fs_state = aaline_bind_fs_state;
    pipe->delete_fs_state = aaline_delete_fs_state;
 
-   pipe->bind_sampler_states = aaline_bind_sampler_states;
-   pipe->set_sampler_textures = aaline_set_sampler_textures;
+   pipe->bind_fragment_sampler_states = aaline_bind_sampler_states;
+   pipe->set_fragment_sampler_textures = aaline_set_sampler_textures;
    
    /* Install once everything is known to be OK:
     */
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
index ae1712fe12..d86717e518 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -131,22 +131,22 @@ aa_transform_decl(struct tgsi_transform_context *ctx,
    struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
 
    if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
-       decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR &&
-       decl->Semantic.SemanticIndex == 0) {
-      aactx->colorOutput = decl->DeclarationRange.First;
+       decl->Semantic.Name == TGSI_SEMANTIC_COLOR &&
+       decl->Semantic.Index == 0) {
+      aactx->colorOutput = decl->Range.First;
    }
    else if (decl->Declaration.File == TGSI_FILE_INPUT) {
-      if ((int) decl->DeclarationRange.Last > aactx->maxInput)
-         aactx->maxInput = decl->DeclarationRange.Last;
-      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC &&
-           (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) {
-         aactx->maxGeneric = decl->Semantic.SemanticIndex;
+      if ((int) decl->Range.Last > aactx->maxInput)
+         aactx->maxInput = decl->Range.Last;
+      if (decl->Semantic.Name == TGSI_SEMANTIC_GENERIC &&
+           (int) decl->Semantic.Index > aactx->maxGeneric) {
+         aactx->maxGeneric = decl->Semantic.Index;
       }
    }
    else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
       uint i;
-      for (i = decl->DeclarationRange.First;
-           i <= decl->DeclarationRange.Last; i++) {
+      for (i = decl->Range.First;
+           i <= decl->Range.Last; i++) {
          aactx->tempsUsed |= (1 << i);
       }
    }
@@ -198,23 +198,23 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       /* XXX this could be linear... */
       decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
       decl.Declaration.Semantic = 1;
-      decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-      decl.Semantic.SemanticIndex = aactx->maxGeneric + 1;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = texInput;
+      decl.Semantic.Name = TGSI_SEMANTIC_GENERIC;
+      decl.Semantic.Index = aactx->maxGeneric + 1;
+      decl.Range.First = 
+      decl.Range.Last = texInput;
       ctx->emit_declaration(ctx, &decl);
 
       /* declare new temp regs */
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_TEMPORARY;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = tmp0;
+      decl.Range.First = 
+      decl.Range.Last = tmp0;
       ctx->emit_declaration(ctx, &decl);
 
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_TEMPORARY;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = aactx->colorTemp;
+      decl.Range.First = 
+      decl.Range.Last = aactx->colorTemp;
       ctx->emit_declaration(ctx, &decl);
 
       aactx->firstInstruction = FALSE;
@@ -234,30 +234,30 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XY;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_XY;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.Src[0].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[0].Register.Index = texInput;
+      newInst.Src[1].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[1].Register.Index = texInput;
       ctx->emit_instruction(ctx, &newInst);
 
       /* ADD t0.x, t0.x, t0.y;  # x^2 + y^2 */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_ADD;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_X;
+      newInst.Src[1].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[1].Register.Index = tmp0;
+      newInst.Src[1].Register.SwizzleX = TGSI_SWIZZLE_Y;
       ctx->emit_instruction(ctx, &newInst);
 
 #if NORMALIZE  /* OPTIONAL normalization of length */
@@ -265,24 +265,24 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_RSQ;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
       ctx->emit_instruction(ctx, &newInst);
 
       /* RCP t0.x, t0.x; */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_RCP;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
       ctx->emit_instruction(ctx, &newInst);
 #endif
 
@@ -290,16 +290,16 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_SGT;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_Y;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.Src[1].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[1].Register.Index = texInput;
+      newInst.Src[1].Register.SwizzleY = TGSI_SWIZZLE_W;
       ctx->emit_instruction(ctx, &newInst);
 
       /* KIL -tmp0.yyyy;   # if -tmp0.y < 0, KILL */
@@ -307,13 +307,13 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst.Instruction.Opcode = TGSI_OPCODE_KIL;
       newInst.Instruction.NumDstRegs = 0;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.Negate = 1;
       ctx->emit_instruction(ctx, &newInst);
 
 
@@ -323,77 +323,77 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_SUB;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_Z;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z;
+      newInst.Src[0].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[0].Register.Index = texInput;
+      newInst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_W;
+      newInst.Src[1].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[1].Register.Index = texInput;
+      newInst.Src[1].Register.SwizzleZ = TGSI_SWIZZLE_Z;
       ctx->emit_instruction(ctx, &newInst);
 
       /* RCP t0.z, t0.z;  # t0.z = 1 / m */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_RCP;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_Z;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_Z;
       ctx->emit_instruction(ctx, &newInst);
 
       /* SUB t0.y, 1, t0.x;  # d = 1 - d */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_SUB;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_Y;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.Src[0].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[0].Register.Index = texInput;
+      newInst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.Src[1].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[1].Register.Index = tmp0;
+      newInst.Src[1].Register.SwizzleY = TGSI_SWIZZLE_X;
       ctx->emit_instruction(ctx, &newInst);
 
       /* MUL t0.w, t0.y, t0.z;   # coverage = d * m */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_W;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.Src[1].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[1].Register.Index = tmp0;
+      newInst.Src[1].Register.SwizzleW = TGSI_SWIZZLE_Z;
       ctx->emit_instruction(ctx, &newInst);
 
       /* SLE t0.y, t0.x, tex.z;  # bool b = distance <= k */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_SLE;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_Y;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.Src[1].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[1].Register.Index = texInput;
+      newInst.Src[1].Register.SwizzleY = TGSI_SWIZZLE_Z;
       ctx->emit_instruction(ctx, &newInst);
 
       /* CMP t0.w, -t0.y, tex.w, t0.w;
@@ -405,29 +405,29 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_CMP;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = tmp0;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_W;
       newInst.Instruction.NumSrcRegs = 3;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[2].SrcRegister.Index = tmp0;
-      newInst.FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
-      newInst.FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = tmp0;
+      newInst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.Src[0].Register.Negate = 1;
+      newInst.Src[1].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[1].Register.Index = texInput;
+      newInst.Src[1].Register.SwizzleX = TGSI_SWIZZLE_W;
+      newInst.Src[1].Register.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.Src[1].Register.SwizzleZ = TGSI_SWIZZLE_W;
+      newInst.Src[1].Register.SwizzleW = TGSI_SWIZZLE_W;
+      newInst.Src[2].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[2].Register.Index = tmp0;
+      newInst.Src[2].Register.SwizzleX = TGSI_SWIZZLE_W;
+      newInst.Src[2].Register.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.Src[2].Register.SwizzleZ = TGSI_SWIZZLE_W;
+      newInst.Src[2].Register.SwizzleW = TGSI_SWIZZLE_W;
       ctx->emit_instruction(ctx, &newInst);
 
    }
@@ -439,26 +439,26 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ;
+      newInst.Dst[0].Register.File = TGSI_FILE_OUTPUT;
+      newInst.Dst[0].Register.Index = aactx->colorOutput;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_XYZ;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = aactx->colorTemp;
       ctx->emit_instruction(ctx, &newInst);
 
       /* MUL result.color.w, colorTemp, tmp0.w; */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
-      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Dst[0].Register.File = TGSI_FILE_OUTPUT;
+      newInst.Dst[0].Register.Index = aactx->colorOutput;
+      newInst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_W;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->tmp0;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = aactx->colorTemp;
+      newInst.Src[1].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[1].Register.Index = aactx->tmp0;
       ctx->emit_instruction(ctx, &newInst);
    }
    else {
@@ -468,11 +468,11 @@ aa_transform_inst(struct tgsi_transform_context *ctx,
       uint i;
 
       for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
-         struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
-         if (dst->DstRegister.File == TGSI_FILE_OUTPUT &&
-             dst->DstRegister.Index == aactx->colorOutput) {
-            dst->DstRegister.File = TGSI_FILE_TEMPORARY;
-            dst->DstRegister.Index = aactx->colorTemp;
+         struct tgsi_full_dst_register *dst = &inst->Dst[i];
+         if (dst->Register.File == TGSI_FILE_OUTPUT &&
+             dst->Register.Index == aactx->colorOutput) {
+            dst->Register.File = TGSI_FILE_TEMPORARY;
+            dst->Register.Index = aactx->colorTemp;
          }
       }
    }
@@ -687,14 +687,14 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
    bind_aapoint_fragment_shader(aapoint);
 
    /* update vertex attrib info */
-   aapoint->tex_slot = draw->vs.num_vs_outputs;
+   aapoint->tex_slot = draw_current_shader_outputs(draw);
    assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */
 
-   aapoint->pos_slot = draw->vs.position_output;
+   aapoint->pos_slot = draw_current_shader_position_output(draw);
 
-   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-   draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib;
-   draw->extra_vp_outputs.slot = aapoint->tex_slot;
+   draw->extra_shader_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+   draw->extra_shader_outputs.semantic_index = aapoint->fs->generic_attrib;
+   draw->extra_shader_outputs.slot = aapoint->tex_slot;
 
    /* find psize slot in post-transform vertex */
    aapoint->psize_slot = -1;
@@ -731,7 +731,7 @@ aapoint_flush(struct draw_stage *stage, unsigned flags)
    aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs);
    draw->suspend_flushing = FALSE;
 
-   draw->extra_vp_outputs.slot = 0;
+   draw->extra_shader_outputs.slot = 0;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c
index 0670268a19..205cda5eab 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_clip.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c
@@ -114,8 +114,8 @@ static void interp( const struct clipper *clip,
 		    const struct vertex_header *out, 
 		    const struct vertex_header *in )
 {
-   const unsigned nr_attrs = clip->stage.draw->vs.num_vs_outputs;
-   const unsigned pos_attr = clip->stage.draw->vs.position_output;
+   const unsigned nr_attrs = draw_current_shader_outputs(clip->stage.draw);
+   const unsigned pos_attr = draw_current_shader_position_output(clip->stage.draw);
    unsigned j;
 
    /* Vertex header.
diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c
index 0a70483858..11b39db599 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_cull.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c
@@ -55,7 +55,7 @@ static INLINE struct cull_stage *cull_stage( struct draw_stage *stage )
 static void cull_tri( struct draw_stage *stage,
 		      struct prim_header *header )
 {
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
 
    /* Window coords: */
    const float *v0 = header->v[0]->data[pos];
diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c
index 40798a5d6e..e829492423 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_offset.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c
@@ -63,7 +63,7 @@ static INLINE struct offset_stage *offset_stage( struct draw_stage *stage )
 static void do_offset_tri( struct draw_stage *stage,
 			   struct prim_header *header )
 {
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    struct offset_stage *offset = offset_stage(stage);   
    float inv_det = 1.0f / header->det;
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
index 283502cdf3..0cc2b71864 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -38,6 +38,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
 
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
@@ -133,20 +134,20 @@ pstip_transform_decl(struct tgsi_transform_context *ctx,
 
    if (decl->Declaration.File == TGSI_FILE_SAMPLER) {
       uint i;
-      for (i = decl->DeclarationRange.First;
-           i <= decl->DeclarationRange.Last; i++) {
+      for (i = decl->Range.First;
+           i <= decl->Range.Last; i++) {
          pctx->samplersUsed |= 1 << i;
       }
    }
    else if (decl->Declaration.File == TGSI_FILE_INPUT) {
-      pctx->maxInput = MAX2(pctx->maxInput, (int) decl->DeclarationRange.Last);
-      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION)
-         pctx->wincoordInput = (int) decl->DeclarationRange.First;
+      pctx->maxInput = MAX2(pctx->maxInput, (int) decl->Range.Last);
+      if (decl->Semantic.Name == TGSI_SEMANTIC_POSITION)
+         pctx->wincoordInput = (int) decl->Range.First;
    }
    else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
       uint i;
-      for (i = decl->DeclarationRange.First;
-           i <= decl->DeclarationRange.Last; i++) {
+      for (i = decl->Range.First;
+           i <= decl->Range.Last; i++) {
          pctx->tempsUsed |= (1 << i);
       }
    }
@@ -226,25 +227,25 @@ pstip_transform_inst(struct tgsi_transform_context *ctx,
          decl.Declaration.File = TGSI_FILE_INPUT;
          decl.Declaration.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */
          decl.Declaration.Semantic = 1;
-         decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION;
-         decl.Semantic.SemanticIndex = 0;
-         decl.DeclarationRange.First = 
-            decl.DeclarationRange.Last = wincoordInput;
+         decl.Semantic.Name = TGSI_SEMANTIC_POSITION;
+         decl.Semantic.Index = 0;
+         decl.Range.First = 
+            decl.Range.Last = wincoordInput;
          ctx->emit_declaration(ctx, &decl);
       }
 
       /* declare new sampler */
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_SAMPLER;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = pctx->freeSampler;
+      decl.Range.First = 
+      decl.Range.Last = pctx->freeSampler;
       ctx->emit_declaration(ctx, &decl);
 
       /* declare new temp regs */
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_TEMPORARY;
-      decl.DeclarationRange.First = 
-      decl.DeclarationRange.Last = pctx->texTemp;
+      decl.Range.First = 
+      decl.Range.Last = pctx->texTemp;
       ctx->emit_declaration(ctx, &decl);
 
       /* emit immediate = {1/32, 1/32, 1, 1}
@@ -280,27 +281,28 @@ pstip_transform_inst(struct tgsi_transform_context *ctx,
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = pctx->texTemp;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = wincoordInput;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_IMMEDIATE;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->numImmed;
+      newInst.Src[0].Register.File = TGSI_FILE_INPUT;
+      newInst.Src[0].Register.Index = wincoordInput;
+      newInst.Src[1].Register.File = TGSI_FILE_IMMEDIATE;
+      newInst.Src[1].Register.Index = pctx->numImmed;
       ctx->emit_instruction(ctx, &newInst);
 
       /* TEX texTemp, texTemp, sampler; */
       newInst = tgsi_default_full_instruction();
       newInst.Instruction.Opcode = TGSI_OPCODE_TEX;
       newInst.Instruction.NumDstRegs = 1;
-      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp;
+      newInst.Dst[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Dst[0].Register.Index = pctx->texTemp;
       newInst.Instruction.NumSrcRegs = 2;
-      newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp;
-      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-      newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->freeSampler;
+      newInst.Instruction.Texture = TRUE;
+      newInst.Texture.Texture = TGSI_TEXTURE_2D;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = pctx->texTemp;
+      newInst.Src[1].Register.File = TGSI_FILE_SAMPLER;
+      newInst.Src[1].Register.Index = pctx->freeSampler;
       ctx->emit_instruction(ctx, &newInst);
 
       /* KIL -texTemp;   # if -texTemp < 0, KILL fragment */
@@ -308,9 +310,9 @@ pstip_transform_inst(struct tgsi_transform_context *ctx,
       newInst.Instruction.Opcode = TGSI_OPCODE_KIL;
       newInst.Instruction.NumDstRegs = 0;
       newInst.Instruction.NumSrcRegs = 1;
-      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-      newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp;
-      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      newInst.Src[0].Register.File = TGSI_FILE_TEMPORARY;
+      newInst.Src[0].Register.Index = pctx->texTemp;
+      newInst.Src[0].Register.Negate = 1;
       ctx->emit_instruction(ctx, &newInst);
    }
 
@@ -427,10 +429,9 @@ pstip_create_texture(struct pstip_stage *pstip)
    texTemp.target = PIPE_TEXTURE_2D;
    texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */
    texTemp.last_level = 0;
-   texTemp.width[0] = 32;
-   texTemp.height[0] = 32;
-   texTemp.depth[0] = 1;
-   pf_get_block(texTemp.format, &texTemp.block);
+   texTemp.width0 = 32;
+   texTemp.height0 = 32;
+   texTemp.depth0 = 1;
 
    pstip->texture = screen->texture_create(screen, &texTemp);
    if (pstip->texture == NULL)
@@ -754,8 +755,8 @@ draw_install_pstipple_stage(struct draw_context *draw,
    pstip->driver_bind_fs_state = pipe->bind_fs_state;
    pstip->driver_delete_fs_state = pipe->delete_fs_state;
 
-   pstip->driver_bind_sampler_states = pipe->bind_sampler_states;
-   pstip->driver_set_sampler_textures = pipe->set_sampler_textures;
+   pstip->driver_bind_sampler_states = pipe->bind_fragment_sampler_states;
+   pstip->driver_set_sampler_textures = pipe->set_fragment_sampler_textures;
    pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple;
 
    /* override the driver's functions */
@@ -763,8 +764,8 @@ draw_install_pstipple_stage(struct draw_context *draw,
    pipe->bind_fs_state = pstip_bind_fs_state;
    pipe->delete_fs_state = pstip_delete_fs_state;
 
-   pipe->bind_sampler_states = pstip_bind_sampler_states;
-   pipe->set_sampler_textures = pstip_set_sampler_textures;
+   pipe->bind_fragment_sampler_states = pstip_bind_sampler_states;
+   pipe->set_fragment_sampler_textures = pstip_set_sampler_textures;
    pipe->set_polygon_stipple = pstip_set_polygon_stipple;
 
    return TRUE;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
index 6e921bac27..70fbab9ea7 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
@@ -73,7 +73,8 @@ screen_interp( struct draw_context *draw,
                const struct vertex_header *v1 )
 {
    uint attr;
-   for (attr = 0; attr < draw->vs.num_vs_outputs; attr++) {
+   int num_outputs = draw_current_shader_outputs(draw);
+   for (attr = 0; attr < num_outputs; attr++) {
       const float *val0 = v0->data[attr];
       const float *val1 = v1->data[attr];
       float *newv = dst->data[attr];
@@ -121,7 +122,7 @@ stipple_line(struct draw_stage *stage, struct prim_header *header)
    struct stipple_stage *stipple = stipple_stage(stage);
    struct vertex_header *v0 = header->v[0];
    struct vertex_header *v1 = header->v[1];
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    const float *pos0 = v0->data[pos];
    const float *pos1 = v1->data[pos];
    float start = 0;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
index f32cbef983..3073c87082 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
@@ -59,7 +59,7 @@ static void wideline_line( struct draw_stage *stage,
                            struct prim_header *header )
 {
    /*const struct wideline_stage *wide = wideline_stage(stage);*/
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    const float half_width = 0.5f * stage->draw->rasterizer->line_width;
 
    struct prim_header tri;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index 7d76a7dbf3..8dc50c0ab4 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -112,7 +112,7 @@ static void set_texcoords(const struct widepoint_stage *wide,
 
    if (wide->point_coord_fs_input >= 0) {
       /* put gl_PointCoord into the extra vertex slot */
-      uint slot = wide->stage.draw->extra_vp_outputs.slot;
+      uint slot = wide->stage.draw->extra_shader_outputs.slot;
       v->data[slot][0] = tc[0];
       v->data[slot][1] = tc[1];
       v->data[slot][2] = 0.0F;
@@ -130,7 +130,7 @@ static void widepoint_point( struct draw_stage *stage,
                              struct prim_header *header )
 {
    const struct widepoint_stage *wide = widepoint_stage(stage);
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite;
    float half_size;
    float left_adj, right_adj, bot_adj, top_adj;
@@ -257,13 +257,13 @@ static void widepoint_first_point( struct draw_stage *stage,
       wide->point_coord_fs_input = find_pntc_input_attrib(draw);
 
       /* setup extra vp output (point coord implemented as a texcoord) */
-      draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-      draw->extra_vp_outputs.semantic_index = 0;
-      draw->extra_vp_outputs.slot = draw->vs.num_vs_outputs;
+      draw->extra_shader_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+      draw->extra_shader_outputs.semantic_index = 0;
+      draw->extra_shader_outputs.slot = draw_current_shader_outputs(draw);
    }
    else {
       wide->point_coord_fs_input = -1;
-      draw->extra_vp_outputs.slot = 0;
+      draw->extra_shader_outputs.slot = 0;
    }
 
    wide->psize_slot = -1;
@@ -287,7 +287,7 @@ static void widepoint_flush( struct draw_stage *stage, unsigned flags )
 {
    stage->point = widepoint_first_point;
    stage->next->flush( stage->next, flags );
-   stage->draw->extra_vp_outputs.slot = 0;
+   stage->draw->extra_shader_outputs.slot = 0;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 41fcb16a0a..e49041556b 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -142,8 +142,6 @@ struct draw_context
 
       /* user-space vertex data, buffers */
       struct {
-         const unsigned *edgeflag;
-
          /** vertex element/index buffer (ex: glDrawElements) */
          const void *elts;
          /** bytes per index (0, 1, 2 or 4) */
@@ -154,8 +152,9 @@ struct draw_context
          /** vertex arrays */
          const void *vbuffer[PIPE_MAX_ATTRIBS];
          
-         /** constant buffer (for vertex shader) */
-         const void *constants;
+         /** constant buffer (for vertex/geometry shader) */
+         const void *vs_constants;
+         const void *gs_constants;
       } user;
 
       boolean test_fse;         /* enable FSE even though its not correct (eg for softpipe) */
@@ -184,6 +183,7 @@ struct draw_context
       struct draw_vertex_shader *vertex_shader;
       uint num_vs_outputs;  /**< convenience, from vertex_shader */
       uint position_output;
+      uint edgeflag_output;
 
       /** TGSI program interpreter runtime state */
       struct tgsi_exec_machine *machine;
@@ -212,6 +212,18 @@ struct draw_context
       struct translate_cache *emit_cache;
    } vs;
 
+   struct {
+      struct draw_geometry_shader *geometry_shader;
+      uint num_gs_outputs;  /**< convenience, from geometry_shader */
+      uint position_output;
+
+      /** TGSI program interpreter runtime state */
+      struct tgsi_exec_machine *machine;
+
+      uint num_samplers;
+      struct tgsi_sampler **samplers;
+   } gs;
+
    /* Clip derived state:
     */
    float plane[12][4];
@@ -223,7 +235,7 @@ struct draw_context
       uint semantic_name;
       uint semantic_index;
       int slot;
-   } extra_vp_outputs;
+   } extra_shader_outputs;
 
    unsigned reduced_prim;
 
@@ -246,6 +258,19 @@ void draw_vs_set_constants( struct draw_context *,
 
 
 
+/*******************************************************************************
+ * Geometry shading code:
+ */
+boolean draw_gs_init( struct draw_context *draw );
+void draw_gs_set_constants( struct draw_context *,
+                            const float (*constants)[4],
+                            unsigned size );
+
+/*******************************************************************************
+ * Common shading code:
+ */
+int draw_current_shader_outputs(struct draw_context *draw);
+int draw_current_shader_position_output(struct draw_context *draw);
 
 /*******************************************************************************
  * Vertex processing (was passthrough) code:
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index dbb5ac7182..2801dbafe4 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -192,7 +192,8 @@ draw_print_arrays(struct draw_context *draw, uint prim, int start, uint count)
                 prim, start, count);
 
    for (i = 0; i < count; i++) {
-      uint ii, j;
+      uint ii = 0;
+      uint j;
 
       if (draw->pt.user.elts) {
          /* indexed arrays */
@@ -313,12 +314,3 @@ draw_arrays(struct draw_context *draw, unsigned prim,
    /* drawing done here: */
    draw_pt_arrays(draw, prim, start, count);
 }
-
-boolean draw_pt_get_edgeflag( struct draw_context *draw,
-                              unsigned idx )
-{
-   if (draw->pt.user.edgeflag)
-      return (draw->pt.user.edgeflag[idx/32] & (1 << (idx%32))) != 0;
-   else
-      return 1;
-}
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index 7a17a9fb6b..20edf7a227 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -149,11 +149,6 @@ struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw );
 struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw);
 
 
-/* More helpers:
- */
-boolean draw_pt_get_edgeflag( struct draw_context *draw,
-                              unsigned idx );
-
 
 /*******************************************************************************
  * HW vertex emit:
@@ -217,7 +212,8 @@ boolean draw_pt_post_vs_run( struct pt_post_vs *pvs,
 void draw_pt_post_vs_prepare( struct pt_post_vs *pvs,
 			      boolean bypass_clipping,
 			      boolean bypass_viewport,
-			      boolean opengl );
+			      boolean opengl,
+			      boolean need_edgeflags );
 
 struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw );
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
index 65c3a34c34..305bfef435 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -42,11 +42,11 @@ struct pt_fetch {
    struct translate *translate;
 
    unsigned vertex_size;
-   boolean need_edgeflags;
 
    struct translate_cache *cache;
 };
 
+
 /* Perform the fetch from API vertex elements & vertex buffers, to a
  * contiguous set of float[4] attributes as required for the
  * vertex_shader->run_linear() method.
@@ -120,7 +120,12 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch,
       fetch->translate = translate_cache_find(fetch->cache, &key);
 
       {
-         static struct vertex_header vh = { 0, 1, 0, UNDEFINED_VERTEX_ID, { .0f, .0f, .0f, .0f } };
+         static struct vertex_header vh = { 0,
+                                            1,
+                                            0,
+                                            UNDEFINED_VERTEX_ID,
+                                            { .0f, .0f, .0f, .0f } };
+
 	 fetch->translate->set_buffer(fetch->translate,
 				      draw->pt.nr_vertex_buffers,
 				      &vh,
@@ -128,9 +133,6 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch,
       }
    }
 
-   fetch->need_edgeflags = ((draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
-                             draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) &&
-                            draw->pt.user.edgeflag);
 }
 
 
@@ -158,17 +160,6 @@ void draw_pt_fetch_run( struct pt_fetch *fetch,
 			count,
 			verts );
 
-   /* Edgeflags are hard to fit into a translate program, populate
-    * them separately if required.  In the setup above they are
-    * defaulted to one, so only need this if there is reason to change
-    * that default:
-    */
-   if (fetch->need_edgeflags) {
-      for (i = 0; i < count; i++) {
-         struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size);
-         vh->edgeflag = draw_pt_get_edgeflag( draw, elts[i] );
-      }
-   }
 }
 
 
@@ -193,18 +184,6 @@ void draw_pt_fetch_run_linear( struct pt_fetch *fetch,
                    start,
                    count,
                    verts );
-
-   /* Edgeflags are hard to fit into a translate program, populate
-    * them separately if required.  In the setup above they are
-    * defaulted to one, so only need this if there is reason to change
-    * that default:
-    */
-   if (fetch->need_edgeflags) {
-      for (i = 0; i < count; i++) {
-         struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size);
-         vh->edgeflag = draw_pt_get_edgeflag( draw, start + i );
-      }
-   }
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
index df6c265b7e..1a9df4cac5 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
@@ -32,6 +32,7 @@
 #include "draw/draw_vertex.h"
 #include "draw/draw_pt.h"
 #include "draw/draw_vs.h"
+#include "draw/draw_gs.h"
 #include "translate/translate.h"
 
 
@@ -85,9 +86,9 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
    draw_pt_post_vs_prepare( fpme->post_vs,
 			    (boolean)draw->bypass_clipping,
 			    (boolean)(draw->identity_viewport ||
-                                      draw->rasterizer->bypass_vs_clip_and_viewport),
-			    (boolean)draw->rasterizer->gl_rasterization_rules );
-			    
+			    draw->rasterizer->bypass_vs_clip_and_viewport),
+			    (boolean)draw->rasterizer->gl_rasterization_rules,
+			    (draw->vs.edgeflag_output ? true : false) );    
 
    if (!(opt & PT_PIPELINE)) {
       draw_pt_emit_prepare( fpme->emit, 
@@ -119,7 +120,8 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle,
 {
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
-   struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   struct draw_vertex_shader *vshader = draw->vs.vertex_shader;
+   struct draw_geometry_shader *gshader = draw->gs.geometry_shader;
    unsigned opt = fpme->opt;
    unsigned alloc_count = align( fetch_count, 4 );
 
@@ -147,13 +149,21 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle,
     */
    if (opt & PT_SHADE)
    {
-      shader->run_linear(shader, 
-			 (const float (*)[4])pipeline_verts->data,
-			 (      float (*)[4])pipeline_verts->data,
-			 (const float (*)[4])draw->pt.user.constants,
-			 fetch_count,
-			 fpme->vertex_size,
-			 fpme->vertex_size);
+      vshader->run_linear(vshader,
+                          (const float (*)[4])pipeline_verts->data,
+                          (      float (*)[4])pipeline_verts->data,
+                          (const float (*)[4])draw->pt.user.vs_constants,
+                          fetch_count,
+                          fpme->vertex_size,
+                          fpme->vertex_size);
+      if (gshader)
+         draw_geometry_shader_run(gshader,
+                                  (const float (*)[4])pipeline_verts->data,
+                                  (      float (*)[4])pipeline_verts->data,
+                                  (const float (*)[4])draw->pt.user.gs_constants,
+                                  fetch_count,
+                                  fpme->vertex_size,
+                                  fpme->vertex_size);
    }
 
    if (draw_pt_post_vs_run( fpme->post_vs,
@@ -196,6 +206,7 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle,
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
    struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   struct draw_geometry_shader *geometry_shader = draw->gs.geometry_shader;
    unsigned opt = fpme->opt;
    unsigned alloc_count = align( count, 4 );
 
@@ -226,10 +237,19 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle,
       shader->run_linear(shader,
 			 (const float (*)[4])pipeline_verts->data,
 			 (      float (*)[4])pipeline_verts->data,
-			 (const float (*)[4])draw->pt.user.constants,
+			 (const float (*)[4])draw->pt.user.vs_constants,
 			 count,
 			 fpme->vertex_size,
 			 fpme->vertex_size);
+
+      if (geometry_shader)
+         draw_geometry_shader_run(geometry_shader,
+                                  (const float (*)[4])pipeline_verts->data,
+                                  (      float (*)[4])pipeline_verts->data,
+                                  (const float (*)[4])draw->pt.user.gs_constants,
+                                  count,
+                                  fpme->vertex_size,
+                                  fpme->vertex_size);
    }
 
    if (draw_pt_post_vs_run( fpme->post_vs,
@@ -270,6 +290,7 @@ static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
    struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   struct draw_geometry_shader *geometry_shader = draw->gs.geometry_shader;
    unsigned opt = fpme->opt;
    unsigned alloc_count = align( count, 4 );
 
@@ -296,10 +317,19 @@ static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle
       shader->run_linear(shader,
 			 (const float (*)[4])pipeline_verts->data,
 			 (      float (*)[4])pipeline_verts->data,
-			 (const float (*)[4])draw->pt.user.constants,
+			 (const float (*)[4])draw->pt.user.vs_constants,
 			 count,
 			 fpme->vertex_size,
 			 fpme->vertex_size);
+
+      if (geometry_shader)
+         draw_geometry_shader_run(geometry_shader,
+                                  (const float (*)[4])pipeline_verts->data,
+                                  (      float (*)[4])pipeline_verts->data,
+                                  (const float (*)[4])draw->pt.user.gs_constants,
+                                  count,
+                                  fpme->vertex_size,
+                                  fpme->vertex_size);
    }
 
    if (draw_pt_post_vs_run( fpme->post_vs,
diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
index 6c1cb48e8b..55151823a1 100644
--- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c
+++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
@@ -100,7 +100,7 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs,
    struct vertex_header *out = vertices;
    const float *scale = pvs->draw->viewport.scale;
    const float *trans = pvs->draw->viewport.translate;
-   const unsigned pos = pvs->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(pvs->draw);
    unsigned clipped = 0;
    unsigned j;
 
@@ -147,6 +147,39 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs,
 
 
 
+/* As above plus edgeflags
+ */
+static boolean 
+post_vs_cliptest_viewport_gl_edgeflag(struct pt_post_vs *pvs,
+                                      struct vertex_header *vertices,
+                                      unsigned count,
+                                      unsigned stride )
+{
+   unsigned j;
+   boolean needpipe;
+
+   needpipe = post_vs_cliptest_viewport_gl( pvs, vertices, count, stride);
+
+   /* If present, copy edgeflag VS output into vertex header.
+    * Otherwise, leave header as is.
+    */
+   if (pvs->draw->vs.edgeflag_output) {
+      struct vertex_header *out = vertices;
+      int ef = pvs->draw->vs.edgeflag_output;
+
+      for (j = 0; j < count; j++) {
+         const float *edgeflag = out->data[ef];
+         out->edgeflag = !(edgeflag[0] != 1.0f);
+         needpipe |= !out->edgeflag;
+         out = (struct vertex_header *)( (char *)out + stride );
+      }
+   }
+   return needpipe;
+}
+
+
+
+
 /* If bypass_clipping is set, skip cliptest and rhw divide.
  */
 static boolean post_vs_viewport( struct pt_post_vs *pvs,
@@ -157,7 +190,7 @@ static boolean post_vs_viewport( struct pt_post_vs *pvs,
    struct vertex_header *out = vertices;
    const float *scale = pvs->draw->viewport.scale;
    const float *trans = pvs->draw->viewport.translate;
-   const unsigned pos = pvs->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(pvs->draw);
    unsigned j;
 
    if (0) debug_printf("%s\n", __FUNCTION__);
@@ -201,17 +234,29 @@ boolean draw_pt_post_vs_run( struct pt_post_vs *pvs,
 void draw_pt_post_vs_prepare( struct pt_post_vs *pvs,
 			      boolean bypass_clipping,
 			      boolean bypass_viewport,
-			      boolean opengl )
+			      boolean opengl,
+			      boolean need_edgeflags )
 {
-   if (bypass_clipping) {
-      if (bypass_viewport)
-	 pvs->run = post_vs_none;
-      else
-	 pvs->run = post_vs_viewport;
+   if (!need_edgeflags) {
+      if (bypass_clipping) {
+         if (bypass_viewport)
+            pvs->run = post_vs_none;
+         else
+            pvs->run = post_vs_viewport;
+      }
+      else {
+         /* if (opengl) */
+         pvs->run = post_vs_cliptest_viewport_gl;
+      }
    }
    else {
-      /* if (opengl) */
-      pvs->run = post_vs_cliptest_viewport_gl;
+      /* If we need to copy edgeflags to the vertex header, it should
+       * mean we're running the primitive pipeline.  Hence the bypass
+       * flags should be false.
+       */
+      assert(!bypass_clipping);
+      assert(!bypass_viewport);
+      pvs->run = post_vs_cliptest_viewport_gl_edgeflag;
    }
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c
index b61fa29143..17c3b8cec2 100644
--- a/src/gallium/auxiliary/draw/draw_pt_util.c
+++ b/src/gallium/auxiliary/draw/draw_pt_util.c
@@ -50,16 +50,32 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       *first = 2;
       *incr = 1;
       break;
+   case PIPE_PRIM_LINES_ADJACENCY:
+      *first = 4;
+      *incr = 2;
+      break;
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      *first = 4;
+      *incr = 1;
+      break;
    case PIPE_PRIM_TRIANGLES:
       *first = 3;
       *incr = 3;
       break;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+      *first = 6;
+      *incr = 3;
+      break;
    case PIPE_PRIM_TRIANGLE_STRIP:
    case PIPE_PRIM_TRIANGLE_FAN:
    case PIPE_PRIM_POLYGON:
       *first = 3;
       *incr = 1;
       break;
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      *first = 6;
+      *incr = 1;
+      break;
    case PIPE_PRIM_QUADS:
       *first = 4;
       *incr = 4;
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
index 010c7a18a7..f0aec5feba 100644
--- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
+++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
@@ -36,6 +36,10 @@ static void FUNC(struct draw_pt_front_end *frontend,
    case PIPE_PRIM_TRIANGLE_STRIP:
    case PIPE_PRIM_QUADS:
    case PIPE_PRIM_QUAD_STRIP:
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
       for (j = 0; j < count;) {
          unsigned remaining = count - j;
          unsigned nr = trim( MIN2(varray->driver_fetch_max, remaining), first, incr );
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
index 1a0527be63..757c487454 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -346,7 +346,8 @@ vcache_check_run( struct draw_pt_front_end *frontend,
                        vcache->fetch_max,
                        draw_count);
       
-   if (max_index == 0xffffffff ||
+   if (max_index >= DRAW_PIPE_MAX_VERTICES ||
+       fetch_count >= UNDEFINED_VERTEX_ID ||
        fetch_count > draw_count) {
       if (0) debug_printf("fail\n");
       goto fail;
@@ -394,6 +395,7 @@ vcache_check_run( struct draw_pt_front_end *frontend,
 
          default:
             assert(0);
+            FREE(storage);
             return;
          }
       }
@@ -422,6 +424,7 @@ vcache_check_run( struct draw_pt_front_end *frontend,
 
          default:
             assert(0);
+            FREE(storage);
             return;
          }
       }
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index 790e89ed82..3553689532 100644
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -101,6 +101,9 @@ draw_create_vertex_shader(struct draw_context *draw,
          if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION &&
              vs->info.output_semantic_index[i] == 0)
             vs->position_output = i;
+         else if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_EDGEFLAG &&
+             vs->info.output_semantic_index[i] == 0)
+            vs->edgeflag_output = i;
       }
    }
 
@@ -120,6 +123,7 @@ draw_bind_vertex_shader(struct draw_context *draw,
       draw->vs.vertex_shader = dvs;
       draw->vs.num_vs_outputs = dvs->info.num_outputs;
       draw->vs.position_output = dvs->position_output;
+      draw->vs.edgeflag_output = dvs->edgeflag_output;
       dvs->prepare( dvs, draw );
    }
    else {
diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h
index 89ae158751..e3b807ebd0 100644
--- a/src/gallium/auxiliary/draw/draw_vs.h
+++ b/src/gallium/auxiliary/draw/draw_vs.h
@@ -107,6 +107,7 @@ struct draw_vertex_shader {
 
    struct tgsi_shader_info info;
    unsigned position_output;
+   unsigned edgeflag_output;
 
    /* Extracted from shader:
     */
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c
index 645d7cccba..1aaae4ab7a 100644
--- a/src/gallium/auxiliary/draw/draw_vs_aos.c
+++ b/src/gallium/auxiliary/draw/draw_vs_aos.c
@@ -361,8 +361,8 @@ static struct x86_reg aos_get_shader_reg_ptr( struct aos_compilation *cp,
 static struct x86_reg get_dst_ptr( struct aos_compilation *cp, 
                                    const struct tgsi_full_dst_register *dst )
 {
-   unsigned file = dst->DstRegister.File;
-   unsigned idx = dst->DstRegister.Index;
+   unsigned file = dst->Register.File;
+   unsigned idx = dst->Register.Index;
    unsigned i;
    
 
@@ -529,27 +529,18 @@ static struct x86_reg fetch_src( struct aos_compilation *cp,
                                  const struct tgsi_full_src_register *src ) 
 {
    struct x86_reg arg0 = aos_get_shader_reg(cp, 
-                                            src->SrcRegister.File, 
-                                            src->SrcRegister.Index);
+                                            src->Register.File, 
+                                            src->Register.Index);
    unsigned i;
    ubyte swz = 0;
    unsigned negs = 0;
    unsigned abs = 0;
 
    for (i = 0; i < 4; i++) {
-      unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( src, i );
+      unsigned swizzle = tgsi_util_get_full_src_register_swizzle( src, i );
       unsigned neg = tgsi_util_get_full_src_register_sign_mode( src, i );
 
-      switch (swizzle) {
-      case TGSI_EXTSWIZZLE_ZERO:
-      case TGSI_EXTSWIZZLE_ONE:
-         AOS_ERROR(cp, "not supporting full swizzles yet in tgsi_aos_sse2");
-         break;
-
-      default:
-         swz |= (swizzle & 0x3) << (i * 2);
-         break;
-      }
+      swz |= (swizzle & 0x3) << (i * 2);
 
       switch (neg) {
       case TGSI_UTIL_SIGN_TOGGLE:
@@ -629,26 +620,13 @@ static void x87_fld_src( struct aos_compilation *cp,
                          unsigned channel ) 
 {
    struct x86_reg arg0 = aos_get_shader_reg_ptr(cp, 
-                                                src->SrcRegister.File, 
-                                                src->SrcRegister.Index);
+                                                src->Register.File, 
+                                                src->Register.Index);
 
-   unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( src, channel );
+   unsigned swizzle = tgsi_util_get_full_src_register_swizzle( src, channel );
    unsigned neg = tgsi_util_get_full_src_register_sign_mode( src, channel );
 
-   switch (swizzle) {
-   case TGSI_EXTSWIZZLE_ZERO:
-      x87_fldz( cp->func );
-      break;
-
-   case TGSI_EXTSWIZZLE_ONE:
-      x87_fld1( cp->func );
-      break;
-
-   default:
-      x87_fld( cp->func, x86_make_disp(arg0, (swizzle & 3) * sizeof(float)) );
-      break;
-   }
-   
+   x87_fld( cp->func, x86_make_disp(arg0, (swizzle & 3) * sizeof(float)) );
 
    switch (neg) {
    case TGSI_UTIL_SIGN_TOGGLE:
@@ -691,15 +669,15 @@ static void store_dest( struct aos_compilation *cp,
 {
    struct x86_reg dst;
 
-   switch (reg->DstRegister.WriteMask) {
+   switch (reg->Register.WriteMask) {
    case 0:
       return;
    
    case TGSI_WRITEMASK_XYZW:
       aos_adopt_xmm_reg(cp, 
                         get_xmm_writable(cp, result), 
-                        reg->DstRegister.File,
-                        reg->DstRegister.Index,
+                        reg->Register.File,
+                        reg->Register.Index,
                         TRUE);
       return;
    default: 
@@ -707,10 +685,10 @@ static void store_dest( struct aos_compilation *cp,
    }
 
    dst = aos_get_shader_reg_xmm(cp, 
-                                reg->DstRegister.File,
-                                reg->DstRegister.Index);
+                                reg->Register.File,
+                                reg->Register.Index);
 
-   switch (reg->DstRegister.WriteMask) {
+   switch (reg->Register.WriteMask) {
    case TGSI_WRITEMASK_X:
       sse_movss(cp->func, dst, get_xmm(cp, result));
       break;
@@ -732,14 +710,14 @@ static void store_dest( struct aos_compilation *cp,
       break;
 
    default:
-      mask_write(cp, dst, result, reg->DstRegister.WriteMask);
+      mask_write(cp, dst, result, reg->Register.WriteMask);
       break;
    }
 
    aos_adopt_xmm_reg(cp, 
                      dst, 
-                     reg->DstRegister.File,
-                     reg->DstRegister.Index,
+                     reg->Register.File,
+                     reg->Register.Index,
                      TRUE);
 
 }
@@ -759,7 +737,7 @@ static void store_scalar_dest( struct aos_compilation *cp,
                                const struct tgsi_full_dst_register *reg,
                                struct x86_reg result )
 {
-   unsigned writemask = reg->DstRegister.WriteMask;
+   unsigned writemask = reg->Register.WriteMask;
    struct x86_reg dst;
 
    if (writemask != TGSI_WRITEMASK_X &&
@@ -776,12 +754,12 @@ static void store_scalar_dest( struct aos_compilation *cp,
 
    result = get_xmm(cp, result);
    dst = aos_get_shader_reg_xmm(cp, 
-                                reg->DstRegister.File,
-                                reg->DstRegister.Index);
+                                reg->Register.File,
+                                reg->Register.Index);
 
 
 
-   switch (reg->DstRegister.WriteMask) {
+   switch (reg->Register.WriteMask) {
    case TGSI_WRITEMASK_X:
       sse_movss(cp->func, dst, result);
       break;
@@ -804,8 +782,8 @@ static void store_scalar_dest( struct aos_compilation *cp,
 
    aos_adopt_xmm_reg(cp, 
                      dst, 
-                     reg->DstRegister.File,
-                     reg->DstRegister.Index,
+                     reg->Register.File,
+                     reg->Register.Index,
                      TRUE);
 }
    
@@ -841,7 +819,7 @@ static void x87_fstp_dest4( struct aos_compilation *cp,
                             const struct tgsi_full_dst_register *dst )
 {
    struct x86_reg ptr = get_dst_ptr(cp, dst); 
-   unsigned writemask = dst->DstRegister.WriteMask;
+   unsigned writemask = dst->Register.WriteMask;
 
    x87_fst_or_nop(cp->func, writemask, 0, ptr);
    x87_fst_or_nop(cp->func, writemask, 1, ptr);
@@ -978,7 +956,7 @@ static void emit_print( struct aos_compilation *cp,
 
 static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
    struct x86_reg neg = aos_get_internal(cp, IMM_NEGS);
    struct x86_reg tmp = aos_get_xmm_reg(cp);
 
@@ -986,27 +964,27 @@ static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_inst
    sse_mulps(cp->func, tmp, neg);
    sse_maxps(cp->func, tmp, arg0);
    
-   store_dest(cp, &op->FullDstRegisters[0], tmp);
+   store_dest(cp, &op->Dst[0], tmp);
    return TRUE;
 }
 
 static boolean emit_ADD( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    sse_addps(cp->func, dst, arg1);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_COS( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);
+   x87_fld_src(cp, &op->Src[0], 0);
    x87_fcos(cp->func);
-   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   x87_fstp_dest4(cp, &op->Dst[0]);
    return TRUE;
 }
 
@@ -1015,8 +993,8 @@ static boolean emit_COS( struct aos_compilation *cp, const struct tgsi_full_inst
  */
 static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg tmp = aos_get_xmm_reg(cp); 
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
@@ -1029,14 +1007,14 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst
    sse_addss(cp->func, dst, tmp);
    
    aos_release_xmm_reg(cp, tmp.idx);
-   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   store_scalar_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg tmp = aos_get_xmm_reg(cp);      
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
@@ -1050,14 +1028,14 @@ static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_inst
    sse_addss(cp->func, dst, tmp);
 
    aos_release_xmm_reg(cp, tmp.idx);
-   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   store_scalar_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg tmp = aos_get_xmm_reg(cp);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
@@ -1073,14 +1051,14 @@ static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_inst
    sse_addss(cp->func, dst, tmp);
 
    aos_release_xmm_reg(cp, tmp.idx);
-   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   store_scalar_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_DST( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-    struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-    struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+    struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+    struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
     struct x86_reg dst = aos_get_xmm_reg(cp);
     struct x86_reg tmp = aos_get_xmm_reg(cp);
     struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
@@ -1095,25 +1073,25 @@ static boolean emit_DST( struct aos_compilation *cp, const struct tgsi_full_inst
     sse_mulps(cp->func, dst, tmp);
 
     aos_release_xmm_reg(cp, tmp.idx);
-    store_dest(cp, &op->FullDstRegisters[0], dst);
+    store_dest(cp, &op->Dst[0], dst);
     return TRUE;
 }
 
 static boolean emit_LG2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
    x87_fld1(cp->func);		/* 1 */
-   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);	/* a0 1 */
+   x87_fld_src(cp, &op->Src[0], 0);	/* a0 1 */
    x87_fyl2x(cp->func);	/* log2(a0) */
-   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   x87_fstp_dest4(cp, &op->Dst[0]);
    return TRUE;
 }
 
 #if 0
 static boolean emit_EX2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);
+   x87_fld_src(cp, &op->Src[0], 0);
    x87_emit_ex2(cp);
-   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   x87_fstp_dest4(cp, &op->Dst[0]);
    return TRUE;
 }
 #endif
@@ -1121,8 +1099,8 @@ static boolean emit_EX2( struct aos_compilation *cp, const struct tgsi_full_inst
 
 static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
-   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   struct x86_reg dst = get_dst_ptr(cp, &op->Dst[0]); 
+   unsigned writemask = op->Dst[0].Register.WriteMask;
    int i;
 
    set_fpu_round_neg_inf( cp );
@@ -1131,7 +1109,7 @@ static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_inst
     */
    for (i = 3; i >= 0; i--) {
       if (writemask & (1<<i)) {
-         x87_fld_src(cp, &op->FullSrcRegisters[0], i);   
+         x87_fld_src(cp, &op->Src[0], i);   
       }
    }
 
@@ -1148,8 +1126,8 @@ static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_inst
 
 static boolean emit_RND( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
-   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   struct x86_reg dst = get_dst_ptr(cp, &op->Dst[0]); 
+   unsigned writemask = op->Dst[0].Register.WriteMask;
    int i;
 
    set_fpu_round_nearest( cp );
@@ -1158,7 +1136,7 @@ static boolean emit_RND( struct aos_compilation *cp, const struct tgsi_full_inst
     */
    for (i = 3; i >= 0; i--) {
       if (writemask & (1<<i)) {
-         x87_fld_src(cp, &op->FullSrcRegisters[0], i);   
+         x87_fld_src(cp, &op->Src[0], i);   
       }
    }
 
@@ -1175,10 +1153,10 @@ static boolean emit_RND( struct aos_compilation *cp, const struct tgsi_full_inst
 
 static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
+   struct x86_reg dst = get_dst_ptr(cp, &op->Dst[0]); 
    struct x86_reg st0 = x86_make_reg(file_x87, 0);
    struct x86_reg st1 = x86_make_reg(file_x87, 1);
-   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   unsigned writemask = op->Dst[0].Register.WriteMask;
    int i;
 
    set_fpu_round_neg_inf( cp );
@@ -1188,7 +1166,7 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst
     */
    for (i = 3; i >= 0; i--) {
       if (writemask & (1<<i)) {
-         x87_fld_src(cp, &op->FullSrcRegisters[0], i);   
+         x87_fld_src(cp, &op->Src[0], i);   
       }
    }
 
@@ -1212,7 +1190,7 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst
 static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
    struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
-   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   unsigned writemask = op->Dst[0].Register.WriteMask;
    unsigned lit_count = cp->lit_count++;
    struct x86_reg result, arg0;
    unsigned i;
@@ -1231,10 +1209,10 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst
    if (writemask != TGSI_WRITEMASK_XYZW) 
       result = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, tmp[0]));
    else 
-      result = get_dst_ptr(cp, &op->FullDstRegisters[0]);    
+      result = get_dst_ptr(cp, &op->Dst[0]);    
 
    
-   arg0 = fetch_src( cp, &op->FullSrcRegisters[0] );
+   arg0 = fetch_src( cp, &op->Src[0] );
    if (arg0.file == file_XMM) {
       struct x86_reg tmp = x86_make_disp(cp->machine_EDX, 
                                          Offset(struct aos_machine, tmp[1]));
@@ -1281,7 +1259,7 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst
 
    if (writemask != TGSI_WRITEMASK_XYZW) {
       store_dest( cp, 
-                  &op->FullDstRegisters[0],
+                  &op->Dst[0],
                   get_xmm_writable( cp, result ) );
    }
 
@@ -1291,8 +1269,8 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst
 #if 0   
 static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
-   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   struct x86_reg dst = get_dst_ptr(cp, &op->Dst[0]); 
+   unsigned writemask = op->Dst[0].Register.WriteMask;
 
    if (writemask & TGSI_WRITEMASK_YZ) {
       struct x86_reg st1 = x86_make_reg(file_x87, 1);
@@ -1308,13 +1286,13 @@ static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_fu
        */
       x87_fldz(cp->func);                           /* 1 0  */
 #endif
-      x87_fld_src(cp, &op->FullSrcRegisters[0], 1); /* a1 1 0  */
+      x87_fld_src(cp, &op->Src[0], 1); /* a1 1 0  */
       x87_fcomi(cp->func, st2);	                    /* a1 1 0  */
       x87_fcmovb(cp->func, st1);                    /* a1' 1 0  */
       x87_fstp(cp->func, st1);                      /* a1' 0  */
       x87_fstp(cp->func, st1);                      /* a1'  */
 
-      x87_fld_src(cp, &op->FullSrcRegisters[0], 3); /* a3 a1'  */
+      x87_fld_src(cp, &op->Src[0], 3); /* a3 a1'  */
       x87_fxch(cp->func, st1);                      /* a1' a3  */
       
 
@@ -1327,7 +1305,7 @@ static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_fu
       /* a0' = max2(a0, 0):
        */
       x87_fldz(cp->func);                           /* 0 r2 */
-      x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0 0 r2 */
+      x87_fld_src(cp, &op->Src[0], 0); /* a0 0 r2 */
       x87_fcomi(cp->func, st1);	
       x87_fcmovb(cp->func, st1);                    /* a0' 0 r2 */
 
@@ -1355,58 +1333,58 @@ static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_fu
 
 static boolean emit_MAX( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    sse_maxps(cp->func, dst, arg1);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 
 static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    sse_minps(cp->func, dst, arg1);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_MOV( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    /* potentially nothing to do */
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_MUL( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    sse_mulps(cp->func, dst, arg1);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 
 static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
-   struct x86_reg arg2 = fetch_src(cp, &op->FullSrcRegisters[2]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
+   struct x86_reg arg2 = fetch_src(cp, &op->Src[2]);
 
    /* If we can't clobber old contents of arg0, get a temporary & copy
     * it there, then clobber it...
@@ -1415,7 +1393,7 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst
 
    sse_mulps(cp->func, arg0, arg1);
    sse_addps(cp->func, arg0, arg2);
-   store_dest(cp, &op->FullDstRegisters[0], arg0);
+   store_dest(cp, &op->Dst[0], arg0);
    return TRUE;
 }
 
@@ -1447,13 +1425,13 @@ static float PIPE_CDECL _exp2(float x)
 static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
 #if 0
-   x87_fld_src(cp, &op->FullSrcRegisters[1], 0);  /* a1.x */
-   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);	/* a0.x a1.x */
+   x87_fld_src(cp, &op->Src[1], 0);  /* a1.x */
+   x87_fld_src(cp, &op->Src[0], 0);	/* a0.x a1.x */
    x87_fyl2x(cp->func);	                                /* a1*log2(a0) */
 
    x87_emit_ex2( cp );		/* 2^(a1*log2(a0)) */
 
-   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   x87_fstp_dest4(cp, &op->Dst[0]);
 #else
    uint i;
 
@@ -1472,9 +1450,9 @@ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_inst
 
    x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, -8) );
 
-   x87_fld_src( cp, &op->FullSrcRegisters[1], 0 );
+   x87_fld_src( cp, &op->Src[1], 0 );
    x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 4 ) );
-   x87_fld_src( cp, &op->FullSrcRegisters[0], 0 );
+   x87_fld_src( cp, &op->Src[0], 0 );
    x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) );
 
    /* tmp_EAX has been pushed & will be restored below */
@@ -1489,7 +1467,7 @@ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_inst
     */
    cp->func->x87_stack++;
 
-   x87_fstp_dest4( cp, &op->FullDstRegisters[0] );
+   x87_fstp_dest4( cp, &op->Dst[0] );
 #endif
    return TRUE;
 }
@@ -1515,7 +1493,7 @@ static boolean emit_EXPBASE2( struct aos_compilation *cp, const struct tgsi_full
 
    x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, -4) );
 
-   x87_fld_src( cp, &op->FullSrcRegisters[0], 0 );
+   x87_fld_src( cp, &op->Src[0], 0 );
    x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) );
 
    /* tmp_EAX has been pushed & will be restored below */
@@ -1530,7 +1508,7 @@ static boolean emit_EXPBASE2( struct aos_compilation *cp, const struct tgsi_full
     */
    cp->func->x87_stack++;
 
-   x87_fstp_dest4( cp, &op->FullDstRegisters[0] );
+   x87_fstp_dest4( cp, &op->Dst[0] );
 
    return TRUE;
 }
@@ -1539,7 +1517,7 @@ static boolean emit_EXPBASE2( struct aos_compilation *cp, const struct tgsi_full
 
 static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
    struct x86_reg dst = aos_get_xmm_reg(cp);
 
    if (cp->have_sse2) {
@@ -1553,7 +1531,7 @@ static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_inst
       sse_divss(cp->func, dst, arg0);
    }
 
-   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   store_scalar_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
@@ -1573,14 +1551,14 @@ static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_inst
 static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
    if (0) {
-      struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+      struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
       struct x86_reg r = aos_get_xmm_reg(cp);
       sse_rsqrtss(cp->func, r, arg0);
-      store_scalar_dest(cp, &op->FullDstRegisters[0], r);
+      store_scalar_dest(cp, &op->Dst[0], r);
       return TRUE;
    }
    else {
-      struct x86_reg arg0           = fetch_src(cp, &op->FullSrcRegisters[0]);
+      struct x86_reg arg0           = fetch_src(cp, &op->Src[0]);
       struct x86_reg r              = aos_get_xmm_reg(cp);
 
       struct x86_reg neg_half       = get_reg_ptr( cp, AOS_FILE_INTERNAL, IMM_RSQ );
@@ -1600,7 +1578,7 @@ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_inst
       sse_addss(   cp->func, tmp, one_point_five ); /* 1.5 - .5 * a * r * r */
       sse_mulss(   cp->func, r,  tmp );             /* r * (1.5 - .5 * a * r * r) */
 
-      store_scalar_dest(cp, &op->FullDstRegisters[0], r);
+      store_scalar_dest(cp, &op->Dst[0], r);
 
       aos_release_xmm_reg(cp, tmp.idx);
 
@@ -1611,23 +1589,23 @@ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_inst
 
 static boolean emit_SGE( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    sse_cmpps(cp->func, dst, arg1, cc_NotLessThan);
    sse_andps(cp->func, dst, ones);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_SIN( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);
+   x87_fld_src(cp, &op->Src[0], 0);
    x87_fsin(cp->func);
-   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   x87_fstp_dest4(cp, &op->Dst[0]);
    return TRUE;
 }
 
@@ -1635,46 +1613,46 @@ static boolean emit_SIN( struct aos_compilation *cp, const struct tgsi_full_inst
 
 static boolean emit_SLT( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
    
    sse_cmpps(cp->func, dst, arg1, cc_LessThan);
    sse_andps(cp->func, dst, ones);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_SUB( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg dst = get_xmm_writable(cp, arg0);
 
    sse_subps(cp->func, dst, arg1);
 
-   store_dest(cp, &op->FullDstRegisters[0], dst);
+   store_dest(cp, &op->Dst[0], dst);
    return TRUE;
 }
 
 static boolean emit_TRUNC( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
    struct x86_reg tmp0 = aos_get_xmm_reg(cp);
 
    sse2_cvttps2dq(cp->func, tmp0, arg0);
    sse2_cvtdq2ps(cp->func, tmp0, tmp0);
 
-   store_dest(cp, &op->FullDstRegisters[0], tmp0);
+   store_dest(cp, &op->Dst[0], tmp0);
    return TRUE;
 }
 
 static boolean emit_XPD( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
 {
-   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
-   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg0 = fetch_src(cp, &op->Src[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->Src[1]);
    struct x86_reg tmp0 = aos_get_xmm_reg(cp);
    struct x86_reg tmp1 = aos_get_xmm_reg(cp);
 
@@ -1692,7 +1670,7 @@ static boolean emit_XPD( struct aos_compilation *cp, const struct tgsi_full_inst
 
 
    aos_release_xmm_reg(cp, tmp0.idx);
-   store_dest(cp, &op->FullDstRegisters[0], tmp1);
+   store_dest(cp, &op->Dst[0], tmp1);
    return TRUE;
 }
 
@@ -1919,10 +1897,10 @@ static void find_last_write_outputs( struct aos_compilation *cp )
          continue;
 
       for (i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++) {
-         if (parse.FullToken.FullInstruction.FullDstRegisters[i].DstRegister.File ==
+         if (parse.FullToken.FullInstruction.Dst[i].Register.File ==
              TGSI_FILE_OUTPUT) 
          {
-            unsigned idx = parse.FullToken.FullInstruction.FullDstRegisters[i].DstRegister.Index;
+            unsigned idx = parse.FullToken.FullInstruction.Dst[i].Register.Index;
             cp->output_last_write[idx] = this_instruction;
          }
       }
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index 7ee567d478..d16692584e 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -147,11 +147,12 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs, 
                               temp_buffer,
                               temp_buffer,
-                              (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
+                              (const float (*)[4])vsvg->base.vs->draw->pt.user.vs_constants,
                               count,
                               temp_vertex_stride, 
                               temp_vertex_stride);
 
+   /* FIXME: geometry shading? */
 
    if (vsvg->base.key.clip) {
       /* not really handling clipping, just do the rhw so we can
@@ -207,7 +208,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs, 
                               temp_buffer,
                               temp_buffer,
-                              (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
+                              (const float (*)[4])vsvg->base.vs->draw->pt.user.vs_constants,
                               count,
                               temp_vertex_stride, 
                               temp_vertex_stride);
diff --git a/src/gallium/auxiliary/gallivm/Makefile b/src/gallium/auxiliary/gallivm/Makefile
deleted file mode 100644
index 5a96d94ec3..0000000000
--- a/src/gallium/auxiliary/gallivm/Makefile
+++ /dev/null
@@ -1,92 +0,0 @@
-# -*-makefile-*-
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = gallivm
-
-
-GALLIVM_SOURCES = \
-        gallivm.cpp  \
-        gallivm_cpu.cpp \
-        instructions.cpp  \
-        loweringpass.cpp \
-        tgsitollvm.cpp \
-        storage.cpp \
-        storagesoa.cpp \
-        instructionssoa.cpp
-
-INC_SOURCES = gallivm_builtins.cpp gallivmsoabuiltins.cpp
-
-CPP_SOURCES = \
-	$(GALLIVM_SOURCES)
-
-C_SOURCES =
-ASM_SOURCES =
-
-OBJECTS = $(C_SOURCES:.c=.o) \
-          $(CPP_SOURCES:.cpp=.o) \
-	  $(ASM_SOURCES:.S=.o)
-
-### Include directories
-INCLUDES = \
-	-I. \
-	-I$(TOP)/src/gallium/drivers \
-	-I$(TOP)/src/gallium/auxiliary \
-	-I$(TOP)/src/gallium/include \
-	-I$(TOP)/src/mesa \
-	-I$(TOP)/include
-
-
-##### RULES #####
-
-.c.o:
-	$(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
-
-.cpp.o:
-	$(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@
-
-.S.o:
-	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
-
-##### TARGETS #####
-
-default:: depend symlinks $(LIBNAME)
-
-
-$(LIBNAME): $(OBJECTS) Makefile
-	$(TOP)/bin/mklib -o $@ -static $(OBJECTS)
-
-
-depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES)
-	rm -f depend
-	touch depend
-	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \
-		$(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null
-
-
-gallivm_builtins.cpp: llvm_builtins.c
-	clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp1.bin
-	(echo "static const unsigned char llvm_builtins_data[] = {"; od -txC temp1.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/,0x00};/") >$@
-	rm temp1.bin
-
-gallivmsoabuiltins.cpp: soabuiltins.c
-	clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp2.bin
-	(echo "static const unsigned char soabuiltins_data[] = {"; od -txC temp2.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/,0x00};/") >$@
-	rm temp2.bin
-
-# Emacs tags
-tags:
-	etags `find . -name \*.[ch]` `find ../include`
-
-
-# Remove .o and backup files
-clean:
-	-rm -f *.o */*.o *~ *.so *~ server/*.o
-	-rm -f depend depend.bak
-	-rm -f gallivm_builtins.cpp
-	-rm -f gallivmsoabuiltins.cpp
-
-symlinks:
-
-
-include depend
diff --git a/src/gallium/auxiliary/gallivm/SConscript b/src/gallium/auxiliary/gallivm/SConscript
deleted file mode 100644
index c0aa51b90a..0000000000
--- a/src/gallium/auxiliary/gallivm/SConscript
+++ /dev/null
@@ -1,16 +0,0 @@
-Import('*')
-
-gallivm = env.ConvenienceLibrary(
-	target = 'gallivm',
-	source = [
-        'gallivm.cpp',
-        'gallivm_cpu.cpp',
-        'instructions.cpp',
-        'loweringpass.cpp',
-        'tgsitollvm.cpp',
-        'storage.cpp',
-        'storagesoa.cpp',
-        'instructionssoa.cpp',
-	])
-
-auxiliaries.insert(0, gallivm)
diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
index bf84401e11..8f7d3b7100 100644
--- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
+++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
@@ -94,8 +94,8 @@ translate_declaration(struct gallivm_ir *prog,
       unsigned first, last, mask;
       uint interp_method;
 
-      first = decl->DeclarationRange.First;
-      last = decl->DeclarationRange.Last;
+      first = decl->Range.First;
+      last = decl->Range.Last;
       mask = decl->Declaration.UsageMask;
 
       /* Do not touch WPOS.xy */
@@ -149,7 +149,7 @@ translate_declarationir(struct gallivm_ir *,
                       struct tgsi_full_declaration *)
 {
    if (decl->Declaration.File == TGSI_FILE_ADDRESS) {
-      int idx = decl->DeclarationRange.First;
+      int idx = decl->Range.First;
       storage->addAddress(idx);
    }
 }
@@ -234,26 +234,26 @@ translate_instruction(llvm::Module *module,
    inputs[3] = 0;
 
    for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) {
-      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      struct tgsi_full_src_register *src = &inst->Src[i];
       llvm::Value *val = 0;
       llvm::Value *indIdx = 0;
 
-      if (src->SrcRegister.Indirect) {
-         indIdx = storage->addrElement(src->SrcRegisterInd.Index);
+      if (src->Register.Indirect) {
+         indIdx = storage->addrElement(src->Indirect.Index);
          indIdx = storage->extractIndex(indIdx);
       }
-      if (src->SrcRegister.File == TGSI_FILE_CONSTANT) {
-         val = storage->constElement(src->SrcRegister.Index, indIdx);
-      } else if (src->SrcRegister.File == TGSI_FILE_INPUT) {
-         val = storage->inputElement(src->SrcRegister.Index, indIdx);
-      } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) {
-         val = storage->tempElement(src->SrcRegister.Index);
-      } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) {
-         val = storage->outputElement(src->SrcRegister.Index, indIdx);
-      } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) {
-         val = storage->immediateElement(src->SrcRegister.Index);
+      if (src->Register.File == TGSI_FILE_CONSTANT) {
+         val = storage->constElement(src->Register.Index, indIdx);
+      } else if (src->Register.File == TGSI_FILE_INPUT) {
+         val = storage->inputElement(src->Register.Index, indIdx);
+      } else if (src->Register.File == TGSI_FILE_TEMPORARY) {
+         val = storage->tempElement(src->Register.Index);
+      } else if (src->Register.File == TGSI_FILE_OUTPUT) {
+         val = storage->outputElement(src->Register.Index, indIdx);
+      } else if (src->Register.File == TGSI_FILE_IMMEDIATE) {
+         val = storage->immediateElement(src->Register.Index);
       } else {
-         fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File);
+         fprintf(stderr, "ERROR: not supported llvm source %d\n", src->Register.File);
          return;
       }
 
@@ -552,7 +552,7 @@ translate_instruction(llvm::Module *module,
       break;
    case TGSI_OPCODE_SHL:
       break;
-   case TGSI_OPCODE_SHR:
+   case TGSI_OPCODE_ISHR:
       break;
    case TGSI_OPCODE_AND:
       break;
@@ -656,14 +656,14 @@ translate_instruction(llvm::Module *module,
 
    /* store results  */
    for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) {
-      struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
-
-      if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
-         storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
-      } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) {
-         storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
-      } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) {
-         storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+      struct tgsi_full_dst_register *dst = &inst->Dst[i];
+
+      if (dst->Register.File == TGSI_FILE_OUTPUT) {
+         storage->setOutputElement(dst->Register.Index, out, dst->Register.WriteMask);
+      } else if (dst->Register.File == TGSI_FILE_TEMPORARY) {
+         storage->setTempElement(dst->Register.Index, out, dst->Register.WriteMask);
+      } else if (dst->Register.File == TGSI_FILE_ADDRESS) {
+         storage->setAddrElement(dst->Register.Index, out, dst->Register.WriteMask);
       } else {
          fprintf(stderr, "ERROR: unsupported LLVM destination!");
          assert(!"wrong destination");
@@ -683,16 +683,16 @@ translate_instructionir(llvm::Module *module,
    std::vector< std::vector<llvm::Value*> > inputs(inst->Instruction.NumSrcRegs);
 
    for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) {
-      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      struct tgsi_full_src_register *src = &inst->Src[i];
       std::vector<llvm::Value*> val;
       llvm::Value *indIdx = 0;
       int swizzle = swizzleInt(src);
 
-      if (src->SrcRegister.Indirect) {
-         indIdx = storage->addrElement(src->SrcRegisterInd.Index);
+      if (src->Register.Indirect) {
+         indIdx = storage->addrElement(src->Indirect.Index);
       }
-      val = storage->load((enum tgsi_file_type)src->SrcRegister.File,
-                          src->SrcRegister.Index, swizzle, instr->getIRBuilder(), indIdx);
+      val = storage->load((enum tgsi_file_type)src->Register.File,
+                          src->Register.Index, swizzle, instr->getIRBuilder(), indIdx);
 
       inputs[i] = val;
    }
@@ -919,7 +919,7 @@ translate_instructionir(llvm::Module *module,
       break;
    case TGSI_OPCODE_SHL:
       break;
-   case TGSI_OPCODE_SHR:
+   case TGSI_OPCODE_ISHR:
       break;
    case TGSI_OPCODE_AND:
       break;
@@ -993,9 +993,9 @@ translate_instructionir(llvm::Module *module,
 
    /* store results  */
    for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) {
-      struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
-      storage->store((enum tgsi_file_type)dst->DstRegister.File,
-                     dst->DstRegister.Index, out, dst->DstRegister.WriteMask,
+      struct tgsi_full_dst_register *dst = &inst->Dst[i];
+      storage->store((enum tgsi_file_type)dst->Register.File,
+                     dst->Register.Index, out, dst->Register.WriteMask,
 		     instr->getIRBuilder() );
    }
 }
diff --git a/src/gallium/auxiliary/indices/Makefile b/src/gallium/auxiliary/indices/Makefile
deleted file mode 100644
index f2ebc3f410..0000000000
--- a/src/gallium/auxiliary/indices/Makefile
+++ /dev/null
@@ -1,16 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = indices
-
-C_SOURCES = \
-	u_indices_gen.c \
-	u_unfilled_gen.c 
-
-include ../../Makefile.template
-
-u_indices_gen.c: u_indices_gen.py
-	python $< > $@
-
-u_unfilled_gen.c: u_unfilled_gen.py
-	python $< > $@
diff --git a/src/gallium/auxiliary/indices/SConscript b/src/gallium/auxiliary/indices/SConscript
deleted file mode 100644
index 712e215534..0000000000
--- a/src/gallium/auxiliary/indices/SConscript
+++ /dev/null
@@ -1,28 +0,0 @@
-Import('*')
-
-from sys import executable as python_cmd
-
-env.CodeGenerate(
-	target = 'u_indices_gen.c', 
-	script = 'u_indices_gen.py', 
-	source = [],
-	command = python_cmd + ' $SCRIPT > $TARGET'
-)
-
-env.CodeGenerate(
-	target = 'u_unfilled_gen.c', 
-	script = 'u_unfilled_gen.py', 
-	source = [],
-	command = python_cmd + ' $SCRIPT > $TARGET'
-)
-
-indices = env.ConvenienceLibrary(
-	target = 'indices',
-	source = [
-#               'u_indices.c',
-#               'u_unfilled_indices.c',
-               'u_indices_gen.c',
-               'u_unfilled_gen.c',
-	])
-
-auxiliaries.insert(0, indices)
diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile
deleted file mode 100644
index 1c00ba8d98..0000000000
--- a/src/gallium/auxiliary/pipebuffer/Makefile
+++ /dev/null
@@ -1,19 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = pipebuffer
-
-C_SOURCES = \
-	pb_buffer_fenced.c \
-	pb_buffer_malloc.c \
-	pb_bufmgr_alt.c \
-	pb_bufmgr_cache.c \
-	pb_bufmgr_debug.c \
-	pb_bufmgr_fenced.c \
-	pb_bufmgr_mm.c \
-	pb_bufmgr_ondemand.c \
-	pb_bufmgr_pool.c \
-	pb_bufmgr_slab.c \
-	pb_validate.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript
deleted file mode 100644
index 8e9f06abe4..0000000000
--- a/src/gallium/auxiliary/pipebuffer/SConscript
+++ /dev/null
@@ -1,19 +0,0 @@
-Import('*')
-
-pipebuffer = env.ConvenienceLibrary(
-	target = 'pipebuffer',
-	source = [
-		'pb_buffer_fenced.c',
-		'pb_buffer_malloc.c',
-		'pb_bufmgr_alt.c',
-		'pb_bufmgr_cache.c',
-		'pb_bufmgr_debug.c',
-		'pb_bufmgr_fenced.c',
-		'pb_bufmgr_mm.c',
-		'pb_bufmgr_ondemand.c',
-		'pb_bufmgr_pool.c',
-		'pb_bufmgr_slab.c',
-		'pb_validate.c',
-	])
-
-auxiliaries.insert(0, pipebuffer)
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
index 2590546cb4..eb7e84be84 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
@@ -46,7 +46,7 @@
 
 #include "pipe/p_compiler.h"
 #include "util/u_debug.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 #include "pipe/p_state.h"
 
 
@@ -237,8 +237,9 @@ pb_reference(struct pb_buffer **dst,
 {
    struct pb_buffer *old = *dst;
 
-   if (pipe_reference((struct pipe_reference**)dst, &src->base.reference))
+   if (pipe_reference(&(*dst)->base.reference, &src->base.reference))
       pb_destroy( old );
+   *dst = src;
 }
 
 
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
index d31476228c..ba6f7b15f9 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007-2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -18,7 +18,7 @@
  * 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
+ * IN NO EVENT SHALL VMWARE 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.
@@ -42,7 +42,7 @@
 #endif
 
 #include "pipe/p_compiler.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 #include "util/u_debug.h"
 #include "pipe/p_thread.h"
 #include "util/u_memory.h"
@@ -80,11 +80,27 @@ struct fenced_buffer_list
  */
 struct fenced_buffer
 {
+   /*
+    * Immutable members.
+    */
+
    struct pb_buffer base;
-   
    struct pb_buffer *buffer;
+   struct fenced_buffer_list *list;
+
+   /**
+    * Protected by fenced_buffer_list::mutex
+    */
+   struct list_head head;
 
-   /* FIXME: protect access with mutex */
+   /**
+    * Following members are mutable and protected by this mutex.
+    * 
+    * You may lock this mutex alone, or lock it with fenced_buffer_list::mutex
+    * held, but in order to prevent deadlocks you must never lock 
+    * fenced_buffer_list::mutex with this mutex held.
+    */
+   pipe_mutex mutex;
 
    /**
     * A bitmask of PIPE_BUFFER_USAGE_CPU/GPU_READ/WRITE describing the current
@@ -96,9 +112,6 @@ struct fenced_buffer
    struct pb_validate *vl;
    unsigned validation_flags;
    struct pipe_fence_handle *fence;
-
-   struct list_head head;
-   struct fenced_buffer_list *list;
 };
 
 
@@ -110,15 +123,24 @@ fenced_buffer(struct pb_buffer *buf)
 }
 
 
+/**
+ * Add the buffer to the fenced list.
+ * 
+ * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this
+ * order, before calling this function.
+ * 
+ * Reference count should be incremented before calling this function.
+ */
 static INLINE void
-_fenced_buffer_add(struct fenced_buffer *fenced_buf)
+fenced_buffer_add_locked(struct fenced_buffer_list *fenced_list, 
+                         struct fenced_buffer *fenced_buf)
 {
-   struct fenced_buffer_list *fenced_list = fenced_buf->list;
-
    assert(pipe_is_referenced(&fenced_buf->base.base.reference));
    assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
    assert(fenced_buf->fence);
 
+   /* TODO: Move the reference count increment here */
+   
 #ifdef DEBUG
    LIST_DEL(&fenced_buf->head);
    assert(fenced_list->numUnfenced);
@@ -130,32 +152,16 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf)
 
 
 /**
- * Actually destroy the buffer.
+ * Remove the buffer from the fenced list.
+ * 
+ * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this 
+ * order before calling this function.
+ * 
+ * Reference count should be decremented after calling this function.
  */
 static INLINE void
-_fenced_buffer_destroy(struct fenced_buffer *fenced_buf)
-{
-   struct fenced_buffer_list *fenced_list = fenced_buf->list;
-   
-   assert(!pipe_is_referenced(&fenced_buf->base.base.reference));
-   assert(!fenced_buf->fence);
-#ifdef DEBUG
-   assert(fenced_buf->head.prev);
-   assert(fenced_buf->head.next);
-   LIST_DEL(&fenced_buf->head);
-   assert(fenced_list->numUnfenced);
-   --fenced_list->numUnfenced;
-#else
-   (void)fenced_list;
-#endif
-   pb_reference(&fenced_buf->buffer, NULL);
-   FREE(fenced_buf);
-}
-
-
-static INLINE void
-_fenced_buffer_remove(struct fenced_buffer_list *fenced_list,
-                      struct fenced_buffer *fenced_buf)
+fenced_buffer_remove_locked(struct fenced_buffer_list *fenced_list,
+                            struct fenced_buffer *fenced_buf)
 {
    struct pb_fence_ops *ops = fenced_list->ops;
 
@@ -177,37 +183,56 @@ _fenced_buffer_remove(struct fenced_buffer_list *fenced_list,
    ++fenced_list->numUnfenced;
 #endif
    
-   /**
-    * FIXME!!!
-    */
-
-   if(!pipe_is_referenced(&fenced_buf->base.base.reference))
-      _fenced_buffer_destroy(fenced_buf);
+   /* TODO: Move the reference count decrement and destruction here */
 }
 
 
+/**
+ * Wait for the fence to expire, and remove it from the fenced list.
+ * 
+ * fenced_buffer::mutex must be held. fenced_buffer_list::mutex must not be 
+ * held -- it will be acquired internally.
+ */
 static INLINE enum pipe_error
-_fenced_buffer_finish(struct fenced_buffer *fenced_buf)
+fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list,
+                              struct fenced_buffer *fenced_buf)
 {
-   struct fenced_buffer_list *fenced_list = fenced_buf->list;
    struct pb_fence_ops *ops = fenced_list->ops;
+   enum pipe_error ret = PIPE_ERROR;
 
 #if 0
    debug_warning("waiting for GPU");
 #endif
 
+   assert(pipe_is_referenced(&fenced_buf->base.base.reference));
    assert(fenced_buf->fence);
+
+   /*
+    * Acquire the global lock. Must release buffer mutex first to preserve
+    * lock order.
+    */
+   pipe_mutex_unlock(fenced_buf->mutex);
+   pipe_mutex_lock(fenced_list->mutex);
+   pipe_mutex_lock(fenced_buf->mutex);
+
    if(fenced_buf->fence) {
-      if(ops->fence_finish(ops, fenced_buf->fence, 0) != 0) {
-	 return PIPE_ERROR;
+      if(ops->fence_finish(ops, fenced_buf->fence, 0) == 0) {
+         /* Remove from the fenced list */
+         /* TODO: remove consequents */
+         fenced_buffer_remove_locked(fenced_list, fenced_buf);
+
+         p_atomic_dec(&fenced_buf->base.base.reference.count);
+         assert(pipe_is_referenced(&fenced_buf->base.base.reference));
+
+         fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+
+         ret = PIPE_OK;
       }
-      /* Remove from the fenced list */
-      /* TODO: remove consequents */
-      _fenced_buffer_remove(fenced_list, fenced_buf);
    }
 
-   fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
-   return PIPE_OK;
+   pipe_mutex_unlock(fenced_list->mutex);
+
+   return ret;
 }
 
 
@@ -215,12 +240,13 @@ _fenced_buffer_finish(struct fenced_buffer *fenced_buf)
  * Free as many fenced buffers from the list head as possible. 
  */
 static void
-_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, 
-                               int wait)
+fenced_buffer_list_check_free_locked(struct fenced_buffer_list *fenced_list, 
+                                     int wait)
 {
    struct pb_fence_ops *ops = fenced_list->ops;
    struct list_head *curr, *next;
    struct fenced_buffer *fenced_buf;
+   struct pb_buffer *pb_buf;
    struct pipe_fence_handle *prev_fence = NULL;
 
    curr = fenced_list->delayed.next;
@@ -228,21 +254,29 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list,
    while(curr != &fenced_list->delayed) {
       fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
 
+      pipe_mutex_lock(fenced_buf->mutex);
+
       if(fenced_buf->fence != prev_fence) {
 	 int signaled;
 	 if (wait)
 	    signaled = ops->fence_finish(ops, fenced_buf->fence, 0);
 	 else
 	    signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
-	 if (signaled != 0)
+	 if (signaled != 0) {
+            pipe_mutex_unlock(fenced_buf->mutex);
 	    break;
+         }
 	 prev_fence = fenced_buf->fence;
       }
       else {
 	 assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
       }
 
-      _fenced_buffer_remove(fenced_list, fenced_buf);
+      fenced_buffer_remove_locked(fenced_list, fenced_buf);
+      pipe_mutex_unlock(fenced_buf->mutex);
+
+      pb_buf = &fenced_buf->base;
+      pb_reference(&pb_buf, NULL);
 
       curr = next; 
       next = curr->next;
@@ -256,30 +290,25 @@ fenced_buffer_destroy(struct pb_buffer *buf)
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);   
    struct fenced_buffer_list *fenced_list = fenced_buf->list;
 
