Merge branch 'master' into instanced-arrays

Conflicts:
	src/gallium/auxiliary/tgsi/tgsi_dump.c
	src/gallium/include/pipe/p_shader_tokens.h

77 files changed, 3060 insertions, 2408 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 80bd0c91db..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;
@@ -85,8 +86,8 @@ struct cso_context {
    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;
@@ -1027,3 +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 e5b92177cf..b9e313e32d 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.h
+++ b/src/gallium/auxiliary/cso_cache/cso_context.h
@@ -146,6 +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 cc5f7f0105..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;
 }
 
 
@@ -421,3 +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 c0f6a61411..8a64c06efc 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
@@ -140,6 +152,7 @@ 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 );
 
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 14375426ed..4585dcdb48 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -660,13 +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 */
@@ -707,7 +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;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
index 75130a8fb0..d86717e518 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -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_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 129d919a84..2a11b8c545 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -152,8 +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) */
@@ -211,6 +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];
@@ -222,7 +235,7 @@ struct draw_context
       uint semantic_name;
       uint semantic_index;
       int slot;
-   } extra_vp_outputs;
+   } extra_shader_outputs;
 
    unsigned reduced_prim;
 
@@ -247,6 +260,19 @@ 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_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
index 0238f2e234..279f4eec63 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"
 
 
@@ -131,7 +132,8 @@ 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 );
 
@@ -159,13 +161,21 @@ 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,
@@ -208,6 +218,7 @@ 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 );
 
@@ -238,10 +249,19 @@ 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,
@@ -282,6 +302,7 @@ 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 );
 
@@ -308,10 +329,19 @@ 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 9dfb47837e..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;
 
@@ -190,7 +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__);
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_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index 60b7a3ea36..9f40030f39 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -148,11 +148,12 @@ 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
@@ -210,7 +211,7 @@ 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 5cafe8c3f0..8f7d3b7100 100644
--- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
+++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
@@ -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;
@@ -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;
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/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/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_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_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c
index 92903fe57f..de9cbc8630 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.c
@@ -1025,7 +1025,7 @@ tgsi_build_full_property(
    size++;
 
    *property = tgsi_build_property(
-      TGSI_PROPERTY_GS_INPUT_PRIM,
+      full_prop->Property.PropertyName,
       header );
 
    assert( full_prop->Property.NrTokens <= 8 + 1 );
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index 4391ca75d1..d7ff262f30 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -123,12 +123,15 @@ static const char *semantic_names[] =
    "NORMAL",
    "FACE",
    "EDGEFLAG",
+   "PRIM_ID",
    "INSTANCEID"
 };
 
 static const char *immediate_type_names[] =
 {
-   "FLT32"
+   "FLT32",
+   "UINT32",
+   "INT32"
 };
 
 static const char *swizzle_names[] =
@@ -175,13 +178,19 @@ static const char *primitive_names[] =
 
 
 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) {
@@ -192,6 +201,52 @@ _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,
@@ -245,7 +300,7 @@ iter_declaration(
 
    TXT( "DCL " );
 
-   _dump_register(
+   _dump_register_decl(
       ctx,
       decl->Declaration.File,
       decl->Range.First,
@@ -360,6 +415,12 @@ 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 );
       }
@@ -436,10 +497,9 @@ iter_instruction(
             dst->Indirect.SwizzleX );
       }
       else {
-         _dump_register(
+         _dump_register_dst(
             ctx,
             dst->Register.File,
-            dst->Register.Index,
             dst->Register.Index );
       }
       _dump_writemask( ctx, dst->Register.WriteMask );
@@ -455,26 +515,11 @@ iter_instruction(
       CHR( ' ' );
 
       if (src->Register.Negate)
-         TXT( "-(" );
+         CHR( '-' );
       if (src->Register.Absolute)
          CHR( '|' );
 
-      if (src->Register.Indirect) {
-         _dump_register_ind(
-            ctx,
-            src->Register.File,
-            src->Register.Index,
-            src->Indirect.File,
-            src->Indirect.Index,
-            src->Indirect.SwizzleX );
-      }
-      else {
-         _dump_register(
-            ctx,
-            src->Register.File,
-            src->Register.Index,
-            src->Register.Index );
-      }
+      _dump_register_src(ctx, src);
 
       if (src->Register.SwizzleX != TGSI_SWIZZLE_X ||
           src->Register.SwizzleY != TGSI_SWIZZLE_Y ||
@@ -489,8 +534,6 @@ iter_instruction(
 
       if (src->Register.Absolute)
          CHR( '|' );
-      if (src->Register.Negate)
-         CHR( ')' );
 
       first_reg = FALSE;
    }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 123117cb0a..f43233bdb4 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
@@ -60,6 +61,7 @@
 #include "util/u_memory.h"
 #include "util/u_math.h"
 