-   pipe_mutex_lock(fenced_list->mutex);
    assert(!pipe_is_referenced(&fenced_buf->base.base.reference));
-   if (fenced_buf->fence) {
-      struct pb_fence_ops *ops = fenced_list->ops;
-      if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) {
-	 struct list_head *curr, *prev;
-	 curr = &fenced_buf->head;
-	 prev = curr->prev;
-	 do {
-	    fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
-	    assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
-	    _fenced_buffer_remove(fenced_list, fenced_buf);
-	    curr = prev;
-	    prev = curr->prev;
-	 } while (curr != &fenced_list->delayed);
-      }	  
-      else {
-	 /* delay destruction */
-      }
-   }
-   else {
-      _fenced_buffer_destroy(fenced_buf);
-   }
+   assert(!fenced_buf->fence);
+
+#ifdef DEBUG
+   pipe_mutex_lock(fenced_list->mutex);
+   assert(fenced_buf->head.prev);
+   assert(fenced_buf->head.next);
+   LIST_DEL(&fenced_buf->head);
+   assert(fenced_list->numUnfenced);
+   --fenced_list->numUnfenced;
    pipe_mutex_unlock(fenced_list->mutex);
+#else
+   (void)fenced_list;
+#endif
+
+   pb_reference(&fenced_buf->buffer, NULL);
+
+   pipe_mutex_destroy(fenced_buf->mutex);
+   FREE(fenced_buf);
 }
 
 
@@ -290,24 +319,23 @@ fenced_buffer_map(struct pb_buffer *buf,
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);
    struct fenced_buffer_list *fenced_list = fenced_buf->list;
    struct pb_fence_ops *ops = fenced_list->ops;
-   void *map;
+   void *map = NULL;
+
+   pipe_mutex_lock(fenced_buf->mutex);
 
    assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE));
    
    /* Serialize writes */
    if((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) ||
       ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) && (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) {
-      if(flags & PIPE_BUFFER_USAGE_DONTBLOCK) {
+      if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) &&
+          ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) {
          /* Don't wait for the GPU to finish writing */
-         if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0)
-            _fenced_buffer_remove(fenced_list, fenced_buf);
-         else
-            return NULL;
-      }
-      else {
-         /* Wait for the GPU to finish writing */
-         _fenced_buffer_finish(fenced_buf);
+         goto done;
       }
+
+      /* Wait for the GPU to finish writing */
+      fenced_buffer_finish_locked(fenced_list, fenced_buf);
    }
 
 #if 0
@@ -324,6 +352,9 @@ fenced_buffer_map(struct pb_buffer *buf,
       fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE;
    }
 
+done:
+   pipe_mutex_unlock(fenced_buf->mutex);
+   
    return map;
 }
 
@@ -332,6 +363,9 @@ static void
 fenced_buffer_unmap(struct pb_buffer *buf)
 {
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+   
+   pipe_mutex_lock(fenced_buf->mutex);
+   
    assert(fenced_buf->mapcount);
    if(fenced_buf->mapcount) {
       pb_unmap(fenced_buf->buffer);
@@ -339,6 +373,8 @@ fenced_buffer_unmap(struct pb_buffer *buf)
       if(!fenced_buf->mapcount)
 	 fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE;
    }
+   
+   pipe_mutex_unlock(fenced_buf->mutex);
 }
 
 
@@ -350,11 +386,14 @@ fenced_buffer_validate(struct pb_buffer *buf,
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);
    enum pipe_error ret;
    
+   pipe_mutex_lock(fenced_buf->mutex);
+
    if(!vl) {
       /* invalidate */
       fenced_buf->vl = NULL;
       fenced_buf->validation_flags = 0;
-      return PIPE_OK;
+      ret = PIPE_OK;
+      goto done;
    }
    
    assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
@@ -362,14 +401,17 @@ fenced_buffer_validate(struct pb_buffer *buf,
    flags &= PIPE_BUFFER_USAGE_GPU_READ_WRITE;
 
    /* Buffer cannot be validated in two different lists */ 
-   if(fenced_buf->vl && fenced_buf->vl != vl)
-      return PIPE_ERROR_RETRY;
+   if(fenced_buf->vl && fenced_buf->vl != vl) {
+      ret = PIPE_ERROR_RETRY;
+      goto done;
+   }
    
 #if 0
    /* Do not validate if buffer is still mapped */
    if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
       /* TODO: wait for the thread that mapped the buffer to unmap it */
-      return PIPE_ERROR_RETRY;
+      ret = PIPE_ERROR_RETRY;
+      goto done;
    }
    /* Final sanity checking */
    assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE));
@@ -379,17 +421,21 @@ fenced_buffer_validate(struct pb_buffer *buf,
    if(fenced_buf->vl == vl &&
       (fenced_buf->validation_flags & flags) == flags) {
       /* Nothing to do -- buffer already validated */
-      return PIPE_OK;
+      ret = PIPE_OK;
+      goto done;
    }
    
    ret = pb_validate(fenced_buf->buffer, vl, flags);
    if (ret != PIPE_OK)
-      return ret;
+      goto done;
    
    fenced_buf->vl = vl;
    fenced_buf->validation_flags |= flags;
    
-   return PIPE_OK;
+done:
+   pipe_mutex_unlock(fenced_buf->mutex);
+
+   return ret;
 }
 
 
@@ -404,29 +450,36 @@ fenced_buffer_fence(struct pb_buffer *buf,
    fenced_buf = fenced_buffer(buf);
    fenced_list = fenced_buf->list;
    ops = fenced_list->ops;
-   
-   if(fence == fenced_buf->fence) {
-      /* Nothing to do */
-      return;
-   }
 
-   assert(fenced_buf->vl);
-   assert(fenced_buf->validation_flags);
-   
    pipe_mutex_lock(fenced_list->mutex);
-   if (fenced_buf->fence)
-      _fenced_buffer_remove(fenced_list, fenced_buf);
-   if (fence) {
-      ops->fence_reference(ops, &fenced_buf->fence, fence);
-      fenced_buf->flags |= fenced_buf->validation_flags;
-      _fenced_buffer_add(fenced_buf);
-   }
-   pipe_mutex_unlock(fenced_list->mutex);
+   pipe_mutex_lock(fenced_buf->mutex);
+
+   assert(pipe_is_referenced(&fenced_buf->base.base.reference));
+
+   if(fence != fenced_buf->fence) {
+      assert(fenced_buf->vl);
+      assert(fenced_buf->validation_flags);
+      
+      if (fenced_buf->fence) {
+         fenced_buffer_remove_locked(fenced_list, fenced_buf);
+         p_atomic_dec(&fenced_buf->base.base.reference.count);
+         assert(pipe_is_referenced(&fenced_buf->base.base.reference));
+      }
+      if (fence) {
+         ops->fence_reference(ops, &fenced_buf->fence, fence);
+         fenced_buf->flags |= fenced_buf->validation_flags;
+         p_atomic_inc(&fenced_buf->base.base.reference.count);
+         fenced_buffer_add_locked(fenced_list, fenced_buf);
+      }
+
+      pb_fence(fenced_buf->buffer, fence);
    
-   pb_fence(fenced_buf->buffer, fence);
+      fenced_buf->vl = NULL;
+      fenced_buf->validation_flags = 0;
+   }
 
-   fenced_buf->vl = NULL;
-   fenced_buf->validation_flags = 0;
+   pipe_mutex_unlock(fenced_buf->mutex);
+   pipe_mutex_unlock(fenced_list->mutex);
 }
 
 
@@ -436,6 +489,7 @@ fenced_buffer_get_base_buffer(struct pb_buffer *buf,
                               pb_size *offset)
 {
    struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+   /* NOTE: accesses immutable members only -- mutex not necessary */
    pb_get_base_buffer(fenced_buf->buffer, base_buf, offset);
 }
 
@@ -475,6 +529,8 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list,
    buf->buffer = buffer;
    buf->list = fenced_list;
    
+   pipe_mutex_init(buf->mutex);
+
 #ifdef DEBUG
    pipe_mutex_lock(fenced_list->mutex);
    LIST_ADDTAIL(&buf->head, &fenced_list->unfenced);
@@ -516,7 +572,7 @@ fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list,
                               int wait)
 {
    pipe_mutex_lock(fenced_list->mutex);
-   _fenced_buffer_list_check_free(fenced_list, wait);
+   fenced_buffer_list_check_free_locked(fenced_list, wait);
    pipe_mutex_unlock(fenced_list->mutex);
 }
 
@@ -538,11 +594,13 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list)
    next = curr->next;
    while(curr != &fenced_list->unfenced) {
       fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+      pipe_mutex_lock(fenced_buf->mutex);
       assert(!fenced_buf->fence);
       debug_printf("%10p %7u %7u\n",
-                   fenced_buf,
+                   (void *) fenced_buf,
                    fenced_buf->base.base.size,
                    p_atomic_read(&fenced_buf->base.base.reference.count));
+      pipe_mutex_unlock(fenced_buf->mutex);
       curr = next; 
       next = curr->next;
    }
@@ -552,13 +610,15 @@ fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list)
    while(curr != &fenced_list->delayed) {
       int signaled;
       fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+      pipe_mutex_lock(fenced_buf->mutex);
       signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
       debug_printf("%10p %7u %7u %10p %s\n",
-                   fenced_buf,
+                   (void *) fenced_buf,
                    fenced_buf->base.base.size,
                    p_atomic_read(&fenced_buf->base.base.reference.count),
-                   fenced_buf->fence,
+                   (void *) fenced_buf->fence,
                    signaled == 0 ? "y" : "n");
+      pipe_mutex_unlock(fenced_buf->mutex);
       curr = next; 
       next = curr->next;
    }
@@ -579,8 +639,8 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list)
 #if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
       sched_yield();
 #endif
-      _fenced_buffer_list_check_free(fenced_list, 1);
       pipe_mutex_lock(fenced_list->mutex);
+      fenced_buffer_list_check_free_locked(fenced_list, 1);
    }
 
 #ifdef DEBUG
@@ -588,6 +648,7 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list)
 #endif
       
    pipe_mutex_unlock(fenced_list->mutex);
+   pipe_mutex_destroy(fenced_list->mutex);
    
    fenced_list->ops->destroy(fenced_list->ops);
    
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
index 39ab8e722c..8c8d713078 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
@@ -51,7 +51,7 @@
 
 
 #include "pipe/p_compiler.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 
 
 #ifdef __cplusplus
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c
index 57d1ede45a..7b34c8e357 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c
@@ -294,7 +294,7 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr,
       LIST_DEL(&buf->head);
       pipe_mutex_unlock(mgr->mutex);
       /* Increase refcount */
-      pb_reference((struct pb_buffer**)&buf, &buf->base);
+      pipe_reference(NULL, &buf->base.base.reference);
       return &buf->base;
    }
    
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
index e7352e90db..d21910d0bf 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
@@ -37,7 +37,6 @@
  */
 
 #include "pipe/p_compiler.h"
-#include "pipe/p_error.h"
 #include "util/u_debug.h"
 #include "pipe/p_thread.h"
 #include "pipe/p_defines.h"
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c
index 150fd50618..ce40c0cf0e 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_validate.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c
@@ -34,7 +34,7 @@
 
 
 #include "pipe/p_compiler.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 #include "util/u_memory.h"
 #include "util/u_debug.h"
 
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.h b/src/gallium/auxiliary/pipebuffer/pb_validate.h
index dfb84df1ce..3c93f30f20 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_validate.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.h
@@ -37,7 +37,7 @@
 
 
 #include "pipe/p_compiler.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 
 #ifdef __cplusplus
 extern "C" {
diff --git a/src/gallium/auxiliary/rbug/Makefile b/src/gallium/auxiliary/rbug/Makefile
deleted file mode 100644
index cd12e8468f..0000000000
--- a/src/gallium/auxiliary/rbug/Makefile
+++ /dev/null
@@ -1,14 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = rbug
-
-C_SOURCES = \
-	rbug_connection.c \
-	rbug_core.c \
-	rbug_texture.c \
-	rbug_context.c \
-	rbug_shader.c \
-	rbug_demarshal.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/rbug/README b/src/gallium/auxiliary/rbug/README
index 33d76371de..d984067893 100644
--- a/src/gallium/auxiliary/rbug/README
+++ b/src/gallium/auxiliary/rbug/README
@@ -16,6 +16,10 @@ for information about applications look in:
 
 progs/rbug/README
 
+for a GUI see:
+
+  http://cgit.freedesktop.org/mesa/rbug-gui
+
 
 --
 Jakob Bornecrantz <jakob@vmware.com>
diff --git a/src/gallium/auxiliary/rbug/SConscript b/src/gallium/auxiliary/rbug/SConscript
deleted file mode 100644
index 4a9afb45d3..0000000000
--- a/src/gallium/auxiliary/rbug/SConscript
+++ /dev/null
@@ -1,14 +0,0 @@
-Import('*')
-
-rbug = env.ConvenienceLibrary(
-	target = 'rbug',
-	source = [
-		'rbug_core.c',
-		'rbug_shader.c',
-		'rbug_context.c',
-		'rbug_texture.c',
-		'rbug_demarshal.c',
-		'rbug_connection.c',
-	])
-
-auxiliaries.insert(0, rbug)
diff --git a/src/gallium/auxiliary/rbug/rbug_connection.c b/src/gallium/auxiliary/rbug/rbug_connection.c
index 52acb700af..ae4e27f9f6 100644
--- a/src/gallium/auxiliary/rbug/rbug_connection.c
+++ b/src/gallium/auxiliary/rbug/rbug_connection.c
@@ -87,6 +87,7 @@ rbug_get_message(struct rbug_connection *c, uint32_t *serial)
    if (!data) {
       return NULL;
    }
+   data->opcode = 0;
 
    do {
       uint8_t *ptr = ((uint8_t*)data) + read;
diff --git a/src/gallium/auxiliary/rbug/rbug_context.h b/src/gallium/auxiliary/rbug/rbug_context.h
index da61c2365b..03126d6b12 100644
--- a/src/gallium/auxiliary/rbug/rbug_context.h
+++ b/src/gallium/auxiliary/rbug/rbug_context.h
@@ -46,7 +46,7 @@ typedef enum
 	RBUG_BLOCK_BEFORE = 1,
 	RBUG_BLOCK_AFTER = 2,
 	RBUG_BLOCK_RULE = 4,
-	RBUG_BLOCK_MASK = 7,
+	RBUG_BLOCK_MASK = 7
 } rbug_block_t;
 
 struct rbug_proto_context_list
diff --git a/src/gallium/auxiliary/rbug/rbug_proto.h b/src/gallium/auxiliary/rbug/rbug_proto.h
index d273be0166..4f3eb75dc4 100644
--- a/src/gallium/auxiliary/rbug/rbug_proto.h
+++ b/src/gallium/auxiliary/rbug/rbug_proto.h
@@ -65,7 +65,7 @@ enum rbug_opcode
 	RBUG_OP_SHADER_DISABLE = 770,
 	RBUG_OP_SHADER_REPLACE = 771,
 	RBUG_OP_SHADER_LIST_REPLY = -768,
-	RBUG_OP_SHADER_INFO_REPLY = -769,
+	RBUG_OP_SHADER_INFO_REPLY = -769
 };
 
 /**
diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile
deleted file mode 100644
index ab8ea464c6..0000000000
--- a/src/gallium/auxiliary/rtasm/Makefile
+++ /dev/null
@@ -1,13 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = rtasm
-
-C_SOURCES = \
-	rtasm_cpu.c \
-	rtasm_execmem.c \
-	rtasm_x86sse.c \
-	rtasm_ppc.c \
-	rtasm_ppc_spe.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript
deleted file mode 100644
index eb48368acc..0000000000
--- a/src/gallium/auxiliary/rtasm/SConscript
+++ /dev/null
@@ -1,13 +0,0 @@
-Import('*')
-
-rtasm = env.ConvenienceLibrary(
-	target = 'rtasm',
-	source = [
-		'rtasm_cpu.c',
-		'rtasm_execmem.c',
-		'rtasm_x86sse.c',
-		'rtasm_ppc.c',
-		'rtasm_ppc_spe.c',
-	])
-
-auxiliaries.insert(0, rtasm)
diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c
index 01811d5011..ffed768f97 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_execmem.c
+++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c
@@ -41,6 +41,12 @@
 #define MAP_ANONYMOUS MAP_ANON
 #endif
 
+#if defined(PIPE_OS_WINDOWS)
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN 1
+#endif
+#include <windows.h>
+#endif
 
 #if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
 
@@ -118,7 +124,29 @@ rtasm_exec_free(void *addr)
 }
 
 
-#else /* PIPE_OS_LINUX || PIPE_OS_BSD || PIPE_OS_SOLARIS */
+#elif defined(PIPE_OS_WINDOWS)
+
+
+/*
+ * Avoid Data Execution Prevention.
+ */
+
+void *
+rtasm_exec_malloc(size_t size)
+{
+   return VirtualAlloc(NULL, size, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
+}
+
+
+void
+rtasm_exec_free(void *addr)
+{
+   VirtualFree(addr, 0, MEM_RELEASE);
+}
+
+
+#else
+
 
 /*
  * Just use regular memory.
@@ -138,4 +166,4 @@ rtasm_exec_free(void *addr)
 }
 
 
-#endif /* PIPE_OS_LINUX || PIPE_OS_BSD || PIPE_OS_SOLARIS */
+#endif
diff --git a/src/gallium/auxiliary/sct/Makefile b/src/gallium/auxiliary/sct/Makefile
deleted file mode 100644
index a7d111b689..0000000000
--- a/src/gallium/auxiliary/sct/Makefile
+++ /dev/null
@@ -1,9 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = sct
-
-C_SOURCES = \
-	sct.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/sct/SConscript b/src/gallium/auxiliary/sct/SConscript
deleted file mode 100644
index 76927d973f..0000000000
--- a/src/gallium/auxiliary/sct/SConscript
+++ /dev/null
@@ -1,9 +0,0 @@
-Import('*')
-
-sct = env.ConvenienceLibrary(
-	target = 'sct',
-	source = [
-		'sct.c'
-	])
-
-auxiliaries.insert(0, sct)
diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c
deleted file mode 100644
index 722d2b7e66..0000000000
--- a/src/gallium/auxiliary/sct/sct.c
+++ /dev/null
@@ -1,453 +0,0 @@
-/**************************************************************************
- * 
- * 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 "util/u_memory.h"
-#include "pipe/p_state.h"
-#include "sct.h"
-
-
-struct texture_list
-{
-   struct pipe_texture *texture;
-   struct texture_list *next;
-};
-
-
-
-#define MAX_SURFACES  ((PIPE_MAX_COLOR_BUFS) + 1)
-
-struct sct_context
-{
-   const struct pipe_context *context;
-
-   /** surfaces the context is drawing into */
-   struct pipe_surface *surfaces[MAX_SURFACES];
-
-   /** currently bound textures */
-   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
-
-   /** previously bound textures, used but not flushed */
-   struct texture_list *textures_used;
-
-   boolean needs_flush;
-
-   struct sct_context *next;
-};
-
-
-
-struct sct_surface
-{
-   const struct pipe_surface *surface;
-
-   /** list of contexts drawing to this surface */
-   struct sct_context_list *contexts;
-
-   struct sct_surface *next;
-};
-
-
-
-/**
- * Find the surface_info for the given pipe_surface
- */
-static struct sct_surface *
-find_surface_info(struct surface_context_tracker *sct,
-                  const struct pipe_surface *surface)
-{
-   struct sct_surface *si;
-   for (si = sct->surfaces; si; si = si->next)
-      if (si->surface == surface)
-         return si;
-   return NULL;
-}
-
-
-/**
- * As above, but create new surface_info if surface is new.
- */
-static struct sct_surface *
-find_create_surface_info(struct surface_context_tracker *sct,
-                         const struct pipe_surface *surface)
-{
-   struct sct_surface *si = find_surface_info(sct, surface);
-   if (si)
-      return si;
-
-   /* alloc new */
-   si = CALLOC_STRUCT(sct_surface);
-   if (si) {
-      si->surface = surface;
-
-      /* insert at head */
-      si->next = sct->surfaces;
-      sct->surfaces = si;
-   }
-
-   return si;
-}
-
-
-/**
- * Find a context_info for the given context.
- */
-static struct sct_context *
-find_context_info(struct surface_context_tracker *sct,
-                  const struct pipe_context *context)
-{
-   struct sct_context *ci;
-   for (ci = sct->contexts; ci; ci = ci->next)
-      if (ci->context == context)
-         return ci;
-   return NULL;
-}
-
-
-/**
- * As above, but create new context_info if context is new.
- */
-static struct sct_context *
-find_create_context_info(struct surface_context_tracker *sct,
-                         const struct pipe_context *context)
-{
-   struct sct_context *ci = find_context_info(sct, context);
-   if (ci)
-      return ci;
-
-   /* alloc new */
-   ci = CALLOC_STRUCT(sct_context);
-   if (ci) {
-      ci->context = context;
-
-      /* insert at head */
-      ci->next = sct->contexts;
-      sct->contexts = ci;
-   }
-
-   return ci;
-}
-
-
-/**
- * Is the context already bound to the surface?
- */
-static boolean
-find_surface_context(const struct sct_surface *si,
-                     const struct pipe_context *context)
-{
-   const struct sct_context_list *cl;
-   for (cl = si->contexts; cl; cl = cl->next) {
-      if (cl->context == context) {
-         return TRUE;
-      }
-   }
-   return FALSE;
-}
-
-
-/**
- * Add a context to the list of contexts associated with a surface.
- */
-static void
-add_context_to_surface(struct sct_surface *si,
-                       const struct pipe_context *context)
-{
-   struct sct_context_list *cl = CALLOC_STRUCT(sct_context_list);
-   if (cl) {
-      cl->context = context;
-      /* insert at head of list of contexts */
-      cl->next = si->contexts;
-      si->contexts = cl;
-   }
-}
-
-
-/**
- * Remove a context from the list of contexts associated with a surface.
- */
-static void
-remove_context_from_surface(struct sct_surface *si,
-                            const struct pipe_context *context)
-{
-   struct sct_context_list *prev = NULL, *curr, *next;
-
-   for (curr = si->contexts; curr; curr = next) {
-      if (curr->context == context) {
-         /* remove */
-         if (prev)
-            prev->next = curr->next;
-         else
-            si->contexts = curr->next;
-         next = curr->next;
-         FREE(curr);
-      }
-      else {
-         prev = curr;
-         next = curr->next;
-      }
-   }
-}
-
-
-/**
- * Unbind context from surface.
- */
-static void
-unbind_context_surface(struct surface_context_tracker *sct,
-                       struct pipe_context *context,
-                       struct pipe_surface *surface)
-{
-   struct sct_surface *si = find_surface_info(sct, surface);
-   if (si) {
-      remove_context_from_surface(si, context);
-   }
-}
-
-
-/**
- * Bind context to a set of surfaces (color + Z).
- * Like MakeCurrent().
- */
-void
-sct_bind_surfaces(struct surface_context_tracker *sct,
-                  struct pipe_context *context,
-                  uint num_surf,
-                  struct pipe_surface **surfaces)
-{
-   struct sct_context *ci = find_create_context_info(sct, context);
-   uint i;
-
-   if (!ci) {
-      return; /* out of memory */
-   }
-
-   /* unbind currently bound surfaces */
-   for (i = 0; i < MAX_SURFACES; i++) {
-      if (ci->surfaces[i]) {
-         unbind_context_surface(sct, context, ci->surfaces[i]);
-      }
-   }
-
-   /* bind new surfaces */
-   for (i = 0; i < num_surf; i++) {
-      struct sct_surface *si = find_create_surface_info(sct, surfaces[i]);
-      if (!find_surface_context(si, context)) {
-         add_context_to_surface(si, context);
-      }
-   }
-}
-
-
-/**
- * Return list of contexts bound to a surface.
- */
-const struct sct_context_list *
-sct_get_surface_contexts(struct surface_context_tracker *sct,
-                         const struct pipe_surface *surface)
-{
-   const struct sct_surface *si = find_surface_info(sct, surface);
-   return si->contexts;
-}
-
-
-
-static boolean
-find_texture(const struct sct_context *ci,
-             const struct pipe_texture *texture)
-{
-   const struct texture_list *tl;
-
-   for (tl = ci->textures_used; tl; tl = tl->next) {
-      if (tl->texture == texture) {
-         return TRUE;
-      }
-   }
-   return FALSE;
-}
-
-
-/**
- * Add the given texture to the context's list of used textures.
- */
-static void
-add_texture_used(struct sct_context *ci,
-                 struct pipe_texture *texture)
-{
-   if (!find_texture(ci, texture)) {
-      /* add to list */
-      struct texture_list *tl = CALLOC_STRUCT(texture_list);
-      if (tl) {
-         pipe_texture_reference(&tl->texture, texture);
-         /* insert at head */
-         tl->next = ci->textures_used;
-         ci->textures_used = tl;
-      }
-   }
-}
-
-
-/**
- * Bind a texture to a rendering context.
- */
-void
-sct_bind_texture(struct surface_context_tracker *sct,
-                 struct pipe_context *context,
-                 uint unit,
-                 struct pipe_texture *tex)
-{
-   struct sct_context *ci = find_context_info(sct, context);
-
-   if (ci->textures[unit] != tex) {
-      /* put texture on the 'used' list */
-      add_texture_used(ci, tex);
-      /* bind new */
-      pipe_texture_reference(&ci->textures[unit], tex);
-   }
-}
-
-
-/**
- * Check if the given texture has been used by the rendering context
- * since the last call to sct_flush_textures().
- */
-boolean
-sct_is_texture_used(struct surface_context_tracker *sct,
-                    const struct pipe_context *context,
-                    const struct pipe_texture *texture)
-{
-   const struct sct_context *ci = find_context_info(sct, context);
-   return find_texture(ci, texture);
-}
-
-
-/**
- * To be called when the image contents of a texture are changed, such
- * as for gl[Copy]TexSubImage().
- * XXX this may not be needed
- */
-void
-sct_update_texture(struct pipe_texture *tex)
-{
-
-}
-
-
-/**
- * When a scene is flushed/rendered we can release the list of
- * used textures.
- */
-void
-sct_flush_textures(struct surface_context_tracker *sct,
-                   struct pipe_context *context)
-{
-   struct sct_context *ci = find_context_info(sct, context);
-   struct texture_list *tl, *next;
-   uint i;
-
-   for (tl = ci->textures_used; tl; tl = next) {
-      next = tl->next;
-      pipe_texture_reference(&tl->texture, NULL);
-      FREE(tl);
-   }
-   ci->textures_used = NULL;
-
-   /* put the currently bound textures on the 'used' list */
-   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
-      add_texture_used(ci, ci->textures[i]);
-   }
-}
-
-
-
-void
-sct_destroy_context(struct surface_context_tracker *sct,
-                    struct pipe_context *context)
-{
-   /* XXX should we require an unbinding first? */
-   {
-      struct sct_surface *si;
-      for (si = sct->surfaces; si; si = si->next) {
-         remove_context_from_surface(si, context);
-      }
-   }
-
-   /* remove context from context_info list */
-   {
-      struct sct_context *ci, *next, *prev = NULL;
-      for (ci = sct->contexts; ci; ci = next) {
-         next = ci->next;
-         if (ci->context == context) {
-            if (prev)
-               prev->next = ci->next;
-            else
-               sct->contexts = ci->next;
-            FREE(ci);
-         }
-         else {
-            prev = ci;
-         }
-      }
-   }
-
-}
-
-
-void
-sct_destroy_surface(struct surface_context_tracker *sct,
-                    struct pipe_surface *surface)
-{
-   if (1) {
-      /* debug/sanity: no context should be bound to surface */
-      struct sct_context *ci;
-      uint i;
-      for (ci = sct->contexts; ci; ci = ci->next) {
-         for (i = 0; i < MAX_SURFACES; i++) {
-            assert(ci->surfaces[i] != surface);
-         }
-      }
-   }
-
-   /* remove surface from sct_surface list */
-   {
-      struct sct_surface *si, *next, *prev = NULL;
-      for (si = sct->surfaces; si; si = next) {
-         next = si->next;
-         if (si->surface == surface) {
-            /* unlink */
-            if (prev)
-               prev->next = si->next;
-            else
-               sct->surfaces = si->next;
-            FREE(si);
-         }
-         else {
-            prev = si;
-         }
-      }
-   }
-}
diff --git a/src/gallium/auxiliary/sct/sct.h b/src/gallium/auxiliary/sct/sct.h
deleted file mode 100644
index cf7c4d3bdf..0000000000
--- a/src/gallium/auxiliary/sct/sct.h
+++ /dev/null
@@ -1,123 +0,0 @@
-/**************************************************************************
- * 
- * 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.
- * 
- **************************************************************************/
-
-/**
- * Surface/Context Tracking
- *
- * For some drivers, we need to monitor the binding between contexts and
- * surfaces/textures.
- * This code may evolve quite a bit...
- */
-
-
-#ifndef SCT_H
-#define SCT_H
-
-
-#ifdef	__cplusplus
-extern "C" {
-#endif
-
-
-struct pipe_context;
-struct pipe_surface;
-
-struct sct_context;
-struct sct_surface;
-
-
-/**
- * Per-device info, basically
- */
-struct surface_context_tracker
-{
-   struct sct_context *contexts;
-   struct sct_surface *surfaces;
-};
-
-
-
-/**
- * Simple linked list of contexts
- */
-struct sct_context_list
-{
-   const struct pipe_context *context;
-   struct sct_context_list *next;
-};
-
-
-
-extern void
-sct_bind_surfaces(struct surface_context_tracker *sct,
-                  struct pipe_context *context,
-                  uint num_surf,
-                  struct pipe_surface **surfaces);
-
-
-extern void
-sct_bind_texture(struct surface_context_tracker *sct,
-                 struct pipe_context *context,
-                 uint unit,
-                 struct pipe_texture *texture);
-
-
-extern void
-sct_update_texture(struct pipe_texture *tex);
-
-
-extern boolean
-sct_is_texture_used(struct surface_context_tracker *sct,
-                    const struct pipe_context *context,
-                    const struct pipe_texture *texture);
-
-extern void
-sct_flush_textures(struct surface_context_tracker *sct,
-                   struct pipe_context *context);
-
-
-extern const struct sct_context_list *
-sct_get_surface_contexts(struct surface_context_tracker *sct,
-                         const struct pipe_surface *surf);
-
-
-extern void
-sct_destroy_context(struct surface_context_tracker *sct,
-                    struct pipe_context *context);
-
-
-extern void
-sct_destroy_surface(struct surface_context_tracker *sct,
-                    struct pipe_surface *surface);
-
-
-
-#ifdef	__cplusplus
-}
-#endif
-
-#endif /* SCT_H */
diff --git a/src/gallium/auxiliary/sct/usage.c b/src/gallium/auxiliary/sct/usage.c
deleted file mode 100644
index 6227f19962..0000000000
--- a/src/gallium/auxiliary/sct/usage.c
+++ /dev/null
@@ -1,61 +0,0 @@
-/* surface / context tracking */
-
-
-/*
-
-context A:
-  render to texture T
-
-context B:
-  texture from T
-
------------------------
-
-flush surface:
-  which contexts are bound to the surface?
-
------------------------
-
-glTexSubImage():
-  which contexts need to be flushed?
-
- */
-
-
-/*
-
-in MakeCurrent():
-
-  call sct_bind_surfaces(context, list of surfaces) to update the
-  dependencies between context and surfaces
-
-
-in SurfaceFlush(), or whatever it is in D3D:
-
-  call sct_get_surface_contexts(surface) to get a list of contexts
-  which are currently bound to the surface.
-
-
-
-in BindTexture():
-
-  call sct_bind_texture(context, texture) to indicate that the texture
-  is used in the scene.
-
-
-in glTexSubImage() or RenderToTexture():
-
-  call sct_is_texture_used(context, texture) to determine if the texture
-  has been used in the scene, but the scene's not flushed.  If TRUE is
-  returned it means the scene has to be rendered/flushed before the contents
-  of the texture can be changed.
-
-
-in psb_scene_flush/terminate():
-
-  call sct_flush_textures(context) to tell the SCT that the textures which
-  were used in the scene can be released.
-
-
-
-*/
diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile
deleted file mode 100644
index 5f0a580b09..0000000000
--- a/src/gallium/auxiliary/tgsi/Makefile
+++ /dev/null
@@ -1,22 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = tgsi
-
-C_SOURCES = \
-	tgsi_sanity.c \
-	tgsi_build.c \
-	tgsi_dump.c \
-	tgsi_exec.c \
-	tgsi_info.c \
-	tgsi_iterate.c \
-	tgsi_parse.c \
-	tgsi_ppc.c \
-	tgsi_scan.c \
-	tgsi_sse2.c \
-	tgsi_text.c \
-	tgsi_transform.c \
-	tgsi_ureg.c \
-	tgsi_util.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript
deleted file mode 100644
index b6bc2924f0..0000000000
--- a/src/gallium/auxiliary/tgsi/SConscript
+++ /dev/null
@@ -1,23 +0,0 @@
-Import('*')
-
-tgsi = env.ConvenienceLibrary(
-	target = 'tgsi',
-	source = [
-		'tgsi_build.c',
-		'tgsi_dump.c',
-		'tgsi_dump_c.c',
-		'tgsi_exec.c',
-		'tgsi_info.c',
-		'tgsi_iterate.c',
-		'tgsi_parse.c',
-		'tgsi_sanity.c',
-		'tgsi_scan.c',
-		'tgsi_ppc.c',
-		'tgsi_sse2.c',
-		'tgsi_text.c',
-		'tgsi_transform.c',
-		'tgsi_ureg.c',
-		'tgsi_util.c',
-	])
-
-auxiliaries.insert(0, tgsi)
diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt
index eb492076b7..080fd4c731 100644
--- a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt
+++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt
@@ -1129,3 +1129,35 @@ TGSI Instruction Specification
 
   target            Label of target instruction.
 
+
+3  Other tokens
+===============
+
+
+3.1  Declaration Semantic
+-------------------------
+
+
+  Follows Declaration token if Semantic bit is set.
+
+  Since its purpose is to link a shader with other stages of the pipeline,
+  it is valid to follow only those Declaration tokens that declare a register
+  either in INPUT or OUTPUT file.
+
+  SemanticName field contains the semantic name of the register being declared.
+  There is no default value.
+
+  SemanticIndex is an optional subscript that can be used to distinguish
+  different register declarations with the same semantic name. The default value
+  is 0.
+
+  The meanings of the individual semantic names are explained in the following
+  sections.
+
+
+3.1.1  FACE
+
+  Valid only in a fragment shader INPUT declaration.
+
+  FACE.x is negative when the primitive is back facing. FACE.x is positive
+  when the primitive is front facing.
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c
index e0cfc54420..de9cbc8630 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.c
@@ -30,21 +30,6 @@
 #include "tgsi_build.h"
 #include "tgsi_parse.h"
 
-/*
- * version
- */
-
-struct tgsi_version
-tgsi_build_version( void )
-{
-   struct tgsi_version  version;
-
-   version.MajorVersion = 1;
-   version.MinorVersion = 1;
-   version.Padding = 0;
-
-   return version;
-}
 
 /*
  * header
@@ -122,7 +107,6 @@ tgsi_default_declaration( void )
    declaration.Centroid = 0;
    declaration.Invariant = 0;
    declaration.Padding = 0;
-   declaration.Extended = 0;
 
    return declaration;
 }
@@ -173,7 +157,7 @@ tgsi_default_full_declaration( void )
    struct tgsi_full_declaration  full_declaration;
 
    full_declaration.Declaration  = tgsi_default_declaration();
-   full_declaration.DeclarationRange = tgsi_default_declaration_range();
+   full_declaration.Range = tgsi_default_declaration_range();
    full_declaration.Semantic = tgsi_default_declaration_semantic();
 
    return full_declaration;
@@ -210,8 +194,8 @@ tgsi_build_full_declaration(
    size++;
 
    *dr = tgsi_build_declaration_range(
-      full_decl->DeclarationRange.First,
-      full_decl->DeclarationRange.Last,
+      full_decl->Range.First,
+      full_decl->Range.Last,
       declaration,
       header );
 
@@ -224,8 +208,8 @@ tgsi_build_full_declaration(
       size++;
 
       *ds = tgsi_build_declaration_semantic(
-         full_decl->Semantic.SemanticName,
-         full_decl->Semantic.SemanticIndex,
+         full_decl->Semantic.Name,
+         full_decl->Semantic.Index,
          declaration,
          header );
    }
@@ -270,8 +254,8 @@ tgsi_default_declaration_semantic( void )
 {
    struct tgsi_declaration_semantic ds;
 
-   ds.SemanticName = TGSI_SEMANTIC_POSITION;
-   ds.SemanticIndex = 0;
+   ds.Name = TGSI_SEMANTIC_POSITION;
+   ds.Index = 0;
    ds.Padding = 0;
 
    return ds;
@@ -290,8 +274,8 @@ tgsi_build_declaration_semantic(
    assert( semantic_index <= 0xFFFF );
 
    ds = tgsi_default_declaration_semantic();
-   ds.SemanticName = semantic_name;
-   ds.SemanticIndex = semantic_index;
+   ds.Name = semantic_name;
+   ds.Index = semantic_index;
 
    declaration_grow( declaration, header );
 
@@ -311,7 +295,6 @@ tgsi_default_immediate( void )
    immediate.NrTokens = 1;
    immediate.DataType = TGSI_IMM_FLOAT32;
    immediate.Padding = 0;
-   immediate.Extended = 0;
 
    return immediate;
 }
@@ -416,24 +399,26 @@ tgsi_default_instruction( void )
    struct tgsi_instruction instruction;
 
    instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION;
-   instruction.NrTokens = 1;
+   instruction.NrTokens = 0;
    instruction.Opcode = TGSI_OPCODE_MOV;
    instruction.Saturate = TGSI_SAT_NONE;
+   instruction.Predicate = 0;
    instruction.NumDstRegs = 1;
    instruction.NumSrcRegs = 1;
+   instruction.Label = 0;
+   instruction.Texture = 0;
    instruction.Padding  = 0;
-   instruction.Extended = 0;
 
    return instruction;
 }
 
 struct tgsi_instruction
-tgsi_build_instruction(
-   unsigned opcode,
-   unsigned saturate,
-   unsigned num_dst_regs,
-   unsigned num_src_regs,
-   struct tgsi_header *header )
+tgsi_build_instruction(unsigned opcode,
+                       unsigned saturate,
+                       unsigned predicate,
+                       unsigned num_dst_regs,
+                       unsigned num_src_regs,
+                       struct tgsi_header *header)
 {
    struct tgsi_instruction instruction;
 
@@ -445,6 +430,7 @@ tgsi_build_instruction(
    instruction = tgsi_default_instruction();
    instruction.Opcode = opcode;
    instruction.Saturate = saturate;
+   instruction.Predicate = predicate;
    instruction.NumDstRegs = num_dst_regs;
    instruction.NumSrcRegs = num_src_regs;
 
@@ -472,18 +458,16 @@ tgsi_default_full_instruction( void )
    unsigned i;
 
    full_instruction.Instruction = tgsi_default_instruction();
-   full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv();
-   full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label();
-   full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture();
+   full_instruction.Predicate = tgsi_default_instruction_predicate();
+   full_instruction.Label = tgsi_default_instruction_label();
+   full_instruction.Texture = tgsi_default_instruction_texture();
    for( i = 0;  i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) {
-      full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register();
+      full_instruction.Dst[i] = tgsi_default_full_dst_register();
    }
    for( i = 0;  i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) {
-      full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register();
+      full_instruction.Src[i] = tgsi_default_full_src_register();
    }
 
-   full_instruction.Flags = 0x0;
-
    return full_instruction;
 }
 
@@ -504,82 +488,70 @@ tgsi_build_full_instruction(
    instruction = (struct tgsi_instruction *) &tokens[size];
    size++;
 
-   *instruction = tgsi_build_instruction(
-      full_inst->Instruction.Opcode,
-      full_inst->Instruction.Saturate,
-      full_inst->Instruction.NumDstRegs,
-      full_inst->Instruction.NumSrcRegs,
-      header );
+   *instruction = tgsi_build_instruction(full_inst->Instruction.Opcode,
+                                         full_inst->Instruction.Saturate,
+                                         full_inst->Instruction.Predicate,
+                                         full_inst->Instruction.NumDstRegs,
+                                         full_inst->Instruction.NumSrcRegs,
+                                         header);
    prev_token = (struct tgsi_token  *) instruction;
 
-   if( tgsi_compare_instruction_ext_nv(
-         full_inst->InstructionExtNv,
-         tgsi_default_instruction_ext_nv() ) ) {
-      struct tgsi_instruction_ext_nv *instruction_ext_nv;
+   if (full_inst->Instruction.Predicate) {
+      struct tgsi_instruction_predicate *instruction_predicate;
 
-      if( maxsize <= size )
+      if (maxsize <= size) {
          return 0;
-      instruction_ext_nv =
-         (struct  tgsi_instruction_ext_nv *) &tokens[size];
+      }
+      instruction_predicate = (struct tgsi_instruction_predicate *)&tokens[size];
       size++;
 
-      *instruction_ext_nv  = tgsi_build_instruction_ext_nv(
-         full_inst->InstructionExtNv.Precision,
-         full_inst->InstructionExtNv.CondDstIndex,
-         full_inst->InstructionExtNv.CondFlowIndex,
-         full_inst->InstructionExtNv.CondMask,
-         full_inst->InstructionExtNv.CondSwizzleX,
-         full_inst->InstructionExtNv.CondSwizzleY,
-         full_inst->InstructionExtNv.CondSwizzleZ,
-         full_inst->InstructionExtNv.CondSwizzleW,
-         full_inst->InstructionExtNv.CondDstUpdate,
-         full_inst->InstructionExtNv.CondFlowEnable,
-         prev_token,
-         instruction,
-         header );
-      prev_token = (struct tgsi_token  *) instruction_ext_nv;
+      *instruction_predicate =
+         tgsi_build_instruction_predicate(full_inst->Predicate.Index,
+                                          full_inst->Predicate.Negate,
+                                          full_inst->Predicate.SwizzleX,
+                                          full_inst->Predicate.SwizzleY,
+                                          full_inst->Predicate.SwizzleZ,
+                                          full_inst->Predicate.SwizzleW,
+                                          instruction,
+                                          header);
    }
 
-   if( tgsi_compare_instruction_ext_label(
-         full_inst->InstructionExtLabel,
-         tgsi_default_instruction_ext_label() ) ) {
-      struct tgsi_instruction_ext_label *instruction_ext_label;
+   if (full_inst->Instruction.Label) {
+      struct tgsi_instruction_label *instruction_label;
 
       if( maxsize <= size )
          return 0;
-      instruction_ext_label =
-         (struct  tgsi_instruction_ext_label *) &tokens[size];
+      instruction_label =
+         (struct  tgsi_instruction_label *) &tokens[size];
       size++;
 
-      *instruction_ext_label = tgsi_build_instruction_ext_label(
-         full_inst->InstructionExtLabel.Label,
+      *instruction_label = tgsi_build_instruction_label(
+         full_inst->Label.Label,
          prev_token,
          instruction,
          header );
-      prev_token = (struct tgsi_token  *) instruction_ext_label;
+      prev_token = (struct tgsi_token  *) instruction_label;
    }
 
-   if( tgsi_compare_instruction_ext_texture(
-         full_inst->InstructionExtTexture,
-         tgsi_default_instruction_ext_texture() ) ) {
-      struct tgsi_instruction_ext_texture *instruction_ext_texture;
+   if (full_inst->Instruction.Texture) {
+      struct tgsi_instruction_texture *instruction_texture;
 
       if( maxsize <= size )
          return 0;
-      instruction_ext_texture =
-         (struct  tgsi_instruction_ext_texture *) &tokens[size];
+      instruction_texture =
+         (struct  tgsi_instruction_texture *) &tokens[size];
       size++;
 
-      *instruction_ext_texture = tgsi_build_instruction_ext_texture(
-         full_inst->InstructionExtTexture.Texture,
+      *instruction_texture = tgsi_build_instruction_texture(
+         full_inst->Texture.Texture,
          prev_token,
          instruction,
          header   );
-      prev_token = (struct tgsi_token  *) instruction_ext_texture;
+      prev_token = (struct tgsi_token  *) instruction_texture;
    }
 
    for( i = 0;  i <   full_inst->Instruction.NumDstRegs; i++ ) {
-      const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i];
+      const struct tgsi_full_dst_register *reg = &full_inst->Dst[i];
       struct tgsi_dst_register *dst_register;
       struct tgsi_token *prev_token;
 
@@ -589,58 +561,15 @@ tgsi_build_full_instruction(
       size++;
 
       *dst_register = tgsi_build_dst_register(
-         reg->DstRegister.File,
-         reg->DstRegister.WriteMask,
-         reg->DstRegister.Indirect,
-         reg->DstRegister.Index,
+         reg->Register.File,
+         reg->Register.WriteMask,
+         reg->Register.Indirect,
+         reg->Register.Index,
          instruction,
          header );
       prev_token = (struct tgsi_token  *) dst_register;
 
-      if( tgsi_compare_dst_register_ext_concode(
-            reg->DstRegisterExtConcode,
-            tgsi_default_dst_register_ext_concode() ) ) {
-         struct tgsi_dst_register_ext_concode *dst_register_ext_concode;
-
-         if( maxsize <= size )
-            return 0;
-         dst_register_ext_concode =
-            (struct  tgsi_dst_register_ext_concode *) &tokens[size];
-         size++;
-
-         *dst_register_ext_concode =   tgsi_build_dst_register_ext_concode(
-            reg->DstRegisterExtConcode.CondMask,
-            reg->DstRegisterExtConcode.CondSwizzleX,
-            reg->DstRegisterExtConcode.CondSwizzleY,
-            reg->DstRegisterExtConcode.CondSwizzleZ,
-            reg->DstRegisterExtConcode.CondSwizzleW,
-            reg->DstRegisterExtConcode.CondSrcIndex,
-            prev_token,
-            instruction,
-            header );
-         prev_token = (struct tgsi_token  *) dst_register_ext_concode;
-      }
-
-      if( tgsi_compare_dst_register_ext_modulate(
-            reg->DstRegisterExtModulate,
-            tgsi_default_dst_register_ext_modulate() ) ) {
-         struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate;
-
-         if( maxsize <= size )
-            return 0;
-         dst_register_ext_modulate =
-            (struct  tgsi_dst_register_ext_modulate *) &tokens[size];
-         size++;
-
-         *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate(
-            reg->DstRegisterExtModulate.Modulate,
-            prev_token,
-            instruction,
-            header );
-         prev_token = (struct tgsi_token  *) dst_register_ext_modulate;
-      }
-
-      if( reg->DstRegister.Indirect ) {
+      if( reg->Register.Indirect ) {
          struct tgsi_src_register *ind;
 
          if( maxsize <= size )
@@ -649,22 +578,23 @@ tgsi_build_full_instruction(
          size++;
 
          *ind = tgsi_build_src_register(
-            reg->DstRegisterInd.File,
-            reg->DstRegisterInd.SwizzleX,
-            reg->DstRegisterInd.SwizzleY,
-            reg->DstRegisterInd.SwizzleZ,
-            reg->DstRegisterInd.SwizzleW,
-            reg->DstRegisterInd.Negate,
-            reg->DstRegisterInd.Indirect,
-            reg->DstRegisterInd.Dimension,
-            reg->DstRegisterInd.Index,
+            reg->Indirect.File,
+            reg->Indirect.SwizzleX,
+            reg->Indirect.SwizzleY,
+            reg->Indirect.SwizzleZ,
+            reg->Indirect.SwizzleW,
+            reg->Indirect.Negate,
+            reg->Indirect.Absolute,
+            reg->Indirect.Indirect,
+            reg->Indirect.Dimension,
+            reg->Indirect.Index,
             instruction,
             header );
       }
    }
 
    for( i = 0;  i < full_inst->Instruction.NumSrcRegs; i++ ) {
-      const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i];
+      const struct tgsi_full_src_register *reg = &full_inst->Src[i];
       struct tgsi_src_register *src_register;
       struct tgsi_token *prev_token;
 
@@ -674,77 +604,21 @@ tgsi_build_full_instruction(
       size++;
 
       *src_register = tgsi_build_src_register(
-         reg->SrcRegister.File,
-         reg->SrcRegister.SwizzleX,
-         reg->SrcRegister.SwizzleY,
-         reg->SrcRegister.SwizzleZ,
-         reg->SrcRegister.SwizzleW,
-         reg->SrcRegister.Negate,
-         reg->SrcRegister.Indirect,
-         reg->SrcRegister.Dimension,
-         reg->SrcRegister.Index,
+         reg->Register.File,
+         reg->Register.SwizzleX,
+         reg->Register.SwizzleY,
+         reg->Register.SwizzleZ,
+         reg->Register.SwizzleW,
+         reg->Register.Negate,
+         reg->Register.Absolute,
+         reg->Register.Indirect,
+         reg->Register.Dimension,
+         reg->Register.Index,
          instruction,
          header );
       prev_token = (struct tgsi_token  *) src_register;
 
-      if( tgsi_compare_src_register_ext_swz(
-            reg->SrcRegisterExtSwz,
-            tgsi_default_src_register_ext_swz() ) ) {
-         struct tgsi_src_register_ext_swz *src_register_ext_swz;
-
-         /* Use of the extended swizzle requires the simple swizzle to be identity.
-          */
-         assert( reg->SrcRegister.SwizzleX == TGSI_SWIZZLE_X );
-         assert( reg->SrcRegister.SwizzleY == TGSI_SWIZZLE_Y );
-         assert( reg->SrcRegister.SwizzleZ == TGSI_SWIZZLE_Z );
-         assert( reg->SrcRegister.SwizzleW == TGSI_SWIZZLE_W );
-         assert( reg->SrcRegister.Negate == FALSE );
-
-         if( maxsize <= size )
-            return 0;
-         src_register_ext_swz =
-            (struct  tgsi_src_register_ext_swz *) &tokens[size];
-         size++;
-
-         *src_register_ext_swz = tgsi_build_src_register_ext_swz(
-            reg->SrcRegisterExtSwz.ExtSwizzleX,
-            reg->SrcRegisterExtSwz.ExtSwizzleY,
-            reg->SrcRegisterExtSwz.ExtSwizzleZ,
-            reg->SrcRegisterExtSwz.ExtSwizzleW,
-            reg->SrcRegisterExtSwz.NegateX,
-            reg->SrcRegisterExtSwz.NegateY,
-            reg->SrcRegisterExtSwz.NegateZ,
-            reg->SrcRegisterExtSwz.NegateW,
-            prev_token,
-            instruction,
-            header );
-         prev_token = (struct tgsi_token  *) src_register_ext_swz;
-      }
-
-      if( tgsi_compare_src_register_ext_mod(
-            reg->SrcRegisterExtMod,
-            tgsi_default_src_register_ext_mod() ) ) {
-         struct tgsi_src_register_ext_mod *src_register_ext_mod;
-
-         if( maxsize <= size )
-            return 0;
-         src_register_ext_mod =
-            (struct  tgsi_src_register_ext_mod *) &tokens[size];
-         size++;
-
-         *src_register_ext_mod = tgsi_build_src_register_ext_mod(
-            reg->SrcRegisterExtMod.Complement,
-            reg->SrcRegisterExtMod.Bias,
-            reg->SrcRegisterExtMod.Scale2X,
-            reg->SrcRegisterExtMod.Absolute,
-            reg->SrcRegisterExtMod.Negate,
-            prev_token,
-            instruction,
-            header );
-         prev_token = (struct tgsi_token  *) src_register_ext_mod;
-      }
-
-      if( reg->SrcRegister.Indirect ) {
+      if( reg->Register.Indirect ) {
          struct  tgsi_src_register *ind;
 
          if( maxsize <= size )
@@ -753,23 +627,24 @@ tgsi_build_full_instruction(
          size++;
 
          *ind = tgsi_build_src_register(
-            reg->SrcRegisterInd.File,
-            reg->SrcRegisterInd.SwizzleX,
-            reg->SrcRegisterInd.SwizzleY,
-            reg->SrcRegisterInd.SwizzleZ,
-            reg->SrcRegisterInd.SwizzleW,
-            reg->SrcRegisterInd.Negate,
-            reg->SrcRegisterInd.Indirect,
-            reg->SrcRegisterInd.Dimension,
-            reg->SrcRegisterInd.Index,
+            reg->Indirect.File,
+            reg->Indirect.SwizzleX,
+            reg->Indirect.SwizzleY,
+            reg->Indirect.SwizzleZ,
+            reg->Indirect.SwizzleW,
+            reg->Indirect.Negate,
+            reg->Indirect.Absolute,
+            reg->Indirect.Indirect,
+            reg->Indirect.Dimension,
+            reg->Indirect.Index,
             instruction,
             header );
       }
 
-      if( reg->SrcRegister.Dimension ) {
+      if( reg->Register.Dimension ) {
          struct  tgsi_dimension *dim;
 
-         assert( !reg->SrcRegisterDim.Dimension );
+         assert( !reg->Dimension.Dimension );
 
          if( maxsize <= size )
             return 0;
@@ -777,12 +652,12 @@ tgsi_build_full_instruction(
          size++;
 
          *dim = tgsi_build_dimension(
-            reg->SrcRegisterDim.Indirect,
-            reg->SrcRegisterDim.Index,
+            reg->Dimension.Indirect,
+            reg->Dimension.Index,
             instruction,
             header );
 
-         if( reg->SrcRegisterDim.Indirect ) {
+         if( reg->Dimension.Indirect ) {
             struct tgsi_src_register *ind;
 
             if( maxsize <= size )
@@ -791,15 +666,16 @@ tgsi_build_full_instruction(
             size++;
 
             *ind = tgsi_build_src_register(
-               reg->SrcRegisterDimInd.File,
-               reg->SrcRegisterDimInd.SwizzleX,
-               reg->SrcRegisterDimInd.SwizzleY,
-               reg->SrcRegisterDimInd.SwizzleZ,
-               reg->SrcRegisterDimInd.SwizzleW,
-               reg->SrcRegisterDimInd.Negate,
-               reg->SrcRegisterDimInd.Indirect,
-               reg->SrcRegisterDimInd.Dimension,
-               reg->SrcRegisterDimInd.Index,
+               reg->DimIndirect.File,
+               reg->DimIndirect.SwizzleX,
+               reg->DimIndirect.SwizzleY,
+               reg->DimIndirect.SwizzleZ,
+               reg->DimIndirect.SwizzleW,
+               reg->DimIndirect.Negate,
+               reg->DimIndirect.Absolute,
+               reg->DimIndirect.Indirect,
+               reg->DimIndirect.Dimension,
+               reg->DimIndirect.Index,
                instruction,
                header );
          }
@@ -809,163 +685,103 @@ tgsi_build_full_instruction(
    return size;
 }
 
-struct tgsi_instruction_ext_nv
-tgsi_default_instruction_ext_nv( void )
+struct tgsi_instruction_predicate
+tgsi_default_instruction_predicate(void)
 {
-   struct tgsi_instruction_ext_nv instruction_ext_nv;
-
-   instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV;
-   instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT;
-   instruction_ext_nv.CondDstIndex = 0;
-   instruction_ext_nv.CondFlowIndex = 0;
-   instruction_ext_nv.CondMask = TGSI_CC_TR;
-   instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X;
-   instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y;
-   instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z;
-   instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W;
-   instruction_ext_nv.CondDstUpdate = 0;
-   instruction_ext_nv.CondFlowEnable = 0;
-   instruction_ext_nv.Padding = 0;
-   instruction_ext_nv.Extended = 0;
-
-   return instruction_ext_nv;
-}
+   struct tgsi_instruction_predicate instruction_predicate;
 
+   instruction_predicate.SwizzleX = TGSI_SWIZZLE_X;
+   instruction_predicate.SwizzleY = TGSI_SWIZZLE_Y;
+   instruction_predicate.SwizzleZ = TGSI_SWIZZLE_Z;
+   instruction_predicate.SwizzleW = TGSI_SWIZZLE_W;
+   instruction_predicate.Negate = 0;
+   instruction_predicate.Index = 0;
+   instruction_predicate.Padding = 0;
 
-/** test for inequality of 32-bit values pointed to by a and b */
-static INLINE boolean
-compare32(const void *a, const void *b)
-{
-   return *((uint32_t *) a) != *((uint32_t *) b);
+   return instruction_predicate;
 }
 
-
-unsigned
-tgsi_compare_instruction_ext_nv(
-   struct tgsi_instruction_ext_nv a,
-   struct tgsi_instruction_ext_nv b )
+struct tgsi_instruction_predicate
+tgsi_build_instruction_predicate(int index,
+                                 unsigned negate,
+                                 unsigned swizzleX,
+                                 unsigned swizzleY,
+                                 unsigned swizzleZ,
+                                 unsigned swizzleW,
+                                 struct tgsi_instruction *instruction,
+                                 struct tgsi_header *header)
 {
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
-}
+   struct tgsi_instruction_predicate instruction_predicate;
 
-struct tgsi_instruction_ext_nv
-tgsi_build_instruction_ext_nv(
-   unsigned precision,
-   unsigned cond_dst_index,
-   unsigned cond_flow_index,
-   unsigned cond_mask,
-   unsigned cond_swizzle_x,
-   unsigned cond_swizzle_y,
-   unsigned cond_swizzle_z,
-   unsigned cond_swizzle_w,
-   unsigned cond_dst_update,
-   unsigned cond_flow_enable,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_instruction_ext_nv instruction_ext_nv;
-
-   instruction_ext_nv = tgsi_default_instruction_ext_nv();
-   instruction_ext_nv.Precision = precision;
-   instruction_ext_nv.CondDstIndex = cond_dst_index;
-   instruction_ext_nv.CondFlowIndex = cond_flow_index;
-   instruction_ext_nv.CondMask = cond_mask;
-   instruction_ext_nv.CondSwizzleX = cond_swizzle_x;
-   instruction_ext_nv.CondSwizzleY = cond_swizzle_y;
-   instruction_ext_nv.CondSwizzleZ = cond_swizzle_z;
-   instruction_ext_nv.CondSwizzleW = cond_swizzle_w;
-   instruction_ext_nv.CondDstUpdate = cond_dst_update;
-   instruction_ext_nv.CondFlowEnable = cond_flow_enable;
-
-   prev_token->Extended = 1;
-   instruction_grow( instruction, header );
+   instruction_predicate = tgsi_default_instruction_predicate();
+   instruction_predicate.SwizzleX = swizzleX;
+   instruction_predicate.SwizzleY = swizzleY;
+   instruction_predicate.SwizzleZ = swizzleZ;
+   instruction_predicate.SwizzleW = swizzleW;
+   instruction_predicate.Negate = negate;
+   instruction_predicate.Index = index;
+
+   instruction_grow(instruction, header);
 
-   return instruction_ext_nv;
+   return instruction_predicate;
 }
 
-struct tgsi_instruction_ext_label
-tgsi_default_instruction_ext_label( void )
+struct tgsi_instruction_label
+tgsi_default_instruction_label( void )
 {
-   struct tgsi_instruction_ext_label instruction_ext_label;
+   struct tgsi_instruction_label instruction_label;
 
-   instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL;
-   instruction_ext_label.Label = 0;
-   instruction_ext_label.Padding = 0;
-   instruction_ext_label.Extended = 0;
-
-   return instruction_ext_label;
-}
+   instruction_label.Label = 0;
+   instruction_label.Padding = 0;
 
-unsigned
-tgsi_compare_instruction_ext_label(
-   struct tgsi_instruction_ext_label a,
-   struct tgsi_instruction_ext_label b )
-{
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
+   return instruction_label;
 }
 
-struct tgsi_instruction_ext_label
-tgsi_build_instruction_ext_label(
+struct tgsi_instruction_label
+tgsi_build_instruction_label(
    unsigned label,
    struct tgsi_token  *prev_token,
    struct tgsi_instruction *instruction,
    struct tgsi_header *header )
 {
-   struct tgsi_instruction_ext_label instruction_ext_label;
+   struct tgsi_instruction_label instruction_label;
 
-   instruction_ext_label = tgsi_default_instruction_ext_label();
-   instruction_ext_label.Label = label;
+   instruction_label = tgsi_default_instruction_label();
+   instruction_label.Label = label;
+   instruction->Label = 1;
 
-   prev_token->Extended = 1;
    instruction_grow( instruction, header );
 
-   return instruction_ext_label;
+   return instruction_label;
 }
 
-struct tgsi_instruction_ext_texture
-tgsi_default_instruction_ext_texture( void )
+struct tgsi_instruction_texture
+tgsi_default_instruction_texture( void )
 {
-   struct tgsi_instruction_ext_texture instruction_ext_texture;
+   struct tgsi_instruction_texture instruction_texture;
 
-   instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE;
-   instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN;
-   instruction_ext_texture.Padding = 0;
-   instruction_ext_texture.Extended = 0;
+   instruction_texture.Texture = TGSI_TEXTURE_UNKNOWN;
+   instruction_texture.Padding = 0;
 
-   return instruction_ext_texture;
-}
-
-unsigned
-tgsi_compare_instruction_ext_texture(
-   struct tgsi_instruction_ext_texture a,
-   struct tgsi_instruction_ext_texture b )
-{
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
+   return instruction_texture;
 }
 
-struct tgsi_instruction_ext_texture
-tgsi_build_instruction_ext_texture(
+struct tgsi_instruction_texture
+tgsi_build_instruction_texture(
    unsigned texture,
    struct tgsi_token *prev_token,
    struct tgsi_instruction *instruction,
    struct tgsi_header *header )
 {
-   struct tgsi_instruction_ext_texture instruction_ext_texture;
+   struct tgsi_instruction_texture instruction_texture;
 
-   instruction_ext_texture = tgsi_default_instruction_ext_texture();
-   instruction_ext_texture.Texture = texture;
+   instruction_texture = tgsi_default_instruction_texture();
+   instruction_texture.Texture = texture;
+   instruction->Texture = 1;
 
-   prev_token->Extended = 1;
    instruction_grow( instruction, header );
 
-   return instruction_ext_texture;
+   return instruction_texture;
 }
 
 struct tgsi_src_register
@@ -979,10 +795,10 @@ tgsi_default_src_register( void )
    src_register.SwizzleZ = TGSI_SWIZZLE_Z;
    src_register.SwizzleW = TGSI_SWIZZLE_W;
    src_register.Negate = 0;
+   src_register.Absolute = 0;
    src_register.Indirect = 0;
    src_register.Dimension = 0;
    src_register.Index = 0;
-   src_register.Extended = 0;
 
    return src_register;
 }
@@ -995,6 +811,7 @@ tgsi_build_src_register(
    unsigned swizzle_z,
    unsigned swizzle_w,
    unsigned negate,
+   unsigned absolute,
    unsigned indirect,
    unsigned dimension,
    int index,
@@ -1018,6 +835,7 @@ tgsi_build_src_register(
    src_register.SwizzleZ = swizzle_z;
    src_register.SwizzleW = swizzle_w;
    src_register.Negate = negate;
+   src_register.Absolute = absolute;
    src_register.Indirect = indirect;
    src_register.Dimension = dimension;
    src_register.Index = index;
@@ -1032,145 +850,14 @@ tgsi_default_full_src_register( void )
 {
    struct tgsi_full_src_register full_src_register;
 
-   full_src_register.SrcRegister = tgsi_default_src_register();
-   full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz();
-   full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod();
-   full_src_register.SrcRegisterInd = tgsi_default_src_register();
-   full_src_register.SrcRegisterDim = tgsi_default_dimension();
-   full_src_register.SrcRegisterDimInd = tgsi_default_src_register();
+   full_src_register.Register = tgsi_default_src_register();
+   full_src_register.Indirect = tgsi_default_src_register();
+   full_src_register.Dimension = tgsi_default_dimension();
+   full_src_register.DimIndirect = tgsi_default_src_register();
 
    return full_src_register;
 }
 
-struct tgsi_src_register_ext_swz
-tgsi_default_src_register_ext_swz( void )
-{
-   struct tgsi_src_register_ext_swz src_register_ext_swz;
-
-   src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ;
-   src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X;
-   src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y;
-   src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z;
-   src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W;
-   src_register_ext_swz.NegateX = 0;
-   src_register_ext_swz.NegateY = 0;
-   src_register_ext_swz.NegateZ = 0;
-   src_register_ext_swz.NegateW = 0;
-   src_register_ext_swz.Padding = 0;
-   src_register_ext_swz.Extended = 0;
-
-   return src_register_ext_swz;
-}
-
-unsigned
-tgsi_compare_src_register_ext_swz(
-   struct tgsi_src_register_ext_swz a,
-   struct tgsi_src_register_ext_swz b )
-{
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
-}
-
-struct tgsi_src_register_ext_swz
-tgsi_build_src_register_ext_swz(
-   unsigned ext_swizzle_x,
-   unsigned ext_swizzle_y,
-   unsigned ext_swizzle_z,
-   unsigned ext_swizzle_w,
-   unsigned negate_x,
-   unsigned negate_y,
-   unsigned negate_z,
-   unsigned negate_w,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_src_register_ext_swz src_register_ext_swz;
-
-   assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE );
-   assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE );
-   assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE );
-   assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE );
-   assert( negate_x <= 1 );
-   assert( negate_y <= 1 );
-   assert( negate_z <= 1 );
-   assert( negate_w <= 1 );
-
-   src_register_ext_swz = tgsi_default_src_register_ext_swz();
-   src_register_ext_swz.ExtSwizzleX = ext_swizzle_x;
-   src_register_ext_swz.ExtSwizzleY = ext_swizzle_y;
-   src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z;
-   src_register_ext_swz.ExtSwizzleW = ext_swizzle_w;
-   src_register_ext_swz.NegateX = negate_x;
-   src_register_ext_swz.NegateY = negate_y;
-   src_register_ext_swz.NegateZ = negate_z;
-   src_register_ext_swz.NegateW = negate_w;
-
-   prev_token->Extended = 1;
-   instruction_grow( instruction, header );
-
-   return src_register_ext_swz;
-}
-
-struct tgsi_src_register_ext_mod
-tgsi_default_src_register_ext_mod( void )
-{
-   struct tgsi_src_register_ext_mod src_register_ext_mod;
-
-   src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD;
-   src_register_ext_mod.Complement = 0;
-   src_register_ext_mod.Bias = 0;
-   src_register_ext_mod.Scale2X = 0;
-   src_register_ext_mod.Absolute = 0;
-   src_register_ext_mod.Negate = 0;
-   src_register_ext_mod.Padding = 0;
-   src_register_ext_mod.Extended = 0;
-
-   return src_register_ext_mod;
-}
-
-unsigned
-tgsi_compare_src_register_ext_mod(
-   struct tgsi_src_register_ext_mod a,
-   struct tgsi_src_register_ext_mod b )
-{
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
-}
-
-struct tgsi_src_register_ext_mod
-tgsi_build_src_register_ext_mod(
-   unsigned complement,
-   unsigned bias,
-   unsigned scale_2x,
-   unsigned absolute,
-   unsigned negate,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_src_register_ext_mod src_register_ext_mod;
-
-   assert( complement <= 1 );
-   assert( bias <= 1 );
-   assert( scale_2x <= 1 );
-   assert( absolute <= 1 );
-   assert( negate <= 1 );
-
-   src_register_ext_mod = tgsi_default_src_register_ext_mod();
-   src_register_ext_mod.Complement = complement;
-   src_register_ext_mod.Bias = bias;
-   src_register_ext_mod.Scale2X = scale_2x;
-   src_register_ext_mod.Absolute = absolute;
-   src_register_ext_mod.Negate = negate;
-
-   prev_token->Extended = 1;
-   instruction_grow( instruction, header );
-
-   return src_register_ext_mod;
-}
 
 struct tgsi_dimension
 tgsi_default_dimension( void )
@@ -1181,7 +868,6 @@ tgsi_default_dimension( void )
    dimension.Dimension = 0;
    dimension.Padding = 0;
    dimension.Index = 0;
-   dimension.Extended = 0;
 
    return dimension;
 }
@@ -1215,7 +901,6 @@ tgsi_default_dst_register( void )
    dst_register.Dimension = 0;
    dst_register.Index = 0;
    dst_register.Padding = 0;
-   dst_register.Extended = 0;
 
    return dst_register;
 }
@@ -1251,118 +936,113 @@ tgsi_default_full_dst_register( void )
 {
    struct tgsi_full_dst_register full_dst_register;
 
-   full_dst_register.DstRegister = tgsi_default_dst_register();
-   full_dst_register.DstRegisterInd = tgsi_default_src_register();
-   full_dst_register.DstRegisterExtConcode =
-      tgsi_default_dst_register_ext_concode();
-   full_dst_register.DstRegisterExtModulate =
-      tgsi_default_dst_register_ext_modulate();
+   full_dst_register.Register = tgsi_default_dst_register();
+   full_dst_register.Indirect = tgsi_default_src_register();
 
    return full_dst_register;
 }
 
-struct tgsi_dst_register_ext_concode
-tgsi_default_dst_register_ext_concode( void )
+struct tgsi_property
+tgsi_default_property( void )
 {
-   struct tgsi_dst_register_ext_concode dst_register_ext_concode;
-
-   dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE;
-   dst_register_ext_concode.CondMask = TGSI_CC_TR;
-   dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X;
-   dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y;
-   dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z;
-   dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W;
-   dst_register_ext_concode.CondSrcIndex = 0;
-   dst_register_ext_concode.Padding = 0;
-   dst_register_ext_concode.Extended = 0;
-
-   return dst_register_ext_concode;
+   struct tgsi_property property;
+
+   property.Type = TGSI_TOKEN_TYPE_PROPERTY;
+   property.NrTokens = 1;
+   property.PropertyName = TGSI_PROPERTY_GS_INPUT_PRIM;
+   property.Padding = 0;
+
+   return property;
 }
 
-unsigned
-tgsi_compare_dst_register_ext_concode(
-   struct tgsi_dst_register_ext_concode a,
-   struct tgsi_dst_register_ext_concode b )
+struct tgsi_property
+tgsi_build_property(unsigned property_name,
+                    struct tgsi_header *header)
 {
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
+   struct tgsi_property property;
+
+   property = tgsi_default_property();
+   property.PropertyName = property_name;
+
+   header_bodysize_grow( header );
+
+   return property;
 }
 
-struct tgsi_dst_register_ext_concode
-tgsi_build_dst_register_ext_concode(
-   unsigned cc,
-   unsigned swizzle_x,
-   unsigned swizzle_y,
-   unsigned swizzle_z,
-   unsigned swizzle_w,
-   int index,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
+
+struct tgsi_full_property
+tgsi_default_full_property( void )
 {
-   struct tgsi_dst_register_ext_concode dst_register_ext_concode;
+   struct tgsi_full_property  full_property;
 
-   assert( cc <= TGSI_CC_FL );
-   assert( swizzle_x <= TGSI_SWIZZLE_W );
-   assert( swizzle_y <= TGSI_SWIZZLE_W );
-   assert( swizzle_z <= TGSI_SWIZZLE_W );
-   assert( swizzle_w <= TGSI_SWIZZLE_W );
-   assert( index >= -32768 && index <= 32767 );
+   full_property.Property  = tgsi_default_property();
+   memset(full_property.u, 0,
+          sizeof(struct tgsi_property_data) * 8);
 
-   dst_register_ext_concode = tgsi_default_dst_register_ext_concode();
-   dst_register_ext_concode.CondMask = cc;
-   dst_register_ext_concode.CondSwizzleX = swizzle_x;
-   dst_register_ext_concode.CondSwizzleY = swizzle_y;
-   dst_register_ext_concode.CondSwizzleZ = swizzle_z;
-   dst_register_ext_concode.CondSwizzleW = swizzle_w;
-   dst_register_ext_concode.CondSrcIndex = index;
+   return full_property;
+}
 
-   prev_token->Extended = 1;
-   instruction_grow( instruction, header );
+static void
+property_grow(
+   struct tgsi_property *property,
+   struct tgsi_header *header )
+{
+   assert( property->NrTokens < 0xFF );
 
-   return dst_register_ext_concode;
+   property->NrTokens++;
+
+   header_bodysize_grow( header );
 }
 
-struct tgsi_dst_register_ext_modulate
-tgsi_default_dst_register_ext_modulate( void )
+struct tgsi_property_data
+tgsi_build_property_data(
+   unsigned value,
+   struct tgsi_property *property,
+   struct tgsi_header *header )
 {
-   struct tgsi_dst_register_ext_modulate dst_register_ext_modulate;
+   struct tgsi_property_data property_data;
+
+   property_data.Data = value;
 
-   dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE;
-   dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X;
-   dst_register_ext_modulate.Padding = 0;
-   dst_register_ext_modulate.Extended = 0;
+   property_grow( property, header );
 
-   return dst_register_ext_modulate;
+   return property_data;
 }
 
 unsigned
-tgsi_compare_dst_register_ext_modulate(
-   struct tgsi_dst_register_ext_modulate a,
-   struct tgsi_dst_register_ext_modulate b )
+tgsi_build_full_property(
+   const struct tgsi_full_property *full_prop,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize )
 {
-   a.Padding = b.Padding = 0;
-   a.Extended = b.Extended = 0;
-   return compare32(&a, &b);
-}
+   unsigned size = 0, i;
+   struct tgsi_property *property;
 
-struct tgsi_dst_register_ext_modulate
-tgsi_build_dst_register_ext_modulate(
-   unsigned modulate,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header )
-{
-   struct tgsi_dst_register_ext_modulate dst_register_ext_modulate;
+   if( maxsize <= size )
+      return 0;
+   property = (struct tgsi_property *) &tokens[size];
+   size++;
 
-   assert( modulate <= TGSI_MODULATE_EIGHTH );
+   *property = tgsi_build_property(
+      full_prop->Property.PropertyName,
+      header );
 
-   dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate();
-   dst_register_ext_modulate.Modulate = modulate;
+   assert( full_prop->Property.NrTokens <= 8 + 1 );
 
-   prev_token->Extended = 1;
-   instruction_grow( instruction, header );
+   for( i = 0; i < full_prop->Property.NrTokens - 1; i++ ) {
+      struct tgsi_property_data *data;
+
+      if( maxsize <= size )
+         return  0;
+      data = (struct tgsi_property_data *) &tokens[size];
+      size++;
 
-   return dst_register_ext_modulate;
+      *data = tgsi_build_property_data(
+         full_prop->u[i].Data,
+         property,
+         header );
+   }
+
+   return size;
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h
index 17d977b059..9de2757fe4 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.h
@@ -36,12 +36,6 @@ struct tgsi_token;
 extern "C" {
 #endif
 
-/*
- * version
- */
-
-struct tgsi_version
-tgsi_build_version( void );
 
 /*
  * header
@@ -133,6 +127,34 @@ tgsi_build_full_immediate(
    unsigned maxsize );
 
 /*
+ * properties
+ */
+
+struct tgsi_property
+tgsi_default_property( void );
+
+struct tgsi_property
+tgsi_build_property(
+   unsigned property_name,
+   struct tgsi_header *header );
+
+struct tgsi_full_property
+tgsi_default_full_property( void );
+
+struct tgsi_property_data
+tgsi_build_property_data(
+   unsigned value,
+   struct tgsi_property *property,
+   struct tgsi_header *header );
+
+unsigned
+tgsi_build_full_property(
+   const struct tgsi_full_property *full_prop,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize );
+
+/*
  * instruction
  */
 
@@ -143,6 +165,7 @@ struct tgsi_instruction
 tgsi_build_instruction(
    unsigned opcode,
    unsigned saturate,
+   unsigned predicate,
    unsigned num_dst_regs,
    unsigned num_src_regs,
    struct tgsi_header *header );
@@ -157,55 +180,34 @@ tgsi_build_full_instruction(
    struct tgsi_header *header,
    unsigned maxsize );
 
-struct tgsi_instruction_ext_nv
-tgsi_default_instruction_ext_nv( void );
-
-unsigned
-tgsi_compare_instruction_ext_nv(
-   struct tgsi_instruction_ext_nv a,
-   struct tgsi_instruction_ext_nv b );
-
-struct tgsi_instruction_ext_nv
-tgsi_build_instruction_ext_nv(
-   unsigned precision,
-   unsigned cond_dst_index,
-   unsigned cond_flow_index,
-   unsigned cond_mask,
-   unsigned cond_swizzle_x,
-   unsigned cond_swizzle_y,
-   unsigned cond_swizzle_z,
-   unsigned cond_swizzle_w,
-   unsigned cond_dst_update,
-   unsigned cond_flow_enable,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
+struct tgsi_instruction_predicate
+tgsi_default_instruction_predicate(void);
 
-struct tgsi_instruction_ext_label
-tgsi_default_instruction_ext_label( void );
+struct tgsi_instruction_predicate
+tgsi_build_instruction_predicate(int index,
+                                 unsigned negate,
+                                 unsigned swizzleX,
+                                 unsigned swizzleY,
+                                 unsigned swizzleZ,
+                                 unsigned swizzleW,
+                                 struct tgsi_instruction *instruction,
+                                 struct tgsi_header *header);
 
-unsigned
-tgsi_compare_instruction_ext_label(
-   struct tgsi_instruction_ext_label a,
-   struct tgsi_instruction_ext_label b );
+struct tgsi_instruction_label
+tgsi_default_instruction_label( void );
 
-struct tgsi_instruction_ext_label
-tgsi_build_instruction_ext_label(
+struct tgsi_instruction_label
+tgsi_build_instruction_label(
    unsigned label,
    struct tgsi_token *prev_token,
    struct tgsi_instruction *instruction,
    struct tgsi_header *header );
 
-struct tgsi_instruction_ext_texture
-tgsi_default_instruction_ext_texture( void );
-
-unsigned
-tgsi_compare_instruction_ext_texture(
-   struct tgsi_instruction_ext_texture a,
-   struct tgsi_instruction_ext_texture b );
+struct tgsi_instruction_texture
+tgsi_default_instruction_texture( void );
 
-struct tgsi_instruction_ext_texture
-tgsi_build_instruction_ext_texture(
+struct tgsi_instruction_texture
+tgsi_build_instruction_texture(
    unsigned texture,
    struct tgsi_token *prev_token,
    struct tgsi_instruction *instruction,
@@ -222,6 +224,7 @@ tgsi_build_src_register(
    unsigned swizzle_z,
    unsigned swizzle_w,
    unsigned negate,
+   unsigned absolute,
    unsigned indirect,
    unsigned dimension,
    int index,
@@ -231,46 +234,6 @@ tgsi_build_src_register(
 struct tgsi_full_src_register
 tgsi_default_full_src_register( void );
 
-struct tgsi_src_register_ext_swz
-tgsi_default_src_register_ext_swz( void );
-
-unsigned
-tgsi_compare_src_register_ext_swz(
-   struct tgsi_src_register_ext_swz a,
-   struct tgsi_src_register_ext_swz b );
-
-struct tgsi_src_register_ext_swz
-tgsi_build_src_register_ext_swz(
-   unsigned ext_swizzle_x,
-   unsigned ext_swizzle_y,
-   unsigned ext_swizzle_z,
-   unsigned ext_swizzle_w,
-   unsigned negate_x,
-   unsigned negate_y,
-   unsigned negate_z,
-   unsigned negate_w,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_src_register_ext_mod
-tgsi_default_src_register_ext_mod( void );
-
-unsigned
-tgsi_compare_src_register_ext_mod(
-   struct tgsi_src_register_ext_mod a,
-   struct tgsi_src_register_ext_mod b );
-
-struct tgsi_src_register_ext_mod
-tgsi_build_src_register_ext_mod(
-   unsigned complement,
-   unsigned bias,
-   unsigned scale_2x,
-   unsigned absolute,
-   unsigned negate,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
 
 struct tgsi_dimension
 tgsi_default_dimension( void );
@@ -297,40 +260,6 @@ tgsi_build_dst_register(
 struct tgsi_full_dst_register
 tgsi_default_full_dst_register( void );
 
-struct tgsi_dst_register_ext_concode
-tgsi_default_dst_register_ext_concode( void );
-
-unsigned
-tgsi_compare_dst_register_ext_concode(
-   struct tgsi_dst_register_ext_concode a,
-   struct tgsi_dst_register_ext_concode b );
-
-struct tgsi_dst_register_ext_concode
-tgsi_build_dst_register_ext_concode(
-   unsigned cc,
-   unsigned swizzle_x,
-   unsigned swizzle_y,
-   unsigned swizzle_z,
-   unsigned swizzle_w,
-   int index,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
-
-struct tgsi_dst_register_ext_modulate
-tgsi_default_dst_register_ext_modulate( void );
-
-unsigned
-tgsi_compare_dst_register_ext_modulate(
-   struct tgsi_dst_register_ext_modulate a,
-   struct tgsi_dst_register_ext_modulate b );
-
-struct tgsi_dst_register_ext_modulate
-tgsi_build_dst_register_ext_modulate(
-   unsigned modulate,
-   struct tgsi_token *prev_token,
-   struct tgsi_instruction *instruction,
-   struct tgsi_header *header );
 
 #if defined __cplusplus
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index 111d95b666..e2e5394f86 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -100,7 +100,9 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "SAMP",
    "ADDR",
    "IMM",
-   "LOOP"
+   "LOOP",
+   "PRED",
+   "SV"
 };
 
 static const char *interpolate_names[] =
@@ -119,12 +121,16 @@ static const char *semantic_names[] =
    "PSIZE",
    "GENERIC",
    "NORMAL",
-   "FACE"
+   "FACE",
+   "EDGEFLAG",
+   "PRIM_ID"
 };
 
 static const char *immediate_type_names[] =
 {
-   "FLT32"
+   "FLT32",
+   "UINT32",
+   "INT32"
 };
 
 static const char *swizzle_names[] =
@@ -148,35 +154,42 @@ static const char *texture_names[] =
    "SHADOWRECT"
 };
 
-static const char *extswizzle_names[] =
+static const char *property_names[] =
 {
-   "x",
-   "y",
-   "z",
-   "w",
-   "0",
-   "1"
+   "GS_INPUT_PRIMITIVE",
+   "GS_OUTPUT_PRIMITIVE",
+   "GS_MAX_OUTPUT_VERTICES"
 };
 
-static const char *modulate_names[TGSI_MODULATE_COUNT] =
+static const char *primitive_names[] =
 {
-   "",
-   "_2X",
-   "_4X",
-   "_8X",
-   "_D2",
-   "_D4",
-   "_D8"
+   "POINTS",
+   "LINES",
+   "LINE_LOOP",
+   "LINE_STRIP",
+   "TRIANGLES",
+   "TRIANGLE_STRIP",
+   "TRIANGLE_FAN",
+   "QUADS",
+   "QUAD_STRIP",
+   "POLYGON"
 };
 
+
 static void
-_dump_register(
+_dump_register_decl(
    struct dump_ctx *ctx,
    uint file,
    int first,
    int last )
 {
    ENM( file, file_names );
+
+   /* all geometry shader inputs are two dimensional */
+   if (file == TGSI_FILE_INPUT &&
+       ctx->iter.processor.Processor == TGSI_PROCESSOR_GEOMETRY)
+      TXT("[]");
+
    CHR( '[' );
    SID( first );
    if (first != last) {
@@ -187,6 +200,52 @@ _dump_register(
 }
 
 static void
+_dump_register_dst(
+   struct dump_ctx *ctx,
+   uint file,
+   int index)
+{
+   ENM( file, file_names );
+
+   CHR( '[' );
+   SID( index );
+   CHR( ']' );
+}
+
+
+static void
+_dump_register_src(
+   struct dump_ctx *ctx,
+   const struct tgsi_full_src_register *src )
+{
+   if (src->Register.Indirect) {
+      ENM( src->Register.File, file_names );
+      CHR( '[' );
+      ENM( src->Indirect.File, file_names );
+      CHR( '[' );
+      SID( src->Indirect.Index );
+      TXT( "]." );
+      ENM( src->Indirect.SwizzleX, swizzle_names );
+      if (src->Register.Index != 0) {
+         if (src->Register.Index > 0)
+            CHR( '+' );
+         SID( src->Register.Index );
+      }
+      CHR( ']' );
+   } else {
+      ENM( src->Register.File, file_names );
+      CHR( '[' );
+      SID( src->Register.Index );
+      CHR( ']' );
+   }
+   if (src->Register.Dimension) {
+      CHR( '[' );
+      SID( src->Dimension.Index );
+      CHR( ']' );
+   }
+}
+
+static void
 _dump_register_ind(
    struct dump_ctx *ctx,
    uint file,
@@ -240,22 +299,22 @@ iter_declaration(
 
    TXT( "DCL " );
 
-   _dump_register(
+   _dump_register_decl(
       ctx,
       decl->Declaration.File,
-      decl->DeclarationRange.First,
-      decl->DeclarationRange.Last );
+      decl->Range.First,
+      decl->Range.Last );
    _dump_writemask(
       ctx,
       decl->Declaration.UsageMask );
 
    if (decl->Declaration.Semantic) {
       TXT( ", " );
-      ENM( decl->Semantic.SemanticName, semantic_names );
-      if (decl->Semantic.SemanticIndex != 0 ||
-          decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC) {
+      ENM( decl->Semantic.Name, semantic_names );
+      if (decl->Semantic.Index != 0 ||
+          decl->Semantic.Name == TGSI_SEMANTIC_GENERIC) {
          CHR( '[' );
-         UID( decl->Semantic.SemanticIndex );
+         UID( decl->Semantic.Index );
          CHR( ']' );
       }
    }
@@ -292,6 +351,50 @@ tgsi_dump_declaration(
 }
 
 static boolean
+iter_property(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_property *prop )
+{
+   int i;
+   struct dump_ctx *ctx = (struct dump_ctx *)iter;
+
+   assert(Elements(property_names) == TGSI_PROPERTY_COUNT);
+
+   TXT( "PROPERTY " );
+   ENM(prop->Property.PropertyName, property_names);
+
+   if (prop->Property.NrTokens > 1)
+      TXT(" ");
+
+   for (i = 0; i < prop->Property.NrTokens - 1; ++i) {
+      switch (prop->Property.PropertyName) {
+      case TGSI_PROPERTY_GS_INPUT_PRIM:
+      case TGSI_PROPERTY_GS_OUTPUT_PRIM:
+         ENM(prop->u[i].Data, primitive_names);
+         break;
+      default:
+         SID( prop->u[i].Data );
+         break;
+      }
+      if (i < prop->Property.NrTokens - 2)
+         TXT( ", " );
+   }
+   EOL();
+
+   return TRUE;
+}
+
+void tgsi_dump_property(
+   const struct tgsi_full_property *prop )
+{
+   struct dump_ctx ctx;
+
+   ctx.printf = dump_ctx_printf;
+
+   iter_property( &ctx.iter, (struct tgsi_full_property *)prop );
+}
+
+static boolean
 iter_immediate(
    struct tgsi_iterate_context *iter,
    struct tgsi_full_immediate *imm )
@@ -311,6 +414,12 @@ iter_immediate(
       case TGSI_IMM_FLOAT32:
          FLT( imm->u[i].Float );
          break;
+      case TGSI_IMM_UINT32:
+         UID(imm->u[i].Uint);
+         break;
+      case TGSI_IMM_INT32:
+         SID(imm->u[i].Int);
+         break;
       default:
          assert( 0 );
       }
@@ -371,117 +480,66 @@ iter_instruction(
    }
 
    for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
-      const struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+      const struct tgsi_full_dst_register *dst = &inst->Dst[i];
 
       if (!first_reg)
          CHR( ',' );
       CHR( ' ' );
 
-      if (dst->DstRegister.Indirect) {
+      if (dst->Register.Indirect) {
          _dump_register_ind(
             ctx,
-            dst->DstRegister.File,
-            dst->DstRegister.Index,
-            dst->DstRegisterInd.File,
-            dst->DstRegisterInd.Index,
-            dst->DstRegisterInd.SwizzleX );
+            dst->Register.File,
+            dst->Register.Index,
+            dst->Indirect.File,
+            dst->Indirect.Index,
+            dst->Indirect.SwizzleX );
       }
       else {
-         _dump_register(
+         _dump_register_dst(
             ctx,
-            dst->DstRegister.File,
-            dst->DstRegister.Index,
-            dst->DstRegister.Index );
+            dst->Register.File,
+            dst->Register.Index );
       }
-      ENM( dst->DstRegisterExtModulate.Modulate, modulate_names );
-      _dump_writemask( ctx, dst->DstRegister.WriteMask );
+      _dump_writemask( ctx, dst->Register.WriteMask );
 
       first_reg = FALSE;
    }
 
    for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
-      const struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      const struct tgsi_full_src_register *src = &inst->Src[i];
 
       if (!first_reg)
          CHR( ',' );
       CHR( ' ' );
 
-      if (src->SrcRegisterExtMod.Negate)
-         TXT( "-(" );
-      if (src->SrcRegisterExtMod.Absolute)
-         CHR( '|' );
-      if (src->SrcRegisterExtMod.Scale2X)
-         TXT( "2*(" );
-      if (src->SrcRegisterExtMod.Bias)
-         CHR( '(' );
-      if (src->SrcRegisterExtMod.Complement)
-         TXT( "1-(" );
-      if (src->SrcRegister.Negate)
+      if (src->Register.Negate)
          CHR( '-' );
+      if (src->Register.Absolute)
+         CHR( '|' );
 
-      if (src->SrcRegister.Indirect) {
-         _dump_register_ind(
-            ctx,
-            src->SrcRegister.File,
-            src->SrcRegister.Index,
-            src->SrcRegisterInd.File,
-            src->SrcRegisterInd.Index,
-            src->SrcRegisterInd.SwizzleX );
-      }
-      else {
-         _dump_register(
-            ctx,
-            src->SrcRegister.File,
-            src->SrcRegister.Index,
-            src->SrcRegister.Index );
-      }
+      _dump_register_src(ctx, src);
 
-      if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X ||
-          src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y ||
-          src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z ||
-          src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) {
-         CHR( '.' );
-         ENM( src->SrcRegister.SwizzleX, swizzle_names );
-         ENM( src->SrcRegister.SwizzleY, swizzle_names );
-         ENM( src->SrcRegister.SwizzleZ, swizzle_names );
-         ENM( src->SrcRegister.SwizzleW, swizzle_names );
-      }
-      if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X ||
-          src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y ||
-          src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z ||
-          src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) {
+      if (src->Register.SwizzleX != TGSI_SWIZZLE_X ||
+          src->Register.SwizzleY != TGSI_SWIZZLE_Y ||
+          src->Register.SwizzleZ != TGSI_SWIZZLE_Z ||
+          src->Register.SwizzleW != TGSI_SWIZZLE_W) {
          CHR( '.' );
-         if (src->SrcRegisterExtSwz.NegateX)
-            TXT("-");
-         ENM( src->SrcRegisterExtSwz.ExtSwizzleX, extswizzle_names );
-         if (src->SrcRegisterExtSwz.NegateY)
-            TXT("-");
-         ENM( src->SrcRegisterExtSwz.ExtSwizzleY, extswizzle_names );
-         if (src->SrcRegisterExtSwz.NegateZ)
-            TXT("-");
-         ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, extswizzle_names );
-         if (src->SrcRegisterExtSwz.NegateW)
-            TXT("-");
-         ENM( src->SrcRegisterExtSwz.ExtSwizzleW, extswizzle_names );
+         ENM( src->Register.SwizzleX, swizzle_names );
+         ENM( src->Register.SwizzleY, swizzle_names );
+         ENM( src->Register.SwizzleZ, swizzle_names );
+         ENM( src->Register.SwizzleW, swizzle_names );
       }
 
-      if (src->SrcRegisterExtMod.Complement)
-         CHR( ')' );
-      if (src->SrcRegisterExtMod.Bias)
-         TXT( ")-.5" );
-      if (src->SrcRegisterExtMod.Scale2X)
-         CHR( ')' );
-      if (src->SrcRegisterExtMod.Absolute)
+      if (src->Register.Absolute)
          CHR( '|' );
-      if (src->SrcRegisterExtMod.Negate)
-         CHR( ')' );
 
       first_reg = FALSE;
    }
 
-   if (inst->InstructionExtTexture.Texture != TGSI_TEXTURE_UNKNOWN) {
+   if (inst->Instruction.Texture) {
       TXT( ", " );
-      ENM( inst->InstructionExtTexture.Texture, texture_names );
+      ENM( inst->Texture.Texture, texture_names );
    }
 
    switch (inst->Instruction.Opcode) {
@@ -491,7 +549,7 @@ iter_instruction(
    case TGSI_OPCODE_ENDLOOP:
    case TGSI_OPCODE_CAL:
       TXT( " :" );
-      UID( inst->InstructionExtLabel.Label );
+      UID( inst->Label.Label );
       break;
    }
 
@@ -529,9 +587,6 @@ prolog(
 {
    struct dump_ctx *ctx = (struct dump_ctx *) iter;
    ENM( iter->processor.Processor, processor_type_names );
-   UID( iter->version.MajorVersion );
-   CHR( '.' );
-   UID( iter->version.MinorVersion );
    EOL();
    return TRUE;
 }
@@ -547,6 +602,7 @@ tgsi_dump(
    ctx.iter.iterate_instruction = iter_instruction;
    ctx.iter.iterate_declaration = iter_declaration;
    ctx.iter.iterate_immediate = iter_immediate;
+   ctx.iter.iterate_property = iter_property;
    ctx.iter.epilog = NULL;
 
    ctx.instno = 0;
@@ -601,6 +657,7 @@ tgsi_dump_str(
    ctx.base.iter.iterate_instruction = iter_instruction;
    ctx.base.iter.iterate_declaration = iter_declaration;
    ctx.base.iter.iterate_immediate = iter_immediate;
+   ctx.base.iter.iterate_property = iter_property;
    ctx.base.iter.epilog = NULL;
 
    ctx.base.instno = 0;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.h b/src/gallium/auxiliary/tgsi/tgsi_dump.h
index ad1e647ec9..4cd27317b3 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.h
@@ -49,6 +49,7 @@ tgsi_dump(
 struct tgsi_full_immediate;
 struct tgsi_full_instruction;
 struct tgsi_full_declaration;
+struct tgsi_full_property;
 
 void
 tgsi_dump_immediate(
@@ -63,6 +64,10 @@ void
 tgsi_dump_declaration(
    const struct tgsi_full_declaration *decl );
 
+void
+tgsi_dump_property(
+   const struct tgsi_full_property *prop );
+
 #if defined __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
index 4a9c02b141..47fd1dd590 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
@@ -79,7 +79,8 @@ static const char *TGSI_FILES[TGSI_FILE_COUNT] =
    "FILE_SAMPLER",
    "FILE_ADDRESS",
    "FILE_IMMEDIATE",
-   "FILE_LOOP"
+   "FILE_LOOP",
+   "FILE_PREDICATE"
 };
 
 static const char *TGSI_INTERPOLATES[] =
@@ -112,34 +113,6 @@ static const char *TGSI_SATS[] =
    "SAT_MINUS_PLUS_ONE"
 };
 
-static const char *TGSI_INSTRUCTION_EXTS[] =
-{
-   "INSTRUCTION_EXT_TYPE_NV",
-   "INSTRUCTION_EXT_TYPE_LABEL",
-   "INSTRUCTION_EXT_TYPE_TEXTURE"
-};
-
-static const char *TGSI_PRECISIONS[] =
-{
-   "PRECISION_DEFAULT",
-   "PRECISION_FLOAT32",
-   "PRECISION_FLOAT16",
-   "PRECISION_FIXED12"
-};
-
-static const char *TGSI_CCS[] =
-{
-   "CC_GT",
-   "CC_EQ",
-   "CC_LT",
-   "CC_UN",
-   "CC_GE",
-   "CC_LE",
-   "CC_NE",
-   "CC_TR",
-   "CC_FL"
-};
-
 static const char *TGSI_SWIZZLES[] =
 {
    "SWIZZLE_X",
@@ -161,22 +134,6 @@ static const char *TGSI_TEXTURES[] =
    "TEXTURE_SHADOWRECT"
 };
 
-static const char *TGSI_SRC_REGISTER_EXTS[] =
-{
-   "SRC_REGISTER_EXT_TYPE_SWZ",
-   "SRC_REGISTER_EXT_TYPE_MOD"
-};
-
-static const char *TGSI_EXTSWIZZLES[] =
-{
-   "EXTSWIZZLE_X",
-   "EXTSWIZZLE_Y",
-   "EXTSWIZZLE_Z",
-   "EXTSWIZZLE_W",
-   "EXTSWIZZLE_ZERO",
-   "EXTSWIZZLE_ONE"
-};
-
 static const char *TGSI_WRITEMASKS[] =
 {
    "0",
@@ -197,23 +154,6 @@ static const char *TGSI_WRITEMASKS[] =
    "WRITEMASK_XYZW"
 };
 
-static const char *TGSI_DST_REGISTER_EXTS[] =
-{
-   "DST_REGISTER_EXT_TYPE_CONDCODE",
-   "DST_REGISTER_EXT_TYPE_MODULATE"
-};
-
-static const char *TGSI_MODULATES[] =
-{
-   "MODULATE_1X",
-   "MODULATE_2X",
-   "MODULATE_4X",
-   "MODULATE_8X",
-   "MODULATE_HALF",
-   "MODULATE_QUARTER",
-   "MODULATE_EIGHTH"
-};
-
 static void
 dump_declaration_verbose(
    struct tgsi_full_declaration  *decl,
@@ -246,6 +186,14 @@ dump_declaration_verbose(
       TXT( "\nSemantic   : " );
       UID( decl->Declaration.Semantic );
    }
+   if (deflt || fd->Declaration.Centroid != decl->Declaration.Centroid) {
+      TXT("\nCentroid   : ");
+      UID(decl->Declaration.Centroid);
+   }
+   if (deflt || fd->Declaration.Invariant != decl->Declaration.Invariant) {
+      TXT("\nInvariant  : ");
+      UID(decl->Declaration.Invariant);
+   }
    if( ignored ) {
       TXT( "\nPadding    : " );
       UIX( decl->Declaration.Padding );
@@ -253,16 +201,16 @@ dump_declaration_verbose(
 
    EOL();
    TXT( "\nFirst: " );
-   UID( decl->DeclarationRange.First );
+   UID( decl->Range.First );
    TXT( "\nLast : " );
-   UID( decl->DeclarationRange.Last );
+   UID( decl->Range.Last );
 
    if( decl->Declaration.Semantic ) {
       EOL();
-      TXT( "\nSemanticName : " );
-      ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS );
-      TXT( "\nSemanticIndex: " );
-      UID( decl->Semantic.SemanticIndex );
+      TXT( "\nName : " );
+      ENM( decl->Semantic.Name, TGSI_SEMANTICS );
+      TXT( "\nIndex: " );
+      UID( decl->Semantic.Index );
       if( ignored ) {
          TXT( "\nPadding      : " );
          UIX( decl->Semantic.Padding );
@@ -322,318 +270,122 @@ dump_instruction_verbose(
       TXT( "\nNumSrcRegs : " );
       UID( inst->Instruction.NumSrcRegs );
    }
+   if (deflt || fi->Instruction.Predicate != inst->Instruction.Predicate) {
+      TXT("\nPredicate  : ");
+      UID(inst->Instruction.Predicate);
+   }
+   if (deflt || fi->Instruction.Label != inst->Instruction.Label) {
+      TXT("\nLabel      : ");
+      UID(inst->Instruction.Label);
+   }
+   if (deflt || fi->Instruction.Texture != inst->Instruction.Texture) {
+      TXT("\nTexture    : ");
+      UID(inst->Instruction.Texture);
+   }
    if( ignored ) {
       TXT( "\nPadding    : " );
       UIX( inst->Instruction.Padding );
    }
 
-   if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) {
-      EOL();
-      TXT( "\nType          : " );
-      ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS );
-      if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) {
-         TXT( "\nPrecision     : " );
-         ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS );
-      }
-      if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) {
-         TXT( "\nCondDstIndex  : " );
-         UID( inst->InstructionExtNv.CondDstIndex );
-      }
-      if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) {
-         TXT( "\nCondFlowIndex : " );
-         UID( inst->InstructionExtNv.CondFlowIndex );
-      }
-      if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) {
-         TXT( "\nCondMask      : " );
-         ENM( inst->InstructionExtNv.CondMask, TGSI_CCS );
-      }
-      if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) {
-         TXT( "\nCondSwizzleX  : " );
-         ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES );
-      }
-      if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) {
-         TXT( "\nCondSwizzleY  : " );
-         ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES );
-      }
-      if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) {
-         TXT( "\nCondSwizzleZ  : " );
-         ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES );
-      }
-      if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) {
-         TXT( "\nCondSwizzleW  : " );
-         ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES );
-      }
-      if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) {
-         TXT( "\nCondDstUpdate : " );
-         UID( inst->InstructionExtNv.CondDstUpdate );
-      }
-      if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) {
-         TXT( "\nCondFlowEnable: " );
-         UID( inst->InstructionExtNv.CondFlowEnable );
-      }
-      if( ignored ) {
-         TXT( "\nPadding       : " );
-         UIX( inst->InstructionExtNv.Padding );
-         if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) {
-            TXT( "\nExtended      : " );
-            UID( inst->InstructionExtNv.Extended );
-         }
-      }
-   }
-
-   if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) {
+   if (deflt || inst->Instruction.Label) {
       EOL();
-      TXT( "\nType    : " );
-      ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS );
-      if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) {
+      if (deflt || fi->Label.Label != inst->Label.Label) {
          TXT( "\nLabel   : " );
-         UID( inst->InstructionExtLabel.Label );
+         UID(inst->Label.Label);
       }
       if( ignored ) {
          TXT( "\nPadding : " );
-         UIX( inst->InstructionExtLabel.Padding );
-         if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) {
-            TXT( "\nExtended: " );
-            UID( inst->InstructionExtLabel.Extended );
-         }
+         UIX(inst->Label.Padding);
       }
    }
 
-   if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) {
+   if (deflt || inst->Instruction.Texture) {
       EOL();
-      TXT( "\nType    : " );
-      ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS );
-      if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) {
+      if (deflt || fi->Texture.Texture != inst->Texture.Texture) {
          TXT( "\nTexture : " );
-         ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES );
+         ENM(inst->Texture.Texture, TGSI_TEXTURES);
       }
       if( ignored ) {
          TXT( "\nPadding : " );
-         UIX( inst->InstructionExtTexture.Padding );
-         if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) {
-            TXT( "\nExtended: " );
-            UID( inst->InstructionExtTexture.Extended );
-         }
+         UIX(inst->Texture.Padding);
       }
    }
 
    for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) {
-      struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
-      struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i];
+      struct tgsi_full_dst_register *dst = &inst->Dst[i];
+      struct tgsi_full_dst_register *fd = &fi->Dst[i];
 
       EOL();
       TXT( "\nFile     : " );
-      ENM( dst->DstRegister.File, TGSI_FILES );
-      if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) {
+      ENM( dst->Register.File, TGSI_FILES );
+      if( deflt || fd->Register.WriteMask != dst->Register.WriteMask ) {
          TXT( "\nWriteMask: " );
-         ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS );
+         ENM( dst->Register.WriteMask, TGSI_WRITEMASKS );
       }
       if( ignored ) {
-         if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) {
+         if( deflt || fd->Register.Indirect != dst->Register.Indirect ) {
             TXT( "\nIndirect : " );
-            UID( dst->DstRegister.Indirect );
+            UID( dst->Register.Indirect );
          }
-         if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) {
+         if( deflt || fd->Register.Dimension != dst->Register.Dimension ) {
             TXT( "\nDimension: " );
-            UID( dst->DstRegister.Dimension );
+            UID( dst->Register.Dimension );
          }
       }
-      if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) {
+      if( deflt || fd->Register.Index != dst->Register.Index ) {
          TXT( "\nIndex    : " );
-         SID( dst->DstRegister.Index );
+         SID( dst->Register.Index );
       }
       if( ignored ) {
          TXT( "\nPadding  : " );
-         UIX( dst->DstRegister.Padding );
-         if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) {
-            TXT( "\nExtended : " );
-            UID( dst->DstRegister.Extended );
-         }
-      }
-
-      if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) {
-         EOL();
-         TXT( "\nType        : " );
-         ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS );
-         if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) {
-            TXT( "\nCondMask    : " );
-            ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS );
-         }
-         if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) {
-            TXT( "\nCondSwizzleX: " );
-            ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES );
-         }
-         if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) {
-            TXT( "\nCondSwizzleY: " );
-            ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES );
-         }
-         if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) {
-            TXT( "\nCondSwizzleZ: " );
-            ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES );
-         }
-         if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) {
-            TXT( "\nCondSwizzleW: " );
-            ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES );
-         }
-         if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) {
-            TXT( "\nCondSrcIndex: " );
-            UID( dst->DstRegisterExtConcode.CondSrcIndex );
-         }
-         if( ignored ) {
-            TXT( "\nPadding     : " );
-            UIX( dst->DstRegisterExtConcode.Padding );
-            if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) {
-               TXT( "\nExtended    : " );
-               UID( dst->DstRegisterExtConcode.Extended );
-            }
-         }
-      }
-
-      if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) {
-         EOL();
-         TXT( "\nType    : " );
-         ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS );
-         if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) {
-            TXT( "\nModulate: " );
-            ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES );
-         }
-         if( ignored ) {
-            TXT( "\nPadding : " );
-            UIX( dst->DstRegisterExtModulate.Padding );
-            if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) {
-               TXT( "\nExtended: " );
-               UID( dst->DstRegisterExtModulate.Extended );
-            }
-         }
+         UIX( dst->Register.Padding );
       }
    }
 
    for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) {
-      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
-      struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i];
+      struct tgsi_full_src_register *src = &inst->Src[i];
+      struct tgsi_full_src_register *fs = &fi->Src[i];
 
       EOL();
       TXT( "\nFile     : ");
-      ENM( src->SrcRegister.File, TGSI_FILES );
-      if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) {
+      ENM( src->Register.File, TGSI_FILES );
+      if( deflt || fs->Register.SwizzleX != src->Register.SwizzleX ) {
          TXT( "\nSwizzleX : " );
-         ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES );
+         ENM( src->Register.SwizzleX, TGSI_SWIZZLES );
       }
-      if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) {
+      if( deflt || fs->Register.SwizzleY != src->Register.SwizzleY ) {
          TXT( "\nSwizzleY : " );
-         ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES );
+         ENM( src->Register.SwizzleY, TGSI_SWIZZLES );
       }
-      if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) {
+      if( deflt || fs->Register.SwizzleZ != src->Register.SwizzleZ ) {
          TXT( "\nSwizzleZ : " );
-         ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES );
+         ENM( src->Register.SwizzleZ, TGSI_SWIZZLES );
       }
-      if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) {
+      if( deflt || fs->Register.SwizzleW != src->Register.SwizzleW ) {
          TXT( "\nSwizzleW : " );
-         ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES );
+         ENM( src->Register.SwizzleW, TGSI_SWIZZLES );
       }
-      if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) {
+      if (deflt || fs->Register.Absolute != src->Register.Absolute) {
+         TXT("\nAbsolute : ");
+         UID(src->Register.Absolute);
+      }
+      if( deflt || fs->Register.Negate != src->Register.Negate ) {
          TXT( "\nNegate   : " );
-         UID( src->SrcRegister.Negate );
+         UID( src->Register.Negate );
       }
       if( ignored ) {
-         if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) {
+         if( deflt || fs->Register.Indirect != src->Register.Indirect ) {
             TXT( "\nIndirect : " );
-            UID( src->SrcRegister.Indirect );
+            UID( src->Register.Indirect );
          }
-         if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) {
+         if( deflt || fs->Register.Dimension != src->Register.Dimension ) {
             TXT( "\nDimension: " );
-            UID( src->SrcRegister.Dimension );
+            UID( src->Register.Dimension );
          }
       }
-      if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) {
+      if( deflt || fs->Register.Index != src->Register.Index ) {
          TXT( "\nIndex    : " );
-         SID( src->SrcRegister.Index );
-      }
-      if( ignored ) {
-         if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) {
-            TXT( "\nExtended : " );
-            UID( src->SrcRegister.Extended );
-         }
-      }
-
-      if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) {
-         EOL();
-         TXT( "\nType       : " );
-         ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS );
-         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) {
-            TXT( "\nExtSwizzleX: " );
-            ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) {
-            TXT( "\nExtSwizzleY: " );
-            ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) {
-            TXT( "\nExtSwizzleZ: " );
-            ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) {
-            TXT( "\nExtSwizzleW: " );
-            ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) {
-            TXT( "\nNegateX   : " );
-            UID( src->SrcRegisterExtSwz.NegateX );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) {
-            TXT( "\nNegateY   : " );
-            UID( src->SrcRegisterExtSwz.NegateY );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) {
-            TXT( "\nNegateZ   : " );
-            UID( src->SrcRegisterExtSwz.NegateZ );
-         }
-         if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) {
-            TXT( "\nNegateW   : " );
-            UID( src->SrcRegisterExtSwz.NegateW );
-         }
-         if( ignored ) {
-            TXT( "\nPadding   : " );
-            UIX( src->SrcRegisterExtSwz.Padding );
-            if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) {
-               TXT( "\nExtended   : " );
-               UID( src->SrcRegisterExtSwz.Extended );
-            }
-         }
-      }
-
-      if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) {
-         EOL();
-         TXT( "\nType     : " );
-         ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS );
-         if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) {
-            TXT( "\nComplement: " );
-            UID( src->SrcRegisterExtMod.Complement );
-         }
-         if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) {
-            TXT( "\nBias     : " );
-            UID( src->SrcRegisterExtMod.Bias );
-         }
-         if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) {
-            TXT( "\nScale2X   : " );
-            UID( src->SrcRegisterExtMod.Scale2X );
-         }
-         if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) {
-            TXT( "\nAbsolute  : " );
-            UID( src->SrcRegisterExtMod.Absolute );
-         }
-         if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) {
-            TXT( "\nNegate   : " );
-            UID( src->SrcRegisterExtMod.Negate );
-         }
-         if( ignored ) {
-            TXT( "\nPadding   : " );
-            UIX( src->SrcRegisterExtMod.Padding );
-            if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) {
-               TXT( "\nExtended  : " );
-               UID( src->SrcRegisterExtMod.Extended );
-            }
-         }
+         SID( src->Register.Index );
       }
    }
 }
@@ -653,12 +405,6 @@ tgsi_dump_c(
 
    TXT( "tgsi-dump begin -----------------" );
 
-   TXT( "\nMajorVersion: " );
-   UID( parse.FullVersion.Version.MajorVersion );
-   TXT( "\nMinorVersion: " );
-   UID( parse.FullVersion.Version.MinorVersion );
-   EOL();
-
    TXT( "\nHeaderSize: " );
    UID( parse.FullHeader.Header.HeaderSize );
    TXT( "\nBodySize  : " );
@@ -678,10 +424,6 @@ tgsi_dump_c(
       if( ignored ) {
          TXT( "\nSize       : " );
          UID( parse.FullToken.Token.NrTokens );
-         if( deflt || parse.FullToken.Token.Extended ) {
-            TXT( "\nExtended   : " );
-            UID( parse.FullToken.Token.Extended );
-         }
       }
 
       switch( parse.FullToken.Token.Type ) {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index c79c56debd..2bcb33392a 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -2,6 +2,7 @@
  * 
  * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
+ * Copyright 2009-2010 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
@@ -60,21 +61,337 @@
 #include "util/u_memory.h"
 #include "util/u_math.h"
 
-#define FAST_MATH 1
 
-/** for tgsi_full_instruction::Flags */
-#define SOA_DEPENDENCY_FLAG 0x1
+#define FAST_MATH 1
 
 #define TILE_TOP_LEFT     0
 #define TILE_TOP_RIGHT    1
 #define TILE_BOTTOM_LEFT  2
 #define TILE_BOTTOM_RIGHT 3
 
+static void
+micro_abs(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = fabsf(src->f[0]);
+   dst->f[1] = fabsf(src->f[1]);
+   dst->f[2] = fabsf(src->f[2]);
+   dst->f[3] = fabsf(src->f[3]);
+}
+
+static void
+micro_arl(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->i[0] = (int)floorf(src->f[0]);
+   dst->i[1] = (int)floorf(src->f[1]);
+   dst->i[2] = (int)floorf(src->f[2]);
+   dst->i[3] = (int)floorf(src->f[3]);
+}
+
+static void
+micro_arr(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->i[0] = (int)floorf(src->f[0] + 0.5f);
+   dst->i[1] = (int)floorf(src->f[1] + 0.5f);
+   dst->i[2] = (int)floorf(src->f[2] + 0.5f);
+   dst->i[3] = (int)floorf(src->f[3] + 0.5f);
+}
+
+static void
+micro_ceil(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->f[0] = ceilf(src->f[0]);
+   dst->f[1] = ceilf(src->f[1]);
+   dst->f[2] = ceilf(src->f[2]);
+   dst->f[3] = ceilf(src->f[3]);
+}
+
+static void
+micro_cos(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = cosf(src->f[0]);
+   dst->f[1] = cosf(src->f[1]);
+   dst->f[2] = cosf(src->f[2]);
+   dst->f[3] = cosf(src->f[3]);
+}
+
+static void
+micro_ddx(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] =
+   dst->f[1] =
+   dst->f[2] =
+   dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT];
+}
+
+static void
+micro_ddy(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] =
+   dst->f[1] =
+   dst->f[2] =
+   dst->f[3] = src->f[TILE_BOTTOM_LEFT] - src->f[TILE_TOP_LEFT];
+}
+
+static void
+micro_exp2(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+#if FAST_MATH
+   dst->f[0] = util_fast_exp2(src->f[0]);
+   dst->f[1] = util_fast_exp2(src->f[1]);
+   dst->f[2] = util_fast_exp2(src->f[2]);
+   dst->f[3] = util_fast_exp2(src->f[3]);
+#else
+#if DEBUG
+   /* Inf is okay for this instruction, so clamp it to silence assertions. */
+   uint i;
+   union tgsi_exec_channel clamped;
+
+   for (i = 0; i < 4; i++) {
+      if (src->f[i] > 127.99999f) {
+         clamped.f[i] = 127.99999f;
+      } else if (src->f[i] < -126.99999f) {
+         clamped.f[i] = -126.99999f;
+      } else {
+         clamped.f[i] = src->f[i];
+      }
+   }
+   src = &clamped;
+#endif /* DEBUG */
+
+   dst->f[0] = powf(2.0f, src->f[0]);
+   dst->f[1] = powf(2.0f, src->f[1]);
+   dst->f[2] = powf(2.0f, src->f[2]);
+   dst->f[3] = powf(2.0f, src->f[3]);
+#endif /* FAST_MATH */
+}
+
+static void
+micro_flr(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = floorf(src->f[0]);
+   dst->f[1] = floorf(src->f[1]);
+   dst->f[2] = floorf(src->f[2]);
+   dst->f[3] = floorf(src->f[3]);
+}
+
+static void
+micro_frc(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src->f[0] - floorf(src->f[0]);
+   dst->f[1] = src->f[1] - floorf(src->f[1]);
+   dst->f[2] = src->f[2] - floorf(src->f[2]);
+   dst->f[3] = src->f[3] - floorf(src->f[3]);
+}
+
+static void
+micro_iabs(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src->i[0] >= 0 ? src->i[0] : -src->i[0];
+   dst->i[1] = src->i[1] >= 0 ? src->i[1] : -src->i[1];
+   dst->i[2] = src->i[2] >= 0 ? src->i[2] : -src->i[2];
+   dst->i[3] = src->i[3] >= 0 ? src->i[3] : -src->i[3];
+}
+
+static void
+micro_ineg(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = -src->i[0];
+   dst->i[1] = -src->i[1];
+   dst->i[2] = -src->i[2];
+   dst->i[3] = -src->i[3];
+}
+
+static void
+micro_lg2(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+#if FAST_MATH
+   dst->f[0] = util_fast_log2(src->f[0]);
+   dst->f[1] = util_fast_log2(src->f[1]);
+   dst->f[2] = util_fast_log2(src->f[2]);
+   dst->f[3] = util_fast_log2(src->f[3]);
+#else
+   dst->f[0] = logf(src->f[0]) * 1.442695f;
+   dst->f[1] = logf(src->f[1]) * 1.442695f;
+   dst->f[2] = logf(src->f[2]) * 1.442695f;
+   dst->f[3] = logf(src->f[3]) * 1.442695f;
+#endif
+}
+
+static void
+micro_lrp(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] * (src[1].f[0] - src[2].f[0]) + src[2].f[0];
+   dst->f[1] = src[0].f[1] * (src[1].f[1] - src[2].f[1]) + src[2].f[1];
+   dst->f[2] = src[0].f[2] * (src[1].f[2] - src[2].f[2]) + src[2].f[2];
+   dst->f[3] = src[0].f[3] * (src[1].f[3] - src[2].f[3]) + src[2].f[3];
+}
+
+static void
+micro_mad(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] * src[1].f[0] + src[2].f[0];
+   dst->f[1] = src[0].f[1] * src[1].f[1] + src[2].f[1];
+   dst->f[2] = src[0].f[2] * src[1].f[2] + src[2].f[2];
+   dst->f[3] = src[0].f[3] * src[1].f[3] + src[2].f[3];
+}
+
+static void
+micro_mov(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src->u[0];
+   dst->u[1] = src->u[1];
+   dst->u[2] = src->u[2];
+   dst->u[3] = src->u[3];
+}
+
+static void
+micro_rcp(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = 1.0f / src->f[0];
+   dst->f[1] = 1.0f / src->f[1];
+   dst->f[2] = 1.0f / src->f[2];
+   dst->f[3] = 1.0f / src->f[3];
+}
+
+static void
+micro_rnd(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = floorf(src->f[0] + 0.5f);
+   dst->f[1] = floorf(src->f[1] + 0.5f);
+   dst->f[2] = floorf(src->f[2] + 0.5f);
+   dst->f[3] = floorf(src->f[3] + 0.5f);
+}
+
+static void
+micro_rsq(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = 1.0f / sqrtf(fabsf(src->f[0]));
+   dst->f[1] = 1.0f / sqrtf(fabsf(src->f[1]));
+   dst->f[2] = 1.0f / sqrtf(fabsf(src->f[2]));
+   dst->f[3] = 1.0f / sqrtf(fabsf(src->f[3]));
+}
+
+static void
+micro_seq(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] == src[1].f[0] ? 1.0f : 0.0f;
+   dst->f[1] = src[0].f[1] == src[1].f[1] ? 1.0f : 0.0f;
+   dst->f[2] = src[0].f[2] == src[1].f[2] ? 1.0f : 0.0f;
+   dst->f[3] = src[0].f[3] == src[1].f[3] ? 1.0f : 0.0f;
+}
+
+static void
+micro_sge(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] >= src[1].f[0] ? 1.0f : 0.0f;
+   dst->f[1] = src[0].f[1] >= src[1].f[1] ? 1.0f : 0.0f;
+   dst->f[2] = src[0].f[2] >= src[1].f[2] ? 1.0f : 0.0f;
+   dst->f[3] = src[0].f[3] >= src[1].f[3] ? 1.0f : 0.0f;
+}
+
+static void
+micro_sgn(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src->f[0] < 0.0f ? -1.0f : src->f[0] > 0.0f ? 1.0f : 0.0f;
+   dst->f[1] = src->f[1] < 0.0f ? -1.0f : src->f[1] > 0.0f ? 1.0f : 0.0f;
+   dst->f[2] = src->f[2] < 0.0f ? -1.0f : src->f[2] > 0.0f ? 1.0f : 0.0f;
+   dst->f[3] = src->f[3] < 0.0f ? -1.0f : src->f[3] > 0.0f ? 1.0f : 0.0f;
+}
+
+static void
+micro_sgt(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] > src[1].f[0] ? 1.0f : 0.0f;
+   dst->f[1] = src[0].f[1] > src[1].f[1] ? 1.0f : 0.0f;
+   dst->f[2] = src[0].f[2] > src[1].f[2] ? 1.0f : 0.0f;
+   dst->f[3] = src[0].f[3] > src[1].f[3] ? 1.0f : 0.0f;
+}
+
+static void
+micro_sin(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = sinf(src->f[0]);
+   dst->f[1] = sinf(src->f[1]);
+   dst->f[2] = sinf(src->f[2]);
+   dst->f[3] = sinf(src->f[3]);
+}
+
+static void
+micro_sle(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] <= src[1].f[0] ? 1.0f : 0.0f;
+   dst->f[1] = src[0].f[1] <= src[1].f[1] ? 1.0f : 0.0f;
+   dst->f[2] = src[0].f[2] <= src[1].f[2] ? 1.0f : 0.0f;
+   dst->f[3] = src[0].f[3] <= src[1].f[3] ? 1.0f : 0.0f;
+}
+
+static void
+micro_slt(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] < src[1].f[0] ? 1.0f : 0.0f;
+   dst->f[1] = src[0].f[1] < src[1].f[1] ? 1.0f : 0.0f;
+   dst->f[2] = src[0].f[2] < src[1].f[2] ? 1.0f : 0.0f;
+   dst->f[3] = src[0].f[3] < src[1].f[3] ? 1.0f : 0.0f;
+}
+
+static void
+micro_sne(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = src[0].f[0] != src[1].f[0] ? 1.0f : 0.0f;
+   dst->f[1] = src[0].f[1] != src[1].f[1] ? 1.0f : 0.0f;
+   dst->f[2] = src[0].f[2] != src[1].f[2] ? 1.0f : 0.0f;
+   dst->f[3] = src[0].f[3] != src[1].f[3] ? 1.0f : 0.0f;
+}
+
+static void
+micro_trunc(union tgsi_exec_channel *dst,
+            const union tgsi_exec_channel *src)
+{
+   dst->f[0] = (float)(int)src->f[0];
+   dst->f[1] = (float)(int)src->f[1];
+   dst->f[2] = (float)(int)src->f[2];
+   dst->f[3] = (float)(int)src->f[3];
+}
+
+
 #define CHAN_X  0
 #define CHAN_Y  1
 #define CHAN_Z  2
 #define CHAN_W  3
 
+enum tgsi_exec_datatype {
+   TGSI_EXEC_DATA_FLOAT,
+   TGSI_EXEC_DATA_INT,
+   TGSI_EXEC_DATA_UINT
+};
+
 /*
  * Shorthand locations of various utility registers (_I = Index, _C = Channel)
  */
@@ -107,12 +424,13 @@
 #define TEMP_HALF_I        TGSI_EXEC_TEMP_HALF_I
 #define TEMP_HALF_C        TGSI_EXEC_TEMP_HALF_C
 #define TEMP_R0            TGSI_EXEC_TEMP_R0
+#define TEMP_P0            TGSI_EXEC_TEMP_P0
 
 #define IS_CHANNEL_ENABLED(INST, CHAN)\
-   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+   ((INST).Dst[0].Register.WriteMask & (1 << (CHAN)))
 
 #define IS_CHANNEL_ENABLED2(INST, CHAN)\
-   ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN)))
+   ((INST).Dst[1].Register.WriteMask & (1 << (CHAN)))
 
 #define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\
    for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\
@@ -125,23 +443,19 @@
 
 /** The execution mask depends on the conditional mask and the loop mask */
 #define UPDATE_EXEC_MASK(MACH) \
-      MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask
+      MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->Switch.mask & MACH->FuncMask
 
 
 static const union tgsi_exec_channel ZeroVec =
    { { 0.0, 0.0, 0.0, 0.0 } };
 
 
-#ifdef DEBUG
-static void
-check_inf_or_nan(const union tgsi_exec_channel *chan)
-{
-   assert(!util_is_inf_or_nan(chan->f[0]));
-   assert(!util_is_inf_or_nan(chan->f[1]));
-   assert(!util_is_inf_or_nan(chan->f[2]));
-   assert(!util_is_inf_or_nan(chan->f[3]));
-}
-#endif
+#define CHECK_INF_OR_NAN(chan) do {\
+      assert(!util_is_inf_or_nan((chan)->f[0]));\
+      assert(!util_is_inf_or_nan((chan)->f[1]));\
+      assert(!util_is_inf_or_nan((chan)->f[2]));\
+      assert(!util_is_inf_or_nan((chan)->f[3]));\
+   } while (0)
 
 
 #ifdef DEBUG
@@ -190,7 +504,7 @@ tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst)
 {
    uint i, chan;
 
-   uint writemask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   uint writemask = inst->Dst[0].Register.WriteMask;
    if (writemask == TGSI_WRITEMASK_X ||
        writemask == TGSI_WRITEMASK_Y ||
        writemask == TGSI_WRITEMASK_Z ||
@@ -202,17 +516,16 @@ tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst)
 
    /* loop over src regs */
    for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
-      if ((inst->FullSrcRegisters[i].SrcRegister.File ==
-           inst->FullDstRegisters[0].DstRegister.File) &&
-          (inst->FullSrcRegisters[i].SrcRegister.Index ==
-           inst->FullDstRegisters[0].DstRegister.Index)) {
+      if ((inst->Src[i].Register.File ==
+           inst->Dst[0].Register.File) &&
+          (inst->Src[i].Register.Index ==
+           inst->Dst[0].Register.Index)) {
          /* loop over dest channels */
          uint channelsWritten = 0x0;
          FOR_EACH_ENABLED_CHANNEL(*inst, chan) {
             /* check if we're reading a channel that's been written */
-            uint swizzle = tgsi_util_get_full_src_register_extswizzle(&inst->FullSrcRegisters[i], chan);
-            if (swizzle <= TGSI_SWIZZLE_W &&
-                (channelsWritten & (1 << swizzle))) {
+            uint swizzle = tgsi_util_get_full_src_register_swizzle(&inst->Src[i], chan);
+            if (channelsWritten & (1 << swizzle)) {
                return TRUE;
             }
 
@@ -295,6 +608,14 @@ tgsi_exec_machine_bind_shader(
                                    * sizeof(struct tgsi_full_declaration));
             maxDeclarations += 10;
          }
+         if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_OUTPUT) {
+            unsigned reg;
+            for (reg = parse.FullToken.FullDeclaration.Range.First;
+                 reg <= parse.FullToken.FullDeclaration.Range.Last;
+                 ++reg) {
+               ++mach->NumOutputs;
+            }
+         }
          memcpy(declarations + numDeclarations,
                 &parse.FullToken.FullDeclaration,
                 sizeof(declarations[0]));
@@ -332,20 +653,6 @@ tgsi_exec_machine_bind_shader(
             maxInstructions += 10;
          }
 
-         if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) {
-            uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode;
-            parse.FullToken.FullInstruction.Flags = SOA_DEPENDENCY_FLAG;
-            /* XXX we only handle SOA dependencies properly for MOV/SWZ
-             * at this time!
-             */
-            if (opcode != TGSI_OPCODE_MOV && opcode != TGSI_OPCODE_SWZ) {
-               debug_printf("Warning: SOA dependency in instruction"
-                            " is not handled:\n");
-               tgsi_dump_instruction(&parse.FullToken.FullInstruction,
-                                     numInstructions);
-            }
-         }
-
          memcpy(instructions + numInstructions,
                 &parse.FullToken.FullInstruction,
                 sizeof(instructions[0]));
@@ -353,6 +660,9 @@ tgsi_exec_machine_bind_shader(
          numInstructions++;
          break;
 
+      case TGSI_TOKEN_TYPE_PROPERTY:
+         break;
+
       default:
          assert( 0 );
       }
@@ -386,6 +696,8 @@ tgsi_exec_machine_create( void )
    memset(mach, 0, sizeof(*mach));
 
    mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR];
+   mach->MaxGeometryShaderOutputs = TGSI_MAX_TOTAL_VERTICES;
+   mach->Predicates = &mach->Temps[TGSI_EXEC_TEMP_P0];
 
    /* Setup constants. */
    for( i = 0; i < 4; i++ ) {
@@ -426,18 +738,6 @@ tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach)
    align_free(mach);
 }
 
-
-static void
-micro_abs(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = fabsf( src->f[0] );
-   dst->f[1] = fabsf( src->f[1] );
-   dst->f[2] = fabsf( src->f[2] );
-   dst->f[3] = fabsf( src->f[3] );
-}
-
 static void
 micro_add(
    union tgsi_exec_channel *dst,
@@ -450,76 +750,6 @@ micro_add(
    dst->f[3] = src0->f[3] + src1->f[3];
 }
 
-#if 0
-static void
-micro_iadd(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->i[0] = src0->i[0] + src1->i[0];
-   dst->i[1] = src0->i[1] + src1->i[1];
-   dst->i[2] = src0->i[2] + src1->i[2];
-   dst->i[3] = src0->i[3] + src1->i[3];
-}
-#endif
-
-static void
-micro_and(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] & src1->u[0];
-   dst->u[1] = src0->u[1] & src1->u[1];
-   dst->u[2] = src0->u[2] & src1->u[2];
-   dst->u[3] = src0->u[3] & src1->u[3];
-}
-
-static void
-micro_ceil(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = ceilf( src->f[0] );
-   dst->f[1] = ceilf( src->f[1] );
-   dst->f[2] = ceilf( src->f[2] );
-   dst->f[3] = ceilf( src->f[3] );
-}
-
-static void
-micro_cos(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = cosf( src->f[0] );
-   dst->f[1] = cosf( src->f[1] );
-   dst->f[2] = cosf( src->f[2] );
-   dst->f[3] = cosf( src->f[3] );
-}
-
-static void
-micro_ddx(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] =
-   dst->f[1] =
-   dst->f[2] =
-   dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT];
-}
-
-static void
-micro_ddy(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] =
-   dst->f[1] =
-   dst->f[2] =
-   dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT];
-}
-
 static void
 micro_div(
    union tgsi_exec_channel *dst,
@@ -540,81 +770,6 @@ micro_div(
    }
 }
 
-#if 0
-static void
-micro_udiv(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] / src1->u[0];
-   dst->u[1] = src0->u[1] / src1->u[1];
-   dst->u[2] = src0->u[2] / src1->u[2];
-   dst->u[3] = src0->u[3] / src1->u[3];
-}
-#endif
-
-static void
-micro_eq(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1,
-   const union tgsi_exec_channel *src2,
-   const union tgsi_exec_channel *src3 )
-{
-   dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0];
-   dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1];
-   dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2];
-   dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3];
-}
-
-#if 0
-static void
-micro_ieq(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1,
-   const union tgsi_exec_channel *src2,
-   const union tgsi_exec_channel *src3 )
-{
-   dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0];
-   dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1];
-   dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2];
-   dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3];
-}
-#endif
-
-static void
-micro_exp2(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src)
-{
-#if FAST_MATH
-   dst->f[0] = util_fast_exp2( src->f[0] );
-   dst->f[1] = util_fast_exp2( src->f[1] );
-   dst->f[2] = util_fast_exp2( src->f[2] );
-   dst->f[3] = util_fast_exp2( src->f[3] );
-#else
-   dst->f[0] = powf( 2.0f, src->f[0] );
-   dst->f[1] = powf( 2.0f, src->f[1] );
-   dst->f[2] = powf( 2.0f, src->f[2] );
-   dst->f[3] = powf( 2.0f, src->f[3] );
-#endif
-}
-
-#if 0
-static void
-micro_f2ut(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->u[0] = (uint) src->f[0];
-   dst->u[1] = (uint) src->f[1];
-   dst->u[2] = (uint) src->f[2];
-   dst->u[3] = (uint) src->f[3];
-}
-#endif
-
 static void
 micro_float_clamp(union tgsi_exec_channel *dst,
                   const union tgsi_exec_channel *src)
@@ -642,71 +797,6 @@ micro_float_clamp(union tgsi_exec_channel *dst,
 }
 
 static void
-micro_flr(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = floorf( src->f[0] );
-   dst->f[1] = floorf( src->f[1] );
-   dst->f[2] = floorf( src->f[2] );
-   dst->f[3] = floorf( src->f[3] );
-}
-
-static void
-micro_frc(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = src->f[0] - floorf( src->f[0] );
-   dst->f[1] = src->f[1] - floorf( src->f[1] );
-   dst->f[2] = src->f[2] - floorf( src->f[2] );
-   dst->f[3] = src->f[3] - floorf( src->f[3] );
-}
-
-static void
-micro_i2f(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = (float) src->i[0];
-   dst->f[1] = (float) src->i[1];
-   dst->f[2] = (float) src->i[2];
-   dst->f[3] = (float) src->i[3];
-}
-
-static void
-micro_lg2(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-#if FAST_MATH
-   dst->f[0] = util_fast_log2( src->f[0] );
-   dst->f[1] = util_fast_log2( src->f[1] );
-   dst->f[2] = util_fast_log2( src->f[2] );
-   dst->f[3] = util_fast_log2( src->f[3] );
-#else
-   dst->f[0] = logf( src->f[0] ) * 1.442695f;
-   dst->f[1] = logf( src->f[1] ) * 1.442695f;
-   dst->f[2] = logf( src->f[2] ) * 1.442695f;
-   dst->f[3] = logf( src->f[3] ) * 1.442695f;
-#endif
-}
-
-static void
-micro_le(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1,
-   const union tgsi_exec_channel *src2,
-   const union tgsi_exec_channel *src3 )
-{
-   dst->f[0] = src0->f[0] <= src1->f[0] ? src2->f[0] : src3->f[0];
-   dst->f[1] = src0->f[1] <= src1->f[1] ? src2->f[1] : src3->f[1];
-   dst->f[2] = src0->f[2] <= src1->f[2] ? src2->f[2] : src3->f[2];
-   dst->f[3] = src0->f[3] <= src1->f[3] ? src2->f[3] : src3->f[3];
-}
-
-static void
 micro_lt(
    union tgsi_exec_channel *dst,
    const union tgsi_exec_channel *src0,
@@ -720,38 +810,6 @@ micro_lt(
    dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3];
 }
 
-#if 0
-static void
-micro_ilt(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1,
-   const union tgsi_exec_channel *src2,
-   const union tgsi_exec_channel *src3 )
-{
-   dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0];
-   dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1];
-   dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2];
-   dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3];
-}
-#endif
-
-#if 0
-static void
-micro_ult(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1,
-   const union tgsi_exec_channel *src2,
-   const union tgsi_exec_channel *src3 )
-{
-   dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0];
-   dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1];
-   dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2];
-   dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3];
-}
-#endif
-
 static void
 micro_max(
    union tgsi_exec_channel *dst,
@@ -764,34 +822,6 @@ micro_max(
    dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3];
 }
 
-#if 0
-static void
-micro_imax(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0];
-   dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1];
-   dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2];
-   dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3];
-}
-#endif
-
-#if 0
-static void
-micro_umax(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0];
-   dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1];
-   dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2];
-   dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3];
-}
-#endif
-
 static void
 micro_min(
    union tgsi_exec_channel *dst,
@@ -804,48 +834,6 @@ micro_min(
    dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3];
 }
 
-#if 0
-static void
-micro_imin(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0];
-   dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1];
-   dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2];
-   dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3];
-}
-#endif
-
-#if 0
-static void
-micro_umin(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0];
-   dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1];
-   dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2];
-   dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3];
-}
-#endif
-
-#if 0
-static void
-micro_umod(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] % src1->u[0];
-   dst->u[1] = src0->u[1] % src1->u[1];
-   dst->u[2] = src0->u[2] % src1->u[2];
-   dst->u[3] = src0->u[3] % src1->u[3];
-}
-#endif
-
 static void
 micro_mul(
    union tgsi_exec_channel *dst,
@@ -860,20 +848,6 @@ micro_mul(
 
 #if 0
 static void
-micro_imul(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->i[0] = src0->i[0] * src1->i[0];
-   dst->i[1] = src0->i[1] * src1->i[1];
-   dst->i[2] = src0->i[2] * src1->i[2];
-   dst->i[3] = src0->i[3] * src1->i[3];
-}
-#endif
-
-#if 0
-static void
 micro_imul64(
    union tgsi_exec_channel *dst0,
    union tgsi_exec_channel *dst1,
@@ -937,42 +911,6 @@ micro_neg(
    dst->f[3] = -src->f[3];
 }
 
-#if 0
-static void
-micro_ineg(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->i[0] = -src->i[0];
-   dst->i[1] = -src->i[1];
-   dst->i[2] = -src->i[2];
-   dst->i[3] = -src->i[3];
-}
-#endif
-
-static void
-micro_not(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->u[0] = ~src->u[0];
-   dst->u[1] = ~src->u[1];
-   dst->u[2] = ~src->u[2];
-   dst->u[3] = ~src->u[3];
-}
-
-static void
-micro_or(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] | src1->u[0];
-   dst->u[1] = src0->u[1] | src1->u[1];
-   dst->u[2] = src0->u[2] | src1->u[2];
-   dst->u[3] = src0->u[3] | src1->u[3];
-}
-
 static void
 micro_pow(
    union tgsi_exec_channel *dst,
@@ -993,88 +931,6 @@ micro_pow(
 }
 
 static void
-micro_rnd(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = floorf( src->f[0] + 0.5f );
-   dst->f[1] = floorf( src->f[1] + 0.5f );
-   dst->f[2] = floorf( src->f[2] + 0.5f );
-   dst->f[3] = floorf( src->f[3] + 0.5f );
-}
-
-static void
-micro_sgn(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = src->f[0] < 0.0f ? -1.0f : src->f[0] > 0.0f ? 1.0f : 0.0f;
-   dst->f[1] = src->f[1] < 0.0f ? -1.0f : src->f[1] > 0.0f ? 1.0f : 0.0f;
-   dst->f[2] = src->f[2] < 0.0f ? -1.0f : src->f[2] > 0.0f ? 1.0f : 0.0f;
-   dst->f[3] = src->f[3] < 0.0f ? -1.0f : src->f[3] > 0.0f ? 1.0f : 0.0f;
-}
-
-static void
-micro_shl(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->i[0] = src0->i[0] << src1->i[0];
-   dst->i[1] = src0->i[1] << src1->i[1];
-   dst->i[2] = src0->i[2] << src1->i[2];
-   dst->i[3] = src0->i[3] << src1->i[3];
-}
-
-static void
-micro_ishr(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->i[0] = src0->i[0] >> src1->i[0];
-   dst->i[1] = src0->i[1] >> src1->i[1];
-   dst->i[2] = src0->i[2] >> src1->i[2];
-   dst->i[3] = src0->i[3] >> src1->i[3];
-}
-
-static void
-micro_trunc(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0 )
-{
-   dst->f[0] = (float) (int) src0->f[0];
-   dst->f[1] = (float) (int) src0->f[1];
-   dst->f[2] = (float) (int) src0->f[2];
-   dst->f[3] = (float) (int) src0->f[3];
-}
-
-#if 0
-static void
-micro_ushr(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] >> src1->u[0];
-   dst->u[1] = src0->u[1] >> src1->u[1];
-   dst->u[2] = src0->u[2] >> src1->u[2];
-   dst->u[3] = src0->u[3] >> src1->u[3];
-}
-#endif
-
-static void
-micro_sin(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = sinf( src->f[0] );
-   dst->f[1] = sinf( src->f[1] );
-   dst->f[2] = sinf( src->f[2] );
-   dst->f[3] = sinf( src->f[3] );
-}
-
-static void
 micro_sqrt( union tgsi_exec_channel *dst,
             const union tgsi_exec_channel *src )
 {
@@ -1096,31 +952,6 @@ micro_sub(
    dst->f[3] = src0->f[3] - src1->f[3];
 }
 
-#if 0
-static void
-micro_u2f(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src )
-{
-   dst->f[0] = (float) src->u[0];
-   dst->f[1] = (float) src->u[1];
-   dst->f[2] = (float) src->u[2];
-   dst->f[3] = (float) src->u[3];
-}
-#endif
-
-static void
-micro_xor(
-   union tgsi_exec_channel *dst,
-   const union tgsi_exec_channel *src0,
-   const union tgsi_exec_channel *src1 )
-{
-   dst->u[0] = src0->u[0] ^ src1->u[0];
-   dst->u[1] = src0->u[1] ^ src1->u[1];
-   dst->u[2] = src0->u[2] ^ src1->u[2];
-   dst->u[3] = src0->u[3] ^ src1->u[3];
-}
-
 static void
 fetch_src_file_channel(
    const struct tgsi_exec_machine *mach,
@@ -1130,10 +961,10 @@ fetch_src_file_channel(
    union tgsi_exec_channel *chan )
 {
    switch( swizzle ) {
-   case TGSI_EXTSWIZZLE_X:
-   case TGSI_EXTSWIZZLE_Y:
-   case TGSI_EXTSWIZZLE_Z:
-   case TGSI_EXTSWIZZLE_W:
+   case TGSI_SWIZZLE_X:
+   case TGSI_SWIZZLE_Y:
+   case TGSI_SWIZZLE_Z:
+   case TGSI_SWIZZLE_W:
       switch( file ) {
       case TGSI_FILE_CONSTANT:
          assert(mach->Consts);
@@ -1156,6 +987,7 @@ fetch_src_file_channel(
          break;
 
       case TGSI_FILE_INPUT:
+      case TGSI_FILE_SYSTEM_VALUE:
          chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0];
          chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1];
          chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2];
@@ -1188,6 +1020,17 @@ fetch_src_file_channel(
          chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3];
          break;
 
+      case TGSI_FILE_PREDICATE:
+         assert(index->i[0] < TGSI_EXEC_NUM_PREDS);
+         assert(index->i[1] < TGSI_EXEC_NUM_PREDS);
+         assert(index->i[2] < TGSI_EXEC_NUM_PREDS);
+         assert(index->i[3] < TGSI_EXEC_NUM_PREDS);
+         chan->u[0] = mach->Predicates[0].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Predicates[0].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Predicates[0].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Predicates[0].xyzw[swizzle].u[3];
+         break;
+
       case TGSI_FILE_OUTPUT:
          /* vertex/fragment output vars can be read too */
          chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0];
@@ -1201,25 +1044,17 @@ fetch_src_file_channel(
       }
       break;
 
-   case TGSI_EXTSWIZZLE_ZERO:
-      *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C];
-      break;
-
-   case TGSI_EXTSWIZZLE_ONE:
-      *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C];
-      break;
-
    default:
       assert( 0 );
    }
 }
 
 static void
-fetch_source(
-   const struct tgsi_exec_machine *mach,
-   union tgsi_exec_channel *chan,
-   const struct tgsi_full_src_register *reg,
-   const uint chan_index )
+fetch_source(const struct tgsi_exec_machine *mach,
+             union tgsi_exec_channel *chan,
+             const struct tgsi_full_src_register *reg,
+             const uint chan_index,
+             enum tgsi_exec_datatype src_datatype)
 {
    union tgsi_exec_channel index;
    uint swizzle;
@@ -1228,13 +1063,13 @@ fetch_source(
     *
     *    file[1],
     *    where:
-    *       file = SrcRegister.File
-    *       [1] = SrcRegister.Index
+    *       file = Register.File
+    *       [1] = Register.Index
     */
    index.i[0] =
    index.i[1] =
    index.i[2] =
-   index.i[3] = reg->SrcRegister.Index;
+   index.i[3] = reg->Register.Index;
 
    /* There is an extra source register that indirectly subscripts
     * a register file. The direct index now becomes an offset
@@ -1242,11 +1077,11 @@ fetch_source(
     *
     *    file[ind[2].x+1],
     *    where:
-    *       ind = SrcRegisterInd.File
-    *       [2] = SrcRegisterInd.Index
-    *       .x = SrcRegisterInd.SwizzleX
+    *       ind = Indirect.File
+    *       [2] = Indirect.Index
+    *       .x = Indirect.SwizzleX
     */
-   if (reg->SrcRegister.Indirect) {
+   if (reg->Register.Indirect) {
       union tgsi_exec_channel index2;
       union tgsi_exec_channel indir_index;
       const uint execmask = mach->ExecMask;
@@ -1256,22 +1091,22 @@ fetch_source(
       index2.i[0] =
       index2.i[1] =
       index2.i[2] =
-      index2.i[3] = reg->SrcRegisterInd.Index;
+      index2.i[3] = reg->Indirect.Index;
 
       /* get current value of address register[swizzle] */
-      swizzle = tgsi_util_get_src_register_swizzle( &reg->SrcRegisterInd, CHAN_X );
+      swizzle = tgsi_util_get_src_register_swizzle( &reg->Indirect, CHAN_X );
       fetch_src_file_channel(
          mach,
-         reg->SrcRegisterInd.File,
+         reg->Indirect.File,
          swizzle,
          &index2,
          &indir_index );
 
       /* add value of address register to the offset */
-      index.i[0] += (int) indir_index.f[0];
-      index.i[1] += (int) indir_index.f[1];
-      index.i[2] += (int) indir_index.f[2];
-      index.i[3] += (int) indir_index.f[3];
+      index.i[0] += indir_index.i[0];
+      index.i[1] += indir_index.i[1];
+      index.i[2] += indir_index.i[2];
+      index.i[3] += indir_index.i[3];
 
       /* for disabled execution channels, zero-out the index to
        * avoid using a potential garbage value.
@@ -1288,15 +1123,16 @@ fetch_source(
     *
     *    file[1][3] == file[1*sizeof(file[1])+3],
     *    where:
-    *       [3] = SrcRegisterDim.Index
+    *       [3] = Dimension.Index
     */
-   if (reg->SrcRegister.Dimension) {
+   if (reg->Register.Dimension) {
       /* The size of the first-order array depends on the register file type.
        * We need to multiply the index to the first array to get an effective,
        * "flat" index that points to the beginning of the second-order array.
        */
-      switch (reg->SrcRegister.File) {
+      switch (reg->Register.File) {
       case TGSI_FILE_INPUT:
+      case TGSI_FILE_SYSTEM_VALUE:
          index.i[0] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
          index.i[1] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
          index.i[2] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
@@ -1312,10 +1148,10 @@ fetch_source(
          assert( 0 );
       }
 
-      index.i[0] += reg->SrcRegisterDim.Index;
-      index.i[1] += reg->SrcRegisterDim.Index;
-      index.i[2] += reg->SrcRegisterDim.Index;
-      index.i[3] += reg->SrcRegisterDim.Index;
+      index.i[0] += reg->Dimension.Index;
+      index.i[1] += reg->Dimension.Index;
+      index.i[2] += reg->Dimension.Index;
+      index.i[3] += reg->Dimension.Index;
 
       /* Again, the second subscript index can be addressed indirectly
        * identically to the first one.
@@ -1324,11 +1160,11 @@ fetch_source(
        *
        *    file[1][ind[4].y+3],
        *    where:
-       *       ind = SrcRegisterDimInd.File
-       *       [4] = SrcRegisterDimInd.Index
-       *       .y = SrcRegisterDimInd.SwizzleX
+       *       ind = DimIndirect.File
+       *       [4] = DimIndirect.Index
+       *       .y = DimIndirect.SwizzleX
        */
-      if (reg->SrcRegisterDim.Indirect) {
+      if (reg->Dimension.Indirect) {
          union tgsi_exec_channel index2;
          union tgsi_exec_channel indir_index;
          const uint execmask = mach->ExecMask;
@@ -1337,20 +1173,20 @@ fetch_source(
          index2.i[0] =
          index2.i[1] =
          index2.i[2] =
-         index2.i[3] = reg->SrcRegisterDimInd.Index;
+         index2.i[3] = reg->DimIndirect.Index;
 
-         swizzle = tgsi_util_get_src_register_swizzle( &reg->SrcRegisterDimInd, CHAN_X );
+         swizzle = tgsi_util_get_src_register_swizzle( &reg->DimIndirect, CHAN_X );
          fetch_src_file_channel(
             mach,
-            reg->SrcRegisterDimInd.File,
+            reg->DimIndirect.File,
             swizzle,
             &index2,
             &indir_index );
 
-         index.i[0] += (int) indir_index.f[0];
-         index.i[1] += (int) indir_index.f[1];
-         index.i[2] += (int) indir_index.f[2];
-         index.i[3] += (int) indir_index.f[3];
+         index.i[0] += indir_index.i[0];
+         index.i[1] += indir_index.i[1];
+         index.i[2] += indir_index.i[2];
+         index.i[3] += indir_index.i[3];
 
          /* for disabled execution channels, zero-out the index to
           * avoid using a potential garbage value.
@@ -1362,49 +1198,43 @@ fetch_source(
       }
 
       /* If by any chance there was a need for a 3D array of register
-       * files, we would have to check whether SrcRegisterDim is followed
+       * files, we would have to check whether Dimension is followed
        * by a dimension register and continue the saga.
        */
    }
 
-   swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+   swizzle = tgsi_util_get_full_src_register_swizzle( reg, chan_index );
    fetch_src_file_channel(
       mach,
-      reg->SrcRegister.File,
+      reg->Register.File,
       swizzle,
       &index,
       chan );
 
-   switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) {
-   case TGSI_UTIL_SIGN_CLEAR:
-      micro_abs( chan, chan );
-      break;
-
-   case TGSI_UTIL_SIGN_SET:
-      micro_abs( chan, chan );
-      micro_neg( chan, chan );
-      break;
-
-   case TGSI_UTIL_SIGN_TOGGLE:
-      micro_neg( chan, chan );
-      break;
-
-   case TGSI_UTIL_SIGN_KEEP:
-      break;
+   if (reg->Register.Absolute) {
+      if (src_datatype == TGSI_EXEC_DATA_FLOAT) {
+         micro_abs(chan, chan);
+      } else {
+         micro_iabs(chan, chan);
+      }
    }
 
-   if (reg->SrcRegisterExtMod.Complement) {
-      micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan );
+   if (reg->Register.Negate) {
+      if (src_datatype == TGSI_EXEC_DATA_FLOAT) {
+         micro_neg(chan, chan);
+      } else {
+         micro_ineg(chan, chan);
+      }
    }
 }
 
 static void
-store_dest(
-   struct tgsi_exec_machine *mach,
-   const union tgsi_exec_channel *chan,
-   const struct tgsi_full_dst_register *reg,
-   const struct tgsi_full_instruction *inst,
-   uint chan_index )
+store_dest(struct tgsi_exec_machine *mach,
+           const union tgsi_exec_channel *chan,
+           const struct tgsi_full_dst_register *reg,
+           const struct tgsi_full_instruction *inst,
+           uint chan_index,
+           enum tgsi_exec_datatype dst_datatype)
 {
    uint i;
    union tgsi_exec_channel null;
@@ -1413,9 +1243,9 @@ store_dest(
    int offset = 0;  /* indirection offset */
    int index;
 
-#ifdef DEBUG
-   check_inf_or_nan(chan);
-#endif
+   if (dst_datatype == TGSI_EXEC_DATA_FLOAT) {
+      CHECK_INF_OR_NAN(chan);
+   }
 
    /* There is an extra source register that indirectly subscripts
     * a register file. The direct index now becomes an offset
@@ -1423,11 +1253,11 @@ store_dest(
     *
     *    file[ind[2].x+1],
     *    where:
-    *       ind = DstRegisterInd.File
-    *       [2] = DstRegisterInd.Index
-    *       .x = DstRegisterInd.SwizzleX
+    *       ind = Indirect.File
+    *       [2] = Indirect.Index
+    *       .x = Indirect.SwizzleX
     */
-   if (reg->DstRegister.Indirect) {
+   if (reg->Register.Indirect) {
       union tgsi_exec_channel index;
       union tgsi_exec_channel indir_index;
       uint swizzle;
@@ -1436,155 +1266,110 @@ store_dest(
       index.i[0] =
       index.i[1] =
       index.i[2] =
-      index.i[3] = reg->DstRegisterInd.Index;
+      index.i[3] = reg->Indirect.Index;
 
       /* get current value of address register[swizzle] */
-      swizzle = tgsi_util_get_src_register_swizzle( &reg->DstRegisterInd, CHAN_X );
+      swizzle = tgsi_util_get_src_register_swizzle( &reg->Indirect, CHAN_X );
 
       /* fetch values from the address/indirection register */
       fetch_src_file_channel(
          mach,
-         reg->DstRegisterInd.File,
+         reg->Indirect.File,
          swizzle,
          &index,
          &indir_index );
 
       /* save indirection offset */
-      offset = (int) indir_index.f[0];
+      offset = indir_index.i[0];
    }
 
-   switch (reg->DstRegister.File) {
+   switch (reg->Register.File) {
    case TGSI_FILE_NULL:
       dst = &null;
       break;
 
    case TGSI_FILE_OUTPUT:
       index = mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0]
-         + reg->DstRegister.Index;
+         + reg->Register.Index;
       dst = &mach->Outputs[offset + index].xyzw[chan_index];
+#if 0
+      if (TGSI_PROCESSOR_GEOMETRY == mach->Processor) {
+         fprintf(stderr, "STORING OUT[%d] mask(%d), = (", offset + index, execmask);
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (execmask & (1 << i))
+               fprintf(stderr, "%f, ", chan->f[i]);
+         fprintf(stderr, ")\n");
+      }
+#endif
       break;
 
    case TGSI_FILE_TEMPORARY:
-      index = reg->DstRegister.Index;
+      index = reg->Register.Index;
       assert( index < TGSI_EXEC_NUM_TEMPS );
       dst = &mach->Temps[offset + index].xyzw[chan_index];
       break;
 
    case TGSI_FILE_ADDRESS:
-      index = reg->DstRegister.Index;
+      index = reg->Register.Index;
       dst = &mach->Addrs[index].xyzw[chan_index];
       break;
 
+   case TGSI_FILE_LOOP:
+      assert(reg->Register.Index == 0);
+      assert(mach->LoopCounterStackTop > 0);
+      assert(chan_index == CHAN_X);
+      dst = &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[chan_index];
+      break;
+
+   case TGSI_FILE_PREDICATE:
+      index = reg->Register.Index;
+      assert(index < TGSI_EXEC_NUM_PREDS);
+      dst = &mach->Predicates[index].xyzw[chan_index];
+      break;
+
    default:
       assert( 0 );
       return;
    }
 
-   if (inst->InstructionExtNv.CondFlowEnable) {
-      union tgsi_exec_channel *cc = &mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C];
+   if (inst->Instruction.Predicate) {
       uint swizzle;
-      uint shift;
-      uint mask;
-      uint test;
-
-      /* Only CC0 supported.
-       */
-      assert( inst->InstructionExtNv.CondFlowIndex < 1 );
+      union tgsi_exec_channel *pred;
 
       switch (chan_index) {
       case CHAN_X:
-         swizzle = inst->InstructionExtNv.CondSwizzleX;
+         swizzle = inst->Predicate.SwizzleX;
          break;
       case CHAN_Y:
-         swizzle = inst->InstructionExtNv.CondSwizzleY;
+         swizzle = inst->Predicate.SwizzleY;
          break;
       case CHAN_Z:
-         swizzle = inst->InstructionExtNv.CondSwizzleZ;
+         swizzle = inst->Predicate.SwizzleZ;
          break;
       case CHAN_W:
-         swizzle = inst->InstructionExtNv.CondSwizzleW;
+         swizzle = inst->Predicate.SwizzleW;
          break;
       default:
-         assert( 0 );
+         assert(0);
          return;
       }
 
-      switch (swizzle) {
-      case TGSI_SWIZZLE_X:
-         shift = TGSI_EXEC_CC_X_SHIFT;
-         mask = TGSI_EXEC_CC_X_MASK;
-         break;
-      case TGSI_SWIZZLE_Y:
-         shift = TGSI_EXEC_CC_Y_SHIFT;
-         mask = TGSI_EXEC_CC_Y_MASK;
-         break;
-      case TGSI_SWIZZLE_Z:
-         shift = TGSI_EXEC_CC_Z_SHIFT;
-         mask = TGSI_EXEC_CC_Z_MASK;
-         break;
-      case TGSI_SWIZZLE_W:
-         shift = TGSI_EXEC_CC_W_SHIFT;
-         mask = TGSI_EXEC_CC_W_MASK;
-         break;
-      default:
-         assert( 0 );
-         return;
-      }
+      assert(inst->Predicate.Index == 0);
 
-      switch (inst->InstructionExtNv.CondMask) {
-      case TGSI_CC_GT:
-         test = ~(TGSI_EXEC_CC_GT << shift) & mask;
-         for (i = 0; i < QUAD_SIZE; i++)
-            if (cc->u[i] & test)
-               execmask &= ~(1 << i);
-         break;
+      pred = &mach->Predicates[inst->Predicate.Index].xyzw[swizzle];
 
-      case TGSI_CC_EQ:
-         test = ~(TGSI_EXEC_CC_EQ << shift) & mask;
-         for (i = 0; i < QUAD_SIZE; i++)
-            if (cc->u[i] & test)
-               execmask &= ~(1 << i);
-         break;
-
-      case TGSI_CC_LT:
-         test = ~(TGSI_EXEC_CC_LT << shift) & mask;
-         for (i = 0; i < QUAD_SIZE; i++)
-            if (cc->u[i] & test)
-               execmask &= ~(1 << i);
-         break;
-
-      case TGSI_CC_GE:
-         test = ~((TGSI_EXEC_CC_GT | TGSI_EXEC_CC_EQ) << shift) & mask;
-         for (i = 0; i < QUAD_SIZE; i++)
-            if (cc->u[i] & test)
-               execmask &= ~(1 << i);
-         break;
-
-      case TGSI_CC_LE:
-         test = ~((TGSI_EXEC_CC_LT | TGSI_EXEC_CC_EQ) << shift) & mask;
-         for (i = 0; i < QUAD_SIZE; i++)
-            if (cc->u[i] & test)
+      if (inst->Predicate.Negate) {
+         for (i = 0; i < QUAD_SIZE; i++) {
+            if (pred->u[i]) {
                execmask &= ~(1 << i);
-         break;
-
-      case TGSI_CC_NE:
-         test = ~((TGSI_EXEC_CC_GT | TGSI_EXEC_CC_LT | TGSI_EXEC_CC_UN) << shift) & mask;
-         for (i = 0; i < QUAD_SIZE; i++)
-            if (cc->u[i] & test)
+            }
+         }
+      } else {
+         for (i = 0; i < QUAD_SIZE; i++) {
+            if (!pred->u[i]) {
                execmask &= ~(1 << i);
-         break;
-
-      case TGSI_CC_TR:
-         break;
-
-      case TGSI_CC_FL:
-         for (i = 0; i < QUAD_SIZE; i++)
-            execmask &= ~(1 << i);
-         break;
-
-      default:
-         assert( 0 );
-         return;
+            }
+         }
       }
    }
 
@@ -1622,58 +1407,13 @@ store_dest(
    default:
       assert( 0 );
    }
-
-   if (inst->InstructionExtNv.CondDstUpdate) {
-      union tgsi_exec_channel *cc = &mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C];
-      uint shift;
-      uint mask;
-
-      /* Only CC0 supported.
-       */
-      assert( inst->InstructionExtNv.CondDstIndex < 1 );
-
-      switch (chan_index) {
-      case CHAN_X:
-         shift = TGSI_EXEC_CC_X_SHIFT;
-         mask = ~TGSI_EXEC_CC_X_MASK;
-         break;
-      case CHAN_Y:
-         shift = TGSI_EXEC_CC_Y_SHIFT;
-         mask = ~TGSI_EXEC_CC_Y_MASK;
-         break;
-      case CHAN_Z:
-         shift = TGSI_EXEC_CC_Z_SHIFT;
-         mask = ~TGSI_EXEC_CC_Z_MASK;
-         break;
-      case CHAN_W:
-         shift = TGSI_EXEC_CC_W_SHIFT;
-         mask = ~TGSI_EXEC_CC_W_MASK;
-         break;
-      default:
-         assert( 0 );
-         return;
-      }
-
-      for (i = 0; i < QUAD_SIZE; i++)
-         if (execmask & (1 << i)) {
-            cc->u[i] &= mask;
-            if (dst->f[i] < 0.0f)
-               cc->u[i] |= TGSI_EXEC_CC_LT << shift;
-            else if (dst->f[i] > 0.0f)
-               cc->u[i] |= TGSI_EXEC_CC_GT << shift;
-            else if (dst->f[i] == 0.0f)
-               cc->u[i] |= TGSI_EXEC_CC_EQ << shift;
-            else
-               cc->u[i] |= TGSI_EXEC_CC_UN << shift;
-         }
-   }
 }
 
 #define FETCH(VAL,INDEX,CHAN)\
-    fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN)
+    fetch_source(mach, VAL, &inst->Src[INDEX], CHAN, TGSI_EXEC_DATA_FLOAT)
 
 #define STORE(VAL,INDEX,CHAN)\
-    store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN )
+   store_dest(mach, VAL, &inst->Dst[INDEX], inst, CHAN, TGSI_EXEC_DATA_FLOAT)
 
 
 /**
@@ -1689,10 +1429,8 @@ exec_kil(struct tgsi_exec_machine *mach,
    uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */
    union tgsi_exec_channel r[1];
 
-   /* This mask stores component bits that were already tested. Note that
-    * we test if the value is less than zero, so 1.0 and 0.0 need not to be
-    * tested. */
-   uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE);
+   /* This mask stores component bits that were already tested. */
+   uniquemask = 0;
 
    for (chan_index = 0; chan_index < 4; chan_index++)
    {
@@ -1700,8 +1438,8 @@ exec_kil(struct tgsi_exec_machine *mach,
       uint i;
 
       /* unswizzle channel */
-      swizzle = tgsi_util_get_full_src_register_extswizzle (
-                        &inst->FullSrcRegisters[0],
+      swizzle = tgsi_util_get_full_src_register_swizzle (
+                        &inst->Src[0],
                         chan_index);
 
       /* check if the component has not been already tested */
@@ -1728,35 +1466,40 @@ exec_kilp(struct tgsi_exec_machine *mach,
 {
    uint kilmask; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */
 
-   if (inst->InstructionExtNv.CondFlowEnable) {
-      uint swizzle[4];
-      uint chan_index;
-
-      kilmask = 0x0;
-
-      swizzle[0] = inst->InstructionExtNv.CondSwizzleX;
-      swizzle[1] = inst->InstructionExtNv.CondSwizzleY;
-      swizzle[2] = inst->InstructionExtNv.CondSwizzleZ;
-      swizzle[3] = inst->InstructionExtNv.CondSwizzleW;
-
-      for (chan_index = 0; chan_index < 4; chan_index++)
-      {
-         uint i;
+   /* "unconditional" kil */
+   kilmask = mach->ExecMask;
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
+}
 
-         for (i = 0; i < 4; i++) {
-            /* TODO: evaluate the condition code */
-            if (0)
-               kilmask |= 1 << i;
-         }
-      }
-   }
-   else {
-      /* "unconditional" kil */
-      kilmask = mach->ExecMask;
+static void
+emit_vertex(struct tgsi_exec_machine *mach)
+{
+   /* FIXME: check for exec mask correctly
+   unsigned i;
+   for (i = 0; i < QUAD_SIZE; ++i) {
+         if ((mach->ExecMask & (1 << i)))
+   */
+   if (mach->ExecMask) {
+      mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += mach->NumOutputs;
+      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++;
    }
-   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
 }
 