+
 #define FAST_MATH 1
 
 #define TILE_TOP_LEFT     0
@@ -67,11 +69,329 @@
 #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)
  */
@@ -123,23 +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
@@ -292,6 +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]));
@@ -372,6 +696,7 @@ 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. */
@@ -413,18 +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,
@@ -437,76 +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_BOTTOM_LEFT] - src->f[TILE_TOP_LEFT];
-}
-
 static void
 micro_div(
    union tgsi_exec_channel *dst,
@@ -527,99 +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
-
-#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
-
-   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)
@@ -647,71 +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,
@@ -725,38 +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,
@@ -769,34 +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,
@@ -809,48 +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,
@@ -865,20 +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,
@@ -942,42 +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,
@@ -998,88 +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 )
 {
@@ -1101,31 +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,
@@ -1224,11 +1050,11 @@ fetch_src_file_channel(
 }
 
 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;
@@ -1277,10 +1103,10 @@ fetch_source(
          &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.
@@ -1357,10 +1183,10 @@ fetch_source(
             &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.
@@ -1385,32 +1211,30 @@ fetch_source(
       &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;
+   if (reg->Register.Absolute) {
+      if (src_datatype == TGSI_EXEC_DATA_FLOAT) {
+         micro_abs(chan, chan);
+      } else {
+         micro_iabs(chan, chan);
+      }
+   }
 
-   case TGSI_UTIL_SIGN_KEEP:
-      break;
+   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;
@@ -1419,9 +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
@@ -1456,7 +1280,7 @@ store_dest(
          &indir_index );
 
       /* save indirection offset */
-      offset = (int) indir_index.f[0];
+      offset = indir_index.i[0];
    }
 
    switch (reg->Register.File) {
@@ -1468,6 +1292,15 @@ store_dest(
       index = mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0]
          + 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:
@@ -1577,10 +1410,10 @@ store_dest(
 }
 
 #define FETCH(VAL,INDEX,CHAN)\
-    fetch_source (mach, VAL, &inst->Src[INDEX], CHAN)
+    fetch_source(mach, VAL, &inst->Src[INDEX], CHAN, TGSI_EXEC_DATA_FLOAT)
 
 #define STORE(VAL,INDEX,CHAN)\
-    store_dest (mach, VAL, &inst->Dst[INDEX], inst, CHAN )
+   store_dest(mach, VAL, &inst->Dst[INDEX], inst, CHAN, TGSI_EXEC_DATA_FLOAT)
 
 
 /**
@@ -1638,6 +1471,35 @@ exec_kilp(struct tgsi_exec_machine *mach,
    mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
 }
 
+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]]++;
+   }
+}
+
+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.
@@ -1908,7 +1770,7 @@ exec_declaration(struct tgsi_exec_machine *mach,
          if (decl->Semantic.Name == TGSI_SEMANTIC_POSITION) {
             assert(decl->Semantic.Index == 0);
             assert(first == last);
-            assert(mask = TGSI_WRITEMASK_XYZW);
+            assert(mask == TGSI_WRITEMASK_XYZW);
 
             mach->Inputs[first] = mach->QuadPos;
          } else if (decl->Semantic.Name == TGSI_SEMANTIC_FACE) {
@@ -1954,6 +1816,461 @@ exec_declaration(struct tgsi_exec_machine *mach,
    }
 }
 
+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_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,
@@ -1968,23 +2285,11 @@ exec_instruction(
 