+static void
+emit_primitive(struct tgsi_exec_machine *mach)
+{
+   unsigned *prim_count = &mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0];
+   /* FIXME: check for exec mask correctly
+   unsigned i;
+   for (i = 0; i < QUAD_SIZE; ++i) {
+         if ((mach->ExecMask & (1 << i)))
+   */
+   if (mach->ExecMask) {
+      ++(*prim_count);
+      debug_assert((*prim_count * mach->NumOutputs) < mach->MaxGeometryShaderOutputs);
+      mach->Primitives[*prim_count] = 0;
+   }
+}
 
 /*
  * Fetch a four texture samples using STR texture coordinates.
@@ -1766,7 +1509,8 @@ fetch_texel( struct tgsi_sampler *sampler,
              const union tgsi_exec_channel *s,
              const union tgsi_exec_channel *t,
              const union tgsi_exec_channel *p,
-             float lodbias,  /* XXX should be float[4] */
+             const union tgsi_exec_channel *c0,
+             enum tgsi_sampler_control control,
              union tgsi_exec_channel *r,
              union tgsi_exec_channel *g,
              union tgsi_exec_channel *b,
@@ -1775,7 +1519,7 @@ fetch_texel( struct tgsi_sampler *sampler,
    uint j;
    float rgba[NUM_CHANNELS][QUAD_SIZE];
 
-   sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba);
+   sampler->get_samples(sampler, s->f, t->f, p->f, c0->f, control, rgba);
 
    for (j = 0; j < 4; j++) {
       r->f[j] = rgba[0][j];
@@ -1786,102 +1530,156 @@ fetch_texel( struct tgsi_sampler *sampler,
 }
 
 
+#define TEX_MODIFIER_NONE           0
+#define TEX_MODIFIER_PROJECTED      1
+#define TEX_MODIFIER_LOD_BIAS       2
+#define TEX_MODIFIER_EXPLICIT_LOD   3
+
+
 static void
 exec_tex(struct tgsi_exec_machine *mach,
          const struct tgsi_full_instruction *inst,
-         boolean biasLod,
-         boolean projected)
+         uint modifier)
 {
-   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   const uint unit = inst->Src[1].Register.Index;
    union tgsi_exec_channel r[4];
+   const union tgsi_exec_channel *lod = &ZeroVec;
+   enum tgsi_sampler_control control;
    uint chan_index;
-   float lodBias;
 
-   /*   debug_printf("Sampler %u unit %u\n", sampler, unit); */
+   if (modifier != TEX_MODIFIER_NONE) {
+      FETCH(&r[3], 0, CHAN_W);
+      if (modifier != TEX_MODIFIER_PROJECTED) {
+         lod = &r[3];
+      }
+   }
+
+   if (modifier == TEX_MODIFIER_EXPLICIT_LOD) {
+      control = tgsi_sampler_lod_explicit;
+   } else {
+      control = tgsi_sampler_lod_bias;
+   }
 
-   switch (inst->InstructionExtTexture.Texture) {
+   switch (inst->Texture.Texture) {
    case TGSI_TEXTURE_1D:
    case TGSI_TEXTURE_SHADOW1D:
-
       FETCH(&r[0], 0, CHAN_X);
 
-      if (projected) {
-         FETCH(&r[1], 0, CHAN_W);
-         micro_div( &r[0], &r[0], &r[1] );
-      }
-
-      if (biasLod) {
-         FETCH(&r[1], 0, CHAN_W);
-         lodBias = r[2].f[0];
+      if (modifier == TEX_MODIFIER_PROJECTED) {
+         micro_div(&r[0], &r[0], &r[3]);
       }
-      else
-         lodBias = 0.0;
 
       fetch_texel(mach->Samplers[unit],
-                  &r[0], &ZeroVec, &ZeroVec, lodBias,  /* S, T, P, BIAS */
-                  &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */
+                  &r[0], &ZeroVec, &ZeroVec, lod,  /* S, T, P, LOD */
+                  control,
+                  &r[0], &r[1], &r[2], &r[3]);     /* R, G, B, A */
       break;
 
    case TGSI_TEXTURE_2D:
    case TGSI_TEXTURE_RECT:
    case TGSI_TEXTURE_SHADOW2D:
    case TGSI_TEXTURE_SHADOWRECT:
-
       FETCH(&r[0], 0, CHAN_X);
       FETCH(&r[1], 0, CHAN_Y);
       FETCH(&r[2], 0, CHAN_Z);
 
-      if (projected) {
-         FETCH(&r[3], 0, CHAN_W);
-         micro_div( &r[0], &r[0], &r[3] );
-         micro_div( &r[1], &r[1], &r[3] );
-         micro_div( &r[2], &r[2], &r[3] );
-      }
-
-      if (biasLod) {
-         FETCH(&r[3], 0, CHAN_W);
-         lodBias = r[3].f[0];
+      if (modifier == TEX_MODIFIER_PROJECTED) {
+         micro_div(&r[0], &r[0], &r[3]);
+         micro_div(&r[1], &r[1], &r[3]);
+         micro_div(&r[2], &r[2], &r[3]);
       }
-      else
-         lodBias = 0.0;
 
       fetch_texel(mach->Samplers[unit],
-                  &r[0], &r[1], &r[2], lodBias,  /* inputs */
+                  &r[0], &r[1], &r[2], lod,     /* S, T, P, LOD */
+                  control,
                   &r[0], &r[1], &r[2], &r[3]);  /* outputs */
       break;
 
    case TGSI_TEXTURE_3D:
    case TGSI_TEXTURE_CUBE:
-
       FETCH(&r[0], 0, CHAN_X);
       FETCH(&r[1], 0, CHAN_Y);
       FETCH(&r[2], 0, CHAN_Z);
 
-      if (projected) {
-         FETCH(&r[3], 0, CHAN_W);
-         micro_div( &r[0], &r[0], &r[3] );
-         micro_div( &r[1], &r[1], &r[3] );
-         micro_div( &r[2], &r[2], &r[3] );
+      if (modifier == TEX_MODIFIER_PROJECTED) {
+         micro_div(&r[0], &r[0], &r[3]);
+         micro_div(&r[1], &r[1], &r[3]);
+         micro_div(&r[2], &r[2], &r[3]);
       }
 
-      if (biasLod) {
-         FETCH(&r[3], 0, CHAN_W);
-         lodBias = r[3].f[0];
-      }
-      else
-         lodBias = 0.0;
+      fetch_texel(mach->Samplers[unit],
+                  &r[0], &r[1], &r[2], lod,
+                  control,
+                  &r[0], &r[1], &r[2], &r[3]);
+      break;
+
+   default:
+      assert(0);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+      STORE(&r[chan_index], 0, chan_index);
+   }
+}
+
+static void
+exec_txd(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   const uint unit = inst->Src[3].Register.Index;
+   union tgsi_exec_channel r[4];
+   uint chan_index;
+
+   /*
+    * XXX: This is fake TXD -- the derivatives are not taken into account, yet.
+    */
+
+   switch (inst->Texture.Texture) {
+   case TGSI_TEXTURE_1D:
+   case TGSI_TEXTURE_SHADOW1D:
+
+      FETCH(&r[0], 0, CHAN_X);
+
+      fetch_texel(mach->Samplers[unit],
+                  &r[0], &ZeroVec, &ZeroVec, &ZeroVec,   /* S, T, P, BIAS */
+                  tgsi_sampler_lod_bias,
+                  &r[0], &r[1], &r[2], &r[3]);           /* R, G, B, A */
+      break;
+
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+   case TGSI_TEXTURE_SHADOW2D:
+   case TGSI_TEXTURE_SHADOWRECT:
+
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 0, CHAN_Z);
 
       fetch_texel(mach->Samplers[unit],
-                  &r[0], &r[1], &r[2], lodBias,
+                  &r[0], &r[1], &r[2], &ZeroVec,   /* inputs */
+                  tgsi_sampler_lod_bias,
+                  &r[0], &r[1], &r[2], &r[3]);     /* outputs */
+      break;
+
+   case TGSI_TEXTURE_3D:
+   case TGSI_TEXTURE_CUBE:
+
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 0, CHAN_Z);
+
+      fetch_texel(mach->Samplers[unit],
+                  &r[0], &r[1], &r[2], &ZeroVec,
+                  tgsi_sampler_lod_bias,
                   &r[0], &r[1], &r[2], &r[3]);
       break;
 
    default:
-      assert (0);
+      assert(0);
    }
 
-   FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-      STORE( &r[chan_index], 0, chan_index );
+   FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+      STORE(&r[chan_index], 0, chan_index);
    }
 }
 
@@ -1954,53 +1752,59 @@ typedef void (* eval_coef_func)(
    unsigned chan );
 
 static void
-exec_declaration(
-   struct tgsi_exec_machine *mach,
-   const struct tgsi_full_declaration *decl )
+exec_declaration(struct tgsi_exec_machine *mach,
+                 const struct tgsi_full_declaration *decl)
 {
-   if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) {
-      if( decl->Declaration.File == TGSI_FILE_INPUT ) {
-         unsigned first, last, mask;
-         eval_coef_func eval;
+   if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) {
+      if (decl->Declaration.File == TGSI_FILE_INPUT ||
+          decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE) {
+         uint first, last, mask;
 
-         first = decl->DeclarationRange.First;
-         last = decl->DeclarationRange.Last;
+         first = decl->Range.First;
+         last = decl->Range.Last;
          mask = decl->Declaration.UsageMask;
 
-         switch( decl->Declaration.Interpolate ) {
-         case TGSI_INTERPOLATE_CONSTANT:
-            eval = eval_constant_coef;
-            break;
-
-         case TGSI_INTERPOLATE_LINEAR:
-            eval = eval_linear_coef;
-            break;
+         if (decl->Semantic.Name == TGSI_SEMANTIC_POSITION) {
+            assert(decl->Semantic.Index == 0);
+            assert(first == last);
+            assert(mask == TGSI_WRITEMASK_XYZW);
 
-         case TGSI_INTERPOLATE_PERSPECTIVE:
-            eval = eval_perspective_coef;
-            break;
+            mach->Inputs[first] = mach->QuadPos;
+         } else if (decl->Semantic.Name == TGSI_SEMANTIC_FACE) {
+            uint i;
 
-         default:
-            eval = NULL;
-            assert( 0 );
-         }
+            assert(decl->Semantic.Index == 0);
+            assert(first == last);
 
-         if( mask == TGSI_WRITEMASK_XYZW ) {
-            unsigned i, j;
-
-            for( i = first; i <= last; i++ ) {
-               for( j = 0; j < NUM_CHANNELS; j++ ) {
-                  eval( mach, i, j );
-               }
+            for (i = 0; i < QUAD_SIZE; i++) {
+               mach->Inputs[first].xyzw[0].f[i] = mach->Face;
+            }
+         } else {
+            eval_coef_func eval;
+            uint i, j;
+
+            switch (decl->Declaration.Interpolate) {
+            case TGSI_INTERPOLATE_CONSTANT:
+               eval = eval_constant_coef;
+               break;
+
+            case TGSI_INTERPOLATE_LINEAR:
+               eval = eval_linear_coef;
+               break;
+
+            case TGSI_INTERPOLATE_PERSPECTIVE:
+               eval = eval_perspective_coef;
+               break;
+
+            default:
+               assert(0);
+               return;
             }
-         }
-         else {
-            unsigned i, j;
 
-            for( j = 0; j < NUM_CHANNELS; j++ ) {
-               if( mask & (1 << j) ) {
-                  for( i = first; i <= last; i++ ) {
-                     eval( mach, i, j );
+            for (j = 0; j < NUM_CHANNELS; j++) {
+               if (mask & (1 << j)) {
+                  for (i = first; i <= last; i++) {
+                     eval(mach, i, j);
                   }
                }
             }
@@ -2009,6 +1813,585 @@ exec_declaration(
    }
 }
 
+typedef void (* micro_op)(union tgsi_exec_channel *dst,
+                          const union tgsi_exec_channel *src);
+
+static void
+exec_scalar_unary(struct tgsi_exec_machine *mach,
+                  const struct tgsi_full_instruction *inst,
+                  micro_op op,
+                  enum tgsi_exec_datatype dst_datatype,
+                  enum tgsi_exec_datatype src_datatype)
+{
+   unsigned int chan;
+   union tgsi_exec_channel src;
+   union tgsi_exec_channel dst;
+
+   fetch_source(mach, &src, &inst->Src[0], CHAN_X, src_datatype);
+   op(&dst, &src);
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst, &inst->Dst[0], inst, chan, dst_datatype);
+      }
+   }
+}
+
+static void
+exec_vector_unary(struct tgsi_exec_machine *mach,
+                  const struct tgsi_full_instruction *inst,
+                  micro_op op,
+                  enum tgsi_exec_datatype dst_datatype,
+                  enum tgsi_exec_datatype src_datatype)
+{
+   unsigned int chan;
+   struct tgsi_exec_vector dst;
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         union tgsi_exec_channel src;
+
+         fetch_source(mach, &src, &inst->Src[0], chan, src_datatype);
+         op(&dst.xyzw[chan], &src);
+      }
+   }
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst.xyzw[chan], &inst->Dst[0], inst, chan, dst_datatype);
+      }
+   }
+}
+
+static void
+exec_vector_binary(struct tgsi_exec_machine *mach,
+                   const struct tgsi_full_instruction *inst,
+                   micro_op op,
+                   enum tgsi_exec_datatype dst_datatype,
+                   enum tgsi_exec_datatype src_datatype)
+{
+   unsigned int chan;
+   struct tgsi_exec_vector dst;
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         union tgsi_exec_channel src[2];
+
+         fetch_source(mach, &src[0], &inst->Src[0], chan, src_datatype);
+         fetch_source(mach, &src[1], &inst->Src[1], chan, src_datatype);
+         op(&dst.xyzw[chan], src);
+      }
+   }
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst.xyzw[chan], &inst->Dst[0], inst, chan, dst_datatype);
+      }
+   }
+}
+
+static void
+exec_vector_trinary(struct tgsi_exec_machine *mach,
+                    const struct tgsi_full_instruction *inst,
+                    micro_op op,
+                    enum tgsi_exec_datatype dst_datatype,
+                    enum tgsi_exec_datatype src_datatype)
+{
+   unsigned int chan;
+   struct tgsi_exec_vector dst;
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         union tgsi_exec_channel src[3];
+
+         fetch_source(mach, &src[0], &inst->Src[0], chan, src_datatype);
+         fetch_source(mach, &src[1], &inst->Src[1], chan, src_datatype);
+         fetch_source(mach, &src[2], &inst->Src[2], chan, src_datatype);
+         op(&dst.xyzw[chan], src);
+      }
+   }
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &dst.xyzw[chan], &inst->Dst[0], inst, chan, dst_datatype);
+      }
+   }
+}
+
+static void
+exec_dp3(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg[3];
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_mul(&arg[2], &arg[0], &arg[1]);
+
+   for (chan = CHAN_Y; chan <= CHAN_Z; chan++) {
+      fetch_source(mach, &arg[0], &inst->Src[0], chan, TGSI_EXEC_DATA_FLOAT);
+      fetch_source(mach, &arg[1], &inst->Src[1], chan, TGSI_EXEC_DATA_FLOAT);
+      micro_mad(&arg[2], arg);
+   }
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &arg[2], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+}
+
+static void
+exec_dp4(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg[3];
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_mul(&arg[2], &arg[0], &arg[1]);
+
+   for (chan = CHAN_Y; chan <= CHAN_W; chan++) {
+      fetch_source(mach, &arg[0], &inst->Src[0], chan, TGSI_EXEC_DATA_FLOAT);
+      fetch_source(mach, &arg[1], &inst->Src[1], chan, TGSI_EXEC_DATA_FLOAT);
+      micro_mad(&arg[2], arg);
+   }
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &arg[2], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+}
+
+static void
+exec_dp2a(struct tgsi_exec_machine *mach,
+          const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg[3];
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_mul(&arg[2], &arg[0], &arg[1]);
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   micro_mad(&arg[0], arg);
+
+   fetch_source(mach, &arg[1], &inst->Src[2], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_add(&arg[0], &arg[0], &arg[1]);
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &arg[0], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+}
+
+static void
+exec_dph(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg[3];
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_mul(&arg[2], &arg[0], &arg[1]);
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   micro_mad(&arg[2], arg);
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Z, TGSI_EXEC_DATA_FLOAT);
+   micro_mad(&arg[0], arg);
+
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_W, TGSI_EXEC_DATA_FLOAT);
+   micro_add(&arg[0], &arg[0], &arg[1]);
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &arg[0], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+}
+
+static void
+exec_dp2(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   unsigned int chan;
+   union tgsi_exec_channel arg[3];
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT);
+   micro_mul(&arg[2], &arg[0], &arg[1]);
+
+   fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT);
+   micro_mad(&arg[2], arg);
+
+   for (chan = 0; chan < NUM_CHANNELS; chan++) {
+      if (inst->Dst[0].Register.WriteMask & (1 << chan)) {
+         store_dest(mach, &arg[2], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT);
+      }
+   }
+}
+
+static void
+exec_break(struct tgsi_exec_machine *mach)
+{
+   if (mach->BreakType == TGSI_EXEC_BREAK_INSIDE_LOOP) {
+      /* turn off loop channels for each enabled exec channel */
+      mach->LoopMask &= ~mach->ExecMask;
+      /* Todo: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
+   } else {
+      assert(mach->BreakType == TGSI_EXEC_BREAK_INSIDE_SWITCH);
+
+      mach->Switch.mask = 0x0;
+
+      UPDATE_EXEC_MASK(mach);
+   }
+}
+
+static void
+exec_switch(struct tgsi_exec_machine *mach,
+            const struct tgsi_full_instruction *inst)
+{
+   assert(mach->SwitchStackTop < TGSI_EXEC_MAX_SWITCH_NESTING);
+   assert(mach->BreakStackTop < TGSI_EXEC_MAX_BREAK_STACK);
+
+   mach->SwitchStack[mach->SwitchStackTop++] = mach->Switch;
+   fetch_source(mach, &mach->Switch.selector, &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_UINT);
+   mach->Switch.mask = 0x0;
+   mach->Switch.defaultMask = 0x0;
+
+   mach->BreakStack[mach->BreakStackTop++] = mach->BreakType;
+   mach->BreakType = TGSI_EXEC_BREAK_INSIDE_SWITCH;
+
+   UPDATE_EXEC_MASK(mach);
+}
+
+static void
+exec_case(struct tgsi_exec_machine *mach,
+          const struct tgsi_full_instruction *inst)
+{
+   uint prevMask = mach->SwitchStack[mach->SwitchStackTop - 1].mask;
+   union tgsi_exec_channel src;
+   uint mask = 0;
+
+   fetch_source(mach, &src, &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_UINT);
+
+   if (mach->Switch.selector.u[0] == src.u[0]) {
+      mask |= 0x1;
+   }
+   if (mach->Switch.selector.u[1] == src.u[1]) {
+      mask |= 0x2;
+   }
+   if (mach->Switch.selector.u[2] == src.u[2]) {
+      mask |= 0x4;
+   }
+   if (mach->Switch.selector.u[3] == src.u[3]) {
+      mask |= 0x8;
+   }
+
+   mach->Switch.defaultMask |= mask;
+
+   mach->Switch.mask |= mask & prevMask;
+
+   UPDATE_EXEC_MASK(mach);
+}
+
+static void
+exec_default(struct tgsi_exec_machine *mach)
+{
+   uint prevMask = mach->SwitchStack[mach->SwitchStackTop - 1].mask;
+
+   mach->Switch.mask |= ~mach->Switch.defaultMask & prevMask;
+
+   UPDATE_EXEC_MASK(mach);
+}
+
+static void
+exec_endswitch(struct tgsi_exec_machine *mach)
+{
+   mach->Switch = mach->SwitchStack[--mach->SwitchStackTop];
+   mach->BreakType = mach->BreakStack[--mach->BreakStackTop];
+
+   UPDATE_EXEC_MASK(mach);
+}
+
+static void
+micro_i2f(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = (float)src->i[0];
+   dst->f[1] = (float)src->i[1];
+   dst->f[2] = (float)src->i[2];
+   dst->f[3] = (float)src->i[3];
+}
+
+static void
+micro_not(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->u[0] = ~src->u[0];
+   dst->u[1] = ~src->u[1];
+   dst->u[2] = ~src->u[2];
+   dst->u[3] = ~src->u[3];
+}
+
+static void
+micro_shl(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] << src[1].u[0];
+   dst->u[1] = src[0].u[1] << src[1].u[1];
+   dst->u[2] = src[0].u[2] << src[1].u[2];
+   dst->u[3] = src[0].u[3] << src[1].u[3];
+}
+
+static void
+micro_and(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] & src[1].u[0];
+   dst->u[1] = src[0].u[1] & src[1].u[1];
+   dst->u[2] = src[0].u[2] & src[1].u[2];
+   dst->u[3] = src[0].u[3] & src[1].u[3];
+}
+
+static void
+micro_or(union tgsi_exec_channel *dst,
+         const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] | src[1].u[0];
+   dst->u[1] = src[0].u[1] | src[1].u[1];
+   dst->u[2] = src[0].u[2] | src[1].u[2];
+   dst->u[3] = src[0].u[3] | src[1].u[3];
+}
+
+static void
+micro_xor(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] ^ src[1].u[0];
+   dst->u[1] = src[0].u[1] ^ src[1].u[1];
+   dst->u[2] = src[0].u[2] ^ src[1].u[2];
+   dst->u[3] = src[0].u[3] ^ src[1].u[3];
+}
+
+static void
+micro_f2i(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->i[0] = (int)src->f[0];
+   dst->i[1] = (int)src->f[1];
+   dst->i[2] = (int)src->f[2];
+   dst->i[3] = (int)src->f[3];
+}
+
+static void
+micro_idiv(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src[0].i[0] / src[1].i[0];
+   dst->i[1] = src[0].i[1] / src[1].i[1];
+   dst->i[2] = src[0].i[2] / src[1].i[2];
+   dst->i[3] = src[0].i[3] / src[1].i[3];
+}
+
+static void
+micro_imax(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src[0].i[0] > src[1].i[0] ? src[0].i[0] : src[1].i[0];
+   dst->i[1] = src[0].i[1] > src[1].i[1] ? src[0].i[1] : src[1].i[1];
+   dst->i[2] = src[0].i[2] > src[1].i[2] ? src[0].i[2] : src[1].i[2];
+   dst->i[3] = src[0].i[3] > src[1].i[3] ? src[0].i[3] : src[1].i[3];
+}
+
+static void
+micro_imin(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src[0].i[0] < src[1].i[0] ? src[0].i[0] : src[1].i[0];
+   dst->i[1] = src[0].i[1] < src[1].i[1] ? src[0].i[1] : src[1].i[1];
+   dst->i[2] = src[0].i[2] < src[1].i[2] ? src[0].i[2] : src[1].i[2];
+   dst->i[3] = src[0].i[3] < src[1].i[3] ? src[0].i[3] : src[1].i[3];
+}
+
+static void
+micro_isge(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src[0].i[0] >= src[1].i[0] ? -1 : 0;
+   dst->i[1] = src[0].i[1] >= src[1].i[1] ? -1 : 0;
+   dst->i[2] = src[0].i[2] >= src[1].i[2] ? -1 : 0;
+   dst->i[3] = src[0].i[3] >= src[1].i[3] ? -1 : 0;
+}
+
+static void
+micro_ishr(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src[0].i[0] >> src[1].i[0];
+   dst->i[1] = src[0].i[1] >> src[1].i[1];
+   dst->i[2] = src[0].i[2] >> src[1].i[2];
+   dst->i[3] = src[0].i[3] >> src[1].i[3];
+}
+
+static void
+micro_islt(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->i[0] = src[0].i[0] < src[1].i[0] ? -1 : 0;
+   dst->i[1] = src[0].i[1] < src[1].i[1] ? -1 : 0;
+   dst->i[2] = src[0].i[2] < src[1].i[2] ? -1 : 0;
+   dst->i[3] = src[0].i[3] < src[1].i[3] ? -1 : 0;
+}
+
+static void
+micro_f2u(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->u[0] = (uint)src->f[0];
+   dst->u[1] = (uint)src->f[1];
+   dst->u[2] = (uint)src->f[2];
+   dst->u[3] = (uint)src->f[3];
+}
+
+static void
+micro_u2f(union tgsi_exec_channel *dst,
+          const union tgsi_exec_channel *src)
+{
+   dst->f[0] = (float)src->u[0];
+   dst->f[1] = (float)src->u[1];
+   dst->f[2] = (float)src->u[2];
+   dst->f[3] = (float)src->u[3];
+}
+
+static void
+micro_uadd(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] + src[1].u[0];
+   dst->u[1] = src[0].u[1] + src[1].u[1];
+   dst->u[2] = src[0].u[2] + src[1].u[2];
+   dst->u[3] = src[0].u[3] + src[1].u[3];
+}
+
+static void
+micro_udiv(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] / src[1].u[0];
+   dst->u[1] = src[0].u[1] / src[1].u[1];
+   dst->u[2] = src[0].u[2] / src[1].u[2];
+   dst->u[3] = src[0].u[3] / src[1].u[3];
+}
+
+static void
+micro_umad(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] * src[1].u[0] + src[2].u[0];
+   dst->u[1] = src[0].u[1] * src[1].u[1] + src[2].u[1];
+   dst->u[2] = src[0].u[2] * src[1].u[2] + src[2].u[2];
+   dst->u[3] = src[0].u[3] * src[1].u[3] + src[2].u[3];
+}
+
+static void
+micro_umax(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] > src[1].u[0] ? src[0].u[0] : src[1].u[0];
+   dst->u[1] = src[0].u[1] > src[1].u[1] ? src[0].u[1] : src[1].u[1];
+   dst->u[2] = src[0].u[2] > src[1].u[2] ? src[0].u[2] : src[1].u[2];
+   dst->u[3] = src[0].u[3] > src[1].u[3] ? src[0].u[3] : src[1].u[3];
+}
+
+static void
+micro_umin(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] < src[1].u[0] ? src[0].u[0] : src[1].u[0];
+   dst->u[1] = src[0].u[1] < src[1].u[1] ? src[0].u[1] : src[1].u[1];
+   dst->u[2] = src[0].u[2] < src[1].u[2] ? src[0].u[2] : src[1].u[2];
+   dst->u[3] = src[0].u[3] < src[1].u[3] ? src[0].u[3] : src[1].u[3];
+}
+
+static void
+micro_umod(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] % src[1].u[0];
+   dst->u[1] = src[0].u[1] % src[1].u[1];
+   dst->u[2] = src[0].u[2] % src[1].u[2];
+   dst->u[3] = src[0].u[3] % src[1].u[3];
+}
+
+static void
+micro_umul(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] * src[1].u[0];
+   dst->u[1] = src[0].u[1] * src[1].u[1];
+   dst->u[2] = src[0].u[2] * src[1].u[2];
+   dst->u[3] = src[0].u[3] * src[1].u[3];
+}
+
+static void
+micro_useq(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] == src[1].u[0] ? ~0 : 0;
+   dst->u[1] = src[0].u[1] == src[1].u[1] ? ~0 : 0;
+   dst->u[2] = src[0].u[2] == src[1].u[2] ? ~0 : 0;
+   dst->u[3] = src[0].u[3] == src[1].u[3] ? ~0 : 0;
+}
+
+static void
+micro_usge(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] >= src[1].u[0] ? ~0 : 0;
+   dst->u[1] = src[0].u[1] >= src[1].u[1] ? ~0 : 0;
+   dst->u[2] = src[0].u[2] >= src[1].u[2] ? ~0 : 0;
+   dst->u[3] = src[0].u[3] >= src[1].u[3] ? ~0 : 0;
+}
+
+static void
+micro_ushr(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] >> src[1].u[0];
+   dst->u[1] = src[0].u[1] >> src[1].u[1];
+   dst->u[2] = src[0].u[2] >> src[1].u[2];
+   dst->u[3] = src[0].u[3] >> src[1].u[3];
+}
+
+static void
+micro_uslt(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] < src[1].u[0] ? ~0 : 0;
+   dst->u[1] = src[0].u[1] < src[1].u[1] ? ~0 : 0;
+   dst->u[2] = src[0].u[2] < src[1].u[2] ? ~0 : 0;
+   dst->u[3] = src[0].u[3] < src[1].u[3] ? ~0 : 0;
+}
+
+static void
+micro_usne(union tgsi_exec_channel *dst,
+           const union tgsi_exec_channel *src)
+{
+   dst->u[0] = src[0].u[0] != src[1].u[0] ? ~0 : 0;
+   dst->u[1] = src[0].u[1] != src[1].u[1] ? ~0 : 0;
+   dst->u[2] = src[0].u[2] != src[1].u[2] ? ~0 : 0;
+   dst->u[3] = src[0].u[3] != src[1].u[3] ? ~0 : 0;
+}
+
 static void
 exec_instruction(
    struct tgsi_exec_machine *mach,
@@ -2017,51 +2400,24 @@ exec_instruction(
 {
    uint chan_index;
    union tgsi_exec_channel r[10];
+   union tgsi_exec_channel d[8];
 
    (*pc)++;
 
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_ARL:
-   case TGSI_OPCODE_FLR:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_flr( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_arl, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_MOV:
-   case TGSI_OPCODE_SWZ:
-      if (inst->Flags & SOA_DEPENDENCY_FLAG) {
-         /* Do all fetches into temp regs, then do all stores to avoid
-          * intermediate/accidental clobbering.  This could be done all the
-          * time for MOV but for other instructions we'll need more temps...
-          */
-         FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-            FETCH( &r[chan_index], 0, chan_index );
-         }
-         FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-            STORE( &r[chan_index], 0, chan_index );
-         }
-      }
-      else {
-         FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-            FETCH( &r[0], 0, chan_index );
-            STORE( &r[0], 0, chan_index );
-         }
-      }
+      exec_vector_unary(mach, inst, micro_mov, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_LIT:
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
-      }
-
       if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
          FETCH( &r[0], 0, CHAN_X );
          if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
-            micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-            STORE( &r[0], 0, CHAN_Y );
+            micro_max(&d[CHAN_Y], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
          }
 
          if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
@@ -2072,34 +2428,30 @@ exec_instruction(
             micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] );
             micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] );
             micro_pow( &r[1], &r[1], &r[2] );
-            micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-            STORE( &r[0], 0, CHAN_Z );
+            micro_lt(&d[CHAN_Z], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
          }
-      }
 
+         if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+            STORE(&d[CHAN_Y], 0, CHAN_Y);
+         }
+         if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+            STORE(&d[CHAN_Z], 0, CHAN_Z);
+         }
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
+      }
       if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
          STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
       }
       break;
 
    case TGSI_OPCODE_RCP:
-   /* TGSI_OPCODE_RECIP */
-      FETCH( &r[0], 0, CHAN_X );
-      micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] );
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_unary(mach, inst, micro_rcp, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_RSQ:
-   /* TGSI_OPCODE_RECIPSQRT */
-      FETCH( &r[0], 0, CHAN_X );
-      micro_abs( &r[0], &r[0] );
-      micro_sqrt( &r[0], &r[0] );
-      micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] );
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_unary(mach, inst, micro_rsq, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_EXP:
@@ -2144,14 +2496,13 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_MUL:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index )
-      {
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
          FETCH(&r[0], 0, chan_index);
          FETCH(&r[1], 1, chan_index);
-
-         micro_mul( &r[0], &r[0], &r[1] );
-
-         STORE(&r[0], 0, chan_index);
+         micro_mul(&d[chan_index], &r[0], &r[1]);
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
@@ -2159,82 +2510,45 @@ exec_instruction(
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( &r[0], 0, chan_index );
          FETCH( &r[1], 1, chan_index );
-         micro_add( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
+         micro_add(&d[chan_index], &r[0], &r[1]);
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
    case TGSI_OPCODE_DP3:
-   /* TGSI_OPCODE_DOT3 */
-      FETCH( &r[0], 0, CHAN_X );
-      FETCH( &r[1], 1, CHAN_X );
-      micro_mul( &r[0], &r[0], &r[1] );
-
-      FETCH( &r[1], 0, CHAN_Y );
-      FETCH( &r[2], 1, CHAN_Y );
-      micro_mul( &r[1], &r[1], &r[2] );
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FETCH( &r[1], 0, CHAN_Z );
-      FETCH( &r[2], 1, CHAN_Z );
-      micro_mul( &r[1], &r[1], &r[2] );
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_dp3(mach, inst);
       break;
 
-    case TGSI_OPCODE_DP4:
-    /* TGSI_OPCODE_DOT4 */
-       FETCH(&r[0], 0, CHAN_X);
-       FETCH(&r[1], 1, CHAN_X);
-
-       micro_mul( &r[0], &r[0], &r[1] );
-
-       FETCH(&r[1], 0, CHAN_Y);
-       FETCH(&r[2], 1, CHAN_Y);
-
-       micro_mul( &r[1], &r[1], &r[2] );
-       micro_add( &r[0], &r[0], &r[1] );
-
-       FETCH(&r[1], 0, CHAN_Z);
-       FETCH(&r[2], 1, CHAN_Z);
-
-       micro_mul( &r[1], &r[1], &r[2] );
-       micro_add( &r[0], &r[0], &r[1] );
-
-       FETCH(&r[1], 0, CHAN_W);
-       FETCH(&r[2], 1, CHAN_W);
-
-       micro_mul( &r[1], &r[1], &r[2] );
-       micro_add( &r[0], &r[0], &r[1] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+   case TGSI_OPCODE_DP4:
+      exec_dp4(mach, inst);
       break;
 
    case TGSI_OPCODE_DST:
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
-         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
-      }
-
       if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
          FETCH( &r[0], 0, CHAN_Y );
          FETCH( &r[1], 1, CHAN_Y);
-         micro_mul( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, CHAN_Y );
+         micro_mul(&d[CHAN_Y], &r[0], &r[1]);
       }
-
       if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-         FETCH( &r[0], 0, CHAN_Z );
-         STORE( &r[0], 0, CHAN_Z );
+         FETCH(&d[CHAN_Z], 0, CHAN_Z);
       }
-
       if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
-         FETCH( &r[0], 1, CHAN_W );
-         STORE( &r[0], 0, CHAN_W );
+         FETCH(&d[CHAN_W], 1, CHAN_W);
+      }
+
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
+         STORE(&mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+         STORE(&d[CHAN_Y], 0, CHAN_Y);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+         STORE(&d[CHAN_Z], 0, CHAN_Z);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
+         STORE(&d[CHAN_W], 0, CHAN_W);
       }
       break;
 
@@ -2244,9 +2558,10 @@ exec_instruction(
          FETCH(&r[1], 1, chan_index);
 
          /* XXX use micro_min()?? */
-         micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] );
-
-         STORE(&r[0], 0, chan_index);
+         micro_lt(&d[chan_index], &r[0], &r[1], &r[0], &r[1]);
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
@@ -2256,67 +2571,38 @@ exec_instruction(
          FETCH(&r[1], 1, chan_index);
 
          /* XXX use micro_max()?? */
-         micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] );
-
-         STORE(&r[0], 0, chan_index );
+         micro_lt(&d[chan_index], &r[0], &r[1], &r[1], &r[0] );
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
    case TGSI_OPCODE_SLT:
-   /* TGSI_OPCODE_SETLT */
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_slt, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SGE:
-   /* TGSI_OPCODE_SETGE */
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_le( &r[0], &r[1], &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_sge, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_MAD:
-   /* TGSI_OPCODE_MADD */
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_mul( &r[0], &r[0], &r[1] );
-         FETCH( &r[1], 2, chan_index );
-         micro_add( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_trinary(mach, inst, micro_mad, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SUB:
       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH(&r[0], 0, chan_index);
          FETCH(&r[1], 1, chan_index);
-
-         micro_sub( &r[0], &r[0], &r[1] );
-
-         STORE(&r[0], 0, chan_index);
+         micro_sub(&d[chan_index], &r[0], &r[1]);
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
    case TGSI_OPCODE_LRP:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH(&r[0], 0, chan_index);
-         FETCH(&r[1], 1, chan_index);
-         FETCH(&r[2], 2, chan_index);
-
-         micro_sub( &r[1], &r[1], &r[2] );
-         micro_mul( &r[0], &r[0], &r[1] );
-         micro_add( &r[0], &r[0], &r[2] );
-
-         STORE(&r[0], 0, chan_index);
-      }
+      exec_vector_trinary(mach, inst, micro_lrp, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_CND:
@@ -2324,35 +2610,19 @@ exec_instruction(
          FETCH(&r[0], 0, chan_index);
          FETCH(&r[1], 1, chan_index);
          FETCH(&r[2], 2, chan_index);
-         micro_lt(&r[0], &mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C], &r[2], &r[0], &r[1]);
-         STORE(&r[0], 0, chan_index);
+         micro_lt(&d[chan_index], &mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C], &r[2], &r[0], &r[1]);
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
    case TGSI_OPCODE_DP2A:
-      FETCH( &r[0], 0, CHAN_X );
-      FETCH( &r[1], 1, CHAN_X );
-      micro_mul( &r[0], &r[0], &r[1] );
-
-      FETCH( &r[1], 0, CHAN_Y );
-      FETCH( &r[2], 1, CHAN_Y );
-      micro_mul( &r[1], &r[1], &r[2] );
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FETCH( &r[2], 2, CHAN_X );
-      micro_add( &r[0], &r[0], &r[2] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_dp2a(mach, inst);
       break;
 
    case TGSI_OPCODE_FRC:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_frc( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_frc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_CLAMP:
@@ -2361,40 +2631,27 @@ exec_instruction(
          FETCH(&r[1], 1, chan_index);
          micro_max(&r[0], &r[0], &r[1]);
          FETCH(&r[1], 2, chan_index);
-         micro_min(&r[0], &r[0], &r[1]);
-         STORE(&r[0], 0, chan_index);
+         micro_min(&d[chan_index], &r[0], &r[1]);
       }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_FLR:
+      exec_vector_unary(mach, inst, micro_flr, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_ROUND:
-   case TGSI_OPCODE_ARR:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_rnd( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_rnd, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_EX2:
-      FETCH(&r[0], 0, CHAN_X);
-
-#if FAST_MATH
-      micro_exp2( &r[0], &r[0] );
-#else
-      micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] );
-#endif
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_unary(mach, inst, micro_exp2, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_LG2:
-      FETCH( &r[0], 0, CHAN_X );
-      micro_lg2( &r[0], &r[0] );
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_unary(mach, inst, micro_lg2, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_POW:
@@ -2418,11 +2675,7 @@ exec_instruction(
       FETCH(&r[4], 1, CHAN_Y);
 
       micro_mul( &r[5], &r[3], &r[4] );
-      micro_sub( &r[2], &r[2], &r[5] );
-
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
-         STORE( &r[2], 0, CHAN_X );
-      }
+      micro_sub(&d[CHAN_X], &r[2], &r[5]);
 
       FETCH(&r[2], 1, CHAN_X);
 
@@ -2431,34 +2684,29 @@ exec_instruction(
       FETCH(&r[5], 0, CHAN_X);
 
       micro_mul( &r[1], &r[1], &r[5] );
-      micro_sub( &r[3], &r[3], &r[1] );
-
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
-         STORE( &r[3], 0, CHAN_Y );
-      }
+      micro_sub(&d[CHAN_Y], &r[3], &r[1]);
 
       micro_mul( &r[5], &r[5], &r[4] );
       micro_mul( &r[0], &r[0], &r[2] );
-      micro_sub( &r[5], &r[5], &r[0] );
+      micro_sub(&d[CHAN_Z], &r[5], &r[0]);
 
-      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
-         STORE( &r[5], 0, CHAN_Z );
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
+         STORE(&d[CHAN_X], 0, CHAN_X);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+         STORE(&d[CHAN_Y], 0, CHAN_Y);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+         STORE(&d[CHAN_Z], 0, CHAN_Z);
       }
-
       if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
          STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
       }
       break;
 
-    case TGSI_OPCODE_ABS:
-       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-          FETCH(&r[0], 0, chan_index);
-
-          micro_abs( &r[0], &r[0] );
-
-          STORE(&r[0], 0, chan_index);
-       }
-       break;
+   case TGSI_OPCODE_ABS:
+      exec_vector_unary(mach, inst, micro_abs, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
+      break;
 
    case TGSI_OPCODE_RCC:
       FETCH(&r[0], 0, CHAN_X);
@@ -2470,56 +2718,19 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_DPH:
-      FETCH(&r[0], 0, CHAN_X);
-      FETCH(&r[1], 1, CHAN_X);
-
-      micro_mul( &r[0], &r[0], &r[1] );
-
-      FETCH(&r[1], 0, CHAN_Y);
-      FETCH(&r[2], 1, CHAN_Y);
-
-      micro_mul( &r[1], &r[1], &r[2] );
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FETCH(&r[1], 0, CHAN_Z);
-      FETCH(&r[2], 1, CHAN_Z);
-
-      micro_mul( &r[1], &r[1], &r[2] );
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FETCH(&r[1], 1, CHAN_W);
-
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_dph(mach, inst);
       break;
 
    case TGSI_OPCODE_COS:
-      FETCH(&r[0], 0, CHAN_X);
-
-      micro_cos( &r[0], &r[0] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_unary(mach, inst, micro_cos, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_DDX:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_ddx( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_ddx, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_DDY:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_ddy( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_ddy, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_KILP:
@@ -2596,14 +2807,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_SEQ:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_eq( &r[0], &r[0], &r[1],
-                   &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C],
-                   &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_seq, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SFL:
@@ -2613,38 +2817,19 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_SGT:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_sgt, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SIN:
-      FETCH( &r[0], 0, CHAN_X );
-      micro_sin( &r[0], &r[0] );
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_scalar_unary(mach, inst, micro_sin, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SLE:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_sle, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SNE:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_sne, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_STR:
@@ -2657,14 +2842,14 @@ exec_instruction(
       /* simple texture lookup */
       /* src[0] = texcoord */
       /* src[1] = sampler unit */
-      exec_tex(mach, inst, FALSE, FALSE);
+      exec_tex(mach, inst, TEX_MODIFIER_NONE);
       break;
 
    case TGSI_OPCODE_TXB:
       /* Texture lookup with lod bias */
       /* src[0] = texcoord (src[0].w = LOD bias) */
       /* src[1] = sampler unit */
-      exec_tex(mach, inst, TRUE, FALSE);
+      exec_tex(mach, inst, TEX_MODIFIER_LOD_BIAS);
       break;
 
    case TGSI_OPCODE_TXD:
@@ -2673,21 +2858,21 @@ exec_instruction(
       /* src[1] = d[strq]/dx */
       /* src[2] = d[strq]/dy */
       /* src[3] = sampler unit */
-      assert (0);
+      exec_txd(mach, inst);
       break;
 
    case TGSI_OPCODE_TXL:
       /* Texture lookup with explit LOD */
       /* src[0] = texcoord (src[0].w = LOD) */
       /* src[1] = sampler unit */
-      exec_tex(mach, inst, TRUE, FALSE);
+      exec_tex(mach, inst, TEX_MODIFIER_EXPLICIT_LOD);
       break;
 
    case TGSI_OPCODE_TXP:
       /* Texture lookup with projection */
       /* src[0] = texcoord (src[0].w = projection) */
       /* src[1] = sampler unit */
-      exec_tex(mach, inst, FALSE, TRUE);
+      exec_tex(mach, inst, TEX_MODIFIER_PROJECTED);
       break;
 
    case TGSI_OPCODE_UP2H:
@@ -2717,13 +2902,8 @@ exec_instruction(
          micro_mul(&r[3], &r[3], &r[1]);
          micro_add(&r[2], &r[2], &r[3]);
          FETCH(&r[3], 0, CHAN_X);
-         micro_add(&r[2], &r[2], &r[3]);
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
-            STORE(&r[2], 0, CHAN_X);
-         }
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
-            STORE(&r[2], 0, CHAN_Z);
-         }
+         micro_add(&d[CHAN_X], &r[2], &r[3]);
+         
       }
       if (IS_CHANNEL_ENABLED(*inst, CHAN_Y) ||
           IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
@@ -2733,13 +2913,20 @@ exec_instruction(
          micro_mul(&r[3], &r[3], &r[1]);
          micro_add(&r[2], &r[2], &r[3]);
          FETCH(&r[3], 0, CHAN_Y);
-         micro_add(&r[2], &r[2], &r[3]);
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
-            STORE(&r[2], 0, CHAN_Y);
-         }
-         if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
-            STORE(&r[2], 0, CHAN_W);
-         }
+         micro_add(&d[CHAN_Y], &r[2], &r[3]);
+         
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_X)) {
+         STORE(&d[CHAN_X], 0, CHAN_X);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+         STORE(&d[CHAN_Y], 0, CHAN_Y);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+         STORE(&d[CHAN_X], 0, CHAN_Z);
+      }
+      if (IS_CHANNEL_ENABLED(*inst, CHAN_W)) {
+         STORE(&d[CHAN_Y], 0, CHAN_W);
       }
       break;
 
@@ -2747,6 +2934,10 @@ exec_instruction(
       assert (0);
       break;
 
+   case TGSI_OPCODE_ARR:
+      exec_vector_unary(mach, inst, micro_arr, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
    case TGSI_OPCODE_BRA:
       assert (0);
       break;
@@ -2766,6 +2957,8 @@ exec_instruction(
          mach->CallStack[mach->CallStackTop].CondStackTop = mach->CondStackTop;
          mach->CallStack[mach->CallStackTop].LoopStackTop = mach->LoopStackTop;
          mach->CallStack[mach->CallStackTop].ContStackTop = mach->ContStackTop;
+         mach->CallStack[mach->CallStackTop].SwitchStackTop = mach->SwitchStackTop;
+         mach->CallStack[mach->CallStackTop].BreakStackTop = mach->BreakStackTop;
          /* note that PC was already incremented above */
          mach->CallStack[mach->CallStackTop].ReturnAddr = *pc;
 
@@ -2773,16 +2966,21 @@ exec_instruction(
 
          /* Second, push the Cond, Loop, Cont, Func stacks */
          assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING);
-         mach->CondStack[mach->CondStackTop++] = mach->CondMask;
          assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
-         mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
          assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
-         mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+         assert(mach->SwitchStackTop < TGSI_EXEC_MAX_SWITCH_NESTING);
+         assert(mach->BreakStackTop < TGSI_EXEC_MAX_BREAK_STACK);
          assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING);
+
+         mach->CondStack[mach->CondStackTop++] = mach->CondMask;
+         mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
+         mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+         mach->SwitchStack[mach->SwitchStackTop++] = mach->Switch;
+         mach->BreakStack[mach->BreakStackTop++] = mach->BreakType;
          mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask;
 
          /* Finally, jump to the subroutine */
-         *pc = inst->InstructionExtLabel.Label;
+         *pc = inst->Label.Label;
       }
       break;
 
@@ -2811,6 +3009,12 @@ exec_instruction(
          mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop;
          mach->ContMask = mach->ContStack[mach->ContStackTop];
 
+         mach->SwitchStackTop = mach->CallStack[mach->CallStackTop].SwitchStackTop;
+         mach->Switch = mach->SwitchStack[mach->SwitchStackTop];
+
+         mach->BreakStackTop = mach->CallStack[mach->CallStackTop].BreakStackTop;
+         mach->BreakType = mach->BreakStack[mach->BreakStackTop];
+
          assert(mach->FuncStackTop > 0);
          mach->FuncMask = mach->FuncStack[--mach->FuncStackTop];
 
@@ -2821,12 +3025,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_SSG:
-   /* TGSI_OPCODE_SGN */
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_sgn( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_sgn, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_CMP:
@@ -2834,10 +3033,10 @@ exec_instruction(
          FETCH(&r[0], 0, chan_index);
          FETCH(&r[1], 1, chan_index);
          FETCH(&r[2], 2, chan_index);
-
-         micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] );
-
-         STORE(&r[0], 0, chan_index);
+         micro_lt(&d[chan_index], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2]);
+      }
+      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
+         STORE(&d[chan_index], 0, chan_index);
       }
       break;
 
@@ -2933,18 +3132,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_DP2:
-      FETCH( &r[0], 0, CHAN_X );
-      FETCH( &r[1], 1, CHAN_X );
-      micro_mul( &r[0], &r[0], &r[1] );
-
-      FETCH( &r[1], 0, CHAN_Y );
-      FETCH( &r[2], 1, CHAN_Y );
-      micro_mul( &r[1], &r[1], &r[2] );
-      micro_add( &r[0], &r[0], &r[1] );
-
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_dp2(mach, inst);
       break;
 
    case TGSI_OPCODE_IF:
@@ -3010,71 +3198,31 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_CEIL:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_ceil( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_ceil, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_I2F:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_i2f( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_i2f, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_INT);
       break;
 
    case TGSI_OPCODE_NOT:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_not( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_not, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_TRUNC:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_trunc( &r[0], &r[0] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_unary(mach, inst, micro_trunc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SHL:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_shl( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
-      }
-      break;
-
-   case TGSI_OPCODE_SHR:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_ishr( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_shl, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_AND:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_and( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_and, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_OR:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_or( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_or, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_MOD:
@@ -3082,12 +3230,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_XOR:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_xor( &r[0], &r[0], &r[1] );
-         STORE( &r[0], 0, chan_index );
-      }
+      exec_vector_binary(mach, inst, micro_xor, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_SAD:
@@ -3103,13 +3246,11 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_EMIT:
-      mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16;
-      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++;
+      emit_vertex(mach);
       break;
 
    case TGSI_OPCODE_ENDPRIM:
-      mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++;
-      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0;
+      emit_primitive(mach);
       break;
 
    case TGSI_OPCODE_BGNFOR:
@@ -3117,43 +3258,62 @@ exec_instruction(
       for (chan_index = 0; chan_index < 3; chan_index++) {
          FETCH( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[chan_index], 0, chan_index );
       }
-      STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X );
       ++mach->LoopCounterStackTop;
+      STORE(&mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], 0, CHAN_X);
+      /* update LoopMask */
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[0] <= 0.0f) {
+         mach->LoopMask &= ~0x1;
+      }
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[1] <= 0.0f) {
+         mach->LoopMask &= ~0x2;
+      }
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[2] <= 0.0f) {
+         mach->LoopMask &= ~0x4;
+      }
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[3] <= 0.0f) {
+         mach->LoopMask &= ~0x8;
+      }
+      /* TODO: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
       /* fall-through (for now) */
    case TGSI_OPCODE_BGNLOOP:
       /* push LoopMask and ContMasks */
       assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
-      mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
       assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
-      mach->ContStack[mach->ContStackTop++] = mach->ContMask;
       assert(mach->LoopLabelStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      assert(mach->BreakStackTop < TGSI_EXEC_MAX_BREAK_STACK);
+
+      mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
+      mach->ContStack[mach->ContStackTop++] = mach->ContMask;
       mach->LoopLabelStack[mach->LoopLabelStackTop++] = *pc - 1;
+      mach->BreakStack[mach->BreakStackTop++] = mach->BreakType;
+      mach->BreakType = TGSI_EXEC_BREAK_INSIDE_LOOP;
       break;
 
    case TGSI_OPCODE_ENDFOR:
       assert(mach->LoopCounterStackTop > 0);
-      micro_sub( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], 
-                 &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X],
-                 &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] );
+      micro_sub(&mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], 
+                &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y],
+                &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]);
       /* update LoopMask */
-      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[0] <= 0) {
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[0] <= 0.0f) {
          mach->LoopMask &= ~0x1;
       }
-      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[1] <= 0 ) {
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[1] <= 0.0f) {
          mach->LoopMask &= ~0x2;
       }
-      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[2] <= 0 ) {
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[2] <= 0.0f) {
          mach->LoopMask &= ~0x4;
       }
-      if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[3] <= 0 ) {
+      if (mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y].f[3] <= 0.0f) {
          mach->LoopMask &= ~0x8;
       }
-      micro_add( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], 
-                 &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], 
-                 &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]);
+      micro_add(&mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], 
+                &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], 
+                &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]);
       assert(mach->LoopLabelStackTop > 0);
       inst = mach->Instructions + mach->LoopLabelStack[mach->LoopLabelStackTop - 1];
-      STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X );
+      STORE(&mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_X], 0, CHAN_X);
       /* Restore ContMask, but don't pop */
       assert(mach->ContStackTop > 0);
       mach->ContMask = mach->ContStack[mach->ContStackTop - 1];
@@ -3174,6 +3334,8 @@ exec_instruction(
          --mach->LoopLabelStackTop;
          assert(mach->LoopCounterStackTop > 0);
          --mach->LoopCounterStackTop;
+
+         mach->BreakType = mach->BreakStack[--mach->BreakStackTop];
       }
       UPDATE_EXEC_MASK(mach);
       break;
@@ -3197,15 +3359,14 @@ exec_instruction(
          mach->ContMask = mach->ContStack[--mach->ContStackTop];
          assert(mach->LoopLabelStackTop > 0);
          --mach->LoopLabelStackTop;
+
+         mach->BreakType = mach->BreakStack[--mach->BreakStackTop];
       }
       UPDATE_EXEC_MASK(mach);
       break;
 
    case TGSI_OPCODE_BRK:
-      /* turn off loop channels for each enabled exec channel */
-      mach->LoopMask &= ~mach->ExecMask;
-      /* Todo: if mach->LoopMask == 0, jump to end of loop */
-      UPDATE_EXEC_MASK(mach);
+      exec_break(mach);
       break;
 
    case TGSI_OPCODE_CONT:
@@ -3220,26 +3381,160 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_ENDSUB:
-      /* no-op */
+      /*
+       * XXX: This really should be a no-op. We should never reach this opcode.
+       */
+
+      assert(mach->CallStackTop > 0);
+      mach->CallStackTop--;
+
+      mach->CondStackTop = mach->CallStack[mach->CallStackTop].CondStackTop;
+      mach->CondMask = mach->CondStack[mach->CondStackTop];
+
+      mach->LoopStackTop = mach->CallStack[mach->CallStackTop].LoopStackTop;
+      mach->LoopMask = mach->LoopStack[mach->LoopStackTop];
+
+      mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop;
+      mach->ContMask = mach->ContStack[mach->ContStackTop];
+
+      mach->SwitchStackTop = mach->CallStack[mach->CallStackTop].SwitchStackTop;
+      mach->Switch = mach->SwitchStack[mach->SwitchStackTop];
+
+      mach->BreakStackTop = mach->CallStack[mach->CallStackTop].BreakStackTop;
+      mach->BreakType = mach->BreakStack[mach->BreakStackTop];
+
+      assert(mach->FuncStackTop > 0);
+      mach->FuncMask = mach->FuncStack[--mach->FuncStackTop];
+
+      *pc = mach->CallStack[mach->CallStackTop].ReturnAddr;
+
+      UPDATE_EXEC_MASK(mach);
       break;
 
-   case TGSI_OPCODE_NOISE1:
-      assert( 0 );
+   case TGSI_OPCODE_NOP:
       break;
 
-   case TGSI_OPCODE_NOISE2:
-      assert( 0 );
+   case TGSI_OPCODE_BREAKC:
+      FETCH(&r[0], 0, CHAN_X);
+      /* update CondMask */
+      if (r[0].u[0] && (mach->ExecMask & 0x1)) {
+         mach->LoopMask &= ~0x1;
+      }
+      if (r[0].u[1] && (mach->ExecMask & 0x2)) {
+         mach->LoopMask &= ~0x2;
+      }
+      if (r[0].u[2] && (mach->ExecMask & 0x4)) {
+         mach->LoopMask &= ~0x4;
+      }
+      if (r[0].u[3] && (mach->ExecMask & 0x8)) {
+         mach->LoopMask &= ~0x8;
+      }
+      /* Todo: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
       break;
 
-   case TGSI_OPCODE_NOISE3:
-      assert( 0 );
+   case TGSI_OPCODE_F2I:
+      exec_vector_unary(mach, inst, micro_f2i, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT);
       break;
 
-   case TGSI_OPCODE_NOISE4:
-      assert( 0 );
+   case TGSI_OPCODE_IDIV:
+      exec_vector_binary(mach, inst, micro_idiv, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
       break;
 
-   case TGSI_OPCODE_NOP:
+   case TGSI_OPCODE_IMAX:
+      exec_vector_binary(mach, inst, micro_imax, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   case TGSI_OPCODE_IMIN:
+      exec_vector_binary(mach, inst, micro_imin, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   case TGSI_OPCODE_INEG:
+      exec_vector_unary(mach, inst, micro_ineg, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   case TGSI_OPCODE_ISGE:
+      exec_vector_binary(mach, inst, micro_isge, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   case TGSI_OPCODE_ISHR:
+      exec_vector_binary(mach, inst, micro_ishr, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   case TGSI_OPCODE_ISLT:
+      exec_vector_binary(mach, inst, micro_islt, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   case TGSI_OPCODE_F2U:
+      exec_vector_unary(mach, inst, micro_f2u, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
+   case TGSI_OPCODE_U2F:
+      exec_vector_unary(mach, inst, micro_u2f, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UADD:
+      exec_vector_binary(mach, inst, micro_uadd, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UDIV:
+      exec_vector_binary(mach, inst, micro_udiv, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UMAD:
+      exec_vector_trinary(mach, inst, micro_umad, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UMAX:
+      exec_vector_binary(mach, inst, micro_umax, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UMIN:
+      exec_vector_binary(mach, inst, micro_umin, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UMOD:
+      exec_vector_binary(mach, inst, micro_umod, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_UMUL:
+      exec_vector_binary(mach, inst, micro_umul, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_USEQ:
+      exec_vector_binary(mach, inst, micro_useq, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_USGE:
+      exec_vector_binary(mach, inst, micro_usge, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_USHR:
+      exec_vector_binary(mach, inst, micro_ushr, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_USLT:
+      exec_vector_binary(mach, inst, micro_uslt, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_USNE:
+      exec_vector_binary(mach, inst, micro_usne, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
+      break;
+
+   case TGSI_OPCODE_SWITCH:
+      exec_switch(mach, inst);
+      break;
+
+   case TGSI_OPCODE_CASE:
+      exec_case(mach, inst);
+      break;
+
+   case TGSI_OPCODE_DEFAULT:
+      exec_default(mach);
+      break;
+
+   case TGSI_OPCODE_ENDSWITCH:
+      exec_endswitch(mach);
       break;
 
    default:
@@ -3248,6 +3543,9 @@ exec_instruction(
 }
 
 
+#define DEBUG_EXECUTION 0
+
+
 /**
  * Run TGSI interpreter.
  * \return bitmask of "alive" quad components
@@ -3264,9 +3562,13 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
    mach->FuncMask = 0xf;
    mach->ExecMask = 0xf;
 
+   mach->Switch.mask = 0xf;
+
    assert(mach->CondStackTop == 0);
    assert(mach->LoopStackTop == 0);
    assert(mach->ContStackTop == 0);
+   assert(mach->SwitchStackTop == 0);
+   assert(mach->BreakStackTop == 0);
    assert(mach->CallStackTop == 0);
 
    mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0;
@@ -3290,10 +3592,67 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
       exec_declaration( mach, mach->Declarations+i );
    }
 
-   /* execute instructions, until pc is set to -1 */
-   while (pc != -1) {
-      assert(pc < (int) mach->NumInstructions);
-      exec_instruction( mach, mach->Instructions + pc, &pc );
+   {
+#if DEBUG_EXECUTION
+      struct tgsi_exec_vector temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_TEMP_EXTRAS];
+      struct tgsi_exec_vector outputs[PIPE_MAX_ATTRIBS];
+      uint inst = 1;
+
+      memcpy(temps, mach->Temps, sizeof(temps));
+      memcpy(outputs, mach->Outputs, sizeof(outputs));
+#endif
+
+      /* execute instructions, until pc is set to -1 */
+      while (pc != -1) {
+
+#if DEBUG_EXECUTION
+         uint i;
+
+         tgsi_dump_instruction(&mach->Instructions[pc], inst++);
+#endif
+
+         assert(pc < (int) mach->NumInstructions);
+         exec_instruction(mach, mach->Instructions + pc, &pc);
+
+#if DEBUG_EXECUTION
+         for (i = 0; i < TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_TEMP_EXTRAS; i++) {
+            if (memcmp(&temps[i], &mach->Temps[i], sizeof(temps[i]))) {
+               uint j;
+
+               memcpy(&temps[i], &mach->Temps[i], sizeof(temps[i]));
+               debug_printf("TEMP[%2u] = ", i);
+               for (j = 0; j < 4; j++) {
+                  if (j > 0) {
+                     debug_printf("           ");
+                  }
+                  debug_printf("(%6f %u, %6f %u, %6f %u, %6f %u)\n",
+                               temps[i].xyzw[0].f[j], temps[i].xyzw[0].u[j],
+                               temps[i].xyzw[1].f[j], temps[i].xyzw[1].u[j],
+                               temps[i].xyzw[2].f[j], temps[i].xyzw[2].u[j],
+                               temps[i].xyzw[3].f[j], temps[i].xyzw[3].u[j]);
+               }
+            }
+         }
+         for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+            if (memcmp(&outputs[i], &mach->Outputs[i], sizeof(outputs[i]))) {
+               uint j;
+
+               memcpy(&outputs[i], &mach->Outputs[i], sizeof(outputs[i]));
+               debug_printf("OUT[%2u] =  ", i);
+               for (j = 0; j < 4; j++) {
+                  if (j > 0) {
+                     debug_printf("           ");
+                  }
+                  debug_printf("(%6f %u, %6f %u, %6f %u, %6f %u)\n",
+                               outputs[i].xyzw[0].f[j], outputs[i].xyzw[0].u[j],
+                               outputs[i].xyzw[1].f[j], outputs[i].xyzw[1].u[j],
+                               outputs[i].xyzw[2].f[j], outputs[i].xyzw[2].u[j],
+                               outputs[i].xyzw[3].f[j], outputs[i].xyzw[3].u[j]);
+               }
+            }
+         }
+#endif
+      }
    }
 
 #if 0
@@ -3307,5 +3666,12 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
    }
 #endif
 
+   assert(mach->CondStackTop == 0);
+   assert(mach->LoopStackTop == 0);
+   assert(mach->ContStackTop == 0);
+   assert(mach->SwitchStackTop == 0);
+   assert(mach->BreakStackTop == 0);
+   assert(mach->CallStackTop == 0);
+
    return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0];
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index c72f76809d..59e3b445cc 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -2,6 +2,7 @@
  * 
  * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
+ * Copyright 2009-2010 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
@@ -35,11 +36,13 @@
 extern "C" {
 #endif
 
+
 #define MAX_LABELS (4 * 1024)  /**< basically, max instructions */
 
 #define NUM_CHANNELS 4  /* R,G,B,A */
 #define QUAD_SIZE    4  /* 4 pixel/quad */
 
+
 /**
   * Registers may be treated as float, signed int or unsigned int.
   */
@@ -69,6 +72,11 @@ struct tgsi_interp_coef
    float dady[NUM_CHANNELS];
 };
 
+enum tgsi_sampler_control {
+   tgsi_sampler_lod_bias,
+   tgsi_sampler_lod_explicit
+};
+
 /**
  * Information for sampling textures, which must be implemented
  * by code outside the TGSI executor.
@@ -80,7 +88,8 @@ struct tgsi_sampler
                        const float s[QUAD_SIZE],
                        const float t[QUAD_SIZE],
                        const float p[QUAD_SIZE],
-                       float lodbias,
+                       const float c0[QUAD_SIZE],
+                       enum tgsi_sampler_control control,
                        float rgba[NUM_CHANNELS][QUAD_SIZE]);
 };
 
@@ -168,13 +177,19 @@ struct tgsi_exec_labels
 
 #define TGSI_EXEC_TEMP_ADDR         (TGSI_EXEC_NUM_TEMPS + 8)
 #define TGSI_EXEC_NUM_ADDRS         1
-#define TGSI_EXEC_NUM_TEMP_EXTRAS   9
 
+/* predicate register */
+#define TGSI_EXEC_TEMP_P0           (TGSI_EXEC_NUM_TEMPS + 9)
+#define TGSI_EXEC_NUM_PREDS         1
+
+#define TGSI_EXEC_NUM_TEMP_EXTRAS   10
 
 
-#define TGSI_EXEC_MAX_COND_NESTING  20
-#define TGSI_EXEC_MAX_LOOP_NESTING  20
-#define TGSI_EXEC_MAX_CALL_NESTING  20
+
+#define TGSI_EXEC_MAX_COND_NESTING  32
+#define TGSI_EXEC_MAX_LOOP_NESTING  32
+#define TGSI_EXEC_MAX_SWITCH_NESTING 32
+#define TGSI_EXEC_MAX_CALL_NESTING  32
 
 /* The maximum number of input attributes per vertex. For 2D
  * input register files, this is the stride between two 1D
@@ -186,6 +201,14 @@ struct tgsi_exec_labels
  */
 #define TGSI_EXEC_MAX_CONST_BUFFER  4096
 
+/* The maximum number of vertices per primitive */
+#define TGSI_MAX_PRIM_VERTICES 6
+
+/* The maximum number of primitives to be generated */
+#define TGSI_MAX_PRIMITIVES 64
+
+/* The maximum total number of vertices */
+#define TGSI_MAX_TOTAL_VERTICES (TGSI_MAX_PRIM_VERTICES * TGSI_MAX_PRIMITIVES * PIPE_MAX_ATTRIBS)
 
 /** function call/activation record */
 struct tgsi_call_record
@@ -193,10 +216,29 @@ struct tgsi_call_record
    uint CondStackTop;
    uint LoopStackTop;
    uint ContStackTop;
+   int SwitchStackTop;
+   int BreakStackTop;
    uint ReturnAddr;
 };
 
 
+/* Switch-case block state. */
+struct tgsi_switch_record {
+   uint mask;                          /**< execution mask */
+   union tgsi_exec_channel selector;   /**< a value case statements are compared to */
+   uint defaultMask;                   /**< non-execute mask for default case */
+};
+
+
+enum tgsi_break_type {
+   TGSI_EXEC_BREAK_INSIDE_LOOP,
+   TGSI_EXEC_BREAK_INSIDE_SWITCH
+};
+
+
+#define TGSI_EXEC_MAX_BREAK_STACK (TGSI_EXEC_MAX_LOOP_NESTING + TGSI_EXEC_MAX_SWITCH_NESTING)
+
+
 /**
  * Run-time virtual machine state for executing TGSI shader.
  */
@@ -209,10 +251,11 @@ struct tgsi_exec_machine
 
    float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
 
-   struct tgsi_exec_vector       Inputs[PIPE_MAX_ATTRIBS];
-   struct tgsi_exec_vector       Outputs[PIPE_MAX_ATTRIBS];
+   struct tgsi_exec_vector       Inputs[TGSI_MAX_PRIM_VERTICES * PIPE_MAX_ATTRIBS];
+   struct tgsi_exec_vector       Outputs[TGSI_MAX_TOTAL_VERTICES];
 
    struct tgsi_exec_vector       *Addrs;
+   struct tgsi_exec_vector       *Predicates;
 
    struct tgsi_sampler           **Samplers;
 
@@ -223,10 +266,13 @@ struct tgsi_exec_machine
 
    /* GEOMETRY processor only. */
    unsigned                      *Primitives;
+   unsigned                       NumOutputs;
+   unsigned                       MaxGeometryShaderOutputs;
 
    /* FRAGMENT processor only. */
    const struct tgsi_interp_coef *InterpCoefs;
    struct tgsi_exec_vector       QuadPos;
+   float                         Face;    /**< +1 if front facing, -1 if back facing */
 
    /* Conditional execution masks */
    uint CondMask;  /**< For IF/ELSE/ENDIF */
@@ -235,6 +281,12 @@ struct tgsi_exec_machine
    uint FuncMask;  /**< For function calls */
    uint ExecMask;  /**< = CondMask & LoopMask */
 
+   /* Current switch-case state. */
+   struct tgsi_switch_record Switch;
+
+   /* Current break type. */
+   enum tgsi_break_type BreakType;
+
    /** Condition mask stack (for nested conditionals) */
    uint CondStack[TGSI_EXEC_MAX_COND_NESTING];
    int CondStackTop;
@@ -247,7 +299,7 @@ struct tgsi_exec_machine
    uint LoopLabelStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int LoopLabelStackTop;
 
-   /** Loop counter stack (x = count, y = current, z = step) */
+   /** Loop counter stack (x = index, y = counter, z = step) */
    struct tgsi_exec_vector LoopCounterStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int LoopCounterStackTop;
    
@@ -255,6 +307,13 @@ struct tgsi_exec_machine
    uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int ContStackTop;
 
+   /** Switch case stack */
+   struct tgsi_switch_record SwitchStack[TGSI_EXEC_MAX_SWITCH_NESTING];
+   int SwitchStackTop;
+
+   enum tgsi_break_type BreakStack[TGSI_EXEC_MAX_BREAK_STACK];
+   int BreakStackTop;
+
    /** Function execution mask stack (for executing subroutine code) */
    uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING];
    int FuncStackTop;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c
index 17af4cb7ad..de0e09cdba 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_info.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_info.c
@@ -119,7 +119,7 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
    { 1, 1, 0, 0, 0, 0, "NOT", TGSI_OPCODE_NOT },
    { 1, 1, 0, 0, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC },
    { 1, 2, 0, 0, 0, 0, "SHL", TGSI_OPCODE_SHL },
-   { 1, 2, 0, 0, 0, 0, "SHR", TGSI_OPCODE_SHR },
+   { 0, 0, 0, 0, 0, 0, "", 88 },      /* removed */
    { 1, 2, 0, 0, 0, 0, "AND", TGSI_OPCODE_AND },
    { 1, 2, 0, 0, 0, 0, "OR", TGSI_OPCODE_OR },
    { 1, 2, 0, 0, 0, 0, "MOD", TGSI_OPCODE_MOD },
@@ -134,10 +134,10 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
    { 0, 0, 0, 0, 0, 1, "BGNSUB", TGSI_OPCODE_BGNSUB },
    { 0, 0, 0, 1, 1, 0, "ENDLOOP", TGSI_OPCODE_ENDLOOP },
    { 0, 0, 0, 0, 1, 0, "ENDSUB", TGSI_OPCODE_ENDSUB },
-   { 1, 1, 0, 0, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 },
-   { 1, 1, 0, 0, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 },
-   { 1, 1, 0, 0, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 },
-   { 1, 1, 0, 0, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 },
+   { 0, 0, 0, 0, 0, 0, "", 103 },     /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 104 },     /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 105 },     /* removed */
+   { 0, 0, 0, 0, 0, 0, "", 106 },     /* removed */
    { 0, 0, 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP },
    { 0, 0, 0, 0, 0, 0, "", 108 },     /* removed */
    { 0, 0, 0, 0, 0, 0, "", 109 },     /* removed */
@@ -149,7 +149,33 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
    { 0, 1, 0, 0, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC },
    { 0, 1, 0, 0, 0, 0, "KIL", TGSI_OPCODE_KIL },
    { 0, 0, 0, 0, 0, 0, "END", TGSI_OPCODE_END },
-   { 1, 1, 0, 0, 0, 0, "SWZ", TGSI_OPCODE_SWZ }
+   { 0, 0, 0, 0, 0, 0, "", 118 },     /* removed */
+   { 1, 1, 0, 0, 0, 0, "F2I", TGSI_OPCODE_F2I },
+   { 1, 2, 0, 0, 0, 0, "IDIV", TGSI_OPCODE_IDIV },
+   { 1, 2, 0, 0, 0, 0, "IMAX", TGSI_OPCODE_IMAX },
+   { 1, 2, 0, 0, 0, 0, "IMIN", TGSI_OPCODE_IMIN },
+   { 1, 1, 0, 0, 0, 0, "INEG", TGSI_OPCODE_INEG },
+   { 1, 2, 0, 0, 0, 0, "ISGE", TGSI_OPCODE_ISGE },
+   { 1, 2, 0, 0, 0, 0, "ISHR", TGSI_OPCODE_ISHR },
+   { 1, 2, 0, 0, 0, 0, "ISLT", TGSI_OPCODE_ISLT },
+   { 1, 1, 0, 0, 0, 0, "F2U", TGSI_OPCODE_F2U },
+   { 1, 1, 0, 0, 0, 0, "U2F", TGSI_OPCODE_U2F },
+   { 1, 2, 0, 0, 0, 0, "UADD", TGSI_OPCODE_UADD },
+   { 1, 2, 0, 0, 0, 0, "UDIV", TGSI_OPCODE_UDIV },
+   { 1, 3, 0, 0, 0, 0, "UMAD", TGSI_OPCODE_UMAD },
+   { 1, 2, 0, 0, 0, 0, "UMAX", TGSI_OPCODE_UMAX },
+   { 1, 2, 0, 0, 0, 0, "UMIN", TGSI_OPCODE_UMIN },
+   { 1, 2, 0, 0, 0, 0, "UMOD", TGSI_OPCODE_UMOD },
+   { 1, 2, 0, 0, 0, 0, "UMUL", TGSI_OPCODE_UMUL },
+   { 1, 2, 0, 0, 0, 0, "USEQ", TGSI_OPCODE_USEQ },
+   { 1, 2, 0, 0, 0, 0, "USGE", TGSI_OPCODE_USGE },
+   { 1, 2, 0, 0, 0, 0, "USHR", TGSI_OPCODE_USHR },
+   { 1, 2, 0, 0, 0, 0, "USLT", TGSI_OPCODE_USLT },
+   { 1, 2, 0, 0, 0, 0, "USNE", TGSI_OPCODE_USNE },
+   { 0, 1, 0, 0, 0, 0, "SWITCH", TGSI_OPCODE_SWITCH },
+   { 0, 1, 0, 0, 0, 0, "CASE", TGSI_OPCODE_CASE },
+   { 0, 0, 0, 0, 0, 0, "DEFAULT", TGSI_OPCODE_DEFAULT },
+   { 0, 0, 0, 0, 0, 0, "ENDSWITCH", TGSI_OPCODE_ENDSWITCH }
 };
 
 const struct tgsi_opcode_info *
diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/tgsi_iterate.c
index d88c2558d8..0ba5fe4841 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_iterate.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.c
@@ -39,7 +39,6 @@ tgsi_iterate_shader(
       return FALSE;
 
    ctx->processor = parse.FullHeader.Processor;
-   ctx->version = parse.FullVersion.Version;
 
    if (ctx->prolog)
       if (!ctx->prolog( ctx ))
@@ -67,6 +66,12 @@ tgsi_iterate_shader(
                goto fail;
          break;
 
+      case TGSI_TOKEN_TYPE_PROPERTY:
+         if (ctx->iterate_property)
+            if (!ctx->iterate_property( ctx,  &parse.FullToken.FullProperty ))
+               goto fail;
+         break;
+
       default:
          assert( 0 );
       }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/tgsi_iterate.h
index ec7b85bf63..8d67f22c42 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_iterate.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.h
@@ -57,11 +57,15 @@ struct tgsi_iterate_context
       struct tgsi_full_immediate *imm );
 
    boolean
+   (* iterate_property)(
+      struct tgsi_iterate_context *ctx,
+      struct tgsi_full_property *prop );
+
+   boolean
    (* epilog)(
       struct tgsi_iterate_context *ctx );
 
    struct tgsi_processor processor;
-   struct tgsi_version version;
 };
 
 boolean
diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
index e7bcf4bf75..e4af15c156 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
@@ -124,7 +124,6 @@ OP11(I2F)
 OP11(NOT)
 OP11(TRUNC)
 OP12(SHL)
-OP12(SHR)
 OP12(AND)
 OP12(OR)
 OP12(MOD)
@@ -139,10 +138,6 @@ OP00_LBL(BGNLOOP)
 OP00(BGNSUB)
 OP00_LBL(ENDLOOP)
 OP00(ENDSUB)
-OP11(NOISE1)
-OP11(NOISE2)
-OP11(NOISE3)
-OP11(NOISE4)
 OP00(NOP)
 OP11(NRM4)
 OP01(CALLNZ)
@@ -150,7 +145,28 @@ OP01(IFC)
 OP01(BREAKC)
 OP01(KIL)
 OP00(END)
-OP11(SWZ)
+OP11(F2I)
+OP12(IDIV)
+OP12(IMAX)
+OP12(IMIN)
+OP11(INEG)
+OP12(ISGE)
+OP12(ISHR)
+OP12(ISLT)
+OP11(F2U)
+OP11(U2F)
+OP12(UADD)
+OP12(UDIV)
+OP13(UMAD)
+OP12(UMAX)
+OP12(UMIN)
+OP12(UMOD)
+OP12(UMUL)
+OP12(USEQ)
+OP12(USGE)
+OP12(USHR)
+OP12(USLT)
+OP12(USNE)
 
 
 #undef OP00
diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c
index 4870f82b6b..8c7062d850 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_parse.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c
@@ -28,44 +28,23 @@
 #include "util/u_debug.h"
 #include "pipe/p_shader_tokens.h"
 #include "tgsi_parse.h"
-#include "tgsi_build.h"
 #include "util/u_memory.h"
 
-void
-tgsi_full_token_init(
-   union tgsi_full_token *full_token )
-{
-   full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION;
-}
-
-void
-tgsi_full_token_free(
-   union tgsi_full_token *full_token )
-{
-}
-
 unsigned
 tgsi_parse_init(
    struct tgsi_parse_context *ctx,
    const struct tgsi_token *tokens )
 {
-   ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0];
-   if( ctx->FullVersion.Version.MajorVersion > 1 ) {
-      return TGSI_PARSE_ERROR;
-   }
-
-   ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1];
+   ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[0];
    if( ctx->FullHeader.Header.HeaderSize >= 2 ) {
-      ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2];
+      ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[1];
    }
    else {
-      ctx->FullHeader.Processor = tgsi_default_processor();
+      return TGSI_PARSE_ERROR;
    }
 
    ctx->Tokens = tokens;
-   ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize;
-
-   tgsi_full_token_init( &ctx->FullToken );
+   ctx->Position = ctx->FullHeader.Header.HeaderSize;
 
    return TGSI_PARSE_OK;
 }
@@ -74,7 +53,6 @@ void
 tgsi_parse_free(
    struct tgsi_parse_context *ctx )
 {
-   tgsi_full_token_free( &ctx->FullToken );
 }
 
 boolean
@@ -82,7 +60,7 @@ tgsi_parse_end_of_tokens(
    struct tgsi_parse_context *ctx )
 {
    return ctx->Position >=
-      1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize;
+      ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize;
 }
 
 
@@ -119,9 +97,6 @@ tgsi_parse_token(
    struct tgsi_token token;
    unsigned i;
 
-   tgsi_full_token_free( &ctx->FullToken );
-   tgsi_full_token_init( &ctx->FullToken );
-
    next_token( ctx, &token );
 
    switch( token.Type ) {
@@ -129,10 +104,10 @@ tgsi_parse_token(
    {
       struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration;
 
-      *decl = tgsi_default_full_declaration();
+      memset(decl, 0, sizeof *decl);
       copy_token(&decl->Declaration, &token);
 
-      next_token( ctx, &decl->DeclarationRange );
+      next_token( ctx, &decl->Range );
 
       if( decl->Declaration.Semantic ) {
          next_token( ctx, &decl->Semantic );
@@ -144,18 +119,29 @@ tgsi_parse_token(
    case TGSI_TOKEN_TYPE_IMMEDIATE:
    {
       struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate;
+      uint imm_count;
 
-      *imm = tgsi_default_full_immediate();
+      memset(imm, 0, sizeof *imm);
       copy_token(&imm->Immediate, &token);
-      assert( !imm->Immediate.Extended );
+
+      imm_count = imm->Immediate.NrTokens - 1;
 
       switch (imm->Immediate.DataType) {
       case TGSI_IMM_FLOAT32:
-         {
-            uint imm_count = imm->Immediate.NrTokens - 1;
-            for (i = 0; i < imm_count; i++) {
-               next_token(ctx, &imm->u[i]);
-            }
+         for (i = 0; i < imm_count; i++) {
+            next_token(ctx, &imm->u[i].Float);
+         }
+         break;
+
+      case TGSI_IMM_UINT32:
+         for (i = 0; i < imm_count; i++) {
+            next_token(ctx, &imm->u[i].Uint);
+         }
+         break;
+
+      case TGSI_IMM_INT32:
+         for (i = 0; i < imm_count; i++) {
+            next_token(ctx, &imm->u[i].Int);
          }
          break;
 
@@ -169,147 +155,76 @@ tgsi_parse_token(
    case TGSI_TOKEN_TYPE_INSTRUCTION:
    {
       struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction;
-      unsigned extended;
 
-      *inst = tgsi_default_full_instruction();
+      memset(inst, 0, sizeof *inst);
       copy_token(&inst->Instruction, &token);
-      extended = inst->Instruction.Extended;
-
-      while( extended ) {
-         struct tgsi_src_register_ext token;
-
-         next_token( ctx, &token );
-
-         switch( token.Type ) {
-         case TGSI_INSTRUCTION_EXT_TYPE_NV:
-            copy_token(&inst->InstructionExtNv, &token);
-            break;
 
-         case TGSI_INSTRUCTION_EXT_TYPE_LABEL:
-            copy_token(&inst->InstructionExtLabel, &token);
-            break;
-
-         case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE:
-            copy_token(&inst->InstructionExtTexture, &token);
-            break;
+      if (inst->Instruction.Predicate) {
+         next_token(ctx, &inst->Predicate);
+      }
 