    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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_unary(mach, inst, micro_arl, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_MOV:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH(&d[chan_index], 0, chan_index);
-      }
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_unary(mach, inst, micro_mov, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_LIT:
@@ -2021,23 +2326,11 @@ exec_instruction(
       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:
@@ -2208,41 +2501,15 @@ exec_instruction(
       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(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[1], &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0], &r[1]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_trinary(mach, inst, micro_mad, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SUB:
@@ -2257,17 +2524,7 @@ exec_instruction(
       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(&d[chan_index], &r[0], &r[2]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_trinary(mach, inst, micro_lrp, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_CND:
@@ -2301,13 +2558,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_FRC:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_frc(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_unary(mach, inst, micro_frc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_CLAMP:
@@ -2323,33 +2574,20 @@ exec_instruction(
       }
       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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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);
-
-      micro_exp2( &r[0], &r[0] );
-
-      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:
@@ -2402,15 +2640,9 @@ exec_instruction(
       }
       break;
 
-    case TGSI_OPCODE_ABS:
-       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-          FETCH(&r[0], 0, chan_index);
-          micro_abs(&d[chan_index], &r[0]);
-       }
-       FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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);
@@ -2449,33 +2681,15 @@ exec_instruction(
       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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_unary(mach, inst, micro_ddy, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_KILP:
@@ -2552,14 +2766,7 @@ 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(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_binary(mach, inst, micro_seq, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_SFL:
@@ -2569,44 +2776,19 @@ 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(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_binary(mach, inst, micro_sne, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_STR:
@@ -2711,6 +2893,10 @@ 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;
@@ -2730,6 +2916,8 @@ 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;
 
@@ -2737,12 +2925,17 @@ 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 */
@@ -2775,6 +2968,12 @@ 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];
 
@@ -2785,14 +2984,7 @@ 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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_unary(mach, inst, micro_sgn, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT);
       break;
 
    case TGSI_OPCODE_CMP:
@@ -2976,87 +3168,31 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_CEIL:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         micro_ceil(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&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_SHR:
-      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
-         FETCH( &r[0], 0, chan_index );
-         FETCH( &r[1], 1, chan_index );
-         micro_ishr(&d[chan_index], &r[0], &r[1]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0], &r[1]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 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(&d[chan_index], &r[0], &r[1]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_binary(mach, inst, micro_or, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_MOD:
@@ -3064,14 +3200,7 @@ 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(&d[chan_index], &r[0], &r[1]);
-      }
-      FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) {
-         STORE(&d[chan_index], 0, chan_index);
-      }
+      exec_vector_binary(mach, inst, micro_xor, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT);
       break;
 
    case TGSI_OPCODE_SAD:
@@ -3087,13 +3216,11 @@ 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:
@@ -3122,11 +3249,15 @@ exec_instruction(
    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:
@@ -3173,6 +3304,8 @@ exec_instruction(
          --mach->LoopLabelStackTop;
          assert(mach->LoopCounterStackTop > 0);
          --mach->LoopCounterStackTop;
+
+         mach->BreakType = mach->BreakStack[--mach->BreakStackTop];
       }
       UPDATE_EXEC_MASK(mach);
       break;
@@ -3196,15 +3329,14 @@ 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:
@@ -3235,6 +3367,12 @@ 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];
 
@@ -3265,11 +3403,116 @@ exec_instruction(
       UPDATE_EXEC_MASK(mach);
       break;
 
+   case TGSI_OPCODE_F2I:
+      exec_vector_unary(mach, inst, micro_f2i, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT);
+      break;
+
+   case TGSI_OPCODE_IDIV:
+      exec_vector_binary(mach, inst, micro_idiv, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT);
+      break;
+
+   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:
       assert( 0 );
    }
 }
 
+
 #define DEBUG_EXECUTION 0
 
 
@@ -3289,9 +3532,13 @@ 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;
@@ -3348,11 +3595,11 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
                   if (j > 0) {
                      debug_printf("           ");
                   }
-                  debug_printf("(%6f, %6f, %6f, %6f)\n",
-                               temps[i].xyzw[0].f[j],
-                               temps[i].xyzw[1].f[j],
-                               temps[i].xyzw[2].f[j],
-                               temps[i].xyzw[3].f[j]);
+                  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]);
                }
             }
          }
@@ -3366,11 +3613,11 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
                   if (j > 0) {
                      debug_printf("           ");
                   }
-                  debug_printf("{%6f, %6f, %6f, %6f}\n",
-                               outputs[i].xyzw[0].f[j],
-                               outputs[i].xyzw[1].f[j],
-                               outputs[i].xyzw[2].f[j],
-                               outputs[i].xyzw[3].f[j]);
+                  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]);
                }
             }
          }
@@ -3392,6 +3639,8 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
    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 fd94c1bc44..aa3a98d7f1 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -179,6 +179,7 @@ struct tgsi_exec_labels
 
 #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
@@ -191,6 +192,14 @@ 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
@@ -198,10 +207,29 @@ 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.
  */
@@ -214,8 +242,8 @@ 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;
@@ -229,6 +257,8 @@ struct tgsi_exec_machine
 
    /* GEOMETRY processor only. */
    unsigned                      *Primitives;
+   unsigned                       NumOutputs;
+   unsigned                       MaxGeometryShaderOutputs;
 
    /* FRAGMENT processor only. */
    const struct tgsi_interp_coef *InterpCoefs;
@@ -242,6 +272,12 @@ 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;
@@ -262,6 +298,13 @@ 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 be375cabb8..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 },
@@ -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 },
-   { 0, 0, 0, 0, 0, 0, "", 118 }      /* removed */
+   { 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_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h
index b34263da48..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)
@@ -146,6 +145,28 @@ OP01(IFC)
 OP01(BREAKC)
 OP01(KIL)
 OP00(END)
+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 fa65ecb997..8c7062d850 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_parse.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c
@@ -119,17 +119,29 @@ tgsi_parse_token(
    case TGSI_TOKEN_TYPE_IMMEDIATE:
    {
       struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate;
+      uint imm_count;
 
       memset(imm, 0, sizeof *imm);
       copy_token(&imm->Immediate, &token);
 
+      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;
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index c27579e794..9b0644465a 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 1024
-#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] =
@@ -148,31 +242,40 @@ static const char *file_names[TGSI_FILE_COUNT] =
 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, 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_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_used( ctx, reg ))
+         cso_hash_insert(ctx->regs_used, scan_register_key(reg), reg);
+      else
+         FREE(reg);
    }
    return TRUE;
 }
@@ -210,33 +313,33 @@ 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->Dst[i].Register.File,
-         inst->Dst[i].Register.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->Src[i].Register.File,
-         inst->Src[i].Register.Index,
+         reg,
          "source",
          (boolean)inst->Src[i].Register.Indirect );
       if (inst->Src[i].Register.Indirect) {
-         uint file;
-         int index;
+         scan_register *ind_reg = MALLOC(sizeof(scan_register));
 
-         file = inst->Src[i].Indirect.File;
-         index = inst->Src[i].Indirect.Index;
+         fill_scan_register1d(ind_reg,
+                              inst->Src[i].Indirect.File,
+                              inst->Src[i].Indirect.Index);
          check_register_usage(
             ctx,
-            file,
-            index,
+            reg,
             "indirect",
             FALSE );
-         if (!(file == TGSI_FILE_ADDRESS || file == TGSI_FILE_LOOP) || index != 0) {
+         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]");
          }
       }
@@ -266,6 +369,19 @@ 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,
@@ -287,9 +403,21 @@ iter_declaration(
    if (!check_file_name( ctx, file ))
       return TRUE;
    for (i = decl->Range.First; i <= decl->Range.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));
+      /* 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;
@@ -301,8 +429,7 @@ 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.
     */
@@ -311,12 +438,16 @@ 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;
    }
@@ -330,8 +461,12 @@ iter_property(
    struct tgsi_iterate_context *iter,
    struct tgsi_full_property *prop )
 {
-   /*struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;*/
+   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;
 }
 
@@ -340,7 +475,6 @@ 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.
     */
@@ -350,13 +484,17 @@ 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);
       }
    }
 
@@ -368,6 +506,18 @@ 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);
+      FREE(reg);
+   }
+   cso_hash_delete(hash);
+}
+
 boolean
 tgsi_sanity_check(
    const struct tgsi_token *tokens )
@@ -381,18 +531,23 @@ tgsi_sanity_check(
    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 0f48b0dc3a..a6cc773003 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
@@ -229,8 +229,8 @@ tgsi_is_passthrough_shader(const struct tgsi_token *tokens)
 
             /* Do a whole bunch of checks for a simple move */
             if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV ||
-                src->Register.File != TGSI_FILE_INPUT ||
-                src->Register.File != TGSI_FILE_SYSTEM_VALUE ||
+                (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 ||
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
index d63c75dafb..118059ace9 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -2578,7 +2578,7 @@ emit_instruction(
       return 0;
       break;
 