-         default:
-            assert( 0 );
-         }
+      if (inst->Instruction.Label) {
+         next_token( ctx, &inst->Label);
+      }
 
-         extended = token.Extended;
+      if (inst->Instruction.Texture) {
+         next_token( ctx, &inst->Texture);
       }
 
       assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS );
 
       for(  i = 0; i < inst->Instruction.NumDstRegs; i++ ) {
-         unsigned extended;
 
-         next_token( ctx, &inst->FullDstRegisters[i].DstRegister );
+         next_token( ctx, &inst->Dst[i].Register );
 
          /*
           * No support for indirect or multi-dimensional addressing.
           */
-         assert( !inst->FullDstRegisters[i].DstRegister.Dimension );
-
-         extended = inst->FullDstRegisters[i].DstRegister.Extended;
-
-         while( extended ) {
-            struct tgsi_src_register_ext token;
-
-            next_token( ctx, &token );
-
-            switch( token.Type ) {
-            case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE:
-               copy_token(&inst->FullDstRegisters[i].DstRegisterExtConcode,
-                          &token);
-               break;
-
-            case TGSI_DST_REGISTER_EXT_TYPE_MODULATE:
-               copy_token(&inst->FullDstRegisters[i].DstRegisterExtModulate,
-                          &token);
-               break;
-
-            default:
-               assert( 0 );
-            }
-
-            extended = token.Extended;
-         }
+         assert( !inst->Dst[i].Register.Dimension );
 
-         if( inst->FullDstRegisters[i].DstRegister.Indirect ) {
-            next_token( ctx, &inst->FullDstRegisters[i].DstRegisterInd );
+         if( inst->Dst[i].Register.Indirect ) {
+            next_token( ctx, &inst->Dst[i].Indirect );
 
             /*
              * No support for indirect or multi-dimensional addressing.
              */
-            assert( !inst->FullDstRegisters[i].DstRegisterInd.Indirect );
-            assert( !inst->FullDstRegisters[i].DstRegisterInd.Dimension );
-            assert( !inst->FullDstRegisters[i].DstRegisterInd.Extended );
+            assert( !inst->Dst[i].Indirect.Dimension );
+            assert( !inst->Dst[i].Indirect.Indirect );
          }
       }
 
       assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS );
 
       for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) {
-         unsigned extended;
-
-         next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister );
-
-         extended = inst->FullSrcRegisters[i].SrcRegister.Extended;
-
-         while( extended ) {
-            struct tgsi_src_register_ext token;
-
-            next_token( ctx, &token );
 
-            switch( token.Type ) {
-            case TGSI_SRC_REGISTER_EXT_TYPE_SWZ:
-               copy_token(&inst->FullSrcRegisters[i].SrcRegisterExtSwz,
-                          &token);
-               break;
+         next_token( ctx, &inst->Src[i].Register );
 
-            case TGSI_SRC_REGISTER_EXT_TYPE_MOD:
-               copy_token(&inst->FullSrcRegisters[i].SrcRegisterExtMod,
-                          &token);
-               break;
-
-            default:
-               assert( 0 );
-            }
-
-            extended = token.Extended;
-         }
-
-         if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) {
-            next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd );
+         if( inst->Src[i].Register.Indirect ) {
+            next_token( ctx, &inst->Src[i].Indirect );
 
             /*
              * No support for indirect or multi-dimensional addressing.
              */
-            assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect );
-            assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension );
-            assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended );
+            assert( !inst->Src[i].Indirect.Indirect );
+            assert( !inst->Src[i].Indirect.Dimension );
          }
 
-         if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) {
-            next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim );
+         if( inst->Src[i].Register.Dimension ) {
+            next_token( ctx, &inst->Src[i].Dimension );
 
             /*
              * No support for multi-dimensional addressing.
              */
-            assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension );
-            assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended );
+            assert( !inst->Src[i].Dimension.Dimension );
 
-            if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) {
-               next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd );
+            if( inst->Src[i].Dimension.Indirect ) {
+               next_token( ctx, &inst->Src[i].DimIndirect );
 
                /*
                * No support for indirect or multi-dimensional addressing.
                */
-               assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect );
-               assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension );
-               assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended );
+               assert( !inst->Src[i].Indirect.Indirect );
+               assert( !inst->Src[i].Indirect.Dimension );
             }
          }
       }
@@ -317,6 +232,22 @@ tgsi_parse_token(
       break;
    }
 
+   case TGSI_TOKEN_TYPE_PROPERTY:
+   {
+      struct tgsi_full_property *prop = &ctx->FullToken.FullProperty;
+      uint prop_count;
+
+      memset(prop, 0, sizeof *prop);
+      copy_token(&prop->Property, &token);
+
+      prop_count = prop->Property.NrTokens - 1;
+      for (i = 0; i < prop_count; i++) {
+         next_token(ctx, &prop->u[i]);
+      }
+
+      break;
+   }
+
    default:
       assert( 0 );
    }
@@ -329,8 +260,7 @@ tgsi_num_tokens(const struct tgsi_token *tokens)
    struct tgsi_parse_context ctx;
    if (tgsi_parse_init(&ctx, tokens) == TGSI_PARSE_OK) {
       unsigned len = (ctx.FullHeader.Header.HeaderSize +
-                      ctx.FullHeader.Header.BodySize +
-                      1);
+                      ctx.FullHeader.Header.BodySize);
       return len;
    }
    return 0;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h
index a26ee5ba86..439a57269b 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_parse.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h
@@ -34,11 +34,6 @@
 extern "C" {
 #endif
 
-struct tgsi_full_version
-{
-   struct tgsi_version  Version;
-};
-
 struct tgsi_full_header
 {
    struct tgsi_header      Header;
@@ -47,26 +42,22 @@ struct tgsi_full_header
 
 struct tgsi_full_dst_register
 {
-   struct tgsi_dst_register               DstRegister;
-   struct tgsi_src_register               DstRegisterInd;
-   struct tgsi_dst_register_ext_concode   DstRegisterExtConcode;
-   struct tgsi_dst_register_ext_modulate  DstRegisterExtModulate;
+   struct tgsi_dst_register               Register;
+   struct tgsi_src_register               Indirect;
 };
 
 struct tgsi_full_src_register
 {
-   struct tgsi_src_register         SrcRegister;
-   struct tgsi_src_register_ext_swz SrcRegisterExtSwz;
-   struct tgsi_src_register_ext_mod SrcRegisterExtMod;
-   struct tgsi_src_register         SrcRegisterInd;
-   struct tgsi_dimension            SrcRegisterDim;
-   struct tgsi_src_register         SrcRegisterDimInd;
+   struct tgsi_src_register         Register;
+   struct tgsi_src_register         Indirect;
+   struct tgsi_dimension            Dimension;
+   struct tgsi_src_register         DimIndirect;
 };
 
 struct tgsi_full_declaration
 {
    struct tgsi_declaration Declaration;
-   struct tgsi_declaration_range DeclarationRange;
+   struct tgsi_declaration_range Range;
    struct tgsi_declaration_semantic Semantic;
 };
 
@@ -76,18 +67,23 @@ struct tgsi_full_immediate
    union tgsi_immediate_data u[4];
 };
 
+struct tgsi_full_property
+{
+   struct tgsi_property   Property;
+   struct tgsi_property_data u[8];
+};
+
 #define TGSI_FULL_MAX_DST_REGISTERS 2
 #define TGSI_FULL_MAX_SRC_REGISTERS 4 /* TXD has 4 */
 
 struct tgsi_full_instruction
 {
    struct tgsi_instruction             Instruction;
-   struct tgsi_instruction_ext_nv      InstructionExtNv;
-   struct tgsi_instruction_ext_label   InstructionExtLabel;
-   struct tgsi_instruction_ext_texture InstructionExtTexture;
-   struct tgsi_full_dst_register       FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS];
-   struct tgsi_full_src_register       FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS];
-   uint Flags;  /**< user-defined usage */
+   struct tgsi_instruction_predicate   Predicate;
+   struct tgsi_instruction_label       Label;
+   struct tgsi_instruction_texture     Texture;
+   struct tgsi_full_dst_register       Dst[TGSI_FULL_MAX_DST_REGISTERS];
+   struct tgsi_full_src_register       Src[TGSI_FULL_MAX_SRC_REGISTERS];
 };
 
 union tgsi_full_token
@@ -96,21 +92,13 @@ union tgsi_full_token
    struct tgsi_full_declaration  FullDeclaration;
    struct tgsi_full_immediate    FullImmediate;
    struct tgsi_full_instruction  FullInstruction;
+   struct tgsi_full_property     FullProperty;
 };
 
-void
-tgsi_full_token_init(
-   union tgsi_full_token *full_token );
-
-void
-tgsi_full_token_free(
-   union tgsi_full_token *full_token );
-
 struct tgsi_parse_context
 {
    const struct tgsi_token    *Tokens;
    unsigned                   Position;
-   struct tgsi_full_version   FullVersion;
    struct tgsi_full_header    FullHeader;
    union tgsi_full_token      FullToken;
 };
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
index 4b1c7d4e01..138d2d095b 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
@@ -60,7 +60,7 @@ const float ppc_builtin_constants[] ALIGN16_ATTRIB = {
    for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)
 
 #define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
-   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+   ((INST).Dst[0].Register.WriteMask & (1 << (CHAN)))
 
 #define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
    if (IS_DST0_CHANNEL_ENABLED( INST, CHAN ))
@@ -156,8 +156,8 @@ init_gen_context(struct gen_context *gen, struct ppc_function *func)
 static boolean
 is_ppc_vec_temporary(const struct tgsi_full_src_register *reg)
 {
-   return (reg->SrcRegister.File == TGSI_FILE_TEMPORARY &&
-           reg->SrcRegister.Index < MAX_PPC_TEMPS);
+   return (reg->Register.File == TGSI_FILE_TEMPORARY &&
+           reg->Register.Index < MAX_PPC_TEMPS);
 }
 
 
@@ -167,8 +167,8 @@ is_ppc_vec_temporary(const struct tgsi_full_src_register *reg)
 static boolean
 is_ppc_vec_temporary_dst(const struct tgsi_full_dst_register *reg)
 {
-   return (reg->DstRegister.File == TGSI_FILE_TEMPORARY &&
-           reg->DstRegister.Index < MAX_PPC_TEMPS);
+   return (reg->Register.File == TGSI_FILE_TEMPORARY &&
+           reg->Register.Index < MAX_PPC_TEMPS);
 }
 
 
@@ -283,18 +283,19 @@ emit_fetch(struct gen_context *gen,
            const struct tgsi_full_src_register *reg,
            const unsigned chan_index)
 {
-   uint swizzle = tgsi_util_get_full_src_register_extswizzle(reg, chan_index);
+   uint swizzle = tgsi_util_get_full_src_register_swizzle(reg, chan_index);
    int dst_vec = -1;
 
    switch (swizzle) {
-   case TGSI_EXTSWIZZLE_X:
-   case TGSI_EXTSWIZZLE_Y:
-   case TGSI_EXTSWIZZLE_Z:
-   case TGSI_EXTSWIZZLE_W:
-      switch (reg->SrcRegister.File) {
+   case TGSI_SWIZZLE_X:
+   case TGSI_SWIZZLE_Y:
+   case TGSI_SWIZZLE_Z:
+   case TGSI_SWIZZLE_W:
+      switch (reg->Register.File) {
       case TGSI_FILE_INPUT:
+      case TGSI_FILE_SYSTEM_VALUE:
          {
-            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 16;
+            int offset = (reg->Register.Index * 4 + swizzle) * 16;
             int offset_reg = emit_li_offset(gen, offset);
             dst_vec = ppc_allocate_vec_register(gen->f);
             ppc_lvx(gen->f, dst_vec, gen->inputs_reg, offset_reg);
@@ -303,11 +304,11 @@ emit_fetch(struct gen_context *gen,
       case TGSI_FILE_TEMPORARY:
          if (is_ppc_vec_temporary(reg)) {
             /* use PPC vec register */
-            dst_vec = gen->temps_map[reg->SrcRegister.Index][swizzle];
+            dst_vec = gen->temps_map[reg->Register.Index][swizzle];
          }
          else {
             /* use memory-based temp register "file" */
-            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 16;
+            int offset = (reg->Register.Index * 4 + swizzle) * 16;
             int offset_reg = emit_li_offset(gen, offset);
             dst_vec = ppc_allocate_vec_register(gen->f);
             ppc_lvx(gen->f, dst_vec, gen->temps_reg, offset_reg);
@@ -315,7 +316,7 @@ emit_fetch(struct gen_context *gen,
          break;
       case TGSI_FILE_IMMEDIATE:
          {
-            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 4;
+            int offset = (reg->Register.Index * 4 + swizzle) * 4;
             int offset_reg = emit_li_offset(gen, offset);
             dst_vec = ppc_allocate_vec_register(gen->f);
             /* Load 4-byte word into vector register.
@@ -331,7 +332,7 @@ emit_fetch(struct gen_context *gen,
          break;
       case TGSI_FILE_CONSTANT:
          {
-            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 4;
+            int offset = (reg->Register.Index * 4 + swizzle) * 4;
             int offset_reg = emit_li_offset(gen, offset);
             dst_vec = ppc_allocate_vec_register(gen->f);
             /* Load 4-byte word into vector register.
@@ -349,16 +350,6 @@ emit_fetch(struct gen_context *gen,
          assert( 0 );
       }
       break;
-   case TGSI_EXTSWIZZLE_ZERO:
-      ppc_vzero(gen->f, dst_vec);
-      break;
-   case TGSI_EXTSWIZZLE_ONE:
-      {
-         int one_vec = gen_one_vec(gen);
-         dst_vec = ppc_allocate_vec_register(gen->f);
-         ppc_vmove(gen->f, dst_vec, one_vec);
-      }
-      break;
    default:
       assert( 0 );
    }
@@ -414,12 +405,12 @@ equal_src_locs(const struct tgsi_full_src_register *a, uint chan_a,
 {
    int swz_a, swz_b;
    int sign_a, sign_b;
-   if (a->SrcRegister.File != b->SrcRegister.File)
+   if (a->Register.File != b->Register.File)
       return FALSE;
-   if (a->SrcRegister.Index != b->SrcRegister.Index)
+   if (a->Register.Index != b->Register.Index)
       return FALSE;
-   swz_a = tgsi_util_get_full_src_register_extswizzle(a, chan_a);
-   swz_b = tgsi_util_get_full_src_register_extswizzle(b, chan_b);
+   swz_a = tgsi_util_get_full_src_register_swizzle(a, chan_a);
+   swz_b = tgsi_util_get_full_src_register_swizzle(b, chan_b);
    if (swz_a != swz_b)
       return FALSE;
    sign_a = tgsi_util_get_full_src_register_sign_mode(a, chan_a);
@@ -441,7 +432,7 @@ get_src_vec(struct gen_context *gen,
             struct tgsi_full_instruction *inst, int src_reg, uint chan)
 {
    const const struct tgsi_full_src_register *src = 
-      &inst->FullSrcRegisters[src_reg];
+      &inst->Src[src_reg];
    int vec;
    uint i;
 
@@ -492,10 +483,10 @@ get_dst_vec(struct gen_context *gen,
             const struct tgsi_full_instruction *inst,
             unsigned chan_index)
 {
-   const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[0];
+   const struct tgsi_full_dst_register *reg = &inst->Dst[0];
 
    if (is_ppc_vec_temporary_dst(reg)) {
-      int vec = gen->temps_map[reg->DstRegister.Index][chan_index];
+      int vec = gen->temps_map[reg->Register.Index][chan_index];
       return vec;
    }
    else {
@@ -515,12 +506,12 @@ emit_store(struct gen_context *gen,
            unsigned chan_index,
            boolean free_vec)
 {
-   const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[0];
+   const struct tgsi_full_dst_register *reg = &inst->Dst[0];
 
-   switch (reg->DstRegister.File) {
+   switch (reg->Register.File) {
    case TGSI_FILE_OUTPUT:
       {
-         int offset = (reg->DstRegister.Index * 4 + chan_index) * 16;
+         int offset = (reg->Register.Index * 4 + chan_index) * 16;
          int offset_reg = emit_li_offset(gen, offset);
          ppc_stvx(gen->f, src_vec, gen->outputs_reg, offset_reg);
       }
@@ -528,14 +519,14 @@ emit_store(struct gen_context *gen,
    case TGSI_FILE_TEMPORARY:
       if (is_ppc_vec_temporary_dst(reg)) {
          if (!free_vec) {
-            int dst_vec = gen->temps_map[reg->DstRegister.Index][chan_index];
+            int dst_vec = gen->temps_map[reg->Register.Index][chan_index];
             if (dst_vec != src_vec)
                ppc_vmove(gen->f, dst_vec, src_vec);
          }
          free_vec = FALSE;
       }
       else {
-         int offset = (reg->DstRegister.Index * 4 + chan_index) * 16;
+         int offset = (reg->Register.Index * 4 + chan_index) * 16;
          int offset_reg = emit_li_offset(gen, offset);
          ppc_stvx(gen->f, src_vec, gen->temps_reg, offset_reg);
       }
@@ -545,7 +536,7 @@ emit_store(struct gen_context *gen,
       emit_addrs(
          func,
          xmm,
-         reg->DstRegister.Index,
+         reg->Register.Index,
          chan_index );
       break;
 #endif
@@ -635,7 +626,6 @@ emit_unaryop(struct gen_context *gen, struct tgsi_full_instruction *inst)
          ppc_vlogefp(gen->f, v1, v0);      /* v1 = log2(v0) */
          break;
       case TGSI_OPCODE_MOV:
-      case TGSI_OPCODE_SWZ:
          if (v0 != v1)
             ppc_vmove(gen->f, v1, v0);
          break;
@@ -1119,7 +1109,6 @@ emit_instruction(struct gen_context *gen,
 
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_MOV:
-   case TGSI_OPCODE_SWZ:
    case TGSI_OPCODE_ABS:
    case TGSI_OPCODE_FLR:
    case TGSI_OPCODE_FRC:
@@ -1185,13 +1174,14 @@ emit_declaration(
    struct ppc_function *func,
    struct tgsi_full_declaration *decl )
 {
-   if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+   if( decl->Declaration.File == TGSI_FILE_INPUT ||
+       decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE ) {
 #if 0
       unsigned first, last, mask;
       unsigned i, j;
 
-      first = decl->DeclarationRange.First;
-      last = decl->DeclarationRange.Last;
+      first = decl->Range.First;
+      last = decl->Range.Last;
       mask = decl->Declaration.UsageMask;
 
       for( i = first; i <= last; i++ ) {
@@ -1351,6 +1341,9 @@ tgsi_emit_ppc(const struct tgsi_token *tokens,
          }
          break;
 
+      case TGSI_TOKEN_TYPE_PROPERTY:
+         break;
+
       default:
 	 ok = 0;
          assert( 0 );
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index 53e13b30e6..7f1c8e5dd6 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -26,32 +26,112 @@
  **************************************************************************/
 
 #include "util/u_debug.h"
+#include "util/u_memory.h"
+#include "util/u_prim.h"
+#include "cso_cache/cso_hash.h"
 #include "tgsi_sanity.h"
 #include "tgsi_info.h"
 #include "tgsi_iterate.h"
 
-typedef uint reg_flag;
-
-#define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8)
-
-#define MAX_REGISTERS 256
-#define MAX_REG_FLAGS ((MAX_REGISTERS + BITS_IN_REG_FLAG - 1) / BITS_IN_REG_FLAG)
+typedef struct {
+   uint file : 28;
+   /* max 2 dimensions */
+   uint dimensions : 4;
+   uint indices[2];
+} scan_register;
 
 struct sanity_check_ctx
 {
    struct tgsi_iterate_context iter;
+   struct cso_hash *regs_decl;
+   struct cso_hash *regs_used;
+   struct cso_hash *regs_ind_used;
 
-   reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REG_FLAGS];
-   reg_flag regs_used[TGSI_FILE_COUNT][MAX_REG_FLAGS];
-   boolean regs_ind_used[TGSI_FILE_COUNT];
    uint num_imms;
    uint num_instructions;
    uint index_of_END;
 
    uint errors;
    uint warnings;
+   uint implied_array_size;
 };
 
+static INLINE unsigned
+scan_register_key(const scan_register *reg)
+{
+   unsigned key = reg->file;
+   key |= (reg->indices[0] << 4);
+   key |= (reg->indices[1] << 18);
+
+   return key;
+}
+
+static void
+fill_scan_register1d(scan_register *reg,
+                     uint file, uint index)
+{
+   reg->file = file;
+   reg->dimensions = 1;
+   reg->indices[0] = index;
+   reg->indices[1] = 0;
+}
+
+static void
+fill_scan_register2d(scan_register *reg,
+                     uint file, uint index1, uint index2)
+{
+   reg->file = file;
+   reg->dimensions = 2;
+   reg->indices[0] = index1;
+   reg->indices[1] = index2;
+}
+
+static void
+scan_register_dst(scan_register *reg,
+                  struct tgsi_full_dst_register *dst)
+{
+   fill_scan_register1d(reg,
+                        dst->Register.File,
+                        dst->Register.Index);
+}
+
+static void
+scan_register_src(scan_register *reg,
+                  struct tgsi_full_src_register *src)
+{
+   if (src->Register.Dimension) {
+      /*FIXME: right now we don't support indirect
+       * multidimensional addressing */
+      debug_assert(!src->Dimension.Indirect);
+      fill_scan_register2d(reg,
+                           src->Register.File,
+                           src->Register.Index,
+                           src->Dimension.Index);
+   } else {
+      fill_scan_register1d(reg,
+                           src->Register.File,
+                           src->Register.Index);
+   }
+}
+
+static scan_register *
+create_scan_register_src(struct tgsi_full_src_register *src)
+{
+   scan_register *reg = MALLOC(sizeof(scan_register));
+   scan_register_src(reg, src);
+
+   return reg;
+}
+
+static scan_register *
+create_scan_register_dst(struct tgsi_full_dst_register *dst)
+{
+   scan_register *reg = MALLOC(sizeof(scan_register));
+   scan_register_dst(reg, dst);
+
+   return reg;
+}
+
 static void
 report_error(
    struct sanity_check_ctx *ctx,
@@ -99,12 +179,12 @@ check_file_name(
 static boolean
 is_register_declared(
    struct sanity_check_ctx *ctx,
-   uint file,
-   int index )
+   const scan_register *reg)
 {
-   assert( index >= 0 && index < MAX_REGISTERS );
-
-   return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
+   void *data = cso_hash_find_data_from_template(
+      ctx->regs_decl, scan_register_key(reg),
+      (void*)reg, sizeof(scan_register));
+   return  data ? TRUE : FALSE;
 }
 
 static boolean
@@ -112,23 +192,37 @@ is_any_register_declared(
    struct sanity_check_ctx *ctx,
    uint file )
 {
-   uint i;
+   struct cso_hash_iter iter =
+      cso_hash_first_node(ctx->regs_decl);
 
-   for (i = 0; i < MAX_REG_FLAGS; i++)
-      if (ctx->regs_decl[file][i])
+   while (!cso_hash_iter_is_null(iter)) {
+      scan_register *reg = (scan_register *)cso_hash_iter_data(iter);
+      if (reg->file == file)
          return TRUE;
+      iter = cso_hash_iter_next(iter);
+   }
+
    return FALSE;
 }
 
 static boolean
 is_register_used(
    struct sanity_check_ctx *ctx,
-   uint file,
-   int index )
+   scan_register *reg)
 {
-   assert( index < MAX_REGISTERS );
+   void *data = cso_hash_find_data_from_template(
+      ctx->regs_used, scan_register_key(reg),
+      reg, sizeof(scan_register));
+   return  data ? TRUE : FALSE;
+}
 
-   return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
+
+static boolean
+is_ind_register_used(
+   struct sanity_check_ctx *ctx,
+   scan_register *reg)
+{
+   return cso_hash_contains(ctx->regs_ind_used, reg->file);
 }
 
 static const char *file_names[TGSI_FILE_COUNT] =
@@ -141,37 +235,49 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "SAMP",
    "ADDR",
    "IMM",
-   "LOOP"
+   "LOOP",
+   "PRED"
 };
 
 static boolean
 check_register_usage(
    struct sanity_check_ctx *ctx,
-   uint file,
-   int index,
+   scan_register *reg,
    const char *name,
    boolean indirect_access )
 {
-   if (!check_file_name( ctx, file ))
+   if (!check_file_name( ctx, reg->file )) {
+      FREE(reg);
       return FALSE;
+   }
 
    if (indirect_access) {
       /* Note that 'index' is an offset relative to the value of the
-       * address register.  No range checking done here.
-       */
-      if (!is_any_register_declared( ctx, file ))
-         report_error( ctx, "%s: Undeclared %s register", file_names[file], name );
-      ctx->regs_ind_used[file] = TRUE;
+       * address register.  No range checking done here.*/
+      reg->indices[0] = 0;
+      reg->indices[1] = 0;
+      if (!is_any_register_declared( ctx, reg->file ))
+         report_error( ctx, "%s: Undeclared %s register", file_names[reg->file], name );
+      if (!is_ind_register_used(ctx, reg))
+         cso_hash_insert(ctx->regs_ind_used, reg->file, reg);
+      else
+         FREE(reg);
    }
    else {
-      if (index < 0 || index >= MAX_REGISTERS) {
-         report_error( ctx, "%s[%d]: Invalid %s index", file_names[file], index, name );
-         return FALSE;
+      if (!is_register_declared( ctx, reg )) {
+         if (reg->dimensions == 2) {
+            report_error( ctx, "%s[%d][%d]: Undeclared %s register", file_names[reg->file],
+                          reg->indices[0], reg->indices[1], name );
+         }
+         else {
+            report_error( ctx, "%s[%d]: Undeclared %s register", file_names[reg->file],
+                          reg->indices[0], name );
+         }
       }
-
-      if (!is_register_declared( ctx, file, index ))
-         report_error( ctx, "%s[%d]: Undeclared %s register", file_names[file], index, name );
-      ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG));
+      if (!is_register_used( ctx, reg ))
+         cso_hash_insert(ctx->regs_used, scan_register_key(reg), reg);
+      else
+         FREE(reg);
    }
    return TRUE;
 }
@@ -209,43 +315,43 @@ iter_instruction(
     * Mark the registers as used.
     */
    for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+      scan_register *reg = create_scan_register_dst(&inst->Dst[i]);
       check_register_usage(
          ctx,
-         inst->FullDstRegisters[i].DstRegister.File,
-         inst->FullDstRegisters[i].DstRegister.Index,
+         reg,
          "destination",
          FALSE );
    }
    for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      scan_register *reg = create_scan_register_src(&inst->Src[i]);
       check_register_usage(
          ctx,
-         inst->FullSrcRegisters[i].SrcRegister.File,
-         inst->FullSrcRegisters[i].SrcRegister.Index,
+         reg,
          "source",
-         (boolean)inst->FullSrcRegisters[i].SrcRegister.Indirect );
-      if (inst->FullSrcRegisters[i].SrcRegister.Indirect) {
-         uint file;
-         int index;
-
-         file = inst->FullSrcRegisters[i].SrcRegisterInd.File;
-         index = inst->FullSrcRegisters[i].SrcRegisterInd.Index;
+         (boolean)inst->Src[i].Register.Indirect );
+      if (inst->Src[i].Register.Indirect) {
+         scan_register *ind_reg = MALLOC(sizeof(scan_register));
+
+         fill_scan_register1d(ind_reg,
+                              inst->Src[i].Indirect.File,
+                              inst->Src[i].Indirect.Index);
+         if (!(reg->file == TGSI_FILE_ADDRESS || reg->file == TGSI_FILE_LOOP) ||
+             reg->indices[0] != 0) {
+            report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]");
+         }
          check_register_usage(
             ctx,
-            file,
-            index,
+            reg,
             "indirect",
             FALSE );
-         if (!(file == TGSI_FILE_ADDRESS || file == TGSI_FILE_LOOP) || index != 0) {
-            report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]");
-         }
       }
    }
 
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_BGNFOR:
    case TGSI_OPCODE_ENDFOR:
-      if (inst->FullDstRegisters[0].DstRegister.File != TGSI_FILE_LOOP ||
-          inst->FullDstRegisters[0].DstRegister.Index != 0) {
+      if (inst->Dst[0].Register.File != TGSI_FILE_LOOP ||
+          inst->Dst[0].Register.Index != 0) {
          report_error(ctx, "Destination register must be LOOP[0]");
       }
       break;
@@ -253,8 +359,8 @@ iter_instruction(
 
    switch (inst->Instruction.Opcode) {
    case TGSI_OPCODE_BGNFOR:
-      if (inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_CONSTANT &&
-          inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_IMMEDIATE) {
+      if (inst->Src[0].Register.File != TGSI_FILE_CONSTANT &&
+          inst->Src[0].Register.File != TGSI_FILE_IMMEDIATE) {
          report_error(ctx, "Source register file must be either CONST or IMM");
       }
       break;
@@ -265,6 +371,19 @@ iter_instruction(
    return TRUE;
 }
 
+static void
+check_and_declare(struct sanity_check_ctx *ctx,
+                  scan_register *reg)
+{
+   if (is_register_declared( ctx, reg))
+      report_error( ctx, "%s[%u]: The same register declared more than once",
+                    file_names[reg->file], reg->indices[0] );
+   cso_hash_insert(ctx->regs_decl,
+                   scan_register_key(reg),
+                   reg);
+}
+
+
 static boolean
 iter_declaration(
    struct tgsi_iterate_context *iter,
@@ -285,10 +404,22 @@ iter_declaration(
    file = decl->Declaration.File;
    if (!check_file_name( ctx, file ))
       return TRUE;
-   for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) {
-      if (is_register_declared( ctx, file, i ))
-         report_error( ctx, "%s[%u]: The same register declared more than once", file_names[file], i );
-      ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG));
+   for (i = decl->Range.First; i <= decl->Range.Last; i++) {
+      /* declared TGSI_FILE_INPUT's for geometry processor
+       * have an implied second dimension */
+      if (file == TGSI_FILE_INPUT &&
+          ctx->iter.processor.Processor == TGSI_PROCESSOR_GEOMETRY) {
+         uint vert;
+         for (vert = 0; vert < ctx->implied_array_size; ++vert) {
+            scan_register *reg = MALLOC(sizeof(scan_register));
+            fill_scan_register2d(reg, file, vert, i);
+            check_and_declare(ctx, reg);
+         }
+      } else {
+         scan_register *reg = MALLOC(sizeof(scan_register));
+         fill_scan_register1d(reg, file, i);
+         check_and_declare(ctx, reg);
+      }
    }
 
    return TRUE;
@@ -300,8 +431,7 @@ iter_immediate(
    struct tgsi_full_immediate *imm )
 {
    struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
-
-   assert( ctx->num_imms < MAX_REGISTERS );
+   scan_register *reg;
 
    /* No immediates allowed after the first instruction.
     */
@@ -310,12 +440,16 @@ iter_immediate(
 
    /* Mark the register as declared.
     */
-   ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG));
+   reg = MALLOC(sizeof(scan_register));
+   fill_scan_register1d(reg, TGSI_FILE_IMMEDIATE, ctx->num_imms);
+   cso_hash_insert(ctx->regs_decl, scan_register_key(reg), reg);
    ctx->num_imms++;
 
    /* Check data type validity.
     */
-   if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) {
+   if (imm->Immediate.DataType != TGSI_IMM_FLOAT32 &&
+       imm->Immediate.DataType != TGSI_IMM_UINT32 &&
+       imm->Immediate.DataType != TGSI_IMM_INT32) {
       report_error( ctx, "(%u): Invalid immediate data type", imm->Immediate.DataType );
       return TRUE;
    }
@@ -323,12 +457,26 @@ iter_immediate(
    return TRUE;
 }
 
+
+static boolean
+iter_property(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_property *prop )
+{
+   struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
+
+   if (iter->processor.Processor == TGSI_PROCESSOR_GEOMETRY &&
+       prop->Property.PropertyName == TGSI_PROPERTY_GS_INPUT_PRIM) {
+      ctx->implied_array_size = u_vertices_per_prim(prop->u[0].Data);
+   }
+   return TRUE;
+}
+
 static boolean
 epilog(
    struct tgsi_iterate_context *iter )
 {
    struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
-   uint file;
 
    /* There must be an END instruction somewhere.
     */
@@ -338,13 +486,17 @@ epilog(
 
    /* Check if all declared registers were used.
     */
-   for (file = TGSI_FILE_NULL; file < TGSI_FILE_COUNT; file++) {
-      uint i;
-
-      for (i = 0; i < MAX_REGISTERS; i++) {
-         if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) {
-            report_warning( ctx, "%s[%u]: Register never used", file_names[file], i );
+   {
+      struct cso_hash_iter iter =
+         cso_hash_first_node(ctx->regs_decl);
+
+      while (!cso_hash_iter_is_null(iter)) {
+         scan_register *reg = (scan_register *)cso_hash_iter_data(iter);
+         if (!is_register_used(ctx, reg) && !is_ind_register_used(ctx, reg)) {
+            report_warning( ctx, "%s[%u]: Register never used",
+                            file_names[reg->file], reg->indices[0] );
          }
+         iter = cso_hash_iter_next(iter);
       }
    }
 
@@ -356,6 +508,19 @@ epilog(
    return TRUE;
 }
 
+static void
+regs_hash_destroy(struct cso_hash *hash)
+{
+   struct cso_hash_iter iter = cso_hash_first_node(hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      scan_register *reg = (scan_register *)cso_hash_iter_data(iter);
+      iter = cso_hash_erase(hash, iter);
+      assert(reg->file < TGSI_FILE_COUNT);
+      FREE(reg);
+   }
+   cso_hash_delete(hash);
+}
+
 boolean
 tgsi_sanity_check(
    const struct tgsi_token *tokens )
@@ -366,20 +531,26 @@ tgsi_sanity_check(
    ctx.iter.iterate_instruction = iter_instruction;
    ctx.iter.iterate_declaration = iter_declaration;
    ctx.iter.iterate_immediate = iter_immediate;
+   ctx.iter.iterate_property = iter_property;
    ctx.iter.epilog = epilog;
 
-   memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) );
-   memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) );
-   memset( ctx.regs_ind_used, 0, sizeof( ctx.regs_ind_used ) );
+   ctx.regs_decl = cso_hash_create();
+   ctx.regs_used = cso_hash_create();
+   ctx.regs_ind_used = cso_hash_create();
+
    ctx.num_imms = 0;
    ctx.num_instructions = 0;
    ctx.index_of_END = ~0;
 
    ctx.errors = 0;
    ctx.warnings = 0;
+   ctx.implied_array_size = 0;
 
    if (!tgsi_iterate_shader( tokens, &ctx.iter ))
       return FALSE;
 
+   regs_hash_destroy(ctx.regs_decl);
+   regs_hash_destroy(ctx.regs_used);
+   regs_hash_destroy(ctx.regs_ind_used);
    return ctx.errors == 0;
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c
index 0db4481a3d..a6cc773003 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
@@ -35,7 +35,6 @@
 
 
 #include "util/u_math.h"
-#include "tgsi/tgsi_build.h"
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_scan.h"
 
@@ -97,14 +96,15 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                uint i;
                for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) {
                   const struct tgsi_full_src_register *src =
-                     &fullinst->FullSrcRegisters[i];
-                  if (src->SrcRegister.File == TGSI_FILE_INPUT) {
-                     const int ind = src->SrcRegister.Index;
+                     &fullinst->Src[i];
+                  if (src->Register.File == TGSI_FILE_INPUT ||
+                      src->Register.File == TGSI_FILE_SYSTEM_VALUE) {
+                     const int ind = src->Register.Index;
                      if (info->input_semantic_name[ind] == TGSI_SEMANTIC_FOG) {
-                        if (src->SrcRegister.SwizzleX == TGSI_SWIZZLE_X) {
+                        if (src->Register.SwizzleX == TGSI_SWIZZLE_X) {
                            info->uses_fogcoord = TRUE;
                         }
-                        else if (src->SrcRegister.SwizzleX == TGSI_SWIZZLE_Y) {
+                        else if (src->Register.SwizzleX == TGSI_SWIZZLE_Y) {
                            info->uses_frontfacing = TRUE;
                         }
                      }
@@ -120,8 +120,8 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                = &parse.FullToken.FullDeclaration;
             const uint file = fulldecl->Declaration.File;
             uint reg;
-            for (reg = fulldecl->DeclarationRange.First;
-                 reg <= fulldecl->DeclarationRange.Last;
+            for (reg = fulldecl->Range.First;
+                 reg <= fulldecl->Range.Last;
                  reg++) {
 
                /* only first 32 regs will appear in this bitfield */
@@ -129,25 +129,29 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                info->file_count[file]++;
                info->file_max[file] = MAX2(info->file_max[file], (int)reg);
 
-               if (file == TGSI_FILE_INPUT) {
-                  info->input_semantic_name[reg] = (ubyte)fulldecl->Semantic.SemanticName;
-                  info->input_semantic_index[reg] = (ubyte)fulldecl->Semantic.SemanticIndex;
+               if (file == TGSI_FILE_INPUT || file == TGSI_FILE_SYSTEM_VALUE) {
+                  info->input_semantic_name[reg] = (ubyte)fulldecl->Semantic.Name;
+                  info->input_semantic_index[reg] = (ubyte)fulldecl->Semantic.Index;
                   info->input_interpolate[reg] = (ubyte)fulldecl->Declaration.Interpolate;
                   info->num_inputs++;
                }
                else if (file == TGSI_FILE_OUTPUT) {
-                  info->output_semantic_name[reg] = (ubyte)fulldecl->Semantic.SemanticName;
-                  info->output_semantic_index[reg] = (ubyte)fulldecl->Semantic.SemanticIndex;
+                  info->output_semantic_name[reg] = (ubyte)fulldecl->Semantic.Name;
+                  info->output_semantic_index[reg] = (ubyte)fulldecl->Semantic.Index;
                   info->num_outputs++;
-               }
 
-               /* special case */
-               if (procType == TGSI_PROCESSOR_FRAGMENT &&
-                   file == TGSI_FILE_OUTPUT &&
-                   fulldecl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) {
-                  info->writes_z = TRUE;
+                  /* extra info for special outputs */
+                  if (procType == TGSI_PROCESSOR_FRAGMENT &&
+                      fulldecl->Semantic.Name == TGSI_SEMANTIC_POSITION) {
+                     info->writes_z = TRUE;
+                  }
+                  if (procType == TGSI_PROCESSOR_VERTEX &&
+                      fulldecl->Semantic.Name == TGSI_SEMANTIC_EDGEFLAG) {
+                     info->writes_edgeflag = TRUE;
+                  }
                }
-            }
+
+             }
          }
          break;
 
@@ -161,6 +165,19 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
             info->file_max[file] = MAX2(info->file_max[file], (int)reg);
          }
          break;
+      case TGSI_TOKEN_TYPE_PROPERTY:
+      {
+         const struct tgsi_full_property *fullprop
+            = &parse.FullToken.FullProperty;
+
+         info->properties[info->num_properties].name =
+            fullprop->Property.PropertyName;
+         memcpy(info->properties[info->num_properties].data,
+                fullprop->u, 8 * sizeof(unsigned));;
+
+         ++info->num_properties;
+      }
+      break;
 
       default:
          assert( 0 );
@@ -206,34 +223,26 @@ tgsi_is_passthrough_shader(const struct tgsi_token *tokens)
             struct tgsi_full_instruction *fullinst =
                &parse.FullToken.FullInstruction;
             const struct tgsi_full_src_register *src =
-               &fullinst->FullSrcRegisters[0];
+               &fullinst->Src[0];
             const struct tgsi_full_dst_register *dst =
-               &fullinst->FullDstRegisters[0];
+               &fullinst->Dst[0];
 
             /* Do a whole bunch of checks for a simple move */
             if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV ||
-                src->SrcRegister.File != TGSI_FILE_INPUT ||
-                dst->DstRegister.File != TGSI_FILE_OUTPUT ||
-                src->SrcRegister.Index != dst->DstRegister.Index ||
-
-                src->SrcRegister.Negate ||
-                src->SrcRegisterExtMod.Negate ||
-                src->SrcRegisterExtMod.Absolute ||
-                src->SrcRegisterExtMod.Scale2X ||
-                src->SrcRegisterExtMod.Bias ||
-                src->SrcRegisterExtMod.Complement ||
-
-                src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X ||
-                src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y ||
-                src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z ||
-                src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ||
-
-                src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X ||
-                src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y ||
-                src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z ||
-                src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W ||
-
-                dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW)
+                (src->Register.File != TGSI_FILE_INPUT &&
+                 src->Register.File != TGSI_FILE_SYSTEM_VALUE) ||
+                dst->Register.File != TGSI_FILE_OUTPUT ||
+                src->Register.Index != dst->Register.Index ||
+
+                src->Register.Negate ||
+                src->Register.Absolute ||
+
+                src->Register.SwizzleX != TGSI_SWIZZLE_X ||
+                src->Register.SwizzleY != TGSI_SWIZZLE_Y ||
+                src->Register.SwizzleZ != TGSI_SWIZZLE_Z ||
+                src->Register.SwizzleW != TGSI_SWIZZLE_W ||
+
+                dst->Register.WriteMask != TGSI_WRITEMASK_XYZW)
             {
                tgsi_parse_free(&parse);
                return FALSE;
@@ -245,6 +254,8 @@ tgsi_is_passthrough_shader(const struct tgsi_token *tokens)
          /* fall-through */
       case TGSI_TOKEN_TYPE_IMMEDIATE:
          /* fall-through */
+      case TGSI_TOKEN_TYPE_PROPERTY:
+         /* fall-through */
       default:
          ; /* no-op */
       }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h
index 8a7ee0c7e4..dae5376c24 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h
@@ -33,7 +33,6 @@
 #include "pipe/p_state.h"
 #include "pipe/p_shader_tokens.h"
 
-
 /**
  * Shader summary info
  */
@@ -58,11 +57,17 @@ struct tgsi_shader_info
    uint opcode_count[TGSI_OPCODE_LAST];  /**< opcode histogram */
 
    boolean writes_z;  /**< does fragment shader write Z value? */
+   boolean writes_edgeflag; /**< vertex shader outputs edgeflag */
    boolean uses_kill;  /**< KIL or KILP instruction used? */
    boolean uses_fogcoord; /**< fragment shader uses fog coord? */
    boolean uses_frontfacing; /**< fragment shader uses front/back-face flag? */
-};
 
+   struct {
+      unsigned name;
+      unsigned data[8];
+   } properties[TGSI_PROPERTY_COUNT];
+   uint num_properties;
+};
 
 extern void
 tgsi_scan_shader(const struct tgsi_token *tokens,
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
index 5f6b83b236..2e13a7aaf9 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -2,6 +2,7 @@
  * 
  * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
+ * Copyright 2009-2010 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
@@ -58,7 +59,7 @@
    for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)
 
 #define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
-   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+   ((INST).Dst[0].Register.WriteMask & (1 << (CHAN)))
 
 #define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
    if (IS_DST0_CHANNEL_ENABLED( INST, CHAN ))
@@ -1260,38 +1261,39 @@ emit_fetch(
    const struct tgsi_full_src_register *reg,
    const unsigned chan_index )
 {
-   unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+   unsigned swizzle = tgsi_util_get_full_src_register_swizzle( reg, chan_index );
 
    switch (swizzle) {
-   case TGSI_EXTSWIZZLE_X:
-   case TGSI_EXTSWIZZLE_Y:
-   case TGSI_EXTSWIZZLE_Z:
-   case TGSI_EXTSWIZZLE_W:
-      switch (reg->SrcRegister.File) {
+   case TGSI_SWIZZLE_X:
+   case TGSI_SWIZZLE_Y:
+   case TGSI_SWIZZLE_Z:
+   case TGSI_SWIZZLE_W:
+      switch (reg->Register.File) {
       case TGSI_FILE_CONSTANT:
          emit_const(
             func,
             xmm,
-            reg->SrcRegister.Index,
+            reg->Register.Index,
             swizzle,
-            reg->SrcRegister.Indirect,
-            reg->SrcRegisterInd.File,
-            reg->SrcRegisterInd.Index );
+            reg->Register.Indirect,
+            reg->Indirect.File,
+            reg->Indirect.Index );
          break;
 
       case TGSI_FILE_IMMEDIATE:
          emit_immediate(
             func,
             xmm,
-            reg->SrcRegister.Index,
+            reg->Register.Index,
             swizzle );
          break;
 
       case TGSI_FILE_INPUT:
+      case TGSI_FILE_SYSTEM_VALUE:
          emit_inputf(
             func,
             xmm,
-            reg->SrcRegister.Index,
+            reg->Register.Index,
             swizzle );
          break;
 
@@ -1299,7 +1301,7 @@ emit_fetch(
          emit_tempf(
             func,
             xmm,
-            reg->SrcRegister.Index,
+            reg->Register.Index,
             swizzle );
          break;
 
@@ -1308,22 +1310,6 @@ emit_fetch(
       }
       break;
 
-   case TGSI_EXTSWIZZLE_ZERO:
-      emit_tempf(
-         func,
-         xmm,
-         TGSI_EXEC_TEMP_00000000_I,
-         TGSI_EXEC_TEMP_00000000_C );
-      break;
-
-   case TGSI_EXTSWIZZLE_ONE:
-      emit_tempf(
-         func,
-         xmm,
-         TEMP_ONE_I,
-         TEMP_ONE_C );
-      break;
-
    default:
       assert( 0 );
    }
@@ -1347,7 +1333,7 @@ emit_fetch(
 }
 
 #define FETCH( FUNC, INST, XMM, INDEX, CHAN )\
-   emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN )
+   emit_fetch( FUNC, XMM, &(INST).Src[INDEX], CHAN )
 
 /**
  * Register store.
@@ -1387,12 +1373,12 @@ emit_store(
    }
 
 
-   switch( reg->DstRegister.File ) {
+   switch( reg->Register.File ) {
    case TGSI_FILE_OUTPUT:
       emit_output(
          func,
          xmm,
-         reg->DstRegister.Index,
+         reg->Register.Index,
          chan_index );
       break;
 
@@ -1400,7 +1386,7 @@ emit_store(
       emit_temps(
          func,
          xmm,
-         reg->DstRegister.Index,
+         reg->Register.Index,
          chan_index );
       break;
 
@@ -1408,7 +1394,7 @@ emit_store(
       emit_addrs(
          func,
          xmm,
-         reg->DstRegister.Index,
+         reg->Register.Index,
          chan_index );
       break;
 
@@ -1418,7 +1404,7 @@ emit_store(
 }
 
 #define STORE( FUNC, INST, XMM, INDEX, CHAN )\
-   emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN )
+   emit_store( FUNC, XMM, &(INST).Dst[INDEX], &(INST), CHAN )
 
 
 static void PIPE_CDECL
@@ -1433,13 +1419,13 @@ fetch_texel( struct tgsi_sampler **sampler,
                 sampler, *sampler,
                 store );
 
-   debug_printf("lodbias %f\n", store[12]);
-
    for (j = 0; j < 4; j++)
-      debug_printf("sample %d texcoord %f %f\n", 
+      debug_printf("sample %d texcoord %f %f %f lodbias %f\n",
                    j, 
                    store[0+j],
-                   store[4+j]);
+                   store[4+j],
+                   store[8 + j],
+                   store[12 + j]);
 #endif
 
    {
@@ -1448,7 +1434,8 @@ fetch_texel( struct tgsi_sampler **sampler,
                               &store[0],  /* s */
                               &store[4],  /* t */
                               &store[8],  /* r */
-                              store[12],  /* lodbias */
+                              &store[12], /* lodbias */
+                              tgsi_sampler_lod_bias,
                               rgba);      /* results */
 
       memcpy( store, rgba, 16 * sizeof(float));
@@ -1475,12 +1462,13 @@ emit_tex( struct x86_function *func,
           boolean lodbias,
           boolean projected)
 {
-   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   const uint unit = inst->Src[1].Register.Index;
    struct x86_reg args[2];
    unsigned count;
    unsigned i;
 
-   switch (inst->InstructionExtTexture.Texture) {
+   assert(inst->Instruction.Texture);
+   switch (inst->Texture.Texture) {
    case TGSI_TEXTURE_1D:
       count = 1;
       break;
@@ -1582,13 +1570,13 @@ emit_kil(
    /* This mask stores component bits that were already tested. Note that
     * we test if the value is less than zero, so 1.0 and 0.0 need not to be
     * tested. */
-   uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE);
+   uniquemask = 0;
 
    FOR_EACH_CHANNEL( chan_index ) {
       unsigned swizzle;
 
       /* unswizzle channel */
-      swizzle = tgsi_util_get_full_src_register_extswizzle(
+      swizzle = tgsi_util_get_full_src_register_swizzle(
          reg,
          chan_index );
 
@@ -1735,15 +1723,15 @@ indirect_temp_reference(const struct tgsi_full_instruction *inst)
 {
    uint i;
    for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
-      const struct tgsi_full_src_register *reg = &inst->FullSrcRegisters[i];
-      if (reg->SrcRegister.File == TGSI_FILE_TEMPORARY &&
-          reg->SrcRegister.Indirect)
+      const struct tgsi_full_src_register *reg = &inst->Src[i];
+      if (reg->Register.File == TGSI_FILE_TEMPORARY &&
+          reg->Register.Indirect)
          return TRUE;
    }
    for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
-      const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[i];
-      if (reg->DstRegister.File == TGSI_FILE_TEMPORARY &&
-          reg->DstRegister.Indirect)
+      const struct tgsi_full_dst_register *reg = &inst->Dst[i];
+      if (reg->Register.File == TGSI_FILE_TEMPORARY &&
+          reg->Register.Indirect)
          return TRUE;
    }
    return FALSE;
@@ -1772,7 +1760,6 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_MOV:
-   case TGSI_OPCODE_SWZ:
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
          FETCH( func, *inst, 4 + chan_index, 0, chan_index );
       }
@@ -2260,7 +2247,7 @@ emit_instruction(
 
    case TGSI_OPCODE_KIL:
       /* conditional kill */
-      emit_kil( func, &inst->FullSrcRegisters[0] );
+      emit_kil( func, &inst->Src[0] );
       break;
 
    case TGSI_OPCODE_PK2H:
@@ -2521,7 +2508,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXL:
-      emit_tex( func, inst, TRUE, FALSE );
+      return 0;
       break;
 
    case TGSI_OPCODE_TXP:
@@ -2593,7 +2580,7 @@ emit_instruction(
       return 0;
       break;
 
-   case TGSI_OPCODE_SHR:
+   case TGSI_OPCODE_ISHR:
       return 0;
       break;
 
@@ -2649,12 +2636,13 @@ emit_declaration(
    struct x86_function *func,
    struct tgsi_full_declaration *decl )
 {
-   if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+   if( decl->Declaration.File == TGSI_FILE_INPUT ||
+       decl->Declaration.File == TGSI_FILE_SYSTEM_VALUE ) {
       unsigned first, last, mask;
       unsigned i, j;
 
-      first = decl->DeclarationRange.First;
-      last = decl->DeclarationRange.Last;
+      first = decl->Range.First;
+      last = decl->Range.Last;
       mask = decl->Declaration.UsageMask;
 
       for( i = first; i <= last; i++ ) {
@@ -2938,8 +2926,7 @@ tgsi_emit_sse2(
              * the result in the cases where the code is too opaque to
              * fix.
              */
-            if (opcode != TGSI_OPCODE_MOV &&
-                opcode != TGSI_OPCODE_SWZ) {
+            if (opcode != TGSI_OPCODE_MOV) {
                debug_printf("Warning: src/dst aliasing in instruction"
                             " is not handled:\n");
                tgsi_dump_instruction(&parse.FullToken.FullInstruction, 1);
@@ -2969,6 +2956,9 @@ tgsi_emit_sse2(
             num_immediates++;
          }
          break;
+      case TGSI_TOKEN_TYPE_PROPERTY:
+         /* we just ignore them for now */
+         break;
 
       default:
 	 ok = 0;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index d438450b1e..9fcffeda36 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -27,6 +27,9 @@
 
 #include "util/u_debug.h"
 #include "util/u_memory.h"
+#include "util/u_prim.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
 #include "tgsi_text.h"
 #include "tgsi_build.h"
 #include "tgsi_info.h"
@@ -59,6 +62,23 @@ static boolean uprcase( char c )
    return c;
 }
 
+/*
+ * Ignore case of str1 and assume str1 is already uppercase.
+ * Return TRUE iff str1 and str2 are equal.
+ */
+static int
+streq_nocase_uprcase(const char *str1,
+                     const char *str2)
+{
+   while (*str1 && *str2) {
+      if (*str1 != uprcase(*str2))
+         return FALSE;
+      str1++;
+      str2++;
+   }
+   return TRUE;
+}
+
 static boolean str_match_no_case( const char **pcur, const char *str )
 {
    const char *cur = *pcur;
@@ -110,6 +130,20 @@ static boolean parse_uint( const char **pcur, uint *val )
    return FALSE;
 }
 
+static boolean parse_identifier( const char **pcur, char *ret )
+{
+   const char *cur = *pcur;
+   int i = 0;
+   if (is_alpha_underscore( cur )) {
+      ret[i++] = *cur++;
+      while (is_alpha_underscore( cur ))
+         ret[i++] = *cur++;
+      *pcur = cur;
+      return TRUE;
+   }
+   return FALSE;
+}
+
 /* Parse floating point.
  */
 static boolean parse_float( const char **pcur, float *val )
@@ -163,28 +197,43 @@ struct translate_ctx
    struct tgsi_token *tokens_cur;
    struct tgsi_token *tokens_end;
    struct tgsi_header *header;
+   unsigned processor : 4;
+   int implied_array_size : 5;
 };
 
 static void report_error( struct translate_ctx *ctx, const char *msg )
 {
-   debug_printf( "\nError: %s", msg );
+   int line = 1;
+   int column = 1;
+   const char *itr = ctx->text;
+
+   while (itr != ctx->cur) {
+      if (*itr == '\n') {
+         column = 1;
+         ++line;
+      }
+      ++column;
+      ++itr;
+   }
+
+   debug_printf( "\nTGSI asm error: %s [%d : %d] \n", msg, line, column );
 }
 
 /* Parse shader header.
  * Return TRUE for one of the following headers.
- *    FRAG1.1
- *    GEOM1.1
- *    VERT1.1
+ *    FRAG
+ *    GEOM
+ *    VERT
  */
 static boolean parse_header( struct translate_ctx *ctx )
 {
    uint processor;
 
-   if (str_match_no_case( &ctx->cur, "FRAG1.1" ))
+   if (str_match_no_case( &ctx->cur, "FRAG" ))
       processor = TGSI_PROCESSOR_FRAGMENT;
-   else if (str_match_no_case( &ctx->cur, "VERT1.1" ))
+   else if (str_match_no_case( &ctx->cur, "VERT" ))
       processor = TGSI_PROCESSOR_VERTEX;
-   else if (str_match_no_case( &ctx->cur, "GEOM1.1" ))
+   else if (str_match_no_case( &ctx->cur, "GEOM" ))
       processor = TGSI_PROCESSOR_GEOMETRY;
    else {
       report_error( ctx, "Unknown header" );
@@ -193,16 +242,13 @@ static boolean parse_header( struct translate_ctx *ctx )
 
    if (ctx->tokens_cur >= ctx->tokens_end)
       return FALSE;
-   *(struct tgsi_version *) ctx->tokens_cur++ = tgsi_build_version();
-
-   if (ctx->tokens_cur >= ctx->tokens_end)
-      return FALSE;
    ctx->header = (struct tgsi_header *) ctx->tokens_cur++;
    *ctx->header = tgsi_build_header();
 
    if (ctx->tokens_cur >= ctx->tokens_end)
       return FALSE;
    *(struct tgsi_processor *) ctx->tokens_cur++ = tgsi_build_processor( processor, ctx->header );
+   ctx->processor = processor;
 
    return TRUE;
 }
@@ -232,7 +278,9 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "SAMP",
    "ADDR",
    "IMM",
-   "LOOP"
+   "LOOP",
+   "PRED",
+   "SV"
 };
 
 static boolean
@@ -297,92 +345,36 @@ parse_opt_writemask(
    return TRUE;
 }
 
-/* <register_file_bracket> ::= <file> `['
- */
 static boolean
-parse_register_file_bracket(
-   struct translate_ctx *ctx,
-   uint *file )
-{
-   if (!parse_file( &ctx->cur, file )) {
-      report_error( ctx, "Unknown register file" );
-      return FALSE;
-   }
-   eat_opt_white( &ctx->cur );
-   if (*ctx->cur != '[') {
-      report_error( ctx, "Expected `['" );
-      return FALSE;
-   }
-   ctx->cur++;
-   return TRUE;
-}
+parse_register_dst( struct translate_ctx *ctx,
+                    uint *file,
+                    int *index );
 
-/* <register_file_bracket_index> ::= <register_file_bracket> <uint>
- */
-static boolean
-parse_register_file_bracket_index(
-   struct translate_ctx *ctx,
-   uint *file,
-   int *index )
-{
-   uint uindex;
+struct parsed_src_bracket {
+   int index;
 
-   if (!parse_register_file_bracket( ctx, file ))
-      return FALSE;
-   eat_opt_white( &ctx->cur );
-   if (!parse_uint( &ctx->cur, &uindex )) {
-      report_error( ctx, "Expected literal unsigned integer" );
-      return FALSE;
-   }
-   *index = (int) uindex;
-   return TRUE;
-}
+   uint ind_file;
+   int ind_index;
+   uint ind_comp;
+};
 
-/* Parse destination register operand.
- *    <register_dst> ::= <register_file_bracket_index> `]'
- */
-static boolean
-parse_register_dst(
-   struct translate_ctx *ctx,
-   uint *file,
-   int *index )
-{
-   if (!parse_register_file_bracket_index( ctx, file, index ))
-      return FALSE;
-   eat_opt_white( &ctx->cur );
-   if (*ctx->cur != ']') {
-      report_error( ctx, "Expected `]'" );
-      return FALSE;
-   }
-   ctx->cur++;
-   return TRUE;
-}
 
-/* Parse source register operand.
- *    <register_src> ::= <register_file_bracket_index> `]' |
- *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `]' |
- *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `+' <uint> `]' |
- *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `-' <uint> `]'
- */
 static boolean
-parse_register_src(
+parse_register_src_bracket(
    struct translate_ctx *ctx,
-   uint *file,
-   int *index,
-   uint *ind_file,
-   int *ind_index,
-   uint *ind_comp)
+   struct parsed_src_bracket *brackets)
 {
    const char *cur;
    uint uindex;
 
-   *ind_comp = TGSI_SWIZZLE_X;
-   if (!parse_register_file_bracket( ctx, file ))
-      return FALSE;
+   memset(brackets, 0, sizeof(struct parsed_src_bracket));
+
    eat_opt_white( &ctx->cur );
+
    cur = ctx->cur;
-   if (parse_file( &cur, ind_file )) {
-      if (!parse_register_dst( ctx, ind_file, ind_index ))
+   if (parse_file( &cur, &brackets->ind_file )) {
+      if (!parse_register_dst( ctx, &brackets->ind_file,
+                               &brackets->ind_index ))
          return FALSE;
       eat_opt_white( &ctx->cur );
 
@@ -392,16 +384,16 @@ parse_register_src(
 
          switch (uprcase(*ctx->cur)) {
          case 'X':
-            *ind_comp = TGSI_SWIZZLE_X;
+            brackets->ind_comp = TGSI_SWIZZLE_X;
             break;
          case 'Y':
-            *ind_comp = TGSI_SWIZZLE_Y;
+            brackets->ind_comp = TGSI_SWIZZLE_Y;
             break;
          case 'Z':
-            *ind_comp = TGSI_SWIZZLE_Z;
+            brackets->ind_comp = TGSI_SWIZZLE_Z;
             break;
          case 'W':
-            *ind_comp = TGSI_SWIZZLE_W;
+            brackets->ind_comp = TGSI_SWIZZLE_W;
             break;
          default:
             report_error(ctx, "Expected indirect register swizzle component `x', `y', `z' or `w'");
@@ -422,12 +414,12 @@ parse_register_src(
             return FALSE;
          }
          if (negate)
-            *index = -(int) uindex;
+            brackets->index = -(int) uindex;
          else
-            *index = (int) uindex;
+            brackets->index = (int) uindex;
       }
       else {
-         *index = 0;
+         brackets->index = 0;
       }
    }
    else {
@@ -435,9 +427,9 @@ parse_register_src(
          report_error( ctx, "Expected literal unsigned integer" );
          return FALSE;
       }
-      *index = (int) uindex;
-      *ind_file = TGSI_FILE_NULL;
-      *ind_index = 0;
+      brackets->index = (int) uindex;
+      brackets->ind_file = TGSI_FILE_NULL;
+      brackets->ind_index = 0;
    }
    eat_opt_white( &ctx->cur );
    if (*ctx->cur != ']') {
@@ -448,20 +440,123 @@ parse_register_src(
    return TRUE;
 }
 
-/* Parse register declaration.
- *    <register_dcl> ::= <register_file_bracket_index> `]' |
- *                       <register_file_bracket_index> `..' <index> `]'
+static boolean
+parse_opt_register_src_bracket(
+   struct translate_ctx *ctx,
+   struct parsed_src_bracket *brackets,
+   int *parsed_brackets)
+{
+   const char *cur = ctx->cur;
+
+   *parsed_brackets = 0;
+
+   eat_opt_white( &cur );
+   if (cur[0] == '[') {
+      ++cur;
+      ctx->cur = cur;
+
+      if (!parse_register_src_bracket(ctx, brackets))
+         return FALSE;
+
+      *parsed_brackets = 1;
+   }
+
+   return TRUE;
+}
+
+/* <register_file_bracket> ::= <file> `['
  */
 static boolean
-parse_register_dcl(
+parse_register_file_bracket(
+   struct translate_ctx *ctx,
+   uint *file )
+{
+   if (!parse_file( &ctx->cur, file )) {
+      report_error( ctx, "Unknown register file" );
+      return FALSE;
+   }
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != '[') {
+      report_error( ctx, "Expected `['" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
+
+/* <register_file_bracket_index> ::= <register_file_bracket> <uint>
+ */
+static boolean
+parse_register_file_bracket_index(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *index )
+{
+   uint uindex;
+
+   if (!parse_register_file_bracket( ctx, file ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   if (!parse_uint( &ctx->cur, &uindex )) {
+      report_error( ctx, "Expected literal unsigned integer" );
+      return FALSE;
+   }
+   *index = (int) uindex;
+   return TRUE;
+}
+
+/* Parse source register operand.
+ *    <register_src> ::= <register_file_bracket_index> `]' |
+ *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `]' |
+ *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `+' <uint> `]' |
+ *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `-' <uint> `]'
+ */
+static boolean
+parse_register_src(
    struct translate_ctx *ctx,
    uint *file,
-   int *first,
-   int *last )
+   struct parsed_src_bracket *brackets)
 {
-   if (!parse_register_file_bracket_index( ctx, file, first ))
+
+   brackets->ind_comp = TGSI_SWIZZLE_X;
+   if (!parse_register_file_bracket( ctx, file ))
       return FALSE;
+   if (!parse_register_src_bracket( ctx, brackets ))
+       return FALSE;
+
+   return TRUE;
+}
+
+struct parsed_dcl_bracket {
+   uint first;
+   uint last;
+};
+
+static boolean
+parse_register_dcl_bracket(
+   struct translate_ctx *ctx,
+   struct parsed_dcl_bracket *bracket)
+{
+   uint uindex;
+   memset(bracket, 0, sizeof(struct parsed_dcl_bracket));
+
    eat_opt_white( &ctx->cur );
+
+   if (!parse_uint( &ctx->cur, &uindex )) {
+      /* it can be an empty bracket [] which means its range
+       * is from 0 to some implied size */
+      if (ctx->cur[0] == ']' && ctx->implied_array_size != 0) {
+         bracket->first = 0;
+         bracket->last = ctx->implied_array_size - 1;
+         goto cleanup;
+      }
+      report_error( ctx, "Expected literal unsigned integer" );
+      return FALSE;
+   }
+   bracket->first = (int) uindex;
+
+   eat_opt_white( &ctx->cur );
+
    if (ctx->cur[0] == '.' && ctx->cur[1] == '.') {
       uint uindex;
 
@@ -471,12 +566,14 @@ parse_register_dcl(
          report_error( ctx, "Expected literal integer" );
          return FALSE;
       }
-      *last = (int) uindex;
+      bracket->last = (int) uindex;
       eat_opt_white( &ctx->cur );
    }
    else {
-      *last = *first;
+      bracket->last = bracket->first;
    }
+
+cleanup:
    if (*ctx->cur != ']') {
       report_error( ctx, "Expected `]' or `..'" );
       return FALSE;
@@ -485,16 +582,70 @@ parse_register_dcl(
    return TRUE;
 }
 
-static const char *modulate_names[TGSI_MODULATE_COUNT] =
+/* Parse register declaration.
+ *    <register_dcl> ::= <register_file_bracket_index> `]' |
+ *                       <register_file_bracket_index> `..' <index> `]'
+ */
+static boolean
+parse_register_dcl(
+   struct translate_ctx *ctx,
+   uint *file,
+   struct parsed_dcl_bracket *brackets,
+   int *num_brackets)
 {
-   "_1X",
-   "_2X",
-   "_4X",
-   "_8X",
-   "_D2",
-   "_D4",
-   "_D8"
-};
+   const char *cur;
+
+   *num_brackets = 0;
+
+   if (!parse_register_file_bracket( ctx, file ))
+      return FALSE;
+   if (!parse_register_dcl_bracket( ctx, &brackets[0] ))
+      return FALSE;
+
+   *num_brackets = 1;
+
+   cur = ctx->cur;
+   eat_opt_white( &cur );
+
+   if (cur[0] == '[') {
+      ++cur;
+      ctx->cur = cur;
+      if (!parse_register_dcl_bracket( ctx, &brackets[1] ))
+         return FALSE;
+      /* for geometry shader we don't really care about
+       * the first brackets it's always the size of the
+       * input primitive. so we want to declare just
+       * the index relevant to the semantics which is in
+       * the second bracket */
+      if (ctx->processor == TGSI_PROCESSOR_GEOMETRY) {
+         brackets[0] = brackets[1];
+      }
+      *num_brackets = 2;
+   }
+
+   return TRUE;
+}
+
+
+/* Parse destination register operand.
+ *    <register_dst> ::= <register_file_bracket_index> `]'
+ */
+static boolean
+parse_register_dst(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *index )
+{
+   if (!parse_register_file_bracket_index( ctx, file, index ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != ']') {
+      report_error( ctx, "Expected `]'" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
 
 static boolean
 parse_dst_operand(
@@ -511,26 +662,13 @@ parse_dst_operand(
 
    cur = ctx->cur;
    eat_opt_white( &cur );
-   if (*cur == '_') {
-      uint i;
-
-      for (i = 0; i < TGSI_MODULATE_COUNT; i++) {
-         if (str_match_no_case( &cur, modulate_names[i] )) {
-            if (!is_digit_alpha_underscore( cur )) {
-               dst->DstRegisterExtModulate.Modulate = i;
-               ctx->cur = cur;
-               break;
-            }
-         }
-      }
-   }
 
    if (!parse_opt_writemask( ctx, &writemask ))
       return FALSE;
 
-   dst->DstRegister.File = file;
-   dst->DstRegister.Index = index;
-   dst->DstRegister.WriteMask = writemask;
+   dst->Register.File = file;
+   dst->Register.Index = index;
+   dst->Register.WriteMask = writemask;
    return TRUE;
 }
 
@@ -538,13 +676,11 @@ static boolean
 parse_optional_swizzle(
    struct translate_ctx *ctx,
    uint swizzle[4],
-   boolean *parsed_swizzle,
-   boolean *parsed_extswizzle )
+   boolean *parsed_swizzle )
 {
    const char *cur = ctx->cur;
 
    *parsed_swizzle = FALSE;
-   *parsed_extswizzle = FALSE;
 
    eat_opt_white( &cur );
    if (*cur == '.') {
@@ -562,15 +698,8 @@ parse_optional_swizzle(
          else if (uprcase( *cur ) == 'W')
             swizzle[i] = TGSI_SWIZZLE_W;
          else {
-            if (*cur == '0')
-               swizzle[i] = TGSI_EXTSWIZZLE_ZERO;
-            else if (*cur == '1')
-               swizzle[i] = TGSI_EXTSWIZZLE_ONE;
-            else {
-               report_error( ctx, "Expected register swizzle component `x', `y', `z', `w', `0' or `1'" );
-               return FALSE;
-            }
-            *parsed_extswizzle = TRUE;
+	    report_error( ctx, "Expected register swizzle component `x', `y', `z', `w', `0' or `1'" );
+	    return FALSE;
          }
          cur++;
       }
@@ -585,171 +714,59 @@ parse_src_operand(
    struct translate_ctx *ctx,
    struct tgsi_full_src_register *src )
 {
-   const char *cur;
-   float value;
    uint file;
-   int index;
-   uint ind_file;
-   int ind_index;
-   uint ind_comp;
    uint swizzle[4];
-   boolean parsed_ext_negate_paren = FALSE;
    boolean parsed_swizzle;
-   boolean parsed_extswizzle;
+   struct parsed_src_bracket bracket[2];
+   int parsed_opt_brackets;
 
    if (*ctx->cur == '-') {
-      cur = ctx->cur;
-      cur++;
-      eat_opt_white( &cur );
-      if (*cur == '(') {
-         cur++;
-         src->SrcRegisterExtMod.Negate = 1;
-         eat_opt_white( &cur );
-         ctx->cur = cur;
-         parsed_ext_negate_paren = TRUE;
-      }
-      else if (*cur == '|') {
-         cur++;
-         src->SrcRegisterExtMod.Negate = 1;
-         src->SrcRegisterExtMod.Absolute = 1;
-         eat_opt_white(&cur);
-         ctx->cur = cur;
-      }
-   }
-   else if (*ctx->cur == '|') {
       ctx->cur++;
       eat_opt_white( &ctx->cur );
-      src->SrcRegisterExtMod.Absolute = 1;
+      src->Register.Negate = 1;
    }
 
-   if (*ctx->cur == '-') {
+   if (*ctx->cur == '|') {
       ctx->cur++;
       eat_opt_white( &ctx->cur );
-      src->SrcRegister.Negate = 1;
+      src->Register.Absolute = 1;
    }
 
-   cur = ctx->cur;
-   if (parse_float( &cur, &value )) {
-      if (value == 2.0f) {
-         eat_opt_white( &cur );
-         if (*cur != '*') {
-            report_error( ctx, "Expected `*'" );
-            return FALSE;
-         }
-         cur++;
-         if (*cur != '(') {
-            report_error( ctx, "Expected `('" );
-            return FALSE;
-         }
-         cur++;
-         src->SrcRegisterExtMod.Scale2X = 1;
-         eat_opt_white( &cur );
-         ctx->cur = cur;
-      }
-   }
-
-   if (*ctx->cur == '(') {
-      ctx->cur++;
-      eat_opt_white( &ctx->cur );
-      src->SrcRegisterExtMod.Bias = 1;
-   }
+   if (!parse_register_src(ctx, &file, &bracket[0]))
+      return FALSE;
+   if (!parse_opt_register_src_bracket(ctx, &bracket[1], &parsed_opt_brackets))
+      return FALSE;
 
-   cur = ctx->cur;
-   if (parse_float( &cur, &value )) {
-      if (value == 1.0f) {
-         eat_opt_white( &cur );
-         if (*cur != '-') {
-            report_error( ctx, "Expected `-'" );
-            return FALSE;
-         }
-         cur++;
-         if (*cur != '(') {
-            report_error( ctx, "Expected `('" );
-            return FALSE;
-         }
-         cur++;
-         src->SrcRegisterExtMod.Complement = 1;
-         eat_opt_white( &cur );
-         ctx->cur = cur;
-      }
+   src->Register.File = file;
+   src->Register.Index = bracket[0].index;
+   if (bracket[0].ind_file != TGSI_FILE_NULL) {
+      src->Register.Indirect = 1;
+      src->Indirect.File = bracket[0].ind_file;
+      src->Indirect.Index = bracket[0].ind_index;
+      src->Indirect.SwizzleX = bracket[0].ind_comp;
+      src->Indirect.SwizzleY = bracket[0].ind_comp;
+      src->Indirect.SwizzleZ = bracket[0].ind_comp;
+      src->Indirect.SwizzleW = bracket[0].ind_comp;
    }
-
-   if (!parse_register_src(ctx, &file, &index, &ind_file, &ind_index, &ind_comp))
-      return FALSE;
-   src->SrcRegister.File = file;
-   src->SrcRegister.Index = index;
-   if (ind_file != TGSI_FILE_NULL) {
-      src->SrcRegister.Indirect = 1;
-      src->SrcRegisterInd.File = ind_file;
-      src->SrcRegisterInd.Index = ind_index;
-      src->SrcRegisterInd.SwizzleX = ind_comp;
-      src->SrcRegisterInd.SwizzleY = ind_comp;
-      src->SrcRegisterInd.SwizzleZ = ind_comp;
-      src->SrcRegisterInd.SwizzleW = ind_comp;
+   if (parsed_opt_brackets) {
+      src->Register.Dimension = 1;
+      src->Dimension.Indirect = 0;
+      src->Dimension.Dimension = 0;
+      src->Dimension.Index = bracket[1].index;
    }
 
    /* Parse optional swizzle.
     */
-   if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle, &parsed_extswizzle )) {
-      if (parsed_extswizzle) {
-         assert( parsed_swizzle );
-
-         src->SrcRegisterExtSwz.ExtSwizzleX = swizzle[0];
-         src->SrcRegisterExtSwz.ExtSwizzleY = swizzle[1];
-         src->SrcRegisterExtSwz.ExtSwizzleZ = swizzle[2];
-         src->SrcRegisterExtSwz.ExtSwizzleW = swizzle[3];
-      }
-      else if (parsed_swizzle) {
-         src->SrcRegister.SwizzleX = swizzle[0];
-         src->SrcRegister.SwizzleY = swizzle[1];
-         src->SrcRegister.SwizzleZ = swizzle[2];
-         src->SrcRegister.SwizzleW = swizzle[3];
-      }
-   }
-
-   if (src->SrcRegisterExtMod.Complement) {
-      eat_opt_white( &ctx->cur );
-      if (*ctx->cur != ')') {
-         report_error( ctx, "Expected `)'" );
-         return FALSE;
-      }
-      ctx->cur++;
-   }
-
-   if (src->SrcRegisterExtMod.Bias) {
-      eat_opt_white( &ctx->cur );
-      if (*ctx->cur != ')') {
-         report_error( ctx, "Expected `)'" );
-         return FALSE;
-      }
-      ctx->cur++;
-      eat_opt_white( &ctx->cur );
-      if (*ctx->cur != '-') {
-         report_error( ctx, "Expected `-'" );
-         return FALSE;
-      }
-      ctx->cur++;
-      eat_opt_white( &ctx->cur );
-      if (!parse_float( &ctx->cur, &value )) {
-         report_error( ctx, "Expected literal floating point" );
-         return FALSE;
-      }
-      if (value != 0.5f) {
-         report_error( ctx, "Expected 0.5" );
-         return FALSE;
+   if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle )) {
+      if (parsed_swizzle) {
+         src->Register.SwizzleX = swizzle[0];
+         src->Register.SwizzleY = swizzle[1];
+         src->Register.SwizzleZ = swizzle[2];
+         src->Register.SwizzleW = swizzle[3];
       }
    }
 
-   if (src->SrcRegisterExtMod.Scale2X) {
-      eat_opt_white( &ctx->cur );
-      if (*ctx->cur != ')') {
-         report_error( ctx, "Expected `)'" );
-         return FALSE;
-      }
-      ctx->cur++;
-   }
-
-   if (src->SrcRegisterExtMod.Absolute) {
+   if (src->Register.Absolute) {
       eat_opt_white( &ctx->cur );
       if (*ctx->cur != '|') {
          report_error( ctx, "Expected `|'" );
@@ -758,14 +775,6 @@ parse_src_operand(
       ctx->cur++;
    }
 
-   if (parsed_ext_negate_paren) {
-      eat_opt_white( &ctx->cur );
-      if (*ctx->cur != ')') {
-         report_error( ctx, "Expected `)'" );
-         return FALSE;
-      }
-      ctx->cur++;
-   }
 
    return TRUE;
 }
@@ -857,11 +866,11 @@ parse_instruction(
       }
 
       if (i < info->num_dst) {
-         if (!parse_dst_operand( ctx, &inst.FullDstRegisters[i] ))
+         if (!parse_dst_operand( ctx, &inst.Dst[i] ))
             return FALSE;
       }
       else if (i < info->num_dst + info->num_src) {
-         if (!parse_src_operand( ctx, &inst.FullSrcRegisters[i - info->num_dst] ))
+         if (!parse_src_operand( ctx, &inst.Src[i - info->num_dst] ))
             return FALSE;
       }
       else {
@@ -870,7 +879,8 @@ parse_instruction(
          for (j = 0; j < TGSI_TEXTURE_COUNT; j++) {
             if (str_match_no_case( &ctx->cur, texture_names[j] )) {
                if (!is_digit_alpha_underscore( ctx->cur )) {
-                  inst.InstructionExtTexture.Texture = j;
+                  inst.Instruction.Texture = 1;
+                  inst.Texture.Texture = j;
                   break;
                }
             }
@@ -896,7 +906,8 @@ parse_instruction(
          report_error( ctx, "Expected a label" );
          return FALSE;
       }
-      inst.InstructionExtLabel.Label = target;
+      inst.Instruction.Label = 1;
+      inst.Label.Label = target;
    }
 
    advance = tgsi_build_full_instruction(
@@ -920,7 +931,9 @@ static const char *semantic_names[TGSI_SEMANTIC_COUNT] =
    "PSIZE",
    "GENERIC",
    "NORMAL",
-   "FACE"
+   "FACE",
+   "EDGEFLAG",
+   "PRIM_ID"
 };
 
 static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] =
@@ -934,8 +947,8 @@ static boolean parse_declaration( struct translate_ctx *ctx )
 {
    struct tgsi_full_declaration decl;
    uint file;
-   int first;
-   int last;
+   struct parsed_dcl_bracket brackets[2];
+   int num_brackets;
    uint writemask;
    const char *cur;
    uint advance;
@@ -947,7 +960,7 @@ static boolean parse_declaration( struct translate_ctx *ctx )
       report_error( ctx, "Syntax error" );
       return FALSE;
    }
-   if (!parse_register_dcl( ctx, &file, &first, &last ))
+   if (!parse_register_dcl( ctx, &file, brackets, &num_brackets))
       return FALSE;
    if (!parse_opt_writemask( ctx, &writemask ))
       return FALSE;
@@ -955,8 +968,8 @@ static boolean parse_declaration( struct translate_ctx *ctx )
    decl = tgsi_default_full_declaration();
    decl.Declaration.File = file;
    decl.Declaration.UsageMask = writemask;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Range.First = brackets[0].first;
+   decl.Range.Last = brackets[0].last;
 
    cur = ctx->cur;
    eat_opt_white( &cur );
@@ -987,13 +1000,13 @@ static boolean parse_declaration( struct translate_ctx *ctx )
                }
                cur2++;
 
-               decl.Semantic.SemanticIndex = index;
+               decl.Semantic.Index = index;
 
                cur = cur2;
             }
 
             decl.Declaration.Semantic = 1;
-            decl.Semantic.SemanticName = i;
+            decl.Semantic.Name = i;
 
             ctx->cur = cur;
             break;
@@ -1099,6 +1112,110 @@ static boolean parse_immediate( struct translate_ctx *ctx )
    return TRUE;
 }
 