-   case TGSI_OPCODE_SHR:
+   case TGSI_OPCODE_ISHR:
       return 0;
       break;
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index f000958bfc..9fcffeda36 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -27,7 +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"
@@ -61,18 +63,20 @@ static boolean uprcase( char c )
 }
 
 /*
- * Ignore case of str1 and assume str2 is already uppercase.
+ * 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 && uprcase(*str1) == *str2) {
+   while (*str1 && *str2) {
+      if (*str1 != uprcase(*str2))
+         return FALSE;
       str1++;
       str2++;
    }
-   return *str1 == *str2;
+   return TRUE;
 }
 
 static boolean str_match_no_case( const char **pcur, const char *str )
@@ -193,11 +197,26 @@ 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.
@@ -229,6 +248,7 @@ static boolean parse_header( struct translate_ctx *ctx )
    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;
 }
@@ -325,92 +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 );
 
@@ -420,16 +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'");
@@ -450,12 +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 {
@@ -463,9 +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 != ']') {
@@ -476,20 +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)
+{
+
+   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)
 {
-   if (!parse_register_file_bracket_index( ctx, file, first ))
+   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;
 
@@ -499,12 +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;
@@ -513,6 +582,70 @@ parse_register_dcl(
    return TRUE;
 }
 
+/* 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)
+{
+   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(
@@ -582,37 +715,44 @@ parse_src_operand(
    struct tgsi_full_src_register *src )
 {
    uint file;
-   int index;
-   uint ind_file;
-   int ind_index;
-   uint ind_comp;
    uint swizzle[4];
    boolean parsed_swizzle;
+   struct parsed_src_bracket bracket[2];
+   int parsed_opt_brackets;
 
    if (*ctx->cur == '-') {
       ctx->cur++;
       eat_opt_white( &ctx->cur );
       src->Register.Negate = 1;
    }
-   
+
    if (*ctx->cur == '|') {
       ctx->cur++;
       eat_opt_white( &ctx->cur );
       src->Register.Absolute = 1;
    }
 
-   if (!parse_register_src(ctx, &file, &index, &ind_file, &ind_index, &ind_comp))
+   if (!parse_register_src(ctx, &file, &bracket[0]))
+      return FALSE;
+   if (!parse_opt_register_src_bracket(ctx, &bracket[1], &parsed_opt_brackets))
       return FALSE;
+
    src->Register.File = file;
-   src->Register.Index = index;
-   if (ind_file != TGSI_FILE_NULL) {
+   src->Register.Index = bracket[0].index;
+   if (bracket[0].ind_file != TGSI_FILE_NULL) {
       src->Register.Indirect = 1;
-      src->Indirect.File = ind_file;
-      src->Indirect.Index = ind_index;
-      src->Indirect.SwizzleX = ind_comp;
-      src->Indirect.SwizzleY = ind_comp;
-      src->Indirect.SwizzleZ = ind_comp;
-      src->Indirect.SwizzleW = ind_comp;
+      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 (parsed_opt_brackets) {
+      src->Register.Dimension = 1;
+      src->Dimension.Indirect = 0;
+      src->Dimension.Dimension = 0;
+      src->Dimension.Index = bracket[1].index;
    }
 
    /* Parse optional swizzle.
@@ -791,7 +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] =
@@ -805,8 +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;
@@ -818,7 +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;
@@ -826,8 +968,8 @@ static boolean parse_declaration( struct translate_ctx *ctx )
    decl = tgsi_default_full_declaration();
    decl.Declaration.File = file;
    decl.Declaration.UsageMask = writemask;
-   decl.Range.First = first;
-   decl.Range.Last = last;
+   decl.Range.First = brackets[0].first;
+   decl.Range.Last = brackets[0].last;
 
    cur = ctx->cur;
    eat_opt_white( &cur );
@@ -1027,7 +1169,7 @@ static boolean parse_property( struct translate_ctx *ctx )
    }
    for (property_name = 0; property_name < TGSI_PROPERTY_COUNT;
         ++property_name) {
-      if (streq_nocase_uprcase(id, property_names[property_name])) {
+      if (streq_nocase_uprcase(property_names[property_name], id)) {
          break;
       }
    }
@@ -1044,6 +1186,10 @@ static boolean parse_property( struct translate_ctx *ctx )
          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] )) {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
index 2713372b05..2b51672b8e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
@@ -109,8 +109,13 @@ struct ureg_program
    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;
 