+static const char *property_names[] =
+{
+   "GS_INPUT_PRIMITIVE",
+   "GS_OUTPUT_PRIMITIVE",
+   "GS_MAX_OUTPUT_VERTICES"
+};
+
+static const char *primitive_names[] =
+{
+   "POINTS",
+   "LINES",
+   "LINE_LOOP",
+   "LINE_STRIP",
+   "TRIANGLES",
+   "TRIANGLE_STRIP",
+   "TRIANGLE_FAN",
+   "QUADS",
+   "QUAD_STRIP",
+   "POLYGON"
+};
+
+static boolean
+parse_primitive( const char **pcur, uint *primitive )
+{
+   uint i;
+
+   for (i = 0; i < PIPE_PRIM_MAX; i++) {
+      const char *cur = *pcur;
+
+      if (str_match_no_case( &cur, primitive_names[i])) {
+         *primitive = i;
+         *pcur = cur;
+         return TRUE;
+      }
+   }
+   return FALSE;
+}
+
+
+static boolean parse_property( struct translate_ctx *ctx )
+{
+   struct tgsi_full_property prop;
+   uint property_name;
+   uint values[8];
+   uint advance;
+   char id[64];
+
+   if (!eat_white( &ctx->cur )) {
+      report_error( ctx, "Syntax error" );
+      return FALSE;
+   }
+   if (!parse_identifier( &ctx->cur, id )) {
+      report_error( ctx, "Syntax error" );
+      return FALSE;
+   }
+   for (property_name = 0; property_name < TGSI_PROPERTY_COUNT;
+        ++property_name) {
+      if (streq_nocase_uprcase(property_names[property_name], id)) {
+         break;
+      }
+   }
+   if (property_name >= TGSI_PROPERTY_COUNT) {
+      debug_printf( "\nError: Unknown property : '%s'", id );
+      return FALSE;
+   }
+
+   eat_opt_white( &ctx->cur );
+   switch(property_name) {
+   case TGSI_PROPERTY_GS_INPUT_PRIM:
+   case TGSI_PROPERTY_GS_OUTPUT_PRIM:
+      if (!parse_primitive(&ctx->cur, &values[0] )) {
+         report_error( ctx, "Unknown primitive name as property!" );
+         return FALSE;
+      }
+      if (property_name == TGSI_PROPERTY_GS_INPUT_PRIM &&
+          ctx->processor == TGSI_PROCESSOR_GEOMETRY) {
+         ctx->implied_array_size = u_vertices_per_prim(values[0]);
+      }
+      break;
+   default:
+      if (!parse_uint(&ctx->cur, &values[0] )) {
+         report_error( ctx, "Expected unsigned integer as property!" );
+         return FALSE;
+      }
+   }
+
+   prop = tgsi_default_full_property();
+   prop.Property.PropertyName = property_name;
+   prop.Property.NrTokens += 1;
+   prop.u[0].Data = values[0];
+
+   advance = tgsi_build_full_property(
+      &prop,
+      ctx->tokens_cur,
+      ctx->header,
+      (uint) (ctx->tokens_end - ctx->tokens_cur) );
+   if (advance == 0)
+      return FALSE;
+   ctx->tokens_cur += advance;
+
+   return TRUE;
+}
+
+
 static boolean translate( struct translate_ctx *ctx )
 {
    eat_opt_white( &ctx->cur );
@@ -1107,7 +1224,6 @@ static boolean translate( struct translate_ctx *ctx )
 
    while (*ctx->cur != '\0') {
       uint label_val = 0;
-
       if (!eat_white( &ctx->cur )) {
          report_error( ctx, "Syntax error" );
          return FALSE;
@@ -1115,7 +1231,6 @@ static boolean translate( struct translate_ctx *ctx )
 
       if (*ctx->cur == '\0')
          break;
-
       if (parse_label( ctx, &label_val )) {
          if (!parse_instruction( ctx, TRUE ))
             return FALSE;
@@ -1128,6 +1243,10 @@ static boolean translate( struct translate_ctx *ctx )
          if (!parse_immediate( ctx ))
             return FALSE;
       }
+      else if (str_match_no_case( &ctx->cur, "PROPERTY" )) {
+         if (!parse_property( ctx ))
+            return FALSE;
+      }
       else if (!parse_instruction( ctx, FALSE )) {
          return FALSE;
       }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c
index bc9c18fd4a..ae875f29ab 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_transform.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c
@@ -79,6 +79,19 @@ emit_immediate(struct tgsi_transform_context *ctx,
 }
 
 
+static void
+emit_property(struct tgsi_transform_context *ctx,
+              const struct tgsi_full_property *prop)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_property(prop,
+                                  ctx->tokens_out + ti,
+                                  ctx->header,
+                                  ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
 
 /**
  * Apply user-defined transformations to the input shader to produce
@@ -110,6 +123,7 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in,
    ctx->emit_instruction = emit_instruction;
    ctx->emit_declaration = emit_declaration;
    ctx->emit_immediate = emit_immediate;
+   ctx->emit_property = emit_property;
    ctx->tokens_out = tokens_out;
    ctx->max_tokens_out = max_tokens_out;
 
@@ -130,15 +144,13 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in,
    /**
     **  Setup output shader
     **/
-   *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version();
-
-   ctx->header = (struct tgsi_header *) (tokens_out + 1);
+   ctx->header = (struct tgsi_header *)tokens_out;
    *ctx->header = tgsi_build_header();
 
-   processor = (struct tgsi_processor *) (tokens_out + 2);
+   processor = (struct tgsi_processor *) (tokens_out + 1);
    *processor = tgsi_build_processor( procType, ctx->header );
 
-   ctx->ti = 3;
+   ctx->ti = 2;
 
 
    /**
@@ -184,6 +196,17 @@ tgsi_transform_shader(const struct tgsi_token *tokens_in,
                ctx->emit_immediate(ctx, fullimm);
          }
          break;
+      case TGSI_TOKEN_TYPE_PROPERTY:
+         {
+            struct tgsi_full_property *fullprop
+               = &parse.FullToken.FullProperty;
+
+            if (ctx->transform_property)
+               ctx->transform_property(ctx, fullprop);
+            else
+               ctx->emit_property(ctx, fullprop);
+         }
+         break;
 
       default:
          assert( 0 );
@@ -215,7 +238,7 @@ tgsi_transform_foo( struct tgsi_token *tokens_out,
                     uint max_tokens_out )
 {
    const char *text = 
-      "FRAG1.1\n"
+      "FRAG\n"
       "DCL IN[0], COLOR, CONSTANT\n"
       "DCL OUT[0], COLOR\n"
       "  0: MOV OUT[0], IN[0]\n"
diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.h b/src/gallium/auxiliary/tgsi/tgsi_transform.h
index a121adbaef..818478e277 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_transform.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_transform.h
@@ -53,6 +53,8 @@ struct tgsi_transform_context
 
    void (*transform_immediate)(struct tgsi_transform_context *ctx,
                                struct tgsi_full_immediate *imm);
+   void (*transform_property)(struct tgsi_transform_context *ctx,
+                              struct tgsi_full_property *prop);
 
    /**
     * Called at end of input program to allow caller to append extra
@@ -73,6 +75,8 @@ struct tgsi_transform_context
                             const struct tgsi_full_declaration *decl);
    void (*emit_immediate)(struct tgsi_transform_context *ctx,
                           const struct tgsi_full_immediate *imm);
+   void (*emit_property)(struct tgsi_transform_context *ctx,
+                         const struct tgsi_full_property *prop);
 
    struct tgsi_header *header;
    uint max_tokens_out;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
index 654426a903..e64e2b731d 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
@@ -29,6 +29,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
 #include "tgsi/tgsi_ureg.h"
+#include "tgsi/tgsi_build.h"
 #include "tgsi/tgsi_info.h"
 #include "tgsi/tgsi_dump.h"
 #include "tgsi/tgsi_sanity.h"
@@ -36,7 +37,6 @@
 #include "util/u_math.h"
 
 union tgsi_any_token {
-   struct tgsi_version version;
    struct tgsi_header header;
    struct tgsi_processor processor;
    struct tgsi_token token;
@@ -46,18 +46,12 @@ union tgsi_any_token {
    struct tgsi_immediate imm;
    union  tgsi_immediate_data imm_data;
    struct tgsi_instruction insn;
-   struct tgsi_instruction_ext_nv insn_ext_nv;
-   struct tgsi_instruction_ext_label insn_ext_label;
-   struct tgsi_instruction_ext_texture insn_ext_texture;
-   struct tgsi_instruction_ext_predicate insn_ext_predicate;
+   struct tgsi_instruction_predicate insn_predicate;
+   struct tgsi_instruction_label insn_label;
+   struct tgsi_instruction_texture insn_texture;
    struct tgsi_src_register src;
-   struct tgsi_src_register_ext_swz src_ext_swz;
-   struct tgsi_src_register_ext_mod src_ext_mod;
    struct tgsi_dimension dim;
    struct tgsi_dst_register dst;
-   struct tgsi_dst_register_ext_concode dst_ext_code;
-   struct tgsi_dst_register_ext_modulate dst_ext_mod;
-   struct tgsi_dst_register_ext_predicate dst_ext_pred;
    unsigned value;
 };
 
@@ -75,6 +69,8 @@ struct ureg_tokens {
 #define UREG_MAX_IMMEDIATE 32
 #define UREG_MAX_TEMP 256
 #define UREG_MAX_ADDR 2
+#define UREG_MAX_LOOP 1
+#define UREG_MAX_PRED 1
 
 #define DOMAIN_DECL 0
 #define DOMAIN_INSN 1
@@ -94,14 +90,24 @@ struct ureg_program
    unsigned vs_inputs[UREG_MAX_INPUT/32];
 
    struct {
+      unsigned index;
+   } gs_input[UREG_MAX_INPUT];
+   unsigned nr_gs_inputs;
+
+   struct {
       unsigned semantic_name;
       unsigned semantic_index;
    } output[UREG_MAX_OUTPUT];
    unsigned nr_outputs;
 
    struct {
-      float v[4];
+      union {
+         float f[4];
+         unsigned u[4];
+         int i[4];
+      } value;
       unsigned nr;
+      unsigned type;
    } immediate[UREG_MAX_IMMEDIATE];
    unsigned nr_immediates;
 
@@ -118,6 +124,8 @@ struct ureg_program
    unsigned nr_constant_ranges;
 
    unsigned nr_addrs;
+   unsigned nr_preds;
+   unsigned nr_loops;
    unsigned nr_instructions;
 
    struct ureg_tokens domain[2];
@@ -141,8 +149,9 @@ static void tokens_expand( struct ureg_tokens *tokens,
 {
    unsigned old_size = tokens->size * sizeof(unsigned);
 
-   if (tokens->tokens == error_tokens)
-      goto fail;
+   if (tokens->tokens == error_tokens) {
+      return;
+   }
 
    while (tokens->count + count > tokens->size) {
       tokens->size = (1 << ++tokens->order);
@@ -151,13 +160,9 @@ static void tokens_expand( struct ureg_tokens *tokens,
    tokens->tokens = REALLOC(tokens->tokens, 
                             old_size,
                             tokens->size * sizeof(unsigned));
-   if (tokens->tokens == NULL) 
-      goto fail;
-
-   return;
-          
-fail:
-   tokens_error(tokens);
+   if (tokens->tokens == NULL) {
+      tokens_error(tokens);
+   }
 }
 
 static void set_bad( struct ureg_program *ureg )
@@ -207,9 +212,13 @@ ureg_dst_register( unsigned file,
    dst.IndirectIndex = 0;
    dst.IndirectSwizzle = 0;
    dst.Saturate  = 0;
+   dst.Predicate = 0;
+   dst.PredNegate = 0;
+   dst.PredSwizzleX = TGSI_SWIZZLE_X;
+   dst.PredSwizzleY = TGSI_SWIZZLE_Y;
+   dst.PredSwizzleZ = TGSI_SWIZZLE_Z;
+   dst.PredSwizzleW = TGSI_SWIZZLE_W;
    dst.Index     = index;
-   dst.Pad1      = 0;
-   dst.Pad2      = 0;
 
    return dst;
 }
@@ -279,6 +288,22 @@ ureg_DECL_vs_input( struct ureg_program *ureg,
 }
 
 
+struct ureg_src
+ureg_DECL_gs_input(struct ureg_program *ureg,
+                   unsigned index)
+{
+   if (ureg->nr_gs_inputs < UREG_MAX_INPUT) {
+      ureg->gs_input[ureg->nr_gs_inputs].index = index;
+      ureg->nr_gs_inputs++;
+   } else {
+      set_bad(ureg);
+   }
+
+   /* XXX: Add suport for true 2D input registers. */
+   return ureg_src_register(TGSI_FILE_INPUT, index);
+}
+
+
 struct ureg_dst 
 ureg_DECL_output( struct ureg_program *ureg,
                   unsigned name,
@@ -350,6 +375,7 @@ struct ureg_src ureg_DECL_constant(struct ureg_program *ureg,
       i = ureg->nr_constant_ranges++;
       ureg->constant_range[i].first = index;
       ureg->constant_range[i].last = index;
+      goto out;
    }
 
    /* Collapse all ranges down to one:
@@ -417,6 +443,32 @@ struct ureg_dst ureg_DECL_address( struct ureg_program *ureg )
    return ureg_dst_register( TGSI_FILE_ADDRESS, 0 );
 }
 
+/* Allocate a new loop register.
+ */
+struct ureg_dst
+ureg_DECL_loop(struct ureg_program *ureg)
+{
+   if (ureg->nr_loops < UREG_MAX_LOOP) {
+      return ureg_dst_register(TGSI_FILE_LOOP, ureg->nr_loops++);
+   }
+
+   assert(0);
+   return ureg_dst_register(TGSI_FILE_LOOP, 0);
+}
+
+/* Allocate a new predicate register.
+ */
+struct ureg_dst
+ureg_DECL_predicate(struct ureg_program *ureg)
+{
+   if (ureg->nr_preds < UREG_MAX_PRED) {
+      return ureg_dst_register(TGSI_FILE_PREDICATE, ureg->nr_preds++);
+   }
+
+   assert(0);
+   return ureg_dst_register(TGSI_FILE_PREDICATE, 0);
+}
+
 /* Allocate a new sampler.
  */
 struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg,
@@ -439,22 +491,22 @@ struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg,
 }
 
 
-
-
-static int match_or_expand_immediate( const float *v,
-                                      unsigned nr,
-                                      float *v2,
-                                      unsigned *nr2,
-                                      unsigned *swizzle )
+static int
+match_or_expand_immediate( const unsigned *v,
+                           unsigned nr,
+                           unsigned *v2,
+                           unsigned *pnr2,
+                           unsigned *swizzle )
 {
+   unsigned nr2 = *pnr2;
    unsigned i, j;
-   
+
    *swizzle = 0;
 
    for (i = 0; i < nr; i++) {
       boolean found = FALSE;
 
-      for (j = 0; j < *nr2 && !found; j++) {
+      for (j = 0; j < nr2 && !found; j++) {
          if (v[i] == v2[j]) {
             *swizzle |= j << (i * 2);
             found = TRUE;
@@ -462,24 +514,28 @@ static int match_or_expand_immediate( const float *v,
       }
 
       if (!found) {
-         if (*nr2 >= 4) 
+         if (nr2 >= 4) {
             return FALSE;
+         }
 
-         v2[*nr2] = v[i];
-         *swizzle |= *nr2 << (i * 2);
-         (*nr2)++;
+         v2[nr2] = v[i];
+         *swizzle |= nr2 << (i * 2);
+         nr2++;
       }
    }
 
+   /* Actually expand immediate only when fully succeeded.
+    */
+   *pnr2 = nr2;
    return TRUE;
 }
 
 
-
-
-struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, 
-                                     const float *v,
-                                     unsigned nr )
+static struct ureg_src
+decl_immediate( struct ureg_program *ureg,
+                const unsigned *v,
+                unsigned nr,
+                unsigned type )
 {
    unsigned i, j;
    unsigned swizzle;
@@ -489,38 +545,82 @@ struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg,
     */
 
    for (i = 0; i < ureg->nr_immediates; i++) {
-      if (match_or_expand_immediate( v, 
-                                     nr,
-                                     ureg->immediate[i].v,
-                                     &ureg->immediate[i].nr, 
-                                     &swizzle ))
+      if (ureg->immediate[i].type != type) {
+         continue;
+      }
+      if (match_or_expand_immediate(v,
+                                    nr,
+                                    ureg->immediate[i].value.u,
+                                    &ureg->immediate[i].nr,
+                                    &swizzle)) {
          goto out;
+      }
    }
 
    if (ureg->nr_immediates < UREG_MAX_IMMEDIATE) {
       i = ureg->nr_immediates++;
-      if (match_or_expand_immediate( v,
-                                     nr,
-                                     ureg->immediate[i].v,
-                                     &ureg->immediate[i].nr, 
-                                     &swizzle ))
+      ureg->immediate[i].type = type;
+      if (match_or_expand_immediate(v,
+                                    nr,
+                                    ureg->immediate[i].value.u,
+                                    &ureg->immediate[i].nr,
+                                    &swizzle)) {
          goto out;
+      }
    }
 
-   set_bad( ureg );
+   set_bad(ureg);
 
 out:
    /* Make sure that all referenced elements are from this immediate.
     * Has the effect of making size-one immediates into scalars.
     */
-   for (j = nr; j < 4; j++)
+   for (j = nr; j < 4; j++) {
       swizzle |= (swizzle & 0x3) << (j * 2);
+   }
+
+   return ureg_swizzle(ureg_src_register(TGSI_FILE_IMMEDIATE, i),
+                       (swizzle >> 0) & 0x3,
+                       (swizzle >> 2) & 0x3,
+                       (swizzle >> 4) & 0x3,
+                       (swizzle >> 6) & 0x3);
+}
+
+
+struct ureg_src
+ureg_DECL_immediate( struct ureg_program *ureg,
+                     const float *v,
+                     unsigned nr )
+{
+   union {
+      float f[4];
+      unsigned u[4];
+   } fu;
+   unsigned int i;
+
+   for (i = 0; i < nr; i++) {
+      fu.f[i] = v[i];
+   }
+
+   return decl_immediate(ureg, fu.u, nr, TGSI_IMM_FLOAT32);
+}
 
-   return ureg_swizzle( ureg_src_register( TGSI_FILE_IMMEDIATE, i ),
-                        (swizzle >> 0) & 0x3,
-                        (swizzle >> 2) & 0x3,
-                        (swizzle >> 4) & 0x3,
-                        (swizzle >> 6) & 0x3);
+
+struct ureg_src
+ureg_DECL_immediate_uint( struct ureg_program *ureg,
+                          const unsigned *v,
+                          unsigned nr )
+{
+   return decl_immediate(ureg, v, nr, TGSI_IMM_UINT32);
+}
+
+
+struct ureg_src
+ureg_DECL_immediate_int( struct ureg_program *ureg,
+                         const int *v,
+                         unsigned nr )
+{
+   return decl_immediate(ureg, (const unsigned *)v, nr, TGSI_IMM_INT32);
 }
 
 
@@ -528,9 +628,7 @@ void
 ureg_emit_src( struct ureg_program *ureg,
                struct ureg_src src )
 {
-   unsigned size = (1 + 
-                    (src.Absolute ? 1 : 0) +
-                    (src.Indirect ? 1 : 0));
+   unsigned size = 1 + (src.Indirect ? 1 : 0);
 
    union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size );
    unsigned n = 0;
@@ -547,17 +645,8 @@ ureg_emit_src( struct ureg_program *ureg,
    out[n].src.SwizzleW = src.SwizzleW;
    out[n].src.Index = src.Index;
    out[n].src.Negate = src.Negate;
+   out[0].src.Absolute = src.Absolute;
    n++;
-   
-   if (src.Absolute) {
-      out[0].src.Extended = 1;
-      out[0].src.Negate = 0;
-      out[n].value = 0;
-      out[n].src_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD;
-      out[n].src_ext_mod.Absolute = 1;
-      out[n].src_ext_mod.Negate = src.Negate;
-      n++;
-   }
 
    if (src.Indirect) {
       out[0].src.Indirect = 1;
@@ -628,37 +717,54 @@ static void validate( unsigned opcode,
 #endif
 }
 
-unsigned
+struct ureg_emit_insn_result
 ureg_emit_insn(struct ureg_program *ureg,
                unsigned opcode,
                boolean saturate,
+               boolean predicate,
+               boolean pred_negate,
+               unsigned pred_swizzle_x,
+               unsigned pred_swizzle_y,
+               unsigned pred_swizzle_z,
+               unsigned pred_swizzle_w,
                unsigned num_dst,
                unsigned num_src )
 {
    union tgsi_any_token *out;
+   uint count = predicate ? 2 : 1;
+   struct ureg_emit_insn_result result;
 
    validate( opcode, num_dst, num_src );
    
-   out = get_tokens( ureg, DOMAIN_INSN, 1 );
-   out[0].value = 0;
-   out[0].insn.Type = TGSI_TOKEN_TYPE_INSTRUCTION;
-   out[0].insn.NrTokens = 0;
+   out = get_tokens( ureg, DOMAIN_INSN, count );
+   out[0].insn = tgsi_default_instruction();
    out[0].insn.Opcode = opcode;
    out[0].insn.Saturate = saturate;
    out[0].insn.NumDstRegs = num_dst;
    out[0].insn.NumSrcRegs = num_src;
-   out[0].insn.Padding = 0;
-   out[0].insn.Extended = 0;
-   
+
+   result.insn_token = ureg->domain[DOMAIN_INSN].count - count;
+   result.extended_token = result.insn_token;
+
+   if (predicate) {
+      out[0].insn.Predicate = 1;
+      out[1].insn_predicate = tgsi_default_instruction_predicate();
+      out[1].insn_predicate.Negate = pred_negate;
+      out[1].insn_predicate.SwizzleX = pred_swizzle_x;
+      out[1].insn_predicate.SwizzleY = pred_swizzle_y;
+      out[1].insn_predicate.SwizzleZ = pred_swizzle_z;
+      out[1].insn_predicate.SwizzleW = pred_swizzle_w;
+   }
+
    ureg->nr_instructions++;
-   
-   return ureg->domain[DOMAIN_INSN].count - 1;
+
+   return result;
 }
 
 
 void
 ureg_emit_label(struct ureg_program *ureg,
-                unsigned insn_token,
+                unsigned extended_token,
                 unsigned *label_token )
 {
    union tgsi_any_token *out, *insn;
@@ -667,13 +773,11 @@ ureg_emit_label(struct ureg_program *ureg,
       return;
 
    out = get_tokens( ureg, DOMAIN_INSN, 1 );
-   insn = retrieve_token( ureg, DOMAIN_INSN, insn_token );
+   out[0].value = 0;
 
-   insn->insn.Extended = 1;
+   insn = retrieve_token( ureg, DOMAIN_INSN, extended_token );
+   insn->insn.Label = 1;
 
-   out[0].value = 0;
-   out[0].insn_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL;
-   
    *label_token = ureg->domain[DOMAIN_INSN].count - 1;
 }
 
@@ -696,26 +800,24 @@ ureg_fixup_label(struct ureg_program *ureg,
 {
    union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, label_token );
 
-   assert(out->insn_ext_label.Type == TGSI_INSTRUCTION_EXT_TYPE_LABEL);
-   out->insn_ext_label.Label = instruction_number;
+   out->insn_label.Label = instruction_number;
 }
 
 
 void
 ureg_emit_texture(struct ureg_program *ureg,
-                  unsigned insn_token,
+                  unsigned extended_token,
                   unsigned target )
 {
    union tgsi_any_token *out, *insn;
 
    out = get_tokens( ureg, DOMAIN_INSN, 1 );
-   insn = retrieve_token( ureg, DOMAIN_INSN, insn_token );
+   insn = retrieve_token( ureg, DOMAIN_INSN, extended_token );
 
-   insn->insn.Extended = 1;
+   insn->insn.Texture = 1;
 
    out[0].value = 0;
-   out[0].insn_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE;
-   out[0].insn_ext_texture.Texture = target;
+   out[0].insn_texture.Texture = target;
 }
 
 
@@ -738,12 +840,34 @@ ureg_insn(struct ureg_program *ureg,
           const struct ureg_src *src,
           unsigned nr_src )
 {
-   unsigned insn, i;
+   struct ureg_emit_insn_result insn;
+   unsigned i;
    boolean saturate;
+   boolean predicate;
+   boolean negate = FALSE;
+   unsigned swizzle[4] = { 0 };
 
    saturate = nr_dst ? dst[0].Saturate : FALSE;
+   predicate = nr_dst ? dst[0].Predicate : FALSE;
+   if (predicate) {
+      negate = dst[0].PredNegate;
+      swizzle[0] = dst[0].PredSwizzleX;
+      swizzle[1] = dst[0].PredSwizzleY;
+      swizzle[2] = dst[0].PredSwizzleZ;
+      swizzle[3] = dst[0].PredSwizzleW;
+   }
 
-   insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src );
+   insn = ureg_emit_insn(ureg,
+                         opcode,
+                         saturate,
+                         predicate,
+                         negate,
+                         swizzle[0],
+                         swizzle[1],
+                         swizzle[2],
+                         swizzle[3],
+                         nr_dst,
+                         nr_src);
 
    for (i = 0; i < nr_dst; i++)
       ureg_emit_dst( ureg, dst[i] );
@@ -751,7 +875,7 @@ ureg_insn(struct ureg_program *ureg,
    for (i = 0; i < nr_src; i++)
       ureg_emit_src( ureg, src[i] );
 
-   ureg_fixup_insn_size( ureg, insn );
+   ureg_fixup_insn_size( ureg, insn.insn_token );
 }
 
 void
@@ -763,14 +887,36 @@ ureg_tex_insn(struct ureg_program *ureg,
               const struct ureg_src *src,
               unsigned nr_src )
 {
-   unsigned insn, i;
+   struct ureg_emit_insn_result insn;
+   unsigned i;
    boolean saturate;
+   boolean predicate;
+   boolean negate = FALSE;
+   unsigned swizzle[4] = { 0 };
 
    saturate = nr_dst ? dst[0].Saturate : FALSE;
+   predicate = nr_dst ? dst[0].Predicate : FALSE;
+   if (predicate) {
+      negate = dst[0].PredNegate;
+      swizzle[0] = dst[0].PredSwizzleX;
+      swizzle[1] = dst[0].PredSwizzleY;
+      swizzle[2] = dst[0].PredSwizzleZ;
+      swizzle[3] = dst[0].PredSwizzleW;
+   }
 
-   insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src );
+   insn = ureg_emit_insn(ureg,
+                         opcode,
+                         saturate,
+                         predicate,
+                         negate,
+                         swizzle[0],
+                         swizzle[1],
+                         swizzle[2],
+                         swizzle[3],
+                         nr_dst,
+                         nr_src);
 
-   ureg_emit_texture( ureg, insn, target );                             \
+   ureg_emit_texture( ureg, insn.extended_token, target );
 
    for (i = 0; i < nr_dst; i++)
       ureg_emit_dst( ureg, dst[i] );
@@ -778,7 +924,7 @@ ureg_tex_insn(struct ureg_program *ureg,
    for (i = 0; i < nr_src; i++)
       ureg_emit_src( ureg, src[i] );
 
-   ureg_fixup_insn_size( ureg, insn );
+   ureg_fixup_insn_size( ureg, insn.insn_token );
 }
 
 
@@ -789,16 +935,27 @@ ureg_label_insn(struct ureg_program *ureg,
                 unsigned nr_src,
                 unsigned *label_token )
 {
-   unsigned insn, i;
+   struct ureg_emit_insn_result insn;
+   unsigned i;
 
-   insn = ureg_emit_insn( ureg, opcode, FALSE, 0, nr_src );
+   insn = ureg_emit_insn(ureg,
+                         opcode,
+                         FALSE,
+                         FALSE,
+                         FALSE,
+                         TGSI_SWIZZLE_X,
+                         TGSI_SWIZZLE_Y,
+                         TGSI_SWIZZLE_Z,
+                         TGSI_SWIZZLE_W,
+                         0,
+                         nr_src);
 
-   ureg_emit_label( ureg, insn, label_token );                  \
+   ureg_emit_label( ureg, insn.extended_token, label_token );
 
    for (i = 0; i < nr_src; i++)
       ureg_emit_src( ureg, src[i] );
 
-   ureg_fixup_insn_size( ureg, insn );
+   ureg_fixup_insn_size( ureg, insn.insn_token );
 }
 
 
@@ -825,8 +982,8 @@ static void emit_decl( struct ureg_program *ureg,
       out[1].decl_range.Last = index;
 
    out[2].value = 0;
-   out[2].decl_semantic.SemanticName = semantic_name;
-   out[2].decl_semantic.SemanticIndex = semantic_index;
+   out[2].decl_semantic.Name = semantic_name;
+   out[2].decl_semantic.Index = semantic_index;
 
 }
 
@@ -851,22 +1008,23 @@ static void emit_decl_range( struct ureg_program *ureg,
    out[1].decl_range.Last = first + count - 1;
 }
 
-static void emit_immediate( struct ureg_program *ureg,
-                            const float *v )
+static void
+emit_immediate( struct ureg_program *ureg,
+                const unsigned *v,
+                unsigned type )
 {
    union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 5 );
 
    out[0].value = 0;
    out[0].imm.Type = TGSI_TOKEN_TYPE_IMMEDIATE;
    out[0].imm.NrTokens = 5;
-   out[0].imm.DataType = TGSI_IMM_FLOAT32;
+   out[0].imm.DataType = type;
    out[0].imm.Padding = 0;
-   out[0].imm.Extended = 0;
 
-   out[1].imm_data.Float = v[0];
-   out[2].imm_data.Float = v[1];
-   out[3].imm_data.Float = v[2];
-   out[4].imm_data.Float = v[3];
+   out[1].imm_data.Uint = v[0];
+   out[2].imm_data.Uint = v[1];
+   out[3].imm_data.Uint = v[2];
+   out[4].imm_data.Uint = v[3];
 }
 
 
@@ -882,8 +1040,7 @@ static void emit_decls( struct ureg_program *ureg )
             emit_decl_range( ureg, TGSI_FILE_INPUT, i, 1 );
          }
       }
-   }
-   else {
+   } else if (ureg->processor == TGSI_PROCESSOR_FRAGMENT) {
       for (i = 0; i < ureg->nr_fs_inputs; i++) {
          emit_decl( ureg, 
                     TGSI_FILE_INPUT, 
@@ -892,6 +1049,13 @@ static void emit_decls( struct ureg_program *ureg )
                     ureg->fs_input[i].semantic_index,
                     ureg->fs_input[i].interp );
       }
+   } else {
+      for (i = 0; i < ureg->nr_gs_inputs; i++) {
+         emit_decl_range(ureg, 
+                         TGSI_FILE_INPUT, 
+                         ureg->gs_input[i].index,
+                         1);
+      }
    }
 
    for (i = 0; i < ureg->nr_outputs; i++) {
@@ -930,9 +1094,24 @@ static void emit_decls( struct ureg_program *ureg )
                        0, ureg->nr_addrs );
    }
 
+   if (ureg->nr_loops) {
+      emit_decl_range(ureg,
+                      TGSI_FILE_LOOP,
+                      0,
+                      ureg->nr_loops);
+   }
+
+   if (ureg->nr_preds) {
+      emit_decl_range(ureg,
+                      TGSI_FILE_PREDICATE,
+                      0,
+                      ureg->nr_preds);
+   }
+
    for (i = 0; i < ureg->nr_immediates; i++) {
       emit_immediate( ureg,
-                      ureg->immediate[i].v );
+                      ureg->immediate[i].value.u,
+                      ureg->immediate[i].type );
    }
 }
 
@@ -955,26 +1134,22 @@ static void copy_instructions( struct ureg_program *ureg )
 static void
 fixup_header_size(struct ureg_program *ureg)
 {
-   union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_DECL, 1 );
+   union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_DECL, 0 );
 
-   out->header.BodySize = ureg->domain[DOMAIN_DECL].count - 3;
+   out->header.BodySize = ureg->domain[DOMAIN_DECL].count - 2;
 }
 
 
 static void
 emit_header( struct ureg_program *ureg )
 {
-   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 );
-
-   out[0].version.MajorVersion = 1;
-   out[0].version.MinorVersion = 1;
-   out[0].version.Padding = 0;
+   union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 2 );
 
-   out[1].header.HeaderSize = 2;
-   out[1].header.BodySize = 0;
+   out[0].header.HeaderSize = 2;
+   out[0].header.BodySize = 0;
 
-   out[2].processor.Processor = ureg->processor;
-   out[2].processor.Padding = 0;
+   out[1].processor.Processor = ureg->processor;
+   out[1].processor.Padding = 0;
 }
 
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
index f04f443b9e..6f11273320 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
@@ -67,9 +67,13 @@ struct ureg_dst
    unsigned WriteMask   : 4;  /* TGSI_WRITEMASK_ */
    unsigned Indirect    : 1;  /* BOOL */
    unsigned Saturate    : 1;  /* BOOL */
+   unsigned Predicate   : 1;
+   unsigned PredNegate  : 1;  /* BOOL */
+   unsigned PredSwizzleX: 2;  /* TGSI_SWIZZLE_ */
+   unsigned PredSwizzleY: 2;  /* TGSI_SWIZZLE_ */
+   unsigned PredSwizzleZ: 2;  /* TGSI_SWIZZLE_ */
+   unsigned PredSwizzleW: 2;  /* TGSI_SWIZZLE_ */
    int      Index       : 16; /* SINT */
-   unsigned Pad1        : 5;
-   unsigned Pad2        : 1;  /* BOOL */
    int      IndirectIndex   : 16; /* SINT */
    int      IndirectSwizzle : 2;  /* TGSI_SWIZZLE_ */
 };
@@ -129,6 +133,10 @@ struct ureg_src
 ureg_DECL_vs_input( struct ureg_program *,
                     unsigned index );
 
+struct ureg_src
+ureg_DECL_gs_input(struct ureg_program *,
+                   unsigned index);
+
 struct ureg_dst
 ureg_DECL_output( struct ureg_program *,
                   unsigned semantic_name,
@@ -140,6 +148,16 @@ ureg_DECL_immediate( struct ureg_program *,
                      unsigned nr );
 
 struct ureg_src
+ureg_DECL_immediate_uint( struct ureg_program *,
+                          const unsigned *v,
+                          unsigned nr );
+
+struct ureg_src
+ureg_DECL_immediate_int( struct ureg_program *,
+                         const int *v,
+                         unsigned nr );
+
+struct ureg_src
 ureg_DECL_constant( struct ureg_program *,
                     unsigned index );
 
@@ -153,6 +171,12 @@ ureg_release_temporary( struct ureg_program *ureg,
 struct ureg_dst
 ureg_DECL_address( struct ureg_program * );
 
+struct ureg_dst
+ureg_DECL_loop( struct ureg_program * );
+
+struct ureg_dst
+ureg_DECL_predicate(struct ureg_program *);
+
 /* Supply an index to the sampler declaration as this is the hook to
  * the external pipe_sampler state.  Users of this function probably
  * don't want just any sampler, but a specific one which they've set
@@ -207,6 +231,90 @@ ureg_imm1f( struct ureg_program *ureg,
    return ureg_DECL_immediate( ureg, v, 1 );
 }
 
+static INLINE struct ureg_src
+ureg_imm4u( struct ureg_program *ureg,
+            unsigned a, unsigned b,
+            unsigned c, unsigned d)
+{
+   unsigned v[4];
+   v[0] = a;
+   v[1] = b;
+   v[2] = c;
+   v[3] = d;
+   return ureg_DECL_immediate_uint( ureg, v, 4 );
+}
+
+static INLINE struct ureg_src
+ureg_imm3u( struct ureg_program *ureg,
+            unsigned a, unsigned b,
+            unsigned c)
+{
+   unsigned v[3];
+   v[0] = a;
+   v[1] = b;
+   v[2] = c;
+   return ureg_DECL_immediate_uint( ureg, v, 3 );
+}
+
+static INLINE struct ureg_src
+ureg_imm2u( struct ureg_program *ureg,
+            unsigned a, unsigned b)
+{
+   unsigned v[2];
+   v[0] = a;
+   v[1] = b;
+   return ureg_DECL_immediate_uint( ureg, v, 2 );
+}
+
+static INLINE struct ureg_src
+ureg_imm1u( struct ureg_program *ureg,
+            unsigned a)
+{
+   return ureg_DECL_immediate_uint( ureg, &a, 1 );
+}
+
+static INLINE struct ureg_src
+ureg_imm4i( struct ureg_program *ureg,
+            int a, int b,
+            int c, int d)
+{
+   int v[4];
+   v[0] = a;
+   v[1] = b;
+   v[2] = c;
+   v[3] = d;
+   return ureg_DECL_immediate_int( ureg, v, 4 );
+}
+
+static INLINE struct ureg_src
+ureg_imm3i( struct ureg_program *ureg,
+            int a, int b,
+            int c)
+{
+   int v[3];
+   v[0] = a;
+   v[1] = b;
+   v[2] = c;
+   return ureg_DECL_immediate_int( ureg, v, 3 );
+}
+
+static INLINE struct ureg_src
+ureg_imm2i( struct ureg_program *ureg,
+            int a, int b)
+{
+   int v[2];
+   v[0] = a;
+   v[1] = b;
+   return ureg_DECL_immediate_int( ureg, v, 2 );
+}
+
+static INLINE struct ureg_src
+ureg_imm1i( struct ureg_program *ureg,
+            int a)
+{
+   return ureg_DECL_immediate_int( ureg, &a, 1 );
+}
+
 /***********************************************************************
  * Functions for patching up labels
  */
@@ -266,10 +374,21 @@ ureg_label_insn(struct ureg_program *ureg,
  * Internal instruction helpers, don't call these directly:
  */
 
-unsigned
+struct ureg_emit_insn_result {
+   unsigned insn_token;       /*< Used to fixup insn size. */
+   unsigned extended_token;   /*< Used to set the Extended bit, usually the same as insn_token. */
+};
+
+struct ureg_emit_insn_result
 ureg_emit_insn(struct ureg_program *ureg,
                unsigned opcode,
                boolean saturate,
+               boolean predicate,
+               boolean pred_negate,
+               unsigned pred_swizzle_x,
+               unsigned pred_swizzle_y,
+               unsigned pred_swizzle_z,
+               unsigned pred_swizzle_w,
                unsigned num_dst,
                unsigned num_src );
 
@@ -300,7 +419,17 @@ ureg_fixup_insn_size(struct ureg_program *ureg,
 static INLINE void ureg_##op( struct ureg_program *ureg )       \
 {                                                               \
    unsigned opcode = TGSI_OPCODE_##op;                          \
-   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \
+   unsigned insn = ureg_emit_insn(ureg,                         \
+                                  opcode,                       \
+                                  FALSE,                        \
+                                  FALSE,                        \
+                                  FALSE,                        \
+                                  TGSI_SWIZZLE_X,               \
+                                  TGSI_SWIZZLE_Y,               \
+                                  TGSI_SWIZZLE_Z,               \
+                                  TGSI_SWIZZLE_W,               \
+                                  0,                            \
+                                  0).insn_token;                \
    ureg_fixup_insn_size( ureg, insn );                          \
 }
 
@@ -309,7 +438,17 @@ static INLINE void ureg_##op( struct ureg_program *ureg,        \
                               struct ureg_src src )             \
 {                                                               \
    unsigned opcode = TGSI_OPCODE_##op;                          \
-   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \
+   unsigned insn = ureg_emit_insn(ureg,                         \
+                                  opcode,                       \
+                                  FALSE,                        \
+                                  FALSE,                        \
+                                  FALSE,                        \
+                                  TGSI_SWIZZLE_X,               \
+                                  TGSI_SWIZZLE_Y,               \
+                                  TGSI_SWIZZLE_Z,               \
+                                  TGSI_SWIZZLE_W,               \
+                                  0,                            \
+                                  1).insn_token;                \
    ureg_emit_src( ureg, src );                                  \
    ureg_fixup_insn_size( ureg, insn );                          \
 }
@@ -319,9 +458,20 @@ static INLINE void ureg_##op( struct ureg_program *ureg,        \
                               unsigned *label_token )           \
 {                                                               \
    unsigned opcode = TGSI_OPCODE_##op;                          \
-   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \
-   ureg_emit_label( ureg, insn, label_token );                  \
-   ureg_fixup_insn_size( ureg, insn );                          \
+   struct ureg_emit_insn_result insn;                           \
+   insn = ureg_emit_insn(ureg,                                  \
+                         opcode,                                \
+                         FALSE,                                 \
+                         FALSE,                                 \
+                         FALSE,                                 \
+                         TGSI_SWIZZLE_X,                        \
+                         TGSI_SWIZZLE_Y,                        \
+                         TGSI_SWIZZLE_Z,                        \
+                         TGSI_SWIZZLE_W,                        \
+                         0,                                     \
+                         0);                                    \
+   ureg_emit_label( ureg, insn.extended_token, label_token );   \
+   ureg_fixup_insn_size( ureg, insn.insn_token );               \
 }
 
 #define OP01_LBL( op )                                          \
@@ -330,10 +480,21 @@ static INLINE void ureg_##op( struct ureg_program *ureg,        \
                               unsigned *label_token )          \
 {                                                               \
    unsigned opcode = TGSI_OPCODE_##op;                          \
-   unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \
-   ureg_emit_label( ureg, insn, label_token );                  \
+   struct ureg_emit_insn_result insn;                           \
+   insn = ureg_emit_insn(ureg,                                  \
+                         opcode,                                \
+                         FALSE,                                 \
+                         FALSE,                                 \
+                         FALSE,                                 \
+                         TGSI_SWIZZLE_X,                        \
+                         TGSI_SWIZZLE_Y,                        \
+                         TGSI_SWIZZLE_Z,                        \
+                         TGSI_SWIZZLE_W,                        \
+                         0,                                     \
+                         1);                                    \
+   ureg_emit_label( ureg, insn.extended_token, label_token );   \
    ureg_emit_src( ureg, src );                                  \
-   ureg_fixup_insn_size( ureg, insn );                          \
+   ureg_fixup_insn_size( ureg, insn.insn_token );               \
 }
 
 #define OP10( op )                                                      \
@@ -341,7 +502,17 @@ static INLINE void ureg_##op( struct ureg_program *ureg,                \
                               struct ureg_dst dst )                     \
 {                                                                       \
    unsigned opcode = TGSI_OPCODE_##op;                                  \
-   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 0 );  \
+   unsigned insn = ureg_emit_insn(ureg,                                 \
+                                  opcode,                               \
+                                  dst.Saturate,                         \
+                                  dst.Predicate,                        \
+                                  dst.PredNegate,                       \
+                                  dst.PredSwizzleX,                     \
+                                  dst.PredSwizzleY,                     \
+                                  dst.PredSwizzleZ,                     \
+                                  dst.PredSwizzleW,                     \
+                                  1,                                    \
+                                  0).insn_token;                        \
    ureg_emit_dst( ureg, dst );                                          \
    ureg_fixup_insn_size( ureg, insn );                                  \
 }
@@ -353,7 +524,17 @@ static INLINE void ureg_##op( struct ureg_program *ureg,                \
                               struct ureg_src src )                     \
 {                                                                       \
    unsigned opcode = TGSI_OPCODE_##op;                                  \
-   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 1 );  \
+   unsigned insn = ureg_emit_insn(ureg,                                 \
+                                  opcode,                               \
+                                  dst.Saturate,                         \
+                                  dst.Predicate,                        \
+                                  dst.PredNegate,                       \
+                                  dst.PredSwizzleX,                     \
+                                  dst.PredSwizzleY,                     \
+                                  dst.PredSwizzleZ,                     \
+                                  dst.PredSwizzleW,                     \
+                                  1,                                    \
+                                  1).insn_token;                        \
    ureg_emit_dst( ureg, dst );                                          \
    ureg_emit_src( ureg, src );                                          \
    ureg_fixup_insn_size( ureg, insn );                                  \
@@ -366,7 +547,17 @@ static INLINE void ureg_##op( struct ureg_program *ureg,                \
                               struct ureg_src src1 )                    \
 {                                                                       \
    unsigned opcode = TGSI_OPCODE_##op;                                  \
-   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 );  \
+   unsigned insn = ureg_emit_insn(ureg,                                 \
+                                  opcode,                               \
+                                  dst.Saturate,                         \
+                                  dst.Predicate,                        \
+                                  dst.PredNegate,                       \
+                                  dst.PredSwizzleX,                     \
+                                  dst.PredSwizzleY,                     \
+                                  dst.PredSwizzleZ,                     \
+                                  dst.PredSwizzleW,                     \
+                                  1,                                    \
+                                  2).insn_token;                        \
    ureg_emit_dst( ureg, dst );                                          \
    ureg_emit_src( ureg, src0 );                                         \
    ureg_emit_src( ureg, src1 );                                         \
@@ -381,12 +572,23 @@ static INLINE void ureg_##op( struct ureg_program *ureg,                \
                               struct ureg_src src1 )                    \
 {                                                                       \
    unsigned opcode = TGSI_OPCODE_##op;                                  \
-   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 );  \
-   ureg_emit_texture( ureg, insn, target );                             \
+   struct ureg_emit_insn_result insn;                                   \
+   insn = ureg_emit_insn(ureg,                                          \
+                         opcode,                                        \
+                         dst.Saturate,                                  \
+                         dst.Predicate,                                 \
+                         dst.PredNegate,                                \
+                         dst.PredSwizzleX,                              \
+                         dst.PredSwizzleY,                              \
+                         dst.PredSwizzleZ,                              \
+                         dst.PredSwizzleW,                              \
+                         1,                                             \
+                         2);                                            \
+   ureg_emit_texture( ureg, insn.extended_token, target );              \
    ureg_emit_dst( ureg, dst );                                          \
    ureg_emit_src( ureg, src0 );                                         \
    ureg_emit_src( ureg, src1 );                                         \
-   ureg_fixup_insn_size( ureg, insn );                                  \
+   ureg_fixup_insn_size( ureg, insn.insn_token );                       \
 }
 
 #define OP13( op )                                                      \
@@ -397,7 +599,17 @@ static INLINE void ureg_##op( struct ureg_program *ureg,                \
                               struct ureg_src src2 )                    \
 {                                                                       \
    unsigned opcode = TGSI_OPCODE_##op;                                  \
-   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 3 );  \
+   unsigned insn = ureg_emit_insn(ureg,                                 \
+                                  opcode,                               \
+                                  dst.Saturate,                         \
+                                  dst.Predicate,                        \
+                                  dst.PredNegate,                       \
+                                  dst.PredSwizzleX,                     \
+                                  dst.PredSwizzleY,                     \
+                                  dst.PredSwizzleZ,                     \
+                                  dst.PredSwizzleW,                     \
+                                  1,                                    \
+                                  3).insn_token;                        \
    ureg_emit_dst( ureg, dst );                                          \
    ureg_emit_src( ureg, src0 );                                         \
    ureg_emit_src( ureg, src1 );                                         \
@@ -415,14 +627,25 @@ static INLINE void ureg_##op( struct ureg_program *ureg,                \
                               struct ureg_src src3 )                    \
 {                                                                       \
    unsigned opcode = TGSI_OPCODE_##op;                                  \
-   unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 4 );  \
-   ureg_emit_texture( ureg, insn, target );                             \
+   struct ureg_emit_insn_result insn;                                   \
+   insn = ureg_emit_insn(ureg,                                          \
+                         opcode,                                        \
+                         dst.Saturate,                                  \
+                         dst.Predicate,                                 \
+                         dst.PredNegate,                                \
+                         dst.PredSwizzleX,                              \
+                         dst.PredSwizzleY,                              \
+                         dst.PredSwizzleZ,                              \
+                         dst.PredSwizzleW,                              \
+                         1,                                             \
+                         4);                                            \
+   ureg_emit_texture( ureg, insn.extended_token, target );              \
    ureg_emit_dst( ureg, dst );                                          \
    ureg_emit_src( ureg, src0 );                                         \
    ureg_emit_src( ureg, src1 );                                         \
    ureg_emit_src( ureg, src2 );                                         \
    ureg_emit_src( ureg, src3 );                                         \
-   ureg_fixup_insn_size( ureg, insn );                                  \
+   ureg_fixup_insn_size( ureg, insn.insn_token );                       \
 }
 
 
@@ -497,6 +720,24 @@ ureg_saturate( struct ureg_dst reg )
    return reg;
 }
 
+static INLINE struct ureg_dst
+ureg_predicate(struct ureg_dst reg,
+               boolean negate,
+               unsigned swizzle_x,
+               unsigned swizzle_y,
+               unsigned swizzle_z,
+               unsigned swizzle_w)
+{
+   assert(reg.File != TGSI_FILE_NULL);
+   reg.Predicate = 1;
+   reg.PredNegate = negate;
+   reg.PredSwizzleX = swizzle_x;
+   reg.PredSwizzleY = swizzle_y;
+   reg.PredSwizzleZ = swizzle_z;
+   reg.PredSwizzleW = swizzle_w;
+   return reg;
+}
+
 static INLINE struct ureg_dst 
 ureg_dst_indirect( struct ureg_dst reg, struct ureg_src addr )
 {
@@ -530,9 +771,13 @@ ureg_dst( struct ureg_src src )
    dst.IndirectIndex = src.IndirectIndex;
    dst.IndirectSwizzle = src.IndirectSwizzle;
    dst.Saturate  = 0;
+   dst.Predicate = 0;
+   dst.PredNegate = 0;
+   dst.PredSwizzleX = TGSI_SWIZZLE_X;
+   dst.PredSwizzleY = TGSI_SWIZZLE_Y;
+   dst.PredSwizzleZ = TGSI_SWIZZLE_Z;
+   dst.PredSwizzleW = TGSI_SWIZZLE_W;
    dst.Index     = src.Index;
-   dst.Pad1      = 0;
-   dst.Pad2      = 0;
 
    return dst;
 }
@@ -571,9 +816,13 @@ ureg_dst_undef( void )
    dst.IndirectIndex = 0;
    dst.IndirectSwizzle = 0;
    dst.Saturate  = 0;
+   dst.Predicate = 0;
+   dst.PredNegate = 0;
+   dst.PredSwizzleX = TGSI_SWIZZLE_X;
+   dst.PredSwizzleY = TGSI_SWIZZLE_Y;
+   dst.PredSwizzleZ = TGSI_SWIZZLE_Z;
+   dst.PredSwizzleW = TGSI_SWIZZLE_W;
    dst.Index     = 0;
-   dst.Pad1      = 0;
-   dst.Pad2      = 0;
 
    return dst;
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c
index 71f8a6ca40..f4ca9e21ed 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_util.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.c
@@ -69,59 +69,15 @@ tgsi_util_get_src_register_swizzle(
    return 0;
 }
 
-unsigned
-tgsi_util_get_src_register_extswizzle(
-   const struct tgsi_src_register_ext_swz *reg,
-   unsigned component )
-{
-   switch( component ) {
-   case 0:
-      return reg->ExtSwizzleX;
-   case 1:
-      return reg->ExtSwizzleY;
-   case 2:
-      return reg->ExtSwizzleZ;
-   case 3:
-      return reg->ExtSwizzleW;
-   default:
-      assert( 0 );
-   }
-   return 0;
-}
 
 unsigned
-tgsi_util_get_full_src_register_extswizzle(
+tgsi_util_get_full_src_register_swizzle(
    const struct tgsi_full_src_register  *reg,
    unsigned component )
 {
-   unsigned swizzle;
-
-   /*
-    * First, calculate  the   extended swizzle for a given channel. This will give
-    * us either a channel index into the simple swizzle or  a constant 1 or   0.
-    */
-   swizzle = tgsi_util_get_src_register_extswizzle(
-      &reg->SrcRegisterExtSwz,
+   return tgsi_util_get_src_register_swizzle(
+      &reg->Register,
       component );
-
-   assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X);
-   assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y);
-   assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z);
-   assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W);
-   assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W);
-   assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W);
-
-   /*
-    * Second, calculate the simple  swizzle  for   the   unswizzled channel index.
-    * Leave the constants intact, they are   not   affected by the   simple swizzle.
-    */
-   if( swizzle <= TGSI_SWIZZLE_W ) {
-      swizzle = tgsi_util_get_src_register_swizzle(
-         &reg->SrcRegister,
-         swizzle );
-   }
-
-   return swizzle;
 }
 
 void
@@ -148,74 +104,6 @@ tgsi_util_set_src_register_swizzle(
    }
 }
 
-void
-tgsi_util_set_src_register_extswizzle(
-   struct tgsi_src_register_ext_swz *reg,
-   unsigned swizzle,
-   unsigned component )
-{
-   switch( component ) {
-   case 0:
-      reg->ExtSwizzleX = swizzle;
-      break;
-   case 1:
-      reg->ExtSwizzleY = swizzle;
-      break;
-   case 2:
-      reg->ExtSwizzleZ = swizzle;
-      break;
-   case 3:
-      reg->ExtSwizzleW = swizzle;
-      break;
-   default:
-      assert( 0 );
-   }
-}
-
-unsigned
-tgsi_util_get_src_register_extnegate(
-   const  struct tgsi_src_register_ext_swz *reg,
-   unsigned component )
-{
-   switch( component ) {
-   case 0:
-      return reg->NegateX;
-   case 1:
-      return reg->NegateY;
-   case 2:
-      return reg->NegateZ;
-   case 3:
-      return reg->NegateW;
-   default:
-      assert( 0 );
-   }
-   return 0;
-}
-
-void
-tgsi_util_set_src_register_extnegate(
-   struct tgsi_src_register_ext_swz *reg,
-   unsigned negate,
-   unsigned component )
-{
-   switch( component ) {
-   case 0:
-      reg->NegateX = negate;
-      break;
-   case 1:
-      reg->NegateY = negate;
-      break;
-   case 2:
-      reg->NegateZ = negate;
-      break;
-   case 3:
-      reg->NegateW = negate;
-      break;
-   default:
-      assert( 0 );
-   }
-}
-
 unsigned
 tgsi_util_get_full_src_register_sign_mode(
    const struct  tgsi_full_src_register *reg,
@@ -223,10 +111,10 @@ tgsi_util_get_full_src_register_sign_mode(
 {
    unsigned sign_mode;
 
-   if( reg->SrcRegisterExtMod.Absolute ) {
+   if( reg->Register.Absolute ) {
       /* Consider only the post-abs negation. */
 
-      if( reg->SrcRegisterExtMod.Negate ) {
+      if( reg->Register.Negate ) {
          sign_mode = TGSI_UTIL_SIGN_SET;
       }
       else {
@@ -234,19 +122,7 @@ tgsi_util_get_full_src_register_sign_mode(
       }
    }
    else {
-      /* Accumulate the three negations. */
-
-      unsigned negate;
-
-      negate = reg->SrcRegister.Negate;
-      if( tgsi_util_get_src_register_extnegate( &reg->SrcRegisterExtSwz, component ) ) {
-         negate = !negate;
-      }
-      if( reg->SrcRegisterExtMod.Negate ) {
-         negate = !negate;
-      }
-
-      if( negate ) {
+      if( reg->Register.Negate ) {
          sign_mode = TGSI_UTIL_SIGN_TOGGLE;
       }
       else {
@@ -262,35 +138,26 @@ tgsi_util_set_full_src_register_sign_mode(
    struct tgsi_full_src_register *reg,
    unsigned sign_mode )
 {
-   reg->SrcRegisterExtSwz.NegateX = 0;
-   reg->SrcRegisterExtSwz.NegateY = 0;
-   reg->SrcRegisterExtSwz.NegateZ = 0;
-   reg->SrcRegisterExtSwz.NegateW = 0;
-
    switch (sign_mode)
    {
    case TGSI_UTIL_SIGN_CLEAR:
-      reg->SrcRegister.Negate = 0;
-      reg->SrcRegisterExtMod.Absolute = 1;
-      reg->SrcRegisterExtMod.Negate = 0;
+      reg->Register.Negate = 0;
+      reg->Register.Absolute = 1;
       break;
 
    case TGSI_UTIL_SIGN_SET:
-      reg->SrcRegister.Negate = 0;
-      reg->SrcRegisterExtMod.Absolute = 1;
-      reg->SrcRegisterExtMod.Negate = 1;
+      reg->Register.Absolute = 1;
+      reg->Register.Negate = 1;
       break;
 
    case TGSI_UTIL_SIGN_TOGGLE:
-      reg->SrcRegister.Negate = 1;
-      reg->SrcRegisterExtMod.Absolute = 0;
-      reg->SrcRegisterExtMod.Negate = 0;
+      reg->Register.Negate = 1;
+      reg->Register.Absolute = 0;
       break;
 
    case TGSI_UTIL_SIGN_KEEP:
-      reg->SrcRegister.Negate = 0;
-      reg->SrcRegisterExtMod.Absolute = 0;
-      reg->SrcRegisterExtMod.Negate = 0;
+      reg->Register.Negate = 0;
+      reg->Register.Absolute = 0;
       break;
 
    default:
diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h
index 21eb656327..19ee2e7cf2 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_util.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.h
@@ -33,7 +33,6 @@ extern "C" {
 #endif
 
 struct tgsi_src_register;
-struct tgsi_src_register_ext_swz;
 struct tgsi_full_src_register;
 
 void *
@@ -45,13 +44,9 @@ tgsi_util_get_src_register_swizzle(
    const struct tgsi_src_register *reg,
    unsigned component );
 
-unsigned
-tgsi_util_get_src_register_extswizzle(
-   const struct tgsi_src_register_ext_swz *reg,
-   unsigned component);
 
 unsigned
-tgsi_util_get_full_src_register_extswizzle(
+tgsi_util_get_full_src_register_swizzle(
    const struct tgsi_full_src_register *reg,
    unsigned component );
 
@@ -61,23 +56,6 @@ tgsi_util_set_src_register_swizzle(
    unsigned swizzle,
    unsigned component );
 
-void
-tgsi_util_set_src_register_extswizzle(
-   struct tgsi_src_register_ext_swz *reg,
-   unsigned swizzle,
-   unsigned component );
-
-unsigned
-tgsi_util_get_src_register_extnegate(
-   const struct tgsi_src_register_ext_swz *reg,
-   unsigned component );
-
-void
-tgsi_util_set_src_register_extnegate(
-   struct tgsi_src_register_ext_swz *reg,
-   unsigned negate,
-   unsigned component );
-
 #define TGSI_UTIL_SIGN_CLEAR    0   /* Force positive */
 #define TGSI_UTIL_SIGN_SET      1   /* Force negative */
 #define TGSI_UTIL_SIGN_TOGGLE   2   /* Negate */
diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile
deleted file mode 100644
index 3c82f8ae03..0000000000
--- a/src/gallium/auxiliary/translate/Makefile
+++ /dev/null
@@ -1,12 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = translate
-
-C_SOURCES = \
-	translate_generic.c \
-	translate_sse.c \
-	translate.c \
-        translate_cache.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript
deleted file mode 100644
index 9553a67537..0000000000
--- a/src/gallium/auxiliary/translate/SConscript
+++ /dev/null
@@ -1,12 +0,0 @@
-Import('*')
-
-translate = env.ConvenienceLibrary(
-	target = 'translate',
-	source = [
-		'translate_generic.c',
-		'translate_sse.c',
-		'translate.c',
-		'translate_cache.c',
-	])
-
-auxiliaries.insert(0, translate)
diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile
deleted file mode 100644
index 1d8bb55bbd..0000000000
--- a/src/gallium/auxiliary/util/Makefile
+++ /dev/null
@@ -1,45 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = util
-
-C_SOURCES = \
-	u_debug.c \
-	u_debug_dump.c \
-	u_debug_symbol.c \
-	u_debug_stack.c \
-	u_blit.c \
-	u_cache.c \
-	u_cpu_detect.c \
-	u_draw_quad.c \
-	u_format.c \
-	u_format_access.c \
-	u_format_table.c \
-	u_gen_mipmap.c \
-	u_handle_table.c \
-	u_hash_table.c \
-	u_hash.c \
-	u_keymap.c \
-	u_linear.c \
-	u_network.c \
-	u_math.c \
-	u_mm.c \
-	u_rect.c \
-	u_simple_shaders.c \
-	u_snprintf.c \
-	u_stream_stdc.c \
-	u_stream_wd.c \
-	u_surface.c \
-	u_tile.c \
-	u_time.c \
-	u_timed_winsys.c \
-	u_upload_mgr.c \
-	u_simple_screen.c
-
-include ../../Makefile.template
-
-u_format_table.c: u_format_table.py u_format_parse.py u_format.csv
-	python u_format_table.py u_format.csv > $@
-
-u_format_access.c: u_format_access.py u_format_parse.py u_format.csv
-	python u_format_access.py u_format.csv > $@
diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript
deleted file mode 100644
index 2187935fa4..0000000000
--- a/src/gallium/auxiliary/util/SConscript
+++ /dev/null
@@ -1,59 +0,0 @@
-Import('*')
-
-env.Clone()
-
-env.Append(CPPPATH = ['.'])
-
-env.CodeGenerate(
-	target = 'u_format_table.c',
-	script = 'u_format_table.py',
-	source = ['u_format.csv'],
-	command = 'python $SCRIPT $SOURCE > $TARGET'
-)
-
-env.CodeGenerate(
-	target = 'u_format_access.c',
-	script = 'u_format_access.py',
-	source = ['u_format.csv'],
-	command = 'python $SCRIPT $SOURCE > $TARGET'
-)
-
-util = env.ConvenienceLibrary(
-	target = 'util',
-	source = [
-		'u_bitmask.c',
-		'u_blit.c',
-		'u_cache.c',
-		'u_cpu_detect.c',
-		'u_debug.c',
-		'u_debug_dump.c',
-		'u_debug_memory.c',
-		'u_debug_profile.c',
-		'u_debug_stack.c',
-		'u_debug_symbol.c',
-		'u_draw_quad.c',
-		'u_format.c',
-		'u_format_access.c',
-		'u_format_table.c',
-		'u_gen_mipmap.c',
-		'u_handle_table.c',
-		'u_hash.c',
-		'u_hash_table.c',
-		'u_keymap.c',
-		'u_network.c',
-		'u_math.c',
-		'u_mm.c',
-		'u_rect.c',
-		'u_simple_shaders.c',
-		'u_snprintf.c',
-		'u_stream_stdc.c',
-		'u_stream_wd.c',
-		'u_surface.c',
-		'u_tile.c',
-		'u_time.c',
-		'u_timed_winsys.c',
-		'u_upload_mgr.c',
-		'u_simple_screen.c',
-	])
-
-auxiliaries.insert(0, util)
diff --git a/src/gallium/auxiliary/util/u_bitmask.c b/src/gallium/auxiliary/util/u_bitmask.c
index 77587c07ec..23c93a3ebc 100644
--- a/src/gallium/auxiliary/util/u_bitmask.c
+++ b/src/gallium/auxiliary/util/u_bitmask.c
@@ -97,12 +97,12 @@ util_bitmask_resize(struct util_bitmask *bm,
    if(!minimum_size)
       return FALSE;
       
-   if(bm->size > minimum_size)
+   if(bm->size >= minimum_size)
       return TRUE;
 
    assert(bm->size % UTIL_BITMASK_BITS_PER_WORD == 0);
    new_size = bm->size;
-   while(!(new_size > minimum_size)) {
+   while(new_size < minimum_size) {
       new_size *= 2;
       /* Check integer overflow */
       if(new_size < bm->size)
@@ -136,7 +136,7 @@ util_bitmask_filled_set(struct util_bitmask *bm,
                         unsigned index)
 {
    assert(bm->filled <= bm->size);
-   assert(index <= bm->size);
+   assert(index < bm->size);
    
    if(index == bm->filled) {
       ++bm->filled;
@@ -149,7 +149,7 @@ util_bitmask_filled_unset(struct util_bitmask *bm,
                           unsigned index)
 {
    assert(bm->filled <= bm->size);
-   assert(index <= bm->size);
+   assert(index < bm->size);
    
    if(index < bm->filled)
       bm->filled = index;
@@ -182,7 +182,7 @@ util_bitmask_add(struct util_bitmask *bm)
       mask = 1;
    }
 found:
-   
+
    /* grow the bitmask if necessary */
    if(!util_bitmask_resize(bm, bm->filled))
       return UTIL_BITMASK_INVALID_INDEX;
@@ -198,9 +198,9 @@ unsigned
 util_bitmask_set(struct util_bitmask *bm, 
                  unsigned index)
 {
-   unsigned word = index / UTIL_BITMASK_BITS_PER_WORD;
-   unsigned bit  = index % UTIL_BITMASK_BITS_PER_WORD;
-   util_bitmask_word mask = 1 << bit;
+   unsigned word;
+   unsigned bit;
+   util_bitmask_word mask;
    
    assert(bm);
    
@@ -208,6 +208,10 @@ util_bitmask_set(struct util_bitmask *bm,
    if(!util_bitmask_resize(bm, index))
       return UTIL_BITMASK_INVALID_INDEX;
 
+   word = index / UTIL_BITMASK_BITS_PER_WORD;
+   bit  = index % UTIL_BITMASK_BITS_PER_WORD;
+   mask = 1 << bit;
+
    bm->words[word] |= mask;
 
    util_bitmask_filled_set(bm, index);
@@ -220,15 +224,19 @@ void
 util_bitmask_clear(struct util_bitmask *bm, 
                    unsigned index)
 {
-   unsigned word = index / UTIL_BITMASK_BITS_PER_WORD;
-   unsigned bit  = index % UTIL_BITMASK_BITS_PER_WORD;
-   util_bitmask_word mask = 1 << bit;
+   unsigned word;
+   unsigned bit;
+   util_bitmask_word mask;
    
    assert(bm);
    
    if(index >= bm->size)
       return;
 
+   word = index / UTIL_BITMASK_BITS_PER_WORD;
+   bit  = index % UTIL_BITMASK_BITS_PER_WORD;
+   mask = 1 << bit;
+
    bm->words[word] &= ~mask;
    
    util_bitmask_filled_unset(bm, index);
@@ -250,7 +258,7 @@ util_bitmask_get(struct util_bitmask *bm,
       return TRUE;
    }
 
-   if(index > bm->size)
+   if(index >= bm->size)
       return FALSE;
 
    if(bm->words[word] & mask) {
diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c
index fb00c3abe8..3f74e2aa8b 100644
--- a/src/gallium/auxiliary/util/u_blit.c
+++ b/src/gallium/auxiliary/util/u_blit.c
@@ -42,10 +42,12 @@
 
 #include "util/u_blit.h"
 #include "util/u_draw_quad.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "util/u_simple_shaders.h"
 #include "util/u_surface.h"
+#include "util/u_rect.h"
 
 #include "cso_cache/cso_context.h"
 
@@ -125,7 +127,8 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso)
    }
 
    /* fragment shader */
-   ctx->fs[TGSI_WRITEMASK_XYZW] = util_make_fragment_tex_shader(pipe);
+   ctx->fs[TGSI_WRITEMASK_XYZW] =
+      util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D);
    ctx->vbuf = NULL;
 
    /* init vertex data that doesn't change */
@@ -301,7 +304,8 @@ util_blit_pixels_writemask(struct blit_state *ctx,
     * no overlapping.
     * Filter mode should not matter since there's no stretching.
     */
-   if (dst->format == src->format &&
+   if (pipe->surface_copy &&
+       dst->format == src->format &&
        srcX0 < srcX1 &&
        dstX0 < dstX1 &&
        srcY0 < srcY1 &&
@@ -352,10 +356,9 @@ util_blit_pixels_writemask(struct blit_state *ctx,
       texTemp.target = PIPE_TEXTURE_2D;
       texTemp.format = src->format;
       texTemp.last_level = 0;
-      texTemp.width[0] = srcW;
-      texTemp.height[0] = srcH;
-      texTemp.depth[0] = 1;
-      pf_get_block(src->format, &texTemp.block);
+      texTemp.width0 = srcW;
+      texTemp.height0 = srcH;
+      texTemp.depth0 = 1;
 
       tex = screen->texture_create(screen, &texTemp);
       if (!tex)
@@ -365,10 +368,17 @@ util_blit_pixels_writemask(struct blit_state *ctx,
                                         PIPE_BUFFER_USAGE_GPU_WRITE);
 
       /* load temp texture */
-      pipe->surface_copy(pipe,
-                         texSurf, 0, 0,   /* dest */
-                         src, srcLeft, srcTop, /* src */
-                         srcW, srcH);     /* size */
+      if (pipe->surface_copy) {
+         pipe->surface_copy(pipe,
+                            texSurf, 0, 0,   /* dest */
+                            src, srcLeft, srcTop, /* src */
+                            srcW, srcH);     /* size */
+      } else {
+         util_surface_copy(pipe, FALSE,
+                           texSurf, 0, 0,   /* dest */
+                           src, srcLeft, srcTop, /* src */
+                           srcW, srcH);     /* size */
+      }
 
       /* free the surface, update the texture if necessary.
        */
@@ -380,10 +390,10 @@ util_blit_pixels_writemask(struct blit_state *ctx,
    }
    else {
       pipe_texture_reference(&tex, src->texture);
-      s0 = srcX0 / (float)tex->width[0];
-      s1 = srcX1 / (float)tex->width[0];
-      t0 = srcY0 / (float)tex->height[0];
-      t1 = srcY1 / (float)tex->height[0];
+      s0 = srcX0 / (float)tex->width0;
+      s1 = srcX1 / (float)tex->width0;
+      t0 = srcY0 / (float)tex->height0;
+      t1 = srcY1 / (float)tex->height0;
    }
 
 
@@ -412,7 +422,9 @@ util_blit_pixels_writemask(struct blit_state *ctx,
    cso_set_sampler_textures(ctx->cso, 1, &tex);
 
    if (ctx->fs[writemask] == NULL)
-      ctx->fs[writemask] = util_make_fragment_tex_shader_writemask(pipe, writemask);
+      ctx->fs[writemask] =
+         util_make_fragment_tex_shader_writemask(pipe, TGSI_TEXTURE_2D,
+                                                 writemask);
 
    /* shaders */
    cso_set_fragment_shader_handle(ctx->cso, ctx->fs[writemask]);
@@ -509,13 +521,13 @@ util_blit_pixels_tex(struct blit_state *ctx,
    assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
           filter == PIPE_TEX_MIPFILTER_LINEAR);
 
-   assert(tex->width[0] != 0);
-   assert(tex->height[0] != 0);
+   assert(tex->width0 != 0);
+   assert(tex->height0 != 0);
 
-   s0 = srcX0 / (float)tex->width[0];
-   s1 = srcX1 / (float)tex->width[0];
-   t0 = srcY0 / (float)tex->height[0];
-   t1 = srcY1 / (float)tex->height[0];
+   s0 = srcX0 / (float)tex->width0;
+   s1 = srcX1 / (float)tex->width0;
+   t0 = srcY0 / (float)tex->height0;
+   t1 = srcY1 / (float)tex->height0;
 
    assert(ctx->pipe->screen->is_format_supported(ctx->pipe->screen, dst->format,
                                                  PIPE_TEXTURE_2D,
diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c
new file mode 100644
index 0000000000..cef3b69e46
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_blitter.c
@@ -0,0 +1,726 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Marek Olšák <maraeo@gmail.com>
+ *
+ * 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.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Blitter utility to facilitate acceleration of the clear, surface_copy,
+ * and surface_fill functions.
+ *
+ * @author Marek Olšák
+ */
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_state.h"
+
+#include "util/u_format.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "util/u_blitter.h"
+#include "util/u_draw_quad.h"
+#include "util/u_pack_color.h"
+#include "util/u_rect.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_texture.h"
+
+#define INVALID_PTR ((void*)~0)
+
+struct blitter_context_priv
+{
+   struct blitter_context blitter;
+
+   struct pipe_context *pipe; /**< pipe context */
+   struct pipe_buffer *vbuf;  /**< quad */
+
+   float vertices[4][2][4];   /**< {pos, color} or {pos, texcoord} */
+
+   /* Templates for various state objects. */
+   struct pipe_depth_stencil_alpha_state template_dsa;
+   struct pipe_sampler_state template_sampler_state;
+
+   /* Constant state objects. */
+   /* Vertex shaders. */
+   void *vs_col; /**< Vertex shader which passes {pos, color} to the output */
+   void *vs_tex; /**<Vertex shader which passes {pos, texcoord} to the output.*/
+
+   /* Fragment shaders. */
+   /* FS which outputs a color to multiple color buffers. */
+   void *fs_col[PIPE_MAX_COLOR_BUFS];
+
+   /* FS which outputs a color from a texture,
+      where the index is PIPE_TEXTURE_* to be sampled. */
+   void *fs_texfetch_col[PIPE_MAX_TEXTURE_TYPES];
+
+   /* FS which outputs a depth from a texture,
+      where the index is PIPE_TEXTURE_* to be sampled. */
+   void *fs_texfetch_depth[PIPE_MAX_TEXTURE_TYPES];
+
+   /* Blend state. */
+   void *blend_write_color;   /**< blend state with writemask of RGBA */
+   void *blend_keep_color;    /**< blend state with writemask of 0 */
+
+   /* Depth stencil alpha state. */
+   void *dsa_write_depth_stencil[0xff]; /**< indices are stencil clear values */
+   void *dsa_write_depth_keep_stencil;
+   void *dsa_keep_depth_stencil;
+
+   /* Sampler state for clamping to a miplevel. */
+   void *sampler_state[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* Rasterizer state. */
+   void *rs_state;
+};
+
+struct blitter_context *util_blitter_create(struct pipe_context *pipe)
+{
+   struct blitter_context_priv *ctx;
+   struct pipe_blend_state blend;
+   struct pipe_depth_stencil_alpha_state *dsa;
+   struct pipe_rasterizer_state rs_state;
+   struct pipe_sampler_state *sampler_state;
+   unsigned i;
+
+   ctx = CALLOC_STRUCT(blitter_context_priv);
+   if (!ctx)
+      return NULL;
+
+   ctx->pipe = pipe;
+
+   /* init state objects for them to be considered invalid */
+   ctx->blitter.saved_blend_state = INVALID_PTR;
+   ctx->blitter.saved_dsa_state = INVALID_PTR;
+   ctx->blitter.saved_rs_state = INVALID_PTR;
+   ctx->blitter.saved_fs = INVALID_PTR;
+   ctx->blitter.saved_vs = INVALID_PTR;
+   ctx->blitter.saved_fb_state.nr_cbufs = ~0;
+   ctx->blitter.saved_num_textures = ~0;
+   ctx->blitter.saved_num_sampler_states = ~0;
+
+   /* blend state objects */
+   memset(&blend, 0, sizeof(blend));
+   ctx->blend_keep_color = pipe->create_blend_state(pipe, &blend);
+
+   blend.colormask = PIPE_MASK_RGBA;
+   ctx->blend_write_color = pipe->create_blend_state(pipe, &blend);
+
+   /* depth stencil alpha state objects */
+   dsa = &ctx->template_dsa;
+   ctx->dsa_keep_depth_stencil =
+      pipe->create_depth_stencil_alpha_state(pipe, dsa);
+
+   dsa->depth.enabled = 1;
+   dsa->depth.writemask = 1;
+   dsa->depth.func = PIPE_FUNC_ALWAYS;
+   ctx->dsa_write_depth_keep_stencil =
+      pipe->create_depth_stencil_alpha_state(pipe, dsa);
+
+   dsa->stencil[0].enabled = 1;
+   dsa->stencil[0].func = PIPE_FUNC_ALWAYS;
+   dsa->stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa->stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+   dsa->stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+   dsa->stencil[0].valuemask = 0xff;
+   dsa->stencil[0].writemask = 0xff;
+   /* The DSA state objects which write depth and stencil are created
+    * on-demand. */
+
+   /* sampler state */
+   sampler_state = &ctx->template_sampler_state;
+   sampler_state->wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler_state->wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler_state->wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   /* The sampler state objects which sample from a specified mipmap level
+    * are created on-demand. */
+
+   /* rasterizer state */
+   memset(&rs_state, 0, sizeof(rs_state));
+   rs_state.front_winding = PIPE_WINDING_CW;
+   rs_state.cull_mode = PIPE_WINDING_NONE;
+   rs_state.bypass_vs_clip_and_viewport = 1;
+   rs_state.gl_rasterization_rules = 1;
+   ctx->rs_state = pipe->create_rasterizer_state(pipe, &rs_state);
+
+   /* fragment shaders are created on-demand */
+
+   /* vertex shaders */
+   {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR };
+      const uint semantic_indices[] = { 0, 0 };
+      ctx->vs_col =
+         util_make_vertex_passthrough_shader(pipe, 2, semantic_names,
+                                             semantic_indices);
+   }
+   {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indices[] = { 0, 0 };
+      ctx->vs_tex =
+         util_make_vertex_passthrough_shader(pipe, 2, semantic_names,
+                                             semantic_indices);
+   }
+
+   /* set invariant vertex coordinates */
+   for (i = 0; i < 4; i++)
+      ctx->vertices[i][0][3] = 1; /*v.w*/
+
+   /* create the vertex buffer */
+   ctx->vbuf = pipe_buffer_create(ctx->pipe->screen,
+                                  32,
+                                  PIPE_BUFFER_USAGE_VERTEX,
+                                  sizeof(ctx->vertices));
+
+   return &ctx->blitter;
+}
+
+void util_blitter_destroy(struct blitter_context *blitter)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->pipe;
+   int i;
+
+   pipe->delete_blend_state(pipe, ctx->blend_write_color);
+   pipe->delete_blend_state(pipe, ctx->blend_keep_color);
+   pipe->delete_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil);
+   pipe->delete_depth_stencil_alpha_state(pipe,
+                                          ctx->dsa_write_depth_keep_stencil);
+
+   for (i = 0; i < 0xff; i++)
+      if (ctx->dsa_write_depth_stencil[i])
+         pipe->delete_depth_stencil_alpha_state(pipe,
+            ctx->dsa_write_depth_stencil[i]);
+
+   pipe->delete_rasterizer_state(pipe, ctx->rs_state);
+   pipe->delete_vs_state(pipe, ctx->vs_col);
+   pipe->delete_vs_state(pipe, ctx->vs_tex);
+
+   for (i = 0; i < PIPE_MAX_TEXTURE_TYPES; i++) {
+      if (ctx->fs_texfetch_col[i])
+         pipe->delete_fs_state(pipe, ctx->fs_texfetch_col[i]);
+      if (ctx->fs_texfetch_depth[i])
+         pipe->delete_fs_state(pipe, ctx->fs_texfetch_depth[i]);
+   }
+
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS && ctx->fs_col[i]; i++)
+      if (ctx->fs_col[i])
+         pipe->delete_fs_state(pipe, ctx->fs_col[i]);
+
+   for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++)
+      if (ctx->sampler_state[i])
+         pipe->delete_sampler_state(pipe, ctx->sampler_state[i]);
+
+   pipe_buffer_reference(&ctx->vbuf, NULL);
+   FREE(ctx);
+}
+
+static void blitter_check_saved_CSOs(struct blitter_context_priv *ctx)
+{
+   /* make sure these CSOs have been saved */
+   assert(ctx->blitter.saved_blend_state != INVALID_PTR &&
+          ctx->blitter.saved_dsa_state != INVALID_PTR &&
+          ctx->blitter.saved_rs_state != INVALID_PTR &&
+          ctx->blitter.saved_fs != INVALID_PTR &&
+          ctx->blitter.saved_vs != INVALID_PTR);
+}
+
+static void blitter_restore_CSOs(struct blitter_context_priv *ctx)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   /* restore the state objects which are always required to be saved */
+   pipe->bind_blend_state(pipe, ctx->blitter.saved_blend_state);
+   pipe->bind_depth_stencil_alpha_state(pipe, ctx->blitter.saved_dsa_state);
+   pipe->bind_rasterizer_state(pipe, ctx->blitter.saved_rs_state);
+   pipe->bind_fs_state(pipe, ctx->blitter.saved_fs);
+   pipe->bind_vs_state(pipe, ctx->blitter.saved_vs);
+
+   ctx->blitter.saved_blend_state = INVALID_PTR;
+   ctx->blitter.saved_dsa_state = INVALID_PTR;
+   ctx->blitter.saved_rs_state = INVALID_PTR;
+   ctx->blitter.saved_fs = INVALID_PTR;
+   ctx->blitter.saved_vs = INVALID_PTR;
+
+   /* restore the state objects which are required to be saved before copy/fill
+    */
+   if (ctx->blitter.saved_fb_state.nr_cbufs != ~0) {
+      pipe->set_framebuffer_state(pipe, &ctx->blitter.saved_fb_state);
+      ctx->blitter.saved_fb_state.nr_cbufs = ~0;
+   }
+
+   if (ctx->blitter.saved_num_sampler_states != ~0) {
+      pipe->bind_fragment_sampler_states(pipe,
+                                         ctx->blitter.saved_num_sampler_states,
+                                         ctx->blitter.saved_sampler_states);
+      ctx->blitter.saved_num_sampler_states = ~0;
+   }
+
+   if (ctx->blitter.saved_num_textures != ~0) {
+      pipe->set_fragment_sampler_textures(pipe,
+                                          ctx->blitter.saved_num_textures,
+                                          ctx->blitter.saved_textures);
+      ctx->blitter.saved_num_textures = ~0;
+   }
+}
+
+static void blitter_set_rectangle(struct blitter_context_priv *ctx,
+                                  unsigned x1, unsigned y1,
+                                  unsigned x2, unsigned y2,
+                                  float depth)
+{
+   int i;
+
+   /* set vertex positions */
+   ctx->vertices[0][0][0] = x1; /*v0.x*/
+   ctx->vertices[0][0][1] = y1; /*v0.y*/
+
+   ctx->vertices[1][0][0] = x2; /*v1.x*/
+   ctx->vertices[1][0][1] = y1; /*v1.y*/
+
+   ctx->vertices[2][0][0] = x2; /*v2.x*/
+   ctx->vertices[2][0][1] = y2; /*v2.y*/
+
+   ctx->vertices[3][0][0] = x1; /*v3.x*/
+   ctx->vertices[3][0][1] = y2; /*v3.y*/
+
+   for (i = 0; i < 4; i++)
+      ctx->vertices[i][0][2] = depth; /*z*/
+}
+
+static void blitter_set_clear_color(struct blitter_context_priv *ctx,
+                                    const float *rgba)
+{
+   int i;
+
+   for (i = 0; i < 4; i++) {
+      ctx->vertices[i][1][0] = rgba[0];
+      ctx->vertices[i][1][1] = rgba[1];
+      ctx->vertices[i][1][2] = rgba[2];
+      ctx->vertices[i][1][3] = rgba[3];
+   }
+}
+
+static void blitter_set_texcoords_2d(struct blitter_context_priv *ctx,
+                                     struct pipe_surface *surf,
+                                     unsigned x1, unsigned y1,
+                                     unsigned x2, unsigned y2)
+{
+   int i;
+   float s1 = x1 / (float)surf->width;
+   float t1 = y1 / (float)surf->height;
+   float s2 = x2 / (float)surf->width;
+   float t2 = y2 / (float)surf->height;
+
+   ctx->vertices[0][1][0] = s1; /*t0.s*/
+   ctx->vertices[0][1][1] = t1; /*t0.t*/
+
+   ctx->vertices[1][1][0] = s2; /*t1.s*/
+   ctx->vertices[1][1][1] = t1; /*t1.t*/
+
+   ctx->vertices[2][1][0] = s2; /*t2.s*/
+   ctx->vertices[2][1][1] = t2; /*t2.t*/
+
+   ctx->vertices[3][1][0] = s1; /*t3.s*/
+   ctx->vertices[3][1][1] = t2; /*t3.t*/
+
+   for (i = 0; i < 4; i++) {
+      ctx->vertices[i][1][2] = 0; /*r*/
+      ctx->vertices[i][1][3] = 1; /*q*/
+   }
+}
+
+static void blitter_set_texcoords_3d(struct blitter_context_priv *ctx,
+                                     struct pipe_surface *surf,
+                                     unsigned x1, unsigned y1,
+                                     unsigned x2, unsigned y2)
+{
+   int i;
+   float depth = u_minify(surf->texture->depth0, surf->level);
+   float r = surf->zslice / depth;
+
+   blitter_set_texcoords_2d(ctx, surf, x1, y1, x2, y2);
+
+   for (i = 0; i < 4; i++)
+      ctx->vertices[i][1][2] = r; /*r*/
+}
+
+static void blitter_set_texcoords_cube(struct blitter_context_priv *ctx,
+                                       struct pipe_surface *surf,
+                                       unsigned x1, unsigned y1,
+                                       unsigned x2, unsigned y2)
+{
+   int i;
+   float s1 = x1 / (float)surf->width;
+   float t1 = y1 / (float)surf->height;
+   float s2 = x2 / (float)surf->width;
+   float t2 = y2 / (float)surf->height;
+   const float st[4][2] = {
+      {s1, t1}, {s2, t1}, {s2, t2}, {s1, t2}
+   };
+
+   util_map_texcoords2d_onto_cubemap(surf->face,
+                                     /* pointer, stride in floats */
+                                     &st[0][0], 2,
+                                     &ctx->vertices[0][1][0], 8);
+
+   for (i = 0; i < 4; i++)
+      ctx->vertices[i][1][3] = 1; /*q*/
+}
+
+static void blitter_draw_quad(struct blitter_context_priv *ctx)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   /* write vertices and draw them */
+   pipe_buffer_write(pipe->screen, ctx->vbuf,
+                     0, sizeof(ctx->vertices), ctx->vertices);
+
+   util_draw_vertex_buffer(pipe, ctx->vbuf, 0, PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+}
+
+static INLINE
+void *blitter_get_state_write_depth_stencil(
+               struct blitter_context_priv *ctx,
+               unsigned stencil)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   stencil &= 0xff;
+
+   /* Create the DSA state on-demand. */
+   if (!ctx->dsa_write_depth_stencil[stencil]) {
+      ctx->template_dsa.stencil[0].ref_value = stencil;
+
+      ctx->dsa_write_depth_stencil[stencil] =
+         pipe->create_depth_stencil_alpha_state(pipe, &ctx->template_dsa);
+   }
+
+   return ctx->dsa_write_depth_stencil[stencil];
+}
+
+static INLINE
+void **blitter_get_sampler_state(struct blitter_context_priv *ctx,
+                                 int miplevel)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_sampler_state *sampler_state = &ctx->template_sampler_state;
+
+   assert(miplevel < PIPE_MAX_TEXTURE_LEVELS);
+
+   /* Create the sampler state on-demand. */
+   if (!ctx->sampler_state[miplevel]) {
+      sampler_state->lod_bias = miplevel;
+      sampler_state->min_lod = miplevel;
+      sampler_state->max_lod = miplevel;
+
+      ctx->sampler_state[miplevel] = pipe->create_sampler_state(pipe,
+                                                                sampler_state);
+   }
+
+   /* Return void** so that it can be passed to bind_fragment_sampler_states
+    * directly. */
+   return &ctx->sampler_state[miplevel];
+}
+
+static INLINE
+void *blitter_get_fs_col(struct blitter_context_priv *ctx, unsigned num_cbufs)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   unsigned index = num_cbufs ? num_cbufs - 1 : 0;
+
+   assert(num_cbufs <= PIPE_MAX_COLOR_BUFS);
+
+   if (!ctx->fs_col[index])
+      ctx->fs_col[index] =
+         util_make_fragment_clonecolor_shader(pipe, num_cbufs);
+
+   return ctx->fs_col[index];
+}
+
+static INLINE
+void *blitter_get_fs_texfetch_col(struct blitter_context_priv *ctx,
+                                  unsigned tex_target)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   assert(tex_target < PIPE_MAX_TEXTURE_TYPES);
+
+   /* Create the fragment shader on-demand. */
+   if (!ctx->fs_texfetch_col[tex_target]) {
+      switch (tex_target) {
+         case PIPE_TEXTURE_1D:
+            ctx->fs_texfetch_col[PIPE_TEXTURE_1D] =
+               util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_1D);
+            break;
+         case PIPE_TEXTURE_2D:
+            ctx->fs_texfetch_col[PIPE_TEXTURE_2D] =
+               util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D);
+            break;
+         case PIPE_TEXTURE_3D:
+            ctx->fs_texfetch_col[PIPE_TEXTURE_3D] =
+               util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_3D);
+            break;
+         case PIPE_TEXTURE_CUBE:
+            ctx->fs_texfetch_col[PIPE_TEXTURE_CUBE] =
+               util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_CUBE);
+            break;
+         default:;
+      }
+   }
+
+   return ctx->fs_texfetch_col[tex_target];
+}
+
+static INLINE
+void *blitter_get_fs_texfetch_depth(struct blitter_context_priv *ctx,
+                                    unsigned tex_target)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   assert(tex_target < PIPE_MAX_TEXTURE_TYPES);
+
+   /* Create the fragment shader on-demand. */
+   if (!ctx->fs_texfetch_depth[tex_target]) {
+      switch (tex_target) {
+         case PIPE_TEXTURE_1D:
+            ctx->fs_texfetch_depth[PIPE_TEXTURE_1D] =
+               util_make_fragment_tex_shader_writedepth(pipe, TGSI_TEXTURE_1D);
+            break;
+         case PIPE_TEXTURE_2D:
+            ctx->fs_texfetch_depth[PIPE_TEXTURE_2D] =
+               util_make_fragment_tex_shader_writedepth(pipe, TGSI_TEXTURE_2D);
+            break;
+         case PIPE_TEXTURE_3D:
+            ctx->fs_texfetch_depth[PIPE_TEXTURE_3D] =
+               util_make_fragment_tex_shader_writedepth(pipe, TGSI_TEXTURE_3D);
+            break;
+         case PIPE_TEXTURE_CUBE:
+            ctx->fs_texfetch_depth[PIPE_TEXTURE_CUBE] =
+               util_make_fragment_tex_shader_writedepth(pipe,TGSI_TEXTURE_CUBE);
+            break;
+         default:;
+      }
+   }
+
+   return ctx->fs_texfetch_depth[tex_target];
+}
+
+void util_blitter_clear(struct blitter_context *blitter,
+                        unsigned width, unsigned height,
+                        unsigned num_cbufs,
+                        unsigned clear_buffers,
+                        const float *rgba,
+                        double depth, unsigned stencil)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->pipe;
+
+   assert(num_cbufs <= PIPE_MAX_COLOR_BUFS);
+
+   blitter_check_saved_CSOs(ctx);
+
+   /* bind CSOs */
+   if (clear_buffers & PIPE_CLEAR_COLOR)
+      pipe->bind_blend_state(pipe, ctx->blend_write_color);
+   else
+      pipe->bind_blend_state(pipe, ctx->blend_keep_color);
+
+   if (clear_buffers & PIPE_CLEAR_DEPTHSTENCIL)
+      pipe->bind_depth_stencil_alpha_state(pipe,
+         blitter_get_state_write_depth_stencil(ctx, stencil));
+   else
+      pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil);
+
+   pipe->bind_rasterizer_state(pipe, ctx->rs_state);
+   pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, num_cbufs));
+   pipe->bind_vs_state(pipe, ctx->vs_col);
+
+   blitter_set_clear_color(ctx, rgba);
+   blitter_set_rectangle(ctx, 0, 0, width, height, depth);
+   blitter_draw_quad(ctx);
+   blitter_restore_CSOs(ctx);
+}
+
+void util_blitter_copy(struct blitter_context *blitter,
+                       struct pipe_surface *dst,
+                       unsigned dstx, unsigned dsty,
+                       struct pipe_surface *src,
+                       unsigned srcx, unsigned srcy,
+                       unsigned width, unsigned height,
+                       boolean ignore_stencil)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_framebuffer_state fb_state;
+   boolean is_stencil, is_depth;
+   unsigned dst_tex_usage;
+
+   /* give up if textures are not set */
+   assert(dst->texture && src->texture);
+   if (!dst->texture || !src->texture)
+      return;
+
+   is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0;
+   is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0;
+   dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
+                                            PIPE_TEXTURE_USAGE_RENDER_TARGET;
+
+   /* check if we can sample from and render to the surfaces */
+   /* (assuming copying a stencil buffer is not possible) */
+   if ((!ignore_stencil && is_stencil) ||
+       !screen->is_format_supported(screen, dst->format, dst->texture->target,
+                                    dst_tex_usage, 0) ||
+       !screen->is_format_supported(screen, src->format, src->texture->target,
+                                    PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      util_surface_copy(pipe, FALSE, dst, dstx, dsty, src, srcx, srcy,
+                        width, height);
+      return;
+   }
+
+   /* check whether the states are properly saved */
+   blitter_check_saved_CSOs(ctx);
+   assert(blitter->saved_fb_state.nr_cbufs != ~0);
+   assert(blitter->saved_num_textures != ~0);
+   assert(blitter->saved_num_sampler_states != ~0);
+   assert(src->texture->target < PIPE_MAX_TEXTURE_TYPES);
+
+   /* bind CSOs */
+   fb_state.width = dst->width;
+   fb_state.height = dst->height;
+
+   if (is_depth) {
+      pipe->bind_blend_state(pipe, ctx->blend_keep_color);
+      pipe->bind_depth_stencil_alpha_state(pipe,
+                                           ctx->dsa_write_depth_keep_stencil);
+      pipe->bind_fs_state(pipe,
+         blitter_get_fs_texfetch_depth(ctx, src->texture->target));
+
+      fb_state.nr_cbufs = 0;
+      fb_state.zsbuf = dst;
+   } else {
+      pipe->bind_blend_state(pipe, ctx->blend_write_color);
+      pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil);
+      pipe->bind_fs_state(pipe,
+         blitter_get_fs_texfetch_col(ctx, src->texture->target));
+
+      fb_state.nr_cbufs = 1;
+      fb_state.cbufs[0] = dst;
+      fb_state.zsbuf = 0;
+   }
+
+   pipe->bind_rasterizer_state(pipe, ctx->rs_state);
+   pipe->bind_vs_state(pipe, ctx->vs_tex);
+   pipe->bind_fragment_sampler_states(pipe, 1,
+      blitter_get_sampler_state(ctx, src->level));
+   pipe->set_fragment_sampler_textures(pipe, 1, &src->texture);
+   pipe->set_framebuffer_state(pipe, &fb_state);
+
+   /* set texture coordinates */
+   switch (src->texture->target) {
+      case PIPE_TEXTURE_1D:
+      case PIPE_TEXTURE_2D:
+         blitter_set_texcoords_2d(ctx, src, srcx, srcy,
+                                  srcx+width, srcy+height);
+         break;
+      case PIPE_TEXTURE_3D:
+         blitter_set_texcoords_3d(ctx, src, srcx, srcy,
+                                  srcx+width, srcy+height);
+         break;
+      case PIPE_TEXTURE_CUBE:
+         blitter_set_texcoords_cube(ctx, src, srcx, srcy,
+                                    srcx+width, srcy+height);
+         break;
+      default:
+         assert(0);
+   }
+
+   blitter_set_rectangle(ctx, dstx, dsty, dstx+width, dsty+height, 0);
+   blitter_draw_quad(ctx);
+   blitter_restore_CSOs(ctx);
+}
+
+void util_blitter_fill(struct blitter_context *blitter,
+                       struct pipe_surface *dst,
+                       unsigned dstx, unsigned dsty,
+                       unsigned width, unsigned height,
+                       unsigned value)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_framebuffer_state fb_state;
+   float rgba[4];
+   ubyte ub_rgba[4] = {0};
+   union util_color color;
+   int i;
+
+   assert(dst->texture);
+   if (!dst->texture)
+      return;
+
+   /* check if we can render to the surface */
+   if (util_format_is_depth_or_stencil(dst->format) || /* unlikely, but you never know */
+       !screen->is_format_supported(screen, dst->format, dst->texture->target,
+                                    PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
+      util_surface_fill(pipe, dst, dstx, dsty, width, height, value);
+      return;
+   }
+
+   /* unpack the color */
+   color.ui = value;
+   util_unpack_color_ub(dst->format, &color,
+                        ub_rgba, ub_rgba+1, ub_rgba+2, ub_rgba+3);
+   for (i = 0; i < 4; i++)
+      rgba[i] = ubyte_to_float(ub_rgba[i]);
+
+   /* check the saved state */
+   blitter_check_saved_CSOs(ctx);
+   assert(blitter->saved_fb_state.nr_cbufs != ~0);
+
+   /* bind CSOs */
+   pipe->bind_blend_state(pipe, ctx->blend_write_color);
+   pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_keep_depth_stencil);
+   pipe->bind_rasterizer_state(pipe, ctx->rs_state);
+   pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, 1));
+   pipe->bind_vs_state(pipe, ctx->vs_col);
+
+   /* set a framebuffer state */
+   fb_state.width = dst->width;
+   fb_state.height = dst->height;
+   fb_state.nr_cbufs = 1;
+   fb_state.cbufs[0] = dst;
+   fb_state.zsbuf = 0;
+   pipe->set_framebuffer_state(pipe, &fb_state);
+
+   blitter_set_clear_color(ctx, rgba);
+   blitter_set_rectangle(ctx, 0, 0, width, height, 0);
+   blitter_draw_quad(ctx);
+   blitter_restore_CSOs(ctx);
+}
diff --git a/src/gallium/auxiliary/util/u_blitter.h b/src/gallium/auxiliary/util/u_blitter.h
new file mode 100644
index 0000000000..3da5a6ca52
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_blitter.h
@@ -0,0 +1,230 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Marek Olšák <maraeo@gmail.com>
+ *
+ * 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 U_BLITTER_H
+#define U_BLITTER_H
+
+#include "util/u_memory.h"
+
+#include "pipe/p_state.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct pipe_context;
+
+struct blitter_context
+{
+   /* Private members, really. */
+   void *saved_blend_state;   /**< blend state */
+   void *saved_dsa_state;     /**< depth stencil alpha state */
+   void *saved_rs_state;      /**< rasterizer state */
+   void *saved_fs, *saved_vs; /**< fragment shader, vertex shader */
+
+   struct pipe_framebuffer_state saved_fb_state;  /**< framebuffer state */
+
+   int saved_num_sampler_states;
+   void *saved_sampler_states[32];
+
+   int saved_num_textures;
+   struct pipe_texture *saved_textures[32]; /* is 32 enough? */
+};
+
+/**
+ * Create a blitter context.
+ */
+struct blitter_context *util_blitter_create(struct pipe_context *pipe);
+
+/**
+ * Destroy a blitter context.
+ */
+void util_blitter_destroy(struct blitter_context *blitter);
+
+/*
+ * These CSOs must be saved before any of the following functions is called:
+ * - blend state
+ * - depth stencil alpha state
+ * - rasterizer state
+ * - vertex shader
+ * - fragment shader
+ */
+
+/**
+ * Clear a specified set of currently bound buffers to specified values.
+ */
+void util_blitter_clear(struct blitter_context *blitter,
+                        unsigned width, unsigned height,
+                        unsigned num_cbufs,
+                        unsigned clear_buffers,
+                        const float *rgba,
+                        double depth, unsigned stencil);
+
+/**
+ * Copy a block of pixels from one surface to another.
+ *
+ * You can copy from any color format to any other color format provided
+ * the former can be sampled and the latter can be rendered to. Otherwise,
+ * a software fallback path is taken and both surfaces must be of the same
+ * format.
+ *
+ * The same holds for depth-stencil formats with the exception that stencil
+ * cannot be copied unless you set ignore_stencil to FALSE. In that case,
+ * a software fallback path is taken and both surfaces must be of the same
+ * format.
+ *
+ * Use pipe_screen->is_format_supported to know your options.
+ *
+ * These states must be saved in the blitter in addition to the state objects
+ * already required to be saved:
+ * - framebuffer state
+ * - fragment sampler states
+ * - fragment sampler textures
+ */
+void util_blitter_copy(struct blitter_context *blitter,
+                       struct pipe_surface *dst,
+                       unsigned dstx, unsigned dsty,
+                       struct pipe_surface *src,
+                       unsigned srcx, unsigned srcy,
+                       unsigned width, unsigned height,
+                       boolean ignore_stencil);
+
+/**
+ * Fill a region of a surface with a constant value.
+ *
+ * If the surface cannot be rendered to or it's a depth-stencil format,
+ * a software fallback path is taken.
+ *
+ * These states must be saved in the blitter in addition to the state objects
+ * already required to be saved:
+ * - framebuffer state
+ */
+void util_blitter_fill(struct blitter_context *blitter,
+                       struct pipe_surface *dst,
+                       unsigned dstx, unsigned dsty,
+                       unsigned width, unsigned height,
+                       unsigned value);
+
+/**
+ * Copy all pixels from one surface to another.
+ *
+ * The rules are the same as in util_blitter_copy with the addition that
+ * surfaces must have the same size.
+ */
+static INLINE
+void util_blitter_copy_surface(struct blitter_context *blitter,
+                               struct pipe_surface *dst,
+                               struct pipe_surface *src,
+                               boolean ignore_stencil)
+{
+   assert(dst->width == src->width && dst->height == src->height);
+
+   util_blitter_copy(blitter, dst, 0, 0, src, 0, 0, src->width, src->height,
+                     ignore_stencil);
+}
+
+
+/* The functions below should be used to save currently bound constant state
+ * objects inside a driver. The objects are automatically restored at the end
+ * of the util_blitter_{clear, fill, copy, copy_surface} functions and then
+ * forgotten.
+ *
+ * CSOs not listed here are not affected by util_blitter. */
+
+static INLINE
+void util_blitter_save_blend(struct blitter_context *blitter,
+                             void *state)
+{
+   blitter->saved_blend_state = state;
+}
+
+static INLINE
+void util_blitter_save_depth_stencil_alpha(struct blitter_context *blitter,
+                                           void *state)
+{
+   blitter->saved_dsa_state = state;
+}
+
+static INLINE
+void util_blitter_save_rasterizer(struct blitter_context *blitter,
+                                  void *state)
+{
+   blitter->saved_rs_state = state;
+}
+
+static INLINE
+void util_blitter_save_fragment_shader(struct blitter_context *blitter,
+                                       void *fs)
+{
+   blitter->saved_fs = fs;
+}
+
+static INLINE
+void util_blitter_save_vertex_shader(struct blitter_context *blitter,
+                                     void *vs)
+{
+   blitter->saved_vs = vs;
+}
+
+static INLINE
+void util_blitter_save_framebuffer(struct blitter_context *blitter,
+                                   struct pipe_framebuffer_state *state)
+{
+   blitter->saved_fb_state = *state;
+}
+
+static INLINE
+void util_blitter_save_fragment_sampler_states(
+                  struct blitter_context *blitter,
+                  int num_sampler_states,
+                  void **sampler_states)
+{
+   assert(num_sampler_states <= Elements(blitter->saved_sampler_states));
+
+   blitter->saved_num_sampler_states = num_sampler_states;
+   memcpy(blitter->saved_sampler_states, sampler_states,
+          num_sampler_states * sizeof(void *));
+}
+
+static INLINE
+void util_blitter_save_fragment_sampler_textures(
+                  struct blitter_context *blitter,
+                  int num_textures,
+                  struct pipe_texture **textures)
+{
+   assert(num_textures <= Elements(blitter->saved_textures));
+
+   blitter->saved_num_textures = num_textures;
+   memcpy(blitter->saved_textures, textures,
+          num_textures * sizeof(struct pipe_texture *));
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_clear.h b/src/gallium/auxiliary/util/u_clear.h
index 7c16b32cf9..2c32db6175 100644
--- a/src/gallium/auxiliary/util/u_clear.h
+++ b/src/gallium/auxiliary/util/u_clear.h
@@ -32,6 +32,7 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
 #include "util/u_pack_color.h"
+#include "util/u_rect.h"
 
 
 /**
@@ -45,16 +46,25 @@ util_clear(struct pipe_context *pipe,
 {
    if (buffers & PIPE_CLEAR_COLOR) {
       struct pipe_surface *ps = framebuffer->cbufs[0];
-      unsigned color;
+      union util_color uc;
 
-      util_pack_color(rgba, ps->format, &color);
-      pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, color);
+      util_pack_color(rgba, ps->format, &uc);
+      if (pipe->surface_fill) {
+         pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, uc.ui);
+      } else {
+         util_surface_fill(pipe, ps, 0, 0, ps->width, ps->height, uc.ui);
+      }
    }
 
    if (buffers & PIPE_CLEAR_DEPTHSTENCIL) {
       struct pipe_surface *ps = framebuffer->zsbuf;
 
-      pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height,
-                         util_pack_z_stencil(ps->format, depth, stencil));
+      if (pipe->surface_fill) {
+         pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height,
+                            util_pack_z_stencil(ps->format, depth, stencil));
+      } else {
+         util_surface_fill(pipe, ps, 0, 0, ps->width, ps->height,
+                           util_pack_z_stencil(ps->format, depth, stencil));
+      }
    }
 }
diff --git a/src/gallium/auxiliary/util/u_cpu_detect.c b/src/gallium/auxiliary/util/u_cpu_detect.c
index ded361ce70..a08241971c 100644
--- a/src/gallium/auxiliary/util/u_cpu_detect.c
+++ b/src/gallium/auxiliary/util/u_cpu_detect.c
@@ -67,18 +67,20 @@
 
 #if defined(PIPE_OS_WINDOWS)
 #include <windows.h>
+#if defined(MSVC)
+#include <intrin.h>
+#endif
 #endif
 
 
 struct util_cpu_caps util_cpu_caps;
 
 static int has_cpuid(void);
-static int cpuid(uint32_t ax, uint32_t *p);
 
 #if defined(PIPE_ARCH_X86)
 
 /* The sigill handlers */
-#if defined(PIPE_OS_LINUX) //&& defined(_POSIX_SOURCE) && defined(X86_FXSR_MAGIC)
+#if defined(PIPE_OS_LINUX) /*&& defined(_POSIX_SOURCE) && defined(X86_FXSR_MAGIC)*/
 static void
 sigill_handler_sse(int signal, struct sigcontext sc)
 {
@@ -131,7 +133,7 @@ win32_sig_handler_sse(EXCEPTION_POINTERS* ep)
 
 
 #if defined(PIPE_ARCH_PPC) && !defined(PIPE_OS_DARWIN)
-static sigjmp_buf __lv_powerpc_jmpbuf;
+static jmp_buf  __lv_powerpc_jmpbuf;
 static volatile sig_atomic_t __lv_powerpc_canjump = 0;
 
 static void
@@ -143,9 +145,11 @@ sigill_handler(int sig)
    }
 
    __lv_powerpc_canjump = 0;
-   siglongjmp(__lv_powerpc_jmpbuf, 1);
+   longjmp(__lv_powerpc_jmpbuf, 1);
 }
+#endif
 
+#if defined(PIPE_ARCH_PPC)
 static void
 check_os_altivec_support(void)
 {
@@ -166,7 +170,7 @@ check_os_altivec_support(void)
    /* no Darwin, do it the brute-force way */
    /* this is borrowed from the libmpeg2 library */
    signal(SIGILL, sigill_handler);
-   if (sigsetjmp(__lv_powerpc_jmpbuf, 1)) {
+   if (setjmp(__lv_powerpc_jmpbuf)) {
       signal(SIGILL, SIG_DFL);
    } else {
       __lv_powerpc_canjump = 1;
@@ -180,9 +184,9 @@ check_os_altivec_support(void)
       signal(SIGILL, SIG_DFL);
       util_cpu_caps.has_altivec = 1;
    }
-#endif
+#endif /* PIPE_OS_DARWIN */
 }
-#endif
+#endif /* PIPE_ARCH_PPC */
 
 /* If we're running on a processor that can do SSE, let's see if we
  * are allowed to or not.  This will catch 2.4.0 or later kernels that
@@ -190,6 +194,7 @@ check_os_altivec_support(void)
  * and RedHat patched 2.2 kernels that have broken exception handling
  * support for user space apps that do SSE.
  */
+#if defined(PIPE_ARCH_X86) || defined (PIPE_ARCH_X86_64)
 static void
 check_os_katmai_support(void)
 {
@@ -235,7 +240,7 @@ check_os_katmai_support(void)
       __asm __volatile ("xorps %xmm0, %xmm0");
 #elif defined(PIPE_CC_MSVC)
       __asm {
-          xorps xmm0, xmm0        // executing SSE instruction
+          xorps xmm0, xmm0        /* executing SSE instruction */
       }
 #else
 #error Unsupported compiler
@@ -278,7 +283,7 @@ check_os_katmai_support(void)
     * and therefore to be safe I'm going to leave this test in here.
     */
    if (util_cpu_caps.has_sse) {
-      //      test_os_katmai_exception_support();
+      /* test_os_katmai_exception_support(); */
    }
 
    /* Restore the original signal handlers.
@@ -334,12 +339,11 @@ static int has_cpuid(void)
 
 /**
  * @sa cpuid.h included in gcc-4.3 onwards.
+ * @sa http://msdn.microsoft.com/en-us/library/hskdteyh.aspx
  */
-static INLINE int
+static INLINE void
 cpuid(uint32_t ax, uint32_t *p)
 {
-   int ret = -1;
-
 #if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86)
    __asm __volatile (
      "xchgl %%ebx, %1\n\t"
@@ -351,7 +355,6 @@ cpuid(uint32_t ax, uint32_t *p)
        "=d" (p[3])
      : "0" (ax)
    );
-   ret = 0;
 #elif defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86_64)
    __asm __volatile (
      "cpuid\n\t"
@@ -361,15 +364,16 @@ cpuid(uint32_t ax, uint32_t *p)
        "=d" (p[3])
      : "0" (ax)
    );
-   ret = 0;
 #elif defined(PIPE_CC_MSVC)
-   __cpuid(ax, p);
-
-   ret = 0;
+   __cpuid(p, ax);
+#else
+   p[0] = 0;
+   p[1] = 0;
+   p[2] = 0;
+   p[3] = 0;
 #endif
-
-   return ret;
 }
+#endif /* X86 or X86_64 */
 
 void
 util_cpu_detect(void)
@@ -390,19 +394,26 @@ util_cpu_detect(void)
    util_cpu_caps.arch = UTIL_CPU_ARCH_SPARC;
 #elif defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
    util_cpu_caps.arch = UTIL_CPU_ARCH_X86;
+   util_cpu_caps.little_endian = 1;
 #elif defined(PIPE_ARCH_PPC)
    util_cpu_caps.arch = UTIL_CPU_ARCH_POWERPC;
+   util_cpu_caps.little_endian = 0;
 #else
    util_cpu_caps.arch = UTIL_CPU_ARCH_UNKNOWN;
 #endif
 
    /* Count the number of CPUs in system */
-#if !defined(PIPE_OS_WINDOWS) && !defined(PIPE_OS_UNKNOWN) && defined(_SC_NPROCESSORS_ONLN)
+#if defined(PIPE_OS_WINDOWS)
+   {
+      SYSTEM_INFO system_info;
+      GetSystemInfo(&system_info);
+      util_cpu_caps.nr_cpus = system_info.dwNumberOfProcessors;
+   }
+#elif defined(PIPE_OS_UNIX) && defined(_SC_NPROCESSORS_ONLN)
    util_cpu_caps.nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
    if (util_cpu_caps.nr_cpus == -1)
       util_cpu_caps.nr_cpus = 1;
-
-#elif defined(PIPE_OS_NETBSD) || defined(PIPE_OS_FREEBSD) || defined(PIPE_OS_OPENBSD)
+#elif defined(PIPE_OS_BSD)
    {
       int mib[2], ncpu;
       int len;
@@ -469,7 +480,6 @@ util_cpu_detect(void)
          util_cpu_caps.cacheline = regs2[2] & 0xFF;
       }
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_FREEBSD) || defined(PIPE_OS_NETBSD) || defined(PIPE_OS_CYGWIN) || defined(PIPE_OS_OPENBSD)
       if (util_cpu_caps.has_sse)
          check_os_katmai_support();
 
@@ -477,13 +487,8 @@ util_cpu_detect(void)
          util_cpu_caps.has_sse2 = 0;
          util_cpu_caps.has_sse3 = 0;
          util_cpu_caps.has_ssse3 = 0;
+         util_cpu_caps.has_sse4_1 = 0;
       }
-#else
-      util_cpu_caps.has_sse = 0;
-      util_cpu_caps.has_sse2 = 0;
-      util_cpu_caps.has_sse3 = 0;
-      util_cpu_caps.has_ssse3 = 0;
-#endif
    }
 #endif /* PIPE_ARCH_X86 || PIPE_ARCH_X86_64 */
 
diff --git a/src/gallium/auxiliary/util/u_cpu_detect.h b/src/gallium/auxiliary/util/u_cpu_detect.h
index 7ea0121c07..4b3dc39c34 100644
--- a/src/gallium/auxiliary/util/u_cpu_detect.h
+++ b/src/gallium/auxiliary/util/u_cpu_detect.h
@@ -54,6 +54,8 @@ struct util_cpu_caps {
    int x86_cpu_type;
    unsigned cacheline;
 
+   unsigned little_endian:1;
+
    unsigned has_tsc:1;
    unsigned has_mmx:1;
    unsigned has_mmx2:1;
diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c
index 96d400c839..9b4e6ca2a7 100644
--- a/src/gallium/auxiliary/util/u_debug.c
+++ b/src/gallium/auxiliary/util/u_debug.c
@@ -64,11 +64,13 @@
 #include "pipe/p_format.h" 
 #include "pipe/p_state.h" 
 #include "pipe/p_inlines.h" 
+#include "util/u_format.h"
 #include "util/u_memory.h" 
 #include "util/u_string.h" 
 #include "util/u_stream.h" 
 #include "util/u_math.h" 
 #include "util/u_tile.h" 
+#include "util/u_prim.h" 
 
 
 #ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
@@ -452,7 +454,8 @@ debug_dump_flags(const struct debug_named_value *names,
 	    util_strncat(output, "|", sizeof(output));
 	 else
 	    first = 0;
-	 util_strncat(output, names->name, sizeof(output));
+	 util_strncat(output, names->name, sizeof(output) - 1);
+	 output[sizeof(output) - 1] = '\0';
 	 value &= ~names->value;
       }
       ++names;
@@ -465,7 +468,8 @@ debug_dump_flags(const struct debug_named_value *names,
 	 first = 0;
       
       util_snprintf(rest, sizeof(rest), "0x%08lx", value);
-      util_strncat(output, rest, sizeof(output));
+      util_strncat(output, rest, sizeof(output) - 1);
+      output[sizeof(output) - 1] = '\0';
    }
    
    if(first)
@@ -600,6 +604,32 @@ const char *pf_name( enum pipe_format format )
 }
 
 
+
+static const struct debug_named_value pipe_prim_names[] = {
+#ifdef DEBUG
+   DEBUG_NAMED_VALUE(PIPE_PRIM_POINTS),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_LINES),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_LINE_LOOP),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_LINE_STRIP),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_TRIANGLES),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_TRIANGLE_STRIP),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_TRIANGLE_FAN),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_QUADS),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_QUAD_STRIP),
+   DEBUG_NAMED_VALUE(PIPE_PRIM_POLYGON),
+#endif
+   DEBUG_NAMED_VALUE_END
+};
+
+
+const char *u_prim_name( unsigned prim )
+{
+   return debug_dump_enum(pipe_prim_names, prim);
+}
+
+
+
+
 #ifdef DEBUG
 void debug_dump_image(const char *prefix,
                       unsigned format, unsigned cpp,
@@ -669,10 +699,10 @@ void debug_dump_surface(const char *prefix,
       goto error;
    
    debug_dump_image(prefix, 
-                    transfer->format,
-                    transfer->block.size, 
-                    transfer->nblocksx,
-                    transfer->nblocksy,
+                    texture->format,
+                    util_format_get_blocksize(texture->format), 
+                    util_format_get_nblocksx(texture->format, transfer->width),
+                    util_format_get_nblocksy(texture->format, transfer->height),
                     transfer->stride,
                     data);
    
diff --git a/src/gallium/auxiliary/util/u_debug.h b/src/gallium/auxiliary/util/u_debug.h
index b8c56fd600..facc30a553 100644
--- a/src/gallium/auxiliary/util/u_debug.h
+++ b/src/gallium/auxiliary/util/u_debug.h
@@ -188,7 +188,7 @@ void _debug_assert_fail(const char *expr,
 #ifdef DEBUG
 #define debug_assert(expr) ((expr) ? (void)0 : _debug_assert_fail(#expr, __FILE__, __LINE__, __FUNCTION__))
 #else
-#define debug_assert(expr) ((void)0)
+#define debug_assert(expr) do { } while (0 && (expr))
 #endif
 
 
@@ -351,17 +351,6 @@ void
 debug_memory_end(unsigned long beginning);
 
 
-#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
-
-void
-debug_profile_start(void);
-
-void 
-debug_profile_stop(void);
-
-#endif
-
-
 #ifdef DEBUG
 struct pipe_surface;
 struct pipe_transfer;
diff --git a/src/gallium/auxiliary/util/u_debug_dump.c b/src/gallium/auxiliary/util/u_debug_dump.c
index 6bdecde048..61624d05c0 100644
--- a/src/gallium/auxiliary/util/u_debug_dump.c
+++ b/src/gallium/auxiliary/util/u_debug_dump.c
@@ -187,3 +187,81 @@ debug_dump_func_short_names[] = {
 };
 
 DEFINE_DEBUG_DUMP_CONTINUOUS(func)
+
+
+static const char *
+debug_dump_tex_target_names[] = {
+   "PIPE_TEXTURE_1D",
+   "PIPE_TEXTURE_2D",
+   "PIPE_TEXTURE_3D",
+   "PIPE_TEXTURE_CUBE"
+};
+
+static const char *
+debug_dump_tex_target_short_names[] = {
+   "1d",
+   "2d",
+   "3d",
+   "cube"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(tex_target)
+
+
+static const char *
+debug_dump_tex_wrap_names[] = {
+   "PIPE_TEX_WRAP_REPEAT",
+   "PIPE_TEX_WRAP_CLAMP",
+   "PIPE_TEX_WRAP_CLAMP_TO_EDGE",
+   "PIPE_TEX_WRAP_CLAMP_TO_BORDER",
+   "PIPE_TEX_WRAP_MIRROR_REPEAT",
+   "PIPE_TEX_WRAP_MIRROR_CLAMP",
+   "PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE",
+   "PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER"
+};
+
+static const char *
+debug_dump_tex_wrap_short_names[] = {
+   "repeat",
+   "clamp",
+   "clamp_to_edge",
+   "clamp_to_border",
+   "mirror_repeat",
+   "mirror_clamp",
+   "mirror_clamp_to_edge",
+   "mirror_clamp_to_border"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(tex_wrap)
+
+
+static const char *
+debug_dump_tex_mipfilter_names[] = {
+   "PIPE_TEX_MIPFILTER_NEAREST",
+   "PIPE_TEX_MIPFILTER_LINEAR",
+   "PIPE_TEX_MIPFILTER_NONE"
+};
+
+static const char *
+debug_dump_tex_mipfilter_short_names[] = {
+   "nearest",
+   "linear",
+   "none"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(tex_mipfilter)
+
+
+static const char *
+debug_dump_tex_filter_names[] = {
+   "PIPE_TEX_FILTER_NEAREST",
+   "PIPE_TEX_FILTER_LINEAR"
+};
+
+static const char *
+debug_dump_tex_filter_short_names[] = {
+   "nearest",
+   "linear"
+};
+
+DEFINE_DEBUG_DUMP_CONTINUOUS(tex_filter)
diff --git a/src/gallium/auxiliary/util/u_debug_dump.h b/src/gallium/auxiliary/util/u_debug_dump.h
index 102935559c..19b130ad18 100644
--- a/src/gallium/auxiliary/util/u_debug_dump.h
+++ b/src/gallium/auxiliary/util/u_debug_dump.h
@@ -54,6 +54,18 @@ debug_dump_blend_func(unsigned value, boolean shortened);
 const char *
 debug_dump_func(unsigned value, boolean shortened);
 
+const char *
+debug_dump_tex_target(unsigned value, boolean shortened);
+
+const char *
+debug_dump_tex_wrap(unsigned value, boolean shortened);
+
+const char *
+debug_dump_tex_mipfilter(unsigned value, boolean shortened);
+
+const char *
+debug_dump_tex_filter(unsigned value, boolean shortened);
+
 
 /* FIXME: Move the other debug_dump_xxx functions out of u_debug.h into here. */
 
diff --git a/src/gallium/auxiliary/util/u_debug_memory.c b/src/gallium/auxiliary/util/u_debug_memory.c
index 7623cb9398..d6484f4ad5 100644
--- a/src/gallium/auxiliary/util/u_debug_memory.c
+++ b/src/gallium/auxiliary/util/u_debug_memory.c
@@ -297,9 +297,9 @@ debug_memory_end(unsigned long start_no)
 
       if((start_no <= hdr->no && hdr->no < last_no) ||
 	 (last_no < start_no && (hdr->no < last_no || start_no <= hdr->no))) {
-	 debug_printf("%s:%u:%s: %u bytes at %p not freed\n",
+	 debug_printf("%s:%u:%s: %lu bytes at %p not freed\n",
 		      hdr->file, hdr->line, hdr->function,
-		      hdr->size, ptr);
+		      (unsigned long) hdr->size, ptr);
 #if DEBUG_MEMORY_STACK
 	 debug_backtrace_dump(hdr->backtrace, DEBUG_MEMORY_STACK);
 #endif
@@ -315,8 +315,8 @@ debug_memory_end(unsigned long start_no)
    }
 
    if(total_size) {
-      debug_printf("Total of %u KB of system memory apparently leaked\n",
-		   (total_size + 1023)/1024);
+      debug_printf("Total of %lu KB of system memory apparently leaked\n",
+		   (unsigned long) (total_size + 1023)/1024);
    }
    else {
       debug_printf("No memory leaks detected.\n");
diff --git a/src/gallium/auxiliary/util/u_debug_profile.c b/src/gallium/auxiliary/util/u_debug_profile.c
deleted file mode 100644
index d765b50144..0000000000
--- a/src/gallium/auxiliary/util/u_debug_profile.c
+++ /dev/null
@@ -1,320 +0,0 @@
-/**************************************************************************
- * 
- * 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.
- * 
- **************************************************************************/
-
-/**
- * @file
- * Poor-man profiling.
- * 
- * @author José Fonseca <jrfonseca@tungstengraphics.com>
- * 
- * @sa http://blogs.msdn.com/joshpoley/archive/2008/03/12/poor-man-s-profiler.aspx
- * @sa http://www.johnpanzer.com/aci_cuj/index.html
- */
-
-#include "pipe/p_config.h" 
-
-#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
-
-#include <windows.h>
-#include <winddi.h>
-
-#include "util/u_debug.h" 
-#include "util/u_string.h" 
-
-
-#define PROFILE_TABLE_SIZE (1024*1024)
-#define FILE_NAME_SIZE 256
-
-struct debug_profile_entry
-{
-   uintptr_t caller;
-   uintptr_t callee;
-   uint64_t samples;
-};
-
-static unsigned long enabled = 0;
-
-static WCHAR wFileName[FILE_NAME_SIZE] = L"\\??\\c:\\00000000.prof";
-static ULONG_PTR iFile = 0;
-
-static struct debug_profile_entry *table = NULL;
-static unsigned long free_table_entries = 0;
-static unsigned long max_table_entries = 0;
-
-uint64_t start_stamp = 0;
-uint64_t end_stamp = 0;
-
-
-static void
-debug_profile_entry(uintptr_t caller, uintptr_t callee, uint64_t samples)
-{
-   unsigned hash = ( caller + callee ) & PROFILE_TABLE_SIZE - 1;
-   
-   while(1) {
-      if(table[hash].caller == 0 && table[hash].callee == 0) {
-         table[hash].caller = caller;
-         table[hash].callee = callee;
-         table[hash].samples = samples;
-         --free_table_entries;
-         break;
-      }
-      else if(table[hash].caller == caller && table[hash].callee == callee) {
-         table[hash].samples += samples;
-         break;
-      }
-      else {
-         ++hash;
-      }
-   }
-}
-
-
-static uintptr_t caller_stack[1024];
-static unsigned last_caller = 0;
-
-
-static int64_t delta(void) {
-   int64_t result = end_stamp - start_stamp;
-   if(result > UINT64_C(0xffffffff))
-      result = 0;
-   return result;
-}
-
-
-static void __cdecl 
-debug_profile_enter(uintptr_t callee)
-{
-   uintptr_t caller = last_caller ? caller_stack[last_caller - 1] : 0;
-                
-   if (caller)
-      debug_profile_entry(caller, 0, delta());
-   debug_profile_entry(caller, callee, 1);
-   caller_stack[last_caller++] = callee;
-}
-
-
-static void __cdecl
-debug_profile_exit(uintptr_t callee)
-{
-   debug_profile_entry(callee, 0, delta());
-   if(last_caller)
-      --last_caller;
-}
-   
-   
-/**
- * Called at the start of every method or function.
- * 
- * @sa http://msdn.microsoft.com/en-us/library/c63a9b7h.aspx
- */
-void __declspec(naked) __cdecl 
-_penter(void) {
-   _asm {
-      push eax
-      mov eax, [enabled]
-      test eax, eax
-      jz skip
-
-      push edx
-      
-      rdtsc
-      mov dword ptr [end_stamp], eax
-      mov dword ptr [end_stamp+4], edx
-
-      xor eax, eax
-      mov [enabled], eax
-
-      mov eax, [esp+8]
-
-      push ebx
-      push ecx
-      push ebp
-      push edi
-      push esi
-
-      push eax
-      call debug_profile_enter
-      add esp, 4
-
-      pop esi
-      pop edi
-      pop ebp
-      pop ecx
-      pop ebx
-
-      mov eax, 1
-      mov [enabled], eax 
-
-      rdtsc
-      mov dword ptr [start_stamp], eax
-      mov dword ptr [start_stamp+4], edx
-      
-      pop edx
-skip:
-      pop eax
-      ret
-   }
-}
-
-
-/**
- * Called at the end of Calls the end of every method or function.
- * 
- * @sa http://msdn.microsoft.com/en-us/library/xc11y76y.aspx
- */
-void __declspec(naked) __cdecl 
-_pexit(void) {
-   _asm {
-      push eax
-      mov eax, [enabled]
-      test eax, eax
-      jz skip
-
-      push edx
-      
-      rdtsc
-      mov dword ptr [end_stamp], eax
-      mov dword ptr [end_stamp+4], edx
-
-      xor eax, eax
-      mov [enabled], eax
-
-      mov eax, [esp+8]
-
-      push ebx
-      push ecx
-      push ebp
-      push edi
-      push esi
-
-      push eax
-      call debug_profile_exit
-      add esp, 4
-
-      pop esi
-      pop edi
-      pop ebp
-      pop ecx
-      pop ebx
-
-      mov eax, 1
-      mov [enabled], eax 
-
-      rdtsc
-      mov dword ptr [start_stamp], eax
-      mov dword ptr [start_stamp+4], edx
-      
-      pop edx
-skip:
-      pop eax
-      ret
-   }
-}
-
-
-/**
- * Reference function for calibration. 
- */
-void __declspec(naked) 
-__debug_profile_reference(void) {
-   _asm {
-      call _penter
-      call _pexit
-      ret
-   }
-}
-
-
-void
-debug_profile_start(void)
-{
-   WCHAR *p;
-
-   /* increment starting from the less significant digit */
-   p = &wFileName[14];
-   while(1) {
-      if(*p == '9') {
-         *p-- = '0';
-      }
-      else {
-         *p += 1;
-         break;
-      }
-   }
-
-   table = EngMapFile(wFileName, 
-                      PROFILE_TABLE_SIZE*sizeof(struct debug_profile_entry), 
-                      &iFile);
-   if(table) {
-      unsigned i;
-      
-      free_table_entries = max_table_entries = PROFILE_TABLE_SIZE;
-      memset(table, 0, PROFILE_TABLE_SIZE*sizeof(struct debug_profile_entry));
-      
-      table[0].caller = (uintptr_t)&__debug_profile_reference;
-      table[0].callee = 0;
-      table[0].samples = 0;
-      --free_table_entries;
-
-      _asm {
-         push edx
-         push eax
-      
-         rdtsc
-         mov dword ptr [start_stamp], eax
-         mov dword ptr [start_stamp+4], edx
-         
-         pop edx
-         pop eax
-      }
-
-      last_caller = 0;
-      
-      enabled = 1;
-
-      for(i = 0; i < 8; ++i) {
-         _asm {
-            call __debug_profile_reference
-         }
-      }
-   }
-}
-
-
-void 
-debug_profile_stop(void)
-{
-   enabled = 0;
-
-   if(iFile)
-      EngUnmapFile(iFile);
-   iFile = 0;
-   table = NULL;
-   free_table_entries = max_table_entries = 0;
-}
-
-#endif /* PROFILE */
diff --git a/src/gallium/auxiliary/util/u_dl.c b/src/gallium/auxiliary/util/u_dl.c
new file mode 100644
index 0000000000..b42b429d4d
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_dl.c
@@ -0,0 +1,79 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * Copyright 1999-2008  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, 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 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
+ * THE COPYRIGHT HOLDERS, AUTHORS 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.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_OS_UNIX)
+#include <dlfcn.h>
+#endif
+#if defined(PIPE_OS_WINDOWS)
+#include <windows.h>
+#endif
+
+#include "u_dl.h"
+
+
+struct util_dl_library *
+util_dl_open(const char *filename)
+{
+#if defined(PIPE_OS_UNIX)
+   return (struct util_dl_library *)dlopen(filename, RTLD_LAZY | RTLD_GLOBAL);
+#elif defined(PIPE_OS_WINDOWS)
+   return (struct util_dl_library *)LoadLibraryA(filename);
+#else
+   return NULL;
+#endif
+}
+
+
+util_dl_proc
+util_dl_get_proc_address(struct util_dl_library *library,
+                         const char *procname)
+{
+#if defined(PIPE_OS_UNIX)
+   return (util_dl_proc)dlsym((void *)library, procname);
+#elif defined(PIPE_OS_WINDOWS)
+   return (util_dl_proc)GetProcAddress((HMODULE)library, procname);
+#else
+   return (util_dl_proc)NULL;
+#endif
+}
+
+
+void
+util_dl_close(struct util_dl_library *library)
+{
+#if defined(PIPE_OS_UNIX)
+   dlclose((void *)library);
+#elif defined(PIPE_OS_WINDOWS)
+   FreeLibrary((HMODULE)library);
+#else
+   (void)library;
+#endif
+}
diff --git a/src/gallium/auxiliary/util/u_dl.h b/src/gallium/auxiliary/util/u_dl.h
new file mode 100644
index 0000000000..85296c58af
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_dl.h
@@ -0,0 +1,73 @@
+/**************************************************************************
+ *
+ * Copyright 2009 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 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
+ * THE COPYRIGHT HOLDERS, AUTHORS 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.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ **************************************************************************/
+
+
+#ifndef U_DL_H_
+#define U_DL_H_
+
+
+#include "pipe/p_config.h"
+
+
+#if defined(PIPE_OS_WINDOWS)
+#  define UTIL_DL_EXT ".dll"
+#elif defined(PIPE_OS_APPLE)
+#  define UTIL_DL_EXT ".dylib"
+#else
+#  define UTIL_DL_EXT ".so"
+#endif
+
+
+struct util_dl_library;
+
+
+typedef void (*util_dl_proc)(void);
+
+
+/**
+ * Open a library dynamically.
+ */
+struct util_dl_library *
+util_dl_open(const char *filename);
+
+
+/**
+ * Lookup a function in a library.
+ */
+util_dl_proc
+util_dl_get_proc_address(struct util_dl_library *library,
+                         const char *procname);
+
+
+/**
+ * Close a library.
+ */
+void
+util_dl_close(struct util_dl_library *library);
+
+
+#endif /* U_DL_H_ */
diff --git a/src/gallium/auxiliary/util/u_format.c b/src/gallium/auxiliary/util/u_format.c
index 98ea13b60b..e0724a1a8b 100644
--- a/src/gallium/auxiliary/util/u_format.c
+++ b/src/gallium/auxiliary/util/u_format.c
@@ -32,15 +32,14 @@
 const struct util_format_description *
 util_format_description(enum pipe_format format)
 {
-   const struct util_format_description *desc = util_format_description_table;
+   const struct util_format_description *desc;
 
-   while(TRUE) {
-      if(desc->format == format)
-         return desc;
+   if (format >= PIPE_FORMAT_COUNT) {
+      return NULL;
+   }
 
-      if(desc->format == PIPE_FORMAT_NONE)
-         return NULL;
+   desc = &util_format_description_table[format];
+   assert(desc->format == format);
 
-      ++desc;
-   };
+   return desc;
 }
diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv
index f1bf94f17d..9f16b42944 100644
--- a/src/gallium/auxiliary/util/u_format.csv
+++ b/src/gallium/auxiliary/util/u_format.csv
@@ -11,6 +11,8 @@ PIPE_FORMAT_A8_UNORM              , arith , 1, 1, un8 ,     ,     ,     , 000x,
 PIPE_FORMAT_I8_UNORM              , arith , 1, 1, un8 ,     ,     ,     , xxxx, rgb
 PIPE_FORMAT_A8L8_UNORM            , arith , 1, 1, un8 , un8 ,     ,     , xxxy, rgb
 PIPE_FORMAT_L16_UNORM             , arith , 1, 1, un16,     ,     ,     , xxx1, rgb
+PIPE_FORMAT_YCBCR                 , yuv   , 2, 1, x32 ,     ,     ,     , xyz1, yuv
+PIPE_FORMAT_YCBCR_REV             , yuv   , 2, 1, x32 ,     ,     ,     , xyz1, yuv
 PIPE_FORMAT_Z16_UNORM             , array , 1, 1, un16,     ,     ,     , x___, zs 
 PIPE_FORMAT_Z32_UNORM             , array , 1, 1, un32,     ,     ,     , x___, zs 
 PIPE_FORMAT_Z32_FLOAT             , array , 1, 1, f32 ,     ,     ,     , x___, zs 
@@ -74,9 +76,9 @@ PIPE_FORMAT_R8G8_SNORM            , array , 1, 1, sn8 , sn8 ,     ,     , xy01,
 PIPE_FORMAT_R8G8B8_SNORM          , array , 1, 1, sn8 , sn8 , sn8 ,     , xyz1, rgb
 PIPE_FORMAT_R8G8B8A8_SNORM        , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyzw, rgb
 PIPE_FORMAT_R8G8B8X8_SNORM        , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyz1, rgb
-PIPE_FORMAT_B6G5R5_SNORM          , arith , 1, 1, sn5 , sn5 , sn6 ,     , zyx1, rgb
-PIPE_FORMAT_A8B8G8R8_SNORM        , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyxw, rgb
-PIPE_FORMAT_X8B8G8R8_SNORM        , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyx1, rgb
+PIPE_FORMAT_B6G5R5_SNORM          , arith , 1, 1, sn5 , sn5 , sn6 ,     , xyz1, rgb
+PIPE_FORMAT_A8B8G8R8_SNORM        , array , 1, 1, sn8 , sn8 , sn8 , sn8 , wzyx, rgb
+PIPE_FORMAT_X8B8G8R8_SNORM        , array , 1, 1, sn8 , sn8 , sn8 , sn8 , wzy1, rgb
 PIPE_FORMAT_R8_SSCALED            , array , 1, 1, s8  ,     ,     ,     , x001, rgb
 PIPE_FORMAT_R8G8_SSCALED          , array , 1, 1, s8  , s8  ,     ,     , xy01, rgb
 PIPE_FORMAT_R8G8B8_SSCALED        , array , 1, 1, s8  , s8  , s8  ,     , xyz1, rgb
@@ -88,12 +90,20 @@ PIPE_FORMAT_R32G32B32_FIXED       , array , 1, 1, h32 , h32 , h32 ,     , xyz1,
 PIPE_FORMAT_R32G32B32A32_FIXED    , array , 1, 1, h32 , h32 , h32 , h32 , xyzw, rgb
 PIPE_FORMAT_L8_SRGB               , arith , 1, 1, u8  ,     ,     ,     , xxx1, srgb 
 PIPE_FORMAT_A8L8_SRGB             , arith , 1, 1, u8  , u8  ,     ,     , xxxy, srgb 
-PIPE_FORMAT_R8G8B8_SRGB           , arith , 1, 1, u8  , u8  , u8  ,     , xyz1, srgb 
-PIPE_FORMAT_R8G8B8A8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , xyzw, srgb 
-PIPE_FORMAT_R8G8B8X8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , xyz1, srgb 
-PIPE_FORMAT_A8R8G8B8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , wxyz, srgb 
-PIPE_FORMAT_X8R8G8B8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , 1xyz, srgb 
-PIPE_FORMAT_B8G8R8A8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , zyxw, srgb 
-PIPE_FORMAT_B8G8R8X8_SRGB         , arith , 1, 1, u8  , u8  , u8  , u8  , zyx1, srgb 
-PIPE_FORMAT_X8UB8UG8SR8S_NORM     , arith , 1, 1, sn8 , sn8 , un8 , x8  , 1zyx, rgb
+PIPE_FORMAT_R8G8B8_SRGB           , array , 1, 1, u8  , u8  , u8  ,     , xyz1, srgb 
+PIPE_FORMAT_R8G8B8A8_SRGB         , array , 1, 1, u8  , u8  , u8  , u8  , xyzw, srgb 
+PIPE_FORMAT_R8G8B8X8_SRGB         , array , 1, 1, u8  , u8  , u8  , u8  , xyz1, srgb 
+PIPE_FORMAT_A8R8G8B8_SRGB         , array , 1, 1, u8  , u8  , u8  , u8  , yzwx, srgb 
+PIPE_FORMAT_X8R8G8B8_SRGB         , array , 1, 1, u8  , u8  , u8  , u8  , yzw1, srgb 
+PIPE_FORMAT_B8G8R8A8_SRGB         , array , 1, 1, u8  , u8  , u8  , u8  , zyxw, srgb 
+PIPE_FORMAT_B8G8R8X8_SRGB         , array , 1, 1, u8  , u8  , u8  , u8  , zyx1, srgb 
+PIPE_FORMAT_X8UB8UG8SR8S_NORM     , array , 1, 1, sn8 , sn8 , un8 , x8  , wzy1, rgb
 PIPE_FORMAT_B6UG5SR5S_NORM        , arith , 1, 1, sn5 , sn5 , un6 ,     , xyz1, rgb
+PIPE_FORMAT_DXT1_RGB              , dxt   , 4, 4, x64 ,     ,     ,     , xyz1, rgb
+PIPE_FORMAT_DXT1_RGBA             , dxt   , 4, 4, x64 ,     ,     ,     , xyzw, rgb
+PIPE_FORMAT_DXT3_RGBA             , dxt   , 4, 4, x128,     ,     ,     , xyzw, rgb
+PIPE_FORMAT_DXT5_RGBA             , dxt   , 4, 4, x128,     ,     ,     , xyzw, rgb
+PIPE_FORMAT_DXT1_SRGB             , dxt   , 4, 4, x64 ,     ,     ,     , xyz1, srgb
+PIPE_FORMAT_DXT1_SRGBA            , dxt   , 4, 4, x64 ,     ,     ,     , xyzw, srgb
+PIPE_FORMAT_DXT3_SRGBA            , dxt   , 4, 4, x128,     ,     ,     , xyzw, srgb
+PIPE_FORMAT_DXT5_SRGBA            , dxt   , 4, 4, x128,     ,     ,     , xyzw, srgb
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index 7b5b7fcda5..a558923b2e 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -32,11 +32,51 @@
 
 #include "pipe/p_format.h"
 
+#ifdef __cplusplus
+extern "C" {
+#endif
 
+
+/**
+ * Describe how to best pack/unpack pixels into/from the prescribed format.
+ *
+ * These are used for automatic code generation of pixel packing and unpacking
+ * routines (in compile time, e.g., u_format_access.py, or in runtime, like
+ * llvmpipe does).
+ *
+ * Thumb rule is: if you're not code generating pixel packing/unpacking then
+ * these are irrelevant for you.
+ *
+ * Note that this can be deduced from other values in util_format_description
+ * structure. This is by design, to make code generation of pixel
+ * packing/unpacking/sampling routines simple and efficient.
+ *
+ * XXX: This should be renamed to something like util_format_pack.
+ */
 enum util_format_layout {
+   /**
+    * Single scalar component.
+    */
    UTIL_FORMAT_LAYOUT_SCALAR = 0,
+
+   /**
+    * One or more components of mixed integer formats, arithmetically encoded
+    * in a word up to 32bits.
+    */
    UTIL_FORMAT_LAYOUT_ARITH = 1,
+
+   /**
+    * One or more components, no mixed formats, each with equal power of two
+    * number of bytes.
+    */
    UTIL_FORMAT_LAYOUT_ARRAY = 2,
+
+   /**
+    * XXX: Not used yet. These might go away and be replaced by a single entry,
+    * for formats where multiple pixels have to be
+    * read in order to determine a single pixel value (i.e., block.width > 1
+    * || block.height > 1)
+    */
    UTIL_FORMAT_LAYOUT_YUV = 3,
    UTIL_FORMAT_LAYOUT_DXT = 4
 };
@@ -50,7 +90,7 @@ struct util_format_block
    /** Block height in pixels */
    unsigned height;
 
-   /** Block size in bytes */
+   /** Block size in bits */
    unsigned bits;
 };
 
@@ -79,7 +119,7 @@ enum util_format_colorspace {
    UTIL_FORMAT_COLORSPACE_RGB = 0,
    UTIL_FORMAT_COLORSPACE_SRGB = 1,
    UTIL_FORMAT_COLORSPACE_YUV = 2,
-   UTIL_FORMAT_COLORSPACE_ZS = 3,
+   UTIL_FORMAT_COLORSPACE_ZS = 3
 };
 
 
@@ -111,6 +151,250 @@ const struct util_format_description *
 util_format_description(enum pipe_format format);
 
 
+/*
+ * Format query functions.
+ */
+
+static INLINE boolean 
+util_format_is_compressed(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return FALSE;
+   }
+
+   return desc->layout == UTIL_FORMAT_LAYOUT_DXT ? TRUE : FALSE;
+}
+
+static INLINE boolean 
+util_format_is_depth_or_stencil(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return FALSE;
+   }
+
+   return desc->colorspace == UTIL_FORMAT_COLORSPACE_ZS ? TRUE : FALSE;
+}
+
+static INLINE boolean 
+util_format_is_depth_and_stencil(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return FALSE;
+   }
+
+   if (desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS) {
+      return FALSE;
+   }
+
+   return (desc->swizzle[0] != UTIL_FORMAT_SWIZZLE_NONE &&
+           desc->swizzle[1] != UTIL_FORMAT_SWIZZLE_NONE) ? TRUE : FALSE;
+}
+
+
+/**
+ * Return total bits needed for the pixel format per block.
+ */
+static INLINE uint
+util_format_get_blocksizebits(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return 0;
+   }
+
+   return desc->block.bits;
+}
+
+/**
+ * Return bytes per block (not pixel) for the given format.
+ */
+static INLINE uint
+util_format_get_blocksize(enum pipe_format format)
+{
+   uint bits = util_format_get_blocksizebits(format);
+
+   assert(bits % 8 == 0);
+
+   return bits / 8;
+}
+
+static INLINE uint
+util_format_get_blockwidth(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return 1;
+   }
+
+   switch (desc->layout) {
+   case UTIL_FORMAT_LAYOUT_YUV:
+      return 2;
+   case UTIL_FORMAT_LAYOUT_DXT:
+      return 4;
+   default:
+      return 1;
+   }
+}
+
+static INLINE uint
+util_format_get_blockheight(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return 1;
+   }
+
+   switch (desc->layout) {
+   case UTIL_FORMAT_LAYOUT_DXT:
+      return 4;
+   default:
+      return 1;
+   }
+}
+
+static INLINE unsigned
+util_format_get_nblocksx(enum pipe_format format,
+                         unsigned x)
+{
+   unsigned blockwidth = util_format_get_blockwidth(format);
+   return (x + blockwidth - 1) / blockwidth;
+}
+
+static INLINE unsigned
+util_format_get_nblocksy(enum pipe_format format,
+                         unsigned y)
+{
+   unsigned blockheight = util_format_get_blockheight(format);
+   return (y + blockheight - 1) / blockheight;
+}
+
+static INLINE unsigned
+util_format_get_nblocks(enum pipe_format format,
+                        unsigned width,
+                        unsigned height)
+{
+   return util_format_get_nblocksx(format, width) * util_format_get_nblocksy(format, height);
+}
+
+static INLINE size_t
+util_format_get_stride(enum pipe_format format,
+                       unsigned width)
+{
+   return util_format_get_nblocksx(format, width) * util_format_get_blocksize(format);
+}
+
+static INLINE size_t
+util_format_get_2d_size(enum pipe_format format,
+                        size_t stride,
+                        unsigned height)
+{
+   return util_format_get_nblocksy(format, height) * stride;
+}
+
+static INLINE uint
+util_format_get_component_bits(enum pipe_format format,
+                               enum util_format_colorspace colorspace,
+                               uint component)
+{
+   const struct util_format_description *desc = util_format_description(format);
+   enum util_format_colorspace desc_colorspace;
+
+   assert(format);
+   if (!format) {
+      return 0;
+   }
+
+   assert(component < 4);
+
+   /* Treat RGB and SRGB as equivalent. */
+   if (colorspace == UTIL_FORMAT_COLORSPACE_SRGB) {
+      colorspace = UTIL_FORMAT_COLORSPACE_RGB;
+   }
+   if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) {
+      desc_colorspace = UTIL_FORMAT_COLORSPACE_RGB;
+   } else {
+      desc_colorspace = desc->colorspace;
+   }
+
+   if (desc_colorspace != colorspace) {
+      return 0;
+   }
+
+   switch (desc->swizzle[component]) {
+   case UTIL_FORMAT_SWIZZLE_X:
+      return desc->channel[0].size;
+   case UTIL_FORMAT_SWIZZLE_Y:
+      return desc->channel[1].size;
+   case UTIL_FORMAT_SWIZZLE_Z:
+      return desc->channel[2].size;
+   case UTIL_FORMAT_SWIZZLE_W:
+      return desc->channel[3].size;
+   default:
+      return 0;
+   }
+}
+
+static INLINE boolean
+util_format_has_alpha(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(format);
+   if (!format) {
+      return FALSE;
+   }
+
+   switch (desc->layout) {
+   case UTIL_FORMAT_LAYOUT_SCALAR:
+   case UTIL_FORMAT_LAYOUT_ARITH:
+   case UTIL_FORMAT_LAYOUT_ARRAY:
+      /* FIXME: pf_get_component_bits( PIPE_FORMAT_A8L8_UNORM, PIPE_FORMAT_COMP_A ) should not return 0 right? */
+      if (format == PIPE_FORMAT_A8_UNORM ||
+          format == PIPE_FORMAT_A8L8_UNORM ||
+          format == PIPE_FORMAT_A8L8_SRGB) {
+         return TRUE;
+      }
+      return util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3) != 0;
+   case UTIL_FORMAT_LAYOUT_YUV:
+      return FALSE;
+   case UTIL_FORMAT_LAYOUT_DXT:
+      switch (format) {
+      case PIPE_FORMAT_DXT1_RGBA:
+      case PIPE_FORMAT_DXT3_RGBA:
+      case PIPE_FORMAT_DXT5_RGBA:
+      case PIPE_FORMAT_DXT1_SRGBA:
+      case PIPE_FORMAT_DXT3_SRGBA:
+      case PIPE_FORMAT_DXT5_SRGBA:
+         return TRUE;
+      default:
+         return FALSE;
+      }
+   default:
+      assert(0);
+      return FALSE;
+   }
+}
+
+
+/*
+ * Format access functions.
+ */
+
 void
 util_format_read_4f(enum pipe_format format,
                     float *dst, unsigned dst_stride, 
@@ -135,4 +419,8 @@ util_format_write_4ub(enum pipe_format format,
                       void *dst, unsigned dst_stride, 
                       unsigned x, unsigned y, unsigned w, unsigned h);
 
+#ifdef __cplusplus
+} // extern "C" {
+#endif
+
 #endif /* ! U_FORMAT_H */
diff --git a/src/gallium/auxiliary/util/u_format_access.py b/src/gallium/auxiliary/util/u_format_access.py
index eeb1a9657f..0b05ddb931 100644
--- a/src/gallium/auxiliary/util/u_format_access.py
+++ b/src/gallium/auxiliary/util/u_format_access.py
@@ -325,14 +325,14 @@ def generate_format_read(format, dst_type, dst_native_type, dst_suffix):
             elif swizzle == SWIZZLE_0:
                 value = '0'
             elif swizzle == SWIZZLE_1:
-                value = '1'
+                value = get_one(dst_type)
             else:
                 assert False
         elif format.colorspace == 'zs':
             if i < 3:
                 value = 'z'
             else:
-                value = '1'
+                value = get_one(dst_type)
         else:
             assert False
         print '         *dst_pixel++ = %s; /* %s */' % (value, 'rgba'[i])
diff --git a/src/gallium/auxiliary/util/u_format_table.py b/src/gallium/auxiliary/util/u_format_table.py
index 8834568e8e..571cab55dc 100755
--- a/src/gallium/auxiliary/util/u_format_table.py
+++ b/src/gallium/auxiliary/util/u_format_table.py
@@ -44,11 +44,10 @@ def colorspace_map(colorspace):
 
 
 colorspace_channels_map = {
-    'rgb': 'rgba',
-    'rgba': 'rgba',
-    'zs': 'zs',
-    'yuv': ['y1', 'y2', 'u', 'v'],
-    'dxt': []
+    'rgb': ['r', 'g', 'b', 'a'],
+    'srgb': ['sr', 'sg', 'sb', 'a'],
+    'zs': ['z', 's'],
+    'yuv': ['y', 'u', 'v'],
 }
 
 
@@ -90,11 +89,20 @@ def write_format_table(formats):
     print 'const struct util_format_description'
     print 'util_format_description_table[] = '
     print "{"
+    print "   {"
+    print "      PIPE_FORMAT_NONE,"
+    print "      \"PIPE_FORMAT_NONE\","
+    print "      {0, 0, 0},"
+    print "      0,"
+    print "      {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}},"
+    print "      {0, 0, 0, 0},"
+    print "      0"
+    print "   },"
     for format in formats:
         print "   {"
         print "      %s," % (format.name,)
         print "      \"%s\"," % (format.name,)
-        print "      {%u, %u, %u}, /* block */" % (format.block_width, format.block_height, format.block_size())
+        print "      {%u, %u, %u},\t/* block */" % (format.block_width, format.block_height, format.block_size())
         print "      %s," % (layout_map(format.layout),)
         print "      {"
         for i in range(4):
@@ -103,7 +111,7 @@ def write_format_table(formats):
                 sep = ","
             else:
                 sep = ""
-            print "         {%s, %s, %u}%s /* %s */" % (kind_map[type.kind], bool_map(type.norm), type.size, sep, "xyzw"[i])
+            print "         {%s, %s, %u}%s\t/* %s */" % (kind_map[type.kind], bool_map(type.norm), type.size, sep, "xyzw"[i])
         print "      },"
         print "      {"
         for i in range(4):
@@ -113,22 +121,13 @@ def write_format_table(formats):
             else:
                 sep = ""
             try:
-                comment = layout_channels_map[format.layout][i]
-            except:
+                comment = colorspace_channels_map[format.colorspace][i]
+            except (KeyError, IndexError):
                 comment = 'ignored'
-            print "         %s%s /* %s */" % (swizzle_map[swizzle], sep, comment)
+            print "         %s%s\t/* %s */" % (swizzle_map[swizzle], sep, comment)
         print "      },"
         print "      %s," % (colorspace_map(format.colorspace),)
         print "   },"
-    print "   {"
-    print "      PIPE_FORMAT_NONE,"
-    print "      \"PIPE_FORMAT_NONE\","
-    print "      {0, 0, 0},"
-    print "      0,"
-    print "      {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}},"
-    print "      {0, 0, 0, 0},"
-    print "      0"
-    print "   },"
     print "};"
 
 
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index aa823aa218..76023794dc 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -41,10 +41,13 @@
 #include "pipe/p_shader_tokens.h"
 #include "pipe/p_state.h"
 
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_draw_quad.h"
 #include "util/u_gen_mipmap.h"
 #include "util/u_simple_shaders.h"
+#include "util/u_math.h"
+#include "util/u_texture.h"
 
 #include "cso_cache/cso_context.h"
 
@@ -60,7 +63,7 @@ struct gen_mipmap_state
    struct pipe_sampler_state sampler;
 
    void *vs;
-   void *fs;
+   void *fs2d, *fsCube;
 
    struct pipe_buffer *vbuf;  /**< quad vertices */
    unsigned vbuf_slot;
@@ -995,7 +998,7 @@ reduce_2d(enum pipe_format pformat,
 {
    enum dtype datatype;
    uint comps;
-   const int bpt = pf_get_size(pformat);
+   const int bpt = util_format_get_blocksize(pformat);
    const ubyte *srcA, *srcB;
    ubyte *dst;
    int row;
@@ -1034,7 +1037,7 @@ reduce_3d(enum pipe_format pformat,
           int dstWidth, int dstHeight, int dstDepth,
           int dstRowStride, ubyte *dstPtr)
 {
-   const int bpt = pf_get_size(pformat);
+   const int bpt = util_format_get_blocksize(pformat);
    const int border = 0;
    int img, row;
    int bytesPerSrcImage, bytesPerDstImage;
@@ -1125,12 +1128,12 @@ make_1d_mipmap(struct gen_mipmap_state *ctx,
       
       srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
                                           PIPE_TRANSFER_READ, 0, 0,
-                                          pt->width[srcLevel],
-                                          pt->height[srcLevel]);
+                                          u_minify(pt->width0, srcLevel),
+                                          u_minify(pt->height0, srcLevel));
       dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
                                           PIPE_TRANSFER_WRITE, 0, 0,
-                                          pt->width[dstLevel],
-                                          pt->height[dstLevel]);
+                                          u_minify(pt->width0, dstLevel),
+                                          u_minify(pt->height0, dstLevel));
 
       srcMap = (ubyte *) screen->transfer_map(screen, srcTrans);
       dstMap = (ubyte *) screen->transfer_map(screen, dstTrans);
@@ -1158,8 +1161,8 @@ make_2d_mipmap(struct gen_mipmap_state *ctx,
    const uint zslice = 0;
    uint dstLevel;
    
-   assert(pt->block.width == 1);
-   assert(pt->block.height == 1);
+   assert(util_format_get_blockwidth(pt->format) == 1);
+   assert(util_format_get_blockheight(pt->format) == 1);
 
    for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
       const uint srcLevel = dstLevel - 1;
@@ -1168,12 +1171,12 @@ make_2d_mipmap(struct gen_mipmap_state *ctx,
       
       srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
                                           PIPE_TRANSFER_READ, 0, 0,
-                                          pt->width[srcLevel],
-                                          pt->height[srcLevel]);
+                                          u_minify(pt->width0, srcLevel),
+                                          u_minify(pt->height0, srcLevel));
       dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
                                           PIPE_TRANSFER_WRITE, 0, 0,
-                                          pt->width[dstLevel],
-                                          pt->height[dstLevel]);
+                                          u_minify(pt->width0, dstLevel),
+                                          u_minify(pt->height0, dstLevel));
 
       srcMap = (ubyte *) screen->transfer_map(screen, srcTrans);
       dstMap = (ubyte *) screen->transfer_map(screen, dstTrans);
@@ -1203,8 +1206,8 @@ make_3d_mipmap(struct gen_mipmap_state *ctx,
    struct pipe_screen *screen = pipe->screen;
    uint dstLevel, zslice = 0;
 
-   assert(pt->block.width == 1);
-   assert(pt->block.height == 1);
+   assert(util_format_get_blockwidth(pt->format) == 1);
+   assert(util_format_get_blockheight(pt->format) == 1);
 
    for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
       const uint srcLevel = dstLevel - 1;
@@ -1213,12 +1216,12 @@ make_3d_mipmap(struct gen_mipmap_state *ctx,
       
       srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
                                           PIPE_TRANSFER_READ, 0, 0,
-                                          pt->width[srcLevel],
-                                          pt->height[srcLevel]);
+                                          u_minify(pt->width0, srcLevel),
+                                          u_minify(pt->height0, srcLevel));
       dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
                                           PIPE_TRANSFER_WRITE, 0, 0,
-                                          pt->width[dstLevel],
-                                          pt->height[dstLevel]);
+                                          u_minify(pt->width0, dstLevel),
+                                          u_minify(pt->height0, dstLevel));
 
       srcMap = (ubyte *) screen->transfer_map(screen, srcTrans);
       dstMap = (ubyte *) screen->transfer_map(screen, dstTrans);
@@ -1316,7 +1319,8 @@ util_create_gen_mipmap(struct pipe_context *pipe,
    }
 
    /* fragment shader */
-   ctx->fs = util_make_fragment_tex_shader(pipe);
+   ctx->fs2d = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_2D);
+   ctx->fsCube = util_make_fragment_tex_shader(pipe, TGSI_TEXTURE_CUBE);
 
    /* vertex data that doesn't change */
    for (i = 0; i < 4; i++) {
@@ -1382,59 +1386,9 @@ set_vertex_data(struct gen_mipmap_state *ctx,
       static const float st[4][2] = {
          {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
       };
-      float rx, ry, rz;
-      uint i;
-
-      /* loop over quad verts */
-      for (i = 0; i < 4; i++) {
-         /* Compute sc = +/-scale and tc = +/-scale.
-          * Not +/-1 to avoid cube face selection ambiguity near the edges,
-          * though that can still sometimes happen with this scale factor...
-          */
-         const float scale = 0.9999f;
-         const float sc = (2.0f * st[i][0] - 1.0f) * scale;
-         const float tc = (2.0f * st[i][1] - 1.0f) * scale;
-
-         switch (face) {
-         case PIPE_TEX_FACE_POS_X:
-            rx = 1.0f;
-            ry = -tc;
-            rz = -sc;
-            break;
-         case PIPE_TEX_FACE_NEG_X:
-            rx = -1.0f;
-            ry = -tc;
-            rz = sc;
-            break;
-         case PIPE_TEX_FACE_POS_Y:
-            rx = sc;
-            ry = 1.0f;
-            rz = tc;
-            break;
-         case PIPE_TEX_FACE_NEG_Y:
-            rx = sc;
-            ry = -1.0f;
-            rz = -tc;
-            break;
-         case PIPE_TEX_FACE_POS_Z:
-            rx = sc;
-            ry = -tc;
-            rz = 1.0f;
-            break;
-         case PIPE_TEX_FACE_NEG_Z:
-            rx = -sc;
-            ry = -tc;
-            rz = -1.0f;
-            break;
-         default:
-            rx = ry = rz = 0.0f;
-            assert(0);
-         }
 
-         ctx->vertices[i][1][0] = rx; /*s*/
-         ctx->vertices[i][1][1] = ry; /*t*/
-         ctx->vertices[i][1][2] = rz; /*r*/
-      }
+      util_map_texcoords2d_onto_cubemap(face, &st[0][0], 2,
+                                        &ctx->vertices[0][1][0], 8);
    }
    else {
       /* 1D/2D */
@@ -1474,7 +1428,8 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx)
    struct pipe_context *pipe = ctx->pipe;
 
    pipe->delete_vs_state(pipe, ctx->vs);
-   pipe->delete_fs_state(pipe, ctx->fs);
+   pipe->delete_fs_state(pipe, ctx->fs2d);
+   pipe->delete_fs_state(pipe, ctx->fsCube);
 
    pipe_buffer_reference(&ctx->vbuf, NULL);
 
@@ -1512,6 +1467,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
    struct pipe_context *pipe = ctx->pipe;
    struct pipe_screen *screen = pipe->screen;
    struct pipe_framebuffer_state fb;
+   void *fs = (pt->target == PIPE_TEXTURE_CUBE) ? ctx->fsCube : ctx->fs2d;
    uint dstLevel;
    uint zslice = 0;
    uint offset;
@@ -1549,7 +1505,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
    cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
    cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
 
-   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+   cso_set_fragment_shader_handle(ctx->cso, fs);
    cso_set_vertex_shader_handle(ctx->cso, ctx->vs);
 
    /* init framebuffer state */
@@ -1575,8 +1531,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
        * Setup framebuffer / dest surface
        */
       fb.cbufs[0] = surf;
-      fb.width = pt->width[dstLevel];
-      fb.height = pt->height[dstLevel];
+      fb.width = u_minify(pt->width0, dstLevel);
+      fb.height = u_minify(pt->height0, dstLevel);
       cso_set_framebuffer(ctx->cso, &fb);
 
       /*
@@ -1597,8 +1553,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
       offset = set_vertex_data(ctx,
                                pt->target,
                                face,
-                               (float) pt->width[dstLevel],
-                               (float) pt->height[dstLevel]);
+                               (float) u_minify(pt->width0, dstLevel),
+                               (float) u_minify(pt->height0, dstLevel));
 
       util_draw_vertex_buffer(ctx->pipe, 
                               ctx->vbuf,
diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c
index 8c2a8f454c..5604e3ac37 100644
--- a/src/gallium/auxiliary/util/u_hash_table.c
+++ b/src/gallium/auxiliary/util/u_hash_table.c
@@ -47,7 +47,7 @@
 #include "util/u_hash_table.h"
 
 
-struct hash_table
+struct util_hash_table
 {
    struct cso_hash *cso;   
    
@@ -61,27 +61,27 @@ struct hash_table
 };
 
 
-struct hash_table_item
+struct util_hash_table_item
 {
    void *key;
    void *value;
 };
 
 
-static INLINE struct hash_table_item *
-hash_table_item(struct cso_hash_iter iter)
+static INLINE struct util_hash_table_item *
+util_hash_table_item(struct cso_hash_iter iter)
 {
-   return (struct hash_table_item *)cso_hash_iter_data(iter);
+   return (struct util_hash_table_item *)cso_hash_iter_data(iter);
 }
 
 
-struct hash_table *
-hash_table_create(unsigned (*hash)(void *key),
-                  int (*compare)(void *key1, void *key2))
+struct util_hash_table *
+util_hash_table_create(unsigned (*hash)(void *key),
+                       int (*compare)(void *key1, void *key2))
 {
-   struct hash_table *ht;
+   struct util_hash_table *ht;
    
-   ht = MALLOC_STRUCT(hash_table);
+   ht = MALLOC_STRUCT(util_hash_table);
    if(!ht)
       return NULL;
    
@@ -99,16 +99,16 @@ hash_table_create(unsigned (*hash)(void *key),
 
 
 static INLINE struct cso_hash_iter
-hash_table_find_iter(struct hash_table *ht,
-                     void *key, 
-                     unsigned key_hash)
+util_hash_table_find_iter(struct util_hash_table *ht,
+                          void *key,
+                          unsigned key_hash)
 {
    struct cso_hash_iter iter;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
    
    iter = cso_hash_find(ht->cso, key_hash);
    while (!cso_hash_iter_is_null(iter)) {
-      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      item = (struct util_hash_table_item *)cso_hash_iter_data(iter);
       if (!ht->compare(item->key, key))
          break;
       iter = cso_hash_iter_next(iter);
@@ -118,17 +118,17 @@ hash_table_find_iter(struct hash_table *ht,
 }
 
 
-static INLINE struct hash_table_item *
-hash_table_find_item(struct hash_table *ht,
-                     void *key, 
-                     unsigned key_hash)
+static INLINE struct util_hash_table_item *
+util_hash_table_find_item(struct util_hash_table *ht,
+                          void *key,
+                          unsigned key_hash)
 {
    struct cso_hash_iter iter;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
    
    iter = cso_hash_find(ht->cso, key_hash);
    while (!cso_hash_iter_is_null(iter)) {
-      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      item = (struct util_hash_table_item *)cso_hash_iter_data(iter);
       if (!ht->compare(item->key, key))
          return item;
       iter = cso_hash_iter_next(iter);
@@ -139,12 +139,12 @@ hash_table_find_item(struct hash_table *ht,
 
 
 enum pipe_error
-hash_table_set(struct hash_table *ht,
-               void *key,
-               void *value)
+util_hash_table_set(struct util_hash_table *ht,
+                    void *key,
+                    void *value)
 {
    unsigned key_hash;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
    struct cso_hash_iter iter;
 
    assert(ht);
@@ -153,14 +153,14 @@ hash_table_set(struct hash_table *ht,
 
    key_hash = ht->hash(key);
 