@@ -513,22 +518,22 @@ 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;
@@ -536,24 +541,28 @@ 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;
@@ -563,38 +572,82 @@ 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);
 }
 
 
@@ -818,8 +871,8 @@ ureg_insn(struct ureg_program *ureg,
    unsigned i;
    boolean saturate;
    boolean predicate;
-   boolean negate;
-   unsigned swizzle[4];
+   boolean negate = FALSE;
+   unsigned swizzle[4] = { 0 };
 
    saturate = nr_dst ? dst[0].Saturate : FALSE;
    predicate = nr_dst ? dst[0].Predicate : FALSE;
@@ -865,8 +918,8 @@ ureg_tex_insn(struct ureg_program *ureg,
    unsigned i;
    boolean saturate;
    boolean predicate;
-   boolean negate;
-   unsigned swizzle[4];
+   boolean negate = FALSE;
+   unsigned swizzle[4] = { 0 };
 
    saturate = nr_dst ? dst[0].Saturate : FALSE;
    predicate = nr_dst ? dst[0].Predicate : FALSE;
@@ -982,21 +1035,23 @@ 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[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];
 }
 
 
@@ -1091,7 +1146,8 @@ static void emit_decls( struct ureg_program *ureg )
 
    for (i = 0; i < ureg->nr_immediates; i++) {
       emit_immediate( ureg,
-                      ureg->immediate[i].v );
+                      ureg->immediate[i].value.u,
+                      ureg->immediate[i].type );
    }
 }
 
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
index 179862d4e2..38e2fd8d0a 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
@@ -154,6 +154,16 @@ 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 );
 
@@ -227,6 +237,90 @@ 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
  */
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 3ed90fd1b7..0000000000
--- a/src/gallium/auxiliary/util/Makefile
+++ /dev/null
@@ -1,48 +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_blitter.c \
-	u_cache.c \
-	u_cpu_detect.c \
-	u_dl.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_texture.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 2a546d19dc..0000000000
--- a/src/gallium/auxiliary/util/SConscript
+++ /dev/null
@@ -1,61 +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_blitter.c',
-		'u_cache.c',
-		'u_cpu_detect.c',
-		'u_debug.c',
-		'u_debug_dump.c',
-		'u_debug_memory.c',
-		'u_debug_stack.c',
-		'u_debug_symbol.c',
-		'u_dl.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_texture.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_debug.c b/src/gallium/auxiliary/util/u_debug.c
index 4e01123fff..9b4e6ca2a7 100644
--- a/src/gallium/auxiliary/util/u_debug.c
+++ b/src/gallium/auxiliary/util/u_debug.c
@@ -454,7 +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;
@@ -467,7 +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)
diff --git a/src/gallium/auxiliary/util/u_debug.h b/src/gallium/auxiliary/util/u_debug.h
index abd834c741..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
 
 
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_format.h b/src/gallium/auxiliary/util/u_format.h
index 090183fb17..a558923b2e 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -119,7 +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
 };
 
 
diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h
index 81aeb83cbb..b2969a210a 100644
--- a/src/gallium/auxiliary/util/u_math.h
+++ b/src/gallium/auxiliary/util/u_math.h
@@ -585,13 +585,12 @@ do {                                     \
 
 static INLINE uint32_t util_unsigned_fixed(float value, unsigned frac_bits)
 {
-   value *= (1<<frac_bits);
-   return value < 0 ? 0 : value;
+   return value < 0 ? 0 : (uint32_t)(value * (1<<frac_bits));
 }
 
 static INLINE int32_t util_signed_fixed(float value, unsigned frac_bits)
 {
-   return value * (1<<frac_bits);
+   return (int32_t)(value * (1<<frac_bits));
 }
 
 
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_prim.h b/src/gallium/auxiliary/util/u_prim.h
index 7434329962..10a874f341 100644
--- a/src/gallium/auxiliary/util/u_prim.h
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -135,6 +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_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/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_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
index ab196c21f8..caf581aca6 100644
--- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
+++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
@@ -334,11 +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)
@@ -442,11 +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)
@@ -532,11 +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)
@@ -658,11 +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)
@@ -1081,6 +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:
       {