-   item = hash_table_find_item(ht, key, key_hash);
+   item = util_hash_table_find_item(ht, key, key_hash);
    if(item) {
       /* TODO: key/value destruction? */
       item->value = value;
       return PIPE_OK;
    }
    
-   item = MALLOC_STRUCT(hash_table_item);
+   item = MALLOC_STRUCT(util_hash_table_item);
    if(!item)
       return PIPE_ERROR_OUT_OF_MEMORY;
    
@@ -178,11 +178,11 @@ hash_table_set(struct hash_table *ht,
 
 
 void *
-hash_table_get(struct hash_table *ht, 
-               void *key)
+util_hash_table_get(struct util_hash_table *ht,
+                    void *key)
 {
    unsigned key_hash;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
 
    assert(ht);
    if (!ht)
@@ -190,7 +190,7 @@ hash_table_get(struct hash_table *ht,
 
    key_hash = ht->hash(key);
 
-   item = hash_table_find_item(ht, key, key_hash);
+   item = util_hash_table_find_item(ht, key, key_hash);
    if(!item)
       return NULL;
    
@@ -199,12 +199,12 @@ hash_table_get(struct hash_table *ht,
 
 
 void
-hash_table_remove(struct hash_table *ht, 
-                  void *key)
+util_hash_table_remove(struct util_hash_table *ht,
+                       void *key)
 {
    unsigned key_hash;
    struct cso_hash_iter iter;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
 
    assert(ht);
    if (!ht)
@@ -212,11 +212,11 @@ hash_table_remove(struct hash_table *ht,
 
    key_hash = ht->hash(key);
 
-   iter = hash_table_find_iter(ht, key, key_hash);
+   iter = util_hash_table_find_iter(ht, key, key_hash);
    if(cso_hash_iter_is_null(iter))
       return;
    
-   item = hash_table_item(iter);
+   item = util_hash_table_item(iter);
    assert(item);
    FREE(item);
    
@@ -225,10 +225,10 @@ hash_table_remove(struct hash_table *ht,
 
 
 void 
-hash_table_clear(struct hash_table *ht)
+util_hash_table_clear(struct util_hash_table *ht)
 {
    struct cso_hash_iter iter;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
 
    assert(ht);
    if (!ht)
@@ -236,7 +236,7 @@ hash_table_clear(struct hash_table *ht)
 
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
-      item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter));
+      item = (struct util_hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter));
       FREE(item);
       iter = cso_hash_first_node(ht->cso);
    }
@@ -244,12 +244,13 @@ hash_table_clear(struct hash_table *ht)
 
 
 enum pipe_error
-hash_table_foreach(struct hash_table *ht,
-                   enum pipe_error (*callback)(void *key, void *value, void *data),
-                   void *data)
+util_hash_table_foreach(struct util_hash_table *ht,
+                     enum pipe_error (*callback)
+                        (void *key, void *value, void *data),
+                     void *data)
 {
    struct cso_hash_iter iter;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
    enum pipe_error result;
 
    assert(ht);
@@ -258,7 +259,7 @@ hash_table_foreach(struct hash_table *ht,
 
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
-      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      item = (struct util_hash_table_item *)cso_hash_iter_data(iter);
       result = callback(item->key, item->value, data);
       if(result != PIPE_OK)
 	 return result;
@@ -270,10 +271,10 @@ hash_table_foreach(struct hash_table *ht,
 
 
 void
-hash_table_destroy(struct hash_table *ht)
+util_hash_table_destroy(struct util_hash_table *ht)
 {
    struct cso_hash_iter iter;
-   struct hash_table_item *item;
+   struct util_hash_table_item *item;
 
    assert(ht);
    if (!ht)
@@ -281,7 +282,7 @@ hash_table_destroy(struct hash_table *ht)
 
    iter = cso_hash_first_node(ht->cso);
    while (!cso_hash_iter_is_null(iter)) {
-      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      item = (struct util_hash_table_item *)cso_hash_iter_data(iter);
       FREE(item);
       iter = cso_hash_iter_next(iter);
    }
diff --git a/src/gallium/auxiliary/util/u_hash_table.h b/src/gallium/auxiliary/util/u_hash_table.h
index feee881582..51ec10a804 100644
--- a/src/gallium/auxiliary/util/u_hash_table.h
+++ b/src/gallium/auxiliary/util/u_hash_table.h
@@ -35,7 +35,7 @@
 #define U_HASH_TABLE_H_
 
 
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 
 
 #ifdef __cplusplus
@@ -46,7 +46,7 @@ extern "C" {
 /**
  * Generic purpose hash table.
  */
-struct hash_table;
+struct util_hash_table;
 
 
 /**
@@ -55,37 +55,38 @@ struct hash_table;
  * @param hash hash function
  * @param compare should return 0 for two equal keys.
  */
-struct hash_table *
-hash_table_create(unsigned (*hash)(void *key),
-                  int (*compare)(void *key1, void *key2));
+struct util_hash_table *
+util_hash_table_create(unsigned (*hash)(void *key),
+                       int (*compare)(void *key1, void *key2));
 
 
 enum pipe_error
-hash_table_set(struct hash_table *ht,
-               void *key,
-               void *value);
+util_hash_table_set(struct util_hash_table *ht,
+                    void *key,
+                    void *value);
 
 void *
-hash_table_get(struct hash_table *ht, 
-               void *key);
+util_hash_table_get(struct util_hash_table *ht,
+                    void *key);
 
 
 void
-hash_table_remove(struct hash_table *ht, 
-                  void *key);
+util_hash_table_remove(struct util_hash_table *ht,
+                       void *key);
 
 
 void
-hash_table_clear(struct hash_table *ht);
+util_hash_table_clear(struct util_hash_table *ht);
 
 
 enum pipe_error
-hash_table_foreach(struct hash_table *ht,
-                   enum pipe_error (*callback)(void *key, void *value, void *data),
-                   void *data);
+util_hash_table_foreach(struct util_hash_table *ht,
+                        enum pipe_error (*callback)
+                        (void *key, void *value, void *data),
+                        void *data);
 
 void
-hash_table_destroy(struct hash_table *ht);
+util_hash_table_destroy(struct util_hash_table *ht);
 
 
 #ifdef __cplusplus
diff --git a/src/gallium/auxiliary/util/u_keymap.c b/src/gallium/auxiliary/util/u_keymap.c
index 508a2ee063..c4b9eb3d9b 100644
--- a/src/gallium/auxiliary/util/u_keymap.c
+++ b/src/gallium/auxiliary/util/u_keymap.c
@@ -28,7 +28,7 @@
 /**
  * Key lookup/associative container.
  *
- * Like Jose's u_hash_table, based on CSO cache code for now.
+ * Like Jose's util_hash_table, based on CSO cache code for now.
  *
  * Author: Brian Paul
  */
@@ -36,7 +36,7 @@
 
 #include "pipe/p_compiler.h"
 #include "util/u_debug.h"
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 
 #include "cso_cache/cso_hash.h"
 
diff --git a/src/gallium/auxiliary/util/u_linear.c b/src/gallium/auxiliary/util/u_linear.c
index a1dce3f5cf..f1aef21677 100644
--- a/src/gallium/auxiliary/util/u_linear.c
+++ b/src/gallium/auxiliary/util/u_linear.c
@@ -82,7 +82,7 @@ void pipe_linear_from_tile(struct pipe_tile_info *t, const void *src_ptr,
 
 void
 pipe_linear_fill_info(struct pipe_tile_info *t,
-		      const struct pipe_format_block *block,
+		      const struct u_linear_format_block *block,
 		      unsigned tile_width, unsigned tile_height,
 		      unsigned tiles_x, unsigned tiles_y)
 {
diff --git a/src/gallium/auxiliary/util/u_linear.h b/src/gallium/auxiliary/util/u_linear.h
index b74308ffa3..42c40b2aa7 100644
--- a/src/gallium/auxiliary/util/u_linear.h
+++ b/src/gallium/auxiliary/util/u_linear.h
@@ -35,6 +35,19 @@
 
 #include "pipe/p_format.h"
 
+struct u_linear_format_block
+{
+   /** Block size in bytes */
+   unsigned size;
+   
+   /** Block width in pixels */
+   unsigned width;
+   
+   /** Block height in pixels */
+   unsigned height;
+};
+
+
 struct pipe_tile_info
 {
    unsigned size;
@@ -49,10 +62,10 @@ struct pipe_tile_info
    unsigned rows;
 
    /* Describe the tile in pixels */
-   struct pipe_format_block tile;
+   struct u_linear_format_block tile;
 
    /* Describe each block within the tile */
-   struct pipe_format_block block;
+   struct u_linear_format_block block;
 };
 
 void pipe_linear_to_tile(size_t src_stride, const void *src_ptr,
@@ -71,7 +84,7 @@ void pipe_linear_from_tile(struct pipe_tile_info *t, const void *src_ptr,
  * @tiles_y number of tiles in y axis
  */
 void pipe_linear_fill_info(struct pipe_tile_info *t,
-			   const struct pipe_format_block *block,
+			   const struct u_linear_format_block *block,
 			   unsigned tile_width, unsigned tile_height,
 			   unsigned tiles_x, unsigned tiles_y);
 
diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h
index 75b075f160..b2969a210a 100644
--- a/src/gallium/auxiliary/util/u_math.h
+++ b/src/gallium/auxiliary/util/u_math.h
@@ -491,6 +491,47 @@ util_next_power_of_two(unsigned x)
 
 
 /**
+ * Return number of bits set in n.
+ */
+static INLINE unsigned
+util_bitcount(unsigned n)
+{
+#if defined(PIPE_CC_GCC)
+   return __builtin_popcount(n);
+#else
+   /* K&R classic bitcount.
+    *
+    * For each iteration, clear the LSB from the bitfield.
+    * Requires only one iteration per set bit, instead of
+    * one iteration per bit less than highest set bit.
+    */
+   unsigned bits = 0;
+   for (bits; n; bits++) {
+      n &= n - 1;
+   }
+   return bits;
+#endif
+}
+
+
+/**
+ * Reverse byte order of a 32 bit word.
+ */
+static INLINE uint32_t
+util_bswap32(uint32_t n)
+{
+#if defined(PIPE_CC_GCC) && (PIPE_CC_GCC_VERSION >= 403)
+   return __builtin_bswap32(n);
+#else
+   return (n >> 24) |
+          ((n >> 8) & 0x0000ff00) |
+          ((n << 8) & 0x00ff0000) |
+          (n << 24);
+#endif
+}
+
+
+/**
  * Clamp X to [MIN, MAX].
  * This is a macro to allow float, int, uint, etc. types.
  */
@@ -499,6 +540,9 @@ util_next_power_of_two(unsigned x)
 #define MIN2( A, B )   ( (A)<(B) ? (A) : (B) )
 #define MAX2( A, B )   ( (A)>(B) ? (A) : (B) )
 
+#define MIN3( A, B, C ) MIN2( MIN2( A, B ), C )
+#define MAX3( A, B, C ) MAX2( MAX2( A, B ), C )
+
 
 static INLINE int
 align(int value, int alignment)
@@ -507,9 +551,9 @@ align(int value, int alignment)
 }
 
 static INLINE unsigned
-minify(unsigned value)
+u_minify(unsigned value, unsigned levels)
 {
-    return MAX2(1, value >> 1);
+    return MAX2(1, value >> levels);
 }
 
 #ifndef COPY_4V
@@ -539,6 +583,18 @@ do {                                     \
 #endif
 
 
+static INLINE uint32_t util_unsigned_fixed(float value, unsigned frac_bits)
+{
+   return value < 0 ? 0 : (uint32_t)(value * (1<<frac_bits));
+}
+
+static INLINE int32_t util_signed_fixed(float value, unsigned frac_bits)
+{
+   return (int32_t)(value * (1<<frac_bits));
+}
+
+
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c
index 4b75d4ba1d..82f83702d1 100644
--- a/src/gallium/auxiliary/util/u_mm.c
+++ b/src/gallium/auxiliary/util/u_mm.c
@@ -39,13 +39,20 @@ u_mmDumpMemInfo(const struct mem_block *heap)
    }
    else {
       const struct mem_block *p;
+      int total_used = 0, total_free = 0;
 
       for (p = heap->next; p != heap; p = p->next) {
 	 debug_printf("  Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size,
                       p->free ? 'F':'.',
                       p->reserved ? 'R':'.');
+         if (p->free)
+            total_free += p->size;
+         else
+            total_used += p->size;
       }
 
+      debug_printf("'\nMemory stats: total = %d, used = %d, free = %d\n",
+                   total_used + total_free, total_used, total_free);
       debug_printf("\nFree list:\n");
 
       for (p = heap->next_free; p != heap; p = p->next_free) {
diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c
index 6269c72e12..87ee0e4768 100644
--- a/src/gallium/auxiliary/util/u_network.c
+++ b/src/gallium/auxiliary/util/u_network.c
@@ -6,7 +6,7 @@
 #if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
 #  include <winsock2.h>
 #  include <windows.h>
-#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_BSD)
+#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE)
 #  include <sys/socket.h>
 #  include <netinet/in.h>
 #  include <unistd.h>
@@ -54,7 +54,7 @@ u_socket_close(int s)
    if (s < 0)
       return;
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_BSD)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE)
    shutdown(s, SHUT_RDWR);
    close(s);
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
@@ -117,7 +117,7 @@ u_socket_connect(const char *hostname, uint16_t port)
    if (!host)
       return -1;
 
-   memcpy((char *)&sa.sin_addr,host->h_addr,host->h_length);
+   memcpy((char *)&sa.sin_addr,host->h_addr_list[0],host->h_length);
    sa.sin_family= host->h_addrtype;
    sa.sin_port = htons(port);
 
@@ -169,7 +169,7 @@ u_socket_listen_on_port(uint16_t portnum)
 void
 u_socket_block(int s, boolean block)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_BSD)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE)
    int old = fcntl(s, F_GETFL, 0);
    if (old == -1)
       return;
diff --git a/src/gallium/auxiliary/util/u_network.h b/src/gallium/auxiliary/util/u_network.h
index 0aa898b967..187dcab86e 100644
--- a/src/gallium/auxiliary/util/u_network.h
+++ b/src/gallium/auxiliary/util/u_network.h
@@ -6,7 +6,7 @@
 
 #if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
 #  define PIPE_HAVE_SOCKETS
-#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_BSD)
+#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE)
 #  define PIPE_HAVE_SOCKETS
 #endif
 
diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h
index eda883b3b9..43eb0153ee 100644
--- a/src/gallium/auxiliary/util/u_pack_color.h
+++ b/src/gallium/auxiliary/util/u_pack_color.h
@@ -37,106 +37,110 @@
 
 #include "pipe/p_compiler.h"
 #include "pipe/p_format.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 
 
+
+union util_color {
+   ubyte ub;
+   ushort us;
+   uint ui;
+   float f[4];
+};
+
 /**
  * Pack ubyte R,G,B,A into dest pixel.
  */
 static INLINE void
 util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a,
-                   enum pipe_format format, void *dest)
+                   enum pipe_format format, union util_color *uc)
 {
    switch (format) {
    case PIPE_FORMAT_R8G8B8A8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (r << 24) | (g << 16) | (b << 8) | a;
+         uc->ui = (r << 24) | (g << 16) | (b << 8) | a;
       }
       return;
    case PIPE_FORMAT_R8G8B8X8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (r << 24) | (g << 16) | (b << 8) | 0xff;
+         uc->ui = (r << 24) | (g << 16) | (b << 8) | 0xff;
       }
       return;
    case PIPE_FORMAT_A8R8G8B8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (a << 24) | (r << 16) | (g << 8) | b;
+         uc->ui = (a << 24) | (r << 16) | (g << 8) | b;
       }
       return;
    case PIPE_FORMAT_X8R8G8B8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (0xff << 24) | (r << 16) | (g << 8) | b;
+         uc->ui = (0xff << 24) | (r << 16) | (g << 8) | b;
       }
       return;
    case PIPE_FORMAT_B8G8R8A8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (b << 24) | (g << 16) | (r << 8) | a;
+         uc->ui = (b << 24) | (g << 16) | (r << 8) | a;
       }
       return;
    case PIPE_FORMAT_B8G8R8X8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (b << 24) | (g << 16) | (r << 8) | 0xff;
+         uc->ui = (b << 24) | (g << 16) | (r << 8) | 0xff;
       }
       return;
    case PIPE_FORMAT_R5G6B5_UNORM:
       {
-         ushort *d = (ushort *) dest;
-         *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3);
+         uc->us = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3);
       }
       return;
    case PIPE_FORMAT_A1R5G5B5_UNORM:
       {
-         ushort *d = (ushort *) dest;
-         *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3);
+         uc->us = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3);
       }
       return;
    case PIPE_FORMAT_A4R4G4B4_UNORM:
       {
-         ushort *d = (ushort *) dest;
-         *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
+         uc->us = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
       }
       return;
    case PIPE_FORMAT_A8_UNORM:
       {
-         ubyte *d = (ubyte *) dest;
-         *d = a;
+         uc->ub = a;
       }
       return;
    case PIPE_FORMAT_L8_UNORM:
    case PIPE_FORMAT_I8_UNORM:
       {
-         ubyte *d = (ubyte *) dest;
-         *d = r;
+         uc->ub = a;
       }
       return;
    case PIPE_FORMAT_R32G32B32A32_FLOAT:
       {
-         float *d = (float *) dest;
-         d[0] = (float)r / 255.0f;
-         d[1] = (float)g / 255.0f;
-         d[2] = (float)b / 255.0f;
-         d[3] = (float)a / 255.0f;
+         uc->f[0] = (float)r / 255.0f;
+         uc->f[1] = (float)g / 255.0f;
+         uc->f[2] = (float)b / 255.0f;
+         uc->f[3] = (float)a / 255.0f;
       }
       return;
    case PIPE_FORMAT_R32G32B32_FLOAT:
       {
-         float *d = (float *) dest;
-         d[0] = (float)r / 255.0f;
-         d[1] = (float)g / 255.0f;
-         d[2] = (float)b / 255.0f;
+         uc->f[0] = (float)r / 255.0f;
+         uc->f[1] = (float)g / 255.0f;
+         uc->f[2] = (float)b / 255.0f;
       }
       return;
 
-   /* XXX lots more cases to add */
+   /* Handle other cases with a generic function.
+    */
    default:
-      debug_print_format("gallium: unhandled format in util_pack_color_ub()", format);
-      assert(0);
+      {
+         ubyte src[4];
+
+         src[0] = r;
+         src[1] = g;
+         src[2] = b;
+         src[3] = a;
+         util_format_write_4ub(format, src, 0, uc, 0, 0, 0, 1, 1);
+      }
    }
 }
  
@@ -145,13 +149,13 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a,
  * Unpack RGBA from a packed pixel, returning values as ubytes in [0,255].
  */
 static INLINE void
-util_unpack_color_ub(enum pipe_format format, const void *src,
+util_unpack_color_ub(enum pipe_format format, union util_color *uc,
                      ubyte *r, ubyte *g, ubyte *b, ubyte *a)
 {
    switch (format) {
    case PIPE_FORMAT_R8G8B8A8_UNORM:
       {
-         uint p = ((const uint *) src)[0];
+         uint p = uc->ui;
          *r = (ubyte) ((p >> 24) & 0xff);
          *g = (ubyte) ((p >> 16) & 0xff);
          *b = (ubyte) ((p >>  8) & 0xff);
@@ -160,7 +164,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_R8G8B8X8_UNORM:
       {
-         uint p = ((const uint *) src)[0];
+         uint p = uc->ui;
          *r = (ubyte) ((p >> 24) & 0xff);
          *g = (ubyte) ((p >> 16) & 0xff);
          *b = (ubyte) ((p >>  8) & 0xff);
@@ -169,7 +173,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_A8R8G8B8_UNORM:
       {
-         uint p = ((const uint *) src)[0];
+         uint p = uc->ui;
          *r = (ubyte) ((p >> 16) & 0xff);
          *g = (ubyte) ((p >>  8) & 0xff);
          *b = (ubyte) ((p >>  0) & 0xff);
@@ -178,7 +182,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_X8R8G8B8_UNORM:
       {
-         uint p = ((const uint *) src)[0];
+         uint p = uc->ui;
          *r = (ubyte) ((p >> 16) & 0xff);
          *g = (ubyte) ((p >>  8) & 0xff);
          *b = (ubyte) ((p >>  0) & 0xff);
@@ -187,7 +191,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_B8G8R8A8_UNORM:
       {
-         uint p = ((const uint *) src)[0];
+         uint p = uc->ui;
          *r = (ubyte) ((p >>  8) & 0xff);
          *g = (ubyte) ((p >> 16) & 0xff);
          *b = (ubyte) ((p >> 24) & 0xff);
@@ -196,7 +200,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_B8G8R8X8_UNORM:
       {
-         uint p = ((const uint *) src)[0];
+         uint p = uc->ui;
          *r = (ubyte) ((p >>  8) & 0xff);
          *g = (ubyte) ((p >> 16) & 0xff);
          *b = (ubyte) ((p >> 24) & 0xff);
@@ -205,7 +209,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_R5G6B5_UNORM:
       {
-         ushort p = ((const ushort *) src)[0];
+         ushort p = uc->us;
          *r = (ubyte) (((p >> 8) & 0xf8) | ((p >> 13) & 0x7));
          *g = (ubyte) (((p >> 3) & 0xfc) | ((p >>  9) & 0x3));
          *b = (ubyte) (((p << 3) & 0xf8) | ((p >>  2) & 0x7));
@@ -214,7 +218,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_A1R5G5B5_UNORM:
       {
-         ushort p = ((const ushort *) src)[0];
+         ushort p = uc->us;
          *r = (ubyte) (((p >>  7) & 0xf8) | ((p >> 12) & 0x7));
          *g = (ubyte) (((p >>  2) & 0xf8) | ((p >>  7) & 0x7));
          *b = (ubyte) (((p <<  3) & 0xf8) | ((p >>  2) & 0x7));
@@ -223,7 +227,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_A4R4G4B4_UNORM:
       {
-         ushort p = ((const ushort *) src)[0];
+         ushort p = uc->us;
          *r = (ubyte) (((p >> 4) & 0xf0) | ((p >>  8) & 0xf));
          *g = (ubyte) (((p >> 0) & 0xf0) | ((p >>  4) & 0xf));
          *b = (ubyte) (((p << 4) & 0xf0) | ((p >>  0) & 0xf));
@@ -232,27 +236,27 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_A8_UNORM:
       {
-         ubyte p = ((const ubyte *) src)[0];
+         ubyte p = uc->ub;
          *r = *g = *b = (ubyte) 0xff;
          *a = p;
       }
       return;
    case PIPE_FORMAT_L8_UNORM:
       {
-         ubyte p = ((const ubyte *) src)[0];
+         ubyte p = uc->ub;
          *r = *g = *b = p;
          *a = (ubyte) 0xff;
       }
       return;
    case PIPE_FORMAT_I8_UNORM:
       {
-         ubyte p = ((const ubyte *) src)[0];
+         ubyte p = uc->ub;
          *r = *g = *b = *a = p;
       }
       return;
    case PIPE_FORMAT_R32G32B32A32_FLOAT:
       {
-         const float *p = (const float *) src;
+         const float *p = &uc->f[0];
          *r = float_to_ubyte(p[0]);
          *g = float_to_ubyte(p[1]);
          *b = float_to_ubyte(p[2]);
@@ -261,7 +265,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
       return;
    case PIPE_FORMAT_R32G32B32_FLOAT:
       {
-         const float *p = (const float *) src;
+         const float *p = &uc->f[0];
          *r = float_to_ubyte(p[0]);
          *g = float_to_ubyte(p[1]);
          *b = float_to_ubyte(p[2]);
@@ -271,7 +275,7 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
 
    case PIPE_FORMAT_R32G32_FLOAT:
       {
-         const float *p = (const float *) src;
+         const float *p = &uc->f[0];
          *r = float_to_ubyte(p[0]);
          *g = float_to_ubyte(p[1]);
          *b = *a = (ubyte) 0xff;
@@ -280,31 +284,40 @@ util_unpack_color_ub(enum pipe_format format, const void *src,
 
    case PIPE_FORMAT_R32_FLOAT:
       {
-         const float *p = (const float *) src;
+         const float *p = &uc->f[0];
          *r = float_to_ubyte(p[0]);
          *g = *b = *a = (ubyte) 0xff;
       }
       return;
 
-   /* XXX lots more cases to add */
+   /* Handle other cases with a generic function.
+    */
    default:
-      debug_print_format("gallium: unhandled format in util_unpack_color_ub()",
-                         format);
-      assert(0);
+      {
+         ubyte dst[4];
+
+         util_format_read_4ub(format, dst, 0, uc, 0, 0, 0, 1, 1);
+         *r = dst[0];
+         *g = dst[1];
+         *b = dst[2];
+         *a = dst[3];
+      }
    }
 }
- 
 
 
 /**
  * Note rgba outside [0,1] will be clamped for int pixel formats.
  */
 static INLINE void
-util_pack_color(const float rgba[4], enum pipe_format format, void *dest)
+util_pack_color(const float rgba[4], enum pipe_format format, union util_color *uc)
 {
-   ubyte r, g, b, a;
+   ubyte r = 0;
+   ubyte g = 0;
+   ubyte b = 0;
+   ubyte a = 0;
 
-   if (pf_size_x(format) <= 8) {
+   if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) <= 8) {
       /* format uses 8-bit components or less */
       r = float_to_ubyte(rgba[0]);
       g = float_to_ubyte(rgba[1]);
@@ -315,92 +328,80 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest)
    switch (format) {
    case PIPE_FORMAT_R8G8B8A8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (r << 24) | (g << 16) | (b << 8) | a;
+         uc->ui = (r << 24) | (g << 16) | (b << 8) | a;
       }
       return;
    case PIPE_FORMAT_R8G8B8X8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (r << 24) | (g << 16) | (b << 8) | 0xff;
+         uc->ui = (r << 24) | (g << 16) | (b << 8) | 0xff;
       }
       return;
    case PIPE_FORMAT_A8R8G8B8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (a << 24) | (r << 16) | (g << 8) | b;
+         uc->ui = (a << 24) | (r << 16) | (g << 8) | b;
       }
       return;
    case PIPE_FORMAT_X8R8G8B8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (0xff << 24) | (r << 16) | (g << 8) | b;
+         uc->ui = (0xff << 24) | (r << 16) | (g << 8) | b;
       }
       return;
    case PIPE_FORMAT_B8G8R8A8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (b << 24) | (g << 16) | (r << 8) | a;
+         uc->ui = (b << 24) | (g << 16) | (r << 8) | a;
       }
       return;
    case PIPE_FORMAT_B8G8R8X8_UNORM:
       {
-         uint *d = (uint *) dest;
-         *d = (b << 24) | (g << 16) | (r << 8) | 0xff;
+         uc->ui = (b << 24) | (g << 16) | (r << 8) | 0xff;
       }
       return;
    case PIPE_FORMAT_R5G6B5_UNORM:
       {
-         ushort *d = (ushort *) dest;
-         *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3);
+         uc->us = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3);
       }
       return;
    case PIPE_FORMAT_A1R5G5B5_UNORM:
       {
-         ushort *d = (ushort *) dest;
-         *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3);
+         uc->us = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3);
       }
       return;
    case PIPE_FORMAT_A4R4G4B4_UNORM:
       {
-         ushort *d = (ushort *) dest;
-         *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
+         uc->ub = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
       }
       return;
    case PIPE_FORMAT_A8_UNORM:
       {
-         ubyte *d = (ubyte *) dest;
-         *d = a;
+         uc->ub = a;
       }
       return;
    case PIPE_FORMAT_L8_UNORM:
    case PIPE_FORMAT_I8_UNORM:
       {
-         ubyte *d = (ubyte *) dest;
-         *d = r;
+         uc->ub = r;
       }
       return;
    case PIPE_FORMAT_R32G32B32A32_FLOAT:
       {
-         float *d = (float *) dest;
-         d[0] = rgba[0];
-         d[1] = rgba[1];
-         d[2] = rgba[2];
-         d[3] = rgba[3];
+         uc->f[0] = rgba[0];
+         uc->f[1] = rgba[1];
+         uc->f[2] = rgba[2];
+         uc->f[3] = rgba[3];
       }
       return;
    case PIPE_FORMAT_R32G32B32_FLOAT:
       {
-         float *d = (float *) dest;
-         d[0] = rgba[0];
-         d[1] = rgba[1];
-         d[2] = rgba[2];
+         uc->f[0] = rgba[0];
+         uc->f[1] = rgba[1];
+         uc->f[2] = rgba[2];
       }
       return;
-   /* XXX lots more cases to add */
+
+   /* Handle other cases with a generic function.
+    */
    default:
-      debug_print_format("gallium: unhandled format in util_pack_color()", format);
-      assert(0);
+      util_format_write_4f(format, rgba, 0, uc, 0, 0, 0, 1, 1);
    }
 }
  
diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h
index a9b533eea7..10a874f341 100644
--- a/src/gallium/auxiliary/util/u_prim.h
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -135,4 +135,39 @@ static INLINE unsigned u_reduced_prim( unsigned pipe_prim )
    }
 }
 
+static INLINE unsigned
+u_vertices_per_prim(int primitive)
+{
+   switch(primitive) {
+   case PIPE_PRIM_POINTS:
+      return 1;
+   case PIPE_PRIM_LINES:
+   case PIPE_PRIM_LINE_LOOP:
+   case PIPE_PRIM_LINE_STRIP:
+      return 2;
+   case PIPE_PRIM_TRIANGLES:
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+      return 3;
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      return 4;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      return 6;
+
+   /* following primitives should never be used
+    * with geometry shaders abd their size is
+    * undefined */
+   case PIPE_PRIM_POLYGON:
+   case PIPE_PRIM_QUADS:
+   case PIPE_PRIM_QUAD_STRIP:
+   default:
+      debug_printf("Unrecognized geometry shader primitive");
+      return 3;
+   }
+}
+
+const char *u_prim_name( unsigned pipe_prim );
+
 #endif
diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c
index 9866b6fc8a..8479161c74 100644
--- a/src/gallium/auxiliary/util/u_rect.c
+++ b/src/gallium/auxiliary/util/u_rect.c
@@ -34,17 +34,18 @@
 #include "pipe/p_format.h"
 #include "pipe/p_context.h"
 #include "pipe/p_screen.h"
+#include "util/u_format.h"
 #include "util/u_rect.h"
 
 
 /**
  * Copy 2D rect from one place to another.
  * Position and sizes are in pixels.
- * src_pitch may be negative to do vertical flip of pixels from source.
+ * src_stride may be negative to do vertical flip of pixels from source.
  */
 void
 util_copy_rect(ubyte * dst,
-               const struct pipe_format_block *block,
+               enum pipe_format format,
                unsigned dst_stride,
                unsigned dst_x,
                unsigned dst_y,
@@ -53,31 +54,30 @@ util_copy_rect(ubyte * dst,
                const ubyte * src,
                int src_stride,
                unsigned src_x, 
-               int src_y)
+               unsigned src_y)
 {
    unsigned i;
    int src_stride_pos = src_stride < 0 ? -src_stride : src_stride;
-
-   assert(block->size > 0);
-   assert(block->width > 0);
-   assert(block->height > 0);
-   assert(src_x >= 0);
-   assert(src_y >= 0);
-   assert(dst_x >= 0);
-   assert(dst_y >= 0);
-
-   dst_x /= block->width;
-   dst_y /= block->height;
-   width = (width + block->width - 1)/block->width;
-   height = (height + block->height - 1)/block->height;
-   src_x /= block->width;
-   src_y /= block->height;
+   int blocksize = util_format_get_blocksize(format);
+   int blockwidth = util_format_get_blockwidth(format);
+   int blockheight = util_format_get_blockheight(format);
+
+   assert(blocksize > 0);
+   assert(blockwidth > 0);
+   assert(blockheight > 0);
+
+   dst_x /= blockwidth;
+   dst_y /= blockheight;
+   width = (width + blockwidth - 1)/blockwidth;
+   height = (height + blockheight - 1)/blockheight;
+   src_x /= blockwidth;
+   src_y /= blockheight;
    
-   dst += dst_x * block->size;
-   src += src_x * block->size;
+   dst += dst_x * blocksize;
+   src += src_x * blocksize;
    dst += dst_y * dst_stride;
    src += src_y * src_stride_pos;
-   width *= block->size;
+   width *= blocksize;
 
    if (width == dst_stride && width == src_stride)
       memcpy(dst, src, height * width);
@@ -92,7 +92,7 @@ util_copy_rect(ubyte * dst,
 
 void
 util_fill_rect(ubyte * dst,
-               const struct pipe_format_block *block,
+               enum pipe_format format,
                unsigned dst_stride,
                unsigned dst_x,
                unsigned dst_y,
@@ -102,23 +102,24 @@ util_fill_rect(ubyte * dst,
 {
    unsigned i, j;
    unsigned width_size;
-
-   assert(block->size > 0);
-   assert(block->width > 0);
-   assert(block->height > 0);
-   assert(dst_x >= 0);
-   assert(dst_y >= 0);
-
-   dst_x /= block->width;
-   dst_y /= block->height;
-   width = (width + block->width - 1)/block->width;
-   height = (height + block->height - 1)/block->height;
+   int blocksize = util_format_get_blocksize(format);
+   int blockwidth = util_format_get_blockwidth(format);
+   int blockheight = util_format_get_blockheight(format);
+
+   assert(blocksize > 0);
+   assert(blockwidth > 0);
+   assert(blockheight > 0);
+
+   dst_x /= blockwidth;
+   dst_y /= blockheight;
+   width = (width + blockwidth - 1)/blockwidth;
+   height = (height + blockheight - 1)/blockheight;
    
-   dst += dst_x * block->size;
+   dst += dst_x * blocksize;
    dst += dst_y * dst_stride;
-   width_size = width * block->size;
+   width_size = width * blocksize;
    
-   switch (block->size) {
+   switch (blocksize) {
    case 1:
       if(dst_stride == width_size)
 	 memset(dst, (ubyte) value, height * width_size);
@@ -172,10 +173,15 @@ util_surface_copy(struct pipe_context *pipe,
    struct pipe_transfer *src_trans, *dst_trans;
    void *dst_map;
    const void *src_map;
+   enum pipe_format src_format, dst_format;
 
    assert(src->texture && dst->texture);
    if (!src->texture || !dst->texture)
       return;
+
+   src_format = src->texture->format;
+   dst_format = dst->texture->format;
+
    src_trans = screen->get_tex_transfer(screen,
                                         src->texture,
                                         src->face,
@@ -192,9 +198,9 @@ util_surface_copy(struct pipe_context *pipe,
                                         PIPE_TRANSFER_WRITE,
                                         dst_x, dst_y, w, h);
 
-   assert(dst_trans->block.size == src_trans->block.size);
-   assert(dst_trans->block.width == src_trans->block.width);
-   assert(dst_trans->block.height == src_trans->block.height);
+   assert(util_format_get_blocksize(dst_format) == util_format_get_blocksize(src_format));
+   assert(util_format_get_blockwidth(dst_format) == util_format_get_blockwidth(src_format));
+   assert(util_format_get_blockheight(dst_format) == util_format_get_blockheight(src_format));
 
    src_map = pipe->screen->transfer_map(screen, src_trans);
    dst_map = pipe->screen->transfer_map(screen, dst_trans);
@@ -205,7 +211,7 @@ util_surface_copy(struct pipe_context *pipe,
    if (src_map && dst_map) {
       /* If do_flip, invert src_y position and pass negative src stride */
       util_copy_rect(dst_map,
-                     &dst_trans->block,
+                     dst_format,
                      dst_trans->stride,
                      0, 0,
                      w, h,
@@ -259,11 +265,11 @@ util_surface_fill(struct pipe_context *pipe,
    if (dst_map) {
       assert(dst_trans->stride > 0);
 
-      switch (dst_trans->block.size) {
+      switch (util_format_get_blocksize(dst_trans->texture->format)) {
       case 1:
       case 2:
       case 4:
-         util_fill_rect(dst_map, &dst_trans->block, dst_trans->stride,
+         util_fill_rect(dst_map, dst_trans->texture->format, dst_trans->stride,
                         0, 0, width, height, value);
          break;
       case 8:
diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h
index daa50834d3..b44d821904 100644
--- a/src/gallium/auxiliary/util/u_rect.h
+++ b/src/gallium/auxiliary/util/u_rect.h
@@ -42,13 +42,13 @@ struct pipe_surface;
 
 
 extern void
-util_copy_rect(ubyte * dst, const struct pipe_format_block *block,
+util_copy_rect(ubyte * dst, enum pipe_format format,
                unsigned dst_stride, unsigned dst_x, unsigned dst_y,
                unsigned width, unsigned height, const ubyte * src,
-               int src_stride, unsigned src_x, int src_y);
+               int src_stride, unsigned src_x, unsigned src_y);
 
 extern void
-util_fill_rect(ubyte * dst, const struct pipe_format_block *block,
+util_fill_rect(ubyte * dst, enum pipe_format format,
                unsigned dst_stride, unsigned dst_x, unsigned dst_y,
                unsigned width, unsigned height, uint32_t value);
 
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c
index 1c8b157d91..b751e29ab6 100644
--- a/src/gallium/auxiliary/util/u_simple_shaders.c
+++ b/src/gallium/auxiliary/util/u_simple_shaders.c
@@ -2,6 +2,7 @@
  *
  * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
+ * Copyright 2009 Marek Olšák <maraeo@gmail.com>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the
@@ -30,6 +31,7 @@
  * Simple vertex/fragment shader generators.
  *  
  * @author Brian Paul
+           Marek Olšák
  */
 
 
@@ -42,13 +44,15 @@
 
 /**
  * Make simple vertex pass-through shader.
+ * \param num_attribs  number of attributes to pass through
+ * \param semantic_names  array of semantic names for each attribute
+ * \param semantic_indexes  array of semantic indexes for each attribute
  */
 void *
 util_make_vertex_passthrough_shader(struct pipe_context *pipe,
                                     uint num_attribs,
                                     const uint *semantic_names,
                                     const uint *semantic_indexes)
-                                    
 {
    struct ureg_program *ureg;
    uint i;
@@ -76,8 +80,6 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
 }
 
 
-
-
 /**
  * Make simple fragment texture shader:
  *  IMM {0,0,0,1}                         // (if writemask != 0xf)
@@ -87,6 +89,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
  */
 void *
 util_make_fragment_tex_shader_writemask(struct pipe_context *pipe,
+                                        unsigned tex_target,
                                         unsigned writemask )
 {
    struct ureg_program *ureg;
@@ -116,20 +119,71 @@ util_make_fragment_tex_shader_writemask(struct pipe_context *pipe,
 
    ureg_TEX( ureg, 
              ureg_writemask(out, writemask),
-             TGSI_TEXTURE_2D, tex, sampler );
+             tex_target, tex, sampler );
    ureg_END( ureg );
 
    return ureg_create_shader_and_destroy( ureg, pipe );
 }
 
+
+/**
+ * Make a simple fragment shader that sets the output color to a color
+ * taken from a texture.
+ * \param tex_target  one of PIPE_TEXTURE_x
+ */
 void *
-util_make_fragment_tex_shader(struct pipe_context *pipe )
+util_make_fragment_tex_shader(struct pipe_context *pipe, unsigned tex_target )
 {
    return util_make_fragment_tex_shader_writemask( pipe,
+                                                   tex_target,
                                                    TGSI_WRITEMASK_XYZW );
 }
 
 
+/**
+ * Make a simple fragment texture shader which reads an X component from
+ * a texture and writes it as depth.
+ */
+void *
+util_make_fragment_tex_shader_writedepth(struct pipe_context *pipe,
+                                         unsigned tex_target)
+{
+   struct ureg_program *ureg;
+   struct ureg_src sampler;
+   struct ureg_src tex;
+   struct ureg_dst out, depth;
+   struct ureg_src imm;
+
+   ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
+   if (ureg == NULL)
+      return NULL;
+
+   sampler = ureg_DECL_sampler( ureg, 0 );
+
+   tex = ureg_DECL_fs_input( ureg,
+                             TGSI_SEMANTIC_GENERIC, 0,
+                             TGSI_INTERPOLATE_PERSPECTIVE );
+
+   out = ureg_DECL_output( ureg,
+                           TGSI_SEMANTIC_COLOR,
+                           0 );
+
+   depth = ureg_DECL_output( ureg,
+                             TGSI_SEMANTIC_POSITION,
+                             0 );
+
+   imm = ureg_imm4f( ureg, 0, 0, 0, 1 );
+
+   ureg_MOV( ureg, out, imm );
+
+   ureg_TEX( ureg,
+             ureg_writemask(depth, TGSI_WRITEMASK_Z),
+             tex_target, tex, sampler );
+   ureg_END( ureg );
+
+   return ureg_create_shader_and_destroy( ureg, pipe );
+}
+
 
 /**
  * Make simple fragment color pass-through shader.
@@ -137,9 +191,22 @@ util_make_fragment_tex_shader(struct pipe_context *pipe )
 void *
 util_make_fragment_passthrough_shader(struct pipe_context *pipe)
 {
+   return util_make_fragment_clonecolor_shader(pipe, 1);
+}
+
+
+/**
+ * Make a fragment shader that copies the input color to N output colors.
+ */
+void *
+util_make_fragment_clonecolor_shader(struct pipe_context *pipe, int num_cbufs)
+{
    struct ureg_program *ureg;
    struct ureg_src src;
-   struct ureg_dst dst;
+   struct ureg_dst dst[PIPE_MAX_COLOR_BUFS];
+   int i;
+
+   assert(num_cbufs <= PIPE_MAX_COLOR_BUFS);
 
    ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT );
    if (ureg == NULL)
@@ -148,12 +215,13 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe)
    src = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_COLOR, 0, 
                              TGSI_INTERPOLATE_PERSPECTIVE );
 
-   dst = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 );
+   for (i = 0; i < num_cbufs; i++)
+      dst[i] = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, i );
+
+   for (i = 0; i < num_cbufs; i++)
+      ureg_MOV( ureg, dst[i], src );
 
-   ureg_MOV( ureg, dst, src );
    ureg_END( ureg );
 
    return ureg_create_shader_and_destroy( ureg, pipe );
 }
-
-
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h
index d2e80d6eb4..6e760942e2 100644
--- a/src/gallium/auxiliary/util/u_simple_shaders.h
+++ b/src/gallium/auxiliary/util/u_simple_shaders.h
@@ -51,16 +51,25 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
 
 extern void *
 util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, 
-                                        unsigned writemask );
+                                        unsigned tex_target,
+                                        unsigned writemask);
 
 extern void *
-util_make_fragment_tex_shader(struct pipe_context *pipe);
+util_make_fragment_tex_shader(struct pipe_context *pipe, unsigned tex_target);
+
+
+extern void *
+util_make_fragment_tex_shader_writedepth(struct pipe_context *pipe,
+                                         unsigned tex_target);
 
 
 extern void *
 util_make_fragment_passthrough_shader(struct pipe_context *pipe);
 
 
+extern void *
+util_make_fragment_clonecolor_shader(struct pipe_context *pipe, int num_cbufs);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/auxiliary/util/u_stream_stdc.c b/src/gallium/auxiliary/util/u_stream_stdc.c
index 5cd05b2904..4d976d6dca 100644
--- a/src/gallium/auxiliary/util/u_stream_stdc.c
+++ b/src/gallium/auxiliary/util/u_stream_stdc.c
@@ -32,7 +32,7 @@
 
 #include "pipe/p_config.h"
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE)
 
 #include <stdio.h>
 
diff --git a/src/gallium/auxiliary/util/u_surface.c b/src/gallium/auxiliary/util/u_surface.c
index cfdf7ab8f8..f66376ad75 100644
--- a/src/gallium/auxiliary/util/u_surface.c
+++ b/src/gallium/auxiliary/util/u_surface.c
@@ -37,6 +37,7 @@
 #include "pipe/p_defines.h"
 
 #include "util/u_memory.h"
+#include "util/u_format.h"
 #include "util/u_surface.h"
 
 
@@ -80,10 +81,9 @@ util_create_rgba_surface(struct pipe_screen *screen,
    templ.target = target;
    templ.format = format;
    templ.last_level = 0;
-   templ.width[0] = width;
-   templ.height[0] = height;
-   templ.depth[0] = 1;
-   pf_get_block(format, &templ.block);
+   templ.width0 = width;
+   templ.height0 = height;
+   templ.depth0 = 1;
    templ.tex_usage = usage;
 
    *textureOut = screen->texture_create(screen, &templ);
diff --git a/src/gallium/auxiliary/util/u_texture.c b/src/gallium/auxiliary/util/u_texture.c
new file mode 100644
index 0000000000..cd477ab640
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_texture.c
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights reserved.
+ * Copyright 2009 Marek Olšák <maraeo@gmail.com>
+ *
+ * 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.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Texture mapping utility functions.
+ *
+ * @author Brian Paul
+ *         Marek Olšák
+ */
+
+#include "pipe/p_defines.h"
+
+#include "util/u_texture.h"
+
+void util_map_texcoords2d_onto_cubemap(unsigned face,
+                                       const float *in_st, unsigned in_stride,
+                                       float *out_str, unsigned out_stride)
+{
+   int i;
+   float rx, ry, rz;
+
+   /* loop over quad verts */
+   for (i = 0; i < 4; i++) {
+      /* Compute sc = +/-scale and tc = +/-scale.
+       * Not +/-1 to avoid cube face selection ambiguity near the edges,
+       * though that can still sometimes happen with this scale factor...
+       */
+      const float scale = 0.9999f;
+      const float sc = (2 * in_st[0] - 1) * scale;
+      const float tc = (2 * in_st[1] - 1) * scale;
+
+      switch (face) {
+         case PIPE_TEX_FACE_POS_X:
+            rx = 1;
+            ry = -tc;
+            rz = -sc;
+            break;
+         case PIPE_TEX_FACE_NEG_X:
+            rx = -1;
+            ry = -tc;
+            rz = sc;
+            break;
+         case PIPE_TEX_FACE_POS_Y:
+            rx = sc;
+            ry = 1;
+            rz = tc;
+            break;
+         case PIPE_TEX_FACE_NEG_Y:
+            rx = sc;
+            ry = -1;
+            rz = -tc;
+            break;
+         case PIPE_TEX_FACE_POS_Z:
+            rx = sc;
+            ry = -tc;
+            rz = 1;
+            break;
+         case PIPE_TEX_FACE_NEG_Z:
+            rx = -sc;
+            ry = -tc;
+            rz = -1;
+            break;
+         default:
+            rx = ry = rz = 0;
+            assert(0);
+      }
+
+      out_str[0] = rx; /*s*/
+      out_str[1] = ry; /*t*/
+      out_str[2] = rz; /*r*/
+
+      in_st += in_stride;
+      out_str += out_stride;
+   }
+}
diff --git a/src/gallium/auxiliary/util/u_texture.h b/src/gallium/auxiliary/util/u_texture.h
new file mode 100644
index 0000000000..93b2f1e4c9
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_texture.h
@@ -0,0 +1,54 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Marek Olšák <maraeo@gmail.com>
+ *
+ * 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 U_TEXTURE_H
+#define U_TEXTURE_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Convert 2D texture coordinates of 4 vertices into cubemap coordinates
+ * in the given face.
+ * Coordinates must be in the range [0,1].
+ *
+ * \param face          Cubemap face.
+ * \param in_st         4 pairs of 2D texture coordinates to convert.
+ * \param in_stride     Stride of in_st in floats.
+ * \param out_str       STR cubemap texture coordinates to compute.
+ * \param out_stride    Stride of out_str in floats.
+ */
+void util_map_texcoords2d_onto_cubemap(unsigned face,
+                                       const float *in_st, unsigned in_stride,
+                                       float *out_str, unsigned out_stride);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c
index 8a22f584be..1ba82bb21f 100644
--- a/src/gallium/auxiliary/util/u_tile.c
+++ b/src/gallium/auxiliary/util/u_tile.c
@@ -34,6 +34,7 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "util/u_rect.h"
@@ -52,7 +53,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt,
    const void *src;
 
    if (dst_stride == 0)
-      dst_stride = pf_get_nblocksx(&pt->block, w) * pt->block.size;
+      dst_stride = util_format_get_stride(pt->texture->format, w);
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
@@ -62,7 +63,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt,
    if(!src)
       return;
 
-   util_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y);
+   util_copy_rect(dst, pt->texture->format, dst_stride, 0, 0, w, h, src, pt->stride, x, y);
 
    screen->transfer_unmap(screen, pt);
 }
@@ -78,9 +79,10 @@ pipe_put_tile_raw(struct pipe_transfer *pt,
 {
    struct pipe_screen *screen = pt->texture->screen;
    void *dst;
+   enum pipe_format format = pt->texture->format;
 
    if (src_stride == 0)
-      src_stride = pf_get_nblocksx(&pt->block, w) * pt->block.size;
+      src_stride = util_format_get_stride(format, w);
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
@@ -90,7 +92,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt,
    if(!dst)
       return;
 
-   util_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0);
+   util_copy_rect(dst, format, pt->stride, x, y, w, h, src, src_stride, 0, 0);
 
    screen->transfer_unmap(screen, pt);
 }
@@ -246,6 +248,53 @@ b8g8r8a8_put_tile_rgba(unsigned *dst,
 }
 
 
+/*** PIPE_FORMAT_R8G8B8A8_UNORM ***/
+
+static void
+r8g8b8a8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       float *p,
+                       unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const unsigned pixel = *src++;
+         pRow[0] = ubyte_to_float((pixel >> 24) & 0xff);
+         pRow[1] = ubyte_to_float((pixel >> 16) & 0xff);
+         pRow[2] = ubyte_to_float((pixel >>  8) & 0xff);
+         pRow[3] = ubyte_to_float((pixel >>  0) & 0xff);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+r8g8b8a8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const float *p,
+                       unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, g, b, a;
+         r = float_to_ubyte(pRow[0]);
+         g = float_to_ubyte(pRow[1]);
+         b = float_to_ubyte(pRow[2]);
+         a = float_to_ubyte(pRow[3]);
+         *dst++ = (r << 24) | (g << 16) | (b << 8) | a;
+      }
+      p += src_stride;
+   }
+}
+
+
 /*** PIPE_FORMAT_A1R5G5B5_UNORM ***/
 
 static void
@@ -1106,27 +1155,6 @@ ycbcr_get_tile_rgba(const ushort *src,
 }
 
 
-static void
-fake_get_tile_rgba(const ushort *src,
-                   unsigned w, unsigned h,
-                   float *p,
-                   unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         pRow[0] =
-         pRow[1] =
-         pRow[2] =
-         pRow[3] = (i ^ j) & 1 ? 1.0f : 0.0f;
-      }
-      p += dst_stride;
-   }
-}
-
-
 void
 pipe_tile_raw_to_rgba(enum pipe_format format,
                       void *src,
@@ -1143,6 +1171,9 @@ pipe_tile_raw_to_rgba(enum pipe_format format,
    case PIPE_FORMAT_B8G8R8A8_UNORM:
       b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
       break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      r8g8b8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
    case PIPE_FORMAT_A1R5G5B5_UNORM:
       a1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
       break;
@@ -1206,8 +1237,10 @@ pipe_tile_raw_to_rgba(enum pipe_format format,
       ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE);
       break;
    default:
-      debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(format));
-      fake_get_tile_rgba(src, w, h, dst, dst_stride);
+      util_format_read_4f(format,
+                          dst, dst_stride * sizeof(float),
+                          src, util_format_get_stride(format, w),
+                          0, 0, w, h);
    }
 }
 
@@ -1219,21 +1252,22 @@ pipe_get_tile_rgba(struct pipe_transfer *pt,
 {
    unsigned dst_stride = w * 4;
    void *packed;
+   enum pipe_format format = pt->texture->format;
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
 
-   packed = MALLOC(pf_get_nblocks(&pt->block, w, h) * pt->block.size);
+   packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format));
 
    if (!packed)
       return;
 
-   if(pt->format == PIPE_FORMAT_YCBCR || pt->format == PIPE_FORMAT_YCBCR_REV)
+   if(format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV)
       assert((x & 1) == 0);
 
    pipe_get_tile_raw(pt, x, y, w, h, packed, 0);
 
-   pipe_tile_raw_to_rgba(pt->format, packed, w, h, p, dst_stride);
+   pipe_tile_raw_to_rgba(format, packed, w, h, p, dst_stride);
 
    FREE(packed);
 }
@@ -1246,16 +1280,17 @@ pipe_put_tile_rgba(struct pipe_transfer *pt,
 {
    unsigned src_stride = w * 4;
    void *packed;
+   enum pipe_format format = pt->texture->format;
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
 
-   packed = MALLOC(pf_get_nblocks(&pt->block, w, h) * pt->block.size);
+   packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format));
 
    if (!packed)
       return;
 
-   switch (pt->format) {
+   switch (format) {
    case PIPE_FORMAT_A8R8G8B8_UNORM:
       a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
       break;
@@ -1265,6 +1300,9 @@ pipe_put_tile_rgba(struct pipe_transfer *pt,
    case PIPE_FORMAT_B8G8R8A8_UNORM:
       b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
       break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      r8g8b8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
+      break;
    case PIPE_FORMAT_A1R5G5B5_UNORM:
       a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
       break;
@@ -1274,9 +1312,6 @@ pipe_put_tile_rgba(struct pipe_transfer *pt,
    case PIPE_FORMAT_R8G8B8_UNORM:
       r8g8b8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
       break;
-   case PIPE_FORMAT_R8G8B8A8_UNORM:
-      assert(0);
-      break;
    case PIPE_FORMAT_A4R4G4B4_UNORM:
       a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
       break;
@@ -1322,7 +1357,7 @@ pipe_put_tile_rgba(struct pipe_transfer *pt,
       /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
       break;
    default:
-      debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(pt->format));
+      debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(format));
    }
 
    pipe_put_tile_raw(pt, x, y, w, h, packed, 0);
@@ -1344,6 +1379,7 @@ pipe_get_tile_z(struct pipe_transfer *pt,
    ubyte *map;
    uint *pDest = z;
    uint i, j;
+   enum pipe_format format = pt->texture->format;
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
@@ -1354,7 +1390,7 @@ pipe_get_tile_z(struct pipe_transfer *pt,
       return;
    }
 
-   switch (pt->format) {
+   switch (format) {
    case PIPE_FORMAT_Z32_UNORM:
       {
          const uint *ptrc
@@ -1428,6 +1464,7 @@ pipe_put_tile_z(struct pipe_transfer *pt,
    const uint *ptrc = zSrc;
    ubyte *map;
    uint i, j;
+   enum pipe_format format = pt->texture->format;
 
    if (pipe_clip_tile(x, y, &w, &h, pt))
       return;
@@ -1438,7 +1475,7 @@ pipe_put_tile_z(struct pipe_transfer *pt,
       return;
    }
 
-   switch (pt->format) {
+   switch (format) {
    case PIPE_FORMAT_Z32_UNORM:
       {
          uint *pDest = (uint *) (map + y * pt->stride + x*4);
diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c
index eb635c9f14..975ee89c45 100644
--- a/src/gallium/auxiliary/util/u_upload_mgr.c
+++ b/src/gallium/auxiliary/util/u_upload_mgr.c
@@ -29,7 +29,7 @@
  * coalescing small buffers into larger ones.
  */
 
-#include "pipe/p_error.h"
+#include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "pipe/p_screen.h"
 #include "util/u_memory.h"
diff --git a/src/gallium/auxiliary/util/u_upload_mgr.h b/src/gallium/auxiliary/util/u_upload_mgr.h
index 745b5834af..e158bed9d0 100644
--- a/src/gallium/auxiliary/util/u_upload_mgr.h
+++ b/src/gallium/auxiliary/util/u_upload_mgr.h
@@ -32,6 +32,8 @@
 #ifndef U_UPLOAD_MGR_H
 #define U_UPLOAD_MGR_H
 
+#include "pipe/p_defines.h"
+
 struct pipe_screen;
 struct pipe_buffer;
 struct u_upload_mgr;
diff --git a/src/gallium/auxiliary/vl/Makefile b/src/gallium/auxiliary/vl/Makefile
deleted file mode 100644
index 4314c1e8d6..0000000000
--- a/src/gallium/auxiliary/vl/Makefile
+++ /dev/null
@@ -1,13 +0,0 @@
-TOP = ../../../..
-include $(TOP)/configs/current
-
-LIBNAME = vl
-
-C_SOURCES = \
-	vl_bitstream_parser.c \
-	vl_mpeg12_mc_renderer.c \
-	vl_compositor.c \
-        vl_csc.c \
-	vl_shader_build.c
-
-include ../../Makefile.template
diff --git a/src/gallium/auxiliary/vl/SConscript b/src/gallium/auxiliary/vl/SConscript
deleted file mode 100644
index aed69f5efe..0000000000
--- a/src/gallium/auxiliary/vl/SConscript
+++ /dev/null
@@ -1,13 +0,0 @@
-Import('*')
-
-vl = env.ConvenienceLibrary(
-	target = 'vl',
-	source = [
-		'vl_bitstream_parser.c',
-		'vl_mpeg12_mc_renderer.c',
-		'vl_compositor.c',
-                'vl_csc.c',
-		'vl_shader_build.c',
-	])
-
-auxiliaries.insert(0, vl)
diff --git a/src/gallium/auxiliary/vl/vl_compositor.c b/src/gallium/auxiliary/vl/vl_compositor.c
index b36dbeb208..fc2a1c59a6 100644
--- a/src/gallium/auxiliary/vl/vl_compositor.c
+++ b/src/gallium/auxiliary/vl/vl_compositor.c
@@ -95,12 +95,11 @@ create_vert_shader(struct vl_compositor *c)
    assert(c);
 
    tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header*)&tokens[1];
+   header = (struct tgsi_header*)&tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+   *(struct tgsi_processor*)&tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0             ; Vertex pos
@@ -172,12 +171,11 @@ create_frag_shader(struct vl_compositor *c)
    assert(c);
 
    tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header*)&tokens[1];
+   header = (struct tgsi_header*)&tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+   *(struct tgsi_processor*)&tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
-   ti = 3;
+   ti = 2;
 
    /* decl i0             ; Texcoords for s0 */
    decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
@@ -213,7 +211,7 @@ create_frag_shader(struct vl_compositor *c)
     */
    for (i = 0; i < 4; ++i) {
       inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i);
-      inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+      inst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X << i;
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
    }
 
@@ -455,8 +453,8 @@ void vl_compositor_render(struct vl_compositor          *compositor,
    assert(dst_area);
    assert(picture_type == PIPE_MPEG12_PICTURE_TYPE_FRAME);
 
-   compositor->fb_state.width = dst_surface->width[0];
-   compositor->fb_state.height = dst_surface->height[0];
+   compositor->fb_state.width = dst_surface->width0;
+   compositor->fb_state.height = dst_surface->height0;
    compositor->fb_state.cbufs[0] = compositor->pipe->screen->get_tex_surface
    (
       compositor->pipe->screen,
@@ -473,10 +471,14 @@ void vl_compositor_render(struct vl_compositor          *compositor,
    compositor->viewport.translate[2] = 0;
    compositor->viewport.translate[3] = 0;
 
+   compositor->scissor.maxx = compositor->fb_state.width;
+   compositor->scissor.maxy = compositor->fb_state.height;
+
    compositor->pipe->set_framebuffer_state(compositor->pipe, &compositor->fb_state);
    compositor->pipe->set_viewport_state(compositor->pipe, &compositor->viewport);
-   compositor->pipe->bind_sampler_states(compositor->pipe, 1, &compositor->sampler);
-   compositor->pipe->set_sampler_textures(compositor->pipe, 1, &src_surface);
+   compositor->pipe->set_scissor_state(compositor->pipe, &compositor->scissor);
+   compositor->pipe->bind_fragment_sampler_states(compositor->pipe, 1, &compositor->sampler);
+   compositor->pipe->set_fragment_sampler_textures(compositor->pipe, 1, &src_surface);
    compositor->pipe->bind_vs_state(compositor->pipe, compositor->vertex_shader);
    compositor->pipe->bind_fs_state(compositor->pipe, compositor->fragment_shader);
    compositor->pipe->set_vertex_buffers(compositor->pipe, 2, compositor->vertex_bufs);
@@ -500,12 +502,12 @@ void vl_compositor_render(struct vl_compositor          *compositor,
    vs_consts->dst_trans.z = 0;
    vs_consts->dst_trans.w = 0;
 
-   vs_consts->src_scale.x = src_area->w / (float)src_surface->width[0];
-   vs_consts->src_scale.y = src_area->h / (float)src_surface->height[0];
+   vs_consts->src_scale.x = src_area->w / (float)src_surface->width0;
+   vs_consts->src_scale.y = src_area->h / (float)src_surface->height0;
    vs_consts->src_scale.z = 1;
    vs_consts->src_scale.w = 1;
-   vs_consts->src_trans.x = src_area->x / (float)src_surface->width[0];
-   vs_consts->src_trans.y = src_area->y / (float)src_surface->height[0];
+   vs_consts->src_trans.x = src_area->x / (float)src_surface->width0;
+   vs_consts->src_trans.y = src_area->y / (float)src_surface->height0;
    vs_consts->src_trans.z = 0;
    vs_consts->src_trans.w = 0;
 
diff --git a/src/gallium/auxiliary/vl/vl_compositor.h b/src/gallium/auxiliary/vl/vl_compositor.h
index 17e2afd353..f441901a75 100644
--- a/src/gallium/auxiliary/vl/vl_compositor.h
+++ b/src/gallium/auxiliary/vl/vl_compositor.h
@@ -44,6 +44,7 @@ struct vl_compositor
    void *vertex_shader;
    void *fragment_shader;
    struct pipe_viewport_state viewport;
+   struct pipe_scissor_state scissor;
    struct pipe_vertex_buffer vertex_bufs[2];
    struct pipe_vertex_element vertex_elems[2];
    struct pipe_constant_buffer vs_const_buf, fs_const_buf;
diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
index 6b3614821c..caf581aca6 100644
--- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
+++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
@@ -29,6 +29,7 @@
 #include <assert.h>
 #include <pipe/p_context.h>
 #include <pipe/p_inlines.h>
+#include <util/u_format.h>
 #include <util/u_math.h>
 #include <util/u_memory.h>
 #include <tgsi/tgsi_parse.h>
@@ -115,12 +116,11 @@ create_intra_vert_shader(struct vl_mpeg12_mc_renderer *r)
    assert(r);
 
    tokens = (struct tgsi_token *) malloc(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header *) &tokens[1];
+   header = (struct tgsi_header *) &tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor *) &tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+   *(struct tgsi_processor *) &tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0              ; Vertex pos
@@ -185,12 +185,11 @@ create_intra_frag_shader(struct vl_mpeg12_mc_renderer *r)
    assert(r);
 
    tokens = (struct tgsi_token *) malloc(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header *) &tokens[1];
+   header = (struct tgsi_header *) &tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor *) &tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+   *(struct tgsi_processor *) &tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0                      ; Luma texcoords
@@ -237,10 +236,10 @@ create_intra_frag_shader(struct vl_mpeg12_mc_renderer *r)
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
       inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-      inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+      inst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_X;
+      inst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_X;
+      inst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_X;
+      inst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X << i;
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
    }
 
@@ -278,12 +277,11 @@ create_frame_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
    assert(r);
 
    tokens = (struct tgsi_token *) malloc(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header *) &tokens[1];
+   header = (struct tgsi_header *) &tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor *) &tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+   *(struct tgsi_processor *) &tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0              ; Vertex pos
@@ -336,11 +334,13 @@ create_frame_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
    free(tokens);
 }
 
+#if 0
 static void
 create_field_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
 {
    assert(false);
 }
+#endif
 
 static void
 create_frame_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
@@ -361,12 +361,11 @@ create_frame_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
    assert(r);
 
    tokens = (struct tgsi_token *) malloc(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header *) &tokens[1];
+   header = (struct tgsi_header *) &tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor *) &tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+   *(struct tgsi_processor *) &tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0                      ; Luma texcoords
@@ -415,10 +414,10 @@ create_frame_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
       inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-      inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+      inst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_X;
+      inst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_X;
+      inst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_X;
+      inst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X << i;
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
    }
 
@@ -445,11 +444,13 @@ create_frame_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
    free(tokens);
 }
 
+#if 0
 static void
 create_field_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
 {
    assert(false);
 }
+#endif
 
 static void
 create_frame_bi_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
@@ -470,12 +471,11 @@ create_frame_bi_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
    assert(r);
 
    tokens = (struct tgsi_token *) malloc(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header *) &tokens[1];
+   header = (struct tgsi_header *) &tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor *) &tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+   *(struct tgsi_processor *) &tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0              ; Vertex pos
@@ -536,11 +536,13 @@ create_frame_bi_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
    free(tokens);
 }
 
+#if 0
 static void
 create_field_bi_pred_vert_shader(struct vl_mpeg12_mc_renderer *r)
 {
    assert(false);
 }
+#endif
 
 static void
 create_frame_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
@@ -561,12 +563,11 @@ create_frame_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
    assert(r);
 
    tokens = (struct tgsi_token *) malloc(max_tokens * sizeof(struct tgsi_token));
-   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
-   header = (struct tgsi_header *) &tokens[1];
+   header = (struct tgsi_header *) &tokens[0];
    *header = tgsi_build_header();
-   *(struct tgsi_processor *) &tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+   *(struct tgsi_processor *) &tokens[1] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
-   ti = 3;
+   ti = 2;
 
    /*
     * decl i0                      ; Luma texcoords
@@ -620,10 +621,10 @@ create_frame_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
       inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-      inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-      inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+      inst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_X;
+      inst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_X;
+      inst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_X;
+      inst.Dst[0].Register.WriteMask = TGSI_WRITEMASK_X << i;
       ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
    }
 
@@ -642,10 +643,10 @@ create_frame_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
 
    /* lerp t1, c1.x, t1, t2        ; Blend past and future texels */
    inst = vl_inst4(TGSI_OPCODE_LRP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-   inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-   inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-   inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-   inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+   inst.Src[0].Register.SwizzleX = TGSI_SWIZZLE_X;
+   inst.Src[0].Register.SwizzleY = TGSI_SWIZZLE_X;
+   inst.Src[0].Register.SwizzleZ = TGSI_SWIZZLE_X;
+   inst.Src[0].Register.SwizzleW = TGSI_SWIZZLE_X;
    ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
    /* add o0, t0, t1               ; Add past/future ref and differential to form final output */
@@ -663,11 +664,13 @@ create_frame_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
    free(tokens);
 }
 
+#if 0
 static void
 create_field_bi_pred_frag_shader(struct vl_mpeg12_mc_renderer *r)
 {
    assert(false);
 }
+#endif
 
 static void
 xfer_buffers_map(struct vl_mpeg12_mc_renderer *r)
@@ -681,7 +684,7 @@ xfer_buffers_map(struct vl_mpeg12_mc_renderer *r)
       (
          r->pipe->screen, r->textures.all[i],
          0, 0, 0, PIPE_TRANSFER_WRITE, 0, 0,
-         r->textures.all[i]->width[0], r->textures.all[i]->height[0]
+         r->textures.all[i]->width0, r->textures.all[i]->height0
       );
 
       r->texels[i] = r->pipe->screen->transfer_map(r->pipe->screen, r->tex_transfer[i]);
@@ -721,6 +724,11 @@ init_pipe_state(struct vl_mpeg12_mc_renderer *r)
    r->viewport.translate[2] = 0;
    r->viewport.translate[3] = 0;
 
+   r->scissor.maxx = r->pot_buffers ?
+      util_next_power_of_two(r->picture_width) : r->picture_width;
+   r->scissor.maxy = r->pot_buffers ?
+      util_next_power_of_two(r->picture_height) : r->picture_height;
+
    r->fb_state.width = r->pot_buffers ?
       util_next_power_of_two(r->picture_width) : r->picture_width;
    r->fb_state.height = r->pot_buffers ?
@@ -830,26 +838,25 @@ init_buffers(struct vl_mpeg12_mc_renderer *r)
    /* TODO: Accomodate HW that can't do this and also for cases when this isn't precise enough */
    template.format = PIPE_FORMAT_R16_SNORM;
    template.last_level = 0;
-   template.width[0] = r->pot_buffers ?
+   template.width0 = r->pot_buffers ?
       util_next_power_of_two(r->picture_width) : r->picture_width;
-   template.height[0] = r->pot_buffers ?
+   template.height0 = r->pot_buffers ?
       util_next_power_of_two(r->picture_height) : r->picture_height;
-   template.depth[0] = 1;
-   pf_get_block(template.format, &template.block);
+   template.depth0 = 1;
    template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_DYNAMIC;
 
    r->textures.individual.y = r->pipe->screen->texture_create(r->pipe->screen, &template);
 
    if (r->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
-      template.width[0] = r->pot_buffers ?
+      template.width0 = r->pot_buffers ?
          util_next_power_of_two(r->picture_width / 2) :
          r->picture_width / 2;
-      template.height[0] = r->pot_buffers ?
+      template.height0 = r->pot_buffers ?
          util_next_power_of_two(r->picture_height / 2) :
          r->picture_height / 2;
    }
    else if (r->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422)
-      template.height[0] = r->pot_buffers ?
+      template.height0 = r->pot_buffers ?
          util_next_power_of_two(r->picture_height / 2) :
          r->picture_height / 2;
 
@@ -1082,6 +1089,9 @@ gen_macroblock_verts(struct vl_mpeg12_mc_renderer *r,
    assert(ycbcr_vb);
    assert(pos < r->macroblocks_per_batch);
 
+   mo_vec[1].x = 0;
+   mo_vec[1].y = 0;
+
    switch (mb->mb_type) {
       case PIPE_MPEG12_MACROBLOCK_TYPE_BI:
       {
@@ -1270,6 +1280,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
 
    r->pipe->set_framebuffer_state(r->pipe, &r->fb_state);
    r->pipe->set_viewport_state(r->pipe, &r->viewport);
+   r->pipe->set_scissor_state(r->pipe, &r->scissor);
 
    vs_consts = pipe_buffer_map
    (
@@ -1277,8 +1288,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
    );
 
-   vs_consts->denorm.x = r->surface->width[0];
-   vs_consts->denorm.y = r->surface->height[0];
+   vs_consts->denorm.x = r->surface->width0;
+   vs_consts->denorm.y = r->surface->height0;
 
    pipe_buffer_unmap(r->pipe->screen, r->vs_const_buf.buffer);
 
@@ -1290,8 +1301,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
    if (num_macroblocks[MACROBLOCK_TYPE_INTRA] > 0) {
       r->pipe->set_vertex_buffers(r->pipe, 1, r->vertex_bufs.all);
       r->pipe->set_vertex_elements(r->pipe, 4, r->vertex_elems);
-      r->pipe->set_sampler_textures(r->pipe, 3, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 3, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 3, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 3, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->i_vs);
       r->pipe->bind_fs_state(r->pipe, r->i_fs);
 
@@ -1304,8 +1315,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
       r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems);
       r->textures.individual.ref[0] = r->past;
-      r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->p_vs[0]);
       r->pipe->bind_fs_state(r->pipe, r->p_fs[0]);
 
@@ -1318,8 +1329,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
       r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems);
       r->textures.individual.ref[0] = r->past;
-      r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->p_vs[1]);
       r->pipe->bind_fs_state(r->pipe, r->p_fs[1]);
 
@@ -1332,8 +1343,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
       r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems);
       r->textures.individual.ref[0] = r->future;
-      r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->p_vs[0]);
       r->pipe->bind_fs_state(r->pipe, r->p_fs[0]);
 
@@ -1346,8 +1357,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
       r->pipe->set_vertex_elements(r->pipe, 6, r->vertex_elems);
       r->textures.individual.ref[0] = r->future;
-      r->pipe->set_sampler_textures(r->pipe, 4, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 4, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 4, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 4, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->p_vs[1]);
       r->pipe->bind_fs_state(r->pipe, r->p_fs[1]);
 
@@ -1361,8 +1372,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       r->pipe->set_vertex_elements(r->pipe, 8, r->vertex_elems);
       r->textures.individual.ref[0] = r->past;
       r->textures.individual.ref[1] = r->future;
-      r->pipe->set_sampler_textures(r->pipe, 5, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 5, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 5, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 5, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->b_vs[0]);
       r->pipe->bind_fs_state(r->pipe, r->b_fs[0]);
 
@@ -1376,8 +1387,8 @@ flush(struct vl_mpeg12_mc_renderer *r)
       r->pipe->set_vertex_elements(r->pipe, 8, r->vertex_elems);
       r->textures.individual.ref[0] = r->past;
       r->textures.individual.ref[1] = r->future;
-      r->pipe->set_sampler_textures(r->pipe, 5, r->textures.all);
-      r->pipe->bind_sampler_states(r->pipe, 5, r->samplers.all);
+      r->pipe->set_fragment_sampler_textures(r->pipe, 5, r->textures.all);
+      r->pipe->bind_fragment_sampler_states(r->pipe, 5, r->samplers.all);
       r->pipe->bind_vs_state(r->pipe, r->b_vs[1]);
       r->pipe->bind_fs_state(r->pipe, r->b_fs[1]);
 
@@ -1444,7 +1455,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby,
    assert(r);
    assert(blocks);
 
-   tex_pitch = r->tex_transfer[0]->stride / r->tex_transfer[0]->block.size;
+   tex_pitch = r->tex_transfer[0]->stride / util_format_get_blocksize(r->tex_transfer[0]->texture->format);
    texels = r->texels[0] + mbpy * tex_pitch + mbpx;
 
    for (y = 0; y < 2; ++y) {
@@ -1483,7 +1494,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby,
    mbpy /= 2;
 
    for (tb = 0; tb < 2; ++tb) {
-      tex_pitch = r->tex_transfer[tb + 1]->stride / r->tex_transfer[tb + 1]->block.size;
+      tex_pitch = r->tex_transfer[tb + 1]->stride / util_format_get_blocksize(r->tex_transfer[tb + 1]->texture->format);
       texels = r->texels[tb + 1] + mbpy * tex_pitch + mbpx;
 
       if ((cbp >> (1 - tb)) & 1) {
@@ -1627,8 +1638,8 @@ vl_mpeg12_mc_renderer_render_macroblocks(struct vl_mpeg12_mc_renderer
       renderer->past = past;
       renderer->future = future;
       renderer->fence = fence;
-      renderer->surface_tex_inv_size.x = 1.0f / surface->width[0];
-      renderer->surface_tex_inv_size.y = 1.0f / surface->height[0];
+      renderer->surface_tex_inv_size.x = 1.0f / surface->width0;
+      renderer->surface_tex_inv_size.y = 1.0f / surface->height0;
    }
 
    while (num_macroblocks) {
diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h
index 5d2c1273ee..64184337a0 100644
--- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h
+++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h
@@ -62,6 +62,7 @@ struct vl_mpeg12_mc_renderer
    unsigned macroblocks_per_batch;
 
    struct pipe_viewport_state viewport;
+   struct pipe_scissor_state scissor;
    struct pipe_constant_buffer vs_const_buf;
    struct pipe_constant_buffer fs_const_buf;
    struct pipe_framebuffer_state fb_state;
diff --git a/src/gallium/auxiliary/vl/vl_shader_build.c b/src/gallium/auxiliary/vl/vl_shader_build.c
index faa20a903c..d011ef97bd 100644
--- a/src/gallium/auxiliary/vl/vl_shader_build.c
+++ b/src/gallium/auxiliary/vl/vl_shader_build.c
@@ -36,10 +36,10 @@ struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index
 
    decl.Declaration.File = TGSI_FILE_INPUT;
    decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = name;
-   decl.Semantic.SemanticIndex = index;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Semantic.Name = name;
+   decl.Semantic.Index = index;
+   decl.Range.First = first;
+   decl.Range.Last = last;
 
    return decl;
 }
@@ -64,11 +64,11 @@ struct tgsi_full_declaration vl_decl_interpolated_input
 
    decl.Declaration.File = TGSI_FILE_INPUT;
    decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = name;
-   decl.Semantic.SemanticIndex = index;
+   decl.Semantic.Name = name;
+   decl.Semantic.Index = index;
    decl.Declaration.Interpolate = interpolation;;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Range.First = first;
+   decl.Range.Last = last;
 
    return decl;
 }
@@ -79,10 +79,10 @@ struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int i
 
    decl.Declaration.File = TGSI_FILE_CONSTANT;
    decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = name;
-   decl.Semantic.SemanticIndex = index;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Semantic.Name = name;
+   decl.Semantic.Index = index;
+   decl.Range.First = first;
+   decl.Range.Last = last;
 
    return decl;
 }
@@ -93,10 +93,10 @@ struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int inde
 
    decl.Declaration.File = TGSI_FILE_OUTPUT;
    decl.Declaration.Semantic = 1;
-   decl.Semantic.SemanticName = name;
-   decl.Semantic.SemanticIndex = index;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Semantic.Name = name;
+   decl.Semantic.Index = index;
+   decl.Range.First = first;
+   decl.Range.Last = last;
 
    return decl;
 }
@@ -107,8 +107,8 @@ struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last
 
    decl = tgsi_default_full_declaration();
    decl.Declaration.File = TGSI_FILE_TEMPORARY;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Range.First = first;
+   decl.Range.Last = last;
 
    return decl;
 }
@@ -119,8 +119,8 @@ struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int l
 
    decl = tgsi_default_full_declaration();
    decl.Declaration.File = TGSI_FILE_SAMPLER;
-   decl.DeclarationRange.First = first;
-   decl.DeclarationRange.Last = last;
+   decl.Range.First = first;
+   decl.Range.Last = last;
 
    return decl;
 }
@@ -138,11 +138,11 @@ struct tgsi_full_instruction vl_inst2
 
    inst.Instruction.Opcode = opcode;
    inst.Instruction.NumDstRegs = 1;
-   inst.FullDstRegisters[0].DstRegister.File = dst_file;
-   inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+   inst.Dst[0].Register.File = dst_file;
+   inst.Dst[0].Register.Index = dst_index;
    inst.Instruction.NumSrcRegs = 1;
-   inst.FullSrcRegisters[0].SrcRegister.File = src_file;
-   inst.FullSrcRegisters[0].SrcRegister.Index = src_index;
+   inst.Src[0].Register.File = src_file;
+   inst.Src[0].Register.Index = src_index;
 
    return inst;
 }
@@ -162,13 +162,13 @@ struct tgsi_full_instruction vl_inst3
 
    inst.Instruction.Opcode = opcode;
    inst.Instruction.NumDstRegs = 1;
-   inst.FullDstRegisters[0].DstRegister.File = dst_file;
-   inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+   inst.Dst[0].Register.File = dst_file;
+   inst.Dst[0].Register.Index = dst_index;
    inst.Instruction.NumSrcRegs = 2;
-   inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
-   inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
-   inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
-   inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+   inst.Src[0].Register.File = src1_file;
+   inst.Src[0].Register.Index = src1_index;
+   inst.Src[1].Register.File = src2_file;
+   inst.Src[1].Register.Index = src2_index;
 
    return inst;
 }
@@ -188,14 +188,15 @@ struct tgsi_full_instruction vl_tex
 
    inst.Instruction.Opcode = TGSI_OPCODE_TEX;
    inst.Instruction.NumDstRegs = 1;
-   inst.FullDstRegisters[0].DstRegister.File = dst_file;
-   inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+   inst.Dst[0].Register.File = dst_file;
+   inst.Dst[0].Register.Index = dst_index;
    inst.Instruction.NumSrcRegs = 2;
-   inst.InstructionExtTexture.Texture = tex;
-   inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
-   inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
-   inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
-   inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+   inst.Instruction.Texture = 1;
+   inst.Texture.Texture = tex;
+   inst.Src[0].Register.File = src1_file;
+   inst.Src[0].Register.Index = src1_index;
+   inst.Src[1].Register.File = src2_file;
+   inst.Src[1].Register.Index = src2_index;
 
    return inst;
 }
@@ -217,15 +218,15 @@ struct tgsi_full_instruction vl_inst4
 
    inst.Instruction.Opcode = opcode;
    inst.Instruction.NumDstRegs = 1;
-   inst.FullDstRegisters[0].DstRegister.File = dst_file;
-   inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+   inst.Dst[0].Register.File = dst_file;
+   inst.Dst[0].Register.Index = dst_index;
    inst.Instruction.NumSrcRegs = 3;
-   inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
-   inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
-   inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
-   inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
-   inst.FullSrcRegisters[2].SrcRegister.File = src3_file;
-   inst.FullSrcRegisters[2].SrcRegister.Index = src3_index;
+   inst.Src[0].Register.File = src1_file;
+   inst.Src[0].Register.Index = src1_index;
+   inst.Src[1].Register.File = src2_file;
+   inst.Src[1].Register.Index = src2_index;
+   inst.Src[2].Register.File = src3_file;
+   inst.Src[2].Register.Index = src3_index;
 
    return inst;
 }